diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
146 files changed, 11639 insertions, 6383 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Kconfig b/drivers/gpu/drm/amd/amdgpu/Kconfig index e8af1f5e8a79..9221e5489069 100644 --- a/drivers/gpu/drm/amd/amdgpu/Kconfig +++ b/drivers/gpu/drm/amd/amdgpu/Kconfig @@ -42,3 +42,4 @@ config DRM_AMDGPU_GART_DEBUGFS source "drivers/gpu/drm/amd/acp/Kconfig" source "drivers/gpu/drm/amd/display/Kconfig" +source "drivers/gpu/drm/amd/amdkfd/Kconfig" diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index bfd332c95b61..138cb787d27e 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -35,7 +35,8 @@ ccflags-y := -I$(FULL_AMD_PATH)/include/asic_reg \ -I$(FULL_AMD_DISPLAY_PATH) \ -I$(FULL_AMD_DISPLAY_PATH)/include \ -I$(FULL_AMD_DISPLAY_PATH)/dc \ - -I$(FULL_AMD_DISPLAY_PATH)/amdgpu_dm + -I$(FULL_AMD_DISPLAY_PATH)/amdgpu_dm \ + -I$(FULL_AMD_PATH)/amdkfd amdgpu-y := amdgpu_drv.o @@ -51,8 +52,8 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ amdgpu_prime.o amdgpu_vm.o amdgpu_ib.o amdgpu_pll.o \ amdgpu_ucode.o amdgpu_bo_list.o amdgpu_ctx.o amdgpu_sync.o \ amdgpu_gtt_mgr.o amdgpu_vram_mgr.o amdgpu_virt.o amdgpu_atomfirmware.o \ - amdgpu_queue_mgr.o amdgpu_vf_error.o amdgpu_sched.o amdgpu_debugfs.o \ - amdgpu_ids.o + amdgpu_vf_error.o amdgpu_sched.o amdgpu_debugfs.o amdgpu_ids.o \ + amdgpu_gmc.o amdgpu_xgmi.o # add asic specific block amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \ @@ -62,7 +63,7 @@ amdgpu-$(CONFIG_DRM_AMDGPU_SI)+= si.o gmc_v6_0.o gfx_v6_0.o si_ih.o si_dma.o dce amdgpu-y += \ vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o \ - vega20_reg_init.o + vega20_reg_init.o nbio_v7_4.o # add DF block amdgpu-y += \ @@ -73,7 +74,7 @@ amdgpu-y += \ amdgpu-y += \ gmc_v7_0.o \ gmc_v8_0.o \ - gfxhub_v1_0.o mmhub_v1_0.o gmc_v9_0.o + gfxhub_v1_0.o mmhub_v1_0.o gmc_v9_0.o gfxhub_v1_1.o # add IH block amdgpu-y += \ @@ -88,7 +89,8 @@ amdgpu-y += \ amdgpu-y += \ amdgpu_psp.o \ psp_v3_1.o \ - psp_v10_0.o + psp_v10_0.o \ + psp_v11_0.o # add SMC block amdgpu-y += \ @@ -108,6 +110,7 @@ amdgpu-y += \ # add async DMA block amdgpu-y += \ + amdgpu_sdma.o \ sdma_v2_4.o \ sdma_v3_0.o \ sdma_v4_0.o @@ -134,6 +137,9 @@ amdgpu-y += \ amdgpu-y += amdgpu_amdkfd.o ifneq ($(CONFIG_HSA_AMD),) +AMDKFD_PATH := ../amdkfd +include $(FULL_AMD_PATH)/amdkfd/Makefile +amdgpu-y += $(AMDKFD_FILES) amdgpu-y += \ amdgpu_amdkfd_fence.o \ amdgpu_amdkfd_gpuvm.o \ diff --git a/drivers/gpu/drm/amd/amdgpu/ObjectID.h b/drivers/gpu/drm/amd/amdgpu/ObjectID.h index 06192698bd96..5b393622f592 100644 --- a/drivers/gpu/drm/amd/amdgpu/ObjectID.h +++ b/drivers/gpu/drm/amd/amdgpu/ObjectID.h @@ -136,6 +136,7 @@ #define GENERIC_OBJECT_ID_PX2_NON_DRIVABLE 0x02 #define GENERIC_OBJECT_ID_MXM_OPM 0x03 #define GENERIC_OBJECT_ID_STEREO_PIN 0x04 //This object could show up from Misc Object table, it follows ATOM_OBJECT format, and contains one ATOM_OBJECT_GPIO_CNTL_RECORD for the stereo pin +#define GENERIC_OBJECT_ID_BRACKET_LAYOUT 0x05 /****************************************************/ /* Graphics Object ENUM ID Definition */ @@ -714,6 +715,13 @@ GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ GENERIC_OBJECT_ID_STEREO_PIN << OBJECT_ID_SHIFT) +#define GENERICOBJECT_BRACKET_LAYOUT_ENUM_ID1 (GRAPH_OBJECT_TYPE_GENERIC << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID1 << ENUM_ID_SHIFT |\ + GENERIC_OBJECT_ID_BRACKET_LAYOUT << OBJECT_ID_SHIFT) + +#define GENERICOBJECT_BRACKET_LAYOUT_ENUM_ID2 (GRAPH_OBJECT_TYPE_GENERIC << OBJECT_TYPE_SHIFT |\ + GRAPH_OBJECT_ENUM_ID2 << ENUM_ID_SHIFT |\ + GENERIC_OBJECT_ID_BRACKET_LAYOUT << OBJECT_ID_SHIFT) /****************************************************/ /* Object Cap definition - Shared with BIOS */ /****************************************************/ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index 7dcbac8af9a7..d0102cfc8efb 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -28,6 +28,8 @@ #ifndef __AMDGPU_H__ #define __AMDGPU_H__ +#include "amdgpu_ctx.h" + #include <linux/atomic.h> #include <linux/wait.h> #include <linux/list.h> @@ -69,10 +71,32 @@ #include "amdgpu_vcn.h" #include "amdgpu_mn.h" #include "amdgpu_gmc.h" +#include "amdgpu_gfx.h" +#include "amdgpu_sdma.h" #include "amdgpu_dm.h" #include "amdgpu_virt.h" #include "amdgpu_gart.h" #include "amdgpu_debugfs.h" +#include "amdgpu_job.h" +#include "amdgpu_bo_list.h" +#include "amdgpu_gem.h" + +#define MAX_GPU_INSTANCE 16 + +struct amdgpu_gpu_instance +{ + struct amdgpu_device *adev; + int mgpu_fan_enabled; +}; + +struct amdgpu_mgpu_info +{ + struct amdgpu_gpu_instance gpu_ins[MAX_GPU_INSTANCE]; + struct mutex mutex; + uint32_t num_gpu; + uint32_t num_dgpu; + uint32_t num_apu; +}; /* * Modules parameters. @@ -105,11 +129,8 @@ extern int amdgpu_vm_fault_stop; extern int amdgpu_vm_debug; extern int amdgpu_vm_update_mode; extern int amdgpu_dc; -extern int amdgpu_dc_log; extern int amdgpu_sched_jobs; extern int amdgpu_sched_hw_submission; -extern int amdgpu_no_evict; -extern int amdgpu_direct_gma_size; extern uint amdgpu_pcie_gen_cap; extern uint amdgpu_pcie_lane_cap; extern uint amdgpu_cg_mask; @@ -130,6 +151,7 @@ extern int amdgpu_compute_multipipe; extern int amdgpu_gpu_recovery; extern int amdgpu_emu_mode; extern uint amdgpu_smu_memory_pool_size; +extern struct amdgpu_mgpu_info mgpu_info; #ifdef CONFIG_DRM_AMDGPU_SI extern int amdgpu_si_support; @@ -149,9 +171,6 @@ extern int amdgpu_cik_support; #define AMDGPUFB_CONN_LIMIT 4 #define AMDGPU_BIOS_NUM_SCRATCH 16 -/* max number of IP instances */ -#define AMDGPU_MAX_SDMA_INSTANCES 2 - /* hard reset data */ #define AMDGPU_ASIC_RESET_DATA 0x39d5e86b @@ -172,13 +191,6 @@ extern int amdgpu_cik_support; #define AMDGPU_RESET_VCE (1 << 13) #define AMDGPU_RESET_VCE1 (1 << 14) -/* GFX current status */ -#define AMDGPU_GFX_NORMAL_MODE 0x00000000L -#define AMDGPU_GFX_SAFE_MODE 0x00000001L -#define AMDGPU_GFX_PG_DISABLED_MODE 0x00000002L -#define AMDGPU_GFX_CG_DISABLED_MODE 0x00000004L -#define AMDGPU_GFX_LBPW_DISABLED_MODE 0x00000008L - /* max cursor sizes (in pixels) */ #define CIK_CURSOR_WIDTH 128 #define CIK_CURSOR_HEIGHT 128 @@ -206,13 +218,6 @@ enum amdgpu_cp_irq { AMDGPU_CP_IRQ_LAST }; -enum amdgpu_sdma_irq { - AMDGPU_SDMA_IRQ_TRAP0 = 0, - AMDGPU_SDMA_IRQ_TRAP1, - - AMDGPU_SDMA_IRQ_LAST -}; - enum amdgpu_thermal_irq { AMDGPU_THERMAL_IRQ_LOW_TO_HIGH = 0, AMDGPU_THERMAL_IRQ_HIGH_TO_LOW, @@ -225,6 +230,10 @@ enum amdgpu_kiq_irq { AMDGPU_CP_KIQ_IRQ_LAST }; +#define MAX_KIQ_REG_WAIT 5000 /* in usecs, 5ms */ +#define MAX_KIQ_REG_BAILOUT_INTERVAL 5 /* in msecs, 5ms */ +#define MAX_KIQ_REG_TRY 20 + int amdgpu_device_ip_set_clockgating_state(void *dev, enum amd_ip_block_type block_type, enum amd_clockgating_state state); @@ -272,70 +281,6 @@ amdgpu_device_ip_get_ip_block(struct amdgpu_device *adev, int amdgpu_device_ip_block_add(struct amdgpu_device *adev, const struct amdgpu_ip_block_version *ip_block_version); -/* provided by hw blocks that can move/clear data. e.g., gfx or sdma */ -struct amdgpu_buffer_funcs { - /* maximum bytes in a single operation */ - uint32_t copy_max_bytes; - - /* number of dw to reserve per operation */ - unsigned copy_num_dw; - - /* used for buffer migration */ - void (*emit_copy_buffer)(struct amdgpu_ib *ib, - /* src addr in bytes */ - uint64_t src_offset, - /* dst addr in bytes */ - uint64_t dst_offset, - /* number of byte to transfer */ - uint32_t byte_count); - - /* maximum bytes in a single operation */ - uint32_t fill_max_bytes; - - /* number of dw to reserve per operation */ - unsigned fill_num_dw; - - /* used for buffer clearing */ - void (*emit_fill_buffer)(struct amdgpu_ib *ib, - /* value to write to memory */ - uint32_t src_data, - /* dst addr in bytes */ - uint64_t dst_offset, - /* number of byte to fill */ - uint32_t byte_count); -}; - -/* provided by hw blocks that can write ptes, e.g., sdma */ -struct amdgpu_vm_pte_funcs { - /* number of dw to reserve per operation */ - unsigned copy_pte_num_dw; - - /* copy pte entries from GART */ - void (*copy_pte)(struct amdgpu_ib *ib, - uint64_t pe, uint64_t src, - unsigned count); - - /* write pte one entry at a time with addr mapping */ - void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe, - uint64_t value, unsigned count, - uint32_t incr); - /* for linear pte/pde updates without addr mapping */ - void (*set_pte_pde)(struct amdgpu_ib *ib, - uint64_t pe, - uint64_t addr, unsigned count, - uint32_t incr, uint64_t flags); -}; - -/* provided by the ih block */ -struct amdgpu_ih_funcs { - /* ring read/write ptr handling, called from interrupt context */ - u32 (*get_wptr)(struct amdgpu_device *adev); - bool (*prescreen_iv)(struct amdgpu_device *adev); - void (*decode_iv)(struct amdgpu_device *adev, - struct amdgpu_iv_entry *entry); - void (*set_rptr)(struct amdgpu_device *adev); -}; - /* * BIOS. */ @@ -361,34 +306,6 @@ struct amdgpu_clock { uint32_t max_pixel_clock; }; -/* - * GEM. - */ - -#define AMDGPU_GEM_DOMAIN_MAX 0x3 -#define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, gem_base) - -void amdgpu_gem_object_free(struct drm_gem_object *obj); -int amdgpu_gem_object_open(struct drm_gem_object *obj, - struct drm_file *file_priv); -void amdgpu_gem_object_close(struct drm_gem_object *obj, - struct drm_file *file_priv); -unsigned long amdgpu_gem_timeout(uint64_t timeout_ns); -struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj); -struct drm_gem_object * -amdgpu_gem_prime_import_sg_table(struct drm_device *dev, - struct dma_buf_attachment *attach, - struct sg_table *sg); -struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev, - struct drm_gem_object *gobj, - int flags); -struct drm_gem_object *amdgpu_gem_prime_import(struct drm_device *dev, - struct dma_buf *dma_buf); -struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *); -void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj); -void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); -int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma); - /* sub-allocation manager, it has to be protected by another lock. * By conception this is an helper for other part of the driver * like the indirect buffer or semaphore, which both have their @@ -438,22 +355,6 @@ struct amdgpu_sa_bo { struct dma_fence *fence; }; -/* - * GEM objects. - */ -void amdgpu_gem_force_release(struct amdgpu_device *adev); -int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, - int alignment, u32 initial_domain, - u64 flags, enum ttm_bo_type type, - struct reservation_object *resv, - struct drm_gem_object **obj); - -int amdgpu_mode_dumb_create(struct drm_file *file_priv, - struct drm_device *dev, - struct drm_mode_create_dumb *args); -int amdgpu_mode_dumb_mmap(struct drm_file *filp, - struct drm_device *dev, - uint32_t handle, uint64_t *offset_p); int amdgpu_fence_slab_init(void); void amdgpu_fence_slab_fini(void); @@ -526,16 +427,25 @@ typedef enum _AMDGPU_DOORBELL64_ASSIGNMENT AMDGPU_DOORBELL64_GFX_RING0 = 0x8b, /* - * Other graphics doorbells can be allocated here: from 0x8c to 0xef + * Other graphics doorbells can be allocated here: from 0x8c to 0xdf * Graphics voltage island aperture 1 - * default non-graphics QWORD index is 0xF0 - 0xFF inclusive + * default non-graphics QWORD index is 0xe0 - 0xFF inclusive */ - /* sDMA engines */ - AMDGPU_DOORBELL64_sDMA_ENGINE0 = 0xF0, - AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0 = 0xF1, - AMDGPU_DOORBELL64_sDMA_ENGINE1 = 0xF2, - AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1 = 0xF3, + /* sDMA engines reserved from 0xe0 -oxef */ + AMDGPU_DOORBELL64_sDMA_ENGINE0 = 0xE0, + AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE0 = 0xE1, + AMDGPU_DOORBELL64_sDMA_ENGINE1 = 0xE8, + AMDGPU_DOORBELL64_sDMA_HI_PRI_ENGINE1 = 0xE9, + + /* For vega10 sriov, the sdma doorbell must be fixed as follow + * to keep the same setting with host driver, or it will + * happen conflicts + */ + AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 = 0xF0, + AMDGPU_VEGA10_DOORBELL64_sDMA_HI_PRI_ENGINE0 = 0xF1, + AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 = 0xF2, + AMDGPU_VEGA10_DOORBELL64_sDMA_HI_PRI_ENGINE1 = 0xF3, /* Interrupt handler */ AMDGPU_DOORBELL64_IH = 0xF4, /* For legacy interrupt ring buffer */ @@ -600,95 +510,6 @@ struct amdgpu_ib { extern const struct drm_sched_backend_ops amdgpu_sched_ops; -int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, - struct amdgpu_job **job, struct amdgpu_vm *vm); -int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, - struct amdgpu_job **job); - -void amdgpu_job_free_resources(struct amdgpu_job *job); -void amdgpu_job_free(struct amdgpu_job *job); -int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring, - struct drm_sched_entity *entity, void *owner, - struct dma_fence **f); - -/* - * Queue manager - */ -struct amdgpu_queue_mapper { - int hw_ip; - struct mutex lock; - /* protected by lock */ - struct amdgpu_ring *queue_map[AMDGPU_MAX_RINGS]; -}; - -struct amdgpu_queue_mgr { - struct amdgpu_queue_mapper mapper[AMDGPU_MAX_IP_NUM]; -}; - -int amdgpu_queue_mgr_init(struct amdgpu_device *adev, - struct amdgpu_queue_mgr *mgr); -int amdgpu_queue_mgr_fini(struct amdgpu_device *adev, - struct amdgpu_queue_mgr *mgr); -int amdgpu_queue_mgr_map(struct amdgpu_device *adev, - struct amdgpu_queue_mgr *mgr, - u32 hw_ip, u32 instance, u32 ring, - struct amdgpu_ring **out_ring); - -/* - * context related structures - */ - -struct amdgpu_ctx_ring { - uint64_t sequence; - struct dma_fence **fences; - struct drm_sched_entity entity; -}; - -struct amdgpu_ctx { - struct kref refcount; - struct amdgpu_device *adev; - struct amdgpu_queue_mgr queue_mgr; - unsigned reset_counter; - unsigned reset_counter_query; - uint32_t vram_lost_counter; - spinlock_t ring_lock; - struct dma_fence **fences; - struct amdgpu_ctx_ring rings[AMDGPU_MAX_RINGS]; - bool preamble_presented; - enum drm_sched_priority init_priority; - enum drm_sched_priority override_priority; - struct mutex lock; - atomic_t guilty; -}; - -struct amdgpu_ctx_mgr { - struct amdgpu_device *adev; - struct mutex lock; - /* protected by lock */ - struct idr ctx_handles; -}; - -struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id); -int amdgpu_ctx_put(struct amdgpu_ctx *ctx); - -int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, - struct dma_fence *fence, uint64_t *seq); -struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, - struct amdgpu_ring *ring, uint64_t seq); -void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, - enum drm_sched_priority priority); - -int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); - -int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id); - -void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr); -void amdgpu_ctx_mgr_entity_cleanup(struct amdgpu_ctx_mgr *mgr); -void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr); -void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); - - /* * file private structure */ @@ -702,293 +523,6 @@ struct amdgpu_fpriv { struct amdgpu_ctx_mgr ctx_mgr; }; -/* - * residency list - */ -struct amdgpu_bo_list_entry { - struct amdgpu_bo *robj; - struct ttm_validate_buffer tv; - struct amdgpu_bo_va *bo_va; - uint32_t priority; - struct page **user_pages; - int user_invalidated; -}; - -struct amdgpu_bo_list { - struct mutex lock; - struct rcu_head rhead; - struct kref refcount; - struct amdgpu_bo *gds_obj; - struct amdgpu_bo *gws_obj; - struct amdgpu_bo *oa_obj; - unsigned first_userptr; - unsigned num_entries; - struct amdgpu_bo_list_entry *array; -}; - -struct amdgpu_bo_list * -amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id); -void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, - struct list_head *validated); -void amdgpu_bo_list_put(struct amdgpu_bo_list *list); -void amdgpu_bo_list_free(struct amdgpu_bo_list *list); - -/* - * GFX stuff - */ -#include "clearstate_defs.h" - -struct amdgpu_rlc_funcs { - void (*enter_safe_mode)(struct amdgpu_device *adev); - void (*exit_safe_mode)(struct amdgpu_device *adev); -}; - -struct amdgpu_rlc { - /* for power gating */ - struct amdgpu_bo *save_restore_obj; - uint64_t save_restore_gpu_addr; - volatile uint32_t *sr_ptr; - const u32 *reg_list; - u32 reg_list_size; - /* for clear state */ - struct amdgpu_bo *clear_state_obj; - uint64_t clear_state_gpu_addr; - volatile uint32_t *cs_ptr; - const struct cs_section_def *cs_data; - u32 clear_state_size; - /* for cp tables */ - struct amdgpu_bo *cp_table_obj; - uint64_t cp_table_gpu_addr; - volatile uint32_t *cp_table_ptr; - u32 cp_table_size; - - /* safe mode for updating CG/PG state */ - bool in_safe_mode; - const struct amdgpu_rlc_funcs *funcs; - - /* for firmware data */ - u32 save_and_restore_offset; - u32 clear_state_descriptor_offset; - u32 avail_scratch_ram_locations; - u32 reg_restore_list_size; - u32 reg_list_format_start; - u32 reg_list_format_separate_start; - u32 starting_offsets_start; - u32 reg_list_format_size_bytes; - u32 reg_list_size_bytes; - u32 reg_list_format_direct_reg_list_length; - u32 save_restore_list_cntl_size_bytes; - u32 save_restore_list_gpm_size_bytes; - u32 save_restore_list_srm_size_bytes; - - u32 *register_list_format; - u32 *register_restore; - u8 *save_restore_list_cntl; - u8 *save_restore_list_gpm; - u8 *save_restore_list_srm; - - bool is_rlc_v2_1; -}; - -#define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES - -struct amdgpu_mec { - struct amdgpu_bo *hpd_eop_obj; - u64 hpd_eop_gpu_addr; - struct amdgpu_bo *mec_fw_obj; - u64 mec_fw_gpu_addr; - u32 num_mec; - u32 num_pipe_per_mec; - u32 num_queue_per_pipe; - void *mqd_backup[AMDGPU_MAX_COMPUTE_RINGS + 1]; - - /* These are the resources for which amdgpu takes ownership */ - DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); -}; - -struct amdgpu_kiq { - u64 eop_gpu_addr; - struct amdgpu_bo *eop_obj; - spinlock_t ring_lock; - struct amdgpu_ring ring; - struct amdgpu_irq_src irq; -}; - -/* - * GPU scratch registers structures, functions & helpers - */ -struct amdgpu_scratch { - unsigned num_reg; - uint32_t reg_base; - uint32_t free_mask; -}; - -/* - * GFX configurations - */ -#define AMDGPU_GFX_MAX_SE 4 -#define AMDGPU_GFX_MAX_SH_PER_SE 2 - -struct amdgpu_rb_config { - uint32_t rb_backend_disable; - uint32_t user_rb_backend_disable; - uint32_t raster_config; - uint32_t raster_config_1; -}; - -struct gb_addr_config { - uint16_t pipe_interleave_size; - uint8_t num_pipes; - uint8_t max_compress_frags; - uint8_t num_banks; - uint8_t num_se; - uint8_t num_rb_per_se; -}; - -struct amdgpu_gfx_config { - unsigned max_shader_engines; - unsigned max_tile_pipes; - unsigned max_cu_per_sh; - unsigned max_sh_per_se; - unsigned max_backends_per_se; - unsigned max_texture_channel_caches; - unsigned max_gprs; - unsigned max_gs_threads; - unsigned max_hw_contexts; - unsigned sc_prim_fifo_size_frontend; - unsigned sc_prim_fifo_size_backend; - unsigned sc_hiz_tile_fifo_size; - unsigned sc_earlyz_tile_fifo_size; - - unsigned num_tile_pipes; - unsigned backend_enable_mask; - unsigned mem_max_burst_length_bytes; - unsigned mem_row_size_in_kb; - unsigned shader_engine_tile_size; - unsigned num_gpus; - unsigned multi_gpu_tile_size; - unsigned mc_arb_ramcfg; - unsigned gb_addr_config; - unsigned num_rbs; - unsigned gs_vgt_table_depth; - unsigned gs_prim_buffer_depth; - - uint32_t tile_mode_array[32]; - uint32_t macrotile_mode_array[16]; - - struct gb_addr_config gb_addr_config_fields; - struct amdgpu_rb_config rb_config[AMDGPU_GFX_MAX_SE][AMDGPU_GFX_MAX_SH_PER_SE]; - - /* gfx configure feature */ - uint32_t double_offchip_lds_buf; - /* cached value of DB_DEBUG2 */ - uint32_t db_debug2; -}; - -struct amdgpu_cu_info { - uint32_t simd_per_cu; - uint32_t max_waves_per_simd; - uint32_t wave_front_size; - uint32_t max_scratch_slots_per_cu; - uint32_t lds_size; - - /* total active CU number */ - uint32_t number; - uint32_t ao_cu_mask; - uint32_t ao_cu_bitmap[4][4]; - uint32_t bitmap[4][4]; -}; - -struct amdgpu_gfx_funcs { - /* get the gpu clock counter */ - uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev); - void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num, u32 sh_num, u32 instance); - void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields); - void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t thread, uint32_t start, uint32_t size, uint32_t *dst); - void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t start, uint32_t size, uint32_t *dst); - void (*select_me_pipe_q)(struct amdgpu_device *adev, u32 me, u32 pipe, u32 queue); -}; - -struct amdgpu_ngg_buf { - struct amdgpu_bo *bo; - uint64_t gpu_addr; - uint32_t size; - uint32_t bo_size; -}; - -enum { - NGG_PRIM = 0, - NGG_POS, - NGG_CNTL, - NGG_PARAM, - NGG_BUF_MAX -}; - -struct amdgpu_ngg { - struct amdgpu_ngg_buf buf[NGG_BUF_MAX]; - uint32_t gds_reserve_addr; - uint32_t gds_reserve_size; - bool init; -}; - -struct amdgpu_gfx { - struct mutex gpu_clock_mutex; - struct amdgpu_gfx_config config; - struct amdgpu_rlc rlc; - struct amdgpu_mec mec; - struct amdgpu_kiq kiq; - struct amdgpu_scratch scratch; - const struct firmware *me_fw; /* ME firmware */ - uint32_t me_fw_version; - const struct firmware *pfp_fw; /* PFP firmware */ - uint32_t pfp_fw_version; - const struct firmware *ce_fw; /* CE firmware */ - uint32_t ce_fw_version; - const struct firmware *rlc_fw; /* RLC firmware */ - uint32_t rlc_fw_version; - const struct firmware *mec_fw; /* MEC firmware */ - uint32_t mec_fw_version; - const struct firmware *mec2_fw; /* MEC2 firmware */ - uint32_t mec2_fw_version; - uint32_t me_feature_version; - uint32_t ce_feature_version; - uint32_t pfp_feature_version; - uint32_t rlc_feature_version; - uint32_t rlc_srlc_fw_version; - uint32_t rlc_srlc_feature_version; - uint32_t rlc_srlg_fw_version; - uint32_t rlc_srlg_feature_version; - uint32_t rlc_srls_fw_version; - uint32_t rlc_srls_feature_version; - uint32_t mec_feature_version; - uint32_t mec2_feature_version; - struct amdgpu_ring gfx_ring[AMDGPU_MAX_GFX_RINGS]; - unsigned num_gfx_rings; - struct amdgpu_ring compute_ring[AMDGPU_MAX_COMPUTE_RINGS]; - unsigned num_compute_rings; - struct amdgpu_irq_src eop_irq; - struct amdgpu_irq_src priv_reg_irq; - struct amdgpu_irq_src priv_inst_irq; - /* gfx status */ - uint32_t gfx_current_status; - /* ce ram size*/ - unsigned ce_ram_size; - struct amdgpu_cu_info cu_info; - const struct amdgpu_gfx_funcs *funcs; - - /* reset mask */ - uint32_t grbm_soft_reset; - uint32_t srbm_soft_reset; - /* s3/s4 mask */ - bool in_suspend; - /* NGG */ - struct amdgpu_ngg ngg; - - /* pipe reservation */ - struct mutex pipe_reserve_mutex; - DECLARE_BITMAP (pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); -}; - int amdgpu_ib_get(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned size, struct amdgpu_ib *ib); void amdgpu_ib_free(struct amdgpu_device *adev, struct amdgpu_ib *ib, @@ -1020,6 +554,7 @@ struct amdgpu_cs_parser { /* scheduler job object */ struct amdgpu_job *job; + struct drm_sched_entity *entity; /* buffer objects */ struct ww_acquire_ctx ticket; @@ -1041,40 +576,6 @@ struct amdgpu_cs_parser { struct drm_syncobj **post_dep_syncobjs; }; -#define AMDGPU_PREAMBLE_IB_PRESENT (1 << 0) /* bit set means command submit involves a preamble IB */ -#define AMDGPU_PREAMBLE_IB_PRESENT_FIRST (1 << 1) /* bit set means preamble IB is first presented in belonging context */ -#define AMDGPU_HAVE_CTX_SWITCH (1 << 2) /* bit set means context switch occured */ - -struct amdgpu_job { - struct drm_sched_job base; - struct amdgpu_device *adev; - struct amdgpu_vm *vm; - struct amdgpu_ring *ring; - struct amdgpu_sync sync; - struct amdgpu_sync sched_sync; - struct amdgpu_ib *ibs; - struct dma_fence *fence; /* the hw fence */ - uint32_t preamble_status; - uint32_t num_ibs; - void *owner; - uint64_t fence_ctx; /* the fence_context this job uses */ - bool vm_needs_flush; - uint64_t vm_pd_addr; - unsigned vmid; - unsigned pasid; - uint32_t gds_base, gds_size; - uint32_t gws_base, gws_size; - uint32_t oa_base, oa_size; - uint32_t vram_lost_counter; - - /* user fence handling */ - uint64_t uf_addr; - uint64_t uf_sequence; - -}; -#define to_amdgpu_job(sched_job) \ - container_of((sched_job), struct amdgpu_job, base) - static inline u32 amdgpu_get_ib_value(struct amdgpu_cs_parser *p, uint32_t ib_idx, int idx) { @@ -1105,58 +606,6 @@ int amdgpu_device_wb_get(struct amdgpu_device *adev, u32 *wb); void amdgpu_device_wb_free(struct amdgpu_device *adev, u32 wb); /* - * SDMA - */ -struct amdgpu_sdma_instance { - /* SDMA firmware */ - const struct firmware *fw; - uint32_t fw_version; - uint32_t feature_version; - - struct amdgpu_ring ring; - bool burst_nop; -}; - -struct amdgpu_sdma { - struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES]; -#ifdef CONFIG_DRM_AMDGPU_SI - //SI DMA has a difference trap irq number for the second engine - struct amdgpu_irq_src trap_irq_1; -#endif - struct amdgpu_irq_src trap_irq; - struct amdgpu_irq_src illegal_inst_irq; - int num_instances; - uint32_t srbm_soft_reset; -}; - -/* - * Firmware - */ -enum amdgpu_firmware_load_type { - AMDGPU_FW_LOAD_DIRECT = 0, - AMDGPU_FW_LOAD_SMU, - AMDGPU_FW_LOAD_PSP, -}; - -struct amdgpu_firmware { - struct amdgpu_firmware_info ucode[AMDGPU_UCODE_ID_MAXIMUM]; - enum amdgpu_firmware_load_type load_type; - struct amdgpu_bo *fw_buf; - unsigned int fw_size; - unsigned int max_ucodes; - /* firmwares are loaded by psp instead of smu from vega10 */ - const struct amdgpu_psp_funcs *funcs; - struct amdgpu_bo *rbuf; - struct mutex mutex; - - /* gpu info firmware data pointer */ - const struct firmware *gpu_info_fw; - - void *fw_buf_ptr; - uint64_t fw_buf_mc; -}; - -/* * Benchmarking */ void amdgpu_benchmark(struct amdgpu_device *adev, int test_number); @@ -1167,31 +616,6 @@ void amdgpu_benchmark(struct amdgpu_device *adev, int test_number); */ void amdgpu_test_moves(struct amdgpu_device *adev); - -/* - * amdgpu smumgr functions - */ -struct amdgpu_smumgr_funcs { - int (*check_fw_load_finish)(struct amdgpu_device *adev, uint32_t fwtype); - int (*request_smu_load_fw)(struct amdgpu_device *adev); - int (*request_smu_specific_fw)(struct amdgpu_device *adev, uint32_t fwtype); -}; - -/* - * amdgpu smumgr - */ -struct amdgpu_smumgr { - struct amdgpu_bo *toc_buf; - struct amdgpu_bo *smu_buf; - /* asic priv smu data */ - void *priv; - spinlock_t smu_lock; - /* smumgr functions */ - const struct amdgpu_smumgr_funcs *smumgr_funcs; - /* ucode loading complete flag */ - uint32_t fw_flags; -}; - /* * ASIC specific register table accessible by UMD */ @@ -1233,23 +657,9 @@ struct amdgpu_asic_funcs { /* * IOCTL. */ -int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); -int amdgpu_gem_info_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); -int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); -int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); -int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); -int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); -int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); int amdgpu_cs_fence_to_handle_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); @@ -1257,9 +667,6 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *fi int amdgpu_cs_wait_fences_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); -int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp); - /* VRAM scratch page for HDP bug, default vram page */ struct amdgpu_vram_scratch { struct amdgpu_bo *robj; @@ -1389,6 +796,7 @@ enum amd_hw_ip_block_type { PWR_HWIP, NBIF_HWIP, THM_HWIP, + CLK_HWIP, MAX_HWIP }; @@ -1543,9 +951,6 @@ struct amdgpu_device { u32 cg_flags; u32 pg_flags; - /* amdgpu smumgr */ - struct amdgpu_smumgr smu; - /* gfx */ struct amdgpu_gfx gfx; @@ -1579,9 +984,9 @@ struct amdgpu_device { DECLARE_HASHTABLE(mn_hash, 7); /* tracking pinned memory */ - u64 vram_pin_size; - u64 invisible_pin_size; - u64 gart_pin_size; + atomic64_t vram_pin_size; + atomic64_t visible_pin_size; + atomic64_t gart_pin_size; /* amdkfd interface */ struct kfd_dev *kfd; @@ -1610,6 +1015,9 @@ struct amdgpu_device { bool has_hw_reset; u8 reset_magic[AMDGPU_RESET_MAGIC_NUM]; + /* s3/s4 mask */ + bool in_suspend; + /* record last mm index being written through WREG32*/ unsigned long last_mm_index; bool in_gpu_reset; @@ -1732,22 +1140,6 @@ int emu_soc_asic_init(struct amdgpu_device *adev); #define RBIOS16(i) (RBIOS8(i) | (RBIOS8((i)+1) << 8)) #define RBIOS32(i) ((RBIOS16(i)) | (RBIOS16((i)+2) << 16)) -static inline struct amdgpu_sdma_instance * -amdgpu_get_sdma_instance(struct amdgpu_ring *ring) -{ - struct amdgpu_device *adev = ring->adev; - int i; - - for (i = 0; i < adev->sdma.num_instances; i++) - if (&adev->sdma.instance[i].ring == ring) - break; - - if (i < AMDGPU_MAX_SDMA_INSTANCES) - return &adev->sdma.instance[i]; - else - return NULL; -} - /* * ASICs macro. */ @@ -1766,75 +1158,16 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_asic_flush_hdp(adev, r) (adev)->asic_funcs->flush_hdp((adev), (r)) #define amdgpu_asic_invalidate_hdp(adev, r) (adev)->asic_funcs->invalidate_hdp((adev), (r)) #define amdgpu_asic_need_full_reset(adev) (adev)->asic_funcs->need_full_reset((adev)) -#define amdgpu_gmc_flush_gpu_tlb(adev, vmid) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid)) -#define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr)) -#define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid)) -#define amdgpu_gmc_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gmc.gmc_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags)) -#define amdgpu_gmc_get_vm_pde(adev, level, dst, flags) (adev)->gmc.gmc_funcs->get_vm_pde((adev), (level), (dst), (flags)) -#define amdgpu_gmc_get_pte_flags(adev, flags) (adev)->gmc.gmc_funcs->get_vm_pte_flags((adev),(flags)) -#define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count))) -#define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr))) -#define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags))) -#define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib))) -#define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r)) -#define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t)) -#define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r)) -#define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r)) -#define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r)) -#define amdgpu_ring_emit_ib(r, ib, vmid, c) (r)->funcs->emit_ib((r), (ib), (vmid), (c)) -#define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r)) -#define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr)) -#define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags)) -#define amdgpu_ring_emit_gds_switch(r, v, db, ds, wb, ws, ab, as) (r)->funcs->emit_gds_switch((r), (v), (db), (ds), (wb), (ws), (ab), (as)) -#define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) -#define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) -#define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d)) -#define amdgpu_ring_emit_rreg(r, d) (r)->funcs->emit_rreg((r), (d)) -#define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v)) -#define amdgpu_ring_emit_reg_wait(r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m)) -#define amdgpu_ring_emit_reg_write_reg_wait(r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) -#define amdgpu_ring_emit_tmz(r, b) (r)->funcs->emit_tmz((r), (b)) -#define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib))) -#define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r)) -#define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o)) -#define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev)) -#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev)) -#define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv)) -#define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev)) -#define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc)) -#define amdgpu_display_backlight_set_level(adev, e, l) (adev)->mode_info.funcs->backlight_set_level((e), (l)) -#define amdgpu_display_backlight_get_level(adev, e) (adev)->mode_info.funcs->backlight_get_level((e)) -#define amdgpu_display_hpd_sense(adev, h) (adev)->mode_info.funcs->hpd_sense((adev), (h)) -#define amdgpu_display_hpd_set_polarity(adev, h) (adev)->mode_info.funcs->hpd_set_polarity((adev), (h)) -#define amdgpu_display_hpd_get_gpio_reg(adev) (adev)->mode_info.funcs->hpd_get_gpio_reg((adev)) -#define amdgpu_display_bandwidth_update(adev) (adev)->mode_info.funcs->bandwidth_update((adev)) -#define amdgpu_display_page_flip(adev, crtc, base, async) (adev)->mode_info.funcs->page_flip((adev), (crtc), (base), (async)) -#define amdgpu_display_page_flip_get_scanoutpos(adev, crtc, vbl, pos) (adev)->mode_info.funcs->page_flip_get_scanoutpos((adev), (crtc), (vbl), (pos)) -#define amdgpu_display_add_encoder(adev, e, s, c) (adev)->mode_info.funcs->add_encoder((adev), (e), (s), (c)) -#define amdgpu_display_add_connector(adev, ci, sd, ct, ib, coi, h, r) (adev)->mode_info.funcs->add_connector((adev), (ci), (sd), (ct), (ib), (coi), (h), (r)) -#define amdgpu_emit_copy_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_copy_buffer((ib), (s), (d), (b)) -#define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b)) -#define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev)) -#define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance)) -#define amdgpu_gds_switch(adev, r, v, d, w, a) (adev)->gds.funcs->patch_gds_switch((r), (v), (d), (w), (a)) -#define amdgpu_psp_check_fw_loading_status(adev, i) (adev)->firmware.funcs->check_fw_loading_status((adev), (i)) -#define amdgpu_gfx_select_me_pipe_q(adev, me, pipe, q) (adev)->gfx.funcs->select_me_pipe_q((adev), (me), (pipe), (q)) /* Common functions */ +bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev); int amdgpu_device_gpu_recover(struct amdgpu_device *adev, - struct amdgpu_job* job, bool force); + struct amdgpu_job* job); void amdgpu_device_pci_config_reset(struct amdgpu_device *adev); bool amdgpu_device_need_post(struct amdgpu_device *adev); -void amdgpu_display_update_priority(struct amdgpu_device *adev); void amdgpu_cs_report_moved_bytes(struct amdgpu_device *adev, u64 num_bytes, u64 num_vis_bytes); -void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain); -bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo); -void amdgpu_device_vram_location(struct amdgpu_device *adev, - struct amdgpu_gmc *mc, u64 base); -void amdgpu_device_gart_location(struct amdgpu_device *adev, - struct amdgpu_gmc *mc); int amdgpu_device_resize_fb_bar(struct amdgpu_device *adev); void amdgpu_device_program_register_sequence(struct amdgpu_device *adev, const u32 *registers, @@ -1885,6 +1218,12 @@ void amdgpu_disable_vblank_kms(struct drm_device *dev, unsigned int pipe); long amdgpu_kms_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); + +/* + * functions used by amdgpu_xgmi.c + */ +int amdgpu_xgmi_add_device(struct amdgpu_device *adev); + /* * functions used by amdgpu_encoder.c */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c index f4c474a95875..297a5490ad8c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c @@ -57,6 +57,10 @@ #define ACP_I2S_COMP2_CAP_REG_OFFSET 0xa8 #define ACP_I2S_COMP1_PLAY_REG_OFFSET 0x6c #define ACP_I2S_COMP2_PLAY_REG_OFFSET 0x68 +#define ACP_BT_PLAY_REGS_START 0x14970 +#define ACP_BT_PLAY_REGS_END 0x14a24 +#define ACP_BT_COMP1_REG_OFFSET 0xac +#define ACP_BT_COMP2_REG_OFFSET 0xa8 #define mmACP_PGFSM_RETAIN_REG 0x51c9 #define mmACP_PGFSM_CONFIG_REG 0x51ca @@ -77,7 +81,7 @@ #define ACP_SOFT_RESET_DONE_TIME_OUT_VALUE 0x000000FF #define ACP_TIMEOUT_LOOP 0x000000FF -#define ACP_DEVS 3 +#define ACP_DEVS 4 #define ACP_SRC_ID 162 enum { @@ -112,136 +116,47 @@ static int acp_sw_fini(void *handle) return 0; } -/* power off a tile/block within ACP */ -static int acp_suspend_tile(void *cgs_dev, int tile) -{ - u32 val = 0; - u32 count = 0; - - if ((tile < ACP_TILE_P1) || (tile > ACP_TILE_DSP2)) { - pr_err("Invalid ACP tile : %d to suspend\n", tile); - return -1; - } - - val = cgs_read_register(cgs_dev, mmACP_PGFSM_READ_REG_0 + tile); - val &= ACP_TILE_ON_MASK; - - if (val == 0x0) { - val = cgs_read_register(cgs_dev, mmACP_PGFSM_RETAIN_REG); - val = val | (1 << tile); - cgs_write_register(cgs_dev, mmACP_PGFSM_RETAIN_REG, val); - cgs_write_register(cgs_dev, mmACP_PGFSM_CONFIG_REG, - 0x500 + tile); - - count = ACP_TIMEOUT_LOOP; - while (true) { - val = cgs_read_register(cgs_dev, mmACP_PGFSM_READ_REG_0 - + tile); - val = val & ACP_TILE_ON_MASK; - if (val == ACP_TILE_OFF_MASK) - break; - if (--count == 0) { - pr_err("Timeout reading ACP PGFSM status\n"); - return -ETIMEDOUT; - } - udelay(100); - } - - val = cgs_read_register(cgs_dev, mmACP_PGFSM_RETAIN_REG); - - val |= ACP_TILE_OFF_RETAIN_REG_MASK; - cgs_write_register(cgs_dev, mmACP_PGFSM_RETAIN_REG, val); - } - return 0; -} - -/* power on a tile/block within ACP */ -static int acp_resume_tile(void *cgs_dev, int tile) -{ - u32 val = 0; - u32 count = 0; - - if ((tile < ACP_TILE_P1) || (tile > ACP_TILE_DSP2)) { - pr_err("Invalid ACP tile to resume\n"); - return -1; - } - - val = cgs_read_register(cgs_dev, mmACP_PGFSM_READ_REG_0 + tile); - val = val & ACP_TILE_ON_MASK; - - if (val != 0x0) { - cgs_write_register(cgs_dev, mmACP_PGFSM_CONFIG_REG, - 0x600 + tile); - count = ACP_TIMEOUT_LOOP; - while (true) { - val = cgs_read_register(cgs_dev, mmACP_PGFSM_READ_REG_0 - + tile); - val = val & ACP_TILE_ON_MASK; - if (val == 0x0) - break; - if (--count == 0) { - pr_err("Timeout reading ACP PGFSM status\n"); - return -ETIMEDOUT; - } - udelay(100); - } - val = cgs_read_register(cgs_dev, mmACP_PGFSM_RETAIN_REG); - if (tile == ACP_TILE_P1) - val = val & (ACP_TILE_P1_MASK); - else if (tile == ACP_TILE_P2) - val = val & (ACP_TILE_P2_MASK); - - cgs_write_register(cgs_dev, mmACP_PGFSM_RETAIN_REG, val); - } - return 0; -} - struct acp_pm_domain { - void *cgs_dev; + void *adev; struct generic_pm_domain gpd; }; static int acp_poweroff(struct generic_pm_domain *genpd) { - int i, ret; struct acp_pm_domain *apd; + struct amdgpu_device *adev; apd = container_of(genpd, struct acp_pm_domain, gpd); if (apd != NULL) { - /* Donot return abruptly if any of power tile fails to suspend. - * Log it and continue powering off other tile - */ - for (i = 4; i >= 0 ; i--) { - ret = acp_suspend_tile(apd->cgs_dev, ACP_TILE_P1 + i); - if (ret) - pr_err("ACP tile %d tile suspend failed\n", i); - } + adev = apd->adev; + /* call smu to POWER GATE ACP block + * smu will + * 1. turn off the acp clock + * 2. power off the acp tiles + * 3. check and enter ulv state + */ + if (adev->powerplay.pp_funcs->set_powergating_by_smu) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true); } return 0; } static int acp_poweron(struct generic_pm_domain *genpd) { - int i, ret; struct acp_pm_domain *apd; + struct amdgpu_device *adev; apd = container_of(genpd, struct acp_pm_domain, gpd); if (apd != NULL) { - for (i = 0; i < 2; i++) { - ret = acp_resume_tile(apd->cgs_dev, ACP_TILE_P1 + i); - if (ret) { - pr_err("ACP tile %d resume failed\n", i); - break; - } - } - - /* Disable DSPs which are not going to be used */ - for (i = 0; i < 3; i++) { - ret = acp_suspend_tile(apd->cgs_dev, ACP_TILE_DSP0 + i); - /* Continue suspending other DSP, even if one fails */ - if (ret) - pr_err("ACP DSP %d suspend failed\n", i); - } + adev = apd->adev; + /* call smu to UNGATE ACP block + * smu will + * 1. exit ulv + * 2. turn on acp clock + * 3. power on acp tiles + */ + if (adev->powerplay.pp_funcs->set_powergating_by_smu) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false); } return 0; } @@ -285,30 +200,31 @@ static int acp_hw_init(void *handle) r = amd_acp_hw_init(adev->acp.cgs_device, ip_block->version->major, ip_block->version->minor); /* -ENODEV means board uses AZ rather than ACP */ - if (r == -ENODEV) + if (r == -ENODEV) { + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true); return 0; - else if (r) + } else if (r) { return r; + } if (adev->rmmio_size == 0 || adev->rmmio_size < 0x5289) return -EINVAL; acp_base = adev->rmmio_base; - if (adev->asic_type != CHIP_STONEY) { - adev->acp.acp_genpd = kzalloc(sizeof(struct acp_pm_domain), GFP_KERNEL); - if (adev->acp.acp_genpd == NULL) - return -ENOMEM; - adev->acp.acp_genpd->gpd.name = "ACP_AUDIO"; - adev->acp.acp_genpd->gpd.power_off = acp_poweroff; - adev->acp.acp_genpd->gpd.power_on = acp_poweron; + adev->acp.acp_genpd = kzalloc(sizeof(struct acp_pm_domain), GFP_KERNEL); + if (adev->acp.acp_genpd == NULL) + return -ENOMEM; + adev->acp.acp_genpd->gpd.name = "ACP_AUDIO"; + adev->acp.acp_genpd->gpd.power_off = acp_poweroff; + adev->acp.acp_genpd->gpd.power_on = acp_poweron; - adev->acp.acp_genpd->cgs_dev = adev->acp.cgs_device; - pm_genpd_init(&adev->acp.acp_genpd->gpd, NULL, false); - } + adev->acp.acp_genpd->adev = adev; + + pm_genpd_init(&adev->acp.acp_genpd->gpd, NULL, false); adev->acp.acp_cell = kcalloc(ACP_DEVS, sizeof(struct mfd_cell), GFP_KERNEL); @@ -316,14 +232,13 @@ static int acp_hw_init(void *handle) if (adev->acp.acp_cell == NULL) return -ENOMEM; - adev->acp.acp_res = kcalloc(4, sizeof(struct resource), GFP_KERNEL); - + adev->acp.acp_res = kcalloc(5, sizeof(struct resource), GFP_KERNEL); if (adev->acp.acp_res == NULL) { kfree(adev->acp.acp_cell); return -ENOMEM; } - i2s_pdata = kcalloc(2, sizeof(struct i2s_platform_data), GFP_KERNEL); + i2s_pdata = kcalloc(3, sizeof(struct i2s_platform_data), GFP_KERNEL); if (i2s_pdata == NULL) { kfree(adev->acp.acp_res); kfree(adev->acp.acp_cell); @@ -358,6 +273,20 @@ static int acp_hw_init(void *handle) i2s_pdata[1].i2s_reg_comp1 = ACP_I2S_COMP1_CAP_REG_OFFSET; i2s_pdata[1].i2s_reg_comp2 = ACP_I2S_COMP2_CAP_REG_OFFSET; + i2s_pdata[2].quirks = DW_I2S_QUIRK_COMP_REG_OFFSET; + switch (adev->asic_type) { + case CHIP_STONEY: + i2s_pdata[2].quirks |= DW_I2S_QUIRK_16BIT_IDX_OVERRIDE; + break; + default: + break; + } + + i2s_pdata[2].cap = DWC_I2S_PLAY | DWC_I2S_RECORD; + i2s_pdata[2].snd_rates = SNDRV_PCM_RATE_8000_96000; + i2s_pdata[2].i2s_reg_comp1 = ACP_BT_COMP1_REG_OFFSET; + i2s_pdata[2].i2s_reg_comp2 = ACP_BT_COMP2_REG_OFFSET; + adev->acp.acp_res[0].name = "acp2x_dma"; adev->acp.acp_res[0].flags = IORESOURCE_MEM; adev->acp.acp_res[0].start = acp_base; @@ -373,13 +302,18 @@ static int acp_hw_init(void *handle) adev->acp.acp_res[2].start = acp_base + ACP_I2S_CAP_REGS_START; adev->acp.acp_res[2].end = acp_base + ACP_I2S_CAP_REGS_END; - adev->acp.acp_res[3].name = "acp2x_dma_irq"; - adev->acp.acp_res[3].flags = IORESOURCE_IRQ; - adev->acp.acp_res[3].start = amdgpu_irq_create_mapping(adev, 162); - adev->acp.acp_res[3].end = adev->acp.acp_res[3].start; + adev->acp.acp_res[3].name = "acp2x_dw_bt_i2s_play_cap"; + adev->acp.acp_res[3].flags = IORESOURCE_MEM; + adev->acp.acp_res[3].start = acp_base + ACP_BT_PLAY_REGS_START; + adev->acp.acp_res[3].end = acp_base + ACP_BT_PLAY_REGS_END; + + adev->acp.acp_res[4].name = "acp2x_dma_irq"; + adev->acp.acp_res[4].flags = IORESOURCE_IRQ; + adev->acp.acp_res[4].start = amdgpu_irq_create_mapping(adev, 162); + adev->acp.acp_res[4].end = adev->acp.acp_res[4].start; adev->acp.acp_cell[0].name = "acp_audio_dma"; - adev->acp.acp_cell[0].num_resources = 4; + adev->acp.acp_cell[0].num_resources = 5; adev->acp.acp_cell[0].resources = &adev->acp.acp_res[0]; adev->acp.acp_cell[0].platform_data = &adev->asic_type; adev->acp.acp_cell[0].pdata_size = sizeof(adev->asic_type); @@ -396,22 +330,27 @@ static int acp_hw_init(void *handle) adev->acp.acp_cell[2].platform_data = &i2s_pdata[1]; adev->acp.acp_cell[2].pdata_size = sizeof(struct i2s_platform_data); + adev->acp.acp_cell[3].name = "designware-i2s"; + adev->acp.acp_cell[3].num_resources = 1; + adev->acp.acp_cell[3].resources = &adev->acp.acp_res[3]; + adev->acp.acp_cell[3].platform_data = &i2s_pdata[2]; + adev->acp.acp_cell[3].pdata_size = sizeof(struct i2s_platform_data); + r = mfd_add_hotplug_devices(adev->acp.parent, adev->acp.acp_cell, ACP_DEVS); if (r) return r; - if (adev->asic_type != CHIP_STONEY) { - for (i = 0; i < ACP_DEVS ; i++) { - dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); - r = pm_genpd_add_device(&adev->acp.acp_genpd->gpd, dev); - if (r) { - dev_err(dev, "Failed to add dev to genpd\n"); - return r; - } + for (i = 0; i < ACP_DEVS ; i++) { + dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); + r = pm_genpd_add_device(&adev->acp.acp_genpd->gpd, dev); + if (r) { + dev_err(dev, "Failed to add dev to genpd\n"); + return r; } } + /* Assert Soft reset of ACP */ val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); @@ -451,7 +390,6 @@ static int acp_hw_init(void *handle) val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); val &= ~ACP_SOFT_RESET__SoftResetAud_MASK; cgs_write_register(adev->acp.cgs_device, mmACP_SOFT_RESET, val); - return 0; } @@ -470,8 +408,10 @@ static int acp_hw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* return early if no ACP */ - if (!adev->acp.acp_cell) + if (!adev->acp.acp_genpd) { + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false); return 0; + } /* Assert Soft reset of ACP */ val = cgs_read_register(adev->acp.cgs_device, mmACP_SOFT_RESET); @@ -509,19 +449,17 @@ static int acp_hw_fini(void *handle) udelay(100); } - if (adev->acp.acp_genpd) { - for (i = 0; i < ACP_DEVS ; i++) { - dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); - ret = pm_genpd_remove_device(dev); - /* If removal fails, dont giveup and try rest */ - if (ret) - dev_err(dev, "remove dev from genpd failed\n"); - } - kfree(adev->acp.acp_genpd); + for (i = 0; i < ACP_DEVS ; i++) { + dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); + ret = pm_genpd_remove_device(dev); + /* If removal fails, dont giveup and try rest */ + if (ret) + dev_err(dev, "remove dev from genpd failed\n"); } mfd_remove_devices(adev->acp.parent); kfree(adev->acp.acp_res); + kfree(adev->acp.acp_genpd); kfree(adev->acp.acp_cell); return 0; @@ -529,11 +467,21 @@ static int acp_hw_fini(void *handle) static int acp_suspend(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* power up on suspend */ + if (!adev->acp.acp_cell) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, false); return 0; } static int acp_resume(void *handle) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + /* power down again on resume */ + if (!adev->acp.acp_cell) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, true); return 0; } @@ -566,6 +514,12 @@ static int acp_set_clockgating_state(void *handle, static int acp_set_powergating_state(void *handle, enum amd_powergating_state state) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool enable = state == AMD_PG_STATE_GATE ? true : false; + + if (adev->powerplay.pp_funcs->set_powergating_by_smu) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_ACP, enable); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c index 0d8c3fc6eace..7f0afc526419 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acpi.c @@ -31,6 +31,7 @@ #include <drm/drm_crtc_helper.h> #include "amdgpu.h" #include "amdgpu_pm.h" +#include "amdgpu_display.h" #include "amd_acpi.h" #include "atom.h" @@ -358,13 +359,14 @@ out: * * Checks the acpi event and if it matches an atif event, * handles it. - * Returns NOTIFY code + * + * Returns: + * NOTIFY_BAD or NOTIFY_DONE, depending on the event. */ static int amdgpu_atif_handler(struct amdgpu_device *adev, struct acpi_bus_event *event) { struct amdgpu_atif *atif = adev->atif; - struct atif_sbios_requests req; int count; DRM_DEBUG_DRIVER("event, device_class = %s, type = %#x\n", @@ -373,48 +375,59 @@ static int amdgpu_atif_handler(struct amdgpu_device *adev, if (strcmp(event->device_class, ACPI_VIDEO_CLASS) != 0) return NOTIFY_DONE; + /* Is this actually our event? */ if (!atif || !atif->notification_cfg.enabled || - event->type != atif->notification_cfg.command_code) - /* Not our event */ - return NOTIFY_DONE; + event->type != atif->notification_cfg.command_code) { + /* These events will generate keypresses otherwise */ + if (event->type == ACPI_VIDEO_NOTIFY_PROBE) + return NOTIFY_BAD; + else + return NOTIFY_DONE; + } - /* Check pending SBIOS requests */ - count = amdgpu_atif_get_sbios_requests(atif, &req); + if (atif->functions.sbios_requests) { + struct atif_sbios_requests req; - if (count <= 0) - return NOTIFY_DONE; + /* Check pending SBIOS requests */ + count = amdgpu_atif_get_sbios_requests(atif, &req); - DRM_DEBUG_DRIVER("ATIF: %d pending SBIOS requests\n", count); + if (count <= 0) + return NOTIFY_BAD; - if (req.pending & ATIF_PANEL_BRIGHTNESS_CHANGE_REQUEST) { - struct amdgpu_encoder *enc = atif->encoder_for_bl; + DRM_DEBUG_DRIVER("ATIF: %d pending SBIOS requests\n", count); - if (enc) { - struct amdgpu_encoder_atom_dig *dig = enc->enc_priv; + /* todo: add DC handling */ + if ((req.pending & ATIF_PANEL_BRIGHTNESS_CHANGE_REQUEST) && + !amdgpu_device_has_dc_support(adev)) { + struct amdgpu_encoder *enc = atif->encoder_for_bl; - DRM_DEBUG_DRIVER("Changing brightness to %d\n", - req.backlight_level); + if (enc) { + struct amdgpu_encoder_atom_dig *dig = enc->enc_priv; + + DRM_DEBUG_DRIVER("Changing brightness to %d\n", + req.backlight_level); - amdgpu_display_backlight_set_level(adev, enc, req.backlight_level); + amdgpu_display_backlight_set_level(adev, enc, req.backlight_level); #if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) || defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) - backlight_force_update(dig->bl_dev, - BACKLIGHT_UPDATE_HOTKEY); + backlight_force_update(dig->bl_dev, + BACKLIGHT_UPDATE_HOTKEY); #endif + } } - } - if (req.pending & ATIF_DGPU_DISPLAY_EVENT) { - if ((adev->flags & AMD_IS_PX) && - amdgpu_atpx_dgpu_req_power_for_displays()) { - pm_runtime_get_sync(adev->ddev->dev); - /* Just fire off a uevent and let userspace tell us what to do */ - drm_helper_hpd_irq_event(adev->ddev); - pm_runtime_mark_last_busy(adev->ddev->dev); - pm_runtime_put_autosuspend(adev->ddev->dev); + if (req.pending & ATIF_DGPU_DISPLAY_EVENT) { + if ((adev->flags & AMD_IS_PX) && + amdgpu_atpx_dgpu_req_power_for_displays()) { + pm_runtime_get_sync(adev->ddev->dev); + /* Just fire off a uevent and let userspace tell us what to do */ + drm_helper_hpd_irq_event(adev->ddev); + pm_runtime_mark_last_busy(adev->ddev->dev); + pm_runtime_put_autosuspend(adev->ddev->dev); + } } + /* TODO: check other events */ } - /* TODO: check other events */ /* We've handled the event, stop the notifier chain. The ACPI interface * overloads ACPI_VIDEO_NOTIFY_PROBE, we don't want to send that to diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 305143fcc1ce..c31a8849e9f8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -28,7 +28,6 @@ #include <linux/module.h> const struct kgd2kfd_calls *kgd2kfd; -bool (*kgd2kfd_init_p)(unsigned int, const struct kgd2kfd_calls**); static const unsigned int compute_vmid_bitmap = 0xFF00; @@ -36,45 +35,23 @@ int amdgpu_amdkfd_init(void) { int ret; -#if defined(CONFIG_HSA_AMD_MODULE) - int (*kgd2kfd_init_p)(unsigned int, const struct kgd2kfd_calls**); - - kgd2kfd_init_p = symbol_request(kgd2kfd_init); - - if (kgd2kfd_init_p == NULL) - return -ENOENT; - - ret = kgd2kfd_init_p(KFD_INTERFACE_VERSION, &kgd2kfd); - if (ret) { - symbol_put(kgd2kfd_init); - kgd2kfd = NULL; - } - - -#elif defined(CONFIG_HSA_AMD) - +#ifdef CONFIG_HSA_AMD ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd); if (ret) kgd2kfd = NULL; - + amdgpu_amdkfd_gpuvm_init_mem_limits(); #else kgd2kfd = NULL; ret = -ENOENT; #endif -#if defined(CONFIG_HSA_AMD_MODULE) || defined(CONFIG_HSA_AMD) - amdgpu_amdkfd_gpuvm_init_mem_limits(); -#endif - return ret; } void amdgpu_amdkfd_fini(void) { - if (kgd2kfd) { + if (kgd2kfd) kgd2kfd->exit(); - symbol_put(kgd2kfd_init); - } } void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) @@ -99,6 +76,7 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions(); break; case CHIP_VEGA10: + case CHIP_VEGA20: case CHIP_RAVEN: kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions(); break; @@ -146,7 +124,7 @@ static void amdgpu_doorbell_get_kfd_info(struct amdgpu_device *adev, void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) { - int i; + int i, n; int last_valid_bit; if (adev->kfd) { struct kgd2kfd_shared_resources gpu_resources = { @@ -155,7 +133,7 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) .num_queue_per_pipe = adev->gfx.mec.num_queue_per_pipe, .gpuvm_size = min(adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT, - AMDGPU_VA_HOLE_START), + AMDGPU_GMC_HOLE_START), .drm_render_minor = adev->ddev->render->index }; @@ -185,7 +163,15 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) &gpu_resources.doorbell_physical_address, &gpu_resources.doorbell_aperture_size, &gpu_resources.doorbell_start_offset); - if (adev->asic_type >= CHIP_VEGA10) { + + if (adev->asic_type < CHIP_VEGA10) { + kgd2kfd->device_init(adev->kfd, &gpu_resources); + return; + } + + n = (adev->asic_type < CHIP_VEGA20) ? 2 : 8; + + for (i = 0; i < n; i += 2) { /* On SOC15 the BIF is involved in routing * doorbells using the low 12 bits of the * address. Communicate the assignments to @@ -193,20 +179,31 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) * process in case of 64-bit doorbells so we * can use each doorbell assignment twice. */ - gpu_resources.sdma_doorbell[0][0] = - AMDGPU_DOORBELL64_sDMA_ENGINE0; - gpu_resources.sdma_doorbell[0][1] = - AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200; - gpu_resources.sdma_doorbell[1][0] = - AMDGPU_DOORBELL64_sDMA_ENGINE1; - gpu_resources.sdma_doorbell[1][1] = - AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200; - /* Doorbells 0x0f0-0ff and 0x2f0-2ff are reserved for - * SDMA, IH and VCN. So don't use them for the CP. - */ - gpu_resources.reserved_doorbell_mask = 0x1f0; - gpu_resources.reserved_doorbell_val = 0x0f0; + if (adev->asic_type == CHIP_VEGA10) { + gpu_resources.sdma_doorbell[0][i] = + AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 + (i >> 1); + gpu_resources.sdma_doorbell[0][i+1] = + AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 + 0x200 + (i >> 1); + gpu_resources.sdma_doorbell[1][i] = + AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 + (i >> 1); + gpu_resources.sdma_doorbell[1][i+1] = + AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 + 0x200 + (i >> 1); + } else { + gpu_resources.sdma_doorbell[0][i] = + AMDGPU_DOORBELL64_sDMA_ENGINE0 + (i >> 1); + gpu_resources.sdma_doorbell[0][i+1] = + AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200 + (i >> 1); + gpu_resources.sdma_doorbell[1][i] = + AMDGPU_DOORBELL64_sDMA_ENGINE1 + (i >> 1); + gpu_resources.sdma_doorbell[1][i+1] = + AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200 + (i >> 1); + } } + /* Doorbells 0x0e0-0ff and 0x2e0-2ff are reserved for + * SDMA, IH and VCN. So don't use them for the CP. + */ + gpu_resources.reserved_doorbell_mask = 0x1e0; + gpu_resources.reserved_doorbell_val = 0x0e0; kgd2kfd->device_init(adev->kfd, &gpu_resources); } @@ -243,15 +240,42 @@ int amdgpu_amdkfd_resume(struct amdgpu_device *adev) return r; } +int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev) +{ + int r = 0; + + if (adev->kfd) + r = kgd2kfd->pre_reset(adev->kfd); + + return r; +} + +int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev) +{ + int r = 0; + + if (adev->kfd) + r = kgd2kfd->post_reset(adev->kfd); + + return r; +} + +void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + if (amdgpu_device_should_recover_gpu(adev)) + amdgpu_device_gpu_recover(adev, NULL); +} + int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, void **mem_obj, uint64_t *gpu_addr, - void **cpu_ptr) + void **cpu_ptr, bool mqd_gfx9) { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; struct amdgpu_bo *bo = NULL; struct amdgpu_bo_param bp; int r; - uint64_t gpu_addr_tmp = 0; void *cpu_ptr_tmp = NULL; memset(&bp, 0, sizeof(bp)); @@ -261,6 +285,10 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; bp.type = ttm_bo_type_kernel; bp.resv = NULL; + + if (mqd_gfx9) + bp.flags |= AMDGPU_GEM_CREATE_MQD_GFX9; + r = amdgpu_bo_create(adev, &bp, &bo); if (r) { dev_err(adev->dev, @@ -275,13 +303,18 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, goto allocate_mem_reserve_bo_failed; } - r = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT, - &gpu_addr_tmp); + r = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT); if (r) { dev_err(adev->dev, "(%d) failed to pin bo for amdkfd\n", r); goto allocate_mem_pin_bo_failed; } + r = amdgpu_ttm_alloc_gart(&bo->tbo); + if (r) { + dev_err(adev->dev, "%p bind failed\n", bo); + goto allocate_mem_kmap_bo_failed; + } + r = amdgpu_bo_kmap(bo, &cpu_ptr_tmp); if (r) { dev_err(adev->dev, @@ -290,7 +323,7 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, } *mem_obj = bo; - *gpu_addr = gpu_addr_tmp; + *gpu_addr = amdgpu_bo_gpu_offset(bo); *cpu_ptr = cpu_ptr_tmp; amdgpu_bo_unreserve(bo); @@ -402,6 +435,13 @@ uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd) return amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); } +uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + return adev->gmc.xgmi.hive_id; +} + int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine, uint32_t vmid, uint64_t gpu_addr, uint32_t *ib_cmd, uint32_t ib_len) @@ -457,6 +497,14 @@ err: return ret; } +void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + amdgpu_dpm_switch_power_profile(adev, + PP_SMC_POWER_PROFILE_COMPUTE, !idle); +} + bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid) { if (adev->kfd) { @@ -467,7 +515,7 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid) return false; } -#if !defined(CONFIG_HSA_AMD_MODULE) && !defined(CONFIG_HSA_AMD) +#ifndef CONFIG_HSA_AMD bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) { return false; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index a8418a3f4e9d..8e0d4f7196b4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -119,6 +119,7 @@ int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm); int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine, uint32_t vmid, uint64_t gpu_addr, uint32_t *ib_cmd, uint32_t ib_len); +void amdgpu_amdkfd_set_compute_idle(struct kgd_dev *kgd, bool idle); struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void); struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void); @@ -126,10 +127,16 @@ struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void); bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid); +int amdgpu_amdkfd_pre_reset(struct amdgpu_device *adev); + +int amdgpu_amdkfd_post_reset(struct amdgpu_device *adev); + +void amdgpu_amdkfd_gpu_reset(struct kgd_dev *kgd); + /* Shared API */ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, void **mem_obj, uint64_t *gpu_addr, - void **cpu_ptr); + void **cpu_ptr, bool mqd_gfx9); void free_gtt_mem(struct kgd_dev *kgd, void *mem_obj); void get_local_mem_info(struct kgd_dev *kgd, struct kfd_local_mem_info *mem_info); @@ -138,6 +145,7 @@ uint64_t get_gpu_clock_counter(struct kgd_dev *kgd); uint32_t get_max_engine_clock_in_mhz(struct kgd_dev *kgd); void get_cu_info(struct kgd_dev *kgd, struct kfd_cu_info *cu_info); uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd); +uint64_t amdgpu_amdkfd_get_hive_id(struct kgd_dev *kgd); #define read_user_wptr(mmptr, wptr, dst) \ ({ \ @@ -155,17 +163,18 @@ uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd); }) /* GPUVM API */ -int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm, - void **process_info, +int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, unsigned int pasid, + void **vm, void **process_info, struct dma_fence **ef); int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, - struct file *filp, + struct file *filp, unsigned int pasid, void **vm, void **process_info, struct dma_fence **ef); void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm); -uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm); +void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm); +uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm); int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( struct kgd_dev *kgd, uint64_t va, uint64_t size, void *vm, struct kgd_mem **mem, @@ -183,6 +192,9 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, int amdgpu_amdkfd_gpuvm_restore_process_bos(void *process_info, struct dma_fence **ef); +int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, + struct kfd_vm_fault_info *info); + void amdgpu_amdkfd_gpuvm_init_mem_limits(void); void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c index 2c14025e5e76..574c1181ae9a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_fence.c @@ -173,7 +173,5 @@ static const struct dma_fence_ops amdkfd_fence_ops = { .get_driver_name = amdkfd_fence_get_driver_name, .get_timeline_name = amdkfd_fence_get_timeline_name, .enable_signaling = amdkfd_fence_enable_signaling, - .signaled = NULL, - .wait = dma_fence_default_wait, .release = amdkfd_fence_release, }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index ea79908dac4c..244d9834a381 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -142,9 +142,10 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type); static void set_scratch_backing_va(struct kgd_dev *kgd, uint64_t va, uint32_t vmid); static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint32_t page_table_base); + uint64_t page_table_base); static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid); static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid); +static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd); /* Because of REG_GET_FIELD() being used, we put this function in the * asic specific file. @@ -204,6 +205,7 @@ static const struct kfd2kgd_calls kfd2kgd = { .create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm, .acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm, .destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm, + .release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm, .get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir, .set_vm_context_page_table_base = set_vm_context_page_table_base, .alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu, @@ -216,6 +218,10 @@ static const struct kfd2kgd_calls kfd2kgd = { .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, .submit_ib = amdgpu_amdkfd_submit_ib, + .get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info, + .read_vmid_from_vmfault_reg = read_vmid_from_vmfault_reg, + .gpu_recover = amdgpu_amdkfd_gpu_reset, + .set_compute_idle = amdgpu_amdkfd_set_compute_idle }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void) @@ -571,6 +577,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, unsigned long flags, end_jiffies; int retry; + if (adev->in_gpu_reset) + return -EIO; + acquire_queue(kgd, pipe_id, queue_id); WREG32(mmCP_HQD_PQ_DOORBELL_CONTROL, 0); @@ -677,7 +686,7 @@ static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, while (true) { temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); - if (temp & SDMA0_STATUS_REG__RB_CMD_IDLE__SHIFT) + if (temp & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK) break; if (time_after(jiffies, end_jiffies)) return -ETIME; @@ -865,7 +874,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) } static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint32_t page_table_base) + uint64_t page_table_base) { struct amdgpu_device *adev = get_amdgpu_device(kgd); @@ -873,7 +882,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, pr_err("trying to set page table base for wrong VMID\n"); return; } - WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, page_table_base); + WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, + lower_32_bits(page_table_base)); } static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) @@ -882,6 +892,9 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) int vmid; unsigned int tmp; + if (adev->in_gpu_reset) + return -EIO; + for (vmid = 0; vmid < 16; vmid++) { if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) continue; @@ -911,3 +924,19 @@ static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid) RREG32(mmVM_INVALIDATE_RESPONSE); return 0; } + + /** + * read_vmid_from_vmfault_reg - read vmid from register + * + * adev: amdgpu_device pointer + * @vmid: vmid pointer + * read vmid from register (CIK). + */ +static uint32_t read_vmid_from_vmfault_reg(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + uint32_t status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS); + + return REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, VMID); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 19dd665e7307..9f149914ad6c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -45,8 +45,6 @@ enum hqd_dequeue_request_type { RESET_WAVES }; -struct vi_sdma_mqd; - /* * Register access functions */ @@ -100,7 +98,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type); static void set_scratch_backing_va(struct kgd_dev *kgd, uint64_t va, uint32_t vmid); static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint32_t page_table_base); + uint64_t page_table_base); static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid); static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid); @@ -164,6 +162,7 @@ static const struct kfd2kgd_calls kfd2kgd = { .create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm, .acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm, .destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm, + .release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm, .get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir, .set_vm_context_page_table_base = set_vm_context_page_table_base, .alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu, @@ -176,6 +175,9 @@ static const struct kfd2kgd_calls kfd2kgd = { .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, .submit_ib = amdgpu_amdkfd_submit_ib, + .get_vm_fault_info = amdgpu_amdkfd_gpuvm_get_vm_fault_info, + .gpu_recover = amdgpu_amdkfd_gpu_reset, + .set_compute_idle = amdgpu_amdkfd_set_compute_idle }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void) @@ -278,7 +280,8 @@ static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) lock_srbm(kgd, mec, pipe, 0, 0); - WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK); + WREG32(mmCPC_INT_CNTL, CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | + CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); unlock_srbm(kgd); @@ -568,6 +571,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, int retry; struct vi_mqd *m = get_mqd(mqd); + if (adev->in_gpu_reset) + return -EIO; + acquire_queue(kgd, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) @@ -827,7 +833,7 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) } static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint32_t page_table_base) + uint64_t page_table_base) { struct amdgpu_device *adev = get_amdgpu_device(kgd); @@ -835,7 +841,8 @@ static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, pr_err("trying to set page table base for wrong VMID\n"); return; } - WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, page_table_base); + WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8, + lower_32_bits(page_table_base)); } static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) @@ -844,6 +851,9 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) int vmid; unsigned int tmp; + if (adev->in_gpu_reset) + return -EIO; + for (vmid = 0; vmid < 16; vmid++) { if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) continue; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c index 1db60aa5b7f0..42cb4c4e0929 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -138,7 +138,7 @@ static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd, uint8_t vmid); static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint32_t page_table_base); + uint64_t page_table_base); static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type); static void set_scratch_backing_va(struct kgd_dev *kgd, uint64_t va, uint32_t vmid); @@ -201,6 +201,7 @@ static const struct kfd2kgd_calls kfd2kgd = { .create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm, .acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm, .destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm, + .release_process_vm = amdgpu_amdkfd_gpuvm_release_process_vm, .get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir, .set_vm_context_page_table_base = set_vm_context_page_table_base, .alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu, @@ -213,6 +214,9 @@ static const struct kfd2kgd_calls kfd2kgd = { .invalidate_tlbs = invalidate_tlbs, .invalidate_tlbs_vmid = invalidate_tlbs_vmid, .submit_ib = amdgpu_amdkfd_submit_ib, + .gpu_recover = amdgpu_amdkfd_gpu_reset, + .set_compute_idle = amdgpu_amdkfd_set_compute_idle, + .get_hive_id = amdgpu_amdkfd_get_hive_id, }; struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void) @@ -679,6 +683,9 @@ static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, uint32_t temp; struct v9_mqd *m = get_mqd(mqd); + if (adev->in_gpu_reset) + return -EIO; + acquire_queue(kgd, pipe_id, queue_id); if (m->cp_hqd_vmid == 0) @@ -866,6 +873,9 @@ static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) int vmid; struct amdgpu_ring *ring = &adev->gfx.kiq.ring; + if (adev->in_gpu_reset) + return -EIO; + if (ring->ready) return invalidate_tlbs_with_kiq(adev, pasid); @@ -1003,11 +1013,10 @@ static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) } static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, - uint32_t page_table_base) + uint64_t page_table_base) { struct amdgpu_device *adev = get_amdgpu_device(kgd); - uint64_t base = (uint64_t)page_table_base << PAGE_SHIFT | - AMDGPU_PTE_VALID; + uint64_t base = page_table_base | AMDGPU_PTE_VALID; if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { pr_err("trying to set page table base for wrong VMID %u\n", diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index ff8fd75f7ca5..df0a059565f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -206,11 +206,9 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo, struct amdgpu_amdkfd_fence ***ef_list, unsigned int *ef_count) { - struct reservation_object_list *fobj; - struct reservation_object *resv; - unsigned int i = 0, j = 0, k = 0, shared_count; - unsigned int count = 0; - struct amdgpu_amdkfd_fence **fence_list; + struct reservation_object *resv = bo->tbo.resv; + struct reservation_object_list *old, *new; + unsigned int i, j, k; if (!ef && !ef_list) return -EINVAL; @@ -220,76 +218,67 @@ static int amdgpu_amdkfd_remove_eviction_fence(struct amdgpu_bo *bo, *ef_count = 0; } - resv = bo->tbo.resv; - fobj = reservation_object_get_list(resv); - - if (!fobj) + old = reservation_object_get_list(resv); + if (!old) return 0; - preempt_disable(); - write_seqcount_begin(&resv->seq); + new = kmalloc(offsetof(typeof(*new), shared[old->shared_max]), + GFP_KERNEL); + if (!new) + return -ENOMEM; - /* Go through all the shared fences in the resevation object. If - * ef is specified and it exists in the list, remove it and reduce the - * count. If ef is not specified, then get the count of eviction fences - * present. + /* Go through all the shared fences in the resevation object and sort + * the interesting ones to the end of the list. */ - shared_count = fobj->shared_count; - for (i = 0; i < shared_count; ++i) { + for (i = 0, j = old->shared_count, k = 0; i < old->shared_count; ++i) { struct dma_fence *f; - f = rcu_dereference_protected(fobj->shared[i], + f = rcu_dereference_protected(old->shared[i], reservation_object_held(resv)); - if (ef) { - if (f->context == ef->base.context) { - dma_fence_put(f); - fobj->shared_count--; - } else { - RCU_INIT_POINTER(fobj->shared[j++], f); - } - } else if (to_amdgpu_amdkfd_fence(f)) - count++; + if ((ef && f->context == ef->base.context) || + (!ef && to_amdgpu_amdkfd_fence(f))) + RCU_INIT_POINTER(new->shared[--j], f); + else + RCU_INIT_POINTER(new->shared[k++], f); } - write_seqcount_end(&resv->seq); - preempt_enable(); + new->shared_max = old->shared_max; + new->shared_count = k; - if (ef || !count) - return 0; + if (!ef) { + unsigned int count = old->shared_count - j; - /* Alloc memory for count number of eviction fence pointers. Fill the - * ef_list array and ef_count - */ - fence_list = kcalloc(count, sizeof(struct amdgpu_amdkfd_fence *), - GFP_KERNEL); - if (!fence_list) - return -ENOMEM; - - preempt_disable(); - write_seqcount_begin(&resv->seq); - - j = 0; - for (i = 0; i < shared_count; ++i) { - struct dma_fence *f; - struct amdgpu_amdkfd_fence *efence; - - f = rcu_dereference_protected(fobj->shared[i], - reservation_object_held(resv)); + /* Alloc memory for count number of eviction fence pointers. + * Fill the ef_list array and ef_count + */ + *ef_list = kcalloc(count, sizeof(**ef_list), GFP_KERNEL); + *ef_count = count; - efence = to_amdgpu_amdkfd_fence(f); - if (efence) { - fence_list[k++] = efence; - fobj->shared_count--; - } else { - RCU_INIT_POINTER(fobj->shared[j++], f); + if (!*ef_list) { + kfree(new); + return -ENOMEM; } } + /* Install the new fence list, seqcount provides the barriers */ + preempt_disable(); + write_seqcount_begin(&resv->seq); + RCU_INIT_POINTER(resv->fence, new); write_seqcount_end(&resv->seq); preempt_enable(); - *ef_list = fence_list; - *ef_count = k; + /* Drop the references to the removed fences or move them to ef_list */ + for (i = j, k = 0; i < old->shared_count; ++i) { + struct dma_fence *f; + + f = rcu_dereference_protected(new->shared[i], + reservation_object_held(resv)); + if (!ef) + (*ef_list)[k++] = to_amdgpu_amdkfd_fence(f); + else + dma_fence_put(f); + } + kfree_rcu(old, rcu); return 0; } @@ -334,7 +323,7 @@ static int amdgpu_amdkfd_bo_validate(struct amdgpu_bo *bo, uint32_t domain, "Called with userptr BO")) return -EINVAL; - amdgpu_ttm_placement_from_domain(bo, domain); + amdgpu_bo_placement_from_domain(bo, domain); ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (ret) @@ -375,7 +364,6 @@ static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm) struct amdgpu_bo *pd = vm->root.base.bo; struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev); struct amdgpu_vm_parser param; - uint64_t addr, flags = AMDGPU_PTE_VALID; int ret; param.domain = AMDGPU_GEM_DOMAIN_VRAM; @@ -394,9 +382,7 @@ static int vm_validate_pt_pd_bos(struct amdgpu_vm *vm) return ret; } - addr = amdgpu_bo_gpu_offset(vm->root.base.bo); - amdgpu_gmc_get_vm_pde(adev, -1, &addr, &flags); - vm->pd_phys_addr = addr; + vm->pd_phys_addr = amdgpu_gmc_pd_addr(vm->root.base.bo); if (vm->use_cpu_for_update) { ret = amdgpu_bo_kmap(pd, NULL); @@ -622,7 +608,7 @@ static int init_user_pages(struct kgd_mem *mem, struct mm_struct *mm, pr_err("%s: Failed to reserve BO\n", __func__); goto release_out; } - amdgpu_ttm_placement_from_domain(bo, mem->domain); + amdgpu_bo_placement_from_domain(bo, mem->domain); ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (ret) pr_err("%s: failed to validate BO\n", __func__); @@ -689,7 +675,6 @@ static int reserve_bo_and_vm(struct kgd_mem *mem, if (!ctx->vm_pd) return -ENOMEM; - ctx->kfd_bo.robj = bo; ctx->kfd_bo.priority = 0; ctx->kfd_bo.tv.bo = &bo->tbo; ctx->kfd_bo.tv.shared = true; @@ -754,7 +739,6 @@ static int reserve_bo_and_cond_vms(struct kgd_mem *mem, return -ENOMEM; } - ctx->kfd_bo.robj = bo; ctx->kfd_bo.priority = 0; ctx->kfd_bo.tv.bo = &bo->tbo; ctx->kfd_bo.tv.shared = true; @@ -1014,8 +998,8 @@ create_evict_fence_fail: return ret; } -int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm, - void **process_info, +int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, unsigned int pasid, + void **vm, void **process_info, struct dma_fence **ef) { struct amdgpu_device *adev = get_amdgpu_device(kgd); @@ -1027,7 +1011,7 @@ int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm, return -ENOMEM; /* Initialize AMDGPU part of the VM */ - ret = amdgpu_vm_init(adev, new_vm, AMDGPU_VM_CONTEXT_COMPUTE, 0); + ret = amdgpu_vm_init(adev, new_vm, AMDGPU_VM_CONTEXT_COMPUTE, pasid); if (ret) { pr_err("Failed init vm ret %d\n", ret); goto amdgpu_vm_init_fail; @@ -1050,7 +1034,7 @@ amdgpu_vm_init_fail: } int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, - struct file *filp, + struct file *filp, unsigned int pasid, void **vm, void **process_info, struct dma_fence **ef) { @@ -1065,7 +1049,7 @@ int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, return -EINVAL; /* Convert VM into a compute VM */ - ret = amdgpu_vm_make_compute(adev, avm); + ret = amdgpu_vm_make_compute(adev, avm, pasid); if (ret) return ret; @@ -1128,11 +1112,34 @@ void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm) kfree(vm); } -uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm) +void amdgpu_amdkfd_gpuvm_release_process_vm(struct kgd_dev *kgd, void *vm) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + + if (WARN_ON(!kgd || !vm)) + return; + + pr_debug("Releasing process vm %p\n", vm); + + /* The original pasid of amdgpu vm has already been + * released during making a amdgpu vm to a compute vm + * The current pasid is managed by kfd and will be + * released on kfd process destroy. Set amdgpu pasid + * to 0 to avoid duplicate release. + */ + amdgpu_vm_release_compute(adev, avm); +} + +uint64_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm) { struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + struct amdgpu_bo *pd = avm->root.base.bo; + struct amdgpu_device *adev = amdgpu_ttm_adev(pd->tbo.bdev); - return avm->pd_phys_addr >> AMDGPU_GPU_PAGE_SHIFT; + if (adev->asic_type < CHIP_VEGA10) + return avm->pd_phys_addr >> AMDGPU_GPU_PAGE_SHIFT; + return avm->pd_phys_addr; } int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( @@ -1587,7 +1594,7 @@ int amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel(struct kgd_dev *kgd, goto bo_reserve_failed; } - ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT, NULL); + ret = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT); if (ret) { pr_err("Failed to pin bo. ret %d\n", ret); goto pin_failed; @@ -1621,6 +1628,20 @@ bo_reserve_failed: return ret; } +int amdgpu_amdkfd_gpuvm_get_vm_fault_info(struct kgd_dev *kgd, + struct kfd_vm_fault_info *mem) +{ + struct amdgpu_device *adev; + + adev = (struct amdgpu_device *)kgd; + if (atomic_read(&adev->gmc.vm_fault_info_updated) == 1) { + *mem = *adev->gmc.vm_fault_info; + mb(); + atomic_set(&adev->gmc.vm_fault_info_updated, 0); + } + return 0; +} + /* Evict a userptr BO by stopping the queues if necessary * * Runs in MMU notifier, may be in RECLAIM_FS context. This means it @@ -1680,7 +1701,7 @@ static int update_invalid_user_pages(struct amdkfd_process_info *process_info, if (amdgpu_bo_reserve(bo, true)) return -EAGAIN; - amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); amdgpu_bo_unreserve(bo); if (ret) { @@ -1824,7 +1845,7 @@ static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) if (mem->user_pages[0]) { amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, mem->user_pages); - amdgpu_ttm_placement_from_domain(bo, mem->domain); + amdgpu_bo_placement_from_domain(bo, mem->domain); ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (ret) { pr_err("%s: failed to validate BO\n", __func__); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c index bf872f694f50..e02781b37e73 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atombios.c @@ -29,6 +29,7 @@ #include "amdgpu_atombios.h" #include "amdgpu_atomfirmware.h" #include "amdgpu_i2c.h" +#include "amdgpu_display.h" #include "atom.h" #include "atom-bits.h" diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c index 236915849cfe..b61e1dc61b4c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c @@ -117,6 +117,10 @@ union igp_info { union umc_info { struct atom_umc_info_v3_1 v31; }; + +union vram_info { + struct atom_vram_info_header_v2_3 v23; +}; /* * Return vram width from integrated system info table, if available, * or 0 if not. @@ -174,7 +178,7 @@ static int convert_atom_mem_type_to_vram_type (struct amdgpu_device *adev, case ATOM_DGPU_VRAM_TYPE_GDDR5: vram_type = AMDGPU_VRAM_TYPE_GDDR5; break; - case ATOM_DGPU_VRAM_TYPE_HBM: + case ATOM_DGPU_VRAM_TYPE_HBM2: vram_type = AMDGPU_VRAM_TYPE_HBM; break; default: @@ -195,7 +199,7 @@ int amdgpu_atomfirmware_get_vram_type(struct amdgpu_device *adev) int index; u16 data_offset, size; union igp_info *igp_info; - union umc_info *umc_info; + union vram_info *vram_info; u8 frev, crev; u8 mem_type; @@ -204,7 +208,7 @@ int amdgpu_atomfirmware_get_vram_type(struct amdgpu_device *adev) integratedsysteminfo); else index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, - umc_info); + vram_info); if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, &size, &frev, &crev, &data_offset)) { @@ -219,11 +223,11 @@ int amdgpu_atomfirmware_get_vram_type(struct amdgpu_device *adev) return 0; } } else { - umc_info = (union umc_info *) + vram_info = (union vram_info *) (mode_info->atom_context->bios + data_offset); switch (crev) { - case 1: - mem_type = umc_info->v31.vram_type; + case 3: + mem_type = vram_info->v23.vram_module[0].memory_type; return convert_atom_mem_type_to_vram_type(adev, mem_type); default: return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c index 9ab89371d9e8..a028661d9e20 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c @@ -32,7 +32,7 @@ struct amdgpu_atpx_functions { bool switch_start; bool switch_end; bool disp_connectors_mapping; - bool disp_detetion_ports; + bool disp_detection_ports; }; struct amdgpu_atpx { @@ -162,7 +162,7 @@ static void amdgpu_atpx_parse_functions(struct amdgpu_atpx_functions *f, u32 mas f->switch_start = mask & ATPX_GRAPHICS_DEVICE_SWITCH_START_NOTIFICATION_SUPPORTED; f->switch_end = mask & ATPX_GRAPHICS_DEVICE_SWITCH_END_NOTIFICATION_SUPPORTED; f->disp_connectors_mapping = mask & ATPX_GET_DISPLAY_CONNECTORS_MAPPING_SUPPORTED; - f->disp_detetion_ports = mask & ATPX_GET_DISPLAY_DETECTION_PORTS_SUPPORTED; + f->disp_detection_ports = mask & ATPX_GET_DISPLAY_DETECTION_PORTS_SUPPORTED; } /** @@ -575,6 +575,7 @@ static const struct amdgpu_px_quirk amdgpu_px_quirk_list[] = { { 0x1002, 0x6900, 0x1002, 0x0124, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0x1002, 0x6900, 0x1028, 0x0812, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0x1002, 0x6900, 0x1028, 0x0813, AMDGPU_PX_QUIRK_FORCE_ATPX }, + { 0x1002, 0x6900, 0x1025, 0x125A, AMDGPU_PX_QUIRK_FORCE_ATPX }, { 0, 0, 0, 0, 0 }, }; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 19cfff31f2e1..3079ea8523c5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -95,11 +95,17 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, r = amdgpu_bo_reserve(sobj, false); if (unlikely(r != 0)) goto out_cleanup; - r = amdgpu_bo_pin(sobj, sdomain, &saddr); + r = amdgpu_bo_pin(sobj, sdomain); + if (r) { + amdgpu_bo_unreserve(sobj); + goto out_cleanup; + } + r = amdgpu_ttm_alloc_gart(&sobj->tbo); amdgpu_bo_unreserve(sobj); if (r) { goto out_cleanup; } + saddr = amdgpu_bo_gpu_offset(sobj); bp.domain = ddomain; r = amdgpu_bo_create(adev, &bp, &dobj); if (r) { @@ -108,11 +114,17 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, r = amdgpu_bo_reserve(dobj, false); if (unlikely(r != 0)) goto out_cleanup; - r = amdgpu_bo_pin(dobj, ddomain, &daddr); + r = amdgpu_bo_pin(dobj, ddomain); + if (r) { + amdgpu_bo_unreserve(sobj); + goto out_cleanup; + } + r = amdgpu_ttm_alloc_gart(&dobj->tbo); amdgpu_bo_unreserve(dobj); if (r) { goto out_cleanup; } + daddr = amdgpu_bo_gpu_offset(dobj); if (adev->mman.buffer_funcs) { time = amdgpu_benchmark_do_move(adev, size, saddr, daddr, n); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c index 92be7f6de197..14d2982a47cc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.c @@ -35,92 +35,57 @@ #define AMDGPU_BO_LIST_MAX_PRIORITY 32u #define AMDGPU_BO_LIST_NUM_BUCKETS (AMDGPU_BO_LIST_MAX_PRIORITY + 1) -static int amdgpu_bo_list_set(struct amdgpu_device *adev, - struct drm_file *filp, - struct amdgpu_bo_list *list, - struct drm_amdgpu_bo_list_entry *info, - unsigned num_entries); - -static void amdgpu_bo_list_release_rcu(struct kref *ref) +static void amdgpu_bo_list_free_rcu(struct rcu_head *rcu) { - unsigned i; - struct amdgpu_bo_list *list = container_of(ref, struct amdgpu_bo_list, - refcount); + struct amdgpu_bo_list *list = container_of(rcu, struct amdgpu_bo_list, + rhead); - for (i = 0; i < list->num_entries; ++i) - amdgpu_bo_unref(&list->array[i].robj); - - mutex_destroy(&list->lock); - kvfree(list->array); - kfree_rcu(list, rhead); + kvfree(list); } -static int amdgpu_bo_list_create(struct amdgpu_device *adev, - struct drm_file *filp, - struct drm_amdgpu_bo_list_entry *info, - unsigned num_entries, - int *id) +static void amdgpu_bo_list_free(struct kref *ref) { - int r; - struct amdgpu_fpriv *fpriv = filp->driver_priv; - struct amdgpu_bo_list *list; + struct amdgpu_bo_list *list = container_of(ref, struct amdgpu_bo_list, + refcount); + struct amdgpu_bo_list_entry *e; - list = kzalloc(sizeof(struct amdgpu_bo_list), GFP_KERNEL); - if (!list) - return -ENOMEM; + amdgpu_bo_list_for_each_entry(e, list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); - /* initialize bo list*/ - mutex_init(&list->lock); - kref_init(&list->refcount); - r = amdgpu_bo_list_set(adev, filp, list, info, num_entries); - if (r) { - kfree(list); - return r; + amdgpu_bo_unref(&bo); } - /* idr alloc should be called only after initialization of bo list. */ - mutex_lock(&fpriv->bo_list_lock); - r = idr_alloc(&fpriv->bo_list_handles, list, 1, 0, GFP_KERNEL); - mutex_unlock(&fpriv->bo_list_lock); - if (r < 0) { - amdgpu_bo_list_free(list); - return r; - } - *id = r; - - return 0; + call_rcu(&list->rhead, amdgpu_bo_list_free_rcu); } -static void amdgpu_bo_list_destroy(struct amdgpu_fpriv *fpriv, int id) +int amdgpu_bo_list_create(struct amdgpu_device *adev, struct drm_file *filp, + struct drm_amdgpu_bo_list_entry *info, + unsigned num_entries, struct amdgpu_bo_list **result) { - struct amdgpu_bo_list *list; - - mutex_lock(&fpriv->bo_list_lock); - list = idr_remove(&fpriv->bo_list_handles, id); - mutex_unlock(&fpriv->bo_list_lock); - if (list) - kref_put(&list->refcount, amdgpu_bo_list_release_rcu); -} - -static int amdgpu_bo_list_set(struct amdgpu_device *adev, - struct drm_file *filp, - struct amdgpu_bo_list *list, - struct drm_amdgpu_bo_list_entry *info, - unsigned num_entries) -{ - struct amdgpu_bo_list_entry *array; - struct amdgpu_bo *gds_obj = adev->gds.gds_gfx_bo; - struct amdgpu_bo *gws_obj = adev->gds.gws_gfx_bo; - struct amdgpu_bo *oa_obj = adev->gds.oa_gfx_bo; - unsigned last_entry = 0, first_userptr = num_entries; + struct amdgpu_bo_list_entry *array; + struct amdgpu_bo_list *list; + uint64_t total_size = 0; + size_t size; unsigned i; int r; - unsigned long total_size = 0; - array = kvmalloc_array(num_entries, sizeof(struct amdgpu_bo_list_entry), GFP_KERNEL); - if (!array) + if (num_entries > (SIZE_MAX - sizeof(struct amdgpu_bo_list)) + / sizeof(struct amdgpu_bo_list_entry)) + return -EINVAL; + + size = sizeof(struct amdgpu_bo_list); + size += num_entries * sizeof(struct amdgpu_bo_list_entry); + list = kvmalloc(size, GFP_KERNEL); + if (!list) return -ENOMEM; + + kref_init(&list->refcount); + list->gds_obj = adev->gds.gds_gfx_bo; + list->gws_obj = adev->gds.gws_gfx_bo; + list->oa_obj = adev->gds.oa_gfx_bo; + + array = amdgpu_bo_list_array_entry(list, 0); memset(array, 0, num_entries * sizeof(struct amdgpu_bo_list_entry)); for (i = 0; i < num_entries; ++i) { @@ -150,66 +115,65 @@ static int amdgpu_bo_list_set(struct amdgpu_device *adev, entry = &array[last_entry++]; } - entry->robj = bo; entry->priority = min(info[i].bo_priority, AMDGPU_BO_LIST_MAX_PRIORITY); - entry->tv.bo = &entry->robj->tbo; - entry->tv.shared = !entry->robj->prime_shared_count; - - if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_GDS) - gds_obj = entry->robj; - if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_GWS) - gws_obj = entry->robj; - if (entry->robj->preferred_domains == AMDGPU_GEM_DOMAIN_OA) - oa_obj = entry->robj; - - total_size += amdgpu_bo_size(entry->robj); - trace_amdgpu_bo_list_set(list, entry->robj); + entry->tv.bo = &bo->tbo; + entry->tv.shared = !bo->prime_shared_count; + + if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GDS) + list->gds_obj = bo; + if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GWS) + list->gws_obj = bo; + if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_OA) + list->oa_obj = bo; + + total_size += amdgpu_bo_size(bo); + trace_amdgpu_bo_list_set(list, bo); } - for (i = 0; i < list->num_entries; ++i) - amdgpu_bo_unref(&list->array[i].robj); - - kvfree(list->array); - - list->gds_obj = gds_obj; - list->gws_obj = gws_obj; - list->oa_obj = oa_obj; list->first_userptr = first_userptr; - list->array = array; list->num_entries = num_entries; trace_amdgpu_cs_bo_status(list->num_entries, total_size); + + *result = list; return 0; error_free: - while (i--) - amdgpu_bo_unref(&array[i].robj); - kvfree(array); + while (i--) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(array[i].tv.bo); + + amdgpu_bo_unref(&bo); + } + kvfree(list); return r; + } -struct amdgpu_bo_list * -amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id) +static void amdgpu_bo_list_destroy(struct amdgpu_fpriv *fpriv, int id) { - struct amdgpu_bo_list *result; + struct amdgpu_bo_list *list; + + mutex_lock(&fpriv->bo_list_lock); + list = idr_remove(&fpriv->bo_list_handles, id); + mutex_unlock(&fpriv->bo_list_lock); + if (list) + kref_put(&list->refcount, amdgpu_bo_list_free); +} +int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id, + struct amdgpu_bo_list **result) +{ rcu_read_lock(); - result = idr_find(&fpriv->bo_list_handles, id); + *result = idr_find(&fpriv->bo_list_handles, id); - if (result) { - if (kref_get_unless_zero(&result->refcount)) { - rcu_read_unlock(); - mutex_lock(&result->lock); - } else { - rcu_read_unlock(); - result = NULL; - } - } else { + if (*result && kref_get_unless_zero(&(*result)->refcount)) { rcu_read_unlock(); + return 0; } - return result; + rcu_read_unlock(); + return -ENOENT; } void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, @@ -220,6 +184,7 @@ void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, * concatenated in descending order. */ struct list_head bucket[AMDGPU_BO_LIST_NUM_BUCKETS]; + struct amdgpu_bo_list_entry *e; unsigned i; for (i = 0; i < AMDGPU_BO_LIST_NUM_BUCKETS; i++) @@ -230,14 +195,14 @@ void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, * in the list, the sort mustn't change the ordering of buffers * with the same priority, i.e. it must be stable. */ - for (i = 0; i < list->num_entries; i++) { - unsigned priority = list->array[i].priority; + amdgpu_bo_list_for_each_entry(e, list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + unsigned priority = e->priority; - if (!list->array[i].robj->parent) - list_add_tail(&list->array[i].tv.head, - &bucket[priority]); + if (!bo->parent) + list_add_tail(&e->tv.head, &bucket[priority]); - list->array[i].user_pages = NULL; + e->user_pages = NULL; } /* Connect the sorted buckets in the output list. */ @@ -247,71 +212,82 @@ void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, void amdgpu_bo_list_put(struct amdgpu_bo_list *list) { - mutex_unlock(&list->lock); - kref_put(&list->refcount, amdgpu_bo_list_release_rcu); + kref_put(&list->refcount, amdgpu_bo_list_free); } -void amdgpu_bo_list_free(struct amdgpu_bo_list *list) -{ - unsigned i; - - for (i = 0; i < list->num_entries; ++i) - amdgpu_bo_unref(&list->array[i].robj); - - mutex_destroy(&list->lock); - kvfree(list->array); - kfree(list); -} - -int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, - struct drm_file *filp) +int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in, + struct drm_amdgpu_bo_list_entry **info_param) { + const void __user *uptr = u64_to_user_ptr(in->bo_info_ptr); const uint32_t info_size = sizeof(struct drm_amdgpu_bo_list_entry); - - struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_fpriv *fpriv = filp->driver_priv; - union drm_amdgpu_bo_list *args = data; - uint32_t handle = args->in.list_handle; - const void __user *uptr = u64_to_user_ptr(args->in.bo_info_ptr); - struct drm_amdgpu_bo_list_entry *info; - struct amdgpu_bo_list *list; - int r; - info = kvmalloc_array(args->in.bo_number, - sizeof(struct drm_amdgpu_bo_list_entry), GFP_KERNEL); + info = kvmalloc_array(in->bo_number, info_size, GFP_KERNEL); if (!info) return -ENOMEM; /* copy the handle array from userspace to a kernel buffer */ r = -EFAULT; - if (likely(info_size == args->in.bo_info_size)) { - unsigned long bytes = args->in.bo_number * - args->in.bo_info_size; + if (likely(info_size == in->bo_info_size)) { + unsigned long bytes = in->bo_number * + in->bo_info_size; if (copy_from_user(info, uptr, bytes)) goto error_free; } else { - unsigned long bytes = min(args->in.bo_info_size, info_size); + unsigned long bytes = min(in->bo_info_size, info_size); unsigned i; - memset(info, 0, args->in.bo_number * info_size); - for (i = 0; i < args->in.bo_number; ++i) { + memset(info, 0, in->bo_number * info_size); + for (i = 0; i < in->bo_number; ++i) { if (copy_from_user(&info[i], uptr, bytes)) goto error_free; - uptr += args->in.bo_info_size; + uptr += in->bo_info_size; } } + *info_param = info; + return 0; + +error_free: + kvfree(info); + return r; +} + +int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp) +{ + struct amdgpu_device *adev = dev->dev_private; + struct amdgpu_fpriv *fpriv = filp->driver_priv; + union drm_amdgpu_bo_list *args = data; + uint32_t handle = args->in.list_handle; + struct drm_amdgpu_bo_list_entry *info = NULL; + struct amdgpu_bo_list *list, *old; + int r; + + r = amdgpu_bo_create_list_entry_array(&args->in, &info); + if (r) + goto error_free; + switch (args->in.operation) { case AMDGPU_BO_LIST_OP_CREATE: r = amdgpu_bo_list_create(adev, filp, info, args->in.bo_number, - &handle); + &list); if (r) goto error_free; + + mutex_lock(&fpriv->bo_list_lock); + r = idr_alloc(&fpriv->bo_list_handles, list, 1, 0, GFP_KERNEL); + mutex_unlock(&fpriv->bo_list_lock); + if (r < 0) { + amdgpu_bo_list_put(list); + return r; + } + + handle = r; break; case AMDGPU_BO_LIST_OP_DESTROY: @@ -320,17 +296,22 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, break; case AMDGPU_BO_LIST_OP_UPDATE: - r = -ENOENT; - list = amdgpu_bo_list_get(fpriv, handle); - if (!list) + r = amdgpu_bo_list_create(adev, filp, info, args->in.bo_number, + &list); + if (r) goto error_free; - r = amdgpu_bo_list_set(adev, filp, list, info, - args->in.bo_number); - amdgpu_bo_list_put(list); - if (r) + mutex_lock(&fpriv->bo_list_lock); + old = idr_replace(&fpriv->bo_list_handles, list, handle); + mutex_unlock(&fpriv->bo_list_lock); + + if (IS_ERR(old)) { + amdgpu_bo_list_put(list); + r = PTR_ERR(old); goto error_free; + } + amdgpu_bo_list_put(old); break; default: @@ -345,6 +326,7 @@ int amdgpu_bo_list_ioctl(struct drm_device *dev, void *data, return 0; error_free: - kvfree(info); + if (info) + kvfree(info); return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h new file mode 100644 index 000000000000..7c5f5d1601e6 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_bo_list.h @@ -0,0 +1,84 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __AMDGPU_BO_LIST_H__ +#define __AMDGPU_BO_LIST_H__ + +#include <drm/ttm/ttm_execbuf_util.h> +#include <drm/amdgpu_drm.h> + +struct amdgpu_device; +struct amdgpu_bo; +struct amdgpu_bo_va; +struct amdgpu_fpriv; + +struct amdgpu_bo_list_entry { + struct ttm_validate_buffer tv; + struct amdgpu_bo_va *bo_va; + uint32_t priority; + struct page **user_pages; + int user_invalidated; +}; + +struct amdgpu_bo_list { + struct rcu_head rhead; + struct kref refcount; + struct amdgpu_bo *gds_obj; + struct amdgpu_bo *gws_obj; + struct amdgpu_bo *oa_obj; + unsigned first_userptr; + unsigned num_entries; +}; + +int amdgpu_bo_list_get(struct amdgpu_fpriv *fpriv, int id, + struct amdgpu_bo_list **result); +void amdgpu_bo_list_get_list(struct amdgpu_bo_list *list, + struct list_head *validated); +void amdgpu_bo_list_put(struct amdgpu_bo_list *list); +int amdgpu_bo_create_list_entry_array(struct drm_amdgpu_bo_list_in *in, + struct drm_amdgpu_bo_list_entry **info_param); + +int amdgpu_bo_list_create(struct amdgpu_device *adev, + struct drm_file *filp, + struct drm_amdgpu_bo_list_entry *info, + unsigned num_entries, + struct amdgpu_bo_list **list); + +static inline struct amdgpu_bo_list_entry * +amdgpu_bo_list_array_entry(struct amdgpu_bo_list *list, unsigned index) +{ + struct amdgpu_bo_list_entry *array = (void *)&list[1]; + + return &array[index]; +} + +#define amdgpu_bo_list_for_each_entry(e, list) \ + for (e = amdgpu_bo_list_array_entry(list, 0); \ + e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \ + ++e) + +#define amdgpu_bo_list_for_each_userptr_entry(e, list) \ + for (e = amdgpu_bo_list_array_entry(list, (list)->first_userptr); \ + e != amdgpu_bo_list_array_entry(list, (list)->num_entries); \ + ++e) + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index e950730f1933..8816c697b205 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -314,17 +314,17 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, (adev->pdev->revision == 0x81) || (adev->pdev->device == 0x665f)) { info->is_kicker = true; - strcpy(fw_name, "radeon/bonaire_k_smc.bin"); + strcpy(fw_name, "amdgpu/bonaire_k_smc.bin"); } else { - strcpy(fw_name, "radeon/bonaire_smc.bin"); + strcpy(fw_name, "amdgpu/bonaire_smc.bin"); } break; case CHIP_HAWAII: if (adev->pdev->revision == 0x80) { info->is_kicker = true; - strcpy(fw_name, "radeon/hawaii_k_smc.bin"); + strcpy(fw_name, "amdgpu/hawaii_k_smc.bin"); } else { - strcpy(fw_name, "radeon/hawaii_smc.bin"); + strcpy(fw_name, "amdgpu/hawaii_smc.bin"); } break; case CHIP_TOPAZ: @@ -367,12 +367,14 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, break; case CHIP_POLARIS10: if (type == CGS_UCODE_ID_SMU) { - if ((adev->pdev->device == 0x67df) && - ((adev->pdev->revision == 0xe0) || - (adev->pdev->revision == 0xe3) || - (adev->pdev->revision == 0xe4) || - (adev->pdev->revision == 0xe5) || - (adev->pdev->revision == 0xe7) || + if (((adev->pdev->device == 0x67df) && + ((adev->pdev->revision == 0xe0) || + (adev->pdev->revision == 0xe3) || + (adev->pdev->revision == 0xe4) || + (adev->pdev->revision == 0xe5) || + (adev->pdev->revision == 0xe7) || + (adev->pdev->revision == 0xef))) || + ((adev->pdev->device == 0x6fdf) && (adev->pdev->revision == 0xef))) { info->is_kicker = true; strcpy(fw_name, "amdgpu/polaris10_k_smc.bin"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index 8e66851eb427..69ad6ec0a4f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -34,6 +34,7 @@ #include "atombios_dp.h" #include "amdgpu_connectors.h" #include "amdgpu_i2c.h" +#include "amdgpu_display.h" #include <linux/pm_runtime.h> @@ -212,30 +213,21 @@ static void amdgpu_connector_update_scratch_regs(struct drm_connector *connector, enum drm_connector_status status) { - struct drm_encoder *best_encoder = NULL; - struct drm_encoder *encoder = NULL; + struct drm_encoder *best_encoder; + struct drm_encoder *encoder; const struct drm_connector_helper_funcs *connector_funcs = connector->helper_private; bool connected; int i; best_encoder = connector_funcs->best_encoder(connector); - for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { - if (connector->encoder_ids[i] == 0) - break; - - encoder = drm_encoder_find(connector->dev, NULL, - connector->encoder_ids[i]); - if (!encoder) - continue; - + drm_connector_for_each_possible_encoder(connector, encoder, i) { if ((encoder == best_encoder) && (status == connector_status_connected)) connected = true; else connected = false; amdgpu_atombios_encoder_set_bios_scratch_regs(connector, encoder, connected); - } } @@ -246,17 +238,11 @@ amdgpu_connector_find_encoder(struct drm_connector *connector, struct drm_encoder *encoder; int i; - for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { - if (connector->encoder_ids[i] == 0) - break; - encoder = drm_encoder_find(connector->dev, NULL, - connector->encoder_ids[i]); - if (!encoder) - continue; - + drm_connector_for_each_possible_encoder(connector, encoder, i) { if (encoder->encoder_type == encoder_type) return encoder; } + return NULL; } @@ -349,22 +335,24 @@ static int amdgpu_connector_ddc_get_modes(struct drm_connector *connector) int ret; if (amdgpu_connector->edid) { - drm_mode_connector_update_edid_property(connector, amdgpu_connector->edid); + drm_connector_update_edid_property(connector, amdgpu_connector->edid); ret = drm_add_edid_modes(connector, amdgpu_connector->edid); return ret; } - drm_mode_connector_update_edid_property(connector, NULL); + drm_connector_update_edid_property(connector, NULL); return 0; } static struct drm_encoder * amdgpu_connector_best_single_encoder(struct drm_connector *connector) { - int enc_id = connector->encoder_ids[0]; + struct drm_encoder *encoder; + int i; + + /* pick the first one */ + drm_connector_for_each_possible_encoder(connector, encoder, i) + return encoder; - /* pick the encoder ids */ - if (enc_id) - return drm_encoder_find(connector->dev, NULL, enc_id); return NULL; } @@ -985,9 +973,8 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force) struct drm_device *dev = connector->dev; struct amdgpu_device *adev = dev->dev_private; struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); - struct drm_encoder *encoder = NULL; const struct drm_encoder_helper_funcs *encoder_funcs; - int i, r; + int r; enum drm_connector_status ret = connector_status_disconnected; bool dret = false, broken_edid = false; @@ -1077,14 +1064,10 @@ amdgpu_connector_dvi_detect(struct drm_connector *connector, bool force) /* find analog encoder */ if (amdgpu_connector->dac_load_detect) { - for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { - if (connector->encoder_ids[i] == 0) - break; - - encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); - if (!encoder) - continue; + struct drm_encoder *encoder; + int i; + drm_connector_for_each_possible_encoder(connector, encoder, i) { if (encoder->encoder_type != DRM_MODE_ENCODER_DAC && encoder->encoder_type != DRM_MODE_ENCODER_TVDAC) continue; @@ -1132,18 +1115,11 @@ exit: static struct drm_encoder * amdgpu_connector_dvi_encoder(struct drm_connector *connector) { - int enc_id = connector->encoder_ids[0]; struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); struct drm_encoder *encoder; int i; - for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { - if (connector->encoder_ids[i] == 0) - break; - - encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); - if (!encoder) - continue; + drm_connector_for_each_possible_encoder(connector, encoder, i) { if (amdgpu_connector->use_digital == true) { if (encoder->encoder_type == DRM_MODE_ENCODER_TMDS) return encoder; @@ -1158,8 +1134,9 @@ amdgpu_connector_dvi_encoder(struct drm_connector *connector) /* then check use digitial */ /* pick the first one */ - if (enc_id) - return drm_encoder_find(connector->dev, NULL, enc_id); + drm_connector_for_each_possible_encoder(connector, encoder, i) + return encoder; + return NULL; } @@ -1296,15 +1273,7 @@ u16 amdgpu_connector_encoder_get_dp_bridge_encoder_id(struct drm_connector *conn struct amdgpu_encoder *amdgpu_encoder; int i; - for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { - if (connector->encoder_ids[i] == 0) - break; - - encoder = drm_encoder_find(connector->dev, NULL, - connector->encoder_ids[i]); - if (!encoder) - continue; - + drm_connector_for_each_possible_encoder(connector, encoder, i) { amdgpu_encoder = to_amdgpu_encoder(encoder); switch (amdgpu_encoder->encoder_id) { @@ -1326,14 +1295,7 @@ static bool amdgpu_connector_encoder_is_hbr2(struct drm_connector *connector) int i; bool found = false; - for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { - if (connector->encoder_ids[i] == 0) - break; - encoder = drm_encoder_find(connector->dev, NULL, - connector->encoder_ids[i]); - if (!encoder) - continue; - + drm_connector_for_each_possible_encoder(connector, encoder, i) { amdgpu_encoder = to_amdgpu_encoder(encoder); if (amdgpu_encoder->caps & ATOM_ENCODER_CAP_RECORD_HBR2) found = true; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 9c85a90be293..663043c8f0f5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -31,45 +31,79 @@ #include <drm/drm_syncobj.h> #include "amdgpu.h" #include "amdgpu_trace.h" +#include "amdgpu_gmc.h" +#include "amdgpu_gem.h" static int amdgpu_cs_user_fence_chunk(struct amdgpu_cs_parser *p, struct drm_amdgpu_cs_chunk_fence *data, uint32_t *offset) { struct drm_gem_object *gobj; + struct amdgpu_bo *bo; unsigned long size; + int r; gobj = drm_gem_object_lookup(p->filp, data->handle); if (gobj == NULL) return -EINVAL; - p->uf_entry.robj = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj)); + bo = amdgpu_bo_ref(gem_to_amdgpu_bo(gobj)); p->uf_entry.priority = 0; - p->uf_entry.tv.bo = &p->uf_entry.robj->tbo; + p->uf_entry.tv.bo = &bo->tbo; p->uf_entry.tv.shared = true; p->uf_entry.user_pages = NULL; - size = amdgpu_bo_size(p->uf_entry.robj); - if (size != PAGE_SIZE || (data->offset + 8) > size) - return -EINVAL; + drm_gem_object_put_unlocked(gobj); + + size = amdgpu_bo_size(bo); + if (size != PAGE_SIZE || (data->offset + 8) > size) { + r = -EINVAL; + goto error_unref; + } + + if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) { + r = -EINVAL; + goto error_unref; + } *offset = data->offset; - drm_gem_object_put_unlocked(gobj); + return 0; - if (amdgpu_ttm_tt_get_usermm(p->uf_entry.robj->tbo.ttm)) { - amdgpu_bo_unref(&p->uf_entry.robj); - return -EINVAL; - } +error_unref: + amdgpu_bo_unref(&bo); + return r; +} + +static int amdgpu_cs_bo_handles_chunk(struct amdgpu_cs_parser *p, + struct drm_amdgpu_bo_list_in *data) +{ + int r; + struct drm_amdgpu_bo_list_entry *info = NULL; + + r = amdgpu_bo_create_list_entry_array(data, &info); + if (r) + return r; + r = amdgpu_bo_list_create(p->adev, p->filp, info, data->bo_number, + &p->bo_list); + if (r) + goto error_free; + + kvfree(info); return 0; + +error_free: + if (info) + kvfree(info); + + return r; } -static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) +static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, union drm_amdgpu_cs *cs) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; struct amdgpu_vm *vm = &fpriv->vm; - union drm_amdgpu_cs *cs = data; uint64_t *chunk_array_user; uint64_t *chunk_array; unsigned size, num_ibs = 0; @@ -163,6 +197,19 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) break; + case AMDGPU_CHUNK_ID_BO_HANDLES: + size = sizeof(struct drm_amdgpu_bo_list_in); + if (p->chunks[i].length_dw * sizeof(uint32_t) < size) { + ret = -EINVAL; + goto free_partial_kdata; + } + + ret = amdgpu_cs_bo_handles_chunk(p, p->chunks[i].kdata); + if (ret) + goto free_partial_kdata; + + break; + case AMDGPU_CHUNK_ID_DEPENDENCIES: case AMDGPU_CHUNK_ID_SYNCOBJ_IN: case AMDGPU_CHUNK_ID_SYNCOBJ_OUT: @@ -183,9 +230,13 @@ static int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data) goto free_all_kdata; } - if (p->uf_entry.robj) + if (p->uf_entry.tv.bo) p->job->uf_addr = uf_offset; kfree(chunk_array); + + /* Use this opportunity to fill in task info for the vm */ + amdgpu_vm_set_task_info(vm); + return 0; free_all_kdata: @@ -257,7 +308,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev, return; } - total_vram = adev->gmc.real_vram_size - adev->vram_pin_size; + total_vram = adev->gmc.real_vram_size - atomic64_read(&adev->vram_pin_size); used_vram = amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); free_vram = used_vram >= total_vram ? 0 : total_vram - used_vram; @@ -302,7 +353,7 @@ static void amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev, *max_bytes = us_to_bytes(adev, adev->mm_stats.accum_us); /* Do the same for visible VRAM if half of it is free */ - if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size) { + if (!amdgpu_gmc_vram_full_visible(&adev->gmc)) { u64 total_vis_vram = adev->gmc.visible_vram_size; u64 used_vis_vram = amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); @@ -359,7 +410,7 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, * to move it. Don't move anything if the threshold is zero. */ if (p->bytes_moved < p->bytes_moved_threshold) { - if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size && + if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && (bo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) { /* And don't move a CPU_ACCESS_REQUIRED BO to limited * visible VRAM if we've depleted our allowance to do @@ -377,11 +428,11 @@ static int amdgpu_cs_bo_validate(struct amdgpu_cs_parser *p, } retry: - amdgpu_ttm_placement_from_domain(bo, domain); + amdgpu_bo_placement_from_domain(bo, domain); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); p->bytes_moved += ctx.bytes_moved; - if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size && + if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && amdgpu_bo_in_cpu_visible_vram(bo)) p->bytes_moved_vis += ctx.bytes_moved; @@ -408,13 +459,13 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, p->evictable = list_prev_entry(p->evictable, tv.head)) { struct amdgpu_bo_list_entry *candidate = p->evictable; - struct amdgpu_bo *bo = candidate->robj; + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(candidate->tv.bo); struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); bool update_bytes_moved_vis; uint32_t other; /* If we reached our current BO we can forget it */ - if (candidate->robj == validated) + if (bo == validated) break; /* We can't move pinned BOs here */ @@ -434,9 +485,9 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, /* Good we can try to move this BO somewhere else */ update_bytes_moved_vis = - adev->gmc.visible_vram_size < adev->gmc.real_vram_size && - amdgpu_bo_in_cpu_visible_vram(bo); - amdgpu_ttm_placement_from_domain(bo, other); + !amdgpu_gmc_vram_full_visible(&adev->gmc) && + amdgpu_bo_in_cpu_visible_vram(bo); + amdgpu_bo_placement_from_domain(bo, other); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); p->bytes_moved += ctx.bytes_moved; if (update_bytes_moved_vis) @@ -479,7 +530,7 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, int r; list_for_each_entry(lobj, validated, tv.head) { - struct amdgpu_bo *bo = lobj->robj; + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(lobj->tv.bo); bool binding_userptr = false; struct mm_struct *usermm; @@ -490,8 +541,8 @@ static int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p, /* Check if we have user pages and nobody bound the BO already */ if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm) && lobj->user_pages) { - amdgpu_ttm_placement_from_domain(bo, - AMDGPU_GEM_DOMAIN_CPU); + amdgpu_bo_placement_from_domain(bo, + AMDGPU_GEM_DOMAIN_CPU); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) return r; @@ -519,32 +570,46 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, union drm_amdgpu_cs *cs) { struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_vm *vm = &fpriv->vm; struct amdgpu_bo_list_entry *e; struct list_head duplicates; - unsigned i, tries = 10; struct amdgpu_bo *gds; struct amdgpu_bo *gws; struct amdgpu_bo *oa; + unsigned tries = 10; int r; INIT_LIST_HEAD(&p->validated); - p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle); - if (p->bo_list) { - amdgpu_bo_list_get_list(p->bo_list, &p->validated); - if (p->bo_list->first_userptr != p->bo_list->num_entries) - p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX); + /* p->bo_list could already be assigned if AMDGPU_CHUNK_ID_BO_HANDLES is present */ + if (cs->in.bo_list_handle) { + if (p->bo_list) + return -EINVAL; + + r = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle, + &p->bo_list); + if (r) + return r; + } else if (!p->bo_list) { + /* Create a empty bo_list when no handle is provided */ + r = amdgpu_bo_list_create(p->adev, p->filp, NULL, 0, + &p->bo_list); + if (r) + return r; } + amdgpu_bo_list_get_list(p->bo_list, &p->validated); + if (p->bo_list->first_userptr != p->bo_list->num_entries) + p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX); + INIT_LIST_HEAD(&duplicates); amdgpu_vm_get_pd_bo(&fpriv->vm, &p->validated, &p->vm_pd); - if (p->uf_entry.robj && !p->uf_entry.robj->parent) + if (p->uf_entry.tv.bo && !ttm_to_amdgpu_bo(p->uf_entry.tv.bo)->parent) list_add(&p->uf_entry.tv.head, &p->validated); while (1) { struct list_head need_pages; - unsigned i; r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true, &duplicates); @@ -554,17 +619,9 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, goto error_free_pages; } - /* Without a BO list we don't have userptr BOs */ - if (!p->bo_list) - break; - INIT_LIST_HEAD(&need_pages); - for (i = p->bo_list->first_userptr; - i < p->bo_list->num_entries; ++i) { - struct amdgpu_bo *bo; - - e = &p->bo_list->array[i]; - bo = e->robj; + amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); if (amdgpu_ttm_tt_userptr_invalidated(bo->tbo.ttm, &e->user_invalidated) && e->user_pages) { @@ -583,7 +640,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, list_del(&e->tv.head); list_add(&e->tv.head, &need_pages); - amdgpu_bo_unreserve(e->robj); + amdgpu_bo_unreserve(bo); } } @@ -602,7 +659,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, /* Fill the page arrays for all userptrs. */ list_for_each_entry(e, &need_pages, tv.head) { - struct ttm_tt *ttm = e->robj->tbo.ttm; + struct ttm_tt *ttm = e->tv.bo->ttm; e->user_pages = kvmalloc_array(ttm->num_pages, sizeof(struct page*), @@ -656,39 +713,28 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, amdgpu_cs_report_moved_bytes(p->adev, p->bytes_moved, p->bytes_moved_vis); - if (p->bo_list) { - struct amdgpu_vm *vm = &fpriv->vm; - unsigned i; - - gds = p->bo_list->gds_obj; - gws = p->bo_list->gws_obj; - oa = p->bo_list->oa_obj; - for (i = 0; i < p->bo_list->num_entries; i++) { - struct amdgpu_bo *bo = p->bo_list->array[i].robj; + gds = p->bo_list->gds_obj; + gws = p->bo_list->gws_obj; + oa = p->bo_list->oa_obj; - p->bo_list->array[i].bo_va = amdgpu_vm_bo_find(vm, bo); - } - } else { - gds = p->adev->gds.gds_gfx_bo; - gws = p->adev->gds.gws_gfx_bo; - oa = p->adev->gds.oa_gfx_bo; - } + amdgpu_bo_list_for_each_entry(e, p->bo_list) + e->bo_va = amdgpu_vm_bo_find(vm, ttm_to_amdgpu_bo(e->tv.bo)); if (gds) { - p->job->gds_base = amdgpu_bo_gpu_offset(gds); - p->job->gds_size = amdgpu_bo_size(gds); + p->job->gds_base = amdgpu_bo_gpu_offset(gds) >> PAGE_SHIFT; + p->job->gds_size = amdgpu_bo_size(gds) >> PAGE_SHIFT; } if (gws) { - p->job->gws_base = amdgpu_bo_gpu_offset(gws); - p->job->gws_size = amdgpu_bo_size(gws); + p->job->gws_base = amdgpu_bo_gpu_offset(gws) >> PAGE_SHIFT; + p->job->gws_size = amdgpu_bo_size(gws) >> PAGE_SHIFT; } if (oa) { - p->job->oa_base = amdgpu_bo_gpu_offset(oa); - p->job->oa_size = amdgpu_bo_size(oa); + p->job->oa_base = amdgpu_bo_gpu_offset(oa) >> PAGE_SHIFT; + p->job->oa_size = amdgpu_bo_size(oa) >> PAGE_SHIFT; } - if (!r && p->uf_entry.robj) { - struct amdgpu_bo *uf = p->uf_entry.robj; + if (!r && p->uf_entry.tv.bo) { + struct amdgpu_bo *uf = ttm_to_amdgpu_bo(p->uf_entry.tv.bo); r = amdgpu_ttm_alloc_gart(&uf->tbo); p->job->uf_addr += amdgpu_bo_gpu_offset(uf); @@ -700,18 +746,12 @@ error_validate: error_free_pages: - if (p->bo_list) { - for (i = p->bo_list->first_userptr; - i < p->bo_list->num_entries; ++i) { - e = &p->bo_list->array[i]; - - if (!e->user_pages) - continue; + amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { + if (!e->user_pages) + continue; - release_pages(e->user_pages, - e->robj->tbo.ttm->num_pages); - kvfree(e->user_pages); - } + release_pages(e->user_pages, e->tv.bo->ttm->num_pages); + kvfree(e->user_pages); } return r; @@ -723,9 +763,11 @@ static int amdgpu_cs_sync_rings(struct amdgpu_cs_parser *p) int r; list_for_each_entry(e, &p->validated, tv.head) { - struct reservation_object *resv = e->robj->tbo.resv; + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + struct reservation_object *resv = bo->tbo.resv; + r = amdgpu_sync_resv(p->adev, &p->job->sync, resv, p->filp, - amdgpu_bo_explicit_sync(e->robj)); + amdgpu_bo_explicit_sync(bo)); if (r) return r; @@ -768,17 +810,88 @@ static void amdgpu_cs_parser_fini(struct amdgpu_cs_parser *parser, int error, kfree(parser->chunks); if (parser->job) amdgpu_job_free(parser->job); - amdgpu_bo_unref(&parser->uf_entry.robj); + if (parser->uf_entry.tv.bo) { + struct amdgpu_bo *uf = ttm_to_amdgpu_bo(parser->uf_entry.tv.bo); + + amdgpu_bo_unref(&uf); + } } -static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) +static int amdgpu_cs_vm_handling(struct amdgpu_cs_parser *p) { - struct amdgpu_device *adev = p->adev; + struct amdgpu_ring *ring = to_amdgpu_ring(p->entity->rq->sched); struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct amdgpu_device *adev = p->adev; struct amdgpu_vm *vm = &fpriv->vm; + struct amdgpu_bo_list_entry *e; struct amdgpu_bo_va *bo_va; struct amdgpu_bo *bo; - int i, r; + int r; + + /* Only for UVD/VCE VM emulation */ + if (ring->funcs->parse_cs || ring->funcs->patch_cs_in_place) { + unsigned i, j; + + for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) { + struct drm_amdgpu_cs_chunk_ib *chunk_ib; + struct amdgpu_bo_va_mapping *m; + struct amdgpu_bo *aobj = NULL; + struct amdgpu_cs_chunk *chunk; + uint64_t offset, va_start; + struct amdgpu_ib *ib; + uint8_t *kptr; + + chunk = &p->chunks[i]; + ib = &p->job->ibs[j]; + chunk_ib = chunk->kdata; + + if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB) + continue; + + va_start = chunk_ib->va_start & AMDGPU_GMC_HOLE_MASK; + r = amdgpu_cs_find_mapping(p, va_start, &aobj, &m); + if (r) { + DRM_ERROR("IB va_start is invalid\n"); + return r; + } + + if ((va_start + chunk_ib->ib_bytes) > + (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) { + DRM_ERROR("IB va_start+ib_bytes is invalid\n"); + return -EINVAL; + } + + /* the IB should be reserved at this point */ + r = amdgpu_bo_kmap(aobj, (void **)&kptr); + if (r) { + return r; + } + + offset = m->start * AMDGPU_GPU_PAGE_SIZE; + kptr += va_start - offset; + + if (ring->funcs->parse_cs) { + memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); + amdgpu_bo_kunmap(aobj); + + r = amdgpu_ring_parse_cs(ring, p, j); + if (r) + return r; + } else { + ib->ptr = (uint32_t *)kptr; + r = amdgpu_ring_patch_cs_in_place(ring, p, j); + amdgpu_bo_kunmap(aobj); + if (r) + return r; + } + + j++; + } + } + + if (!p->job->vm) + return amdgpu_cs_sync_rings(p); + r = amdgpu_vm_clear_freed(adev, vm, NULL); if (r) @@ -808,29 +921,26 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) return r; } - if (p->bo_list) { - for (i = 0; i < p->bo_list->num_entries; i++) { - struct dma_fence *f; - - /* ignore duplicates */ - bo = p->bo_list->array[i].robj; - if (!bo) - continue; + amdgpu_bo_list_for_each_entry(e, p->bo_list) { + struct dma_fence *f; - bo_va = p->bo_list->array[i].bo_va; - if (bo_va == NULL) - continue; + /* ignore duplicates */ + bo = ttm_to_amdgpu_bo(e->tv.bo); + if (!bo) + continue; - r = amdgpu_vm_bo_update(adev, bo_va, false); - if (r) - return r; + bo_va = e->bo_va; + if (bo_va == NULL) + continue; - f = bo_va->last_pt_update; - r = amdgpu_sync_fence(adev, &p->job->sync, f, false); - if (r) - return r; - } + r = amdgpu_vm_bo_update(adev, bo_va, false); + if (r) + return r; + f = bo_va->last_pt_update; + r = amdgpu_sync_fence(adev, &p->job->sync, f, false); + if (r) + return r; } r = amdgpu_vm_handle_moved(adev, vm); @@ -845,11 +955,18 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) if (r) return r; - if (amdgpu_vm_debug && p->bo_list) { + r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv); + if (r) + return r; + + p->job->vm_pd_addr = amdgpu_gmc_pd_addr(vm->root.base.bo); + + if (amdgpu_vm_debug) { /* Invalidate all BOs to test for userspace bugs */ - for (i = 0; i < p->bo_list->num_entries; i++) { + amdgpu_bo_list_for_each_entry(e, p->bo_list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + /* ignore duplicates */ - bo = p->bo_list->array[i].robj; if (!bo) continue; @@ -857,82 +974,6 @@ static int amdgpu_bo_vm_update_pte(struct amdgpu_cs_parser *p) } } - return r; -} - -static int amdgpu_cs_ib_vm_chunk(struct amdgpu_device *adev, - struct amdgpu_cs_parser *p) -{ - struct amdgpu_fpriv *fpriv = p->filp->driver_priv; - struct amdgpu_vm *vm = &fpriv->vm; - struct amdgpu_ring *ring = p->job->ring; - int r; - - /* Only for UVD/VCE VM emulation */ - if (p->job->ring->funcs->parse_cs) { - unsigned i, j; - - for (i = 0, j = 0; i < p->nchunks && j < p->job->num_ibs; i++) { - struct drm_amdgpu_cs_chunk_ib *chunk_ib; - struct amdgpu_bo_va_mapping *m; - struct amdgpu_bo *aobj = NULL; - struct amdgpu_cs_chunk *chunk; - uint64_t offset, va_start; - struct amdgpu_ib *ib; - uint8_t *kptr; - - chunk = &p->chunks[i]; - ib = &p->job->ibs[j]; - chunk_ib = chunk->kdata; - - if (chunk->chunk_id != AMDGPU_CHUNK_ID_IB) - continue; - - va_start = chunk_ib->va_start & AMDGPU_VA_HOLE_MASK; - r = amdgpu_cs_find_mapping(p, va_start, &aobj, &m); - if (r) { - DRM_ERROR("IB va_start is invalid\n"); - return r; - } - - if ((va_start + chunk_ib->ib_bytes) > - (m->last + 1) * AMDGPU_GPU_PAGE_SIZE) { - DRM_ERROR("IB va_start+ib_bytes is invalid\n"); - return -EINVAL; - } - - /* the IB should be reserved at this point */ - r = amdgpu_bo_kmap(aobj, (void **)&kptr); - if (r) { - return r; - } - - offset = m->start * AMDGPU_GPU_PAGE_SIZE; - kptr += va_start - offset; - - memcpy(ib->ptr, kptr, chunk_ib->ib_bytes); - amdgpu_bo_kunmap(aobj); - - r = amdgpu_ring_parse_cs(ring, p, j); - if (r) - return r; - - j++; - } - } - - if (p->job->vm) { - p->job->vm_pd_addr = amdgpu_bo_gpu_offset(vm->root.base.bo); - - r = amdgpu_bo_vm_update_pte(p); - if (r) - return r; - - r = reservation_object_reserve_shared(vm->root.base.bo->tbo.resv); - if (r) - return r; - } - return amdgpu_cs_sync_rings(p); } @@ -941,14 +982,15 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, { struct amdgpu_fpriv *fpriv = parser->filp->driver_priv; struct amdgpu_vm *vm = &fpriv->vm; - int i, j; int r, ce_preempt = 0, de_preempt = 0; + struct amdgpu_ring *ring; + int i, j; for (i = 0, j = 0; i < parser->nchunks && j < parser->job->num_ibs; i++) { struct amdgpu_cs_chunk *chunk; struct amdgpu_ib *ib; struct drm_amdgpu_cs_chunk_ib *chunk_ib; - struct amdgpu_ring *ring; + struct drm_sched_entity *entity; chunk = &parser->chunks[i]; ib = &parser->job->ibs[j]; @@ -970,27 +1012,24 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, return -EINVAL; } - r = amdgpu_queue_mgr_map(adev, &parser->ctx->queue_mgr, chunk_ib->ip_type, - chunk_ib->ip_instance, chunk_ib->ring, &ring); + r = amdgpu_ctx_get_entity(parser->ctx, chunk_ib->ip_type, + chunk_ib->ip_instance, chunk_ib->ring, + &entity); if (r) return r; - if (chunk_ib->flags & AMDGPU_IB_FLAG_PREAMBLE) { - parser->job->preamble_status |= AMDGPU_PREAMBLE_IB_PRESENT; - if (!parser->ctx->preamble_presented) { - parser->job->preamble_status |= AMDGPU_PREAMBLE_IB_PRESENT_FIRST; - parser->ctx->preamble_presented = true; - } - } + if (chunk_ib->flags & AMDGPU_IB_FLAG_PREAMBLE) + parser->job->preamble_status |= + AMDGPU_PREAMBLE_IB_PRESENT; - if (parser->job->ring && parser->job->ring != ring) + if (parser->entity && parser->entity != entity) return -EINVAL; - parser->job->ring = ring; + parser->entity = entity; - r = amdgpu_ib_get(adev, vm, - ring->funcs->parse_cs ? chunk_ib->ib_bytes : 0, - ib); + ring = to_amdgpu_ring(entity->rq->sched); + r = amdgpu_ib_get(adev, vm, ring->funcs->parse_cs ? + chunk_ib->ib_bytes : 0, ib); if (r) { DRM_ERROR("Failed to get ib !\n"); return r; @@ -1004,12 +1043,13 @@ static int amdgpu_cs_ib_fill(struct amdgpu_device *adev, } /* UVD & VCE fw doesn't support user fences */ + ring = to_amdgpu_ring(parser->entity->rq->sched); if (parser->job->uf_addr && ( - parser->job->ring->funcs->type == AMDGPU_RING_TYPE_UVD || - parser->job->ring->funcs->type == AMDGPU_RING_TYPE_VCE)) + ring->funcs->type == AMDGPU_RING_TYPE_UVD || + ring->funcs->type == AMDGPU_RING_TYPE_VCE)) return -EINVAL; - return amdgpu_ctx_wait_prev_fence(parser->ctx, parser->job->ring->idx); + return amdgpu_ctx_wait_prev_fence(parser->ctx, parser->entity); } static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p, @@ -1025,24 +1065,23 @@ static int amdgpu_cs_process_fence_dep(struct amdgpu_cs_parser *p, sizeof(struct drm_amdgpu_cs_chunk_dep); for (i = 0; i < num_deps; ++i) { - struct amdgpu_ring *ring; struct amdgpu_ctx *ctx; + struct drm_sched_entity *entity; struct dma_fence *fence; ctx = amdgpu_ctx_get(fpriv, deps[i].ctx_id); if (ctx == NULL) return -EINVAL; - r = amdgpu_queue_mgr_map(p->adev, &ctx->queue_mgr, - deps[i].ip_type, - deps[i].ip_instance, - deps[i].ring, &ring); + r = amdgpu_ctx_get_entity(ctx, deps[i].ip_type, + deps[i].ip_instance, + deps[i].ring, &entity); if (r) { amdgpu_ctx_put(ctx); return r; } - fence = amdgpu_ctx_get_fence(ctx, ring, + fence = amdgpu_ctx_get_fence(ctx, entity, deps[i].handle); if (IS_ERR(fence)) { r = PTR_ERR(fence); @@ -1065,7 +1104,7 @@ static int amdgpu_syncobj_lookup_and_add_to_sync(struct amdgpu_cs_parser *p, { int r; struct dma_fence *fence; - r = drm_syncobj_find_fence(p->filp, handle, &fence); + r = drm_syncobj_find_fence(p->filp, handle, 0, &fence); if (r) return r; @@ -1154,71 +1193,80 @@ static void amdgpu_cs_post_dependencies(struct amdgpu_cs_parser *p) int i; for (i = 0; i < p->num_post_dep_syncobjs; ++i) - drm_syncobj_replace_fence(p->post_dep_syncobjs[i], p->fence); + drm_syncobj_replace_fence(p->post_dep_syncobjs[i], 0, p->fence); } static int amdgpu_cs_submit(struct amdgpu_cs_parser *p, union drm_amdgpu_cs *cs) { - struct amdgpu_ring *ring = p->job->ring; - struct drm_sched_entity *entity = &p->ctx->rings[ring->idx].entity; + struct amdgpu_fpriv *fpriv = p->filp->driver_priv; + struct drm_sched_entity *entity = p->entity; + enum drm_sched_priority priority; + struct amdgpu_ring *ring; + struct amdgpu_bo_list_entry *e; struct amdgpu_job *job; - unsigned i; uint64_t seq; int r; - amdgpu_mn_lock(p->mn); - if (p->bo_list) { - for (i = p->bo_list->first_userptr; - i < p->bo_list->num_entries; ++i) { - struct amdgpu_bo *bo = p->bo_list->array[i].robj; - - if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) { - amdgpu_mn_unlock(p->mn); - return -ERESTARTSYS; - } - } - } - job = p->job; p->job = NULL; - r = drm_sched_job_init(&job->base, &ring->sched, entity, p->filp); - if (r) { - amdgpu_job_free(job); - amdgpu_mn_unlock(p->mn); - return r; + r = drm_sched_job_init(&job->base, entity, p->filp); + if (r) + goto error_unlock; + + /* No memory allocation is allowed while holding the mn lock */ + amdgpu_mn_lock(p->mn); + amdgpu_bo_list_for_each_userptr_entry(e, p->bo_list) { + struct amdgpu_bo *bo = ttm_to_amdgpu_bo(e->tv.bo); + + if (amdgpu_ttm_tt_userptr_needs_pages(bo->tbo.ttm)) { + r = -ERESTARTSYS; + goto error_abort; + } } job->owner = p->filp; - job->fence_ctx = entity->fence_context; p->fence = dma_fence_get(&job->base.s_fence->finished); - r = amdgpu_ctx_add_fence(p->ctx, ring, p->fence, &seq); - if (r) { - dma_fence_put(p->fence); - dma_fence_put(&job->base.s_fence->finished); - amdgpu_job_free(job); - amdgpu_mn_unlock(p->mn); - return r; - } - + amdgpu_ctx_add_fence(p->ctx, entity, p->fence, &seq); amdgpu_cs_post_dependencies(p); + if ((job->preamble_status & AMDGPU_PREAMBLE_IB_PRESENT) && + !p->ctx->preamble_presented) { + job->preamble_status |= AMDGPU_PREAMBLE_IB_PRESENT_FIRST; + p->ctx->preamble_presented = true; + } + cs->out.handle = seq; job->uf_sequence = seq; amdgpu_job_free_resources(job); - amdgpu_ring_priority_get(job->ring, job->base.s_priority); trace_amdgpu_cs_ioctl(job); + amdgpu_vm_bo_trace_cs(&fpriv->vm, &p->ticket); + priority = job->base.s_priority; drm_sched_entity_push_job(&job->base, entity); + ring = to_amdgpu_ring(entity->rq->sched); + amdgpu_ring_priority_get(ring, priority); + + amdgpu_vm_move_to_lru_tail(p->adev, &fpriv->vm); + ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence); amdgpu_mn_unlock(p->mn); return 0; + +error_abort: + dma_fence_put(&job->base.s_fence->finished); + job->base.s_fence = NULL; + amdgpu_mn_unlock(p->mn); + +error_unlock: + amdgpu_job_free(job); + return r; } int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) @@ -1245,6 +1293,12 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) if (r) goto out; + r = amdgpu_cs_dependencies(adev, &parser); + if (r) { + DRM_ERROR("Failed in the dependencies handling %d!\n", r); + goto out; + } + r = amdgpu_cs_parser_bos(&parser, data); if (r) { if (r == -ENOMEM) @@ -1256,16 +1310,10 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) reserved_buffers = true; - r = amdgpu_cs_dependencies(adev, &parser); - if (r) { - DRM_ERROR("Failed in the dependencies handling %d!\n", r); - goto out; - } - for (i = 0; i < parser.job->num_ibs; i++) trace_amdgpu_cs(&parser, i); - r = amdgpu_cs_ib_vm_chunk(adev, &parser); + r = amdgpu_cs_vm_handling(&parser); if (r) goto out; @@ -1289,9 +1337,8 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { union drm_amdgpu_wait_cs *wait = data; - struct amdgpu_device *adev = dev->dev_private; unsigned long timeout = amdgpu_gem_timeout(wait->in.timeout); - struct amdgpu_ring *ring = NULL; + struct drm_sched_entity *entity; struct amdgpu_ctx *ctx; struct dma_fence *fence; long r; @@ -1300,15 +1347,14 @@ int amdgpu_cs_wait_ioctl(struct drm_device *dev, void *data, if (ctx == NULL) return -EINVAL; - r = amdgpu_queue_mgr_map(adev, &ctx->queue_mgr, - wait->in.ip_type, wait->in.ip_instance, - wait->in.ring, &ring); + r = amdgpu_ctx_get_entity(ctx, wait->in.ip_type, wait->in.ip_instance, + wait->in.ring, &entity); if (r) { amdgpu_ctx_put(ctx); return r; } - fence = amdgpu_ctx_get_fence(ctx, ring, wait->in.handle); + fence = amdgpu_ctx_get_fence(ctx, entity, wait->in.handle); if (IS_ERR(fence)) r = PTR_ERR(fence); else if (fence) { @@ -1340,7 +1386,7 @@ static struct dma_fence *amdgpu_cs_get_fence(struct amdgpu_device *adev, struct drm_file *filp, struct drm_amdgpu_fence *user) { - struct amdgpu_ring *ring; + struct drm_sched_entity *entity; struct amdgpu_ctx *ctx; struct dma_fence *fence; int r; @@ -1349,14 +1395,14 @@ static struct dma_fence *amdgpu_cs_get_fence(struct amdgpu_device *adev, if (ctx == NULL) return ERR_PTR(-EINVAL); - r = amdgpu_queue_mgr_map(adev, &ctx->queue_mgr, user->ip_type, - user->ip_instance, user->ring, &ring); + r = amdgpu_ctx_get_entity(ctx, user->ip_type, user->ip_instance, + user->ring, &entity); if (r) { amdgpu_ctx_put(ctx); return ERR_PTR(r); } - fence = amdgpu_ctx_get_fence(ctx, ring, user->seq_no); + fence = amdgpu_ctx_get_fence(ctx, entity, user->seq_no); amdgpu_ctx_put(ctx); return fence; @@ -1605,7 +1651,7 @@ int amdgpu_cs_find_mapping(struct amdgpu_cs_parser *parser, if (!((*bo)->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)) { (*bo)->flags |= AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; - amdgpu_ttm_placement_from_domain(*bo, (*bo)->allowed_domains); + amdgpu_bo_placement_from_domain(*bo, (*bo)->allowed_domains); r = ttm_bo_validate(&(*bo)->tbo, &(*bo)->placement, &ctx); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index c5bb36275e93..f9b54236102d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -27,6 +27,30 @@ #include "amdgpu.h" #include "amdgpu_sched.h" +#define to_amdgpu_ctx_entity(e) \ + container_of((e), struct amdgpu_ctx_entity, entity) + +const unsigned int amdgpu_ctx_num_entities[AMDGPU_HW_IP_NUM] = { + [AMDGPU_HW_IP_GFX] = 1, + [AMDGPU_HW_IP_COMPUTE] = 4, + [AMDGPU_HW_IP_DMA] = 2, + [AMDGPU_HW_IP_UVD] = 1, + [AMDGPU_HW_IP_VCE] = 1, + [AMDGPU_HW_IP_UVD_ENC] = 1, + [AMDGPU_HW_IP_VCN_DEC] = 1, + [AMDGPU_HW_IP_VCN_ENC] = 1, +}; + +static int amdgput_ctx_total_num_entities(void) +{ + unsigned i, num_entities = 0; + + for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) + num_entities += amdgpu_ctx_num_entities[i]; + + return num_entities; +} + static int amdgpu_ctx_priority_permit(struct drm_file *filp, enum drm_sched_priority priority) { @@ -48,6 +72,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, struct drm_file *filp, struct amdgpu_ctx *ctx) { + unsigned num_entities = amdgput_ctx_total_num_entities(); unsigned i, j; int r; @@ -60,19 +85,33 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, memset(ctx, 0, sizeof(*ctx)); ctx->adev = adev; - kref_init(&ctx->refcount); - spin_lock_init(&ctx->ring_lock); - ctx->fences = kcalloc(amdgpu_sched_jobs * AMDGPU_MAX_RINGS, + + ctx->fences = kcalloc(amdgpu_sched_jobs * num_entities, sizeof(struct dma_fence*), GFP_KERNEL); if (!ctx->fences) return -ENOMEM; - mutex_init(&ctx->lock); + ctx->entities[0] = kcalloc(num_entities, + sizeof(struct amdgpu_ctx_entity), + GFP_KERNEL); + if (!ctx->entities[0]) { + r = -ENOMEM; + goto error_free_fences; + } - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { - ctx->rings[i].sequence = 1; - ctx->rings[i].fences = &ctx->fences[amdgpu_sched_jobs * i]; + for (i = 0; i < num_entities; ++i) { + struct amdgpu_ctx_entity *entity = &ctx->entities[0][i]; + + entity->sequence = 1; + entity->fences = &ctx->fences[amdgpu_sched_jobs * i]; } + for (i = 1; i < AMDGPU_HW_IP_NUM; ++i) + ctx->entities[i] = ctx->entities[i - 1] + + amdgpu_ctx_num_entities[i - 1]; + + kref_init(&ctx->refcount); + spin_lock_init(&ctx->ring_lock); + mutex_init(&ctx->lock); ctx->reset_counter = atomic_read(&adev->gpu_reset_counter); ctx->reset_counter_query = ctx->reset_counter; @@ -80,32 +119,70 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, ctx->init_priority = priority; ctx->override_priority = DRM_SCHED_PRIORITY_UNSET; - /* create context entity for each ring */ - for (i = 0; i < adev->num_rings; i++) { - struct amdgpu_ring *ring = adev->rings[i]; - struct drm_sched_rq *rq; - - rq = &ring->sched.sched_rq[priority]; + for (i = 0; i < AMDGPU_HW_IP_NUM; ++i) { + struct amdgpu_ring *rings[AMDGPU_MAX_RINGS]; + struct drm_sched_rq *rqs[AMDGPU_MAX_RINGS]; + unsigned num_rings; + + switch (i) { + case AMDGPU_HW_IP_GFX: + rings[0] = &adev->gfx.gfx_ring[0]; + num_rings = 1; + break; + case AMDGPU_HW_IP_COMPUTE: + for (j = 0; j < adev->gfx.num_compute_rings; ++j) + rings[j] = &adev->gfx.compute_ring[j]; + num_rings = adev->gfx.num_compute_rings; + break; + case AMDGPU_HW_IP_DMA: + for (j = 0; j < adev->sdma.num_instances; ++j) + rings[j] = &adev->sdma.instance[j].ring; + num_rings = adev->sdma.num_instances; + break; + case AMDGPU_HW_IP_UVD: + rings[0] = &adev->uvd.inst[0].ring; + num_rings = 1; + break; + case AMDGPU_HW_IP_VCE: + rings[0] = &adev->vce.ring[0]; + num_rings = 1; + break; + case AMDGPU_HW_IP_UVD_ENC: + rings[0] = &adev->uvd.inst[0].ring_enc[0]; + num_rings = 1; + break; + case AMDGPU_HW_IP_VCN_DEC: + rings[0] = &adev->vcn.ring_dec; + num_rings = 1; + break; + case AMDGPU_HW_IP_VCN_ENC: + rings[0] = &adev->vcn.ring_enc[0]; + num_rings = 1; + break; + case AMDGPU_HW_IP_VCN_JPEG: + rings[0] = &adev->vcn.ring_jpeg; + num_rings = 1; + break; + } - if (ring == &adev->gfx.kiq.ring) - continue; + for (j = 0; j < num_rings; ++j) + rqs[j] = &rings[j]->sched.sched_rq[priority]; - r = drm_sched_entity_init(&ring->sched, &ctx->rings[i].entity, - rq, &ctx->guilty); + for (j = 0; j < amdgpu_ctx_num_entities[i]; ++j) + r = drm_sched_entity_init(&ctx->entities[i][j].entity, + rqs, num_rings, &ctx->guilty); if (r) - goto failed; + goto error_cleanup_entities; } - r = amdgpu_queue_mgr_init(adev, &ctx->queue_mgr); - if (r) - goto failed; - return 0; -failed: - for (j = 0; j < i; j++) - drm_sched_entity_fini(&adev->rings[j]->sched, - &ctx->rings[j].entity); +error_cleanup_entities: + for (i = 0; i < num_entities; ++i) + drm_sched_entity_destroy(&ctx->entities[0][i].entity); + kfree(ctx->entities[0]); + +error_free_fences: kfree(ctx->fences); ctx->fences = NULL; return r; @@ -114,25 +191,47 @@ failed: static void amdgpu_ctx_fini(struct kref *ref) { struct amdgpu_ctx *ctx = container_of(ref, struct amdgpu_ctx, refcount); + unsigned num_entities = amdgput_ctx_total_num_entities(); struct amdgpu_device *adev = ctx->adev; unsigned i, j; if (!adev) return; - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) + for (i = 0; i < num_entities; ++i) for (j = 0; j < amdgpu_sched_jobs; ++j) - dma_fence_put(ctx->rings[i].fences[j]); + dma_fence_put(ctx->entities[0][i].fences[j]); kfree(ctx->fences); - ctx->fences = NULL; - - amdgpu_queue_mgr_fini(adev, &ctx->queue_mgr); + kfree(ctx->entities[0]); mutex_destroy(&ctx->lock); kfree(ctx); } +int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance, + u32 ring, struct drm_sched_entity **entity) +{ + if (hw_ip >= AMDGPU_HW_IP_NUM) { + DRM_ERROR("unknown HW IP type: %d\n", hw_ip); + return -EINVAL; + } + + /* Right now all IPs have only one instance - multiple rings. */ + if (instance != 0) { + DRM_DEBUG("invalid ip instance: %d\n", instance); + return -EINVAL; + } + + if (ring >= amdgpu_ctx_num_entities[hw_ip]) { + DRM_DEBUG("invalid ring: %d %d\n", hw_ip, ring); + return -EINVAL; + } + + *entity = &ctx->entities[hw_ip][ring].entity; + return 0; +} + static int amdgpu_ctx_alloc(struct amdgpu_device *adev, struct amdgpu_fpriv *fpriv, struct drm_file *filp, @@ -169,18 +268,17 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev, static void amdgpu_ctx_do_release(struct kref *ref) { struct amdgpu_ctx *ctx; + unsigned num_entities; u32 i; ctx = container_of(ref, struct amdgpu_ctx, refcount); - for (i = 0; i < ctx->adev->num_rings; i++) { + num_entities = 0; + for (i = 0; i < AMDGPU_HW_IP_NUM; i++) + num_entities += amdgpu_ctx_num_entities[i]; - if (ctx->adev->rings[i] == &ctx->adev->gfx.kiq.ring) - continue; - - drm_sched_entity_fini(&ctx->adev->rings[i]->sched, - &ctx->rings[i].entity); - } + for (i = 0; i < num_entities; i++) + drm_sched_entity_destroy(&ctx->entities[0][i].entity); amdgpu_ctx_fini(ref); } @@ -336,56 +434,56 @@ int amdgpu_ctx_put(struct amdgpu_ctx *ctx) return 0; } -int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring, - struct dma_fence *fence, uint64_t* handler) +void amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, + struct drm_sched_entity *entity, + struct dma_fence *fence, uint64_t* handle) { - struct amdgpu_ctx_ring *cring = & ctx->rings[ring->idx]; - uint64_t seq = cring->sequence; - unsigned idx = 0; + struct amdgpu_ctx_entity *centity = to_amdgpu_ctx_entity(entity); + uint64_t seq = centity->sequence; struct dma_fence *other = NULL; + unsigned idx = 0; idx = seq & (amdgpu_sched_jobs - 1); - other = cring->fences[idx]; + other = centity->fences[idx]; if (other) BUG_ON(!dma_fence_is_signaled(other)); dma_fence_get(fence); spin_lock(&ctx->ring_lock); - cring->fences[idx] = fence; - cring->sequence++; + centity->fences[idx] = fence; + centity->sequence++; spin_unlock(&ctx->ring_lock); dma_fence_put(other); - if (handler) - *handler = seq; - - return 0; + if (handle) + *handle = seq; } struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, - struct amdgpu_ring *ring, uint64_t seq) + struct drm_sched_entity *entity, + uint64_t seq) { - struct amdgpu_ctx_ring *cring = & ctx->rings[ring->idx]; + struct amdgpu_ctx_entity *centity = to_amdgpu_ctx_entity(entity); struct dma_fence *fence; spin_lock(&ctx->ring_lock); if (seq == ~0ull) - seq = ctx->rings[ring->idx].sequence - 1; + seq = centity->sequence - 1; - if (seq >= cring->sequence) { + if (seq >= centity->sequence) { spin_unlock(&ctx->ring_lock); return ERR_PTR(-EINVAL); } - if (seq + amdgpu_sched_jobs < cring->sequence) { + if (seq + amdgpu_sched_jobs < centity->sequence) { spin_unlock(&ctx->ring_lock); return NULL; } - fence = dma_fence_get(cring->fences[seq & (amdgpu_sched_jobs - 1)]); + fence = dma_fence_get(centity->fences[seq & (amdgpu_sched_jobs - 1)]); spin_unlock(&ctx->ring_lock); return fence; @@ -394,35 +492,28 @@ struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, enum drm_sched_priority priority) { - int i; - struct amdgpu_device *adev = ctx->adev; - struct drm_sched_rq *rq; - struct drm_sched_entity *entity; - struct amdgpu_ring *ring; + unsigned num_entities = amdgput_ctx_total_num_entities(); enum drm_sched_priority ctx_prio; + unsigned i; ctx->override_priority = priority; ctx_prio = (ctx->override_priority == DRM_SCHED_PRIORITY_UNSET) ? ctx->init_priority : ctx->override_priority; - for (i = 0; i < adev->num_rings; i++) { - ring = adev->rings[i]; - entity = &ctx->rings[i].entity; - rq = &ring->sched.sched_rq[ctx_prio]; + for (i = 0; i < num_entities; i++) { + struct drm_sched_entity *entity = &ctx->entities[0][i].entity; - if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) - continue; - - drm_sched_entity_set_rq(entity, rq); + drm_sched_entity_set_priority(entity, ctx_prio); } } -int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id) +int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, + struct drm_sched_entity *entity) { - struct amdgpu_ctx_ring *cring = &ctx->rings[ring_id]; - unsigned idx = cring->sequence & (amdgpu_sched_jobs - 1); - struct dma_fence *other = cring->fences[idx]; + struct amdgpu_ctx_entity *centity = to_amdgpu_ctx_entity(entity); + unsigned idx = centity->sequence & (amdgpu_sched_jobs - 1); + struct dma_fence *other = centity->fences[idx]; if (other) { signed long r; @@ -444,35 +535,37 @@ void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr) idr_init(&mgr->ctx_handles); } -void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr) +void amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr) { + unsigned num_entities = amdgput_ctx_total_num_entities(); struct amdgpu_ctx *ctx; struct idr *idp; uint32_t id, i; + long max_wait = MAX_WAIT_SCHED_ENTITY_Q_EMPTY; idp = &mgr->ctx_handles; + mutex_lock(&mgr->lock); idr_for_each_entry(idp, ctx, id) { - if (!ctx->adev) + if (!ctx->adev) { + mutex_unlock(&mgr->lock); return; + } - for (i = 0; i < ctx->adev->num_rings; i++) { - - if (ctx->adev->rings[i] == &ctx->adev->gfx.kiq.ring) - continue; + for (i = 0; i < num_entities; i++) { + struct drm_sched_entity *entity; - if (kref_read(&ctx->refcount) == 1) - drm_sched_entity_do_release(&ctx->adev->rings[i]->sched, - &ctx->rings[i].entity); - else - DRM_ERROR("ctx %p is still alive\n", ctx); + entity = &ctx->entities[0][i].entity; + max_wait = drm_sched_entity_flush(entity, max_wait); } } + mutex_unlock(&mgr->lock); } -void amdgpu_ctx_mgr_entity_cleanup(struct amdgpu_ctx_mgr *mgr) +void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr) { + unsigned num_entities = amdgput_ctx_total_num_entities(); struct amdgpu_ctx *ctx; struct idr *idp; uint32_t id, i; @@ -484,17 +577,13 @@ void amdgpu_ctx_mgr_entity_cleanup(struct amdgpu_ctx_mgr *mgr) if (!ctx->adev) return; - for (i = 0; i < ctx->adev->num_rings; i++) { - - if (ctx->adev->rings[i] == &ctx->adev->gfx.kiq.ring) - continue; - - if (kref_read(&ctx->refcount) == 1) - drm_sched_entity_cleanup(&ctx->adev->rings[i]->sched, - &ctx->rings[i].entity); - else - DRM_ERROR("ctx %p is still alive\n", ctx); + if (kref_read(&ctx->refcount) != 1) { + DRM_ERROR("ctx %p is still alive\n", ctx); + continue; } + + for (i = 0; i < num_entities; i++) + drm_sched_entity_fini(&ctx->entities[0][i].entity); } } @@ -504,7 +593,7 @@ void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr) struct idr *idp; uint32_t id; - amdgpu_ctx_mgr_entity_cleanup(mgr); + amdgpu_ctx_mgr_entity_fini(mgr); idp = &mgr->ctx_handles; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h new file mode 100644 index 000000000000..b3b012c0a7da --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h @@ -0,0 +1,88 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __AMDGPU_CTX_H__ +#define __AMDGPU_CTX_H__ + +#include "amdgpu_ring.h" + +struct drm_device; +struct drm_file; +struct amdgpu_fpriv; + +struct amdgpu_ctx_entity { + uint64_t sequence; + struct dma_fence **fences; + struct drm_sched_entity entity; +}; + +struct amdgpu_ctx { + struct kref refcount; + struct amdgpu_device *adev; + unsigned reset_counter; + unsigned reset_counter_query; + uint32_t vram_lost_counter; + spinlock_t ring_lock; + struct dma_fence **fences; + struct amdgpu_ctx_entity *entities[AMDGPU_HW_IP_NUM]; + bool preamble_presented; + enum drm_sched_priority init_priority; + enum drm_sched_priority override_priority; + struct mutex lock; + atomic_t guilty; +}; + +struct amdgpu_ctx_mgr { + struct amdgpu_device *adev; + struct mutex lock; + /* protected by lock */ + struct idr ctx_handles; +}; + +extern const unsigned int amdgpu_ctx_num_entities[AMDGPU_HW_IP_NUM]; + +struct amdgpu_ctx *amdgpu_ctx_get(struct amdgpu_fpriv *fpriv, uint32_t id); +int amdgpu_ctx_put(struct amdgpu_ctx *ctx); + +int amdgpu_ctx_get_entity(struct amdgpu_ctx *ctx, u32 hw_ip, u32 instance, + u32 ring, struct drm_sched_entity **entity); +void amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, + struct drm_sched_entity *entity, + struct dma_fence *fence, uint64_t *seq); +struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx, + struct drm_sched_entity *entity, + uint64_t seq); +void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx, + enum drm_sched_priority priority); + +int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); + +int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, + struct drm_sched_entity *entity); + +void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_mgr_entity_flush(struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index f5fb93795a69..dd9a4fb9ce39 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -826,21 +826,13 @@ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev) { struct drm_minor *minor = adev->ddev->primary; struct dentry *ent, *root = minor->debugfs_root; - unsigned i, j; + unsigned int i; for (i = 0; i < ARRAY_SIZE(debugfs_regs); i++) { ent = debugfs_create_file(debugfs_regs_names[i], S_IFREG | S_IRUGO, root, adev, debugfs_regs[i]); - if (IS_ERR(ent)) { - for (j = 0; j < i; j++) { - debugfs_remove(adev->debugfs_regs[i]); - adev->debugfs_regs[i] = NULL; - } - return PTR_ERR(ent); - } - - if (!i) + if (!i && !IS_ERR_OR_NULL(ent)) i_size_write(ent->d_inode, adev->rmmio_size); adev->debugfs_regs[i] = ent; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 6e5284e6c028..1e4dd09a5072 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -25,6 +25,7 @@ * Alex Deucher * Jerome Glisse */ +#include <linux/power_supply.h> #include <linux/kthread.h> #include <linux/console.h> #include <linux/slab.h> @@ -61,6 +62,8 @@ MODULE_FIRMWARE("amdgpu/vega10_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/vega12_gpu_info.bin"); MODULE_FIRMWARE("amdgpu/raven_gpu_info.bin"); +MODULE_FIRMWARE("amdgpu/picasso_gpu_info.bin"); +MODULE_FIRMWARE("amdgpu/raven2_gpu_info.bin"); #define AMDGPU_RESUME_MS 2000 @@ -651,73 +654,6 @@ void amdgpu_device_wb_free(struct amdgpu_device *adev, u32 wb) } /** - * amdgpu_device_vram_location - try to find VRAM location - * - * @adev: amdgpu device structure holding all necessary informations - * @mc: memory controller structure holding memory informations - * @base: base address at which to put VRAM - * - * Function will try to place VRAM at base address provided - * as parameter. - */ -void amdgpu_device_vram_location(struct amdgpu_device *adev, - struct amdgpu_gmc *mc, u64 base) -{ - uint64_t limit = (uint64_t)amdgpu_vram_limit << 20; - - mc->vram_start = base; - mc->vram_end = mc->vram_start + mc->mc_vram_size - 1; - if (limit && limit < mc->real_vram_size) - mc->real_vram_size = limit; - dev_info(adev->dev, "VRAM: %lluM 0x%016llX - 0x%016llX (%lluM used)\n", - mc->mc_vram_size >> 20, mc->vram_start, - mc->vram_end, mc->real_vram_size >> 20); -} - -/** - * amdgpu_device_gart_location - try to find GTT location - * - * @adev: amdgpu device structure holding all necessary informations - * @mc: memory controller structure holding memory informations - * - * Function will place try to place GTT before or after VRAM. - * - * If GTT size is bigger than space left then we ajust GTT size. - * Thus function will never fails. - * - * FIXME: when reducing GTT size align new size on power of 2. - */ -void amdgpu_device_gart_location(struct amdgpu_device *adev, - struct amdgpu_gmc *mc) -{ - u64 size_af, size_bf; - - mc->gart_size += adev->pm.smu_prv_buffer_size; - - size_af = adev->gmc.mc_mask - mc->vram_end; - size_bf = mc->vram_start; - if (size_bf > size_af) { - if (mc->gart_size > size_bf) { - dev_warn(adev->dev, "limiting GTT\n"); - mc->gart_size = size_bf; - } - mc->gart_start = 0; - } else { - if (mc->gart_size > size_af) { - dev_warn(adev->dev, "limiting GTT\n"); - mc->gart_size = size_af; - } - /* VCE doesn't like it when BOs cross a 4GB segment, so align - * the GART base on a 4GB boundary as well. - */ - mc->gart_start = ALIGN(mc->vram_end + 1, 0x100000000ULL); - } - mc->gart_end = mc->gart_start + mc->gart_size - 1; - dev_info(adev->dev, "GTT: %lluM 0x%016llX - 0x%016llX\n", - mc->gart_size >> 20, mc->gart_start, mc->gart_end); -} - -/** * amdgpu_device_resize_fb_bar - try to resize FB BAR * * @adev: amdgpu_device pointer @@ -1077,7 +1013,7 @@ static const struct vga_switcheroo_client_ops amdgpu_switcheroo_ops = { /** * amdgpu_device_ip_set_clockgating_state - set the CG state * - * @adev: amdgpu_device pointer + * @dev: amdgpu_device pointer * @block_type: Type of hardware IP (SMU, GFX, UVD, etc.) * @state: clockgating state (gate or ungate) * @@ -1111,7 +1047,7 @@ int amdgpu_device_ip_set_clockgating_state(void *dev, /** * amdgpu_device_ip_set_powergating_state - set the PG state * - * @adev: amdgpu_device pointer + * @dev: amdgpu_device pointer * @block_type: Type of hardware IP (SMU, GFX, UVD, etc.) * @state: powergating state (gate or ungate) * @@ -1222,7 +1158,7 @@ bool amdgpu_device_ip_is_idle(struct amdgpu_device *adev, * amdgpu_device_ip_get_ip_block - get a hw IP pointer * * @adev: amdgpu_device pointer - * @block_type: Type of hardware IP (SMU, GFX, UVD, etc.) + * @type: Type of hardware IP (SMU, GFX, UVD, etc.) * * Returns a pointer to the hardware IP block structure * if it exists for the asic, otherwise NULL. @@ -1398,7 +1334,12 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) chip_name = "vega12"; break; case CHIP_RAVEN: - chip_name = "raven"; + if (adev->rev_id >= 8) + chip_name = "raven2"; + else if (adev->pdev->device == 0x15d8) + chip_name = "picasso"; + else + chip_name = "raven"; break; } @@ -1552,6 +1493,8 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) } adev->powerplay.pp_feature = amdgpu_pp_feature_mask; + if (amdgpu_sriov_vf(adev)) + adev->powerplay.pp_feature &= ~PP_GFXOFF_MASK; for (i = 0; i < adev->num_ip_blocks; i++) { if ((amdgpu_ip_block_mask & (1 << i)) == 0) { @@ -1582,6 +1525,92 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) return 0; } +static int amdgpu_device_ip_hw_init_phase1(struct amdgpu_device *adev) +{ + int i, r; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_blocks[i].status.sw) + continue; + if (adev->ip_blocks[i].status.hw) + continue; + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON || + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) { + r = adev->ip_blocks[i].version->funcs->hw_init(adev); + if (r) { + DRM_ERROR("hw_init of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + adev->ip_blocks[i].status.hw = true; + } + } + + return 0; +} + +static int amdgpu_device_ip_hw_init_phase2(struct amdgpu_device *adev) +{ + int i, r; + + for (i = 0; i < adev->num_ip_blocks; i++) { + if (!adev->ip_blocks[i].status.sw) + continue; + if (adev->ip_blocks[i].status.hw) + continue; + r = adev->ip_blocks[i].version->funcs->hw_init(adev); + if (r) { + DRM_ERROR("hw_init of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + adev->ip_blocks[i].status.hw = true; + } + + return 0; +} + +static int amdgpu_device_fw_loading(struct amdgpu_device *adev) +{ + int r = 0; + int i; + + if (adev->asic_type >= CHIP_VEGA10) { + for (i = 0; i < adev->num_ip_blocks; i++) { + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) { + if (adev->in_gpu_reset || adev->in_suspend) { + if (amdgpu_sriov_vf(adev) && adev->in_gpu_reset) + break; /* sriov gpu reset, psp need to do hw_init before IH because of hw limit */ + r = adev->ip_blocks[i].version->funcs->resume(adev); + if (r) { + DRM_ERROR("resume of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + } else { + r = adev->ip_blocks[i].version->funcs->hw_init(adev); + if (r) { + DRM_ERROR("hw_init of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + } + adev->ip_blocks[i].status.hw = true; + } + } + } + + if (adev->powerplay.pp_funcs->load_firmware) { + r = adev->powerplay.pp_funcs->load_firmware(adev->powerplay.pp_handle); + if (r) { + pr_err("firmware loading failed\n"); + return r; + } + } + + return 0; +} + /** * amdgpu_device_ip_init - run init for hardware IPs * @@ -1638,20 +1667,23 @@ static int amdgpu_device_ip_init(struct amdgpu_device *adev) } } - for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_blocks[i].status.sw) - continue; - if (adev->ip_blocks[i].status.hw) - continue; - r = adev->ip_blocks[i].version->funcs->hw_init((void *)adev); - if (r) { - DRM_ERROR("hw_init of IP block <%s> failed %d\n", - adev->ip_blocks[i].version->funcs->name, r); - return r; - } - adev->ip_blocks[i].status.hw = true; - } + r = amdgpu_ucode_create_bo(adev); /* create ucode bo when sw_init complete*/ + if (r) + return r; + + r = amdgpu_device_ip_hw_init_phase1(adev); + if (r) + return r; + + r = amdgpu_device_fw_loading(adev); + if (r) + return r; + + r = amdgpu_device_ip_hw_init_phase2(adev); + if (r) + return r; + amdgpu_xgmi_add_device(adev); amdgpu_amdkfd_device_init(adev); if (amdgpu_sriov_vf(adev)) @@ -1691,29 +1723,28 @@ static bool amdgpu_device_check_vram_lost(struct amdgpu_device *adev) } /** - * amdgpu_device_ip_late_set_cg_state - late init for clockgating + * amdgpu_device_set_cg_state - set clockgating for amdgpu device * * @adev: amdgpu_device pointer * - * Late initialization pass enabling clockgating for hardware IPs. * The list of all the hardware IPs that make up the asic is walked and the - * set_clockgating_state callbacks are run. This stage is run late - * in the init process. + * set_clockgating_state callbacks are run. + * Late initialization pass enabling clockgating for hardware IPs. + * Fini or suspend, pass disabling clockgating for hardware IPs. * Returns 0 on success, negative error code on failure. */ -static int amdgpu_device_ip_late_set_cg_state(struct amdgpu_device *adev) + +static int amdgpu_device_set_cg_state(struct amdgpu_device *adev, + enum amd_clockgating_state state) { - int i = 0, r; + int i, j, r; if (amdgpu_emu_mode == 1) return 0; - r = amdgpu_ib_ring_tests(adev); - if (r) - DRM_ERROR("ib ring test failed (%d).\n", r); - - for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_blocks[i].status.valid) + for (j = 0; j < adev->num_ip_blocks; j++) { + i = state == AMD_CG_STATE_GATE ? j : adev->num_ip_blocks - j - 1; + if (!adev->ip_blocks[i].status.late_initialized) continue; /* skip CG for VCE/UVD, it's handled specially */ if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD && @@ -1722,7 +1753,7 @@ static int amdgpu_device_ip_late_set_cg_state(struct amdgpu_device *adev) adev->ip_blocks[i].version->funcs->set_clockgating_state) { /* enable clockgating to save power */ r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev, - AMD_CG_STATE_GATE); + state); if (r) { DRM_ERROR("set_clockgating_state(gate) of IP block <%s> failed %d\n", adev->ip_blocks[i].version->funcs->name, r); @@ -1731,17 +1762,35 @@ static int amdgpu_device_ip_late_set_cg_state(struct amdgpu_device *adev) } } - if (adev->powerplay.pp_feature & PP_GFXOFF_MASK) { - /* enable gfx powergating */ - amdgpu_device_ip_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_GFX, - AMD_PG_STATE_GATE); - /* enable gfxoff */ - amdgpu_device_ip_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_SMC, - AMD_PG_STATE_GATE); - } + return 0; +} +static int amdgpu_device_set_pg_state(struct amdgpu_device *adev, enum amd_powergating_state state) +{ + int i, j, r; + + if (amdgpu_emu_mode == 1) + return 0; + + for (j = 0; j < adev->num_ip_blocks; j++) { + i = state == AMD_PG_STATE_GATE ? j : adev->num_ip_blocks - j - 1; + if (!adev->ip_blocks[i].status.late_initialized) + continue; + /* skip CG for VCE/UVD, it's handled specially */ + if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD && + adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE && + adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCN && + adev->ip_blocks[i].version->funcs->set_powergating_state) { + /* enable powergating to save power */ + r = adev->ip_blocks[i].version->funcs->set_powergating_state((void *)adev, + state); + if (r) { + DRM_ERROR("set_powergating_state(gate) of IP block <%s> failed %d\n", + adev->ip_blocks[i].version->funcs->name, r); + return r; + } + } + } return 0; } @@ -1762,7 +1811,7 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev) int i = 0, r; for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_blocks[i].status.valid) + if (!adev->ip_blocks[i].status.hw) continue; if (adev->ip_blocks[i].version->funcs->late_init) { r = adev->ip_blocks[i].version->funcs->late_init((void *)adev); @@ -1771,10 +1820,13 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev) adev->ip_blocks[i].version->funcs->name, r); return r; } - adev->ip_blocks[i].status.late_initialized = true; } + adev->ip_blocks[i].status.late_initialized = true; } + amdgpu_device_set_cg_state(adev, AMD_CG_STATE_GATE); + amdgpu_device_set_pg_state(adev, AMD_PG_STATE_GATE); + queue_delayed_work(system_wq, &adev->late_init_work, msecs_to_jiffies(AMDGPU_RESUME_MS)); @@ -1799,20 +1851,15 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) int i, r; amdgpu_amdkfd_device_fini(adev); + + amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE); + amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE); + /* need to disable SMC first */ for (i = 0; i < adev->num_ip_blocks; i++) { if (!adev->ip_blocks[i].status.hw) continue; - if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC && - adev->ip_blocks[i].version->funcs->set_clockgating_state) { - /* ungate blocks before hw fini so that we can shutdown the blocks safely */ - r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev, - AMD_CG_STATE_UNGATE); - if (r) { - DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n", - adev->ip_blocks[i].version->funcs->name, r); - return r; - } + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_SMC) { r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev); /* XXX handle errors */ if (r) { @@ -1828,20 +1875,6 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) if (!adev->ip_blocks[i].status.hw) continue; - if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD && - adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE && - adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCN && - adev->ip_blocks[i].version->funcs->set_clockgating_state) { - /* ungate blocks before hw fini so that we can shutdown the blocks safely */ - r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev, - AMD_CG_STATE_UNGATE); - if (r) { - DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n", - adev->ip_blocks[i].version->funcs->name, r); - return r; - } - } - r = adev->ip_blocks[i].version->funcs->hw_fini((void *)adev); /* XXX handle errors */ if (r) { @@ -1858,6 +1891,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) continue; if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC) { + amdgpu_ucode_free_bo(adev); amdgpu_free_static_csa(adev); amdgpu_device_wb_fini(adev); amdgpu_device_vram_scratch_fini(adev); @@ -1888,24 +1922,78 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) return 0; } +static int amdgpu_device_enable_mgpu_fan_boost(void) +{ + struct amdgpu_gpu_instance *gpu_ins; + struct amdgpu_device *adev; + int i, ret = 0; + + mutex_lock(&mgpu_info.mutex); + + /* + * MGPU fan boost feature should be enabled + * only when there are two or more dGPUs in + * the system + */ + if (mgpu_info.num_dgpu < 2) + goto out; + + for (i = 0; i < mgpu_info.num_dgpu; i++) { + gpu_ins = &(mgpu_info.gpu_ins[i]); + adev = gpu_ins->adev; + if (!(adev->flags & AMD_IS_APU) && + !gpu_ins->mgpu_fan_enabled && + adev->powerplay.pp_funcs && + adev->powerplay.pp_funcs->enable_mgpu_fan_boost) { + ret = amdgpu_dpm_enable_mgpu_fan_boost(adev); + if (ret) + break; + + gpu_ins->mgpu_fan_enabled = 1; + } + } + +out: + mutex_unlock(&mgpu_info.mutex); + + return ret; +} + /** - * amdgpu_device_ip_late_init_func_handler - work handler for clockgating + * amdgpu_device_ip_late_init_func_handler - work handler for ib test * - * @work: work_struct - * - * Work handler for amdgpu_device_ip_late_set_cg_state. We put the - * clockgating setup into a worker thread to speed up driver init and - * resume from suspend. + * @work: work_struct. */ static void amdgpu_device_ip_late_init_func_handler(struct work_struct *work) { struct amdgpu_device *adev = container_of(work, struct amdgpu_device, late_init_work.work); - amdgpu_device_ip_late_set_cg_state(adev); + int r; + + r = amdgpu_ib_ring_tests(adev); + if (r) + DRM_ERROR("ib ring test failed (%d).\n", r); + + r = amdgpu_device_enable_mgpu_fan_boost(); + if (r) + DRM_ERROR("enable mgpu fan boost failed (%d).\n", r); +} + +static void amdgpu_device_delay_enable_gfx_off(struct work_struct *work) +{ + struct amdgpu_device *adev = + container_of(work, struct amdgpu_device, gfx.gfx_off_delay_work.work); + + mutex_lock(&adev->gfx.gfx_off_mutex); + if (!adev->gfx.gfx_off_state && !adev->gfx.gfx_off_req_count) { + if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, true)) + adev->gfx.gfx_off_state = true; + } + mutex_unlock(&adev->gfx.gfx_off_mutex); } /** - * amdgpu_device_ip_suspend - run suspend for hardware IPs + * amdgpu_device_ip_suspend_phase1 - run suspend for hardware IPs (phase 1) * * @adev: amdgpu_device pointer * @@ -1915,39 +2003,52 @@ static void amdgpu_device_ip_late_init_func_handler(struct work_struct *work) * in each IP into a state suitable for suspend. * Returns 0 on success, negative error code on failure. */ -int amdgpu_device_ip_suspend(struct amdgpu_device *adev) +static int amdgpu_device_ip_suspend_phase1(struct amdgpu_device *adev) { int i, r; - if (amdgpu_sriov_vf(adev)) - amdgpu_virt_request_full_gpu(adev, false); - - /* ungate SMC block powergating */ - if (adev->powerplay.pp_feature & PP_GFXOFF_MASK) - amdgpu_device_ip_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_SMC, - AMD_CG_STATE_UNGATE); - - /* ungate SMC block first */ - r = amdgpu_device_ip_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_SMC, - AMD_CG_STATE_UNGATE); - if (r) { - DRM_ERROR("set_clockgating_state(ungate) SMC failed %d\n", r); - } + amdgpu_device_set_pg_state(adev, AMD_PG_STATE_UNGATE); + amdgpu_device_set_cg_state(adev, AMD_CG_STATE_UNGATE); for (i = adev->num_ip_blocks - 1; i >= 0; i--) { if (!adev->ip_blocks[i].status.valid) continue; - /* ungate blocks so that suspend can properly shut them down */ - if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_SMC && - adev->ip_blocks[i].version->funcs->set_clockgating_state) { - r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev, - AMD_CG_STATE_UNGATE); + /* displays are handled separately */ + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) { + /* XXX handle errors */ + r = adev->ip_blocks[i].version->funcs->suspend(adev); + /* XXX handle errors */ if (r) { - DRM_ERROR("set_clockgating_state(ungate) of IP block <%s> failed %d\n", + DRM_ERROR("suspend of IP block <%s> failed %d\n", adev->ip_blocks[i].version->funcs->name, r); } } + } + + return 0; +} + +/** + * amdgpu_device_ip_suspend_phase2 - run suspend for hardware IPs (phase 2) + * + * @adev: amdgpu_device pointer + * + * Main suspend function for hardware IPs. The list of all the hardware + * IPs that make up the asic is walked, clockgating is disabled and the + * suspend callbacks are run. suspend puts the hardware and software state + * in each IP into a state suitable for suspend. + * Returns 0 on success, negative error code on failure. + */ +static int amdgpu_device_ip_suspend_phase2(struct amdgpu_device *adev) +{ + int i, r; + + for (i = adev->num_ip_blocks - 1; i >= 0; i--) { + if (!adev->ip_blocks[i].status.valid) + continue; + /* displays are handled in phase1 */ + if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_DCE) + continue; /* XXX handle errors */ r = adev->ip_blocks[i].version->funcs->suspend(adev); /* XXX handle errors */ @@ -1957,10 +2058,36 @@ int amdgpu_device_ip_suspend(struct amdgpu_device *adev) } } + return 0; +} + +/** + * amdgpu_device_ip_suspend - run suspend for hardware IPs + * + * @adev: amdgpu_device pointer + * + * Main suspend function for hardware IPs. The list of all the hardware + * IPs that make up the asic is walked, clockgating is disabled and the + * suspend callbacks are run. suspend puts the hardware and software state + * in each IP into a state suitable for suspend. + * Returns 0 on success, negative error code on failure. + */ +int amdgpu_device_ip_suspend(struct amdgpu_device *adev) +{ + int r; + + if (amdgpu_sriov_vf(adev)) + amdgpu_virt_request_full_gpu(adev, false); + + r = amdgpu_device_ip_suspend_phase1(adev); + if (r) + return r; + r = amdgpu_device_ip_suspend_phase2(adev); + if (amdgpu_sriov_vf(adev)) amdgpu_virt_release_full_gpu(adev, false); - return 0; + return r; } static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) @@ -1970,6 +2097,7 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) static enum amd_ip_block_type ip_order[] = { AMD_IP_BLOCK_TYPE_GMC, AMD_IP_BLOCK_TYPE_COMMON, + AMD_IP_BLOCK_TYPE_PSP, AMD_IP_BLOCK_TYPE_IH, }; @@ -1985,7 +2113,7 @@ static int amdgpu_device_ip_reinit_early_sriov(struct amdgpu_device *adev) continue; r = block->version->funcs->hw_init(adev); - DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"successed"); + DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); if (r) return r; } @@ -2000,7 +2128,6 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev) static enum amd_ip_block_type ip_order[] = { AMD_IP_BLOCK_TYPE_SMC, - AMD_IP_BLOCK_TYPE_PSP, AMD_IP_BLOCK_TYPE_DCE, AMD_IP_BLOCK_TYPE_GFX, AMD_IP_BLOCK_TYPE_SDMA, @@ -2020,7 +2147,7 @@ static int amdgpu_device_ip_reinit_late_sriov(struct amdgpu_device *adev) continue; r = block->version->funcs->hw_init(adev); - DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"successed"); + DRM_INFO("RE-INIT: %s %s\n", block->version->funcs->name, r?"failed":"succeeded"); if (r) return r; } @@ -2085,7 +2212,8 @@ static int amdgpu_device_ip_resume_phase2(struct amdgpu_device *adev) continue; if (adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_COMMON || adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_GMC || - adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH) + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_IH || + adev->ip_blocks[i].version->type == AMD_IP_BLOCK_TYPE_PSP) continue; r = adev->ip_blocks[i].version->funcs->resume(adev); if (r) { @@ -2117,6 +2245,11 @@ static int amdgpu_device_ip_resume(struct amdgpu_device *adev) r = amdgpu_device_ip_resume_phase1(adev); if (r) return r; + + r = amdgpu_device_fw_loading(adev); + if (r) + return r; + r = amdgpu_device_ip_resume_phase2(adev); return r; @@ -2210,7 +2343,7 @@ bool amdgpu_device_has_dc_support(struct amdgpu_device *adev) * amdgpu_device_init - initialize the driver * * @adev: amdgpu_device pointer - * @pdev: drm dev pointer + * @ddev: drm dev pointer * @pdev: pci dev pointer * @flags: driver flags * @@ -2242,7 +2375,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->mman.buffer_funcs = NULL; adev->mman.buffer_funcs_ring = NULL; adev->vm_manager.vm_pte_funcs = NULL; - adev->vm_manager.vm_pte_num_rings = 0; + adev->vm_manager.vm_pte_num_rqs = 0; adev->gmc.gmc_funcs = NULL; adev->fence_context = dma_fence_context_alloc(AMDGPU_MAX_RINGS); bitmap_zero(adev->gfx.pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); @@ -2274,6 +2407,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, mutex_init(&adev->gfx.gpu_clock_mutex); mutex_init(&adev->srbm_mutex); mutex_init(&adev->gfx.pipe_reserve_mutex); + mutex_init(&adev->gfx.gfx_off_mutex); mutex_init(&adev->grbm_idx_mutex); mutex_init(&adev->mn_lock); mutex_init(&adev->virt.vf_errors.lock); @@ -2300,6 +2434,11 @@ int amdgpu_device_init(struct amdgpu_device *adev, INIT_DELAYED_WORK(&adev->late_init_work, amdgpu_device_ip_late_init_func_handler); + INIT_DELAYED_WORK(&adev->gfx.gfx_off_delay_work, + amdgpu_device_delay_enable_gfx_off); + + adev->gfx.gfx_off_req_count = 1; + adev->pm.ac_power = power_supply_is_system_supplied() > 0 ? true : false; /* Registers mapping */ /* TODO: block userspace mapping of io register */ @@ -2581,8 +2720,9 @@ void amdgpu_device_fini(struct amdgpu_device *adev) /** * amdgpu_device_suspend - initiate device suspend * - * @pdev: drm dev pointer - * @state: suspend state + * @dev: drm dev pointer + * @suspend: suspend state + * @fbcon : notify the fbdev of suspend * * Puts the hw in the suspend state (all asics). * Returns 0 for success or an error on failure. @@ -2604,8 +2744,14 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; + adev->in_suspend = true; drm_kms_helper_poll_disable(dev); + if (fbcon) + amdgpu_fbdev_set_suspend(adev, 1); + + cancel_delayed_work_sync(&adev->late_init_work); + if (!amdgpu_device_has_dc_support(adev)) { /* turn off display hw */ drm_modeset_lock_all(dev); @@ -2613,44 +2759,46 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF); } drm_modeset_unlock_all(dev); - } - - amdgpu_amdkfd_suspend(adev); - - /* unpin the front buffers and cursors */ - list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - struct drm_framebuffer *fb = crtc->primary->fb; - struct amdgpu_bo *robj; - - if (amdgpu_crtc->cursor_bo) { - struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo); - r = amdgpu_bo_reserve(aobj, true); - if (r == 0) { - amdgpu_bo_unpin(aobj); - amdgpu_bo_unreserve(aobj); + /* unpin the front buffers and cursors */ + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + struct drm_framebuffer *fb = crtc->primary->fb; + struct amdgpu_bo *robj; + + if (amdgpu_crtc->cursor_bo) { + struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo); + r = amdgpu_bo_reserve(aobj, true); + if (r == 0) { + amdgpu_bo_unpin(aobj); + amdgpu_bo_unreserve(aobj); + } } - } - if (fb == NULL || fb->obj[0] == NULL) { - continue; - } - robj = gem_to_amdgpu_bo(fb->obj[0]); - /* don't unpin kernel fb objects */ - if (!amdgpu_fbdev_robj_is_fb(adev, robj)) { - r = amdgpu_bo_reserve(robj, true); - if (r == 0) { - amdgpu_bo_unpin(robj); - amdgpu_bo_unreserve(robj); + if (fb == NULL || fb->obj[0] == NULL) { + continue; + } + robj = gem_to_amdgpu_bo(fb->obj[0]); + /* don't unpin kernel fb objects */ + if (!amdgpu_fbdev_robj_is_fb(adev, robj)) { + r = amdgpu_bo_reserve(robj, true); + if (r == 0) { + amdgpu_bo_unpin(robj); + amdgpu_bo_unreserve(robj); + } } } } + + amdgpu_amdkfd_suspend(adev); + + r = amdgpu_device_ip_suspend_phase1(adev); + /* evict vram memory */ amdgpu_bo_evict_vram(adev); amdgpu_fence_driver_suspend(adev); - r = amdgpu_device_ip_suspend(adev); + r = amdgpu_device_ip_suspend_phase2(adev); /* evict remaining vram memory * This second call to evict vram is to evict the gart page table @@ -2669,18 +2817,15 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) DRM_ERROR("amdgpu asic reset failed\n"); } - if (fbcon) { - console_lock(); - amdgpu_fbdev_set_suspend(adev, 1); - console_unlock(); - } return 0; } /** * amdgpu_device_resume - initiate device resume * - * @pdev: drm dev pointer + * @dev: drm dev pointer + * @resume: resume state + * @fbcon : notify the fbdev of resume * * Bring the hw back to operating state (all asics). * Returns 0 for success or an error on failure. @@ -2696,15 +2841,12 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) if (dev->switch_power_state == DRM_SWITCH_POWER_OFF) return 0; - if (fbcon) - console_lock(); - if (resume) { pci_set_power_state(dev->pdev, PCI_D0); pci_restore_state(dev->pdev); r = pci_enable_device(dev->pdev); if (r) - goto unlock; + return r; } /* post card */ @@ -2717,29 +2859,30 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) r = amdgpu_device_ip_resume(adev); if (r) { DRM_ERROR("amdgpu_device_ip_resume failed (%d).\n", r); - goto unlock; + return r; } amdgpu_fence_driver_resume(adev); r = amdgpu_device_ip_late_init(adev); if (r) - goto unlock; - - /* pin cursors */ - list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { - struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - - if (amdgpu_crtc->cursor_bo) { - struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo); - r = amdgpu_bo_reserve(aobj, true); - if (r == 0) { - r = amdgpu_bo_pin(aobj, - AMDGPU_GEM_DOMAIN_VRAM, - &amdgpu_crtc->cursor_addr); - if (r != 0) - DRM_ERROR("Failed to pin cursor BO (%d)\n", r); - amdgpu_bo_unreserve(aobj); + return r; + + if (!amdgpu_device_has_dc_support(adev)) { + /* pin cursors */ + list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); + + if (amdgpu_crtc->cursor_bo) { + struct amdgpu_bo *aobj = gem_to_amdgpu_bo(amdgpu_crtc->cursor_bo); + r = amdgpu_bo_reserve(aobj, true); + if (r == 0) { + r = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM); + if (r != 0) + DRM_ERROR("Failed to pin cursor BO (%d)\n", r); + amdgpu_crtc->cursor_addr = amdgpu_bo_gpu_offset(aobj); + amdgpu_bo_unreserve(aobj); + } } } } @@ -2747,6 +2890,9 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) if (r) return r; + /* Make sure IB tests flushed */ + flush_delayed_work(&adev->late_init_work); + /* blat the mode back in */ if (fbcon) { if (!amdgpu_device_has_dc_support(adev)) { @@ -2760,6 +2906,7 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) } drm_modeset_unlock_all(dev); } + amdgpu_fbdev_set_suspend(adev, 0); } drm_kms_helper_poll_enable(dev); @@ -2783,15 +2930,9 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) #ifdef CONFIG_PM dev->dev->power.disable_depth--; #endif + adev->in_suspend = false; - if (fbcon) - amdgpu_fbdev_set_suspend(adev, 0); - -unlock: - if (fbcon) - console_unlock(); - - return r; + return 0; } /** @@ -2949,117 +3090,64 @@ static int amdgpu_device_ip_post_soft_reset(struct amdgpu_device *adev) } /** - * amdgpu_device_recover_vram_from_shadow - restore shadowed VRAM buffers - * - * @adev: amdgpu_device pointer - * @ring: amdgpu_ring for the engine handling the buffer operations - * @bo: amdgpu_bo buffer whose shadow is being restored - * @fence: dma_fence associated with the operation - * - * Restores the VRAM buffer contents from the shadow in GTT. Used to - * restore things like GPUVM page tables after a GPU reset where - * the contents of VRAM might be lost. - * Returns 0 on success, negative error code on failure. - */ -static int amdgpu_device_recover_vram_from_shadow(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_bo *bo, - struct dma_fence **fence) -{ - uint32_t domain; - int r; - - if (!bo->shadow) - return 0; - - r = amdgpu_bo_reserve(bo, true); - if (r) - return r; - domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); - /* if bo has been evicted, then no need to recover */ - if (domain == AMDGPU_GEM_DOMAIN_VRAM) { - r = amdgpu_bo_validate(bo->shadow); - if (r) { - DRM_ERROR("bo validate failed!\n"); - goto err; - } - - r = amdgpu_bo_restore_from_shadow(adev, ring, bo, - NULL, fence, true); - if (r) { - DRM_ERROR("recover page table failed!\n"); - goto err; - } - } -err: - amdgpu_bo_unreserve(bo); - return r; -} - -/** - * amdgpu_device_handle_vram_lost - Handle the loss of VRAM contents + * amdgpu_device_recover_vram - Recover some VRAM contents * * @adev: amdgpu_device pointer * * Restores the contents of VRAM buffers from the shadows in GTT. Used to * restore things like GPUVM page tables after a GPU reset where * the contents of VRAM might be lost. - * Returns 0 on success, 1 on failure. + * + * Returns: + * 0 on success, negative error code on failure. */ -static int amdgpu_device_handle_vram_lost(struct amdgpu_device *adev) +static int amdgpu_device_recover_vram(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; - struct amdgpu_bo *bo, *tmp; struct dma_fence *fence = NULL, *next = NULL; - long r = 1; - int i = 0; - long tmo; + struct amdgpu_bo *shadow; + long r = 1, tmo; if (amdgpu_sriov_runtime(adev)) - tmo = msecs_to_jiffies(amdgpu_lockup_timeout); + tmo = msecs_to_jiffies(8000); else tmo = msecs_to_jiffies(100); DRM_INFO("recover vram bo from shadow start\n"); mutex_lock(&adev->shadow_list_lock); - list_for_each_entry_safe(bo, tmp, &adev->shadow_list, shadow_list) { - next = NULL; - amdgpu_device_recover_vram_from_shadow(adev, ring, bo, &next); + list_for_each_entry(shadow, &adev->shadow_list, shadow_list) { + + /* No need to recover an evicted BO */ + if (shadow->tbo.mem.mem_type != TTM_PL_TT || + shadow->parent->tbo.mem.mem_type != TTM_PL_VRAM) + continue; + + r = amdgpu_bo_restore_shadow(shadow, &next); + if (r) + break; + if (fence) { r = dma_fence_wait_timeout(fence, false, tmo); - if (r == 0) - pr_err("wait fence %p[%d] timeout\n", fence, i); - else if (r < 0) - pr_err("wait fence %p[%d] interrupted\n", fence, i); - if (r < 1) { - dma_fence_put(fence); - fence = next; + dma_fence_put(fence); + fence = next; + if (r <= 0) break; - } - i++; + } else { + fence = next; } - - dma_fence_put(fence); - fence = next; } mutex_unlock(&adev->shadow_list_lock); - if (fence) { - r = dma_fence_wait_timeout(fence, false, tmo); - if (r == 0) - pr_err("wait fence %p[%d] timeout\n", fence, i); - else if (r < 0) - pr_err("wait fence %p[%d] interrupted\n", fence, i); - - } + if (fence) + tmo = dma_fence_wait_timeout(fence, false, tmo); dma_fence_put(fence); - if (r > 0) - DRM_INFO("recover vram bo from shadow done\n"); - else + if (r <= 0 || tmo <= 0) { DRM_ERROR("recover vram bo from shadow failed\n"); + return -EIO; + } - return (r > 0) ? 0 : 1; + DRM_INFO("recover vram bo from shadow done\n"); + return 0; } /** @@ -3068,7 +3156,7 @@ static int amdgpu_device_handle_vram_lost(struct amdgpu_device *adev) * @adev: amdgpu device pointer * * attempt to do soft-reset or full-reset and reinitialize Asic - * return 0 means successed otherwise failed + * return 0 means succeeded otherwise failed */ static int amdgpu_device_reset(struct amdgpu_device *adev) { @@ -3112,6 +3200,10 @@ retry: if (r) goto out; + r = amdgpu_device_fw_loading(adev); + if (r) + return r; + r = amdgpu_device_ip_resume_phase2(adev); if (r) goto out; @@ -3133,8 +3225,8 @@ out: } } - if (!r && ((need_full_reset && !(adev->flags & AMD_IS_APU)) || vram_lost)) - r = amdgpu_device_handle_vram_lost(adev); + if (!r) + r = amdgpu_device_recover_vram(adev); return r; } @@ -3143,9 +3235,10 @@ out: * amdgpu_device_reset_sriov - reset ASIC for SR-IOV vf * * @adev: amdgpu device pointer + * @from_hypervisor: request from hypervisor * * do VF FLR and reinitialize Asic - * return 0 means successed otherwise failed + * return 0 means succeeded otherwise failed */ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, bool from_hypervisor) @@ -3167,6 +3260,10 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, /* we need recover gart prior to run SMC/CP/SDMA resume */ amdgpu_gtt_mgr_recover(&adev->mman.bdev.man[TTM_PL_TT]); + r = amdgpu_device_fw_loading(adev); + if (r) + return r; + /* now we are okay to resume SMC/CP/SDMA */ r = amdgpu_device_ip_reinit_late_sriov(adev); if (r) @@ -3179,44 +3276,59 @@ error: amdgpu_virt_release_full_gpu(adev, true); if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) { atomic_inc(&adev->vram_lost_counter); - r = amdgpu_device_handle_vram_lost(adev); + r = amdgpu_device_recover_vram(adev); } return r; } /** + * amdgpu_device_should_recover_gpu - check if we should try GPU recovery + * + * @adev: amdgpu device pointer + * + * Check amdgpu_gpu_recovery and SRIOV status to see if we should try to recover + * a hung GPU. + */ +bool amdgpu_device_should_recover_gpu(struct amdgpu_device *adev) +{ + if (!amdgpu_device_ip_check_soft_reset(adev)) { + DRM_INFO("Timeout, but no hardware hang detected.\n"); + return false; + } + + if (amdgpu_gpu_recovery == 0 || (amdgpu_gpu_recovery == -1 && + !amdgpu_sriov_vf(adev))) { + DRM_INFO("GPU recovery disabled.\n"); + return false; + } + + return true; +} + +/** * amdgpu_device_gpu_recover - reset the asic and recover scheduler * * @adev: amdgpu device pointer * @job: which job trigger hang - * @force forces reset regardless of amdgpu_gpu_recovery * * Attempt to reset the GPU if it has hung (all asics). * Returns 0 for success or an error on failure. */ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, - struct amdgpu_job *job, bool force) + struct amdgpu_job *job) { int i, r, resched; - if (!force && !amdgpu_device_ip_check_soft_reset(adev)) { - DRM_INFO("No hardware hang detected. Did some blocks stall?\n"); - return 0; - } - - if (!force && (amdgpu_gpu_recovery == 0 || - (amdgpu_gpu_recovery == -1 && !amdgpu_sriov_vf(adev)))) { - DRM_INFO("GPU recovery disabled.\n"); - return 0; - } - dev_info(adev->dev, "GPU reset begin!\n"); mutex_lock(&adev->lock_reset); atomic_inc(&adev->gpu_reset_counter); adev->in_gpu_reset = 1; + /* Block kfd */ + amdgpu_amdkfd_pre_reset(adev); + /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); @@ -3229,10 +3341,10 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, kthread_park(ring->sched.thread); - if (job && job->ring->idx != i) + if (job && job->base.sched == &ring->sched) continue; - drm_sched_hw_job_reset(&ring->sched, &job->base); + drm_sched_hw_job_reset(&ring->sched, job ? &job->base : NULL); /* after all hw jobs are reset, hw fence is meaningless, so force_completion */ amdgpu_fence_driver_force_completion(ring); @@ -3253,7 +3365,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, * or all rings (in the case @job is NULL) * after above amdgpu_reset accomplished */ - if ((!job || job->ring->idx == i) && !r) + if ((!job || job->base.sched == &ring->sched) && !r) drm_sched_job_recovery(&ring->sched); kthread_unpark(ring->sched.thread); @@ -3270,9 +3382,11 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, dev_info(adev->dev, "GPU reset(%d) failed\n", atomic_read(&adev->gpu_reset_counter)); amdgpu_vf_error_put(adev, AMDGIM_ERROR_VF_GPU_RESET_FAIL, 0, r); } else { - dev_info(adev->dev, "GPU reset(%d) successed!\n",atomic_read(&adev->gpu_reset_counter)); + dev_info(adev->dev, "GPU reset(%d) succeeded!\n",atomic_read(&adev->gpu_reset_counter)); } + /*unlock kfd */ + amdgpu_amdkfd_post_reset(adev); amdgpu_vf_error_trans_all(adev); adev->in_gpu_reset = 0; mutex_unlock(&adev->lock_reset); @@ -3290,8 +3404,9 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, */ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) { - u32 mask; - int ret; + struct pci_dev *pdev; + enum pci_bus_speed speed_cap; + enum pcie_link_width link_width; if (amdgpu_pcie_gen_cap) adev->pm.pcie_gen_mask = amdgpu_pcie_gen_cap; @@ -3309,27 +3424,61 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) } if (adev->pm.pcie_gen_mask == 0) { - ret = drm_pcie_get_speed_cap_mask(adev->ddev, &mask); - if (!ret) { - adev->pm.pcie_gen_mask = (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 | + /* asic caps */ + pdev = adev->pdev; + speed_cap = pcie_get_speed_cap(pdev); + if (speed_cap == PCI_SPEED_UNKNOWN) { + adev->pm.pcie_gen_mask |= (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 | CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2 | CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN3); - - if (mask & DRM_PCIE_SPEED_25) - adev->pm.pcie_gen_mask |= CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1; - if (mask & DRM_PCIE_SPEED_50) - adev->pm.pcie_gen_mask |= CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2; - if (mask & DRM_PCIE_SPEED_80) - adev->pm.pcie_gen_mask |= CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3; } else { - adev->pm.pcie_gen_mask = AMDGPU_DEFAULT_PCIE_GEN_MASK; + if (speed_cap == PCIE_SPEED_16_0GT) + adev->pm.pcie_gen_mask |= (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 | + CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2 | + CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN3 | + CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN4); + else if (speed_cap == PCIE_SPEED_8_0GT) + adev->pm.pcie_gen_mask |= (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 | + CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2 | + CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN3); + else if (speed_cap == PCIE_SPEED_5_0GT) + adev->pm.pcie_gen_mask |= (CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1 | + CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN2); + else + adev->pm.pcie_gen_mask |= CAIL_ASIC_PCIE_LINK_SPEED_SUPPORT_GEN1; + } + /* platform caps */ + pdev = adev->ddev->pdev->bus->self; + speed_cap = pcie_get_speed_cap(pdev); + if (speed_cap == PCI_SPEED_UNKNOWN) { + adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2); + } else { + if (speed_cap == PCIE_SPEED_16_0GT) + adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN4); + else if (speed_cap == PCIE_SPEED_8_0GT) + adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3); + else if (speed_cap == PCIE_SPEED_5_0GT) + adev->pm.pcie_gen_mask |= (CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1 | + CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2); + else + adev->pm.pcie_gen_mask |= CAIL_PCIE_LINK_SPEED_SUPPORT_GEN1; + } } if (adev->pm.pcie_mlw_mask == 0) { - ret = drm_pcie_get_max_link_width(adev->ddev, &mask); - if (!ret) { - switch (mask) { - case 32: + pdev = adev->ddev->pdev->bus->self; + link_width = pcie_get_width_cap(pdev); + if (link_width == PCIE_LNK_WIDTH_UNKNOWN) { + adev->pm.pcie_mlw_mask |= AMDGPU_DEFAULT_PCIE_MLW_MASK; + } else { + switch (link_width) { + case PCIE_LNK_X32: adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X32 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 | @@ -3338,7 +3487,7 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X1); break; - case 16: + case PCIE_LNK_X16: adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X16 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 | @@ -3346,36 +3495,34 @@ static void amdgpu_device_get_pcie_info(struct amdgpu_device *adev) CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X1); break; - case 12: + case PCIE_LNK_X12: adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X12 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X1); break; - case 8: + case PCIE_LNK_X8: adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X8 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X1); break; - case 4: + case PCIE_LNK_X4: adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X4 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X1); break; - case 2: + case PCIE_LNK_X2: adev->pm.pcie_mlw_mask = (CAIL_PCIE_LINK_WIDTH_SUPPORT_X2 | CAIL_PCIE_LINK_WIDTH_SUPPORT_X1); break; - case 1: + case PCIE_LNK_X1: adev->pm.pcie_mlw_mask = CAIL_PCIE_LINK_WIDTH_SUPPORT_X1; break; default: break; } - } else { - adev->pm.pcie_mlw_mask = AMDGPU_DEFAULT_PCIE_MLW_MASK; } } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 76ee8e04ff11..6748cd7fc129 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -157,7 +157,6 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, struct amdgpu_bo *new_abo; unsigned long flags; u64 tiling_flags; - u64 base; int i, r; work = kzalloc(sizeof *work, GFP_KERNEL); @@ -189,12 +188,18 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, goto cleanup; } - r = amdgpu_bo_pin(new_abo, amdgpu_display_supported_domains(adev), &base); + r = amdgpu_bo_pin(new_abo, amdgpu_display_supported_domains(adev)); if (unlikely(r != 0)) { DRM_ERROR("failed to pin new abo buffer before flip\n"); goto unreserve; } + r = amdgpu_ttm_alloc_gart(&new_abo->tbo); + if (unlikely(r != 0)) { + DRM_ERROR("%p bind failed\n", new_abo); + goto unpin; + } + r = reservation_object_get_fences_rcu(new_abo->tbo.resv, &work->excl, &work->shared_count, &work->shared); @@ -206,7 +211,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, amdgpu_bo_get_tiling_flags(new_abo, &tiling_flags); amdgpu_bo_unreserve(new_abo); - work->base = base; + work->base = amdgpu_bo_gpu_offset(new_abo); work->target_vblank = target - (uint32_t)drm_crtc_vblank_count(crtc) + amdgpu_get_vblank_counter_kms(dev, work->crtc_id); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h index f66e3e3fef0a..06b922fe0d42 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h @@ -23,6 +23,21 @@ #ifndef __AMDGPU_DISPLAY_H__ #define __AMDGPU_DISPLAY_H__ +#define amdgpu_display_vblank_get_counter(adev, crtc) (adev)->mode_info.funcs->vblank_get_counter((adev), (crtc)) +#define amdgpu_display_backlight_set_level(adev, e, l) (adev)->mode_info.funcs->backlight_set_level((e), (l)) +#define amdgpu_display_backlight_get_level(adev, e) (adev)->mode_info.funcs->backlight_get_level((e)) +#define amdgpu_display_hpd_sense(adev, h) (adev)->mode_info.funcs->hpd_sense((adev), (h)) +#define amdgpu_display_hpd_set_polarity(adev, h) (adev)->mode_info.funcs->hpd_set_polarity((adev), (h)) +#define amdgpu_display_hpd_get_gpio_reg(adev) (adev)->mode_info.funcs->hpd_get_gpio_reg((adev)) +#define amdgpu_display_bandwidth_update(adev) (adev)->mode_info.funcs->bandwidth_update((adev)) +#define amdgpu_display_page_flip(adev, crtc, base, async) (adev)->mode_info.funcs->page_flip((adev), (crtc), (base), (async)) +#define amdgpu_display_page_flip_get_scanoutpos(adev, crtc, vbl, pos) (adev)->mode_info.funcs->page_flip_get_scanoutpos((adev), (crtc), (vbl), (pos)) +#define amdgpu_display_add_encoder(adev, e, s, c) (adev)->mode_info.funcs->add_encoder((adev), (e), (s), (c)) +#define amdgpu_display_add_connector(adev, ci, sd, ct, ib, coi, h, r) (adev)->mode_info.funcs->add_connector((adev), (ci), (sd), (ct), (ib), (coi), (h), (r)) + +int amdgpu_display_freesync_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +void amdgpu_display_update_priority(struct amdgpu_device *adev); uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev); struct drm_framebuffer * amdgpu_display_user_framebuffer_create(struct drm_device *dev, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c index 77ad59ade85c..1c4595562f8f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c @@ -28,6 +28,7 @@ #include "amdgpu_i2c.h" #include "amdgpu_dpm.h" #include "atom.h" +#include "amd_pcie.h" void amdgpu_dpm_print_class_info(u32 class, u32 class2) { @@ -936,9 +937,11 @@ enum amdgpu_pcie_gen amdgpu_get_pcie_gen_support(struct amdgpu_device *adev, case AMDGPU_PCIE_GEN3: return AMDGPU_PCIE_GEN3; default: - if ((sys_mask & DRM_PCIE_SPEED_80) && (default_gen == AMDGPU_PCIE_GEN3)) + if ((sys_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN3) && + (default_gen == AMDGPU_PCIE_GEN3)) return AMDGPU_PCIE_GEN3; - else if ((sys_mask & DRM_PCIE_SPEED_50) && (default_gen == AMDGPU_PCIE_GEN2)) + else if ((sys_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_GEN2) && + (default_gen == AMDGPU_PCIE_GEN2)) return AMDGPU_PCIE_GEN2; else return AMDGPU_PCIE_GEN1; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index dd6203a0a6b7..f972cd156795 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -278,6 +278,9 @@ enum amdgpu_pcie_gen { #define amdgpu_dpm_get_fan_speed_rpm(adev, s) \ ((adev)->powerplay.pp_funcs->get_fan_speed_rpm)((adev)->powerplay.pp_handle, (s)) +#define amdgpu_dpm_set_fan_speed_rpm(adev, s) \ + ((adev)->powerplay.pp_funcs->set_fan_speed_rpm)((adev)->powerplay.pp_handle, (s)) + #define amdgpu_dpm_get_sclk(adev, l) \ ((adev)->powerplay.pp_funcs->get_sclk((adev)->powerplay.pp_handle, (l))) @@ -287,12 +290,6 @@ enum amdgpu_pcie_gen { #define amdgpu_dpm_force_performance_level(adev, l) \ ((adev)->powerplay.pp_funcs->force_performance_level((adev)->powerplay.pp_handle, (l))) -#define amdgpu_dpm_powergate_uvd(adev, g) \ - ((adev)->powerplay.pp_funcs->powergate_uvd((adev)->powerplay.pp_handle, (g))) - -#define amdgpu_dpm_powergate_vce(adev, g) \ - ((adev)->powerplay.pp_funcs->powergate_vce((adev)->powerplay.pp_handle, (g))) - #define amdgpu_dpm_get_current_power_state(adev) \ ((adev)->powerplay.pp_funcs->get_current_power_state((adev)->powerplay.pp_handle)) @@ -347,6 +344,10 @@ enum amdgpu_pcie_gen { ((adev)->powerplay.pp_funcs->set_clockgating_by_smu(\ (adev)->powerplay.pp_handle, msg_id)) +#define amdgpu_dpm_set_powergating_by_smu(adev, block_type, gate) \ + ((adev)->powerplay.pp_funcs->set_powergating_by_smu(\ + (adev)->powerplay.pp_handle, block_type, gate)) + #define amdgpu_dpm_get_power_profile_mode(adev, buf) \ ((adev)->powerplay.pp_funcs->get_power_profile_mode(\ (adev)->powerplay.pp_handle, buf)) @@ -359,9 +360,9 @@ enum amdgpu_pcie_gen { ((adev)->powerplay.pp_funcs->odn_edit_dpm_table(\ (adev)->powerplay.pp_handle, type, parameter, size)) -#define amdgpu_dpm_set_mmhub_powergating_by_smu(adev) \ - ((adev)->powerplay.pp_funcs->set_mmhub_powergating_by_smu( \ - (adev)->powerplay.pp_handle)) +#define amdgpu_dpm_enable_mgpu_fan_boost(adev) \ + ((adev)->powerplay.pp_funcs->enable_mgpu_fan_boost(\ + (adev)->powerplay.pp_handle)) struct amdgpu_dpm { struct amdgpu_ps *ps; @@ -402,7 +403,6 @@ struct amdgpu_dpm { u32 tdp_adjustment; u16 load_line_slope; bool power_control; - bool ac_power; /* special states active */ bool thermal_active; bool uvd_active; @@ -439,6 +439,7 @@ struct amdgpu_pm { struct amd_pp_display_configuration pm_display_cfg;/* set by dc */ uint32_t smu_prv_buffer_size; struct amdgpu_bo *smu_prv_buffer; + bool ac_power; }; #define R600_SSTU_DFLT 0 diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index b0bf2f24da48..28781414d71c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -1,10 +1,3 @@ -/** - * \file amdgpu_drv.c - * AMD Amdgpu driver - * - * \author Gareth Hughes <[email protected]> - */ - /* * Copyright 2000 VA Linux Systems, Inc., Sunnyvale, California. * All Rights Reserved. @@ -43,6 +36,7 @@ #include "amdgpu.h" #include "amdgpu_irq.h" +#include "amdgpu_gem.h" #include "amdgpu_amdkfd.h" @@ -76,9 +70,10 @@ * - 3.24.0 - Add high priority compute support for gfx9 * - 3.25.0 - Add support for sensor query info (stable pstate sclk/mclk). * - 3.26.0 - GFX9: Process AMDGPU_IB_FLAG_TC_WB_NOT_INVALIDATE. + * - 3.27.0 - Add new chunk to to AMDGPU_CS to enable BO_LIST creation. */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 26 +#define KMS_DRIVER_MINOR 27 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; @@ -110,11 +105,8 @@ int amdgpu_vram_page_split = 512; int amdgpu_vm_update_mode = -1; int amdgpu_exp_hw_support = 0; int amdgpu_dc = -1; -int amdgpu_dc_log = 0; int amdgpu_sched_jobs = 32; int amdgpu_sched_hw_submission = 2; -int amdgpu_no_evict = 0; -int amdgpu_direct_gma_size = 0; uint amdgpu_pcie_gen_cap = 0; uint amdgpu_pcie_lane_cap = 0; uint amdgpu_cg_mask = 0xffffffff; @@ -122,7 +114,8 @@ uint amdgpu_pg_mask = 0xffffffff; uint amdgpu_sdma_phase_quantum = 32; char *amdgpu_disable_cu = NULL; char *amdgpu_virtual_display = NULL; -uint amdgpu_pp_feature_mask = 0xffff3fff; /* gfxoff (bit 15) disabled by default */ +/* OverDrive(bit 14) disabled by default*/ +uint amdgpu_pp_feature_mask = 0xffffbfff; int amdgpu_ngg = 0; int amdgpu_prim_buf_per_se = 0; int amdgpu_pos_buf_per_se = 0; @@ -134,164 +127,372 @@ int amdgpu_compute_multipipe = -1; int amdgpu_gpu_recovery = -1; /* auto */ int amdgpu_emu_mode = 0; uint amdgpu_smu_memory_pool_size = 0; +struct amdgpu_mgpu_info mgpu_info = { + .mutex = __MUTEX_INITIALIZER(mgpu_info.mutex), +}; +/** + * DOC: vramlimit (int) + * Restrict the total amount of VRAM in MiB for testing. The default is 0 (Use full VRAM). + */ MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes"); module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); +/** + * DOC: vis_vramlimit (int) + * Restrict the amount of CPU visible VRAM in MiB for testing. The default is 0 (Use full CPU visible VRAM). + */ MODULE_PARM_DESC(vis_vramlimit, "Restrict visible VRAM for testing, in megabytes"); module_param_named(vis_vramlimit, amdgpu_vis_vram_limit, int, 0444); +/** + * DOC: gartsize (uint) + * Restrict the size of GART in Mib (32, 64, etc.) for testing. The default is -1 (The size depends on asic). + */ MODULE_PARM_DESC(gartsize, "Size of GART to setup in megabytes (32, 64, etc., -1=auto)"); module_param_named(gartsize, amdgpu_gart_size, uint, 0600); +/** + * DOC: gttsize (int) + * Restrict the size of GTT domain in MiB for testing. The default is -1 (It's VRAM size if 3GB < VRAM < 3/4 RAM, + * otherwise 3/4 RAM size). + */ MODULE_PARM_DESC(gttsize, "Size of the GTT domain in megabytes (-1 = auto)"); module_param_named(gttsize, amdgpu_gtt_size, int, 0600); +/** + * DOC: moverate (int) + * Set maximum buffer migration rate in MB/s. The default is -1 (8 MB/s). + */ MODULE_PARM_DESC(moverate, "Maximum buffer migration rate in MB/s. (32, 64, etc., -1=auto, 0=1=disabled)"); module_param_named(moverate, amdgpu_moverate, int, 0600); +/** + * DOC: benchmark (int) + * Run benchmarks. The default is 0 (Skip benchmarks). + */ MODULE_PARM_DESC(benchmark, "Run benchmark"); module_param_named(benchmark, amdgpu_benchmarking, int, 0444); +/** + * DOC: test (int) + * Test BO GTT->VRAM and VRAM->GTT GPU copies. The default is 0 (Skip test, only set 1 to run test). + */ MODULE_PARM_DESC(test, "Run tests"); module_param_named(test, amdgpu_testing, int, 0444); +/** + * DOC: audio (int) + * Set HDMI/DPAudio. Only affects non-DC display handling. The default is -1 (Enabled), set 0 to disabled it. + */ MODULE_PARM_DESC(audio, "Audio enable (-1 = auto, 0 = disable, 1 = enable)"); module_param_named(audio, amdgpu_audio, int, 0444); +/** + * DOC: disp_priority (int) + * Set display Priority (1 = normal, 2 = high). Only affects non-DC display handling. The default is 0 (auto). + */ MODULE_PARM_DESC(disp_priority, "Display Priority (0 = auto, 1 = normal, 2 = high)"); module_param_named(disp_priority, amdgpu_disp_priority, int, 0444); +/** + * DOC: hw_i2c (int) + * To enable hw i2c engine. Only affects non-DC display handling. The default is 0 (Disabled). + */ MODULE_PARM_DESC(hw_i2c, "hw i2c engine enable (0 = disable)"); module_param_named(hw_i2c, amdgpu_hw_i2c, int, 0444); +/** + * DOC: pcie_gen2 (int) + * To disable PCIE Gen2/3 mode (0 = disable, 1 = enable). The default is -1 (auto, enabled). + */ MODULE_PARM_DESC(pcie_gen2, "PCIE Gen2 mode (-1 = auto, 0 = disable, 1 = enable)"); module_param_named(pcie_gen2, amdgpu_pcie_gen2, int, 0444); +/** + * DOC: msi (int) + * To disable Message Signaled Interrupts (MSI) functionality (1 = enable, 0 = disable). The default is -1 (auto, enabled). + */ MODULE_PARM_DESC(msi, "MSI support (1 = enable, 0 = disable, -1 = auto)"); module_param_named(msi, amdgpu_msi, int, 0444); +/** + * DOC: lockup_timeout (int) + * Set GPU scheduler timeout value in ms. Value 0 is invalidated, will be adjusted to 10000. + * Negative values mean 'infinite timeout' (MAX_JIFFY_OFFSET). The default is 10000. + */ MODULE_PARM_DESC(lockup_timeout, "GPU lockup timeout in ms > 0 (default 10000)"); module_param_named(lockup_timeout, amdgpu_lockup_timeout, int, 0444); +/** + * DOC: dpm (int) + * Override for dynamic power management setting (1 = enable, 0 = disable). The default is -1 (auto). + */ MODULE_PARM_DESC(dpm, "DPM support (1 = enable, 0 = disable, -1 = auto)"); module_param_named(dpm, amdgpu_dpm, int, 0444); +/** + * DOC: fw_load_type (int) + * Set different firmware loading type for debugging (0 = direct, 1 = SMU, 2 = PSP). The default is -1 (auto). + */ MODULE_PARM_DESC(fw_load_type, "firmware loading type (0 = direct, 1 = SMU, 2 = PSP, -1 = auto)"); module_param_named(fw_load_type, amdgpu_fw_load_type, int, 0444); +/** + * DOC: aspm (int) + * To disable ASPM (1 = enable, 0 = disable). The default is -1 (auto, enabled). + */ MODULE_PARM_DESC(aspm, "ASPM support (1 = enable, 0 = disable, -1 = auto)"); module_param_named(aspm, amdgpu_aspm, int, 0444); +/** + * DOC: runpm (int) + * Override for runtime power management control for dGPUs in PX/HG laptops. The amdgpu driver can dynamically power down + * the dGPU on PX/HG laptops when it is idle. The default is -1 (auto enable). Setting the value to 0 disables this functionality. + */ MODULE_PARM_DESC(runpm, "PX runtime pm (1 = force enable, 0 = disable, -1 = PX only default)"); module_param_named(runpm, amdgpu_runtime_pm, int, 0444); +/** + * DOC: ip_block_mask (uint) + * Override what IP blocks are enabled on the GPU. Each GPU is a collection of IP blocks (gfx, display, video, etc.). + * Use this parameter to disable specific blocks. Note that the IP blocks do not have a fixed index. Some asics may not have + * some IPs or may include multiple instances of an IP so the ordering various from asic to asic. See the driver output in + * the kernel log for the list of IPs on the asic. The default is 0xffffffff (enable all blocks on a device). + */ MODULE_PARM_DESC(ip_block_mask, "IP Block Mask (all blocks enabled (default))"); module_param_named(ip_block_mask, amdgpu_ip_block_mask, uint, 0444); +/** + * DOC: bapm (int) + * Bidirectional Application Power Management (BAPM) used to dynamically share TDP between CPU and GPU. Set value 0 to disable it. + * The default -1 (auto, enabled) + */ MODULE_PARM_DESC(bapm, "BAPM support (1 = enable, 0 = disable, -1 = auto)"); module_param_named(bapm, amdgpu_bapm, int, 0444); +/** + * DOC: deep_color (int) + * Set 1 to enable Deep Color support. Only affects non-DC display handling. The default is 0 (disabled). + */ MODULE_PARM_DESC(deep_color, "Deep Color support (1 = enable, 0 = disable (default))"); module_param_named(deep_color, amdgpu_deep_color, int, 0444); +/** + * DOC: vm_size (int) + * Override the size of the GPU's per client virtual address space in GiB. The default is -1 (automatic for each asic). + */ MODULE_PARM_DESC(vm_size, "VM address space size in gigabytes (default 64GB)"); module_param_named(vm_size, amdgpu_vm_size, int, 0444); +/** + * DOC: vm_fragment_size (int) + * Override VM fragment size in bits (4, 5, etc. 4 = 64K, 9 = 2M). The default is -1 (automatic for each asic). + */ MODULE_PARM_DESC(vm_fragment_size, "VM fragment size in bits (4, 5, etc. 4 = 64K (default), Max 9 = 2M)"); module_param_named(vm_fragment_size, amdgpu_vm_fragment_size, int, 0444); +/** + * DOC: vm_block_size (int) + * Override VM page table size in bits (default depending on vm_size and hw setup). The default is -1 (automatic for each asic). + */ MODULE_PARM_DESC(vm_block_size, "VM page table size in bits (default depending on vm_size)"); module_param_named(vm_block_size, amdgpu_vm_block_size, int, 0444); +/** + * DOC: vm_fault_stop (int) + * Stop on VM fault for debugging (0 = never, 1 = print first, 2 = always). The default is 0 (No stop). + */ MODULE_PARM_DESC(vm_fault_stop, "Stop on VM fault (0 = never (default), 1 = print first, 2 = always)"); module_param_named(vm_fault_stop, amdgpu_vm_fault_stop, int, 0444); +/** + * DOC: vm_debug (int) + * Debug VM handling (0 = disabled, 1 = enabled). The default is 0 (Disabled). + */ MODULE_PARM_DESC(vm_debug, "Debug VM handling (0 = disabled (default), 1 = enabled)"); module_param_named(vm_debug, amdgpu_vm_debug, int, 0644); +/** + * DOC: vm_update_mode (int) + * Override VM update mode. VM updated by using CPU (0 = never, 1 = Graphics only, 2 = Compute only, 3 = Both). The default + * is -1 (Only in large BAR(LB) systems Compute VM tables will be updated by CPU, otherwise 0, never). + */ MODULE_PARM_DESC(vm_update_mode, "VM update using CPU (0 = never (default except for large BAR(LB)), 1 = Graphics only, 2 = Compute only (default for LB), 3 = Both"); module_param_named(vm_update_mode, amdgpu_vm_update_mode, int, 0444); +/** + * DOC: vram_page_split (int) + * Override the number of pages after we split VRAM allocations (default 512, -1 = disable). The default is 512. + */ MODULE_PARM_DESC(vram_page_split, "Number of pages after we split VRAM allocations (default 512, -1 = disable)"); module_param_named(vram_page_split, amdgpu_vram_page_split, int, 0444); +/** + * DOC: exp_hw_support (int) + * Enable experimental hw support (1 = enable). The default is 0 (disabled). + */ MODULE_PARM_DESC(exp_hw_support, "experimental hw support (1 = enable, 0 = disable (default))"); module_param_named(exp_hw_support, amdgpu_exp_hw_support, int, 0444); +/** + * DOC: dc (int) + * Disable/Enable Display Core driver for debugging (1 = enable, 0 = disable). The default is -1 (automatic for each asic). + */ MODULE_PARM_DESC(dc, "Display Core driver (1 = enable, 0 = disable, -1 = auto (default))"); module_param_named(dc, amdgpu_dc, int, 0444); -MODULE_PARM_DESC(dc_log, "Display Core Log Level (0 = minimal (default), 1 = chatty"); -module_param_named(dc_log, amdgpu_dc_log, int, 0444); - +/** + * DOC: sched_jobs (int) + * Override the max number of jobs supported in the sw queue. The default is 32. + */ MODULE_PARM_DESC(sched_jobs, "the max number of jobs supported in the sw queue (default 32)"); module_param_named(sched_jobs, amdgpu_sched_jobs, int, 0444); +/** + * DOC: sched_hw_submission (int) + * Override the max number of HW submissions. The default is 2. + */ MODULE_PARM_DESC(sched_hw_submission, "the max number of HW submissions (default 2)"); module_param_named(sched_hw_submission, amdgpu_sched_hw_submission, int, 0444); +/** + * DOC: ppfeaturemask (uint) + * Override power features enabled. See enum PP_FEATURE_MASK in drivers/gpu/drm/amd/include/amd_shared.h. + * The default is the current set of stable power features. + */ MODULE_PARM_DESC(ppfeaturemask, "all power features enabled (default))"); module_param_named(ppfeaturemask, amdgpu_pp_feature_mask, uint, 0444); -MODULE_PARM_DESC(no_evict, "Support pinning request from user space (1 = enable, 0 = disable (default))"); -module_param_named(no_evict, amdgpu_no_evict, int, 0444); - -MODULE_PARM_DESC(direct_gma_size, "Direct GMA size in megabytes (max 96MB)"); -module_param_named(direct_gma_size, amdgpu_direct_gma_size, int, 0444); - +/** + * DOC: pcie_gen_cap (uint) + * Override PCIE gen speed capabilities. See the CAIL flags in drivers/gpu/drm/amd/include/amd_pcie.h. + * The default is 0 (automatic for each asic). + */ MODULE_PARM_DESC(pcie_gen_cap, "PCIE Gen Caps (0: autodetect (default))"); module_param_named(pcie_gen_cap, amdgpu_pcie_gen_cap, uint, 0444); +/** + * DOC: pcie_lane_cap (uint) + * Override PCIE lanes capabilities. See the CAIL flags in drivers/gpu/drm/amd/include/amd_pcie.h. + * The default is 0 (automatic for each asic). + */ MODULE_PARM_DESC(pcie_lane_cap, "PCIE Lane Caps (0: autodetect (default))"); module_param_named(pcie_lane_cap, amdgpu_pcie_lane_cap, uint, 0444); +/** + * DOC: cg_mask (uint) + * Override Clockgating features enabled on GPU (0 = disable clock gating). See the AMD_CG_SUPPORT flags in + * drivers/gpu/drm/amd/include/amd_shared.h. The default is 0xffffffff (all enabled). + */ MODULE_PARM_DESC(cg_mask, "Clockgating flags mask (0 = disable clock gating)"); module_param_named(cg_mask, amdgpu_cg_mask, uint, 0444); +/** + * DOC: pg_mask (uint) + * Override Powergating features enabled on GPU (0 = disable power gating). See the AMD_PG_SUPPORT flags in + * drivers/gpu/drm/amd/include/amd_shared.h. The default is 0xffffffff (all enabled). + */ MODULE_PARM_DESC(pg_mask, "Powergating flags mask (0 = disable power gating)"); module_param_named(pg_mask, amdgpu_pg_mask, uint, 0444); +/** + * DOC: sdma_phase_quantum (uint) + * Override SDMA context switch phase quantum (x 1K GPU clock cycles, 0 = no change). The default is 32. + */ MODULE_PARM_DESC(sdma_phase_quantum, "SDMA context switch phase quantum (x 1K GPU clock cycles, 0 = no change (default 32))"); module_param_named(sdma_phase_quantum, amdgpu_sdma_phase_quantum, uint, 0444); +/** + * DOC: disable_cu (charp) + * Set to disable CUs (It's set like se.sh.cu,...). The default is NULL. + */ MODULE_PARM_DESC(disable_cu, "Disable CUs (se.sh.cu,...)"); module_param_named(disable_cu, amdgpu_disable_cu, charp, 0444); +/** + * DOC: virtual_display (charp) + * Set to enable virtual display feature. This feature provides a virtual display hardware on headless boards + * or in virtualized environments. It will be set like xxxx:xx:xx.x,x;xxxx:xx:xx.x,x. It's the pci address of + * the device, plus the number of crtcs to expose. E.g., 0000:26:00.0,4 would enable 4 virtual crtcs on the pci + * device at 26:00.0. The default is NULL. + */ MODULE_PARM_DESC(virtual_display, "Enable virtual display feature (the virtual_display will be set like xxxx:xx:xx.x,x;xxxx:xx:xx.x,x)"); module_param_named(virtual_display, amdgpu_virtual_display, charp, 0444); +/** + * DOC: ngg (int) + * Set to enable Next Generation Graphics (1 = enable). The default is 0 (disabled). + */ MODULE_PARM_DESC(ngg, "Next Generation Graphics (1 = enable, 0 = disable(default depending on gfx))"); module_param_named(ngg, amdgpu_ngg, int, 0444); +/** + * DOC: prim_buf_per_se (int) + * Override the size of Primitive Buffer per Shader Engine in Byte. The default is 0 (depending on gfx). + */ MODULE_PARM_DESC(prim_buf_per_se, "the size of Primitive Buffer per Shader Engine (default depending on gfx)"); module_param_named(prim_buf_per_se, amdgpu_prim_buf_per_se, int, 0444); +/** + * DOC: pos_buf_per_se (int) + * Override the size of Position Buffer per Shader Engine in Byte. The default is 0 (depending on gfx). + */ MODULE_PARM_DESC(pos_buf_per_se, "the size of Position Buffer per Shader Engine (default depending on gfx)"); module_param_named(pos_buf_per_se, amdgpu_pos_buf_per_se, int, 0444); +/** + * DOC: cntl_sb_buf_per_se (int) + * Override the size of Control Sideband per Shader Engine in Byte. The default is 0 (depending on gfx). + */ MODULE_PARM_DESC(cntl_sb_buf_per_se, "the size of Control Sideband per Shader Engine (default depending on gfx)"); module_param_named(cntl_sb_buf_per_se, amdgpu_cntl_sb_buf_per_se, int, 0444); +/** + * DOC: param_buf_per_se (int) + * Override the size of Off-Chip Pramater Cache per Shader Engine in Byte. The default is 0 (depending on gfx). + */ MODULE_PARM_DESC(param_buf_per_se, "the size of Off-Chip Pramater Cache per Shader Engine (default depending on gfx)"); module_param_named(param_buf_per_se, amdgpu_param_buf_per_se, int, 0444); +/** + * DOC: job_hang_limit (int) + * Set how much time allow a job hang and not drop it. The default is 0. + */ MODULE_PARM_DESC(job_hang_limit, "how much time allow a job hang and not drop it (default 0)"); module_param_named(job_hang_limit, amdgpu_job_hang_limit, int ,0444); +/** + * DOC: lbpw (int) + * Override Load Balancing Per Watt (LBPW) support (1 = enable, 0 = disable). The default is -1 (auto, enabled). + */ MODULE_PARM_DESC(lbpw, "Load Balancing Per Watt (LBPW) support (1 = enable, 0 = disable, -1 = auto)"); module_param_named(lbpw, amdgpu_lbpw, int, 0444); MODULE_PARM_DESC(compute_multipipe, "Force compute queues to be spread across pipes (1 = enable, 0 = disable, -1 = auto)"); module_param_named(compute_multipipe, amdgpu_compute_multipipe, int, 0444); +/** + * DOC: gpu_recovery (int) + * Set to enable GPU recovery mechanism (1 = enable, 0 = disable). The default is -1 (auto, disabled except SRIOV). + */ MODULE_PARM_DESC(gpu_recovery, "Enable GPU recovery mechanism, (1 = enable, 0 = disable, -1 = auto)"); module_param_named(gpu_recovery, amdgpu_gpu_recovery, int, 0444); +/** + * DOC: emu_mode (int) + * Set value 1 to enable emulation mode. This is only needed when running on an emulator. The default is 0 (disabled). + */ MODULE_PARM_DESC(emu_mode, "Emulation mode, (1 = enable, 0 = disable)"); module_param_named(emu_mode, amdgpu_emu_mode, int, 0444); +/** + * DOC: si_support (int) + * Set SI support driver. This parameter works after set config CONFIG_DRM_AMDGPU_SI. For SI asic, when radeon driver is enabled, + * set value 0 to use radeon driver, while set value 1 to use amdgpu driver. The default is using radeon driver when it available, + * otherwise using amdgpu driver. + */ #ifdef CONFIG_DRM_AMDGPU_SI #if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE) @@ -305,6 +506,12 @@ MODULE_PARM_DESC(si_support, "SI support (1 = enabled (default), 0 = disabled)") module_param_named(si_support, amdgpu_si_support, int, 0444); #endif +/** + * DOC: cik_support (int) + * Set CIK support driver. This parameter works after set config CONFIG_DRM_AMDGPU_CIK. For CIK asic, when radeon driver is enabled, + * set value 0 to use radeon driver, while set value 1 to use amdgpu driver. The default is using radeon driver when it available, + * otherwise using amdgpu driver. + */ #ifdef CONFIG_DRM_AMDGPU_CIK #if defined(CONFIG_DRM_RADEON) || defined(CONFIG_DRM_RADEON_MODULE) @@ -318,11 +525,112 @@ MODULE_PARM_DESC(cik_support, "CIK support (1 = enabled (default), 0 = disabled) module_param_named(cik_support, amdgpu_cik_support, int, 0444); #endif +/** + * DOC: smu_memory_pool_size (uint) + * It is used to reserve gtt for smu debug usage, setting value 0 to disable it. The actual size is value * 256MiB. + * E.g. 0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte. The default is 0 (disabled). + */ MODULE_PARM_DESC(smu_memory_pool_size, "reserve gtt for smu debug usage, 0 = disable," "0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte"); module_param_named(smu_memory_pool_size, amdgpu_smu_memory_pool_size, uint, 0444); +#ifdef CONFIG_HSA_AMD +/** + * DOC: sched_policy (int) + * Set scheduling policy. Default is HWS(hardware scheduling) with over-subscription. + * Setting 1 disables over-subscription. Setting 2 disables HWS and statically + * assigns queues to HQDs. + */ +int sched_policy = KFD_SCHED_POLICY_HWS; +module_param(sched_policy, int, 0444); +MODULE_PARM_DESC(sched_policy, + "Scheduling policy (0 = HWS (Default), 1 = HWS without over-subscription, 2 = Non-HWS (Used for debugging only)"); + +/** + * DOC: hws_max_conc_proc (int) + * Maximum number of processes that HWS can schedule concurrently. The maximum is the + * number of VMIDs assigned to the HWS, which is also the default. + */ +int hws_max_conc_proc = 8; +module_param(hws_max_conc_proc, int, 0444); +MODULE_PARM_DESC(hws_max_conc_proc, + "Max # processes HWS can execute concurrently when sched_policy=0 (0 = no concurrency, #VMIDs for KFD = Maximum(default))"); + +/** + * DOC: cwsr_enable (int) + * CWSR(compute wave store and resume) allows the GPU to preempt shader execution in + * the middle of a compute wave. Default is 1 to enable this feature. Setting 0 + * disables it. + */ +int cwsr_enable = 1; +module_param(cwsr_enable, int, 0444); +MODULE_PARM_DESC(cwsr_enable, "CWSR enable (0 = Off, 1 = On (Default))"); + +/** + * DOC: max_num_of_queues_per_device (int) + * Maximum number of queues per device. Valid setting is between 1 and 4096. Default + * is 4096. + */ +int max_num_of_queues_per_device = KFD_MAX_NUM_OF_QUEUES_PER_DEVICE_DEFAULT; +module_param(max_num_of_queues_per_device, int, 0444); +MODULE_PARM_DESC(max_num_of_queues_per_device, + "Maximum number of supported queues per device (1 = Minimum, 4096 = default)"); + +/** + * DOC: send_sigterm (int) + * Send sigterm to HSA process on unhandled exceptions. Default is not to send sigterm + * but just print errors on dmesg. Setting 1 enables sending sigterm. + */ +int send_sigterm; +module_param(send_sigterm, int, 0444); +MODULE_PARM_DESC(send_sigterm, + "Send sigterm to HSA process on unhandled exception (0 = disable, 1 = enable)"); + +/** + * DOC: debug_largebar (int) + * Set debug_largebar as 1 to enable simulating large-bar capability on non-large bar + * system. This limits the VRAM size reported to ROCm applications to the visible + * size, usually 256MB. + * Default value is 0, diabled. + */ +int debug_largebar; +module_param(debug_largebar, int, 0444); +MODULE_PARM_DESC(debug_largebar, + "Debug large-bar flag used to simulate large-bar capability on non-large bar machine (0 = disable, 1 = enable)"); + +/** + * DOC: ignore_crat (int) + * Ignore CRAT table during KFD initialization. By default, KFD uses the ACPI CRAT + * table to get information about AMD APUs. This option can serve as a workaround on + * systems with a broken CRAT table. + */ +int ignore_crat; +module_param(ignore_crat, int, 0444); +MODULE_PARM_DESC(ignore_crat, + "Ignore CRAT table during KFD initialization (0 = use CRAT (default), 1 = ignore CRAT)"); + +/** + * DOC: noretry (int) + * This parameter sets sh_mem_config.retry_disable. Default value, 0, enables retry. + * Setting 1 disables retry. + * Retry is needed for recoverable page faults. + */ +int noretry; +module_param(noretry, int, 0644); +MODULE_PARM_DESC(noretry, + "Set sh_mem_config.retry_disable on Vega10 (0 = retry enabled (default), 1 = retry disabled)"); + +/** + * DOC: halt_if_hws_hang (int) + * Halt if HWS hang is detected. Default value, 0, disables the halt on hang. + * Setting 1 enables halt on hang. + */ +int halt_if_hws_hang; +module_param(halt_if_hws_hang, int, 0644); +MODULE_PARM_DESC(halt_if_hws_hang, "Halt if HWS hang is detected (0 = off (default), 1 = on)"); +#endif + static const struct pci_device_id pciidlist[] = { #ifdef CONFIG_DRM_AMDGPU_SI {0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI}, @@ -532,6 +840,7 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x67CA, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS10}, {0x1002, 0x67CC, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS10}, {0x1002, 0x67CF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS10}, + {0x1002, 0x6FDF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS10}, /* Polaris12 */ {0x1002, 0x6980, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12}, {0x1002, 0x6981, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12}, @@ -561,14 +870,15 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x69A3, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA12}, {0x1002, 0x69AF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA12}, /* Vega 20 */ - {0x1002, 0x66A0, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x66A1, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x66A2, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x66A3, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x66A7, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, - {0x1002, 0x66AF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x66A0, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20}, + {0x1002, 0x66A1, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20}, + {0x1002, 0x66A2, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20}, + {0x1002, 0x66A3, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20}, + {0x1002, 0x66A7, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20}, + {0x1002, 0x66AF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20}, /* Raven */ {0x1002, 0x15dd, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RAVEN|AMD_IS_APU}, + {0x1002, 0x15d8, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RAVEN|AMD_IS_APU}, {0, 0, 0} }; @@ -577,28 +887,6 @@ MODULE_DEVICE_TABLE(pci, pciidlist); static struct drm_driver kms_driver; -static int amdgpu_kick_out_firmware_fb(struct pci_dev *pdev) -{ - struct apertures_struct *ap; - bool primary = false; - - ap = alloc_apertures(1); - if (!ap) - return -ENOMEM; - - ap->ranges[0].base = pci_resource_start(pdev, 0); - ap->ranges[0].size = pci_resource_len(pdev, 0); - -#ifdef CONFIG_X86 - primary = pdev->resource[PCI_ROM_RESOURCE].flags & IORESOURCE_ROM_SHADOW; -#endif - drm_fb_helper_remove_conflicting_framebuffers(ap, "amdgpudrmfb", primary); - kfree(ap); - - return 0; -} - - static int amdgpu_pci_probe(struct pci_dev *pdev, const struct pci_device_id *ent) { @@ -617,30 +905,18 @@ static int amdgpu_pci_probe(struct pci_dev *pdev, return -ENODEV; } - /* - * Initialize amdkfd before starting radeon. If it was not loaded yet, - * defer radeon probing - */ - ret = amdgpu_amdkfd_init(); - if (ret == -EPROBE_DEFER) - return ret; - /* Get rid of things like offb */ - ret = amdgpu_kick_out_firmware_fb(pdev); + ret = drm_fb_helper_remove_conflicting_pci_framebuffers(pdev, 0, "amdgpudrmfb"); if (ret) return ret; - /* warn the user if they mix atomic and non-atomic capable GPUs */ - if ((kms_driver.driver_features & DRIVER_ATOMIC) && !supports_atomic) - DRM_ERROR("Mixing atomic and non-atomic capable GPUs!\n"); - /* support atomic early so the atomic debugfs stuff gets created */ - if (supports_atomic) - kms_driver.driver_features |= DRIVER_ATOMIC; - dev = drm_dev_alloc(&kms_driver, &pdev->dev); if (IS_ERR(dev)) return PTR_ERR(dev); + if (!supports_atomic) + dev->driver_features &= ~DRIVER_ATOMIC; + ret = pci_enable_device(pdev); if (ret) goto err_free; @@ -664,7 +940,7 @@ retry_init: err_pci: pci_disable_device(pdev); err_free: - drm_dev_unref(dev); + drm_dev_put(dev); return ret; } @@ -673,8 +949,8 @@ amdgpu_pci_remove(struct pci_dev *pdev) { struct drm_device *dev = pci_get_drvdata(pdev); - drm_dev_unregister(dev); - drm_dev_unref(dev); + DRM_ERROR("Device removal is currently not supported outside of fbcon\n"); + drm_dev_unplug(dev); pci_disable_device(pdev); pci_set_drvdata(pdev, NULL); } @@ -855,9 +1131,21 @@ static const struct dev_pm_ops amdgpu_pm_ops = { .runtime_idle = amdgpu_pmops_runtime_idle, }; +static int amdgpu_flush(struct file *f, fl_owner_t id) +{ + struct drm_file *file_priv = f->private_data; + struct amdgpu_fpriv *fpriv = file_priv->driver_priv; + + amdgpu_ctx_mgr_entity_flush(&fpriv->ctx_mgr); + + return 0; +} + + static const struct file_operations amdgpu_driver_kms_fops = { .owner = THIS_MODULE, .open = drm_open, + .flush = amdgpu_flush, .release = drm_release, .unlocked_ioctl = amdgpu_drm_ioctl, .mmap = amdgpu_mmap, @@ -880,7 +1168,7 @@ amdgpu_get_crtc_scanout_position(struct drm_device *dev, unsigned int pipe, static struct drm_driver kms_driver = { .driver_features = - DRIVER_USE_AGP | + DRIVER_USE_AGP | DRIVER_ATOMIC | DRIVER_HAVE_IRQ | DRIVER_IRQ_SHARED | DRIVER_GEM | DRIVER_PRIME | DRIVER_RENDER | DRIVER_MODESET | DRIVER_SYNCOBJ, .load = amdgpu_driver_load_kms, @@ -957,6 +1245,10 @@ static int __init amdgpu_init(void) pdriver = &amdgpu_kms_pci_driver; driver->num_ioctls = amdgpu_max_kms_ioctl; amdgpu_register_atpx_handler(); + + /* Ignore KFD init failures. Normal when CONFIG_HSA_AMD is not set. */ + amdgpu_amdkfd_init(); + /* let modprobe override vga console setting */ return pci_register_driver(pdriver); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c index 94138abe093b..ec78e2b2015c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_encoders.c @@ -28,6 +28,7 @@ #include <drm/amdgpu_drm.h> #include "amdgpu.h" #include "amdgpu_connectors.h" +#include "amdgpu_display.h" #include "atom.h" #include "atombios_encoders.h" @@ -46,7 +47,7 @@ amdgpu_link_encoder_connector(struct drm_device *dev) list_for_each_entry(encoder, &dev->mode_config.encoder_list, head) { amdgpu_encoder = to_amdgpu_encoder(encoder); if (amdgpu_encoder->devices & amdgpu_connector->devices) { - drm_mode_connector_attach_encoder(connector, encoder); + drm_connector_attach_encoder(connector, encoder); if (amdgpu_encoder->devices & (ATOM_DEVICE_LCD_SUPPORT)) { amdgpu_atombios_encoder_init_backlight(amdgpu_encoder, connector); adev->mode_info.bl_encoder = amdgpu_encoder; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index bc5fd8ebab5d..5cbde74b97dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -33,6 +33,7 @@ #include <drm/amdgpu_drm.h> #include "amdgpu.h" #include "cikd.h" +#include "amdgpu_gem.h" #include <drm/drm_fb_helper.h> @@ -146,7 +147,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | AMDGPU_GEM_CREATE_VRAM_CLEARED, - true, NULL, &gobj); + ttm_bo_type_kernel, NULL, &gobj); if (ret) { pr_err("failed to allocate framebuffer (%d)\n", aligned_size); return -ENOMEM; @@ -168,11 +169,19 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev, } - ret = amdgpu_bo_pin(abo, domain, NULL); + ret = amdgpu_bo_pin(abo, domain); if (ret) { amdgpu_bo_unreserve(abo); goto out_unref; } + + ret = amdgpu_ttm_alloc_gart(&abo->tbo); + if (ret) { + amdgpu_bo_unreserve(abo); + dev_err(adev->dev, "%p bind failed\n", abo); + goto out_unref; + } + ret = amdgpu_bo_kmap(abo, NULL); amdgpu_bo_unreserve(abo); if (ret) { @@ -365,8 +374,8 @@ void amdgpu_fbdev_fini(struct amdgpu_device *adev) void amdgpu_fbdev_set_suspend(struct amdgpu_device *adev, int state) { if (adev->mode_info.rfbdev) - drm_fb_helper_set_suspend(&adev->mode_info.rfbdev->helper, - state); + drm_fb_helper_set_suspend_unlocked(&adev->mode_info.rfbdev->helper, + state); } int amdgpu_fbdev_total_size(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index e74d620d9699..5448cf27654e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -216,8 +216,10 @@ static void amdgpu_fence_schedule_fallback(struct amdgpu_ring *ring) * Checks the current fence value and calculates the last * signalled fence value. Wakes the fence queue if the * sequence number has increased. + * + * Returns true if fence was processed */ -void amdgpu_fence_process(struct amdgpu_ring *ring) +bool amdgpu_fence_process(struct amdgpu_ring *ring) { struct amdgpu_fence_driver *drv = &ring->fence_drv; uint32_t seq, last_seq; @@ -229,11 +231,12 @@ void amdgpu_fence_process(struct amdgpu_ring *ring) } while (atomic_cmpxchg(&drv->last_seq, last_seq, seq) != last_seq); - if (seq != ring->fence_drv.sync_seq) + if (del_timer(&ring->fence_drv.fallback_timer) && + seq != ring->fence_drv.sync_seq) amdgpu_fence_schedule_fallback(ring); if (unlikely(seq == last_seq)) - return; + return false; last_seq &= drv->num_fences_mask; seq &= drv->num_fences_mask; @@ -260,6 +263,8 @@ void amdgpu_fence_process(struct amdgpu_ring *ring) dma_fence_put(fence); } while (last_seq != seq); + + return true; } /** @@ -274,7 +279,8 @@ static void amdgpu_fence_fallback(struct timer_list *t) struct amdgpu_ring *ring = from_timer(ring, t, fence_drv.fallback_timer); - amdgpu_fence_process(ring); + if (amdgpu_fence_process(ring)) + DRM_WARN("Fence fallback timer expired on ring %s\n", ring->name); } /** @@ -646,7 +652,6 @@ static const struct dma_fence_ops amdgpu_fence_ops = { .get_driver_name = amdgpu_fence_get_driver_name, .get_timeline_name = amdgpu_fence_get_timeline_name, .enable_signaling = amdgpu_fence_enable_signaling, - .wait = dma_fence_default_wait, .release = amdgpu_fence_release, }; @@ -702,7 +707,7 @@ static int amdgpu_debugfs_gpu_recover(struct seq_file *m, void *data) struct amdgpu_device *adev = dev->dev_private; seq_printf(m, "gpu recover\n"); - amdgpu_device_gpu_recover(adev, NULL, true); + amdgpu_device_gpu_recover(adev, NULL); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index dd11b7313ca0..11fea28f8ad3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -112,7 +112,7 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) { int r; - if (adev->gart.robj == NULL) { + if (adev->gart.bo == NULL) { struct amdgpu_bo_param bp; memset(&bp, 0, sizeof(bp)); @@ -123,7 +123,7 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; bp.type = ttm_bo_type_kernel; bp.resv = NULL; - r = amdgpu_bo_create(adev, &bp, &adev->gart.robj); + r = amdgpu_bo_create(adev, &bp, &adev->gart.bo); if (r) { return r; } @@ -143,23 +143,20 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) */ int amdgpu_gart_table_vram_pin(struct amdgpu_device *adev) { - uint64_t gpu_addr; int r; - r = amdgpu_bo_reserve(adev->gart.robj, false); + r = amdgpu_bo_reserve(adev->gart.bo, false); if (unlikely(r != 0)) return r; - r = amdgpu_bo_pin(adev->gart.robj, - AMDGPU_GEM_DOMAIN_VRAM, &gpu_addr); + r = amdgpu_bo_pin(adev->gart.bo, AMDGPU_GEM_DOMAIN_VRAM); if (r) { - amdgpu_bo_unreserve(adev->gart.robj); + amdgpu_bo_unreserve(adev->gart.bo); return r; } - r = amdgpu_bo_kmap(adev->gart.robj, &adev->gart.ptr); + r = amdgpu_bo_kmap(adev->gart.bo, &adev->gart.ptr); if (r) - amdgpu_bo_unpin(adev->gart.robj); - amdgpu_bo_unreserve(adev->gart.robj); - adev->gart.table_addr = gpu_addr; + amdgpu_bo_unpin(adev->gart.bo); + amdgpu_bo_unreserve(adev->gart.bo); return r; } @@ -175,14 +172,14 @@ void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev) { int r; - if (adev->gart.robj == NULL) { + if (adev->gart.bo == NULL) { return; } - r = amdgpu_bo_reserve(adev->gart.robj, true); + r = amdgpu_bo_reserve(adev->gart.bo, true); if (likely(r == 0)) { - amdgpu_bo_kunmap(adev->gart.robj); - amdgpu_bo_unpin(adev->gart.robj); - amdgpu_bo_unreserve(adev->gart.robj); + amdgpu_bo_kunmap(adev->gart.bo); + amdgpu_bo_unpin(adev->gart.bo); + amdgpu_bo_unreserve(adev->gart.bo); adev->gart.ptr = NULL; } } @@ -198,10 +195,10 @@ void amdgpu_gart_table_vram_unpin(struct amdgpu_device *adev) */ void amdgpu_gart_table_vram_free(struct amdgpu_device *adev) { - if (adev->gart.robj == NULL) { + if (adev->gart.bo == NULL) { return; } - amdgpu_bo_unref(&adev->gart.robj); + amdgpu_bo_unref(&adev->gart.bo); } /* @@ -234,7 +231,7 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, } t = offset / AMDGPU_GPU_PAGE_SIZE; - p = t / (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); + p = t / AMDGPU_GPU_PAGES_IN_CPU_PAGE; for (i = 0; i < pages; i++, p++) { #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS adev->gart.pages[p] = NULL; @@ -243,7 +240,7 @@ int amdgpu_gart_unbind(struct amdgpu_device *adev, uint64_t offset, if (!adev->gart.ptr) continue; - for (j = 0; j < (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); j++, t++) { + for (j = 0; j < AMDGPU_GPU_PAGES_IN_CPU_PAGE; j++, t++) { amdgpu_gmc_set_pte_pde(adev, adev->gart.ptr, t, page_base, flags); page_base += AMDGPU_GPU_PAGE_SIZE; @@ -282,7 +279,7 @@ int amdgpu_gart_map(struct amdgpu_device *adev, uint64_t offset, for (i = 0; i < pages; i++) { page_base = dma_addr[i]; - for (j = 0; j < (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); j++, t++) { + for (j = 0; j < AMDGPU_GPU_PAGES_IN_CPU_PAGE; j++, t++) { amdgpu_gmc_set_pte_pde(adev, dst, t, page_base, flags); page_base += AMDGPU_GPU_PAGE_SIZE; } @@ -319,7 +316,7 @@ int amdgpu_gart_bind(struct amdgpu_device *adev, uint64_t offset, #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS t = offset / AMDGPU_GPU_PAGE_SIZE; - p = t / (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); + p = t / AMDGPU_GPU_PAGES_IN_CPU_PAGE; for (i = 0; i < pages; i++, p++) adev->gart.pages[p] = pagelist ? pagelist[i] : NULL; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h index 456295c00291..9ff62887e4e3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.h @@ -37,9 +37,10 @@ struct amdgpu_bo; #define AMDGPU_GPU_PAGE_SHIFT 12 #define AMDGPU_GPU_PAGE_ALIGN(a) (((a) + AMDGPU_GPU_PAGE_MASK) & ~AMDGPU_GPU_PAGE_MASK) +#define AMDGPU_GPU_PAGES_IN_CPU_PAGE (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE) + struct amdgpu_gart { - u64 table_addr; - struct amdgpu_bo *robj; + struct amdgpu_bo *bo; void *ptr; unsigned num_gpu_pages; unsigned num_cpu_pages; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h index e73728d90388..ecbcefe49a98 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gds.h @@ -24,13 +24,6 @@ #ifndef __AMDGPU_GDS_H__ #define __AMDGPU_GDS_H__ -/* Because TTM request that alloacted buffer should be PAGE_SIZE aligned, - * we should report GDS/GWS/OA size as PAGE_SIZE aligned - * */ -#define AMDGPU_GDS_SHIFT 2 -#define AMDGPU_GWS_SHIFT PAGE_SHIFT -#define AMDGPU_OA_SHIFT PAGE_SHIFT - struct amdgpu_ring; struct amdgpu_bo; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 5fb156a01774..7b3d1ebda9df 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -244,16 +244,10 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, return -EINVAL; } flags |= AMDGPU_GEM_CREATE_NO_CPU_ACCESS; - if (args->in.domains == AMDGPU_GEM_DOMAIN_GDS) - size = size << AMDGPU_GDS_SHIFT; - else if (args->in.domains == AMDGPU_GEM_DOMAIN_GWS) - size = size << AMDGPU_GWS_SHIFT; - else if (args->in.domains == AMDGPU_GEM_DOMAIN_OA) - size = size << AMDGPU_OA_SHIFT; - else - return -EINVAL; + /* GDS allocations must be DW aligned */ + if (args->in.domains & AMDGPU_GEM_DOMAIN_GDS) + size = ALIGN(size, 4); } - size = roundup(size, PAGE_SIZE); if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) { r = amdgpu_bo_reserve(vm->root.base.bo, false); @@ -265,7 +259,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, r = amdgpu_gem_object_create(adev, size, args->in.alignment, (u32)(0xffffffff & args->in.domains), - flags, false, resv, &gobj); + flags, ttm_bo_type_device, resv, &gobj); if (flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID) { if (!r) { struct amdgpu_bo *abo = gem_to_amdgpu_bo(gobj); @@ -317,7 +311,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, /* create a gem object to contain this object in */ r = amdgpu_gem_object_create(adev, args->size, 0, AMDGPU_GEM_DOMAIN_CPU, - 0, 0, NULL, &gobj); + 0, ttm_bo_type_device, NULL, &gobj); if (r) return r; @@ -344,7 +338,7 @@ int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, if (r) goto free_pages; - amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); amdgpu_bo_unreserve(bo); if (r) @@ -510,7 +504,6 @@ out: * @adev: amdgpu_device pointer * @vm: vm to update * @bo_va: bo_va to update - * @list: validation list * @operation: map, unmap or clear * * Update the bo_va directly after setting its address. Errors are not @@ -519,7 +512,6 @@ out: static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct amdgpu_bo_va *bo_va, - struct list_head *list, uint32_t operation) { int r; @@ -574,16 +566,16 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, return -EINVAL; } - if (args->va_address >= AMDGPU_VA_HOLE_START && - args->va_address < AMDGPU_VA_HOLE_END) { + if (args->va_address >= AMDGPU_GMC_HOLE_START && + args->va_address < AMDGPU_GMC_HOLE_END) { dev_dbg(&dev->pdev->dev, "va_address 0x%LX is in VA hole 0x%LX-0x%LX\n", - args->va_address, AMDGPU_VA_HOLE_START, - AMDGPU_VA_HOLE_END); + args->va_address, AMDGPU_GMC_HOLE_START, + AMDGPU_GMC_HOLE_END); return -EINVAL; } - args->va_address &= AMDGPU_VA_HOLE_MASK; + args->va_address &= AMDGPU_GMC_HOLE_MASK; if ((args->flags & ~valid_flags) && (args->flags & ~prt_flags)) { dev_dbg(&dev->pdev->dev, "invalid flags combination 0x%08X\n", @@ -612,7 +604,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, return -ENOENT; abo = gem_to_amdgpu_bo(gobj); tv.bo = &abo->tbo; - tv.shared = false; + tv.shared = !!(abo->flags & AMDGPU_GEM_CREATE_VM_ALWAYS_VALID); list_add(&tv.head, &list); } else { gobj = NULL; @@ -673,7 +665,7 @@ int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, break; } if (!r && !(args->flags & AMDGPU_VM_DELAY_UPDATE) && !amdgpu_vm_debug) - amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va, &list, + amdgpu_gem_va_update_vm(adev, &fpriv->vm, bo_va, args->operation); error_backoff: @@ -768,7 +760,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv, amdgpu_display_supported_domains(adev)); r = amdgpu_gem_object_create(adev, args->size, 0, domain, AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED, - false, NULL, &gobj); + ttm_bo_type_device, NULL, &gobj); if (r) return -ENOMEM; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h new file mode 100644 index 000000000000..d63daba9b17c --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.h @@ -0,0 +1,92 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __AMDGPU_GEM_H__ +#define __AMDGPU_GEM_H__ + +#include <drm/amdgpu_drm.h> +#include <drm/drm_gem.h> + +/* + * GEM. + */ + +#define AMDGPU_GEM_DOMAIN_MAX 0x3 +#define gem_to_amdgpu_bo(gobj) container_of((gobj), struct amdgpu_bo, gem_base) + +void amdgpu_gem_object_free(struct drm_gem_object *obj); +int amdgpu_gem_object_open(struct drm_gem_object *obj, + struct drm_file *file_priv); +void amdgpu_gem_object_close(struct drm_gem_object *obj, + struct drm_file *file_priv); +unsigned long amdgpu_gem_timeout(uint64_t timeout_ns); +struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj); +struct drm_gem_object * +amdgpu_gem_prime_import_sg_table(struct drm_device *dev, + struct dma_buf_attachment *attach, + struct sg_table *sg); +struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev, + struct drm_gem_object *gobj, + int flags); +struct drm_gem_object *amdgpu_gem_prime_import(struct drm_device *dev, + struct dma_buf *dma_buf); +struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *); +void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj); +void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr); +int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma); + +/* + * GEM objects. + */ +void amdgpu_gem_force_release(struct amdgpu_device *adev); +int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, + int alignment, u32 initial_domain, + u64 flags, enum ttm_bo_type type, + struct reservation_object *resv, + struct drm_gem_object **obj); + +int amdgpu_mode_dumb_create(struct drm_file *file_priv, + struct drm_device *dev, + struct drm_mode_create_dumb *args); +int amdgpu_mode_dumb_mmap(struct drm_file *filp, + struct drm_device *dev, + uint32_t handle, uint64_t *offset_p); + +int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +int amdgpu_gem_info_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +int amdgpu_gem_userptr_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +int amdgpu_gem_mmap_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +int amdgpu_gem_wait_idle_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +int amdgpu_gem_va_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); +int amdgpu_gem_op_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); + +int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data, + struct drm_file *filp); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 239bf2a4b3c6..790fd5408ddf 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -26,9 +26,44 @@ #include "amdgpu.h" #include "amdgpu_gfx.h" +/* delay 0.1 second to enable gfx off feature */ +#define GFX_OFF_DELAY_ENABLE msecs_to_jiffies(100) + /* - * GPU scratch registers helpers function. + * GPU GFX IP block helpers function. */ + +int amdgpu_gfx_queue_to_bit(struct amdgpu_device *adev, int mec, + int pipe, int queue) +{ + int bit = 0; + + bit += mec * adev->gfx.mec.num_pipe_per_mec + * adev->gfx.mec.num_queue_per_pipe; + bit += pipe * adev->gfx.mec.num_queue_per_pipe; + bit += queue; + + return bit; +} + +void amdgpu_gfx_bit_to_queue(struct amdgpu_device *adev, int bit, + int *mec, int *pipe, int *queue) +{ + *queue = bit % adev->gfx.mec.num_queue_per_pipe; + *pipe = (bit / adev->gfx.mec.num_queue_per_pipe) + % adev->gfx.mec.num_pipe_per_mec; + *mec = (bit / adev->gfx.mec.num_queue_per_pipe) + / adev->gfx.mec.num_pipe_per_mec; + +} + +bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, + int mec, int pipe, int queue) +{ + return test_bit(amdgpu_gfx_queue_to_bit(adev, mec, pipe, queue), + adev->gfx.mec.queue_bitmap); +} + /** * amdgpu_gfx_scratch_get - Allocate a scratch register * @@ -340,3 +375,40 @@ void amdgpu_gfx_compute_mqd_sw_fini(struct amdgpu_device *adev) &ring->mqd_gpu_addr, &ring->mqd_ptr); } + +/* amdgpu_gfx_off_ctrl - Handle gfx off feature enable/disable + * + * @adev: amdgpu_device pointer + * @bool enable true: enable gfx off feature, false: disable gfx off feature + * + * 1. gfx off feature will be enabled by gfx ip after gfx cg gp enabled. + * 2. other client can send request to disable gfx off feature, the request should be honored. + * 3. other client can cancel their request of disable gfx off feature + * 4. other client should not send request to enable gfx off feature before disable gfx off feature. + */ + +void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable) +{ + if (!(adev->powerplay.pp_feature & PP_GFXOFF_MASK)) + return; + + if (!adev->powerplay.pp_funcs->set_powergating_by_smu) + return; + + + mutex_lock(&adev->gfx.gfx_off_mutex); + + if (!enable) + adev->gfx.gfx_off_req_count++; + else if (adev->gfx.gfx_off_req_count > 0) + adev->gfx.gfx_off_req_count--; + + if (enable && !adev->gfx.gfx_off_state && !adev->gfx.gfx_off_req_count) { + schedule_delayed_work(&adev->gfx.gfx_off_delay_work, GFX_OFF_DELAY_ENABLE); + } else if (!enable && adev->gfx.gfx_off_state) { + if (!amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, false)) + adev->gfx.gfx_off_state = false; + } + + mutex_unlock(&adev->gfx.gfx_off_mutex); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index 1f279050d334..b61b5c11aead 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -24,28 +24,297 @@ #ifndef __AMDGPU_GFX_H__ #define __AMDGPU_GFX_H__ -int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg); -void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg); +/* + * GFX stuff + */ +#include "clearstate_defs.h" +#include "amdgpu_ring.h" -void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, - unsigned max_sh); +/* GFX current status */ +#define AMDGPU_GFX_NORMAL_MODE 0x00000000L +#define AMDGPU_GFX_SAFE_MODE 0x00000001L +#define AMDGPU_GFX_PG_DISABLED_MODE 0x00000002L +#define AMDGPU_GFX_CG_DISABLED_MODE 0x00000004L +#define AMDGPU_GFX_LBPW_DISABLED_MODE 0x00000008L -void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev); -int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_irq_src *irq); +struct amdgpu_rlc_funcs { + void (*enter_safe_mode)(struct amdgpu_device *adev); + void (*exit_safe_mode)(struct amdgpu_device *adev); +}; -void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring, - struct amdgpu_irq_src *irq); +struct amdgpu_rlc { + /* for power gating */ + struct amdgpu_bo *save_restore_obj; + uint64_t save_restore_gpu_addr; + volatile uint32_t *sr_ptr; + const u32 *reg_list; + u32 reg_list_size; + /* for clear state */ + struct amdgpu_bo *clear_state_obj; + uint64_t clear_state_gpu_addr; + volatile uint32_t *cs_ptr; + const struct cs_section_def *cs_data; + u32 clear_state_size; + /* for cp tables */ + struct amdgpu_bo *cp_table_obj; + uint64_t cp_table_gpu_addr; + volatile uint32_t *cp_table_ptr; + u32 cp_table_size; -void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev); -int amdgpu_gfx_kiq_init(struct amdgpu_device *adev, - unsigned hpd_size); + /* safe mode for updating CG/PG state */ + bool in_safe_mode; + const struct amdgpu_rlc_funcs *funcs; -int amdgpu_gfx_compute_mqd_sw_init(struct amdgpu_device *adev, - unsigned mqd_size); -void amdgpu_gfx_compute_mqd_sw_fini(struct amdgpu_device *adev); + /* for firmware data */ + u32 save_and_restore_offset; + u32 clear_state_descriptor_offset; + u32 avail_scratch_ram_locations; + u32 reg_restore_list_size; + u32 reg_list_format_start; + u32 reg_list_format_separate_start; + u32 starting_offsets_start; + u32 reg_list_format_size_bytes; + u32 reg_list_size_bytes; + u32 reg_list_format_direct_reg_list_length; + u32 save_restore_list_cntl_size_bytes; + u32 save_restore_list_gpm_size_bytes; + u32 save_restore_list_srm_size_bytes; + + u32 *register_list_format; + u32 *register_restore; + u8 *save_restore_list_cntl; + u8 *save_restore_list_gpm; + u8 *save_restore_list_srm; + + bool is_rlc_v2_1; +}; + +#define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES + +struct amdgpu_mec { + struct amdgpu_bo *hpd_eop_obj; + u64 hpd_eop_gpu_addr; + struct amdgpu_bo *mec_fw_obj; + u64 mec_fw_gpu_addr; + u32 num_mec; + u32 num_pipe_per_mec; + u32 num_queue_per_pipe; + void *mqd_backup[AMDGPU_MAX_COMPUTE_RINGS + 1]; + + /* These are the resources for which amdgpu takes ownership */ + DECLARE_BITMAP(queue_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); +}; + +struct amdgpu_kiq { + u64 eop_gpu_addr; + struct amdgpu_bo *eop_obj; + spinlock_t ring_lock; + struct amdgpu_ring ring; + struct amdgpu_irq_src irq; +}; + +/* + * GPU scratch registers structures, functions & helpers + */ +struct amdgpu_scratch { + unsigned num_reg; + uint32_t reg_base; + uint32_t free_mask; +}; + +/* + * GFX configurations + */ +#define AMDGPU_GFX_MAX_SE 4 +#define AMDGPU_GFX_MAX_SH_PER_SE 2 + +struct amdgpu_rb_config { + uint32_t rb_backend_disable; + uint32_t user_rb_backend_disable; + uint32_t raster_config; + uint32_t raster_config_1; +}; + +struct gb_addr_config { + uint16_t pipe_interleave_size; + uint8_t num_pipes; + uint8_t max_compress_frags; + uint8_t num_banks; + uint8_t num_se; + uint8_t num_rb_per_se; +}; + +struct amdgpu_gfx_config { + unsigned max_shader_engines; + unsigned max_tile_pipes; + unsigned max_cu_per_sh; + unsigned max_sh_per_se; + unsigned max_backends_per_se; + unsigned max_texture_channel_caches; + unsigned max_gprs; + unsigned max_gs_threads; + unsigned max_hw_contexts; + unsigned sc_prim_fifo_size_frontend; + unsigned sc_prim_fifo_size_backend; + unsigned sc_hiz_tile_fifo_size; + unsigned sc_earlyz_tile_fifo_size; + + unsigned num_tile_pipes; + unsigned backend_enable_mask; + unsigned mem_max_burst_length_bytes; + unsigned mem_row_size_in_kb; + unsigned shader_engine_tile_size; + unsigned num_gpus; + unsigned multi_gpu_tile_size; + unsigned mc_arb_ramcfg; + unsigned gb_addr_config; + unsigned num_rbs; + unsigned gs_vgt_table_depth; + unsigned gs_prim_buffer_depth; + + uint32_t tile_mode_array[32]; + uint32_t macrotile_mode_array[16]; + + struct gb_addr_config gb_addr_config_fields; + struct amdgpu_rb_config rb_config[AMDGPU_GFX_MAX_SE][AMDGPU_GFX_MAX_SH_PER_SE]; + + /* gfx configure feature */ + uint32_t double_offchip_lds_buf; + /* cached value of DB_DEBUG2 */ + uint32_t db_debug2; +}; + +struct amdgpu_cu_info { + uint32_t simd_per_cu; + uint32_t max_waves_per_simd; + uint32_t wave_front_size; + uint32_t max_scratch_slots_per_cu; + uint32_t lds_size; + + /* total active CU number */ + uint32_t number; + uint32_t ao_cu_mask; + uint32_t ao_cu_bitmap[4][4]; + uint32_t bitmap[4][4]; +}; + +struct amdgpu_gfx_funcs { + /* get the gpu clock counter */ + uint64_t (*get_gpu_clock_counter)(struct amdgpu_device *adev); + void (*select_se_sh)(struct amdgpu_device *adev, u32 se_num, + u32 sh_num, u32 instance); + void (*read_wave_data)(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t *dst, int *no_fields); + void (*read_wave_vgprs)(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t thread, uint32_t start, + uint32_t size, uint32_t *dst); + void (*read_wave_sgprs)(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t start, uint32_t size, + uint32_t *dst); + void (*select_me_pipe_q)(struct amdgpu_device *adev, u32 me, u32 pipe, + u32 queue); +}; + +struct amdgpu_ngg_buf { + struct amdgpu_bo *bo; + uint64_t gpu_addr; + uint32_t size; + uint32_t bo_size; +}; + +enum { + NGG_PRIM = 0, + NGG_POS, + NGG_CNTL, + NGG_PARAM, + NGG_BUF_MAX +}; + +struct amdgpu_ngg { + struct amdgpu_ngg_buf buf[NGG_BUF_MAX]; + uint32_t gds_reserve_addr; + uint32_t gds_reserve_size; + bool init; +}; + +struct sq_work { + struct work_struct work; + unsigned ih_data; +}; + +struct amdgpu_gfx { + struct mutex gpu_clock_mutex; + struct amdgpu_gfx_config config; + struct amdgpu_rlc rlc; + struct amdgpu_mec mec; + struct amdgpu_kiq kiq; + struct amdgpu_scratch scratch; + const struct firmware *me_fw; /* ME firmware */ + uint32_t me_fw_version; + const struct firmware *pfp_fw; /* PFP firmware */ + uint32_t pfp_fw_version; + const struct firmware *ce_fw; /* CE firmware */ + uint32_t ce_fw_version; + const struct firmware *rlc_fw; /* RLC firmware */ + uint32_t rlc_fw_version; + const struct firmware *mec_fw; /* MEC firmware */ + uint32_t mec_fw_version; + const struct firmware *mec2_fw; /* MEC2 firmware */ + uint32_t mec2_fw_version; + uint32_t me_feature_version; + uint32_t ce_feature_version; + uint32_t pfp_feature_version; + uint32_t rlc_feature_version; + uint32_t rlc_srlc_fw_version; + uint32_t rlc_srlc_feature_version; + uint32_t rlc_srlg_fw_version; + uint32_t rlc_srlg_feature_version; + uint32_t rlc_srls_fw_version; + uint32_t rlc_srls_feature_version; + uint32_t mec_feature_version; + uint32_t mec2_feature_version; + bool mec_fw_write_wait; + bool me_fw_write_wait; + struct amdgpu_ring gfx_ring[AMDGPU_MAX_GFX_RINGS]; + unsigned num_gfx_rings; + struct amdgpu_ring compute_ring[AMDGPU_MAX_COMPUTE_RINGS]; + unsigned num_compute_rings; + struct amdgpu_irq_src eop_irq; + struct amdgpu_irq_src priv_reg_irq; + struct amdgpu_irq_src priv_inst_irq; + struct amdgpu_irq_src cp_ecc_error_irq; + struct amdgpu_irq_src sq_irq; + struct sq_work sq_work; + + /* gfx status */ + uint32_t gfx_current_status; + /* ce ram size*/ + unsigned ce_ram_size; + struct amdgpu_cu_info cu_info; + const struct amdgpu_gfx_funcs *funcs; + + /* reset mask */ + uint32_t grbm_soft_reset; + uint32_t srbm_soft_reset; + + /* NGG */ + struct amdgpu_ngg ngg; + + /* gfx off */ + bool gfx_off_state; /* true: enabled, false: disabled */ + struct mutex gfx_off_mutex; + uint32_t gfx_off_req_count; /* default 1, enable gfx off: dec 1, disable gfx off: add 1 */ + struct delayed_work gfx_off_delay_work; + + /* pipe reservation */ + struct mutex pipe_reserve_mutex; + DECLARE_BITMAP (pipe_reserve_bitmap, AMDGPU_MAX_COMPUTE_QUEUES); +}; + +#define amdgpu_gfx_get_gpu_clock_counter(adev) (adev)->gfx.funcs->get_gpu_clock_counter((adev)) +#define amdgpu_gfx_select_se_sh(adev, se, sh, instance) (adev)->gfx.funcs->select_se_sh((adev), (se), (sh), (instance)) +#define amdgpu_gfx_select_me_pipe_q(adev, me, pipe, q) (adev)->gfx.funcs->select_me_pipe_q((adev), (me), (pipe), (q)) /** * amdgpu_gfx_create_bitmask - create a bitmask @@ -60,34 +329,34 @@ static inline u32 amdgpu_gfx_create_bitmask(u32 bit_width) return (u32)((1ULL << bit_width) - 1); } -static inline int amdgpu_gfx_queue_to_bit(struct amdgpu_device *adev, - int mec, int pipe, int queue) -{ - int bit = 0; +int amdgpu_gfx_scratch_get(struct amdgpu_device *adev, uint32_t *reg); +void amdgpu_gfx_scratch_free(struct amdgpu_device *adev, uint32_t reg); - bit += mec * adev->gfx.mec.num_pipe_per_mec - * adev->gfx.mec.num_queue_per_pipe; - bit += pipe * adev->gfx.mec.num_queue_per_pipe; - bit += queue; +void amdgpu_gfx_parse_disable_cu(unsigned *mask, unsigned max_se, + unsigned max_sh); - return bit; -} +int amdgpu_gfx_kiq_init_ring(struct amdgpu_device *adev, + struct amdgpu_ring *ring, + struct amdgpu_irq_src *irq); -static inline void amdgpu_gfx_bit_to_queue(struct amdgpu_device *adev, int bit, - int *mec, int *pipe, int *queue) -{ - *queue = bit % adev->gfx.mec.num_queue_per_pipe; - *pipe = (bit / adev->gfx.mec.num_queue_per_pipe) - % adev->gfx.mec.num_pipe_per_mec; - *mec = (bit / adev->gfx.mec.num_queue_per_pipe) - / adev->gfx.mec.num_pipe_per_mec; +void amdgpu_gfx_kiq_free_ring(struct amdgpu_ring *ring, + struct amdgpu_irq_src *irq); -} -static inline bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, - int mec, int pipe, int queue) -{ - return test_bit(amdgpu_gfx_queue_to_bit(adev, mec, pipe, queue), - adev->gfx.mec.queue_bitmap); -} +void amdgpu_gfx_kiq_fini(struct amdgpu_device *adev); +int amdgpu_gfx_kiq_init(struct amdgpu_device *adev, + unsigned hpd_size); + +int amdgpu_gfx_compute_mqd_sw_init(struct amdgpu_device *adev, + unsigned mqd_size); +void amdgpu_gfx_compute_mqd_sw_fini(struct amdgpu_device *adev); + +void amdgpu_gfx_compute_queue_acquire(struct amdgpu_device *adev); +int amdgpu_gfx_queue_to_bit(struct amdgpu_device *adev, int mec, + int pipe, int queue); +void amdgpu_gfx_bit_to_queue(struct amdgpu_device *adev, int bit, + int *mec, int *pipe, int *queue); +bool amdgpu_gfx_is_mec_queue_enabled(struct amdgpu_device *adev, int mec, + int pipe, int queue); +void amdgpu_gfx_off_ctrl(struct amdgpu_device *adev, bool enable); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c new file mode 100644 index 000000000000..d73367cab4f3 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -0,0 +1,215 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + */ + +#include "amdgpu.h" + +/** + * amdgpu_gmc_get_pde_for_bo - get the PDE for a BO + * + * @bo: the BO to get the PDE for + * @level: the level in the PD hirarchy + * @addr: resulting addr + * @flags: resulting flags + * + * Get the address and flags to be used for a PDE (Page Directory Entry). + */ +void amdgpu_gmc_get_pde_for_bo(struct amdgpu_bo *bo, int level, + uint64_t *addr, uint64_t *flags) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + struct ttm_dma_tt *ttm; + + switch (bo->tbo.mem.mem_type) { + case TTM_PL_TT: + ttm = container_of(bo->tbo.ttm, struct ttm_dma_tt, ttm); + *addr = ttm->dma_address[0]; + break; + case TTM_PL_VRAM: + *addr = amdgpu_bo_gpu_offset(bo); + break; + default: + *addr = 0; + break; + } + *flags = amdgpu_ttm_tt_pde_flags(bo->tbo.ttm, &bo->tbo.mem); + amdgpu_gmc_get_vm_pde(adev, level, addr, flags); +} + +/** + * amdgpu_gmc_pd_addr - return the address of the root directory + * + */ +uint64_t amdgpu_gmc_pd_addr(struct amdgpu_bo *bo) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + uint64_t pd_addr; + + /* TODO: move that into ASIC specific code */ + if (adev->asic_type >= CHIP_VEGA10) { + uint64_t flags = AMDGPU_PTE_VALID; + + amdgpu_gmc_get_pde_for_bo(bo, -1, &pd_addr, &flags); + pd_addr |= flags; + } else { + pd_addr = amdgpu_bo_gpu_offset(bo); + } + return pd_addr; +} + +/** + * amdgpu_gmc_agp_addr - return the address in the AGP address space + * + * @tbo: TTM BO which needs the address, must be in GTT domain + * + * Tries to figure out how to access the BO through the AGP aperture. Returns + * AMDGPU_BO_INVALID_OFFSET if that is not possible. + */ +uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); + struct ttm_dma_tt *ttm; + + if (bo->num_pages != 1 || bo->ttm->caching_state == tt_cached) + return AMDGPU_BO_INVALID_OFFSET; + + ttm = container_of(bo->ttm, struct ttm_dma_tt, ttm); + if (ttm->dma_address[0] + PAGE_SIZE >= adev->gmc.agp_size) + return AMDGPU_BO_INVALID_OFFSET; + + return adev->gmc.agp_start + ttm->dma_address[0]; +} + +/** + * amdgpu_gmc_vram_location - try to find VRAM location + * + * @adev: amdgpu device structure holding all necessary informations + * @mc: memory controller structure holding memory informations + * @base: base address at which to put VRAM + * + * Function will try to place VRAM at base address provided + * as parameter. + */ +void amdgpu_gmc_vram_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc, + u64 base) +{ + uint64_t limit = (uint64_t)amdgpu_vram_limit << 20; + + mc->vram_start = base; + mc->vram_end = mc->vram_start + mc->mc_vram_size - 1; + if (limit && limit < mc->real_vram_size) + mc->real_vram_size = limit; + + if (mc->xgmi.num_physical_nodes == 0) { + mc->fb_start = mc->vram_start; + mc->fb_end = mc->vram_end; + } + dev_info(adev->dev, "VRAM: %lluM 0x%016llX - 0x%016llX (%lluM used)\n", + mc->mc_vram_size >> 20, mc->vram_start, + mc->vram_end, mc->real_vram_size >> 20); +} + +/** + * amdgpu_gmc_gart_location - try to find GART location + * + * @adev: amdgpu device structure holding all necessary informations + * @mc: memory controller structure holding memory informations + * + * Function will place try to place GART before or after VRAM. + * + * If GART size is bigger than space left then we ajust GART size. + * Thus function will never fails. + */ +void amdgpu_gmc_gart_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc) +{ + const uint64_t four_gb = 0x100000000ULL; + u64 size_af, size_bf; + /*To avoid the hole, limit the max mc address to AMDGPU_GMC_HOLE_START*/ + u64 max_mc_address = min(adev->gmc.mc_mask, AMDGPU_GMC_HOLE_START - 1); + + mc->gart_size += adev->pm.smu_prv_buffer_size; + + /* VCE doesn't like it when BOs cross a 4GB segment, so align + * the GART base on a 4GB boundary as well. + */ + size_bf = mc->fb_start; + size_af = max_mc_address + 1 - ALIGN(mc->fb_end + 1, four_gb); + + if (mc->gart_size > max(size_bf, size_af)) { + dev_warn(adev->dev, "limiting GART\n"); + mc->gart_size = max(size_bf, size_af); + } + + if ((size_bf >= mc->gart_size && size_bf < size_af) || + (size_af < mc->gart_size)) + mc->gart_start = 0; + else + mc->gart_start = max_mc_address - mc->gart_size + 1; + + mc->gart_start &= ~(four_gb - 1); + mc->gart_end = mc->gart_start + mc->gart_size - 1; + dev_info(adev->dev, "GART: %lluM 0x%016llX - 0x%016llX\n", + mc->gart_size >> 20, mc->gart_start, mc->gart_end); +} + +/** + * amdgpu_gmc_agp_location - try to find AGP location + * @adev: amdgpu device structure holding all necessary informations + * @mc: memory controller structure holding memory informations + * + * Function will place try to find a place for the AGP BAR in the MC address + * space. + * + * AGP BAR will be assigned the largest available hole in the address space. + * Should be called after VRAM and GART locations are setup. + */ +void amdgpu_gmc_agp_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc) +{ + const uint64_t sixteen_gb = 1ULL << 34; + const uint64_t sixteen_gb_mask = ~(sixteen_gb - 1); + u64 size_af, size_bf; + + if (mc->fb_start > mc->gart_start) { + size_bf = (mc->fb_start & sixteen_gb_mask) - + ALIGN(mc->gart_end + 1, sixteen_gb); + size_af = mc->mc_mask + 1 - ALIGN(mc->fb_end + 1, sixteen_gb); + } else { + size_bf = mc->fb_start & sixteen_gb_mask; + size_af = (mc->gart_start & sixteen_gb_mask) - + ALIGN(mc->fb_end + 1, sixteen_gb); + } + + if (size_bf > size_af) { + mc->agp_start = (mc->fb_start - size_bf) & sixteen_gb_mask; + mc->agp_size = size_bf; + } else { + mc->agp_start = ALIGN(mc->fb_end + 1, sixteen_gb); + mc->agp_size = size_af; + } + + mc->agp_end = mc->agp_start + mc->agp_size - 1; + dev_info(adev->dev, "AGP: %lluM 0x%016llX - 0x%016llX\n", + mc->agp_size >> 20, mc->agp_start, mc->agp_end); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index 893c2490b783..6fa7ef446e46 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -30,6 +30,19 @@ #include "amdgpu_irq.h" +/* VA hole for 48bit addresses on Vega10 */ +#define AMDGPU_GMC_HOLE_START 0x0000800000000000ULL +#define AMDGPU_GMC_HOLE_END 0xffff800000000000ULL + +/* + * Hardware is programmed as if the hole doesn't exists with start and end + * address values. + * + * This mask is used to remove the upper 16bits of the VA and so come up with + * the linear addr value. + */ +#define AMDGPU_GMC_HOLE_MASK 0x0000ffffffffffffULL + struct firmware; /* @@ -74,6 +87,20 @@ struct amdgpu_gmc_funcs { u64 *dst, u64 *flags); }; +struct amdgpu_xgmi { + /* from psp */ + u64 device_id; + u64 hive_id; + /* fixed per family */ + u64 node_segment_size; + /* physical node (0-3) */ + unsigned physical_node_id; + /* number of nodes (0-4) */ + unsigned num_physical_nodes; + /* gpu list in the same hive */ + struct list_head head; +}; + struct amdgpu_gmc { resource_size_t aper_size; resource_size_t aper_base; @@ -81,11 +108,22 @@ struct amdgpu_gmc { * about vram size near mc fb location */ u64 mc_vram_size; u64 visible_vram_size; + u64 agp_size; + u64 agp_start; + u64 agp_end; u64 gart_size; u64 gart_start; u64 gart_end; u64 vram_start; u64 vram_end; + /* FB region , it's same as local vram region in single GPU, in XGMI + * configuration, this region covers all GPUs in the same hive , + * each GPU in the hive has the same view of this FB region . + * GPU0's vram starts at offset (0 * segment size) , + * GPU1 starts at offset (1 * segment size), etc. + */ + u64 fb_start; + u64 fb_end; unsigned vram_width; u64 real_vram_size; int vram_mtrr; @@ -105,8 +143,58 @@ struct amdgpu_gmc { /* protects concurrent invalidation */ spinlock_t invalidate_lock; bool translate_further; + struct kfd_vm_fault_info *vm_fault_info; + atomic_t vm_fault_info_updated; const struct amdgpu_gmc_funcs *gmc_funcs; + + struct amdgpu_xgmi xgmi; }; +#define amdgpu_gmc_flush_gpu_tlb(adev, vmid) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid)) +#define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr)) +#define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid)) +#define amdgpu_gmc_set_pte_pde(adev, pt, idx, addr, flags) (adev)->gmc.gmc_funcs->set_pte_pde((adev), (pt), (idx), (addr), (flags)) +#define amdgpu_gmc_get_vm_pde(adev, level, dst, flags) (adev)->gmc.gmc_funcs->get_vm_pde((adev), (level), (dst), (flags)) +#define amdgpu_gmc_get_pte_flags(adev, flags) (adev)->gmc.gmc_funcs->get_vm_pte_flags((adev),(flags)) + +/** + * amdgpu_gmc_vram_full_visible - Check if full VRAM is visible through the BAR + * + * @adev: amdgpu_device pointer + * + * Returns: + * True if full VRAM is visible through the BAR + */ +static inline bool amdgpu_gmc_vram_full_visible(struct amdgpu_gmc *gmc) +{ + WARN_ON(gmc->real_vram_size < gmc->visible_vram_size); + + return (gmc->real_vram_size == gmc->visible_vram_size); +} + +/** + * amdgpu_gmc_sign_extend - sign extend the given gmc address + * + * @addr: address to extend + */ +static inline uint64_t amdgpu_gmc_sign_extend(uint64_t addr) +{ + if (addr >= AMDGPU_GMC_HOLE_START) + addr |= AMDGPU_GMC_HOLE_END; + + return addr; +} + +void amdgpu_gmc_get_pde_for_bo(struct amdgpu_bo *bo, int level, + uint64_t *addr, uint64_t *flags); +uint64_t amdgpu_gmc_pd_addr(struct amdgpu_bo *bo); +uint64_t amdgpu_gmc_agp_addr(struct ttm_buffer_object *bo); +void amdgpu_gmc_vram_location(struct amdgpu_device *adev, struct amdgpu_gmc *mc, + u64 base); +void amdgpu_gmc_gart_location(struct amdgpu_device *adev, + struct amdgpu_gmc *mc); +void amdgpu_gmc_agp_location(struct amdgpu_device *adev, + struct amdgpu_gmc *mc); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 7aaa263ad8c7..b8963b725dfa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -32,6 +32,7 @@ #include <drm/amdgpu_drm.h> #include "amdgpu.h" #include "atom.h" +#include "amdgpu_trace.h" #define AMDGPU_IB_TEST_TIMEOUT msecs_to_jiffies(1000) @@ -139,7 +140,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, /* ring tests don't use a job */ if (job) { vm = job->vm; - fence_ctx = job->fence_ctx; + fence_ctx = job->base.s_fence->scheduled.context; } else { vm = NULL; fence_ctx = 0; @@ -164,10 +165,16 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, return r; } + need_ctx_switch = ring->current_ctx != fence_ctx; if (ring->funcs->emit_pipeline_sync && job && ((tmp = amdgpu_sync_get_fence(&job->sched_sync, NULL)) || + (amdgpu_sriov_vf(adev) && need_ctx_switch) || amdgpu_vm_need_pipeline_sync(ring, job))) { need_pipe_sync = true; + + if (tmp) + trace_amdgpu_ib_pipe_sync(job, tmp); + dma_fence_put(tmp); } @@ -196,7 +203,6 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, } skip_preamble = ring->current_ctx == fence_ctx; - need_ctx_switch = ring->current_ctx != fence_ctx; if (job && ring->funcs->emit_cntxcntl) { if (need_ctx_switch) status |= AMDGPU_HAVE_CTX_SWITCH; @@ -348,12 +354,21 @@ int amdgpu_ib_ring_tests(struct amdgpu_device *adev) if (!ring || !ring->ready) continue; + /* skip IB tests for KIQ in general for the below reasons: + * 1. We never submit IBs to the KIQ + * 2. KIQ doesn't use the EOP interrupts, + * we use some other CP interrupt. + */ + if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) + continue; + /* MM engine need more time */ if (ring->funcs->type == AMDGPU_RING_TYPE_UVD || ring->funcs->type == AMDGPU_RING_TYPE_VCE || ring->funcs->type == AMDGPU_RING_TYPE_UVD_ENC || ring->funcs->type == AMDGPU_RING_TYPE_VCN_DEC || - ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) + ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC || + ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) tmo = tmo_mm; else tmo = tmo_gfx; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c index a1c78f90eadf..df9b173c3d0b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ids.c @@ -574,15 +574,10 @@ void amdgpu_vmid_mgr_init(struct amdgpu_device *adev) /* skip over VMID 0, since it is the system VM */ for (j = 1; j < id_mgr->num_ids; ++j) { amdgpu_vmid_reset(adev, i, j); - amdgpu_sync_create(&id_mgr->ids[i].active); + amdgpu_sync_create(&id_mgr->ids[j].active); list_add_tail(&id_mgr->ids[j].list, &id_mgr->ids_lru); } } - - adev->vm_manager.fence_context = - dma_fence_context_alloc(AMDGPU_MAX_RINGS); - for (i = 0; i < AMDGPU_MAX_RINGS; ++i) - adev->vm_manager.seqno[i] = 0; } /** diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c index 06373d44b3da..8af67f649660 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c @@ -24,46 +24,21 @@ #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_ih.h" -#include "amdgpu_amdkfd.h" - -/** - * amdgpu_ih_ring_alloc - allocate memory for the IH ring - * - * @adev: amdgpu_device pointer - * - * Allocate a ring buffer for the interrupt controller. - * Returns 0 for success, errors for failure. - */ -static int amdgpu_ih_ring_alloc(struct amdgpu_device *adev) -{ - int r; - - /* Allocate ring buffer */ - if (adev->irq.ih.ring_obj == NULL) { - r = amdgpu_bo_create_kernel(adev, adev->irq.ih.ring_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, - &adev->irq.ih.ring_obj, - &adev->irq.ih.gpu_addr, - (void **)&adev->irq.ih.ring); - if (r) { - DRM_ERROR("amdgpu: failed to create ih ring buffer (%d).\n", r); - return r; - } - } - return 0; -} /** * amdgpu_ih_ring_init - initialize the IH state * * @adev: amdgpu_device pointer + * @ih: ih ring to initialize + * @ring_size: ring size to allocate + * @use_bus_addr: true when we can use dma_alloc_coherent * * Initializes the IH state and allocates a buffer * for the IH ring buffer. * Returns 0 for success, errors for failure. */ -int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size, - bool use_bus_addr) +int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, + unsigned ring_size, bool use_bus_addr) { u32 rb_bufsz; int r; @@ -71,70 +46,76 @@ int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size, /* Align ring size */ rb_bufsz = order_base_2(ring_size / 4); ring_size = (1 << rb_bufsz) * 4; - adev->irq.ih.ring_size = ring_size; - adev->irq.ih.ptr_mask = adev->irq.ih.ring_size - 1; - adev->irq.ih.rptr = 0; - adev->irq.ih.use_bus_addr = use_bus_addr; + ih->ring_size = ring_size; + ih->ptr_mask = ih->ring_size - 1; + ih->rptr = 0; + ih->use_bus_addr = use_bus_addr; - if (adev->irq.ih.use_bus_addr) { - if (!adev->irq.ih.ring) { - /* add 8 bytes for the rptr/wptr shadows and - * add them to the end of the ring allocation. - */ - adev->irq.ih.ring = pci_alloc_consistent(adev->pdev, - adev->irq.ih.ring_size + 8, - &adev->irq.ih.rb_dma_addr); - if (adev->irq.ih.ring == NULL) - return -ENOMEM; - memset((void *)adev->irq.ih.ring, 0, adev->irq.ih.ring_size + 8); - adev->irq.ih.wptr_offs = (adev->irq.ih.ring_size / 4) + 0; - adev->irq.ih.rptr_offs = (adev->irq.ih.ring_size / 4) + 1; - } - return 0; + if (use_bus_addr) { + if (ih->ring) + return 0; + + /* add 8 bytes for the rptr/wptr shadows and + * add them to the end of the ring allocation. + */ + ih->ring = dma_alloc_coherent(adev->dev, ih->ring_size + 8, + &ih->rb_dma_addr, GFP_KERNEL); + if (ih->ring == NULL) + return -ENOMEM; + + memset((void *)ih->ring, 0, ih->ring_size + 8); + ih->wptr_offs = (ih->ring_size / 4) + 0; + ih->rptr_offs = (ih->ring_size / 4) + 1; } else { - r = amdgpu_device_wb_get(adev, &adev->irq.ih.wptr_offs); + r = amdgpu_device_wb_get(adev, &ih->wptr_offs); + if (r) + return r; + + r = amdgpu_device_wb_get(adev, &ih->rptr_offs); if (r) { - dev_err(adev->dev, "(%d) ih wptr_offs wb alloc failed\n", r); + amdgpu_device_wb_free(adev, ih->wptr_offs); return r; } - r = amdgpu_device_wb_get(adev, &adev->irq.ih.rptr_offs); + r = amdgpu_bo_create_kernel(adev, ih->ring_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_GTT, + &ih->ring_obj, &ih->gpu_addr, + (void **)&ih->ring); if (r) { - amdgpu_device_wb_free(adev, adev->irq.ih.wptr_offs); - dev_err(adev->dev, "(%d) ih rptr_offs wb alloc failed\n", r); + amdgpu_device_wb_free(adev, ih->rptr_offs); + amdgpu_device_wb_free(adev, ih->wptr_offs); return r; } - - return amdgpu_ih_ring_alloc(adev); } + return 0; } /** * amdgpu_ih_ring_fini - tear down the IH state * * @adev: amdgpu_device pointer + * @ih: ih ring to tear down * * Tears down the IH state and frees buffer * used for the IH ring buffer. */ -void amdgpu_ih_ring_fini(struct amdgpu_device *adev) +void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih) { - if (adev->irq.ih.use_bus_addr) { - if (adev->irq.ih.ring) { - /* add 8 bytes for the rptr/wptr shadows and - * add them to the end of the ring allocation. - */ - pci_free_consistent(adev->pdev, adev->irq.ih.ring_size + 8, - (void *)adev->irq.ih.ring, - adev->irq.ih.rb_dma_addr); - adev->irq.ih.ring = NULL; - } + if (ih->use_bus_addr) { + if (!ih->ring) + return; + + /* add 8 bytes for the rptr/wptr shadows and + * add them to the end of the ring allocation. + */ + dma_free_coherent(adev->dev, ih->ring_size + 8, + (void *)ih->ring, ih->rb_dma_addr); + ih->ring = NULL; } else { - amdgpu_bo_free_kernel(&adev->irq.ih.ring_obj, - &adev->irq.ih.gpu_addr, - (void **)&adev->irq.ih.ring); - amdgpu_device_wb_free(adev, adev->irq.ih.wptr_offs); - amdgpu_device_wb_free(adev, adev->irq.ih.rptr_offs); + amdgpu_bo_free_kernel(&ih->ring_obj, &ih->gpu_addr, + (void **)&ih->ring); + amdgpu_device_wb_free(adev, ih->wptr_offs); + amdgpu_device_wb_free(adev, ih->rptr_offs); } } @@ -142,133 +123,45 @@ void amdgpu_ih_ring_fini(struct amdgpu_device *adev) * amdgpu_ih_process - interrupt handler * * @adev: amdgpu_device pointer + * @ih: ih ring to process * * Interrupt hander (VI), walk the IH ring. * Returns irq process return code. */ -int amdgpu_ih_process(struct amdgpu_device *adev) +int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, + void (*callback)(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih)) { - struct amdgpu_iv_entry entry; u32 wptr; - if (!adev->irq.ih.enabled || adev->shutdown) + if (!ih->enabled || adev->shutdown) return IRQ_NONE; wptr = amdgpu_ih_get_wptr(adev); restart_ih: /* is somebody else already processing irqs? */ - if (atomic_xchg(&adev->irq.ih.lock, 1)) + if (atomic_xchg(&ih->lock, 1)) return IRQ_NONE; - DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, adev->irq.ih.rptr, wptr); + DRM_DEBUG("%s: rptr %d, wptr %d\n", __func__, ih->rptr, wptr); /* Order reading of wptr vs. reading of IH ring data */ rmb(); - while (adev->irq.ih.rptr != wptr) { - u32 ring_index = adev->irq.ih.rptr >> 2; - - /* Prescreening of high-frequency interrupts */ - if (!amdgpu_ih_prescreen_iv(adev)) { - adev->irq.ih.rptr &= adev->irq.ih.ptr_mask; - continue; - } - - /* Before dispatching irq to IP blocks, send it to amdkfd */ - amdgpu_amdkfd_interrupt(adev, - (const void *) &adev->irq.ih.ring[ring_index]); - - entry.iv_entry = (const uint32_t *) - &adev->irq.ih.ring[ring_index]; - amdgpu_ih_decode_iv(adev, &entry); - adev->irq.ih.rptr &= adev->irq.ih.ptr_mask; - - amdgpu_irq_dispatch(adev, &entry); + while (ih->rptr != wptr) { + callback(adev, ih); + ih->rptr &= ih->ptr_mask; } + amdgpu_ih_set_rptr(adev); - atomic_set(&adev->irq.ih.lock, 0); + atomic_set(&ih->lock, 0); /* make sure wptr hasn't changed while processing */ wptr = amdgpu_ih_get_wptr(adev); - if (wptr != adev->irq.ih.rptr) + if (wptr != ih->rptr) goto restart_ih; return IRQ_HANDLED; } -/** - * amdgpu_ih_add_fault - Add a page fault record - * - * @adev: amdgpu device pointer - * @key: 64-bit encoding of PASID and address - * - * This should be called when a retry page fault interrupt is - * received. If this is a new page fault, it will be added to a hash - * table. The return value indicates whether this is a new fault, or - * a fault that was already known and is already being handled. - * - * If there are too many pending page faults, this will fail. Retry - * interrupts should be ignored in this case until there is enough - * free space. - * - * Returns 0 if the fault was added, 1 if the fault was already known, - * -ENOSPC if there are too many pending faults. - */ -int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key) -{ - unsigned long flags; - int r = -ENOSPC; - - if (WARN_ON_ONCE(!adev->irq.ih.faults)) - /* Should be allocated in <IP>_ih_sw_init on GPUs that - * support retry faults and require retry filtering. - */ - return r; - - spin_lock_irqsave(&adev->irq.ih.faults->lock, flags); - - /* Only let the hash table fill up to 50% for best performance */ - if (adev->irq.ih.faults->count >= (1 << (AMDGPU_PAGEFAULT_HASH_BITS-1))) - goto unlock_out; - - r = chash_table_copy_in(&adev->irq.ih.faults->hash, key, NULL); - if (!r) - adev->irq.ih.faults->count++; - - /* chash_table_copy_in should never fail unless we're losing count */ - WARN_ON_ONCE(r < 0); - -unlock_out: - spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags); - return r; -} - -/** - * amdgpu_ih_clear_fault - Remove a page fault record - * - * @adev: amdgpu device pointer - * @key: 64-bit encoding of PASID and address - * - * This should be called when a page fault has been handled. Any - * future interrupt with this key will be processed as a new - * page fault. - */ -void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key) -{ - unsigned long flags; - int r; - - if (!adev->irq.ih.faults) - return; - - spin_lock_irqsave(&adev->irq.ih.faults->lock, flags); - - r = chash_table_remove(&adev->irq.ih.faults->hash, key, NULL); - if (!WARN_ON_ONCE(r < 0)) { - adev->irq.ih.faults->count--; - WARN_ON_ONCE(adev->irq.ih.faults->count < 0); - } - - spin_unlock_irqrestore(&adev->irq.ih.faults->lock, flags); -} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h index 0e01f115bbe5..9ce8c93ec19b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h @@ -24,20 +24,8 @@ #ifndef __AMDGPU_IH_H__ #define __AMDGPU_IH_H__ -#include <linux/chash.h> -#include "soc15_ih_clientid.h" - struct amdgpu_device; - -#define AMDGPU_IH_CLIENTID_LEGACY 0 -#define AMDGPU_IH_CLIENTID_MAX SOC15_IH_CLIENTID_MAX - -#define AMDGPU_PAGEFAULT_HASH_BITS 8 -struct amdgpu_retryfault_hashtable { - DECLARE_CHASH_TABLE(hash, AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); - spinlock_t lock; - int count; -}; +struct amdgpu_iv_entry; /* * R6xx+ IH ring @@ -57,30 +45,28 @@ struct amdgpu_ih_ring { bool use_doorbell; bool use_bus_addr; dma_addr_t rb_dma_addr; /* only used when use_bus_addr = true */ - struct amdgpu_retryfault_hashtable *faults; }; -#define AMDGPU_IH_SRC_DATA_MAX_SIZE_DW 4 - -struct amdgpu_iv_entry { - unsigned client_id; - unsigned src_id; - unsigned ring_id; - unsigned vmid; - unsigned vmid_src; - uint64_t timestamp; - unsigned timestamp_src; - unsigned pasid; - unsigned pasid_src; - unsigned src_data[AMDGPU_IH_SRC_DATA_MAX_SIZE_DW]; - const uint32_t *iv_entry; +/* provided by the ih block */ +struct amdgpu_ih_funcs { + /* ring read/write ptr handling, called from interrupt context */ + u32 (*get_wptr)(struct amdgpu_device *adev); + bool (*prescreen_iv)(struct amdgpu_device *adev); + void (*decode_iv)(struct amdgpu_device *adev, + struct amdgpu_iv_entry *entry); + void (*set_rptr)(struct amdgpu_device *adev); }; -int amdgpu_ih_ring_init(struct amdgpu_device *adev, unsigned ring_size, - bool use_bus_addr); -void amdgpu_ih_ring_fini(struct amdgpu_device *adev); -int amdgpu_ih_process(struct amdgpu_device *adev); -int amdgpu_ih_add_fault(struct amdgpu_device *adev, u64 key); -void amdgpu_ih_clear_fault(struct amdgpu_device *adev, u64 key); +#define amdgpu_ih_get_wptr(adev) (adev)->irq.ih_funcs->get_wptr((adev)) +#define amdgpu_ih_prescreen_iv(adev) (adev)->irq.ih_funcs->prescreen_iv((adev)) +#define amdgpu_ih_decode_iv(adev, iv) (adev)->irq.ih_funcs->decode_iv((adev), (iv)) +#define amdgpu_ih_set_rptr(adev) (adev)->irq.ih_funcs->set_rptr((adev)) + +int amdgpu_ih_ring_init(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, + unsigned ring_size, bool use_bus_addr); +void amdgpu_ih_ring_fini(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih); +int amdgpu_ih_process(struct amdgpu_device *adev, struct amdgpu_ih_ring *ih, + void (*callback)(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih)); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c index 3a5ca462abf0..52c17f6219a7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c @@ -25,6 +25,23 @@ * Alex Deucher * Jerome Glisse */ + +/** + * DOC: Interrupt Handling + * + * Interrupts generated within GPU hardware raise interrupt requests that are + * passed to amdgpu IRQ handler which is responsible for detecting source and + * type of the interrupt and dispatching matching handlers. If handling an + * interrupt requires calling kernel functions that may sleep processing is + * dispatched to work handlers. + * + * If MSI functionality is not disabled by module parameter then MSI + * support will be enabled. + * + * For GPU interrupt sources that may be driven by another driver, IRQ domain + * support is used (with mapping between virtual and hardware IRQs). + */ + #include <linux/irq.h> #include <drm/drmP.h> #include <drm/drm_crtc_helper.h> @@ -34,6 +51,7 @@ #include "atom.h" #include "amdgpu_connectors.h" #include "amdgpu_trace.h" +#include "amdgpu_amdkfd.h" #include <linux/pm_runtime.h> @@ -43,19 +61,21 @@ #define AMDGPU_WAIT_IDLE_TIMEOUT 200 -/* - * Handle hotplug events outside the interrupt handler proper. - */ /** - * amdgpu_hotplug_work_func - display hotplug work handler + * amdgpu_hotplug_work_func - work handler for display hotplug event * - * @work: work struct + * @work: work struct pointer * - * This is the hot plug event work handler (all asics). - * The work gets scheduled from the irq handler if there - * was a hot plug interrupt. It walks the connector table - * and calls the hotplug handler for each one, then sends - * a drm hotplug event to alert userspace. + * This is the hotplug event work handler (all ASICs). + * The work gets scheduled from the IRQ handler if there + * was a hotplug interrupt. It walks through the connector table + * and calls hotplug handler for each connector. After this, it sends + * a DRM hotplug event to alert userspace. + * + * This design approach is required in order to defer hotplug event handling + * from the IRQ handler to a work handler because hotplug handler has to use + * mutexes which cannot be locked in an IRQ handler (since &mutex_lock may + * sleep). */ static void amdgpu_hotplug_work_func(struct work_struct *work) { @@ -74,24 +94,29 @@ static void amdgpu_hotplug_work_func(struct work_struct *work) } /** - * amdgpu_irq_reset_work_func - execute gpu reset + * amdgpu_irq_reset_work_func - execute GPU reset * - * @work: work struct + * @work: work struct pointer * - * Execute scheduled gpu reset (cayman+). - * This function is called when the irq handler - * thinks we need a gpu reset. + * Execute scheduled GPU reset (Cayman+). + * This function is called when the IRQ handler thinks we need a GPU reset. */ static void amdgpu_irq_reset_work_func(struct work_struct *work) { struct amdgpu_device *adev = container_of(work, struct amdgpu_device, reset_work); - if (!amdgpu_sriov_vf(adev)) - amdgpu_device_gpu_recover(adev, NULL, false); + if (!amdgpu_sriov_vf(adev) && amdgpu_device_should_recover_gpu(adev)) + amdgpu_device_gpu_recover(adev, NULL); } -/* Disable *all* interrupts */ +/** + * amdgpu_irq_disable_all - disable *all* interrupts + * + * @adev: amdgpu device pointer + * + * Disable all types of interrupts from all sources. + */ void amdgpu_irq_disable_all(struct amdgpu_device *adev) { unsigned long irqflags; @@ -99,7 +124,7 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev) int r; spin_lock_irqsave(&adev->irq.lock, irqflags); - for (i = 0; i < AMDGPU_IH_CLIENTID_MAX; ++i) { + for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) { if (!adev->irq.client[i].sources) continue; @@ -123,11 +148,43 @@ void amdgpu_irq_disable_all(struct amdgpu_device *adev) } /** - * amdgpu_irq_handler - irq handler + * amdgpu_irq_callback - callback from the IH ring + * + * @adev: amdgpu device pointer + * @ih: amdgpu ih ring + * + * Callback from IH ring processing to handle the entry at the current position + * and advance the read pointer. + */ +static void amdgpu_irq_callback(struct amdgpu_device *adev, + struct amdgpu_ih_ring *ih) +{ + u32 ring_index = ih->rptr >> 2; + struct amdgpu_iv_entry entry; + + /* Prescreening of high-frequency interrupts */ + if (!amdgpu_ih_prescreen_iv(adev)) + return; + + /* Before dispatching irq to IP blocks, send it to amdkfd */ + amdgpu_amdkfd_interrupt(adev, (const void *) &ih->ring[ring_index]); + + entry.iv_entry = (const uint32_t *)&ih->ring[ring_index]; + amdgpu_ih_decode_iv(adev, &entry); + + amdgpu_irq_dispatch(adev, &entry); +} + +/** + * amdgpu_irq_handler - IRQ handler + * + * @irq: IRQ number (unused) + * @arg: pointer to DRM device * - * @int irq, void *arg: args + * IRQ handler for amdgpu driver (all ASICs). * - * This is the irq handler for the amdgpu driver (all asics). + * Returns: + * result of handling the IRQ, as defined by &irqreturn_t */ irqreturn_t amdgpu_irq_handler(int irq, void *arg) { @@ -135,25 +192,25 @@ irqreturn_t amdgpu_irq_handler(int irq, void *arg) struct amdgpu_device *adev = dev->dev_private; irqreturn_t ret; - ret = amdgpu_ih_process(adev); + ret = amdgpu_ih_process(adev, &adev->irq.ih, amdgpu_irq_callback); if (ret == IRQ_HANDLED) pm_runtime_mark_last_busy(dev->dev); return ret; } /** - * amdgpu_msi_ok - asic specific msi checks + * amdgpu_msi_ok - check whether MSI functionality is enabled * - * @adev: amdgpu device pointer + * @adev: amdgpu device pointer (unused) * - * Handles asic specific MSI checks to determine if - * MSIs should be enabled on a particular chip (all asics). - * Returns true if MSIs should be enabled, false if MSIs - * should not be enabled. + * Checks whether MSI functionality has been disabled via module parameter + * (all ASICs). + * + * Returns: + * *true* if MSIs are allowed to be enabled or *false* otherwise */ static bool amdgpu_msi_ok(struct amdgpu_device *adev) { - /* force MSI on */ if (amdgpu_msi == 1) return true; else if (amdgpu_msi == 0) @@ -163,12 +220,15 @@ static bool amdgpu_msi_ok(struct amdgpu_device *adev) } /** - * amdgpu_irq_init - init driver interrupt info + * amdgpu_irq_init - initialize interrupt handling * * @adev: amdgpu device pointer * - * Sets up the work irq handlers, vblank init, MSIs, etc. (all asics). - * Returns 0 for success, error for failure. + * Sets up work functions for hotplug and reset interrupts, enables MSI + * functionality, initializes vblank, hotplug and reset interrupt handling. + * + * Returns: + * 0 on success or error code on failure */ int amdgpu_irq_init(struct amdgpu_device *adev) { @@ -176,7 +236,7 @@ int amdgpu_irq_init(struct amdgpu_device *adev) spin_lock_init(&adev->irq.lock); - /* enable msi */ + /* Enable MSI if not disabled by module parameter */ adev->irq.msi_enabled = false; if (amdgpu_msi_ok(adev)) { @@ -189,7 +249,7 @@ int amdgpu_irq_init(struct amdgpu_device *adev) if (!amdgpu_device_has_dc_support(adev)) { if (!adev->enable_virtual_display) - /* Disable vblank irqs aggressively for power-saving */ + /* Disable vblank IRQs aggressively for power-saving */ /* XXX: can this be enabled for DC? */ adev->ddev->vblank_disable_immediate = true; @@ -197,7 +257,7 @@ int amdgpu_irq_init(struct amdgpu_device *adev) if (r) return r; - /* pre DCE11 */ + /* Pre-DCE11 */ INIT_WORK(&adev->hotplug_work, amdgpu_hotplug_work_func); } @@ -220,11 +280,13 @@ int amdgpu_irq_init(struct amdgpu_device *adev) } /** - * amdgpu_irq_fini - tear down driver interrupt info + * amdgpu_irq_fini - shut down interrupt handling * * @adev: amdgpu device pointer * - * Tears down the work irq handlers, vblank handlers, MSIs, etc. (all asics). + * Tears down work functions for hotplug and reset interrupts, disables MSI + * functionality, shuts down vblank, hotplug and reset interrupt handling, + * turns off interrupts from all sources (all ASICs). */ void amdgpu_irq_fini(struct amdgpu_device *adev) { @@ -240,7 +302,7 @@ void amdgpu_irq_fini(struct amdgpu_device *adev) cancel_work_sync(&adev->reset_work); } - for (i = 0; i < AMDGPU_IH_CLIENTID_MAX; ++i) { + for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) { if (!adev->irq.client[i].sources) continue; @@ -264,18 +326,23 @@ void amdgpu_irq_fini(struct amdgpu_device *adev) } /** - * amdgpu_irq_add_id - register irq source + * amdgpu_irq_add_id - register IRQ source * * @adev: amdgpu device pointer - * @src_id: source id for this source - * @source: irq source + * @client_id: client id + * @src_id: source id + * @source: IRQ source pointer + * + * Registers IRQ source on a client. * + * Returns: + * 0 on success or error code otherwise */ int amdgpu_irq_add_id(struct amdgpu_device *adev, unsigned client_id, unsigned src_id, struct amdgpu_irq_src *source) { - if (client_id >= AMDGPU_IH_CLIENTID_MAX) + if (client_id >= AMDGPU_IRQ_CLIENTID_MAX) return -EINVAL; if (src_id >= AMDGPU_MAX_IRQ_SRC_ID) @@ -312,12 +379,12 @@ int amdgpu_irq_add_id(struct amdgpu_device *adev, } /** - * amdgpu_irq_dispatch - dispatch irq to IP blocks + * amdgpu_irq_dispatch - dispatch IRQ to IP blocks * * @adev: amdgpu device pointer - * @entry: interrupt vector + * @entry: interrupt vector pointer * - * Dispatches the irq to the different IP blocks + * Dispatches IRQ to IP blocks. */ void amdgpu_irq_dispatch(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) @@ -329,7 +396,7 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev, trace_amdgpu_iv(entry); - if (client_id >= AMDGPU_IH_CLIENTID_MAX) { + if (client_id >= AMDGPU_IRQ_CLIENTID_MAX) { DRM_DEBUG("Invalid client_id in IV: %d\n", client_id); return; } @@ -361,13 +428,13 @@ void amdgpu_irq_dispatch(struct amdgpu_device *adev, } /** - * amdgpu_irq_update - update hw interrupt state + * amdgpu_irq_update - update hardware interrupt state * * @adev: amdgpu device pointer - * @src: interrupt src you want to enable - * @type: type of interrupt you want to update + * @src: interrupt source pointer + * @type: type of interrupt * - * Updates the interrupt state for a specific src (all asics). + * Updates interrupt state for the specific source (all ASICs). */ int amdgpu_irq_update(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type) @@ -378,7 +445,7 @@ int amdgpu_irq_update(struct amdgpu_device *adev, spin_lock_irqsave(&adev->irq.lock, irqflags); - /* we need to determine after taking the lock, otherwise + /* We need to determine after taking the lock, otherwise we might disable just enabled interrupts again */ if (amdgpu_irq_enabled(adev, src, type)) state = AMDGPU_IRQ_STATE_ENABLE; @@ -390,11 +457,19 @@ int amdgpu_irq_update(struct amdgpu_device *adev, return r; } +/** + * amdgpu_irq_gpu_reset_resume_helper - update interrupt states on all sources + * + * @adev: amdgpu device pointer + * + * Updates state of all types of interrupts on all sources on resume after + * reset. + */ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev) { int i, j, k; - for (i = 0; i < AMDGPU_IH_CLIENTID_MAX; ++i) { + for (i = 0; i < AMDGPU_IRQ_CLIENTID_MAX; ++i) { if (!adev->irq.client[i].sources) continue; @@ -413,10 +488,13 @@ void amdgpu_irq_gpu_reset_resume_helper(struct amdgpu_device *adev) * amdgpu_irq_get - enable interrupt * * @adev: amdgpu device pointer - * @src: interrupt src you want to enable - * @type: type of interrupt you want to enable + * @src: interrupt source pointer + * @type: type of interrupt + * + * Enables specified type of interrupt on the specified source (all ASICs). * - * Enables the interrupt type for a specific src (all asics). + * Returns: + * 0 on success or error code otherwise */ int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type) @@ -440,10 +518,13 @@ int amdgpu_irq_get(struct amdgpu_device *adev, struct amdgpu_irq_src *src, * amdgpu_irq_put - disable interrupt * * @adev: amdgpu device pointer - * @src: interrupt src you want to disable - * @type: type of interrupt you want to disable + * @src: interrupt source pointer + * @type: type of interrupt * - * Disables the interrupt type for a specific src (all asics). + * Enables specified type of interrupt on the specified source (all ASICs). + * + * Returns: + * 0 on success or error code otherwise */ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type) @@ -464,12 +545,17 @@ int amdgpu_irq_put(struct amdgpu_device *adev, struct amdgpu_irq_src *src, } /** - * amdgpu_irq_enabled - test if irq is enabled or not + * amdgpu_irq_enabled - check whether interrupt is enabled or not * * @adev: amdgpu device pointer - * @idx: interrupt src you want to test + * @src: interrupt source pointer + * @type: type of interrupt + * + * Checks whether the given type of interrupt is enabled on the given source. * - * Tests if the given interrupt source is enabled or not + * Returns: + * *true* if interrupt is enabled, *false* if interrupt is disabled or on + * invalid parameters */ bool amdgpu_irq_enabled(struct amdgpu_device *adev, struct amdgpu_irq_src *src, unsigned type) @@ -486,7 +572,7 @@ bool amdgpu_irq_enabled(struct amdgpu_device *adev, struct amdgpu_irq_src *src, return !!atomic_read(&src->enabled_types[type]); } -/* gen irq */ +/* XXX: Generic IRQ handling */ static void amdgpu_irq_mask(struct irq_data *irqd) { /* XXX */ @@ -497,12 +583,26 @@ static void amdgpu_irq_unmask(struct irq_data *irqd) /* XXX */ } +/* amdgpu hardware interrupt chip descriptor */ static struct irq_chip amdgpu_irq_chip = { .name = "amdgpu-ih", .irq_mask = amdgpu_irq_mask, .irq_unmask = amdgpu_irq_unmask, }; +/** + * amdgpu_irqdomain_map - create mapping between virtual and hardware IRQ numbers + * + * @d: amdgpu IRQ domain pointer (unused) + * @irq: virtual IRQ number + * @hwirq: hardware irq number + * + * Current implementation assigns simple interrupt handler to the given virtual + * IRQ. + * + * Returns: + * 0 on success or error code otherwise + */ static int amdgpu_irqdomain_map(struct irq_domain *d, unsigned int irq, irq_hw_number_t hwirq) { @@ -514,17 +614,21 @@ static int amdgpu_irqdomain_map(struct irq_domain *d, return 0; } +/* Implementation of methods for amdgpu IRQ domain */ static const struct irq_domain_ops amdgpu_hw_irqdomain_ops = { .map = amdgpu_irqdomain_map, }; /** - * amdgpu_irq_add_domain - create a linear irq domain + * amdgpu_irq_add_domain - create a linear IRQ domain * * @adev: amdgpu device pointer * - * Create an irq domain for GPU interrupt sources + * Creates an IRQ domain for GPU interrupt sources * that may be driven by another driver (e.g., ACP). + * + * Returns: + * 0 on success or error code otherwise */ int amdgpu_irq_add_domain(struct amdgpu_device *adev) { @@ -539,11 +643,11 @@ int amdgpu_irq_add_domain(struct amdgpu_device *adev) } /** - * amdgpu_irq_remove_domain - remove the irq domain + * amdgpu_irq_remove_domain - remove the IRQ domain * * @adev: amdgpu device pointer * - * Remove the irq domain for GPU interrupt sources + * Removes the IRQ domain for GPU interrupt sources * that may be driven by another driver (e.g., ACP). */ void amdgpu_irq_remove_domain(struct amdgpu_device *adev) @@ -555,16 +659,17 @@ void amdgpu_irq_remove_domain(struct amdgpu_device *adev) } /** - * amdgpu_irq_create_mapping - create a mapping between a domain irq and a - * Linux irq + * amdgpu_irq_create_mapping - create mapping between domain Linux IRQs * * @adev: amdgpu device pointer * @src_id: IH source id * - * Create a mapping between a domain irq (GPU IH src id) and a Linux irq + * Creates mapping between a domain IRQ (GPU IH src id) and a Linux IRQ * Use this for components that generate a GPU interrupt, but are driven * by a different driver (e.g., ACP). - * Returns the Linux irq. + * + * Returns: + * Linux IRQ */ unsigned amdgpu_irq_create_mapping(struct amdgpu_device *adev, unsigned src_id) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h index 3375ad778edc..f6ce171cb8aa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.h @@ -25,19 +25,38 @@ #define __AMDGPU_IRQ_H__ #include <linux/irqdomain.h> +#include "soc15_ih_clientid.h" #include "amdgpu_ih.h" -#define AMDGPU_MAX_IRQ_SRC_ID 0x100 +#define AMDGPU_MAX_IRQ_SRC_ID 0x100 #define AMDGPU_MAX_IRQ_CLIENT_ID 0x100 +#define AMDGPU_IRQ_CLIENTID_LEGACY 0 +#define AMDGPU_IRQ_CLIENTID_MAX SOC15_IH_CLIENTID_MAX + +#define AMDGPU_IRQ_SRC_DATA_MAX_SIZE_DW 4 + struct amdgpu_device; -struct amdgpu_iv_entry; enum amdgpu_interrupt_state { AMDGPU_IRQ_STATE_DISABLE, AMDGPU_IRQ_STATE_ENABLE, }; +struct amdgpu_iv_entry { + unsigned client_id; + unsigned src_id; + unsigned ring_id; + unsigned vmid; + unsigned vmid_src; + uint64_t timestamp; + unsigned timestamp_src; + unsigned pasid; + unsigned pasid_src; + unsigned src_data[AMDGPU_IRQ_SRC_DATA_MAX_SIZE_DW]; + const uint32_t *iv_entry; +}; + struct amdgpu_irq_src { unsigned num_types; atomic_t *enabled_types; @@ -63,7 +82,7 @@ struct amdgpu_irq { bool installed; spinlock_t lock; /* interrupt sources */ - struct amdgpu_irq_client client[AMDGPU_IH_CLIENTID_MAX]; + struct amdgpu_irq_client client[AMDGPU_IRQ_CLIENTID_MAX]; /* status, etc. */ bool msi_enabled; /* msi enabled */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index 2bd56760c744..755f733bf0d9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -30,14 +30,21 @@ static void amdgpu_job_timedout(struct drm_sched_job *s_job) { - struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base); + struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched); + struct amdgpu_job *job = to_amdgpu_job(s_job); - DRM_ERROR("ring %s timeout, last signaled seq=%u, last emitted seq=%u\n", - job->base.sched->name, - atomic_read(&job->ring->fence_drv.last_seq), - job->ring->fence_drv.sync_seq); + if (amdgpu_ring_soft_recovery(ring, job->vmid, s_job->s_fence->parent)) { + DRM_ERROR("ring %s timeout, but soft recovered\n", + s_job->sched->name); + return; + } + + DRM_ERROR("ring %s timeout, signaled seq=%u, emitted seq=%u\n", + job->base.sched->name, atomic_read(&ring->fence_drv.last_seq), + ring->fence_drv.sync_seq); - amdgpu_device_gpu_recover(job->adev, job, false); + if (amdgpu_device_should_recover_gpu(ring->adev)) + amdgpu_device_gpu_recover(ring->adev, job); } int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, @@ -54,7 +61,11 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, if (!*job) return -ENOMEM; - (*job)->adev = adev; + /* + * Initialize the scheduler to at least some ring so that we always + * have a pointer to adev. + */ + (*job)->base.sched = &adev->rings[0]->sched; (*job)->vm = vm; (*job)->ibs = (void *)&(*job)[1]; (*job)->num_ibs = num_ibs; @@ -62,6 +73,7 @@ int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, amdgpu_sync_create(&(*job)->sync); amdgpu_sync_create(&(*job)->sched_sync); (*job)->vram_lost_counter = atomic_read(&adev->vram_lost_counter); + (*job)->vm_pd_addr = AMDGPU_BO_INVALID_OFFSET; return 0; } @@ -78,14 +90,13 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, r = amdgpu_ib_get(adev, NULL, size, &(*job)->ibs[0]); if (r) kfree(*job); - else - (*job)->vm_pd_addr = adev->gart.table_addr; return r; } void amdgpu_job_free_resources(struct amdgpu_job *job) { + struct amdgpu_ring *ring = to_amdgpu_ring(job->base.sched); struct dma_fence *f; unsigned i; @@ -93,14 +104,15 @@ void amdgpu_job_free_resources(struct amdgpu_job *job) f = job->base.s_fence ? &job->base.s_fence->finished : job->fence; for (i = 0; i < job->num_ibs; ++i) - amdgpu_ib_free(job->adev, &job->ibs[i], f); + amdgpu_ib_free(ring->adev, &job->ibs[i], f); } static void amdgpu_job_free_cb(struct drm_sched_job *s_job) { - struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base); + struct amdgpu_ring *ring = to_amdgpu_ring(s_job->sched); + struct amdgpu_job *job = to_amdgpu_job(s_job); - amdgpu_ring_priority_put(job->ring, s_job->s_priority); + amdgpu_ring_priority_put(ring, s_job->s_priority); dma_fence_put(job->fence); amdgpu_sync_free(&job->sync); amdgpu_sync_free(&job->sched_sync); @@ -117,50 +129,68 @@ void amdgpu_job_free(struct amdgpu_job *job) kfree(job); } -int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring, - struct drm_sched_entity *entity, void *owner, - struct dma_fence **f) +int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity, + void *owner, struct dma_fence **f) { + enum drm_sched_priority priority; + struct amdgpu_ring *ring; int r; - job->ring = ring; if (!f) return -EINVAL; - r = drm_sched_job_init(&job->base, &ring->sched, entity, owner); + r = drm_sched_job_init(&job->base, entity, owner); if (r) return r; job->owner = owner; - job->fence_ctx = entity->fence_context; *f = dma_fence_get(&job->base.s_fence->finished); amdgpu_job_free_resources(job); - amdgpu_ring_priority_get(job->ring, job->base.s_priority); + priority = job->base.s_priority; drm_sched_entity_push_job(&job->base, entity); + ring = to_amdgpu_ring(entity->rq->sched); + amdgpu_ring_priority_get(ring, priority); + + return 0; +} + +int amdgpu_job_submit_direct(struct amdgpu_job *job, struct amdgpu_ring *ring, + struct dma_fence **fence) +{ + int r; + + job->base.sched = &ring->sched; + r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, NULL, fence); + job->fence = dma_fence_get(*fence); + if (r) + return r; + + amdgpu_job_free(job); return 0; } static struct dma_fence *amdgpu_job_dependency(struct drm_sched_job *sched_job, struct drm_sched_entity *s_entity) { + struct amdgpu_ring *ring = to_amdgpu_ring(s_entity->rq->sched); struct amdgpu_job *job = to_amdgpu_job(sched_job); struct amdgpu_vm *vm = job->vm; + struct dma_fence *fence; bool explicit = false; int r; - struct dma_fence *fence = amdgpu_sync_get_fence(&job->sync, &explicit); + fence = amdgpu_sync_get_fence(&job->sync, &explicit); if (fence && explicit) { if (drm_sched_dependency_optimized(fence, s_entity)) { - r = amdgpu_sync_fence(job->adev, &job->sched_sync, fence, false); + r = amdgpu_sync_fence(ring->adev, &job->sched_sync, + fence, false); if (r) - DRM_ERROR("Error adding fence to sync (%d)\n", r); + DRM_ERROR("Error adding fence (%d)\n", r); } } while (fence == NULL && vm && !job->vmid) { - struct amdgpu_ring *ring = job->ring; - r = amdgpu_vmid_grab(vm, ring, &job->sync, &job->base.s_fence->finished, job); @@ -175,30 +205,25 @@ static struct dma_fence *amdgpu_job_dependency(struct drm_sched_job *sched_job, static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job) { + struct amdgpu_ring *ring = to_amdgpu_ring(sched_job->sched); struct dma_fence *fence = NULL, *finished; - struct amdgpu_device *adev; struct amdgpu_job *job; int r; - if (!sched_job) { - DRM_ERROR("job is null\n"); - return NULL; - } job = to_amdgpu_job(sched_job); finished = &job->base.s_fence->finished; - adev = job->adev; BUG_ON(amdgpu_sync_peek_fence(&job->sync, NULL)); trace_amdgpu_sched_run_job(job); - if (job->vram_lost_counter != atomic_read(&adev->vram_lost_counter)) + if (job->vram_lost_counter != atomic_read(&ring->adev->vram_lost_counter)) dma_fence_set_error(finished, -ECANCELED);/* skip IB as well if VRAM lost */ if (finished->error < 0) { DRM_INFO("Skip scheduling IBs!\n"); } else { - r = amdgpu_ib_schedule(job->ring, job->num_ibs, job->ibs, job, + r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, job, &fence); if (r) DRM_ERROR("Error scheduling IBs (%d)\n", r); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h new file mode 100644 index 000000000000..57cfe78a262b --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.h @@ -0,0 +1,74 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __AMDGPU_JOB_H__ +#define __AMDGPU_JOB_H__ + +/* bit set means command submit involves a preamble IB */ +#define AMDGPU_PREAMBLE_IB_PRESENT (1 << 0) +/* bit set means preamble IB is first presented in belonging context */ +#define AMDGPU_PREAMBLE_IB_PRESENT_FIRST (1 << 1) +/* bit set means context switch occured */ +#define AMDGPU_HAVE_CTX_SWITCH (1 << 2) + +#define to_amdgpu_job(sched_job) \ + container_of((sched_job), struct amdgpu_job, base) + +struct amdgpu_fence; + +struct amdgpu_job { + struct drm_sched_job base; + struct amdgpu_vm *vm; + struct amdgpu_sync sync; + struct amdgpu_sync sched_sync; + struct amdgpu_ib *ibs; + struct dma_fence *fence; /* the hw fence */ + uint32_t preamble_status; + uint32_t num_ibs; + void *owner; + bool vm_needs_flush; + uint64_t vm_pd_addr; + unsigned vmid; + unsigned pasid; + uint32_t gds_base, gds_size; + uint32_t gws_base, gws_size; + uint32_t oa_base, oa_size; + uint32_t vram_lost_counter; + + /* user fence handling */ + uint64_t uf_addr; + uint64_t uf_sequence; + +}; + +int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs, + struct amdgpu_job **job, struct amdgpu_vm *vm); +int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size, + struct amdgpu_job **job); + +void amdgpu_job_free_resources(struct amdgpu_job *job); +void amdgpu_job_free(struct amdgpu_job *job); +int amdgpu_job_submit(struct amdgpu_job *job, struct drm_sched_entity *entity, + void *owner, struct dma_fence **f); +int amdgpu_job_submit_direct(struct amdgpu_job *job, struct amdgpu_ring *ring, + struct dma_fence **fence); +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 91517b166a3b..81732a84c2ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -37,6 +37,32 @@ #include <linux/slab.h> #include <linux/pm_runtime.h> #include "amdgpu_amdkfd.h" +#include "amdgpu_gem.h" +#include "amdgpu_display.h" + +static void amdgpu_unregister_gpu_instance(struct amdgpu_device *adev) +{ + struct amdgpu_gpu_instance *gpu_instance; + int i; + + mutex_lock(&mgpu_info.mutex); + + for (i = 0; i < mgpu_info.num_gpu; i++) { + gpu_instance = &(mgpu_info.gpu_ins[i]); + if (gpu_instance->adev == adev) { + mgpu_info.gpu_ins[i] = + mgpu_info.gpu_ins[mgpu_info.num_gpu - 1]; + mgpu_info.num_gpu--; + if (adev->flags & AMD_IS_APU) + mgpu_info.num_apu--; + else + mgpu_info.num_dgpu--; + break; + } + } + + mutex_unlock(&mgpu_info.mutex); +} /** * amdgpu_driver_unload_kms - Main unload function for KMS. @@ -53,6 +79,8 @@ void amdgpu_driver_unload_kms(struct drm_device *dev) if (adev == NULL) return; + amdgpu_unregister_gpu_instance(adev); + if (adev->rmmio == NULL) goto done_free; @@ -73,6 +101,31 @@ done_free: dev->dev_private = NULL; } +static void amdgpu_register_gpu_instance(struct amdgpu_device *adev) +{ + struct amdgpu_gpu_instance *gpu_instance; + + mutex_lock(&mgpu_info.mutex); + + if (mgpu_info.num_gpu >= MAX_GPU_INSTANCE) { + DRM_ERROR("Cannot register more gpu instance\n"); + mutex_unlock(&mgpu_info.mutex); + return; + } + + gpu_instance = &(mgpu_info.gpu_ins[mgpu_info.num_gpu]); + gpu_instance->adev = adev; + gpu_instance->mgpu_fan_enabled = 0; + + mgpu_info.num_gpu++; + if (adev->flags & AMD_IS_APU) + mgpu_info.num_apu++; + else + mgpu_info.num_dgpu++; + + mutex_unlock(&mgpu_info.mutex); +} + /** * amdgpu_driver_load_kms - Main load function for KMS. * @@ -167,6 +220,7 @@ int amdgpu_driver_load_kms(struct drm_device *dev, unsigned long flags) pm_runtime_put_autosuspend(dev->dev); } + amdgpu_register_gpu_instance(adev); out: if (r) { /* balance pm_runtime_get_sync in amdgpu_driver_unload_kms */ @@ -255,12 +309,133 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info, fw_info->ver = adev->psp.asd_fw_version; fw_info->feature = adev->psp.asd_feature_version; break; + case AMDGPU_INFO_FW_DMCU: + fw_info->ver = adev->dm.dmcu_fw_version; + fw_info->feature = 0; + break; default: return -EINVAL; } return 0; } +static int amdgpu_hw_ip_info(struct amdgpu_device *adev, + struct drm_amdgpu_info *info, + struct drm_amdgpu_info_hw_ip *result) +{ + uint32_t ib_start_alignment = 0; + uint32_t ib_size_alignment = 0; + enum amd_ip_block_type type; + unsigned int num_rings = 0; + unsigned int i, j; + + if (info->query_hw_ip.ip_instance >= AMDGPU_HW_IP_INSTANCE_MAX_COUNT) + return -EINVAL; + + switch (info->query_hw_ip.type) { + case AMDGPU_HW_IP_GFX: + type = AMD_IP_BLOCK_TYPE_GFX; + for (i = 0; i < adev->gfx.num_gfx_rings; i++) + if (adev->gfx.gfx_ring[i].ready) + ++num_rings; + ib_start_alignment = 32; + ib_size_alignment = 32; + break; + case AMDGPU_HW_IP_COMPUTE: + type = AMD_IP_BLOCK_TYPE_GFX; + for (i = 0; i < adev->gfx.num_compute_rings; i++) + if (adev->gfx.compute_ring[i].ready) + ++num_rings; + ib_start_alignment = 32; + ib_size_alignment = 32; + break; + case AMDGPU_HW_IP_DMA: + type = AMD_IP_BLOCK_TYPE_SDMA; + for (i = 0; i < adev->sdma.num_instances; i++) + if (adev->sdma.instance[i].ring.ready) + ++num_rings; + ib_start_alignment = 256; + ib_size_alignment = 4; + break; + case AMDGPU_HW_IP_UVD: + type = AMD_IP_BLOCK_TYPE_UVD; + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; + + if (adev->uvd.inst[i].ring.ready) + ++num_rings; + } + ib_start_alignment = 64; + ib_size_alignment = 64; + break; + case AMDGPU_HW_IP_VCE: + type = AMD_IP_BLOCK_TYPE_VCE; + for (i = 0; i < adev->vce.num_rings; i++) + if (adev->vce.ring[i].ready) + ++num_rings; + ib_start_alignment = 4; + ib_size_alignment = 1; + break; + case AMDGPU_HW_IP_UVD_ENC: + type = AMD_IP_BLOCK_TYPE_UVD; + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; + + for (j = 0; j < adev->uvd.num_enc_rings; j++) + if (adev->uvd.inst[i].ring_enc[j].ready) + ++num_rings; + } + ib_start_alignment = 64; + ib_size_alignment = 64; + break; + case AMDGPU_HW_IP_VCN_DEC: + type = AMD_IP_BLOCK_TYPE_VCN; + if (adev->vcn.ring_dec.ready) + ++num_rings; + ib_start_alignment = 16; + ib_size_alignment = 16; + break; + case AMDGPU_HW_IP_VCN_ENC: + type = AMD_IP_BLOCK_TYPE_VCN; + for (i = 0; i < adev->vcn.num_enc_rings; i++) + if (adev->vcn.ring_enc[i].ready) + ++num_rings; + ib_start_alignment = 64; + ib_size_alignment = 1; + break; + case AMDGPU_HW_IP_VCN_JPEG: + type = AMD_IP_BLOCK_TYPE_VCN; + if (adev->vcn.ring_jpeg.ready) + ++num_rings; + ib_start_alignment = 16; + ib_size_alignment = 16; + break; + default: + return -EINVAL; + } + + for (i = 0; i < adev->num_ip_blocks; i++) + if (adev->ip_blocks[i].version->type == type && + adev->ip_blocks[i].status.valid) + break; + + if (i == adev->num_ip_blocks) + return 0; + + num_rings = min(amdgpu_ctx_num_entities[info->query_hw_ip.type], + num_rings); + + result->hw_ip_version_major = adev->ip_blocks[i].version->major; + result->hw_ip_version_minor = adev->ip_blocks[i].version->minor; + result->capabilities_flags = 0; + result->available_rings = (1 << num_rings) - 1; + result->ib_start_alignment = ib_start_alignment; + result->ib_size_alignment = ib_size_alignment; + return 0; +} + /* * Userspace get information ioctl */ @@ -286,7 +461,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file struct drm_crtc *crtc; uint32_t ui32 = 0; uint64_t ui64 = 0; - int i, j, found; + int i, found; int ui32_size = sizeof(ui32); if (!info->return_size || !info->return_pointer) @@ -316,91 +491,14 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file return copy_to_user(out, &ui32, min(size, 4u)) ? -EFAULT : 0; case AMDGPU_INFO_HW_IP_INFO: { struct drm_amdgpu_info_hw_ip ip = {}; - enum amd_ip_block_type type; - uint32_t ring_mask = 0; - uint32_t ib_start_alignment = 0; - uint32_t ib_size_alignment = 0; - - if (info->query_hw_ip.ip_instance >= AMDGPU_HW_IP_INSTANCE_MAX_COUNT) - return -EINVAL; + int ret; - switch (info->query_hw_ip.type) { - case AMDGPU_HW_IP_GFX: - type = AMD_IP_BLOCK_TYPE_GFX; - for (i = 0; i < adev->gfx.num_gfx_rings; i++) - ring_mask |= ((adev->gfx.gfx_ring[i].ready ? 1 : 0) << i); - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 8; - break; - case AMDGPU_HW_IP_COMPUTE: - type = AMD_IP_BLOCK_TYPE_GFX; - for (i = 0; i < adev->gfx.num_compute_rings; i++) - ring_mask |= ((adev->gfx.compute_ring[i].ready ? 1 : 0) << i); - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 8; - break; - case AMDGPU_HW_IP_DMA: - type = AMD_IP_BLOCK_TYPE_SDMA; - for (i = 0; i < adev->sdma.num_instances; i++) - ring_mask |= ((adev->sdma.instance[i].ring.ready ? 1 : 0) << i); - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 1; - break; - case AMDGPU_HW_IP_UVD: - type = AMD_IP_BLOCK_TYPE_UVD; - for (i = 0; i < adev->uvd.num_uvd_inst; i++) - ring_mask |= ((adev->uvd.inst[i].ring.ready ? 1 : 0) << i); - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 16; - break; - case AMDGPU_HW_IP_VCE: - type = AMD_IP_BLOCK_TYPE_VCE; - for (i = 0; i < adev->vce.num_rings; i++) - ring_mask |= ((adev->vce.ring[i].ready ? 1 : 0) << i); - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 1; - break; - case AMDGPU_HW_IP_UVD_ENC: - type = AMD_IP_BLOCK_TYPE_UVD; - for (i = 0; i < adev->uvd.num_uvd_inst; i++) - for (j = 0; j < adev->uvd.num_enc_rings; j++) - ring_mask |= - ((adev->uvd.inst[i].ring_enc[j].ready ? 1 : 0) << - (j + i * adev->uvd.num_enc_rings)); - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 1; - break; - case AMDGPU_HW_IP_VCN_DEC: - type = AMD_IP_BLOCK_TYPE_VCN; - ring_mask = adev->vcn.ring_dec.ready ? 1 : 0; - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 16; - break; - case AMDGPU_HW_IP_VCN_ENC: - type = AMD_IP_BLOCK_TYPE_VCN; - for (i = 0; i < adev->vcn.num_enc_rings; i++) - ring_mask |= ((adev->vcn.ring_enc[i].ready ? 1 : 0) << i); - ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; - ib_size_alignment = 1; - break; - default: - return -EINVAL; - } + ret = amdgpu_hw_ip_info(adev, info, &ip); + if (ret) + return ret; - for (i = 0; i < adev->num_ip_blocks; i++) { - if (adev->ip_blocks[i].version->type == type && - adev->ip_blocks[i].status.valid) { - ip.hw_ip_version_major = adev->ip_blocks[i].version->major; - ip.hw_ip_version_minor = adev->ip_blocks[i].version->minor; - ip.capabilities_flags = 0; - ip.available_rings = ring_mask; - ip.ib_start_alignment = ib_start_alignment; - ip.ib_size_alignment = ib_size_alignment; - break; - } - } - return copy_to_user(out, &ip, - min((size_t)size, sizeof(ip))) ? -EFAULT : 0; + ret = copy_to_user(out, &ip, min((size_t)size, sizeof(ip))); + return ret ? -EFAULT : 0; } case AMDGPU_INFO_HW_IP_COUNT: { enum amd_ip_block_type type; @@ -427,6 +525,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file break; case AMDGPU_HW_IP_VCN_DEC: case AMDGPU_HW_IP_VCN_ENC: + case AMDGPU_HW_IP_VCN_JPEG: type = AMD_IP_BLOCK_TYPE_VCN; break; default: @@ -481,26 +580,26 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file struct drm_amdgpu_info_gds gds_info; memset(&gds_info, 0, sizeof(gds_info)); - gds_info.gds_gfx_partition_size = adev->gds.mem.gfx_partition_size >> AMDGPU_GDS_SHIFT; - gds_info.compute_partition_size = adev->gds.mem.cs_partition_size >> AMDGPU_GDS_SHIFT; - gds_info.gds_total_size = adev->gds.mem.total_size >> AMDGPU_GDS_SHIFT; - gds_info.gws_per_gfx_partition = adev->gds.gws.gfx_partition_size >> AMDGPU_GWS_SHIFT; - gds_info.gws_per_compute_partition = adev->gds.gws.cs_partition_size >> AMDGPU_GWS_SHIFT; - gds_info.oa_per_gfx_partition = adev->gds.oa.gfx_partition_size >> AMDGPU_OA_SHIFT; - gds_info.oa_per_compute_partition = adev->gds.oa.cs_partition_size >> AMDGPU_OA_SHIFT; + gds_info.gds_gfx_partition_size = adev->gds.mem.gfx_partition_size; + gds_info.compute_partition_size = adev->gds.mem.cs_partition_size; + gds_info.gds_total_size = adev->gds.mem.total_size; + gds_info.gws_per_gfx_partition = adev->gds.gws.gfx_partition_size; + gds_info.gws_per_compute_partition = adev->gds.gws.cs_partition_size; + gds_info.oa_per_gfx_partition = adev->gds.oa.gfx_partition_size; + gds_info.oa_per_compute_partition = adev->gds.oa.cs_partition_size; return copy_to_user(out, &gds_info, min((size_t)size, sizeof(gds_info))) ? -EFAULT : 0; } case AMDGPU_INFO_VRAM_GTT: { struct drm_amdgpu_info_vram_gtt vram_gtt; - vram_gtt.vram_size = adev->gmc.real_vram_size; - vram_gtt.vram_size -= adev->vram_pin_size; - vram_gtt.vram_cpu_accessible_size = adev->gmc.visible_vram_size; - vram_gtt.vram_cpu_accessible_size -= (adev->vram_pin_size - adev->invisible_pin_size); + vram_gtt.vram_size = adev->gmc.real_vram_size - + atomic64_read(&adev->vram_pin_size); + vram_gtt.vram_cpu_accessible_size = adev->gmc.visible_vram_size - + atomic64_read(&adev->visible_pin_size); vram_gtt.gtt_size = adev->mman.bdev.man[TTM_PL_TT].size; vram_gtt.gtt_size *= PAGE_SIZE; - vram_gtt.gtt_size -= adev->gart_pin_size; + vram_gtt.gtt_size -= atomic64_read(&adev->gart_pin_size); return copy_to_user(out, &vram_gtt, min((size_t)size, sizeof(vram_gtt))) ? -EFAULT : 0; } @@ -509,17 +608,16 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file memset(&mem, 0, sizeof(mem)); mem.vram.total_heap_size = adev->gmc.real_vram_size; - mem.vram.usable_heap_size = - adev->gmc.real_vram_size - adev->vram_pin_size; + mem.vram.usable_heap_size = adev->gmc.real_vram_size - + atomic64_read(&adev->vram_pin_size); mem.vram.heap_usage = amdgpu_vram_mgr_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); mem.vram.max_allocation = mem.vram.usable_heap_size * 3 / 4; mem.cpu_accessible_vram.total_heap_size = adev->gmc.visible_vram_size; - mem.cpu_accessible_vram.usable_heap_size = - adev->gmc.visible_vram_size - - (adev->vram_pin_size - adev->invisible_pin_size); + mem.cpu_accessible_vram.usable_heap_size = adev->gmc.visible_vram_size - + atomic64_read(&adev->visible_pin_size); mem.cpu_accessible_vram.heap_usage = amdgpu_vram_mgr_vis_usage(&adev->mman.bdev.man[TTM_PL_VRAM]); mem.cpu_accessible_vram.max_allocation = @@ -527,8 +625,8 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file mem.gtt.total_heap_size = adev->mman.bdev.man[TTM_PL_TT].size; mem.gtt.total_heap_size *= PAGE_SIZE; - mem.gtt.usable_heap_size = mem.gtt.total_heap_size - - adev->gart_pin_size; + mem.gtt.usable_heap_size = mem.gtt.total_heap_size - + atomic64_read(&adev->gart_pin_size); mem.gtt.heap_usage = amdgpu_gtt_mgr_usage(&adev->mman.bdev.man[TTM_PL_TT]); mem.gtt.max_allocation = mem.gtt.usable_heap_size * 3 / 4; @@ -607,16 +705,17 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file vm_size -= AMDGPU_VA_RESERVED_SIZE; /* Older VCE FW versions are buggy and can handle only 40bits */ - if (adev->vce.fw_version < AMDGPU_VCE_FW_53_45) + if (adev->vce.fw_version && + adev->vce.fw_version < AMDGPU_VCE_FW_53_45) vm_size = min(vm_size, 1ULL << 40); dev_info.virtual_address_offset = AMDGPU_VA_RESERVED_SIZE; dev_info.virtual_address_max = - min(vm_size, AMDGPU_VA_HOLE_START); + min(vm_size, AMDGPU_GMC_HOLE_START); - if (vm_size > AMDGPU_VA_HOLE_START) { - dev_info.high_va_offset = AMDGPU_VA_HOLE_END; - dev_info.high_va_max = AMDGPU_VA_HOLE_END | vm_size; + if (vm_size > AMDGPU_GMC_HOLE_START) { + dev_info.high_va_offset = AMDGPU_GMC_HOLE_END; + dev_info.high_va_max = AMDGPU_GMC_HOLE_END | vm_size; } dev_info.virtual_address_alignment = max((int)PAGE_SIZE, AMDGPU_GPU_PAGE_SIZE); dev_info.pte_fragment_size = (1 << adev->vm_manager.fragment_size) * AMDGPU_GPU_PAGE_SIZE; @@ -930,12 +1029,11 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, return; pm_runtime_get_sync(dev->dev); - amdgpu_ctx_mgr_entity_fini(&fpriv->ctx_mgr); - if (adev->asic_type != CHIP_RAVEN) { + if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_UVD) != NULL) amdgpu_uvd_free_handles(adev, file_priv); + if (amdgpu_device_ip_get_ip_block(adev, AMD_IP_BLOCK_TYPE_VCE) != NULL) amdgpu_vce_free_handles(adev, file_priv); - } amdgpu_vm_bo_rmv(adev, fpriv->prt_va); @@ -958,7 +1056,7 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, amdgpu_bo_unref(&pd); idr_for_each_entry(&fpriv->bo_list_handles, list, handle) - amdgpu_bo_list_free(list); + amdgpu_bo_list_put(list); idr_destroy(&fpriv->bo_list_handles); mutex_destroy(&fpriv->bo_list_lock); @@ -1253,6 +1351,14 @@ static int amdgpu_debugfs_firmware_info(struct seq_file *m, void *data) seq_printf(m, "VCN feature version: %u, firmware version: 0x%08x\n", fw_info.feature, fw_info.ver); + /* DMCU */ + query_fw.fw_type = AMDGPU_INFO_FW_DMCU; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "DMCU feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + seq_printf(m, "VBIOS version: %s\n", ctx->vbios_version); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index 83e344fbb50a..e55508b39496 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -28,6 +28,21 @@ * Christian König <[email protected]> */ +/** + * DOC: MMU Notifier + * + * For coherent userptr handling registers an MMU notifier to inform the driver + * about updates on the page tables of a process. + * + * When somebody tries to invalidate the page tables we block the update until + * all operations on the pages in question are completed, then those pages are + * marked as accessed and also dirty if it wasn't a read only access. + * + * New command submissions using the userptrs in question are delayed until all + * page table invalidation are completed and we once more see a coherent process + * address space. + */ + #include <linux/firmware.h> #include <linux/module.h> #include <linux/mmu_notifier.h> @@ -38,6 +53,22 @@ #include "amdgpu.h" #include "amdgpu_amdkfd.h" +/** + * struct amdgpu_mn + * + * @adev: amdgpu device pointer + * @mm: process address space + * @mn: MMU notifier structure + * @type: type of MMU notifier + * @work: destruction work item + * @node: hash table node to find structure by adev and mn + * @lock: rw semaphore protecting the notifier nodes + * @objects: interval tree containing amdgpu_mn_nodes + * @read_lock: mutex for recursive locking of @lock + * @recursion: depth of recursion + * + * Data for each amdgpu device and process address space. + */ struct amdgpu_mn { /* constant after initialisation */ struct amdgpu_device *adev; @@ -58,13 +89,21 @@ struct amdgpu_mn { atomic_t recursion; }; +/** + * struct amdgpu_mn_node + * + * @it: interval node defining start-last of the affected address range + * @bos: list of all BOs in the affected address range + * + * Manages all BOs which are affected of a certain range of address space. + */ struct amdgpu_mn_node { struct interval_tree_node it; struct list_head bos; }; /** - * amdgpu_mn_destroy - destroy the rmn + * amdgpu_mn_destroy - destroy the MMU notifier * * @work: previously sheduled work item * @@ -72,47 +111,50 @@ struct amdgpu_mn_node { */ static void amdgpu_mn_destroy(struct work_struct *work) { - struct amdgpu_mn *rmn = container_of(work, struct amdgpu_mn, work); - struct amdgpu_device *adev = rmn->adev; + struct amdgpu_mn *amn = container_of(work, struct amdgpu_mn, work); + struct amdgpu_device *adev = amn->adev; struct amdgpu_mn_node *node, *next_node; struct amdgpu_bo *bo, *next_bo; mutex_lock(&adev->mn_lock); - down_write(&rmn->lock); - hash_del(&rmn->node); + down_write(&amn->lock); + hash_del(&amn->node); rbtree_postorder_for_each_entry_safe(node, next_node, - &rmn->objects.rb_root, it.rb) { + &amn->objects.rb_root, it.rb) { list_for_each_entry_safe(bo, next_bo, &node->bos, mn_list) { bo->mn = NULL; list_del_init(&bo->mn_list); } kfree(node); } - up_write(&rmn->lock); + up_write(&amn->lock); mutex_unlock(&adev->mn_lock); - mmu_notifier_unregister_no_release(&rmn->mn, rmn->mm); - kfree(rmn); + mmu_notifier_unregister_no_release(&amn->mn, amn->mm); + kfree(amn); } /** * amdgpu_mn_release - callback to notify about mm destruction * * @mn: our notifier - * @mn: the mm this callback is about + * @mm: the mm this callback is about * * Shedule a work item to lazy destroy our notifier. */ static void amdgpu_mn_release(struct mmu_notifier *mn, struct mm_struct *mm) { - struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); - INIT_WORK(&rmn->work, amdgpu_mn_destroy); - schedule_work(&rmn->work); + struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn); + + INIT_WORK(&amn->work, amdgpu_mn_destroy); + schedule_work(&amn->work); } /** - * amdgpu_mn_lock - take the write side lock for this mn + * amdgpu_mn_lock - take the write side lock for this notifier + * + * @mn: our notifier */ void amdgpu_mn_lock(struct amdgpu_mn *mn) { @@ -121,7 +163,9 @@ void amdgpu_mn_lock(struct amdgpu_mn *mn) } /** - * amdgpu_mn_unlock - drop the write side lock for this mn + * amdgpu_mn_unlock - drop the write side lock for this notifier + * + * @mn: our notifier */ void amdgpu_mn_unlock(struct amdgpu_mn *mn) { @@ -130,40 +174,44 @@ void amdgpu_mn_unlock(struct amdgpu_mn *mn) } /** - * amdgpu_mn_read_lock - take the rmn read lock + * amdgpu_mn_read_lock - take the read side lock for this notifier * - * @rmn: our notifier - * - * Take the rmn read side lock. + * @amn: our notifier */ -static void amdgpu_mn_read_lock(struct amdgpu_mn *rmn) +static int amdgpu_mn_read_lock(struct amdgpu_mn *amn, bool blockable) { - mutex_lock(&rmn->read_lock); - if (atomic_inc_return(&rmn->recursion) == 1) - down_read_non_owner(&rmn->lock); - mutex_unlock(&rmn->read_lock); + if (blockable) + mutex_lock(&amn->read_lock); + else if (!mutex_trylock(&amn->read_lock)) + return -EAGAIN; + + if (atomic_inc_return(&amn->recursion) == 1) + down_read_non_owner(&amn->lock); + mutex_unlock(&amn->read_lock); + + return 0; } /** - * amdgpu_mn_read_unlock - drop the rmn read lock + * amdgpu_mn_read_unlock - drop the read side lock for this notifier * - * @rmn: our notifier - * - * Drop the rmn read side lock. + * @amn: our notifier */ -static void amdgpu_mn_read_unlock(struct amdgpu_mn *rmn) +static void amdgpu_mn_read_unlock(struct amdgpu_mn *amn) { - if (atomic_dec_return(&rmn->recursion) == 0) - up_read_non_owner(&rmn->lock); + if (atomic_dec_return(&amn->recursion) == 0) + up_read_non_owner(&amn->lock); } /** * amdgpu_mn_invalidate_node - unmap all BOs of a node * * @node: the node with the BOs to unmap + * @start: start of address range affected + * @end: end of address range affected * - * We block for all BOs and unmap them by move them - * into system domain again. + * Block for operations on BOs to finish and mark pages as accessed and + * potentially dirty. */ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, unsigned long start, @@ -190,42 +238,54 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, * amdgpu_mn_invalidate_range_start_gfx - callback to notify about mm change * * @mn: our notifier - * @mn: the mm this callback is about + * @mm: the mm this callback is about * @start: start of updated range * @end: end of updated range * - * We block for all BOs between start and end to be idle and - * unmap them by move them into system domain again. + * Block for operations on BOs to finish and mark pages as accessed and + * potentially dirty. */ -static void amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn, +static int amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn, struct mm_struct *mm, unsigned long start, - unsigned long end) + unsigned long end, + bool blockable) { - struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); + struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn); struct interval_tree_node *it; /* notification is exclusive, but interval is inclusive */ end -= 1; - amdgpu_mn_read_lock(rmn); + /* TODO we should be able to split locking for interval tree and + * amdgpu_mn_invalidate_node + */ + if (amdgpu_mn_read_lock(amn, blockable)) + return -EAGAIN; - it = interval_tree_iter_first(&rmn->objects, start, end); + it = interval_tree_iter_first(&amn->objects, start, end); while (it) { struct amdgpu_mn_node *node; + if (!blockable) { + amdgpu_mn_read_unlock(amn); + return -EAGAIN; + } + node = container_of(it, struct amdgpu_mn_node, it); it = interval_tree_iter_next(it, start, end); amdgpu_mn_invalidate_node(node, start, end); } + + return 0; } /** * amdgpu_mn_invalidate_range_start_hsa - callback to notify about mm change * * @mn: our notifier - * @mn: the mm this callback is about + * @mm: the mm this callback is about * @start: start of updated range * @end: end of updated range * @@ -233,24 +293,31 @@ static void amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn, * necessitates evicting all user-mode queues of the process. The BOs * are restorted in amdgpu_mn_invalidate_range_end_hsa. */ -static void amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn, +static int amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn, struct mm_struct *mm, unsigned long start, - unsigned long end) + unsigned long end, + bool blockable) { - struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); + struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn); struct interval_tree_node *it; /* notification is exclusive, but interval is inclusive */ end -= 1; - amdgpu_mn_read_lock(rmn); + if (amdgpu_mn_read_lock(amn, blockable)) + return -EAGAIN; - it = interval_tree_iter_first(&rmn->objects, start, end); + it = interval_tree_iter_first(&amn->objects, start, end); while (it) { struct amdgpu_mn_node *node; struct amdgpu_bo *bo; + if (!blockable) { + amdgpu_mn_read_unlock(amn); + return -EAGAIN; + } + node = container_of(it, struct amdgpu_mn_node, it); it = interval_tree_iter_next(it, start, end); @@ -262,13 +329,15 @@ static void amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn, amdgpu_amdkfd_evict_userptr(mem, mm); } } + + return 0; } /** * amdgpu_mn_invalidate_range_end - callback to notify about mm change * * @mn: our notifier - * @mn: the mm this callback is about + * @mm: the mm this callback is about * @start: start of updated range * @end: end of updated range * @@ -279,9 +348,9 @@ static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn, unsigned long start, unsigned long end) { - struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); + struct amdgpu_mn *amn = container_of(mn, struct amdgpu_mn, mn); - amdgpu_mn_read_unlock(rmn); + amdgpu_mn_read_unlock(amn); } static const struct mmu_notifier_ops amdgpu_mn_ops[] = { @@ -315,7 +384,7 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev, enum amdgpu_mn_type type) { struct mm_struct *mm = current->mm; - struct amdgpu_mn *rmn; + struct amdgpu_mn *amn; unsigned long key = AMDGPU_MN_KEY(mm, type); int r; @@ -325,41 +394,41 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev, return ERR_PTR(-EINTR); } - hash_for_each_possible(adev->mn_hash, rmn, node, key) - if (AMDGPU_MN_KEY(rmn->mm, rmn->type) == key) + hash_for_each_possible(adev->mn_hash, amn, node, key) + if (AMDGPU_MN_KEY(amn->mm, amn->type) == key) goto release_locks; - rmn = kzalloc(sizeof(*rmn), GFP_KERNEL); - if (!rmn) { - rmn = ERR_PTR(-ENOMEM); + amn = kzalloc(sizeof(*amn), GFP_KERNEL); + if (!amn) { + amn = ERR_PTR(-ENOMEM); goto release_locks; } - rmn->adev = adev; - rmn->mm = mm; - init_rwsem(&rmn->lock); - rmn->type = type; - rmn->mn.ops = &amdgpu_mn_ops[type]; - rmn->objects = RB_ROOT_CACHED; - mutex_init(&rmn->read_lock); - atomic_set(&rmn->recursion, 0); + amn->adev = adev; + amn->mm = mm; + init_rwsem(&amn->lock); + amn->type = type; + amn->mn.ops = &amdgpu_mn_ops[type]; + amn->objects = RB_ROOT_CACHED; + mutex_init(&amn->read_lock); + atomic_set(&amn->recursion, 0); - r = __mmu_notifier_register(&rmn->mn, mm); + r = __mmu_notifier_register(&amn->mn, mm); if (r) - goto free_rmn; + goto free_amn; - hash_add(adev->mn_hash, &rmn->node, AMDGPU_MN_KEY(mm, type)); + hash_add(adev->mn_hash, &amn->node, AMDGPU_MN_KEY(mm, type)); release_locks: up_write(&mm->mmap_sem); mutex_unlock(&adev->mn_lock); - return rmn; + return amn; -free_rmn: +free_amn: up_write(&mm->mmap_sem); mutex_unlock(&adev->mn_lock); - kfree(rmn); + kfree(amn); return ERR_PTR(r); } @@ -379,14 +448,14 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); enum amdgpu_mn_type type = bo->kfd_bo ? AMDGPU_MN_TYPE_HSA : AMDGPU_MN_TYPE_GFX; - struct amdgpu_mn *rmn; + struct amdgpu_mn *amn; struct amdgpu_mn_node *node = NULL, *new_node; struct list_head bos; struct interval_tree_node *it; - rmn = amdgpu_mn_get(adev, type); - if (IS_ERR(rmn)) - return PTR_ERR(rmn); + amn = amdgpu_mn_get(adev, type); + if (IS_ERR(amn)) + return PTR_ERR(amn); new_node = kmalloc(sizeof(*new_node), GFP_KERNEL); if (!new_node) @@ -394,12 +463,12 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) INIT_LIST_HEAD(&bos); - down_write(&rmn->lock); + down_write(&amn->lock); - while ((it = interval_tree_iter_first(&rmn->objects, addr, end))) { + while ((it = interval_tree_iter_first(&amn->objects, addr, end))) { kfree(node); node = container_of(it, struct amdgpu_mn_node, it); - interval_tree_remove(&node->it, &rmn->objects); + interval_tree_remove(&node->it, &amn->objects); addr = min(it->start, addr); end = max(it->last, end); list_splice(&node->bos, &bos); @@ -410,7 +479,7 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) else kfree(new_node); - bo->mn = rmn; + bo->mn = amn; node->it.start = addr; node->it.last = end; @@ -418,9 +487,9 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) list_splice(&bos, &node->bos); list_add(&bo->mn_list, &node->bos); - interval_tree_insert(&node->it, &rmn->objects); + interval_tree_insert(&node->it, &amn->objects); - up_write(&rmn->lock); + up_write(&amn->lock); return 0; } @@ -435,18 +504,18 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) void amdgpu_mn_unregister(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - struct amdgpu_mn *rmn; + struct amdgpu_mn *amn; struct list_head *head; mutex_lock(&adev->mn_lock); - rmn = bo->mn; - if (rmn == NULL) { + amn = bo->mn; + if (amn == NULL) { mutex_unlock(&adev->mn_lock); return; } - down_write(&rmn->lock); + down_write(&amn->lock); /* save the next list entry for later */ head = bo->mn_list.next; @@ -456,12 +525,13 @@ void amdgpu_mn_unregister(struct amdgpu_bo *bo) if (list_empty(head)) { struct amdgpu_mn_node *node; + node = container_of(head, struct amdgpu_mn_node, bos); - interval_tree_remove(&node->it, &rmn->objects); + interval_tree_remove(&node->it, &amn->objects); kfree(node); } - up_write(&rmn->lock); + up_write(&amn->lock); mutex_unlock(&adev->mn_lock); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 3526efa8960e..904014dc5915 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -38,23 +38,48 @@ #include "amdgpu_trace.h" #include "amdgpu_amdkfd.h" -static bool amdgpu_need_backup(struct amdgpu_device *adev) -{ - if (adev->flags & AMD_IS_APU) - return false; +/** + * DOC: amdgpu_object + * + * This defines the interfaces to operate on an &amdgpu_bo buffer object which + * represents memory used by driver (VRAM, system memory, etc.). The driver + * provides DRM/GEM APIs to userspace. DRM/GEM APIs then use these interfaces + * to create/destroy/set buffer object which are then managed by the kernel TTM + * memory manager. + * The interfaces are also used internally by kernel clients, including gfx, + * uvd, etc. for kernel managed allocations used by the GPU. + * + */ - if (amdgpu_gpu_recovery == 0 || - (amdgpu_gpu_recovery == -1 && !amdgpu_sriov_vf(adev))) - return false; +/** + * amdgpu_bo_subtract_pin_size - Remove BO from pin_size accounting + * + * @bo: &amdgpu_bo buffer object + * + * This function is called when a BO stops being pinned, and updates the + * &amdgpu_device pin_size values accordingly. + */ +static void amdgpu_bo_subtract_pin_size(struct amdgpu_bo *bo) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - return true; + if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { + atomic64_sub(amdgpu_bo_size(bo), &adev->vram_pin_size); + atomic64_sub(amdgpu_vram_mgr_bo_visible_size(bo), + &adev->visible_pin_size); + } else if (bo->tbo.mem.mem_type == TTM_PL_TT) { + atomic64_sub(amdgpu_bo_size(bo), &adev->gart_pin_size); + } } -static void amdgpu_ttm_bo_destroy(struct ttm_buffer_object *tbo) +static void amdgpu_bo_destroy(struct ttm_buffer_object *tbo) { struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev); struct amdgpu_bo *bo = ttm_to_amdgpu_bo(tbo); + if (bo->pin_count > 0) + amdgpu_bo_subtract_pin_size(bo); + if (bo->kfd_bo) amdgpu_amdkfd_unreserve_system_memory_limit(bo); @@ -73,14 +98,32 @@ static void amdgpu_ttm_bo_destroy(struct ttm_buffer_object *tbo) kfree(bo); } -bool amdgpu_ttm_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) +/** + * amdgpu_bo_is_amdgpu_bo - check if the buffer object is an &amdgpu_bo + * @bo: buffer object to be checked + * + * Uses destroy function associated with the object to determine if this is + * an &amdgpu_bo. + * + * Returns: + * true if the object belongs to &amdgpu_bo, false if not. + */ +bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo) { - if (bo->destroy == &amdgpu_ttm_bo_destroy) + if (bo->destroy == &amdgpu_bo_destroy) return true; return false; } -void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) +/** + * amdgpu_bo_placement_from_domain - set buffer's placement + * @abo: &amdgpu_bo buffer object whose placement is to be set + * @domain: requested domain + * + * Sets buffer's placement according to requested domain and the buffer's + * flags. + */ +void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain) { struct amdgpu_device *adev = amdgpu_ttm_adev(abo->tbo.bdev); struct ttm_placement *placement = &abo->placement; @@ -108,10 +151,7 @@ void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) if (domain & AMDGPU_GEM_DOMAIN_GTT) { places[c].fpfn = 0; - if (flags & AMDGPU_GEM_CREATE_SHADOW) - places[c].lpfn = adev->gmc.gart_size >> PAGE_SHIFT; - else - places[c].lpfn = 0; + places[c].lpfn = 0; places[c].flags = TTM_PL_FLAG_TT; if (flags & AMDGPU_GEM_CREATE_CPU_GTT_USWC) places[c].flags |= TTM_PL_FLAG_WC | @@ -161,6 +201,8 @@ void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) c++; } + BUG_ON(c >= AMDGPU_BO_MAX_PLACEMENTS); + placement->num_placement = c; placement->placement = places; @@ -184,7 +226,8 @@ void amdgpu_ttm_placement_from_domain(struct amdgpu_bo *abo, u32 domain) * * Note: For bo_ptr new BO is only created if bo_ptr points to NULL. * - * Returns 0 on success, negative error code otherwise. + * Returns: + * 0 on success, negative error code otherwise. */ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, unsigned long size, int align, @@ -195,6 +238,11 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, bool free = false; int r; + if (!size) { + amdgpu_bo_unref(bo_ptr); + return 0; + } + memset(&bp, 0, sizeof(bp)); bp.size = size; bp.byte_align = align; @@ -220,22 +268,33 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, goto error_free; } - r = amdgpu_bo_pin(*bo_ptr, domain, gpu_addr); + r = amdgpu_bo_pin(*bo_ptr, domain); if (r) { dev_err(adev->dev, "(%d) kernel bo pin failed\n", r); goto error_unreserve; } + r = amdgpu_ttm_alloc_gart(&(*bo_ptr)->tbo); + if (r) { + dev_err(adev->dev, "%p bind failed\n", *bo_ptr); + goto error_unpin; + } + + if (gpu_addr) + *gpu_addr = amdgpu_bo_gpu_offset(*bo_ptr); + if (cpu_addr) { r = amdgpu_bo_kmap(*bo_ptr, cpu_addr); if (r) { dev_err(adev->dev, "(%d) kernel bo map failed\n", r); - goto error_unreserve; + goto error_unpin; } } return 0; +error_unpin: + amdgpu_bo_unpin(*bo_ptr); error_unreserve: amdgpu_bo_unreserve(*bo_ptr); @@ -261,7 +320,8 @@ error_free: * * Note: For bo_ptr new BO is only created if bo_ptr points to NULL. * - * Returns 0 on success, negative error code otherwise. + * Returns: + * 0 on success, negative error code otherwise. */ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, unsigned long size, int align, @@ -276,7 +336,8 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, if (r) return r; - amdgpu_bo_unreserve(*bo_ptr); + if (*bo_ptr) + amdgpu_bo_unreserve(*bo_ptr); return 0; } @@ -285,6 +346,8 @@ int amdgpu_bo_create_kernel(struct amdgpu_device *adev, * amdgpu_bo_free_kernel - free BO for kernel use * * @bo: amdgpu BO to free + * @gpu_addr: pointer to where the BO's GPU memory space address was stored + * @cpu_addr: pointer to where the BO's CPU memory space address was stored * * unmaps and unpin a BO for kernel internal use. */ @@ -364,7 +427,11 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, int r; page_align = roundup(bp->byte_align, PAGE_SIZE) >> PAGE_SHIFT; - size = ALIGN(size, PAGE_SIZE); + if (bp->domain & (AMDGPU_GEM_DOMAIN_GDS | AMDGPU_GEM_DOMAIN_GWS | + AMDGPU_GEM_DOMAIN_OA)) + size <<= PAGE_SHIFT; + else + size = ALIGN(size, PAGE_SIZE); if (!amdgpu_bo_validate_size(adev, size, bp->domain)) return -ENOMEM; @@ -379,7 +446,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, return -ENOMEM; drm_gem_private_object_init(adev->ddev, &bo->gem_base, size); INIT_LIST_HEAD(&bo->shadow_list); - INIT_LIST_HEAD(&bo->va); + bo->vm_bo = NULL; bo->preferred_domains = bp->preferred_domain ? bp->preferred_domain : bp->domain; bo->allowed_domains = bo->preferred_domains; @@ -418,17 +485,17 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, #endif bo->tbo.bdev = &adev->mman.bdev; - amdgpu_ttm_placement_from_domain(bo, bp->domain); + amdgpu_bo_placement_from_domain(bo, bp->domain); if (bp->type == ttm_bo_type_kernel) bo->tbo.priority = 1; r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, bp->type, &bo->placement, page_align, &ctx, acc_size, - NULL, bp->resv, &amdgpu_ttm_bo_destroy); + NULL, bp->resv, &amdgpu_bo_destroy); if (unlikely(r != 0)) return r; - if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size && + if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && bo->tbo.mem.mem_type == TTM_PL_VRAM && bo->tbo.mem.start < adev->gmc.visible_vram_size >> PAGE_SHIFT) amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, @@ -469,7 +536,7 @@ fail_unreserve: } static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, - unsigned long size, int byte_align, + unsigned long size, struct amdgpu_bo *bo) { struct amdgpu_bo_param bp; @@ -480,7 +547,6 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, memset(&bp, 0, sizeof(bp)); bp.size = size; - bp.byte_align = byte_align; bp.domain = AMDGPU_GEM_DOMAIN_GTT; bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC | AMDGPU_GEM_CREATE_SHADOW; @@ -491,13 +557,27 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); - list_add_tail(&bo->shadow_list, &adev->shadow_list); + list_add_tail(&bo->shadow->shadow_list, &adev->shadow_list); mutex_unlock(&adev->shadow_list_lock); } return r; } +/** + * amdgpu_bo_create - create an &amdgpu_bo buffer object + * @adev: amdgpu device object + * @bp: parameters to be used for the buffer object + * @bo_ptr: pointer to the buffer object pointer + * + * Creates an &amdgpu_bo buffer object; and if requested, also creates a + * shadow object. + * Shadow object is used to backup the original buffer object, and is always + * in GTT. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_create(struct amdgpu_device *adev, struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr) @@ -510,12 +590,12 @@ int amdgpu_bo_create(struct amdgpu_device *adev, if (r) return r; - if ((flags & AMDGPU_GEM_CREATE_SHADOW) && amdgpu_need_backup(adev)) { + if ((flags & AMDGPU_GEM_CREATE_SHADOW) && !(adev->flags & AMD_IS_APU)) { if (!bp->resv) WARN_ON(reservation_object_lock((*bo_ptr)->tbo.resv, NULL)); - r = amdgpu_bo_create_shadow(adev, bp->size, bp->byte_align, (*bo_ptr)); + r = amdgpu_bo_create_shadow(adev, bp->size, *bo_ptr); if (!bp->resv) reservation_object_unlock((*bo_ptr)->tbo.resv); @@ -527,6 +607,21 @@ int amdgpu_bo_create(struct amdgpu_device *adev, return r; } +/** + * amdgpu_bo_backup_to_shadow - Backs up an &amdgpu_bo buffer object + * @adev: amdgpu device object + * @ring: amdgpu_ring for the engine handling the buffer operations + * @bo: &amdgpu_bo buffer to be backed up + * @resv: reservation object with embedded fence + * @fence: dma_fence associated with the operation + * @direct: whether to submit the job directly + * + * Copies an &amdgpu_bo buffer object to its shadow object. + * Not used for now. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, struct amdgpu_ring *ring, struct amdgpu_bo *bo, @@ -559,6 +654,18 @@ err: return r; } +/** + * amdgpu_bo_validate - validate an &amdgpu_bo buffer object + * @bo: pointer to the buffer object + * + * Sets placement according to domain; and changes placement and caching + * policy of the buffer object according to the placement. + * This is used for validating shadow bos. It calls ttm_bo_validate() to + * make sure the buffer is resident where it needs to be. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_validate(struct amdgpu_bo *bo) { struct ttm_operation_ctx ctx = { false, false }; @@ -571,7 +678,7 @@ int amdgpu_bo_validate(struct amdgpu_bo *bo) domain = bo->preferred_domains; retry: - amdgpu_ttm_placement_from_domain(bo, domain); + amdgpu_bo_placement_from_domain(bo, domain); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) { domain = bo->allowed_domains; @@ -581,38 +688,45 @@ retry: return r; } -int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_bo *bo, - struct reservation_object *resv, - struct dma_fence **fence, - bool direct) +/** + * amdgpu_bo_restore_shadow - restore an &amdgpu_bo shadow + * + * @shadow: &amdgpu_bo shadow to be restored + * @fence: dma_fence associated with the operation + * + * Copies a buffer object's shadow content back to the object. + * This is used for recovering a buffer from its shadow in case of a gpu + * reset where vram context may be lost. + * + * Returns: + * 0 for success or a negative error code on failure. + */ +int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, struct dma_fence **fence) { - struct amdgpu_bo *shadow = bo->shadow; - uint64_t bo_addr, shadow_addr; - int r; + struct amdgpu_device *adev = amdgpu_ttm_adev(shadow->tbo.bdev); + struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; + uint64_t shadow_addr, parent_addr; - if (!shadow) - return -EINVAL; + shadow_addr = amdgpu_bo_gpu_offset(shadow); + parent_addr = amdgpu_bo_gpu_offset(shadow->parent); - bo_addr = amdgpu_bo_gpu_offset(bo); - shadow_addr = amdgpu_bo_gpu_offset(bo->shadow); - - r = reservation_object_reserve_shared(bo->tbo.resv); - if (r) - goto err; - - r = amdgpu_copy_buffer(ring, shadow_addr, bo_addr, - amdgpu_bo_size(bo), resv, fence, - direct, false); - if (!r) - amdgpu_bo_fence(bo, *fence, true); - -err: - return r; + return amdgpu_copy_buffer(ring, shadow_addr, parent_addr, + amdgpu_bo_size(shadow), NULL, fence, + true, false); } +/** + * amdgpu_bo_kmap - map an &amdgpu_bo buffer object + * @bo: &amdgpu_bo buffer object to be mapped + * @ptr: kernel virtual address to be returned + * + * Calls ttm_bo_kmap() to set up the kernel virtual mapping; calls + * amdgpu_bo_kptr() to get the kernel virtual address. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) { void *kptr; @@ -643,6 +757,15 @@ int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr) return 0; } +/** + * amdgpu_bo_kptr - returns a kernel virtual address of the buffer object + * @bo: &amdgpu_bo buffer object + * + * Calls ttm_kmap_obj_virtual() to get the kernel virtual address + * + * Returns: + * the virtual address of a buffer object area. + */ void *amdgpu_bo_kptr(struct amdgpu_bo *bo) { bool is_iomem; @@ -650,21 +773,42 @@ void *amdgpu_bo_kptr(struct amdgpu_bo *bo) return ttm_kmap_obj_virtual(&bo->kmap, &is_iomem); } +/** + * amdgpu_bo_kunmap - unmap an &amdgpu_bo buffer object + * @bo: &amdgpu_bo buffer object to be unmapped + * + * Unmaps a kernel map set up by amdgpu_bo_kmap(). + */ void amdgpu_bo_kunmap(struct amdgpu_bo *bo) { if (bo->kmap.bo) ttm_bo_kunmap(&bo->kmap); } +/** + * amdgpu_bo_ref - reference an &amdgpu_bo buffer object + * @bo: &amdgpu_bo buffer object + * + * References the contained &ttm_buffer_object. + * + * Returns: + * a refcounted pointer to the &amdgpu_bo buffer object. + */ struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo) { if (bo == NULL) return NULL; - ttm_bo_reference(&bo->tbo); + ttm_bo_get(&bo->tbo); return bo; } +/** + * amdgpu_bo_unref - unreference an &amdgpu_bo buffer object + * @bo: &amdgpu_bo buffer object + * + * Unreferences the contained &ttm_buffer_object and clear the pointer + */ void amdgpu_bo_unref(struct amdgpu_bo **bo) { struct ttm_buffer_object *tbo; @@ -673,14 +817,34 @@ void amdgpu_bo_unref(struct amdgpu_bo **bo) return; tbo = &((*bo)->tbo); - ttm_bo_unref(&tbo); - if (tbo == NULL) - *bo = NULL; + ttm_bo_put(tbo); + *bo = NULL; } +/** + * amdgpu_bo_pin_restricted - pin an &amdgpu_bo buffer object + * @bo: &amdgpu_bo buffer object to be pinned + * @domain: domain to be pinned to + * @min_offset: the start of requested address range + * @max_offset: the end of requested address range + * + * Pins the buffer object according to requested domain and address range. If + * the memory is unbound gart memory, binds the pages into gart table. Adjusts + * pin_count and pin_size accordingly. + * + * Pinning means to lock pages in memory along with keeping them at a fixed + * offset. It is required when a buffer can not be moved, for example, when + * a display buffer is being scanned out. + * + * Compared with amdgpu_bo_pin(), this function gives more flexibility on + * where to pin a buffer if there are specific restrictions on where a buffer + * must be located. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, - u64 min_offset, u64 max_offset, - u64 *gpu_addr) + u64 min_offset, u64 max_offset) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct ttm_operation_ctx ctx = { false, false }; @@ -712,8 +876,6 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, return -EINVAL; bo->pin_count++; - if (gpu_addr) - *gpu_addr = amdgpu_bo_gpu_offset(bo); if (max_offset != 0) { u64 domain_start = bo->tbo.bdev->man[mem_type].gpu_offset; @@ -728,7 +890,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, /* force to pin into visible video ram */ if (!(bo->flags & AMDGPU_GEM_CREATE_NO_CPU_ACCESS)) bo->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - amdgpu_ttm_placement_from_domain(bo, domain); + amdgpu_bo_placement_from_domain(bo, domain); for (i = 0; i < bo->placement.num_placement; i++) { unsigned fpfn, lpfn; @@ -749,33 +911,48 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, goto error; } - r = amdgpu_ttm_alloc_gart(&bo->tbo); - if (unlikely(r)) { - dev_err(adev->dev, "%p bind failed\n", bo); - goto error; - } - bo->pin_count = 1; - if (gpu_addr != NULL) - *gpu_addr = amdgpu_bo_gpu_offset(bo); domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); if (domain == AMDGPU_GEM_DOMAIN_VRAM) { - adev->vram_pin_size += amdgpu_bo_size(bo); - adev->invisible_pin_size += amdgpu_vram_mgr_bo_invisible_size(bo); + atomic64_add(amdgpu_bo_size(bo), &adev->vram_pin_size); + atomic64_add(amdgpu_vram_mgr_bo_visible_size(bo), + &adev->visible_pin_size); } else if (domain == AMDGPU_GEM_DOMAIN_GTT) { - adev->gart_pin_size += amdgpu_bo_size(bo); + atomic64_add(amdgpu_bo_size(bo), &adev->gart_pin_size); } error: return r; } -int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain, u64 *gpu_addr) +/** + * amdgpu_bo_pin - pin an &amdgpu_bo buffer object + * @bo: &amdgpu_bo buffer object to be pinned + * @domain: domain to be pinned to + * + * A simple wrapper to amdgpu_bo_pin_restricted(). + * Provides a simpler API for buffers that do not have any strict restrictions + * on where a buffer must be located. + * + * Returns: + * 0 for success or a negative error code on failure. + */ +int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain) { - return amdgpu_bo_pin_restricted(bo, domain, 0, 0, gpu_addr); + return amdgpu_bo_pin_restricted(bo, domain, 0, 0); } +/** + * amdgpu_bo_unpin - unpin an &amdgpu_bo buffer object + * @bo: &amdgpu_bo buffer object to be unpinned + * + * Decreases the pin_count, and clears the flags if pin_count reaches 0. + * Changes placement and pin size accordingly. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_unpin(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); @@ -790,12 +967,7 @@ int amdgpu_bo_unpin(struct amdgpu_bo *bo) if (bo->pin_count) return 0; - if (bo->tbo.mem.mem_type == TTM_PL_VRAM) { - adev->vram_pin_size -= amdgpu_bo_size(bo); - adev->invisible_pin_size -= amdgpu_vram_mgr_bo_invisible_size(bo); - } else if (bo->tbo.mem.mem_type == TTM_PL_TT) { - adev->gart_pin_size -= amdgpu_bo_size(bo); - } + amdgpu_bo_subtract_pin_size(bo); for (i = 0; i < bo->placement.num_placement; i++) { bo->placements[i].lpfn = 0; @@ -808,13 +980,25 @@ int amdgpu_bo_unpin(struct amdgpu_bo *bo) return r; } +/** + * amdgpu_bo_evict_vram - evict VRAM buffers + * @adev: amdgpu device object + * + * Evicts all VRAM buffers on the lru list of the memory type. + * Mainly used for evicting vram at suspend time. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_evict_vram(struct amdgpu_device *adev) { /* late 2.6.33 fix IGP hibernate - we need pm ops to do this correct */ - if (0 && (adev->flags & AMD_IS_APU)) { +#ifndef CONFIG_HIBERNATION + if (adev->flags & AMD_IS_APU) { /* Useless to evict on IGP chips */ return 0; } +#endif return ttm_bo_evict_mm(&adev->mman.bdev, TTM_PL_VRAM); } @@ -830,6 +1014,15 @@ static const char *amdgpu_vram_names[] = { "DDR4", }; +/** + * amdgpu_bo_init - initialize memory manager + * @adev: amdgpu device object + * + * Calls amdgpu_ttm_init() to initialize amdgpu memory manager. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_init(struct amdgpu_device *adev) { /* reserve PAT memory space to WC for VRAM */ @@ -847,6 +1040,16 @@ int amdgpu_bo_init(struct amdgpu_device *adev) return amdgpu_ttm_init(adev); } +/** + * amdgpu_bo_late_init - late init + * @adev: amdgpu device object + * + * Calls amdgpu_ttm_late_init() to free resources used earlier during + * initialization. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_late_init(struct amdgpu_device *adev) { amdgpu_ttm_late_init(adev); @@ -854,6 +1057,12 @@ int amdgpu_bo_late_init(struct amdgpu_device *adev) return 0; } +/** + * amdgpu_bo_fini - tear down memory manager + * @adev: amdgpu device object + * + * Reverses amdgpu_bo_init() to tear down memory manager. + */ void amdgpu_bo_fini(struct amdgpu_device *adev) { amdgpu_ttm_fini(adev); @@ -861,12 +1070,33 @@ void amdgpu_bo_fini(struct amdgpu_device *adev) arch_io_free_memtype_wc(adev->gmc.aper_base, adev->gmc.aper_size); } +/** + * amdgpu_bo_fbdev_mmap - mmap fbdev memory + * @bo: &amdgpu_bo buffer object + * @vma: vma as input from the fbdev mmap method + * + * Calls ttm_fbdev_mmap() to mmap fbdev memory if it is backed by a bo. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_fbdev_mmap(struct amdgpu_bo *bo, struct vm_area_struct *vma) { return ttm_fbdev_mmap(vma, &bo->tbo); } +/** + * amdgpu_bo_set_tiling_flags - set tiling flags + * @bo: &amdgpu_bo buffer object + * @tiling_flags: new flags + * + * Sets buffer object's tiling flags with the new one. Used by GEM ioctl or + * kernel driver to set the tiling flags on a buffer. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_set_tiling_flags(struct amdgpu_bo *bo, u64 tiling_flags) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); @@ -879,6 +1109,14 @@ int amdgpu_bo_set_tiling_flags(struct amdgpu_bo *bo, u64 tiling_flags) return 0; } +/** + * amdgpu_bo_get_tiling_flags - get tiling flags + * @bo: &amdgpu_bo buffer object + * @tiling_flags: returned flags + * + * Gets buffer object's tiling flags. Used by GEM ioctl or kernel driver to + * set the tiling flags on a buffer. + */ void amdgpu_bo_get_tiling_flags(struct amdgpu_bo *bo, u64 *tiling_flags) { lockdep_assert_held(&bo->tbo.resv->lock.base); @@ -887,6 +1125,19 @@ void amdgpu_bo_get_tiling_flags(struct amdgpu_bo *bo, u64 *tiling_flags) *tiling_flags = bo->tiling_flags; } +/** + * amdgpu_bo_set_metadata - set metadata + * @bo: &amdgpu_bo buffer object + * @metadata: new metadata + * @metadata_size: size of the new metadata + * @flags: flags of the new metadata + * + * Sets buffer object's metadata, its size and flags. + * Used via GEM ioctl. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata, uint32_t metadata_size, uint64_t flags) { @@ -916,6 +1167,21 @@ int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata, return 0; } +/** + * amdgpu_bo_get_metadata - get metadata + * @bo: &amdgpu_bo buffer object + * @buffer: returned metadata + * @buffer_size: size of the buffer + * @metadata_size: size of the returned metadata + * @flags: flags of the returned metadata + * + * Gets buffer object's metadata, its size and flags. buffer_size shall not be + * less than metadata_size. + * Used via GEM ioctl. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer, size_t buffer_size, uint32_t *metadata_size, uint64_t *flags) @@ -939,6 +1205,16 @@ int amdgpu_bo_get_metadata(struct amdgpu_bo *bo, void *buffer, return 0; } +/** + * amdgpu_bo_move_notify - notification about a memory move + * @bo: pointer to a buffer object + * @evict: if this move is evicting the buffer from the graphics address space + * @new_mem: new information of the bufer object + * + * Marks the corresponding &amdgpu_bo buffer object as invalid, also performs + * bookkeeping. + * TTM driver callback which is called when ttm moves a buffer. + */ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, bool evict, struct ttm_mem_reg *new_mem) @@ -947,7 +1223,7 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, struct amdgpu_bo *abo; struct ttm_mem_reg *old_mem = &bo->mem; - if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) + if (!amdgpu_bo_is_amdgpu_bo(bo)) return; abo = ttm_to_amdgpu_bo(bo); @@ -964,9 +1240,20 @@ void amdgpu_bo_move_notify(struct ttm_buffer_object *bo, return; /* move_notify is called before move happens */ - trace_amdgpu_ttm_bo_move(abo, new_mem->mem_type, old_mem->mem_type); + trace_amdgpu_bo_move(abo, new_mem->mem_type, old_mem->mem_type); } +/** + * amdgpu_bo_fault_reserve_notify - notification about a memory fault + * @bo: pointer to a buffer object + * + * Notifies the driver we are taking a fault on this BO and have reserved it, + * also performs bookkeeping. + * TTM driver callback for dealing with vm faults. + * + * Returns: + * 0 for success or a negative error code on failure. + */ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); @@ -975,7 +1262,7 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) unsigned long offset, size; int r; - if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) + if (!amdgpu_bo_is_amdgpu_bo(bo)) return 0; abo = ttm_to_amdgpu_bo(bo); @@ -997,8 +1284,8 @@ int amdgpu_bo_fault_reserve_notify(struct ttm_buffer_object *bo) /* hurrah the memory is not visible ! */ atomic64_inc(&adev->num_vram_cpu_page_faults); - amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | - AMDGPU_GEM_DOMAIN_GTT); + amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | + AMDGPU_GEM_DOMAIN_GTT); /* Avoid costly evictions; only set GTT as a busy placement */ abo->placement.num_busy_placement = 1; @@ -1040,25 +1327,32 @@ void amdgpu_bo_fence(struct amdgpu_bo *bo, struct dma_fence *fence, * amdgpu_bo_gpu_offset - return GPU offset of bo * @bo: amdgpu object for which we query the offset * - * Returns current GPU offset of the object. - * * Note: object should either be pinned or reserved when calling this * function, it might be useful to add check for this for debugging. + * + * Returns: + * current GPU offset of the object. */ u64 amdgpu_bo_gpu_offset(struct amdgpu_bo *bo) { WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_SYSTEM); - WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_TT && - !amdgpu_gtt_mgr_has_gart_addr(&bo->tbo.mem)); WARN_ON_ONCE(!ww_mutex_is_locked(&bo->tbo.resv->lock) && - !bo->pin_count); + !bo->pin_count && bo->tbo.type != ttm_bo_type_kernel); WARN_ON_ONCE(bo->tbo.mem.start == AMDGPU_BO_INVALID_OFFSET); WARN_ON_ONCE(bo->tbo.mem.mem_type == TTM_PL_VRAM && !(bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS)); - return bo->tbo.offset; + return amdgpu_gmc_sign_extend(bo->tbo.offset); } +/** + * amdgpu_bo_get_preferred_pin_domain - get preferred domain for scanout + * @adev: amdgpu device object + * @domain: allowed :ref:`memory domains <amdgpu_memory_domains>` + * + * Returns: + * Which of the allowed domains is preferred for pinning the BO for scanout. + */ uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev, uint32_t domain) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 731748033878..7d3312d0da11 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -32,6 +32,7 @@ #include "amdgpu.h" #define AMDGPU_BO_INVALID_OFFSET LONG_MAX +#define AMDGPU_BO_MAX_PLACEMENTS 3 struct amdgpu_bo_param { unsigned long size; @@ -77,7 +78,7 @@ struct amdgpu_bo { /* Protected by tbo.reserved */ u32 preferred_domains; u32 allowed_domains; - struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1]; + struct ttm_place placements[AMDGPU_BO_MAX_PLACEMENTS]; struct ttm_placement placement; struct ttm_buffer_object tbo; struct ttm_bo_kmap_obj kmap; @@ -88,8 +89,8 @@ struct amdgpu_bo { void *metadata; u32 metadata_size; unsigned prime_shared_count; - /* list of all virtual address to which this bo is associated to */ - struct list_head va; + /* per VM structure for page tables and with virtual addresses */ + struct amdgpu_vm_bo_base *vm_bo; /* Constant after initialization */ struct drm_gem_object gem_base; struct amdgpu_bo *parent; @@ -193,19 +194,6 @@ static inline u64 amdgpu_bo_mmap_offset(struct amdgpu_bo *bo) } /** - * amdgpu_bo_gpu_accessible - return whether the bo is currently in memory that - * is accessible to the GPU. - */ -static inline bool amdgpu_bo_gpu_accessible(struct amdgpu_bo *bo) -{ - switch (bo->tbo.mem.mem_type) { - case TTM_PL_TT: return amdgpu_gtt_mgr_has_gart_addr(&bo->tbo.mem); - case TTM_PL_VRAM: return true; - default: return false; - } -} - -/** * amdgpu_bo_in_cpu_visible_vram - check if BO is (partly) in visible VRAM */ static inline bool amdgpu_bo_in_cpu_visible_vram(struct amdgpu_bo *bo) @@ -234,6 +222,9 @@ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC; } +bool amdgpu_bo_is_amdgpu_bo(struct ttm_buffer_object *bo); +void amdgpu_bo_placement_from_domain(struct amdgpu_bo *abo, u32 domain); + int amdgpu_bo_create(struct amdgpu_device *adev, struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr); @@ -252,10 +243,9 @@ void *amdgpu_bo_kptr(struct amdgpu_bo *bo); void amdgpu_bo_kunmap(struct amdgpu_bo *bo); struct amdgpu_bo *amdgpu_bo_ref(struct amdgpu_bo *bo); void amdgpu_bo_unref(struct amdgpu_bo **bo); -int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain, u64 *gpu_addr); +int amdgpu_bo_pin(struct amdgpu_bo *bo, u32 domain); int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, - u64 min_offset, u64 max_offset, - u64 *gpu_addr); + u64 min_offset, u64 max_offset); int amdgpu_bo_unpin(struct amdgpu_bo *bo); int amdgpu_bo_evict_vram(struct amdgpu_device *adev); int amdgpu_bo_init(struct amdgpu_device *adev); @@ -283,12 +273,8 @@ int amdgpu_bo_backup_to_shadow(struct amdgpu_device *adev, struct reservation_object *resv, struct dma_fence **fence, bool direct); int amdgpu_bo_validate(struct amdgpu_bo *bo); -int amdgpu_bo_restore_from_shadow(struct amdgpu_device *adev, - struct amdgpu_ring *ring, - struct amdgpu_bo *bo, - struct reservation_object *resv, - struct dma_fence **fence, - bool direct); +int amdgpu_bo_restore_shadow(struct amdgpu_bo *shadow, + struct dma_fence **fence); uint32_t amdgpu_bo_get_preferred_pin_domain(struct amdgpu_device *adev, uint32_t domain); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index fc818b4d849c..94055a485e01 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -27,11 +27,12 @@ #include "amdgpu_drv.h" #include "amdgpu_pm.h" #include "amdgpu_dpm.h" +#include "amdgpu_display.h" #include "atom.h" #include <linux/power_supply.h> #include <linux/hwmon.h> #include <linux/hwmon-sysfs.h> - +#include <linux/nospec.h> static int amdgpu_debugfs_pm_init(struct amdgpu_device *adev); @@ -68,11 +69,11 @@ void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev) if (adev->pm.dpm_enabled) { mutex_lock(&adev->pm.mutex); if (power_supply_is_system_supplied() > 0) - adev->pm.dpm.ac_power = true; + adev->pm.ac_power = true; else - adev->pm.dpm.ac_power = false; + adev->pm.ac_power = false; if (adev->powerplay.pp_funcs->enable_bapm) - amdgpu_dpm_enable_bapm(adev, adev->pm.dpm.ac_power); + amdgpu_dpm_enable_bapm(adev, adev->pm.ac_power); mutex_unlock(&adev->pm.mutex); } } @@ -80,12 +81,15 @@ void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev) /** * DOC: power_dpm_state * - * This is a legacy interface and is only provided for backwards compatibility. - * The amdgpu driver provides a sysfs API for adjusting certain power - * related parameters. The file power_dpm_state is used for this. + * The power_dpm_state file is a legacy interface and is only provided for + * backwards compatibility. The amdgpu driver provides a sysfs API for adjusting + * certain power related parameters. The file power_dpm_state is used for this. * It accepts the following arguments: + * * - battery + * * - balanced + * * - performance * * battery @@ -169,14 +173,21 @@ fail: * The amdgpu driver provides a sysfs API for adjusting certain power * related parameters. The file power_dpm_force_performance_level is * used for this. It accepts the following arguments: + * * - auto + * * - low + * * - high + * * - manual - * - GPU fan + * * - profile_standard + * * - profile_min_sclk + * * - profile_min_mclk + * * - profile_peak * * auto @@ -393,6 +404,7 @@ static ssize_t amdgpu_set_pp_force_state(struct device *dev, count = -EINVAL; goto fail; } + idx = array_index_nospec(idx, ARRAY_SIZE(data.states)); amdgpu_dpm_get_pp_num_states(adev, &data); state = data.states[idx]; @@ -462,9 +474,14 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * in each power level within a power state. The pp_od_clk_voltage is used for * this. * + * < For Vega10 and previous ASICs > + * * Reading the file will display: + * * - a list of engine clock levels and voltages labeled OD_SCLK + * * - a list of memory clock levels and voltages labeled OD_MCLK + * * - a list of valid ranges for sclk, mclk, and voltage labeled OD_RANGE * * To manually adjust these settings, first select manual using @@ -476,6 +493,44 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, * "c" (commit) to the file to commit your changes. If you want to reset to the * default power levels, write "r" (reset) to the file to reset them. * + * + * < For Vega20 > + * + * Reading the file will display: + * + * - minimum and maximum engine clock labeled OD_SCLK + * + * - maximum memory clock labeled OD_MCLK + * + * - three <frequency, voltage> points labeled OD_VDDC_CURVE. + * They can be used to calibrate the sclk voltage curve. + * + * - a list of valid ranges for sclk, mclk, and voltage curve points + * labeled OD_RANGE + * + * To manually adjust these settings: + * + * - First select manual using power_dpm_force_performance_level + * + * - For clock frequency setting, enter a new value by writing a + * string that contains "s/m index clock" to the file. The index + * should be 0 if to set minimum clock. And 1 if to set maximum + * clock. E.g., "s 0 500" will update minimum sclk to be 500 MHz. + * "m 1 800" will update maximum mclk to be 800Mhz. + * + * For sclk voltage curve, enter the new values by writing a + * string that contains "vc point clock voltage" to the file. The + * points are indexed by 0, 1 and 2. E.g., "vc 0 300 600" will + * update point1 with clock set as 300Mhz and voltage as + * 600mV. "vc 2 1000 1000" will update point3 with clock set + * as 1000Mhz and voltage 1000mV. + * + * - When you have edited all of the states as needed, write "c" (commit) + * to the file to commit your changes + * + * - If you want to reset to the default power levels, write "r" (reset) + * to the file to reset them + * */ static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev, @@ -505,6 +560,8 @@ static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev, type = PP_OD_RESTORE_DEFAULT_TABLE; else if (*buf == 'c') type = PP_OD_COMMIT_DPM_TABLE; + else if (!strncmp(buf, "vc", 2)) + type = PP_OD_EDIT_VDDC_CURVE; else return -EINVAL; @@ -512,6 +569,8 @@ static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev, tmp_str = buf_cpy; + if (type == PP_OD_EDIT_VDDC_CURVE) + tmp_str++; while (isspace(*++tmp_str)); while (tmp_str[0]) { @@ -555,6 +614,7 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, if (adev->powerplay.pp_funcs->print_clock_levels) { size = amdgpu_dpm_print_clock_levels(adev, OD_SCLK, buf); size += amdgpu_dpm_print_clock_levels(adev, OD_MCLK, buf+size); + size += amdgpu_dpm_print_clock_levels(adev, OD_VDDC_CURVE, buf+size); size += amdgpu_dpm_print_clock_levels(adev, OD_RANGE, buf+size); return size; } else { @@ -593,40 +653,59 @@ static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev, return snprintf(buf, PAGE_SIZE, "\n"); } -static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, - struct device_attribute *attr, - const char *buf, - size_t count) +/* + * Worst case: 32 bits individually specified, in octal at 12 characters + * per line (+1 for \n). + */ +#define AMDGPU_MASK_BUF_MAX (32 * 13) + +static ssize_t amdgpu_read_mask(const char *buf, size_t count, uint32_t *mask) { - struct drm_device *ddev = dev_get_drvdata(dev); - struct amdgpu_device *adev = ddev->dev_private; int ret; long level; - uint32_t mask = 0; char *sub_str = NULL; char *tmp; - char buf_cpy[count]; + char buf_cpy[AMDGPU_MASK_BUF_MAX + 1]; const char delimiter[3] = {' ', '\n', '\0'}; + size_t bytes; - memcpy(buf_cpy, buf, count+1); + *mask = 0; + + bytes = min(count, sizeof(buf_cpy) - 1); + memcpy(buf_cpy, buf, bytes); + buf_cpy[bytes] = '\0'; tmp = buf_cpy; while (tmp[0]) { - sub_str = strsep(&tmp, delimiter); + sub_str = strsep(&tmp, delimiter); if (strlen(sub_str)) { ret = kstrtol(sub_str, 0, &level); - - if (ret) { - count = -EINVAL; - goto fail; - } - mask |= 1 << level; + if (ret) + return -EINVAL; + *mask |= 1 << level; } else break; } + + return 0; +} + +static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + int ret; + uint32_t mask = 0; + + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; + if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_SCLK, mask); -fail: return count; } @@ -651,32 +730,15 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; int ret; - long level; uint32_t mask = 0; - char *sub_str = NULL; - char *tmp; - char buf_cpy[count]; - const char delimiter[3] = {' ', '\n', '\0'}; - memcpy(buf_cpy, buf, count+1); - tmp = buf_cpy; - while (tmp[0]) { - sub_str = strsep(&tmp, delimiter); - if (strlen(sub_str)) { - ret = kstrtol(sub_str, 0, &level); + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; - if (ret) { - count = -EINVAL; - goto fail; - } - mask |= 1 << level; - } else - break; - } if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_MCLK, mask); -fail: return count; } @@ -701,33 +763,15 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, struct drm_device *ddev = dev_get_drvdata(dev); struct amdgpu_device *adev = ddev->dev_private; int ret; - long level; uint32_t mask = 0; - char *sub_str = NULL; - char *tmp; - char buf_cpy[count]; - const char delimiter[3] = {' ', '\n', '\0'}; - memcpy(buf_cpy, buf, count+1); - tmp = buf_cpy; - - while (tmp[0]) { - sub_str = strsep(&tmp, delimiter); - if (strlen(sub_str)) { - ret = kstrtol(sub_str, 0, &level); + ret = amdgpu_read_mask(buf, count, &mask); + if (ret) + return ret; - if (ret) { - count = -EINVAL; - goto fail; - } - mask |= 1 << level; - } else - break; - } if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_PCIE, mask); -fail: return count; } @@ -905,6 +949,36 @@ fail: return -EINVAL; } +/** + * DOC: busy_percent + * + * The amdgpu driver provides a sysfs API for reading how busy the GPU + * is as a percentage. The file gpu_busy_percent is used for this. + * The SMU firmware computes a percentage of load based on the + * aggregate activity level in the IP cores. + */ +static ssize_t amdgpu_get_busy_percent(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct drm_device *ddev = dev_get_drvdata(dev); + struct amdgpu_device *adev = ddev->dev_private; + int r, value, size = sizeof(value); + + /* sanity check PP is enabled */ + if (!(adev->powerplay.pp_funcs && + adev->powerplay.pp_funcs->read_sensor)) + return -EINVAL; + + /* read the IP busy sensor */ + r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_LOAD, + (void *)&value, &size); + if (r) + return r; + + return snprintf(buf, PAGE_SIZE, "%d\n", value); +} + static DEVICE_ATTR(power_dpm_state, S_IRUGO | S_IWUSR, amdgpu_get_dpm_state, amdgpu_set_dpm_state); static DEVICE_ATTR(power_dpm_force_performance_level, S_IRUGO | S_IWUSR, amdgpu_get_dpm_forced_performance_level, @@ -938,6 +1012,8 @@ static DEVICE_ATTR(pp_power_profile_mode, S_IRUGO | S_IWUSR, static DEVICE_ATTR(pp_od_clk_voltage, S_IRUGO | S_IWUSR, amdgpu_get_pp_od_clk_voltage, amdgpu_set_pp_od_clk_voltage); +static DEVICE_ATTR(gpu_busy_percent, S_IRUGO, + amdgpu_get_busy_percent, NULL); static ssize_t amdgpu_hwmon_show_temp(struct device *dev, struct device_attribute *attr, @@ -1044,12 +1120,19 @@ static ssize_t amdgpu_hwmon_set_pwm1(struct device *dev, struct amdgpu_device *adev = dev_get_drvdata(dev); int err; u32 value; + u32 pwm_mode; /* Can't adjust fan when the card is off */ if ((adev->flags & AMD_IS_PX) && (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON)) return -EINVAL; + pwm_mode = amdgpu_dpm_get_fan_control_mode(adev); + if (pwm_mode != AMD_FAN_CTRL_MANUAL) { + pr_info("manual fan speed control should be enabled first\n"); + return -EINVAL; + } + err = kstrtou32(buf, 10, &value); if (err) return err; @@ -1111,6 +1194,148 @@ static ssize_t amdgpu_hwmon_get_fan1_input(struct device *dev, return sprintf(buf, "%i\n", speed); } +static ssize_t amdgpu_hwmon_get_fan1_min(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + u32 min_rpm = 0; + u32 size = sizeof(min_rpm); + int r; + + if (!adev->powerplay.pp_funcs->read_sensor) + return -EINVAL; + + r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_MIN_FAN_RPM, + (void *)&min_rpm, &size); + if (r) + return r; + + return snprintf(buf, PAGE_SIZE, "%d\n", min_rpm); +} + +static ssize_t amdgpu_hwmon_get_fan1_max(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + u32 max_rpm = 0; + u32 size = sizeof(max_rpm); + int r; + + if (!adev->powerplay.pp_funcs->read_sensor) + return -EINVAL; + + r = amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_MAX_FAN_RPM, + (void *)&max_rpm, &size); + if (r) + return r; + + return snprintf(buf, PAGE_SIZE, "%d\n", max_rpm); +} + +static ssize_t amdgpu_hwmon_get_fan1_target(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + int err; + u32 rpm = 0; + + /* Can't adjust fan when the card is off */ + if ((adev->flags & AMD_IS_PX) && + (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON)) + return -EINVAL; + + if (adev->powerplay.pp_funcs->get_fan_speed_rpm) { + err = amdgpu_dpm_get_fan_speed_rpm(adev, &rpm); + if (err) + return err; + } + + return sprintf(buf, "%i\n", rpm); +} + +static ssize_t amdgpu_hwmon_set_fan1_target(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + int err; + u32 value; + u32 pwm_mode; + + pwm_mode = amdgpu_dpm_get_fan_control_mode(adev); + if (pwm_mode != AMD_FAN_CTRL_MANUAL) + return -ENODATA; + + /* Can't adjust fan when the card is off */ + if ((adev->flags & AMD_IS_PX) && + (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON)) + return -EINVAL; + + err = kstrtou32(buf, 10, &value); + if (err) + return err; + + if (adev->powerplay.pp_funcs->set_fan_speed_rpm) { + err = amdgpu_dpm_set_fan_speed_rpm(adev, value); + if (err) + return err; + } + + return count; +} + +static ssize_t amdgpu_hwmon_get_fan1_enable(struct device *dev, + struct device_attribute *attr, + char *buf) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + u32 pwm_mode = 0; + + if (!adev->powerplay.pp_funcs->get_fan_control_mode) + return -EINVAL; + + pwm_mode = amdgpu_dpm_get_fan_control_mode(adev); + + return sprintf(buf, "%i\n", pwm_mode == AMD_FAN_CTRL_AUTO ? 0 : 1); +} + +static ssize_t amdgpu_hwmon_set_fan1_enable(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + struct amdgpu_device *adev = dev_get_drvdata(dev); + int err; + int value; + u32 pwm_mode; + + /* Can't adjust fan when the card is off */ + if ((adev->flags & AMD_IS_PX) && + (adev->ddev->switch_power_state != DRM_SWITCH_POWER_ON)) + return -EINVAL; + + if (!adev->powerplay.pp_funcs->set_fan_control_mode) + return -EINVAL; + + err = kstrtoint(buf, 10, &value); + if (err) + return err; + + if (value == 0) + pwm_mode = AMD_FAN_CTRL_AUTO; + else if (value == 1) + pwm_mode = AMD_FAN_CTRL_MANUAL; + else + return -EINVAL; + + amdgpu_dpm_set_fan_control_mode(adev, pwm_mode); + + return count; +} + static ssize_t amdgpu_hwmon_show_vddgfx(struct device *dev, struct device_attribute *attr, char *buf) @@ -1156,7 +1381,7 @@ static ssize_t amdgpu_hwmon_show_vddnb(struct device *dev, int r, size = sizeof(vddnb); /* only APUs have vddnb */ - if (adev->flags & AMD_IS_APU) + if (!(adev->flags & AMD_IS_APU)) return -EINVAL; /* Can't get voltage when the card is off */ @@ -1285,37 +1510,61 @@ static ssize_t amdgpu_hwmon_set_power_cap(struct device *dev, * DOC: hwmon * * The amdgpu driver exposes the following sensor interfaces: + * * - GPU temperature (via the on-die sensor) + * * - GPU voltage + * * - Northbridge voltage (APUs only) + * * - GPU power + * * - GPU fan * * hwmon interfaces for GPU temperature: + * * - temp1_input: the on die GPU temperature in millidegrees Celsius + * * - temp1_crit: temperature critical max value in millidegrees Celsius + * * - temp1_crit_hyst: temperature hysteresis for critical limit in millidegrees Celsius * * hwmon interfaces for GPU voltage: + * * - in0_input: the voltage on the GPU in millivolts + * * - in1_input: the voltage on the Northbridge in millivolts * * hwmon interfaces for GPU power: + * * - power1_average: average power used by the GPU in microWatts + * * - power1_cap_min: minimum cap supported in microWatts + * * - power1_cap_max: maximum cap supported in microWatts + * * - power1_cap: selected power cap in microWatts * * hwmon interfaces for GPU fan: + * * - pwm1: pulse width modulation fan level (0-255) - * - pwm1_enable: pulse width modulation fan control method - * 0: no fan speed control - * 1: manual fan speed control using pwm interface - * 2: automatic fan speed control + * + * - pwm1_enable: pulse width modulation fan control method (0: no fan speed control, 1: manual fan speed control using pwm interface, 2: automatic fan speed control) + * * - pwm1_min: pulse width modulation fan control minimum level (0) + * * - pwm1_max: pulse width modulation fan control maximum level (255) + * + * - fan1_min: an minimum value Unit: revolution/min (RPM) + * + * - fan1_max: an maxmum value Unit: revolution/max (RPM) + * * - fan1_input: fan speed in RPM * + * - fan[1-*]_target: Desired fan speed Unit: revolution/min (RPM) + * + * - fan[1-*]_enable: Enable or disable the sensors.1: Enable 0: Disable + * * You can use hwmon tools like sensors to view this information on your system. * */ @@ -1328,6 +1577,10 @@ static SENSOR_DEVICE_ATTR(pwm1_enable, S_IRUGO | S_IWUSR, amdgpu_hwmon_get_pwm1_ static SENSOR_DEVICE_ATTR(pwm1_min, S_IRUGO, amdgpu_hwmon_get_pwm1_min, NULL, 0); static SENSOR_DEVICE_ATTR(pwm1_max, S_IRUGO, amdgpu_hwmon_get_pwm1_max, NULL, 0); static SENSOR_DEVICE_ATTR(fan1_input, S_IRUGO, amdgpu_hwmon_get_fan1_input, NULL, 0); +static SENSOR_DEVICE_ATTR(fan1_min, S_IRUGO, amdgpu_hwmon_get_fan1_min, NULL, 0); +static SENSOR_DEVICE_ATTR(fan1_max, S_IRUGO, amdgpu_hwmon_get_fan1_max, NULL, 0); +static SENSOR_DEVICE_ATTR(fan1_target, S_IRUGO | S_IWUSR, amdgpu_hwmon_get_fan1_target, amdgpu_hwmon_set_fan1_target, 0); +static SENSOR_DEVICE_ATTR(fan1_enable, S_IRUGO | S_IWUSR, amdgpu_hwmon_get_fan1_enable, amdgpu_hwmon_set_fan1_enable, 0); static SENSOR_DEVICE_ATTR(in0_input, S_IRUGO, amdgpu_hwmon_show_vddgfx, NULL, 0); static SENSOR_DEVICE_ATTR(in0_label, S_IRUGO, amdgpu_hwmon_show_vddgfx_label, NULL, 0); static SENSOR_DEVICE_ATTR(in1_input, S_IRUGO, amdgpu_hwmon_show_vddnb, NULL, 0); @@ -1346,6 +1599,10 @@ static struct attribute *hwmon_attributes[] = { &sensor_dev_attr_pwm1_min.dev_attr.attr, &sensor_dev_attr_pwm1_max.dev_attr.attr, &sensor_dev_attr_fan1_input.dev_attr.attr, + &sensor_dev_attr_fan1_min.dev_attr.attr, + &sensor_dev_attr_fan1_max.dev_attr.attr, + &sensor_dev_attr_fan1_target.dev_attr.attr, + &sensor_dev_attr_fan1_enable.dev_attr.attr, &sensor_dev_attr_in0_input.dev_attr.attr, &sensor_dev_attr_in0_label.dev_attr.attr, &sensor_dev_attr_in1_input.dev_attr.attr, @@ -1364,13 +1621,16 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, struct amdgpu_device *adev = dev_get_drvdata(dev); umode_t effective_mode = attr->mode; - /* Skip fan attributes if fan is not present */ if (adev->pm.no_fan && (attr == &sensor_dev_attr_pwm1.dev_attr.attr || attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr || attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || attr == &sensor_dev_attr_pwm1_min.dev_attr.attr || - attr == &sensor_dev_attr_fan1_input.dev_attr.attr)) + attr == &sensor_dev_attr_fan1_input.dev_attr.attr || + attr == &sensor_dev_attr_fan1_min.dev_attr.attr || + attr == &sensor_dev_attr_fan1_max.dev_attr.attr || + attr == &sensor_dev_attr_fan1_target.dev_attr.attr || + attr == &sensor_dev_attr_fan1_enable.dev_attr.attr)) return 0; /* Skip limit attributes if DPM is not enabled */ @@ -1380,7 +1640,12 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, attr == &sensor_dev_attr_pwm1.dev_attr.attr || attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr || attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || - attr == &sensor_dev_attr_pwm1_min.dev_attr.attr)) + attr == &sensor_dev_attr_pwm1_min.dev_attr.attr || + attr == &sensor_dev_attr_fan1_input.dev_attr.attr || + attr == &sensor_dev_attr_fan1_min.dev_attr.attr || + attr == &sensor_dev_attr_fan1_max.dev_attr.attr || + attr == &sensor_dev_attr_fan1_target.dev_attr.attr || + attr == &sensor_dev_attr_fan1_enable.dev_attr.attr)) return 0; /* mask fan attributes if we have no bindings for this asic to expose */ @@ -1405,10 +1670,18 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, /* hide max/min values if we can't both query and manage the fan */ if ((!adev->powerplay.pp_funcs->set_fan_speed_percent && !adev->powerplay.pp_funcs->get_fan_speed_percent) && + (!adev->powerplay.pp_funcs->set_fan_speed_rpm && + !adev->powerplay.pp_funcs->get_fan_speed_rpm) && (attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || attr == &sensor_dev_attr_pwm1_min.dev_attr.attr)) return 0; + if ((!adev->powerplay.pp_funcs->set_fan_speed_rpm && + !adev->powerplay.pp_funcs->get_fan_speed_rpm) && + (attr == &sensor_dev_attr_fan1_max.dev_attr.attr || + attr == &sensor_dev_attr_fan1_min.dev_attr.attr)) + return 0; + /* only APUs have vddnb */ if (!(adev->flags & AMD_IS_APU) && (attr == &sensor_dev_attr_in1_input.dev_attr.attr || @@ -1668,56 +1941,21 @@ static void amdgpu_dpm_change_power_state_locked(struct amdgpu_device *adev) void amdgpu_dpm_enable_uvd(struct amdgpu_device *adev, bool enable) { - if (adev->powerplay.pp_funcs->powergate_uvd) { + if (adev->powerplay.pp_funcs->set_powergating_by_smu) { /* enable/disable UVD */ mutex_lock(&adev->pm.mutex); - amdgpu_dpm_powergate_uvd(adev, !enable); + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_UVD, !enable); mutex_unlock(&adev->pm.mutex); - } else { - if (enable) { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.uvd_active = true; - adev->pm.dpm.state = POWER_STATE_TYPE_INTERNAL_UVD; - mutex_unlock(&adev->pm.mutex); - } else { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.uvd_active = false; - mutex_unlock(&adev->pm.mutex); - } - amdgpu_pm_compute_clocks(adev); } } void amdgpu_dpm_enable_vce(struct amdgpu_device *adev, bool enable) { - if (adev->powerplay.pp_funcs->powergate_vce) { + if (adev->powerplay.pp_funcs->set_powergating_by_smu) { /* enable/disable VCE */ mutex_lock(&adev->pm.mutex); - amdgpu_dpm_powergate_vce(adev, !enable); + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_VCE, !enable); mutex_unlock(&adev->pm.mutex); - } else { - if (enable) { - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.vce_active = true; - /* XXX select vce level based on ring/task */ - adev->pm.dpm.vce_level = AMD_VCE_LEVEL_AC_ALL; - mutex_unlock(&adev->pm.mutex); - amdgpu_device_ip_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_VCE, - AMD_CG_STATE_UNGATE); - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCE, - AMD_PG_STATE_UNGATE); - amdgpu_pm_compute_clocks(adev); - } else { - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCE, - AMD_PG_STATE_GATE); - amdgpu_device_ip_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_VCE, - AMD_CG_STATE_GATE); - mutex_lock(&adev->pm.mutex); - adev->pm.dpm.vce_active = false; - mutex_unlock(&adev->pm.mutex); - amdgpu_pm_compute_clocks(adev); - } - } } @@ -1825,6 +2063,13 @@ int amdgpu_pm_sysfs_init(struct amdgpu_device *adev) "pp_od_clk_voltage\n"); return ret; } + ret = device_create_file(adev->dev, + &dev_attr_gpu_busy_percent); + if (ret) { + DRM_ERROR("failed to create device file " + "gpu_busy_level\n"); + return ret; + } ret = amdgpu_debugfs_pm_init(adev); if (ret) { DRM_ERROR("Failed to register debugfs file for dpm!\n"); @@ -1860,6 +2105,7 @@ void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev) &dev_attr_pp_power_profile_mode); device_remove_file(adev->dev, &dev_attr_pp_od_clk_voltage); + device_remove_file(adev->dev, &dev_attr_gpu_busy_percent); } void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) @@ -1898,14 +2144,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) } else { mutex_lock(&adev->pm.mutex); amdgpu_dpm_get_active_displays(adev); - /* update battery/ac status */ - if (power_supply_is_system_supplied() > 0) - adev->pm.dpm.ac_power = true; - else - adev->pm.dpm.ac_power = false; - amdgpu_dpm_change_power_state_locked(adev); - mutex_unlock(&adev->pm.mutex); } } @@ -1918,6 +2157,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) static int amdgpu_debugfs_pm_info_pp(struct seq_file *m, struct amdgpu_device *adev) { uint32_t value; + uint64_t value64; uint32_t query = 0; int size; @@ -1956,6 +2196,10 @@ static int amdgpu_debugfs_pm_info_pp(struct seq_file *m, struct amdgpu_device *a seq_printf(m, "GPU Load: %u %%\n", value); seq_printf(m, "\n"); + /* SMC feature mask */ + if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_ENABLED_SMC_FEATURES_MASK, (void *)&value64, &size)) + seq_printf(m, "SMC Feature Mask: 0x%016llx\n", value64); + /* UVD clocks */ if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_UVD_POWER, (void *)&value, &size)) { if (!value) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 4683626b065f..e45e929aaab5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -23,15 +23,32 @@ * * Authors: Alex Deucher */ + +/** + * DOC: PRIME Buffer Sharing + * + * The following callback implementations are used for :ref:`sharing GEM buffer + * objects between different devices via PRIME <prime_buffer_sharing>`. + */ + #include <drm/drmP.h> #include "amdgpu.h" #include "amdgpu_display.h" +#include "amdgpu_gem.h" #include <drm/amdgpu_drm.h> #include <linux/dma-buf.h> static const struct dma_buf_ops amdgpu_dmabuf_ops; +/** + * amdgpu_gem_prime_get_sg_table - &drm_driver.gem_prime_get_sg_table + * implementation + * @obj: GEM buffer object (BO) + * + * Returns: + * A scatter/gather table for the pinned pages of the BO's memory. + */ struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj) { struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); @@ -40,6 +57,15 @@ struct sg_table *amdgpu_gem_prime_get_sg_table(struct drm_gem_object *obj) return drm_prime_pages_to_sg(bo->tbo.ttm->pages, npages); } +/** + * amdgpu_gem_prime_vmap - &dma_buf_ops.vmap implementation + * @obj: GEM BO + * + * Sets up an in-kernel virtual mapping of the BO's memory. + * + * Returns: + * The virtual address of the mapping or an error pointer. + */ void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj) { struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); @@ -53,6 +79,13 @@ void *amdgpu_gem_prime_vmap(struct drm_gem_object *obj) return bo->dma_buf_vmap.virtual; } +/** + * amdgpu_gem_prime_vunmap - &dma_buf_ops.vunmap implementation + * @obj: GEM BO + * @vaddr: Virtual address (unused) + * + * Tears down the in-kernel virtual mapping of the BO's memory. + */ void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) { struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); @@ -60,6 +93,17 @@ void amdgpu_gem_prime_vunmap(struct drm_gem_object *obj, void *vaddr) ttm_bo_kunmap(&bo->dma_buf_vmap); } +/** + * amdgpu_gem_prime_mmap - &drm_driver.gem_prime_mmap implementation + * @obj: GEM BO + * @vma: Virtual memory area + * + * Sets up a userspace mapping of the BO's memory in the given + * virtual memory area. + * + * Returns: + * 0 on success or a negative error code on failure. + */ int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma) { struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); @@ -94,6 +138,19 @@ int amdgpu_gem_prime_mmap(struct drm_gem_object *obj, struct vm_area_struct *vma return ret; } +/** + * amdgpu_gem_prime_import_sg_table - &drm_driver.gem_prime_import_sg_table + * implementation + * @dev: DRM device + * @attach: DMA-buf attachment + * @sg: Scatter/gather table + * + * Imports shared DMA buffer memory exported by another device. + * + * Returns: + * A new GEM BO of the given DRM device, representing the memory + * described by the given DMA-buf attachment and scatter/gather table. + */ struct drm_gem_object * amdgpu_gem_prime_import_sg_table(struct drm_device *dev, struct dma_buf_attachment *attach, @@ -132,8 +189,19 @@ error: return ERR_PTR(ret); } +/** + * amdgpu_gem_map_attach - &dma_buf_ops.attach implementation + * @dma_buf: Shared DMA buffer + * @attach: DMA-buf attachment + * + * Makes sure that the shared DMA buffer can be accessed by the target device. + * For now, simply pins it to the GTT domain, where it should be accessible by + * all DMA devices. + * + * Returns: + * 0 on success or a negative error code on failure. + */ static int amdgpu_gem_map_attach(struct dma_buf *dma_buf, - struct device *target_dev, struct dma_buf_attachment *attach) { struct drm_gem_object *obj = dma_buf->priv; @@ -141,7 +209,7 @@ static int amdgpu_gem_map_attach(struct dma_buf *dma_buf, struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); long r; - r = drm_gem_map_attach(dma_buf, target_dev, attach); + r = drm_gem_map_attach(dma_buf, attach); if (r) return r; @@ -165,7 +233,7 @@ static int amdgpu_gem_map_attach(struct dma_buf *dma_buf, } /* pin buffer into GTT */ - r = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT, NULL); + r = amdgpu_bo_pin(bo, AMDGPU_GEM_DOMAIN_GTT); if (r) goto error_unreserve; @@ -181,6 +249,14 @@ error_detach: return r; } +/** + * amdgpu_gem_map_detach - &dma_buf_ops.detach implementation + * @dma_buf: Shared DMA buffer + * @attach: DMA-buf attachment + * + * This is called when a shared DMA buffer no longer needs to be accessible by + * another device. For now, simply unpins the buffer from GTT. + */ static void amdgpu_gem_map_detach(struct dma_buf *dma_buf, struct dma_buf_attachment *attach) { @@ -202,6 +278,13 @@ error: drm_gem_map_detach(dma_buf, attach); } +/** + * amdgpu_gem_prime_res_obj - &drm_driver.gem_prime_res_obj implementation + * @obj: GEM BO + * + * Returns: + * The BO's reservation object. + */ struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *obj) { struct amdgpu_bo *bo = gem_to_amdgpu_bo(obj); @@ -209,6 +292,18 @@ struct reservation_object *amdgpu_gem_prime_res_obj(struct drm_gem_object *obj) return bo->tbo.resv; } +/** + * amdgpu_gem_begin_cpu_access - &dma_buf_ops.begin_cpu_access implementation + * @dma_buf: Shared DMA buffer + * @direction: Direction of DMA transfer + * + * This is called before CPU access to the shared DMA buffer's memory. If it's + * a read access, the buffer is moved to the GTT domain if possible, for optimal + * CPU read performance. + * + * Returns: + * 0 on success or a negative error code on failure. + */ static int amdgpu_gem_begin_cpu_access(struct dma_buf *dma_buf, enum dma_data_direction direction) { @@ -229,7 +324,7 @@ static int amdgpu_gem_begin_cpu_access(struct dma_buf *dma_buf, return ret; if (!bo->pin_count && (bo->allowed_domains & AMDGPU_GEM_DOMAIN_GTT)) { - amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_GTT); ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); } @@ -245,14 +340,24 @@ static const struct dma_buf_ops amdgpu_dmabuf_ops = { .release = drm_gem_dmabuf_release, .begin_cpu_access = amdgpu_gem_begin_cpu_access, .map = drm_gem_dmabuf_kmap, - .map_atomic = drm_gem_dmabuf_kmap_atomic, .unmap = drm_gem_dmabuf_kunmap, - .unmap_atomic = drm_gem_dmabuf_kunmap_atomic, .mmap = drm_gem_dmabuf_mmap, .vmap = drm_gem_dmabuf_vmap, .vunmap = drm_gem_dmabuf_vunmap, }; +/** + * amdgpu_gem_prime_export - &drm_driver.gem_prime_export implementation + * @dev: DRM device + * @gobj: GEM BO + * @flags: Flags such as DRM_CLOEXEC and DRM_RDWR. + * + * The main work is done by the &drm_gem_prime_export helper, which in turn + * uses &amdgpu_gem_prime_res_obj. + * + * Returns: + * Shared DMA buffer representing the GEM BO from the given device. + */ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev, struct drm_gem_object *gobj, int flags) @@ -273,6 +378,17 @@ struct dma_buf *amdgpu_gem_prime_export(struct drm_device *dev, return buf; } +/** + * amdgpu_gem_prime_import - &drm_driver.gem_prime_import implementation + * @dev: DRM device + * @dma_buf: Shared DMA buffer + * + * The main work is done by the &drm_gem_prime_import helper, which in turn + * uses &amdgpu_gem_prime_import_sg_table. + * + * Returns: + * GEM BO representing the shared DMA buffer for the given device. + */ struct drm_gem_object *amdgpu_gem_prime_import(struct drm_device *dev, struct dma_buf *dma_buf) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index 9f1a5bd39ae8..25d2f3e757f1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -31,6 +31,7 @@ #include "soc15_common.h" #include "psp_v3_1.h" #include "psp_v10_0.h" +#include "psp_v11_0.h" static void psp_set_funcs(struct amdgpu_device *adev); @@ -52,12 +53,14 @@ static int psp_sw_init(void *handle) switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: - case CHIP_VEGA20: psp_v3_1_set_psp_funcs(psp); break; case CHIP_RAVEN: psp_v10_0_set_psp_funcs(psp); break; + case CHIP_VEGA20: + psp_v11_0_set_psp_funcs(psp); + break; default: return -EINVAL; } @@ -131,6 +134,18 @@ psp_cmd_submit_buf(struct psp_context *psp, msleep(1); } + /* the status field must be 0 after FW is loaded */ + if (ucode && psp->cmd_buf_mem->resp.status) { + DRM_ERROR("failed loading with status (%d) and ucode id (%d)\n", + psp->cmd_buf_mem->resp.status, ucode->ucode_id); + return -EINVAL; + } + + if (ucode) { + ucode->tmr_mc_addr_lo = psp->cmd_buf_mem->resp.fw_addr_lo; + ucode->tmr_mc_addr_hi = psp->cmd_buf_mem->resp.fw_addr_hi; + } + return ret; } @@ -155,7 +170,7 @@ static int psp_tmr_init(struct psp_context *psp) * Note: this memory need be reserved till the driver * uninitializes. */ - ret = amdgpu_bo_create_kernel(psp->adev, 0x300000, 0x100000, + ret = amdgpu_bo_create_kernel(psp->adev, PSP_TMR_SIZE, 0x100000, AMDGPU_GEM_DOMAIN_VRAM, &psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf); @@ -171,7 +186,9 @@ static int psp_tmr_load(struct psp_context *psp) if (!cmd) return -ENOMEM; - psp_prep_tmr_cmd_buf(cmd, psp->tmr_mc_addr, 0x300000); + psp_prep_tmr_cmd_buf(cmd, psp->tmr_mc_addr, PSP_TMR_SIZE); + DRM_INFO("reserve 0x%x from 0x%llx for PSP TMR SIZE\n", + PSP_TMR_SIZE, psp->tmr_mc_addr); ret = psp_cmd_submit_buf(psp, NULL, cmd, psp->fence_buf_mc_addr, 1); @@ -435,8 +452,6 @@ static int psp_hw_fini(void *handle) if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) return 0; - amdgpu_ucode_fini_bo(adev); - psp_ring_destroy(psp, PSP_RING_TYPE__KM); amdgpu_bo_free_kernel(&psp->tmr_bo, &psp->tmr_mc_addr, &psp->tmr_buf); @@ -589,3 +604,12 @@ const struct amdgpu_ip_block_version psp_v10_0_ip_block = .rev = 0, .funcs = &psp_ip_funcs, }; + +const struct amdgpu_ip_block_version psp_v11_0_ip_block = +{ + .type = AMD_IP_BLOCK_TYPE_PSP, + .major = 11, + .minor = 0, + .rev = 0, + .funcs = &psp_ip_funcs, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h index 129209686848..8b8720e9c3f0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h @@ -32,8 +32,10 @@ #define PSP_CMD_BUFFER_SIZE 0x1000 #define PSP_ASD_SHARED_MEM_SIZE 0x4000 #define PSP_1_MEG 0x100000 +#define PSP_TMR_SIZE 0x400000 struct psp_context; +struct psp_xgmi_topology_info; enum psp_ring_type { @@ -63,18 +65,27 @@ struct psp_funcs int (*prep_cmd_buf)(struct amdgpu_firmware_info *ucode, struct psp_gfx_cmd_resp *cmd); int (*ring_init)(struct psp_context *psp, enum psp_ring_type ring_type); - int (*ring_create)(struct psp_context *psp, enum psp_ring_type ring_type); + int (*ring_create)(struct psp_context *psp, + enum psp_ring_type ring_type); int (*ring_stop)(struct psp_context *psp, enum psp_ring_type ring_type); int (*ring_destroy)(struct psp_context *psp, enum psp_ring_type ring_type); - int (*cmd_submit)(struct psp_context *psp, struct amdgpu_firmware_info *ucode, - uint64_t cmd_buf_mc_addr, uint64_t fence_mc_addr, int index); + int (*cmd_submit)(struct psp_context *psp, + struct amdgpu_firmware_info *ucode, + uint64_t cmd_buf_mc_addr, uint64_t fence_mc_addr, + int index); bool (*compare_sram_data)(struct psp_context *psp, struct amdgpu_firmware_info *ucode, enum AMDGPU_UCODE_ID ucode_type); bool (*smu_reload_quirk)(struct psp_context *psp); int (*mode1_reset)(struct psp_context *psp); + uint64_t (*xgmi_get_device_id)(struct psp_context *psp); + uint64_t (*xgmi_get_hive_id)(struct psp_context *psp); + int (*xgmi_get_topology_info)(struct psp_context *psp, int number_devices, + struct psp_xgmi_topology_info *topology); + int (*xgmi_set_topology_info)(struct psp_context *psp, int number_devices, + struct psp_xgmi_topology_info *topology); }; struct psp_context @@ -83,11 +94,11 @@ struct psp_context struct psp_ring km_ring; struct psp_gfx_cmd_resp *cmd; - const struct psp_funcs *funcs; + const struct psp_funcs *funcs; /* fence buffer */ - struct amdgpu_bo *fw_pri_bo; - uint64_t fw_pri_mc_addr; + struct amdgpu_bo *fw_pri_bo; + uint64_t fw_pri_mc_addr; void *fw_pri_buf; /* sos firmware */ @@ -100,8 +111,8 @@ struct psp_context uint8_t *sos_start_addr; /* tmr buffer */ - struct amdgpu_bo *tmr_bo; - uint64_t tmr_mc_addr; + struct amdgpu_bo *tmr_bo; + uint64_t tmr_mc_addr; void *tmr_buf; /* asd firmware and buffer */ @@ -110,13 +121,13 @@ struct psp_context uint32_t asd_feature_version; uint32_t asd_ucode_size; uint8_t *asd_start_addr; - struct amdgpu_bo *asd_shared_bo; - uint64_t asd_shared_mc_addr; + struct amdgpu_bo *asd_shared_bo; + uint64_t asd_shared_mc_addr; void *asd_shared_buf; /* fence buffer */ - struct amdgpu_bo *fence_buf_bo; - uint64_t fence_buf_mc_addr; + struct amdgpu_bo *fence_buf_bo; + uint64_t fence_buf_mc_addr; void *fence_buf; /* cmd buffer */ @@ -130,6 +141,23 @@ struct amdgpu_psp_funcs { enum AMDGPU_UCODE_ID); }; +struct psp_xgmi_topology_info { + /* Generated by PSP to identify the GPU instance within xgmi connection */ + uint64_t device_id; + /* + * If all bits set to 0 , driver indicates it wants to retrieve the xgmi + * connection vector topology, but not access enable the connections + * if some or all bits are set to 1, driver indicates it want to retrieve the + * current xgmi topology and access enable the link to GPU[i] associated + * with the bit position in the vector. + * On return,: bits indicated which xgmi links are present/active depending + * on the value passed in. The relative bit offset for the relative GPU index + * within the hive is always marked active. + */ + uint32_t connection_mask; + uint32_t reserved; /* must be 0 */ +}; + #define psp_prep_cmd_buf(ucode, type) (psp)->funcs->prep_cmd_buf((ucode), (type)) #define psp_ring_init(psp, type) (psp)->funcs->ring_init((psp), (type)) #define psp_ring_create(psp, type) (psp)->funcs->ring_create((psp), (type)) @@ -149,6 +177,18 @@ struct amdgpu_psp_funcs { ((psp)->funcs->smu_reload_quirk ? (psp)->funcs->smu_reload_quirk((psp)) : false) #define psp_mode1_reset(psp) \ ((psp)->funcs->mode1_reset ? (psp)->funcs->mode1_reset((psp)) : false) +#define psp_xgmi_get_device_id(psp) \ + ((psp)->funcs->xgmi_get_device_id ? (psp)->funcs->xgmi_get_device_id((psp)) : 0) +#define psp_xgmi_get_hive_id(psp) \ + ((psp)->funcs->xgmi_get_hive_id ? (psp)->funcs->xgmi_get_hive_id((psp)) : 0) +#define psp_xgmi_get_topology_info(psp, num_device, topology) \ + ((psp)->funcs->xgmi_get_topology_info ? \ + (psp)->funcs->xgmi_get_topology_info((psp), (num_device), (topology)) : -EINVAL) +#define psp_xgmi_set_topology_info(psp, num_device, topology) \ + ((psp)->funcs->xgmi_set_topology_info ? \ + (psp)->funcs->xgmi_set_topology_info((psp), (num_device), (topology)) : -EINVAL) + +#define amdgpu_psp_check_fw_loading_status(adev, i) (adev)->firmware.funcs->check_fw_loading_status((adev), (i)) extern const struct amd_ip_funcs psp_ip_funcs; @@ -159,5 +199,6 @@ extern int psp_wait_for(struct psp_context *psp, uint32_t reg_index, extern const struct amdgpu_ip_block_version psp_v10_0_ip_block; int psp_gpu_reset(struct amdgpu_device *adev); +extern const struct amdgpu_ip_block_version psp_v11_0_ip_block; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c deleted file mode 100644 index 8af16e81c7d4..000000000000 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c +++ /dev/null @@ -1,307 +0,0 @@ -/* - * Copyright 2017 Valve Corporation - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: Andres Rodriguez - */ - -#include "amdgpu.h" -#include "amdgpu_ring.h" - -static int amdgpu_queue_mapper_init(struct amdgpu_queue_mapper *mapper, - int hw_ip) -{ - if (!mapper) - return -EINVAL; - - if (hw_ip > AMDGPU_MAX_IP_NUM) - return -EINVAL; - - mapper->hw_ip = hw_ip; - mutex_init(&mapper->lock); - - memset(mapper->queue_map, 0, sizeof(mapper->queue_map)); - - return 0; -} - -static struct amdgpu_ring *amdgpu_get_cached_map(struct amdgpu_queue_mapper *mapper, - int ring) -{ - return mapper->queue_map[ring]; -} - -static int amdgpu_update_cached_map(struct amdgpu_queue_mapper *mapper, - int ring, struct amdgpu_ring *pring) -{ - if (WARN_ON(mapper->queue_map[ring])) { - DRM_ERROR("Un-expected ring re-map\n"); - return -EINVAL; - } - - mapper->queue_map[ring] = pring; - - return 0; -} - -static int amdgpu_identity_map(struct amdgpu_device *adev, - struct amdgpu_queue_mapper *mapper, - u32 ring, - struct amdgpu_ring **out_ring) -{ - u32 instance; - - switch (mapper->hw_ip) { - case AMDGPU_HW_IP_GFX: - *out_ring = &adev->gfx.gfx_ring[ring]; - break; - case AMDGPU_HW_IP_COMPUTE: - *out_ring = &adev->gfx.compute_ring[ring]; - break; - case AMDGPU_HW_IP_DMA: - *out_ring = &adev->sdma.instance[ring].ring; - break; - case AMDGPU_HW_IP_UVD: - instance = ring; - *out_ring = &adev->uvd.inst[instance].ring; - break; - case AMDGPU_HW_IP_VCE: - *out_ring = &adev->vce.ring[ring]; - break; - case AMDGPU_HW_IP_UVD_ENC: - instance = ring / adev->uvd.num_enc_rings; - *out_ring = - &adev->uvd.inst[instance].ring_enc[ring%adev->uvd.num_enc_rings]; - break; - case AMDGPU_HW_IP_VCN_DEC: - *out_ring = &adev->vcn.ring_dec; - break; - case AMDGPU_HW_IP_VCN_ENC: - *out_ring = &adev->vcn.ring_enc[ring]; - break; - default: - *out_ring = NULL; - DRM_ERROR("unknown HW IP type: %d\n", mapper->hw_ip); - return -EINVAL; - } - - return amdgpu_update_cached_map(mapper, ring, *out_ring); -} - -static enum amdgpu_ring_type amdgpu_hw_ip_to_ring_type(int hw_ip) -{ - switch (hw_ip) { - case AMDGPU_HW_IP_GFX: - return AMDGPU_RING_TYPE_GFX; - case AMDGPU_HW_IP_COMPUTE: - return AMDGPU_RING_TYPE_COMPUTE; - case AMDGPU_HW_IP_DMA: - return AMDGPU_RING_TYPE_SDMA; - case AMDGPU_HW_IP_UVD: - return AMDGPU_RING_TYPE_UVD; - case AMDGPU_HW_IP_VCE: - return AMDGPU_RING_TYPE_VCE; - default: - DRM_ERROR("Invalid HW IP specified %d\n", hw_ip); - return -1; - } -} - -static int amdgpu_lru_map(struct amdgpu_device *adev, - struct amdgpu_queue_mapper *mapper, - u32 user_ring, bool lru_pipe_order, - struct amdgpu_ring **out_ring) -{ - int r, i, j; - int ring_type = amdgpu_hw_ip_to_ring_type(mapper->hw_ip); - int ring_blacklist[AMDGPU_MAX_RINGS]; - struct amdgpu_ring *ring; - - /* 0 is a valid ring index, so initialize to -1 */ - memset(ring_blacklist, 0xff, sizeof(ring_blacklist)); - - for (i = 0, j = 0; i < AMDGPU_MAX_RINGS; i++) { - ring = mapper->queue_map[i]; - if (ring) - ring_blacklist[j++] = ring->idx; - } - - r = amdgpu_ring_lru_get(adev, ring_type, ring_blacklist, - j, lru_pipe_order, out_ring); - if (r) - return r; - - return amdgpu_update_cached_map(mapper, user_ring, *out_ring); -} - -/** - * amdgpu_queue_mgr_init - init an amdgpu_queue_mgr struct - * - * @adev: amdgpu_device pointer - * @mgr: amdgpu_queue_mgr structure holding queue information - * - * Initialize the the selected @mgr (all asics). - * - * Returns 0 on success, error on failure. - */ -int amdgpu_queue_mgr_init(struct amdgpu_device *adev, - struct amdgpu_queue_mgr *mgr) -{ - int i, r; - - if (!adev || !mgr) - return -EINVAL; - - memset(mgr, 0, sizeof(*mgr)); - - for (i = 0; i < AMDGPU_MAX_IP_NUM; ++i) { - r = amdgpu_queue_mapper_init(&mgr->mapper[i], i); - if (r) - return r; - } - - return 0; -} - -/** - * amdgpu_queue_mgr_fini - de-initialize an amdgpu_queue_mgr struct - * - * @adev: amdgpu_device pointer - * @mgr: amdgpu_queue_mgr structure holding queue information - * - * De-initialize the the selected @mgr (all asics). - * - * Returns 0 on success, error on failure. - */ -int amdgpu_queue_mgr_fini(struct amdgpu_device *adev, - struct amdgpu_queue_mgr *mgr) -{ - return 0; -} - -/** - * amdgpu_queue_mgr_map - Map a userspace ring id to an amdgpu_ring - * - * @adev: amdgpu_device pointer - * @mgr: amdgpu_queue_mgr structure holding queue information - * @hw_ip: HW IP enum - * @instance: HW instance - * @ring: user ring id - * @our_ring: pointer to mapped amdgpu_ring - * - * Map a userspace ring id to an appropriate kernel ring. Different - * policies are configurable at a HW IP level. - * - * Returns 0 on success, error on failure. - */ -int amdgpu_queue_mgr_map(struct amdgpu_device *adev, - struct amdgpu_queue_mgr *mgr, - u32 hw_ip, u32 instance, u32 ring, - struct amdgpu_ring **out_ring) -{ - int r, ip_num_rings; - struct amdgpu_queue_mapper *mapper = &mgr->mapper[hw_ip]; - - if (!adev || !mgr || !out_ring) - return -EINVAL; - - if (hw_ip >= AMDGPU_MAX_IP_NUM) - return -EINVAL; - - if (ring >= AMDGPU_MAX_RINGS) - return -EINVAL; - - /* Right now all IPs have only one instance - multiple rings. */ - if (instance != 0) { - DRM_DEBUG("invalid ip instance: %d\n", instance); - return -EINVAL; - } - - switch (hw_ip) { - case AMDGPU_HW_IP_GFX: - ip_num_rings = adev->gfx.num_gfx_rings; - break; - case AMDGPU_HW_IP_COMPUTE: - ip_num_rings = adev->gfx.num_compute_rings; - break; - case AMDGPU_HW_IP_DMA: - ip_num_rings = adev->sdma.num_instances; - break; - case AMDGPU_HW_IP_UVD: - ip_num_rings = adev->uvd.num_uvd_inst; - break; - case AMDGPU_HW_IP_VCE: - ip_num_rings = adev->vce.num_rings; - break; - case AMDGPU_HW_IP_UVD_ENC: - ip_num_rings = - adev->uvd.num_enc_rings * adev->uvd.num_uvd_inst; - break; - case AMDGPU_HW_IP_VCN_DEC: - ip_num_rings = 1; - break; - case AMDGPU_HW_IP_VCN_ENC: - ip_num_rings = adev->vcn.num_enc_rings; - break; - default: - DRM_DEBUG("unknown ip type: %d\n", hw_ip); - return -EINVAL; - } - - if (ring >= ip_num_rings) { - DRM_DEBUG("Ring index:%d exceeds maximum:%d for ip:%d\n", - ring, ip_num_rings, hw_ip); - return -EINVAL; - } - - mutex_lock(&mapper->lock); - - *out_ring = amdgpu_get_cached_map(mapper, ring); - if (*out_ring) { - /* cache hit */ - r = 0; - goto out_unlock; - } - - switch (mapper->hw_ip) { - case AMDGPU_HW_IP_GFX: - case AMDGPU_HW_IP_UVD: - case AMDGPU_HW_IP_VCE: - case AMDGPU_HW_IP_UVD_ENC: - case AMDGPU_HW_IP_VCN_DEC: - case AMDGPU_HW_IP_VCN_ENC: - r = amdgpu_identity_map(adev, mapper, ring, out_ring); - break; - case AMDGPU_HW_IP_DMA: - r = amdgpu_lru_map(adev, mapper, ring, false, out_ring); - break; - case AMDGPU_HW_IP_COMPUTE: - r = amdgpu_lru_map(adev, mapper, ring, true, out_ring); - break; - default: - *out_ring = NULL; - r = -EINVAL; - DRM_DEBUG("unknown HW IP type: %d\n", mapper->hw_ip); - } - -out_unlock: - mutex_unlock(&mapper->lock); - return r; -} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index c6850b629d0e..b70e85ec147d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -135,9 +135,6 @@ void amdgpu_ring_commit(struct amdgpu_ring *ring) if (ring->funcs->end_use) ring->funcs->end_use(ring); - - if (ring->funcs->type != AMDGPU_RING_TYPE_KIQ) - amdgpu_ring_lru_touch(ring->adev, ring); } /** @@ -211,7 +208,8 @@ void amdgpu_ring_priority_get(struct amdgpu_ring *ring, if (!ring->funcs->set_priority) return; - atomic_inc(&ring->num_jobs[priority]); + if (atomic_inc_return(&ring->num_jobs[priority]) <= 0) + return; mutex_lock(&ring->priority_mutex); if (priority <= ring->priority) @@ -304,7 +302,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, 0xffffffffffffffff : ring->buf_mask; /* Allocate ring buffer */ if (ring->ring_obj == NULL) { - r = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, + r = amdgpu_bo_create_kernel(adev, ring->ring_size + ring->funcs->extra_dw, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, &ring->ring_obj, &ring->gpu_addr, @@ -319,8 +317,6 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, ring->max_dw = max_dw; ring->priority = DRM_SCHED_PRIORITY_NORMAL; mutex_init(&ring->priority_mutex); - INIT_LIST_HEAD(&ring->lru_list); - amdgpu_ring_lru_touch(adev, ring); for (i = 0; i < DRM_SCHED_PRIORITY_MAX; ++i) atomic_set(&ring->num_jobs[i], 0); @@ -367,99 +363,6 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) ring->adev->rings[ring->idx] = NULL; } -static void amdgpu_ring_lru_touch_locked(struct amdgpu_device *adev, - struct amdgpu_ring *ring) -{ - /* list_move_tail handles the case where ring isn't part of the list */ - list_move_tail(&ring->lru_list, &adev->ring_lru_list); -} - -static bool amdgpu_ring_is_blacklisted(struct amdgpu_ring *ring, - int *blacklist, int num_blacklist) -{ - int i; - - for (i = 0; i < num_blacklist; i++) { - if (ring->idx == blacklist[i]) - return true; - } - - return false; -} - -/** - * amdgpu_ring_lru_get - get the least recently used ring for a HW IP block - * - * @adev: amdgpu_device pointer - * @type: amdgpu_ring_type enum - * @blacklist: blacklisted ring ids array - * @num_blacklist: number of entries in @blacklist - * @lru_pipe_order: find a ring from the least recently used pipe - * @ring: output ring - * - * Retrieve the amdgpu_ring structure for the least recently used ring of - * a specific IP block (all asics). - * Returns 0 on success, error on failure. - */ -int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, - int *blacklist, int num_blacklist, - bool lru_pipe_order, struct amdgpu_ring **ring) -{ - struct amdgpu_ring *entry; - - /* List is sorted in LRU order, find first entry corresponding - * to the desired HW IP */ - *ring = NULL; - spin_lock(&adev->ring_lru_list_lock); - list_for_each_entry(entry, &adev->ring_lru_list, lru_list) { - if (entry->funcs->type != type) - continue; - - if (amdgpu_ring_is_blacklisted(entry, blacklist, num_blacklist)) - continue; - - if (!*ring) { - *ring = entry; - - /* We are done for ring LRU */ - if (!lru_pipe_order) - break; - } - - /* Move all rings on the same pipe to the end of the list */ - if (entry->pipe == (*ring)->pipe) - amdgpu_ring_lru_touch_locked(adev, entry); - } - - /* Move the ring we found to the end of the list */ - if (*ring) - amdgpu_ring_lru_touch_locked(adev, *ring); - - spin_unlock(&adev->ring_lru_list_lock); - - if (!*ring) { - DRM_ERROR("Ring LRU contains no entries for ring type:%d\n", type); - return -EINVAL; - } - - return 0; -} - -/** - * amdgpu_ring_lru_touch - mark a ring as recently being used - * - * @adev: amdgpu_device pointer - * @ring: ring to touch - * - * Move @ring to the tail of the lru list - */ -void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring) -{ - spin_lock(&adev->ring_lru_list_lock); - amdgpu_ring_lru_touch_locked(adev, ring); - spin_unlock(&adev->ring_lru_list_lock); -} - /** * amdgpu_ring_emit_reg_write_reg_wait_helper - ring helper * @@ -480,6 +383,31 @@ void amdgpu_ring_emit_reg_write_reg_wait_helper(struct amdgpu_ring *ring, amdgpu_ring_emit_reg_wait(ring, reg1, mask, mask); } +/** + * amdgpu_ring_soft_recovery - try to soft recover a ring lockup + * + * @ring: ring to try the recovery on + * @vmid: VMID we try to get going again + * @fence: timedout fence + * + * Tries to get a ring proceeding again when it is stuck. + */ +bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid, + struct dma_fence *fence) +{ + ktime_t deadline = ktime_add_us(ktime_get(), 10000); + + if (!ring->funcs->soft_recovery) + return false; + + atomic_inc(&ring->adev->gpu_reset_counter); + while (!dma_fence_is_signaled(fence) && + ktime_to_ns(ktime_sub(deadline, ktime_get())) > 0) + ring->funcs->soft_recovery(ring, vmid); + + return dma_fence_is_signaled(fence); +} + /* * Debugfs info */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 1513124c5659..4caa301ce454 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -44,6 +44,8 @@ #define AMDGPU_FENCE_FLAG_INT (1 << 1) #define AMDGPU_FENCE_FLAG_TC_WB_ONLY (1 << 2) +#define to_amdgpu_ring(s) container_of((s), struct amdgpu_ring, sched) + enum amdgpu_ring_type { AMDGPU_RING_TYPE_GFX, AMDGPU_RING_TYPE_COMPUTE, @@ -53,7 +55,8 @@ enum amdgpu_ring_type { AMDGPU_RING_TYPE_KIQ, AMDGPU_RING_TYPE_UVD_ENC, AMDGPU_RING_TYPE_VCN_DEC, - AMDGPU_RING_TYPE_VCN_ENC + AMDGPU_RING_TYPE_VCN_ENC, + AMDGPU_RING_TYPE_VCN_JPEG }; struct amdgpu_device; @@ -94,7 +97,7 @@ void amdgpu_fence_driver_resume(struct amdgpu_device *adev); int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence, unsigned flags); int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s); -void amdgpu_fence_process(struct amdgpu_ring *ring); +bool amdgpu_fence_process(struct amdgpu_ring *ring); int amdgpu_fence_wait_empty(struct amdgpu_ring *ring); signed long amdgpu_fence_wait_polling(struct amdgpu_ring *ring, uint32_t wait_seq, @@ -112,6 +115,7 @@ struct amdgpu_ring_funcs { u32 nop; bool support_64bit_ptrs; unsigned vmhub; + unsigned extra_dw; /* ring read/write ptr handling */ u64 (*get_rptr)(struct amdgpu_ring *ring); @@ -119,6 +123,7 @@ struct amdgpu_ring_funcs { void (*set_wptr)(struct amdgpu_ring *ring); /* validating and patching of IBs */ int (*parse_cs)(struct amdgpu_cs_parser *p, uint32_t ib_idx); + int (*patch_cs_in_place)(struct amdgpu_cs_parser *p, uint32_t ib_idx); /* constants to calculate how many DW are needed for an emit */ unsigned emit_frame_size; unsigned emit_ib_size; @@ -163,6 +168,8 @@ struct amdgpu_ring_funcs { /* priority functions */ void (*set_priority) (struct amdgpu_ring *ring, enum drm_sched_priority priority); + /* Try to soft recover the ring to make the fence signal */ + void (*soft_recovery)(struct amdgpu_ring *ring, unsigned vmid); }; struct amdgpu_ring { @@ -170,7 +177,6 @@ struct amdgpu_ring { const struct amdgpu_ring_funcs *funcs; struct amdgpu_fence_driver fence_drv; struct drm_gpu_scheduler sched; - struct list_head lru_list; struct amdgpu_bo *ring_obj; volatile uint32_t *ring; @@ -216,6 +222,30 @@ struct amdgpu_ring { #endif }; +#define amdgpu_ring_parse_cs(r, p, ib) ((r)->funcs->parse_cs((p), (ib))) +#define amdgpu_ring_patch_cs_in_place(r, p, ib) ((r)->funcs->patch_cs_in_place((p), (ib))) +#define amdgpu_ring_test_ring(r) (r)->funcs->test_ring((r)) +#define amdgpu_ring_test_ib(r, t) (r)->funcs->test_ib((r), (t)) +#define amdgpu_ring_get_rptr(r) (r)->funcs->get_rptr((r)) +#define amdgpu_ring_get_wptr(r) (r)->funcs->get_wptr((r)) +#define amdgpu_ring_set_wptr(r) (r)->funcs->set_wptr((r)) +#define amdgpu_ring_emit_ib(r, ib, vmid, c) (r)->funcs->emit_ib((r), (ib), (vmid), (c)) +#define amdgpu_ring_emit_pipeline_sync(r) (r)->funcs->emit_pipeline_sync((r)) +#define amdgpu_ring_emit_vm_flush(r, vmid, addr) (r)->funcs->emit_vm_flush((r), (vmid), (addr)) +#define amdgpu_ring_emit_fence(r, addr, seq, flags) (r)->funcs->emit_fence((r), (addr), (seq), (flags)) +#define amdgpu_ring_emit_gds_switch(r, v, db, ds, wb, ws, ab, as) (r)->funcs->emit_gds_switch((r), (v), (db), (ds), (wb), (ws), (ab), (as)) +#define amdgpu_ring_emit_hdp_flush(r) (r)->funcs->emit_hdp_flush((r)) +#define amdgpu_ring_emit_switch_buffer(r) (r)->funcs->emit_switch_buffer((r)) +#define amdgpu_ring_emit_cntxcntl(r, d) (r)->funcs->emit_cntxcntl((r), (d)) +#define amdgpu_ring_emit_rreg(r, d) (r)->funcs->emit_rreg((r), (d)) +#define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v)) +#define amdgpu_ring_emit_reg_wait(r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m)) +#define amdgpu_ring_emit_reg_write_reg_wait(r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) +#define amdgpu_ring_emit_tmz(r, b) (r)->funcs->emit_tmz((r), (b)) +#define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib))) +#define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r)) +#define amdgpu_ring_patch_cond_exec(r,o) (r)->funcs->patch_cond_exec((r),(o)) + int amdgpu_ring_alloc(struct amdgpu_ring *ring, unsigned ndw); void amdgpu_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count); void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib); @@ -229,13 +259,11 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, unsigned ring_size, struct amdgpu_irq_src *irq_src, unsigned irq_type); void amdgpu_ring_fini(struct amdgpu_ring *ring); -int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, - int *blacklist, int num_blacklist, - bool lru_pipe_order, struct amdgpu_ring **ring); -void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring); void amdgpu_ring_emit_reg_write_reg_wait_helper(struct amdgpu_ring *ring, uint32_t reg0, uint32_t val0, uint32_t reg1, uint32_t val1); +bool amdgpu_ring_soft_recovery(struct amdgpu_ring *ring, unsigned int vmid, + struct dma_fence *fence); static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c index fb1667b35daa..12f2bf97611f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c @@ -226,6 +226,8 @@ static bool amdgpu_sa_bo_next_hole(struct amdgpu_sa_manager *sa_manager, for (i = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i) { struct amdgpu_sa_bo *sa_bo; + fences[i] = NULL; + if (list_empty(&sa_manager->flist[i])) continue; @@ -296,10 +298,8 @@ int amdgpu_sa_bo_new(struct amdgpu_sa_manager *sa_manager, spin_lock(&sa_manager->wq.lock); do { - for (i = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i) { - fences[i] = NULL; + for (i = 0; i < AMDGPU_SA_NUM_FENCE_LISTS; ++i) tries[i] = 0; - } do { amdgpu_sa_bo_try_free(sa_manager); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c index 86a0715d9431..1cafe8d83a4d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c @@ -53,9 +53,8 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev, int fd, enum drm_sched_priority priority) { - struct file *filp = fcheck(fd); + struct file *filp = fget(fd); struct drm_file *file; - struct pid *pid; struct amdgpu_fpriv *fpriv; struct amdgpu_ctx *ctx; uint32_t id; @@ -63,20 +62,12 @@ static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev, if (!filp) return -EINVAL; - pid = get_pid(((struct drm_file *)filp->private_data)->pid); + file = filp->private_data; + fpriv = file->driver_priv; + idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id) + amdgpu_ctx_priority_override(ctx, priority); - mutex_lock(&adev->ddev->filelist_mutex); - list_for_each_entry(file, &adev->ddev->filelist, lhead) { - if (file->pid != pid) - continue; - - fpriv = file->driver_priv; - idr_for_each_entry(&fpriv->ctx_mgr.ctx_handles, ctx, id) - amdgpu_ctx_priority_override(ctx, priority); - } - mutex_unlock(&adev->ddev->filelist_mutex); - - put_pid(pid); + fput(filp); return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c new file mode 100644 index 000000000000..bc9244b429ef --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.c @@ -0,0 +1,44 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include <drm/drmP.h> +#include "amdgpu.h" +#include "amdgpu_sdma.h" + +/* + * GPU SDMA IP block helpers function. + */ + +struct amdgpu_sdma_instance * amdgpu_get_sdma_instance(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + int i; + + for (i = 0; i < adev->sdma.num_instances; i++) + if (&adev->sdma.instance[i].ring == ring) + break; + + if (i < AMDGPU_MAX_SDMA_INSTANCES) + return &adev->sdma.instance[i]; + else + return NULL; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h new file mode 100644 index 000000000000..500113ec65ca --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sdma.h @@ -0,0 +1,97 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __AMDGPU_SDMA_H__ +#define __AMDGPU_SDMA_H__ + +/* max number of IP instances */ +#define AMDGPU_MAX_SDMA_INSTANCES 2 + +enum amdgpu_sdma_irq { + AMDGPU_SDMA_IRQ_TRAP0 = 0, + AMDGPU_SDMA_IRQ_TRAP1, + + AMDGPU_SDMA_IRQ_LAST +}; + +struct amdgpu_sdma_instance { + /* SDMA firmware */ + const struct firmware *fw; + uint32_t fw_version; + uint32_t feature_version; + + struct amdgpu_ring ring; + bool burst_nop; +}; + +struct amdgpu_sdma { + struct amdgpu_sdma_instance instance[AMDGPU_MAX_SDMA_INSTANCES]; + struct amdgpu_irq_src trap_irq; + struct amdgpu_irq_src illegal_inst_irq; + int num_instances; + uint32_t srbm_soft_reset; +}; + +/* + * Provided by hw blocks that can move/clear data. e.g., gfx or sdma + * But currently, we use sdma to move data. + */ +struct amdgpu_buffer_funcs { + /* maximum bytes in a single operation */ + uint32_t copy_max_bytes; + + /* number of dw to reserve per operation */ + unsigned copy_num_dw; + + /* used for buffer migration */ + void (*emit_copy_buffer)(struct amdgpu_ib *ib, + /* src addr in bytes */ + uint64_t src_offset, + /* dst addr in bytes */ + uint64_t dst_offset, + /* number of byte to transfer */ + uint32_t byte_count); + + /* maximum bytes in a single operation */ + uint32_t fill_max_bytes; + + /* number of dw to reserve per operation */ + unsigned fill_num_dw; + + /* used for buffer clearing */ + void (*emit_fill_buffer)(struct amdgpu_ib *ib, + /* value to write to memory */ + uint32_t src_data, + /* dst addr in bytes */ + uint64_t dst_offset, + /* number of byte to fill */ + uint32_t byte_count); +}; + +#define amdgpu_emit_copy_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_copy_buffer((ib), (s), (d), (b)) +#define amdgpu_emit_fill_buffer(adev, ib, s, d, b) (adev)->mman.buffer_funcs->emit_fill_buffer((ib), (s), (d), (b)) + +struct amdgpu_sdma_instance * +amdgpu_get_sdma_instance(struct amdgpu_ring *ring); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index e3878256743a..8904e62dca7a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -1,3 +1,4 @@ +// SPDX-License-Identifier: GPL-2.0 OR MIT /* * Copyright 2009 VMware, Inc. * @@ -75,11 +76,12 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) r = amdgpu_bo_reserve(vram_obj, false); if (unlikely(r != 0)) goto out_unref; - r = amdgpu_bo_pin(vram_obj, AMDGPU_GEM_DOMAIN_VRAM, &vram_addr); + r = amdgpu_bo_pin(vram_obj, AMDGPU_GEM_DOMAIN_VRAM); if (r) { DRM_ERROR("Failed to pin VRAM object\n"); goto out_unres; } + vram_addr = amdgpu_bo_gpu_offset(vram_obj); for (i = 0; i < n; i++) { void *gtt_map, *vram_map; void **gart_start, **gart_end; @@ -96,11 +98,17 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) r = amdgpu_bo_reserve(gtt_obj[i], false); if (unlikely(r != 0)) goto out_lclean_unref; - r = amdgpu_bo_pin(gtt_obj[i], AMDGPU_GEM_DOMAIN_GTT, &gart_addr); + r = amdgpu_bo_pin(gtt_obj[i], AMDGPU_GEM_DOMAIN_GTT); if (r) { DRM_ERROR("Failed to pin GTT object %d\n", i); goto out_lclean_unres; } + r = amdgpu_ttm_alloc_gart(>t_obj[i]->tbo); + if (r) { + DRM_ERROR("%p bind failed\n", gtt_obj[i]); + goto out_lclean_unpin; + } + gart_addr = amdgpu_bo_gpu_offset(gtt_obj[i]); r = amdgpu_bo_kmap(gtt_obj[i], >t_map); if (r) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index e96e26d3f3b0..e9bf70e2ac51 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -103,7 +103,7 @@ TRACE_EVENT(amdgpu_iv, __entry->src_data[2] = iv->src_data[2]; __entry->src_data[3] = iv->src_data[3]; ), - TP_printk("client_id:%u src_id:%u ring:%u vmid:%u timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x\n", + TP_printk("client_id:%u src_id:%u ring:%u vmid:%u timestamp: %llu pasid:%u src_data: %08x %08x %08x %08x", __entry->client_id, __entry->src_id, __entry->ring_id, __entry->vmid, __entry->timestamp, __entry->pasid, @@ -150,10 +150,10 @@ TRACE_EVENT(amdgpu_cs, TP_fast_assign( __entry->bo_list = p->bo_list; - __entry->ring = p->job->ring->idx; + __entry->ring = to_amdgpu_ring(p->entity->rq->sched)->idx; __entry->dw = p->job->ibs[i].length_dw; __entry->fences = amdgpu_fence_count_emitted( - p->job->ring); + to_amdgpu_ring(p->entity->rq->sched)); ), TP_printk("bo_list=%p, ring=%u, dw=%u, fences=%u", __entry->bo_list, __entry->ring, __entry->dw, @@ -178,7 +178,7 @@ TRACE_EVENT(amdgpu_cs_ioctl, __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job)) __entry->context = job->base.s_fence->finished.context; __entry->seqno = job->base.s_fence->finished.seqno; - __entry->ring_name = job->ring->name; + __entry->ring_name = to_amdgpu_ring(job->base.sched)->name; __entry->num_ibs = job->num_ibs; ), TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u", @@ -203,7 +203,7 @@ TRACE_EVENT(amdgpu_sched_run_job, __assign_str(timeline, AMDGPU_JOB_GET_TIMELINE_NAME(job)) __entry->context = job->base.s_fence->finished.context; __entry->seqno = job->base.s_fence->finished.seqno; - __entry->ring_name = job->ring->name; + __entry->ring_name = to_amdgpu_ring(job->base.sched)->name; __entry->num_ibs = job->num_ibs; ), TP_printk("sched_job=%llu, timeline=%s, context=%u, seqno=%u, ring_name=%s, num_ibs=%u", @@ -314,6 +314,11 @@ DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_mapping, TP_ARGS(mapping) ); +DEFINE_EVENT(amdgpu_vm_mapping, amdgpu_vm_bo_cs, + TP_PROTO(struct amdgpu_bo_va_mapping *mapping), + TP_ARGS(mapping) +); + TRACE_EVENT(amdgpu_vm_set_ptes, TP_PROTO(uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint64_t flags), @@ -436,7 +441,7 @@ TRACE_EVENT(amdgpu_cs_bo_status, __entry->total_bo, __entry->total_size) ); -TRACE_EVENT(amdgpu_ttm_bo_move, +TRACE_EVENT(amdgpu_bo_move, TP_PROTO(struct amdgpu_bo* bo, uint32_t new_placement, uint32_t old_placement), TP_ARGS(bo, new_placement, old_placement), TP_STRUCT__entry( @@ -457,6 +462,30 @@ TRACE_EVENT(amdgpu_ttm_bo_move, __entry->new_placement, __entry->bo_size) ); +TRACE_EVENT(amdgpu_ib_pipe_sync, + TP_PROTO(struct amdgpu_job *sched_job, struct dma_fence *fence), + TP_ARGS(sched_job, fence), + TP_STRUCT__entry( + __field(const char *,name) + __field(uint64_t, id) + __field(struct dma_fence *, fence) + __field(uint64_t, ctx) + __field(unsigned, seqno) + ), + + TP_fast_assign( + __entry->name = sched_job->base.sched->name; + __entry->id = sched_job->base.id; + __entry->fence = fence; + __entry->ctx = fence->context; + __entry->seqno = fence->seqno; + ), + TP_printk("job ring=%s, id=%llu, need pipe sync to fence=%p, context=%llu, seq=%u", + __entry->name, __entry->id, + __entry->fence, __entry->ctx, + __entry->seqno) +); + #undef AMDGPU_JOB_GET_TIMELINE_NAME #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c index b160b958e5fe..f212402570a5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace_points.c @@ -1,4 +1,4 @@ -// SPDX-License-Identifier: GPL-2.0 +// SPDX-License-Identifier: MIT /* Copyright Red Hat Inc 2010. * * Permission is hereby granted, free of charge, to any person obtaining a diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index e93a0a237dc3..a44fc12ae1f9 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -47,6 +47,7 @@ #include "amdgpu_object.h" #include "amdgpu_trace.h" #include "amdgpu_amdkfd.h" +#include "amdgpu_sdma.h" #include "bif/bif_4_1_d.h" #define DRM_FILE_PAGE_OFFSET (0x100000000ULL >> PAGE_SHIFT) @@ -92,11 +93,9 @@ static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref) } /** - * amdgpu_ttm_global_init - Initialize global TTM memory reference - * structures. + * amdgpu_ttm_global_init - Initialize global TTM memory reference structures. * - * @adev: AMDGPU device for which the global structures need to be - * registered. + * @adev: AMDGPU device for which the global structures need to be registered. * * This is called as part of the AMDGPU ttm init from amdgpu_ttm_init() * during bring up. @@ -104,8 +103,6 @@ static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref) static int amdgpu_ttm_global_init(struct amdgpu_device *adev) { struct drm_global_reference *global_ref; - struct amdgpu_ring *ring; - struct drm_sched_rq *rq; int r; /* ensure reference is false in case init fails */ @@ -138,21 +135,10 @@ static int amdgpu_ttm_global_init(struct amdgpu_device *adev) mutex_init(&adev->mman.gtt_window_lock); - ring = adev->mman.buffer_funcs_ring; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL]; - r = drm_sched_entity_init(&ring->sched, &adev->mman.entity, - rq, NULL); - if (r) { - DRM_ERROR("Failed setting up TTM BO move run queue.\n"); - goto error_entity; - } - adev->mman.mem_global_referenced = true; return 0; -error_entity: - drm_global_item_unref(&adev->mman.bo_global_ref.ref); error_bo: drm_global_item_unref(&adev->mman.mem_global_ref); error_mem: @@ -162,8 +148,6 @@ error_mem: static void amdgpu_ttm_global_fini(struct amdgpu_device *adev) { if (adev->mman.mem_global_referenced) { - drm_sched_entity_fini(adev->mman.entity.sched, - &adev->mman.entity); mutex_destroy(&adev->mman.gtt_window_lock); drm_global_item_unref(&adev->mman.bo_global_ref.ref); drm_global_item_unref(&adev->mman.mem_global_ref); @@ -177,13 +161,12 @@ static int amdgpu_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags) } /** - * amdgpu_init_mem_type - Initialize a memory manager for a specific - * type of memory request. + * amdgpu_init_mem_type - Initialize a memory manager for a specific type of + * memory request. * - * @bdev: The TTM BO device object (contains a reference to - * amdgpu_device) - * @type: The type of memory requested - * @man: + * @bdev: The TTM BO device object (contains a reference to amdgpu_device) + * @type: The type of memory requested + * @man: The memory type manager for each domain * * This is called by ttm_bo_init_mm() when a buffer object is being * initialized. @@ -263,7 +246,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, } /* Object isn't an AMDGPU object so ignore */ - if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) { + if (!amdgpu_bo_is_amdgpu_bo(bo)) { placement->placement = &placements; placement->busy_placement = &placements; placement->num_placement = 1; @@ -273,11 +256,18 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, abo = ttm_to_amdgpu_bo(bo); switch (bo->mem.mem_type) { + case AMDGPU_PL_GDS: + case AMDGPU_PL_GWS: + case AMDGPU_PL_OA: + placement->num_placement = 0; + placement->num_busy_placement = 0; + return; + case TTM_PL_VRAM: if (!adev->mman.buffer_funcs_enabled) { /* Move to system memory */ - amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU); - } else if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size && + amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU); + } else if (!amdgpu_gmc_vram_full_visible(&adev->gmc) && !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) && amdgpu_bo_in_cpu_visible_vram(abo)) { @@ -286,7 +276,7 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, * BO will be evicted to GTT rather than causing other * BOs to be evicted from VRAM */ - amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | + amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT; abo->placements[0].lpfn = 0; @@ -294,12 +284,13 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, abo->placement.num_busy_placement = 1; } else { /* Move to GTT memory */ - amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT); + amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT); } break; case TTM_PL_TT: default: - amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU); + amdgpu_bo_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU); + break; } *placement = abo->placement; } @@ -307,8 +298,8 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, /** * amdgpu_verify_access - Verify access for a mmap call * - * @bo: The buffer object to map - * @filp: The file pointer from the process performing the mmap + * @bo: The buffer object to map + * @filp: The file pointer from the process performing the mmap * * This is called by ttm_bo_mmap() to verify whether a process * has the right to mmap a BO to their process space. @@ -333,11 +324,10 @@ static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp) /** * amdgpu_move_null - Register memory for a buffer object * - * @bo: The bo to assign the memory to - * @new_mem: The memory to be assigned. + * @bo: The bo to assign the memory to + * @new_mem: The memory to be assigned. * - * Assign the memory from new_mem to the memory of the buffer object - * bo. + * Assign the memory from new_mem to the memory of the buffer object bo. */ static void amdgpu_move_null(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem) @@ -350,8 +340,12 @@ static void amdgpu_move_null(struct ttm_buffer_object *bo, } /** - * amdgpu_mm_node_addr - Compute the GPU relative offset of a GTT - * buffer. + * amdgpu_mm_node_addr - Compute the GPU relative offset of a GTT buffer. + * + * @bo: The bo to assign the memory to. + * @mm_node: Memory manager node for drm allocator. + * @mem: The region where the bo resides. + * */ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, struct drm_mm_node *mm_node, @@ -359,7 +353,7 @@ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, { uint64_t addr = 0; - if (mem->mem_type != TTM_PL_TT || amdgpu_gtt_mgr_has_gart_addr(mem)) { + if (mm_node->start != AMDGPU_BO_INVALID_OFFSET) { addr = mm_node->start << PAGE_SHIFT; addr += bo->bdev->man[mem->mem_type].gpu_offset; } @@ -367,10 +361,12 @@ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, } /** - * amdgpu_find_mm_node - Helper function finds the drm_mm_node - * corresponding to @offset. It also modifies - * the offset to be within the drm_mm_node - * returned + * amdgpu_find_mm_node - Helper function finds the drm_mm_node corresponding to + * @offset. It also modifies the offset to be within the drm_mm_node returned + * + * @mem: The region where the bo resides. + * @offset: The offset that drm_mm_node is used for finding. + * */ static struct drm_mm_node *amdgpu_find_mm_node(struct ttm_mem_reg *mem, unsigned long *offset) @@ -445,8 +441,7 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev, /* Map only what needs to be accessed. Map src to window 0 and * dst to window 1 */ - if (src->mem->mem_type == TTM_PL_TT && - !amdgpu_gtt_mgr_has_gart_addr(src->mem)) { + if (src->mem->start == AMDGPU_BO_INVALID_OFFSET) { r = amdgpu_map_buffer(src->bo, src->mem, PFN_UP(cur_size + src_page_offset), src_node_start, 0, ring, @@ -459,8 +454,7 @@ int amdgpu_ttm_copy_mem_to_mem(struct amdgpu_device *adev, from += src_page_offset; } - if (dst->mem->mem_type == TTM_PL_TT && - !amdgpu_gtt_mgr_has_gart_addr(dst->mem)) { + if (dst->mem->start == AMDGPU_BO_INVALID_OFFSET) { r = amdgpu_map_buffer(dst->bo, dst->mem, PFN_UP(cur_size + dst_page_offset), dst_node_start, 1, ring, @@ -512,8 +506,8 @@ error: /** * amdgpu_move_blit - Copy an entire buffer to another buffer * - * This is a helper called by amdgpu_bo_move() and - * amdgpu_move_vram_ram() to help move buffers to and from VRAM. + * This is a helper called by amdgpu_bo_move() and amdgpu_move_vram_ram() to + * help move buffers to and from VRAM. */ static int amdgpu_move_blit(struct ttm_buffer_object *bo, bool evict, bool no_wait_gpu, @@ -538,7 +532,11 @@ static int amdgpu_move_blit(struct ttm_buffer_object *bo, if (r) goto error; - r = ttm_bo_pipeline_move(bo, fence, evict, new_mem); + /* Always block for VM page tables before committing the new location */ + if (bo->type == ttm_bo_type_kernel) + r = ttm_bo_move_accel_cleanup(bo, fence, true, new_mem); + else + r = ttm_bo_pipeline_move(bo, fence, evict, new_mem); dma_fence_put(fence); return r; @@ -595,7 +593,7 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict, } /* blit VRAM to GTT */ - r = amdgpu_move_blit(bo, true, ctx->no_wait_gpu, &tmp_mem, old_mem); + r = amdgpu_move_blit(bo, evict, ctx->no_wait_gpu, &tmp_mem, old_mem); if (unlikely(r)) { goto out_cleanup; } @@ -647,7 +645,7 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, bool evict, } /* copy to VRAM */ - r = amdgpu_move_blit(bo, true, ctx->no_wait_gpu, new_mem, old_mem); + r = amdgpu_move_blit(bo, evict, ctx->no_wait_gpu, new_mem, old_mem); if (unlikely(r)) { goto out_cleanup; } @@ -689,6 +687,16 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict, amdgpu_move_null(bo, new_mem); return 0; } + if (old_mem->mem_type == AMDGPU_PL_GDS || + old_mem->mem_type == AMDGPU_PL_GWS || + old_mem->mem_type == AMDGPU_PL_OA || + new_mem->mem_type == AMDGPU_PL_GDS || + new_mem->mem_type == AMDGPU_PL_GWS || + new_mem->mem_type == AMDGPU_PL_OA) { + /* Nothing to save here */ + amdgpu_move_null(bo, new_mem); + return 0; + } if (!adev->mman.buffer_funcs_enabled) goto memcpy; @@ -809,8 +817,8 @@ struct amdgpu_ttm_tt { }; /** - * amdgpu_ttm_tt_get_user_pages - Pin pages of memory pointed to - * by a USERPTR pointer to memory + * amdgpu_ttm_tt_get_user_pages - Pin pages of memory pointed to by a USERPTR + * pointer to memory * * Called by amdgpu_gem_userptr_ioctl() and amdgpu_cs_parser_bos(). * This provides a wrapper around the get_user_pages() call to provide @@ -833,8 +841,10 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) down_read(&mm->mmap_sem); if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) { - /* check that we only use anonymous memory - to prevent problems with writeback */ + /* + * check that we only use anonymous memory to prevent problems + * with writeback + */ unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE; struct vm_area_struct *vma; @@ -885,10 +895,9 @@ release_pages: } /** - * amdgpu_ttm_tt_set_user_pages - Copy pages in, putting old pages - * as necessary. + * amdgpu_ttm_tt_set_user_pages - Copy pages in, putting old pages as necessary. * - * Called by amdgpu_cs_list_validate(). This creates the page list + * Called by amdgpu_cs_list_validate(). This creates the page list * that backs user memory and will ultimately be mapped into the device * address space. */ @@ -930,8 +939,7 @@ void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm) } /** - * amdgpu_ttm_tt_pin_userptr - prepare the sg table with the - * user pages + * amdgpu_ttm_tt_pin_userptr - prepare the sg table with the user pages * * Called by amdgpu_ttm_backend_bind() **/ @@ -1095,42 +1103,48 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) struct ttm_mem_reg tmp; struct ttm_placement placement; struct ttm_place placements; - uint64_t flags; + uint64_t addr, flags; int r; - if (bo->mem.mem_type != TTM_PL_TT || - amdgpu_gtt_mgr_has_gart_addr(&bo->mem)) + if (bo->mem.start != AMDGPU_BO_INVALID_OFFSET) return 0; - /* allocate GTT space */ - tmp = bo->mem; - tmp.mm_node = NULL; - placement.num_placement = 1; - placement.placement = &placements; - placement.num_busy_placement = 1; - placement.busy_placement = &placements; - placements.fpfn = 0; - placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT; - placements.flags = (bo->mem.placement & ~TTM_PL_MASK_MEM) | - TTM_PL_FLAG_TT; + addr = amdgpu_gmc_agp_addr(bo); + if (addr != AMDGPU_BO_INVALID_OFFSET) { + bo->mem.start = addr >> PAGE_SHIFT; + } else { - r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx); - if (unlikely(r)) - return r; + /* allocate GART space */ + tmp = bo->mem; + tmp.mm_node = NULL; + placement.num_placement = 1; + placement.placement = &placements; + placement.num_busy_placement = 1; + placement.busy_placement = &placements; + placements.fpfn = 0; + placements.lpfn = adev->gmc.gart_size >> PAGE_SHIFT; + placements.flags = (bo->mem.placement & ~TTM_PL_MASK_MEM) | + TTM_PL_FLAG_TT; + + r = ttm_bo_mem_space(bo, &placement, &tmp, &ctx); + if (unlikely(r)) + return r; - /* compute PTE flags for this buffer object */ - flags = amdgpu_ttm_tt_pte_flags(adev, bo->ttm, &tmp); + /* compute PTE flags for this buffer object */ + flags = amdgpu_ttm_tt_pte_flags(adev, bo->ttm, &tmp); - /* Bind pages */ - gtt->offset = (u64)tmp.start << PAGE_SHIFT; - r = amdgpu_ttm_gart_bind(adev, bo, flags); - if (unlikely(r)) { - ttm_bo_mem_put(bo, &tmp); - return r; + /* Bind pages */ + gtt->offset = (u64)tmp.start << PAGE_SHIFT; + r = amdgpu_ttm_gart_bind(adev, bo, flags); + if (unlikely(r)) { + ttm_bo_mem_put(bo, &tmp); + return r; + } + + ttm_bo_mem_put(bo, &bo->mem); + bo->mem = tmp; } - ttm_bo_mem_put(bo, &bo->mem); - bo->mem = tmp; bo->offset = (bo->mem.start << PAGE_SHIFT) + bo->bdev->man[bo->mem.mem_type].gpu_offset; @@ -1310,8 +1324,8 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) } /** - * amdgpu_ttm_tt_set_userptr - Initialize userptr GTT ttm_tt - * for the current task + * amdgpu_ttm_tt_set_userptr - Initialize userptr GTT ttm_tt for the current + * task * * @ttm: The ttm_tt object to bind this userptr object to * @addr: The address in the current tasks VM space to use @@ -1361,9 +1375,8 @@ struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm) } /** - * amdgpu_ttm_tt_affect_userptr - Determine if a ttm_tt object lays - * inside an address range for the - * current task. + * amdgpu_ttm_tt_affect_userptr - Determine if a ttm_tt object lays inside an + * address range for the current task. * */ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, @@ -1401,8 +1414,7 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, } /** - * amdgpu_ttm_tt_userptr_invalidated - Has the ttm_tt object been - * invalidated? + * amdgpu_ttm_tt_userptr_invalidated - Has the ttm_tt object been invalidated? */ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, int *last_invalidated) @@ -1415,10 +1427,8 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, } /** - * amdgpu_ttm_tt_userptr_needs_pages - Have the pages backing this - * ttm_tt object been invalidated - * since the last time they've - * been set? + * amdgpu_ttm_tt_userptr_needs_pages - Have the pages backing this ttm_tt object + * been invalidated since the last time they've been set? */ bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm) { @@ -1444,13 +1454,14 @@ bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm) } /** - * amdgpu_ttm_tt_pte_flags - Compute PTE flags for ttm_tt object + * amdgpu_ttm_tt_pde_flags - Compute PDE flags for ttm_tt object * * @ttm: The ttm_tt object to compute the flags for * @mem: The memory registry backing this ttm_tt object + * + * Figure out the flags to use for a VM PDE (Page Directory Entry). */ -uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, - struct ttm_mem_reg *mem) +uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_mem_reg *mem) { uint64_t flags = 0; @@ -1464,6 +1475,22 @@ uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, flags |= AMDGPU_PTE_SNOOPED; } + return flags; +} + +/** + * amdgpu_ttm_tt_pte_flags - Compute PTE flags for ttm_tt object + * + * @ttm: The ttm_tt object to compute the flags for + * @mem: The memory registry backing this ttm_tt object + + * Figure out the flags to use for a VM PTE (Page Table Entry). + */ +uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, + struct ttm_mem_reg *mem) +{ + uint64_t flags = amdgpu_ttm_tt_pde_flags(ttm, mem); + flags |= adev->gart.gart_pte_flags; flags |= AMDGPU_PTE_READABLE; @@ -1474,13 +1501,12 @@ uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, } /** - * amdgpu_ttm_bo_eviction_valuable - Check to see if we can evict - * a buffer object. + * amdgpu_ttm_bo_eviction_valuable - Check to see if we can evict a buffer + * object. * - * Return true if eviction is sensible. Called by - * ttm_mem_evict_first() on behalf of ttm_bo_mem_force_space() - * which tries to evict buffer objects until it can find space - * for a new object and by ttm_bo_force_list_clean() which is + * Return true if eviction is sensible. Called by ttm_mem_evict_first() on + * behalf of ttm_bo_mem_force_space() which tries to evict buffer objects until + * it can find space for a new object and by ttm_bo_force_list_clean() which is * used to clean out a memory space. */ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, @@ -1530,8 +1556,7 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, } /** - * amdgpu_ttm_access_memory - Read or Write memory that backs a - * buffer object. + * amdgpu_ttm_access_memory - Read or Write memory that backs a buffer object. * * @bo: The buffer object to read/write * @offset: Offset into buffer object @@ -1695,7 +1720,7 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) AMDGPU_GEM_DOMAIN_VRAM, adev->fw_vram_usage.start_offset, (adev->fw_vram_usage.start_offset + - adev->fw_vram_usage.size), NULL); + adev->fw_vram_usage.size)); if (r) goto error_pin; r = amdgpu_bo_kmap(adev->fw_vram_usage.reserved_bo, @@ -1719,8 +1744,8 @@ error_create: return r; } /** - * amdgpu_ttm_init - Init the memory management (ttm) as well as - * various gtt/vram related fields. + * amdgpu_ttm_init - Init the memory management (ttm) as well as various + * gtt/vram related fields. * * This initializes all of the memory space pools that the TTM layer * will need such as the GTT space (system memory mapped to the device), @@ -1788,14 +1813,12 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) * This is used for VGA emulation and pre-OS scanout buffers to * avoid display artifacts while transitioning between pre-OS * and driver. */ - if (adev->gmc.stolen_size) { - r = amdgpu_bo_create_kernel(adev, adev->gmc.stolen_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->stolen_vga_memory, - NULL, NULL); - if (r) - return r; - } + r = amdgpu_bo_create_kernel(adev, adev->gmc.stolen_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->stolen_vga_memory, + NULL, NULL); + if (r) + return r; DRM_INFO("amdgpu: %uM of VRAM memory ready\n", (unsigned) (adev->gmc.real_vram_size / (1024 * 1024))); @@ -1822,45 +1845,45 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) (unsigned)(gtt_size / (1024 * 1024))); /* Initialize various on-chip memory pools */ - adev->gds.mem.total_size = adev->gds.mem.total_size << AMDGPU_GDS_SHIFT; - adev->gds.mem.gfx_partition_size = adev->gds.mem.gfx_partition_size << AMDGPU_GDS_SHIFT; - adev->gds.mem.cs_partition_size = adev->gds.mem.cs_partition_size << AMDGPU_GDS_SHIFT; - adev->gds.gws.total_size = adev->gds.gws.total_size << AMDGPU_GWS_SHIFT; - adev->gds.gws.gfx_partition_size = adev->gds.gws.gfx_partition_size << AMDGPU_GWS_SHIFT; - adev->gds.gws.cs_partition_size = adev->gds.gws.cs_partition_size << AMDGPU_GWS_SHIFT; - adev->gds.oa.total_size = adev->gds.oa.total_size << AMDGPU_OA_SHIFT; - adev->gds.oa.gfx_partition_size = adev->gds.oa.gfx_partition_size << AMDGPU_OA_SHIFT; - adev->gds.oa.cs_partition_size = adev->gds.oa.cs_partition_size << AMDGPU_OA_SHIFT; - /* GDS Memory */ - if (adev->gds.mem.total_size) { - r = ttm_bo_init_mm(&adev->mman.bdev, AMDGPU_PL_GDS, - adev->gds.mem.total_size >> PAGE_SHIFT); - if (r) { - DRM_ERROR("Failed initializing GDS heap.\n"); - return r; - } + r = ttm_bo_init_mm(&adev->mman.bdev, AMDGPU_PL_GDS, + adev->gds.mem.total_size); + if (r) { + DRM_ERROR("Failed initializing GDS heap.\n"); + return r; } - /* GWS */ - if (adev->gds.gws.total_size) { - r = ttm_bo_init_mm(&adev->mman.bdev, AMDGPU_PL_GWS, - adev->gds.gws.total_size >> PAGE_SHIFT); - if (r) { - DRM_ERROR("Failed initializing gws heap.\n"); - return r; - } + r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, + &adev->gds.gds_gfx_bo, NULL, NULL); + if (r) + return r; + + r = ttm_bo_init_mm(&adev->mman.bdev, AMDGPU_PL_GWS, + adev->gds.gws.total_size); + if (r) { + DRM_ERROR("Failed initializing gws heap.\n"); + return r; } - /* OA */ - if (adev->gds.oa.total_size) { - r = ttm_bo_init_mm(&adev->mman.bdev, AMDGPU_PL_OA, - adev->gds.oa.total_size >> PAGE_SHIFT); - if (r) { - DRM_ERROR("Failed initializing oa heap.\n"); - return r; - } + r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, + &adev->gds.gws_gfx_bo, NULL, NULL); + if (r) + return r; + + r = ttm_bo_init_mm(&adev->mman.bdev, AMDGPU_PL_OA, + adev->gds.oa.total_size); + if (r) { + DRM_ERROR("Failed initializing oa heap.\n"); + return r; } + r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, + &adev->gds.oa_gfx_bo, NULL, NULL); + if (r) + return r; + /* Register debugfs entries for amdgpu_ttm */ r = amdgpu_ttm_debugfs_init(adev); if (r) { @@ -1871,8 +1894,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } /** - * amdgpu_ttm_late_init - Handle any late initialization for - * amdgpu_ttm + * amdgpu_ttm_late_init - Handle any late initialization for amdgpu_ttm */ void amdgpu_ttm_late_init(struct amdgpu_device *adev) { @@ -1896,12 +1918,9 @@ void amdgpu_ttm_fini(struct amdgpu_device *adev) ttm_bo_clean_mm(&adev->mman.bdev, TTM_PL_VRAM); ttm_bo_clean_mm(&adev->mman.bdev, TTM_PL_TT); - if (adev->gds.mem.total_size) - ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_GDS); - if (adev->gds.gws.total_size) - ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_GWS); - if (adev->gds.oa.total_size) - ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_OA); + ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_GDS); + ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_GWS); + ttm_bo_clean_mm(&adev->mman.bdev, AMDGPU_PL_OA); ttm_bo_device_release(&adev->mman.bdev); amdgpu_ttm_global_fini(adev); adev->mman.initialized = false; @@ -1921,10 +1940,30 @@ void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable) { struct ttm_mem_type_manager *man = &adev->mman.bdev.man[TTM_PL_VRAM]; uint64_t size; + int r; - if (!adev->mman.initialized || adev->in_gpu_reset) + if (!adev->mman.initialized || adev->in_gpu_reset || + adev->mman.buffer_funcs_enabled == enable) return; + if (enable) { + struct amdgpu_ring *ring; + struct drm_sched_rq *rq; + + ring = adev->mman.buffer_funcs_ring; + rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL]; + r = drm_sched_entity_init(&adev->mman.entity, &rq, 1, NULL); + if (r) { + DRM_ERROR("Failed setting up TTM BO move entity (%d)\n", + r); + return; + } + } else { + drm_sched_entity_destroy(&adev->mman.entity); + dma_fence_put(man->move); + man->move = NULL; + } + /* this just adjusts TTM size idea, which sets lpfn to the correct value */ if (enable) size = adev->gmc.real_vram_size; @@ -1987,7 +2026,7 @@ static int amdgpu_map_buffer(struct ttm_buffer_object *bo, src_addr = num_dw * 4; src_addr += job->ibs[0].gpu_addr; - dst_addr = adev->gart.table_addr; + dst_addr = amdgpu_bo_gpu_offset(adev->gart.bo); dst_addr += window * AMDGPU_GTT_MAX_TRANSFER_SIZE * 8; amdgpu_emit_copy_buffer(adev, &job->ibs[0], src_addr, dst_addr, num_bytes); @@ -2002,7 +2041,7 @@ static int amdgpu_map_buffer(struct ttm_buffer_object *bo, if (r) goto error_free; - r = amdgpu_job_submit(job, ring, &adev->mman.entity, + r = amdgpu_job_submit(job, &adev->mman.entity, AMDGPU_FENCE_OWNER_UNDEFINED, &fence); if (r) goto error_free; @@ -2047,7 +2086,10 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, if (r) return r; - job->vm_needs_flush = vm_needs_flush; + if (vm_needs_flush) { + job->vm_pd_addr = amdgpu_gmc_pd_addr(adev->gart.bo); + job->vm_needs_flush = true; + } if (resv) { r = amdgpu_sync_resv(adev, &job->sync, resv, AMDGPU_FENCE_OWNER_UNDEFINED, @@ -2071,24 +2113,19 @@ int amdgpu_copy_buffer(struct amdgpu_ring *ring, uint64_t src_offset, amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > num_dw); - if (direct_submit) { - r = amdgpu_ib_schedule(ring, job->num_ibs, job->ibs, - NULL, fence); - job->fence = dma_fence_get(*fence); - if (r) - DRM_ERROR("Error scheduling IBs (%d)\n", r); - amdgpu_job_free(job); - } else { - r = amdgpu_job_submit(job, ring, &adev->mman.entity, + if (direct_submit) + r = amdgpu_job_submit_direct(job, ring, fence); + else + r = amdgpu_job_submit(job, &adev->mman.entity, AMDGPU_FENCE_OWNER_UNDEFINED, fence); - if (r) - goto error_free; - } + if (r) + goto error_free; return r; error_free: amdgpu_job_free(job); + DRM_ERROR("Error scheduling IBs (%d)\n", r); return r; } @@ -2171,7 +2208,7 @@ int amdgpu_fill_buffer(struct amdgpu_bo *bo, amdgpu_ring_pad_ib(ring, &job->ibs[0]); WARN_ON(job->ibs[0].length_dw > num_dw); - r = amdgpu_job_submit(job, ring, &adev->mman.entity, + r = amdgpu_job_submit(job, &adev->mman.entity, AMDGPU_FENCE_OWNER_UNDEFINED, fence); if (r) goto error_free; @@ -2188,7 +2225,7 @@ error_free: static int amdgpu_mm_dump_table(struct seq_file *m, void *data) { struct drm_info_node *node = (struct drm_info_node *)m->private; - unsigned ttm_pl = *(int *)node->info_ent->data; + unsigned ttm_pl = (uintptr_t)node->info_ent->data; struct drm_device *dev = node->minor->dev; struct amdgpu_device *adev = dev->dev_private; struct ttm_mem_type_manager *man = &adev->mman.bdev.man[ttm_pl]; @@ -2198,12 +2235,12 @@ static int amdgpu_mm_dump_table(struct seq_file *m, void *data) return 0; } -static int ttm_pl_vram = TTM_PL_VRAM; -static int ttm_pl_tt = TTM_PL_TT; - static const struct drm_info_list amdgpu_ttm_debugfs_list[] = { - {"amdgpu_vram_mm", amdgpu_mm_dump_table, 0, &ttm_pl_vram}, - {"amdgpu_gtt_mm", amdgpu_mm_dump_table, 0, &ttm_pl_tt}, + {"amdgpu_vram_mm", amdgpu_mm_dump_table, 0, (void *)TTM_PL_VRAM}, + {"amdgpu_gtt_mm", amdgpu_mm_dump_table, 0, (void *)TTM_PL_TT}, + {"amdgpu_gds_mm", amdgpu_mm_dump_table, 0, (void *)AMDGPU_PL_GDS}, + {"amdgpu_gws_mm", amdgpu_mm_dump_table, 0, (void *)AMDGPU_PL_GWS}, + {"amdgpu_oa_mm", amdgpu_mm_dump_table, 0, (void *)AMDGPU_PL_OA}, {"ttm_page_pool", ttm_page_alloc_debugfs, 0, NULL}, #ifdef CONFIG_SWIOTLB {"ttm_dma_page_pool", ttm_dma_page_alloc_debugfs, 0, NULL} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index e5da4654b630..fe8f276e9811 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -73,7 +73,7 @@ bool amdgpu_gtt_mgr_has_gart_addr(struct ttm_mem_reg *mem); uint64_t amdgpu_gtt_mgr_usage(struct ttm_mem_type_manager *man); int amdgpu_gtt_mgr_recover(struct ttm_mem_type_manager *man); -u64 amdgpu_vram_mgr_bo_invisible_size(struct amdgpu_bo *bo); +u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo); uint64_t amdgpu_vram_mgr_usage(struct ttm_mem_type_manager *man); uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_mem_type_manager *man); @@ -116,6 +116,7 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, int *last_invalidated); bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm); bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm); +uint64_t amdgpu_ttm_tt_pde_flags(struct ttm_tt *ttm, struct ttm_mem_reg *mem); uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, struct ttm_mem_reg *mem); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index f55f72a37ca8..7b33867036e7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -277,6 +277,7 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) case CHIP_PITCAIRN: case CHIP_VERDE: case CHIP_OLAND: + case CHIP_HAINAN: return AMDGPU_FW_LOAD_DIRECT; #endif #ifdef CONFIG_DRM_AMDGPU_CIK @@ -296,19 +297,15 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) case CHIP_POLARIS11: case CHIP_POLARIS12: case CHIP_VEGAM: - if (!load_type) - return AMDGPU_FW_LOAD_DIRECT; - else - return AMDGPU_FW_LOAD_SMU; + return AMDGPU_FW_LOAD_SMU; case CHIP_VEGA10: case CHIP_RAVEN: case CHIP_VEGA12: + case CHIP_VEGA20: if (!load_type) return AMDGPU_FW_LOAD_DIRECT; else return AMDGPU_FW_LOAD_PSP; - case CHIP_VEGA20: - return AMDGPU_FW_LOAD_DIRECT; default: DRM_ERROR("Unknown firmware load type\n"); } @@ -322,6 +319,7 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, { const struct common_firmware_header *header = NULL; const struct gfx_firmware_header_v1_0 *cp_hdr = NULL; + const struct dmcu_firmware_header_v1_0 *dmcu_hdr = NULL; if (NULL == ucode->fw) return 0; @@ -333,8 +331,8 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, return 0; header = (const struct common_firmware_header *)ucode->fw->data; - cp_hdr = (const struct gfx_firmware_header_v1_0 *)ucode->fw->data; + dmcu_hdr = (const struct dmcu_firmware_header_v1_0 *)ucode->fw->data; if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP || (ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC1 && @@ -343,7 +341,9 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC2_JT && ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL && ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM && - ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM)) { + ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM && + ucode->ucode_id != AMDGPU_UCODE_ID_DMCU_ERAM && + ucode->ucode_id != AMDGPU_UCODE_ID_DMCU_INTV)) { ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes); memcpy(ucode->kaddr, (void *)((uint8_t *)ucode->fw->data + @@ -365,6 +365,20 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, le32_to_cpu(header->ucode_array_offset_bytes) + le32_to_cpu(cp_hdr->jt_offset) * 4), ucode->ucode_size); + } else if (ucode->ucode_id == AMDGPU_UCODE_ID_DMCU_ERAM) { + ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes) - + le32_to_cpu(dmcu_hdr->intv_size_bytes); + + memcpy(ucode->kaddr, (void *)((uint8_t *)ucode->fw->data + + le32_to_cpu(header->ucode_array_offset_bytes)), + ucode->ucode_size); + } else if (ucode->ucode_id == AMDGPU_UCODE_ID_DMCU_INTV) { + ucode->ucode_size = le32_to_cpu(dmcu_hdr->intv_size_bytes); + + memcpy(ucode->kaddr, (void *)((uint8_t *)ucode->fw->data + + le32_to_cpu(header->ucode_array_offset_bytes) + + le32_to_cpu(dmcu_hdr->intv_offset_bytes)), + ucode->ucode_size); } else if (ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL) { ucode->ucode_size = adev->gfx.rlc.save_restore_list_cntl_size_bytes; memcpy(ucode->kaddr, adev->gfx.rlc.save_restore_list_cntl, @@ -406,32 +420,41 @@ static int amdgpu_ucode_patch_jt(struct amdgpu_firmware_info *ucode, return 0; } -int amdgpu_ucode_init_bo(struct amdgpu_device *adev) +int amdgpu_ucode_create_bo(struct amdgpu_device *adev) { - uint64_t fw_offset = 0; - int i, err; - struct amdgpu_firmware_info *ucode = NULL; - const struct common_firmware_header *header = NULL; - - if (!adev->firmware.fw_size) { - dev_warn(adev->dev, "No ip firmware need to load\n"); - return 0; - } - - if (!adev->in_gpu_reset) { - err = amdgpu_bo_create_kernel(adev, adev->firmware.fw_size, PAGE_SIZE, - amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, - &adev->firmware.fw_buf, - &adev->firmware.fw_buf_mc, - &adev->firmware.fw_buf_ptr); - if (err) { + if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) { + amdgpu_bo_create_kernel(adev, adev->firmware.fw_size, PAGE_SIZE, + amdgpu_sriov_vf(adev) ? AMDGPU_GEM_DOMAIN_VRAM : AMDGPU_GEM_DOMAIN_GTT, + &adev->firmware.fw_buf, + &adev->firmware.fw_buf_mc, + &adev->firmware.fw_buf_ptr); + if (!adev->firmware.fw_buf) { dev_err(adev->dev, "failed to create kernel buffer for firmware.fw_buf\n"); - goto failed; + return -ENOMEM; + } else if (amdgpu_sriov_vf(adev)) { + memset(adev->firmware.fw_buf_ptr, 0, adev->firmware.fw_size); } } + return 0; +} - memset(adev->firmware.fw_buf_ptr, 0, adev->firmware.fw_size); +void amdgpu_ucode_free_bo(struct amdgpu_device *adev) +{ + if (adev->firmware.load_type != AMDGPU_FW_LOAD_DIRECT) + amdgpu_bo_free_kernel(&adev->firmware.fw_buf, + &adev->firmware.fw_buf_mc, + &adev->firmware.fw_buf_ptr); +} + +int amdgpu_ucode_init_bo(struct amdgpu_device *adev) +{ + uint64_t fw_offset = 0; + int i; + struct amdgpu_firmware_info *ucode = NULL; + /* for baremetal, the ucode is allocated in gtt, so don't need to fill the bo when reset/suspend */ + if (!amdgpu_sriov_vf(adev) && (adev->in_gpu_reset || adev->in_suspend)) + return 0; /* * if SMU loaded firmware, it needn't add SMC, UVD, and VCE * ucode info here @@ -448,7 +471,6 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev) for (i = 0; i < adev->firmware.max_ucodes; i++) { ucode = &adev->firmware.ucode[i]; if (ucode->fw) { - header = (const struct common_firmware_header *)ucode->fw->data; amdgpu_ucode_init_single_fw(adev, ucode, adev->firmware.fw_buf_mc + fw_offset, adev->firmware.fw_buf_ptr + fw_offset); if (i == AMDGPU_UCODE_ID_CP_MEC1 && @@ -463,33 +485,4 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev) } } return 0; - -failed: - if (err) - adev->firmware.load_type = AMDGPU_FW_LOAD_DIRECT; - - return err; -} - -int amdgpu_ucode_fini_bo(struct amdgpu_device *adev) -{ - int i; - struct amdgpu_firmware_info *ucode = NULL; - - if (!adev->firmware.fw_size) - return 0; - - for (i = 0; i < adev->firmware.max_ucodes; i++) { - ucode = &adev->firmware.ucode[i]; - if (ucode->fw) { - ucode->mc_addr = 0; - ucode->kaddr = NULL; - } - } - - amdgpu_bo_free_kernel(&adev->firmware.fw_buf, - &adev->firmware.fw_buf_mc, - &adev->firmware.fw_buf_ptr); - - return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index 08e38579af24..aa6641b944a0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -157,6 +157,13 @@ struct gpu_info_firmware_header_v1_0 { uint16_t version_minor; /* version */ }; +/* version_major=1, version_minor=0 */ +struct dmcu_firmware_header_v1_0 { + struct common_firmware_header header; + uint32_t intv_offset_bytes; /* interrupt vectors offset from end of header, in bytes */ + uint32_t intv_size_bytes; /* size of interrupt vectors, in bytes */ +}; + /* header is fixed size */ union amdgpu_firmware_header { struct common_firmware_header common; @@ -170,6 +177,7 @@ union amdgpu_firmware_header { struct sdma_firmware_header_v1_0 sdma; struct sdma_firmware_header_v1_1 sdma_v1_1; struct gpu_info_firmware_header_v1_0 gpu_info; + struct dmcu_firmware_header_v1_0 dmcu; uint8_t raw[0x100]; }; @@ -193,7 +201,11 @@ enum AMDGPU_UCODE_ID { AMDGPU_UCODE_ID_STORAGE, AMDGPU_UCODE_ID_SMC, AMDGPU_UCODE_ID_UVD, + AMDGPU_UCODE_ID_UVD1, AMDGPU_UCODE_ID_VCE, + AMDGPU_UCODE_ID_VCN, + AMDGPU_UCODE_ID_DMCU_ERAM, + AMDGPU_UCODE_ID_DMCU_INTV, AMDGPU_UCODE_ID_MAXIMUM, }; @@ -204,6 +216,12 @@ enum AMDGPU_UCODE_STATUS { AMDGPU_UCODE_STATUS_LOADED, }; +enum amdgpu_firmware_load_type { + AMDGPU_FW_LOAD_DIRECT = 0, + AMDGPU_FW_LOAD_SMU, + AMDGPU_FW_LOAD_PSP, +}; + /* conform to smu_ucode_xfer_cz.h */ #define AMDGPU_SDMA0_UCODE_LOADED 0x00000001 #define AMDGPU_SDMA1_UCODE_LOADED 0x00000002 @@ -226,6 +244,27 @@ struct amdgpu_firmware_info { void *kaddr; /* ucode_size_bytes */ uint32_t ucode_size; + /* starting tmr mc address */ + uint32_t tmr_mc_addr_lo; + uint32_t tmr_mc_addr_hi; +}; + +struct amdgpu_firmware { + struct amdgpu_firmware_info ucode[AMDGPU_UCODE_ID_MAXIMUM]; + enum amdgpu_firmware_load_type load_type; + struct amdgpu_bo *fw_buf; + unsigned int fw_size; + unsigned int max_ucodes; + /* firmwares are loaded by psp instead of smu from vega10 */ + const struct amdgpu_psp_funcs *funcs; + struct amdgpu_bo *rbuf; + struct mutex mutex; + + /* gpu info firmware data pointer */ + const struct firmware *gpu_info_fw; + + void *fw_buf_ptr; + uint64_t fw_buf_mc; }; void amdgpu_ucode_print_mc_hdr(const struct common_firmware_header *hdr); @@ -237,8 +276,10 @@ void amdgpu_ucode_print_gpu_info_hdr(const struct common_firmware_header *hdr); int amdgpu_ucode_validate(const struct firmware *fw); bool amdgpu_ucode_hdr_version(union amdgpu_firmware_header *hdr, uint16_t hdr_major, uint16_t hdr_minor); + int amdgpu_ucode_init_bo(struct amdgpu_device *adev); -int amdgpu_ucode_fini_bo(struct amdgpu_device *adev); +int amdgpu_ucode_create_bo(struct amdgpu_device *adev); +void amdgpu_ucode_free_bo(struct amdgpu_device *adev); enum amdgpu_firmware_load_type amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 3ff08e326838..e5a6db6beab7 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -53,11 +53,11 @@ /* Firmware Names */ #ifdef CONFIG_DRM_AMDGPU_CIK -#define FIRMWARE_BONAIRE "radeon/bonaire_uvd.bin" -#define FIRMWARE_KABINI "radeon/kabini_uvd.bin" -#define FIRMWARE_KAVERI "radeon/kaveri_uvd.bin" -#define FIRMWARE_HAWAII "radeon/hawaii_uvd.bin" -#define FIRMWARE_MULLINS "radeon/mullins_uvd.bin" +#define FIRMWARE_BONAIRE "amdgpu/bonaire_uvd.bin" +#define FIRMWARE_KABINI "amdgpu/kabini_uvd.bin" +#define FIRMWARE_KAVERI "amdgpu/kaveri_uvd.bin" +#define FIRMWARE_HAWAII "amdgpu/hawaii_uvd.bin" +#define FIRMWARE_MULLINS "amdgpu/mullins_uvd.bin" #endif #define FIRMWARE_TONGA "amdgpu/tonga_uvd.bin" #define FIRMWARE_CARRIZO "amdgpu/carrizo_uvd.bin" @@ -122,12 +122,10 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work); int amdgpu_uvd_sw_init(struct amdgpu_device *adev) { - struct amdgpu_ring *ring; - struct drm_sched_rq *rq; unsigned long bo_size; const char *fw_name; const struct common_firmware_header *hdr; - unsigned version_major, version_minor, family_id; + unsigned family_id; int i, j, r; INIT_DELAYED_WORK(&adev->uvd.idle_work, amdgpu_uvd_idle_work_handler); @@ -208,29 +206,46 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) hdr = (const struct common_firmware_header *)adev->uvd.fw->data; family_id = le32_to_cpu(hdr->ucode_version) & 0xff; - version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff; - version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff; - DRM_INFO("Found UVD firmware Version: %hu.%hu Family ID: %hu\n", - version_major, version_minor, family_id); - - /* - * Limit the number of UVD handles depending on microcode major - * and minor versions. The firmware version which has 40 UVD - * instances support is 1.80. So all subsequent versions should - * also have the same support. - */ - if ((version_major > 0x01) || - ((version_major == 0x01) && (version_minor >= 0x50))) - adev->uvd.max_handles = AMDGPU_MAX_UVD_HANDLES; - adev->uvd.fw_version = ((version_major << 24) | (version_minor << 16) | - (family_id << 8)); + if (adev->asic_type < CHIP_VEGA20) { + unsigned version_major, version_minor; - if ((adev->asic_type == CHIP_POLARIS10 || - adev->asic_type == CHIP_POLARIS11) && - (adev->uvd.fw_version < FW_1_66_16)) - DRM_ERROR("POLARIS10/11 UVD firmware version %hu.%hu is too old.\n", - version_major, version_minor); + version_major = (le32_to_cpu(hdr->ucode_version) >> 24) & 0xff; + version_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff; + DRM_INFO("Found UVD firmware Version: %hu.%hu Family ID: %hu\n", + version_major, version_minor, family_id); + + /* + * Limit the number of UVD handles depending on microcode major + * and minor versions. The firmware version which has 40 UVD + * instances support is 1.80. So all subsequent versions should + * also have the same support. + */ + if ((version_major > 0x01) || + ((version_major == 0x01) && (version_minor >= 0x50))) + adev->uvd.max_handles = AMDGPU_MAX_UVD_HANDLES; + + adev->uvd.fw_version = ((version_major << 24) | (version_minor << 16) | + (family_id << 8)); + + if ((adev->asic_type == CHIP_POLARIS10 || + adev->asic_type == CHIP_POLARIS11) && + (adev->uvd.fw_version < FW_1_66_16)) + DRM_ERROR("POLARIS10/11 UVD firmware version %hu.%hu is too old.\n", + version_major, version_minor); + } else { + unsigned int enc_major, enc_minor, dec_minor; + + dec_minor = (le32_to_cpu(hdr->ucode_version) >> 8) & 0xff; + enc_minor = (le32_to_cpu(hdr->ucode_version) >> 24) & 0x3f; + enc_major = (le32_to_cpu(hdr->ucode_version) >> 30) & 0x3; + DRM_INFO("Found UVD firmware ENC: %hu.%hu DEC: .%hu Family ID: %hu\n", + enc_major, enc_minor, dec_minor, family_id); + + adev->uvd.max_handles = AMDGPU_MAX_UVD_HANDLES; + + adev->uvd.fw_version = le32_to_cpu(hdr->ucode_version); + } bo_size = AMDGPU_UVD_STACK_SIZE + AMDGPU_UVD_HEAP_SIZE + AMDGPU_UVD_SESSION_SIZE * adev->uvd.max_handles; @@ -238,7 +253,8 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); for (j = 0; j < adev->uvd.num_uvd_inst; j++) { - + if (adev->uvd.harvest_config & (1 << j)) + continue; r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, &adev->uvd.inst[j].vcpu_bo, &adev->uvd.inst[j].gpu_addr, &adev->uvd.inst[j].cpu_addr); @@ -246,21 +262,13 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) dev_err(adev->dev, "(%d) failed to allocate UVD bo\n", r); return r; } + } - ring = &adev->uvd.inst[j].ring; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&ring->sched, &adev->uvd.inst[j].entity, - rq, NULL); - if (r != 0) { - DRM_ERROR("Failed setting up UVD(%d) run queue.\n", j); - return r; - } - - for (i = 0; i < adev->uvd.max_handles; ++i) { - atomic_set(&adev->uvd.inst[j].handles[i], 0); - adev->uvd.inst[j].filp[i] = NULL; - } + for (i = 0; i < adev->uvd.max_handles; ++i) { + atomic_set(&adev->uvd.handles[i], 0); + adev->uvd.filp[i] = NULL; } + /* from uvd v5.0 HW addressing capacity increased to 64 bits */ if (!amdgpu_device_ip_block_version_cmp(adev, AMD_IP_BLOCK_TYPE_UVD, 5, 0)) adev->uvd.address_64_bit = true; @@ -289,10 +297,12 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) { int i, j; - for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { - kfree(adev->uvd.inst[j].saved_bo); + drm_sched_entity_destroy(&adev->uvd.entity); - drm_sched_entity_fini(&adev->uvd.inst[j].ring.sched, &adev->uvd.inst[j].entity); + for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.harvest_config & (1 << j)) + continue; + kvfree(adev->uvd.inst[j].saved_bo); amdgpu_bo_free_kernel(&adev->uvd.inst[j].vcpu_bo, &adev->uvd.inst[j].gpu_addr, @@ -308,6 +318,29 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) return 0; } +/** + * amdgpu_uvd_entity_init - init entity + * + * @adev: amdgpu_device pointer + * + */ +int amdgpu_uvd_entity_init(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + struct drm_sched_rq *rq; + int r; + + ring = &adev->uvd.inst[0].ring; + rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; + r = drm_sched_entity_init(&adev->uvd.entity, &rq, 1, NULL); + if (r) { + DRM_ERROR("Failed setting up UVD kernel entity.\n"); + return r; + } + + return 0; +} + int amdgpu_uvd_suspend(struct amdgpu_device *adev) { unsigned size; @@ -316,24 +349,26 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev) cancel_delayed_work_sync(&adev->uvd.idle_work); + /* only valid for physical mode */ + if (adev->asic_type < CHIP_POLARIS10) { + for (i = 0; i < adev->uvd.max_handles; ++i) + if (atomic_read(&adev->uvd.handles[i])) + break; + + if (i == adev->uvd.max_handles) + return 0; + } + for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.harvest_config & (1 << j)) + continue; if (adev->uvd.inst[j].vcpu_bo == NULL) continue; - /* only valid for physical mode */ - if (adev->asic_type < CHIP_POLARIS10) { - for (i = 0; i < adev->uvd.max_handles; ++i) - if (atomic_read(&adev->uvd.inst[j].handles[i])) - break; - - if (i == adev->uvd.max_handles) - continue; - } - size = amdgpu_bo_size(adev->uvd.inst[j].vcpu_bo); ptr = adev->uvd.inst[j].cpu_addr; - adev->uvd.inst[j].saved_bo = kmalloc(size, GFP_KERNEL); + adev->uvd.inst[j].saved_bo = kvmalloc(size, GFP_KERNEL); if (!adev->uvd.inst[j].saved_bo) return -ENOMEM; @@ -349,6 +384,8 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; if (adev->uvd.inst[i].vcpu_bo == NULL) return -EINVAL; @@ -357,7 +394,7 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) if (adev->uvd.inst[i].saved_bo != NULL) { memcpy_toio(ptr, adev->uvd.inst[i].saved_bo, size); - kfree(adev->uvd.inst[i].saved_bo); + kvfree(adev->uvd.inst[i].saved_bo); adev->uvd.inst[i].saved_bo = NULL; } else { const struct common_firmware_header *hdr; @@ -381,30 +418,27 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp) { - struct amdgpu_ring *ring; - int i, j, r; - - for (j = 0; j < adev->uvd.num_uvd_inst; j++) { - ring = &adev->uvd.inst[j].ring; + struct amdgpu_ring *ring = &adev->uvd.inst[0].ring; + int i, r; - for (i = 0; i < adev->uvd.max_handles; ++i) { - uint32_t handle = atomic_read(&adev->uvd.inst[j].handles[i]); - if (handle != 0 && adev->uvd.inst[j].filp[i] == filp) { - struct dma_fence *fence; - - r = amdgpu_uvd_get_destroy_msg(ring, handle, - false, &fence); - if (r) { - DRM_ERROR("Error destroying UVD(%d) %d!\n", j, r); - continue; - } + for (i = 0; i < adev->uvd.max_handles; ++i) { + uint32_t handle = atomic_read(&adev->uvd.handles[i]); - dma_fence_wait(fence, false); - dma_fence_put(fence); + if (handle != 0 && adev->uvd.filp[i] == filp) { + struct dma_fence *fence; - adev->uvd.inst[j].filp[i] = NULL; - atomic_set(&adev->uvd.inst[j].handles[i], 0); + r = amdgpu_uvd_get_destroy_msg(ring, handle, false, + &fence); + if (r) { + DRM_ERROR("Error destroying UVD %d!\n", r); + continue; } + + dma_fence_wait(fence, false); + dma_fence_put(fence); + + adev->uvd.filp[i] = NULL; + atomic_set(&adev->uvd.handles[i], 0); } } } @@ -459,7 +493,7 @@ static int amdgpu_uvd_cs_pass1(struct amdgpu_uvd_cs_ctx *ctx) if (cmd == 0x0 || cmd == 0x3) { /* yes, force it into VRAM */ uint32_t domain = AMDGPU_GEM_DOMAIN_VRAM; - amdgpu_ttm_placement_from_domain(bo, domain); + amdgpu_bo_placement_from_domain(bo, domain); } amdgpu_uvd_force_into_uvd_segment(bo); @@ -679,16 +713,15 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, void *ptr; long r; int i; - uint32_t ip_instance = ctx->parser->job->ring->me; if (offset & 0x3F) { - DRM_ERROR("UVD(%d) messages must be 64 byte aligned!\n", ip_instance); + DRM_ERROR("UVD messages must be 64 byte aligned!\n"); return -EINVAL; } r = amdgpu_bo_kmap(bo, &ptr); if (r) { - DRM_ERROR("Failed mapping the UVD(%d) message (%ld)!\n", ip_instance, r); + DRM_ERROR("Failed mapping the UVD) message (%ld)!\n", r); return r; } @@ -698,7 +731,7 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, handle = msg[2]; if (handle == 0) { - DRM_ERROR("Invalid UVD(%d) handle!\n", ip_instance); + DRM_ERROR("Invalid UVD handle!\n"); return -EINVAL; } @@ -709,18 +742,19 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, /* try to alloc a new handle */ for (i = 0; i < adev->uvd.max_handles; ++i) { - if (atomic_read(&adev->uvd.inst[ip_instance].handles[i]) == handle) { - DRM_ERROR("(%d)Handle 0x%x already in use!\n", ip_instance, handle); + if (atomic_read(&adev->uvd.handles[i]) == handle) { + DRM_ERROR(")Handle 0x%x already in use!\n", + handle); return -EINVAL; } - if (!atomic_cmpxchg(&adev->uvd.inst[ip_instance].handles[i], 0, handle)) { - adev->uvd.inst[ip_instance].filp[i] = ctx->parser->filp; + if (!atomic_cmpxchg(&adev->uvd.handles[i], 0, handle)) { + adev->uvd.filp[i] = ctx->parser->filp; return 0; } } - DRM_ERROR("No more free UVD(%d) handles!\n", ip_instance); + DRM_ERROR("No more free UVD handles!\n"); return -ENOSPC; case 1: @@ -732,27 +766,27 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, /* validate the handle */ for (i = 0; i < adev->uvd.max_handles; ++i) { - if (atomic_read(&adev->uvd.inst[ip_instance].handles[i]) == handle) { - if (adev->uvd.inst[ip_instance].filp[i] != ctx->parser->filp) { - DRM_ERROR("UVD(%d) handle collision detected!\n", ip_instance); + if (atomic_read(&adev->uvd.handles[i]) == handle) { + if (adev->uvd.filp[i] != ctx->parser->filp) { + DRM_ERROR("UVD handle collision detected!\n"); return -EINVAL; } return 0; } } - DRM_ERROR("Invalid UVD(%d) handle 0x%x!\n", ip_instance, handle); + DRM_ERROR("Invalid UVD handle 0x%x!\n", handle); return -ENOENT; case 2: /* it's a destroy msg, free the handle */ for (i = 0; i < adev->uvd.max_handles; ++i) - atomic_cmpxchg(&adev->uvd.inst[ip_instance].handles[i], handle, 0); + atomic_cmpxchg(&adev->uvd.handles[i], handle, 0); amdgpu_bo_kunmap(bo); return 0; default: - DRM_ERROR("Illegal UVD(%d) message type (%d)!\n", ip_instance, msg_type); + DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type); return -EINVAL; } BUG(); @@ -1000,7 +1034,7 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, if (!ring->adev->uvd.address_64_bit) { struct ttm_operation_ctx ctx = { true, false }; - amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM); + amdgpu_bo_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_VRAM); amdgpu_uvd_force_into_uvd_segment(bo); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); if (r) @@ -1045,19 +1079,16 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, if (r < 0) goto err_free; - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err_free; - - amdgpu_job_free(job); } else { r = amdgpu_sync_resv(adev, &job->sync, bo->tbo.resv, AMDGPU_FENCE_OWNER_UNDEFINED, false); if (r) goto err_free; - r = amdgpu_job_submit(job, ring, &adev->uvd.inst[ring->me].entity, + r = amdgpu_job_submit(job, &adev->uvd.entity, AMDGPU_FENCE_OWNER_UNDEFINED, &f); if (r) goto err_free; @@ -1149,6 +1180,8 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work) unsigned fences = 0, i, j; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring); for (j = 0; j < adev->uvd.num_enc_rings; ++j) { fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]); @@ -1259,7 +1292,7 @@ uint32_t amdgpu_uvd_used_handles(struct amdgpu_device *adev) * necessarily linear. So we need to count * all non-zero handles. */ - if (atomic_read(&adev->uvd.inst->handles[i])) + if (atomic_read(&adev->uvd.handles[i])) used_handles++; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h index 8b23a1b00c76..a3ab1a41060f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h @@ -42,30 +42,34 @@ struct amdgpu_uvd_inst { void *cpu_addr; uint64_t gpu_addr; void *saved_bo; - atomic_t handles[AMDGPU_MAX_UVD_HANDLES]; - struct drm_file *filp[AMDGPU_MAX_UVD_HANDLES]; struct amdgpu_ring ring; struct amdgpu_ring ring_enc[AMDGPU_MAX_UVD_ENC_RINGS]; struct amdgpu_irq_src irq; - struct drm_sched_entity entity; - struct drm_sched_entity entity_enc; uint32_t srbm_soft_reset; }; +#define AMDGPU_UVD_HARVEST_UVD0 (1 << 0) +#define AMDGPU_UVD_HARVEST_UVD1 (1 << 1) + struct amdgpu_uvd { const struct firmware *fw; /* UVD firmware */ unsigned fw_version; unsigned max_handles; unsigned num_enc_rings; - uint8_t num_uvd_inst; + uint8_t num_uvd_inst; bool address_64_bit; bool use_ctx_buf; - struct amdgpu_uvd_inst inst[AMDGPU_MAX_UVD_INSTANCES]; + struct amdgpu_uvd_inst inst[AMDGPU_MAX_UVD_INSTANCES]; + struct drm_file *filp[AMDGPU_MAX_UVD_HANDLES]; + atomic_t handles[AMDGPU_MAX_UVD_HANDLES]; + struct drm_sched_entity entity; struct delayed_work idle_work; + unsigned harvest_config; }; int amdgpu_uvd_sw_init(struct amdgpu_device *adev); int amdgpu_uvd_sw_fini(struct amdgpu_device *adev); +int amdgpu_uvd_entity_init(struct amdgpu_device *adev); int amdgpu_uvd_suspend(struct amdgpu_device *adev); int amdgpu_uvd_resume(struct amdgpu_device *adev); int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index 23d960ec1cf2..5f3f54073818 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -40,11 +40,11 @@ /* Firmware Names */ #ifdef CONFIG_DRM_AMDGPU_CIK -#define FIRMWARE_BONAIRE "radeon/bonaire_vce.bin" -#define FIRMWARE_KABINI "radeon/kabini_vce.bin" -#define FIRMWARE_KAVERI "radeon/kaveri_vce.bin" -#define FIRMWARE_HAWAII "radeon/hawaii_vce.bin" -#define FIRMWARE_MULLINS "radeon/mullins_vce.bin" +#define FIRMWARE_BONAIRE "amdgpu/bonaire_vce.bin" +#define FIRMWARE_KABINI "amdgpu/kabini_vce.bin" +#define FIRMWARE_KAVERI "amdgpu/kaveri_vce.bin" +#define FIRMWARE_HAWAII "amdgpu/hawaii_vce.bin" +#define FIRMWARE_MULLINS "amdgpu/mullins_vce.bin" #endif #define FIRMWARE_TONGA "amdgpu/tonga_vce.bin" #define FIRMWARE_CARRIZO "amdgpu/carrizo_vce.bin" @@ -90,8 +90,6 @@ static void amdgpu_vce_idle_work_handler(struct work_struct *work); */ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size) { - struct amdgpu_ring *ring; - struct drm_sched_rq *rq; const char *fw_name; const struct common_firmware_header *hdr; unsigned ucode_version, version_major, version_minor, binary_id; @@ -188,15 +186,6 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size) return r; } - ring = &adev->vce.ring[0]; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&ring->sched, &adev->vce.entity, - rq, NULL); - if (r != 0) { - DRM_ERROR("Failed setting up VCE run queue.\n"); - return r; - } - for (i = 0; i < AMDGPU_MAX_VCE_HANDLES; ++i) { atomic_set(&adev->vce.handles[i], 0); adev->vce.filp[i] = NULL; @@ -222,7 +211,7 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev) if (adev->vce.vcpu_bo == NULL) return 0; - drm_sched_entity_fini(&adev->vce.ring[0].sched, &adev->vce.entity); + drm_sched_entity_destroy(&adev->vce.entity); amdgpu_bo_free_kernel(&adev->vce.vcpu_bo, &adev->vce.gpu_addr, (void **)&adev->vce.cpu_addr); @@ -237,6 +226,29 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev) } /** + * amdgpu_vce_entity_init - init entity + * + * @adev: amdgpu_device pointer + * + */ +int amdgpu_vce_entity_init(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring; + struct drm_sched_rq *rq; + int r; + + ring = &adev->vce.ring[0]; + rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; + r = drm_sched_entity_init(&adev->vce.entity, &rq, 1, NULL); + if (r != 0) { + DRM_ERROR("Failed setting up VCE run queue.\n"); + return r; + } + + return 0; +} + +/** * amdgpu_vce_suspend - unpin VCE fw memory * * @adev: amdgpu_device pointer @@ -246,6 +258,8 @@ int amdgpu_vce_suspend(struct amdgpu_device *adev) { int i; + cancel_delayed_work_sync(&adev->vce.idle_work); + if (adev->vce.vcpu_bo == NULL) return 0; @@ -256,7 +270,6 @@ int amdgpu_vce_suspend(struct amdgpu_device *adev) if (i == AMDGPU_MAX_VCE_HANDLES) return 0; - cancel_delayed_work_sync(&adev->vce.idle_work); /* TODO: suspending running encoding sessions isn't supported */ return -EINVAL; } @@ -470,12 +483,10 @@ int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; - amdgpu_job_free(job); if (fence) *fence = dma_fence_get(f); dma_fence_put(f); @@ -532,19 +543,13 @@ int amdgpu_vce_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - if (direct) { - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); - if (r) - goto err; - - amdgpu_job_free(job); - } else { - r = amdgpu_job_submit(job, ring, &ring->adev->vce.entity, + if (direct) + r = amdgpu_job_submit_direct(job, ring, &f); + else + r = amdgpu_job_submit(job, &ring->adev->vce.entity, AMDGPU_FENCE_OWNER_UNDEFINED, &f); - if (r) - goto err; - } + if (r) + goto err; if (fence) *fence = dma_fence_get(f); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h index 71781267ee4c..a1f209eed4c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h @@ -55,6 +55,7 @@ struct amdgpu_vce { int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size); int amdgpu_vce_sw_fini(struct amdgpu_device *adev); +int amdgpu_vce_entity_init(struct amdgpu_device *adev); int amdgpu_vce_suspend(struct amdgpu_device *adev); int amdgpu_vce_resume(struct amdgpu_device *adev); int amdgpu_vce_get_create_msg(struct amdgpu_ring *ring, uint32_t handle, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index 1b4ad9b2a755..27da13df2f11 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -36,14 +36,19 @@ #include "soc15_common.h" #include "vcn/vcn_1_0_offset.h" +#include "vcn/vcn_1_0_sh_mask.h" /* 1 second timeout */ #define VCN_IDLE_TIMEOUT msecs_to_jiffies(1000) /* Firmware Names */ #define FIRMWARE_RAVEN "amdgpu/raven_vcn.bin" +#define FIRMWARE_PICASSO "amdgpu/picasso_vcn.bin" +#define FIRMWARE_RAVEN2 "amdgpu/raven2_vcn.bin" MODULE_FIRMWARE(FIRMWARE_RAVEN); +MODULE_FIRMWARE(FIRMWARE_PICASSO); +MODULE_FIRMWARE(FIRMWARE_RAVEN2); static void amdgpu_vcn_idle_work_handler(struct work_struct *work); @@ -59,7 +64,12 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_RAVEN: - fw_name = FIRMWARE_RAVEN; + if (adev->rev_id >= 8) + fw_name = FIRMWARE_RAVEN2; + else if (adev->pdev->device == 0x15d8) + fw_name = FIRMWARE_PICASSO; + else + fw_name = FIRMWARE_RAVEN; break; default: return -EINVAL; @@ -111,9 +121,9 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) version_major, version_minor, family_id); } - bo_size = AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8) - + AMDGPU_VCN_STACK_SIZE + AMDGPU_VCN_HEAP_SIZE - + AMDGPU_VCN_SESSION_SIZE * 40; + bo_size = AMDGPU_VCN_STACK_SIZE + AMDGPU_VCN_CONTEXT_SIZE; + if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) + bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, &adev->vcn.vcpu_bo, &adev->vcn.gpu_addr, &adev->vcn.cpu_addr); @@ -129,7 +139,7 @@ int amdgpu_vcn_sw_fini(struct amdgpu_device *adev) { int i; - kfree(adev->vcn.saved_bo); + kvfree(adev->vcn.saved_bo); amdgpu_bo_free_kernel(&adev->vcn.vcpu_bo, &adev->vcn.gpu_addr, @@ -140,6 +150,8 @@ int amdgpu_vcn_sw_fini(struct amdgpu_device *adev) for (i = 0; i < adev->vcn.num_enc_rings; ++i) amdgpu_ring_fini(&adev->vcn.ring_enc[i]); + amdgpu_ring_fini(&adev->vcn.ring_jpeg); + release_firmware(adev->vcn.fw); return 0; @@ -150,15 +162,15 @@ int amdgpu_vcn_suspend(struct amdgpu_device *adev) unsigned size; void *ptr; + cancel_delayed_work_sync(&adev->vcn.idle_work); + if (adev->vcn.vcpu_bo == NULL) return 0; - cancel_delayed_work_sync(&adev->vcn.idle_work); - size = amdgpu_bo_size(adev->vcn.vcpu_bo); ptr = adev->vcn.cpu_addr; - adev->vcn.saved_bo = kmalloc(size, GFP_KERNEL); + adev->vcn.saved_bo = kvmalloc(size, GFP_KERNEL); if (!adev->vcn.saved_bo) return -ENOMEM; @@ -180,36 +192,184 @@ int amdgpu_vcn_resume(struct amdgpu_device *adev) if (adev->vcn.saved_bo != NULL) { memcpy_toio(ptr, adev->vcn.saved_bo, size); - kfree(adev->vcn.saved_bo); + kvfree(adev->vcn.saved_bo); adev->vcn.saved_bo = NULL; } else { const struct common_firmware_header *hdr; unsigned offset; hdr = (const struct common_firmware_header *)adev->vcn.fw->data; - offset = le32_to_cpu(hdr->ucode_array_offset_bytes); - memcpy_toio(adev->vcn.cpu_addr, adev->vcn.fw->data + offset, - le32_to_cpu(hdr->ucode_size_bytes)); - size -= le32_to_cpu(hdr->ucode_size_bytes); - ptr += le32_to_cpu(hdr->ucode_size_bytes); + if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { + offset = le32_to_cpu(hdr->ucode_array_offset_bytes); + memcpy_toio(adev->vcn.cpu_addr, adev->vcn.fw->data + offset, + le32_to_cpu(hdr->ucode_size_bytes)); + size -= le32_to_cpu(hdr->ucode_size_bytes); + ptr += le32_to_cpu(hdr->ucode_size_bytes); + } memset_io(ptr, 0, size); } return 0; } +static int amdgpu_vcn_pause_dpg_mode(struct amdgpu_device *adev, + struct dpg_pause_state *new_state) +{ + int ret_code; + uint32_t reg_data = 0; + uint32_t reg_data2 = 0; + struct amdgpu_ring *ring; + + /* pause/unpause if state is changed */ + if (adev->vcn.pause_state.fw_based != new_state->fw_based) { + DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d", + adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg, + new_state->fw_based, new_state->jpeg); + + reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) & + (~UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK); + + if (new_state->fw_based == VCN_DPG_STATE__PAUSE) { + ret_code = 0; + + if (!(reg_data & UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK)) + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); + + if (!ret_code) { + /* pause DPG non-jpeg */ + reg_data |= UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; + WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE, + UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK, + UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK, ret_code); + + /* Restore */ + ring = &adev->vcn.ring_enc[0]; + WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr); + WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4); + WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); + + ring = &adev->vcn.ring_enc[1]; + WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr); + WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4); + WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); + WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); + + ring = &adev->vcn.ring_dec; + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, + RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2)); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, + UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); + } + } else { + /* unpause dpg non-jpeg, no need to wait */ + reg_data &= ~UVD_DPG_PAUSE__NJ_PAUSE_DPG_REQ_MASK; + WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); + } + adev->vcn.pause_state.fw_based = new_state->fw_based; + } + + /* pause/unpause if state is changed */ + if (adev->vcn.pause_state.jpeg != new_state->jpeg) { + DRM_DEBUG("dpg pause state changed %d:%d -> %d:%d", + adev->vcn.pause_state.fw_based, adev->vcn.pause_state.jpeg, + new_state->fw_based, new_state->jpeg); + + reg_data = RREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE) & + (~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK); + + if (new_state->jpeg == VCN_DPG_STATE__PAUSE) { + ret_code = 0; + + if (!(reg_data & UVD_DPG_PAUSE__NJ_PAUSE_DPG_ACK_MASK)) + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); + + if (!ret_code) { + /* Make sure JPRG Snoop is disabled before sending the pause */ + reg_data2 = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS); + reg_data2 |= UVD_POWER_STATUS__JRBC_SNOOP_DIS_MASK; + WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, reg_data2); + + /* pause DPG jpeg */ + reg_data |= UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK; + WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_DPG_PAUSE, + UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK, + UVD_DPG_PAUSE__JPEG_PAUSE_DPG_ACK_MASK, ret_code); + + /* Restore */ + ring = &adev->vcn.ring_jpeg; + WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, + UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK | + UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); + WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW, + lower_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH, + upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, ring->wptr); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, ring->wptr); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, + UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); + + ring = &adev->vcn.ring_dec; + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, + RREG32_SOC15(UVD, 0, mmUVD_SCRATCH2)); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, + UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); + } + } else { + /* unpause dpg jpeg, no need to wait */ + reg_data &= ~UVD_DPG_PAUSE__JPEG_PAUSE_DPG_REQ_MASK; + WREG32_SOC15(UVD, 0, mmUVD_DPG_PAUSE, reg_data); + } + adev->vcn.pause_state.jpeg = new_state->jpeg; + } + + return 0; +} + static void amdgpu_vcn_idle_work_handler(struct work_struct *work) { struct amdgpu_device *adev = container_of(work, struct amdgpu_device, vcn.idle_work.work); - unsigned fences = amdgpu_fence_count_emitted(&adev->vcn.ring_dec); - unsigned i; + unsigned int fences = 0; + unsigned int i; for (i = 0; i < adev->vcn.num_enc_rings; ++i) { fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]); } + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + struct dpg_pause_state new_state; + + if (fences) + new_state.fw_based = VCN_DPG_STATE__PAUSE; + else + new_state.fw_based = VCN_DPG_STATE__UNPAUSE; + + if (amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg)) + new_state.jpeg = VCN_DPG_STATE__PAUSE; + else + new_state.jpeg = VCN_DPG_STATE__UNPAUSE; + + amdgpu_vcn_pause_dpg_mode(adev, &new_state); + } + + fences += amdgpu_fence_count_emitted(&adev->vcn.ring_jpeg); + fences += amdgpu_fence_count_emitted(&adev->vcn.ring_dec); + if (fences == 0) { + amdgpu_gfx_off_ctrl(adev, true); if (adev->pm.dpm_enabled) amdgpu_dpm_enable_uvd(adev, false); else @@ -225,13 +385,30 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) struct amdgpu_device *adev = ring->adev; bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work); - if (set_clocks && adev->pm.dpm_enabled) { + if (set_clocks) { + amdgpu_gfx_off_ctrl(adev, false); if (adev->pm.dpm_enabled) amdgpu_dpm_enable_uvd(adev, true); else amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, AMD_PG_STATE_UNGATE); } + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) { + struct dpg_pause_state new_state; + + if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_ENC) + new_state.fw_based = VCN_DPG_STATE__PAUSE; + else + new_state.fw_based = adev->vcn.pause_state.fw_based; + + if (ring->funcs->type == AMDGPU_RING_TYPE_VCN_JPEG) + new_state.jpeg = VCN_DPG_STATE__PAUSE; + else + new_state.jpeg = adev->vcn.pause_state.jpeg; + + amdgpu_vcn_pause_dpg_mode(adev, &new_state); + } } void amdgpu_vcn_ring_end_use(struct amdgpu_ring *ring) @@ -246,7 +423,7 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) unsigned i; int r; - WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0xCAFEDEAD); + WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) { DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", @@ -254,11 +431,11 @@ int amdgpu_vcn_dec_ring_test_ring(struct amdgpu_ring *ring) return r; } amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0)); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID)); + tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9)); if (tmp == 0xDEADBEEF) break; DRM_UDELAY(1); @@ -304,13 +481,10 @@ static int amdgpu_vcn_dec_send_msg(struct amdgpu_ring *ring, } ib->length_dw = 16; - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err_free; - amdgpu_job_free(job); - amdgpu_bo_fence(bo, f, false); amdgpu_bo_unreserve(bo); amdgpu_bo_unref(&bo); @@ -495,12 +669,10 @@ static int amdgpu_vcn_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t hand for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; - amdgpu_job_free(job); if (fence) *fence = dma_fence_get(f); dma_fence_put(f); @@ -549,12 +721,10 @@ static int amdgpu_vcn_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t han for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; - amdgpu_job_free(job); if (fence) *fence = dma_fence_get(f); dma_fence_put(f); @@ -597,3 +767,127 @@ error: dma_fence_put(fence); return r; } + +int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t tmp = 0; + unsigned i; + int r; + + WREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0xCAFEDEAD); + r = amdgpu_ring_alloc(ring, 3); + + if (r) { + DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", + ring->idx, r); + return r; + } + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, 0)); + amdgpu_ring_write(ring, 0xDEADBEEF); + amdgpu_ring_commit(ring); + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9)); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + + if (i < adev->usec_timeout) { + DRM_DEBUG("ring test on %d succeeded in %d usecs\n", + ring->idx, i); + } else { + DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", + ring->idx, tmp); + r = -EINVAL; + } + + return r; +} + +static int amdgpu_vcn_jpeg_set_reg(struct amdgpu_ring *ring, uint32_t handle, + struct dma_fence **fence) +{ + struct amdgpu_device *adev = ring->adev; + struct amdgpu_job *job; + struct amdgpu_ib *ib; + struct dma_fence *f = NULL; + const unsigned ib_size_dw = 16; + int i, r; + + r = amdgpu_job_alloc_with_ib(ring->adev, ib_size_dw * 4, &job); + if (r) + return r; + + ib = &job->ibs[0]; + + ib->ptr[0] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9), 0, 0, PACKETJ_TYPE0); + ib->ptr[1] = 0xDEADBEEF; + for (i = 2; i < 16; i += 2) { + ib->ptr[i] = PACKETJ(0, 0, 0, PACKETJ_TYPE6); + ib->ptr[i+1] = 0; + } + ib->length_dw = 16; + + r = amdgpu_job_submit_direct(job, ring, &f); + if (r) + goto err; + + if (fence) + *fence = dma_fence_get(f); + dma_fence_put(f); + + return 0; + +err: + amdgpu_job_free(job); + return r; +} + +int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t tmp = 0; + unsigned i; + struct dma_fence *fence = NULL; + long r = 0; + + r = amdgpu_vcn_jpeg_set_reg(ring, 1, &fence); + if (r) { + DRM_ERROR("amdgpu: failed to set jpeg register (%ld).\n", r); + goto error; + } + + r = dma_fence_wait_timeout(fence, false, timeout); + if (r == 0) { + DRM_ERROR("amdgpu: IB test timed out.\n"); + r = -ETIMEDOUT; + goto error; + } else if (r < 0) { + DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); + goto error; + } else + r = 0; + + for (i = 0; i < adev->usec_timeout; i++) { + tmp = RREG32(SOC15_REG_OFFSET(UVD, 0, mmUVD_SCRATCH9)); + if (tmp == 0xDEADBEEF) + break; + DRM_UDELAY(1); + } + + if (i < adev->usec_timeout) + DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + else { + DRM_ERROR("ib test failed (0x%08X)\n", tmp); + r = -EINVAL; + } + + dma_fence_put(fence); + +error: + return r; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h index 773010b9ff15..a0ad19af9080 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h @@ -24,9 +24,9 @@ #ifndef __AMDGPU_VCN_H__ #define __AMDGPU_VCN_H__ -#define AMDGPU_VCN_STACK_SIZE (200*1024) -#define AMDGPU_VCN_HEAP_SIZE (256*1024) -#define AMDGPU_VCN_SESSION_SIZE (50*1024) +#define AMDGPU_VCN_STACK_SIZE (128*1024) +#define AMDGPU_VCN_CONTEXT_SIZE (512*1024) + #define AMDGPU_VCN_FIRMWARE_OFFSET 256 #define AMDGPU_VCN_MAX_ENC_RINGS 3 @@ -56,6 +56,16 @@ enum engine_status_constants { UVD_STATUS__RBC_BUSY = 0x1, }; +enum internal_dpg_state { + VCN_DPG_STATE__UNPAUSE = 0, + VCN_DPG_STATE__PAUSE, +}; + +struct dpg_pause_state { + enum internal_dpg_state fw_based; + enum internal_dpg_state jpeg; +}; + struct amdgpu_vcn { struct amdgpu_bo *vcpu_bo; void *cpu_addr; @@ -66,8 +76,11 @@ struct amdgpu_vcn { const struct firmware *fw; /* VCN firmware */ struct amdgpu_ring ring_dec; struct amdgpu_ring ring_enc[AMDGPU_VCN_MAX_ENC_RINGS]; + struct amdgpu_ring ring_jpeg; struct amdgpu_irq_src irq; unsigned num_enc_rings; + enum amd_powergating_state cur_state; + struct dpg_pause_state pause_state; }; int amdgpu_vcn_sw_init(struct amdgpu_device *adev); @@ -83,4 +96,7 @@ int amdgpu_vcn_dec_ring_test_ib(struct amdgpu_ring *ring, long timeout); int amdgpu_vcn_enc_ring_test_ring(struct amdgpu_ring *ring); int amdgpu_vcn_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout); +int amdgpu_vcn_jpeg_ring_test_ring(struct amdgpu_ring *ring); +int amdgpu_vcn_jpeg_ring_test_ib(struct amdgpu_ring *ring, long timeout); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index 21adb1b6e5cb..f2f358aa0597 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -22,18 +22,13 @@ */ #include "amdgpu.h" -#define MAX_KIQ_REG_WAIT 5000 /* in usecs, 5ms */ -#define MAX_KIQ_REG_BAILOUT_INTERVAL 5 /* in msecs, 5ms */ -#define MAX_KIQ_REG_TRY 20 uint64_t amdgpu_csa_vaddr(struct amdgpu_device *adev) { uint64_t addr = adev->vm_manager.max_pfn << AMDGPU_GPU_PAGE_SHIFT; addr -= AMDGPU_VA_RESERVED_SIZE; - - if (addr >= AMDGPU_VA_HOLE_START) - addr |= AMDGPU_VA_HOLE_END; + addr = amdgpu_gmc_sign_extend(addr); return addr; } @@ -76,7 +71,7 @@ void amdgpu_free_static_csa(struct amdgpu_device *adev) { int amdgpu_map_static_csa(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct amdgpu_bo_va **bo_va) { - uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_VA_HOLE_MASK; + uint64_t csa_addr = amdgpu_csa_vaddr(adev) & AMDGPU_GMC_HOLE_MASK; struct ww_acquire_ctx ticket; struct list_head list; struct amdgpu_bo_list_entry pd; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index fdcb498f6d19..6904d794d60a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -33,9 +33,11 @@ #include "amdgpu.h" #include "amdgpu_trace.h" #include "amdgpu_amdkfd.h" +#include "amdgpu_gmc.h" -/* - * GPUVM +/** + * DOC: GPUVM + * * GPUVM is similar to the legacy gart on older asics, however * rather than there being a single global gart table * for the entire GPU, there are multiple VM page tables active @@ -63,74 +65,82 @@ INTERVAL_TREE_DEFINE(struct amdgpu_bo_va_mapping, rb, uint64_t, __subtree_last, #undef START #undef LAST -/* Local structure. Encapsulate some VM table update parameters to reduce +/** + * struct amdgpu_pte_update_params - Local structure + * + * Encapsulate some VM table update parameters to reduce * the number of function parameters + * */ struct amdgpu_pte_update_params { - /* amdgpu device we do this update for */ + + /** + * @adev: amdgpu device we do this update for + */ struct amdgpu_device *adev; - /* optional amdgpu_vm we do this update for */ + + /** + * @vm: optional amdgpu_vm we do this update for + */ struct amdgpu_vm *vm; - /* address where to copy page table entries from */ + + /** + * @src: address where to copy page table entries from + */ uint64_t src; - /* indirect buffer to fill with commands */ + + /** + * @ib: indirect buffer to fill with commands + */ struct amdgpu_ib *ib; - /* Function which actually does the update */ + + /** + * @func: Function which actually does the update + */ void (*func)(struct amdgpu_pte_update_params *params, struct amdgpu_bo *bo, uint64_t pe, uint64_t addr, unsigned count, uint32_t incr, uint64_t flags); - /* The next two are used during VM update by CPU - * DMA addresses to use for mapping - * Kernel pointer of PD/PT BO that needs to be updated + /** + * @pages_addr: + * + * DMA addresses to use for mapping, used during VM update by CPU */ dma_addr_t *pages_addr; + + /** + * @kptr: + * + * Kernel pointer of PD/PT BO that needs to be updated, + * used during VM update by CPU + */ void *kptr; }; -/* Helper to disable partial resident texture feature from a fence callback */ +/** + * struct amdgpu_prt_cb - Helper to disable partial resident texture feature from a fence callback + */ struct amdgpu_prt_cb { + + /** + * @adev: amdgpu device + */ struct amdgpu_device *adev; + + /** + * @cb: callback + */ struct dma_fence_cb cb; }; -static void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, - struct amdgpu_vm *vm, - struct amdgpu_bo *bo) -{ - base->vm = vm; - base->bo = bo; - INIT_LIST_HEAD(&base->bo_list); - INIT_LIST_HEAD(&base->vm_status); - - if (!bo) - return; - list_add_tail(&base->bo_list, &bo->va); - - if (bo->tbo.type == ttm_bo_type_kernel) - list_move(&base->vm_status, &vm->relocated); - - if (bo->tbo.resv != vm->root.base.bo->tbo.resv) - return; - - if (bo->preferred_domains & - amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type)) - return; - - /* - * we checked all the prerequisites, but it looks like this per vm bo - * is currently evicted. add the bo to the evicted list to make sure it - * is validated on next vm use to avoid fault. - * */ - list_move_tail(&base->vm_status, &vm->evicted); -} - /** * amdgpu_vm_level_shift - return the addr shift for each level * * @adev: amdgpu_device pointer + * @level: VMPT level * - * Returns the number of bits the pfn needs to be right shifted for a level. + * Returns: + * The number of bits the pfn needs to be right shifted for a level. */ static unsigned amdgpu_vm_level_shift(struct amdgpu_device *adev, unsigned level) @@ -158,8 +168,10 @@ static unsigned amdgpu_vm_level_shift(struct amdgpu_device *adev, * amdgpu_vm_num_entries - return the number of entries in a PD/PT * * @adev: amdgpu_device pointer + * @level: VMPT level * - * Calculate the number of entries in a page directory or page table. + * Returns: + * The number of entries in a page directory or page table. */ static unsigned amdgpu_vm_num_entries(struct amdgpu_device *adev, unsigned level) @@ -179,11 +191,33 @@ static unsigned amdgpu_vm_num_entries(struct amdgpu_device *adev, } /** + * amdgpu_vm_entries_mask - the mask to get the entry number of a PD/PT + * + * @adev: amdgpu_device pointer + * @level: VMPT level + * + * Returns: + * The mask to extract the entry number of a PD/PT from an address. + */ +static uint32_t amdgpu_vm_entries_mask(struct amdgpu_device *adev, + unsigned int level) +{ + if (level <= adev->vm_manager.root_level) + return 0xffffffff; + else if (level != AMDGPU_VM_PTB) + return 0x1ff; + else + return AMDGPU_VM_PTE_COUNT(adev) - 1; +} + +/** * amdgpu_vm_bo_size - returns the size of the BOs in bytes * * @adev: amdgpu_device pointer + * @level: VMPT level * - * Calculate the size of the BO for a page directory or page table in bytes. + * Returns: + * The size of the BO for a page directory or page table in bytes. */ static unsigned amdgpu_vm_bo_size(struct amdgpu_device *adev, unsigned level) { @@ -191,6 +225,382 @@ static unsigned amdgpu_vm_bo_size(struct amdgpu_device *adev, unsigned level) } /** + * amdgpu_vm_bo_evicted - vm_bo is evicted + * + * @vm_bo: vm_bo which is evicted + * + * State for PDs/PTs and per VM BOs which are not at the location they should + * be. + */ +static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo) +{ + struct amdgpu_vm *vm = vm_bo->vm; + struct amdgpu_bo *bo = vm_bo->bo; + + vm_bo->moved = true; + if (bo->tbo.type == ttm_bo_type_kernel) + list_move(&vm_bo->vm_status, &vm->evicted); + else + list_move_tail(&vm_bo->vm_status, &vm->evicted); +} + +/** + * amdgpu_vm_bo_relocated - vm_bo is reloacted + * + * @vm_bo: vm_bo which is relocated + * + * State for PDs/PTs which needs to update their parent PD. + */ +static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo) +{ + list_move(&vm_bo->vm_status, &vm_bo->vm->relocated); +} + +/** + * amdgpu_vm_bo_moved - vm_bo is moved + * + * @vm_bo: vm_bo which is moved + * + * State for per VM BOs which are moved, but that change is not yet reflected + * in the page tables. + */ +static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo) +{ + list_move(&vm_bo->vm_status, &vm_bo->vm->moved); +} + +/** + * amdgpu_vm_bo_idle - vm_bo is idle + * + * @vm_bo: vm_bo which is now idle + * + * State for PDs/PTs and per VM BOs which have gone through the state machine + * and are now idle. + */ +static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo) +{ + list_move(&vm_bo->vm_status, &vm_bo->vm->idle); + vm_bo->moved = false; +} + +/** + * amdgpu_vm_bo_invalidated - vm_bo is invalidated + * + * @vm_bo: vm_bo which is now invalidated + * + * State for normal BOs which are invalidated and that change not yet reflected + * in the PTs. + */ +static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo) +{ + spin_lock(&vm_bo->vm->invalidated_lock); + list_move(&vm_bo->vm_status, &vm_bo->vm->invalidated); + spin_unlock(&vm_bo->vm->invalidated_lock); +} + +/** + * amdgpu_vm_bo_done - vm_bo is done + * + * @vm_bo: vm_bo which is now done + * + * State for normal BOs which are invalidated and that change has been updated + * in the PTs. + */ +static void amdgpu_vm_bo_done(struct amdgpu_vm_bo_base *vm_bo) +{ + spin_lock(&vm_bo->vm->invalidated_lock); + list_del_init(&vm_bo->vm_status); + spin_unlock(&vm_bo->vm->invalidated_lock); +} + +/** + * amdgpu_vm_bo_base_init - Adds bo to the list of bos associated with the vm + * + * @base: base structure for tracking BO usage in a VM + * @vm: vm to which bo is to be added + * @bo: amdgpu buffer object + * + * Initialize a bo_va_base structure and add it to the appropriate lists + * + */ +static void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, + struct amdgpu_vm *vm, + struct amdgpu_bo *bo) +{ + base->vm = vm; + base->bo = bo; + base->next = NULL; + INIT_LIST_HEAD(&base->vm_status); + + if (!bo) + return; + base->next = bo->vm_bo; + bo->vm_bo = base; + + if (bo->tbo.resv != vm->root.base.bo->tbo.resv) + return; + + vm->bulk_moveable = false; + if (bo->tbo.type == ttm_bo_type_kernel) + amdgpu_vm_bo_relocated(base); + else + amdgpu_vm_bo_idle(base); + + if (bo->preferred_domains & + amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type)) + return; + + /* + * we checked all the prerequisites, but it looks like this per vm bo + * is currently evicted. add the bo to the evicted list to make sure it + * is validated on next vm use to avoid fault. + * */ + amdgpu_vm_bo_evicted(base); +} + +/** + * amdgpu_vm_pt_parent - get the parent page directory + * + * @pt: child page table + * + * Helper to get the parent entry for the child page table. NULL if we are at + * the root page directory. + */ +static struct amdgpu_vm_pt *amdgpu_vm_pt_parent(struct amdgpu_vm_pt *pt) +{ + struct amdgpu_bo *parent = pt->base.bo->parent; + + if (!parent) + return NULL; + + return container_of(parent->vm_bo, struct amdgpu_vm_pt, base); +} + +/** + * amdgpu_vm_pt_cursor - state for for_each_amdgpu_vm_pt + */ +struct amdgpu_vm_pt_cursor { + uint64_t pfn; + struct amdgpu_vm_pt *parent; + struct amdgpu_vm_pt *entry; + unsigned level; +}; + +/** + * amdgpu_vm_pt_start - start PD/PT walk + * + * @adev: amdgpu_device pointer + * @vm: amdgpu_vm structure + * @start: start address of the walk + * @cursor: state to initialize + * + * Initialize a amdgpu_vm_pt_cursor to start a walk. + */ +static void amdgpu_vm_pt_start(struct amdgpu_device *adev, + struct amdgpu_vm *vm, uint64_t start, + struct amdgpu_vm_pt_cursor *cursor) +{ + cursor->pfn = start; + cursor->parent = NULL; + cursor->entry = &vm->root; + cursor->level = adev->vm_manager.root_level; +} + +/** + * amdgpu_vm_pt_descendant - go to child node + * + * @adev: amdgpu_device pointer + * @cursor: current state + * + * Walk to the child node of the current node. + * Returns: + * True if the walk was possible, false otherwise. + */ +static bool amdgpu_vm_pt_descendant(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + unsigned mask, shift, idx; + + if (!cursor->entry->entries) + return false; + + BUG_ON(!cursor->entry->base.bo); + mask = amdgpu_vm_entries_mask(adev, cursor->level); + shift = amdgpu_vm_level_shift(adev, cursor->level); + + ++cursor->level; + idx = (cursor->pfn >> shift) & mask; + cursor->parent = cursor->entry; + cursor->entry = &cursor->entry->entries[idx]; + return true; +} + +/** + * amdgpu_vm_pt_sibling - go to sibling node + * + * @adev: amdgpu_device pointer + * @cursor: current state + * + * Walk to the sibling node of the current node. + * Returns: + * True if the walk was possible, false otherwise. + */ +static bool amdgpu_vm_pt_sibling(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + unsigned shift, num_entries; + + /* Root doesn't have a sibling */ + if (!cursor->parent) + return false; + + /* Go to our parents and see if we got a sibling */ + shift = amdgpu_vm_level_shift(adev, cursor->level - 1); + num_entries = amdgpu_vm_num_entries(adev, cursor->level - 1); + + if (cursor->entry == &cursor->parent->entries[num_entries - 1]) + return false; + + cursor->pfn += 1ULL << shift; + cursor->pfn &= ~((1ULL << shift) - 1); + ++cursor->entry; + return true; +} + +/** + * amdgpu_vm_pt_ancestor - go to parent node + * + * @cursor: current state + * + * Walk to the parent node of the current node. + * Returns: + * True if the walk was possible, false otherwise. + */ +static bool amdgpu_vm_pt_ancestor(struct amdgpu_vm_pt_cursor *cursor) +{ + if (!cursor->parent) + return false; + + --cursor->level; + cursor->entry = cursor->parent; + cursor->parent = amdgpu_vm_pt_parent(cursor->parent); + return true; +} + +/** + * amdgpu_vm_pt_next - get next PD/PT in hieratchy + * + * @adev: amdgpu_device pointer + * @cursor: current state + * + * Walk the PD/PT tree to the next node. + */ +static void amdgpu_vm_pt_next(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + /* First try a newborn child */ + if (amdgpu_vm_pt_descendant(adev, cursor)) + return; + + /* If that didn't worked try to find a sibling */ + while (!amdgpu_vm_pt_sibling(adev, cursor)) { + /* No sibling, go to our parents and grandparents */ + if (!amdgpu_vm_pt_ancestor(cursor)) { + cursor->pfn = ~0ll; + return; + } + } +} + +/** + * amdgpu_vm_pt_first_leaf - get first leaf PD/PT + * + * @adev: amdgpu_device pointer + * @vm: amdgpu_vm structure + * @start: start addr of the walk + * @cursor: state to initialize + * + * Start a walk and go directly to the leaf node. + */ +static void amdgpu_vm_pt_first_leaf(struct amdgpu_device *adev, + struct amdgpu_vm *vm, uint64_t start, + struct amdgpu_vm_pt_cursor *cursor) +{ + amdgpu_vm_pt_start(adev, vm, start, cursor); + while (amdgpu_vm_pt_descendant(adev, cursor)); +} + +/** + * amdgpu_vm_pt_next_leaf - get next leaf PD/PT + * + * @adev: amdgpu_device pointer + * @cursor: current state + * + * Walk the PD/PT tree to the next leaf node. + */ +static void amdgpu_vm_pt_next_leaf(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + amdgpu_vm_pt_next(adev, cursor); + while (amdgpu_vm_pt_descendant(adev, cursor)); +} + +/** + * for_each_amdgpu_vm_pt_leaf - walk over all leaf PDs/PTs in the hierarchy + */ +#define for_each_amdgpu_vm_pt_leaf(adev, vm, start, end, cursor) \ + for (amdgpu_vm_pt_first_leaf((adev), (vm), (start), &(cursor)); \ + (cursor).pfn <= end; amdgpu_vm_pt_next_leaf((adev), &(cursor))) + +/** + * amdgpu_vm_pt_first_dfs - start a deep first search + * + * @adev: amdgpu_device structure + * @vm: amdgpu_vm structure + * @cursor: state to initialize + * + * Starts a deep first traversal of the PD/PT tree. + */ +static void amdgpu_vm_pt_first_dfs(struct amdgpu_device *adev, + struct amdgpu_vm *vm, + struct amdgpu_vm_pt_cursor *cursor) +{ + amdgpu_vm_pt_start(adev, vm, 0, cursor); + while (amdgpu_vm_pt_descendant(adev, cursor)); +} + +/** + * amdgpu_vm_pt_next_dfs - get the next node for a deep first search + * + * @adev: amdgpu_device structure + * @cursor: current state + * + * Move the cursor to the next node in a deep first search. + */ +static void amdgpu_vm_pt_next_dfs(struct amdgpu_device *adev, + struct amdgpu_vm_pt_cursor *cursor) +{ + if (!cursor->entry) + return; + + if (!cursor->parent) + cursor->entry = NULL; + else if (amdgpu_vm_pt_sibling(adev, cursor)) + while (amdgpu_vm_pt_descendant(adev, cursor)); + else + amdgpu_vm_pt_ancestor(cursor); +} + +/** + * for_each_amdgpu_vm_pt_dfs_safe - safe deep first search of all PDs/PTs + */ +#define for_each_amdgpu_vm_pt_dfs_safe(adev, vm, cursor, entry) \ + for (amdgpu_vm_pt_first_dfs((adev), (vm), &(cursor)), \ + (entry) = (cursor).entry, amdgpu_vm_pt_next_dfs((adev), &(cursor));\ + (entry); (entry) = (cursor).entry, \ + amdgpu_vm_pt_next_dfs((adev), &(cursor))) + +/** * amdgpu_vm_get_pd_bo - add the VM PD to a validation list * * @vm: vm providing the BOs @@ -204,15 +614,55 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, struct list_head *validated, struct amdgpu_bo_list_entry *entry) { - entry->robj = vm->root.base.bo; entry->priority = 0; - entry->tv.bo = &entry->robj->tbo; + entry->tv.bo = &vm->root.base.bo->tbo; entry->tv.shared = true; entry->user_pages = NULL; list_add(&entry->tv.head, validated); } /** + * amdgpu_vm_move_to_lru_tail - move all BOs to the end of LRU + * + * @adev: amdgpu device pointer + * @vm: vm providing the BOs + * + * Move all BOs to the end of LRU and remember their positions to put them + * together. + */ +void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev, + struct amdgpu_vm *vm) +{ + struct ttm_bo_global *glob = adev->mman.bdev.glob; + struct amdgpu_vm_bo_base *bo_base; + + if (vm->bulk_moveable) { + spin_lock(&glob->lru_lock); + ttm_bo_bulk_move_lru_tail(&vm->lru_bulk_move); + spin_unlock(&glob->lru_lock); + return; + } + + memset(&vm->lru_bulk_move, 0, sizeof(vm->lru_bulk_move)); + + spin_lock(&glob->lru_lock); + list_for_each_entry(bo_base, &vm->idle, vm_status) { + struct amdgpu_bo *bo = bo_base->bo; + + if (!bo->parent) + continue; + + ttm_bo_move_to_lru_tail(&bo->tbo, &vm->lru_bulk_move); + if (bo->shadow) + ttm_bo_move_to_lru_tail(&bo->shadow->tbo, + &vm->lru_bulk_move); + } + spin_unlock(&glob->lru_lock); + + vm->bulk_moveable = true; +} + +/** * amdgpu_vm_validate_pt_bos - validate the page table BOs * * @adev: amdgpu device pointer @@ -221,52 +671,44 @@ void amdgpu_vm_get_pd_bo(struct amdgpu_vm *vm, * @param: parameter for the validation callback * * Validate the page table BOs on command submission if neccessary. + * + * Returns: + * Validation result. */ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, int (*validate)(void *p, struct amdgpu_bo *bo), void *param) { - struct ttm_bo_global *glob = adev->mman.bdev.glob; struct amdgpu_vm_bo_base *bo_base, *tmp; int r = 0; + vm->bulk_moveable &= list_empty(&vm->evicted); + list_for_each_entry_safe(bo_base, tmp, &vm->evicted, vm_status) { struct amdgpu_bo *bo = bo_base->bo; - if (bo->parent) { - r = validate(param, bo); - if (r) - break; - - spin_lock(&glob->lru_lock); - ttm_bo_move_to_lru_tail(&bo->tbo); - if (bo->shadow) - ttm_bo_move_to_lru_tail(&bo->shadow->tbo); - spin_unlock(&glob->lru_lock); - } + r = validate(param, bo); + if (r) + break; if (bo->tbo.type != ttm_bo_type_kernel) { - spin_lock(&vm->moved_lock); - list_move(&bo_base->vm_status, &vm->moved); - spin_unlock(&vm->moved_lock); + amdgpu_vm_bo_moved(bo_base); } else { - list_move(&bo_base->vm_status, &vm->relocated); + if (vm->use_cpu_for_update) + r = amdgpu_bo_kmap(bo, NULL); + else + r = amdgpu_ttm_alloc_gart(&bo->tbo); + if (r) + break; + if (bo->shadow) { + r = amdgpu_ttm_alloc_gart(&bo->shadow->tbo); + if (r) + break; + } + amdgpu_vm_bo_relocated(bo_base); } } - spin_lock(&glob->lru_lock); - list_for_each_entry(bo_base, &vm->idle, vm_status) { - struct amdgpu_bo *bo = bo_base->bo; - - if (!bo->parent) - continue; - - ttm_bo_move_to_lru_tail(&bo->tbo); - if (bo->shadow) - ttm_bo_move_to_lru_tail(&bo->shadow->tbo); - } - spin_unlock(&glob->lru_lock); - return r; } @@ -276,6 +718,9 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, * @vm: VM to check * * Check if all VM PDs/PTs are ready for updates + * + * Returns: + * True if eviction list is empty. */ bool amdgpu_vm_ready(struct amdgpu_vm *vm) { @@ -286,10 +731,15 @@ bool amdgpu_vm_ready(struct amdgpu_vm *vm) * amdgpu_vm_clear_bo - initially clear the PDs/PTs * * @adev: amdgpu_device pointer + * @vm: VM to clear BO from * @bo: BO to clear * @level: level this BO is at + * @pte_support_ats: indicate ATS support from PTE * * Root PD needs to be reserved when calling this. + * + * Returns: + * 0 on success, errno otherwise. */ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, struct amdgpu_vm *vm, struct amdgpu_bo *bo, @@ -303,14 +753,13 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, uint64_t addr; int r; - addr = amdgpu_bo_gpu_offset(bo); entries = amdgpu_bo_size(bo) / 8; if (pte_support_ats) { if (level == adev->vm_manager.root_level) { ats_entries = amdgpu_vm_level_shift(adev, level); ats_entries += AMDGPU_GPU_PAGE_SHIFT; - ats_entries = AMDGPU_VA_HOLE_START >> ats_entries; + ats_entries = AMDGPU_GMC_HOLE_START >> ats_entries; ats_entries = min(ats_entries, entries); entries -= ats_entries; } else { @@ -321,7 +770,7 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, ats_entries = 0; } - ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); + ring = container_of(vm->entity.rq->sched, struct amdgpu_ring, sched); r = reservation_object_reserve_shared(bo->tbo.resv); if (r) @@ -331,10 +780,15 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, if (r) goto error; + r = amdgpu_ttm_alloc_gart(&bo->tbo); + if (r) + return r; + r = amdgpu_job_alloc_with_ib(adev, 64, &job); if (r) goto error; + addr = amdgpu_bo_gpu_offset(bo); if (ats_entries) { uint64_t ats_value; @@ -359,8 +813,8 @@ static int amdgpu_vm_clear_bo(struct amdgpu_device *adev, if (r) goto error_free; - r = amdgpu_job_submit(job, ring, &vm->entity, - AMDGPU_FENCE_OWNER_UNDEFINED, &fence); + r = amdgpu_job_submit(job, &vm->entity, AMDGPU_FENCE_OWNER_UNDEFINED, + &fence); if (r) goto error_free; @@ -381,110 +835,33 @@ error: } /** - * amdgpu_vm_alloc_levels - allocate the PD/PT levels + * amdgpu_vm_bo_param - fill in parameters for PD/PT allocation * * @adev: amdgpu_device pointer - * @vm: requested vm - * @saddr: start of the address range - * @eaddr: end of the address range - * - * Make sure the page directories and page tables are allocated + * @vm: requesting vm + * @bp: resulting BO allocation parameters */ -static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_vm_pt *parent, - uint64_t saddr, uint64_t eaddr, - unsigned level, bool ats) +static void amdgpu_vm_bo_param(struct amdgpu_device *adev, struct amdgpu_vm *vm, + int level, struct amdgpu_bo_param *bp) { - unsigned shift = amdgpu_vm_level_shift(adev, level); - unsigned pt_idx, from, to; - u64 flags; - int r; - - if (!parent->entries) { - unsigned num_entries = amdgpu_vm_num_entries(adev, level); - - parent->entries = kvmalloc_array(num_entries, - sizeof(struct amdgpu_vm_pt), - GFP_KERNEL | __GFP_ZERO); - if (!parent->entries) - return -ENOMEM; - memset(parent->entries, 0 , sizeof(struct amdgpu_vm_pt)); - } - - from = saddr >> shift; - to = eaddr >> shift; - if (from >= amdgpu_vm_num_entries(adev, level) || - to >= amdgpu_vm_num_entries(adev, level)) - return -EINVAL; - - ++level; - saddr = saddr & ((1 << shift) - 1); - eaddr = eaddr & ((1 << shift) - 1); - - flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + memset(bp, 0, sizeof(*bp)); + + bp->size = amdgpu_vm_bo_size(adev, level); + bp->byte_align = AMDGPU_GPU_PAGE_SIZE; + bp->domain = AMDGPU_GEM_DOMAIN_VRAM; + if (bp->size <= PAGE_SIZE && adev->asic_type >= CHIP_VEGA10 && + adev->flags & AMD_IS_APU) + bp->domain |= AMDGPU_GEM_DOMAIN_GTT; + bp->domain = amdgpu_bo_get_preferred_pin_domain(adev, bp->domain); + bp->flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS | + AMDGPU_GEM_CREATE_CPU_GTT_USWC; if (vm->use_cpu_for_update) - flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - else - flags |= (AMDGPU_GEM_CREATE_NO_CPU_ACCESS | - AMDGPU_GEM_CREATE_SHADOW); - - /* walk over the address space and allocate the page tables */ - for (pt_idx = from; pt_idx <= to; ++pt_idx) { - struct reservation_object *resv = vm->root.base.bo->tbo.resv; - struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; - struct amdgpu_bo *pt; - - if (!entry->base.bo) { - struct amdgpu_bo_param bp; - - memset(&bp, 0, sizeof(bp)); - bp.size = amdgpu_vm_bo_size(adev, level); - bp.byte_align = AMDGPU_GPU_PAGE_SIZE; - bp.domain = AMDGPU_GEM_DOMAIN_VRAM; - bp.flags = flags; - bp.type = ttm_bo_type_kernel; - bp.resv = resv; - r = amdgpu_bo_create(adev, &bp, &pt); - if (r) - return r; - - r = amdgpu_vm_clear_bo(adev, vm, pt, level, ats); - if (r) { - amdgpu_bo_unref(&pt->shadow); - amdgpu_bo_unref(&pt); - return r; - } - - if (vm->use_cpu_for_update) { - r = amdgpu_bo_kmap(pt, NULL); - if (r) { - amdgpu_bo_unref(&pt->shadow); - amdgpu_bo_unref(&pt); - return r; - } - } - - /* Keep a reference to the root directory to avoid - * freeing them up in the wrong order. - */ - pt->parent = amdgpu_bo_ref(parent->base.bo); - - amdgpu_vm_bo_base_init(&entry->base, vm, pt); - } - - if (level < AMDGPU_VM_PTB) { - uint64_t sub_saddr = (pt_idx == from) ? saddr : 0; - uint64_t sub_eaddr = (pt_idx == to) ? eaddr : - ((1 << shift) - 1); - r = amdgpu_vm_alloc_levels(adev, vm, entry, sub_saddr, - sub_eaddr, level, ats); - if (r) - return r; - } - } - - return 0; + bp->flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; + else if (!vm->root.base.bo || vm->root.base.bo->shadow) + bp->flags |= AMDGPU_GEM_CREATE_SHADOW; + bp->type = ttm_bo_type_kernel; + if (vm->root.base.bo) + bp->resv = vm->root.base.bo->tbo.resv; } /** @@ -495,14 +872,20 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, * @saddr: Start address which needs to be allocated * @size: Size from start address we need. * - * Make sure the page tables are allocated. + * Make sure the page directories and page tables are allocated + * + * Returns: + * 0 on success, errno otherwise. */ int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, struct amdgpu_vm *vm, uint64_t saddr, uint64_t size) { - uint64_t eaddr; + struct amdgpu_vm_pt_cursor cursor; + struct amdgpu_bo *pt; bool ats = false; + uint64_t eaddr; + int r; /* validate the parameters */ if (saddr & AMDGPU_GPU_PAGE_MASK || size & AMDGPU_GPU_PAGE_MASK) @@ -511,7 +894,7 @@ int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, eaddr = saddr + size - 1; if (vm->pte_support_ats) - ats = saddr < AMDGPU_VA_HOLE_START; + ats = saddr < AMDGPU_GMC_HOLE_START; saddr /= AMDGPU_GPU_PAGE_SIZE; eaddr /= AMDGPU_GPU_PAGE_SIZE; @@ -522,8 +905,84 @@ int amdgpu_vm_alloc_pts(struct amdgpu_device *adev, return -EINVAL; } - return amdgpu_vm_alloc_levels(adev, vm, &vm->root, saddr, eaddr, - adev->vm_manager.root_level, ats); + for_each_amdgpu_vm_pt_leaf(adev, vm, saddr, eaddr, cursor) { + struct amdgpu_vm_pt *entry = cursor.entry; + struct amdgpu_bo_param bp; + + if (cursor.level < AMDGPU_VM_PTB) { + unsigned num_entries; + + num_entries = amdgpu_vm_num_entries(adev, cursor.level); + entry->entries = kvmalloc_array(num_entries, + sizeof(*entry->entries), + GFP_KERNEL | + __GFP_ZERO); + if (!entry->entries) + return -ENOMEM; + } + + + if (entry->base.bo) + continue; + + amdgpu_vm_bo_param(adev, vm, cursor.level, &bp); + + r = amdgpu_bo_create(adev, &bp, &pt); + if (r) + return r; + + r = amdgpu_vm_clear_bo(adev, vm, pt, cursor.level, ats); + if (r) + goto error_free_pt; + + if (vm->use_cpu_for_update) { + r = amdgpu_bo_kmap(pt, NULL); + if (r) + goto error_free_pt; + } + + /* Keep a reference to the root directory to avoid + * freeing them up in the wrong order. + */ + pt->parent = amdgpu_bo_ref(cursor.parent->base.bo); + + amdgpu_vm_bo_base_init(&entry->base, vm, pt); + } + + return 0; + +error_free_pt: + amdgpu_bo_unref(&pt->shadow); + amdgpu_bo_unref(&pt); + return r; +} + +/** + * amdgpu_vm_free_pts - free PD/PT levels + * + * @adev: amdgpu device structure + * @vm: amdgpu vm structure + * + * Free the page directory or page table level and all sub levels. + */ +static void amdgpu_vm_free_pts(struct amdgpu_device *adev, + struct amdgpu_vm *vm) +{ + struct amdgpu_vm_pt_cursor cursor; + struct amdgpu_vm_pt *entry; + + for_each_amdgpu_vm_pt_dfs_safe(adev, vm, cursor, entry) { + + if (entry->base.bo) { + entry->base.bo->vm_bo = NULL; + list_del(&entry->base.vm_status); + amdgpu_bo_unref(&entry->base.bo->shadow); + amdgpu_bo_unref(&entry->base.bo); + } + kvfree(entry->entries); + } + + BUG_ON(vm->root.base.bo); } /** @@ -561,6 +1020,15 @@ void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev) } } +/** + * amdgpu_vm_need_pipeline_sync - Check if pipe sync is needed for job. + * + * @ring: ring on which the job will be submitted + * @job: job to submit + * + * Returns: + * True if sync is needed. + */ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring, struct amdgpu_job *job) { @@ -588,19 +1056,17 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring, return vm_flush_needed || gds_switch_needed; } -static bool amdgpu_vm_is_large_bar(struct amdgpu_device *adev) -{ - return (adev->gmc.real_vram_size == adev->gmc.visible_vram_size); -} - /** * amdgpu_vm_flush - hardware flush the vm * * @ring: ring to use for flush - * @vmid: vmid number to use - * @pd_addr: address of the page directory + * @job: related job + * @need_pipe_sync: is pipe sync needed * * Emit a VM flush when it is necessary. + * + * Returns: + * 0 on success, errno otherwise. */ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_pipe_sync) { @@ -630,7 +1096,8 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ } gds_switch_needed &= !!ring->funcs->emit_gds_switch; - vm_flush_needed &= !!ring->funcs->emit_vm_flush; + vm_flush_needed &= !!ring->funcs->emit_vm_flush && + job->vm_pd_addr != AMDGPU_BO_INVALID_OFFSET; pasid_mapping_needed &= adev->gmc.gmc_funcs->emit_pasid_mapping && ring->funcs->emit_wreg; @@ -708,16 +1175,20 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ * Returns the found bo_va or NULL if none is found * * Object has to be reserved! + * + * Returns: + * Found bo_va or NULL. */ struct amdgpu_bo_va *amdgpu_vm_bo_find(struct amdgpu_vm *vm, struct amdgpu_bo *bo) { - struct amdgpu_bo_va *bo_va; + struct amdgpu_vm_bo_base *base; - list_for_each_entry(bo_va, &bo->va, base.bo_list) { - if (bo_va->base.vm == vm) { - return bo_va; - } + for (base = bo->vm_bo; base; base = base->next) { + if (base->vm != vm) + continue; + + return container_of(base, struct amdgpu_bo_va, base); } return NULL; } @@ -789,7 +1260,10 @@ static void amdgpu_vm_do_copy_ptes(struct amdgpu_pte_update_params *params, * @addr: the unmapped addr * * Look up the physical address of the page that the pte resolves - * to and return the pointer for the page table entry. + * to. + * + * Returns: + * The pointer for the page table entry. */ static uint64_t amdgpu_vm_map_gart(const dma_addr_t *pages_addr, uint64_t addr) { @@ -842,6 +1316,17 @@ static void amdgpu_vm_cpu_set_ptes(struct amdgpu_pte_update_params *params, } } + +/** + * amdgpu_vm_wait_pd - Wait for PT BOs to be free. + * + * @adev: amdgpu_device pointer + * @vm: related vm + * @owner: fence owner + * + * Returns: + * 0 on success, errno otherwise. + */ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm, void *owner) { @@ -856,6 +1341,22 @@ static int amdgpu_vm_wait_pd(struct amdgpu_device *adev, struct amdgpu_vm *vm, return r; } +/** + * amdgpu_vm_update_func - helper to call update function + * + * Calls the update function for both the given BO as well as its shadow. + */ +static void amdgpu_vm_update_func(struct amdgpu_pte_update_params *params, + struct amdgpu_bo *bo, + uint64_t pe, uint64_t addr, + unsigned count, uint32_t incr, + uint64_t flags) +{ + if (bo->shadow) + params->func(params, bo->shadow, pe, addr, count, incr, flags); + params->func(params, bo, pe, addr, count, incr, flags); +} + /* * amdgpu_vm_update_pde - update a single level in the hierarchy * @@ -883,44 +1384,28 @@ static void amdgpu_vm_update_pde(struct amdgpu_pte_update_params *params, pbo = pbo->parent; level += params->adev->vm_manager.root_level; - pt = amdgpu_bo_gpu_offset(entry->base.bo); - flags = AMDGPU_PTE_VALID; - amdgpu_gmc_get_vm_pde(params->adev, level, &pt, &flags); + amdgpu_gmc_get_pde_for_bo(entry->base.bo, level, &pt, &flags); pde = (entry - parent->entries) * 8; - if (bo->shadow) - params->func(params, bo->shadow, pde, pt, 1, 0, flags); - params->func(params, bo, pde, pt, 1, 0, flags); + amdgpu_vm_update_func(params, bo, pde, pt, 1, 0, flags); } /* - * amdgpu_vm_invalidate_level - mark all PD levels as invalid + * amdgpu_vm_invalidate_pds - mark all PDs as invalid * - * @parent: parent PD + * @adev: amdgpu_device pointer + * @vm: related vm * * Mark all PD level as invalid after an error. */ -static void amdgpu_vm_invalidate_level(struct amdgpu_device *adev, - struct amdgpu_vm *vm, - struct amdgpu_vm_pt *parent, - unsigned level) +static void amdgpu_vm_invalidate_pds(struct amdgpu_device *adev, + struct amdgpu_vm *vm) { - unsigned pt_idx, num_entries; + struct amdgpu_vm_pt_cursor cursor; + struct amdgpu_vm_pt *entry; - /* - * Recurse into the subdirectories. This recursion is harmless because - * we only have a maximum of 5 layers. - */ - num_entries = amdgpu_vm_num_entries(adev, level); - for (pt_idx = 0; pt_idx < num_entries; ++pt_idx) { - struct amdgpu_vm_pt *entry = &parent->entries[pt_idx]; - - if (!entry->base.bo) - continue; - - if (!entry->base.moved) - list_move(&entry->base.vm_status, &vm->relocated); - amdgpu_vm_invalidate_level(adev, vm, entry, level + 1); - } + for_each_amdgpu_vm_pt_dfs_safe(adev, vm, cursor, entry) + if (entry->base.bo && !entry->base.moved) + amdgpu_vm_bo_relocated(&entry->base); } /* @@ -930,7 +1415,9 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_device *adev, * @vm: requested vm * * Makes sure all directories are up to date. - * Returns 0 for success, error for failure. + * + * Returns: + * 0 for success, error for failure. */ int amdgpu_vm_update_directories(struct amdgpu_device *adev, struct amdgpu_vm *vm) @@ -948,14 +1435,6 @@ restart: params.adev = adev; if (vm->use_cpu_for_update) { - struct amdgpu_vm_bo_base *bo_base; - - list_for_each_entry(bo_base, &vm->relocated, vm_status) { - r = amdgpu_bo_kmap(bo_base->bo, NULL); - if (unlikely(r)) - return r; - } - r = amdgpu_vm_wait_pd(adev, vm, AMDGPU_FENCE_OWNER_VM); if (unlikely(r)) return r; @@ -972,25 +1451,16 @@ restart: } while (!list_empty(&vm->relocated)) { - struct amdgpu_vm_bo_base *bo_base, *parent; struct amdgpu_vm_pt *pt, *entry; - struct amdgpu_bo *bo; - bo_base = list_first_entry(&vm->relocated, - struct amdgpu_vm_bo_base, - vm_status); - bo_base->moved = false; - list_move(&bo_base->vm_status, &vm->idle); + entry = list_first_entry(&vm->relocated, struct amdgpu_vm_pt, + base.vm_status); + amdgpu_vm_bo_idle(&entry->base); - bo = bo_base->bo->parent; - if (!bo) + pt = amdgpu_vm_pt_parent(entry); + if (!pt) continue; - parent = list_first_entry(&bo->va, struct amdgpu_vm_bo_base, - bo_list); - pt = container_of(parent, struct amdgpu_vm_pt, base); - entry = container_of(bo_base, struct amdgpu_vm_pt, base); - amdgpu_vm_update_pde(¶ms, vm, pt, entry); if (!vm->use_cpu_for_update && @@ -1009,15 +1479,15 @@ restart: struct amdgpu_ring *ring; struct dma_fence *fence; - ring = container_of(vm->entity.sched, struct amdgpu_ring, + ring = container_of(vm->entity.rq->sched, struct amdgpu_ring, sched); amdgpu_ring_pad_ib(ring, params.ib); amdgpu_sync_resv(adev, &job->sync, root->tbo.resv, AMDGPU_FENCE_OWNER_VM, false); WARN_ON(params.ib->length_dw > ndw); - r = amdgpu_job_submit(job, ring, &vm->entity, - AMDGPU_FENCE_OWNER_VM, &fence); + r = amdgpu_job_submit(job, &vm->entity, AMDGPU_FENCE_OWNER_VM, + &fence); if (r) goto error; @@ -1032,205 +1502,206 @@ restart: return 0; error: - amdgpu_vm_invalidate_level(adev, vm, &vm->root, - adev->vm_manager.root_level); + amdgpu_vm_invalidate_pds(adev, vm); amdgpu_job_free(job); return r; } /** - * amdgpu_vm_find_entry - find the entry for an address - * - * @p: see amdgpu_pte_update_params definition - * @addr: virtual address in question - * @entry: resulting entry or NULL - * @parent: parent entry + * amdgpu_vm_update_huge - figure out parameters for PTE updates * - * Find the vm_pt entry and it's parent for the given address. + * Make sure to set the right flags for the PTEs at the desired level. */ -void amdgpu_vm_get_entry(struct amdgpu_pte_update_params *p, uint64_t addr, - struct amdgpu_vm_pt **entry, - struct amdgpu_vm_pt **parent) -{ - unsigned level = p->adev->vm_manager.root_level; - - *parent = NULL; - *entry = &p->vm->root; - while ((*entry)->entries) { - unsigned shift = amdgpu_vm_level_shift(p->adev, level++); +static void amdgpu_vm_update_huge(struct amdgpu_pte_update_params *params, + struct amdgpu_bo *bo, unsigned level, + uint64_t pe, uint64_t addr, + unsigned count, uint32_t incr, + uint64_t flags) - *parent = *entry; - *entry = &(*entry)->entries[addr >> shift]; - addr &= (1ULL << shift) - 1; +{ + if (level != AMDGPU_VM_PTB) { + flags |= AMDGPU_PDE_PTE; + amdgpu_gmc_get_vm_pde(params->adev, level, &addr, &flags); } - if (level != AMDGPU_VM_PTB) - *entry = NULL; + amdgpu_vm_update_func(params, bo, pe, addr, count, incr, flags); } /** - * amdgpu_vm_handle_huge_pages - handle updating the PD with huge pages + * amdgpu_vm_fragment - get fragment for PTEs * - * @p: see amdgpu_pte_update_params definition - * @entry: vm_pt entry to check - * @parent: parent entry - * @nptes: number of PTEs updated with this operation - * @dst: destination address where the PTEs should point to - * @flags: access flags fro the PTEs + * @params: see amdgpu_pte_update_params definition + * @start: first PTE to handle + * @end: last PTE to handle + * @flags: hw mapping flags + * @frag: resulting fragment size + * @frag_end: end of this fragment * - * Check if we can update the PD with a huge page. + * Returns the first possible fragment for the start and end address. */ -static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p, - struct amdgpu_vm_pt *entry, - struct amdgpu_vm_pt *parent, - unsigned nptes, uint64_t dst, - uint64_t flags) +static void amdgpu_vm_fragment(struct amdgpu_pte_update_params *params, + uint64_t start, uint64_t end, uint64_t flags, + unsigned int *frag, uint64_t *frag_end) { - uint64_t pde; + /** + * The MC L1 TLB supports variable sized pages, based on a fragment + * field in the PTE. When this field is set to a non-zero value, page + * granularity is increased from 4KB to (1 << (12 + frag)). The PTE + * flags are considered valid for all PTEs within the fragment range + * and corresponding mappings are assumed to be physically contiguous. + * + * The L1 TLB can store a single PTE for the whole fragment, + * significantly increasing the space available for translation + * caching. This leads to large improvements in throughput when the + * TLB is under pressure. + * + * The L2 TLB distributes small and large fragments into two + * asymmetric partitions. The large fragment cache is significantly + * larger. Thus, we try to use large fragments wherever possible. + * Userspace can support this by aligning virtual base address and + * allocation size to the fragment size. + * + * Starting with Vega10 the fragment size only controls the L1. The L2 + * is now directly feed with small/huge/giant pages from the walker. + */ + unsigned max_frag; - /* In the case of a mixed PT the PDE must point to it*/ - if (p->adev->asic_type >= CHIP_VEGA10 && !p->src && - nptes == AMDGPU_VM_PTE_COUNT(p->adev)) { - /* Set the huge page flag to stop scanning at this PDE */ - flags |= AMDGPU_PDE_PTE; - } + if (params->adev->asic_type < CHIP_VEGA10) + max_frag = params->adev->vm_manager.fragment_size; + else + max_frag = 31; - if (!(flags & AMDGPU_PDE_PTE)) { - if (entry->huge) { - /* Add the entry to the relocated list to update it. */ - entry->huge = false; - list_move(&entry->base.vm_status, &p->vm->relocated); - } + /* system pages are non continuously */ + if (params->src) { + *frag = 0; + *frag_end = end; return; } - entry->huge = true; - amdgpu_gmc_get_vm_pde(p->adev, AMDGPU_VM_PDB0, &dst, &flags); - - pde = (entry - parent->entries) * 8; - if (parent->base.bo->shadow) - p->func(p, parent->base.bo->shadow, pde, dst, 1, 0, flags); - p->func(p, parent->base.bo, pde, dst, 1, 0, flags); + /* This intentionally wraps around if no bit is set */ + *frag = min((unsigned)ffs(start) - 1, (unsigned)fls64(end - start) - 1); + if (*frag >= max_frag) { + *frag = max_frag; + *frag_end = end & ~((1ULL << max_frag) - 1); + } else { + *frag_end = start + (1 << *frag); + } } /** * amdgpu_vm_update_ptes - make sure that page tables are valid * * @params: see amdgpu_pte_update_params definition - * @vm: requested vm * @start: start of GPU address range * @end: end of GPU address range * @dst: destination address to map to, the next dst inside the function * @flags: mapping flags * * Update the page tables in the range @start - @end. - * Returns 0 for success, -EINVAL for failure. + * + * Returns: + * 0 for success, -EINVAL for failure. */ static int amdgpu_vm_update_ptes(struct amdgpu_pte_update_params *params, - uint64_t start, uint64_t end, - uint64_t dst, uint64_t flags) + uint64_t start, uint64_t end, + uint64_t dst, uint64_t flags) { struct amdgpu_device *adev = params->adev; - const uint64_t mask = AMDGPU_VM_PTE_COUNT(adev) - 1; + struct amdgpu_vm_pt_cursor cursor; + uint64_t frag_start = start, frag_end; + unsigned int frag; - uint64_t addr, pe_start; - struct amdgpu_bo *pt; - unsigned nptes; + /* figure out the initial fragment */ + amdgpu_vm_fragment(params, frag_start, end, flags, &frag, &frag_end); - /* walk over the address space and update the page tables */ - for (addr = start; addr < end; addr += nptes, - dst += nptes * AMDGPU_GPU_PAGE_SIZE) { - struct amdgpu_vm_pt *entry, *parent; + /* walk over the address space and update the PTs */ + amdgpu_vm_pt_start(adev, params->vm, start, &cursor); + while (cursor.pfn < end) { + struct amdgpu_bo *pt = cursor.entry->base.bo; + unsigned shift, parent_shift, mask; + uint64_t incr, entry_end, pe_start; - amdgpu_vm_get_entry(params, addr, &entry, &parent); - if (!entry) + if (!pt) return -ENOENT; - if ((addr & ~mask) == (end & ~mask)) - nptes = end - addr; - else - nptes = AMDGPU_VM_PTE_COUNT(adev) - (addr & mask); - - amdgpu_vm_handle_huge_pages(params, entry, parent, - nptes, dst, flags); - /* We don't need to update PTEs for huge pages */ - if (entry->huge) + /* The root level can't be a huge page */ + if (cursor.level == adev->vm_manager.root_level) { + if (!amdgpu_vm_pt_descendant(adev, &cursor)) + return -ENOENT; continue; + } - pt = entry->base.bo; - pe_start = (addr & mask) * 8; - if (pt->shadow) - params->func(params, pt->shadow, pe_start, dst, nptes, - AMDGPU_GPU_PAGE_SIZE, flags); - params->func(params, pt, pe_start, dst, nptes, - AMDGPU_GPU_PAGE_SIZE, flags); - } - - return 0; -} + /* First check if the entry is already handled */ + if (cursor.pfn < frag_start) { + cursor.entry->huge = true; + amdgpu_vm_pt_next(adev, &cursor); + continue; + } -/* - * amdgpu_vm_frag_ptes - add fragment information to PTEs - * - * @params: see amdgpu_pte_update_params definition - * @vm: requested vm - * @start: first PTE to handle - * @end: last PTE to handle - * @dst: addr those PTEs should point to - * @flags: hw mapping flags - * Returns 0 for success, -EINVAL for failure. - */ -static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, - uint64_t start, uint64_t end, - uint64_t dst, uint64_t flags) -{ - /** - * The MC L1 TLB supports variable sized pages, based on a fragment - * field in the PTE. When this field is set to a non-zero value, page - * granularity is increased from 4KB to (1 << (12 + frag)). The PTE - * flags are considered valid for all PTEs within the fragment range - * and corresponding mappings are assumed to be physically contiguous. - * - * The L1 TLB can store a single PTE for the whole fragment, - * significantly increasing the space available for translation - * caching. This leads to large improvements in throughput when the - * TLB is under pressure. - * - * The L2 TLB distributes small and large fragments into two - * asymmetric partitions. The large fragment cache is significantly - * larger. Thus, we try to use large fragments wherever possible. - * Userspace can support this by aligning virtual base address and - * allocation size to the fragment size. - */ - unsigned max_frag = params->adev->vm_manager.fragment_size; - int r; + /* If it isn't already handled it can't be a huge page */ + if (cursor.entry->huge) { + /* Add the entry to the relocated list to update it. */ + cursor.entry->huge = false; + amdgpu_vm_bo_relocated(&cursor.entry->base); + } - /* system pages are non continuously */ - if (params->src || !(flags & AMDGPU_PTE_VALID)) - return amdgpu_vm_update_ptes(params, start, end, dst, flags); - - while (start != end) { - uint64_t frag_flags, frag_end; - unsigned frag; - - /* This intentionally wraps around if no bit is set */ - frag = min((unsigned)ffs(start) - 1, - (unsigned)fls64(end - start) - 1); - if (frag >= max_frag) { - frag_flags = AMDGPU_PTE_FRAG(max_frag); - frag_end = end & ~((1ULL << max_frag) - 1); - } else { - frag_flags = AMDGPU_PTE_FRAG(frag); - frag_end = start + (1 << frag); + shift = amdgpu_vm_level_shift(adev, cursor.level); + parent_shift = amdgpu_vm_level_shift(adev, cursor.level - 1); + if (adev->asic_type < CHIP_VEGA10) { + /* No huge page support before GMC v9 */ + if (cursor.level != AMDGPU_VM_PTB) { + if (!amdgpu_vm_pt_descendant(adev, &cursor)) + return -ENOENT; + continue; + } + } else if (frag < shift) { + /* We can't use this level when the fragment size is + * smaller than the address shift. Go to the next + * child entry and try again. + */ + if (!amdgpu_vm_pt_descendant(adev, &cursor)) + return -ENOENT; + continue; + } else if (frag >= parent_shift) { + /* If the fragment size is even larger than the parent + * shift we should go up one level and check it again. + */ + if (!amdgpu_vm_pt_ancestor(&cursor)) + return -ENOENT; + continue; } - r = amdgpu_vm_update_ptes(params, start, frag_end, dst, - flags | frag_flags); - if (r) - return r; + /* Looks good so far, calculate parameters for the update */ + incr = AMDGPU_GPU_PAGE_SIZE << shift; + mask = amdgpu_vm_entries_mask(adev, cursor.level); + pe_start = ((cursor.pfn >> shift) & mask) * 8; + entry_end = (mask + 1) << shift; + entry_end += cursor.pfn & ~(entry_end - 1); + entry_end = min(entry_end, end); + + do { + uint64_t upd_end = min(entry_end, frag_end); + unsigned nptes = (upd_end - frag_start) >> shift; + + amdgpu_vm_update_huge(params, pt, cursor.level, + pe_start, dst, nptes, incr, + flags | AMDGPU_PTE_FRAG(frag)); + + pe_start += nptes * 8; + dst += nptes * AMDGPU_GPU_PAGE_SIZE << shift; + + frag_start = upd_end; + if (frag_start >= frag_end) { + /* figure out the next fragment */ + amdgpu_vm_fragment(params, frag_start, end, + flags, &frag, &frag_end); + if (frag < shift) + break; + } + } while (frag_start < entry_end); - dst += (frag_end - start) * AMDGPU_GPU_PAGE_SIZE; - start = frag_end; + if (frag >= shift) + amdgpu_vm_pt_next(adev, &cursor); } return 0; @@ -1250,7 +1721,9 @@ static int amdgpu_vm_frag_ptes(struct amdgpu_pte_update_params *params, * @fence: optional resulting fence * * Fill in the page table entries between @start and @last. - * Returns 0 for success, -EINVAL for failure. + * + * Returns: + * 0 for success, -EINVAL for failure. */ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, struct dma_fence *exclusive, @@ -1290,11 +1763,11 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, params.func = amdgpu_vm_cpu_set_ptes; params.pages_addr = pages_addr; - return amdgpu_vm_frag_ptes(¶ms, start, last + 1, - addr, flags); + return amdgpu_vm_update_ptes(¶ms, start, last + 1, + addr, flags); } - ring = container_of(vm->entity.sched, struct amdgpu_ring, sched); + ring = container_of(vm->entity.rq->sched, struct amdgpu_ring, sched); nptes = last - start + 1; @@ -1326,7 +1799,10 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, ndw += ncmds * 10; /* extra commands for begin/end fragments */ - ndw += 2 * 10 * adev->vm_manager.fragment_size; + if (vm->root.base.bo->shadow) + ndw += 2 * 10 * adev->vm_manager.fragment_size * 2; + else + ndw += 2 * 10 * adev->vm_manager.fragment_size; params.func = amdgpu_vm_do_set_ptes; } @@ -1367,14 +1843,13 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev, if (r) goto error_free; - r = amdgpu_vm_frag_ptes(¶ms, start, last + 1, addr, flags); + r = amdgpu_vm_update_ptes(¶ms, start, last + 1, addr, flags); if (r) goto error_free; amdgpu_ring_pad_ib(ring, params.ib); WARN_ON(params.ib->length_dw > ndw); - r = amdgpu_job_submit(job, ring, &vm->entity, - AMDGPU_FENCE_OWNER_VM, &f); + r = amdgpu_job_submit(job, &vm->entity, AMDGPU_FENCE_OWNER_VM, &f); if (r) goto error_free; @@ -1402,7 +1877,9 @@ error_free: * * Split the mapping into smaller chunks so that each update fits * into a SDMA IB. - * Returns 0 for success, -EINVAL for failure. + * + * Returns: + * 0 for success, -EINVAL for failure. */ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, struct dma_fence *exclusive, @@ -1455,7 +1932,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, if (nodes) { addr = nodes->start << PAGE_SHIFT; max_entries = (nodes->size - pfn) * - (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); + AMDGPU_GPU_PAGES_IN_CPU_PAGE; } else { addr = 0; max_entries = S64_MAX; @@ -1466,7 +1943,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, max_entries = min(max_entries, 16ull * 1024ull); for (count = 1; - count < max_entries / (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); + count < max_entries / AMDGPU_GPU_PAGES_IN_CPU_PAGE; ++count) { uint64_t idx = pfn + count; @@ -1480,7 +1957,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, dma_addr = pages_addr; } else { addr = pages_addr[pfn]; - max_entries = count * (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); + max_entries = count * AMDGPU_GPU_PAGES_IN_CPU_PAGE; } } else if (flags & AMDGPU_PTE_VALID) { @@ -1495,7 +1972,7 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, if (r) return r; - pfn += (last - start + 1) / (PAGE_SIZE / AMDGPU_GPU_PAGE_SIZE); + pfn += (last - start + 1) / AMDGPU_GPU_PAGES_IN_CPU_PAGE; if (nodes && nodes->size == pfn) { pfn = 0; ++nodes; @@ -1515,7 +1992,9 @@ static int amdgpu_vm_bo_split_mapping(struct amdgpu_device *adev, * @clear: if true clear the entries * * Fill in the page table entries for @bo_va. - * Returns 0 for success, -EINVAL for failure. + * + * Returns: + * 0 for success, -EINVAL for failure. */ int amdgpu_vm_bo_update(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va, @@ -1531,18 +2010,17 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, uint64_t flags; int r; - if (clear || !bo_va->base.bo) { + if (clear || !bo) { mem = NULL; nodes = NULL; exclusive = NULL; } else { struct ttm_dma_tt *ttm; - mem = &bo_va->base.bo->tbo.mem; + mem = &bo->tbo.mem; nodes = mem->mm_node; if (mem->mem_type == TTM_PL_TT) { - ttm = container_of(bo_va->base.bo->tbo.ttm, - struct ttm_dma_tt, ttm); + ttm = container_of(bo->tbo.ttm, struct ttm_dma_tt, ttm); pages_addr = ttm->dma_address; } exclusive = reservation_object_get_excl(bo->tbo.resv); @@ -1580,10 +2058,6 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, amdgpu_asic_flush_hdp(adev, NULL); } - spin_lock(&vm->moved_lock); - list_del_init(&bo_va->base.vm_status); - spin_unlock(&vm->moved_lock); - /* If the BO is not in its preferred location add it back to * the evicted list so that it gets validated again on the * next command submission. @@ -1592,9 +2066,11 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, uint32_t mem_type = bo->tbo.mem.mem_type; if (!(bo->preferred_domains & amdgpu_mem_type_to_domain(mem_type))) - list_add_tail(&bo_va->base.vm_status, &vm->evicted); + amdgpu_vm_bo_evicted(&bo_va->base); else - list_add(&bo_va->base.vm_status, &vm->idle); + amdgpu_vm_bo_idle(&bo_va->base); + } else { + amdgpu_vm_bo_done(&bo_va->base); } list_splice_init(&bo_va->invalids, &bo_va->valids); @@ -1610,6 +2086,8 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, /** * amdgpu_vm_update_prt_state - update the global PRT state + * + * @adev: amdgpu_device pointer */ static void amdgpu_vm_update_prt_state(struct amdgpu_device *adev) { @@ -1624,6 +2102,8 @@ static void amdgpu_vm_update_prt_state(struct amdgpu_device *adev) /** * amdgpu_vm_prt_get - add a PRT user + * + * @adev: amdgpu_device pointer */ static void amdgpu_vm_prt_get(struct amdgpu_device *adev) { @@ -1636,6 +2116,8 @@ static void amdgpu_vm_prt_get(struct amdgpu_device *adev) /** * amdgpu_vm_prt_put - drop a PRT user + * + * @adev: amdgpu_device pointer */ static void amdgpu_vm_prt_put(struct amdgpu_device *adev) { @@ -1645,6 +2127,9 @@ static void amdgpu_vm_prt_put(struct amdgpu_device *adev) /** * amdgpu_vm_prt_cb - callback for updating the PRT status + * + * @fence: fence for the callback + * @_cb: the callback function */ static void amdgpu_vm_prt_cb(struct dma_fence *fence, struct dma_fence_cb *_cb) { @@ -1656,6 +2141,9 @@ static void amdgpu_vm_prt_cb(struct dma_fence *fence, struct dma_fence_cb *_cb) /** * amdgpu_vm_add_prt_cb - add callback for updating the PRT status + * + * @adev: amdgpu_device pointer + * @fence: fence for the callback */ static void amdgpu_vm_add_prt_cb(struct amdgpu_device *adev, struct dma_fence *fence) @@ -1747,9 +2235,11 @@ static void amdgpu_vm_prt_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) * or if an error occurred) * * Make sure all freed BOs are cleared in the PT. - * Returns 0 for success. - * * PTs have to be reserved and mutex must be locked! + * + * Returns: + * 0 for success. + * */ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct amdgpu_vm *vm, @@ -1765,7 +2255,8 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, struct amdgpu_bo_va_mapping, list); list_del(&mapping->list); - if (vm->pte_support_ats && mapping->start < AMDGPU_VA_HOLE_START) + if (vm->pte_support_ats && + mapping->start < AMDGPU_GMC_HOLE_START) init_pte_value = AMDGPU_PTE_DEFAULT_ATC; r = amdgpu_vm_bo_update_mapping(adev, NULL, NULL, vm, @@ -1794,10 +2285,11 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, * * @adev: amdgpu_device pointer * @vm: requested vm - * @sync: sync object to add fences to * * Make sure all BOs which are moved are updated in the PTs. - * Returns 0 for success. + * + * Returns: + * 0 for success. * * PTs have to be reserved! */ @@ -1805,40 +2297,40 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm) { struct amdgpu_bo_va *bo_va, *tmp; - struct list_head moved; + struct reservation_object *resv; bool clear; int r; - INIT_LIST_HEAD(&moved); - spin_lock(&vm->moved_lock); - list_splice_init(&vm->moved, &moved); - spin_unlock(&vm->moved_lock); + list_for_each_entry_safe(bo_va, tmp, &vm->moved, base.vm_status) { + /* Per VM BOs never need to bo cleared in the page tables */ + r = amdgpu_vm_bo_update(adev, bo_va, false); + if (r) + return r; + } - list_for_each_entry_safe(bo_va, tmp, &moved, base.vm_status) { - struct reservation_object *resv = bo_va->base.bo->tbo.resv; + spin_lock(&vm->invalidated_lock); + while (!list_empty(&vm->invalidated)) { + bo_va = list_first_entry(&vm->invalidated, struct amdgpu_bo_va, + base.vm_status); + resv = bo_va->base.bo->tbo.resv; + spin_unlock(&vm->invalidated_lock); - /* Per VM BOs never need to bo cleared in the page tables */ - if (resv == vm->root.base.bo->tbo.resv) - clear = false; /* Try to reserve the BO to avoid clearing its ptes */ - else if (!amdgpu_vm_debug && reservation_object_trylock(resv)) + if (!amdgpu_vm_debug && reservation_object_trylock(resv)) clear = false; /* Somebody else is using the BO right now */ else clear = true; r = amdgpu_vm_bo_update(adev, bo_va, clear); - if (r) { - spin_lock(&vm->moved_lock); - list_splice(&moved, &vm->moved); - spin_unlock(&vm->moved_lock); + if (r) return r; - } - if (!clear && resv != vm->root.base.bo->tbo.resv) + if (!clear) reservation_object_unlock(resv); - + spin_lock(&vm->invalidated_lock); } + spin_unlock(&vm->invalidated_lock); return 0; } @@ -1852,7 +2344,9 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, * * Add @bo into the requested vm. * Add @bo to the list of bos associated with the vm - * Returns newly added bo_va or NULL for failure + * + * Returns: + * Newly added bo_va or NULL for failure * * Object has to be reserved! */ @@ -1901,9 +2395,7 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv && !bo_va->base.moved) { - spin_lock(&vm->moved_lock); list_move(&bo_va->base.vm_status, &vm->moved); - spin_unlock(&vm->moved_lock); } trace_amdgpu_vm_bo_map(bo_va, mapping); } @@ -1915,10 +2407,13 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, * @bo_va: bo_va to store the address * @saddr: where to map the BO * @offset: requested offset in the BO + * @size: BO size in bytes * @flags: attributes of pages (read/write/valid/etc.) * * Add a mapping of the BO at the specefied addr into the VM. - * Returns 0 for success, error for failure. + * + * Returns: + * 0 for success, error for failure. * * Object has to be reserved and unreserved outside! */ @@ -1976,11 +2471,14 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev, * @bo_va: bo_va to store the address * @saddr: where to map the BO * @offset: requested offset in the BO + * @size: BO size in bytes * @flags: attributes of pages (read/write/valid/etc.) * * Add a mapping of the BO at the specefied addr into the VM. Replace existing * mappings as we do so. - * Returns 0 for success, error for failure. + * + * Returns: + * 0 for success, error for failure. * * Object has to be reserved and unreserved outside! */ @@ -2037,7 +2535,9 @@ int amdgpu_vm_bo_replace_map(struct amdgpu_device *adev, * @saddr: where to the BO is mapped * * Remove a mapping of the BO at the specefied addr from the VM. - * Returns 0 for success, error for failure. + * + * Returns: + * 0 for success, error for failure. * * Object has to be reserved and unreserved outside! */ @@ -2091,7 +2591,9 @@ int amdgpu_vm_bo_unmap(struct amdgpu_device *adev, * @size: size of the range * * Remove all mappings in a range, split them as appropriate. - * Returns 0 for success, error for failure. + * + * Returns: + * 0 for success, error for failure. */ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, struct amdgpu_vm *vm, @@ -2188,8 +2690,13 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, * amdgpu_vm_bo_lookup_mapping - find mapping by address * * @vm: the requested VM + * @addr: the address * * Find a mapping by it's address. + * + * Returns: + * The amdgpu_bo_va_mapping matching for addr or NULL + * */ struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, uint64_t addr) @@ -2198,6 +2705,35 @@ struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, } /** + * amdgpu_vm_bo_trace_cs - trace all reserved mappings + * + * @vm: the requested vm + * @ticket: CS ticket + * + * Trace all mappings of BOs reserved during a command submission. + */ +void amdgpu_vm_bo_trace_cs(struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket) +{ + struct amdgpu_bo_va_mapping *mapping; + + if (!trace_amdgpu_vm_bo_cs_enabled()) + return; + + for (mapping = amdgpu_vm_it_iter_first(&vm->va, 0, U64_MAX); mapping; + mapping = amdgpu_vm_it_iter_next(mapping, 0, U64_MAX)) { + if (mapping->bo_va && mapping->bo_va->base.bo) { + struct amdgpu_bo *bo; + + bo = mapping->bo_va->base.bo; + if (READ_ONCE(bo->tbo.resv->lock.ctx) != ticket) + continue; + } + + trace_amdgpu_vm_bo_cs(mapping); + } +} + +/** * amdgpu_vm_bo_rmv - remove a bo to a specific vm * * @adev: amdgpu_device pointer @@ -2211,13 +2747,27 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va) { struct amdgpu_bo_va_mapping *mapping, *next; + struct amdgpu_bo *bo = bo_va->base.bo; struct amdgpu_vm *vm = bo_va->base.vm; + struct amdgpu_vm_bo_base **base; + + if (bo) { + if (bo->tbo.resv == vm->root.base.bo->tbo.resv) + vm->bulk_moveable = false; + + for (base = &bo_va->base.bo->vm_bo; *base; + base = &(*base)->next) { + if (*base != &bo_va->base) + continue; - list_del(&bo_va->base.bo_list); + *base = bo_va->base.next; + break; + } + } - spin_lock(&vm->moved_lock); + spin_lock(&vm->invalidated_lock); list_del(&bo_va->base.vm_status); - spin_unlock(&vm->moved_lock); + spin_unlock(&vm->invalidated_lock); list_for_each_entry_safe(mapping, next, &bo_va->valids, list) { list_del(&mapping->list); @@ -2241,8 +2791,8 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, * amdgpu_vm_bo_invalidate - mark the bo as invalid * * @adev: amdgpu_device pointer - * @vm: requested vm * @bo: amdgpu buffer object + * @evicted: is the BO evicted * * Mark @bo as invalid. */ @@ -2255,33 +2805,35 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, if (bo->parent && bo->parent->shadow == bo) bo = bo->parent; - list_for_each_entry(bo_base, &bo->va, bo_list) { + for (bo_base = bo->vm_bo; bo_base; bo_base = bo_base->next) { struct amdgpu_vm *vm = bo_base->vm; - bool was_moved = bo_base->moved; - bo_base->moved = true; if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) { - if (bo->tbo.type == ttm_bo_type_kernel) - list_move(&bo_base->vm_status, &vm->evicted); - else - list_move_tail(&bo_base->vm_status, - &vm->evicted); + amdgpu_vm_bo_evicted(bo_base); continue; } - if (was_moved) + if (bo_base->moved) continue; + bo_base->moved = true; - if (bo->tbo.type == ttm_bo_type_kernel) { - list_move(&bo_base->vm_status, &vm->relocated); - } else { - spin_lock(&bo_base->vm->moved_lock); - list_move(&bo_base->vm_status, &vm->moved); - spin_unlock(&bo_base->vm->moved_lock); - } + if (bo->tbo.type == ttm_bo_type_kernel) + amdgpu_vm_bo_relocated(bo_base); + else if (bo->tbo.resv == vm->root.base.bo->tbo.resv) + amdgpu_vm_bo_moved(bo_base); + else + amdgpu_vm_bo_invalidated(bo_base); } } +/** + * amdgpu_vm_get_block_size - calculate VM page table size as power of two + * + * @vm_size: VM size + * + * Returns: + * VM page table as power of two + */ static uint32_t amdgpu_vm_get_block_size(uint64_t vm_size) { /* Total bits covered by PD + PTs */ @@ -2299,24 +2851,52 @@ static uint32_t amdgpu_vm_get_block_size(uint64_t vm_size) * amdgpu_vm_adjust_size - adjust vm size, block size and fragment size * * @adev: amdgpu_device pointer - * @vm_size: the default vm size if it's set auto + * @min_vm_size: the minimum vm size in GB if it's set auto + * @fragment_size_default: Default PTE fragment size + * @max_level: max VMPT level + * @max_bits: max address space size in bits + * */ -void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t vm_size, +void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size, uint32_t fragment_size_default, unsigned max_level, unsigned max_bits) { + unsigned int max_size = 1 << (max_bits - 30); + unsigned int vm_size; uint64_t tmp; /* adjust vm size first */ if (amdgpu_vm_size != -1) { - unsigned max_size = 1 << (max_bits - 30); - vm_size = amdgpu_vm_size; if (vm_size > max_size) { dev_warn(adev->dev, "VM size (%d) too large, max is %u GB\n", amdgpu_vm_size, max_size); vm_size = max_size; } + } else { + struct sysinfo si; + unsigned int phys_ram_gb; + + /* Optimal VM size depends on the amount of physical + * RAM available. Underlying requirements and + * assumptions: + * + * - Need to map system memory and VRAM from all GPUs + * - VRAM from other GPUs not known here + * - Assume VRAM <= system memory + * - On GFX8 and older, VM space can be segmented for + * different MTYPEs + * - Need to allow room for fragmentation, guard pages etc. + * + * This adds up to a rough guess of system memory x3. + * Round up to power of two to maximize the available + * VM size with the given page table size. + */ + si_meminfo(&si); + phys_ram_gb = ((uint64_t)si.totalram * si.mem_unit + + (1 << 30) - 1) >> 30; + vm_size = roundup_pow_of_two( + min(max(phys_ram_gb * 3, min_vm_size), max_size)); } adev->vm_manager.max_pfn = (uint64_t)vm_size << 18; @@ -2361,27 +2941,40 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t vm_size, adev->vm_manager.fragment_size); } +static struct amdgpu_retryfault_hashtable *init_fault_hash(void) +{ + struct amdgpu_retryfault_hashtable *fault_hash; + + fault_hash = kmalloc(sizeof(*fault_hash), GFP_KERNEL); + if (!fault_hash) + return fault_hash; + + INIT_CHASH_TABLE(fault_hash->hash, + AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); + spin_lock_init(&fault_hash->lock); + fault_hash->count = 0; + + return fault_hash; +} + /** * amdgpu_vm_init - initialize a vm instance * * @adev: amdgpu_device pointer * @vm: requested vm * @vm_context: Indicates if it GFX or Compute context + * @pasid: Process address space identifier * * Init @vm fields. + * + * Returns: + * 0 for success, error for failure. */ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int vm_context, unsigned int pasid) { struct amdgpu_bo_param bp; struct amdgpu_bo *root; - const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE, - AMDGPU_VM_PTE_COUNT(adev) * 8); - unsigned ring_instance; - struct amdgpu_ring *ring; - struct drm_sched_rq *rq; - unsigned long size; - uint64_t flags; int r, i; vm->va = RB_ROOT_CACHED; @@ -2389,19 +2982,15 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->reserved_vmid[i] = NULL; INIT_LIST_HEAD(&vm->evicted); INIT_LIST_HEAD(&vm->relocated); - spin_lock_init(&vm->moved_lock); INIT_LIST_HEAD(&vm->moved); INIT_LIST_HEAD(&vm->idle); + INIT_LIST_HEAD(&vm->invalidated); + spin_lock_init(&vm->invalidated_lock); INIT_LIST_HEAD(&vm->freed); /* create scheduler entity for page table updates */ - - ring_instance = atomic_inc_return(&adev->vm_manager.vm_pte_next_ring); - ring_instance %= adev->vm_manager.vm_pte_num_rings; - ring = adev->vm_manager.vm_pte_rings[ring_instance]; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL]; - r = drm_sched_entity_init(&ring->sched, &vm->entity, - rq, NULL); + r = drm_sched_entity_init(&vm->entity, adev->vm_manager.vm_pte_rqs, + adev->vm_manager.vm_pte_num_rqs, NULL); if (r) return r; @@ -2419,24 +3008,13 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, } DRM_DEBUG_DRIVER("VM update mode is %s\n", vm->use_cpu_for_update ? "CPU" : "SDMA"); - WARN_ONCE((vm->use_cpu_for_update & !amdgpu_vm_is_large_bar(adev)), + WARN_ONCE((vm->use_cpu_for_update & !amdgpu_gmc_vram_full_visible(&adev->gmc)), "CPU update of VM recommended only for large BAR system\n"); vm->last_update = NULL; - flags = AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; - if (vm->use_cpu_for_update) - flags |= AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; - else - flags |= AMDGPU_GEM_CREATE_SHADOW; - - size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level); - memset(&bp, 0, sizeof(bp)); - bp.size = size; - bp.byte_align = align; - bp.domain = AMDGPU_GEM_DOMAIN_VRAM; - bp.flags = flags; - bp.type = ttm_bo_type_kernel; - bp.resv = NULL; + amdgpu_vm_bo_param(adev, vm, adev->vm_manager.root_level, &bp); + if (vm_context == AMDGPU_VM_CONTEXT_COMPUTE) + bp.flags &= ~AMDGPU_GEM_CREATE_SHADOW; r = amdgpu_bo_create(adev, &bp, &root); if (r) goto error_free_sched_entity; @@ -2467,6 +3045,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->pasid = pasid; } + vm->fault_hash = init_fault_hash(); + if (!vm->fault_hash) { + r = -ENOMEM; + goto error_free_root; + } + INIT_KFIFO(vm->faults); vm->fault_credit = 16; @@ -2481,7 +3065,7 @@ error_free_root: vm->root.base.bo = NULL; error_free_sched_entity: - drm_sched_entity_fini(&ring->sched, &vm->entity); + drm_sched_entity_destroy(&vm->entity); return r; } @@ -2489,6 +3073,9 @@ error_free_sched_entity: /** * amdgpu_vm_make_compute - Turn a GFX VM into a compute VM * + * @adev: amdgpu_device pointer + * @vm: requested vm + * * This only works on GFX VMs that don't have any BOs added and no * page tables allocated yet. * @@ -2498,12 +3085,12 @@ error_free_sched_entity: * - pasid (old PASID is released, because compute manages its own PASIDs) * * Reinitializes the page directory to reflect the changed ATS - * setting. May leave behind an unused shadow BO for the page - * directory when switching from SDMA updates to CPU updates. + * setting. * - * Returns 0 for success, -errno for errors. + * Returns: + * 0 for success, -errno for errors. */ -int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) +int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned int pasid) { bool pte_support_ats = (adev->asic_type == CHIP_RAVEN); int r; @@ -2515,7 +3102,20 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) /* Sanity checks */ if (!RB_EMPTY_ROOT(&vm->va.rb_root) || vm->root.entries) { r = -EINVAL; - goto error; + goto unreserve_bo; + } + + if (pasid) { + unsigned long flags; + + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); + r = idr_alloc(&adev->vm_manager.pasid_idr, vm, pasid, pasid + 1, + GFP_ATOMIC); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); + + if (r == -ENOSPC) + goto unreserve_bo; + r = 0; } /* Check if PD needs to be reinitialized and do it before @@ -2526,7 +3126,7 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) adev->vm_manager.root_level, pte_support_ats); if (r) - goto error; + goto free_idr; } /* Update VM state */ @@ -2535,7 +3135,7 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) vm->pte_support_ats = pte_support_ats; DRM_DEBUG_DRIVER("VM update mode is %s\n", vm->use_cpu_for_update ? "CPU" : "SDMA"); - WARN_ONCE((vm->use_cpu_for_update & !amdgpu_vm_is_large_bar(adev)), + WARN_ONCE((vm->use_cpu_for_update & !amdgpu_gmc_vram_full_visible(&adev->gmc)), "CPU update of VM recommended only for large BAR system\n"); if (vm->pasid) { @@ -2545,42 +3145,52 @@ int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) idr_remove(&adev->vm_manager.pasid_idr, vm->pasid); spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); + /* Free the original amdgpu allocated pasid + * Will be replaced with kfd allocated pasid + */ + amdgpu_pasid_free(vm->pasid); vm->pasid = 0; } -error: + /* Free the shadow bo for compute VM */ + amdgpu_bo_unref(&vm->root.base.bo->shadow); + + if (pasid) + vm->pasid = pasid; + + goto unreserve_bo; + +free_idr: + if (pasid) { + unsigned long flags; + + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); + idr_remove(&adev->vm_manager.pasid_idr, pasid); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); + } +unreserve_bo: amdgpu_bo_unreserve(vm->root.base.bo); return r; } /** - * amdgpu_vm_free_levels - free PD/PT levels - * - * @adev: amdgpu device structure - * @parent: PD/PT starting level to free - * @level: level of parent structure + * amdgpu_vm_release_compute - release a compute vm + * @adev: amdgpu_device pointer + * @vm: a vm turned into compute vm by calling amdgpu_vm_make_compute * - * Free the page directory or page table level and all sub levels. + * This is a correspondant of amdgpu_vm_make_compute. It decouples compute + * pasid from vm. Compute should stop use of vm after this call. */ -static void amdgpu_vm_free_levels(struct amdgpu_device *adev, - struct amdgpu_vm_pt *parent, - unsigned level) +void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm) { - unsigned i, num_entries = amdgpu_vm_num_entries(adev, level); + if (vm->pasid) { + unsigned long flags; - if (parent->base.bo) { - list_del(&parent->base.bo_list); - list_del(&parent->base.vm_status); - amdgpu_bo_unref(&parent->base.bo->shadow); - amdgpu_bo_unref(&parent->base.bo); + spin_lock_irqsave(&adev->vm_manager.pasid_lock, flags); + idr_remove(&adev->vm_manager.pasid_idr, vm->pasid); + spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); } - - if (parent->entries) - for (i = 0; i < num_entries; i++) - amdgpu_vm_free_levels(adev, &parent->entries[i], - level + 1); - - kvfree(parent->entries); + vm->pasid = 0; } /** @@ -2604,7 +3214,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) /* Clear pending page faults from IH when the VM is destroyed */ while (kfifo_get(&vm->faults, &fault)) - amdgpu_ih_clear_fault(adev, fault); + amdgpu_vm_clear_fault(vm->fault_hash, fault); if (vm->pasid) { unsigned long flags; @@ -2614,7 +3224,10 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags); } - drm_sched_entity_fini(vm->entity.sched, &vm->entity); + kfree(vm->fault_hash); + vm->fault_hash = NULL; + + drm_sched_entity_destroy(&vm->entity); if (!RB_EMPTY_ROOT(&vm->va.rb_root)) { dev_err(adev->dev, "still active bo inside vm\n"); @@ -2640,8 +3253,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) if (r) { dev_err(adev->dev, "Leaking page tables because BO reservation failed\n"); } else { - amdgpu_vm_free_levels(adev, &vm->root, - adev->vm_manager.root_level); + amdgpu_vm_free_pts(adev, vm); amdgpu_bo_unreserve(root); } amdgpu_bo_unref(&root); @@ -2656,8 +3268,10 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) * @adev: amdgpu_device pointer * @pasid: PASID do identify the VM * - * This function is expected to be called in interrupt context. Returns - * true if there was fault credit, false otherwise + * This function is expected to be called in interrupt context. + * + * Returns: + * True if there was fault credit, false otherwise */ bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, unsigned int pasid) @@ -2702,7 +3316,6 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) for (i = 0; i < AMDGPU_MAX_RINGS; ++i) adev->vm_manager.seqno[i] = 0; - atomic_set(&adev->vm_manager.vm_pte_next_ring, 0); spin_lock_init(&adev->vm_manager.prt_lock); atomic_set(&adev->vm_manager.num_prt_users, 0); @@ -2711,7 +3324,7 @@ void amdgpu_vm_manager_init(struct amdgpu_device *adev) */ #ifdef CONFIG_X86_64 if (amdgpu_vm_update_mode == -1) { - if (amdgpu_vm_is_large_bar(adev)) + if (amdgpu_gmc_vram_full_visible(&adev->gmc)) adev->vm_manager.vm_update_mode = AMDGPU_VM_USE_CPU_FOR_COMPUTE; else @@ -2741,6 +3354,16 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) amdgpu_vmid_mgr_fini(adev); } +/** + * amdgpu_vm_ioctl - Manages VMID reservation for vm hubs. + * + * @dev: drm device pointer + * @data: drm_amdgpu_vm + * @filp: drm file pointer + * + * Returns: + * 0 for success, -errno for errors. + */ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) { union drm_amdgpu_vm *args = data; @@ -2764,3 +3387,117 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) return 0; } + +/** + * amdgpu_vm_get_task_info - Extracts task info for a PASID. + * + * @adev: drm device pointer + * @pasid: PASID identifier for VM + * @task_info: task_info to fill. + */ +void amdgpu_vm_get_task_info(struct amdgpu_device *adev, unsigned int pasid, + struct amdgpu_task_info *task_info) +{ + struct amdgpu_vm *vm; + + spin_lock(&adev->vm_manager.pasid_lock); + + vm = idr_find(&adev->vm_manager.pasid_idr, pasid); + if (vm) + *task_info = vm->task_info; + + spin_unlock(&adev->vm_manager.pasid_lock); +} + +/** + * amdgpu_vm_set_task_info - Sets VMs task info. + * + * @vm: vm for which to set the info + */ +void amdgpu_vm_set_task_info(struct amdgpu_vm *vm) +{ + if (!vm->task_info.pid) { + vm->task_info.pid = current->pid; + get_task_comm(vm->task_info.task_name, current); + + if (current->group_leader->mm == current->mm) { + vm->task_info.tgid = current->group_leader->pid; + get_task_comm(vm->task_info.process_name, current->group_leader); + } + } +} + +/** + * amdgpu_vm_add_fault - Add a page fault record to fault hash table + * + * @fault_hash: fault hash table + * @key: 64-bit encoding of PASID and address + * + * This should be called when a retry page fault interrupt is + * received. If this is a new page fault, it will be added to a hash + * table. The return value indicates whether this is a new fault, or + * a fault that was already known and is already being handled. + * + * If there are too many pending page faults, this will fail. Retry + * interrupts should be ignored in this case until there is enough + * free space. + * + * Returns 0 if the fault was added, 1 if the fault was already known, + * -ENOSPC if there are too many pending faults. + */ +int amdgpu_vm_add_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key) +{ + unsigned long flags; + int r = -ENOSPC; + + if (WARN_ON_ONCE(!fault_hash)) + /* Should be allocated in amdgpu_vm_init + */ + return r; + + spin_lock_irqsave(&fault_hash->lock, flags); + + /* Only let the hash table fill up to 50% for best performance */ + if (fault_hash->count >= (1 << (AMDGPU_PAGEFAULT_HASH_BITS-1))) + goto unlock_out; + + r = chash_table_copy_in(&fault_hash->hash, key, NULL); + if (!r) + fault_hash->count++; + + /* chash_table_copy_in should never fail unless we're losing count */ + WARN_ON_ONCE(r < 0); + +unlock_out: + spin_unlock_irqrestore(&fault_hash->lock, flags); + return r; +} + +/** + * amdgpu_vm_clear_fault - Remove a page fault record + * + * @fault_hash: fault hash table + * @key: 64-bit encoding of PASID and address + * + * This should be called when a page fault has been handled. Any + * future interrupt with this key will be processed as a new + * page fault. + */ +void amdgpu_vm_clear_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key) +{ + unsigned long flags; + int r; + + if (!fault_hash) + return; + + spin_lock_irqsave(&fault_hash->lock, flags); + + r = chash_table_remove(&fault_hash->hash, key, NULL); + if (!WARN_ON_ONCE(r < 0)) { + fault_hash->count--; + WARN_ON_ONCE(fault_hash->count < 0); + } + + spin_unlock_irqrestore(&fault_hash->lock, flags); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 061b99a18cb8..2a8898d19c8b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -29,6 +29,8 @@ #include <linux/rbtree.h> #include <drm/gpu_scheduler.h> #include <drm/drm_file.h> +#include <drm/ttm/ttm_bo_driver.h> +#include <linux/chash.h> #include "amdgpu_sync.h" #include "amdgpu_ring.h" @@ -48,9 +50,6 @@ struct amdgpu_bo_list_entry; /* number of entries in page table */ #define AMDGPU_VM_PTE_COUNT(adev) (1 << (adev)->vm_manager.block_size) -/* PTBs (Page Table Blocks) need to be aligned to 32K */ -#define AMDGPU_VM_PTB_ALIGN_SIZE 32768 - #define AMDGPU_PTE_VALID (1ULL << 0) #define AMDGPU_PTE_SYSTEM (1ULL << 1) #define AMDGPU_PTE_SNOOPED (1ULL << 2) @@ -103,19 +102,6 @@ struct amdgpu_bo_list_entry; /* hardcode that limit for now */ #define AMDGPU_VA_RESERVED_SIZE (1ULL << 20) -/* VA hole for 48bit addresses on Vega10 */ -#define AMDGPU_VA_HOLE_START 0x0000800000000000ULL -#define AMDGPU_VA_HOLE_END 0xffff800000000000ULL - -/* - * Hardware is programmed as if the hole doesn't exists with start and end - * address values. - * - * This mask is used to remove the upper 16bits of the VA and so come up with - * the linear addr value. - */ -#define AMDGPU_VA_HOLE_MASK 0x0000ffffffffffffULL - /* max vmids dedicated for process */ #define AMDGPU_VM_MAX_RESERVED_VMID 1 @@ -143,7 +129,7 @@ struct amdgpu_vm_bo_base { struct amdgpu_bo *bo; /* protected by bo being reserved */ - struct list_head bo_list; + struct amdgpu_vm_bo_base *next; /* protected by spinlock */ struct list_head vm_status; @@ -160,10 +146,46 @@ struct amdgpu_vm_pt { struct amdgpu_vm_pt *entries; }; +/* provided by hw blocks that can write ptes, e.g., sdma */ +struct amdgpu_vm_pte_funcs { + /* number of dw to reserve per operation */ + unsigned copy_pte_num_dw; + + /* copy pte entries from GART */ + void (*copy_pte)(struct amdgpu_ib *ib, + uint64_t pe, uint64_t src, + unsigned count); + + /* write pte one entry at a time with addr mapping */ + void (*write_pte)(struct amdgpu_ib *ib, uint64_t pe, + uint64_t value, unsigned count, + uint32_t incr); + /* for linear pte/pde updates without addr mapping */ + void (*set_pte_pde)(struct amdgpu_ib *ib, + uint64_t pe, + uint64_t addr, unsigned count, + uint32_t incr, uint64_t flags); +}; + #define AMDGPU_VM_FAULT(pasid, addr) (((u64)(pasid) << 48) | (addr)) #define AMDGPU_VM_FAULT_PASID(fault) ((u64)(fault) >> 48) #define AMDGPU_VM_FAULT_ADDR(fault) ((u64)(fault) & 0xfffffffff000ULL) + +struct amdgpu_task_info { + char process_name[TASK_COMM_LEN]; + char task_name[TASK_COMM_LEN]; + pid_t pid; + pid_t tgid; +}; + +#define AMDGPU_PAGEFAULT_HASH_BITS 8 +struct amdgpu_retryfault_hashtable { + DECLARE_CHASH_TABLE(hash, AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); + spinlock_t lock; + int count; +}; + struct amdgpu_vm { /* tree of virtual addresses mapped */ struct rb_root_cached va; @@ -174,13 +196,16 @@ struct amdgpu_vm { /* PT BOs which relocated and their parent need an update */ struct list_head relocated; - /* BOs moved, but not yet updated in the PT */ + /* per VM BOs moved, but not yet updated in the PT */ struct list_head moved; - spinlock_t moved_lock; /* All BOs of this VM not currently in the state machine */ struct list_head idle; + /* regular invalidated BOs, but not yet updated in the PT */ + struct list_head invalidated; + spinlock_t invalidated_lock; + /* BO mappings freed, but not yet updated in the PT */ struct list_head freed; @@ -215,6 +240,15 @@ struct amdgpu_vm { /* Valid while the PD is reserved or fenced */ uint64_t pd_phys_addr; + + /* Some basic info about the task */ + struct amdgpu_task_info task_info; + + /* Store positions of group of BOs */ + struct ttm_lru_bulk_move lru_bulk_move; + /* mark whether can do the bulk move */ + bool bulk_moveable; + struct amdgpu_retryfault_hashtable *fault_hash; }; struct amdgpu_vm_manager { @@ -233,10 +267,9 @@ struct amdgpu_vm_manager { /* vram base address for page table entry */ u64 vram_base_offset; /* vm pte handling */ - const struct amdgpu_vm_pte_funcs *vm_pte_funcs; - struct amdgpu_ring *vm_pte_rings[AMDGPU_MAX_RINGS]; - unsigned vm_pte_num_rings; - atomic_t vm_pte_next_ring; + const struct amdgpu_vm_pte_funcs *vm_pte_funcs; + struct drm_sched_rq *vm_pte_rqs[AMDGPU_MAX_RINGS]; + unsigned vm_pte_num_rqs; /* partial resident texture handling */ spinlock_t prt_lock; @@ -255,11 +288,16 @@ struct amdgpu_vm_manager { spinlock_t pasid_lock; }; +#define amdgpu_vm_copy_pte(adev, ib, pe, src, count) ((adev)->vm_manager.vm_pte_funcs->copy_pte((ib), (pe), (src), (count))) +#define amdgpu_vm_write_pte(adev, ib, pe, value, count, incr) ((adev)->vm_manager.vm_pte_funcs->write_pte((ib), (pe), (value), (count), (incr))) +#define amdgpu_vm_set_pte_pde(adev, ib, pe, addr, count, incr, flags) ((adev)->vm_manager.vm_pte_funcs->set_pte_pde((ib), (pe), (addr), (count), (incr), (flags))) + void amdgpu_vm_manager_init(struct amdgpu_device *adev); void amdgpu_vm_manager_fini(struct amdgpu_device *adev); int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int vm_context, unsigned int pasid); -int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm); +int amdgpu_vm_make_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned int pasid); +void amdgpu_vm_release_compute(struct amdgpu_device *adev, struct amdgpu_vm *vm); void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm); bool amdgpu_vm_pasid_fault_credit(struct amdgpu_device *adev, unsigned int pasid); @@ -307,9 +345,10 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev, uint64_t saddr, uint64_t size); struct amdgpu_bo_va_mapping *amdgpu_vm_bo_lookup_mapping(struct amdgpu_vm *vm, uint64_t addr); +void amdgpu_vm_bo_trace_cs(struct amdgpu_vm *vm, struct ww_acquire_ctx *ticket); void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, struct amdgpu_bo_va *bo_va); -void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t vm_size, +void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size, uint32_t fragment_size_default, unsigned max_level, unsigned max_bits); int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp); @@ -317,4 +356,16 @@ bool amdgpu_vm_need_pipeline_sync(struct amdgpu_ring *ring, struct amdgpu_job *job); void amdgpu_vm_check_compute_bug(struct amdgpu_device *adev); +void amdgpu_vm_get_task_info(struct amdgpu_device *adev, unsigned int pasid, + struct amdgpu_task_info *task_info); + +void amdgpu_vm_set_task_info(struct amdgpu_vm *vm); + +void amdgpu_vm_move_to_lru_tail(struct amdgpu_device *adev, + struct amdgpu_vm *vm); + +int amdgpu_vm_add_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key); + +void amdgpu_vm_clear_fault(struct amdgpu_retryfault_hashtable *fault_hash, u64 key); + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index b6333f92ba45..3f9d5d00c9b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -97,38 +97,56 @@ static u64 amdgpu_vram_mgr_vis_size(struct amdgpu_device *adev, } /** - * amdgpu_vram_mgr_bo_invisible_size - CPU invisible BO size + * amdgpu_vram_mgr_bo_visible_size - CPU visible BO size * * @bo: &amdgpu_bo buffer object (must be in VRAM) * * Returns: - * How much of the given &amdgpu_bo buffer object lies in CPU invisible VRAM. + * How much of the given &amdgpu_bo buffer object lies in CPU visible VRAM. */ -u64 amdgpu_vram_mgr_bo_invisible_size(struct amdgpu_bo *bo) +u64 amdgpu_vram_mgr_bo_visible_size(struct amdgpu_bo *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct ttm_mem_reg *mem = &bo->tbo.mem; struct drm_mm_node *nodes = mem->mm_node; unsigned pages = mem->num_pages; - u64 usage = 0; + u64 usage; - if (adev->gmc.visible_vram_size == adev->gmc.real_vram_size) - return 0; + if (amdgpu_gmc_vram_full_visible(&adev->gmc)) + return amdgpu_bo_size(bo); if (mem->start >= adev->gmc.visible_vram_size >> PAGE_SHIFT) - return amdgpu_bo_size(bo); + return 0; - while (nodes && pages) { - usage += nodes->size << PAGE_SHIFT; - usage -= amdgpu_vram_mgr_vis_size(adev, nodes); - pages -= nodes->size; - ++nodes; - } + for (usage = 0; nodes && pages; pages -= nodes->size, nodes++) + usage += amdgpu_vram_mgr_vis_size(adev, nodes); return usage; } /** + * amdgpu_vram_mgr_virt_start - update virtual start address + * + * @mem: ttm_mem_reg to update + * @node: just allocated node + * + * Calculate a virtual BO start address to easily check if everything is CPU + * accessible. + */ +static void amdgpu_vram_mgr_virt_start(struct ttm_mem_reg *mem, + struct drm_mm_node *node) +{ + unsigned long start; + + start = node->start + node->size; + if (start > mem->num_pages) + start -= mem->num_pages; + else + start = 0; + mem->start = max(mem->start, start); +} + +/** * amdgpu_vram_mgr_new - allocate new ranges * * @man: TTM memory type manager @@ -180,10 +198,25 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man, pages_left = mem->num_pages; spin_lock(&mgr->lock); - for (i = 0; i < num_nodes; ++i) { + for (i = 0; pages_left >= pages_per_node; ++i) { + unsigned long pages = rounddown_pow_of_two(pages_left); + + r = drm_mm_insert_node_in_range(mm, &nodes[i], pages, + pages_per_node, 0, + place->fpfn, lpfn, + mode); + if (unlikely(r)) + break; + + usage += nodes[i].size << PAGE_SHIFT; + vis_usage += amdgpu_vram_mgr_vis_size(adev, &nodes[i]); + amdgpu_vram_mgr_virt_start(mem, &nodes[i]); + pages_left -= pages; + } + + for (; pages_left; ++i) { unsigned long pages = min(pages_left, pages_per_node); uint32_t alignment = mem->page_alignment; - unsigned long start; if (pages == pages_per_node) alignment = pages_per_node; @@ -197,16 +230,7 @@ static int amdgpu_vram_mgr_new(struct ttm_mem_type_manager *man, usage += nodes[i].size << PAGE_SHIFT; vis_usage += amdgpu_vram_mgr_vis_size(adev, &nodes[i]); - - /* Calculate a virtual BO start address to easily check if - * everything is CPU accessible. - */ - start = nodes[i].start + nodes[i].size; - if (start > mem->num_pages) - start -= mem->num_pages; - else - start = 0; - mem->start = max(mem->start, start); + amdgpu_vram_mgr_virt_start(mem, &nodes[i]); pages_left -= pages; } spin_unlock(&mgr->lock); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c new file mode 100644 index 000000000000..897afbb348c1 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi.c @@ -0,0 +1,119 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * + */ +#include <linux/list.h> +#include "amdgpu.h" +#include "amdgpu_psp.h" + + +static DEFINE_MUTEX(xgmi_mutex); + +#define AMDGPU_MAX_XGMI_HIVE 8 +#define AMDGPU_MAX_XGMI_DEVICE_PER_HIVE 4 + +struct amdgpu_hive_info { + uint64_t hive_id; + struct list_head device_list; +}; + +static struct amdgpu_hive_info xgmi_hives[AMDGPU_MAX_XGMI_HIVE]; +static unsigned hive_count = 0; + +static struct amdgpu_hive_info *amdgpu_get_xgmi_hive(struct amdgpu_device *adev) +{ + int i; + struct amdgpu_hive_info *tmp; + + if (!adev->gmc.xgmi.hive_id) + return NULL; + for (i = 0 ; i < hive_count; ++i) { + tmp = &xgmi_hives[i]; + if (tmp->hive_id == adev->gmc.xgmi.hive_id) + return tmp; + } + if (i >= AMDGPU_MAX_XGMI_HIVE) + return NULL; + + /* initialize new hive if not exist */ + tmp = &xgmi_hives[hive_count++]; + tmp->hive_id = adev->gmc.xgmi.hive_id; + INIT_LIST_HEAD(&tmp->device_list); + return tmp; +} + +int amdgpu_xgmi_add_device(struct amdgpu_device *adev) +{ + struct psp_xgmi_topology_info tmp_topology[AMDGPU_MAX_XGMI_DEVICE_PER_HIVE]; + struct amdgpu_hive_info *hive; + struct amdgpu_xgmi *entry; + struct amdgpu_device *tmp_adev; + + int count = 0, ret = -EINVAL; + + if ((adev->asic_type < CHIP_VEGA20) || + (adev->flags & AMD_IS_APU) ) + return 0; + adev->gmc.xgmi.device_id = psp_xgmi_get_device_id(&adev->psp); + adev->gmc.xgmi.hive_id = psp_xgmi_get_hive_id(&adev->psp); + + memset(&tmp_topology[0], 0, sizeof(tmp_topology)); + mutex_lock(&xgmi_mutex); + hive = amdgpu_get_xgmi_hive(adev); + if (!hive) + goto exit; + + list_add_tail(&adev->gmc.xgmi.head, &hive->device_list); + list_for_each_entry(entry, &hive->device_list, head) + tmp_topology[count++].device_id = entry->device_id; + + ret = psp_xgmi_get_topology_info(&adev->psp, count, tmp_topology); + if (ret) { + dev_err(adev->dev, + "XGMI: Get topology failure on device %llx, hive %llx, ret %d", + adev->gmc.xgmi.device_id, + adev->gmc.xgmi.hive_id, ret); + goto exit; + } + /* Each psp need to set the latest topology */ + list_for_each_entry(tmp_adev, &hive->device_list, gmc.xgmi.head) { + ret = psp_xgmi_set_topology_info(&tmp_adev->psp, count, tmp_topology); + if (ret) { + dev_err(tmp_adev->dev, + "XGMI: Set topology failure on device %llx, hive %llx, ret %d", + tmp_adev->gmc.xgmi.device_id, + tmp_adev->gmc.xgmi.hive_id, ret); + /* To do : continue with some node failed or disable the whole hive */ + break; + } + } + if (!ret) + dev_info(adev->dev, "XGMI: Add node %d to hive 0x%llx.\n", + adev->gmc.xgmi.physical_node_id, + adev->gmc.xgmi.hive_id); + +exit: + mutex_unlock(&xgmi_mutex); + return ret; +} + + diff --git a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c index d702fb8e3427..60e2447e12c5 100644 --- a/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c +++ b/drivers/gpu/drm/amd/amdgpu/atombios_encoders.c @@ -28,6 +28,7 @@ #include <drm/amdgpu_drm.h> #include "amdgpu.h" #include "amdgpu_connectors.h" +#include "amdgpu_display.h" #include "atom.h" #include "atombios_encoders.h" #include "atombios_dp.h" diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index 7fbad2f5f0bd..79220a91abe3 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -49,10 +49,10 @@ #include "gmc/gmc_7_1_d.h" #include "gmc/gmc_7_1_sh_mask.h" -MODULE_FIRMWARE("radeon/bonaire_smc.bin"); -MODULE_FIRMWARE("radeon/bonaire_k_smc.bin"); -MODULE_FIRMWARE("radeon/hawaii_smc.bin"); -MODULE_FIRMWARE("radeon/hawaii_k_smc.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_smc.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_k_smc.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_smc.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_k_smc.bin"); #define MC_CG_ARB_FREQ_F0 0x0a #define MC_CG_ARB_FREQ_F1 0x0b @@ -951,12 +951,12 @@ static void ci_apply_state_adjust_rules(struct amdgpu_device *adev, else pi->battery_state = false; - if (adev->pm.dpm.ac_power) + if (adev->pm.ac_power) max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; else max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; - if (adev->pm.dpm.ac_power == false) { + if (adev->pm.ac_power == false) { for (i = 0; i < ps->performance_level_count; i++) { if (ps->performance_levels[i].mclk > max_limits->mclk) ps->performance_levels[i].mclk = max_limits->mclk; @@ -4078,7 +4078,7 @@ static int ci_enable_uvd_dpm(struct amdgpu_device *adev, bool enable) const struct amdgpu_clock_and_voltage_limits *max_limits; int i; - if (adev->pm.dpm.ac_power) + if (adev->pm.ac_power) max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; else max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; @@ -4127,7 +4127,7 @@ static int ci_enable_vce_dpm(struct amdgpu_device *adev, bool enable) const struct amdgpu_clock_and_voltage_limits *max_limits; int i; - if (adev->pm.dpm.ac_power) + if (adev->pm.ac_power) max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; else max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; @@ -4160,7 +4160,7 @@ static int ci_enable_samu_dpm(struct amdgpu_device *adev, bool enable) const struct amdgpu_clock_and_voltage_limits *max_limits; int i; - if (adev->pm.dpm.ac_power) + if (adev->pm.ac_power) max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; else max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; @@ -4191,7 +4191,7 @@ static int ci_enable_acp_dpm(struct amdgpu_device *adev, bool enable) const struct amdgpu_clock_and_voltage_limits *max_limits; int i; - if (adev->pm.dpm.ac_power) + if (adev->pm.ac_power) max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; else max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; @@ -5815,7 +5815,7 @@ static int ci_dpm_init_microcode(struct amdgpu_device *adev) default: BUG(); } - snprintf(fw_name, sizeof(fw_name), "radeon/%s_smc.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_smc.bin", chip_name); err = request_firmware(&adev->pm.fw, fw_name, adev->dev); if (err) goto out; @@ -5846,8 +5846,7 @@ static int ci_dpm_init(struct amdgpu_device *adev) adev->pm.dpm.priv = pi; pi->sys_pcie_mask = - (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_MASK) >> - CAIL_PCIE_LINK_SPEED_SUPPORT_SHIFT; + adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_MASK; pi->force_pcie_gen = AMDGPU_PCIE_GEN_INVALID; @@ -6278,12 +6277,12 @@ static int ci_dpm_sw_init(void *handle) int ret; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 230, + ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 230, &adev->pm.dpm.thermal.irq); if (ret) return ret; - ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 231, + ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 231, &adev->pm.dpm.thermal.irq); if (ret) return ret; @@ -6767,6 +6766,19 @@ static int ci_dpm_read_sensor(void *handle, int idx, } } +static int ci_set_powergating_by_smu(void *handle, + uint32_t block_type, bool gate) +{ + switch (block_type) { + case AMD_IP_BLOCK_TYPE_UVD: + ci_dpm_powergate_uvd(handle, gate); + break; + default: + break; + } + return 0; +} + static const struct amd_ip_funcs ci_dpm_ip_funcs = { .name = "ci_dpm", .early_init = ci_dpm_early_init, @@ -6804,7 +6816,7 @@ static const struct amd_pm_funcs ci_dpm_funcs = { .debugfs_print_current_performance_level = &ci_dpm_debugfs_print_current_performance_level, .force_performance_level = &ci_dpm_force_performance_level, .vblank_too_short = &ci_dpm_vblank_too_short, - .powergate_uvd = &ci_dpm_powergate_uvd, + .set_powergating_by_smu = &ci_set_powergating_by_smu, .set_fan_control_mode = &ci_dpm_set_fan_control_mode, .get_fan_control_mode = &ci_dpm_get_fan_control_mode, .set_fan_speed_percent = &ci_dpm_set_fan_speed_percent, diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index 8ff4c60d1b59..f41f5f57e9f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -1476,7 +1476,7 @@ static void cik_pcie_gen3_enable(struct amdgpu_device *adev) tmp |= PCIE_LC_CNTL4__LC_REDO_EQ_MASK; WREG32_PCIE(ixPCIE_LC_CNTL4, tmp); - mdelay(100); + msleep(100); /* linkctl */ pci_read_config_word(root, bridge_pos + PCI_EXP_LNKCTL, &tmp16); @@ -2002,10 +2002,12 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &cik_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v7_0_ip_block); amdgpu_device_ip_block_add(adev, &cik_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v7_2_ip_block); + amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); if (amdgpu_dpm == -1) - amdgpu_device_ip_block_add(adev, &ci_smu_ip_block); - else amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); + else + amdgpu_device_ip_block_add(adev, &ci_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); #if defined(CONFIG_DRM_AMD_DC) @@ -2014,8 +2016,6 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v8_2_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v7_2_ip_block); - amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); amdgpu_device_ip_block_add(adev, &uvd_v4_2_ip_block); amdgpu_device_ip_block_add(adev, &vce_v2_0_ip_block); break; @@ -2023,10 +2023,12 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &cik_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v7_0_ip_block); amdgpu_device_ip_block_add(adev, &cik_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v7_3_ip_block); + amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); if (amdgpu_dpm == -1) - amdgpu_device_ip_block_add(adev, &ci_smu_ip_block); - else amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); + else + amdgpu_device_ip_block_add(adev, &ci_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); #if defined(CONFIG_DRM_AMD_DC) @@ -2035,8 +2037,6 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v8_5_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v7_3_ip_block); - amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); amdgpu_device_ip_block_add(adev, &uvd_v4_2_ip_block); amdgpu_device_ip_block_add(adev, &vce_v2_0_ip_block); break; @@ -2044,6 +2044,8 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &cik_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v7_0_ip_block); amdgpu_device_ip_block_add(adev, &cik_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v7_1_ip_block); + amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); amdgpu_device_ip_block_add(adev, &kv_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -2053,8 +2055,7 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v8_1_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v7_1_ip_block); - amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); + amdgpu_device_ip_block_add(adev, &uvd_v4_2_ip_block); amdgpu_device_ip_block_add(adev, &vce_v2_0_ip_block); break; @@ -2063,6 +2064,8 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &cik_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v7_0_ip_block); amdgpu_device_ip_block_add(adev, &cik_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v7_2_ip_block); + amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); amdgpu_device_ip_block_add(adev, &kv_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -2072,8 +2075,6 @@ int cik_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v8_3_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v7_2_ip_block); - amdgpu_device_ip_block_add(adev, &cik_sdma_ip_block); amdgpu_device_ip_block_add(adev, &uvd_v4_2_ip_block); amdgpu_device_ip_block_add(adev, &vce_v2_0_ip_block); break; diff --git a/drivers/gpu/drm/amd/amdgpu/cik_ih.c b/drivers/gpu/drm/amd/amdgpu/cik_ih.c index 44d10c2172f6..b5775c6a857b 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_ih.c @@ -276,7 +276,7 @@ static void cik_ih_decode_iv(struct amdgpu_device *adev, dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - entry->client_id = AMDGPU_IH_CLIENTID_LEGACY; + entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; entry->src_data[0] = dw[1] & 0xfffffff; entry->ring_id = dw[2] & 0xff; @@ -318,7 +318,7 @@ static int cik_ih_sw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = amdgpu_ih_ring_init(adev, 64 * 1024, false); + r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false); if (r) return r; @@ -332,7 +332,7 @@ static int cik_ih_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_irq_fini(adev); - amdgpu_ih_ring_fini(adev); + amdgpu_ih_ring_fini(adev, &adev->irq.ih); amdgpu_irq_remove_domain(adev); return 0; @@ -468,8 +468,7 @@ static const struct amdgpu_ih_funcs cik_ih_funcs = { static void cik_ih_set_interrupt_funcs(struct amdgpu_device *adev) { - if (adev->irq.ih_funcs == NULL) - adev->irq.ih_funcs = &cik_ih_funcs; + adev->irq.ih_funcs = &cik_ih_funcs; } const struct amdgpu_ip_block_version cik_ih_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c index a7576255cc30..b918c8886b75 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik_sdma.c +++ b/drivers/gpu/drm/amd/amdgpu/cik_sdma.c @@ -54,16 +54,16 @@ static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev); static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev); static int cik_sdma_soft_reset(void *handle); -MODULE_FIRMWARE("radeon/bonaire_sdma.bin"); -MODULE_FIRMWARE("radeon/bonaire_sdma1.bin"); -MODULE_FIRMWARE("radeon/hawaii_sdma.bin"); -MODULE_FIRMWARE("radeon/hawaii_sdma1.bin"); -MODULE_FIRMWARE("radeon/kaveri_sdma.bin"); -MODULE_FIRMWARE("radeon/kaveri_sdma1.bin"); -MODULE_FIRMWARE("radeon/kabini_sdma.bin"); -MODULE_FIRMWARE("radeon/kabini_sdma1.bin"); -MODULE_FIRMWARE("radeon/mullins_sdma.bin"); -MODULE_FIRMWARE("radeon/mullins_sdma1.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_sdma.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_sdma1.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_sdma.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_sdma1.bin"); +MODULE_FIRMWARE("amdgpu/kaveri_sdma.bin"); +MODULE_FIRMWARE("amdgpu/kaveri_sdma1.bin"); +MODULE_FIRMWARE("amdgpu/kabini_sdma.bin"); +MODULE_FIRMWARE("amdgpu/kabini_sdma1.bin"); +MODULE_FIRMWARE("amdgpu/mullins_sdma.bin"); +MODULE_FIRMWARE("amdgpu/mullins_sdma1.bin"); u32 amdgpu_cik_gpu_check_soft_reset(struct amdgpu_device *adev); @@ -132,9 +132,9 @@ static int cik_sdma_init_microcode(struct amdgpu_device *adev) for (i = 0; i < adev->sdma.num_instances; i++) { if (i == 0) - snprintf(fw_name, sizeof(fw_name), "radeon/%s_sdma.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma.bin", chip_name); else - snprintf(fw_name, sizeof(fw_name), "radeon/%s_sdma1.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sdma1.bin", chip_name); err = request_firmware(&adev->sdma.instance[i].fw, fw_name, adev->dev); if (err) goto out; @@ -177,9 +177,8 @@ static uint64_t cik_sdma_ring_get_rptr(struct amdgpu_ring *ring) static uint64_t cik_sdma_ring_get_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - u32 me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1; - return (RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me]) & 0x3fffc) >> 2; + return (RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me]) & 0x3fffc) >> 2; } /** @@ -192,9 +191,8 @@ static uint64_t cik_sdma_ring_get_wptr(struct amdgpu_ring *ring) static void cik_sdma_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - u32 me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1; - WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me], + WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], (lower_32_bits(ring->wptr) << 2) & 0x3fffc); } @@ -248,7 +246,7 @@ static void cik_sdma_ring_emit_hdp_flush(struct amdgpu_ring *ring) SDMA_POLL_REG_MEM_EXTRA_FUNC(3)); /* == */ u32 ref_and_mask; - if (ring == &ring->adev->sdma.instance[0].ring) + if (ring->me == 0) ref_and_mask = GPU_HDP_FLUSH_DONE__SDMA0_MASK; else ref_and_mask = GPU_HDP_FLUSH_DONE__SDMA1_MASK; @@ -972,19 +970,19 @@ static int cik_sdma_sw_init(void *handle) } /* SDMA trap event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 224, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224, &adev->sdma.trap_irq); if (r) return r; /* SDMA Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 241, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 241, &adev->sdma.illegal_inst_irq); if (r) return r; /* SDMA Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 247, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 247, &adev->sdma.illegal_inst_irq); if (r) return r; @@ -1290,8 +1288,10 @@ static void cik_sdma_set_ring_funcs(struct amdgpu_device *adev) { int i; - for (i = 0; i < adev->sdma.num_instances; i++) + for (i = 0; i < adev->sdma.num_instances; i++) { adev->sdma.instance[i].ring.funcs = &cik_sdma_ring_funcs; + adev->sdma.instance[i].ring.me = i; + } } static const struct amdgpu_irq_src_funcs cik_sdma_trap_irq_funcs = { @@ -1370,10 +1370,8 @@ static const struct amdgpu_buffer_funcs cik_sdma_buffer_funcs = { static void cik_sdma_set_buffer_funcs(struct amdgpu_device *adev) { - if (adev->mman.buffer_funcs == NULL) { - adev->mman.buffer_funcs = &cik_sdma_buffer_funcs; - adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; - } + adev->mman.buffer_funcs = &cik_sdma_buffer_funcs; + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; } static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = { @@ -1386,16 +1384,16 @@ static const struct amdgpu_vm_pte_funcs cik_sdma_vm_pte_funcs = { static void cik_sdma_set_vm_pte_funcs(struct amdgpu_device *adev) { + struct drm_gpu_scheduler *sched; unsigned i; - if (adev->vm_manager.vm_pte_funcs == NULL) { - adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs; - for (i = 0; i < adev->sdma.num_instances; i++) - adev->vm_manager.vm_pte_rings[i] = - &adev->sdma.instance[i].ring; - - adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances; + adev->vm_manager.vm_pte_funcs = &cik_sdma_vm_pte_funcs; + for (i = 0; i < adev->sdma.num_instances; i++) { + sched = &adev->sdma.instance[i].ring.sched; + adev->vm_manager.vm_pte_rqs[i] = + &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; } + adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; } const struct amdgpu_ip_block_version cik_sdma_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/cz_ih.c b/drivers/gpu/drm/amd/amdgpu/cz_ih.c index 960c29e17da6..df5ac4d85a00 100644 --- a/drivers/gpu/drm/amd/amdgpu/cz_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/cz_ih.c @@ -255,7 +255,7 @@ static void cz_ih_decode_iv(struct amdgpu_device *adev, dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - entry->client_id = AMDGPU_IH_CLIENTID_LEGACY; + entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; entry->src_data[0] = dw[1] & 0xfffffff; entry->ring_id = dw[2] & 0xff; @@ -297,7 +297,7 @@ static int cz_ih_sw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = amdgpu_ih_ring_init(adev, 64 * 1024, false); + r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false); if (r) return r; @@ -311,7 +311,7 @@ static int cz_ih_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_irq_fini(adev); - amdgpu_ih_ring_fini(adev); + amdgpu_ih_ring_fini(adev, &adev->irq.ih); amdgpu_irq_remove_domain(adev); return 0; @@ -449,8 +449,7 @@ static const struct amdgpu_ih_funcs cz_ih_funcs = { static void cz_ih_set_interrupt_funcs(struct amdgpu_device *adev) { - if (adev->irq.ih_funcs == NULL) - adev->irq.ih_funcs = &cz_ih_funcs; + adev->irq.ih_funcs = &cz_ih_funcs; } const struct amdgpu_ip_block_version cz_ih_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index ada241bfeee9..4cfecdce29a3 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -31,6 +31,7 @@ #include "atombios_encoders.h" #include "amdgpu_pll.h" #include "amdgpu_connectors.h" +#include "amdgpu_display.h" #include "dce_v10_0.h" #include "dce/dce_10_0_d.h" @@ -41,6 +42,8 @@ #include "gmc/gmc_8_1_d.h" #include "gmc/gmc_8_1_sh_mask.h" +#include "ivsrcid/ivsrcid_vislands30.h" + static void dce_v10_0_set_display_funcs(struct amdgpu_device *adev); static void dce_v10_0_set_irq_funcs(struct amdgpu_device *adev); @@ -1855,15 +1858,14 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, if (unlikely(r != 0)) return r; - if (atomic) { - fb_location = amdgpu_bo_gpu_offset(abo); - } else { - r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location); + if (!atomic) { + r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM); if (unlikely(r != 0)) { amdgpu_bo_unreserve(abo); return -EINVAL; } } + fb_location = amdgpu_bo_gpu_offset(abo); amdgpu_bo_get_tiling_flags(abo, &tiling_flags); amdgpu_bo_unreserve(abo); @@ -1941,6 +1943,17 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, /* Greater 8 bpc fb needs to bypass hw-lut to retain precision */ bypass_lut = true; break; + case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_ABGR8888: + fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2); + fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0); + fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_RED_CROSSBAR, 2); + fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_BLUE_CROSSBAR, 2); +#ifdef __BIG_ENDIAN + fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP, + ENDIAN_8IN32); +#endif + break; default: DRM_ERROR("Unsupported screen format %s\n", drm_get_format_name(target_fb->format->format, &format_name)); @@ -2370,13 +2383,14 @@ static int dce_v10_0_crtc_cursor_set2(struct drm_crtc *crtc, return ret; } - ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM, &amdgpu_crtc->cursor_addr); + ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM); amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); drm_gem_object_put_unlocked(obj); return ret; } + amdgpu_crtc->cursor_addr = amdgpu_bo_gpu_offset(aobj); dce_v10_0_lock_cursor(crtc, true); @@ -2732,19 +2746,19 @@ static int dce_v10_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->mode_info.num_crtc; i++) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); if (r) return r; } - for (i = 8; i < 20; i += 2) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq); + for (i = VISLANDS30_IV_SRCID_D1_GRPH_PFLIP; i < 20; i += 2) { + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq); if (r) return r; } /* HPD hotplug */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 42, &adev->hpd_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq); if (r) return r; @@ -3556,8 +3570,7 @@ static const struct amdgpu_display_funcs dce_v10_0_display_funcs = { static void dce_v10_0_set_display_funcs(struct amdgpu_device *adev) { - if (adev->mode_info.funcs == NULL) - adev->mode_info.funcs = &dce_v10_0_display_funcs; + adev->mode_info.funcs = &dce_v10_0_display_funcs; } static const struct amdgpu_irq_src_funcs dce_v10_0_crtc_irq_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index a5b96eac3033..7c868916d90f 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -31,6 +31,7 @@ #include "atombios_encoders.h" #include "amdgpu_pll.h" #include "amdgpu_connectors.h" +#include "amdgpu_display.h" #include "dce_v11_0.h" #include "dce/dce_11_0_d.h" @@ -41,6 +42,8 @@ #include "gmc/gmc_8_1_d.h" #include "gmc/gmc_8_1_sh_mask.h" +#include "ivsrcid/ivsrcid_vislands30.h" + static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev); static void dce_v11_0_set_irq_funcs(struct amdgpu_device *adev); @@ -1897,15 +1900,14 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, if (unlikely(r != 0)) return r; - if (atomic) { - fb_location = amdgpu_bo_gpu_offset(abo); - } else { - r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location); + if (!atomic) { + r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM); if (unlikely(r != 0)) { amdgpu_bo_unreserve(abo); return -EINVAL; } } + fb_location = amdgpu_bo_gpu_offset(abo); amdgpu_bo_get_tiling_flags(abo, &tiling_flags); amdgpu_bo_unreserve(abo); @@ -1983,6 +1985,17 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, /* Greater 8 bpc fb needs to bypass hw-lut to retain precision */ bypass_lut = true; break; + case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_ABGR8888: + fb_format = REG_SET_FIELD(0, GRPH_CONTROL, GRPH_DEPTH, 2); + fb_format = REG_SET_FIELD(fb_format, GRPH_CONTROL, GRPH_FORMAT, 0); + fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_RED_CROSSBAR, 2); + fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_BLUE_CROSSBAR, 2); +#ifdef __BIG_ENDIAN + fb_swap = REG_SET_FIELD(fb_swap, GRPH_SWAP_CNTL, GRPH_ENDIAN_SWAP, + ENDIAN_8IN32); +#endif + break; default: DRM_ERROR("Unsupported screen format %s\n", drm_get_format_name(target_fb->format->format, &format_name)); @@ -2449,13 +2462,14 @@ static int dce_v11_0_crtc_cursor_set2(struct drm_crtc *crtc, return ret; } - ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM, &amdgpu_crtc->cursor_addr); + ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM); amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); drm_gem_object_put_unlocked(obj); return ret; } + amdgpu_crtc->cursor_addr = amdgpu_bo_gpu_offset(aobj); dce_v11_0_lock_cursor(crtc, true); @@ -2853,19 +2867,19 @@ static int dce_v11_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->mode_info.num_crtc; i++) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); if (r) return r; } - for (i = 8; i < 20; i += 2) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq); + for (i = VISLANDS30_IV_SRCID_D1_GRPH_PFLIP; i < 20; i += 2) { + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq); if (r) return r; } /* HPD hotplug */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 42, &adev->hpd_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_HOTPLUG_DETECT_A, &adev->hpd_irq); if (r) return r; @@ -3688,8 +3702,7 @@ static const struct amdgpu_display_funcs dce_v11_0_display_funcs = { static void dce_v11_0_set_display_funcs(struct amdgpu_device *adev) { - if (adev->mode_info.funcs == NULL) - adev->mode_info.funcs = &dce_v11_0_display_funcs; + adev->mode_info.funcs = &dce_v11_0_display_funcs; } static const struct amdgpu_irq_src_funcs dce_v11_0_crtc_irq_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index 394cc1e8fe20..17eaaba36017 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -30,6 +30,7 @@ #include "atombios_encoders.h" #include "amdgpu_pll.h" #include "amdgpu_connectors.h" +#include "amdgpu_display.h" #include "bif/bif_3_0_d.h" #include "bif/bif_3_0_sh_mask.h" @@ -1811,15 +1812,14 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc, if (unlikely(r != 0)) return r; - if (atomic) { - fb_location = amdgpu_bo_gpu_offset(abo); - } else { - r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location); + if (!atomic) { + r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM); if (unlikely(r != 0)) { amdgpu_bo_unreserve(abo); return -EINVAL; } } + fb_location = amdgpu_bo_gpu_offset(abo); amdgpu_bo_get_tiling_flags(abo, &tiling_flags); amdgpu_bo_unreserve(abo); @@ -1888,6 +1888,16 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc, /* Greater 8 bpc fb needs to bypass hw-lut to retain precision */ bypass_lut = true; break; + case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_ABGR8888: + fb_format = (GRPH_DEPTH(GRPH_DEPTH_32BPP) | + GRPH_FORMAT(GRPH_FORMAT_ARGB8888)); + fb_swap = (GRPH_RED_CROSSBAR(GRPH_RED_SEL_B) | + GRPH_BLUE_CROSSBAR(GRPH_BLUE_SEL_R)); +#ifdef __BIG_ENDIAN + fb_swap |= GRPH_ENDIAN_SWAP(GRPH_ENDIAN_8IN32); +#endif + break; default: DRM_ERROR("Unsupported screen format %s\n", drm_get_format_name(target_fb->format->format, &format_name)); @@ -2263,13 +2273,14 @@ static int dce_v6_0_crtc_cursor_set2(struct drm_crtc *crtc, return ret; } - ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM, &amdgpu_crtc->cursor_addr); + ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM); amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); drm_gem_object_put_unlocked(obj); return ret; } + amdgpu_crtc->cursor_addr = amdgpu_bo_gpu_offset(aobj); dce_v6_0_lock_cursor(crtc, true); @@ -2605,19 +2616,19 @@ static int dce_v6_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->mode_info.num_crtc; i++) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); if (r) return r; } for (i = 8; i < 20; i += 2) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq); if (r) return r; } /* HPD hotplug */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 42, &adev->hpd_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 42, &adev->hpd_irq); if (r) return r; @@ -3365,8 +3376,7 @@ static const struct amdgpu_display_funcs dce_v6_0_display_funcs = { static void dce_v6_0_set_display_funcs(struct amdgpu_device *adev) { - if (adev->mode_info.funcs == NULL) - adev->mode_info.funcs = &dce_v6_0_display_funcs; + adev->mode_info.funcs = &dce_v6_0_display_funcs; } static const struct amdgpu_irq_src_funcs dce_v6_0_crtc_irq_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index c9b9ab8f1b05..8c0576978d36 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -31,6 +31,7 @@ #include "atombios_encoders.h" #include "amdgpu_pll.h" #include "amdgpu_connectors.h" +#include "amdgpu_display.h" #include "dce_v8_0.h" #include "dce/dce_8_0_d.h" @@ -1786,15 +1787,14 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, if (unlikely(r != 0)) return r; - if (atomic) { - fb_location = amdgpu_bo_gpu_offset(abo); - } else { - r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM, &fb_location); + if (!atomic) { + r = amdgpu_bo_pin(abo, AMDGPU_GEM_DOMAIN_VRAM); if (unlikely(r != 0)) { amdgpu_bo_unreserve(abo); return -EINVAL; } } + fb_location = amdgpu_bo_gpu_offset(abo); amdgpu_bo_get_tiling_flags(abo, &tiling_flags); amdgpu_bo_unreserve(abo); @@ -1865,6 +1865,16 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, /* Greater 8 bpc fb needs to bypass hw-lut to retain precision */ bypass_lut = true; break; + case DRM_FORMAT_XBGR8888: + case DRM_FORMAT_ABGR8888: + fb_format = ((GRPH_DEPTH_32BPP << GRPH_CONTROL__GRPH_DEPTH__SHIFT) | + (GRPH_FORMAT_ARGB8888 << GRPH_CONTROL__GRPH_FORMAT__SHIFT)); + fb_swap = ((GRPH_RED_SEL_B << GRPH_SWAP_CNTL__GRPH_RED_CROSSBAR__SHIFT) | + (GRPH_BLUE_SEL_R << GRPH_SWAP_CNTL__GRPH_BLUE_CROSSBAR__SHIFT)); +#ifdef __BIG_ENDIAN + fb_swap |= (GRPH_ENDIAN_8IN32 << GRPH_SWAP_CNTL__GRPH_ENDIAN_SWAP__SHIFT); +#endif + break; default: DRM_ERROR("Unsupported screen format %s\n", drm_get_format_name(target_fb->format->format, &format_name)); @@ -2274,13 +2284,14 @@ static int dce_v8_0_crtc_cursor_set2(struct drm_crtc *crtc, return ret; } - ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM, &amdgpu_crtc->cursor_addr); + ret = amdgpu_bo_pin(aobj, AMDGPU_GEM_DOMAIN_VRAM); amdgpu_bo_unreserve(aobj); if (ret) { DRM_ERROR("Failed to pin new cursor BO (%d)\n", ret); drm_gem_object_put_unlocked(obj); return ret; } + amdgpu_crtc->cursor_addr = amdgpu_bo_gpu_offset(aobj); dce_v8_0_lock_cursor(crtc, true); @@ -2632,19 +2643,19 @@ static int dce_v8_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->mode_info.num_crtc; i++) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + 1, &adev->crtc_irq); if (r) return r; } for (i = 8; i < 20; i += 2) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i, &adev->pageflip_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i, &adev->pageflip_irq); if (r) return r; } /* HPD hotplug */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 42, &adev->hpd_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 42, &adev->hpd_irq); if (r) return r; @@ -3447,8 +3458,7 @@ static const struct amdgpu_display_funcs dce_v8_0_display_funcs = { static void dce_v8_0_set_display_funcs(struct amdgpu_device *adev) { - if (adev->mode_info.funcs == NULL) - adev->mode_info.funcs = &dce_v8_0_display_funcs; + adev->mode_info.funcs = &dce_v8_0_display_funcs; } static const struct amdgpu_irq_src_funcs dce_v8_0_crtc_irq_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c index dbf2ccd0c744..fdace004544d 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -36,6 +36,7 @@ #include "dce_v10_0.h" #include "dce_v11_0.h" #include "dce_virtual.h" +#include "ivsrcid/ivsrcid_vislands30.h" #define DCE_VIRTUAL_VBLANK_PERIOD 16666666 @@ -269,25 +270,18 @@ static int dce_virtual_early_init(void *handle) static struct drm_encoder * dce_virtual_encoder(struct drm_connector *connector) { - int enc_id = connector->encoder_ids[0]; struct drm_encoder *encoder; int i; - for (i = 0; i < DRM_CONNECTOR_MAX_ENCODER; i++) { - if (connector->encoder_ids[i] == 0) - break; - - encoder = drm_encoder_find(connector->dev, NULL, connector->encoder_ids[i]); - if (!encoder) - continue; - + drm_connector_for_each_possible_encoder(connector, encoder, i) { if (encoder->encoder_type == DRM_MODE_ENCODER_VIRTUAL) return encoder; } /* pick the first one */ - if (enc_id) - return drm_encoder_find(connector->dev, NULL, enc_id); + drm_connector_for_each_possible_encoder(connector, encoder, i) + return encoder; + return NULL; } @@ -378,7 +372,7 @@ static int dce_virtual_sw_init(void *handle) int r, i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 229, &adev->crtc_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SMU_DISP_TIMER2_TRIGGER, &adev->crtc_irq); if (r) return r; @@ -634,7 +628,7 @@ static int dce_virtual_connector_encoder_init(struct amdgpu_device *adev, drm_connector_register(connector); /* link them */ - drm_mode_connector_attach_encoder(connector, encoder); + drm_connector_attach_encoder(connector, encoder); return 0; } @@ -655,8 +649,7 @@ static const struct amdgpu_display_funcs dce_virtual_display_funcs = { static void dce_virtual_set_display_funcs(struct amdgpu_device *adev) { - if (adev->mode_info.funcs == NULL) - adev->mode_info.funcs = &dce_virtual_display_funcs; + adev->mode_info.funcs = &dce_virtual_display_funcs; } static int dce_virtual_pageflip(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c index cd6bf291a853..d76eb27945dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v6_0.c @@ -44,30 +44,30 @@ static void gfx_v6_0_set_ring_funcs(struct amdgpu_device *adev); static void gfx_v6_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v6_0_get_cu_info(struct amdgpu_device *adev); -MODULE_FIRMWARE("radeon/tahiti_pfp.bin"); -MODULE_FIRMWARE("radeon/tahiti_me.bin"); -MODULE_FIRMWARE("radeon/tahiti_ce.bin"); -MODULE_FIRMWARE("radeon/tahiti_rlc.bin"); - -MODULE_FIRMWARE("radeon/pitcairn_pfp.bin"); -MODULE_FIRMWARE("radeon/pitcairn_me.bin"); -MODULE_FIRMWARE("radeon/pitcairn_ce.bin"); -MODULE_FIRMWARE("radeon/pitcairn_rlc.bin"); - -MODULE_FIRMWARE("radeon/verde_pfp.bin"); -MODULE_FIRMWARE("radeon/verde_me.bin"); -MODULE_FIRMWARE("radeon/verde_ce.bin"); -MODULE_FIRMWARE("radeon/verde_rlc.bin"); - -MODULE_FIRMWARE("radeon/oland_pfp.bin"); -MODULE_FIRMWARE("radeon/oland_me.bin"); -MODULE_FIRMWARE("radeon/oland_ce.bin"); -MODULE_FIRMWARE("radeon/oland_rlc.bin"); - -MODULE_FIRMWARE("radeon/hainan_pfp.bin"); -MODULE_FIRMWARE("radeon/hainan_me.bin"); -MODULE_FIRMWARE("radeon/hainan_ce.bin"); -MODULE_FIRMWARE("radeon/hainan_rlc.bin"); +MODULE_FIRMWARE("amdgpu/tahiti_pfp.bin"); +MODULE_FIRMWARE("amdgpu/tahiti_me.bin"); +MODULE_FIRMWARE("amdgpu/tahiti_ce.bin"); +MODULE_FIRMWARE("amdgpu/tahiti_rlc.bin"); + +MODULE_FIRMWARE("amdgpu/pitcairn_pfp.bin"); +MODULE_FIRMWARE("amdgpu/pitcairn_me.bin"); +MODULE_FIRMWARE("amdgpu/pitcairn_ce.bin"); +MODULE_FIRMWARE("amdgpu/pitcairn_rlc.bin"); + +MODULE_FIRMWARE("amdgpu/verde_pfp.bin"); +MODULE_FIRMWARE("amdgpu/verde_me.bin"); +MODULE_FIRMWARE("amdgpu/verde_ce.bin"); +MODULE_FIRMWARE("amdgpu/verde_rlc.bin"); + +MODULE_FIRMWARE("amdgpu/oland_pfp.bin"); +MODULE_FIRMWARE("amdgpu/oland_me.bin"); +MODULE_FIRMWARE("amdgpu/oland_ce.bin"); +MODULE_FIRMWARE("amdgpu/oland_rlc.bin"); + +MODULE_FIRMWARE("amdgpu/hainan_pfp.bin"); +MODULE_FIRMWARE("amdgpu/hainan_me.bin"); +MODULE_FIRMWARE("amdgpu/hainan_ce.bin"); +MODULE_FIRMWARE("amdgpu/hainan_rlc.bin"); static u32 gfx_v6_0_get_csb_size(struct amdgpu_device *adev); static void gfx_v6_0_get_csb_buffer(struct amdgpu_device *adev, volatile u32 *buffer); @@ -335,7 +335,7 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev) default: BUG(); } - snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); if (err) goto out; @@ -346,7 +346,7 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev) adev->gfx.pfp_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); adev->gfx.pfp_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); if (err) goto out; @@ -357,7 +357,7 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev) adev->gfx.me_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); adev->gfx.me_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "radeon/%s_ce.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); if (err) goto out; @@ -368,7 +368,7 @@ static int gfx_v6_0_init_microcode(struct amdgpu_device *adev) adev->gfx.ce_fw_version = le32_to_cpu(cp_hdr->header.ucode_version); adev->gfx.ce_feature_version = le32_to_cpu(cp_hdr->ucode_feature_version); - snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); if (err) goto out; @@ -1552,7 +1552,7 @@ static void gfx_v6_0_config_init(struct amdgpu_device *adev) adev->gfx.config.double_offchip_lds_buf = 0; } -static void gfx_v6_0_gpu_init(struct amdgpu_device *adev) +static void gfx_v6_0_constants_init(struct amdgpu_device *adev) { u32 gb_addr_config = 0; u32 mc_shared_chmap, mc_arb_ramcfg; @@ -3094,15 +3094,15 @@ static int gfx_v6_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i, r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq); if (r) return r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 184, &adev->gfx.priv_reg_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 184, &adev->gfx.priv_reg_irq); if (r) return r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 185, &adev->gfx.priv_inst_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 185, &adev->gfx.priv_inst_irq); if (r) return r; @@ -3175,7 +3175,7 @@ static int gfx_v6_0_hw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - gfx_v6_0_gpu_init(adev); + gfx_v6_0_constants_init(adev); r = gfx_v6_0_rlc_resume(adev); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 42b6144c1fd5..0e72bc09939a 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -57,36 +57,36 @@ static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev); static void gfx_v7_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v7_0_set_gds_init(struct amdgpu_device *adev); -MODULE_FIRMWARE("radeon/bonaire_pfp.bin"); -MODULE_FIRMWARE("radeon/bonaire_me.bin"); -MODULE_FIRMWARE("radeon/bonaire_ce.bin"); -MODULE_FIRMWARE("radeon/bonaire_rlc.bin"); -MODULE_FIRMWARE("radeon/bonaire_mec.bin"); - -MODULE_FIRMWARE("radeon/hawaii_pfp.bin"); -MODULE_FIRMWARE("radeon/hawaii_me.bin"); -MODULE_FIRMWARE("radeon/hawaii_ce.bin"); -MODULE_FIRMWARE("radeon/hawaii_rlc.bin"); -MODULE_FIRMWARE("radeon/hawaii_mec.bin"); - -MODULE_FIRMWARE("radeon/kaveri_pfp.bin"); -MODULE_FIRMWARE("radeon/kaveri_me.bin"); -MODULE_FIRMWARE("radeon/kaveri_ce.bin"); -MODULE_FIRMWARE("radeon/kaveri_rlc.bin"); -MODULE_FIRMWARE("radeon/kaveri_mec.bin"); -MODULE_FIRMWARE("radeon/kaveri_mec2.bin"); - -MODULE_FIRMWARE("radeon/kabini_pfp.bin"); -MODULE_FIRMWARE("radeon/kabini_me.bin"); -MODULE_FIRMWARE("radeon/kabini_ce.bin"); -MODULE_FIRMWARE("radeon/kabini_rlc.bin"); -MODULE_FIRMWARE("radeon/kabini_mec.bin"); - -MODULE_FIRMWARE("radeon/mullins_pfp.bin"); -MODULE_FIRMWARE("radeon/mullins_me.bin"); -MODULE_FIRMWARE("radeon/mullins_ce.bin"); -MODULE_FIRMWARE("radeon/mullins_rlc.bin"); -MODULE_FIRMWARE("radeon/mullins_mec.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_pfp.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_me.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_ce.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_rlc.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_mec.bin"); + +MODULE_FIRMWARE("amdgpu/hawaii_pfp.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_me.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_ce.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_rlc.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_mec.bin"); + +MODULE_FIRMWARE("amdgpu/kaveri_pfp.bin"); +MODULE_FIRMWARE("amdgpu/kaveri_me.bin"); +MODULE_FIRMWARE("amdgpu/kaveri_ce.bin"); +MODULE_FIRMWARE("amdgpu/kaveri_rlc.bin"); +MODULE_FIRMWARE("amdgpu/kaveri_mec.bin"); +MODULE_FIRMWARE("amdgpu/kaveri_mec2.bin"); + +MODULE_FIRMWARE("amdgpu/kabini_pfp.bin"); +MODULE_FIRMWARE("amdgpu/kabini_me.bin"); +MODULE_FIRMWARE("amdgpu/kabini_ce.bin"); +MODULE_FIRMWARE("amdgpu/kabini_rlc.bin"); +MODULE_FIRMWARE("amdgpu/kabini_mec.bin"); + +MODULE_FIRMWARE("amdgpu/mullins_pfp.bin"); +MODULE_FIRMWARE("amdgpu/mullins_me.bin"); +MODULE_FIRMWARE("amdgpu/mullins_ce.bin"); +MODULE_FIRMWARE("amdgpu/mullins_rlc.bin"); +MODULE_FIRMWARE("amdgpu/mullins_mec.bin"); static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = { @@ -925,7 +925,7 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev) default: BUG(); } - snprintf(fw_name, sizeof(fw_name), "radeon/%s_pfp.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_pfp.bin", chip_name); err = request_firmware(&adev->gfx.pfp_fw, fw_name, adev->dev); if (err) goto out; @@ -933,7 +933,7 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev) if (err) goto out; - snprintf(fw_name, sizeof(fw_name), "radeon/%s_me.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_me.bin", chip_name); err = request_firmware(&adev->gfx.me_fw, fw_name, adev->dev); if (err) goto out; @@ -941,7 +941,7 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev) if (err) goto out; - snprintf(fw_name, sizeof(fw_name), "radeon/%s_ce.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_ce.bin", chip_name); err = request_firmware(&adev->gfx.ce_fw, fw_name, adev->dev); if (err) goto out; @@ -949,7 +949,7 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev) if (err) goto out; - snprintf(fw_name, sizeof(fw_name), "radeon/%s_mec.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); if (err) goto out; @@ -958,7 +958,7 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev) goto out; if (adev->asic_type == CHIP_KAVERI) { - snprintf(fw_name, sizeof(fw_name), "radeon/%s_mec2.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec2.bin", chip_name); err = request_firmware(&adev->gfx.mec2_fw, fw_name, adev->dev); if (err) goto out; @@ -967,7 +967,7 @@ static int gfx_v7_0_init_microcode(struct amdgpu_device *adev) goto out; } - snprintf(fw_name, sizeof(fw_name), "radeon/%s_rlc.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", chip_name); err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); if (err) goto out; @@ -1886,14 +1886,14 @@ static void gfx_v7_0_config_init(struct amdgpu_device *adev) } /** - * gfx_v7_0_gpu_init - setup the 3D engine + * gfx_v7_0_constants_init - setup the 3D engine * * @adev: amdgpu_device pointer * - * Configures the 3D engine and tiling configuration - * registers so that the 3D engine is usable. + * init the gfx constants such as the 3D engine, tiling configuration + * registers, maximum number of quad pipes, render backends... */ -static void gfx_v7_0_gpu_init(struct amdgpu_device *adev) +static void gfx_v7_0_constants_init(struct amdgpu_device *adev) { u32 sh_mem_cfg, sh_static_mem_cfg, sh_mem_base; u32 tmp; @@ -4170,15 +4170,6 @@ static void gfx_v7_0_ring_emit_gds_switch(struct amdgpu_ring *ring, uint32_t gws_base, uint32_t gws_size, uint32_t oa_base, uint32_t oa_size) { - gds_base = gds_base >> AMDGPU_GDS_SHIFT; - gds_size = gds_size >> AMDGPU_GDS_SHIFT; - - gws_base = gws_base >> AMDGPU_GWS_SHIFT; - gws_size = gws_size >> AMDGPU_GWS_SHIFT; - - oa_base = oa_base >> AMDGPU_OA_SHIFT; - oa_size = oa_size >> AMDGPU_OA_SHIFT; - /* GDS Base */ amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | @@ -4212,6 +4203,18 @@ static void gfx_v7_0_ring_emit_gds_switch(struct amdgpu_ring *ring, amdgpu_ring_write(ring, (1 << (oa_size + oa_base)) - (1 << oa_base)); } +static void gfx_v7_0_ring_soft_recovery(struct amdgpu_ring *ring, unsigned vmid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t value = 0; + + value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03); + value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01); + value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1); + value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid); + WREG32(mmSQ_CMD, value); +} + static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address) { WREG32(mmSQ_IND_INDEX, @@ -4513,18 +4516,18 @@ static int gfx_v7_0_sw_init(void *handle) adev->gfx.mec.num_queue_per_pipe = 8; /* EOP Event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq); if (r) return r; /* Privileged reg */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 184, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 184, &adev->gfx.priv_reg_irq); if (r) return r; /* Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 185, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 185, &adev->gfx.priv_inst_irq); if (r) return r; @@ -4579,25 +4582,6 @@ static int gfx_v7_0_sw_init(void *handle) } } - /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, - &adev->gds.gds_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, - &adev->gds.gws_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, - &adev->gds.oa_gfx_bo, NULL, NULL); - if (r) - return r; - adev->gfx.ce_ram_size = 0x8000; gfx_v7_0_gpu_early_init(adev); @@ -4640,7 +4624,7 @@ static int gfx_v7_0_hw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - gfx_v7_0_gpu_init(adev); + gfx_v7_0_constants_init(adev); /* init rlc */ r = gfx_v7_0_rlc_resume(adev); @@ -5088,6 +5072,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { .pad_ib = amdgpu_ring_generic_pad_ib, .emit_cntxcntl = gfx_v7_ring_emit_cntxcntl, .emit_wreg = gfx_v7_0_ring_emit_wreg, + .soft_recovery = gfx_v7_0_ring_soft_recovery, }; static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index 818874b13c99..3d0f277a6523 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -51,6 +51,8 @@ #include "smu/smu_7_1_3_d.h" +#include "ivsrcid/ivsrcid_vislands30.h" + #define GFX8_NUM_GFX_RINGS 1 #define GFX8_MEC_HPD_SIZE 2048 @@ -704,6 +706,17 @@ static const u32 stoney_mgcg_cgcg_init[] = mmCGTS_SM_CTRL_REG, 0xffffffff, 0x96940200, }; + +static const char * const sq_edc_source_names[] = { + "SQ_EDC_INFO_SOURCE_INVALID: No EDC error has occurred", + "SQ_EDC_INFO_SOURCE_INST: EDC source is Instruction Fetch", + "SQ_EDC_INFO_SOURCE_SGPR: EDC source is SGPR or SQC data return", + "SQ_EDC_INFO_SOURCE_VGPR: EDC source is VGPR", + "SQ_EDC_INFO_SOURCE_LDS: EDC source is LDS", + "SQ_EDC_INFO_SOURCE_GDS: EDC source is GDS", + "SQ_EDC_INFO_SOURCE_TA: EDC source is TA", +}; + static void gfx_v8_0_set_ring_funcs(struct amdgpu_device *adev); static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev); static void gfx_v8_0_set_gds_init(struct amdgpu_device *adev); @@ -866,26 +879,32 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) struct amdgpu_device *adev = ring->adev; struct amdgpu_ib ib; struct dma_fence *f = NULL; - uint32_t scratch; - uint32_t tmp = 0; + + unsigned int index; + uint64_t gpu_addr; + uint32_t tmp; long r; - r = amdgpu_gfx_scratch_get(adev, &scratch); + r = amdgpu_device_wb_get(adev, &index); if (r) { - DRM_ERROR("amdgpu: failed to get scratch reg (%ld).\n", r); + dev_err(adev->dev, "(%ld) failed to allocate wb slot\n", r); return r; } - WREG32(scratch, 0xCAFEDEAD); + + gpu_addr = adev->wb.gpu_addr + (index * 4); + adev->wb.wb[index] = cpu_to_le32(0xCAFEDEAD); memset(&ib, 0, sizeof(ib)); - r = amdgpu_ib_get(adev, NULL, 256, &ib); + r = amdgpu_ib_get(adev, NULL, 16, &ib); if (r) { DRM_ERROR("amdgpu: failed to get ib (%ld).\n", r); goto err1; } - ib.ptr[0] = PACKET3(PACKET3_SET_UCONFIG_REG, 1); - ib.ptr[1] = ((scratch - PACKET3_SET_UCONFIG_REG_START)); - ib.ptr[2] = 0xDEADBEEF; - ib.length_dw = 3; + ib.ptr[0] = PACKET3(PACKET3_WRITE_DATA, 3); + ib.ptr[1] = WRITE_DATA_DST_SEL(5) | WR_CONFIRM; + ib.ptr[2] = lower_32_bits(gpu_addr); + ib.ptr[3] = upper_32_bits(gpu_addr); + ib.ptr[4] = 0xDEADBEEF; + ib.length_dw = 5; r = amdgpu_ib_schedule(ring, 1, &ib, NULL, &f); if (r) @@ -900,20 +919,21 @@ static int gfx_v8_0_ring_test_ib(struct amdgpu_ring *ring, long timeout) DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); goto err2; } - tmp = RREG32(scratch); + + tmp = adev->wb.wb[index]; if (tmp == 0xDEADBEEF) { DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); r = 0; } else { - DRM_ERROR("amdgpu: ib test failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); + DRM_ERROR("ib test on ring %d failed\n", ring->idx); r = -EINVAL; } + err2: amdgpu_ib_free(adev, &ib, NULL); dma_fence_put(f); err1: - amdgpu_gfx_scratch_free(adev, scratch); + amdgpu_device_wb_free(adev, index); return r; } @@ -1094,14 +1114,14 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) tmp = (unsigned int *)((uintptr_t)rlc_hdr + le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_format_size_bytes >> 2); i++) + for (i = 0 ; i < (adev->gfx.rlc.reg_list_format_size_bytes >> 2); i++) adev->gfx.rlc.register_list_format[i] = le32_to_cpu(tmp[i]); adev->gfx.rlc.register_restore = adev->gfx.rlc.register_list_format + i; tmp = (unsigned int *)((uintptr_t)rlc_hdr + le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) + for (i = 0 ; i < (adev->gfx.rlc.reg_list_size_bytes >> 2); i++) adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); if (adev->asic_type >= CHIP_POLARIS10 && adev->asic_type <= CHIP_POLARIS12) { @@ -1153,64 +1173,61 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) } } - if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_PFP]; - info->ucode_id = AMDGPU_UCODE_ID_CP_PFP; - info->fw = adev->gfx.pfp_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_ME]; - info->ucode_id = AMDGPU_UCODE_ID_CP_ME; - info->fw = adev->gfx.me_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_CE]; - info->ucode_id = AMDGPU_UCODE_ID_CP_CE; - info->fw = adev->gfx.ce_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_PFP]; + info->ucode_id = AMDGPU_UCODE_ID_CP_PFP; + info->fw = adev->gfx.pfp_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_ME]; + info->ucode_id = AMDGPU_UCODE_ID_CP_ME; + info->fw = adev->gfx.me_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_CE]; + info->ucode_id = AMDGPU_UCODE_ID_CP_CE; + info->fw = adev->gfx.ce_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_G; + info->fw = adev->gfx.rlc_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; + info->fw = adev->gfx.mec_fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + + /* we need account JT in */ + cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE); - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_G]; - info->ucode_id = AMDGPU_UCODE_ID_RLC_G; - info->fw = adev->gfx.rlc_fw; - header = (const struct common_firmware_header *)info->fw->data; + if (amdgpu_sriov_vf(adev)) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_STORAGE]; + info->ucode_id = AMDGPU_UCODE_ID_STORAGE; + info->fw = adev->gfx.mec_fw; adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + ALIGN(le32_to_cpu(64 * PAGE_SIZE), PAGE_SIZE); + } - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; - info->fw = adev->gfx.mec_fw; + if (adev->gfx.mec2_fw) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2]; + info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2; + info->fw = adev->gfx.mec2_fw; header = (const struct common_firmware_header *)info->fw->data; adev->firmware.fw_size += ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - - /* we need account JT in */ - cp_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(cp_hdr->jt_size) << 2, PAGE_SIZE); - - if (amdgpu_sriov_vf(adev)) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_STORAGE]; - info->ucode_id = AMDGPU_UCODE_ID_STORAGE; - info->fw = adev->gfx.mec_fw; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(64 * PAGE_SIZE), PAGE_SIZE); - } - - if (adev->gfx.mec2_fw) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC2]; - info->ucode_id = AMDGPU_UCODE_ID_CP_MEC2; - info->fw = adev->gfx.mec2_fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - } - } out: @@ -1999,6 +2016,8 @@ static int gfx_v8_0_compute_ring_init(struct amdgpu_device *adev, int ring_id, return 0; } +static void gfx_v8_0_sq_irq_work_func(struct work_struct *work); + static int gfx_v8_0_sw_init(void *handle) { int i, j, k, r, ring_id; @@ -2026,28 +2045,39 @@ static int gfx_v8_0_sw_init(void *handle) adev->gfx.mec.num_pipe_per_mec = 4; adev->gfx.mec.num_queue_per_pipe = 8; - /* KIQ event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 178, &adev->gfx.kiq.irq); - if (r) - return r; - /* EOP Event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 181, &adev->gfx.eop_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_END_OF_PIPE, &adev->gfx.eop_irq); if (r) return r; /* Privileged reg */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 184, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_REG_FAULT, &adev->gfx.priv_reg_irq); if (r) return r; /* Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 185, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_PRIV_INSTR_FAULT, &adev->gfx.priv_inst_irq); if (r) return r; + /* Add CP EDC/ECC irq */ + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_CP_ECC_ERROR, + &adev->gfx.cp_ecc_error_irq); + if (r) + return r; + + /* SQ interrupts. */ + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SQ_INTERRUPT_MSG, + &adev->gfx.sq_irq); + if (r) { + DRM_ERROR("amdgpu_irq_add() for SQ failed: %d\n", r); + return r; + } + + INIT_WORK(&adev->gfx.sq_work.work, gfx_v8_0_sq_irq_work_func); + adev->gfx.gfx_current_status = AMDGPU_GFX_NORMAL_MODE; gfx_v8_0_scratch_init(adev); @@ -2123,25 +2153,6 @@ static int gfx_v8_0_sw_init(void *handle) if (r) return r; - /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, - &adev->gds.gds_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, - &adev->gds.gws_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, - &adev->gds.oa_gfx_bo, NULL, NULL); - if (r) - return r; - adev->gfx.ce_ram_size = 0x8000; r = gfx_v8_0_gpu_early_init(adev); @@ -3816,7 +3827,7 @@ static void gfx_v8_0_config_init(struct amdgpu_device *adev) } } -static void gfx_v8_0_gpu_init(struct amdgpu_device *adev) +static void gfx_v8_0_constants_init(struct amdgpu_device *adev) { u32 tmp, sh_static_mem_cfg; int i; @@ -4162,65 +4173,11 @@ static void gfx_v8_0_rlc_start(struct amdgpu_device *adev) udelay(50); } -static int gfx_v8_0_rlc_load_microcode(struct amdgpu_device *adev) -{ - const struct rlc_firmware_header_v2_0 *hdr; - const __le32 *fw_data; - unsigned i, fw_size; - - if (!adev->gfx.rlc_fw) - return -EINVAL; - - hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; - amdgpu_ucode_print_rlc_hdr(&hdr->header); - - fw_data = (const __le32 *)(adev->gfx.rlc_fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; - - WREG32(mmRLC_GPM_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmRLC_GPM_UCODE_DATA, le32_to_cpup(fw_data++)); - WREG32(mmRLC_GPM_UCODE_ADDR, adev->gfx.rlc_fw_version); - - return 0; -} - static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) { - int r; - u32 tmp; - gfx_v8_0_rlc_stop(adev); - - /* disable CG */ - tmp = RREG32(mmRLC_CGCG_CGLS_CTRL); - tmp &= ~(RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK | - RLC_CGCG_CGLS_CTRL__CGLS_EN_MASK); - WREG32(mmRLC_CGCG_CGLS_CTRL, tmp); - if (adev->asic_type == CHIP_POLARIS11 || - adev->asic_type == CHIP_POLARIS10 || - adev->asic_type == CHIP_POLARIS12 || - adev->asic_type == CHIP_VEGAM) { - tmp = RREG32(mmRLC_CGCG_CGLS_CTRL_3D); - tmp &= ~0x3; - WREG32(mmRLC_CGCG_CGLS_CTRL_3D, tmp); - } - - /* disable PG */ - WREG32(mmRLC_PG_CNTL, 0); - gfx_v8_0_rlc_reset(adev); gfx_v8_0_init_pg(adev); - - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { - /* legacy rlc firmware loading */ - r = gfx_v8_0_rlc_load_microcode(adev); - if (r) - return r; - } - gfx_v8_0_rlc_start(adev); return 0; @@ -4246,63 +4203,6 @@ static void gfx_v8_0_cp_gfx_enable(struct amdgpu_device *adev, bool enable) udelay(50); } -static int gfx_v8_0_cp_gfx_load_microcode(struct amdgpu_device *adev) -{ - const struct gfx_firmware_header_v1_0 *pfp_hdr; - const struct gfx_firmware_header_v1_0 *ce_hdr; - const struct gfx_firmware_header_v1_0 *me_hdr; - const __le32 *fw_data; - unsigned i, fw_size; - - if (!adev->gfx.me_fw || !adev->gfx.pfp_fw || !adev->gfx.ce_fw) - return -EINVAL; - - pfp_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.pfp_fw->data; - ce_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.ce_fw->data; - me_hdr = (const struct gfx_firmware_header_v1_0 *) - adev->gfx.me_fw->data; - - amdgpu_ucode_print_gfx_hdr(&pfp_hdr->header); - amdgpu_ucode_print_gfx_hdr(&ce_hdr->header); - amdgpu_ucode_print_gfx_hdr(&me_hdr->header); - - gfx_v8_0_cp_gfx_enable(adev, false); - - /* PFP */ - fw_data = (const __le32 *) - (adev->gfx.pfp_fw->data + - le32_to_cpu(pfp_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(pfp_hdr->header.ucode_size_bytes) / 4; - WREG32(mmCP_PFP_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_PFP_UCODE_DATA, le32_to_cpup(fw_data++)); - WREG32(mmCP_PFP_UCODE_ADDR, adev->gfx.pfp_fw_version); - - /* CE */ - fw_data = (const __le32 *) - (adev->gfx.ce_fw->data + - le32_to_cpu(ce_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(ce_hdr->header.ucode_size_bytes) / 4; - WREG32(mmCP_CE_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_CE_UCODE_DATA, le32_to_cpup(fw_data++)); - WREG32(mmCP_CE_UCODE_ADDR, adev->gfx.ce_fw_version); - - /* ME */ - fw_data = (const __le32 *) - (adev->gfx.me_fw->data + - le32_to_cpu(me_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(me_hdr->header.ucode_size_bytes) / 4; - WREG32(mmCP_ME_RAM_WADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_ME_RAM_DATA, le32_to_cpup(fw_data++)); - WREG32(mmCP_ME_RAM_WADDR, adev->gfx.me_fw_version); - - return 0; -} - static u32 gfx_v8_0_get_csb_size(struct amdgpu_device *adev) { u32 count = 0; @@ -4502,52 +4402,6 @@ static void gfx_v8_0_cp_compute_enable(struct amdgpu_device *adev, bool enable) udelay(50); } -static int gfx_v8_0_cp_compute_load_microcode(struct amdgpu_device *adev) -{ - const struct gfx_firmware_header_v1_0 *mec_hdr; - const __le32 *fw_data; - unsigned i, fw_size; - - if (!adev->gfx.mec_fw) - return -EINVAL; - - gfx_v8_0_cp_compute_enable(adev, false); - - mec_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec_fw->data; - amdgpu_ucode_print_gfx_hdr(&mec_hdr->header); - - fw_data = (const __le32 *) - (adev->gfx.mec_fw->data + - le32_to_cpu(mec_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(mec_hdr->header.ucode_size_bytes) / 4; - - /* MEC1 */ - WREG32(mmCP_MEC_ME1_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_MEC_ME1_UCODE_DATA, le32_to_cpup(fw_data+i)); - WREG32(mmCP_MEC_ME1_UCODE_ADDR, adev->gfx.mec_fw_version); - - /* Loading MEC2 firmware is only necessary if MEC2 should run different microcode than MEC1. */ - if (adev->gfx.mec2_fw) { - const struct gfx_firmware_header_v1_0 *mec2_hdr; - - mec2_hdr = (const struct gfx_firmware_header_v1_0 *)adev->gfx.mec2_fw->data; - amdgpu_ucode_print_gfx_hdr(&mec2_hdr->header); - - fw_data = (const __le32 *) - (adev->gfx.mec2_fw->data + - le32_to_cpu(mec2_hdr->header.ucode_array_offset_bytes)); - fw_size = le32_to_cpu(mec2_hdr->header.ucode_size_bytes) / 4; - - WREG32(mmCP_MEC_ME2_UCODE_ADDR, 0); - for (i = 0; i < fw_size; i++) - WREG32(mmCP_MEC_ME2_UCODE_DATA, le32_to_cpup(fw_data+i)); - WREG32(mmCP_MEC_ME2_UCODE_ADDR, adev->gfx.mec2_fw_version); - } - - return 0; -} - /* KIQ functions */ static void gfx_v8_0_kiq_setting(struct amdgpu_ring *ring) { @@ -4566,7 +4420,6 @@ static void gfx_v8_0_kiq_setting(struct amdgpu_ring *ring) static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) { struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; - uint32_t scratch, tmp = 0; uint64_t queue_mask = 0; int r, i; @@ -4585,17 +4438,9 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) queue_mask |= (1ull << i); } - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("Failed to get scratch reg (%d).\n", r); - return r; - } - WREG32(scratch, 0xCAFEDEAD); - - r = amdgpu_ring_alloc(kiq_ring, (8 * adev->gfx.num_compute_rings) + 11); + r = amdgpu_ring_alloc(kiq_ring, (8 * adev->gfx.num_compute_rings) + 8); if (r) { DRM_ERROR("Failed to lock KIQ (%d).\n", r); - amdgpu_gfx_scratch_free(adev, scratch); return r; } /* set resources */ @@ -4627,25 +4472,12 @@ static int gfx_v8_0_kiq_kcq_enable(struct amdgpu_device *adev) amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr)); amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); } - /* write to scratch for completion */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); - amdgpu_ring_write(kiq_ring, 0xDEADBEEF); - amdgpu_ring_commit(kiq_ring); - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i >= adev->usec_timeout) { - DRM_ERROR("KCQ enable failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); - r = -EINVAL; + r = amdgpu_ring_test_ring(kiq_ring); + if (r) { + DRM_ERROR("KCQ enable failed\n"); + kiq_ring->ready = false; } - amdgpu_gfx_scratch_free(adev, scratch); - return r; } @@ -4895,7 +4727,7 @@ static int gfx_v8_0_kcq_init_queue(struct amdgpu_ring *ring) struct vi_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->in_gpu_reset && !adev->gfx.in_suspend) { + if (!adev->in_gpu_reset && !adev->in_suspend) { memset((void *)mqd, 0, sizeof(struct vi_mqd_allocation)); ((struct vi_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct vi_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; @@ -4932,26 +4764,33 @@ static void gfx_v8_0_set_mec_doorbell_range(struct amdgpu_device *adev) static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = NULL; - int r = 0, i; - - gfx_v8_0_cp_compute_enable(adev, true); + struct amdgpu_ring *ring; + int r; ring = &adev->gfx.kiq.ring; r = amdgpu_bo_reserve(ring->mqd_obj, false); if (unlikely(r != 0)) - goto done; + return r; r = amdgpu_bo_kmap(ring->mqd_obj, &ring->mqd_ptr); - if (!r) { - r = gfx_v8_0_kiq_init_queue(ring); - amdgpu_bo_kunmap(ring->mqd_obj); - ring->mqd_ptr = NULL; - } + if (unlikely(r != 0)) + return r; + + gfx_v8_0_kiq_init_queue(ring); + amdgpu_bo_kunmap(ring->mqd_obj); + ring->mqd_ptr = NULL; amdgpu_bo_unreserve(ring->mqd_obj); - if (r) - goto done; + ring->ready = true; + return 0; +} + +static int gfx_v8_0_kcq_resume(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = NULL; + int r = 0, i; + + gfx_v8_0_cp_compute_enable(adev, true); for (i = 0; i < adev->gfx.num_compute_rings; i++) { ring = &adev->gfx.compute_ring[i]; @@ -4976,15 +4815,6 @@ static int gfx_v8_0_kiq_resume(struct amdgpu_device *adev) if (r) goto done; - /* Test KIQ */ - ring = &adev->gfx.kiq.ring; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; - goto done; - } - /* Test KCQs */ for (i = 0; i < adev->gfx.num_compute_rings; i++) { ring = &adev->gfx.compute_ring[i]; @@ -5005,25 +4835,17 @@ static int gfx_v8_0_cp_resume(struct amdgpu_device *adev) if (!(adev->flags & AMD_IS_APU)) gfx_v8_0_enable_gui_idle_interrupt(adev, false); - if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { - /* legacy firmware loading */ - r = gfx_v8_0_cp_gfx_load_microcode(adev); - if (r) - return r; - - r = gfx_v8_0_cp_compute_load_microcode(adev); - if (r) - return r; - } + r = gfx_v8_0_kiq_resume(adev); + if (r) + return r; r = gfx_v8_0_cp_gfx_resume(adev); if (r) return r; - r = gfx_v8_0_kiq_resume(adev); + r = gfx_v8_0_kcq_resume(adev); if (r) return r; - gfx_v8_0_enable_gui_idle_interrupt(adev, true); return 0; @@ -5041,7 +4863,7 @@ static int gfx_v8_0_hw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; gfx_v8_0_init_golden_registers(adev); - gfx_v8_0_gpu_init(adev); + gfx_v8_0_constants_init(adev); r = gfx_v8_0_rlc_resume(adev); if (r) @@ -5052,113 +4874,74 @@ static int gfx_v8_0_hw_init(void *handle) return r; } -static int gfx_v8_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring) +static int gfx_v8_0_kcq_disable(struct amdgpu_device *adev) { - struct amdgpu_device *adev = kiq_ring->adev; - uint32_t scratch, tmp = 0; int r, i; + struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("Failed to get scratch reg (%d).\n", r); - return r; - } - WREG32(scratch, 0xCAFEDEAD); - - r = amdgpu_ring_alloc(kiq_ring, 10); - if (r) { + r = amdgpu_ring_alloc(kiq_ring, 6 * adev->gfx.num_compute_rings); + if (r) DRM_ERROR("Failed to lock KIQ (%d).\n", r); - amdgpu_gfx_scratch_free(adev, scratch); - return r; - } - /* unmap queues */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); - amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); - amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); - amdgpu_ring_write(kiq_ring, 0); - amdgpu_ring_write(kiq_ring, 0); - amdgpu_ring_write(kiq_ring, 0); - /* write to scratch for completion */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); - amdgpu_ring_write(kiq_ring, 0xDEADBEEF); - amdgpu_ring_commit(kiq_ring); - - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i >= adev->usec_timeout) { - DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); - r = -EINVAL; + amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); } - amdgpu_gfx_scratch_free(adev, scratch); + r = amdgpu_ring_test_ring(kiq_ring); + if (r) + DRM_ERROR("KCQ disable failed\n"); + return r; } -static int gfx_v8_0_hw_fini(void *handle) +static bool gfx_v8_0_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int i; - - amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); - amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); - - /* disable KCQ to avoid CPC touch memory not valid anymore */ - for (i = 0; i < adev->gfx.num_compute_rings; i++) - gfx_v8_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]); - if (amdgpu_sriov_vf(adev)) { - pr_debug("For SRIOV client, shouldn't do anything.\n"); - return 0; - } - gfx_v8_0_cp_enable(adev, false); - gfx_v8_0_rlc_stop(adev); - - amdgpu_device_ip_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_GFX, - AMD_PG_STATE_UNGATE); - - return 0; + if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE) + || RREG32(mmGRBM_STATUS2) != 0x8) + return false; + else + return true; } -static int gfx_v8_0_suspend(void *handle) +static bool gfx_v8_0_rlc_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - adev->gfx.in_suspend = true; - return gfx_v8_0_hw_fini(adev); -} -static int gfx_v8_0_resume(void *handle) -{ - int r; - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - r = gfx_v8_0_hw_init(adev); - adev->gfx.in_suspend = false; - return r; + if (RREG32(mmGRBM_STATUS2) != 0x8) + return false; + else + return true; } -static bool gfx_v8_0_is_idle(void *handle) +static int gfx_v8_0_wait_for_rlc_idle(void *handle) { + unsigned int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (REG_GET_FIELD(RREG32(mmGRBM_STATUS), GRBM_STATUS, GUI_ACTIVE)) - return false; - else - return true; + for (i = 0; i < adev->usec_timeout; i++) { + if (gfx_v8_0_rlc_is_idle(handle)) + return 0; + + udelay(1); + } + return -ETIMEDOUT; } static int gfx_v8_0_wait_for_idle(void *handle) { - unsigned i; + unsigned int i; struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (i = 0; i < adev->usec_timeout; i++) { @@ -5170,6 +4953,47 @@ static int gfx_v8_0_wait_for_idle(void *handle) return -ETIMEDOUT; } +static int gfx_v8_0_hw_fini(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); + amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); + + amdgpu_irq_put(adev, &adev->gfx.cp_ecc_error_irq, 0); + + amdgpu_irq_put(adev, &adev->gfx.sq_irq, 0); + + /* disable KCQ to avoid CPC touch memory not valid anymore */ + gfx_v8_0_kcq_disable(adev); + + if (amdgpu_sriov_vf(adev)) { + pr_debug("For SRIOV client, shouldn't do anything.\n"); + return 0; + } + adev->gfx.rlc.funcs->enter_safe_mode(adev); + if (!gfx_v8_0_wait_for_idle(adev)) + gfx_v8_0_cp_enable(adev, false); + else + pr_err("cp is busy, skip halt cp\n"); + if (!gfx_v8_0_wait_for_rlc_idle(adev)) + gfx_v8_0_rlc_stop(adev); + else + pr_err("rlc is busy, skip halt rlc\n"); + adev->gfx.rlc.funcs->exit_safe_mode(adev); + return 0; +} + +static int gfx_v8_0_suspend(void *handle) +{ + return gfx_v8_0_hw_fini(handle); +} + +static int gfx_v8_0_resume(void *handle) +{ + return gfx_v8_0_hw_init(handle); +} + static bool gfx_v8_0_check_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -5349,10 +5173,6 @@ static int gfx_v8_0_post_soft_reset(void *handle) srbm_soft_reset = adev->gfx.srbm_soft_reset; if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || - REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) - gfx_v8_0_cp_gfx_resume(adev); - - if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPF) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPC) || REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CPG)) { @@ -5368,7 +5188,13 @@ static int gfx_v8_0_post_soft_reset(void *handle) mutex_unlock(&adev->srbm_mutex); } gfx_v8_0_kiq_resume(adev); + gfx_v8_0_kcq_resume(adev); } + + if (REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_CP) || + REG_GET_FIELD(grbm_soft_reset, GRBM_SOFT_RESET, SOFT_RESET_GFX)) + gfx_v8_0_cp_gfx_resume(adev); + gfx_v8_0_rlc_start(adev); return 0; @@ -5400,15 +5226,6 @@ static void gfx_v8_0_ring_emit_gds_switch(struct amdgpu_ring *ring, uint32_t gws_base, uint32_t gws_size, uint32_t oa_base, uint32_t oa_size) { - gds_base = gds_base >> AMDGPU_GDS_SHIFT; - gds_size = gds_size >> AMDGPU_GDS_SHIFT; - - gws_base = gws_base >> AMDGPU_GWS_SHIFT; - gws_size = gws_size >> AMDGPU_GWS_SHIFT; - - oa_base = oa_base >> AMDGPU_OA_SHIFT; - oa_size = oa_size >> AMDGPU_OA_SHIFT; - /* GDS Base */ amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | @@ -5542,9 +5359,19 @@ static int gfx_v8_0_late_init(void *handle) if (r) return r; - amdgpu_device_ip_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_GFX, - AMD_PG_STATE_GATE); + r = amdgpu_irq_get(adev, &adev->gfx.cp_ecc_error_irq, 0); + if (r) { + DRM_ERROR("amdgpu_irq_get() failed to get IRQ for EDC, r: %d.\n", r); + return r; + } + + r = amdgpu_irq_get(adev, &adev->gfx.sq_irq, 0); + if (r) { + DRM_ERROR( + "amdgpu_irq_get() failed to get IRQ for SQ, r: %d.\n", + r); + return r; + } return 0; } @@ -5552,14 +5379,12 @@ static int gfx_v8_0_late_init(void *handle) static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *adev, bool enable) { - if ((adev->asic_type == CHIP_POLARIS11) || + if (((adev->asic_type == CHIP_POLARIS11) || (adev->asic_type == CHIP_POLARIS12) || - (adev->asic_type == CHIP_VEGAM)) + (adev->asic_type == CHIP_VEGAM)) && + adev->powerplay.pp_funcs->set_powergating_by_smu) /* Send msg to SMU via Powerplay */ - amdgpu_device_ip_set_powergating_state(adev, - AMD_IP_BLOCK_TYPE_SMC, - enable ? - AMD_PG_STATE_GATE : AMD_PG_STATE_UNGATE); + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GFX, enable); WREG32_FIELD(RLC_PG_CNTL, STATIC_PER_CU_PG_ENABLE, enable ? 1 : 0); } @@ -5614,6 +5439,11 @@ static int gfx_v8_0_set_powergating_state(void *handle, if (amdgpu_sriov_vf(adev)) return 0; + if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_RLC_SMU_HS | + AMD_PG_SUPPORT_CP | + AMD_PG_SUPPORT_GFX_DMG)) + adev->gfx.rlc.funcs->enter_safe_mode(adev); switch (adev->asic_type) { case CHIP_CARRIZO: case CHIP_STONEY: @@ -5663,7 +5493,11 @@ static int gfx_v8_0_set_powergating_state(void *handle, default: break; } - + if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_SMG | + AMD_PG_SUPPORT_RLC_SMU_HS | + AMD_PG_SUPPORT_CP | + AMD_PG_SUPPORT_GFX_DMG)) + adev->gfx.rlc.funcs->exit_safe_mode(adev); return 0; } @@ -6668,6 +6502,18 @@ static void gfx_v8_0_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, amdgpu_ring_write(ring, val); } +static void gfx_v8_0_ring_soft_recovery(struct amdgpu_ring *ring, unsigned vmid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t value = 0; + + value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03); + value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01); + value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1); + value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid); + WREG32(mmSQ_CMD, value); +} + static void gfx_v8_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { @@ -6787,6 +6633,77 @@ static int gfx_v8_0_set_eop_interrupt_state(struct amdgpu_device *adev, return 0; } +static int gfx_v8_0_set_cp_ecc_int_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + int enable_flag; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + enable_flag = 0; + break; + + case AMDGPU_IRQ_STATE_ENABLE: + enable_flag = 1; + break; + + default: + return -EINVAL; + } + + WREG32_FIELD(CP_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_INT_CNTL_RING0, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_INT_CNTL_RING1, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_INT_CNTL_RING2, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CPC_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, enable_flag); + WREG32_FIELD(CP_ME1_PIPE0_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME1_PIPE1_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME1_PIPE2_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME1_PIPE3_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE0_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE1_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE2_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + WREG32_FIELD(CP_ME2_PIPE3_INT_CNTL, CP_ECC_ERROR_INT_ENABLE, + enable_flag); + + return 0; +} + +static int gfx_v8_0_set_sq_int_state(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + unsigned int type, + enum amdgpu_interrupt_state state) +{ + int enable_flag; + + switch (state) { + case AMDGPU_IRQ_STATE_DISABLE: + enable_flag = 1; + break; + + case AMDGPU_IRQ_STATE_ENABLE: + enable_flag = 0; + break; + + default: + return -EINVAL; + } + + WREG32_FIELD(SQ_INTERRUPT_MSG_CTRL, STALL, + enable_flag); + + return 0; +} + static int gfx_v8_0_eop_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) @@ -6837,49 +6754,111 @@ static int gfx_v8_0_priv_inst_irq(struct amdgpu_device *adev, return 0; } -static int gfx_v8_0_kiq_set_interrupt_state(struct amdgpu_device *adev, - struct amdgpu_irq_src *src, - unsigned int type, - enum amdgpu_interrupt_state state) +static int gfx_v8_0_cp_ecc_error_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) { - struct amdgpu_ring *ring = &(adev->gfx.kiq.ring); + DRM_ERROR("CP EDC/ECC error detected."); + return 0; +} - switch (type) { - case AMDGPU_CP_KIQ_IRQ_DRIVER0: - WREG32_FIELD(CPC_INT_CNTL, GENERIC2_INT_ENABLE, - state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); - if (ring->me == 1) - WREG32_FIELD_OFFSET(CP_ME1_PIPE0_INT_CNTL, - ring->pipe, - GENERIC2_INT_ENABLE, - state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); - else - WREG32_FIELD_OFFSET(CP_ME2_PIPE0_INT_CNTL, - ring->pipe, - GENERIC2_INT_ENABLE, - state == AMDGPU_IRQ_STATE_DISABLE ? 0 : 1); - break; - default: - BUG(); /* kiq only support GENERIC2_INT now */ - break; +static void gfx_v8_0_parse_sq_irq(struct amdgpu_device *adev, unsigned ih_data) +{ + u32 enc, se_id, sh_id, cu_id; + char type[20]; + int sq_edc_source = -1; + + enc = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_CMN, ENCODING); + se_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_CMN, SE_ID); + + switch (enc) { + case 0: + DRM_INFO("SQ general purpose intr detected:" + "se_id %d, immed_overflow %d, host_reg_overflow %d," + "host_cmd_overflow %d, cmd_timestamp %d," + "reg_timestamp %d, thread_trace_buff_full %d," + "wlt %d, thread_trace %d.\n", + se_id, + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, IMMED_OVERFLOW), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, HOST_REG_OVERFLOW), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, HOST_CMD_OVERFLOW), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, CMD_TIMESTAMP), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, REG_TIMESTAMP), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, THREAD_TRACE_BUF_FULL), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, WLT), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_AUTO, THREAD_TRACE) + ); + break; + case 1: + case 2: + + cu_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, CU_ID); + sh_id = REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, SH_ID); + + /* + * This function can be called either directly from ISR + * or from BH in which case we can access SQ_EDC_INFO + * instance + */ + if (in_task()) { + mutex_lock(&adev->grbm_idx_mutex); + gfx_v8_0_select_se_sh(adev, se_id, sh_id, cu_id); + + sq_edc_source = REG_GET_FIELD(RREG32(mmSQ_EDC_INFO), SQ_EDC_INFO, SOURCE); + + gfx_v8_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + mutex_unlock(&adev->grbm_idx_mutex); + } + + if (enc == 1) + sprintf(type, "instruction intr"); + else + sprintf(type, "EDC/ECC error"); + + DRM_INFO( + "SQ %s detected: " + "se_id %d, sh_id %d, cu_id %d, simd_id %d, wave_id %d, vm_id %d " + "trap %s, sq_ed_info.source %s.\n", + type, se_id, sh_id, cu_id, + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, SIMD_ID), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, WAVE_ID), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, VM_ID), + REG_GET_FIELD(ih_data, SQ_INTERRUPT_WORD_WAVE, PRIV) ? "true" : "false", + (sq_edc_source != -1) ? sq_edc_source_names[sq_edc_source] : "unavailable" + ); + break; + default: + DRM_ERROR("SQ invalid encoding type\n."); } - return 0; } -static int gfx_v8_0_kiq_irq(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - struct amdgpu_iv_entry *entry) +static void gfx_v8_0_sq_irq_work_func(struct work_struct *work) { - u8 me_id, pipe_id, queue_id; - struct amdgpu_ring *ring = &(adev->gfx.kiq.ring); - me_id = (entry->ring_id & 0x0c) >> 2; - pipe_id = (entry->ring_id & 0x03) >> 0; - queue_id = (entry->ring_id & 0x70) >> 4; - DRM_DEBUG("IH: CPC GENERIC2_INT, me:%d, pipe:%d, queue:%d\n", - me_id, pipe_id, queue_id); + struct amdgpu_device *adev = container_of(work, struct amdgpu_device, gfx.sq_work.work); + struct sq_work *sq_work = container_of(work, struct sq_work, work); + + gfx_v8_0_parse_sq_irq(adev, sq_work->ih_data); +} + +static int gfx_v8_0_sq_irq(struct amdgpu_device *adev, + struct amdgpu_irq_src *source, + struct amdgpu_iv_entry *entry) +{ + unsigned ih_data = entry->src_data[0]; + + /* + * Try to submit work so SQ_EDC_INFO can be accessed from + * BH. If previous work submission hasn't finished yet + * just print whatever info is possible directly from the ISR. + */ + if (work_pending(&adev->gfx.sq_work.work)) { + gfx_v8_0_parse_sq_irq(adev, ih_data); + } else { + adev->gfx.sq_work.ih_data = ih_data; + schedule_work(&adev->gfx.sq_work.work); + } - amdgpu_fence_process(ring); return 0; } @@ -6946,6 +6925,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_gfx = { .init_cond_exec = gfx_v8_0_ring_emit_init_cond_exec, .patch_cond_exec = gfx_v8_0_ring_emit_patch_cond_exec, .emit_wreg = gfx_v8_0_ring_emit_wreg, + .soft_recovery = gfx_v8_0_ring_soft_recovery, }; static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { @@ -7032,9 +7012,14 @@ static const struct amdgpu_irq_src_funcs gfx_v8_0_priv_inst_irq_funcs = { .process = gfx_v8_0_priv_inst_irq, }; -static const struct amdgpu_irq_src_funcs gfx_v8_0_kiq_irq_funcs = { - .set = gfx_v8_0_kiq_set_interrupt_state, - .process = gfx_v8_0_kiq_irq, +static const struct amdgpu_irq_src_funcs gfx_v8_0_cp_ecc_error_irq_funcs = { + .set = gfx_v8_0_set_cp_ecc_int_state, + .process = gfx_v8_0_cp_ecc_error_irq, +}; + +static const struct amdgpu_irq_src_funcs gfx_v8_0_sq_irq_funcs = { + .set = gfx_v8_0_set_sq_int_state, + .process = gfx_v8_0_sq_irq, }; static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev) @@ -7048,8 +7033,11 @@ static void gfx_v8_0_set_irq_funcs(struct amdgpu_device *adev) adev->gfx.priv_inst_irq.num_types = 1; adev->gfx.priv_inst_irq.funcs = &gfx_v8_0_priv_inst_irq_funcs; - adev->gfx.kiq.irq.num_types = AMDGPU_CP_KIQ_IRQ_LAST; - adev->gfx.kiq.irq.funcs = &gfx_v8_0_kiq_irq_funcs; + adev->gfx.cp_ecc_error_irq.num_types = 1; + adev->gfx.cp_ecc_error_irq.funcs = &gfx_v8_0_cp_ecc_error_irq_funcs; + + adev->gfx.sq_irq.num_types = 1; + adev->gfx.sq_irq.funcs = &gfx_v8_0_sq_irq_funcs; } static void gfx_v8_0_set_rlc_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index a69153435ea7..6d7baf59d6e1 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -38,6 +38,8 @@ #include "clearstate_gfx9.h" #include "v9_structs.h" +#include "ivsrcid/gfx/irqsrcs_gfx_9_0.h" + #define GFX9_NUM_GFX_RINGS 1 #define GFX9_MEC_HPD_SIZE 2048 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L @@ -78,9 +80,24 @@ MODULE_FIRMWARE("amdgpu/raven_mec.bin"); MODULE_FIRMWARE("amdgpu/raven_mec2.bin"); MODULE_FIRMWARE("amdgpu/raven_rlc.bin"); +MODULE_FIRMWARE("amdgpu/picasso_ce.bin"); +MODULE_FIRMWARE("amdgpu/picasso_pfp.bin"); +MODULE_FIRMWARE("amdgpu/picasso_me.bin"); +MODULE_FIRMWARE("amdgpu/picasso_mec.bin"); +MODULE_FIRMWARE("amdgpu/picasso_mec2.bin"); +MODULE_FIRMWARE("amdgpu/picasso_rlc.bin"); + +MODULE_FIRMWARE("amdgpu/raven2_ce.bin"); +MODULE_FIRMWARE("amdgpu/raven2_pfp.bin"); +MODULE_FIRMWARE("amdgpu/raven2_me.bin"); +MODULE_FIRMWARE("amdgpu/raven2_mec.bin"); +MODULE_FIRMWARE("amdgpu/raven2_mec2.bin"); +MODULE_FIRMWARE("amdgpu/raven2_rlc.bin"); + static const struct soc15_reg_golden golden_settings_gc_9_0[] = { - SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG2, 0xf00fffff, 0x00000420), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG2, 0xf00fffff, 0x00000400), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG3, 0x80000000, 0x80000000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_GPU_ID, 0x0000000f, 0x00000000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_BINNER_EVENT_CNTL_3, 0x00000003, 0x82400024), SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE, 0x3fffffff, 0x00000001), @@ -102,11 +119,25 @@ static const struct soc15_reg_golden golden_settings_gc_9_0_vg10[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL, 0x0000f000, 0x00012107), SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_3, 0x30000000, 0x10000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPC_UTCL1_CNTL, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_UTCL1_CNTL, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPG_UTCL1_CNTL, 0x08000000, 0x08000080), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG, 0xffff77ff, 0x2a114042), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG_READ, 0xffff77ff, 0x2a114042), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmIA_UTCL1_CNTL, 0x08000000, 0x08000080), SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_1, 0x00008000, 0x00048000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_0, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_1, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_2, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_PREWALKER_UTCL1_CNTL, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_UTCL1_CNTL, 0x08000000, 0x08000080), SOC15_REG_GOLDEN_VALUE(GC, 0, mmRMI_UTCL1_CNTL2, 0x00030000, 0x00020000), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x00001800, 0x00000800) + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL_1, 0x0000000f, 0x01000107), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x00001800, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmWD_UTCL1_CNTL, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC1_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC2_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_DEBUG, 0x00000000, 0x00008000) }; static const struct soc15_reg_golden golden_settings_gc_9_0_vg20[] = @@ -146,7 +177,10 @@ static const struct soc15_reg_golden golden_settings_gc_9_1[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_LO, 0xffffffff, 0x00003120), SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_CACHE_INVALIDATION, 0x3fff3af3, 0x19200000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_GS_MAX_WAVE_ID, 0x00000fff, 0x000000ff), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmWD_UTCL1_CNTL, 0x08000000, 0x08000080) + SOC15_REG_GOLDEN_VALUE(GC, 0, mmWD_UTCL1_CNTL, 0x08000000, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC1_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC2_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_DEBUG, 0x00000000, 0x00008000) }; static const struct soc15_reg_golden golden_settings_gc_9_1_rv1[] = @@ -160,6 +194,29 @@ static const struct soc15_reg_golden golden_settings_gc_9_1_rv1[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x01bd9f33, 0x00000800) }; +static const struct soc15_reg_golden golden_settings_gc_9_1_rv2[] = +{ + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_DCC_CONFIG, 0xff7fffff, 0x04000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL, 0xfffdf3cf, 0x00014104), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_2, 0xff7fffff, 0x0a000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPC_UTCL1_CNTL, 0x7f0fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_UTCL1_CNTL, 0xff8fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPG_UTCL1_CNTL, 0x7f8fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG, 0xffff77ff, 0x26013041), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG_READ, 0xffff77ff, 0x26013041), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmIA_UTCL1_CNTL, 0x3f8fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_1, 0xffffffff, 0x04040000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_0, 0xff0fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_1, 0xff0fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_2, 0xff0fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_PREWALKER_UTCL1_CNTL, 0xff0fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_UTCL1_CNTL, 0xff0fffff, 0x08000080), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_HI, 0xffffffff, 0x00000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_LO, 0xffffffff, 0x00000010), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x01bd9f33, 0x01000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmWD_UTCL1_CNTL, 0x3f8fffff, 0x08000080), +}; + static const struct soc15_reg_golden golden_settings_gc_9_x_common[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, mmGRBM_CAM_INDEX, 0xffffffff, 0x00000000), @@ -197,7 +254,10 @@ static const struct soc15_reg_golden golden_settings_gc_9_2_1_vg12[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL_1, 0xffff03ff, 0x01000107), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_HI, 0xffffffff, 0x00000000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_LO, 0xffffffff, 0x76325410), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x01bd9f33, 0x01000000) + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x01bd9f33, 0x01000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC1_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_MEC2_F32_INT_DIS, 0x00000000, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCP_DEBUG, 0x00000000, 0x00008000) }; static const u32 GFX_RLC_SRM_INDEX_CNTL_ADDR_OFFSETS[] = @@ -227,6 +287,7 @@ static const u32 GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[] = #define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 #define VEGA12_GB_ADDR_CONFIG_GOLDEN 0x24104041 #define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042 +#define RAVEN2_GB_ADDR_CONFIG_GOLDEN 0x26013041 static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev); static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev); @@ -266,12 +327,16 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) ARRAY_SIZE(golden_settings_gc_9_0_vg20)); break; case CHIP_RAVEN: - soc15_program_register_sequence(adev, - golden_settings_gc_9_1, - ARRAY_SIZE(golden_settings_gc_9_1)); - soc15_program_register_sequence(adev, - golden_settings_gc_9_1_rv1, - ARRAY_SIZE(golden_settings_gc_9_1_rv1)); + soc15_program_register_sequence(adev, golden_settings_gc_9_1, + ARRAY_SIZE(golden_settings_gc_9_1)); + if (adev->rev_id >= 8) + soc15_program_register_sequence(adev, + golden_settings_gc_9_1_rv2, + ARRAY_SIZE(golden_settings_gc_9_1_rv2)); + else + soc15_program_register_sequence(adev, + golden_settings_gc_9_1_rv1, + ARRAY_SIZE(golden_settings_gc_9_1_rv1)); break; default: break; @@ -469,6 +534,61 @@ static void gfx_v9_0_init_rlc_ext_microcode(struct amdgpu_device *adev) le32_to_cpu(rlc_hdr->reg_list_format_direct_reg_list_length); } +static void gfx_v9_0_check_fw_write_wait(struct amdgpu_device *adev) +{ + adev->gfx.me_fw_write_wait = false; + adev->gfx.mec_fw_write_wait = false; + + switch (adev->asic_type) { + case CHIP_VEGA10: + if ((adev->gfx.me_fw_version >= 0x0000009c) && + (adev->gfx.me_feature_version >= 42) && + (adev->gfx.pfp_fw_version >= 0x000000b1) && + (adev->gfx.pfp_feature_version >= 42)) + adev->gfx.me_fw_write_wait = true; + + if ((adev->gfx.mec_fw_version >= 0x00000193) && + (adev->gfx.mec_feature_version >= 42)) + adev->gfx.mec_fw_write_wait = true; + break; + case CHIP_VEGA12: + if ((adev->gfx.me_fw_version >= 0x0000009c) && + (adev->gfx.me_feature_version >= 44) && + (adev->gfx.pfp_fw_version >= 0x000000b2) && + (adev->gfx.pfp_feature_version >= 44)) + adev->gfx.me_fw_write_wait = true; + + if ((adev->gfx.mec_fw_version >= 0x00000196) && + (adev->gfx.mec_feature_version >= 44)) + adev->gfx.mec_fw_write_wait = true; + break; + case CHIP_VEGA20: + if ((adev->gfx.me_fw_version >= 0x0000009c) && + (adev->gfx.me_feature_version >= 44) && + (adev->gfx.pfp_fw_version >= 0x000000b2) && + (adev->gfx.pfp_feature_version >= 44)) + adev->gfx.me_fw_write_wait = true; + + if ((adev->gfx.mec_fw_version >= 0x00000197) && + (adev->gfx.mec_feature_version >= 44)) + adev->gfx.mec_fw_write_wait = true; + break; + case CHIP_RAVEN: + if ((adev->gfx.me_fw_version >= 0x0000009c) && + (adev->gfx.me_feature_version >= 42) && + (adev->gfx.pfp_fw_version >= 0x000000b1) && + (adev->gfx.pfp_feature_version >= 42)) + adev->gfx.me_fw_write_wait = true; + + if ((adev->gfx.mec_fw_version >= 0x00000192) && + (adev->gfx.mec_feature_version >= 42)) + adev->gfx.mec_fw_write_wait = true; + break; + default: + break; + } +} + static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; @@ -496,7 +616,12 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) chip_name = "vega20"; break; case CHIP_RAVEN: - chip_name = "raven"; + if (adev->rev_id >= 8) + chip_name = "raven2"; + else if (adev->pdev->device == 0x15d8) + chip_name = "picasso"; + else + chip_name = "raven"; break; default: BUG(); @@ -577,14 +702,14 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) tmp = (unsigned int *)((uintptr_t)rlc_hdr + le32_to_cpu(rlc_hdr->reg_list_format_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_format_size_bytes >> 2); i++) + for (i = 0 ; i < (adev->gfx.rlc.reg_list_format_size_bytes >> 2); i++) adev->gfx.rlc.register_list_format[i] = le32_to_cpu(tmp[i]); adev->gfx.rlc.register_restore = adev->gfx.rlc.register_list_format + i; tmp = (unsigned int *)((uintptr_t)rlc_hdr + le32_to_cpu(rlc_hdr->reg_list_array_offset_bytes)); - for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) + for (i = 0 ; i < (adev->gfx.rlc.reg_list_size_bytes >> 2); i++) adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); if (adev->gfx.rlc.is_rlc_v2_1) @@ -648,7 +773,10 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) adev->firmware.fw_size += ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - if (adev->gfx.rlc.is_rlc_v2_1) { + if (adev->gfx.rlc.is_rlc_v2_1 && + adev->gfx.rlc.save_restore_list_cntl_size_bytes && + adev->gfx.rlc.save_restore_list_gpm_size_bytes && + adev->gfx.rlc.save_restore_list_srm_size_bytes) { info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL]; info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL; info->fw = adev->gfx.rlc_fw; @@ -700,6 +828,7 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) } out: + gfx_v9_0_check_fw_write_wait(adev); if (err) { dev_err(adev->dev, "gfx9: Failed to load firmware \"%s\"\n", @@ -789,6 +918,50 @@ static void gfx_v9_0_get_csb_buffer(struct amdgpu_device *adev, buffer[count++] = cpu_to_le32(0); } +static void gfx_v9_0_init_always_on_cu_mask(struct amdgpu_device *adev) +{ + struct amdgpu_cu_info *cu_info = &adev->gfx.cu_info; + uint32_t pg_always_on_cu_num = 2; + uint32_t always_on_cu_num; + uint32_t i, j, k; + uint32_t mask, cu_bitmap, counter; + + if (adev->flags & AMD_IS_APU) + always_on_cu_num = 4; + else if (adev->asic_type == CHIP_VEGA12) + always_on_cu_num = 8; + else + always_on_cu_num = 12; + + mutex_lock(&adev->grbm_idx_mutex); + for (i = 0; i < adev->gfx.config.max_shader_engines; i++) { + for (j = 0; j < adev->gfx.config.max_sh_per_se; j++) { + mask = 1; + cu_bitmap = 0; + counter = 0; + gfx_v9_0_select_se_sh(adev, i, j, 0xffffffff); + + for (k = 0; k < adev->gfx.config.max_cu_per_sh; k ++) { + if (cu_info->bitmap[i][j] & mask) { + if (counter == pg_always_on_cu_num) + WREG32_SOC15(GC, 0, mmRLC_PG_ALWAYS_ON_CU_MASK, cu_bitmap); + if (counter < always_on_cu_num) + cu_bitmap |= mask; + else + break; + counter++; + } + mask <<= 1; + } + + WREG32_SOC15(GC, 0, mmRLC_LB_ALWAYS_ACTIVE_CU_MASK, cu_bitmap); + cu_info->ao_cu_bitmap[i][j] = cu_bitmap; + } + } + gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + mutex_unlock(&adev->grbm_idx_mutex); +} + static void gfx_v9_0_init_lbpw(struct amdgpu_device *adev) { uint32_t data; @@ -822,8 +995,10 @@ static void gfx_v9_0_init_lbpw(struct amdgpu_device *adev) data |= 0x00C00000; WREG32_SOC15(GC, 0, mmRLC_GPM_GENERAL_7, data); - /* set RLC_LB_ALWAYS_ACTIVE_CU_MASK = 0xFFF */ - WREG32_SOC15(GC, 0, mmRLC_LB_ALWAYS_ACTIVE_CU_MASK, 0xFFF); + /* + * RLC_LB_ALWAYS_ACTIVE_CU_MASK = 0xF (4 CUs AON for Raven), + * programmed in gfx_v9_0_init_always_on_cu_mask() + */ /* set RLC_LB_CNTL = 0x8000_0095, 31 bit is reserved, * but used for RLC_LB_CNTL configuration */ @@ -832,6 +1007,57 @@ static void gfx_v9_0_init_lbpw(struct amdgpu_device *adev) data |= REG_SET_FIELD(data, RLC_LB_CNTL, RESERVED, 0x80000); WREG32_SOC15(GC, 0, mmRLC_LB_CNTL, data); mutex_unlock(&adev->grbm_idx_mutex); + + gfx_v9_0_init_always_on_cu_mask(adev); +} + +static void gfx_v9_4_init_lbpw(struct amdgpu_device *adev) +{ + uint32_t data; + + /* set mmRLC_LB_THR_CONFIG_1/2/3/4 */ + WREG32_SOC15(GC, 0, mmRLC_LB_THR_CONFIG_1, 0x0000007F); + WREG32_SOC15(GC, 0, mmRLC_LB_THR_CONFIG_2, 0x033388F8); + WREG32_SOC15(GC, 0, mmRLC_LB_THR_CONFIG_3, 0x00000077); + WREG32_SOC15(GC, 0, mmRLC_LB_THR_CONFIG_4, (0x10 | 0x27 << 8 | 0x02FA << 16)); + + /* set mmRLC_LB_CNTR_INIT = 0x0000_0000 */ + WREG32_SOC15(GC, 0, mmRLC_LB_CNTR_INIT, 0x00000000); + + /* set mmRLC_LB_CNTR_MAX = 0x0000_0500 */ + WREG32_SOC15(GC, 0, mmRLC_LB_CNTR_MAX, 0x00000800); + + mutex_lock(&adev->grbm_idx_mutex); + /* set mmRLC_LB_INIT_CU_MASK thru broadcast mode to enable all SE/SH*/ + gfx_v9_0_select_se_sh(adev, 0xffffffff, 0xffffffff, 0xffffffff); + WREG32_SOC15(GC, 0, mmRLC_LB_INIT_CU_MASK, 0xffffffff); + + /* set mmRLC_LB_PARAMS = 0x003F_1006 */ + data = REG_SET_FIELD(0, RLC_LB_PARAMS, FIFO_SAMPLES, 0x0003); + data |= REG_SET_FIELD(data, RLC_LB_PARAMS, PG_IDLE_SAMPLES, 0x0010); + data |= REG_SET_FIELD(data, RLC_LB_PARAMS, PG_IDLE_SAMPLE_INTERVAL, 0x033F); + WREG32_SOC15(GC, 0, mmRLC_LB_PARAMS, data); + + /* set mmRLC_GPM_GENERAL_7[31-16] = 0x00C0 */ + data = RREG32_SOC15(GC, 0, mmRLC_GPM_GENERAL_7); + data &= 0x0000FFFF; + data |= 0x00C00000; + WREG32_SOC15(GC, 0, mmRLC_GPM_GENERAL_7, data); + + /* + * RLC_LB_ALWAYS_ACTIVE_CU_MASK = 0xFFF (12 CUs AON), + * programmed in gfx_v9_0_init_always_on_cu_mask() + */ + + /* set RLC_LB_CNTL = 0x8000_0095, 31 bit is reserved, + * but used for RLC_LB_CNTL configuration */ + data = RLC_LB_CNTL__LB_CNT_SPIM_ACTIVE_MASK; + data |= REG_SET_FIELD(data, RLC_LB_CNTL, CU_MASK_USED_OFF_HYST, 0x09); + data |= REG_SET_FIELD(data, RLC_LB_CNTL, RESERVED, 0x80000); + WREG32_SOC15(GC, 0, mmRLC_LB_CNTL, data); + mutex_unlock(&adev->grbm_idx_mutex); + + gfx_v9_0_init_always_on_cu_mask(adev); } static void gfx_v9_0_enable_lbpw(struct amdgpu_device *adev, bool enable) @@ -943,6 +1169,7 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) dst_ptr = adev->gfx.rlc.cs_ptr; gfx_v9_0_get_csb_buffer(adev, dst_ptr); amdgpu_bo_kunmap(adev->gfx.rlc.clear_state_obj); + amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); } @@ -964,13 +1191,55 @@ static int gfx_v9_0_rlc_init(struct amdgpu_device *adev) rv_init_cp_jump_table(adev); amdgpu_bo_kunmap(adev->gfx.rlc.cp_table_obj); amdgpu_bo_unreserve(adev->gfx.rlc.cp_table_obj); + } + switch (adev->asic_type) { + case CHIP_RAVEN: gfx_v9_0_init_lbpw(adev); + break; + case CHIP_VEGA20: + gfx_v9_4_init_lbpw(adev); + break; + default: + break; } return 0; } +static int gfx_v9_0_csb_vram_pin(struct amdgpu_device *adev) +{ + int r; + + r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, false); + if (unlikely(r != 0)) + return r; + + r = amdgpu_bo_pin(adev->gfx.rlc.clear_state_obj, + AMDGPU_GEM_DOMAIN_VRAM); + if (!r) + adev->gfx.rlc.clear_state_gpu_addr = + amdgpu_bo_gpu_offset(adev->gfx.rlc.clear_state_obj); + + amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); + + return r; +} + +static void gfx_v9_0_csb_vram_unpin(struct amdgpu_device *adev) +{ + int r; + + if (!adev->gfx.rlc.clear_state_obj) + return; + + r = amdgpu_bo_reserve(adev->gfx.rlc.clear_state_obj, true); + if (likely(r == 0)) { + amdgpu_bo_unpin(adev->gfx.rlc.clear_state_obj); + amdgpu_bo_unreserve(adev->gfx.rlc.clear_state_obj); + } +} + static void gfx_v9_0_mec_fini(struct amdgpu_device *adev) { amdgpu_bo_free_kernel(&adev->gfx.mec.hpd_eop_obj, NULL, NULL); @@ -1160,7 +1429,10 @@ static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) adev->gfx.config.sc_prim_fifo_size_backend = 0x100; adev->gfx.config.sc_hiz_tile_fifo_size = 0x30; adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; - gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN; + if (adev->rev_id >= 8) + gb_addr_config = RAVEN2_GB_ADDR_CONFIG_GOLDEN; + else + gb_addr_config = RAVEN_GB_ADDR_CONFIG_GOLDEN; break; default: BUG(); @@ -1371,8 +1643,7 @@ static int gfx_v9_0_ngg_en(struct amdgpu_device *adev) gfx_v9_0_write_data_to_reg(ring, 0, false, SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_SIZE), (adev->gds.mem.total_size + - adev->gfx.ngg.gds_reserve_size) >> - AMDGPU_GDS_SHIFT); + adev->gfx.ngg.gds_reserve_size)); amdgpu_ring_write(ring, PACKET3(PACKET3_DMA_DATA, 5)); amdgpu_ring_write(ring, (PACKET3_DMA_DATA_CP_SYNC | @@ -1450,24 +1721,19 @@ static int gfx_v9_0_sw_init(void *handle) adev->gfx.mec.num_pipe_per_mec = 4; adev->gfx.mec.num_queue_per_pipe = 8; - /* KIQ event */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, 178, &adev->gfx.kiq.irq); - if (r) - return r; - /* EOP Event */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, 181, &adev->gfx.eop_irq); + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_EOP_INTERRUPT, &adev->gfx.eop_irq); if (r) return r; /* Privileged reg */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, 184, + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_REG_FAULT, &adev->gfx.priv_reg_irq); if (r) return r; /* Privileged inst */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, 185, + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_GRBM_CP, GFX_9_0__SRCID__CP_PRIV_INSTR_FAULT, &adev->gfx.priv_inst_irq); if (r) return r; @@ -1545,25 +1811,6 @@ static int gfx_v9_0_sw_init(void *handle) if (r) return r; - /* reserve GDS, GWS and OA resource for gfx */ - r = amdgpu_bo_create_kernel(adev, adev->gds.mem.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GDS, - &adev->gds.gds_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.gws.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_GWS, - &adev->gds.gws_gfx_bo, NULL, NULL); - if (r) - return r; - - r = amdgpu_bo_create_kernel(adev, adev->gds.oa.gfx_partition_size, - PAGE_SIZE, AMDGPU_GEM_DOMAIN_OA, - &adev->gds.oa_gfx_bo, NULL, NULL); - if (r) - return r; - adev->gfx.ce_ram_size = 0x8000; r = gfx_v9_0_gpu_early_init(adev); @@ -1711,7 +1958,7 @@ static void gfx_v9_0_init_compute_vmid(struct amdgpu_device *adev) mutex_unlock(&adev->srbm_mutex); } -static void gfx_v9_0_gpu_init(struct amdgpu_device *adev) +static void gfx_v9_0_constants_init(struct amdgpu_device *adev) { u32 tmp; int i; @@ -2148,8 +2395,16 @@ static void gfx_v9_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *ad static void gfx_v9_0_init_pg(struct amdgpu_device *adev) { - if (!adev->gfx.rlc.is_rlc_v2_1) - return; + gfx_v9_0_init_csb(adev); + + /* + * Rlc save restore list is workable since v2_1. + * And it's needed by gfxoff feature. + */ + if (adev->gfx.rlc.is_rlc_v2_1) { + gfx_v9_1_init_rlc_save_restore_list(adev); + gfx_v9_0_enable_save_restore_machine(adev); + } if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_GFX_SMG | @@ -2157,10 +2412,6 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_GDS | AMD_PG_SUPPORT_RLC_SMU_HS)) { - gfx_v9_0_init_csb(adev); - gfx_v9_1_init_rlc_save_restore_list(adev); - gfx_v9_0_enable_save_restore_machine(adev); - WREG32(mmRLC_JUMP_TABLE_RESTORE, adev->gfx.rlc.cp_table_gpu_addr >> 8); gfx_v9_0_init_gfx_power_gating(adev); @@ -2252,9 +2503,6 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) /* disable CG */ WREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL, 0); - /* disable PG */ - WREG32_SOC15(GC, 0, mmRLC_PG_CNTL, 0); - gfx_v9_0_rlc_reset(adev); gfx_v9_0_init_pg(adev); @@ -2266,7 +2514,8 @@ static int gfx_v9_0_rlc_resume(struct amdgpu_device *adev) return r; } - if (adev->asic_type == CHIP_RAVEN) { + if (adev->asic_type == CHIP_RAVEN || + adev->asic_type == CHIP_VEGA20) { if (amdgpu_lbpw != 0) gfx_v9_0_enable_lbpw(adev, true); else @@ -2559,7 +2808,6 @@ static void gfx_v9_0_kiq_setting(struct amdgpu_ring *ring) static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev) { struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; - uint32_t scratch, tmp = 0; uint64_t queue_mask = 0; int r, i; @@ -2578,17 +2826,9 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev) queue_mask |= (1ull << i); } - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("Failed to get scratch reg (%d).\n", r); - return r; - } - WREG32(scratch, 0xCAFEDEAD); - - r = amdgpu_ring_alloc(kiq_ring, (7 * adev->gfx.num_compute_rings) + 11); + r = amdgpu_ring_alloc(kiq_ring, (7 * adev->gfx.num_compute_rings) + 8); if (r) { DRM_ERROR("Failed to lock KIQ (%d).\n", r); - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -2625,24 +2865,12 @@ static int gfx_v9_0_kiq_kcq_enable(struct amdgpu_device *adev) amdgpu_ring_write(kiq_ring, lower_32_bits(wptr_addr)); amdgpu_ring_write(kiq_ring, upper_32_bits(wptr_addr)); } - /* write to scratch for completion */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); - amdgpu_ring_write(kiq_ring, 0xDEADBEEF); - amdgpu_ring_commit(kiq_ring); - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i >= adev->usec_timeout) { - DRM_ERROR("KCQ enable failed (scratch(0x%04X)=0x%08X)\n", - scratch, tmp); - r = -EINVAL; + r = amdgpu_ring_test_ring(kiq_ring); + if (r) { + DRM_ERROR("KCQ enable failed\n"); + kiq_ring->ready = false; } - amdgpu_gfx_scratch_free(adev, scratch); return r; } @@ -2975,7 +3203,7 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) struct v9_mqd *mqd = ring->mqd_ptr; int mqd_idx = ring - &adev->gfx.compute_ring[0]; - if (!adev->in_gpu_reset && !adev->gfx.in_suspend) { + if (!adev->in_gpu_reset && !adev->in_suspend) { memset((void *)mqd, 0, sizeof(struct v9_mqd_allocation)); ((struct v9_mqd_allocation *)mqd)->dynamic_cu_mask = 0xFFFFFFFF; ((struct v9_mqd_allocation *)mqd)->dynamic_rb_mask = 0xFFFFFFFF; @@ -3004,26 +3232,33 @@ static int gfx_v9_0_kcq_init_queue(struct amdgpu_ring *ring) static int gfx_v9_0_kiq_resume(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = NULL; - int r = 0, i; - - gfx_v9_0_cp_compute_enable(adev, true); + struct amdgpu_ring *ring; + int r; ring = &adev->gfx.kiq.ring; r = amdgpu_bo_reserve(ring->mqd_obj, false); if (unlikely(r != 0)) - goto done; + return r; r = amdgpu_bo_kmap(ring->mqd_obj, (void **)&ring->mqd_ptr); - if (!r) { - r = gfx_v9_0_kiq_init_queue(ring); - amdgpu_bo_kunmap(ring->mqd_obj); - ring->mqd_ptr = NULL; - } + if (unlikely(r != 0)) + return r; + + gfx_v9_0_kiq_init_queue(ring); + amdgpu_bo_kunmap(ring->mqd_obj); + ring->mqd_ptr = NULL; amdgpu_bo_unreserve(ring->mqd_obj); - if (r) - goto done; + ring->ready = true; + return 0; +} + +static int gfx_v9_0_kcq_resume(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = NULL; + int r = 0, i; + + gfx_v9_0_cp_compute_enable(adev, true); for (i = 0; i < adev->gfx.num_compute_rings; i++) { ring = &adev->gfx.compute_ring[i]; @@ -3066,11 +3301,15 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) return r; } + r = gfx_v9_0_kiq_resume(adev); + if (r) + return r; + r = gfx_v9_0_cp_gfx_resume(adev); if (r) return r; - r = gfx_v9_0_kiq_resume(adev); + r = gfx_v9_0_kcq_resume(adev); if (r) return r; @@ -3081,12 +3320,6 @@ static int gfx_v9_0_cp_resume(struct amdgpu_device *adev) return r; } - ring = &adev->gfx.kiq.ring; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) - ring->ready = false; - for (i = 0; i < adev->gfx.num_compute_rings; i++) { ring = &adev->gfx.compute_ring[i]; @@ -3114,7 +3347,11 @@ static int gfx_v9_0_hw_init(void *handle) gfx_v9_0_init_golden_registers(adev); - gfx_v9_0_gpu_init(adev); + gfx_v9_0_constants_init(adev); + + r = gfx_v9_0_csb_vram_pin(adev); + if (r) + return r; r = gfx_v9_0_rlc_resume(adev); if (r) @@ -3131,71 +3368,45 @@ static int gfx_v9_0_hw_init(void *handle) return r; } -static int gfx_v9_0_kcq_disable(struct amdgpu_ring *kiq_ring,struct amdgpu_ring *ring) +static int gfx_v9_0_kcq_disable(struct amdgpu_device *adev) { - struct amdgpu_device *adev = kiq_ring->adev; - uint32_t scratch, tmp = 0; int r, i; + struct amdgpu_ring *kiq_ring = &adev->gfx.kiq.ring; - r = amdgpu_gfx_scratch_get(adev, &scratch); - if (r) { - DRM_ERROR("Failed to get scratch reg (%d).\n", r); - return r; - } - WREG32(scratch, 0xCAFEDEAD); - - r = amdgpu_ring_alloc(kiq_ring, 10); - if (r) { + r = amdgpu_ring_alloc(kiq_ring, 6 * adev->gfx.num_compute_rings); + if (r) DRM_ERROR("Failed to lock KIQ (%d).\n", r); - amdgpu_gfx_scratch_free(adev, scratch); - return r; - } - /* unmap queues */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); - amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + struct amdgpu_ring *ring = &adev->gfx.compute_ring[i]; + + amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_UNMAP_QUEUES, 4)); + amdgpu_ring_write(kiq_ring, /* Q_sel: 0, vmid: 0, engine: 0, num_Q: 1 */ PACKET3_UNMAP_QUEUES_ACTION(1) | /* RESET_QUEUES */ PACKET3_UNMAP_QUEUES_QUEUE_SEL(0) | PACKET3_UNMAP_QUEUES_ENGINE_SEL(0) | PACKET3_UNMAP_QUEUES_NUM_QUEUES(1)); - amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); - amdgpu_ring_write(kiq_ring, 0); - amdgpu_ring_write(kiq_ring, 0); - amdgpu_ring_write(kiq_ring, 0); - /* write to scratch for completion */ - amdgpu_ring_write(kiq_ring, PACKET3(PACKET3_SET_UCONFIG_REG, 1)); - amdgpu_ring_write(kiq_ring, (scratch - PACKET3_SET_UCONFIG_REG_START)); - amdgpu_ring_write(kiq_ring, 0xDEADBEEF); - amdgpu_ring_commit(kiq_ring); - - for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32(scratch); - if (tmp == 0xDEADBEEF) - break; - DRM_UDELAY(1); - } - if (i >= adev->usec_timeout) { - DRM_ERROR("KCQ disabled failed (scratch(0x%04X)=0x%08X)\n", scratch, tmp); - r = -EINVAL; + amdgpu_ring_write(kiq_ring, PACKET3_UNMAP_QUEUES_DOORBELL_OFFSET0(ring->doorbell_index)); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); + amdgpu_ring_write(kiq_ring, 0); } - amdgpu_gfx_scratch_free(adev, scratch); + r = amdgpu_ring_test_ring(kiq_ring); + if (r) + DRM_ERROR("KCQ disable failed\n"); + return r; } static int gfx_v9_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int i; - - amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_GFX, - AMD_PG_STATE_UNGATE); amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); /* disable KCQ to avoid CPC touch memory not valid anymore */ - for (i = 0; i < adev->gfx.num_compute_rings; i++) - gfx_v9_0_kcq_disable(&adev->gfx.kiq.ring, &adev->gfx.compute_ring[i]); + gfx_v9_0_kcq_disable(adev); if (amdgpu_sriov_vf(adev)) { gfx_v9_0_cp_gfx_enable(adev, false); @@ -3211,7 +3422,7 @@ static int gfx_v9_0_hw_fini(void *handle) /* Use deinitialize sequence from CAIL when unbinding device from driver, * otherwise KIQ is hanging when binding back */ - if (!adev->in_gpu_reset && !adev->gfx.in_suspend) { + if (!adev->in_gpu_reset && !adev->in_suspend) { mutex_lock(&adev->srbm_mutex); soc15_grbm_select(adev, adev->gfx.kiq.ring.me, adev->gfx.kiq.ring.pipe, @@ -3224,25 +3435,19 @@ static int gfx_v9_0_hw_fini(void *handle) gfx_v9_0_cp_enable(adev, false); gfx_v9_0_rlc_stop(adev); + gfx_v9_0_csb_vram_unpin(adev); + return 0; } static int gfx_v9_0_suspend(void *handle) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - - adev->gfx.in_suspend = true; - return gfx_v9_0_hw_fini(adev); + return gfx_v9_0_hw_fini(handle); } static int gfx_v9_0_resume(void *handle) { - struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int r; - - r = gfx_v9_0_hw_init(adev); - adev->gfx.in_suspend = false; - return r; + return gfx_v9_0_hw_init(handle); } static bool gfx_v9_0_is_idle(void *handle) @@ -3351,15 +3556,6 @@ static void gfx_v9_0_ring_emit_gds_switch(struct amdgpu_ring *ring, { struct amdgpu_device *adev = ring->adev; - gds_base = gds_base >> AMDGPU_GDS_SHIFT; - gds_size = gds_size >> AMDGPU_GDS_SHIFT; - - gws_base = gws_base >> AMDGPU_GWS_SHIFT; - gws_size = gws_size >> AMDGPU_GWS_SHIFT; - - oa_base = oa_base >> AMDGPU_OA_SHIFT; - oa_size = oa_size >> AMDGPU_OA_SHIFT; - /* GDS Base */ gfx_v9_0_write_data_to_reg(ring, 0, false, SOC15_REG_OFFSET(GC, 0, mmGDS_VMID0_BASE) + 2 * vmid, @@ -3433,7 +3629,7 @@ static void gfx_v9_0_enter_rlc_safe_mode(struct amdgpu_device *adev) /* wait for RLC_SAFE_MODE */ for (i = 0; i < adev->usec_timeout; i++) { - if (!REG_GET_FIELD(SOC15_REG_OFFSET(GC, 0, mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) + if (!REG_GET_FIELD(RREG32_SOC15(GC, 0, mmRLC_SAFE_MODE), RLC_SAFE_MODE, CMD)) break; udelay(1); } @@ -3510,8 +3706,11 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev if (enable && (adev->cg_flags & AMD_CG_SUPPORT_GFX_MGCG)) { /* 1 - RLC_CGTT_MGCG_OVERRIDE */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); - data &= ~(RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK | - RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK | + + if (adev->asic_type != CHIP_VEGA12) + data &= ~RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK; + + data &= ~(RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK | RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK | RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK); @@ -3541,11 +3740,15 @@ static void gfx_v9_0_update_medium_grain_clock_gating(struct amdgpu_device *adev } else { /* 1 - MGCG_OVERRIDE */ def = data = RREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE); - data |= (RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK | - RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK | + + if (adev->asic_type != CHIP_VEGA12) + data |= RLC_CGTT_MGCG_OVERRIDE__CPF_CGTT_SCLK_OVERRIDE_MASK; + + data |= (RLC_CGTT_MGCG_OVERRIDE__RLC_CGTT_SCLK_OVERRIDE_MASK | RLC_CGTT_MGCG_OVERRIDE__GRBM_CGTT_SCLK_OVERRIDE_MASK | RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGCG_OVERRIDE_MASK | RLC_CGTT_MGCG_OVERRIDE__GFXIP_MGLS_OVERRIDE_MASK); + if (def != data) WREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE, data); @@ -3581,9 +3784,11 @@ static void gfx_v9_0_update_3d_clock_gating(struct amdgpu_device *adev, /* update CGCG and CGLS override bits */ if (def != data) WREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE, data); - /* enable 3Dcgcg FSM(0x0020003f) */ + + /* enable 3Dcgcg FSM(0x0000363f) */ def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL_3D); - data = (0x2000 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | + + data = (0x36 << RLC_CGCG_CGLS_CTRL_3D__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | RLC_CGCG_CGLS_CTRL_3D__CGCG_EN_MASK; if (adev->cg_flags & AMD_CG_SUPPORT_GFX_3D_CGLS) data |= (0x000F << RLC_CGCG_CGLS_CTRL_3D__CGLS_REP_COMPANSAT_DELAY__SHIFT) | @@ -3630,9 +3835,10 @@ static void gfx_v9_0_update_coarse_grain_clock_gating(struct amdgpu_device *adev if (def != data) WREG32_SOC15(GC, 0, mmRLC_CGTT_MGCG_OVERRIDE, data); - /* enable cgcg FSM(0x0020003F) */ + /* enable cgcg FSM(0x0000363F) */ def = RREG32_SOC15(GC, 0, mmRLC_CGCG_CGLS_CTRL); - data = (0x2000 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | + + data = (0x36 << RLC_CGCG_CGLS_CTRL__CGCG_GFX_IDLE_THRESHOLD__SHIFT) | RLC_CGCG_CGLS_CTRL__CGCG_EN_MASK; if (adev->cg_flags & AMD_CG_SUPPORT_GFX_CGLS) data |= (0x000F << RLC_CGCG_CGLS_CTRL__CGLS_REP_COMPANSAT_DELAY__SHIFT) | @@ -3696,6 +3902,10 @@ static int gfx_v9_0_set_powergating_state(void *handle, switch (adev->asic_type) { case CHIP_RAVEN: + if (!enable) { + amdgpu_gfx_off_ctrl(adev, false); + cancel_delayed_work_sync(&adev->gfx.gfx_off_delay_work); + } if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) { gfx_v9_0_enable_sck_slow_down_on_power_up(adev, true); gfx_v9_0_enable_sck_slow_down_on_power_down(adev, true); @@ -3714,6 +3924,17 @@ static int gfx_v9_0_set_powergating_state(void *handle, /* update mgcg state */ gfx_v9_0_update_gfx_mg_power_gating(adev, enable); + + if (enable) + amdgpu_gfx_off_ctrl(adev, true); + break; + case CHIP_VEGA12: + if (!enable) { + amdgpu_gfx_off_ctrl(adev, false); + cancel_delayed_work_sync(&adev->gfx.gfx_off_delay_work); + } else { + amdgpu_gfx_off_ctrl(adev, true); + } break; default: break; @@ -4274,8 +4495,11 @@ static void gfx_v9_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring, uint32_t ref, uint32_t mask) { int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); + struct amdgpu_device *adev = ring->adev; + bool fw_version_ok = (ring->funcs->type == AMDGPU_RING_TYPE_GFX) ? + adev->gfx.me_fw_write_wait : adev->gfx.mec_fw_write_wait; - if (amdgpu_sriov_vf(ring->adev)) + if (fw_version_ok) gfx_v9_0_wait_reg_mem(ring, usepfp, 0, 1, reg0, reg1, ref, mask, 0x20); else @@ -4283,6 +4507,18 @@ static void gfx_v9_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring, ref, mask); } +static void gfx_v9_0_ring_soft_recovery(struct amdgpu_ring *ring, unsigned vmid) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t value = 0; + + value = REG_SET_FIELD(value, SQ_CMD, CMD, 0x03); + value = REG_SET_FIELD(value, SQ_CMD, MODE, 0x01); + value = REG_SET_FIELD(value, SQ_CMD, CHECK_VMID, 1); + value = REG_SET_FIELD(value, SQ_CMD, VM_ID, vmid); + WREG32(mmSQ_CMD, value); +} + static void gfx_v9_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { @@ -4477,68 +4713,6 @@ static int gfx_v9_0_priv_inst_irq(struct amdgpu_device *adev, return 0; } -static int gfx_v9_0_kiq_set_interrupt_state(struct amdgpu_device *adev, - struct amdgpu_irq_src *src, - unsigned int type, - enum amdgpu_interrupt_state state) -{ - uint32_t tmp, target; - struct amdgpu_ring *ring = &(adev->gfx.kiq.ring); - - if (ring->me == 1) - target = SOC15_REG_OFFSET(GC, 0, mmCP_ME1_PIPE0_INT_CNTL); - else - target = SOC15_REG_OFFSET(GC, 0, mmCP_ME2_PIPE0_INT_CNTL); - target += ring->pipe; - - switch (type) { - case AMDGPU_CP_KIQ_IRQ_DRIVER0: - if (state == AMDGPU_IRQ_STATE_DISABLE) { - tmp = RREG32_SOC15(GC, 0, mmCPC_INT_CNTL); - tmp = REG_SET_FIELD(tmp, CPC_INT_CNTL, - GENERIC2_INT_ENABLE, 0); - WREG32_SOC15(GC, 0, mmCPC_INT_CNTL, tmp); - - tmp = RREG32(target); - tmp = REG_SET_FIELD(tmp, CP_ME2_PIPE0_INT_CNTL, - GENERIC2_INT_ENABLE, 0); - WREG32(target, tmp); - } else { - tmp = RREG32_SOC15(GC, 0, mmCPC_INT_CNTL); - tmp = REG_SET_FIELD(tmp, CPC_INT_CNTL, - GENERIC2_INT_ENABLE, 1); - WREG32_SOC15(GC, 0, mmCPC_INT_CNTL, tmp); - - tmp = RREG32(target); - tmp = REG_SET_FIELD(tmp, CP_ME2_PIPE0_INT_CNTL, - GENERIC2_INT_ENABLE, 1); - WREG32(target, tmp); - } - break; - default: - BUG(); /* kiq only support GENERIC2_INT now */ - break; - } - return 0; -} - -static int gfx_v9_0_kiq_irq(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - struct amdgpu_iv_entry *entry) -{ - u8 me_id, pipe_id, queue_id; - struct amdgpu_ring *ring = &(adev->gfx.kiq.ring); - - me_id = (entry->ring_id & 0x0c) >> 2; - pipe_id = (entry->ring_id & 0x03) >> 0; - queue_id = (entry->ring_id & 0x70) >> 4; - DRM_DEBUG("IH: CPC GENERIC2_INT, me:%d, pipe:%d, queue:%d\n", - me_id, pipe_id, queue_id); - - amdgpu_fence_process(ring); - return 0; -} - static const struct amd_ip_funcs gfx_v9_0_ip_funcs = { .name = "gfx_v9_0", .early_init = gfx_v9_0_early_init, @@ -4605,6 +4779,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, + .soft_recovery = gfx_v9_0_ring_soft_recovery, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { @@ -4686,11 +4861,6 @@ static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev) adev->gfx.compute_ring[i].funcs = &gfx_v9_0_ring_funcs_compute; } -static const struct amdgpu_irq_src_funcs gfx_v9_0_kiq_irq_funcs = { - .set = gfx_v9_0_kiq_set_interrupt_state, - .process = gfx_v9_0_kiq_irq, -}; - static const struct amdgpu_irq_src_funcs gfx_v9_0_eop_irq_funcs = { .set = gfx_v9_0_set_eop_interrupt_state, .process = gfx_v9_0_eop_irq, @@ -4716,9 +4886,6 @@ static void gfx_v9_0_set_irq_funcs(struct amdgpu_device *adev) adev->gfx.priv_inst_irq.num_types = 1; adev->gfx.priv_inst_irq.funcs = &gfx_v9_0_priv_inst_irq_funcs; - - adev->gfx.kiq.irq.num_types = AMDGPU_CP_KIQ_IRQ_LAST; - adev->gfx.kiq.irq.funcs = &gfx_v9_0_kiq_irq_funcs; } static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev) @@ -4738,7 +4905,20 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev) static void gfx_v9_0_set_gds_init(struct amdgpu_device *adev) { /* init asci gds info */ - adev->gds.mem.total_size = RREG32_SOC15(GC, 0, mmGDS_VMID0_SIZE); + switch (adev->asic_type) { + case CHIP_VEGA10: + case CHIP_VEGA12: + case CHIP_VEGA20: + adev->gds.mem.total_size = 0x10000; + break; + case CHIP_RAVEN: + adev->gds.mem.total_size = 0x1000; + break; + default: + adev->gds.mem.total_size = 0x10000; + break; + } + adev->gds.gws.total_size = 64; adev->gds.oa.total_size = 16; diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c index acfbd2d749cf..ceb7847b504f 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_0.c @@ -37,13 +37,7 @@ u64 gfxhub_v1_0_get_mc_fb_offset(struct amdgpu_device *adev) static void gfxhub_v1_0_init_gart_pt_regs(struct amdgpu_device *adev) { - uint64_t value; - - BUG_ON(adev->gart.table_addr & (~0x0000FFFFFFFFF000ULL)); - value = adev->gart.table_addr - adev->gmc.vram_start - + adev->vm_manager.vram_base_offset; - value &= 0x0000FFFFFFFFF000ULL; - value |= 0x1; /*valid bit*/ + uint64_t value = amdgpu_gmc_pd_addr(adev->gart.bo); WREG32_SOC15(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, lower_32_bits(value)); @@ -71,16 +65,28 @@ static void gfxhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) { uint64_t value; - /* Disable AGP. */ + /* Program the AGP BAR */ WREG32_SOC15(GC, 0, mmMC_VM_AGP_BASE, 0); - WREG32_SOC15(GC, 0, mmMC_VM_AGP_TOP, 0); - WREG32_SOC15(GC, 0, mmMC_VM_AGP_BOT, 0xFFFFFFFF); + WREG32_SOC15(GC, 0, mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 24); + WREG32_SOC15(GC, 0, mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 24); /* Program the system aperture low logical page number. */ WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, - adev->gmc.vram_start >> 18); - WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, - adev->gmc.vram_end >> 18); + min(adev->gmc.vram_start, adev->gmc.agp_start) >> 18); + + if (adev->asic_type == CHIP_RAVEN && adev->rev_id >= 0x8) + /* + * Raven2 has a HW issue that it is unable to use the vram which + * is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR. So here is the + * workaround that increase system aperture high address (add 1) + * to get rid of the VM fault and hardware hang. + */ + WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + max((adev->gmc.vram_end >> 18) + 0x1, + adev->gmc.agp_end >> 18)); + else + WREG32_SOC15(GC, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + max(adev->gmc.vram_end, adev->gmc.agp_end) >> 18); /* Set default page address. */ value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_1.c b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_1.c new file mode 100644 index 000000000000..5e9ab8eb214a --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_1.c @@ -0,0 +1,53 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "gfxhub_v1_1.h" + +#include "gc/gc_9_2_1_offset.h" +#include "gc/gc_9_2_1_sh_mask.h" + +#include "soc15_common.h" + +int gfxhub_v1_1_get_xgmi_info(struct amdgpu_device *adev) +{ + u32 xgmi_lfb_cntl = RREG32_SOC15(GC, 0, mmMC_VM_XGMI_LFB_CNTL); + u32 max_region = + REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL, PF_MAX_REGION); + + /* PF_MAX_REGION=0 means xgmi is disabled */ + if (max_region) { + adev->gmc.xgmi.num_physical_nodes = max_region + 1; + if (adev->gmc.xgmi.num_physical_nodes > 4) + return -EINVAL; + + adev->gmc.xgmi.physical_node_id = + REG_GET_FIELD(xgmi_lfb_cntl, MC_VM_XGMI_LFB_CNTL, PF_LFB_REGION); + if (adev->gmc.xgmi.physical_node_id > 3) + return -EINVAL; + adev->gmc.xgmi.node_segment_size = REG_GET_FIELD( + RREG32_SOC15(GC, 0, mmMC_VM_XGMI_LFB_SIZE), + MC_VM_XGMI_LFB_SIZE, PF_LFB_SIZE) << 24; + } + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_1.h b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_1.h new file mode 100644 index 000000000000..d753cf28a0a6 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/gfxhub_v1_1.h @@ -0,0 +1,29 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __GFXHUB_V1_1_H__ +#define __GFXHUB_V1_1_H__ + +int gfxhub_v1_1_get_xgmi_info(struct amdgpu_device *adev); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 79f9ac29019b..e1c2b4e9c7b2 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -26,6 +26,7 @@ #include "amdgpu.h" #include "gmc_v6_0.h" #include "amdgpu_ucode.h" +#include "amdgpu_gem.h" #include "bif/bif_3_0_d.h" #include "bif/bif_3_0_sh_mask.h" @@ -41,11 +42,11 @@ static void gmc_v6_0_set_gmc_funcs(struct amdgpu_device *adev); static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev); static int gmc_v6_0_wait_for_idle(void *handle); -MODULE_FIRMWARE("radeon/tahiti_mc.bin"); -MODULE_FIRMWARE("radeon/pitcairn_mc.bin"); -MODULE_FIRMWARE("radeon/verde_mc.bin"); -MODULE_FIRMWARE("radeon/oland_mc.bin"); -MODULE_FIRMWARE("radeon/si58_mc.bin"); +MODULE_FIRMWARE("amdgpu/tahiti_mc.bin"); +MODULE_FIRMWARE("amdgpu/pitcairn_mc.bin"); +MODULE_FIRMWARE("amdgpu/verde_mc.bin"); +MODULE_FIRMWARE("amdgpu/oland_mc.bin"); +MODULE_FIRMWARE("amdgpu/si58_mc.bin"); #define MC_SEQ_MISC0__MT__MASK 0xf0000000 #define MC_SEQ_MISC0__MT__GDDR1 0x10000000 @@ -134,9 +135,9 @@ static int gmc_v6_0_init_microcode(struct amdgpu_device *adev) is_58_fw = true; if (is_58_fw) - snprintf(fw_name, sizeof(fw_name), "radeon/si58_mc.bin"); + snprintf(fw_name, sizeof(fw_name), "amdgpu/si58_mc.bin"); else - snprintf(fw_name, sizeof(fw_name), "radeon/%s_mc.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name); err = request_firmware(&adev->gmc.fw, fw_name, adev->dev); if (err) goto out; @@ -223,8 +224,8 @@ static void gmc_v6_0_vram_gtt_location(struct amdgpu_device *adev, u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF; base <<= 24; - amdgpu_device_vram_location(adev, &adev->gmc, base); - amdgpu_device_gart_location(adev, mc); + amdgpu_gmc_vram_location(adev, &adev->gmc, base); + amdgpu_gmc_gart_location(adev, mc); } static void gmc_v6_0_mc_program(struct amdgpu_device *adev) @@ -493,16 +494,20 @@ static void gmc_v6_0_set_prt(struct amdgpu_device *adev, bool enable) static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) { + uint64_t table_addr; int r, i; u32 field; - if (adev->gart.robj == NULL) { + if (adev->gart.bo == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); return -EINVAL; } r = amdgpu_gart_table_vram_pin(adev); if (r) return r; + + table_addr = amdgpu_bo_gpu_offset(adev->gart.bo); + /* Setup TLB control */ WREG32(mmMC_VM_MX_L1_TLB_CNTL, (0xA << 7) | @@ -531,7 +536,7 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) /* setup context0 */ WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->gmc.gart_start >> 12); WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->gmc.gart_end >> 12); - WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, table_addr >> 12); WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, (u32)(adev->dummy_page_addr >> 12)); WREG32(mmVM_CONTEXT0_CNTL2, 0); @@ -555,10 +560,10 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) for (i = 1; i < 16; i++) { if (i < 8) WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + i, - adev->gart.table_addr >> 12); + table_addr >> 12); else WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + i - 8, - adev->gart.table_addr >> 12); + table_addr >> 12); } /* enable context1-15 */ @@ -578,7 +583,7 @@ static int gmc_v6_0_gart_enable(struct amdgpu_device *adev) gmc_v6_0_flush_gpu_tlb(adev, 0); dev_info(adev->dev, "PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), - (unsigned long long)adev->gart.table_addr); + (unsigned long long)table_addr); adev->gart.ready = true; return 0; } @@ -587,7 +592,7 @@ static int gmc_v6_0_gart_init(struct amdgpu_device *adev) { int r; - if (adev->gart.robj) { + if (adev->gart.bo) { dev_warn(adev->dev, "gmc_v6_0 PCIE GART already initialized\n"); return 0; } @@ -632,12 +637,6 @@ static void gmc_v6_0_gart_disable(struct amdgpu_device *adev) amdgpu_gart_table_vram_unpin(adev); } -static void gmc_v6_0_gart_fini(struct amdgpu_device *adev) -{ - amdgpu_gart_table_vram_free(adev); - amdgpu_gart_fini(adev); -} - static void gmc_v6_0_vm_decode_fault(struct amdgpu_device *adev, u32 status, u32 addr, u32 mc_client) { @@ -860,11 +859,11 @@ static int gmc_v6_0_sw_init(void *handle) adev->gmc.vram_type = gmc_v6_0_convert_vram_type(tmp); } - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 146, &adev->gmc.vm_fault); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 146, &adev->gmc.vm_fault); if (r) return r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 147, &adev->gmc.vm_fault); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 147, &adev->gmc.vm_fault); if (r) return r; @@ -935,8 +934,9 @@ static int gmc_v6_0_sw_fini(void *handle) amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); - gmc_v6_0_gart_fini(adev); + amdgpu_gart_table_vram_free(adev); amdgpu_bo_fini(adev); + amdgpu_gart_fini(adev); release_firmware(adev->gmc.fw); adev->gmc.fw = NULL; @@ -1180,8 +1180,7 @@ static const struct amdgpu_irq_src_funcs gmc_v6_0_irq_funcs = { static void gmc_v6_0_set_gmc_funcs(struct amdgpu_device *adev) { - if (adev->gmc.gmc_funcs == NULL) - adev->gmc.gmc_funcs = &gmc_v6_0_gmc_funcs; + adev->gmc.gmc_funcs = &gmc_v6_0_gmc_funcs; } static void gmc_v6_0_set_irq_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 7147bfe25a23..910c4ce19cb3 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -28,6 +28,8 @@ #include "cik.h" #include "gmc_v7_0.h" #include "amdgpu_ucode.h" +#include "amdgpu_amdkfd.h" +#include "amdgpu_gem.h" #include "bif/bif_4_1_d.h" #include "bif/bif_4_1_sh_mask.h" @@ -43,12 +45,14 @@ #include "amdgpu_atombios.h" +#include "ivsrcid/ivsrcid_vislands30.h" + static void gmc_v7_0_set_gmc_funcs(struct amdgpu_device *adev); static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev); static int gmc_v7_0_wait_for_idle(void *handle); -MODULE_FIRMWARE("radeon/bonaire_mc.bin"); -MODULE_FIRMWARE("radeon/hawaii_mc.bin"); +MODULE_FIRMWARE("amdgpu/bonaire_mc.bin"); +MODULE_FIRMWARE("amdgpu/hawaii_mc.bin"); MODULE_FIRMWARE("amdgpu/topaz_mc.bin"); static const u32 golden_settings_iceland_a11[] = @@ -147,10 +151,7 @@ static int gmc_v7_0_init_microcode(struct amdgpu_device *adev) default: BUG(); } - if (adev->asic_type == CHIP_TOPAZ) - snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name); - else - snprintf(fw_name, sizeof(fw_name), "radeon/%s_mc.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mc.bin", chip_name); err = request_firmware(&adev->gmc.fw, fw_name, adev->dev); if (err) @@ -241,8 +242,8 @@ static void gmc_v7_0_vram_gtt_location(struct amdgpu_device *adev, u64 base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF; base <<= 24; - amdgpu_device_vram_location(adev, &adev->gmc, base); - amdgpu_device_gart_location(adev, mc); + amdgpu_gmc_vram_location(adev, &adev->gmc, base); + amdgpu_gmc_gart_location(adev, mc); } /** @@ -601,16 +602,20 @@ static void gmc_v7_0_set_prt(struct amdgpu_device *adev, bool enable) */ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) { + uint64_t table_addr; int r, i; u32 tmp, field; - if (adev->gart.robj == NULL) { + if (adev->gart.bo == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); return -EINVAL; } r = amdgpu_gart_table_vram_pin(adev); if (r) return r; + + table_addr = amdgpu_bo_gpu_offset(adev->gart.bo); + /* Setup TLB control */ tmp = RREG32(mmMC_VM_MX_L1_TLB_CNTL); tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); @@ -642,7 +647,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) /* setup context0 */ WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->gmc.gart_start >> 12); WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->gmc.gart_end >> 12); - WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, table_addr >> 12); WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, (u32)(adev->dummy_page_addr >> 12)); WREG32(mmVM_CONTEXT0_CNTL2, 0); @@ -666,10 +671,10 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) for (i = 1; i < 16; i++) { if (i < 8) WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + i, - adev->gart.table_addr >> 12); + table_addr >> 12); else WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + i - 8, - adev->gart.table_addr >> 12); + table_addr >> 12); } /* enable context1-15 */ @@ -696,7 +701,7 @@ static int gmc_v7_0_gart_enable(struct amdgpu_device *adev) gmc_v7_0_flush_gpu_tlb(adev, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), - (unsigned long long)adev->gart.table_addr); + (unsigned long long)table_addr); adev->gart.ready = true; return 0; } @@ -705,7 +710,7 @@ static int gmc_v7_0_gart_init(struct amdgpu_device *adev) { int r; - if (adev->gart.robj) { + if (adev->gart.bo) { WARN(1, "R600 PCIE GART already initialized\n"); return 0; } @@ -747,19 +752,6 @@ static void gmc_v7_0_gart_disable(struct amdgpu_device *adev) } /** - * gmc_v7_0_gart_fini - vm fini callback - * - * @adev: amdgpu_device pointer - * - * Tears down the driver GART/VM setup (CIK). - */ -static void gmc_v7_0_gart_fini(struct amdgpu_device *adev) -{ - amdgpu_gart_table_vram_free(adev); - amdgpu_gart_fini(adev); -} - -/** * gmc_v7_0_vm_decode_fault - print human readable fault info * * @adev: amdgpu_device pointer @@ -999,11 +991,11 @@ static int gmc_v7_0_sw_init(void *handle) adev->gmc.vram_type = gmc_v7_0_convert_vram_type(tmp); } - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 146, &adev->gmc.vm_fault); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_PAGE_INV_FAULT, &adev->gmc.vm_fault); if (r) return r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 147, &adev->gmc.vm_fault); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_MEM_PROT_FAULT, &adev->gmc.vm_fault); if (r) return r; @@ -1079,6 +1071,12 @@ static int gmc_v7_0_sw_init(void *handle) adev->vm_manager.vram_base_offset = 0; } + adev->gmc.vm_fault_info = kmalloc(sizeof(struct kfd_vm_fault_info), + GFP_KERNEL); + if (!adev->gmc.vm_fault_info) + return -ENOMEM; + atomic_set(&adev->gmc.vm_fault_info_updated, 0); + return 0; } @@ -1088,8 +1086,10 @@ static int gmc_v7_0_sw_fini(void *handle) amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); - gmc_v7_0_gart_fini(adev); + kfree(adev->gmc.vm_fault_info); + amdgpu_gart_table_vram_free(adev); amdgpu_bo_fini(adev); + amdgpu_gart_fini(adev); release_firmware(adev->gmc.fw); adev->gmc.fw = NULL; @@ -1277,7 +1277,7 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - u32 addr, status, mc_client; + u32 addr, status, mc_client, vmid; addr = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_ADDR); status = RREG32(mmVM_CONTEXT1_PROTECTION_FAULT_STATUS); @@ -1302,6 +1302,29 @@ static int gmc_v7_0_process_interrupt(struct amdgpu_device *adev, entry->pasid); } + vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, + VMID); + if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid) + && !atomic_read(&adev->gmc.vm_fault_info_updated)) { + struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info; + u32 protections = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + PROTECTIONS); + + info->vmid = vmid; + info->mc_id = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + MEMORY_CLIENT_ID); + info->status = status; + info->page_addr = addr; + info->prot_valid = protections & 0x7 ? true : false; + info->prot_read = protections & 0x8 ? true : false; + info->prot_write = protections & 0x10 ? true : false; + info->prot_exec = protections & 0x20 ? true : false; + mb(); + atomic_set(&adev->gmc.vm_fault_info_updated, 1); + } + return 0; } @@ -1365,8 +1388,7 @@ static const struct amdgpu_irq_src_funcs gmc_v7_0_irq_funcs = { static void gmc_v7_0_set_gmc_funcs(struct amdgpu_device *adev) { - if (adev->gmc.gmc_funcs == NULL) - adev->gmc.gmc_funcs = &gmc_v7_0_gmc_funcs; + adev->gmc.gmc_funcs = &gmc_v7_0_gmc_funcs; } static void gmc_v7_0_set_irq_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index 1edbe6b477b5..1d3265c97b70 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -26,6 +26,8 @@ #include "amdgpu.h" #include "gmc_v8_0.h" #include "amdgpu_ucode.h" +#include "amdgpu_amdkfd.h" +#include "amdgpu_gem.h" #include "gmc/gmc_8_1_d.h" #include "gmc/gmc_8_1_sh_mask.h" @@ -44,6 +46,7 @@ #include "amdgpu_atombios.h" +#include "ivsrcid/ivsrcid_vislands30.h" static void gmc_v8_0_set_gmc_funcs(struct amdgpu_device *adev); static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev); @@ -408,8 +411,8 @@ static void gmc_v8_0_vram_gtt_location(struct amdgpu_device *adev, base = RREG32(mmMC_VM_FB_LOCATION) & 0xFFFF; base <<= 24; - amdgpu_device_vram_location(adev, &adev->gmc, base); - amdgpu_device_gart_location(adev, mc); + amdgpu_gmc_vram_location(adev, &adev->gmc, base); + amdgpu_gmc_gart_location(adev, mc); } /** @@ -804,16 +807,20 @@ static void gmc_v8_0_set_prt(struct amdgpu_device *adev, bool enable) */ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) { + uint64_t table_addr; int r, i; u32 tmp, field; - if (adev->gart.robj == NULL) { + if (adev->gart.bo == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); return -EINVAL; } r = amdgpu_gart_table_vram_pin(adev); if (r) return r; + + table_addr = amdgpu_bo_gpu_offset(adev->gart.bo); + /* Setup TLB control */ tmp = RREG32(mmMC_VM_MX_L1_TLB_CNTL); tmp = REG_SET_FIELD(tmp, MC_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, 1); @@ -861,7 +868,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) /* setup context0 */ WREG32(mmVM_CONTEXT0_PAGE_TABLE_START_ADDR, adev->gmc.gart_start >> 12); WREG32(mmVM_CONTEXT0_PAGE_TABLE_END_ADDR, adev->gmc.gart_end >> 12); - WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, adev->gart.table_addr >> 12); + WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR, table_addr >> 12); WREG32(mmVM_CONTEXT0_PROTECTION_FAULT_DEFAULT_ADDR, (u32)(adev->dummy_page_addr >> 12)); WREG32(mmVM_CONTEXT0_CNTL2, 0); @@ -885,10 +892,10 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) for (i = 1; i < 16; i++) { if (i < 8) WREG32(mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + i, - adev->gart.table_addr >> 12); + table_addr >> 12); else WREG32(mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + i - 8, - adev->gart.table_addr >> 12); + table_addr >> 12); } /* enable context1-15 */ @@ -916,7 +923,7 @@ static int gmc_v8_0_gart_enable(struct amdgpu_device *adev) gmc_v8_0_flush_gpu_tlb(adev, 0); DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), - (unsigned long long)adev->gart.table_addr); + (unsigned long long)table_addr); adev->gart.ready = true; return 0; } @@ -925,7 +932,7 @@ static int gmc_v8_0_gart_init(struct amdgpu_device *adev) { int r; - if (adev->gart.robj) { + if (adev->gart.bo) { WARN(1, "R600 PCIE GART already initialized\n"); return 0; } @@ -967,19 +974,6 @@ static void gmc_v8_0_gart_disable(struct amdgpu_device *adev) } /** - * gmc_v8_0_gart_fini - vm fini callback - * - * @adev: amdgpu_device pointer - * - * Tears down the driver GART/VM setup (CIK). - */ -static void gmc_v8_0_gart_fini(struct amdgpu_device *adev) -{ - amdgpu_gart_table_vram_free(adev); - amdgpu_gart_fini(adev); -} - -/** * gmc_v8_0_vm_decode_fault - print human readable fault info * * @adev: amdgpu_device pointer @@ -1101,11 +1095,11 @@ static int gmc_v8_0_sw_init(void *handle) adev->gmc.vram_type = gmc_v8_0_convert_vram_type(tmp); } - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 146, &adev->gmc.vm_fault); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_PAGE_INV_FAULT, &adev->gmc.vm_fault); if (r) return r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 147, &adev->gmc.vm_fault); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_GFX_MEM_PROT_FAULT, &adev->gmc.vm_fault); if (r) return r; @@ -1181,6 +1175,12 @@ static int gmc_v8_0_sw_init(void *handle) adev->vm_manager.vram_base_offset = 0; } + adev->gmc.vm_fault_info = kmalloc(sizeof(struct kfd_vm_fault_info), + GFP_KERNEL); + if (!adev->gmc.vm_fault_info) + return -ENOMEM; + atomic_set(&adev->gmc.vm_fault_info_updated, 0); + return 0; } @@ -1190,8 +1190,10 @@ static int gmc_v8_0_sw_fini(void *handle) amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); - gmc_v8_0_gart_fini(adev); + kfree(adev->gmc.vm_fault_info); + amdgpu_gart_table_vram_free(adev); amdgpu_bo_fini(adev); + amdgpu_gart_fini(adev); release_firmware(adev->gmc.fw); adev->gmc.fw = NULL; @@ -1425,7 +1427,7 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - u32 addr, status, mc_client; + u32 addr, status, mc_client, vmid; if (amdgpu_sriov_vf(adev)) { dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n", @@ -1447,8 +1449,13 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev, gmc_v8_0_set_fault_enable_default(adev, false); if (printk_ratelimit()) { - dev_err(adev->dev, "GPU fault detected: %d 0x%08x\n", - entry->src_id, entry->src_data[0]); + struct amdgpu_task_info task_info = { 0 }; + + amdgpu_vm_get_task_info(adev, entry->pasid, &task_info); + + dev_err(adev->dev, "GPU fault detected: %d 0x%08x for process %s pid %d thread %s pid %d\n", + entry->src_id, entry->src_data[0], task_info.process_name, + task_info.tgid, task_info.task_name, task_info.pid); dev_err(adev->dev, " VM_CONTEXT1_PROTECTION_FAULT_ADDR 0x%08X\n", addr); dev_err(adev->dev, " VM_CONTEXT1_PROTECTION_FAULT_STATUS 0x%08X\n", @@ -1457,6 +1464,29 @@ static int gmc_v8_0_process_interrupt(struct amdgpu_device *adev, entry->pasid); } + vmid = REG_GET_FIELD(status, VM_CONTEXT1_PROTECTION_FAULT_STATUS, + VMID); + if (amdgpu_amdkfd_is_kfd_vmid(adev, vmid) + && !atomic_read(&adev->gmc.vm_fault_info_updated)) { + struct kfd_vm_fault_info *info = adev->gmc.vm_fault_info; + u32 protections = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + PROTECTIONS); + + info->vmid = vmid; + info->mc_id = REG_GET_FIELD(status, + VM_CONTEXT1_PROTECTION_FAULT_STATUS, + MEMORY_CLIENT_ID); + info->status = status; + info->page_addr = addr; + info->prot_valid = protections & 0x7 ? true : false; + info->prot_read = protections & 0x8 ? true : false; + info->prot_write = protections & 0x10 ? true : false; + info->prot_exec = protections & 0x20 ? true : false; + mb(); + atomic_set(&adev->gmc.vm_fault_info_updated, 1); + } + return 0; } @@ -1703,8 +1733,7 @@ static const struct amdgpu_irq_src_funcs gmc_v8_0_irq_funcs = { static void gmc_v8_0_set_gmc_funcs(struct amdgpu_device *adev) { - if (adev->gmc.gmc_funcs == NULL) - adev->gmc.gmc_funcs = &gmc_v8_0_gmc_funcs; + adev->gmc.gmc_funcs = &gmc_v8_0_gmc_funcs; } static void gmc_v8_0_set_irq_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index 3c0a85d4e4ab..f35d7a554ad5 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -25,6 +25,7 @@ #include "amdgpu.h" #include "gmc_v9_0.h" #include "amdgpu_atomfirmware.h" +#include "amdgpu_gem.h" #include "hdp/hdp_4_0_offset.h" #include "hdp/hdp_4_0_sh_mask.h" @@ -42,6 +43,9 @@ #include "gfxhub_v1_0.h" #include "mmhub_v1_0.h" +#include "gfxhub_v1_1.h" + +#include "ivsrcid/vmc/irqsrcs_vmc_1_0.h" /* add these here since we already include dce12 headers and these are for DCN */ #define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION 0x055d @@ -257,12 +261,17 @@ static int gmc_v9_0_process_interrupt(struct amdgpu_device *adev, } if (printk_ratelimit()) { + struct amdgpu_task_info task_info = { 0 }; + + amdgpu_vm_get_task_info(adev, entry->pasid, &task_info); + dev_err(adev->dev, - "[%s] VMC page fault (src_id:%u ring:%u vmid:%u pasid:%u)\n", + "[%s] VMC page fault (src_id:%u ring:%u vmid:%u pasid:%u, for process %s pid %d thread %s pid %d)\n", entry->vmid_src ? "mmhub" : "gfxhub", entry->src_id, entry->ring_id, entry->vmid, - entry->pasid); - dev_err(adev->dev, " at page 0x%016llx from %d\n", + entry->pasid, task_info.process_name, task_info.tgid, + task_info.task_name, task_info.pid); + dev_err(adev->dev, " in page starting at address 0x%016llx from %d\n", addr, entry->client_id); if (!amdgpu_sriov_vf(adev)) dev_err(adev->dev, @@ -303,6 +312,48 @@ static uint32_t gmc_v9_0_get_invalidate_req(unsigned int vmid) return req; } +static signed long amdgpu_kiq_reg_write_reg_wait(struct amdgpu_device *adev, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask) +{ + signed long r, cnt = 0; + unsigned long flags; + uint32_t seq; + struct amdgpu_kiq *kiq = &adev->gfx.kiq; + struct amdgpu_ring *ring = &kiq->ring; + + spin_lock_irqsave(&kiq->ring_lock, flags); + + amdgpu_ring_alloc(ring, 32); + amdgpu_ring_emit_reg_write_reg_wait(ring, reg0, reg1, + ref, mask); + amdgpu_fence_emit_polling(ring, &seq); + amdgpu_ring_commit(ring); + spin_unlock_irqrestore(&kiq->ring_lock, flags); + + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); + + /* don't wait anymore for IRQ context */ + if (r < 1 && in_interrupt()) + goto failed_kiq; + + might_sleep(); + + while (r < 1 && cnt++ < MAX_KIQ_REG_TRY) { + msleep(MAX_KIQ_REG_BAILOUT_INTERVAL); + r = amdgpu_fence_wait_polling(ring, seq, MAX_KIQ_REG_WAIT); + } + + if (cnt > MAX_KIQ_REG_TRY) + goto failed_kiq; + + return 0; + +failed_kiq: + pr_err("failed to invalidate tlb with kiq\n"); + return r; +} + /* * GART * VMID 0 is the physical GPU addresses as used by the kernel. @@ -324,13 +375,23 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, /* Use register 17 for GART */ const unsigned eng = 17; unsigned i, j; - - spin_lock(&adev->gmc.invalidate_lock); + int r; for (i = 0; i < AMDGPU_MAX_VMHUBS; ++i) { struct amdgpu_vmhub *hub = &adev->vmhub[i]; u32 tmp = gmc_v9_0_get_invalidate_req(vmid); + if (adev->gfx.kiq.ring.ready && + (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev)) && + !adev->in_gpu_reset) { + r = amdgpu_kiq_reg_write_reg_wait(adev, hub->vm_inv_eng0_req + eng, + hub->vm_inv_eng0_ack + eng, tmp, 1 << vmid); + if (!r) + continue; + } + + spin_lock(&adev->gmc.invalidate_lock); + WREG32_NO_KIQ(hub->vm_inv_eng0_req + eng, tmp); /* Busy wait for ACK.*/ @@ -341,8 +402,10 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, break; cpu_relax(); } - if (j < 100) + if (j < 100) { + spin_unlock(&adev->gmc.invalidate_lock); continue; + } /* Wait for ACK with a delay.*/ for (j = 0; j < adev->usec_timeout; j++) { @@ -352,13 +415,13 @@ static void gmc_v9_0_flush_gpu_tlb(struct amdgpu_device *adev, break; udelay(1); } - if (j < adev->usec_timeout) + if (j < adev->usec_timeout) { + spin_unlock(&adev->gmc.invalidate_lock); continue; - + } + spin_unlock(&adev->gmc.invalidate_lock); DRM_ERROR("Timeout waiting for VM flush ACK!\n"); } - - spin_unlock(&adev->gmc.invalidate_lock); } static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, @@ -367,12 +430,8 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, struct amdgpu_device *adev = ring->adev; struct amdgpu_vmhub *hub = &adev->vmhub[ring->funcs->vmhub]; uint32_t req = gmc_v9_0_get_invalidate_req(vmid); - uint64_t flags = AMDGPU_PTE_VALID; unsigned eng = ring->vm_inv_eng; - amdgpu_gmc_get_vm_pde(adev, -1, &pd_addr, &flags); - pd_addr |= flags; - amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_lo32 + (2 * vmid), lower_32_bits(pd_addr)); @@ -502,7 +561,7 @@ static uint64_t gmc_v9_0_get_vm_pte_flags(struct amdgpu_device *adev, static void gmc_v9_0_get_vm_pde(struct amdgpu_device *adev, int level, uint64_t *addr, uint64_t *flags) { - if (!(*flags & AMDGPU_PDE_PTE)) + if (!(*flags & AMDGPU_PDE_PTE) && !(*flags & AMDGPU_PTE_SYSTEM)) *addr = adev->vm_manager.vram_base_offset + *addr - adev->gmc.vram_start; BUG_ON(*addr & 0xFFFF00000000003FULL); @@ -534,8 +593,7 @@ static const struct amdgpu_gmc_funcs gmc_v9_0_gmc_funcs = { static void gmc_v9_0_set_gmc_funcs(struct amdgpu_device *adev) { - if (adev->gmc.gmc_funcs == NULL) - adev->gmc.gmc_funcs = &gmc_v9_0_gmc_funcs; + adev->gmc.gmc_funcs = &gmc_v9_0_gmc_funcs; } static int gmc_v9_0_early_init(void *handle) @@ -634,6 +692,29 @@ static int gmc_v9_0_ecc_available(struct amdgpu_device *adev) return lost_sheep == 0; } +static bool gmc_v9_0_keep_stolen_memory(struct amdgpu_device *adev) +{ + + /* + * TODO: + * Currently there is a bug where some memory client outside + * of the driver writes to first 8M of VRAM on S3 resume, + * this overrides GART which by default gets placed in first 8M and + * causes VM_FAULTS once GTT is accessed. + * Keep the stolen memory reservation until the while this is not solved. + * Also check code in gmc_v9_0_get_vbios_fb_size and gmc_v9_0_late_init + */ + switch (adev->asic_type) { + case CHIP_VEGA10: + return true; + case CHIP_RAVEN: + case CHIP_VEGA12: + case CHIP_VEGA20: + default: + return false; + } +} + static int gmc_v9_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -650,10 +731,8 @@ static int gmc_v9_0_late_init(void *handle) unsigned i; int r; - /* - * TODO - Uncomment once GART corruption issue is fixed. - */ - /* amdgpu_bo_late_init(adev); */ + if (!gmc_v9_0_keep_stolen_memory(adev)) + amdgpu_bo_late_init(adev); for(i = 0; i < adev->num_rings; ++i) { struct amdgpu_ring *ring = adev->rings[i]; @@ -691,10 +770,18 @@ static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev, u64 base = 0; if (!amdgpu_sriov_vf(adev)) base = mmhub_v1_0_get_fb_location(adev); - amdgpu_device_vram_location(adev, &adev->gmc, base); - amdgpu_device_gart_location(adev, mc); + /* add the xgmi offset of the physical node */ + base += adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size; + amdgpu_gmc_vram_location(adev, &adev->gmc, base); + amdgpu_gmc_gart_location(adev, mc); + if (!amdgpu_sriov_vf(adev)) + amdgpu_gmc_agp_location(adev, mc); /* base offset of vram pages */ adev->vm_manager.vram_base_offset = gfxhub_v1_0_get_mc_fb_offset(adev); + + /* XXX: add the xgmi offset of the physical node? */ + adev->vm_manager.vram_base_offset += + adev->gmc.xgmi.physical_node_id * adev->gmc.xgmi.node_segment_size; } /** @@ -774,7 +861,7 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev) { int r; - if (adev->gart.robj) { + if (adev->gart.bo) { WARN(1, "VEGA10 PCIE GART already initialized\n"); return 0; } @@ -790,18 +877,16 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev) static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) { -#if 0 u32 d1vga_control = RREG32_SOC15(DCE, 0, mmD1VGA_CONTROL); -#endif unsigned size; /* * TODO Remove once GART corruption is resolved * Check related code in gmc_v9_0_sw_fini * */ - size = 9 * 1024 * 1024; + if (gmc_v9_0_keep_stolen_memory(adev)) + return 9 * 1024 * 1024; -#if 0 if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) { size = 9 * 1024 * 1024; /* reserve 8MB for vga emulator and 1 MB for FB */ } else { @@ -818,6 +903,7 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) break; case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: default: viewport = RREG32_SOC15(DCE, 0, mmSCL0_VIEWPORT_SIZE); size = (REG_GET_FIELD(viewport, SCL0_VIEWPORT_SIZE, VIEWPORT_HEIGHT) * @@ -830,7 +916,6 @@ static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) if ((adev->gmc.real_vram_size - size) < (8 * 1024 * 1024)) return 0; -#endif return size; } @@ -872,9 +957,9 @@ static int gmc_v9_0_sw_init(void *handle) } /* This interrupt is VMC page fault.*/ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VMC, 0, + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VMC, VMC_1_0__SRCID__VM_FAULT, &adev->gmc.vm_fault); - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_UTCL2, 0, + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_UTCL2, UTCL2_1_0__SRCID__FAULT, &adev->gmc.vm_fault); if (r) @@ -906,6 +991,12 @@ static int gmc_v9_0_sw_init(void *handle) } adev->need_swiotlb = drm_get_max_iomem() > ((u64)1 << dma_bits); + if (adev->asic_type == CHIP_VEGA20) { + r = gfxhub_v1_1_get_xgmi_info(adev); + if (r) + return r; + } + r = gmc_v9_0_mc_init(adev); if (r) return r; @@ -935,39 +1026,19 @@ static int gmc_v9_0_sw_init(void *handle) return 0; } -/** - * gmc_v9_0_gart_fini - vm fini callback - * - * @adev: amdgpu_device pointer - * - * Tears down the driver GART/VM setup (CIK). - */ -static void gmc_v9_0_gart_fini(struct amdgpu_device *adev) -{ - amdgpu_gart_table_vram_free(adev); - amdgpu_gart_fini(adev); -} - static int gmc_v9_0_sw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); - gmc_v9_0_gart_fini(adev); - /* - * TODO: - * Currently there is a bug where some memory client outside - * of the driver writes to first 8M of VRAM on S3 resume, - * this overrides GART which by default gets placed in first 8M and - * causes VM_FAULTS once GTT is accessed. - * Keep the stolen memory reservation until the while this is not solved. - * Also check code in gmc_v9_0_get_vbios_fb_size and gmc_v9_0_late_init - */ - amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, NULL); + if (gmc_v9_0_keep_stolen_memory(adev)) + amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, NULL); + amdgpu_gart_table_vram_free(adev); amdgpu_bo_fini(adev); + amdgpu_gart_fini(adev); return 0; } @@ -1012,7 +1083,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) golden_settings_vega10_hdp, ARRAY_SIZE(golden_settings_vega10_hdp)); - if (adev->gart.robj == NULL) { + if (adev->gart.bo == NULL) { dev_err(adev->dev, "No VRAM object for PCIE GART.\n"); return -EINVAL; } @@ -1022,7 +1093,6 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_RAVEN: - mmhub_v1_0_initialize_power_gating(adev); mmhub_v1_0_update_power_gating(adev, true); break; default: @@ -1056,7 +1126,7 @@ static int gmc_v9_0_gart_enable(struct amdgpu_device *adev) DRM_INFO("PCIE GART of %uM enabled (table at 0x%016llX).\n", (unsigned)(adev->gmc.gart_size >> 20), - (unsigned long long)adev->gart.table_addr); + (unsigned long long)amdgpu_bo_gpu_offset(adev->gart.bo)); adev->gart.ready = true; return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c index 842c4b677b4d..cf0fc61aebe6 100644 --- a/drivers/gpu/drm/amd/amdgpu/iceland_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/iceland_ih.c @@ -255,7 +255,7 @@ static void iceland_ih_decode_iv(struct amdgpu_device *adev, dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - entry->client_id = AMDGPU_IH_CLIENTID_LEGACY; + entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; entry->src_data[0] = dw[1] & 0xfffffff; entry->ring_id = dw[2] & 0xff; @@ -297,7 +297,7 @@ static int iceland_ih_sw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = amdgpu_ih_ring_init(adev, 64 * 1024, false); + r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false); if (r) return r; @@ -311,7 +311,7 @@ static int iceland_ih_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_irq_fini(adev); - amdgpu_ih_ring_fini(adev); + amdgpu_ih_ring_fini(adev, &adev->irq.ih); amdgpu_irq_remove_domain(adev); return 0; @@ -447,8 +447,7 @@ static const struct amdgpu_ih_funcs iceland_ih_funcs = { static void iceland_ih_set_interrupt_funcs(struct amdgpu_device *adev) { - if (adev->irq.ih_funcs == NULL) - adev->irq.ih_funcs = &iceland_ih_funcs; + adev->irq.ih_funcs = &iceland_ih_funcs; } const struct amdgpu_ip_block_version iceland_ih_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index 7a1e77c93bf1..d0e478f43443 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -65,8 +65,6 @@ static int kv_set_thermal_temperature_range(struct amdgpu_device *adev, int min_temp, int max_temp); static int kv_init_fps_limits(struct amdgpu_device *adev); -static void kv_dpm_powergate_uvd(void *handle, bool gate); -static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate); static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate); static void kv_dpm_powergate_acp(struct amdgpu_device *adev, bool gate); @@ -1354,8 +1352,6 @@ static int kv_dpm_enable(struct amdgpu_device *adev) return ret; } - kv_update_current_ps(adev, adev->pm.dpm.boot_ps); - if (adev->irq.installed && amdgpu_is_internal_thermal_sensor(adev->pm.int_thermal_type)) { ret = kv_set_thermal_temperature_range(adev, KV_TEMP_RANGE_MIN, KV_TEMP_RANGE_MAX); @@ -1374,6 +1370,8 @@ static int kv_dpm_enable(struct amdgpu_device *adev) static void kv_dpm_disable(struct amdgpu_device *adev) { + struct kv_power_info *pi = kv_get_pi(adev); + amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq, AMDGPU_THERMAL_IRQ_LOW_TO_HIGH); amdgpu_irq_put(adev, &adev->pm.dpm.thermal.irq, @@ -1387,8 +1385,10 @@ static void kv_dpm_disable(struct amdgpu_device *adev) /* powerup blocks */ kv_dpm_powergate_acp(adev, false); kv_dpm_powergate_samu(adev, false); - kv_dpm_powergate_vce(adev, false); - kv_dpm_powergate_uvd(adev, false); + if (pi->caps_vce_pg) /* power on the VCE block */ + amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_VCEPowerON); + if (pi->caps_uvd_pg) /* power on the UVD block */ + amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_UVDPowerON); kv_enable_smc_cac(adev, false); kv_enable_didt(adev, false); @@ -1551,7 +1551,6 @@ static int kv_update_vce_dpm(struct amdgpu_device *adev, int ret; if (amdgpu_new_state->evclk > 0 && amdgpu_current_state->evclk == 0) { - kv_dpm_powergate_vce(adev, false); if (pi->caps_stable_p_state) pi->vce_boot_level = table->count - 1; else @@ -1573,7 +1572,6 @@ static int kv_update_vce_dpm(struct amdgpu_device *adev, kv_enable_vce_dpm(adev, true); } else if (amdgpu_new_state->evclk == 0 && amdgpu_current_state->evclk > 0) { kv_enable_vce_dpm(adev, false); - kv_dpm_powergate_vce(adev, true); } return 0; @@ -1702,24 +1700,32 @@ static void kv_dpm_powergate_uvd(void *handle, bool gate) } } -static void kv_dpm_powergate_vce(struct amdgpu_device *adev, bool gate) +static void kv_dpm_powergate_vce(void *handle, bool gate) { + struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct kv_power_info *pi = kv_get_pi(adev); - - if (pi->vce_power_gated == gate) - return; + int ret; pi->vce_power_gated = gate; - if (!pi->caps_vce_pg) - return; - - if (gate) - amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_VCEPowerOFF); - else - amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_VCEPowerON); + if (gate) { + /* stop the VCE block */ + ret = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCE, + AMD_PG_STATE_GATE); + kv_enable_vce_dpm(adev, false); + if (pi->caps_vce_pg) /* power off the VCE block */ + amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_VCEPowerOFF); + } else { + if (pi->caps_vce_pg) /* power on the VCE block */ + amdgpu_kv_notify_message_to_smu(adev, PPSMC_MSG_VCEPowerON); + kv_enable_vce_dpm(adev, true); + /* re-init the VCE block */ + ret = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCE, + AMD_PG_STATE_UNGATE); + } } + static void kv_dpm_powergate_samu(struct amdgpu_device *adev, bool gate) { struct kv_power_info *pi = kv_get_pi(adev); @@ -1921,7 +1927,7 @@ static int kv_dpm_set_power_state(void *handle) int ret; if (pi->bapm_enable) { - ret = amdgpu_kv_smc_bapm_enable(adev, adev->pm.dpm.ac_power); + ret = amdgpu_kv_smc_bapm_enable(adev, adev->pm.ac_power); if (ret) { DRM_ERROR("amdgpu_kv_smc_bapm_enable failed\n"); return ret; @@ -2989,12 +2995,12 @@ static int kv_dpm_sw_init(void *handle) int ret; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 230, + ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 230, &adev->pm.dpm.thermal.irq); if (ret) return ret; - ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 231, + ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 231, &adev->pm.dpm.thermal.irq); if (ret) return ret; @@ -3061,7 +3067,7 @@ static int kv_dpm_hw_init(void *handle) else adev->pm.dpm_enabled = true; mutex_unlock(&adev->pm.mutex); - + amdgpu_pm_compute_clocks(adev); return ret; } @@ -3306,6 +3312,22 @@ static int kv_dpm_read_sensor(void *handle, int idx, } } +static int kv_set_powergating_by_smu(void *handle, + uint32_t block_type, bool gate) +{ + switch (block_type) { + case AMD_IP_BLOCK_TYPE_UVD: + kv_dpm_powergate_uvd(handle, gate); + break; + case AMD_IP_BLOCK_TYPE_VCE: + kv_dpm_powergate_vce(handle, gate); + break; + default: + break; + } + return 0; +} + static const struct amd_ip_funcs kv_dpm_ip_funcs = { .name = "kv_dpm", .early_init = kv_dpm_early_init, @@ -3342,7 +3364,7 @@ static const struct amd_pm_funcs kv_dpm_funcs = { .print_power_state = &kv_dpm_print_power_state, .debugfs_print_current_performance_level = &kv_dpm_debugfs_print_current_performance_level, .force_performance_level = &kv_dpm_force_performance_level, - .powergate_uvd = &kv_dpm_powergate_uvd, + .set_powergating_by_smu = kv_set_powergating_by_smu, .enable_bapm = &kv_dpm_enable_bapm, .get_vce_clock_state = amdgpu_get_vce_clock_state, .check_state_equal = kv_check_state_equal, diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 3d53c4413f13..14649f8475f3 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -38,22 +38,23 @@ u64 mmhub_v1_0_get_fb_location(struct amdgpu_device *adev) { u64 base = RREG32_SOC15(MMHUB, 0, mmMC_VM_FB_LOCATION_BASE); + u64 top = RREG32_SOC15(MMHUB, 0, mmMC_VM_FB_LOCATION_TOP); base &= MC_VM_FB_LOCATION_BASE__FB_BASE_MASK; base <<= 24; + top &= MC_VM_FB_LOCATION_TOP__FB_TOP_MASK; + top <<= 24; + + adev->gmc.fb_start = base; + adev->gmc.fb_end = top; + return base; } static void mmhub_v1_0_init_gart_pt_regs(struct amdgpu_device *adev) { - uint64_t value; - - BUG_ON(adev->gart.table_addr & (~0x0000FFFFFFFFF000ULL)); - value = adev->gart.table_addr - adev->gmc.vram_start + - adev->vm_manager.vram_base_offset; - value &= 0x0000FFFFFFFFF000ULL; - value |= 0x1; /* valid bit */ + uint64_t value = amdgpu_gmc_pd_addr(adev->gart.bo); WREG32_SOC15(MMHUB, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32, lower_32_bits(value)); @@ -82,16 +83,28 @@ static void mmhub_v1_0_init_system_aperture_regs(struct amdgpu_device *adev) uint64_t value; uint32_t tmp; - /* Disable AGP. */ + /* Program the AGP BAR */ WREG32_SOC15(MMHUB, 0, mmMC_VM_AGP_BASE, 0); - WREG32_SOC15(MMHUB, 0, mmMC_VM_AGP_TOP, 0); - WREG32_SOC15(MMHUB, 0, mmMC_VM_AGP_BOT, 0x00FFFFFF); + WREG32_SOC15(MMHUB, 0, mmMC_VM_AGP_BOT, adev->gmc.agp_start >> 24); + WREG32_SOC15(MMHUB, 0, mmMC_VM_AGP_TOP, adev->gmc.agp_end >> 24); /* Program the system aperture low logical page number. */ WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_LOW_ADDR, - adev->gmc.vram_start >> 18); - WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, - adev->gmc.vram_end >> 18); + min(adev->gmc.vram_start, adev->gmc.agp_start) >> 18); + + if (adev->asic_type == CHIP_RAVEN && adev->rev_id >= 0x8) + /* + * Raven2 has a HW issue that it is unable to use the vram which + * is out of MC_VM_SYSTEM_APERTURE_HIGH_ADDR. So here is the + * workaround that increase system aperture high address (add 1) + * to get rid of the VM fault and hardware hang. + */ + WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + max((adev->gmc.vram_end >> 18) + 0x1, + adev->gmc.agp_end >> 18)); + else + WREG32_SOC15(MMHUB, 0, mmMC_VM_SYSTEM_APERTURE_HIGH_ADDR, + max(adev->gmc.vram_end, adev->gmc.agp_end) >> 18); /* Set default page address. */ value = adev->vram_scratch.gpu_addr - adev->gmc.vram_start + @@ -260,236 +273,16 @@ static void mmhub_v1_0_program_invalidation(struct amdgpu_device *adev) } } -struct pctl_data { - uint32_t index; - uint32_t data; -}; - -static const struct pctl_data pctl0_data[] = { - {0x0, 0x7a640}, - {0x9, 0x2a64a}, - {0xd, 0x2a680}, - {0x11, 0x6a684}, - {0x19, 0xea68e}, - {0x29, 0xa69e}, - {0x2b, 0x0010a6c0}, - {0x3d, 0x83a707}, - {0xc2, 0x8a7a4}, - {0xcc, 0x1a7b8}, - {0xcf, 0xfa7cc}, - {0xe0, 0x17a7dd}, - {0xf9, 0xa7dc}, - {0xfb, 0x12a7f5}, - {0x10f, 0xa808}, - {0x111, 0x12a810}, - {0x125, 0x7a82c} -}; -#define PCTL0_DATA_LEN (ARRAY_SIZE(pctl0_data)) - -#define PCTL0_RENG_EXEC_END_PTR 0x12d -#define PCTL0_STCTRL_REG_SAVE_RANGE0_BASE 0xa640 -#define PCTL0_STCTRL_REG_SAVE_RANGE0_LIMIT 0xa833 - -static const struct pctl_data pctl1_data[] = { - {0x0, 0x39a000}, - {0x3b, 0x44a040}, - {0x81, 0x2a08d}, - {0x85, 0x6ba094}, - {0xf2, 0x18a100}, - {0x10c, 0x4a132}, - {0x112, 0xca141}, - {0x120, 0x2fa158}, - {0x151, 0x17a1d0}, - {0x16a, 0x1a1e9}, - {0x16d, 0x13a1ec}, - {0x182, 0x7a201}, - {0x18b, 0x3a20a}, - {0x190, 0x7a580}, - {0x199, 0xa590}, - {0x19b, 0x4a594}, - {0x1a1, 0x1a59c}, - {0x1a4, 0x7a82c}, - {0x1ad, 0xfa7cc}, - {0x1be, 0x17a7dd}, - {0x1d7, 0x12a810}, - {0x1eb, 0x4000a7e1}, - {0x1ec, 0x5000a7f5}, - {0x1ed, 0x4000a7e2}, - {0x1ee, 0x5000a7dc}, - {0x1ef, 0x4000a7e3}, - {0x1f0, 0x5000a7f6}, - {0x1f1, 0x5000a7e4} -}; -#define PCTL1_DATA_LEN (ARRAY_SIZE(pctl1_data)) - -#define PCTL1_RENG_EXEC_END_PTR 0x1f1 -#define PCTL1_STCTRL_REG_SAVE_RANGE0_BASE 0xa000 -#define PCTL1_STCTRL_REG_SAVE_RANGE0_LIMIT 0xa20d -#define PCTL1_STCTRL_REG_SAVE_RANGE1_BASE 0xa580 -#define PCTL1_STCTRL_REG_SAVE_RANGE1_LIMIT 0xa59d -#define PCTL1_STCTRL_REG_SAVE_RANGE2_BASE 0xa82c -#define PCTL1_STCTRL_REG_SAVE_RANGE2_LIMIT 0xa833 - -static void mmhub_v1_0_power_gating_write_save_ranges(struct amdgpu_device *adev) -{ - uint32_t tmp = 0; - - /* PCTL0_STCTRL_REGISTER_SAVE_RANGE0 */ - tmp = REG_SET_FIELD(tmp, PCTL0_STCTRL_REGISTER_SAVE_RANGE0, - STCTRL_REGISTER_SAVE_BASE, - PCTL0_STCTRL_REG_SAVE_RANGE0_BASE); - tmp = REG_SET_FIELD(tmp, PCTL0_STCTRL_REGISTER_SAVE_RANGE0, - STCTRL_REGISTER_SAVE_LIMIT, - PCTL0_STCTRL_REG_SAVE_RANGE0_LIMIT); - WREG32_SOC15(MMHUB, 0, mmPCTL0_STCTRL_REGISTER_SAVE_RANGE0, tmp); - - /* PCTL1_STCTRL_REGISTER_SAVE_RANGE0 */ - tmp = 0; - tmp = REG_SET_FIELD(tmp, PCTL1_STCTRL_REGISTER_SAVE_RANGE0, - STCTRL_REGISTER_SAVE_BASE, - PCTL1_STCTRL_REG_SAVE_RANGE0_BASE); - tmp = REG_SET_FIELD(tmp, PCTL1_STCTRL_REGISTER_SAVE_RANGE0, - STCTRL_REGISTER_SAVE_LIMIT, - PCTL1_STCTRL_REG_SAVE_RANGE0_LIMIT); - WREG32_SOC15(MMHUB, 0, mmPCTL1_STCTRL_REGISTER_SAVE_RANGE0, tmp); - - /* PCTL1_STCTRL_REGISTER_SAVE_RANGE1 */ - tmp = 0; - tmp = REG_SET_FIELD(tmp, PCTL1_STCTRL_REGISTER_SAVE_RANGE1, - STCTRL_REGISTER_SAVE_BASE, - PCTL1_STCTRL_REG_SAVE_RANGE1_BASE); - tmp = REG_SET_FIELD(tmp, PCTL1_STCTRL_REGISTER_SAVE_RANGE1, - STCTRL_REGISTER_SAVE_LIMIT, - PCTL1_STCTRL_REG_SAVE_RANGE1_LIMIT); - WREG32_SOC15(MMHUB, 0, mmPCTL1_STCTRL_REGISTER_SAVE_RANGE1, tmp); - - /* PCTL1_STCTRL_REGISTER_SAVE_RANGE2 */ - tmp = 0; - tmp = REG_SET_FIELD(tmp, PCTL1_STCTRL_REGISTER_SAVE_RANGE2, - STCTRL_REGISTER_SAVE_BASE, - PCTL1_STCTRL_REG_SAVE_RANGE2_BASE); - tmp = REG_SET_FIELD(tmp, PCTL1_STCTRL_REGISTER_SAVE_RANGE2, - STCTRL_REGISTER_SAVE_LIMIT, - PCTL1_STCTRL_REG_SAVE_RANGE2_LIMIT); - WREG32_SOC15(MMHUB, 0, mmPCTL1_STCTRL_REGISTER_SAVE_RANGE2, tmp); -} - -void mmhub_v1_0_initialize_power_gating(struct amdgpu_device *adev) -{ - uint32_t pctl0_misc = 0; - uint32_t pctl0_reng_execute = 0; - uint32_t pctl1_misc = 0; - uint32_t pctl1_reng_execute = 0; - int i = 0; - - if (amdgpu_sriov_vf(adev)) - return; - - /****************** pctl0 **********************/ - pctl0_misc = RREG32_SOC15(MMHUB, 0, mmPCTL0_MISC); - pctl0_reng_execute = RREG32_SOC15(MMHUB, 0, mmPCTL0_RENG_EXECUTE); - - /* Light sleep must be disabled before writing to pctl0 registers */ - pctl0_misc &= ~PCTL0_MISC__RENG_MEM_LS_ENABLE_MASK; - WREG32_SOC15(MMHUB, 0, mmPCTL0_MISC, pctl0_misc); - - /* Write data used to access ram of register engine */ - for (i = 0; i < PCTL0_DATA_LEN; i++) { - WREG32_SOC15(MMHUB, 0, mmPCTL0_RENG_RAM_INDEX, - pctl0_data[i].index); - WREG32_SOC15(MMHUB, 0, mmPCTL0_RENG_RAM_DATA, - pctl0_data[i].data); - } - - /* Re-enable light sleep */ - pctl0_misc |= PCTL0_MISC__RENG_MEM_LS_ENABLE_MASK; - WREG32_SOC15(MMHUB, 0, mmPCTL0_MISC, pctl0_misc); - - /****************** pctl1 **********************/ - pctl1_misc = RREG32_SOC15(MMHUB, 0, mmPCTL1_MISC); - pctl1_reng_execute = RREG32_SOC15(MMHUB, 0, mmPCTL1_RENG_EXECUTE); - - /* Light sleep must be disabled before writing to pctl1 registers */ - pctl1_misc &= ~PCTL1_MISC__RENG_MEM_LS_ENABLE_MASK; - WREG32_SOC15(MMHUB, 0, mmPCTL1_MISC, pctl1_misc); - - /* Write data used to access ram of register engine */ - for (i = 0; i < PCTL1_DATA_LEN; i++) { - WREG32_SOC15(MMHUB, 0, mmPCTL1_RENG_RAM_INDEX, - pctl1_data[i].index); - WREG32_SOC15(MMHUB, 0, mmPCTL1_RENG_RAM_DATA, - pctl1_data[i].data); - } - - /* Re-enable light sleep */ - pctl1_misc |= PCTL1_MISC__RENG_MEM_LS_ENABLE_MASK; - WREG32_SOC15(MMHUB, 0, mmPCTL1_MISC, pctl1_misc); - - mmhub_v1_0_power_gating_write_save_ranges(adev); - - /* Set the reng execute end ptr for pctl0 */ - pctl0_reng_execute = REG_SET_FIELD(pctl0_reng_execute, - PCTL0_RENG_EXECUTE, - RENG_EXECUTE_END_PTR, - PCTL0_RENG_EXEC_END_PTR); - WREG32_SOC15(MMHUB, 0, mmPCTL0_RENG_EXECUTE, pctl0_reng_execute); - - /* Set the reng execute end ptr for pctl1 */ - pctl1_reng_execute = REG_SET_FIELD(pctl1_reng_execute, - PCTL1_RENG_EXECUTE, - RENG_EXECUTE_END_PTR, - PCTL1_RENG_EXEC_END_PTR); - WREG32_SOC15(MMHUB, 0, mmPCTL1_RENG_EXECUTE, pctl1_reng_execute); -} - void mmhub_v1_0_update_power_gating(struct amdgpu_device *adev, bool enable) { - uint32_t pctl0_reng_execute = 0; - uint32_t pctl1_reng_execute = 0; - if (amdgpu_sriov_vf(adev)) return; - pctl0_reng_execute = RREG32_SOC15(MMHUB, 0, mmPCTL0_RENG_EXECUTE); - pctl1_reng_execute = RREG32_SOC15(MMHUB, 0, mmPCTL1_RENG_EXECUTE); - if (enable && adev->pg_flags & AMD_PG_SUPPORT_MMHUB) { - pctl0_reng_execute = REG_SET_FIELD(pctl0_reng_execute, - PCTL0_RENG_EXECUTE, - RENG_EXECUTE_ON_PWR_UP, 1); - pctl0_reng_execute = REG_SET_FIELD(pctl0_reng_execute, - PCTL0_RENG_EXECUTE, - RENG_EXECUTE_ON_REG_UPDATE, 1); - WREG32_SOC15(MMHUB, 0, mmPCTL0_RENG_EXECUTE, pctl0_reng_execute); - - pctl1_reng_execute = REG_SET_FIELD(pctl1_reng_execute, - PCTL1_RENG_EXECUTE, - RENG_EXECUTE_ON_PWR_UP, 1); - pctl1_reng_execute = REG_SET_FIELD(pctl1_reng_execute, - PCTL1_RENG_EXECUTE, - RENG_EXECUTE_ON_REG_UPDATE, 1); - WREG32_SOC15(MMHUB, 0, mmPCTL1_RENG_EXECUTE, pctl1_reng_execute); - - if (adev->powerplay.pp_funcs->set_mmhub_powergating_by_smu) - amdgpu_dpm_set_mmhub_powergating_by_smu(adev); + if (adev->powerplay.pp_funcs->set_powergating_by_smu) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_GMC, true); - } else { - pctl0_reng_execute = REG_SET_FIELD(pctl0_reng_execute, - PCTL0_RENG_EXECUTE, - RENG_EXECUTE_ON_PWR_UP, 0); - pctl0_reng_execute = REG_SET_FIELD(pctl0_reng_execute, - PCTL0_RENG_EXECUTE, - RENG_EXECUTE_ON_REG_UPDATE, 0); - WREG32_SOC15(MMHUB, 0, mmPCTL0_RENG_EXECUTE, pctl0_reng_execute); - - pctl1_reng_execute = REG_SET_FIELD(pctl1_reng_execute, - PCTL1_RENG_EXECUTE, - RENG_EXECUTE_ON_PWR_UP, 0); - pctl1_reng_execute = REG_SET_FIELD(pctl1_reng_execute, - PCTL1_RENG_EXECUTE, - RENG_EXECUTE_ON_REG_UPDATE, 0); - WREG32_SOC15(MMHUB, 0, mmPCTL1_RENG_EXECUTE, pctl1_reng_execute); } } diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h index 5d38229baf69..bef3d0c0c117 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.h @@ -32,7 +32,6 @@ void mmhub_v1_0_init(struct amdgpu_device *adev); int mmhub_v1_0_set_clockgating(struct amdgpu_device *adev, enum amd_clockgating_state state); void mmhub_v1_0_get_clockgating(struct amdgpu_device *adev, u32 *flags); -void mmhub_v1_0_initialize_power_gating(struct amdgpu_device *adev); void mmhub_v1_0_update_power_gating(struct amdgpu_device *adev, bool enable); diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index 078f70faedcb..8cbb4655896a 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -266,8 +266,8 @@ flr_done: } /* Trigger recovery for world switch failure if no TDR */ - if (amdgpu_lockup_timeout == 0) - amdgpu_device_gpu_recover(adev, NULL, true); + if (amdgpu_device_should_recover_gpu(adev)) + amdgpu_device_gpu_recover(adev, NULL); } static int xgpu_ai_set_mailbox_rcv_irq(struct amdgpu_device *adev, diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c index 9fc1c37344ce..64e875d528dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_vi.c @@ -521,7 +521,8 @@ static void xgpu_vi_mailbox_flr_work(struct work_struct *work) } /* Trigger recovery due to world switch failure */ - amdgpu_device_gpu_recover(adev, NULL, false); + if (amdgpu_device_should_recover_gpu(adev)) + amdgpu_device_gpu_recover(adev, NULL); } static int xgpu_vi_set_mailbox_rcv_irq(struct amdgpu_device *adev, @@ -579,11 +580,11 @@ int xgpu_vi_mailbox_add_irq_id(struct amdgpu_device *adev) { int r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 135, &adev->virt.rcv_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 135, &adev->virt.rcv_irq); if (r) return r; - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 138, &adev->virt.ack_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 138, &adev->virt.ack_irq); if (r) { amdgpu_irq_put(adev, &adev->virt.rcv_irq, 0); return r; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c index 365517c0121e..df34dc79d444 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c @@ -34,19 +34,10 @@ #define smnCPM_CONTROL 0x11180460 #define smnPCIE_CNTL2 0x11180070 -/* vega20 */ -#define mmRCC_DEV0_EPF0_STRAP0_VG20 0x0011 -#define mmRCC_DEV0_EPF0_STRAP0_VG20_BASE_IDX 2 - static u32 nbio_v7_0_get_rev_id(struct amdgpu_device *adev) { u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); - if (adev->asic_type == CHIP_VEGA20) - tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0_VG20); - else - tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); - tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; @@ -84,14 +75,10 @@ static void nbio_v7_0_sdma_doorbell_range(struct amdgpu_device *adev, int instan SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE); u32 doorbell_range = RREG32(reg); - u32 range = 2; - - if (adev->asic_type == CHIP_VEGA20) - range = 8; if (use_doorbell) { doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index); - doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, range); + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 2); } else doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0); @@ -146,9 +133,6 @@ static void nbio_v7_0_update_medium_grain_clock_gating(struct amdgpu_device *ade { uint32_t def, data; - if (adev->asic_type == CHIP_VEGA20) - return; - /* NBIF_MGCG_CTRL_LCLK */ def = data = RREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK); diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c new file mode 100644 index 000000000000..f8cee95d61cc --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.c @@ -0,0 +1,248 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "amdgpu_atombios.h" +#include "nbio_v7_4.h" + +#include "nbio/nbio_7_4_offset.h" +#include "nbio/nbio_7_4_sh_mask.h" + +#define smnNBIF_MGCG_CTRL_LCLK 0x1013a21c + +#define smnCPM_CONTROL 0x11180460 +#define smnPCIE_CNTL2 0x11180070 + +static u32 nbio_v7_4_get_rev_id(struct amdgpu_device *adev) +{ + u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); + + tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; + tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; + + return tmp; +} + +static void nbio_v7_4_mc_access_enable(struct amdgpu_device *adev, bool enable) +{ + if (enable) + WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, + BIF_FB_EN__FB_READ_EN_MASK | BIF_FB_EN__FB_WRITE_EN_MASK); + else + WREG32_SOC15(NBIO, 0, mmBIF_FB_EN, 0); +} + +static void nbio_v7_4_hdp_flush(struct amdgpu_device *adev, + struct amdgpu_ring *ring) +{ + if (!ring || !ring->funcs->emit_wreg) + WREG32_SOC15_NO_KIQ(NBIO, 0, mmHDP_MEM_COHERENCY_FLUSH_CNTL, 0); + else + amdgpu_ring_emit_wreg(ring, SOC15_REG_OFFSET( + NBIO, 0, mmHDP_MEM_COHERENCY_FLUSH_CNTL), 0); +} + +static u32 nbio_v7_4_get_memsize(struct amdgpu_device *adev) +{ + return RREG32_SOC15(NBIO, 0, mmRCC_CONFIG_MEMSIZE); +} + +static void nbio_v7_4_sdma_doorbell_range(struct amdgpu_device *adev, int instance, + bool use_doorbell, int doorbell_index) +{ + u32 reg = instance == 0 ? SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA0_DOORBELL_RANGE) : + SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE); + + u32 doorbell_range = RREG32(reg); + + if (use_doorbell) { + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index); + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 2); + } else + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0); + + WREG32(reg, doorbell_range); +} + +static void nbio_v7_4_enable_doorbell_aperture(struct amdgpu_device *adev, + bool enable) +{ + WREG32_FIELD15(NBIO, 0, RCC_DOORBELL_APER_EN, BIF_DOORBELL_APER_EN, enable ? 1 : 0); +} + +static void nbio_v7_4_enable_doorbell_selfring_aperture(struct amdgpu_device *adev, + bool enable) +{ + +} + +static void nbio_v7_4_ih_doorbell_range(struct amdgpu_device *adev, + bool use_doorbell, int doorbell_index) +{ + u32 ih_doorbell_range = RREG32_SOC15(NBIO, 0 , mmBIF_IH_DOORBELL_RANGE); + + if (use_doorbell) { + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, OFFSET, doorbell_index); + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 2); + } else + ih_doorbell_range = REG_SET_FIELD(ih_doorbell_range, BIF_IH_DOORBELL_RANGE, SIZE, 0); + + WREG32_SOC15(NBIO, 0, mmBIF_IH_DOORBELL_RANGE, ih_doorbell_range); +} + + +static void nbio_v7_4_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + //TODO: Add support for v7.4 +} + +static void nbio_v7_4_update_medium_grain_light_sleep(struct amdgpu_device *adev, + bool enable) +{ + uint32_t def, data; + + def = data = RREG32_PCIE(smnPCIE_CNTL2); + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_BIF_LS)) { + data |= (PCIE_CNTL2__SLV_MEM_LS_EN_MASK | + PCIE_CNTL2__MST_MEM_LS_EN_MASK | + PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK); + } else { + data &= ~(PCIE_CNTL2__SLV_MEM_LS_EN_MASK | + PCIE_CNTL2__MST_MEM_LS_EN_MASK | + PCIE_CNTL2__REPLAY_MEM_LS_EN_MASK); + } + + if (def != data) + WREG32_PCIE(smnPCIE_CNTL2, data); +} + +static void nbio_v7_4_get_clockgating_state(struct amdgpu_device *adev, + u32 *flags) +{ + int data; + + /* AMD_CG_SUPPORT_BIF_MGCG */ + data = RREG32_PCIE(smnCPM_CONTROL); + if (data & CPM_CONTROL__LCLK_DYN_GATE_ENABLE_MASK) + *flags |= AMD_CG_SUPPORT_BIF_MGCG; + + /* AMD_CG_SUPPORT_BIF_LS */ + data = RREG32_PCIE(smnPCIE_CNTL2); + if (data & PCIE_CNTL2__SLV_MEM_LS_EN_MASK) + *flags |= AMD_CG_SUPPORT_BIF_LS; +} + +static void nbio_v7_4_ih_control(struct amdgpu_device *adev) +{ + u32 interrupt_cntl; + + /* setup interrupt control */ + WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL2, adev->dummy_page_addr >> 8); + interrupt_cntl = RREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL); + /* INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=0 - dummy read disabled with msi, enabled without msi + * INTERRUPT_CNTL__IH_DUMMY_RD_OVERRIDE_MASK=1 - dummy read controlled by IH_DUMMY_RD_EN + */ + interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_DUMMY_RD_OVERRIDE, 0); + /* INTERRUPT_CNTL__IH_REQ_NONSNOOP_EN_MASK=1 if ring is in non-cacheable memory, e.g., vram */ + interrupt_cntl = REG_SET_FIELD(interrupt_cntl, INTERRUPT_CNTL, IH_REQ_NONSNOOP_EN, 0); + WREG32_SOC15(NBIO, 0, mmINTERRUPT_CNTL, interrupt_cntl); +} + +static u32 nbio_v7_4_get_hdp_flush_req_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_REQ); +} + +static u32 nbio_v7_4_get_hdp_flush_done_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, mmGPU_HDP_FLUSH_DONE); +} + +static u32 nbio_v7_4_get_pcie_index_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_INDEX2); +} + +static u32 nbio_v7_4_get_pcie_data_offset(struct amdgpu_device *adev) +{ + return SOC15_REG_OFFSET(NBIO, 0, mmPCIE_DATA2); +} + +static const struct nbio_hdp_flush_reg nbio_v7_4_hdp_flush_reg = { + .ref_and_mask_cp0 = GPU_HDP_FLUSH_DONE__CP0_MASK, + .ref_and_mask_cp1 = GPU_HDP_FLUSH_DONE__CP1_MASK, + .ref_and_mask_cp2 = GPU_HDP_FLUSH_DONE__CP2_MASK, + .ref_and_mask_cp3 = GPU_HDP_FLUSH_DONE__CP3_MASK, + .ref_and_mask_cp4 = GPU_HDP_FLUSH_DONE__CP4_MASK, + .ref_and_mask_cp5 = GPU_HDP_FLUSH_DONE__CP5_MASK, + .ref_and_mask_cp6 = GPU_HDP_FLUSH_DONE__CP6_MASK, + .ref_and_mask_cp7 = GPU_HDP_FLUSH_DONE__CP7_MASK, + .ref_and_mask_cp8 = GPU_HDP_FLUSH_DONE__CP8_MASK, + .ref_and_mask_cp9 = GPU_HDP_FLUSH_DONE__CP9_MASK, + .ref_and_mask_sdma0 = GPU_HDP_FLUSH_DONE__SDMA0_MASK, + .ref_and_mask_sdma1 = GPU_HDP_FLUSH_DONE__SDMA1_MASK, +}; + +static void nbio_v7_4_detect_hw_virt(struct amdgpu_device *adev) +{ + uint32_t reg; + + reg = RREG32_SOC15(NBIO, 0, mmRCC_IOV_FUNC_IDENTIFIER); + if (reg & 1) + adev->virt.caps |= AMDGPU_SRIOV_CAPS_IS_VF; + + if (reg & 0x80000000) + adev->virt.caps |= AMDGPU_SRIOV_CAPS_ENABLE_IOV; + + if (!reg) { + if (is_virtual_machine()) /* passthrough mode exclus sriov mod */ + adev->virt.caps |= AMDGPU_PASSTHROUGH_MODE; + } +} + +static void nbio_v7_4_init_registers(struct amdgpu_device *adev) +{ + +} + +const struct amdgpu_nbio_funcs nbio_v7_4_funcs = { + .hdp_flush_reg = &nbio_v7_4_hdp_flush_reg, + .get_hdp_flush_req_offset = nbio_v7_4_get_hdp_flush_req_offset, + .get_hdp_flush_done_offset = nbio_v7_4_get_hdp_flush_done_offset, + .get_pcie_index_offset = nbio_v7_4_get_pcie_index_offset, + .get_pcie_data_offset = nbio_v7_4_get_pcie_data_offset, + .get_rev_id = nbio_v7_4_get_rev_id, + .mc_access_enable = nbio_v7_4_mc_access_enable, + .hdp_flush = nbio_v7_4_hdp_flush, + .get_memsize = nbio_v7_4_get_memsize, + .sdma_doorbell_range = nbio_v7_4_sdma_doorbell_range, + .enable_doorbell_aperture = nbio_v7_4_enable_doorbell_aperture, + .enable_doorbell_selfring_aperture = nbio_v7_4_enable_doorbell_selfring_aperture, + .ih_doorbell_range = nbio_v7_4_ih_doorbell_range, + .update_medium_grain_clock_gating = nbio_v7_4_update_medium_grain_clock_gating, + .update_medium_grain_light_sleep = nbio_v7_4_update_medium_grain_light_sleep, + .get_clockgating_state = nbio_v7_4_get_clockgating_state, + .ih_control = nbio_v7_4_ih_control, + .init_registers = nbio_v7_4_init_registers, + .detect_hw_virt = nbio_v7_4_detect_hw_virt, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h new file mode 100644 index 000000000000..c442865bac4f --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_4.h @@ -0,0 +1,31 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __NBIO_V7_4_H__ +#define __NBIO_V7_4_H__ + +#include "soc15_common.h" + +extern const struct amdgpu_nbio_funcs nbio_v7_4_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index 0cf48d26c676..882bd83a28c4 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -189,7 +189,8 @@ enum psp_gfx_fw_type GFX_FW_TYPE_RLC_RESTORE_LIST_GPM_MEM = 20, GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_MEM = 21, GFX_FW_TYPE_RLC_RESTORE_LIST_CNTL = 22, - GFX_FW_TYPE_MAX = 23 + GFX_FW_TYPE_UVD1 = 23, + GFX_FW_TYPE_MAX = 24 }; /* Command to load HW IP FW. */ diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index 0ff136d02d9b..295c2205485a 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -35,6 +35,8 @@ #include "sdma0/sdma0_4_1_offset.h" MODULE_FIRMWARE("amdgpu/raven_asd.bin"); +MODULE_FIRMWARE("amdgpu/picasso_asd.bin"); +MODULE_FIRMWARE("amdgpu/raven2_asd.bin"); static int psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) @@ -88,6 +90,15 @@ psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type * case AMDGPU_UCODE_ID_VCE: *type = GFX_FW_TYPE_VCE; break; + case AMDGPU_UCODE_ID_VCN: + *type = GFX_FW_TYPE_VCN; + break; + case AMDGPU_UCODE_ID_DMCU_ERAM: + *type = GFX_FW_TYPE_DMCU_ERAM; + break; + case AMDGPU_UCODE_ID_DMCU_INTV: + *type = GFX_FW_TYPE_DMCU_ISR; + break; case AMDGPU_UCODE_ID_MAXIMUM: default: return -EINVAL; @@ -108,7 +119,12 @@ static int psp_v10_0_init_microcode(struct psp_context *psp) switch (adev->asic_type) { case CHIP_RAVEN: - chip_name = "raven"; + if (adev->rev_id >= 0x8) + chip_name = "raven2"; + else if (adev->pdev->device == 0x15d8) + chip_name = "picasso"; + else + chip_name = "raven"; break; default: BUG(); } diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c new file mode 100644 index 000000000000..3f3fac2d50cd --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.c @@ -0,0 +1,595 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include <linux/firmware.h> +#include "amdgpu.h" +#include "amdgpu_psp.h" +#include "amdgpu_ucode.h" +#include "soc15_common.h" +#include "psp_v11_0.h" + +#include "mp/mp_11_0_offset.h" +#include "mp/mp_11_0_sh_mask.h" +#include "gc/gc_9_0_offset.h" +#include "sdma0/sdma0_4_0_offset.h" +#include "nbio/nbio_7_4_offset.h" + +MODULE_FIRMWARE("amdgpu/vega20_sos.bin"); + +/* address block */ +#define smnMP1_FIRMWARE_FLAGS 0x3010024 + +static int +psp_v11_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type *type) +{ + switch (ucode->ucode_id) { + case AMDGPU_UCODE_ID_SDMA0: + *type = GFX_FW_TYPE_SDMA0; + break; + case AMDGPU_UCODE_ID_SDMA1: + *type = GFX_FW_TYPE_SDMA1; + break; + case AMDGPU_UCODE_ID_CP_CE: + *type = GFX_FW_TYPE_CP_CE; + break; + case AMDGPU_UCODE_ID_CP_PFP: + *type = GFX_FW_TYPE_CP_PFP; + break; + case AMDGPU_UCODE_ID_CP_ME: + *type = GFX_FW_TYPE_CP_ME; + break; + case AMDGPU_UCODE_ID_CP_MEC1: + *type = GFX_FW_TYPE_CP_MEC; + break; + case AMDGPU_UCODE_ID_CP_MEC1_JT: + *type = GFX_FW_TYPE_CP_MEC_ME1; + break; + case AMDGPU_UCODE_ID_CP_MEC2: + *type = GFX_FW_TYPE_CP_MEC; + break; + case AMDGPU_UCODE_ID_CP_MEC2_JT: + *type = GFX_FW_TYPE_CP_MEC_ME2; + break; + case AMDGPU_UCODE_ID_RLC_G: + *type = GFX_FW_TYPE_RLC_G; + break; + case AMDGPU_UCODE_ID_SMC: + *type = GFX_FW_TYPE_SMU; + break; + case AMDGPU_UCODE_ID_UVD: + *type = GFX_FW_TYPE_UVD; + break; + case AMDGPU_UCODE_ID_VCE: + *type = GFX_FW_TYPE_VCE; + break; + case AMDGPU_UCODE_ID_UVD1: + *type = GFX_FW_TYPE_UVD1; + break; + case AMDGPU_UCODE_ID_MAXIMUM: + default: + return -EINVAL; + } + + return 0; +} + +static int psp_v11_0_init_microcode(struct psp_context *psp) +{ + struct amdgpu_device *adev = psp->adev; + const char *chip_name; + char fw_name[30]; + int err = 0; + const struct psp_firmware_header_v1_0 *hdr; + + DRM_DEBUG("\n"); + + switch (adev->asic_type) { + case CHIP_VEGA20: + chip_name = "vega20"; + break; + default: + BUG(); + } + + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_sos.bin", chip_name); + err = request_firmware(&adev->psp.sos_fw, fw_name, adev->dev); + if (err) + goto out; + + err = amdgpu_ucode_validate(adev->psp.sos_fw); + if (err) + goto out; + + hdr = (const struct psp_firmware_header_v1_0 *)adev->psp.sos_fw->data; + adev->psp.sos_fw_version = le32_to_cpu(hdr->header.ucode_version); + adev->psp.sos_feature_version = le32_to_cpu(hdr->ucode_feature_version); + adev->psp.sos_bin_size = le32_to_cpu(hdr->sos_size_bytes); + adev->psp.sys_bin_size = le32_to_cpu(hdr->header.ucode_size_bytes) - + le32_to_cpu(hdr->sos_size_bytes); + adev->psp.sys_start_addr = (uint8_t *)hdr + + le32_to_cpu(hdr->header.ucode_array_offset_bytes); + adev->psp.sos_start_addr = (uint8_t *)adev->psp.sys_start_addr + + le32_to_cpu(hdr->sos_offset_bytes); + return 0; +out: + if (err) { + dev_err(adev->dev, + "psp v11.0: Failed to load firmware \"%s\"\n", + fw_name); + release_firmware(adev->psp.sos_fw); + adev->psp.sos_fw = NULL; + } + + return err; +} + +static int psp_v11_0_bootloader_load_sysdrv(struct psp_context *psp) +{ + int ret; + uint32_t psp_gfxdrv_command_reg = 0; + struct amdgpu_device *adev = psp->adev; + uint32_t sol_reg; + + /* Check sOS sign of life register to confirm sys driver and sOS + * are already been loaded. + */ + sol_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81); + if (sol_reg) + return 0; + + /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), + 0x80000000, 0x80000000, false); + if (ret) + return ret; + + memset(psp->fw_pri_buf, 0, PSP_1_MEG); + + /* Copy PSP System Driver binary to memory */ + memcpy(psp->fw_pri_buf, psp->sys_start_addr, psp->sys_bin_size); + + /* Provide the sys driver to bootrom */ + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36, + (uint32_t)(psp->fw_pri_mc_addr >> 20)); + psp_gfxdrv_command_reg = 1 << 16; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_35, + psp_gfxdrv_command_reg); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), + 0x80000000, 0x80000000, false); + + return ret; +} + +static int psp_v11_0_bootloader_load_sos(struct psp_context *psp) +{ + int ret; + unsigned int psp_gfxdrv_command_reg = 0; + struct amdgpu_device *adev = psp->adev; + uint32_t sol_reg; + + /* Check sOS sign of life register to confirm sys driver and sOS + * are already been loaded. + */ + sol_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81); + if (sol_reg) + return 0; + + /* Wait for bootloader to signify that is ready having bit 31 of C2PMSG_35 set to 1 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_35), + 0x80000000, 0x80000000, false); + if (ret) + return ret; + + memset(psp->fw_pri_buf, 0, PSP_1_MEG); + + /* Copy Secure OS binary to PSP memory */ + memcpy(psp->fw_pri_buf, psp->sos_start_addr, psp->sos_bin_size); + + /* Provide the PSP secure OS to bootrom */ + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_36, + (uint32_t)(psp->fw_pri_mc_addr >> 20)); + psp_gfxdrv_command_reg = 2 << 16; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_35, + psp_gfxdrv_command_reg); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_81), + RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_81), + 0, true); + + return ret; +} + +static int psp_v11_0_prep_cmd_buf(struct amdgpu_firmware_info *ucode, + struct psp_gfx_cmd_resp *cmd) +{ + int ret; + uint64_t fw_mem_mc_addr = ucode->mc_addr; + + memset(cmd, 0, sizeof(struct psp_gfx_cmd_resp)); + + cmd->cmd_id = GFX_CMD_ID_LOAD_IP_FW; + cmd->cmd.cmd_load_ip_fw.fw_phy_addr_lo = lower_32_bits(fw_mem_mc_addr); + cmd->cmd.cmd_load_ip_fw.fw_phy_addr_hi = upper_32_bits(fw_mem_mc_addr); + cmd->cmd.cmd_load_ip_fw.fw_size = ucode->ucode_size; + + ret = psp_v11_0_get_fw_type(ucode, &cmd->cmd.cmd_load_ip_fw.fw_type); + if (ret) + DRM_ERROR("Unknown firmware type\n"); + + return ret; +} + +static int psp_v11_0_ring_init(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring; + struct amdgpu_device *adev = psp->adev; + + ring = &psp->km_ring; + + ring->ring_type = ring_type; + + /* allocate 4k Page of Local Frame Buffer memory for ring */ + ring->ring_size = 0x1000; + ret = amdgpu_bo_create_kernel(adev, ring->ring_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->firmware.rbuf, + &ring->ring_mem_mc_addr, + (void **)&ring->ring_mem); + if (ret) { + ring->ring_size = 0; + return ret; + } + + return 0; +} + +static int psp_v11_0_ring_create(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + unsigned int psp_ring_reg = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + /* Write low address of the ring to C2PMSG_69 */ + psp_ring_reg = lower_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_69, psp_ring_reg); + /* Write high address of the ring to C2PMSG_70 */ + psp_ring_reg = upper_32_bits(ring->ring_mem_mc_addr); + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_70, psp_ring_reg); + /* Write size of ring to C2PMSG_71 */ + psp_ring_reg = ring->ring_size; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_71, psp_ring_reg); + /* Write the ring initialization command to C2PMSG_64 */ + psp_ring_reg = ring_type; + psp_ring_reg = psp_ring_reg << 16; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, psp_ring_reg); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + + /* Wait for response flag (bit 31) in C2PMSG_64 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x8000FFFF, false); + + return ret; +} + +static int psp_v11_0_ring_stop(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + struct amdgpu_device *adev = psp->adev; + + /* Write the ring destroy command to C2PMSG_64 */ + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_64, GFX_CTRL_CMD_ID_DESTROY_RINGS); + + /* there might be handshake issue with hardware which needs delay */ + mdelay(20); + + /* Wait for response flag (bit 31) in C2PMSG_64 */ + ret = psp_wait_for(psp, SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64), + 0x80000000, 0x80000000, false); + + return ret; +} + +static int psp_v11_0_ring_destroy(struct psp_context *psp, + enum psp_ring_type ring_type) +{ + int ret = 0; + struct psp_ring *ring = &psp->km_ring; + struct amdgpu_device *adev = psp->adev; + + ret = psp_v11_0_ring_stop(psp, ring_type); + if (ret) + DRM_ERROR("Fail to stop psp ring\n"); + + amdgpu_bo_free_kernel(&adev->firmware.rbuf, + &ring->ring_mem_mc_addr, + (void **)&ring->ring_mem); + + return ret; +} + +static int psp_v11_0_cmd_submit(struct psp_context *psp, + struct amdgpu_firmware_info *ucode, + uint64_t cmd_buf_mc_addr, uint64_t fence_mc_addr, + int index) +{ + unsigned int psp_write_ptr_reg = 0; + struct psp_gfx_rb_frame *write_frame = psp->km_ring.ring_mem; + struct psp_ring *ring = &psp->km_ring; + struct psp_gfx_rb_frame *ring_buffer_start = ring->ring_mem; + struct psp_gfx_rb_frame *ring_buffer_end = ring_buffer_start + + ring->ring_size / sizeof(struct psp_gfx_rb_frame) - 1; + struct amdgpu_device *adev = psp->adev; + uint32_t ring_size_dw = ring->ring_size / 4; + uint32_t rb_frame_size_dw = sizeof(struct psp_gfx_rb_frame) / 4; + + /* KM (GPCOM) prepare write pointer */ + psp_write_ptr_reg = RREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67); + + /* Update KM RB frame pointer to new frame */ + /* write_frame ptr increments by size of rb_frame in bytes */ + /* psp_write_ptr_reg increments by size of rb_frame in DWORDs */ + if ((psp_write_ptr_reg % ring_size_dw) == 0) + write_frame = ring_buffer_start; + else + write_frame = ring_buffer_start + (psp_write_ptr_reg / rb_frame_size_dw); + /* Check invalid write_frame ptr address */ + if ((write_frame < ring_buffer_start) || (ring_buffer_end < write_frame)) { + DRM_ERROR("ring_buffer_start = %p; ring_buffer_end = %p; write_frame = %p\n", + ring_buffer_start, ring_buffer_end, write_frame); + DRM_ERROR("write_frame is pointing to address out of bounds\n"); + return -EINVAL; + } + + /* Initialize KM RB frame */ + memset(write_frame, 0, sizeof(struct psp_gfx_rb_frame)); + + /* Update KM RB frame */ + write_frame->cmd_buf_addr_hi = upper_32_bits(cmd_buf_mc_addr); + write_frame->cmd_buf_addr_lo = lower_32_bits(cmd_buf_mc_addr); + write_frame->fence_addr_hi = upper_32_bits(fence_mc_addr); + write_frame->fence_addr_lo = lower_32_bits(fence_mc_addr); + write_frame->fence_value = index; + + /* Update the write Pointer in DWORDs */ + psp_write_ptr_reg = (psp_write_ptr_reg + rb_frame_size_dw) % ring_size_dw; + WREG32_SOC15(MP0, 0, mmMP0_SMN_C2PMSG_67, psp_write_ptr_reg); + + return 0; +} + +static int +psp_v11_0_sram_map(struct amdgpu_device *adev, + unsigned int *sram_offset, unsigned int *sram_addr_reg_offset, + unsigned int *sram_data_reg_offset, + enum AMDGPU_UCODE_ID ucode_id) +{ + int ret = 0; + + switch (ucode_id) { +/* TODO: needs to confirm */ +#if 0 + case AMDGPU_UCODE_ID_SMC: + *sram_offset = 0; + *sram_addr_reg_offset = 0; + *sram_data_reg_offset = 0; + break; +#endif + + case AMDGPU_UCODE_ID_CP_CE: + *sram_offset = 0x0; + *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_CE_UCODE_ADDR); + *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_CE_UCODE_DATA); + break; + + case AMDGPU_UCODE_ID_CP_PFP: + *sram_offset = 0x0; + *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_PFP_UCODE_ADDR); + *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_PFP_UCODE_DATA); + break; + + case AMDGPU_UCODE_ID_CP_ME: + *sram_offset = 0x0; + *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_ME_UCODE_ADDR); + *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_ME_UCODE_DATA); + break; + + case AMDGPU_UCODE_ID_CP_MEC1: + *sram_offset = 0x10000; + *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME1_UCODE_ADDR); + *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_MEC_ME1_UCODE_DATA); + break; + + case AMDGPU_UCODE_ID_CP_MEC2: + *sram_offset = 0x10000; + *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_MEC2_UCODE_ADDR); + *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmCP_HYP_MEC2_UCODE_DATA); + break; + + case AMDGPU_UCODE_ID_RLC_G: + *sram_offset = 0x2000; + *sram_addr_reg_offset = SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UCODE_ADDR); + *sram_data_reg_offset = SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_UCODE_DATA); + break; + + case AMDGPU_UCODE_ID_SDMA0: + *sram_offset = 0x0; + *sram_addr_reg_offset = SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_UCODE_ADDR); + *sram_data_reg_offset = SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_UCODE_DATA); + break; + +/* TODO: needs to confirm */ +#if 0 + case AMDGPU_UCODE_ID_SDMA1: + *sram_offset = ; + *sram_addr_reg_offset = ; + break; + + case AMDGPU_UCODE_ID_UVD: + *sram_offset = ; + *sram_addr_reg_offset = ; + break; + + case AMDGPU_UCODE_ID_VCE: + *sram_offset = ; + *sram_addr_reg_offset = ; + break; +#endif + + case AMDGPU_UCODE_ID_MAXIMUM: + default: + ret = -EINVAL; + break; + } + + return ret; +} + +static bool psp_v11_0_compare_sram_data(struct psp_context *psp, + struct amdgpu_firmware_info *ucode, + enum AMDGPU_UCODE_ID ucode_type) +{ + int err = 0; + unsigned int fw_sram_reg_val = 0; + unsigned int fw_sram_addr_reg_offset = 0; + unsigned int fw_sram_data_reg_offset = 0; + unsigned int ucode_size; + uint32_t *ucode_mem = NULL; + struct amdgpu_device *adev = psp->adev; + + err = psp_v11_0_sram_map(adev, &fw_sram_reg_val, &fw_sram_addr_reg_offset, + &fw_sram_data_reg_offset, ucode_type); + if (err) + return false; + + WREG32(fw_sram_addr_reg_offset, fw_sram_reg_val); + + ucode_size = ucode->ucode_size; + ucode_mem = (uint32_t *)ucode->kaddr; + while (ucode_size) { + fw_sram_reg_val = RREG32(fw_sram_data_reg_offset); + + if (*ucode_mem != fw_sram_reg_val) + return false; + + ucode_mem++; + /* 4 bytes */ + ucode_size -= 4; + } + + return true; +} + +static int psp_v11_0_mode1_reset(struct psp_context *psp) +{ + int ret; + uint32_t offset; + struct amdgpu_device *adev = psp->adev; + + offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_64); + + ret = psp_wait_for(psp, offset, 0x80000000, 0x8000FFFF, false); + + if (ret) { + DRM_INFO("psp is not working correctly before mode1 reset!\n"); + return -EINVAL; + } + + /*send the mode 1 reset command*/ + WREG32(offset, GFX_CTRL_CMD_ID_MODE1_RST); + + mdelay(1000); + + offset = SOC15_REG_OFFSET(MP0, 0, mmMP0_SMN_C2PMSG_33); + + ret = psp_wait_for(psp, offset, 0x80000000, 0x80000000, false); + + if (ret) { + DRM_INFO("psp mode 1 reset failed!\n"); + return -EINVAL; + } + + DRM_INFO("psp mode1 reset succeed \n"); + + return 0; +} + +/* TODO: Fill in follow functions once PSP firmware interface for XGMI is ready. + * For now, return success and hack the hive_id so high level code can + * start testing + */ +static int psp_v11_0_xgmi_get_topology_info(struct psp_context *psp, + int number_devices, struct psp_xgmi_topology_info *topology) +{ + return 0; +} + +static int psp_v11_0_xgmi_set_topology_info(struct psp_context *psp, + int number_devices, struct psp_xgmi_topology_info *topology) +{ + return 0; +} + +static u64 psp_v11_0_xgmi_get_hive_id(struct psp_context *psp) +{ + u64 hive_id = 0; + + /* Remove me when we can get correct hive_id through PSP */ + if (psp->adev->gmc.xgmi.num_physical_nodes) + hive_id = 0x123456789abcdef; + + return hive_id; +} + +static const struct psp_funcs psp_v11_0_funcs = { + .init_microcode = psp_v11_0_init_microcode, + .bootloader_load_sysdrv = psp_v11_0_bootloader_load_sysdrv, + .bootloader_load_sos = psp_v11_0_bootloader_load_sos, + .prep_cmd_buf = psp_v11_0_prep_cmd_buf, + .ring_init = psp_v11_0_ring_init, + .ring_create = psp_v11_0_ring_create, + .ring_stop = psp_v11_0_ring_stop, + .ring_destroy = psp_v11_0_ring_destroy, + .cmd_submit = psp_v11_0_cmd_submit, + .compare_sram_data = psp_v11_0_compare_sram_data, + .mode1_reset = psp_v11_0_mode1_reset, + .xgmi_get_topology_info = psp_v11_0_xgmi_get_topology_info, + .xgmi_set_topology_info = psp_v11_0_xgmi_set_topology_info, + .xgmi_get_hive_id = psp_v11_0_xgmi_get_hive_id, +}; + +void psp_v11_0_set_psp_funcs(struct psp_context *psp) +{ + psp->funcs = &psp_v11_0_funcs; +} diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v11_0.h b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.h new file mode 100644 index 000000000000..082c16c887bf --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/psp_v11_0.h @@ -0,0 +1,30 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef __PSP_V11_0_H__ +#define __PSP_V11_0_H__ + +#include "amdgpu_psp.h" + +void psp_v11_0_set_psp_funcs(struct psp_context *psp); + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index 727071fee6f6..e1ebf770c303 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -41,8 +41,6 @@ MODULE_FIRMWARE("amdgpu/vega10_sos.bin"); MODULE_FIRMWARE("amdgpu/vega10_asd.bin"); MODULE_FIRMWARE("amdgpu/vega12_sos.bin"); MODULE_FIRMWARE("amdgpu/vega12_asd.bin"); -MODULE_FIRMWARE("amdgpu/vega20_sos.bin"); -MODULE_FIRMWARE("amdgpu/vega20_asd.bin"); #define smnMP1_FIRMWARE_FLAGS 0x3010028 diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c index c7190c39c4f5..2d4770e173dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v2_4.c @@ -44,6 +44,8 @@ #include "iceland_sdma_pkt_open.h" +#include "ivsrcid/ivsrcid_vislands30.h" + static void sdma_v2_4_set_ring_funcs(struct amdgpu_device *adev); static void sdma_v2_4_set_buffer_funcs(struct amdgpu_device *adev); static void sdma_v2_4_set_vm_pte_funcs(struct amdgpu_device *adev); @@ -202,8 +204,7 @@ static uint64_t sdma_v2_4_ring_get_rptr(struct amdgpu_ring *ring) static uint64_t sdma_v2_4_ring_get_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1; - u32 wptr = RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me]) >> 2; + u32 wptr = RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me]) >> 2; return wptr; } @@ -218,9 +219,8 @@ static uint64_t sdma_v2_4_ring_get_wptr(struct amdgpu_ring *ring) static void sdma_v2_4_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1; - WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me], lower_32_bits(ring->wptr) << 2); + WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], lower_32_bits(ring->wptr) << 2); } static void sdma_v2_4_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) @@ -273,7 +273,7 @@ static void sdma_v2_4_ring_emit_hdp_flush(struct amdgpu_ring *ring) { u32 ref_and_mask = 0; - if (ring == &ring->adev->sdma.instance[0].ring) + if (ring->me == 0) ref_and_mask = REG_SET_FIELD(ref_and_mask, GPU_HDP_FLUSH_DONE, SDMA0, 1); else ref_and_mask = REG_SET_FIELD(ref_and_mask, GPU_HDP_FLUSH_DONE, SDMA1, 1); @@ -504,41 +504,6 @@ static int sdma_v2_4_rlc_resume(struct amdgpu_device *adev) return 0; } -/** - * sdma_v2_4_load_microcode - load the sDMA ME ucode - * - * @adev: amdgpu_device pointer - * - * Loads the sDMA0/1 ucode. - * Returns 0 for success, -EINVAL if the ucode is not available. - */ -static int sdma_v2_4_load_microcode(struct amdgpu_device *adev) -{ - const struct sdma_firmware_header_v1_0 *hdr; - const __le32 *fw_data; - u32 fw_size; - int i, j; - - /* halt the MEs */ - sdma_v2_4_enable(adev, false); - - for (i = 0; i < adev->sdma.num_instances; i++) { - if (!adev->sdma.instance[i].fw) - return -EINVAL; - hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma.instance[i].fw->data; - amdgpu_ucode_print_sdma_hdr(&hdr->header); - fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; - fw_data = (const __le32 *) - (adev->sdma.instance[i].fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - WREG32(mmSDMA0_UCODE_ADDR + sdma_offsets[i], 0); - for (j = 0; j < fw_size; j++) - WREG32(mmSDMA0_UCODE_DATA + sdma_offsets[i], le32_to_cpup(fw_data++)); - WREG32(mmSDMA0_UCODE_ADDR + sdma_offsets[i], adev->sdma.instance[i].fw_version); - } - - return 0; -} /** * sdma_v2_4_start - setup and start the async dma engines @@ -552,13 +517,6 @@ static int sdma_v2_4_start(struct amdgpu_device *adev) { int r; - - if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { - r = sdma_v2_4_load_microcode(adev); - if (r) - return r; - } - /* halt the engine before programing */ sdma_v2_4_enable(adev, false); @@ -898,19 +856,19 @@ static int sdma_v2_4_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* SDMA trap event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 224, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP, &adev->sdma.trap_irq); if (r) return r; /* SDMA Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 241, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 241, &adev->sdma.illegal_inst_irq); if (r) return r; /* SDMA Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 247, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_SRBM_WRITE, &adev->sdma.illegal_inst_irq); if (r) return r; @@ -1213,8 +1171,10 @@ static void sdma_v2_4_set_ring_funcs(struct amdgpu_device *adev) { int i; - for (i = 0; i < adev->sdma.num_instances; i++) + for (i = 0; i < adev->sdma.num_instances; i++) { adev->sdma.instance[i].ring.funcs = &sdma_v2_4_ring_funcs; + adev->sdma.instance[i].ring.me = i; + } } static const struct amdgpu_irq_src_funcs sdma_v2_4_trap_irq_funcs = { @@ -1294,10 +1254,8 @@ static const struct amdgpu_buffer_funcs sdma_v2_4_buffer_funcs = { static void sdma_v2_4_set_buffer_funcs(struct amdgpu_device *adev) { - if (adev->mman.buffer_funcs == NULL) { - adev->mman.buffer_funcs = &sdma_v2_4_buffer_funcs; - adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; - } + adev->mman.buffer_funcs = &sdma_v2_4_buffer_funcs; + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; } static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = { @@ -1310,16 +1268,16 @@ static const struct amdgpu_vm_pte_funcs sdma_v2_4_vm_pte_funcs = { static void sdma_v2_4_set_vm_pte_funcs(struct amdgpu_device *adev) { + struct drm_gpu_scheduler *sched; unsigned i; - if (adev->vm_manager.vm_pte_funcs == NULL) { - adev->vm_manager.vm_pte_funcs = &sdma_v2_4_vm_pte_funcs; - for (i = 0; i < adev->sdma.num_instances; i++) - adev->vm_manager.vm_pte_rings[i] = - &adev->sdma.instance[i].ring; - - adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances; + adev->vm_manager.vm_pte_funcs = &sdma_v2_4_vm_pte_funcs; + for (i = 0; i < adev->sdma.num_instances; i++) { + sched = &adev->sdma.instance[i].ring.sched; + adev->vm_manager.vm_pte_rqs[i] = + &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; } + adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v2_4_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index aa9ab299fd32..6fb3edaba0ec 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -44,6 +44,8 @@ #include "tonga_sdma_pkt_open.h" +#include "ivsrcid/ivsrcid_vislands30.h" + static void sdma_v3_0_set_ring_funcs(struct amdgpu_device *adev); static void sdma_v3_0_set_buffer_funcs(struct amdgpu_device *adev); static void sdma_v3_0_set_vm_pte_funcs(struct amdgpu_device *adev); @@ -316,14 +318,13 @@ static int sdma_v3_0_init_microcode(struct amdgpu_device *adev) if (adev->sdma.instance[i].feature_version >= 20) adev->sdma.instance[i].burst_nop = true; - if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) { - info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; - info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i; - info->fw = adev->sdma.instance[i].fw; - header = (const struct common_firmware_header *)info->fw->data; - adev->firmware.fw_size += - ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); - } + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_SDMA0 + i]; + info->ucode_id = AMDGPU_UCODE_ID_SDMA0 + i; + info->fw = adev->sdma.instance[i].fw; + header = (const struct common_firmware_header *)info->fw->data; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + } out: if (err) { @@ -365,9 +366,7 @@ static uint64_t sdma_v3_0_ring_get_wptr(struct amdgpu_ring *ring) /* XXX check if swapping is necessary on BE */ wptr = ring->adev->wb.wb[ring->wptr_offs] >> 2; } else { - int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1; - - wptr = RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me]) >> 2; + wptr = RREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me]) >> 2; } return wptr; @@ -394,9 +393,7 @@ static void sdma_v3_0_ring_set_wptr(struct amdgpu_ring *ring) WRITE_ONCE(*wb, (lower_32_bits(ring->wptr) << 2)); } else { - int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1; - - WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[me], lower_32_bits(ring->wptr) << 2); + WREG32(mmSDMA0_GFX_RB_WPTR + sdma_offsets[ring->me], lower_32_bits(ring->wptr) << 2); } } @@ -450,7 +447,7 @@ static void sdma_v3_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) { u32 ref_and_mask = 0; - if (ring == &ring->adev->sdma.instance[0].ring) + if (ring->me == 0) ref_and_mask = REG_SET_FIELD(ref_and_mask, GPU_HDP_FLUSH_DONE, SDMA0, 1); else ref_and_mask = REG_SET_FIELD(ref_and_mask, GPU_HDP_FLUSH_DONE, SDMA1, 1); @@ -780,42 +777,6 @@ static int sdma_v3_0_rlc_resume(struct amdgpu_device *adev) } /** - * sdma_v3_0_load_microcode - load the sDMA ME ucode - * - * @adev: amdgpu_device pointer - * - * Loads the sDMA0/1 ucode. - * Returns 0 for success, -EINVAL if the ucode is not available. - */ -static int sdma_v3_0_load_microcode(struct amdgpu_device *adev) -{ - const struct sdma_firmware_header_v1_0 *hdr; - const __le32 *fw_data; - u32 fw_size; - int i, j; - - /* halt the MEs */ - sdma_v3_0_enable(adev, false); - - for (i = 0; i < adev->sdma.num_instances; i++) { - if (!adev->sdma.instance[i].fw) - return -EINVAL; - hdr = (const struct sdma_firmware_header_v1_0 *)adev->sdma.instance[i].fw->data; - amdgpu_ucode_print_sdma_hdr(&hdr->header); - fw_size = le32_to_cpu(hdr->header.ucode_size_bytes) / 4; - fw_data = (const __le32 *) - (adev->sdma.instance[i].fw->data + - le32_to_cpu(hdr->header.ucode_array_offset_bytes)); - WREG32(mmSDMA0_UCODE_ADDR + sdma_offsets[i], 0); - for (j = 0; j < fw_size; j++) - WREG32(mmSDMA0_UCODE_DATA + sdma_offsets[i], le32_to_cpup(fw_data++)); - WREG32(mmSDMA0_UCODE_ADDR + sdma_offsets[i], adev->sdma.instance[i].fw_version); - } - - return 0; -} - -/** * sdma_v3_0_start - setup and start the async dma engines * * @adev: amdgpu_device pointer @@ -827,12 +788,6 @@ static int sdma_v3_0_start(struct amdgpu_device *adev) { int r; - if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { - r = sdma_v3_0_load_microcode(adev); - if (r) - return r; - } - /* disable sdma engine before programing it */ sdma_v3_0_ctx_switch_enable(adev, false); sdma_v3_0_enable(adev, false); @@ -1179,19 +1134,19 @@ static int sdma_v3_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* SDMA trap event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 224, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_TRAP, &adev->sdma.trap_irq); if (r) return r; /* SDMA Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 241, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 241, &adev->sdma.illegal_inst_irq); if (r) return r; /* SDMA Privileged inst */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 247, + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_SDMA_SRBM_WRITE, &adev->sdma.illegal_inst_irq); if (r) return r; @@ -1655,8 +1610,10 @@ static void sdma_v3_0_set_ring_funcs(struct amdgpu_device *adev) { int i; - for (i = 0; i < adev->sdma.num_instances; i++) + for (i = 0; i < adev->sdma.num_instances; i++) { adev->sdma.instance[i].ring.funcs = &sdma_v3_0_ring_funcs; + adev->sdma.instance[i].ring.me = i; + } } static const struct amdgpu_irq_src_funcs sdma_v3_0_trap_irq_funcs = { @@ -1736,10 +1693,8 @@ static const struct amdgpu_buffer_funcs sdma_v3_0_buffer_funcs = { static void sdma_v3_0_set_buffer_funcs(struct amdgpu_device *adev) { - if (adev->mman.buffer_funcs == NULL) { - adev->mman.buffer_funcs = &sdma_v3_0_buffer_funcs; - adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; - } + adev->mman.buffer_funcs = &sdma_v3_0_buffer_funcs; + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; } static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = { @@ -1752,16 +1707,16 @@ static const struct amdgpu_vm_pte_funcs sdma_v3_0_vm_pte_funcs = { static void sdma_v3_0_set_vm_pte_funcs(struct amdgpu_device *adev) { + struct drm_gpu_scheduler *sched; unsigned i; - if (adev->vm_manager.vm_pte_funcs == NULL) { - adev->vm_manager.vm_pte_funcs = &sdma_v3_0_vm_pte_funcs; - for (i = 0; i < adev->sdma.num_instances; i++) - adev->vm_manager.vm_pte_rings[i] = - &adev->sdma.instance[i].ring; - - adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances; + adev->vm_manager.vm_pte_funcs = &sdma_v3_0_vm_pte_funcs; + for (i = 0; i < adev->sdma.num_instances; i++) { + sched = &adev->sdma.instance[i].ring.sched; + adev->vm_manager.vm_pte_rqs[i] = + &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; } + adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v3_0_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index ca53b3fba422..04fa3d972636 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -27,10 +27,10 @@ #include "amdgpu_ucode.h" #include "amdgpu_trace.h" -#include "sdma0/sdma0_4_0_offset.h" -#include "sdma0/sdma0_4_0_sh_mask.h" -#include "sdma1/sdma1_4_0_offset.h" -#include "sdma1/sdma1_4_0_sh_mask.h" +#include "sdma0/sdma0_4_2_offset.h" +#include "sdma0/sdma0_4_2_sh_mask.h" +#include "sdma1/sdma1_4_2_offset.h" +#include "sdma1/sdma1_4_2_sh_mask.h" #include "hdp/hdp_4_0_offset.h" #include "sdma0/sdma0_4_1_default.h" @@ -38,6 +38,9 @@ #include "soc15.h" #include "vega10_sdma_pkt_open.h" +#include "ivsrcid/sdma0/irqsrcs_sdma0_4_0.h" +#include "ivsrcid/sdma1/irqsrcs_sdma1_4_0.h" + MODULE_FIRMWARE("amdgpu/vega10_sdma.bin"); MODULE_FIRMWARE("amdgpu/vega10_sdma1.bin"); MODULE_FIRMWARE("amdgpu/vega12_sdma.bin"); @@ -45,6 +48,8 @@ MODULE_FIRMWARE("amdgpu/vega12_sdma1.bin"); MODULE_FIRMWARE("amdgpu/vega20_sdma.bin"); MODULE_FIRMWARE("amdgpu/vega20_sdma1.bin"); MODULE_FIRMWARE("amdgpu/raven_sdma.bin"); +MODULE_FIRMWARE("amdgpu/picasso_sdma.bin"); +MODULE_FIRMWARE("amdgpu/raven2_sdma.bin"); #define SDMA0_POWER_CNTL__ON_OFF_CONDITION_HOLD_TIME_MASK 0x000000F8L #define SDMA0_POWER_CNTL__ON_OFF_STATUS_DURATION_TIME_MASK 0xFC000000L @@ -67,6 +72,7 @@ static const struct soc15_reg_golden golden_settings_sdma_4[] = { SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_IB_CNTL, 0x800f0100, 0x00000100), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_WATERMK, 0xfc000000, 0x00000000), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831f07), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CLK_CTRL, 0xffffffff, 0x3f000100), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GFX_IB_CNTL, 0x800f0100, 0x00000100), @@ -78,7 +84,8 @@ static const struct soc15_reg_golden golden_settings_sdma_4[] = { SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC0_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC1_IB_CNTL, 0x800f0100, 0x00000100), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC1_RB_WPTR_POLL_CNTL, 0x0000fff0, 0x00403000), - SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_PAGE, 0x000003ff, 0x000003c0) + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_PAGE, 0x000003ff, 0x000003c0), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_WATERMK, 0xfc000000, 0x00000000) }; static const struct soc15_reg_golden golden_settings_sdma_vg10[] = { @@ -95,8 +102,7 @@ static const struct soc15_reg_golden golden_settings_sdma_vg12[] = { SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG_READ, 0x0018773f, 0x00104001) }; -static const struct soc15_reg_golden golden_settings_sdma_4_1[] = -{ +static const struct soc15_reg_golden golden_settings_sdma_4_1[] = { SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CHICKEN_BITS, 0xfe931f07, 0x02831d07), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CLK_CTRL, 0xffffffff, 0x3f000100), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GFX_IB_CNTL, 0x800f0111, 0x00000100), @@ -106,29 +112,73 @@ static const struct soc15_reg_golden golden_settings_sdma_4_1[] = SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_IB_CNTL, 0x800f0111, 0x00000100), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), - SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0) + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_WATERMK, 0xfc000000, 0x00000000) }; -static const struct soc15_reg_golden golden_settings_sdma_4_2[] = +static const struct soc15_reg_golden golden_settings_sdma0_4_2_init[] = { + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff0, 0x00403000), +}; + +static const struct soc15_reg_golden golden_settings_sdma0_4_2[] = { SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CHICKEN_BITS, 0xfe931f07, 0x02831d07), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CLK_CTRL, 0xffffffff, 0x3f000100), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG, 0x0000773f, 0x00004002), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x0000773f, 0x00004002), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GFX_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_PAGE_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_PAGE_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RD_BURST_CNTL, 0x0000000f, 0x00000003), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC0_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff0, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC2_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC2_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC3_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC3_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC4_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC4_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC5_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC5_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC6_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC6_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC7_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC7_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_WATERMK, 0xFE000000, 0x00000000), +}; + +static const struct soc15_reg_golden golden_settings_sdma1_4_2[] = { SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831d07), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CLK_CTRL, 0xffffffff, 0x3f000100), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG, 0x0000773f, 0x00004002), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG_READ, 0x0000773f, 0x00004002), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GFX_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GFX_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_PAGE_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_PAGE_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), - SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RD_BURST_CNTL, 0x0000000f, 0x00000003), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC0_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff0, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC1_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC1_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), - SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_PAGE, 0x000003ff, 0x000003c0) + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC2_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC2_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC3_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC3_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC4_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC4_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC5_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC5_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC6_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC6_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC7_RB_RPTR_ADDR_LO, 0xfffffffd, 0x00000001), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC7_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_PAGE, 0x000003ff, 0x000003c0), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_WATERMK, 0xFE000000, 0x00000000), }; static const struct soc15_reg_golden golden_settings_sdma_rv1[] = @@ -137,6 +187,12 @@ static const struct soc15_reg_golden golden_settings_sdma_rv1[] = SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x0018773f, 0x00000002) }; +static const struct soc15_reg_golden golden_settings_sdma_rv2[] = +{ + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG, 0x0018773f, 0x00003001), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x0018773f, 0x00003001) +}; + static u32 sdma_v4_0_get_reg_offset(struct amdgpu_device *adev, u32 instance, u32 offset) { @@ -165,16 +221,27 @@ static void sdma_v4_0_init_golden_registers(struct amdgpu_device *adev) break; case CHIP_VEGA20: soc15_program_register_sequence(adev, - golden_settings_sdma_4_2, - ARRAY_SIZE(golden_settings_sdma_4_2)); + golden_settings_sdma0_4_2_init, + ARRAY_SIZE(golden_settings_sdma0_4_2_init)); + soc15_program_register_sequence(adev, + golden_settings_sdma0_4_2, + ARRAY_SIZE(golden_settings_sdma0_4_2)); + soc15_program_register_sequence(adev, + golden_settings_sdma1_4_2, + ARRAY_SIZE(golden_settings_sdma1_4_2)); break; case CHIP_RAVEN: soc15_program_register_sequence(adev, - golden_settings_sdma_4_1, - ARRAY_SIZE(golden_settings_sdma_4_1)); - soc15_program_register_sequence(adev, - golden_settings_sdma_rv1, - ARRAY_SIZE(golden_settings_sdma_rv1)); + golden_settings_sdma_4_1, + ARRAY_SIZE(golden_settings_sdma_4_1)); + if (adev->rev_id >= 8) + soc15_program_register_sequence(adev, + golden_settings_sdma_rv2, + ARRAY_SIZE(golden_settings_sdma_rv2)); + else + soc15_program_register_sequence(adev, + golden_settings_sdma_rv1, + ARRAY_SIZE(golden_settings_sdma_rv1)); break; default: break; @@ -215,7 +282,12 @@ static int sdma_v4_0_init_microcode(struct amdgpu_device *adev) chip_name = "vega20"; break; case CHIP_RAVEN: - chip_name = "raven"; + if (adev->rev_id >= 8) + chip_name = "raven2"; + else if (adev->pdev->device == 0x15d8) + chip_name = "picasso"; + else + chip_name = "raven"; break; default: BUG(); @@ -296,13 +368,12 @@ static uint64_t sdma_v4_0_ring_get_wptr(struct amdgpu_ring *ring) DRM_DEBUG("wptr/doorbell before shift == 0x%016llx\n", wptr); } else { u32 lowbit, highbit; - int me = (ring == &adev->sdma.instance[0].ring) ? 0 : 1; - lowbit = RREG32(sdma_v4_0_get_reg_offset(adev, me, mmSDMA0_GFX_RB_WPTR)) >> 2; - highbit = RREG32(sdma_v4_0_get_reg_offset(adev, me, mmSDMA0_GFX_RB_WPTR_HI)) >> 2; + lowbit = RREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR)) >> 2; + highbit = RREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI)) >> 2; DRM_DEBUG("wptr [%i]high== 0x%08x low==0x%08x\n", - me, highbit, lowbit); + ring->me, highbit, lowbit); wptr = highbit; wptr = wptr << 32; wptr |= lowbit; @@ -339,17 +410,15 @@ static void sdma_v4_0_ring_set_wptr(struct amdgpu_ring *ring) ring->doorbell_index, ring->wptr << 2); WDOORBELL64(ring->doorbell_index, ring->wptr << 2); } else { - int me = (ring == &ring->adev->sdma.instance[0].ring) ? 0 : 1; - DRM_DEBUG("Not using doorbell -- " "mmSDMA%i_GFX_RB_WPTR == 0x%08x " "mmSDMA%i_GFX_RB_WPTR_HI == 0x%08x\n", - me, + ring->me, lower_32_bits(ring->wptr << 2), - me, + ring->me, upper_32_bits(ring->wptr << 2)); - WREG32(sdma_v4_0_get_reg_offset(adev, me, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2)); - WREG32(sdma_v4_0_get_reg_offset(adev, me, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2)); + WREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR), lower_32_bits(ring->wptr << 2)); + WREG32(sdma_v4_0_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI), upper_32_bits(ring->wptr << 2)); } } @@ -430,7 +499,7 @@ static void sdma_v4_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) u32 ref_and_mask = 0; const struct nbio_hdp_flush_reg *nbio_hf_reg = adev->nbio_funcs->hdp_flush_reg; - if (ring == &ring->adev->sdma.instance[0].ring) + if (ring->me == 0) ref_and_mask = nbio_hf_reg->ref_and_mask_sdma0; else ref_and_mask = nbio_hf_reg->ref_and_mask_sdma1; @@ -751,7 +820,7 @@ sdma_v4_1_update_power_gating(struct amdgpu_device *adev, bool enable) uint32_t def, data; if (enable && (adev->pg_flags & AMD_PG_SUPPORT_SDMA)) { - /* disable idle interrupt */ + /* enable idle interrupt */ def = data = RREG32(SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_CNTL)); data |= SDMA0_CNTL__CTXEMPTY_INT_ENABLE_MASK; @@ -1228,13 +1297,13 @@ static int sdma_v4_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* SDMA trap event */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_SDMA0, 224, + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_SDMA0, SDMA0_4_0__SRCID__SDMA_TRAP, &adev->sdma.trap_irq); if (r) return r; /* SDMA trap event */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_SDMA1, 224, + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_SDMA1, SDMA1_4_0__SRCID__SDMA_TRAP, &adev->sdma.trap_irq); if (r) return r; @@ -1253,9 +1322,15 @@ static int sdma_v4_0_sw_init(void *handle) DRM_INFO("use_doorbell being set to: [%s]\n", ring->use_doorbell?"true":"false"); - ring->doorbell_index = (i == 0) ? - (AMDGPU_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset - : (AMDGPU_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset + if (adev->asic_type == CHIP_VEGA10) + ring->doorbell_index = (i == 0) ? + (AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset + : (AMDGPU_VEGA10_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset + else + ring->doorbell_index = (i == 0) ? + (AMDGPU_DOORBELL64_sDMA_ENGINE0 << 1) //get DWORD offset + : (AMDGPU_DOORBELL64_sDMA_ENGINE1 << 1); // get DWORD offset + sprintf(ring->name, "sdma%d", i); r = amdgpu_ring_init(adev, ring, 1024, @@ -1291,6 +1366,9 @@ static int sdma_v4_0_hw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (adev->asic_type == CHIP_RAVEN && adev->powerplay.pp_funcs->set_powergating_by_smu) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, false); + sdma_v4_0_init_golden_registers(adev); r = sdma_v4_0_start(adev); @@ -1308,6 +1386,9 @@ static int sdma_v4_0_hw_fini(void *handle) sdma_v4_0_ctx_switch_enable(adev, false); sdma_v4_0_enable(adev, false); + if (adev->asic_type == CHIP_RAVEN && adev->powerplay.pp_funcs->set_powergating_by_smu) + amdgpu_dpm_set_powergating_by_smu(adev, AMD_IP_BLOCK_TYPE_SDMA, true); + return 0; } @@ -1651,8 +1732,10 @@ static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev) { int i; - for (i = 0; i < adev->sdma.num_instances; i++) + for (i = 0; i < adev->sdma.num_instances; i++) { adev->sdma.instance[i].ring.funcs = &sdma_v4_0_ring_funcs; + adev->sdma.instance[i].ring.me = i; + } } static const struct amdgpu_irq_src_funcs sdma_v4_0_trap_irq_funcs = { @@ -1732,10 +1815,8 @@ static const struct amdgpu_buffer_funcs sdma_v4_0_buffer_funcs = { static void sdma_v4_0_set_buffer_funcs(struct amdgpu_device *adev) { - if (adev->mman.buffer_funcs == NULL) { - adev->mman.buffer_funcs = &sdma_v4_0_buffer_funcs; - adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; - } + adev->mman.buffer_funcs = &sdma_v4_0_buffer_funcs; + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; } static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = { @@ -1748,16 +1829,16 @@ static const struct amdgpu_vm_pte_funcs sdma_v4_0_vm_pte_funcs = { static void sdma_v4_0_set_vm_pte_funcs(struct amdgpu_device *adev) { + struct drm_gpu_scheduler *sched; unsigned i; - if (adev->vm_manager.vm_pte_funcs == NULL) { - adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs; - for (i = 0; i < adev->sdma.num_instances; i++) - adev->vm_manager.vm_pte_rings[i] = - &adev->sdma.instance[i].ring; - - adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances; + adev->vm_manager.vm_pte_funcs = &sdma_v4_0_vm_pte_funcs; + for (i = 0; i < adev->sdma.num_instances; i++) { + sched = &adev->sdma.instance[i].ring.sched; + adev->vm_manager.vm_pte_rqs[i] = + &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; } + adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; } const struct amdgpu_ip_block_version sdma_v4_0_ip_block = { diff --git a/drivers/gpu/drm/amd/amdgpu/si.c b/drivers/gpu/drm/amd/amdgpu/si.c index c364ef94cc36..f8408f88cd37 100644 --- a/drivers/gpu/drm/amd/amdgpu/si.c +++ b/drivers/gpu/drm/amd/amdgpu/si.c @@ -2057,13 +2057,13 @@ int si_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &si_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v6_0_ip_block); amdgpu_device_ip_block_add(adev, &si_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v6_0_ip_block); + amdgpu_device_ip_block_add(adev, &si_dma_ip_block); amdgpu_device_ip_block_add(adev, &si_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); else amdgpu_device_ip_block_add(adev, &dce_v6_0_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v6_0_ip_block); - amdgpu_device_ip_block_add(adev, &si_dma_ip_block); /* amdgpu_device_ip_block_add(adev, &uvd_v3_1_ip_block); */ /* amdgpu_device_ip_block_add(adev, &vce_v1_0_ip_block); */ break; @@ -2071,13 +2071,14 @@ int si_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &si_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v6_0_ip_block); amdgpu_device_ip_block_add(adev, &si_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v6_0_ip_block); + amdgpu_device_ip_block_add(adev, &si_dma_ip_block); amdgpu_device_ip_block_add(adev, &si_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); else amdgpu_device_ip_block_add(adev, &dce_v6_4_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v6_0_ip_block); - amdgpu_device_ip_block_add(adev, &si_dma_ip_block); + /* amdgpu_device_ip_block_add(adev, &uvd_v3_1_ip_block); */ /* amdgpu_device_ip_block_add(adev, &vce_v1_0_ip_block); */ break; @@ -2085,11 +2086,11 @@ int si_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &si_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v6_0_ip_block); amdgpu_device_ip_block_add(adev, &si_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v6_0_ip_block); + amdgpu_device_ip_block_add(adev, &si_dma_ip_block); amdgpu_device_ip_block_add(adev, &si_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v6_0_ip_block); - amdgpu_device_ip_block_add(adev, &si_dma_ip_block); break; default: BUG(); diff --git a/drivers/gpu/drm/amd/amdgpu/si_dma.c b/drivers/gpu/drm/amd/amdgpu/si_dma.c index b75d901ba3c4..adbaea6da0d7 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dma.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dma.c @@ -502,12 +502,14 @@ static int si_dma_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* DMA0 trap event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 224, &adev->sdma.trap_irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 224, + &adev->sdma.trap_irq); if (r) return r; /* DMA1 trap event */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 244, &adev->sdma.trap_irq_1); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 244, + &adev->sdma.trap_irq); if (r) return r; @@ -649,17 +651,10 @@ static int si_dma_process_trap_irq(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { - amdgpu_fence_process(&adev->sdma.instance[0].ring); - - return 0; -} - -static int si_dma_process_trap_irq_1(struct amdgpu_device *adev, - struct amdgpu_irq_src *source, - struct amdgpu_iv_entry *entry) -{ - amdgpu_fence_process(&adev->sdma.instance[1].ring); - + if (entry->src_id == 224) + amdgpu_fence_process(&adev->sdma.instance[0].ring); + else + amdgpu_fence_process(&adev->sdma.instance[1].ring); return 0; } @@ -786,11 +781,6 @@ static const struct amdgpu_irq_src_funcs si_dma_trap_irq_funcs = { .process = si_dma_process_trap_irq, }; -static const struct amdgpu_irq_src_funcs si_dma_trap_irq_funcs_1 = { - .set = si_dma_set_trap_irq_state, - .process = si_dma_process_trap_irq_1, -}; - static const struct amdgpu_irq_src_funcs si_dma_illegal_inst_irq_funcs = { .process = si_dma_process_illegal_inst_irq, }; @@ -799,7 +789,6 @@ static void si_dma_set_irq_funcs(struct amdgpu_device *adev) { adev->sdma.trap_irq.num_types = AMDGPU_SDMA_IRQ_LAST; adev->sdma.trap_irq.funcs = &si_dma_trap_irq_funcs; - adev->sdma.trap_irq_1.funcs = &si_dma_trap_irq_funcs_1; adev->sdma.illegal_inst_irq.funcs = &si_dma_illegal_inst_irq_funcs; } @@ -863,10 +852,8 @@ static const struct amdgpu_buffer_funcs si_dma_buffer_funcs = { static void si_dma_set_buffer_funcs(struct amdgpu_device *adev) { - if (adev->mman.buffer_funcs == NULL) { - adev->mman.buffer_funcs = &si_dma_buffer_funcs; - adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; - } + adev->mman.buffer_funcs = &si_dma_buffer_funcs; + adev->mman.buffer_funcs_ring = &adev->sdma.instance[0].ring; } static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = { @@ -879,16 +866,16 @@ static const struct amdgpu_vm_pte_funcs si_dma_vm_pte_funcs = { static void si_dma_set_vm_pte_funcs(struct amdgpu_device *adev) { + struct drm_gpu_scheduler *sched; unsigned i; - if (adev->vm_manager.vm_pte_funcs == NULL) { - adev->vm_manager.vm_pte_funcs = &si_dma_vm_pte_funcs; - for (i = 0; i < adev->sdma.num_instances; i++) - adev->vm_manager.vm_pte_rings[i] = - &adev->sdma.instance[i].ring; - - adev->vm_manager.vm_pte_num_rings = adev->sdma.num_instances; + adev->vm_manager.vm_pte_funcs = &si_dma_vm_pte_funcs; + for (i = 0; i < adev->sdma.num_instances; i++) { + sched = &adev->sdma.instance[i].ring.sched; + adev->vm_manager.vm_pte_rqs[i] = + &sched->sched_rq[DRM_SCHED_PRIORITY_KERNEL]; } + adev->vm_manager.vm_pte_num_rqs = adev->sdma.num_instances; } const struct amdgpu_ip_block_version si_dma_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c index 5c97a3671726..da58040fdbdc 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c @@ -56,16 +56,16 @@ #define BIOS_SCRATCH_4 0x5cd -MODULE_FIRMWARE("radeon/tahiti_smc.bin"); -MODULE_FIRMWARE("radeon/pitcairn_smc.bin"); -MODULE_FIRMWARE("radeon/pitcairn_k_smc.bin"); -MODULE_FIRMWARE("radeon/verde_smc.bin"); -MODULE_FIRMWARE("radeon/verde_k_smc.bin"); -MODULE_FIRMWARE("radeon/oland_smc.bin"); -MODULE_FIRMWARE("radeon/oland_k_smc.bin"); -MODULE_FIRMWARE("radeon/hainan_smc.bin"); -MODULE_FIRMWARE("radeon/hainan_k_smc.bin"); -MODULE_FIRMWARE("radeon/banks_k_2_smc.bin"); +MODULE_FIRMWARE("amdgpu/tahiti_smc.bin"); +MODULE_FIRMWARE("amdgpu/pitcairn_smc.bin"); +MODULE_FIRMWARE("amdgpu/pitcairn_k_smc.bin"); +MODULE_FIRMWARE("amdgpu/verde_smc.bin"); +MODULE_FIRMWARE("amdgpu/verde_k_smc.bin"); +MODULE_FIRMWARE("amdgpu/oland_smc.bin"); +MODULE_FIRMWARE("amdgpu/oland_k_smc.bin"); +MODULE_FIRMWARE("amdgpu/hainan_smc.bin"); +MODULE_FIRMWARE("amdgpu/hainan_k_smc.bin"); +MODULE_FIRMWARE("amdgpu/banks_k_2_smc.bin"); static const struct amd_pm_funcs si_dpm_funcs; @@ -3480,7 +3480,7 @@ static void si_apply_state_adjust_rules(struct amdgpu_device *adev, disable_sclk_switching = true; } - if (adev->pm.dpm.ac_power) + if (adev->pm.ac_power) max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_ac; else max_limits = &adev->pm.dpm.dyn_state.max_clock_voltage_on_dc; @@ -3489,7 +3489,7 @@ static void si_apply_state_adjust_rules(struct amdgpu_device *adev, if (ps->performance_levels[i].vddc > ps->performance_levels[i+1].vddc) ps->performance_levels[i].vddc = ps->performance_levels[i+1].vddc; } - if (adev->pm.dpm.ac_power == false) { + if (adev->pm.ac_power == false) { for (i = 0; i < ps->performance_level_count; i++) { if (ps->performance_levels[i].mclk > max_limits->mclk) ps->performance_levels[i].mclk = max_limits->mclk; @@ -6887,7 +6887,6 @@ static int si_dpm_enable(struct amdgpu_device *adev) si_enable_auto_throttle_source(adev, AMDGPU_DPM_AUTO_THROTTLE_SRC_THERMAL, true); si_thermal_start_thermal_controller(adev); - ni_update_current_ps(adev, boot_ps); return 0; } @@ -7318,8 +7317,7 @@ static int si_dpm_init(struct amdgpu_device *adev) pi = &eg_pi->rv7xx; si_pi->sys_pcie_mask = - (adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_MASK) >> - CAIL_PCIE_LINK_SPEED_SUPPORT_SHIFT; + adev->pm.pcie_gen_mask & CAIL_PCIE_LINK_SPEED_SUPPORT_MASK; si_pi->force_pcie_gen = AMDGPU_PCIE_GEN_INVALID; si_pi->boot_pcie_gen = si_get_current_pcie_speed(adev); @@ -7667,7 +7665,7 @@ static int si_dpm_init_microcode(struct amdgpu_device *adev) default: BUG(); } - snprintf(fw_name, sizeof(fw_name), "radeon/%s_smc.bin", chip_name); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_smc.bin", chip_name); err = request_firmware(&adev->pm.fw, fw_name, adev->dev); if (err) goto out; @@ -7689,11 +7687,11 @@ static int si_dpm_sw_init(void *handle) int ret; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 230, &adev->pm.dpm.thermal.irq); + ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 230, &adev->pm.dpm.thermal.irq); if (ret) return ret; - ret = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 231, &adev->pm.dpm.thermal.irq); + ret = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 231, &adev->pm.dpm.thermal.irq); if (ret) return ret; @@ -7764,7 +7762,7 @@ static int si_dpm_hw_init(void *handle) else adev->pm.dpm_enabled = true; mutex_unlock(&adev->pm.mutex); - + amdgpu_pm_compute_clocks(adev); return ret; } diff --git a/drivers/gpu/drm/amd/amdgpu/si_enums.h b/drivers/gpu/drm/amd/amdgpu/si_enums.h index dc9e0e6b4558..790ba46eaebb 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_enums.h +++ b/drivers/gpu/drm/amd/amdgpu/si_enums.h @@ -46,6 +46,26 @@ #define GRPH_ENDIAN_8IN16 1 #define GRPH_ENDIAN_8IN32 2 #define GRPH_ENDIAN_8IN64 3 +#define GRPH_RED_CROSSBAR(x) (((x) & 0x3) << 4) +#define GRPH_RED_SEL_R 0 +#define GRPH_RED_SEL_G 1 +#define GRPH_RED_SEL_B 2 +#define GRPH_RED_SEL_A 3 +#define GRPH_GREEN_CROSSBAR(x) (((x) & 0x3) << 6) +#define GRPH_GREEN_SEL_G 0 +#define GRPH_GREEN_SEL_B 1 +#define GRPH_GREEN_SEL_A 2 +#define GRPH_GREEN_SEL_R 3 +#define GRPH_BLUE_CROSSBAR(x) (((x) & 0x3) << 8) +#define GRPH_BLUE_SEL_B 0 +#define GRPH_BLUE_SEL_A 1 +#define GRPH_BLUE_SEL_R 2 +#define GRPH_BLUE_SEL_G 3 +#define GRPH_ALPHA_CROSSBAR(x) (((x) & 0x3) << 10) +#define GRPH_ALPHA_SEL_A 0 +#define GRPH_ALPHA_SEL_R 1 +#define GRPH_ALPHA_SEL_G 2 +#define GRPH_ALPHA_SEL_B 3 #define GRPH_DEPTH(x) (((x) & 0x3) << 0) #define GRPH_DEPTH_8BPP 0 diff --git a/drivers/gpu/drm/amd/amdgpu/si_ih.c b/drivers/gpu/drm/amd/amdgpu/si_ih.c index 60dad63098a2..b3d7d9f83202 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/si_ih.c @@ -142,7 +142,7 @@ static void si_ih_decode_iv(struct amdgpu_device *adev, dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - entry->client_id = AMDGPU_IH_CLIENTID_LEGACY; + entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; entry->src_data[0] = dw[1] & 0xfffffff; entry->ring_id = dw[2] & 0xff; @@ -170,7 +170,7 @@ static int si_ih_sw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = amdgpu_ih_ring_init(adev, 64 * 1024, false); + r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, false); if (r) return r; @@ -182,7 +182,7 @@ static int si_ih_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_irq_fini(adev); - amdgpu_ih_ring_fini(adev); + amdgpu_ih_ring_fini(adev, &adev->irq.ih); return 0; } @@ -308,8 +308,7 @@ static const struct amdgpu_ih_funcs si_ih_funcs = { static void si_ih_set_interrupt_funcs(struct amdgpu_device *adev) { - if (adev->irq.ih_funcs == NULL) - adev->irq.ih_funcs = &si_ih_funcs; + adev->irq.ih_funcs = &si_ih_funcs; } const struct amdgpu_ip_block_version si_ih_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/sid.h b/drivers/gpu/drm/amd/amdgpu/sid.h index c57eff159374..7cf12adb3915 100644 --- a/drivers/gpu/drm/amd/amdgpu/sid.h +++ b/drivers/gpu/drm/amd/amdgpu/sid.h @@ -2201,6 +2201,26 @@ # define EVERGREEN_GRPH_ENDIAN_8IN16 1 # define EVERGREEN_GRPH_ENDIAN_8IN32 2 # define EVERGREEN_GRPH_ENDIAN_8IN64 3 +#define EVERGREEN_GRPH_RED_CROSSBAR(x) (((x) & 0x3) << 4) +# define EVERGREEN_GRPH_RED_SEL_R 0 +# define EVERGREEN_GRPH_RED_SEL_G 1 +# define EVERGREEN_GRPH_RED_SEL_B 2 +# define EVERGREEN_GRPH_RED_SEL_A 3 +#define EVERGREEN_GRPH_GREEN_CROSSBAR(x) (((x) & 0x3) << 6) +# define EVERGREEN_GRPH_GREEN_SEL_G 0 +# define EVERGREEN_GRPH_GREEN_SEL_B 1 +# define EVERGREEN_GRPH_GREEN_SEL_A 2 +# define EVERGREEN_GRPH_GREEN_SEL_R 3 +#define EVERGREEN_GRPH_BLUE_CROSSBAR(x) (((x) & 0x3) << 8) +# define EVERGREEN_GRPH_BLUE_SEL_B 0 +# define EVERGREEN_GRPH_BLUE_SEL_A 1 +# define EVERGREEN_GRPH_BLUE_SEL_R 2 +# define EVERGREEN_GRPH_BLUE_SEL_G 3 +#define EVERGREEN_GRPH_ALPHA_CROSSBAR(x) (((x) & 0x3) << 10) +# define EVERGREEN_GRPH_ALPHA_SEL_A 0 +# define EVERGREEN_GRPH_ALPHA_SEL_R 1 +# define EVERGREEN_GRPH_ALPHA_SEL_G 2 +# define EVERGREEN_GRPH_ALPHA_SEL_B 3 #define EVERGREEN_D3VGA_CONTROL 0xf8 #define EVERGREEN_D4VGA_CONTROL 0xf9 diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 83f2717fcf81..bf5e6a413dee 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -479,6 +479,11 @@ static const struct amdgpu_ip_block_version vega10_common_ip_block = .funcs = &soc15_common_ip_funcs, }; +static uint32_t soc15_get_rev_id(struct amdgpu_device *adev) +{ + return adev->nbio_funcs->get_rev_id(adev); +} + int soc15_set_ip_blocks(struct amdgpu_device *adev) { /* Set IP register base before any HW register access */ @@ -498,7 +503,7 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) if (adev->flags & AMD_IS_APU) adev->nbio_funcs = &nbio_v7_0_funcs; else if (adev->asic_type == CHIP_VEGA20) - adev->nbio_funcs = &nbio_v7_0_funcs; + adev->nbio_funcs = &nbio_v7_4_funcs; else adev->nbio_funcs = &nbio_v6_1_funcs; @@ -506,6 +511,8 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) adev->df_funcs = &df_v3_6_funcs; else adev->df_funcs = &df_v1_7_funcs; + + adev->rev_id = soc15_get_rev_id(adev); adev->nbio_funcs->detect_hw_virt(adev); if (amdgpu_sriov_vf(adev)) @@ -518,11 +525,14 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vega10_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &vega10_ih_ip_block); - if (adev->asic_type != CHIP_VEGA20) { + if (adev->asic_type == CHIP_VEGA20) + amdgpu_device_ip_block_add(adev, &psp_v11_0_ip_block); + else amdgpu_device_ip_block_add(adev, &psp_v3_1_ip_block); - if (!amdgpu_sriov_vf(adev)) - amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); - } + amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block); + if (!amdgpu_sriov_vf(adev)) + amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); #if defined(CONFIG_DRM_AMD_DC) @@ -531,16 +541,18 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) #else # warning "Enable CONFIG_DRM_AMD_DC for display support on SOC15." #endif - amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block); - amdgpu_device_ip_block_add(adev, &uvd_v7_0_ip_block); - amdgpu_device_ip_block_add(adev, &vce_v4_0_ip_block); + if (!(adev->asic_type == CHIP_VEGA20 && amdgpu_sriov_vf(adev))) { + amdgpu_device_ip_block_add(adev, &uvd_v7_0_ip_block); + amdgpu_device_ip_block_add(adev, &vce_v4_0_ip_block); + } break; case CHIP_RAVEN: amdgpu_device_ip_block_add(adev, &vega10_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &vega10_ih_ip_block); amdgpu_device_ip_block_add(adev, &psp_v10_0_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -550,8 +562,6 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) #else # warning "Enable CONFIG_DRM_AMD_DC for display support on SOC15." #endif - amdgpu_device_ip_block_add(adev, &gfx_v9_0_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v4_0_ip_block); amdgpu_device_ip_block_add(adev, &vcn_v1_0_ip_block); break; default: @@ -561,11 +571,6 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) return 0; } -static uint32_t soc15_get_rev_id(struct amdgpu_device *adev) -{ - return adev->nbio_funcs->get_rev_id(adev); -} - static void soc15_flush_hdp(struct amdgpu_device *adev, struct amdgpu_ring *ring) { adev->nbio_funcs->hdp_flush(adev, ring); @@ -622,7 +627,6 @@ static int soc15_common_early_init(void *handle) adev->asic_funcs = &soc15_asic_funcs; - adev->rev_id = soc15_get_rev_id(adev); adev->external_rev_id = 0xFF; switch (adev->asic_type) { case CHIP_VEGA10: @@ -693,35 +697,79 @@ static int soc15_common_early_init(void *handle) adev->external_rev_id = adev->rev_id + 0x28; break; case CHIP_RAVEN: - adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | - AMD_CG_SUPPORT_GFX_MGLS | - AMD_CG_SUPPORT_GFX_RLC_LS | - AMD_CG_SUPPORT_GFX_CP_LS | - AMD_CG_SUPPORT_GFX_3D_CGCG | - AMD_CG_SUPPORT_GFX_3D_CGLS | - AMD_CG_SUPPORT_GFX_CGCG | - AMD_CG_SUPPORT_GFX_CGLS | - AMD_CG_SUPPORT_BIF_MGCG | - AMD_CG_SUPPORT_BIF_LS | - AMD_CG_SUPPORT_HDP_MGCG | - AMD_CG_SUPPORT_HDP_LS | - AMD_CG_SUPPORT_DRM_MGCG | - AMD_CG_SUPPORT_DRM_LS | - AMD_CG_SUPPORT_ROM_MGCG | - AMD_CG_SUPPORT_MC_MGCG | - AMD_CG_SUPPORT_MC_LS | - AMD_CG_SUPPORT_SDMA_MGCG | - AMD_CG_SUPPORT_SDMA_LS | - AMD_CG_SUPPORT_VCN_MGCG; - - adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_VCN; + if (adev->rev_id >= 0x8) + adev->external_rev_id = adev->rev_id + 0x81; + else if (adev->pdev->device == 0x15d8) + adev->external_rev_id = adev->rev_id + 0x41; + else + adev->external_rev_id = 0x1; + + if (adev->rev_id >= 0x8) { + adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_BIF_LS | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_ROM_MGCG | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_SDMA_MGCG | + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCN_MGCG; + + adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_VCN; + } else if (adev->pdev->device == 0x15d8) { + adev->cg_flags = AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_BIF_LS | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_ROM_MGCG | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_SDMA_MGCG | + AMD_CG_SUPPORT_SDMA_LS; + + adev->pg_flags = AMD_PG_SUPPORT_SDMA | + AMD_PG_SUPPORT_MMHUB | + AMD_PG_SUPPORT_VCN | + AMD_PG_SUPPORT_VCN_DPG; + } else { + adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_RLC_LS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_BIF_MGCG | + AMD_CG_SUPPORT_BIF_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_DRM_MGCG | + AMD_CG_SUPPORT_DRM_LS | + AMD_CG_SUPPORT_ROM_MGCG | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_SDMA_MGCG | + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCN_MGCG; + + adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_VCN; + } if (adev->powerplay.pp_feature & PP_GFXOFF_MASK) adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_CP | AMD_PG_SUPPORT_RLC_SMU_HS; - - adev->external_rev_id = 0x1; break; default: /* FIXME: not supported yet */ diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.h b/drivers/gpu/drm/amd/amdgpu/soc15.h index 1f714b7af520..f8ad7804dc40 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15.h @@ -26,6 +26,7 @@ #include "nbio_v6_1.h" #include "nbio_v7_0.h" +#include "nbio_v7_4.h" #define SOC15_FLUSH_GPU_TLB_NUM_WREG 4 #define SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT 1 diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h index 0942f492d2e1..958b10a57073 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h @@ -56,12 +56,34 @@ tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \ loop--; \ if (!loop) { \ + DRM_ERROR("Register(%d) [%s] failed to reach value 0x%08x != 0x%08x\n", \ + inst, #reg, (unsigned)expected_value, (unsigned)(tmp_ & (mask))); \ ret = -ETIMEDOUT; \ break; \ } \ } \ } while (0) +#define RREG32_SOC15_DPG_MODE(ip, inst, reg, mask, sram_sel) \ + ({ WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_MASK, mask); \ + WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_CTL, \ + UVD_DPG_LMA_CTL__MASK_EN_MASK | \ + ((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) \ + << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ + (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ + RREG32_SOC15(ip, inst, mmUVD_DPG_LMA_DATA); }) + +#define WREG32_SOC15_DPG_MODE(ip, inst, reg, value, mask, sram_sel) \ + do { \ + WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_DATA, value); \ + WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_MASK, mask); \ + WREG32_SOC15(ip, inst, mmUVD_DPG_LMA_CTL, \ + UVD_DPG_LMA_CTL__READ_WRITE_MASK | \ + ((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) \ + << UVD_DPG_LMA_CTL__READ_WRITE_ADDR__SHIFT) | \ + (sram_sel << UVD_DPG_LMA_CTL__SRAM_SEL__SHIFT)); \ + } while (0) + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/soc15d.h b/drivers/gpu/drm/amd/amdgpu/soc15d.h index 8dc29107228f..edfe50821cd9 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15d.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15d.h @@ -53,6 +53,29 @@ #define PACKET3_COMPUTE(op, n) (PACKET3(op, n) | 1 << 1) +#define PACKETJ_CONDITION_CHECK0 0 +#define PACKETJ_CONDITION_CHECK1 1 +#define PACKETJ_CONDITION_CHECK2 2 +#define PACKETJ_CONDITION_CHECK3 3 +#define PACKETJ_CONDITION_CHECK4 4 +#define PACKETJ_CONDITION_CHECK5 5 +#define PACKETJ_CONDITION_CHECK6 6 +#define PACKETJ_CONDITION_CHECK7 7 + +#define PACKETJ_TYPE0 0 +#define PACKETJ_TYPE1 1 +#define PACKETJ_TYPE2 2 +#define PACKETJ_TYPE3 3 +#define PACKETJ_TYPE4 4 +#define PACKETJ_TYPE5 5 +#define PACKETJ_TYPE6 6 +#define PACKETJ_TYPE7 7 + +#define PACKETJ(reg, r, cond, type) ((reg & 0x3FFFF) | \ + ((r & 0x3F) << 18) | \ + ((cond & 0xF) << 24) | \ + ((type & 0xF) << 28)) + /* Packet 3 types */ #define PACKET3_NOP 0x10 #define PACKET3_SET_BASE 0x11 diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c index 52853d8a8fdd..3abffd06b5c7 100644 --- a/drivers/gpu/drm/amd/amdgpu/tonga_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/tonga_ih.c @@ -266,7 +266,7 @@ static void tonga_ih_decode_iv(struct amdgpu_device *adev, dw[2] = le32_to_cpu(adev->irq.ih.ring[ring_index + 2]); dw[3] = le32_to_cpu(adev->irq.ih.ring[ring_index + 3]); - entry->client_id = AMDGPU_IH_CLIENTID_LEGACY; + entry->client_id = AMDGPU_IRQ_CLIENTID_LEGACY; entry->src_id = dw[0] & 0xff; entry->src_data[0] = dw[1] & 0xfffffff; entry->ring_id = dw[2] & 0xff; @@ -317,7 +317,7 @@ static int tonga_ih_sw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = amdgpu_ih_ring_init(adev, 64 * 1024, true); + r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 64 * 1024, true); if (r) return r; @@ -334,7 +334,7 @@ static int tonga_ih_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_irq_fini(adev); - amdgpu_ih_ring_fini(adev); + amdgpu_ih_ring_fini(adev, &adev->irq.ih); amdgpu_irq_remove_domain(adev); return 0; @@ -513,8 +513,7 @@ static const struct amdgpu_ih_funcs tonga_ih_funcs = { static void tonga_ih_set_interrupt_funcs(struct amdgpu_device *adev) { - if (adev->irq.ih_funcs == NULL) - adev->irq.ih_funcs = &tonga_ih_funcs; + adev->irq.ih_funcs = &tonga_ih_funcs; } const struct amdgpu_ip_block_version tonga_ih_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c index 6fed3d7797a8..1fc17bf39fed 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c @@ -108,7 +108,7 @@ static int uvd_v4_2_sw_init(void *handle) int r; /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq); if (r) return r; @@ -123,6 +123,10 @@ static int uvd_v4_2_sw_init(void *handle) ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); + if (r) + return r; + + r = amdgpu_uvd_entity_init(adev); return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index 341ee6d55ce8..fde6ad5ac9ab 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -35,6 +35,7 @@ #include "vi.h" #include "smu/smu_7_1_2_d.h" #include "smu/smu_7_1_2_sh_mask.h" +#include "ivsrcid/ivsrcid_vislands30.h" static void uvd_v5_0_set_ring_funcs(struct amdgpu_device *adev); static void uvd_v5_0_set_irq_funcs(struct amdgpu_device *adev); @@ -104,7 +105,7 @@ static int uvd_v5_0_sw_init(void *handle) int r; /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_UVD_SYSTEM_MESSAGE, &adev->uvd.inst->irq); if (r) return r; @@ -119,6 +120,10 @@ static int uvd_v5_0_sw_init(void *handle) ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); + if (r) + return r; + + r = amdgpu_uvd_entity_init(adev); return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index bfddf97dd13e..7a5b40275e8e 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -36,6 +36,7 @@ #include "bif/bif_5_1_d.h" #include "gmc/gmc_8_1_d.h" #include "vi.h" +#include "ivsrcid/ivsrcid_vislands30.h" /* Polaris10/11/12 firmware version */ #define FW_1_130_16 ((1 << 24) | (130 << 16) | (16 << 8)) @@ -247,12 +248,10 @@ static int uvd_v6_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; - amdgpu_job_free(job); if (fence) *fence = dma_fence_get(f); dma_fence_put(f); @@ -275,7 +274,7 @@ err: */ static int uvd_v6_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, - bool direct, struct dma_fence **fence) + struct dma_fence **fence) { const unsigned ib_size_dw = 16; struct amdgpu_job *job; @@ -311,19 +310,9 @@ static int uvd_v6_0_enc_get_destroy_msg(struct amdgpu_ring *ring, for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - if (direct) { - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); - if (r) - goto err; - - amdgpu_job_free(job); - } else { - r = amdgpu_job_submit(job, ring, &ring->adev->vce.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, &f); - if (r) - goto err; - } + r = amdgpu_job_submit_direct(job, ring, &f); + if (r) + goto err; if (fence) *fence = dma_fence_get(f); @@ -352,7 +341,7 @@ static int uvd_v6_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) goto error; } - r = uvd_v6_0_enc_get_destroy_msg(ring, 1, true, &fence); + r = uvd_v6_0_enc_get_destroy_msg(ring, 1, &fence); if (r) { DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); goto error; @@ -400,14 +389,14 @@ static int uvd_v6_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_UVD_SYSTEM_MESSAGE, &adev->uvd.inst->irq); if (r) return r; /* UVD ENC TRAP */ if (uvd_v6_0_enc_support(adev)) { for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 119, &adev->uvd.inst->irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, i + VISLANDS30_IV_SRCID_UVD_ENC_GEN_PURP, &adev->uvd.inst->irq); if (r) return r; } @@ -425,16 +414,6 @@ static int uvd_v6_0_sw_init(void *handle) adev->uvd.num_enc_rings = 0; DRM_INFO("UVD ENC is disabled\n"); - } else { - struct drm_sched_rq *rq; - ring = &adev->uvd.inst->ring_enc[0]; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&ring->sched, &adev->uvd.inst->entity_enc, - rq, NULL); - if (r) { - DRM_ERROR("Failed setting up UVD ENC run queue.\n"); - return r; - } } r = amdgpu_uvd_resume(adev); @@ -457,6 +436,8 @@ static int uvd_v6_0_sw_init(void *handle) } } + r = amdgpu_uvd_entity_init(adev); + return r; } @@ -470,8 +451,6 @@ static int uvd_v6_0_sw_fini(void *handle) return r; if (uvd_v6_0_enc_support(adev)) { - drm_sched_entity_fini(&adev->uvd.inst->ring_enc[0].sched, &adev->uvd.inst->entity_enc); - for (i = 0; i < adev->uvd.num_enc_rings; ++i) amdgpu_ring_fini(&adev->uvd.inst->ring_enc[i]); } @@ -1569,7 +1548,6 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_phys_funcs = { static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_UVD, .align_mask = 0xf, - .nop = PACKET0(mmUVD_NO_OP, 0), .support_64bit_ptrs = false, .get_rptr = uvd_v6_0_ring_get_rptr, .get_wptr = uvd_v6_0_ring_get_wptr, @@ -1587,7 +1565,7 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .emit_hdp_flush = uvd_v6_0_ring_emit_hdp_flush, .test_ring = uvd_v6_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, - .insert_nop = amdgpu_ring_insert_nop, + .insert_nop = uvd_v6_0_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .begin_use = amdgpu_uvd_ring_begin_use, .end_use = amdgpu_uvd_ring_end_use, diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index 57d32f21b3a6..58b39afcfb86 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -39,6 +39,13 @@ #include "hdp/hdp_4_0_offset.h" #include "mmhub/mmhub_1_0_offset.h" #include "mmhub/mmhub_1_0_sh_mask.h" +#include "ivsrcid/uvd/irqsrcs_uvd_7_0.h" + +#define mmUVD_PG0_CC_UVD_HARVESTING 0x00c7 +#define mmUVD_PG0_CC_UVD_HARVESTING_BASE_IDX 1 +//UVD_PG0_CC_UVD_HARVESTING +#define UVD_PG0_CC_UVD_HARVESTING__UVD_DISABLE__SHIFT 0x1 +#define UVD_PG0_CC_UVD_HARVESTING__UVD_DISABLE_MASK 0x00000002L #define UVD7_MAX_HW_INSTANCES_VEGA20 2 @@ -249,12 +256,10 @@ static int uvd_v7_0_enc_get_create_msg(struct amdgpu_ring *ring, uint32_t handle for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); + r = amdgpu_job_submit_direct(job, ring, &f); if (r) goto err; - amdgpu_job_free(job); if (fence) *fence = dma_fence_get(f); dma_fence_put(f); @@ -275,8 +280,8 @@ err: * * Close up a stream for HW test or if userspace failed to do so */ -int uvd_v7_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, - bool direct, struct dma_fence **fence) +static int uvd_v7_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, + struct dma_fence **fence) { const unsigned ib_size_dw = 16; struct amdgpu_job *job; @@ -312,19 +317,9 @@ int uvd_v7_0_enc_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, for (i = ib->length_dw; i < ib_size_dw; ++i) ib->ptr[i] = 0x0; - if (direct) { - r = amdgpu_ib_schedule(ring, 1, ib, NULL, &f); - job->fence = dma_fence_get(f); - if (r) - goto err; - - amdgpu_job_free(job); - } else { - r = amdgpu_job_submit(job, ring, &ring->adev->vce.entity, - AMDGPU_FENCE_OWNER_UNDEFINED, &f); - if (r) - goto err; - } + r = amdgpu_job_submit_direct(job, ring, &f); + if (r) + goto err; if (fence) *fence = dma_fence_get(f); @@ -353,7 +348,7 @@ static int uvd_v7_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) goto error; } - r = uvd_v7_0_enc_get_destroy_msg(ring, 1, true, &fence); + r = uvd_v7_0_enc_get_destroy_msg(ring, 1, &fence); if (r) { DRM_ERROR("amdgpu: (%d)failed to get destroy ib (%ld).\n", ring->me, r); goto error; @@ -377,10 +372,25 @@ error: static int uvd_v7_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->asic_type == CHIP_VEGA20) + + if (adev->asic_type == CHIP_VEGA20) { + u32 harvest; + int i; + adev->uvd.num_uvd_inst = UVD7_MAX_HW_INSTANCES_VEGA20; - else + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + harvest = RREG32_SOC15(UVD, i, mmUVD_PG0_CC_UVD_HARVESTING); + if (harvest & UVD_PG0_CC_UVD_HARVESTING__UVD_DISABLE_MASK) { + adev->uvd.harvest_config |= 1 << i; + } + } + if (adev->uvd.harvest_config == (AMDGPU_UVD_HARVEST_UVD0 | + AMDGPU_UVD_HARVEST_UVD1)) + /* both instances are harvested, disable the block */ + return -ENOENT; + } else { adev->uvd.num_uvd_inst = 1; + } if (amdgpu_sriov_vf(adev)) adev->uvd.num_enc_rings = 1; @@ -396,19 +406,21 @@ static int uvd_v7_0_early_init(void *handle) static int uvd_v7_0_sw_init(void *handle) { struct amdgpu_ring *ring; - struct drm_sched_rq *rq; + int i, j, r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + if (adev->uvd.harvest_config & (1 << j)) + continue; /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_uvds[j], 124, &adev->uvd.inst[j].irq); + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_uvds[j], UVD_7_0__SRCID__UVD_SYSTEM_MESSAGE_INTERRUPT, &adev->uvd.inst[j].irq); if (r) return r; /* UVD ENC TRAP */ for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_uvds[j], i + 119, &adev->uvd.inst[j].irq); + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_uvds[j], i + UVD_7_0__SRCID__UVD_ENC_GEN_PURP, &adev->uvd.inst[j].irq); if (r) return r; } @@ -425,18 +437,14 @@ static int uvd_v7_0_sw_init(void *handle) adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].fw = adev->uvd.fw; adev->firmware.fw_size += ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE); - DRM_INFO("PSP loading UVD firmware\n"); - } - for (j = 0; j < adev->uvd.num_uvd_inst; j++) { - ring = &adev->uvd.inst[j].ring_enc[0]; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&ring->sched, &adev->uvd.inst[j].entity_enc, - rq, NULL); - if (r) { - DRM_ERROR("(%d)Failed setting up UVD ENC run queue.\n", j); - return r; + if (adev->uvd.num_uvd_inst == UVD7_MAX_HW_INSTANCES_VEGA20) { + adev->firmware.ucode[AMDGPU_UCODE_ID_UVD1].ucode_id = AMDGPU_UCODE_ID_UVD1; + adev->firmware.ucode[AMDGPU_UCODE_ID_UVD1].fw = adev->uvd.fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE); } + DRM_INFO("PSP loading UVD firmware\n"); } r = amdgpu_uvd_resume(adev); @@ -444,6 +452,8 @@ static int uvd_v7_0_sw_init(void *handle) return r; for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + if (adev->uvd.harvest_config & (1 << j)) + continue; if (!amdgpu_sriov_vf(adev)) { ring = &adev->uvd.inst[j].ring; sprintf(ring->name, "uvd<%d>", j); @@ -472,6 +482,10 @@ static int uvd_v7_0_sw_init(void *handle) } } + r = amdgpu_uvd_entity_init(adev); + if (r) + return r; + r = amdgpu_virt_alloc_mm_table(adev); if (r) return r; @@ -491,8 +505,8 @@ static int uvd_v7_0_sw_fini(void *handle) return r; for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { - drm_sched_entity_fini(&adev->uvd.inst[j].ring_enc[0].sched, &adev->uvd.inst[j].entity_enc); - + if (adev->uvd.harvest_config & (1 << j)) + continue; for (i = 0; i < adev->uvd.num_enc_rings; ++i) amdgpu_ring_fini(&adev->uvd.inst[j].ring_enc[i]); } @@ -521,6 +535,8 @@ static int uvd_v7_0_hw_init(void *handle) goto done; for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.harvest_config & (1 << j)) + continue; ring = &adev->uvd.inst[j].ring; if (!amdgpu_sriov_vf(adev)) { @@ -600,8 +616,11 @@ static int uvd_v7_0_hw_fini(void *handle) DRM_DEBUG("For SRIOV client, shouldn't do anything.\n"); } - for (i = 0; i < adev->uvd.num_uvd_inst; ++i) + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; adev->uvd.inst[i].ring.ready = false; + } return 0; } @@ -644,11 +663,18 @@ static void uvd_v7_0_mc_resume(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, - lower_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); + i == 0 ? + adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].tmr_mc_addr_lo: + adev->firmware.ucode[AMDGPU_UCODE_ID_UVD1].tmr_mc_addr_lo); WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, - upper_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); + i == 0 ? + adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].tmr_mc_addr_hi: + adev->firmware.ucode[AMDGPU_UCODE_ID_UVD1].tmr_mc_addr_hi); + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_OFFSET0, 0); offset = 0; } else { WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, @@ -656,10 +682,10 @@ static void uvd_v7_0_mc_resume(struct amdgpu_device *adev) WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, upper_32_bits(adev->uvd.inst[i].gpu_addr)); offset = size; + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_OFFSET0, + AMDGPU_UVD_FIRMWARE_OFFSET >> 3); } - WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_OFFSET0, - AMDGPU_UVD_FIRMWARE_OFFSET >> 3); WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_SIZE0, size); WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, @@ -716,6 +742,8 @@ static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, WREG32_SOC15(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP, 0); for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; WDOORBELL32(adev->uvd.inst[i].ring_enc[0].doorbell_index, 0); adev->wb.wb[adev->uvd.inst[i].ring_enc[0].wptr_offs] = 0; adev->uvd.inst[i].ring_enc[0].wptr = 0; @@ -772,6 +800,8 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) init_table += header->uvd_table_offset; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; ring = &adev->uvd.inst[i].ring; ring->wptr = 0; size = AMDGPU_GPU_PAGE_ALIGN(adev->uvd.fw->size + 4); @@ -911,6 +941,8 @@ static int uvd_v7_0_start(struct amdgpu_device *adev) int i, j, k, r; for (k = 0; k < adev->uvd.num_uvd_inst; ++k) { + if (adev->uvd.harvest_config & (1 << k)) + continue; /* disable DPG */ WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_POWER_STATUS), 0, ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); @@ -923,6 +955,8 @@ static int uvd_v7_0_start(struct amdgpu_device *adev) uvd_v7_0_mc_resume(adev); for (k = 0; k < adev->uvd.num_uvd_inst; ++k) { + if (adev->uvd.harvest_config & (1 << k)) + continue; ring = &adev->uvd.inst[k].ring; /* disable clock gating */ WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_CGC_CTRL), 0, @@ -1090,6 +1124,8 @@ static void uvd_v7_0_stop(struct amdgpu_device *adev) uint8_t i = 0; for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->uvd.harvest_config & (1 << i)) + continue; /* force RBC into idle state */ WREG32_SOC15(UVD, i, mmUVD_RBC_RB_CNTL, 0x11010101); @@ -1227,6 +1263,35 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring) } /** + * uvd_v7_0_ring_patch_cs_in_place - Patch the IB for command submission. + * + * @p: the CS parser with the IBs + * @ib_idx: which IB to patch + * + */ +static int uvd_v7_0_ring_patch_cs_in_place(struct amdgpu_cs_parser *p, + uint32_t ib_idx) +{ + struct amdgpu_ring *ring = to_amdgpu_ring(p->entity->rq->sched); + struct amdgpu_ib *ib = &p->job->ibs[ib_idx]; + unsigned i; + + /* No patching necessary for the first instance */ + if (!ring->me) + return 0; + + for (i = 0; i < ib->length_dw; i += 2) { + uint32_t reg = amdgpu_get_ib_value(p, ib_idx, i); + + reg -= p->adev->reg_offset[UVD_HWIP][0][1]; + reg += p->adev->reg_offset[UVD_HWIP][1][1]; + + amdgpu_set_ib_value(p, ib_idx, i, reg); + } + return 0; +} + +/** * uvd_v7_0_ring_emit_ib - execute indirect buffer * * @ring: amdgpu_ring pointer @@ -1718,6 +1783,7 @@ static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = { .get_rptr = uvd_v7_0_ring_get_rptr, .get_wptr = uvd_v7_0_ring_get_wptr, .set_wptr = uvd_v7_0_ring_set_wptr, + .patch_cs_in_place = uvd_v7_0_ring_patch_cs_in_place, .emit_frame_size = 6 + /* hdp invalidate */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + @@ -1777,6 +1843,8 @@ static void uvd_v7_0_set_ring_funcs(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; adev->uvd.inst[i].ring.funcs = &uvd_v7_0_ring_vm_funcs; adev->uvd.inst[i].ring.me = i; DRM_INFO("UVD(%d) is enabled in VM mode\n", i); @@ -1788,6 +1856,8 @@ static void uvd_v7_0_set_enc_ring_funcs(struct amdgpu_device *adev) int i, j; for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + if (adev->uvd.harvest_config & (1 << j)) + continue; for (i = 0; i < adev->uvd.num_enc_rings; ++i) { adev->uvd.inst[j].ring_enc[i].funcs = &uvd_v7_0_enc_ring_vm_funcs; adev->uvd.inst[j].ring_enc[i].me = j; @@ -1807,6 +1877,8 @@ static void uvd_v7_0_set_irq_funcs(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.harvest_config & (1 << i)) + continue; adev->uvd.inst[i].irq.num_types = adev->uvd.num_enc_rings + 1; adev->uvd.inst[i].irq.funcs = &uvd_v7_0_irq_funcs; } diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c index 47f70827195b..ea28828360d3 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v2_0.c @@ -56,7 +56,7 @@ static uint64_t vce_v2_0_ring_get_rptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) return RREG32(mmVCE_RB_RPTR); else return RREG32(mmVCE_RB_RPTR2); @@ -73,7 +73,7 @@ static uint64_t vce_v2_0_ring_get_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) return RREG32(mmVCE_RB_WPTR); else return RREG32(mmVCE_RB_WPTR2); @@ -90,7 +90,7 @@ static void vce_v2_0_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) WREG32(mmVCE_RB_WPTR, lower_32_bits(ring->wptr)); else WREG32(mmVCE_RB_WPTR2, lower_32_bits(ring->wptr)); @@ -417,7 +417,7 @@ static int vce_v2_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* VCE */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 167, &adev->vce.irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, 167, &adev->vce.irq); if (r) return r; @@ -439,6 +439,8 @@ static int vce_v2_0_sw_init(void *handle) return r; } + r = amdgpu_vce_entity_init(adev); + return r; } @@ -627,8 +629,10 @@ static void vce_v2_0_set_ring_funcs(struct amdgpu_device *adev) { int i; - for (i = 0; i < adev->vce.num_rings; i++) + for (i = 0; i < adev->vce.num_rings; i++) { adev->vce.ring[i].funcs = &vce_v2_0_ring_funcs; + adev->vce.ring[i].me = i; + } } static const struct amdgpu_irq_src_funcs vce_v2_0_irq_funcs = { diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index a71b97519cc0..6dbd39730070 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -39,6 +39,7 @@ #include "smu/smu_7_1_2_sh_mask.h" #include "gca/gfx_8_0_d.h" #include "gca/gfx_8_0_sh_mask.h" +#include "ivsrcid/ivsrcid_vislands30.h" #define GRBM_GFX_INDEX__VCE_INSTANCE__SHIFT 0x04 @@ -86,9 +87,9 @@ static uint64_t vce_v3_0_ring_get_rptr(struct amdgpu_ring *ring) else if (adev->vce.harvest_config == AMDGPU_VCE_HARVEST_VCE0) WREG32(mmGRBM_GFX_INDEX, GET_VCE_INSTANCE(1)); - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) v = RREG32(mmVCE_RB_RPTR); - else if (ring == &adev->vce.ring[1]) + else if (ring->me == 1) v = RREG32(mmVCE_RB_RPTR2); else v = RREG32(mmVCE_RB_RPTR3); @@ -118,9 +119,9 @@ static uint64_t vce_v3_0_ring_get_wptr(struct amdgpu_ring *ring) else if (adev->vce.harvest_config == AMDGPU_VCE_HARVEST_VCE0) WREG32(mmGRBM_GFX_INDEX, GET_VCE_INSTANCE(1)); - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) v = RREG32(mmVCE_RB_WPTR); - else if (ring == &adev->vce.ring[1]) + else if (ring->me == 1) v = RREG32(mmVCE_RB_WPTR2); else v = RREG32(mmVCE_RB_WPTR3); @@ -149,9 +150,9 @@ static void vce_v3_0_ring_set_wptr(struct amdgpu_ring *ring) else if (adev->vce.harvest_config == AMDGPU_VCE_HARVEST_VCE0) WREG32(mmGRBM_GFX_INDEX, GET_VCE_INSTANCE(1)); - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) WREG32(mmVCE_RB_WPTR, lower_32_bits(ring->wptr)); - else if (ring == &adev->vce.ring[1]) + else if (ring->me == 1) WREG32(mmVCE_RB_WPTR2, lower_32_bits(ring->wptr)); else WREG32(mmVCE_RB_WPTR3, lower_32_bits(ring->wptr)); @@ -422,7 +423,7 @@ static int vce_v3_0_sw_init(void *handle) int r, i; /* VCE */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 167, &adev->vce.irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IRQ_CLIENTID_LEGACY, VISLANDS30_IV_SRCID_VCE_TRAP, &adev->vce.irq); if (r) return r; @@ -447,6 +448,8 @@ static int vce_v3_0_sw_init(void *handle) return r; } + r = amdgpu_vce_entity_init(adev); + return r; } @@ -942,12 +945,16 @@ static void vce_v3_0_set_ring_funcs(struct amdgpu_device *adev) int i; if (adev->asic_type >= CHIP_STONEY) { - for (i = 0; i < adev->vce.num_rings; i++) + for (i = 0; i < adev->vce.num_rings; i++) { adev->vce.ring[i].funcs = &vce_v3_0_ring_vm_funcs; + adev->vce.ring[i].me = i; + } DRM_INFO("VCE enabled in VM mode\n"); } else { - for (i = 0; i < adev->vce.num_rings; i++) + for (i = 0; i < adev->vce.num_rings; i++) { adev->vce.ring[i].funcs = &vce_v3_0_ring_phys_funcs; + adev->vce.ring[i].me = i; + } DRM_INFO("VCE enabled in physical mode\n"); } } diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c index 8fd1b742985a..1c9471890bf7 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c @@ -39,6 +39,8 @@ #include "mmhub/mmhub_1_0_offset.h" #include "mmhub/mmhub_1_0_sh_mask.h" +#include "ivsrcid/vce/irqsrcs_vce_4_0.h" + #define VCE_STATUS_VCPU_REPORT_FW_LOADED_MASK 0x02 #define VCE_V4_0_FW_SIZE (384 * 1024) @@ -60,9 +62,9 @@ static uint64_t vce_v4_0_ring_get_rptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) return RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_RPTR)); - else if (ring == &adev->vce.ring[1]) + else if (ring->me == 1) return RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_RPTR2)); else return RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_RPTR3)); @@ -82,9 +84,9 @@ static uint64_t vce_v4_0_ring_get_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) return adev->wb.wb[ring->wptr_offs]; - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) return RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR)); - else if (ring == &adev->vce.ring[1]) + else if (ring->me == 1) return RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR2)); else return RREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR3)); @@ -108,10 +110,10 @@ static void vce_v4_0_ring_set_wptr(struct amdgpu_ring *ring) return; } - if (ring == &adev->vce.ring[0]) + if (ring->me == 0) WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR), lower_32_bits(ring->wptr)); - else if (ring == &adev->vce.ring[1]) + else if (ring->me == 1) WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_RB_WPTR2), lower_32_bits(ring->wptr)); else @@ -417,6 +419,7 @@ static int vce_v4_0_sw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ring *ring; + unsigned size; int r, i; @@ -436,7 +439,7 @@ static int vce_v4_0_sw_init(void *handle) const struct common_firmware_header *hdr; unsigned size = amdgpu_bo_size(adev->vce.vcpu_bo); - adev->vce.saved_bo = kmalloc(size, GFP_KERNEL); + adev->vce.saved_bo = kvmalloc(size, GFP_KERNEL); if (!adev->vce.saved_bo) return -ENOMEM; @@ -472,6 +475,11 @@ static int vce_v4_0_sw_init(void *handle) return r; } + + r = amdgpu_vce_entity_init(adev); + if (r) + return r; + r = amdgpu_virt_alloc_mm_table(adev); if (r) return r; @@ -488,7 +496,7 @@ static int vce_v4_0_sw_fini(void *handle) amdgpu_virt_free_mm_table(adev); if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - kfree(adev->vce.saved_bo); + kvfree(adev->vce.saved_bo); adev->vce.saved_bo = NULL; } @@ -593,6 +601,7 @@ static int vce_v4_0_resume(void *handle) static void vce_v4_0_mc_resume(struct amdgpu_device *adev) { uint32_t offset, size; + uint64_t tmr_mc_addr; WREG32_P(SOC15_REG_OFFSET(VCE, 0, mmVCE_CLOCK_GATING_A), 0, ~(1 << 16)); WREG32_P(SOC15_REG_OFFSET(VCE, 0, mmVCE_UENC_CLOCK_GATING), 0x1FF000, ~0xFF9FF000); @@ -605,21 +614,25 @@ static void vce_v4_0_mc_resume(struct amdgpu_device *adev) WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_SWAP_CNTL1), 0); WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VM_CTRL), 0); + offset = AMDGPU_VCE_FIRMWARE_OFFSET; + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + tmr_mc_addr = (uint64_t)(adev->firmware.ucode[AMDGPU_UCODE_ID_VCE].tmr_mc_addr_hi) << 32 | + adev->firmware.ucode[AMDGPU_UCODE_ID_VCE].tmr_mc_addr_lo; WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR0), - (adev->firmware.ucode[AMDGPU_UCODE_ID_VCE].mc_addr >> 8)); + (tmr_mc_addr >> 8)); WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_64BIT_BAR0), - (adev->firmware.ucode[AMDGPU_UCODE_ID_VCE].mc_addr >> 40) & 0xff); + (tmr_mc_addr >> 40) & 0xff); + WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET0), 0); } else { WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR0), (adev->vce.gpu_addr >> 8)); WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_64BIT_BAR0), (adev->vce.gpu_addr >> 40) & 0xff); + WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET0), offset & ~0x0f000000); } - offset = AMDGPU_VCE_FIRMWARE_OFFSET; size = VCE_V4_0_FW_SIZE; - WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_OFFSET0), offset & ~0x0f000000); WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_VCPU_CACHE_SIZE0), size); WREG32(SOC15_REG_OFFSET(VCE, 0, mmVCE_LMI_VCPU_CACHE_40BIT_BAR1), (adev->vce.gpu_addr >> 8)); @@ -1088,8 +1101,10 @@ static void vce_v4_0_set_ring_funcs(struct amdgpu_device *adev) { int i; - for (i = 0; i < adev->vce.num_rings; i++) + for (i = 0; i < adev->vce.num_rings; i++) { adev->vce.ring[i].funcs = &vce_v4_0_ring_vm_funcs; + adev->vce.ring[i].me = i; + } DRM_INFO("VCE enabled in VM mode\n"); } diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 29684c3ea4ef..eae90922fdbe 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -35,10 +35,19 @@ #include "mmhub/mmhub_9_1_offset.h" #include "mmhub/mmhub_9_1_sh_mask.h" +#include "ivsrcid/vcn/irqsrcs_vcn_1_0.h" + +#define mmUVD_RBC_XX_IB_REG_CHECK 0x05ab +#define mmUVD_RBC_XX_IB_REG_CHECK_BASE_IDX 1 +#define mmUVD_REG_XX_MASK 0x05ac +#define mmUVD_REG_XX_MASK_BASE_IDX 1 + static int vcn_v1_0_stop(struct amdgpu_device *adev); static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev); static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev); +static void vcn_v1_0_set_jpeg_ring_funcs(struct amdgpu_device *adev); static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev); +static void vcn_v1_0_jpeg_ring_set_patch_ring(struct amdgpu_ring *ring, uint32_t ptr); /** * vcn_v1_0_early_init - set function pointers @@ -55,6 +64,7 @@ static int vcn_v1_0_early_init(void *handle) vcn_v1_0_set_dec_ring_funcs(adev); vcn_v1_0_set_enc_ring_funcs(adev); + vcn_v1_0_set_jpeg_ring_funcs(adev); vcn_v1_0_set_irq_funcs(adev); return 0; @@ -74,22 +84,37 @@ static int vcn_v1_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* VCN DEC TRAP */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, 124, &adev->vcn.irq); + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, VCN_1_0__SRCID__UVD_SYSTEM_MESSAGE_INTERRUPT, &adev->vcn.irq); if (r) return r; /* VCN ENC TRAP */ for (i = 0; i < adev->vcn.num_enc_rings; ++i) { - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, i + 119, + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, i + VCN_1_0__SRCID__UVD_ENC_GENERAL_PURPOSE, &adev->vcn.irq); if (r) return r; } + /* VCN JPEG TRAP */ + r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_VCN, 126, &adev->vcn.irq); + if (r) + return r; + r = amdgpu_vcn_sw_init(adev); if (r) return r; + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + const struct common_firmware_header *hdr; + hdr = (const struct common_firmware_header *)adev->vcn.fw->data; + adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].ucode_id = AMDGPU_UCODE_ID_VCN; + adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].fw = adev->vcn.fw; + adev->firmware.fw_size += + ALIGN(le32_to_cpu(hdr->ucode_size_bytes), PAGE_SIZE); + DRM_INFO("PSP loading VCN firmware\n"); + } + r = amdgpu_vcn_resume(adev); if (r) return r; @@ -108,6 +133,12 @@ static int vcn_v1_0_sw_init(void *handle) return r; } + ring = &adev->vcn.ring_jpeg; + sprintf(ring->name, "vcn_jpeg"); + r = amdgpu_ring_init(adev, ring, 512, &adev->vcn.irq, 0); + if (r) + return r; + return r; } @@ -162,9 +193,18 @@ static int vcn_v1_0_hw_init(void *handle) } } + ring = &adev->vcn.ring_jpeg; + ring->ready = true; + r = amdgpu_ring_test_ring(ring); + if (r) { + ring->ready = false; + goto done; + } + done: if (!r) - DRM_INFO("VCN decode and encode initialized successfully.\n"); + DRM_INFO("VCN decode and encode initialized successfully(under %s).\n", + (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG)?"DPG Mode":"SPG Mode"); return r; } @@ -232,38 +272,52 @@ static int vcn_v1_0_resume(void *handle) } /** - * vcn_v1_0_mc_resume - memory controller programming + * vcn_v1_0_mc_resume_spg_mode - memory controller programming * * @adev: amdgpu_device pointer * * Let the VCN memory controller know it's offsets */ -static void vcn_v1_0_mc_resume(struct amdgpu_device *adev) +static void vcn_v1_0_mc_resume_spg_mode(struct amdgpu_device *adev) { uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4); - - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + uint32_t offset; + + /* cache window 0: fw */ + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo)); + WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi)); + WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0); + offset = 0; + } else { + WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, lower_32_bits(adev->vcn.gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, upper_32_bits(adev->vcn.gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, - AMDGPU_UVD_FIRMWARE_OFFSET >> 3); + offset = size; + WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, + AMDGPU_UVD_FIRMWARE_OFFSET >> 3); + } + WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size); + /* cache window 1: stack */ WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, - lower_32_bits(adev->vcn.gpu_addr + size)); + lower_32_bits(adev->vcn.gpu_addr + offset)); WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, - upper_32_bits(adev->vcn.gpu_addr + size)); + upper_32_bits(adev->vcn.gpu_addr + offset)); WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_HEAP_SIZE); + WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE); + /* cache window 2: context */ WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, - lower_32_bits(adev->vcn.gpu_addr + size + AMDGPU_VCN_HEAP_SIZE)); + lower_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, - upper_32_bits(adev->vcn.gpu_addr + size + AMDGPU_VCN_HEAP_SIZE)); + upper_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE)); WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, - AMDGPU_VCN_STACK_SIZE + (AMDGPU_VCN_SESSION_SIZE * 40)); + WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE); WREG32_SOC15(UVD, 0, mmUVD_UDEC_ADDR_CONFIG, adev->gfx.config.gb_addr_config); @@ -271,6 +325,96 @@ static void vcn_v1_0_mc_resume(struct amdgpu_device *adev) adev->gfx.config.gb_addr_config); WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG, adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_JPEG_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); +} + +static void vcn_v1_0_mc_resume_dpg_mode(struct amdgpu_device *adev) +{ + uint32_t size = AMDGPU_GPU_PAGE_ALIGN(adev->vcn.fw->size + 4); + uint32_t offset; + + /* cache window 0: fw */ + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_lo), + 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + (adev->firmware.ucode[AMDGPU_UCODE_ID_VCN].tmr_mc_addr_hi), + 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, 0, + 0xFFFFFFFF, 0); + offset = 0; + } else { + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + lower_32_bits(adev->vcn.gpu_addr), 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + upper_32_bits(adev->vcn.gpu_addr), 0xFFFFFFFF, 0); + offset = size; + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, + AMDGPU_UVD_FIRMWARE_OFFSET >> 3, 0xFFFFFFFF, 0); + } + + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size, 0xFFFFFFFF, 0); + + /* cache window 1: stack */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, + lower_32_bits(adev->vcn.gpu_addr + offset), 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, + upper_32_bits(adev->vcn.gpu_addr + offset), 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, 0, + 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_VCN_STACK_SIZE, + 0xFFFFFFFF, 0); + + /* cache window 2: context */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, + lower_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), + 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, + upper_32_bits(adev->vcn.gpu_addr + offset + AMDGPU_VCN_STACK_SIZE), + 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, 0, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, AMDGPU_VCN_CONTEXT_SIZE, + 0xFFFFFFFF, 0); + + /* VCN global tiling registers */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_UDEC_DBW_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_CURR_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_CURR_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_RECON1_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_RECON1_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_REF_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MIF_REF_UV_ADDR_CONFIG, + adev->gfx.config.gb_addr_config, 0xFFFFFFFF, 0); } /** @@ -473,6 +617,60 @@ static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev) WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data); } +static void vcn_v1_0_clock_gating_dpg_mode(struct amdgpu_device *adev, uint8_t sram_sel) +{ + uint32_t reg_data = 0; + + /* disable JPEG CGC */ + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) + reg_data = 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + else + reg_data = 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + reg_data |= 1 << JPEG_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; + reg_data |= 4 << JPEG_CGC_CTRL__CLK_OFF_DELAY__SHIFT; + WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel); + + WREG32_SOC15_DPG_MODE(UVD, 0, mmJPEG_CGC_GATE, 0, 0xFFFFFFFF, sram_sel); + + /* enable sw clock gating control */ + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) + reg_data = 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + else + reg_data = 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; + reg_data |= 1 << UVD_CGC_CTRL__CLK_GATE_DLY_TIMER__SHIFT; + reg_data |= 4 << UVD_CGC_CTRL__CLK_OFF_DELAY__SHIFT; + reg_data &= ~(UVD_CGC_CTRL__UDEC_RE_MODE_MASK | + UVD_CGC_CTRL__UDEC_CM_MODE_MASK | + UVD_CGC_CTRL__UDEC_IT_MODE_MASK | + UVD_CGC_CTRL__UDEC_DB_MODE_MASK | + UVD_CGC_CTRL__UDEC_MP_MODE_MASK | + UVD_CGC_CTRL__SYS_MODE_MASK | + UVD_CGC_CTRL__UDEC_MODE_MASK | + UVD_CGC_CTRL__MPEG2_MODE_MASK | + UVD_CGC_CTRL__REGS_MODE_MASK | + UVD_CGC_CTRL__RBC_MODE_MASK | + UVD_CGC_CTRL__LMI_MC_MODE_MASK | + UVD_CGC_CTRL__LMI_UMC_MODE_MASK | + UVD_CGC_CTRL__IDCT_MODE_MASK | + UVD_CGC_CTRL__MPRD_MODE_MASK | + UVD_CGC_CTRL__MPC_MODE_MASK | + UVD_CGC_CTRL__LBSI_MODE_MASK | + UVD_CGC_CTRL__LRBBM_MODE_MASK | + UVD_CGC_CTRL__WCB_MODE_MASK | + UVD_CGC_CTRL__VCPU_MODE_MASK | + UVD_CGC_CTRL__SCPU_MODE_MASK); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_CTRL, reg_data, 0xFFFFFFFF, sram_sel); + + /* turn off clock gating */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_CGC_GATE, 0, 0xFFFFFFFF, sram_sel); + + /* turn on SUVD clock gating */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_GATE, 1, 0xFFFFFFFF, sram_sel); + + /* turn on sw mode in UVD_SUVD_CGC_CTRL */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SUVD_CGC_CTRL, 0, 0xFFFFFFFF, sram_sel); +} + static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev) { uint32_t data = 0; @@ -568,7 +766,7 @@ static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev) * * Setup and start the VCN block */ -static int vcn_v1_0_start(struct amdgpu_device *adev) +static int vcn_v1_0_start_spg_mode(struct amdgpu_device *adev) { struct amdgpu_ring *ring = &adev->vcn.ring_dec; uint32_t rb_bufsz, tmp; @@ -578,9 +776,11 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) /* disable byte swapping */ lmi_swap_cntl = 0; - vcn_v1_0_mc_resume(adev); - vcn_1_0_disable_static_power_gating(adev); + + tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) | UVD_STATUS__UVD_BUSY; + WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp); + /* disable clock gating */ vcn_v1_0_disable_clock_gating(adev); @@ -588,32 +788,13 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0, ~UVD_MASTINT_EN__VCPU_EN_MASK); - /* stall UMC and register bus before resetting VCPU */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), - UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - mdelay(1); - - /* put LMI, VCPU, RBC etc... into reset */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, - UVD_SOFT_RESET__LMI_SOFT_RESET_MASK | - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | - UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK | - UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | - UVD_SOFT_RESET__CSM_SOFT_RESET_MASK | - UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | - UVD_SOFT_RESET__TAP_SOFT_RESET_MASK | - UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK); - mdelay(5); - /* initialize VCN memory controller */ - WREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL, - (0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | - UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | - UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | - UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | - UVD_LMI_CTRL__REQ_MODE_MASK | - 0x00100000L); + tmp = RREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL); + WREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL, tmp | + UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__MASK_MC_URGENT_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK); #ifdef __BIG_ENDIAN /* swap (8 in 32) RB and IB */ @@ -621,41 +802,61 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) #endif WREG32_SOC15(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA0, 0x40c2040); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA1, 0x0); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB0, 0x40c2040); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB1, 0x0); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_ALU, 0); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUX, 0x88); + tmp = RREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL); + tmp &= ~UVD_MPC_CNTL__REPLACEMENT_MODE_MASK; + tmp |= 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT; + WREG32_SOC15(UVD, 0, mmUVD_MPC_CNTL, tmp); + + WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA0, + ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT))); + + WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB0, + ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT))); + + WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUX, + ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) | + (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT))); - /* take all subblocks out of reset, except VCPU */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); - mdelay(5); + vcn_v1_0_mc_resume_spg_mode(adev); + + WREG32_SOC15(UVD, 0, mmUVD_REG_XX_MASK, 0x10); + WREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK, + RREG32_SOC15(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK) | 0x3); /* enable VCPU clock */ - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, - UVD_VCPU_CNTL__CLK_EN_MASK); + WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, UVD_VCPU_CNTL__CLK_EN_MASK); + + /* boot up the VCPU */ + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0, + ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); /* enable UMC */ WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0, ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - /* boot up the VCPU */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, 0); - mdelay(10); + tmp = RREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET); + tmp &= ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK; + tmp &= ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK; + WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, tmp); for (i = 0; i < 10; ++i) { uint32_t status; for (j = 0; j < 100; ++j) { status = RREG32_SOC15(UVD, 0, mmUVD_STATUS); - if (status & 2) + if (status & UVD_STATUS__IDLE) break; mdelay(10); } r = 0; - if (status & 2) + if (status & UVD_STATUS__IDLE) break; DRM_ERROR("VCN decode not responding, trying to reset the VCPU!!!\n"); @@ -675,19 +876,22 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) } /* enable master interrupt */ WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), - (UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), - ~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK)); + UVD_MASTINT_EN__VCPU_EN_MASK, ~UVD_MASTINT_EN__VCPU_EN_MASK); - /* clear the bit 4 of VCN_STATUS */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0, - ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT)); + /* enable system interrupt for JRBC, TODO: move to set interrupt*/ + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SYS_INT_EN), + UVD_SYS_INT_EN__UVD_JRBC_EN_MASK, + ~UVD_SYS_INT_EN__UVD_JRBC_EN_MASK); + + /* clear the busy bit of UVD_STATUS */ + tmp = RREG32_SOC15(UVD, 0, mmUVD_STATUS) & ~UVD_STATUS__UVD_BUSY; + WREG32_SOC15(UVD, 0, mmUVD_STATUS, tmp); /* force RBC into idle state */ rb_bufsz = order_base_2(ring->ring_size); tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp); @@ -708,6 +912,8 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) /* Initialize the ring buffer's read and write pointers */ WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0); + WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0); + ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR); WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); @@ -729,9 +935,186 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4); + ring = &adev->vcn.ring_jpeg; + WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_VMID, 0); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, UVD_JRBC_RB_CNTL__RB_NO_FETCH_MASK | + UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); + WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_LOW, lower_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, 0, mmUVD_LMI_JRBC_RB_64BIT_BAR_HIGH, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR, 0); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, 0); + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_CNTL, UVD_JRBC_RB_CNTL__RB_RPTR_WR_EN_MASK); + + /* initialize wptr */ + ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); + + /* copy patch commands to the jpeg ring */ + vcn_v1_0_jpeg_ring_set_patch_ring(ring, + (ring->wptr + ring->max_dw * amdgpu_sched_hw_submission)); + return 0; } +static int vcn_v1_0_start_dpg_mode(struct amdgpu_device *adev) +{ + struct amdgpu_ring *ring = &adev->vcn.ring_dec; + uint32_t rb_bufsz, tmp; + uint32_t lmi_swap_cntl; + + /* disable byte swapping */ + lmi_swap_cntl = 0; + + vcn_1_0_enable_static_power_gating(adev); + + /* enable dynamic power gating mode */ + tmp = RREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS); + tmp |= UVD_POWER_STATUS__UVD_PG_MODE_MASK; + tmp |= UVD_POWER_STATUS__UVD_PG_EN_MASK; + WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, tmp); + + /* enable clock gating */ + vcn_v1_0_clock_gating_dpg_mode(adev, 0); + + /* enable VCPU clock */ + tmp = (0xFF << UVD_VCPU_CNTL__PRB_TIMEOUT_VAL__SHIFT); + tmp |= UVD_VCPU_CNTL__CLK_EN_MASK; + tmp |= UVD_VCPU_CNTL__MIF_WR_LOW_THRESHOLD_BP_MASK; + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_VCPU_CNTL, tmp, 0xFFFFFFFF, 0); + + /* disable interupt */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN, + 0, UVD_MASTINT_EN__VCPU_EN_MASK, 0); + + /* initialize VCN memory controller */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL, + (8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | + UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__REQ_MODE_MASK | + UVD_LMI_CTRL__CRC_RESET_MASK | + UVD_LMI_CTRL__MASK_MC_URGENT_MASK | + 0x00100000L, 0xFFFFFFFF, 0); + +#ifdef __BIG_ENDIAN + /* swap (8 in 32) RB and IB */ + lmi_swap_cntl = 0xa; +#endif + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl, 0xFFFFFFFF, 0); + + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_CNTL, + 0x2 << UVD_MPC_CNTL__REPLACEMENT_MODE__SHIFT, 0xFFFFFFFF, 0); + + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXA0, + ((0x1 << UVD_MPC_SET_MUXA0__VARA_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXA0__VARA_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXA0__VARA_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXA0__VARA_4__SHIFT)), 0xFFFFFFFF, 0); + + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUXB0, + ((0x1 << UVD_MPC_SET_MUXB0__VARB_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUXB0__VARB_2__SHIFT) | + (0x3 << UVD_MPC_SET_MUXB0__VARB_3__SHIFT) | + (0x4 << UVD_MPC_SET_MUXB0__VARB_4__SHIFT)), 0xFFFFFFFF, 0); + + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MPC_SET_MUX, + ((0x0 << UVD_MPC_SET_MUX__SET_0__SHIFT) | + (0x1 << UVD_MPC_SET_MUX__SET_1__SHIFT) | + (0x2 << UVD_MPC_SET_MUX__SET_2__SHIFT)), 0xFFFFFFFF, 0); + + vcn_v1_0_mc_resume_dpg_mode(adev); + + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_REG_XX_MASK, 0x10, 0xFFFFFFFF, 0); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_RBC_XX_IB_REG_CHECK, 0x3, 0xFFFFFFFF, 0); + + /* boot up the VCPU */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SOFT_RESET, 0, 0xFFFFFFFF, 0); + + /* enable UMC */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL2, + 0x1F << UVD_LMI_CTRL2__RE_OFLD_MIF_WR_REQ_NUM__SHIFT, + 0xFFFFFFFF, 0); + + /* enable master interrupt */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_MASTINT_EN, + UVD_MASTINT_EN__VCPU_EN_MASK, UVD_MASTINT_EN__VCPU_EN_MASK, 0); + + vcn_v1_0_clock_gating_dpg_mode(adev, 1); + /* setup mmUVD_LMI_CTRL */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_LMI_CTRL, + (8 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | + UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__REQ_MODE_MASK | + UVD_LMI_CTRL__CRC_RESET_MASK | + UVD_LMI_CTRL__MASK_MC_URGENT_MASK | + 0x00100000L, 0xFFFFFFFF, 1); + + tmp = adev->gfx.config.gb_addr_config; + /* setup VCN global tiling registers */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1); + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_JPEG_UV_ADDR_CONFIG, tmp, 0xFFFFFFFF, 1); + + /* enable System Interrupt for JRBC */ + WREG32_SOC15_DPG_MODE(UVD, 0, mmUVD_SYS_INT_EN, + UVD_SYS_INT_EN__UVD_JRBC_EN_MASK, 0xFFFFFFFF, 1); + + /* force RBC into idle state */ + rb_bufsz = order_base_2(ring->ring_size); + tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp); + + /* set the write pointer delay */ + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0); + + /* set the wb address */ + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR, + (upper_32_bits(ring->gpu_addr) >> 2)); + + /* programm the RB_BASE for ring buffer */ + WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW, + lower_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH, + upper_32_bits(ring->gpu_addr)); + + /* Initialize the ring buffer's read and write pointers */ + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0); + + WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, 0); + + ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR); + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, + lower_32_bits(ring->wptr)); + + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0, + ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); + + /* initialize wptr */ + ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); + + /* copy patch commands to the jpeg ring */ + vcn_v1_0_jpeg_ring_set_patch_ring(ring, + (ring->wptr + ring->max_dw * amdgpu_sched_hw_submission)); + + return 0; +} + +static int vcn_v1_0_start(struct amdgpu_device *adev) +{ + int r; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) + r = vcn_v1_0_start_dpg_mode(adev); + else + r = vcn_v1_0_start_spg_mode(adev); + return r; +} + /** * vcn_v1_0_stop - stop VCN block * @@ -739,41 +1122,90 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) * * stop the VCN block */ -static int vcn_v1_0_stop(struct amdgpu_device *adev) +static int vcn_v1_0_stop_spg_mode(struct amdgpu_device *adev) { - /* force RBC into idle state */ - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, 0x11010101); + int ret_code, tmp; - /* Stall UMC and register bus before resetting VCPU */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), - UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - mdelay(1); + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_STATUS, UVD_STATUS__IDLE, 0x7, ret_code); + + tmp = UVD_LMI_STATUS__VCPU_LMI_WRITE_CLEAN_MASK | + UVD_LMI_STATUS__READ_CLEAN_MASK | + UVD_LMI_STATUS__WRITE_CLEAN_MASK | + UVD_LMI_STATUS__WRITE_CLEAN_RAW_MASK; + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp, ret_code); /* put VCPU into reset */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); - mdelay(5); + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, + ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + + tmp = UVD_LMI_STATUS__UMC_READ_CLEAN_RAW_MASK | + UVD_LMI_STATUS__UMC_WRITE_CLEAN_RAW_MASK; + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_LMI_STATUS, tmp, tmp, ret_code); /* disable VCPU clock */ - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, 0x0); + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL), 0, + ~UVD_VCPU_CNTL__CLK_EN_MASK); - /* Unstall UMC and register bus */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0, - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + /* reset LMI UMC/LMI */ + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), + UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK, + ~UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK); + + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), + UVD_SOFT_RESET__LMI_SOFT_RESET_MASK, + ~UVD_SOFT_RESET__LMI_SOFT_RESET_MASK); - WREG32_SOC15(VCN, 0, mmUVD_STATUS, 0); + WREG32_SOC15(UVD, 0, mmUVD_STATUS, 0); vcn_v1_0_enable_clock_gating(adev); vcn_1_0_enable_static_power_gating(adev); return 0; } +static int vcn_v1_0_stop_dpg_mode(struct amdgpu_device *adev) +{ + int ret_code = 0; + + /* Wait for power status to be UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF */ + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); + + if (!ret_code) { + int tmp = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR) & 0x7FFFFFFF; + /* wait for read ptr to be equal to write ptr */ + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_RBC_RB_RPTR, tmp, 0xFFFFFFFF, ret_code); + + SOC15_WAIT_ON_RREG(UVD, 0, mmUVD_POWER_STATUS, + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF, + UVD_POWER_STATUS__UVD_POWER_STATUS_MASK, ret_code); + } + + /* disable dynamic power gating mode */ + WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0, + ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + + return 0; +} + +static int vcn_v1_0_stop(struct amdgpu_device *adev) +{ + int r; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) + r = vcn_v1_0_stop_dpg_mode(adev); + else + r = vcn_v1_0_stop_spg_mode(adev); + + return r; +} + static bool vcn_v1_0_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == 0x2); + return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == UVD_STATUS__IDLE); } static int vcn_v1_0_wait_for_idle(void *handle) @@ -781,7 +1213,8 @@ static int vcn_v1_0_wait_for_idle(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int ret = 0; - SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, 0x2, 0x2, ret); + SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, UVD_STATUS__IDLE, + UVD_STATUS__IDLE, ret); return ret; } @@ -843,6 +1276,10 @@ static void vcn_v1_0_dec_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; + if (adev->pg_flags & AMD_PG_SUPPORT_VCN_DPG) + WREG32_SOC15(UVD, 0, mmUVD_SCRATCH2, + lower_32_bits(ring->wptr) | 0x80000000); + WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); } @@ -1126,6 +1563,387 @@ static void vcn_v1_0_enc_ring_emit_wreg(struct amdgpu_ring *ring, amdgpu_ring_write(ring, val); } + +/** + * vcn_v1_0_jpeg_ring_get_rptr - get read pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware read pointer + */ +static uint64_t vcn_v1_0_jpeg_ring_get_rptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + return RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_RPTR); +} + +/** + * vcn_v1_0_jpeg_ring_get_wptr - get write pointer + * + * @ring: amdgpu_ring pointer + * + * Returns the current hardware write pointer + */ +static uint64_t vcn_v1_0_jpeg_ring_get_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + return RREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR); +} + +/** + * vcn_v1_0_jpeg_ring_set_wptr - set write pointer + * + * @ring: amdgpu_ring pointer + * + * Commits the write pointer to the hardware + */ +static void vcn_v1_0_jpeg_ring_set_wptr(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + WREG32_SOC15(UVD, 0, mmUVD_JRBC_RB_WPTR, lower_32_bits(ring->wptr)); +} + +/** + * vcn_v1_0_jpeg_ring_insert_start - insert a start command + * + * @ring: amdgpu_ring pointer + * + * Write a start command to the ring. + */ +static void vcn_v1_0_jpeg_ring_insert_start(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x68e04); + + amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x80010000); +} + +/** + * vcn_v1_0_jpeg_ring_insert_end - insert a end command + * + * @ring: amdgpu_ring pointer + * + * Write a end command to the ring. + */ +static void vcn_v1_0_jpeg_ring_insert_end(struct amdgpu_ring *ring) +{ + struct amdgpu_device *adev = ring->adev; + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x68e04); + + amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x00010000); +} + +/** + * vcn_v1_0_jpeg_ring_emit_fence - emit an fence & trap command + * + * @ring: amdgpu_ring pointer + * @fence: fence to emit + * + * Write a fence and a trap command to the ring. + */ +static void vcn_v1_0_jpeg_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, + unsigned flags) +{ + struct amdgpu_device *adev = ring->adev; + + WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_DATA0), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, seq); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_DATA1), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, seq); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, lower_32_bits(addr)); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_WR_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_CMD), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x8); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JPEG_GPCOM_CMD), 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE4)); + amdgpu_ring_write(ring, 0); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x01400200); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, seq); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, lower_32_bits(addr)); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, upper_32_bits(addr)); + + amdgpu_ring_write(ring, + PACKETJ(0, 0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE2)); + amdgpu_ring_write(ring, 0xffffffff); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x3fbc); + + amdgpu_ring_write(ring, + PACKETJ(0, 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x1); + + /* emit trap */ + amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE7)); + amdgpu_ring_write(ring, 0); +} + +/** + * vcn_v1_0_jpeg_ring_emit_ib - execute indirect buffer + * + * @ring: amdgpu_ring pointer + * @ib: indirect buffer to execute + * + * Write ring commands to execute the indirect buffer. + */ +static void vcn_v1_0_jpeg_ring_emit_ib(struct amdgpu_ring *ring, + struct amdgpu_ib *ib, + unsigned vmid, bool ctx_switch) +{ + struct amdgpu_device *adev = ring->adev; + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_VMID), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, (vmid | (vmid << 4))); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JPEG_VMID), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, (vmid | (vmid << 4))); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_IB_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_IB_SIZE), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, ib->length_dw); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, lower_32_bits(ring->gpu_addr)); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, upper_32_bits(ring->gpu_addr)); + + amdgpu_ring_write(ring, + PACKETJ(0, 0, PACKETJ_CONDITION_CHECK0, PACKETJ_TYPE2)); + amdgpu_ring_write(ring, 0); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x01400200); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x2); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_STATUS), 0, PACKETJ_CONDITION_CHECK3, PACKETJ_TYPE3)); + amdgpu_ring_write(ring, 0x2); +} + +static void vcn_v1_0_jpeg_ring_emit_reg_wait(struct amdgpu_ring *ring, + uint32_t reg, uint32_t val, + uint32_t mask) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t reg_offset = (reg << 2); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, 0x01400200); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0)); + amdgpu_ring_write(ring, val); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); + if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || + ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, + PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3)); + } else { + amdgpu_ring_write(ring, reg_offset); + amdgpu_ring_write(ring, + PACKETJ(0, 0, 0, PACKETJ_TYPE3)); + } + amdgpu_ring_write(ring, mask); +} + +static void vcn_v1_0_jpeg_ring_emit_vm_flush(struct amdgpu_ring *ring, + unsigned vmid, uint64_t pd_addr) +{ + struct amdgpu_vmhub *hub = &ring->adev->vmhub[ring->funcs->vmhub]; + uint32_t data0, data1, mask; + + pd_addr = amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); + + /* wait for register write */ + data0 = hub->ctx0_ptb_addr_lo32 + vmid * 2; + data1 = lower_32_bits(pd_addr); + mask = 0xffffffff; + vcn_v1_0_jpeg_ring_emit_reg_wait(ring, data0, data1, mask); +} + +static void vcn_v1_0_jpeg_ring_emit_wreg(struct amdgpu_ring *ring, + uint32_t reg, uint32_t val) +{ + struct amdgpu_device *adev = ring->adev; + uint32_t reg_offset = (reg << 2); + + amdgpu_ring_write(ring, + PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0)); + if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || + ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, + PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0)); + } else { + amdgpu_ring_write(ring, reg_offset); + amdgpu_ring_write(ring, + PACKETJ(0, 0, 0, PACKETJ_TYPE0)); + } + amdgpu_ring_write(ring, val); +} + +static void vcn_v1_0_jpeg_ring_nop(struct amdgpu_ring *ring, uint32_t count) +{ + int i; + + WARN_ON(ring->wptr % 2 || count % 2); + + for (i = 0; i < count / 2; i++) { + amdgpu_ring_write(ring, PACKETJ(0, 0, 0, PACKETJ_TYPE6)); + amdgpu_ring_write(ring, 0); + } +} + +static void vcn_v1_0_jpeg_ring_patch_wreg(struct amdgpu_ring *ring, uint32_t *ptr, uint32_t reg_offset, uint32_t val) +{ + struct amdgpu_device *adev = ring->adev; + ring->ring[(*ptr)++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0); + if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || + ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { + ring->ring[(*ptr)++] = 0; + ring->ring[(*ptr)++] = PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE0); + } else { + ring->ring[(*ptr)++] = reg_offset; + ring->ring[(*ptr)++] = PACKETJ(0, 0, 0, PACKETJ_TYPE0); + } + ring->ring[(*ptr)++] = val; +} + +static void vcn_v1_0_jpeg_ring_set_patch_ring(struct amdgpu_ring *ring, uint32_t ptr) +{ + struct amdgpu_device *adev = ring->adev; + + uint32_t reg, reg_offset, val, mask, i; + + // 1st: program mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW + reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_LOW); + reg_offset = (reg << 2); + val = lower_32_bits(ring->gpu_addr); + vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); + + // 2nd: program mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH + reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_JRBC_RB_MEM_RD_64BIT_BAR_HIGH); + reg_offset = (reg << 2); + val = upper_32_bits(ring->gpu_addr); + vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); + + // 3rd to 5th: issue MEM_READ commands + for (i = 0; i <= 2; i++) { + ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE2); + ring->ring[ptr++] = 0; + } + + // 6th: program mmUVD_JRBC_RB_CNTL register to enable NO_FETCH and RPTR write ability + reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL); + reg_offset = (reg << 2); + val = 0x13; + vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); + + // 7th: program mmUVD_JRBC_RB_REF_DATA + reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA); + reg_offset = (reg << 2); + val = 0x1; + vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); + + // 8th: issue conditional register read mmUVD_JRBC_RB_CNTL + reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL); + reg_offset = (reg << 2); + val = 0x1; + mask = 0x1; + + ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_COND_RD_TIMER), 0, 0, PACKETJ_TYPE0); + ring->ring[ptr++] = 0x01400200; + ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_REF_DATA), 0, 0, PACKETJ_TYPE0); + ring->ring[ptr++] = val; + ring->ring[ptr++] = PACKETJ(SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_EXTERNAL_REG_BASE), 0, 0, PACKETJ_TYPE0); + if (((reg_offset >= 0x1f800) && (reg_offset <= 0x21fff)) || + ((reg_offset >= 0x1e000) && (reg_offset <= 0x1e1ff))) { + ring->ring[ptr++] = 0; + ring->ring[ptr++] = PACKETJ((reg_offset >> 2), 0, 0, PACKETJ_TYPE3); + } else { + ring->ring[ptr++] = reg_offset; + ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE3); + } + ring->ring[ptr++] = mask; + + //9th to 21st: insert no-op + for (i = 0; i <= 12; i++) { + ring->ring[ptr++] = PACKETJ(0, 0, 0, PACKETJ_TYPE6); + ring->ring[ptr++] = 0; + } + + //22nd: reset mmUVD_JRBC_RB_RPTR + reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_RPTR); + reg_offset = (reg << 2); + val = 0; + vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); + + //23rd: program mmUVD_JRBC_RB_CNTL to disable no_fetch + reg = SOC15_REG_OFFSET(UVD, 0, mmUVD_JRBC_RB_CNTL); + reg_offset = (reg << 2); + val = 0x12; + vcn_v1_0_jpeg_ring_patch_wreg(ring, &ptr, reg_offset, val); +} + static int vcn_v1_0_set_interrupt_state(struct amdgpu_device *adev, struct amdgpu_irq_src *source, unsigned type, @@ -1150,6 +1968,9 @@ static int vcn_v1_0_process_interrupt(struct amdgpu_device *adev, case 120: amdgpu_fence_process(&adev->vcn.ring_enc[1]); break; + case 126: + amdgpu_fence_process(&adev->vcn.ring_jpeg); + break; default: DRM_ERROR("Unhandled interrupt: %d %d\n", entry->src_id, entry->src_data[0]); @@ -1182,12 +2003,20 @@ static int vcn_v1_0_set_powergating_state(void *handle, * revisit this when there is a cleaner line between * the smc and the hw blocks */ + int ret; struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if(state == adev->vcn.cur_state) + return 0; + if (state == AMD_PG_STATE_GATE) - return vcn_v1_0_stop(adev); + ret = vcn_v1_0_stop(adev); else - return vcn_v1_0_start(adev); + ret = vcn_v1_0_start(adev); + + if(!ret) + adev->vcn.cur_state = state; + return ret; } static const struct amd_ip_funcs vcn_v1_0_ip_funcs = { @@ -1273,6 +2102,40 @@ static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = { .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; +static const struct amdgpu_ring_funcs vcn_v1_0_jpeg_ring_vm_funcs = { + .type = AMDGPU_RING_TYPE_VCN_JPEG, + .align_mask = 0xf, + .nop = PACKET0(0x81ff, 0), + .support_64bit_ptrs = false, + .vmhub = AMDGPU_MMHUB, + .extra_dw = 64, + .get_rptr = vcn_v1_0_jpeg_ring_get_rptr, + .get_wptr = vcn_v1_0_jpeg_ring_get_wptr, + .set_wptr = vcn_v1_0_jpeg_ring_set_wptr, + .emit_frame_size = + 6 + 6 + /* hdp invalidate / flush */ + SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + + 8 + /* vcn_v1_0_jpeg_ring_emit_vm_flush */ + 26 + 26 + /* vcn_v1_0_jpeg_ring_emit_fence x2 vm fence */ + 6, + .emit_ib_size = 22, /* vcn_v1_0_jpeg_ring_emit_ib */ + .emit_ib = vcn_v1_0_jpeg_ring_emit_ib, + .emit_fence = vcn_v1_0_jpeg_ring_emit_fence, + .emit_vm_flush = vcn_v1_0_jpeg_ring_emit_vm_flush, + .test_ring = amdgpu_vcn_jpeg_ring_test_ring, + .test_ib = amdgpu_vcn_jpeg_ring_test_ib, + .insert_nop = vcn_v1_0_jpeg_ring_nop, + .insert_start = vcn_v1_0_jpeg_ring_insert_start, + .insert_end = vcn_v1_0_jpeg_ring_insert_end, + .pad_ib = amdgpu_ring_generic_pad_ib, + .begin_use = amdgpu_vcn_ring_begin_use, + .end_use = amdgpu_vcn_ring_end_use, + .emit_wreg = vcn_v1_0_jpeg_ring_emit_wreg, + .emit_reg_wait = vcn_v1_0_jpeg_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, +}; + static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev) { adev->vcn.ring_dec.funcs = &vcn_v1_0_dec_ring_vm_funcs; @@ -1289,6 +2152,12 @@ static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev) DRM_INFO("VCN encode is enabled in VM mode\n"); } +static void vcn_v1_0_set_jpeg_ring_funcs(struct amdgpu_device *adev) +{ + adev->vcn.ring_jpeg.funcs = &vcn_v1_0_jpeg_ring_vm_funcs; + DRM_INFO("VCN jpeg decode is enabled in VM mode\n"); +} + static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = { .set = vcn_v1_0_set_interrupt_state, .process = vcn_v1_0_process_interrupt, @@ -1296,7 +2165,7 @@ static const struct amdgpu_irq_src_funcs vcn_v1_0_irq_funcs = { static void vcn_v1_0_set_irq_funcs(struct amdgpu_device *adev) { - adev->vcn.irq.num_types = adev->vcn.num_enc_rings + 1; + adev->vcn.irq.num_types = adev->vcn.num_enc_rings + 2; adev->vcn.irq.funcs = &vcn_v1_0_irq_funcs; } diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c index 5ae5ed2e62d6..a99f71797aa3 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_ih.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_ih.c @@ -265,35 +265,36 @@ static bool vega10_ih_prescreen_iv(struct amdgpu_device *adev) return true; } - addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12); - key = AMDGPU_VM_FAULT(pasid, addr); - r = amdgpu_ih_add_fault(adev, key); - - /* Hash table is full or the fault is already being processed, - * ignore further page faults - */ - if (r != 0) - goto ignore_iv; - /* Track retry faults in per-VM fault FIFO. */ spin_lock(&adev->vm_manager.pasid_lock); vm = idr_find(&adev->vm_manager.pasid_idr, pasid); + addr = ((u64)(dw5 & 0xf) << 44) | ((u64)dw4 << 12); + key = AMDGPU_VM_FAULT(pasid, addr); if (!vm) { /* VM not found, process it normally */ spin_unlock(&adev->vm_manager.pasid_lock); - amdgpu_ih_clear_fault(adev, key); return true; + } else { + r = amdgpu_vm_add_fault(vm->fault_hash, key); + + /* Hash table is full or the fault is already being processed, + * ignore further page faults + */ + if (r != 0) { + spin_unlock(&adev->vm_manager.pasid_lock); + goto ignore_iv; + } } /* No locking required with single writer and single reader */ r = kfifo_put(&vm->faults, key); if (!r) { /* FIFO is full. Ignore it until there is space */ + amdgpu_vm_clear_fault(vm->fault_hash, key); spin_unlock(&adev->vm_manager.pasid_lock); - amdgpu_ih_clear_fault(adev, key); goto ignore_iv; } - spin_unlock(&adev->vm_manager.pasid_lock); + spin_unlock(&adev->vm_manager.pasid_lock); /* It's the first fault for this address, process it normally */ return true; @@ -379,21 +380,13 @@ static int vega10_ih_sw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - r = amdgpu_ih_ring_init(adev, 256 * 1024, true); + r = amdgpu_ih_ring_init(adev, &adev->irq.ih, 256 * 1024, true); if (r) return r; adev->irq.ih.use_doorbell = true; adev->irq.ih.doorbell_index = AMDGPU_DOORBELL64_IH << 1; - adev->irq.ih.faults = kmalloc(sizeof(*adev->irq.ih.faults), GFP_KERNEL); - if (!adev->irq.ih.faults) - return -ENOMEM; - INIT_CHASH_TABLE(adev->irq.ih.faults->hash, - AMDGPU_PAGEFAULT_HASH_BITS, 8, 0); - spin_lock_init(&adev->irq.ih.faults->lock); - adev->irq.ih.faults->count = 0; - r = amdgpu_irq_init(adev); return r; @@ -404,10 +397,7 @@ static int vega10_ih_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_irq_fini(adev); - amdgpu_ih_ring_fini(adev); - - kfree(adev->irq.ih.faults); - adev->irq.ih.faults = NULL; + amdgpu_ih_ring_fini(adev, &adev->irq.ih); return 0; } @@ -504,8 +494,7 @@ static const struct amdgpu_ih_funcs vega10_ih_funcs = { static void vega10_ih_set_interrupt_funcs(struct amdgpu_device *adev) { - if (adev->irq.ih_funcs == NULL) - adev->irq.ih_funcs = &vega10_ih_funcs; + adev->irq.ih_funcs = &vega10_ih_funcs; } const struct amdgpu_ip_block_version vega10_ih_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c b/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c index 45aafca7f315..c5c9b2bc190d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c +++ b/drivers/gpu/drm/amd/amdgpu/vega10_reg_init.c @@ -51,6 +51,7 @@ int vega10_reg_base_init(struct amdgpu_device *adev) adev->reg_offset[PWR_HWIP][i] = (uint32_t *)(&(PWR_BASE.instance[i])); adev->reg_offset[NBIF_HWIP][i] = (uint32_t *)(&(NBIF_BASE.instance[i])); adev->reg_offset[THM_HWIP][i] = (uint32_t *)(&(THM_BASE.instance[i])); + adev->reg_offset[CLK_HWIP][i] = (uint32_t *)(&(CLK_BASE.instance[i])); } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c b/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c index 52778de93ab0..2d4473557b0d 100644 --- a/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c +++ b/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c @@ -38,6 +38,7 @@ int vega20_reg_base_init(struct amdgpu_device *adev) adev->reg_offset[ATHUB_HWIP][i] = (uint32_t *)(&(ATHUB_BASE.instance[i])); adev->reg_offset[NBIO_HWIP][i] = (uint32_t *)(&(NBIO_BASE.instance[i])); adev->reg_offset[MP0_HWIP][i] = (uint32_t *)(&(MP0_BASE.instance[i])); + adev->reg_offset[MP1_HWIP][i] = (uint32_t *)(&(MP1_BASE.instance[i])); adev->reg_offset[UVD_HWIP][i] = (uint32_t *)(&(UVD_BASE.instance[i])); adev->reg_offset[VCE_HWIP][i] = (uint32_t *)(&(VCE_BASE.instance[i])); adev->reg_offset[DF_HWIP][i] = (uint32_t *)(&(DF_BASE.instance[i])); @@ -46,6 +47,8 @@ int vega20_reg_base_init(struct amdgpu_device *adev) adev->reg_offset[SDMA0_HWIP][i] = (uint32_t *)(&(SDMA0_BASE.instance[i])); adev->reg_offset[SDMA1_HWIP][i] = (uint32_t *)(&(SDMA1_BASE.instance[i])); adev->reg_offset[SMUIO_HWIP][i] = (uint32_t *)(&(SMUIO_BASE.instance[i])); + adev->reg_offset[NBIF_HWIP][i] = (uint32_t *)(&(NBIO_BASE.instance[i])); + adev->reg_offset[THM_HWIP][i] = (uint32_t *)(&(THM_BASE.instance[i])); } return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 4ac1288ab7df..07880d35e9de 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -112,8 +112,8 @@ static u32 vi_smc_rreg(struct amdgpu_device *adev, u32 reg) u32 r; spin_lock_irqsave(&adev->smc_idx_lock, flags); - WREG32(mmSMC_IND_INDEX_11, (reg)); - r = RREG32(mmSMC_IND_DATA_11); + WREG32_NO_KIQ(mmSMC_IND_INDEX_11, (reg)); + r = RREG32_NO_KIQ(mmSMC_IND_DATA_11); spin_unlock_irqrestore(&adev->smc_idx_lock, flags); return r; } @@ -1363,11 +1363,11 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, if (adev->cg_flags & (AMD_CG_SUPPORT_MC_LS | AMD_CG_SUPPORT_MC_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_MC_LS) { - pp_support_state = AMD_CG_SUPPORT_MC_LS; + pp_support_state = PP_STATE_SUPPORT_LS; pp_state = PP_STATE_LS; } if (adev->cg_flags & AMD_CG_SUPPORT_MC_MGCG) { - pp_support_state |= AMD_CG_SUPPORT_MC_MGCG; + pp_support_state |= PP_STATE_SUPPORT_CG; pp_state |= PP_STATE_CG; } if (state == AMD_CG_STATE_UNGATE) @@ -1382,11 +1382,11 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, if (adev->cg_flags & (AMD_CG_SUPPORT_SDMA_LS | AMD_CG_SUPPORT_SDMA_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_SDMA_LS) { - pp_support_state = AMD_CG_SUPPORT_SDMA_LS; + pp_support_state = PP_STATE_SUPPORT_LS; pp_state = PP_STATE_LS; } if (adev->cg_flags & AMD_CG_SUPPORT_SDMA_MGCG) { - pp_support_state |= AMD_CG_SUPPORT_SDMA_MGCG; + pp_support_state |= PP_STATE_SUPPORT_CG; pp_state |= PP_STATE_CG; } if (state == AMD_CG_STATE_UNGATE) @@ -1401,11 +1401,11 @@ static int vi_common_set_clockgating_state_by_smu(void *handle, if (adev->cg_flags & (AMD_CG_SUPPORT_HDP_LS | AMD_CG_SUPPORT_HDP_MGCG)) { if (adev->cg_flags & AMD_CG_SUPPORT_HDP_LS) { - pp_support_state = AMD_CG_SUPPORT_HDP_LS; + pp_support_state = PP_STATE_SUPPORT_LS; pp_state = PP_STATE_LS; } if (adev->cg_flags & AMD_CG_SUPPORT_HDP_MGCG) { - pp_support_state |= AMD_CG_SUPPORT_HDP_MGCG; + pp_support_state |= PP_STATE_SUPPORT_CG; pp_state |= PP_STATE_CG; } if (state == AMD_CG_STATE_UNGATE) @@ -1596,16 +1596,18 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vi_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v7_4_ip_block); amdgpu_device_ip_block_add(adev, &iceland_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v2_4_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v2_4_ip_block); break; case CHIP_FIJI: amdgpu_device_ip_block_add(adev, &vi_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v8_5_ip_block); amdgpu_device_ip_block_add(adev, &tonga_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -1615,8 +1617,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v10_1_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); if (!amdgpu_sriov_vf(adev)) { amdgpu_device_ip_block_add(adev, &uvd_v6_0_ip_block); amdgpu_device_ip_block_add(adev, &vce_v3_0_ip_block); @@ -1626,6 +1626,8 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vi_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v8_0_ip_block); amdgpu_device_ip_block_add(adev, &tonga_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -1635,8 +1637,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v10_0_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); if (!amdgpu_sriov_vf(adev)) { amdgpu_device_ip_block_add(adev, &uvd_v5_0_ip_block); amdgpu_device_ip_block_add(adev, &vce_v3_0_ip_block); @@ -1649,6 +1649,8 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vi_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v8_1_ip_block); amdgpu_device_ip_block_add(adev, &tonga_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v3_1_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -1658,8 +1660,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v11_2_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v3_1_ip_block); amdgpu_device_ip_block_add(adev, &uvd_v6_3_ip_block); amdgpu_device_ip_block_add(adev, &vce_v3_4_ip_block); break; @@ -1667,6 +1667,8 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vi_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v8_0_ip_block); amdgpu_device_ip_block_add(adev, &cz_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -1676,8 +1678,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v11_0_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v8_0_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &uvd_v6_0_ip_block); amdgpu_device_ip_block_add(adev, &vce_v3_1_ip_block); #if defined(CONFIG_DRM_AMD_ACP) @@ -1688,6 +1688,8 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vi_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v8_0_ip_block); amdgpu_device_ip_block_add(adev, &cz_ih_ip_block); + amdgpu_device_ip_block_add(adev, &gfx_v8_1_ip_block); + amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); if (adev->enable_virtual_display) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); @@ -1697,8 +1699,6 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) #endif else amdgpu_device_ip_block_add(adev, &dce_v11_0_ip_block); - amdgpu_device_ip_block_add(adev, &gfx_v8_1_ip_block); - amdgpu_device_ip_block_add(adev, &sdma_v3_0_ip_block); amdgpu_device_ip_block_add(adev, &uvd_v6_2_ip_block); amdgpu_device_ip_block_add(adev, &vce_v3_4_ip_block); #if defined(CONFIG_DRM_AMD_ACP) |