diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu')
39 files changed, 653 insertions, 281 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index 9aa952f258cf..6dfdff58bffd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -1057,6 +1057,9 @@ static int amdgpu_cs_patch_ibs(struct amdgpu_cs_parser *p, r = amdgpu_ring_parse_cs(ring, p, job, ib); if (r) return r; + + if (ib->sa_bo) + ib->gpu_addr = amdgpu_sa_bo_gpu_addr(ib->sa_bo); } else { ib->ptr = (uint32_t *)kptr; r = amdgpu_ring_patch_cs_in_place(ring, p, job, ib); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 5cb33ac99f70..c43d1b6e5d66 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -685,16 +685,24 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, switch (args->in.op) { case AMDGPU_CTX_OP_ALLOC_CTX: + if (args->in.flags) + return -EINVAL; r = amdgpu_ctx_alloc(adev, fpriv, filp, priority, &id); args->out.alloc.ctx_id = id; break; case AMDGPU_CTX_OP_FREE_CTX: + if (args->in.flags) + return -EINVAL; r = amdgpu_ctx_free(fpriv, id); break; case AMDGPU_CTX_OP_QUERY_STATE: + if (args->in.flags) + return -EINVAL; r = amdgpu_ctx_query(adev, fpriv, id, &args->out); break; case AMDGPU_CTX_OP_QUERY_STATE2: + if (args->in.flags) + return -EINVAL; r = amdgpu_ctx_query2(adev, fpriv, id, &args->out); break; case AMDGPU_CTX_OP_GET_STABLE_PSTATE: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c index ac108fca64fe..4bd61c169ca8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_discovery.c @@ -278,7 +278,7 @@ static int amdgpu_discovery_read_binary_from_mem(struct amdgpu_device *adev, msg = RREG32(mmMP0_SMN_C2PMSG_33); if (msg & 0x80000000) break; - usleep_range(1000, 1100); + msleep(1); } } @@ -1500,6 +1500,7 @@ union gc_info { struct gc_info_v1_0 v1; struct gc_info_v1_1 v1_1; struct gc_info_v1_2 v1_2; + struct gc_info_v1_3 v1_3; struct gc_info_v2_0 v2; struct gc_info_v2_1 v2_1; }; @@ -1558,6 +1559,16 @@ static int amdgpu_discovery_get_gfx_info(struct amdgpu_device *adev) adev->gfx.config.gc_gl1c_size_per_instance = le32_to_cpu(gc_info->v1_2.gc_gl1c_size_per_instance); adev->gfx.config.gc_gl2c_per_gpu = le32_to_cpu(gc_info->v1_2.gc_gl2c_per_gpu); } + if (le16_to_cpu(gc_info->v1.header.version_minor) >= 3) { + adev->gfx.config.gc_tcp_size_per_cu = le32_to_cpu(gc_info->v1_3.gc_tcp_size_per_cu); + adev->gfx.config.gc_tcp_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_tcp_cache_line_size); + adev->gfx.config.gc_instruction_cache_size_per_sqc = le32_to_cpu(gc_info->v1_3.gc_instruction_cache_size_per_sqc); + adev->gfx.config.gc_instruction_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_instruction_cache_line_size); + adev->gfx.config.gc_scalar_data_cache_size_per_sqc = le32_to_cpu(gc_info->v1_3.gc_scalar_data_cache_size_per_sqc); + adev->gfx.config.gc_scalar_data_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_scalar_data_cache_line_size); + adev->gfx.config.gc_tcc_size = le32_to_cpu(gc_info->v1_3.gc_tcc_size); + adev->gfx.config.gc_tcc_cache_line_size = le32_to_cpu(gc_info->v1_3.gc_tcc_cache_line_size); + } break; case 2: adev->gfx.config.max_shader_engines = le32_to_cpu(gc_info->v2.gc_num_se); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 094498a0964b..d7ef8cbecf6c 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -2908,6 +2908,7 @@ static const struct file_operations amdgpu_driver_kms_fops = { #ifdef CONFIG_PROC_FS .show_fdinfo = drm_show_fdinfo, #endif + .fop_flags = FOP_UNSIGNED_OFFSET, }; int amdgpu_file_to_fpriv(struct file *filp, struct amdgpu_fpriv **fpriv) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index aad2027e5c7c..0e617dff8765 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -348,6 +348,9 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, return -EINVAL; } + /* always clear VRAM */ + flags |= AMDGPU_GEM_CREATE_VRAM_CLEARED; + /* create a gem object to contain this object in */ if (args->in.domains & (AMDGPU_GEM_DOMAIN_GDS | AMDGPU_GEM_DOMAIN_GWS | AMDGPU_GEM_DOMAIN_OA)) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c index 82452606ae6c..1849510a308a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c @@ -509,6 +509,16 @@ int amdgpu_gfx_disable_kcq(struct amdgpu_device *adev, int xcc_id) int i, r = 0; int j; + if (adev->enable_mes) { + for (i = 0; i < adev->gfx.num_compute_rings; i++) { + j = i + xcc_id * adev->gfx.num_compute_rings; + amdgpu_mes_unmap_legacy_queue(adev, + &adev->gfx.compute_ring[j], + RESET_QUEUES, 0, 0); + } + return 0; + } + if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) return -EINVAL; @@ -551,6 +561,18 @@ int amdgpu_gfx_disable_kgq(struct amdgpu_device *adev, int xcc_id) int i, r = 0; int j; + if (adev->enable_mes) { + if (amdgpu_gfx_is_master_xcc(adev, xcc_id)) { + for (i = 0; i < adev->gfx.num_gfx_rings; i++) { + j = i + xcc_id * adev->gfx.num_gfx_rings; + amdgpu_mes_unmap_legacy_queue(adev, + &adev->gfx.gfx_ring[j], + PREEMPT_QUEUES, 0, 0); + } + } + return 0; + } + if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) return -EINVAL; @@ -635,7 +657,7 @@ int amdgpu_gfx_enable_kcq(struct amdgpu_device *adev, int xcc_id) uint64_t queue_mask = 0; int r, i, j; - if (adev->enable_mes) + if (adev->mes.enable_legacy_queue_map) return amdgpu_gfx_mes_enable_kcq(adev, xcc_id); if (!kiq->pmf || !kiq->pmf->kiq_map_queues || !kiq->pmf->kiq_set_resources) @@ -697,7 +719,7 @@ int amdgpu_gfx_enable_kgq(struct amdgpu_device *adev, int xcc_id) amdgpu_device_flush_hdp(adev, NULL); - if (adev->enable_mes) { + if (adev->mes.enable_legacy_queue_map) { for (i = 0; i < adev->gfx.num_gfx_rings; i++) { j = i + xcc_id * adev->gfx.num_gfx_rings; r = amdgpu_mes_map_legacy_queue(adev, @@ -995,7 +1017,7 @@ uint32_t amdgpu_kiq_rreg(struct amdgpu_device *adev, uint32_t reg, uint32_t xcc_ if (amdgpu_device_skip_hw_access(adev)) return 0; - if (adev->mes.ring.sched.ready) + if (adev->mes.ring[0].sched.ready) return amdgpu_mes_rreg(adev, reg); BUG_ON(!ring->funcs->emit_rreg); @@ -1065,7 +1087,7 @@ void amdgpu_kiq_wreg(struct amdgpu_device *adev, uint32_t reg, uint32_t v, uint3 if (amdgpu_device_skip_hw_access(adev)) return; - if (adev->mes.ring.sched.ready) { + if (adev->mes.ring[0].sched.ready) { amdgpu_mes_wreg(adev, reg, v); return; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h index ddda94e49db4..56cc58edbb4e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h @@ -240,6 +240,12 @@ struct amdgpu_gfx_config { uint32_t gc_tcp_size_per_cu; uint32_t gc_num_cu_per_sqc; uint32_t gc_tcc_size; + uint32_t gc_tcp_cache_line_size; + uint32_t gc_instruction_cache_size_per_sqc; + uint32_t gc_instruction_cache_line_size; + uint32_t gc_scalar_data_cache_size_per_sqc; + uint32_t gc_scalar_data_cache_line_size; + uint32_t gc_tcc_cache_line_size; }; struct amdgpu_cu_info { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c index c02659025656..b49b3650fd62 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.c @@ -589,7 +589,8 @@ int amdgpu_gmc_allocate_vm_inv_eng(struct amdgpu_device *adev) ring = adev->rings[i]; vmhub = ring->vm_hub; - if (ring == &adev->mes.ring || + if (ring == &adev->mes.ring[0] || + ring == &adev->mes.ring[1] || ring == &adev->umsch_mm.ring) continue; @@ -761,7 +762,7 @@ void amdgpu_gmc_fw_reg_write_reg_wait(struct amdgpu_device *adev, unsigned long flags; uint32_t seq; - if (adev->mes.ring.sched.ready) { + if (adev->mes.ring[0].sched.ready) { amdgpu_mes_reg_write_reg_wait(adev, reg0, reg1, ref, mask); return; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h index febca3130497..4d951a1baefa 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h @@ -156,6 +156,8 @@ struct amdgpu_gmc_funcs { uint64_t addr, uint64_t *flags); /* get the amount of memory used by the vbios for pre-OS console */ unsigned int (*get_vbios_fb_size)(struct amdgpu_device *adev); + /* get the DCC buffer alignment */ + unsigned int (*get_dcc_alignment)(struct amdgpu_device *adev); enum amdgpu_memory_partition (*query_mem_partition_mode)( struct amdgpu_device *adev); @@ -363,6 +365,10 @@ struct amdgpu_gmc { (adev)->gmc.gmc_funcs->override_vm_pte_flags \ ((adev), (vm), (addr), (pte_flags)) #define amdgpu_gmc_get_vbios_fb_size(adev) (adev)->gmc.gmc_funcs->get_vbios_fb_size((adev)) +#define amdgpu_gmc_get_dcc_alignment(adev) ({ \ + typeof(adev) _adev = (adev); \ + _adev->gmc.gmc_funcs->get_dcc_alignment(_adev); \ +}) /** * amdgpu_gmc_vram_full_visible - Check if full VRAM is visible through the BAR diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c index e238f2832f65..908e13455152 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_job.c @@ -264,9 +264,8 @@ amdgpu_job_prepare_job(struct drm_sched_job *sched_job, struct dma_fence *fence = NULL; int r; - /* Ignore soft recovered fences here */ r = drm_sched_entity_error(s_entity); - if (r && r != -ENODATA) + if (r) goto error; if (!fence && job->gang_submit) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c index dac88d2dd70d..1cb1ec7beefe 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.c @@ -135,9 +135,11 @@ int amdgpu_mes_init(struct amdgpu_device *adev) idr_init(&adev->mes.queue_id_idr); ida_init(&adev->mes.doorbell_ida); spin_lock_init(&adev->mes.queue_id_lock); - spin_lock_init(&adev->mes.ring_lock); mutex_init(&adev->mes.mutex_hidden); + for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) + spin_lock_init(&adev->mes.ring_lock[i]); + adev->mes.total_max_queue = AMDGPU_FENCE_MES_QUEUE_ID_MASK; adev->mes.vmid_mask_mmhub = 0xffffff00; adev->mes.vmid_mask_gfxhub = 0xffffff00; @@ -163,36 +165,38 @@ int amdgpu_mes_init(struct amdgpu_device *adev) adev->mes.sdma_hqd_mask[i] = 0xfc; } - r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs); - if (r) { - dev_err(adev->dev, - "(%d) ring trail_fence_offs wb alloc failed\n", r); - goto error_ids; - } - adev->mes.sch_ctx_gpu_addr = - adev->wb.gpu_addr + (adev->mes.sch_ctx_offs * 4); - adev->mes.sch_ctx_ptr = - (uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs]; + for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) { + r = amdgpu_device_wb_get(adev, &adev->mes.sch_ctx_offs[i]); + if (r) { + dev_err(adev->dev, + "(%d) ring trail_fence_offs wb alloc failed\n", + r); + goto error; + } + adev->mes.sch_ctx_gpu_addr[i] = + adev->wb.gpu_addr + (adev->mes.sch_ctx_offs[i] * 4); + adev->mes.sch_ctx_ptr[i] = + (uint64_t *)&adev->wb.wb[adev->mes.sch_ctx_offs[i]]; - r = amdgpu_device_wb_get(adev, &adev->mes.query_status_fence_offs); - if (r) { - amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); - dev_err(adev->dev, - "(%d) query_status_fence_offs wb alloc failed\n", r); - goto error_ids; + r = amdgpu_device_wb_get(adev, + &adev->mes.query_status_fence_offs[i]); + if (r) { + dev_err(adev->dev, + "(%d) query_status_fence_offs wb alloc failed\n", + r); + goto error; + } + adev->mes.query_status_fence_gpu_addr[i] = adev->wb.gpu_addr + + (adev->mes.query_status_fence_offs[i] * 4); + adev->mes.query_status_fence_ptr[i] = + (uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs[i]]; } - adev->mes.query_status_fence_gpu_addr = - adev->wb.gpu_addr + (adev->mes.query_status_fence_offs * 4); - adev->mes.query_status_fence_ptr = - (uint64_t *)&adev->wb.wb[adev->mes.query_status_fence_offs]; r = amdgpu_device_wb_get(adev, &adev->mes.read_val_offs); if (r) { - amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); - amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); dev_err(adev->dev, "(%d) read_val_offs alloc failed\n", r); - goto error_ids; + goto error; } adev->mes.read_val_gpu_addr = adev->wb.gpu_addr + (adev->mes.read_val_offs * 4); @@ -212,10 +216,16 @@ int amdgpu_mes_init(struct amdgpu_device *adev) error_doorbell: amdgpu_mes_doorbell_free(adev); error: - amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); - amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); - amdgpu_device_wb_free(adev, adev->mes.read_val_offs); -error_ids: + for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) { + if (adev->mes.sch_ctx_ptr[i]) + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]); + if (adev->mes.query_status_fence_ptr[i]) + amdgpu_device_wb_free(adev, + adev->mes.query_status_fence_offs[i]); + } + if (adev->mes.read_val_ptr) + amdgpu_device_wb_free(adev, adev->mes.read_val_offs); + idr_destroy(&adev->mes.pasid_idr); idr_destroy(&adev->mes.gang_id_idr); idr_destroy(&adev->mes.queue_id_idr); @@ -226,13 +236,22 @@ error_ids: void amdgpu_mes_fini(struct amdgpu_device *adev) { + int i; + amdgpu_bo_free_kernel(&adev->mes.event_log_gpu_obj, &adev->mes.event_log_gpu_addr, &adev->mes.event_log_cpu_addr); - amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); - amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); - amdgpu_device_wb_free(adev, adev->mes.read_val_offs); + for (i = 0; i < AMDGPU_MAX_MES_PIPES; i++) { + if (adev->mes.sch_ctx_ptr[i]) + amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs[i]); + if (adev->mes.query_status_fence_ptr[i]) + amdgpu_device_wb_free(adev, + adev->mes.query_status_fence_offs[i]); + } + if (adev->mes.read_val_ptr) + amdgpu_device_wb_free(adev, adev->mes.read_val_offs); + amdgpu_mes_doorbell_free(adev); idr_destroy(&adev->mes.pasid_idr); @@ -1499,7 +1518,7 @@ int amdgpu_mes_init_microcode(struct amdgpu_device *adev, int pipe) amdgpu_ucode_ip_version_decode(adev, GC_HWIP, ucode_prefix, sizeof(ucode_prefix)); - if (adev->enable_uni_mes && pipe == AMDGPU_MES_SCHED_PIPE) { + if (adev->enable_uni_mes) { snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_uni_mes.bin", ucode_prefix); } else if (amdgpu_ip_version(adev, GC_HWIP, 0) >= IP_VERSION(11, 0, 0) && diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h index 2d659c612f03..bcce1add4ef6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mes.h @@ -75,6 +75,7 @@ struct amdgpu_mes { uint32_t sched_version; uint32_t kiq_version; + bool enable_legacy_queue_map; uint32_t total_max_queue; uint32_t max_doorbell_slices; @@ -82,8 +83,8 @@ struct amdgpu_mes { uint64_t default_process_quantum; uint64_t default_gang_quantum; - struct amdgpu_ring ring; - spinlock_t ring_lock; + struct amdgpu_ring ring[AMDGPU_MAX_MES_PIPES]; + spinlock_t ring_lock[AMDGPU_MAX_MES_PIPES]; const struct firmware *fw[AMDGPU_MAX_MES_PIPES]; @@ -112,12 +113,12 @@ struct amdgpu_mes { uint32_t gfx_hqd_mask[AMDGPU_MES_MAX_GFX_PIPES]; uint32_t sdma_hqd_mask[AMDGPU_MES_MAX_SDMA_PIPES]; uint32_t aggregated_doorbells[AMDGPU_MES_PRIORITY_NUM_LEVELS]; - uint32_t sch_ctx_offs; - uint64_t sch_ctx_gpu_addr; - uint64_t *sch_ctx_ptr; - uint32_t query_status_fence_offs; - uint64_t query_status_fence_gpu_addr; - uint64_t *query_status_fence_ptr; + uint32_t sch_ctx_offs[AMDGPU_MAX_MES_PIPES]; + uint64_t sch_ctx_gpu_addr[AMDGPU_MAX_MES_PIPES]; + uint64_t *sch_ctx_ptr[AMDGPU_MAX_MES_PIPES]; + uint32_t query_status_fence_offs[AMDGPU_MAX_MES_PIPES]; + uint64_t query_status_fence_gpu_addr[AMDGPU_MAX_MES_PIPES]; + uint64_t *query_status_fence_ptr[AMDGPU_MAX_MES_PIPES]; uint32_t read_val_offs; uint64_t read_val_gpu_addr; uint32_t *read_val_ptr; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c index 0c856005df6b..38face981c3e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp_ta.c @@ -166,6 +166,9 @@ static ssize_t ta_if_load_debugfs_write(struct file *fp, const char *buf, size_t if (ret) return -EFAULT; + if (ta_bin_len > PSP_1_MEG) + return -EINVAL; + copy_pos += sizeof(uint32_t); ta_bin = kzalloc(ta_bin_len, GFP_KERNEL); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index ad49cecb20b8..e6344a6b0a9f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -212,6 +212,8 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring, */ if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ) sched_hw_submission = max(sched_hw_submission, 256); + if (ring->funcs->type == AMDGPU_RING_TYPE_MES) + sched_hw_submission = 8; else if (ring == &adev->sdma.instance[0].page) sched_hw_submission = 256; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h index 1a5439abd1a0..c87d68d4be53 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h @@ -461,8 +461,11 @@ struct amdgpu_vcn5_fw_shared { struct amdgpu_fw_shared_unified_queue_struct sq; uint8_t pad1[8]; struct amdgpu_fw_shared_fw_logging fw_log; + uint8_t pad2[20]; struct amdgpu_fw_shared_rb_setup rb_setup; - uint8_t pad2[4]; + struct amdgpu_fw_shared_smu_interface_info smu_dpm_interface; + struct amdgpu_fw_shared_drm_key_wa drm_key_wa; + uint8_t pad3[9]; }; #define VCN_BLOCK_ENCODE_DISABLE_MASK 0x80 diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c index 111c380f929b..b287a82e6177 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_virt.c @@ -858,7 +858,7 @@ void amdgpu_virt_post_reset(struct amdgpu_device *adev) adev->gfx.is_poweron = false; } - adev->mes.ring.sched.ready = false; + adev->mes.ring[0].sched.ready = false; } bool amdgpu_virt_fw_load_skip_check(struct amdgpu_device *adev, uint32_t ucode_id) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c index f91cc149d06c..7d26a962f811 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vram_mgr.c @@ -456,6 +456,7 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, u64 vis_usage = 0, max_bytes, min_block_size; struct amdgpu_vram_mgr_resource *vres; u64 size, remaining_size, lpfn, fpfn; + unsigned int adjust_dcc_size = 0; struct drm_buddy *mm = &mgr->mm; struct drm_buddy_block *block; unsigned long pages_per_block; @@ -511,7 +512,19 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, /* Allocate blocks in desired range */ vres->flags |= DRM_BUDDY_RANGE_ALLOCATION; + if (bo->flags & AMDGPU_GEM_CREATE_GFX12_DCC && + adev->gmc.gmc_funcs->get_dcc_alignment) + adjust_dcc_size = amdgpu_gmc_get_dcc_alignment(adev); + remaining_size = (u64)vres->base.size; + if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) { + unsigned int dcc_size; + + dcc_size = roundup_pow_of_two(vres->base.size + adjust_dcc_size); + remaining_size = (u64)dcc_size; + + vres->flags |= DRM_BUDDY_TRIM_DISABLE; + } mutex_lock(&mgr->lock); while (remaining_size) { @@ -521,8 +534,11 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, min_block_size = mgr->default_page_size; size = remaining_size; - if ((size >= (u64)pages_per_block << PAGE_SHIFT) && - !(size & (((u64)pages_per_block << PAGE_SHIFT) - 1))) + + if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) + min_block_size = size; + else if ((size >= (u64)pages_per_block << PAGE_SHIFT) && + !(size & (((u64)pages_per_block << PAGE_SHIFT) - 1))) min_block_size = (u64)pages_per_block << PAGE_SHIFT; BUG_ON(min_block_size < mm->chunk_size); @@ -553,6 +569,22 @@ static int amdgpu_vram_mgr_new(struct ttm_resource_manager *man, } mutex_unlock(&mgr->lock); + if (bo->flags & AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS && adjust_dcc_size) { + struct drm_buddy_block *dcc_block; + unsigned long dcc_start; + u64 trim_start; + + dcc_block = amdgpu_vram_mgr_first_block(&vres->blocks); + /* Adjust the start address for DCC buffers only */ + dcc_start = + roundup((unsigned long)amdgpu_vram_mgr_block_start(dcc_block), + adjust_dcc_size); + trim_start = (u64)dcc_start; + drm_buddy_block_trim(mm, &trim_start, + (u64)vres->base.size, + &vres->blocks); + } + vres->base.start = 0; size = max_t(u64, amdgpu_vram_mgr_blocks_size(&vres->blocks), vres->base.size); diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c index 2957702fca0c..e444e621ddaa 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v10_0.c @@ -4116,6 +4116,7 @@ static void gfx_v10_0_check_gfxoff_flag(struct amdgpu_device *adev) static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) { + char fw_name[53]; char ucode_prefix[30]; const char *wks = ""; int err; @@ -4149,8 +4150,8 @@ static int gfx_v10_0_init_microcode(struct amdgpu_device *adev) amdgpu_gfx_cp_init_microcode(adev, AMDGPU_UCODE_ID_CP_CE); if (!amdgpu_sriov_vf(adev)) { - err = amdgpu_ucode_request(adev, &adev->gfx.rlc_fw, - "amdgpu/%s_rlc.bin", ucode_prefix); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_rlc.bin", ucode_prefix); + err = request_firmware(&adev->gfx.rlc_fw, fw_name, adev->dev); if (err) goto out; diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c index f384be0d1800..e45d23e82878 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v12_0.c @@ -202,6 +202,12 @@ static const struct amdgpu_hwip_reg_entry gc_gfx_queue_reg_list_12[] = { SOC15_REG_ENTRY_STR(GC, 0, regCP_IB1_BUFSZ) }; +static const struct soc15_reg_golden golden_settings_gc_12_0[] = { + SOC15_REG_GOLDEN_VALUE(GC, 0, regDB_MEM_CONFIG, 0x0000000f, 0x0000000f), + SOC15_REG_GOLDEN_VALUE(GC, 0, regCB_HW_CONTROL_1, 0x03000000, 0x03000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, regGL2C_CTRL5, 0x00000070, 0x00000020) +}; + #define DEFAULT_SH_MEM_CONFIG \ ((SH_MEM_ADDRESS_MODE_64 << SH_MEM_CONFIG__ADDRESS_MODE__SHIFT) | \ (SH_MEM_ALIGNMENT_MODE_UNALIGNED << SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT) | \ @@ -2999,7 +3005,7 @@ static int gfx_v12_0_compute_mqd_init(struct amdgpu_device *adev, void *m, (order_base_2(prop->queue_size / 4) - 1)); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, RPTR_BLOCK_SIZE, (order_base_2(AMDGPU_GPU_PAGE_SIZE / 4) - 1)); - tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 0); + tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, UNORD_DISPATCH, 1); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, TUNNEL_DISPATCH, 0); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, PRIV_STATE, 1); tmp = REG_SET_FIELD(tmp, CP_HQD_PQ_CONTROL, KMD_QUEUE, 1); @@ -3432,6 +3438,24 @@ static void gfx_v12_0_disable_gpa_mode(struct amdgpu_device *adev) WREG32_SOC15(GC, 0, regCPG_PSP_DEBUG, data); } +static void gfx_v12_0_init_golden_registers(struct amdgpu_device *adev) +{ + if (amdgpu_sriov_vf(adev)) + return; + + switch (amdgpu_ip_version(adev, GC_HWIP, 0)) { + case IP_VERSION(12, 0, 0): + case IP_VERSION(12, 0, 1): + if (adev->rev_id == 0) + soc15_program_register_sequence(adev, + golden_settings_gc_12_0, + (const u32)ARRAY_SIZE(golden_settings_gc_12_0)); + break; + default: + break; + } +} + static int gfx_v12_0_hw_init(void *handle) { int r; @@ -3472,6 +3496,9 @@ static int gfx_v12_0_hw_init(void *handle) } } + if (!amdgpu_emu_mode) + gfx_v12_0_init_golden_registers(adev); + adev->gfx.is_poweron = true; if (get_gb_addr_config(adev)) @@ -3519,33 +3546,9 @@ static int gfx_v12_0_hw_init(void *handle) return r; } -static int gfx_v12_0_kiq_disable_kgq(struct amdgpu_device *adev) -{ - struct amdgpu_kiq *kiq = &adev->gfx.kiq[0]; - struct amdgpu_ring *kiq_ring = &kiq->ring; - int i, r = 0; - - if (!kiq->pmf || !kiq->pmf->kiq_unmap_queues) - return -EINVAL; - - if (amdgpu_ring_alloc(kiq_ring, kiq->pmf->unmap_queues_size * - adev->gfx.num_gfx_rings)) - return -ENOMEM; - - for (i = 0; i < adev->gfx.num_gfx_rings; i++) - kiq->pmf->kiq_unmap_queues(kiq_ring, &adev->gfx.gfx_ring[i], - PREEMPT_QUEUES, 0, 0); - - if (adev->gfx.kiq[0].ring.sched.ready) - r = amdgpu_ring_test_helper(kiq_ring); - - return r; -} - static int gfx_v12_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - int r; uint32_t tmp; amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); @@ -3553,8 +3556,7 @@ static int gfx_v12_0_hw_fini(void *handle) if (!adev->no_hw_access) { if (amdgpu_async_gfx_ring) { - r = gfx_v12_0_kiq_disable_kgq(adev); - if (r) + if (amdgpu_gfx_disable_kgq(adev, 0)) DRM_ERROR("KGQ disable failed\n"); } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c index b88a6fa173b3..2797fd84432b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v11_0.c @@ -231,7 +231,7 @@ static void gmc_v11_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, /* This is necessary for SRIOV as well as for GFXOFF to function * properly under bare metal */ - if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring.sched.ready) && + if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring[0].sched.ready) && (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) { amdgpu_gmc_fw_reg_write_reg_wait(adev, req, ack, inv_req, 1 << vmid, GET_INST(GC, 0)); diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c index fd3ac483760e..edcb5351f8cc 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v12_0.c @@ -299,7 +299,7 @@ static void gmc_v12_0_flush_gpu_tlb(struct amdgpu_device *adev, uint32_t vmid, /* This is necessary for SRIOV as well as for GFXOFF to function * properly under bare metal */ - if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring.sched.ready) && + if ((adev->gfx.kiq[0].ring.sched.ready || adev->mes.ring[0].sched.ready) && (amdgpu_sriov_runtime(adev) || !amdgpu_sriov_vf(adev))) { struct amdgpu_vmhub *hub = &adev->vmhub[vmhub]; const unsigned eng = 17; @@ -542,6 +542,23 @@ static unsigned gmc_v12_0_get_vbios_fb_size(struct amdgpu_device *adev) return 0; } +static unsigned int gmc_v12_0_get_dcc_alignment(struct amdgpu_device *adev) +{ + unsigned int max_tex_channel_caches, alignment; + + if (amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 0) && + amdgpu_ip_version(adev, GC_HWIP, 0) != IP_VERSION(12, 0, 1)) + return 0; + + max_tex_channel_caches = adev->gfx.config.max_texture_channel_caches; + if (is_power_of_2(max_tex_channel_caches)) + alignment = (unsigned int)(max_tex_channel_caches / SZ_4); + else + alignment = roundup_pow_of_two(max_tex_channel_caches); + + return (unsigned int)(alignment * max_tex_channel_caches * SZ_1K); +} + static const struct amdgpu_gmc_funcs gmc_v12_0_gmc_funcs = { .flush_gpu_tlb = gmc_v12_0_flush_gpu_tlb, .flush_gpu_tlb_pasid = gmc_v12_0_flush_gpu_tlb_pasid, @@ -551,6 +568,7 @@ static const struct amdgpu_gmc_funcs gmc_v12_0_gmc_funcs = { .get_vm_pde = gmc_v12_0_get_vm_pde, .get_vm_pte = gmc_v12_0_get_vm_pte, .get_vbios_fb_size = gmc_v12_0_get_vbios_fb_size, + .get_dcc_alignment = gmc_v12_0_get_dcc_alignment, }; static void gmc_v12_0_set_gmc_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c index 71f43a5c7f72..6e0e88076224 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.c @@ -23,6 +23,7 @@ #include "amdgpu.h" #include "amdgpu_jpeg.h" +#include "amdgpu_cs.h" #include "soc15.h" #include "soc15d.h" #include "vcn_v1_0.h" @@ -34,6 +35,9 @@ static void jpeg_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev); static void jpeg_v1_0_set_irq_funcs(struct amdgpu_device *adev); static void jpeg_v1_0_ring_begin_use(struct amdgpu_ring *ring); +static int jpeg_v1_dec_ring_parse_cs(struct amdgpu_cs_parser *parser, + struct amdgpu_job *job, + struct amdgpu_ib *ib); static void jpeg_v1_0_decode_ring_patch_wreg(struct amdgpu_ring *ring, uint32_t *ptr, uint32_t reg_offset, uint32_t val) { @@ -300,7 +304,10 @@ static void jpeg_v1_0_decode_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, PACKETJ(SOC15_REG_OFFSET(JPEG, 0, mmUVD_LMI_JRBC_IB_VMID), 0, 0, PACKETJ_TYPE0)); - amdgpu_ring_write(ring, (vmid | (vmid << 4))); + if (ring->funcs->parse_cs) + amdgpu_ring_write(ring, 0); + else + amdgpu_ring_write(ring, (vmid | (vmid << 4))); amdgpu_ring_write(ring, PACKETJ(SOC15_REG_OFFSET(JPEG, 0, mmUVD_LMI_JPEG_VMID), 0, 0, PACKETJ_TYPE0)); @@ -554,6 +561,7 @@ static const struct amdgpu_ring_funcs jpeg_v1_0_decode_ring_vm_funcs = { .get_rptr = jpeg_v1_0_decode_ring_get_rptr, .get_wptr = jpeg_v1_0_decode_ring_get_wptr, .set_wptr = jpeg_v1_0_decode_ring_set_wptr, + .parse_cs = jpeg_v1_dec_ring_parse_cs, .emit_frame_size = 6 + 6 + /* hdp invalidate / flush */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + @@ -611,3 +619,69 @@ static void jpeg_v1_0_ring_begin_use(struct amdgpu_ring *ring) vcn_v1_0_set_pg_for_begin_use(ring, set_clocks); } + +/** + * jpeg_v1_dec_ring_parse_cs - command submission parser + * + * @parser: Command submission parser context + * @job: the job to parse + * @ib: the IB to parse + * + * Parse the command stream, return -EINVAL for invalid packet, + * 0 otherwise + */ +static int jpeg_v1_dec_ring_parse_cs(struct amdgpu_cs_parser *parser, + struct amdgpu_job *job, + struct amdgpu_ib *ib) +{ + u32 i, reg, res, cond, type; + int ret = 0; + struct amdgpu_device *adev = parser->adev; + + for (i = 0; i < ib->length_dw ; i += 2) { + reg = CP_PACKETJ_GET_REG(ib->ptr[i]); + res = CP_PACKETJ_GET_RES(ib->ptr[i]); + cond = CP_PACKETJ_GET_COND(ib->ptr[i]); + type = CP_PACKETJ_GET_TYPE(ib->ptr[i]); + + if (res || cond != PACKETJ_CONDITION_CHECK0) /* only allow 0 for now */ + return -EINVAL; + + if (reg >= JPEG_V1_REG_RANGE_START && reg <= JPEG_V1_REG_RANGE_END) + continue; + + switch (type) { + case PACKETJ_TYPE0: + if (reg != JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_HIGH && + reg != JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_LOW && + reg != JPEG_V1_LMI_JPEG_READ_64BIT_BAR_HIGH && + reg != JPEG_V1_LMI_JPEG_READ_64BIT_BAR_LOW && + reg != JPEG_V1_REG_CTX_INDEX && + reg != JPEG_V1_REG_CTX_DATA) { + ret = -EINVAL; + } + break; + case PACKETJ_TYPE1: + if (reg != JPEG_V1_REG_CTX_DATA) + ret = -EINVAL; + break; + case PACKETJ_TYPE3: + if (reg != JPEG_V1_REG_SOFT_RESET) + ret = -EINVAL; + break; + case PACKETJ_TYPE6: + if (ib->ptr[i] != CP_PACKETJ_NOP) + ret = -EINVAL; + break; + default: + ret = -EINVAL; + } + + if (ret) { + dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]); + break; + } + } + + return ret; +} diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h index bbf33a6a3972..9654d22e0376 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v1_0.h @@ -29,4 +29,15 @@ int jpeg_v1_0_sw_init(void *handle); void jpeg_v1_0_sw_fini(void *handle); void jpeg_v1_0_start(struct amdgpu_device *adev, int mode); +#define JPEG_V1_REG_RANGE_START 0x8000 +#define JPEG_V1_REG_RANGE_END 0x803f + +#define JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_HIGH 0x8238 +#define JPEG_V1_LMI_JPEG_WRITE_64BIT_BAR_LOW 0x8239 +#define JPEG_V1_LMI_JPEG_READ_64BIT_BAR_HIGH 0x825a +#define JPEG_V1_LMI_JPEG_READ_64BIT_BAR_LOW 0x825b +#define JPEG_V1_REG_CTX_INDEX 0x8328 +#define JPEG_V1_REG_CTX_DATA 0x8329 +#define JPEG_V1_REG_SOFT_RESET 0x83a0 + #endif /*__JPEG_V1_0_H__*/ diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c index 99adf3625657..41c0f8750dc1 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.c @@ -23,6 +23,7 @@ #include "amdgpu.h" #include "amdgpu_jpeg.h" +#include "amdgpu_cs.h" #include "amdgpu_pm.h" #include "soc15.h" #include "soc15d.h" @@ -538,11 +539,15 @@ void jpeg_v2_0_dec_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET, 0, 0, PACKETJ_TYPE0)); - amdgpu_ring_write(ring, (vmid | (vmid << 4))); + + if (ring->funcs->parse_cs) + amdgpu_ring_write(ring, 0); + else + amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8))); amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JPEG_VMID_INTERNAL_OFFSET, 0, 0, PACKETJ_TYPE0)); - amdgpu_ring_write(ring, (vmid | (vmid << 4))); + amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8))); amdgpu_ring_write(ring, PACKETJ(mmUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET, 0, 0, PACKETJ_TYPE0)); @@ -764,6 +769,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_0_dec_ring_vm_funcs = { .get_rptr = jpeg_v2_0_dec_ring_get_rptr, .get_wptr = jpeg_v2_0_dec_ring_get_wptr, .set_wptr = jpeg_v2_0_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + @@ -810,3 +816,58 @@ const struct amdgpu_ip_block_version jpeg_v2_0_ip_block = { .rev = 0, .funcs = &jpeg_v2_0_ip_funcs, }; + +/** + * jpeg_v2_dec_ring_parse_cs - command submission parser + * + * @parser: Command submission parser context + * @job: the job to parse + * @ib: the IB to parse + * + * Parse the command stream, return -EINVAL for invalid packet, + * 0 otherwise + */ +int jpeg_v2_dec_ring_parse_cs(struct amdgpu_cs_parser *parser, + struct amdgpu_job *job, + struct amdgpu_ib *ib) +{ + u32 i, reg, res, cond, type; + struct amdgpu_device *adev = parser->adev; + + for (i = 0; i < ib->length_dw ; i += 2) { + reg = CP_PACKETJ_GET_REG(ib->ptr[i]); + res = CP_PACKETJ_GET_RES(ib->ptr[i]); + cond = CP_PACKETJ_GET_COND(ib->ptr[i]); + type = CP_PACKETJ_GET_TYPE(ib->ptr[i]); + + if (res) /* only support 0 at the moment */ + return -EINVAL; + + switch (type) { + case PACKETJ_TYPE0: + if (cond != PACKETJ_CONDITION_CHECK0 || reg < JPEG_REG_RANGE_START || + reg > JPEG_REG_RANGE_END) { + dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]); + return -EINVAL; + } + break; + case PACKETJ_TYPE3: + if (cond != PACKETJ_CONDITION_CHECK3 || reg < JPEG_REG_RANGE_START || + reg > JPEG_REG_RANGE_END) { + dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]); + return -EINVAL; + } + break; + case PACKETJ_TYPE6: + if (ib->ptr[i] == CP_PACKETJ_NOP) + continue; + dev_err(adev->dev, "Invalid packet [0x%08x]!\n", ib->ptr[i]); + return -EINVAL; + default: + dev_err(adev->dev, "Unknown packet type %d !\n", type); + return -EINVAL; + } + } + + return 0; +} diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h index 654e43e83e2c..63fadda7a673 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_0.h @@ -45,6 +45,9 @@ #define JRBC_DEC_EXTERNAL_REG_WRITE_ADDR 0x18000 +#define JPEG_REG_RANGE_START 0x4000 +#define JPEG_REG_RANGE_END 0x41c2 + void jpeg_v2_0_dec_ring_insert_start(struct amdgpu_ring *ring); void jpeg_v2_0_dec_ring_insert_end(struct amdgpu_ring *ring); void jpeg_v2_0_dec_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq, @@ -57,6 +60,9 @@ void jpeg_v2_0_dec_ring_emit_vm_flush(struct amdgpu_ring *ring, unsigned vmid, uint64_t pd_addr); void jpeg_v2_0_dec_ring_emit_wreg(struct amdgpu_ring *ring, uint32_t reg, uint32_t val); void jpeg_v2_0_dec_ring_nop(struct amdgpu_ring *ring, uint32_t count); +int jpeg_v2_dec_ring_parse_cs(struct amdgpu_cs_parser *parser, + struct amdgpu_job *job, + struct amdgpu_ib *ib); extern const struct amdgpu_ip_block_version jpeg_v2_0_ip_block; diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c index d8ef95c847c2..eedb9a829d95 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v2_5.c @@ -662,6 +662,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_5_dec_ring_vm_funcs = { .get_rptr = jpeg_v2_5_dec_ring_get_rptr, .get_wptr = jpeg_v2_5_dec_ring_get_wptr, .set_wptr = jpeg_v2_5_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + @@ -691,6 +692,7 @@ static const struct amdgpu_ring_funcs jpeg_v2_6_dec_ring_vm_funcs = { .get_rptr = jpeg_v2_5_dec_ring_get_rptr, .get_wptr = jpeg_v2_5_dec_ring_get_wptr, .set_wptr = jpeg_v2_5_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c index 31cfa3ce6528..b1e7fd25afbc 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v3_0.c @@ -560,6 +560,7 @@ static const struct amdgpu_ring_funcs jpeg_v3_0_dec_ring_vm_funcs = { .get_rptr = jpeg_v3_0_dec_ring_get_rptr, .get_wptr = jpeg_v3_0_dec_ring_get_wptr, .set_wptr = jpeg_v3_0_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c index 3dac8f259d7f..6c5c1a68a9b7 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.c @@ -727,6 +727,7 @@ static const struct amdgpu_ring_funcs jpeg_v4_0_dec_ring_vm_funcs = { .get_rptr = jpeg_v4_0_dec_ring_get_rptr, .get_wptr = jpeg_v4_0_dec_ring_get_wptr, .set_wptr = jpeg_v4_0_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h index 07d36c2abd6b..47638fd4d4e2 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0.h @@ -32,5 +32,4 @@ enum amdgpu_jpeg_v4_0_sub_block { }; extern const struct amdgpu_ip_block_version jpeg_v4_0_ip_block; - #endif /* __JPEG_V4_0_H__ */ diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c index ad524ddc9760..b55041f38cec 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_3.c @@ -25,6 +25,7 @@ #include "amdgpu_jpeg.h" #include "soc15.h" #include "soc15d.h" +#include "jpeg_v2_0.h" #include "jpeg_v4_0_3.h" #include "mmsch_v4_0_3.h" @@ -782,11 +783,15 @@ void jpeg_v4_0_3_dec_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, PACKETJ(regUVD_LMI_JRBC_IB_VMID_INTERNAL_OFFSET, 0, 0, PACKETJ_TYPE0)); - amdgpu_ring_write(ring, (vmid | (vmid << 4))); + + if (ring->funcs->parse_cs) + amdgpu_ring_write(ring, 0); + else + amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8))); amdgpu_ring_write(ring, PACKETJ(regUVD_LMI_JPEG_VMID_INTERNAL_OFFSET, 0, 0, PACKETJ_TYPE0)); - amdgpu_ring_write(ring, (vmid | (vmid << 4))); + amdgpu_ring_write(ring, (vmid | (vmid << 4) | (vmid << 8))); amdgpu_ring_write(ring, PACKETJ(regUVD_LMI_JRBC_IB_64BIT_BAR_LOW_INTERNAL_OFFSET, 0, 0, PACKETJ_TYPE0)); @@ -1084,6 +1089,7 @@ static const struct amdgpu_ring_funcs jpeg_v4_0_3_dec_ring_vm_funcs = { .get_rptr = jpeg_v4_0_3_dec_ring_get_rptr, .get_wptr = jpeg_v4_0_3_dec_ring_get_wptr, .set_wptr = jpeg_v4_0_3_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c index f96ac6bce526..44eeed445ea9 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v4_0_5.c @@ -768,6 +768,7 @@ static const struct amdgpu_ring_funcs jpeg_v4_0_5_dec_ring_vm_funcs = { .get_rptr = jpeg_v4_0_5_dec_ring_get_rptr, .get_wptr = jpeg_v4_0_5_dec_ring_get_wptr, .set_wptr = jpeg_v4_0_5_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + diff --git a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c index d694a276498a..d662aa841f97 100644 --- a/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c +++ b/drivers/gpu/drm/amd/amdgpu/jpeg_v5_0_0.c @@ -26,6 +26,7 @@ #include "amdgpu_pm.h" #include "soc15.h" #include "soc15d.h" +#include "jpeg_v2_0.h" #include "jpeg_v4_0_3.h" #include "vcn/vcn_5_0_0_offset.h" @@ -646,6 +647,7 @@ static const struct amdgpu_ring_funcs jpeg_v5_0_0_dec_ring_vm_funcs = { .get_rptr = jpeg_v5_0_0_dec_ring_get_rptr, .get_wptr = jpeg_v5_0_0_dec_ring_get_wptr, .set_wptr = jpeg_v5_0_0_dec_ring_set_wptr, + .parse_cs = jpeg_v2_dec_ring_parse_cs, .emit_frame_size = SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c index f9343642ae7e..8aded0a67037 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v11_0.c @@ -162,13 +162,13 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, union MESAPI__QUERY_MES_STATUS mes_status_pkt; signed long timeout = 3000000; /* 3000 ms */ struct amdgpu_device *adev = mes->adev; - struct amdgpu_ring *ring = &mes->ring; + struct amdgpu_ring *ring = &mes->ring[0]; struct MES_API_STATUS *api_status; union MESAPI__MISC *x_pkt = pkt; const char *op_str, *misc_op_str; unsigned long flags; u64 status_gpu_addr; - u32 status_offset; + u32 seq, status_offset; u64 *status_ptr; signed long r; int ret; @@ -191,11 +191,18 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, status_ptr = (u64 *)&adev->wb.wb[status_offset]; *status_ptr = 0; - spin_lock_irqsave(&mes->ring_lock, flags); + spin_lock_irqsave(&mes->ring_lock[0], flags); r = amdgpu_ring_alloc(ring, (size + sizeof(mes_status_pkt)) / 4); if (r) goto error_unlock_free; + seq = ++ring->fence_drv.sync_seq; + r = amdgpu_fence_wait_polling(ring, + seq - ring->fence_drv.num_fences_mask, + timeout); + if (r < 1) + goto error_undo; + api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off); api_status->api_completion_fence_addr = status_gpu_addr; api_status->api_completion_fence_value = 1; @@ -208,14 +215,13 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; mes_status_pkt.api_status.api_completion_fence_addr = ring->fence_drv.gpu_addr; - mes_status_pkt.api_status.api_completion_fence_value = - ++ring->fence_drv.sync_seq; + mes_status_pkt.api_status.api_completion_fence_value = seq; amdgpu_ring_write_multiple(ring, &mes_status_pkt, sizeof(mes_status_pkt) / 4); amdgpu_ring_commit(ring); - spin_unlock_irqrestore(&mes->ring_lock, flags); + spin_unlock_irqrestore(&mes->ring_lock[0], flags); op_str = mes_v11_0_get_op_string(x_pkt); misc_op_str = mes_v11_0_get_misc_op_string(x_pkt); @@ -229,7 +235,7 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, dev_dbg(adev->dev, "MES msg=%d was emitted\n", x_pkt->header.opcode); - r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq, timeout); + r = amdgpu_fence_wait_polling(ring, seq, timeout); if (r < 1 || !*status_ptr) { if (misc_op_str) @@ -252,8 +258,12 @@ static int mes_v11_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, amdgpu_device_wb_free(adev, status_offset); return 0; +error_undo: + dev_err(adev->dev, "MES ring buffer is full.\n"); + amdgpu_ring_undo(ring); + error_unlock_free: - spin_unlock_irqrestore(&mes->ring_lock, flags); + spin_unlock_irqrestore(&mes->ring_lock[0], flags); error_wb_free: amdgpu_device_wb_free(adev, status_offset); @@ -512,9 +522,9 @@ static int mes_v11_0_set_hw_resources(struct amdgpu_mes *mes) mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub; mes_set_hw_res_pkt.gds_size = adev->gds.gds_size; mes_set_hw_res_pkt.paging_vmid = 0; - mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr; + mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr[0]; mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr = - mes->query_status_fence_gpu_addr; + mes->query_status_fence_gpu_addr[0]; for (i = 0; i < MAX_COMPUTE_PIPES; i++) mes_set_hw_res_pkt.compute_hqd_mask[i] = @@ -683,6 +693,28 @@ static void mes_v11_0_free_ucode_buffers(struct amdgpu_device *adev, (void **)&adev->mes.ucode_fw_ptr[pipe]); } +static void mes_v11_0_get_fw_version(struct amdgpu_device *adev) +{ + int pipe; + + /* get MES scheduler/KIQ versions */ + mutex_lock(&adev->srbm_mutex); + + for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { + soc21_grbm_select(adev, 3, pipe, 0, 0); + + if (pipe == AMDGPU_MES_SCHED_PIPE) + adev->mes.sched_version = + RREG32_SOC15(GC, 0, regCP_MES_GP3_LO); + else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq) + adev->mes.kiq_version = + RREG32_SOC15(GC, 0, regCP_MES_GP3_LO); + } + + soc21_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); +} + static void mes_v11_0_enable(struct amdgpu_device *adev, bool enable) { uint64_t ucode_addr; @@ -1015,7 +1047,7 @@ static int mes_v11_0_kiq_enable_queue(struct amdgpu_device *adev) return r; } - kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring); + kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring[0]); return amdgpu_ring_test_helper(kiq_ring); } @@ -1029,7 +1061,7 @@ static int mes_v11_0_queue_init(struct amdgpu_device *adev, if (pipe == AMDGPU_MES_KIQ_PIPE) ring = &adev->gfx.kiq[0].ring; else if (pipe == AMDGPU_MES_SCHED_PIPE) - ring = &adev->mes.ring; + ring = &adev->mes.ring[0]; else BUG(); @@ -1052,18 +1084,6 @@ static int mes_v11_0_queue_init(struct amdgpu_device *adev, mes_v11_0_queue_init_register(ring); } - /* get MES scheduler/KIQ versions */ - mutex_lock(&adev->srbm_mutex); - soc21_grbm_select(adev, 3, pipe, 0, 0); - - if (pipe == AMDGPU_MES_SCHED_PIPE) - adev->mes.sched_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO); - else if (pipe == AMDGPU_MES_KIQ_PIPE && adev->enable_mes_kiq) - adev->mes.kiq_version = RREG32_SOC15(GC, 0, regCP_MES_GP3_LO); - - soc21_grbm_select(adev, 0, 0, 0, 0); - mutex_unlock(&adev->srbm_mutex); - return 0; } @@ -1071,7 +1091,7 @@ static int mes_v11_0_ring_init(struct amdgpu_device *adev) { struct amdgpu_ring *ring; - ring = &adev->mes.ring; + ring = &adev->mes.ring[0]; ring->funcs = &mes_v11_0_ring_funcs; @@ -1124,7 +1144,7 @@ static int mes_v11_0_mqd_sw_init(struct amdgpu_device *adev, if (pipe == AMDGPU_MES_KIQ_PIPE) ring = &adev->gfx.kiq[0].ring; else if (pipe == AMDGPU_MES_SCHED_PIPE) - ring = &adev->mes.ring; + ring = &adev->mes.ring[0]; else BUG(); @@ -1200,9 +1220,6 @@ static int mes_v11_0_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int pipe; - amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); - amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); - for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { kfree(adev->mes.mqd_backup[pipe]); @@ -1216,12 +1233,12 @@ static int mes_v11_0_sw_fini(void *handle) &adev->gfx.kiq[0].ring.mqd_gpu_addr, &adev->gfx.kiq[0].ring.mqd_ptr); - amdgpu_bo_free_kernel(&adev->mes.ring.mqd_obj, - &adev->mes.ring.mqd_gpu_addr, - &adev->mes.ring.mqd_ptr); + amdgpu_bo_free_kernel(&adev->mes.ring[0].mqd_obj, + &adev->mes.ring[0].mqd_gpu_addr, + &adev->mes.ring[0].mqd_ptr); amdgpu_ring_fini(&adev->gfx.kiq[0].ring); - amdgpu_ring_fini(&adev->mes.ring); + amdgpu_ring_fini(&adev->mes.ring[0]); if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { mes_v11_0_free_ucode_buffers(adev, AMDGPU_MES_KIQ_PIPE); @@ -1313,15 +1330,24 @@ static int mes_v11_0_kiq_hw_init(struct amdgpu_device *adev) mes_v11_0_enable(adev, true); + mes_v11_0_get_fw_version(adev); + mes_v11_0_kiq_setting(&adev->gfx.kiq[0].ring); r = mes_v11_0_queue_init(adev, AMDGPU_MES_KIQ_PIPE); if (r) goto failure; - r = mes_v11_0_hw_init(adev); - if (r) - goto failure; + if ((adev->mes.sched_version & AMDGPU_MES_VERSION_MASK) >= 0x47) + adev->mes.enable_legacy_queue_map = true; + else + adev->mes.enable_legacy_queue_map = false; + + if (adev->mes.enable_legacy_queue_map) { + r = mes_v11_0_hw_init(adev); + if (r) + goto failure; + } return r; @@ -1332,9 +1358,9 @@ failure: static int mes_v11_0_kiq_hw_fini(struct amdgpu_device *adev) { - if (adev->mes.ring.sched.ready) { - mes_v11_0_kiq_dequeue(&adev->mes.ring); - adev->mes.ring.sched.ready = false; + if (adev->mes.ring[0].sched.ready) { + mes_v11_0_kiq_dequeue(&adev->mes.ring[0]); + adev->mes.ring[0].sched.ready = false; } if (amdgpu_sriov_vf(adev)) { @@ -1352,7 +1378,7 @@ static int mes_v11_0_hw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->mes.ring.sched.ready) + if (adev->mes.ring[0].sched.ready) goto out; if (!adev->enable_mes_kiq) { @@ -1397,7 +1423,7 @@ out: * with MES enabled. */ adev->gfx.kiq[0].ring.sched.ready = false; - adev->mes.ring.sched.ready = true; + adev->mes.ring[0].sched.ready = true; return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c index 0713bc3eb263..a79a8adc3aa5 100644 --- a/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mes_v12_0.c @@ -142,19 +142,20 @@ static const char *mes_v12_0_get_misc_op_string(union MESAPI__MISC *x_pkt) } static int mes_v12_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, - void *pkt, int size, - int api_status_off) + int pipe, void *pkt, int size, + int api_status_off) { union MESAPI__QUERY_MES_STATUS mes_status_pkt; signed long timeout = 3000000; /* 3000 ms */ struct amdgpu_device *adev = mes->adev; - struct amdgpu_ring *ring = &mes->ring; + struct amdgpu_ring *ring = &mes->ring[pipe]; + spinlock_t *ring_lock = &mes->ring_lock[pipe]; struct MES_API_STATUS *api_status; union MESAPI__MISC *x_pkt = pkt; const char *op_str, *misc_op_str; unsigned long flags; u64 status_gpu_addr; - u32 status_offset; + u32 seq, status_offset; u64 *status_ptr; signed long r; int ret; @@ -177,11 +178,18 @@ static int mes_v12_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, status_ptr = (u64 *)&adev->wb.wb[status_offset]; *status_ptr = 0; - spin_lock_irqsave(&mes->ring_lock, flags); + spin_lock_irqsave(ring_lock, flags); r = amdgpu_ring_alloc(ring, (size + sizeof(mes_status_pkt)) / 4); if (r) goto error_unlock_free; + seq = ++ring->fence_drv.sync_seq; + r = amdgpu_fence_wait_polling(ring, + seq - ring->fence_drv.num_fences_mask, + timeout); + if (r < 1) + goto error_undo; + api_status = (struct MES_API_STATUS *)((char *)pkt + api_status_off); api_status->api_completion_fence_addr = status_gpu_addr; api_status->api_completion_fence_value = 1; @@ -194,39 +202,39 @@ static int mes_v12_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; mes_status_pkt.api_status.api_completion_fence_addr = ring->fence_drv.gpu_addr; - mes_status_pkt.api_status.api_completion_fence_value = - ++ring->fence_drv.sync_seq; + mes_status_pkt.api_status.api_completion_fence_value = seq; amdgpu_ring_write_multiple(ring, &mes_status_pkt, sizeof(mes_status_pkt) / 4); amdgpu_ring_commit(ring); - spin_unlock_irqrestore(&mes->ring_lock, flags); + spin_unlock_irqrestore(ring_lock, flags); op_str = mes_v12_0_get_op_string(x_pkt); misc_op_str = mes_v12_0_get_misc_op_string(x_pkt); if (misc_op_str) - dev_dbg(adev->dev, "MES msg=%s (%s) was emitted\n", op_str, - misc_op_str); + dev_dbg(adev->dev, "MES(%d) msg=%s (%s) was emitted\n", + pipe, op_str, misc_op_str); else if (op_str) - dev_dbg(adev->dev, "MES msg=%s was emitted\n", op_str); + dev_dbg(adev->dev, "MES(%d) msg=%s was emitted\n", + pipe, op_str); else - dev_dbg(adev->dev, "MES msg=%d was emitted\n", - x_pkt->header.opcode); + dev_dbg(adev->dev, "MES(%d) msg=%d was emitted\n", + pipe, x_pkt->header.opcode); - r = amdgpu_fence_wait_polling(ring, ring->fence_drv.sync_seq, timeout); + r = amdgpu_fence_wait_polling(ring, seq, timeout); if (r < 1 || !*status_ptr) { if (misc_op_str) - dev_err(adev->dev, "MES failed to respond to msg=%s (%s)\n", - op_str, misc_op_str); + dev_err(adev->dev, "MES(%d) failed to respond to msg=%s (%s)\n", + pipe, op_str, misc_op_str); else if (op_str) - dev_err(adev->dev, "MES failed to respond to msg=%s\n", - op_str); + dev_err(adev->dev, "MES(%d) failed to respond to msg=%s\n", + pipe, op_str); else - dev_err(adev->dev, "MES failed to respond to msg=%d\n", - x_pkt->header.opcode); + dev_err(adev->dev, "MES(%d) failed to respond to msg=%d\n", + pipe, x_pkt->header.opcode); while (halt_if_hws_hang) schedule(); @@ -238,8 +246,12 @@ static int mes_v12_0_submit_pkt_and_poll_completion(struct amdgpu_mes *mes, amdgpu_device_wb_free(adev, status_offset); return 0; +error_undo: + dev_err(adev->dev, "MES ring buffer is full.\n"); + amdgpu_ring_undo(ring); + error_unlock_free: - spin_unlock_irqrestore(&mes->ring_lock, flags); + spin_unlock_irqrestore(ring_lock, flags); error_wb_free: amdgpu_device_wb_free(adev, status_offset); @@ -254,6 +266,8 @@ static int convert_to_mes_queue_type(int queue_type) return MES_QUEUE_TYPE_COMPUTE; else if (queue_type == AMDGPU_RING_TYPE_SDMA) return MES_QUEUE_TYPE_SDMA; + else if (queue_type == AMDGPU_RING_TYPE_MES) + return MES_QUEUE_TYPE_SCHQ; else BUG(); return -1; @@ -311,6 +325,7 @@ static int mes_v12_0_add_hw_queue(struct amdgpu_mes *mes, mes_add_queue_pkt.gds_size = input->queue_size; return mes_v12_0_submit_pkt_and_poll_completion(mes, + AMDGPU_MES_SCHED_PIPE, &mes_add_queue_pkt, sizeof(mes_add_queue_pkt), offsetof(union MESAPI__ADD_QUEUE, api_status)); } @@ -330,6 +345,7 @@ static int mes_v12_0_remove_hw_queue(struct amdgpu_mes *mes, mes_remove_queue_pkt.gang_context_addr = input->gang_context_addr; return mes_v12_0_submit_pkt_and_poll_completion(mes, + AMDGPU_MES_SCHED_PIPE, &mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt), offsetof(union MESAPI__REMOVE_QUEUE, api_status)); } @@ -338,6 +354,7 @@ static int mes_v12_0_map_legacy_queue(struct amdgpu_mes *mes, struct mes_map_legacy_queue_input *input) { union MESAPI__ADD_QUEUE mes_add_queue_pkt; + int pipe; memset(&mes_add_queue_pkt, 0, sizeof(mes_add_queue_pkt)); @@ -354,7 +371,12 @@ static int mes_v12_0_map_legacy_queue(struct amdgpu_mes *mes, convert_to_mes_queue_type(input->queue_type); mes_add_queue_pkt.map_legacy_kq = 1; - return mes_v12_0_submit_pkt_and_poll_completion(mes, + if (mes->adev->enable_uni_mes) + pipe = AMDGPU_MES_KIQ_PIPE; + else + pipe = AMDGPU_MES_SCHED_PIPE; + + return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe, &mes_add_queue_pkt, sizeof(mes_add_queue_pkt), offsetof(union MESAPI__ADD_QUEUE, api_status)); } @@ -363,6 +385,7 @@ static int mes_v12_0_unmap_legacy_queue(struct amdgpu_mes *mes, struct mes_unmap_legacy_queue_input *input) { union MESAPI__REMOVE_QUEUE mes_remove_queue_pkt; + int pipe; memset(&mes_remove_queue_pkt, 0, sizeof(mes_remove_queue_pkt)); @@ -387,7 +410,12 @@ static int mes_v12_0_unmap_legacy_queue(struct amdgpu_mes *mes, convert_to_mes_queue_type(input->queue_type); } - return mes_v12_0_submit_pkt_and_poll_completion(mes, + if (mes->adev->enable_uni_mes) + pipe = AMDGPU_MES_KIQ_PIPE; + else + pipe = AMDGPU_MES_SCHED_PIPE; + + return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe, &mes_remove_queue_pkt, sizeof(mes_remove_queue_pkt), offsetof(union MESAPI__REMOVE_QUEUE, api_status)); } @@ -404,7 +432,7 @@ static int mes_v12_0_resume_gang(struct amdgpu_mes *mes, return 0; } -static int mes_v12_0_query_sched_status(struct amdgpu_mes *mes) +static int mes_v12_0_query_sched_status(struct amdgpu_mes *mes, int pipe) { union MESAPI__QUERY_MES_STATUS mes_status_pkt; @@ -414,7 +442,7 @@ static int mes_v12_0_query_sched_status(struct amdgpu_mes *mes) mes_status_pkt.header.opcode = MES_SCH_API_QUERY_SCHEDULER_STATUS; mes_status_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; - return mes_v12_0_submit_pkt_and_poll_completion(mes, + return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe, &mes_status_pkt, sizeof(mes_status_pkt), offsetof(union MESAPI__QUERY_MES_STATUS, api_status)); } @@ -423,6 +451,7 @@ static int mes_v12_0_misc_op(struct amdgpu_mes *mes, struct mes_misc_op_input *input) { union MESAPI__MISC misc_pkt; + int pipe; memset(&misc_pkt, 0, sizeof(misc_pkt)); @@ -475,12 +504,17 @@ static int mes_v12_0_misc_op(struct amdgpu_mes *mes, return -EINVAL; } - return mes_v12_0_submit_pkt_and_poll_completion(mes, + if (mes->adev->enable_uni_mes) + pipe = AMDGPU_MES_KIQ_PIPE; + else + pipe = AMDGPU_MES_SCHED_PIPE; + + return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe, &misc_pkt, sizeof(misc_pkt), offsetof(union MESAPI__MISC, api_status)); } -static int mes_v12_0_set_hw_resources_1(struct amdgpu_mes *mes) +static int mes_v12_0_set_hw_resources_1(struct amdgpu_mes *mes, int pipe) { union MESAPI_SET_HW_RESOURCES_1 mes_set_hw_res_1_pkt; @@ -491,12 +525,12 @@ static int mes_v12_0_set_hw_resources_1(struct amdgpu_mes *mes) mes_set_hw_res_1_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; mes_set_hw_res_1_pkt.mes_kiq_unmap_timeout = 100; - return mes_v12_0_submit_pkt_and_poll_completion(mes, + return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe, &mes_set_hw_res_1_pkt, sizeof(mes_set_hw_res_1_pkt), offsetof(union MESAPI_SET_HW_RESOURCES_1, api_status)); } -static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes) +static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes, int pipe) { int i; struct amdgpu_device *adev = mes->adev; @@ -508,27 +542,33 @@ static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes) mes_set_hw_res_pkt.header.opcode = MES_SCH_API_SET_HW_RSRC; mes_set_hw_res_pkt.header.dwsize = API_FRAME_SIZE_IN_DWORDS; - mes_set_hw_res_pkt.vmid_mask_mmhub = mes->vmid_mask_mmhub; - mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub; - mes_set_hw_res_pkt.gds_size = adev->gds.gds_size; - mes_set_hw_res_pkt.paging_vmid = 0; - mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = mes->sch_ctx_gpu_addr; - mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr = - mes->query_status_fence_gpu_addr; - - for (i = 0; i < MAX_COMPUTE_PIPES; i++) - mes_set_hw_res_pkt.compute_hqd_mask[i] = - mes->compute_hqd_mask[i]; - - for (i = 0; i < MAX_GFX_PIPES; i++) - mes_set_hw_res_pkt.gfx_hqd_mask[i] = mes->gfx_hqd_mask[i]; - - for (i = 0; i < MAX_SDMA_PIPES; i++) - mes_set_hw_res_pkt.sdma_hqd_mask[i] = mes->sdma_hqd_mask[i]; + if (pipe == AMDGPU_MES_SCHED_PIPE) { + mes_set_hw_res_pkt.vmid_mask_mmhub = mes->vmid_mask_mmhub; + mes_set_hw_res_pkt.vmid_mask_gfxhub = mes->vmid_mask_gfxhub; + mes_set_hw_res_pkt.gds_size = adev->gds.gds_size; + mes_set_hw_res_pkt.paging_vmid = 0; + + for (i = 0; i < MAX_COMPUTE_PIPES; i++) + mes_set_hw_res_pkt.compute_hqd_mask[i] = + mes->compute_hqd_mask[i]; + + for (i = 0; i < MAX_GFX_PIPES; i++) + mes_set_hw_res_pkt.gfx_hqd_mask[i] = + mes->gfx_hqd_mask[i]; + + for (i = 0; i < MAX_SDMA_PIPES; i++) + mes_set_hw_res_pkt.sdma_hqd_mask[i] = + mes->sdma_hqd_mask[i]; + + for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++) + mes_set_hw_res_pkt.aggregated_doorbells[i] = + mes->aggregated_doorbells[i]; + } - for (i = 0; i < AMD_PRIORITY_NUM_LEVELS; i++) - mes_set_hw_res_pkt.aggregated_doorbells[i] = - mes->aggregated_doorbells[i]; + mes_set_hw_res_pkt.g_sch_ctx_gpu_mc_ptr = + mes->sch_ctx_gpu_addr[pipe]; + mes_set_hw_res_pkt.query_status_fence_gpu_mc_ptr = + mes->query_status_fence_gpu_addr[pipe]; for (i = 0; i < 5; i++) { mes_set_hw_res_pkt.gc_base[i] = adev->reg_offset[GC_HWIP][0][i]; @@ -556,7 +596,7 @@ static int mes_v12_0_set_hw_resources(struct amdgpu_mes *mes) mes_set_hw_res_pkt.event_intr_history_gpu_mc_ptr = mes->event_log_gpu_addr; } - return mes_v12_0_submit_pkt_and_poll_completion(mes, + return mes_v12_0_submit_pkt_and_poll_completion(mes, pipe, &mes_set_hw_res_pkt, sizeof(mes_set_hw_res_pkt), offsetof(union MESAPI_SET_HW_RESOURCES, api_status)); } @@ -734,16 +774,11 @@ static void mes_v12_0_enable(struct amdgpu_device *adev, bool enable) if (enable) { data = RREG32_SOC15(GC, 0, regCP_MES_CNTL); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1); - data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, - (!adev->enable_uni_mes && adev->enable_mes_kiq) ? 1 : 0); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, 1); WREG32_SOC15(GC, 0, regCP_MES_CNTL, data); mutex_lock(&adev->srbm_mutex); for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { - if ((!adev->enable_mes_kiq || adev->enable_uni_mes) && - pipe == AMDGPU_MES_KIQ_PIPE) - continue; - soc21_grbm_select(adev, 3, pipe, 0, 0); ucode_addr = adev->mes.uc_start_addr[pipe] >> 2; @@ -757,8 +792,7 @@ static void mes_v12_0_enable(struct amdgpu_device *adev, bool enable) /* unhalt MES and activate pipe0 */ data = REG_SET_FIELD(0, CP_MES_CNTL, MES_PIPE0_ACTIVE, 1); - data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, - (!adev->enable_uni_mes && adev->enable_mes_kiq) ? 1 : 0); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_ACTIVE, 1); WREG32_SOC15(GC, 0, regCP_MES_CNTL, data); if (amdgpu_emu_mode) @@ -774,8 +808,7 @@ static void mes_v12_0_enable(struct amdgpu_device *adev, bool enable) data = REG_SET_FIELD(data, CP_MES_CNTL, MES_INVALIDATE_ICACHE, 1); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE0_RESET, 1); - data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, - (!adev->enable_uni_mes && adev->enable_mes_kiq) ? 1 : 0); + data = REG_SET_FIELD(data, CP_MES_CNTL, MES_PIPE1_RESET, 1); data = REG_SET_FIELD(data, CP_MES_CNTL, MES_HALT, 1); WREG32_SOC15(GC, 0, regCP_MES_CNTL, data); } @@ -790,10 +823,6 @@ static void mes_v12_0_set_ucode_start_addr(struct amdgpu_device *adev) mutex_lock(&adev->srbm_mutex); for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { - if ((!adev->enable_mes_kiq || adev->enable_uni_mes) && - pipe == AMDGPU_MES_KIQ_PIPE) - continue; - /* me=3, queue=0 */ soc21_grbm_select(adev, 3, pipe, 0, 0); @@ -1085,7 +1114,7 @@ static int mes_v12_0_kiq_enable_queue(struct amdgpu_device *adev) return r; } - kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring); + kiq->pmf->kiq_map_queues(kiq_ring, &adev->mes.ring[0]); r = amdgpu_ring_test_ring(kiq_ring); if (r) { @@ -1101,14 +1130,12 @@ static int mes_v12_0_queue_init(struct amdgpu_device *adev, struct amdgpu_ring *ring; int r; - if (pipe == AMDGPU_MES_KIQ_PIPE) + if (!adev->enable_uni_mes && pipe == AMDGPU_MES_KIQ_PIPE) ring = &adev->gfx.kiq[0].ring; - else if (pipe == AMDGPU_MES_SCHED_PIPE) - ring = &adev->mes.ring; else - BUG(); + ring = &adev->mes.ring[pipe]; - if ((pipe == AMDGPU_MES_SCHED_PIPE) && + if ((adev->enable_uni_mes || pipe == AMDGPU_MES_SCHED_PIPE) && (amdgpu_in_reset(adev) || adev->in_suspend)) { *(ring->wptr_cpu_addr) = 0; *(ring->rptr_cpu_addr) = 0; @@ -1120,13 +1147,12 @@ static int mes_v12_0_queue_init(struct amdgpu_device *adev, return r; if (pipe == AMDGPU_MES_SCHED_PIPE) { - if (adev->enable_uni_mes) { - mes_v12_0_queue_init_register(ring); - } else { + if (adev->enable_uni_mes) + r = amdgpu_mes_map_legacy_queue(adev, ring); + else r = mes_v12_0_kiq_enable_queue(adev); - if (r) - return r; - } + if (r) + return r; } else { mes_v12_0_queue_init_register(ring); } @@ -1146,25 +1172,29 @@ static int mes_v12_0_queue_init(struct amdgpu_device *adev, return 0; } -static int mes_v12_0_ring_init(struct amdgpu_device *adev) +static int mes_v12_0_ring_init(struct amdgpu_device *adev, int pipe) { struct amdgpu_ring *ring; - ring = &adev->mes.ring; + ring = &adev->mes.ring[pipe]; ring->funcs = &mes_v12_0_ring_funcs; ring->me = 3; - ring->pipe = 0; + ring->pipe = pipe; ring->queue = 0; ring->ring_obj = NULL; ring->use_doorbell = true; - ring->doorbell_index = adev->doorbell_index.mes_ring0 << 1; - ring->eop_gpu_addr = adev->mes.eop_gpu_addr[AMDGPU_MES_SCHED_PIPE]; + ring->eop_gpu_addr = adev->mes.eop_gpu_addr[pipe]; ring->no_scheduler = true; sprintf(ring->name, "mes_%d.%d.%d", ring->me, ring->pipe, ring->queue); + if (pipe == AMDGPU_MES_SCHED_PIPE) + ring->doorbell_index = adev->doorbell_index.mes_ring0 << 1; + else + ring->doorbell_index = adev->doorbell_index.mes_ring1 << 1; + return amdgpu_ring_init(adev, ring, 1024, NULL, 0, AMDGPU_RING_PRIO_DEFAULT, NULL); } @@ -1178,7 +1208,7 @@ static int mes_v12_0_kiq_ring_init(struct amdgpu_device *adev) ring = &adev->gfx.kiq[0].ring; ring->me = 3; - ring->pipe = adev->enable_uni_mes ? 0 : 1; + ring->pipe = 1; ring->queue = 0; ring->adev = NULL; @@ -1200,12 +1230,10 @@ static int mes_v12_0_mqd_sw_init(struct amdgpu_device *adev, int r, mqd_size = sizeof(struct v12_compute_mqd); struct amdgpu_ring *ring; - if (pipe == AMDGPU_MES_KIQ_PIPE) + if (!adev->enable_uni_mes && pipe == AMDGPU_MES_KIQ_PIPE) ring = &adev->gfx.kiq[0].ring; - else if (pipe == AMDGPU_MES_SCHED_PIPE) - ring = &adev->mes.ring; else - BUG(); + ring = &adev->mes.ring[pipe]; if (ring->mqd_obj) return 0; @@ -1238,6 +1266,7 @@ static int mes_v12_0_sw_init(void *handle) adev->mes.funcs = &mes_v12_0_funcs; adev->mes.kiq_hw_init = &mes_v12_0_kiq_hw_init; adev->mes.kiq_hw_fini = &mes_v12_0_kiq_hw_fini; + adev->mes.enable_legacy_queue_map = true; adev->mes.event_log_size = AMDGPU_MES_LOG_BUFFER_SIZE; @@ -1246,9 +1275,6 @@ static int mes_v12_0_sw_init(void *handle) return r; for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { - if (!adev->enable_mes_kiq && pipe == AMDGPU_MES_KIQ_PIPE) - continue; - r = mes_v12_0_allocate_eop_buf(adev, pipe); if (r) return r; @@ -1256,18 +1282,15 @@ static int mes_v12_0_sw_init(void *handle) r = mes_v12_0_mqd_sw_init(adev, pipe); if (r) return r; - } - if (adev->enable_mes_kiq) { - r = mes_v12_0_kiq_ring_init(adev); + if (!adev->enable_uni_mes && pipe == AMDGPU_MES_KIQ_PIPE) + r = mes_v12_0_kiq_ring_init(adev); + else + r = mes_v12_0_ring_init(adev, pipe); if (r) return r; } - r = mes_v12_0_ring_init(adev); - if (r) - return r; - return 0; } @@ -1276,9 +1299,6 @@ static int mes_v12_0_sw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int pipe; - amdgpu_device_wb_free(adev, adev->mes.sch_ctx_offs); - amdgpu_device_wb_free(adev, adev->mes.query_status_fence_offs); - for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { kfree(adev->mes.mqd_backup[pipe]); @@ -1286,18 +1306,21 @@ static int mes_v12_0_sw_fini(void *handle) &adev->mes.eop_gpu_addr[pipe], NULL); amdgpu_ucode_release(&adev->mes.fw[pipe]); - } - - amdgpu_bo_free_kernel(&adev->gfx.kiq[0].ring.mqd_obj, - &adev->gfx.kiq[0].ring.mqd_gpu_addr, - &adev->gfx.kiq[0].ring.mqd_ptr); - amdgpu_bo_free_kernel(&adev->mes.ring.mqd_obj, - &adev->mes.ring.mqd_gpu_addr, - &adev->mes.ring.mqd_ptr); + if (adev->enable_uni_mes || pipe == AMDGPU_MES_SCHED_PIPE) { + amdgpu_bo_free_kernel(&adev->mes.ring[pipe].mqd_obj, + &adev->mes.ring[pipe].mqd_gpu_addr, + &adev->mes.ring[pipe].mqd_ptr); + amdgpu_ring_fini(&adev->mes.ring[pipe]); + } + } - amdgpu_ring_fini(&adev->gfx.kiq[0].ring); - amdgpu_ring_fini(&adev->mes.ring); + if (!adev->enable_uni_mes) { + amdgpu_bo_free_kernel(&adev->gfx.kiq[0].ring.mqd_obj, + &adev->gfx.kiq[0].ring.mqd_gpu_addr, + &adev->gfx.kiq[0].ring.mqd_ptr); + amdgpu_ring_fini(&adev->gfx.kiq[0].ring); + } if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { mes_v12_0_free_ucode_buffers(adev, AMDGPU_MES_KIQ_PIPE); @@ -1341,7 +1364,7 @@ static void mes_v12_0_kiq_dequeue_sched(struct amdgpu_device *adev) soc21_grbm_select(adev, 0, 0, 0, 0); mutex_unlock(&adev->srbm_mutex); - adev->mes.ring.sched.ready = false; + adev->mes.ring[0].sched.ready = false; } static void mes_v12_0_kiq_setting(struct amdgpu_ring *ring) @@ -1362,10 +1385,10 @@ static int mes_v12_0_kiq_hw_init(struct amdgpu_device *adev) { int r = 0; - mes_v12_0_kiq_setting(&adev->gfx.kiq[0].ring); - if (adev->enable_uni_mes) - return mes_v12_0_hw_init(adev); + mes_v12_0_kiq_setting(&adev->mes.ring[AMDGPU_MES_KIQ_PIPE]); + else + mes_v12_0_kiq_setting(&adev->gfx.kiq[0].ring); if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { @@ -1392,9 +1415,19 @@ static int mes_v12_0_kiq_hw_init(struct amdgpu_device *adev) if (r) goto failure; - r = mes_v12_0_hw_init(adev); - if (r) - goto failure; + if (adev->enable_uni_mes) { + r = mes_v12_0_set_hw_resources(&adev->mes, AMDGPU_MES_KIQ_PIPE); + if (r) + goto failure; + + mes_v12_0_set_hw_resources_1(&adev->mes, AMDGPU_MES_KIQ_PIPE); + } + + if (adev->mes.enable_legacy_queue_map) { + r = mes_v12_0_hw_init(adev); + if (r) + goto failure; + } return r; @@ -1405,9 +1438,15 @@ failure: static int mes_v12_0_kiq_hw_fini(struct amdgpu_device *adev) { - if (adev->mes.ring.sched.ready) { - mes_v12_0_kiq_dequeue_sched(adev); - adev->mes.ring.sched.ready = false; + if (adev->mes.ring[0].sched.ready) { + if (adev->enable_uni_mes) + amdgpu_mes_unmap_legacy_queue(adev, + &adev->mes.ring[AMDGPU_MES_SCHED_PIPE], + RESET_QUEUES, 0, 0); + else + mes_v12_0_kiq_dequeue_sched(adev); + + adev->mes.ring[0].sched.ready = false; } mes_v12_0_enable(adev, false); @@ -1420,10 +1459,10 @@ static int mes_v12_0_hw_init(void *handle) int r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (adev->mes.ring.sched.ready) + if (adev->mes.ring[0].sched.ready) goto out; - if (!adev->enable_mes_kiq || adev->enable_uni_mes) { + if (!adev->enable_mes_kiq) { if (adev->firmware.load_type == AMDGPU_FW_LOAD_DIRECT) { r = mes_v12_0_load_microcode(adev, AMDGPU_MES_SCHED_PIPE, true); @@ -1443,23 +1482,23 @@ static int mes_v12_0_hw_init(void *handle) mes_v12_0_enable(adev, true); } + /* Enable the MES to handle doorbell ring on unmapped queue */ + mes_v12_0_enable_unmapped_doorbell_handling(&adev->mes, true); + r = mes_v12_0_queue_init(adev, AMDGPU_MES_SCHED_PIPE); if (r) goto failure; - r = mes_v12_0_set_hw_resources(&adev->mes); + r = mes_v12_0_set_hw_resources(&adev->mes, AMDGPU_MES_SCHED_PIPE); if (r) goto failure; if (adev->enable_uni_mes) - mes_v12_0_set_hw_resources_1(&adev->mes); + mes_v12_0_set_hw_resources_1(&adev->mes, AMDGPU_MES_SCHED_PIPE); mes_v12_0_init_aggregated_doorbell(&adev->mes); - /* Enable the MES to handle doorbell ring on unmapped queue */ - mes_v12_0_enable_unmapped_doorbell_handling(&adev->mes, true); - - r = mes_v12_0_query_sched_status(&adev->mes); + r = mes_v12_0_query_sched_status(&adev->mes, AMDGPU_MES_SCHED_PIPE); if (r) { DRM_ERROR("MES is busy\n"); goto failure; @@ -1472,7 +1511,7 @@ out: * with MES enabled. */ adev->gfx.kiq[0].ring.sched.ready = false; - adev->mes.ring.sched.ready = true; + adev->mes.ring[0].sched.ready = true; return 0; @@ -1515,17 +1554,7 @@ static int mes_v12_0_early_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int pipe, r; - if (adev->enable_uni_mes) { - r = amdgpu_mes_init_microcode(adev, AMDGPU_MES_SCHED_PIPE); - if (!r) - return 0; - - adev->enable_uni_mes = false; - } - for (pipe = 0; pipe < AMDGPU_MAX_MES_PIPES; pipe++) { - if (!adev->enable_mes_kiq && pipe == AMDGPU_MES_KIQ_PIPE) - continue; r = amdgpu_mes_init_microcode(adev, pipe); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c index 5bbaa2b2caab..0fbc3be81f14 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v4_1_0.c @@ -80,7 +80,8 @@ static uint32_t mmhub_v4_1_0_get_invalidate_req(unsigned int vmid, /* invalidate using legacy mode on vmid*/ req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, PER_VMID_INVALIDATE_REQ, 1 << vmid); - req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, flush_type); + /* Only use legacy inv on mmhub side */ + req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, FLUSH_TYPE, 0); req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PTES, 1); req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE0, 1); req = REG_SET_FIELD(req, MMVM_INVALIDATE_ENG0_REQ, INVALIDATE_L2_PDE1, 1); diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c index af1e90159ce3..2e72d445415f 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v5_2.c @@ -176,14 +176,16 @@ static void sdma_v5_2_ring_set_wptr(struct amdgpu_ring *ring) DRM_DEBUG("calling WDOORBELL64(0x%08x, 0x%016llx)\n", ring->doorbell_index, ring->wptr << 2); WDOORBELL64(ring->doorbell_index, ring->wptr << 2); - /* SDMA seems to miss doorbells sometimes when powergating kicks in. - * Updating the wptr directly will wake it. This is only safe because - * we disallow gfxoff in begin_use() and then allow it again in end_use(). - */ - WREG32(sdma_v5_2_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR), - lower_32_bits(ring->wptr << 2)); - WREG32(sdma_v5_2_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI), - upper_32_bits(ring->wptr << 2)); + if (amdgpu_ip_version(adev, SDMA0_HWIP, 0) == IP_VERSION(5, 2, 1)) { + /* SDMA seems to miss doorbells sometimes when powergating kicks in. + * Updating the wptr directly will wake it. This is only safe because + * we disallow gfxoff in begin_use() and then allow it again in end_use(). + */ + WREG32(sdma_v5_2_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR), + lower_32_bits(ring->wptr << 2)); + WREG32(sdma_v5_2_get_reg_offset(adev, ring->me, mmSDMA0_GFX_RB_WPTR_HI), + upper_32_bits(ring->wptr << 2)); + } } else { DRM_DEBUG("Not using doorbell -- " "mmSDMA%i_GFX_RB_WPTR == 0x%08x " diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c index 41b5e45697dc..ecee9e7d7e4c 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v7_0.c @@ -1575,8 +1575,7 @@ static void sdma_v7_0_emit_copy_buffer(struct amdgpu_ib *ib, ib->ptr[ib->length_dw++] = SDMA_PKT_COPY_LINEAR_HEADER_OP(SDMA_OP_COPY) | SDMA_PKT_COPY_LINEAR_HEADER_SUB_OP(SDMA_SUBOP_COPY_LINEAR) | SDMA_PKT_COPY_LINEAR_HEADER_TMZ((copy_flags & AMDGPU_COPY_FLAGS_TMZ) ? 1 : 0) | - SDMA_PKT_COPY_LINEAR_HEADER_CPV((copy_flags & - (AMDGPU_COPY_FLAGS_READ_DECOMPRESSED | AMDGPU_COPY_FLAGS_WRITE_COMPRESSED)) ? 1 : 0); + SDMA_PKT_COPY_LINEAR_HEADER_CPV(1); ib->ptr[ib->length_dw++] = byte_count - 1; ib->ptr[ib->length_dw++] = 0; /* src/dst endian swap */ @@ -1590,6 +1589,8 @@ static void sdma_v7_0_emit_copy_buffer(struct amdgpu_ib *ib, ((copy_flags & AMDGPU_COPY_FLAGS_READ_DECOMPRESSED) ? SDMA_DCC_READ_CM(2) : 0) | ((copy_flags & AMDGPU_COPY_FLAGS_WRITE_COMPRESSED) ? SDMA_DCC_WRITE_CM(1) : 0) | SDMA_DCC_MAX_COM(max_com) | SDMA_DCC_MAX_UCOM(1); + else + ib->ptr[ib->length_dw++] = 0; } /** @@ -1616,7 +1617,7 @@ static void sdma_v7_0_emit_fill_buffer(struct amdgpu_ib *ib, static const struct amdgpu_buffer_funcs sdma_v7_0_buffer_funcs = { .copy_max_bytes = 0x400000, - .copy_num_dw = 7, + .copy_num_dw = 8, .emit_copy_buffer = sdma_v7_0_emit_copy_buffer, .fill_max_bytes = 0x400000, .fill_num_dw = 5, diff --git a/drivers/gpu/drm/amd/amdgpu/soc15d.h b/drivers/gpu/drm/amd/amdgpu/soc15d.h index 2357ff39323f..e74e1983da53 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15d.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15d.h @@ -76,6 +76,12 @@ ((cond & 0xF) << 24) | \ ((type & 0xF) << 28)) +#define CP_PACKETJ_NOP 0x60000000 +#define CP_PACKETJ_GET_REG(x) ((x) & 0x3FFFF) +#define CP_PACKETJ_GET_RES(x) (((x) >> 18) & 0x3F) +#define CP_PACKETJ_GET_COND(x) (((x) >> 24) & 0xF) +#define CP_PACKETJ_GET_TYPE(x) (((x) >> 28) & 0xF) + /* Packet 3 types */ #define PACKET3_NOP 0x10 #define PACKET3_SET_BASE 0x11 diff --git a/drivers/gpu/drm/amd/amdgpu/soc24.c b/drivers/gpu/drm/amd/amdgpu/soc24.c index 7d641d0dadba..b0c3678cfb31 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc24.c +++ b/drivers/gpu/drm/amd/amdgpu/soc24.c @@ -406,6 +406,7 @@ static int soc24_common_early_init(void *handle) AMD_CG_SUPPORT_ATHUB_MGCG | AMD_CG_SUPPORT_ATHUB_LS | AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_HDP_SD | AMD_CG_SUPPORT_MC_LS; adev->pg_flags = AMD_PG_SUPPORT_VCN | AMD_PG_SUPPORT_JPEG | @@ -424,6 +425,7 @@ static int soc24_common_early_init(void *handle) AMD_CG_SUPPORT_ATHUB_MGCG | AMD_CG_SUPPORT_ATHUB_LS | AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_HDP_SD | AMD_CG_SUPPORT_MC_LS; adev->pg_flags = AMD_PG_SUPPORT_VCN | |