Merge tag 'amd-drm-next-6.5-2023-06-09' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.5-2023-06-02:

amdgpu:
- SR-IOV fixes
- Warning fixes
- Misc code cleanups and spelling fixes
- DCN 3.2 updates
- Improved DC FAMS support for better power management
- Improved DC SubVP support for better power management
- DCN 3.1.x fixes
- Max IB size query
- DC GPU reset fixes
- RAS updates
- DCN 3.0.x fixes
- S/G display fixes
- CP shadow buffer support
- Implement connector force callback
- Z8 power improvements
- PSP 13.0.10 vbflash support
- Mode2 reset fixes
- Store MQDs in VRAM to improve queue switch latency
- VCN 3.x fixes
- JPEG 3.x fixes
- Enable DC_FP on LoongArch
- GFXOFF fixes
- GC 9.4.3 partition support
- SDMA 4.4.2 partition support
- VCN/JPEG 4.0.3 partition support
- VCN 4.0.3 updates
- NBIO 7.9 updates
- GC 9.4.3 updates
- Take NUMA into account when allocating memory
- Handle NUMA for partitions
- SMU 13.0.6 updates
- GC 9.4.3 RAS updates
- Stop including unused swiotlb.h
- SMU 13.0.7 fixes
- Fix clock output ordering on some APUs
- Clean up DC FPGA code
- GFX9 preemption fixes
- Misc irq fixes
- S0ix fixes
- Add new DRM_AMDGPU_WERROR config parameter to help with CI
- PCIe fix for RDNA2
- kdoc fixes
- Documentation updates

amdkfd:
- Query TTM mem limit rather than hardcoding it
- GC 9.4.3 partition support
- Handle NUMA for partitions

radeon:
- Fix possible double free
- Stop including unused swiotlb.h
- Fix possible division by zero

ttm:
- Add query for TTM mem limit
- Add NUMA awareness to pools
- Export ttm_pool_fini()

UAPI:
- Add new ctx query flag to better handle GPU resets
  Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/22290
- Add new interface to query and set shadow buffer for RDNA3
  Mesa MR: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/21986
- Add new INFO query for max IB size
  Proposed userspace: https://gitlab.freedesktop.org/bnieuwenhuizen/mesa/-/commits/ib-rejection-v3

amd-drm-next-6.5-2023-06-09:

amdgpu:
- S0ix fixes
- Initial SMU13 Overdrive support
- kdoc fixes
- Misc clode cleanups
- Flexible array fixes
- Display OTG fixes
- SMU 13.0.6 updates
- Revert some broken clock counter updates
- Misc display fixes
- GFX9 preemption fixes
- Add support for newer EEPROM bad page table format
- Add missing radeon secondary id
- Add support for new colorspace KMS API
- CSA fix
- Stable pstate fixes for APUs
- make vbl interface admin only
- Handle PCI accelerator class

amdkfd:
- Add debugger support for gdb

radeon:
- Fix possible UAF

drm:
- Add Colorspace functionality

UAPI:
- Add debugger interface for enabling gdb
  Proposed userspace: https://github.com/ROCm-Developer-Tools/ROCdbgapi/tree/wip-dbgapi
- Add KMS colorspace API
  Discussion: https://lists.freedesktop.org/archives/dri-devel/2023-June/408128.html

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230609174817.7764-1-alexander.deucher@amd.com

1 files changed, 371 insertions, 52 deletions

diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
index fdbfd725841f..601bb9f68048 100644
--- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
+++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c
@@ -32,6 +32,22 @@
 #include "gc/gc_9_0_sh_mask.h"
 #include "sdma0/sdma0_4_0_sh_mask.h"
 #include "amdgpu_amdkfd.h"
+#include "kfd_device_queue_manager.h"
+
+static void update_mqd(struct mqd_manager *mm, void *mqd,
+		       struct queue_properties *q,
+		       struct mqd_update_info *minfo);
+
+static uint64_t mqd_stride_v9(struct mqd_manager *mm,
+				struct queue_properties *q)
+{
+	if (mm->dev->kfd->cwsr_enabled &&
+	    q->type == KFD_QUEUE_TYPE_COMPUTE)
+		return ALIGN(q->ctl_stack_size, PAGE_SIZE) +
+			ALIGN(sizeof(struct v9_mqd), PAGE_SIZE);
+
+	return mm->mqd_size;
+}
 
 static inline struct v9_mqd *get_mqd(void *mqd)
 {
@@ -49,8 +65,7 @@ static void update_cu_mask(struct mqd_manager *mm, void *mqd,
 	struct v9_mqd *m;
 	uint32_t se_mask[KFD_MAX_NUM_SE] = {0};
 
-	if (!minfo || (minfo->update_flag != UPDATE_FLAG_CU_MASK) ||
-	    !minfo->cu_mask.ptr)
+	if (!minfo || !minfo->cu_mask.ptr)
 		return;
 
 	mqd_symmetrically_map_cu_mask(mm,
@@ -83,7 +98,7 @@ static void set_priority(struct v9_mqd *m, struct queue_properties *q)
 	m->cp_hqd_queue_priority = q->priority;
 }
 
-static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
+static struct kfd_mem_obj *allocate_mqd(struct kfd_node *node,
 		struct queue_properties *q)
 {
 	int retval;
@@ -105,28 +120,30 @@ static struct kfd_mem_obj *allocate_mqd(struct kfd_dev *kfd,
 	 * pass a special bo flag AMDGPU_GEM_CREATE_CP_MQD_GFX9 to instruct
 	 * amdgpu memory functions to do so.
 	 */
-	if (kfd->cwsr_enabled && (q->type == KFD_QUEUE_TYPE_COMPUTE)) {
+	if (node->kfd->cwsr_enabled && (q->type == KFD_QUEUE_TYPE_COMPUTE)) {
 		mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL);
 		if (!mqd_mem_obj)
 			return NULL;
-		retval = amdgpu_amdkfd_alloc_gtt_mem(kfd->adev,
-			ALIGN(q->ctl_stack_size, PAGE_SIZE) +
-				ALIGN(sizeof(struct v9_mqd), PAGE_SIZE),
+		retval = amdgpu_amdkfd_alloc_gtt_mem(node->adev,
+			(ALIGN(q->ctl_stack_size, PAGE_SIZE) +
+			ALIGN(sizeof(struct v9_mqd), PAGE_SIZE)) *
+			NUM_XCC(node->xcc_mask),
 			&(mqd_mem_obj->gtt_mem),
 			&(mqd_mem_obj->gpu_addr),
 			(void *)&(mqd_mem_obj->cpu_ptr), true);
+
+		if (retval) {
+			kfree(mqd_mem_obj);
+			return NULL;
+		}
 	} else {
-		retval = kfd_gtt_sa_allocate(kfd, sizeof(struct v9_mqd),
+		retval = kfd_gtt_sa_allocate(node, sizeof(struct v9_mqd),
 				&mqd_mem_obj);
-	}
-
-	if (retval) {
-		kfree(mqd_mem_obj);
-		return NULL;
+		if (retval)
+			return NULL;
 	}
 
 	return mqd_mem_obj;
-
 }
 
 static void init_mqd(struct mqd_manager *mm, void **mqd,
@@ -135,7 +152,6 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 {
 	uint64_t addr;
 	struct v9_mqd *m;
-	struct amdgpu_device *adev = (struct amdgpu_device *)mm->dev->adev;
 
 	m = (struct v9_mqd *) mqd_mem_obj->cpu_ptr;
 	addr = mqd_mem_obj->gpu_addr;
@@ -165,31 +181,21 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 			1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT |
 			1 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT;
 
-	if (q->format == KFD_QUEUE_FORMAT_AQL) {
+	/* Set cp_hqd_hq_scheduler0 bit 14 to 1 to have the CP set up the
+	 * DISPATCH_PTR.  This is required for the kfd debugger
+	 */
+	m->cp_hqd_hq_status0 = 1 << 14;
+
+	if (q->format == KFD_QUEUE_FORMAT_AQL)
 		m->cp_hqd_aql_control =
 			1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT;
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) {
-			/* On GC 9.4.3, DW 41 is re-purposed as
-			 * compute_tg_chunk_size.
-			 * TODO: review this setting when active CUs in the
-			 * partition play a role
-			 */
-			m->compute_static_thread_mgmt_se6 = 1;
-		}
-	} else {
-		/* PM4 queue */
-		if (adev->ip_versions[GC_HWIP][0] == IP_VERSION(9, 4, 3)) {
-			m->compute_static_thread_mgmt_se6 = 0;
-			/* TODO: program pm4_target_xcc */
-		}
-	}
 
 	if (q->tba_addr) {
 		m->compute_pgm_rsrc2 |=
 			(1 << COMPUTE_PGM_RSRC2__TRAP_PRESENT__SHIFT);
 	}
 
-	if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address) {
+	if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address) {
 		m->cp_hqd_persistent_state |=
 			(1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT);
 		m->cp_hqd_ctx_save_base_addr_lo =
@@ -205,7 +211,7 @@ static void init_mqd(struct mqd_manager *mm, void **mqd,
 	*mqd = m;
 	if (gart_addr)
 		*gart_addr = addr;
-	mm->update_mqd(mm, m, q, NULL);
+	update_mqd(mm, m, q, NULL);
 }
 
 static int load_mqd(struct mqd_manager *mm, void *mqd,
@@ -217,14 +223,13 @@ static int load_mqd(struct mqd_manager *mm, void *mqd,
 
 	return mm->dev->kfd2kgd->hqd_load(mm->dev->adev, mqd, pipe_id, queue_id,
 					  (uint32_t __user *)p->write_ptr,
-					  wptr_shift, 0, mms);
+					  wptr_shift, 0, mms, 0);
 }
 
 static void update_mqd(struct mqd_manager *mm, void *mqd,
 			struct queue_properties *q,
 			struct mqd_update_info *minfo)
 {
-	struct amdgpu_device *adev = (struct amdgpu_device *)mm->dev->adev;
 	struct v9_mqd *m;
 
 	m = get_mqd(mqd);
@@ -257,9 +262,14 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 	 * Limit is 0xFF EOP entries (= 0x7F8 dwords). CP will not submit
 	 * more than (EOP entry count - 1) so a queue size of 0x800 dwords
 	 * is safe, giving a maximum field value of 0xA.
+	 *
+	 * Also, do calculation only if EOP is used (size > 0), otherwise
+	 * the order_base_2 calculation provides incorrect result.
+	 *
 	 */
-	m->cp_hqd_eop_control = min(0xA,
-		order_base_2(q->eop_ring_buffer_size / 4) - 1);
+	m->cp_hqd_eop_control = q->eop_ring_buffer_size ?
+		min(0xA, order_base_2(q->eop_ring_buffer_size / 4) - 1) : 0;
+
 	m->cp_hqd_eop_base_addr_lo =
 			lower_32_bits(q->eop_ring_buffer_address >> 8);
 	m->cp_hqd_eop_base_addr_hi =
@@ -270,17 +280,14 @@ static void update_mqd(struct mqd_manager *mm, void *mqd,
 	m->cp_hqd_vmid = q->vmid;
 
 	if (q->format == KFD_QUEUE_FORMAT_AQL) {
-		m->cp_hqd_pq_control |=
+		m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK |
 				2 << CP_HQD_PQ_CONTROL__SLOT_BASED_WPTR__SHIFT |
 				1 << CP_HQD_PQ_CONTROL__QUEUE_FULL_EN__SHIFT |
 				1 << CP_HQD_PQ_CONTROL__WPP_CLAMP_EN__SHIFT;
-		if (adev->ip_versions[GC_HWIP][0] != IP_VERSION(9, 4, 3))
-			m->cp_hqd_pq_control |=
-				 CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
 		m->cp_hqd_pq_doorbell_control |= 1 <<
 			CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_BIF_DROP__SHIFT;
 	}
-	if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address)
+	if (mm->dev->kfd->cwsr_enabled && q->ctx_save_restore_area_address)
 		m->cp_hqd_ctx_save_control = 0;
 
 	update_cu_mask(mm, mqd, minfo);
@@ -298,11 +305,13 @@ static uint32_t read_doorbell_id(void *mqd)
 }
 
 static int get_wave_state(struct mqd_manager *mm, void *mqd,
+			  struct queue_properties *q,
 			  void __user *ctl_stack,
 			  u32 *ctl_stack_used_size,
 			  u32 *save_area_used_size)
 {
 	struct v9_mqd *m;
+	struct kfd_context_save_area_header header;
 
 	/* Control stack is located one page after MQD. */
 	void *mqd_ctl_stack = (void *)((uintptr_t)mqd + PAGE_SIZE);
@@ -314,7 +323,18 @@ static int get_wave_state(struct mqd_manager *mm, void *mqd,
 	*save_area_used_size = m->cp_hqd_wg_state_offset -
 		m->cp_hqd_cntl_stack_size;
 
-	if (copy_to_user(ctl_stack, mqd_ctl_stack, m->cp_hqd_cntl_stack_size))
+	header.wave_state.control_stack_size = *ctl_stack_used_size;
+	header.wave_state.wave_state_size = *save_area_used_size;
+
+	header.wave_state.wave_state_offset = m->cp_hqd_wg_state_offset;
+	header.wave_state.control_stack_offset = m->cp_hqd_cntl_stack_offset;
+
+	if (copy_to_user(ctl_stack, &header, sizeof(header.wave_state)))
+		return -EFAULT;
+
+	if (copy_to_user(ctl_stack + m->cp_hqd_cntl_stack_offset,
+				mqd_ctl_stack + m->cp_hqd_cntl_stack_offset,
+				*ctl_stack_used_size))
 		return -EFAULT;
 
 	return 0;
@@ -467,6 +487,288 @@ static void restore_mqd_sdma(struct mqd_manager *mm, void **mqd,
 	qp->is_active = 0;
 }
 
+static void init_mqd_hiq_v9_4_3(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *q)
+{
+	struct v9_mqd *m;
+	int xcc = 0;
+	struct kfd_mem_obj xcc_mqd_mem_obj;
+	uint64_t xcc_gart_addr = 0;
+
+	memset(&xcc_mqd_mem_obj, 0x0, sizeof(struct kfd_mem_obj));
+
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		kfd_get_hiq_xcc_mqd(mm->dev, &xcc_mqd_mem_obj, xcc);
+
+		init_mqd(mm, (void **)&m, &xcc_mqd_mem_obj, &xcc_gart_addr, q);
+
+		m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK |
+					1 << CP_HQD_PQ_CONTROL__PRIV_STATE__SHIFT |
+					1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT;
+		m->cp_mqd_stride_size = kfd_hiq_mqd_stride(mm->dev);
+		if (xcc == 0) {
+			/* Set no_update_rptr = 0 in Master XCC */
+			m->cp_hqd_pq_control &= ~CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
+
+			/* Set the MQD pointer and gart address to XCC0 MQD */
+			*mqd = m;
+			*gart_addr = xcc_gart_addr;
+		}
+	}
+}
+
+static int hiq_load_mqd_kiq_v9_4_3(struct mqd_manager *mm, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			struct queue_properties *p, struct mm_struct *mms)
+{
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	void *xcc_mqd;
+	uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev);
+
+	for_each_inst(xcc_id, xcc_mask) {
+		xcc_mqd = mqd + hiq_mqd_size * inst;
+		err = mm->dev->kfd2kgd->hiq_mqd_load(mm->dev->adev, xcc_mqd,
+						     pipe_id, queue_id,
+						     p->doorbell_off, xcc_id);
+		if (err) {
+			pr_debug("Failed to load HIQ MQD for XCC: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static int destroy_hiq_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+			enum kfd_preempt_type type, unsigned int timeout,
+			uint32_t pipe_id, uint32_t queue_id)
+{
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	void *xcc_mqd;
+	uint64_t hiq_mqd_size = kfd_hiq_mqd_stride(mm->dev);
+
+	for_each_inst(xcc_id, xcc_mask) {
+		xcc_mqd = mqd + hiq_mqd_size * inst;
+		err = mm->dev->kfd2kgd->hqd_destroy(mm->dev->adev, xcc_mqd,
+						    type, timeout, pipe_id,
+						    queue_id, xcc_id);
+		if (err) {
+			pr_debug("Destroy MQD failed for xcc: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static void get_xcc_mqd(struct kfd_mem_obj *mqd_mem_obj,
+			       struct kfd_mem_obj *xcc_mqd_mem_obj,
+			       uint64_t offset)
+{
+	xcc_mqd_mem_obj->gtt_mem = (offset == 0) ?
+					mqd_mem_obj->gtt_mem : NULL;
+	xcc_mqd_mem_obj->gpu_addr = mqd_mem_obj->gpu_addr + offset;
+	xcc_mqd_mem_obj->cpu_ptr = (uint32_t *)((uintptr_t)mqd_mem_obj->cpu_ptr
+						+ offset);
+}
+
+static void init_mqd_v9_4_3(struct mqd_manager *mm, void **mqd,
+			struct kfd_mem_obj *mqd_mem_obj, uint64_t *gart_addr,
+			struct queue_properties *q)
+{
+	struct v9_mqd *m;
+	int xcc = 0;
+	struct kfd_mem_obj xcc_mqd_mem_obj;
+	uint64_t xcc_gart_addr = 0;
+	uint64_t xcc_ctx_save_restore_area_address;
+	uint64_t offset = mm->mqd_stride(mm, q);
+	uint32_t local_xcc_start = mm->dev->dqm->current_logical_xcc_start++;
+
+	memset(&xcc_mqd_mem_obj, 0x0, sizeof(struct kfd_mem_obj));
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		get_xcc_mqd(mqd_mem_obj, &xcc_mqd_mem_obj, offset*xcc);
+
+		init_mqd(mm, (void **)&m, &xcc_mqd_mem_obj, &xcc_gart_addr, q);
+
+		m->cp_mqd_stride_size = offset;
+
+		/*
+		 * Update the CWSR address for each XCC if CWSR is enabled
+		 * and CWSR area is allocated in thunk
+		 */
+		if (mm->dev->kfd->cwsr_enabled &&
+		    q->ctx_save_restore_area_address) {
+			xcc_ctx_save_restore_area_address =
+				q->ctx_save_restore_area_address +
+				(xcc * q->ctx_save_restore_area_size);
+
+			m->cp_hqd_ctx_save_base_addr_lo =
+				lower_32_bits(xcc_ctx_save_restore_area_address);
+			m->cp_hqd_ctx_save_base_addr_hi =
+				upper_32_bits(xcc_ctx_save_restore_area_address);
+		}
+
+		if (q->format == KFD_QUEUE_FORMAT_AQL) {
+			m->compute_tg_chunk_size = 1;
+			m->compute_current_logic_xcc_id =
+					(local_xcc_start + xcc) %
+					NUM_XCC(mm->dev->xcc_mask);
+
+			switch (xcc) {
+			case 0:
+				/* Master XCC */
+				m->cp_hqd_pq_control &=
+					~CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
+				break;
+			default:
+				break;
+			}
+		} else {
+			/* PM4 Queue */
+			m->compute_current_logic_xcc_id = 0;
+			m->compute_tg_chunk_size = 0;
+			m->pm4_target_xcc_in_xcp = q->pm4_target_xcc;
+		}
+
+		if (xcc == 0) {
+			/* Set the MQD pointer and gart address to XCC0 MQD */
+			*mqd = m;
+			*gart_addr = xcc_gart_addr;
+		}
+	}
+}
+
+static void update_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+		      struct queue_properties *q, struct mqd_update_info *minfo)
+{
+	struct v9_mqd *m;
+	int xcc = 0;
+	uint64_t size = mm->mqd_stride(mm, q);
+
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		m = get_mqd(mqd + size * xcc);
+		update_mqd(mm, m, q, minfo);
+
+		if (q->format == KFD_QUEUE_FORMAT_AQL) {
+			switch (xcc) {
+			case 0:
+				/* Master XCC */
+				m->cp_hqd_pq_control &=
+					~CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK;
+				break;
+			default:
+				break;
+			}
+			m->compute_tg_chunk_size = 1;
+		} else {
+			/* PM4 Queue */
+			m->compute_current_logic_xcc_id = 0;
+			m->compute_tg_chunk_size = 0;
+			m->pm4_target_xcc_in_xcp = q->pm4_target_xcc;
+		}
+	}
+}
+
+static int destroy_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+		   enum kfd_preempt_type type, unsigned int timeout,
+		   uint32_t pipe_id, uint32_t queue_id)
+{
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	void *xcc_mqd;
+	struct v9_mqd *m;
+	uint64_t mqd_offset;
+
+	m = get_mqd(mqd);
+	mqd_offset = m->cp_mqd_stride_size;
+
+	for_each_inst(xcc_id, xcc_mask) {
+		xcc_mqd = mqd + mqd_offset * inst;
+		err = mm->dev->kfd2kgd->hqd_destroy(mm->dev->adev, xcc_mqd,
+						    type, timeout, pipe_id,
+						    queue_id, xcc_id);
+		if (err) {
+			pr_debug("Destroy MQD failed for xcc: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static int load_mqd_v9_4_3(struct mqd_manager *mm, void *mqd,
+			uint32_t pipe_id, uint32_t queue_id,
+			struct queue_properties *p, struct mm_struct *mms)
+{
+	/* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */
+	uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0);
+	uint32_t xcc_mask = mm->dev->xcc_mask;
+	int xcc_id, err, inst = 0;
+	void *xcc_mqd;
+	uint64_t mqd_stride_size = mm->mqd_stride(mm, p);
+
+	for_each_inst(xcc_id, xcc_mask) {
+		xcc_mqd = mqd + mqd_stride_size * inst;
+		err = mm->dev->kfd2kgd->hqd_load(
+			mm->dev->adev, xcc_mqd, pipe_id, queue_id,
+			(uint32_t __user *)p->write_ptr, wptr_shift, 0, mms,
+			xcc_id);
+		if (err) {
+			pr_debug("Load MQD failed for xcc: %d\n", inst);
+			break;
+		}
+		++inst;
+	}
+
+	return err;
+}
+
+static int get_wave_state_v9_4_3(struct mqd_manager *mm, void *mqd,
+				 struct queue_properties *q,
+				 void __user *ctl_stack,
+				 u32 *ctl_stack_used_size,
+				 u32 *save_area_used_size)
+{
+	int xcc, err = 0;
+	void *xcc_mqd;
+	void __user *xcc_ctl_stack;
+	uint64_t mqd_stride_size = mm->mqd_stride(mm, q);
+	u32 tmp_ctl_stack_used_size = 0, tmp_save_area_used_size = 0;
+
+	for (xcc = 0; xcc < NUM_XCC(mm->dev->xcc_mask); xcc++) {
+		xcc_mqd = mqd + mqd_stride_size * xcc;
+		xcc_ctl_stack = (void __user *)((uintptr_t)ctl_stack +
+					q->ctx_save_restore_area_size * xcc);
+
+		err = get_wave_state(mm, xcc_mqd, q, xcc_ctl_stack,
+				     &tmp_ctl_stack_used_size,
+				     &tmp_save_area_used_size);
+		if (err)
+			break;
+
+		/*
+		 * Set the ctl_stack_used_size and save_area_used_size to
+		 * ctl_stack_used_size and save_area_used_size of XCC 0 when
+		 * passing the info the user-space.
+		 * For multi XCC, user-space would have to look at the header
+		 * info of each Control stack area to determine the control
+		 * stack size and save area used.
+		 */
+		if (xcc == 0) {
+			*ctl_stack_used_size = tmp_ctl_stack_used_size;
+			*save_area_used_size = tmp_save_area_used_size;
+		}
+	}
+
+	return err;
+}
+
 #if defined(CONFIG_DEBUG_FS)
 
 static int debugfs_show_mqd(struct seq_file *m, void *data)
@@ -486,7 +788,7 @@ static int debugfs_show_mqd_sdma(struct seq_file *m, void *data)
 #endif
 
 struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type,
-		struct kfd_dev *dev)
+		struct kfd_node *dev)
 {
 	struct mqd_manager *mqd;
 
@@ -502,34 +804,50 @@ struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type,
 	switch (type) {
 	case KFD_MQD_TYPE_CP:
 		mqd->allocate_mqd = allocate_mqd;
-		mqd->init_mqd = init_mqd;
 		mqd->free_mqd = kfd_free_mqd_cp;
-		mqd->load_mqd = load_mqd;
-		mqd->update_mqd = update_mqd;
-		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
-		mqd->get_wave_state = get_wave_state;
 		mqd->get_checkpoint_info = get_checkpoint_info;
 		mqd->checkpoint_mqd = checkpoint_mqd;
 		mqd->restore_mqd = restore_mqd;
 		mqd->mqd_size = sizeof(struct v9_mqd);
+		mqd->mqd_stride = mqd_stride_v9;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
+		if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3)) {
+			mqd->init_mqd = init_mqd_v9_4_3;
+			mqd->load_mqd = load_mqd_v9_4_3;
+			mqd->update_mqd = update_mqd_v9_4_3;
+			mqd->destroy_mqd = destroy_mqd_v9_4_3;
+			mqd->get_wave_state = get_wave_state_v9_4_3;
+		} else {
+			mqd->init_mqd = init_mqd;
+			mqd->load_mqd = load_mqd;
+			mqd->update_mqd = update_mqd;
+			mqd->destroy_mqd = kfd_destroy_mqd_cp;
+			mqd->get_wave_state = get_wave_state;
+		}
 		break;
 	case KFD_MQD_TYPE_HIQ:
 		mqd->allocate_mqd = allocate_hiq_mqd;
-		mqd->init_mqd = init_mqd_hiq;
 		mqd->free_mqd = free_mqd_hiq_sdma;
-		mqd->load_mqd = kfd_hiq_load_mqd_kiq;
 		mqd->update_mqd = update_mqd;
-		mqd->destroy_mqd = kfd_destroy_mqd_cp;
 		mqd->is_occupied = kfd_is_occupied_cp;
 		mqd->mqd_size = sizeof(struct v9_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd;
 #endif
 		mqd->read_doorbell_id = read_doorbell_id;
+		if (KFD_GC_VERSION(dev) == IP_VERSION(9, 4, 3)) {
+			mqd->init_mqd = init_mqd_hiq_v9_4_3;
+			mqd->load_mqd = hiq_load_mqd_kiq_v9_4_3;
+			mqd->destroy_mqd = destroy_hiq_mqd_v9_4_3;
+		} else {
+			mqd->init_mqd = init_mqd_hiq;
+			mqd->load_mqd = kfd_hiq_load_mqd_kiq;
+			mqd->destroy_mqd = kfd_destroy_mqd_cp;
+		}
 		break;
 	case KFD_MQD_TYPE_DIQ:
 		mqd->allocate_mqd = allocate_mqd;
@@ -555,6 +873,7 @@ struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type,
 		mqd->checkpoint_mqd = checkpoint_mqd_sdma;
 		mqd->restore_mqd = restore_mqd_sdma;
 		mqd->mqd_size = sizeof(struct v9_sdma_mqd);
+		mqd->mqd_stride = kfd_mqd_stride;
 #if defined(CONFIG_DEBUG_FS)
 		mqd->debugfs_show_mqd = debugfs_show_mqd_sdma;
 #endif