Merge tag 'amd-drm-next-6.18-2025-09-19' of https://gitlab.freedesktop.org/agd5f/linux into drm-next

amd-drm-next-6.18-2025-09-19:

amdgpu:
- Fence drv clean up fix
- DPC fixes
- Misc display fixes
- Support the MMIO remap page as a ttm pool
- JPEG parser updates
- UserQ updates
- VCN ctx handling fixes
- Documentation updates
- Misc cleanups
- SMU 13.0.x updates
- SI DPM updates
- GC 11.x cleaner shader updates
- DMCUB updates
- DML fixes
- Improve fallback handling for pixel encoding
- VCN reset improvements
- DCE6 DC updates
- DSC fixes
- Use devm for i2c buses
- GPUVM locking updates
- GPUVM documentation improvements
- Drop non-DC DCE11 code
- S0ix fixes
- Backlight fix
- SR-IOV fixes

amdkfd:
- SVM updates

Signed-off-by: Dave Airlie <airlied@redhat.com>

From: Alex Deucher <alexander.deucher@amd.com>
Link: https://lore.kernel.org/r/20250919193354.2989255-1-alexander.deucher@amd.com

1 files changed, 111 insertions, 94 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index dbda3a38a2b0..112ce584a5ad 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -128,6 +128,17 @@ struct amdgpu_vm_tlb_seq_struct {
 };
 
 /**
+ * amdgpu_vm_assert_locked - check if VM is correctly locked
+ * @vm: the VM which schould be tested
+ *
+ * Asserts that the VM root PD is locked.
+ */
+static void amdgpu_vm_assert_locked(struct amdgpu_vm *vm)
+{
+	dma_resv_assert_held(vm->root.bo->tbo.base.resv);
+}
+
+/**
  * amdgpu_vm_set_pasid - manage pasid and vm ptr mapping
  *
  * @adev: amdgpu_device pointer
@@ -143,6 +154,8 @@ int amdgpu_vm_set_pasid(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 {
 	int r;
 
+	amdgpu_vm_assert_locked(vm);
+
 	if (vm->pasid == pasid)
 		return 0;
 
@@ -181,12 +194,11 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo)
 	struct amdgpu_bo *bo = vm_bo->bo;
 
 	vm_bo->moved = true;
-	spin_lock(&vm_bo->vm->status_lock);
+	amdgpu_vm_assert_locked(vm);
 	if (bo->tbo.type == ttm_bo_type_kernel)
 		list_move(&vm_bo->vm_status, &vm->evicted);
 	else
 		list_move_tail(&vm_bo->vm_status, &vm->evicted);
-	spin_unlock(&vm_bo->vm->status_lock);
 }
 /**
  * amdgpu_vm_bo_moved - vm_bo is moved
@@ -198,9 +210,8 @@ static void amdgpu_vm_bo_evicted(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo)
 {
-	spin_lock(&vm_bo->vm->status_lock);
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->moved);
-	spin_unlock(&vm_bo->vm->status_lock);
 }
 
 /**
@@ -213,9 +224,8 @@ static void amdgpu_vm_bo_moved(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo)
 {
-	spin_lock(&vm_bo->vm->status_lock);
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->idle);
-	spin_unlock(&vm_bo->vm->status_lock);
 	vm_bo->moved = false;
 }
 
@@ -229,9 +239,9 @@ static void amdgpu_vm_bo_idle(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo)
 {
-	spin_lock(&vm_bo->vm->status_lock);
+	spin_lock(&vm_bo->vm->invalidated_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->invalidated);
-	spin_unlock(&vm_bo->vm->status_lock);
+	spin_unlock(&vm_bo->vm->invalidated_lock);
 }
 
 /**
@@ -244,10 +254,9 @@ static void amdgpu_vm_bo_invalidated(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_evicted_user(struct amdgpu_vm_bo_base *vm_bo)
 {
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	vm_bo->moved = true;
-	spin_lock(&vm_bo->vm->status_lock);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->evicted_user);
-	spin_unlock(&vm_bo->vm->status_lock);
 }
 
 /**
@@ -260,13 +269,11 @@ static void amdgpu_vm_bo_evicted_user(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo)
 {
-	if (vm_bo->bo->parent) {
-		spin_lock(&vm_bo->vm->status_lock);
+	amdgpu_vm_assert_locked(vm_bo->vm);
+	if (vm_bo->bo->parent)
 		list_move(&vm_bo->vm_status, &vm_bo->vm->relocated);
-		spin_unlock(&vm_bo->vm->status_lock);
-	} else {
+	else
 		amdgpu_vm_bo_idle(vm_bo);
-	}
 }
 
 /**
@@ -279,9 +286,8 @@ static void amdgpu_vm_bo_relocated(struct amdgpu_vm_bo_base *vm_bo)
  */
 static void amdgpu_vm_bo_done(struct amdgpu_vm_bo_base *vm_bo)
 {
-	spin_lock(&vm_bo->vm->status_lock);
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	list_move(&vm_bo->vm_status, &vm_bo->vm->done);
-	spin_unlock(&vm_bo->vm->status_lock);
 }
 
 /**
@@ -295,10 +301,13 @@ static void amdgpu_vm_bo_reset_state_machine(struct amdgpu_vm *vm)
 {
 	struct amdgpu_vm_bo_base *vm_bo, *tmp;
 
-	spin_lock(&vm->status_lock);
+	spin_lock(&vm->invalidated_lock);
 	list_splice_init(&vm->done, &vm->invalidated);
 	list_for_each_entry(vm_bo, &vm->invalidated, vm_status)
 		vm_bo->moved = true;
+	spin_unlock(&vm->invalidated_lock);
+
+	amdgpu_vm_assert_locked(vm_bo->vm);
 	list_for_each_entry_safe(vm_bo, tmp, &vm->idle, vm_status) {
 		struct amdgpu_bo *bo = vm_bo->bo;
 
@@ -308,14 +317,13 @@ static void amdgpu_vm_bo_reset_state_machine(struct amdgpu_vm *vm)
 		else if (bo->parent)
 			list_move(&vm_bo->vm_status, &vm_bo->vm->relocated);
 	}
-	spin_unlock(&vm->status_lock);
 }
 
 /**
  * amdgpu_vm_update_shared - helper to update shared memory stat
  * @base: base structure for tracking BO usage in a VM
  *
- * Takes the vm status_lock and updates the shared memory stat. If the basic
+ * Takes the vm stats_lock and updates the shared memory stat. If the basic
  * stat changed (e.g. buffer was moved) amdgpu_vm_update_stats need to be called
  * as well.
  */
@@ -327,7 +335,8 @@ static void amdgpu_vm_update_shared(struct amdgpu_vm_bo_base *base)
 	uint32_t bo_memtype = amdgpu_bo_mem_stats_placement(bo);
 	bool shared;
 
-	spin_lock(&vm->status_lock);
+	dma_resv_assert_held(bo->tbo.base.resv);
+	spin_lock(&vm->stats_lock);
 	shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base);
 	if (base->shared != shared) {
 		base->shared = shared;
@@ -339,7 +348,7 @@ static void amdgpu_vm_update_shared(struct amdgpu_vm_bo_base *base)
 			vm->stats[bo_memtype].drm.private += size;
 		}
 	}
-	spin_unlock(&vm->status_lock);
+	spin_unlock(&vm->stats_lock);
 }
 
 /**
@@ -364,11 +373,11 @@ void amdgpu_vm_bo_update_shared(struct amdgpu_bo *bo)
  *        be bo->tbo.resource
  * @sign: if we should add (+1) or subtract (-1) from the stat
  *
- * Caller need to have the vm status_lock held. Useful for when multiple update
+ * Caller need to have the vm stats_lock held. Useful for when multiple update
  * need to happen at the same time.
  */
 static void amdgpu_vm_update_stats_locked(struct amdgpu_vm_bo_base *base,
-			    struct ttm_resource *res, int sign)
+					  struct ttm_resource *res, int sign)
 {
 	struct amdgpu_vm *vm = base->vm;
 	struct amdgpu_bo *bo = base->bo;
@@ -392,7 +401,8 @@ static void amdgpu_vm_update_stats_locked(struct amdgpu_vm_bo_base *base,
 		 */
 		if (bo->flags & AMDGPU_GEM_CREATE_DISCARDABLE)
 			vm->stats[res_memtype].drm.purgeable += size;
-		if (!(bo->preferred_domains & amdgpu_mem_type_to_domain(res_memtype)))
+		if (!(bo->preferred_domains &
+		      amdgpu_mem_type_to_domain(res_memtype)))
 			vm->stats[bo_memtype].evicted += size;
 	}
 }
@@ -411,9 +421,9 @@ void amdgpu_vm_update_stats(struct amdgpu_vm_bo_base *base,
 {
 	struct amdgpu_vm *vm = base->vm;
 
-	spin_lock(&vm->status_lock);
+	spin_lock(&vm->stats_lock);
 	amdgpu_vm_update_stats_locked(base, res, sign);
-	spin_unlock(&vm->status_lock);
+	spin_unlock(&vm->stats_lock);
 }
 
 /**
@@ -439,10 +449,10 @@ void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base,
 	base->next = bo->vm_bo;
 	bo->vm_bo = base;
 
-	spin_lock(&vm->status_lock);
+	spin_lock(&vm->stats_lock);
 	base->shared = drm_gem_object_is_shared_for_memory_stats(&bo->tbo.base);
 	amdgpu_vm_update_stats_locked(base, bo->tbo.resource, +1);
-	spin_unlock(&vm->status_lock);
+	spin_unlock(&vm->stats_lock);
 
 	if (!amdgpu_vm_is_bo_always_valid(vm, bo))
 		return;
@@ -485,6 +495,42 @@ int amdgpu_vm_lock_pd(struct amdgpu_vm *vm, struct drm_exec *exec,
 }
 
 /**
+ * amdgpu_vm_lock_done_list - lock all BOs on the done list
+ * @vm: vm providing the BOs
+ * @exec: drm execution context
+ * @num_fences: number of extra fences to reserve
+ *
+ * Lock the BOs on the done list in the DRM execution context.
+ */
+int amdgpu_vm_lock_done_list(struct amdgpu_vm *vm, struct drm_exec *exec,
+			     unsigned int num_fences)
+{
+	struct list_head *prev = &vm->done;
+	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo *bo;
+	int ret;
+
+	/* We can only trust prev->next while holding the lock */
+	spin_lock(&vm->invalidated_lock);
+	while (!list_is_head(prev->next, &vm->done)) {
+		bo_va = list_entry(prev->next, typeof(*bo_va), base.vm_status);
+		spin_unlock(&vm->invalidated_lock);
+
+		bo = bo_va->base.bo;
+		if (bo) {
+			ret = drm_exec_prepare_obj(exec, &bo->tbo.base, 1);
+			if (unlikely(ret))
+				return ret;
+		}
+		spin_lock(&vm->invalidated_lock);
+		prev = prev->next;
+	}
+	spin_unlock(&vm->invalidated_lock);
+
+	return 0;
+}
+
+/**
  * amdgpu_vm_move_to_lru_tail - move all BOs to the end of LRU
  *
  * @adev: amdgpu device pointer
@@ -575,7 +621,7 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		       void *param)
 {
 	uint64_t new_vm_generation = amdgpu_vm_generation(adev, vm);
-	struct amdgpu_vm_bo_base *bo_base;
+	struct amdgpu_vm_bo_base *bo_base, *tmp;
 	struct amdgpu_bo *bo;
 	int r;
 
@@ -588,13 +634,7 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			return r;
 	}
 
-	spin_lock(&vm->status_lock);
-	while (!list_empty(&vm->evicted)) {
-		bo_base = list_first_entry(&vm->evicted,
-					   struct amdgpu_vm_bo_base,
-					   vm_status);
-		spin_unlock(&vm->status_lock);
-
+	list_for_each_entry_safe(bo_base, tmp, &vm->evicted, vm_status) {
 		bo = bo_base->bo;
 
 		r = validate(param, bo);
@@ -607,37 +647,21 @@ int amdgpu_vm_validate(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 			vm->update_funcs->map_table(to_amdgpu_bo_vm(bo));
 			amdgpu_vm_bo_relocated(bo_base);
 		}
-		spin_lock(&vm->status_lock);
 	}
-	while (ticket && !list_empty(&vm->evicted_user)) {
-		bo_base = list_first_entry(&vm->evicted_user,
-					   struct amdgpu_vm_bo_base,
-					   vm_status);
-		spin_unlock(&vm->status_lock);
-
-		bo = bo_base->bo;
 
-		if (dma_resv_locking_ctx(bo->tbo.base.resv) != ticket) {
-			struct amdgpu_task_info *ti = amdgpu_vm_get_task_info_vm(vm);
+	if (ticket) {
+		list_for_each_entry_safe(bo_base, tmp, &vm->evicted_user,
+					 vm_status) {
+			bo = bo_base->bo;
+			dma_resv_assert_held(bo->tbo.base.resv);
 
-			pr_warn_ratelimited("Evicted user BO is not reserved\n");
-			if (ti) {
-				pr_warn_ratelimited("pid %d\n", ti->task.pid);
-				amdgpu_vm_put_task_info(ti);
-			}
+			r = validate(param, bo);
+			if (r)
+				return r;
 
-			return -EINVAL;
+			amdgpu_vm_bo_invalidated(bo_base);
 		}
-
-		r = validate(param, bo);
-		if (r)
-			return r;
-
-		amdgpu_vm_bo_invalidated(bo_base);
-
-		spin_lock(&vm->status_lock);
 	}
-	spin_unlock(&vm->status_lock);
 
 	amdgpu_vm_eviction_lock(vm);
 	vm->evicting = false;
@@ -660,13 +684,13 @@ bool amdgpu_vm_ready(struct amdgpu_vm *vm)
 {
 	bool ret;
 
+	amdgpu_vm_assert_locked(vm);
+
 	amdgpu_vm_eviction_lock(vm);
 	ret = !vm->evicting;
 	amdgpu_vm_eviction_unlock(vm);
 
-	spin_lock(&vm->status_lock);
 	ret &= list_empty(&vm->evicted);
-	spin_unlock(&vm->status_lock);
 
 	spin_lock(&vm->immediate.lock);
 	ret &= !vm->immediate.stopped;
@@ -957,16 +981,13 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 			  struct amdgpu_vm *vm, bool immediate)
 {
 	struct amdgpu_vm_update_params params;
-	struct amdgpu_vm_bo_base *entry;
+	struct amdgpu_vm_bo_base *entry, *tmp;
 	bool flush_tlb_needed = false;
-	LIST_HEAD(relocated);
 	int r, idx;
 
-	spin_lock(&vm->status_lock);
-	list_splice_init(&vm->relocated, &relocated);
-	spin_unlock(&vm->status_lock);
+	amdgpu_vm_assert_locked(vm);
 
-	if (list_empty(&relocated))
+	if (list_empty(&vm->relocated))
 		return 0;
 
 	if (!drm_dev_enter(adev_to_drm(adev), &idx))
@@ -982,7 +1003,7 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 	if (r)
 		goto error;
 
-	list_for_each_entry(entry, &relocated, vm_status) {
+	list_for_each_entry(entry, &vm->relocated, vm_status) {
 		/* vm_flush_needed after updating moved PDEs */
 		flush_tlb_needed |= entry->moved;
 
@@ -998,9 +1019,7 @@ int amdgpu_vm_update_pdes(struct amdgpu_device *adev,
 	if (flush_tlb_needed)
 		atomic64_inc(&vm->tlb_seq);
 
-	while (!list_empty(&relocated)) {
-		entry = list_first_entry(&relocated, struct amdgpu_vm_bo_base,
-					 vm_status);
+	list_for_each_entry_safe(entry, tmp, &vm->relocated, vm_status) {
 		amdgpu_vm_bo_idle(entry);
 	}
 
@@ -1227,9 +1246,9 @@ error_free:
 void amdgpu_vm_get_memory(struct amdgpu_vm *vm,
 			  struct amdgpu_mem_stats stats[__AMDGPU_PL_NUM])
 {
-	spin_lock(&vm->status_lock);
+	spin_lock(&vm->stats_lock);
 	memcpy(stats, vm->stats, sizeof(*stats) * __AMDGPU_PL_NUM);
-	spin_unlock(&vm->status_lock);
+	spin_unlock(&vm->stats_lock);
 }
 
 /**
@@ -1596,29 +1615,24 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 			   struct amdgpu_vm *vm,
 			   struct ww_acquire_ctx *ticket)
 {
-	struct amdgpu_bo_va *bo_va;
+	struct amdgpu_bo_va *bo_va, *tmp;
 	struct dma_resv *resv;
 	bool clear, unlock;
 	int r;
 
-	spin_lock(&vm->status_lock);
-	while (!list_empty(&vm->moved)) {
-		bo_va = list_first_entry(&vm->moved, struct amdgpu_bo_va,
-					 base.vm_status);
-		spin_unlock(&vm->status_lock);
-
+	list_for_each_entry_safe(bo_va, tmp, &vm->moved, base.vm_status) {
 		/* Per VM BOs never need to bo cleared in the page tables */
 		r = amdgpu_vm_bo_update(adev, bo_va, false);
 		if (r)
 			return r;
-		spin_lock(&vm->status_lock);
 	}
 
+	spin_lock(&vm->invalidated_lock);
 	while (!list_empty(&vm->invalidated)) {
 		bo_va = list_first_entry(&vm->invalidated, struct amdgpu_bo_va,
 					 base.vm_status);
 		resv = bo_va->base.bo->tbo.base.resv;
-		spin_unlock(&vm->status_lock);
+		spin_unlock(&vm->invalidated_lock);
 
 		/* Try to reserve the BO to avoid clearing its ptes */
 		if (!adev->debug_vm && dma_resv_trylock(resv)) {
@@ -1650,9 +1664,9 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
 		     bo_va->base.bo->tbo.resource->mem_type == TTM_PL_SYSTEM))
 			amdgpu_vm_bo_evicted_user(&bo_va->base);
 
-		spin_lock(&vm->status_lock);
+		spin_lock(&vm->invalidated_lock);
 	}
-	spin_unlock(&vm->status_lock);
+	spin_unlock(&vm->invalidated_lock);
 
 	return 0;
 }
@@ -2181,9 +2195,9 @@ void amdgpu_vm_bo_del(struct amdgpu_device *adev,
 		}
 	}
 
-	spin_lock(&vm->status_lock);
+	spin_lock(&vm->invalidated_lock);
 	list_del(&bo_va->base.vm_status);
-	spin_unlock(&vm->status_lock);
+	spin_unlock(&vm->invalidated_lock);
 
 	list_for_each_entry_safe(mapping, next, &bo_va->valids, list) {
 		list_del(&mapping->list);
@@ -2291,10 +2305,10 @@ void amdgpu_vm_bo_move(struct amdgpu_bo *bo, struct ttm_resource *new_mem,
 	for (bo_base = bo->vm_bo; bo_base; bo_base = bo_base->next) {
 		struct amdgpu_vm *vm = bo_base->vm;
 
-		spin_lock(&vm->status_lock);
+		spin_lock(&vm->stats_lock);
 		amdgpu_vm_update_stats_locked(bo_base, bo->tbo.resource, -1);
 		amdgpu_vm_update_stats_locked(bo_base, new_mem, +1);
-		spin_unlock(&vm->status_lock);
+		spin_unlock(&vm->stats_lock);
 	}
 
 	amdgpu_vm_bo_invalidate(bo, evicted);
@@ -2561,11 +2575,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	INIT_LIST_HEAD(&vm->relocated);
 	INIT_LIST_HEAD(&vm->moved);
 	INIT_LIST_HEAD(&vm->idle);
+	spin_lock_init(&vm->invalidated_lock);
 	INIT_LIST_HEAD(&vm->invalidated);
-	spin_lock_init(&vm->status_lock);
 	INIT_LIST_HEAD(&vm->freed);
 	INIT_LIST_HEAD(&vm->done);
 	INIT_KFIFO(vm->faults);
+	spin_lock_init(&vm->stats_lock);
 
 	r = amdgpu_vm_init_entities(adev, vm);
 	if (r)
@@ -3030,7 +3045,8 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m)
 	unsigned int total_done_objs = 0;
 	unsigned int id = 0;
 
-	spin_lock(&vm->status_lock);
+	amdgpu_vm_assert_locked(vm);
+
 	seq_puts(m, "\tIdle BOs:\n");
 	list_for_each_entry_safe(bo_va, tmp, &vm->idle, base.vm_status) {
 		if (!bo_va->base.bo)
@@ -3068,11 +3084,13 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m)
 	id = 0;
 
 	seq_puts(m, "\tInvalidated BOs:\n");
+	spin_lock(&vm->invalidated_lock);
 	list_for_each_entry_safe(bo_va, tmp, &vm->invalidated, base.vm_status) {
 		if (!bo_va->base.bo)
 			continue;
 		total_invalidated += amdgpu_bo_print_info(id++,	bo_va->base.bo, m);
 	}
+	spin_unlock(&vm->invalidated_lock);
 	total_invalidated_objs = id;
 	id = 0;
 
@@ -3082,7 +3100,6 @@ void amdgpu_debugfs_vm_bo_info(struct amdgpu_vm *vm, struct seq_file *m)
 			continue;
 		total_done += amdgpu_bo_print_info(id++, bo_va->base.bo, m);
 	}
-	spin_unlock(&vm->status_lock);
 	total_done_objs = id;
 
 	seq_printf(m, "\tTotal idle size:        %12lld\tobjs:\t%d\n", total_idle,