Merge branch 'drm-fixes-4.3' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

radeon and amdgpu fixes for 4.3.  It's a bit bigger than usual since
it's 3 weeks worth of fixes since I was on vacation, then at XDC.
- lots of stability fixes
- suspend and resume fixes
- GPU scheduler fixes
- Misc other fixes

* 'drm-fixes-4.3' of git://people.freedesktop.org/~agd5f/linux: (31 commits)
  drm/radeon: add quirk for MSI R7 370
  drm/amdgpu: Sprinkle drm_modeset_lock_all to appease locking checks
  drm/radeon: Sprinkle drm_modeset_lock_all to appease locking checks
  drm/amdgpu: sync ce and me with SWITCH_BUFFER(2)
  drm/amdgpu: integer overflow in amdgpu_mode_dumb_create()
  drm/amdgpu: info leak in amdgpu_gem_metadata_ioctl()
  drm/amdgpu: integer overflow in amdgpu_info_ioctl()
  drm/amdgpu: unwind properly in amdgpu_cs_parser_init()
  drm/amdgpu: Fix max_vblank_count value for current display engines
  drm/amdgpu: use kmemdup rather than duplicating its implementation
  drm/amdgpu: fix UVD suspend and resume for VI APU
  drm/amdgpu: fix the UVD suspend sequence order
  drm/amdgpu: make UVD handle checking more strict
  drm/amdgpu: Disable UVD PG
  drm/amdgpu: more scheduler cleanups v2
  drm/amdgpu: cleanup fence queue init v2
  drm/amdgpu: rename fence->scheduler to sched v2
  drm/amdgpu: cleanup entity init
  drm/amdgpu: refine the scheduler job type conversion
  drm/amdgpu: refine the job naming for amdgpu_job and amdgpu_sched_job
  ...

43 files changed, 534 insertions, 466 deletions

diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
index 668939a14206..6647fb26ef25 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h
@@ -82,6 +82,7 @@ extern int amdgpu_vm_block_size;
 extern int amdgpu_enable_scheduler;
 extern int amdgpu_sched_jobs;
 extern int amdgpu_sched_hw_submission;
+extern int amdgpu_enable_semaphores;
 
 #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS	        3000
 #define AMDGPU_MAX_USEC_TIMEOUT			100000	/* 100 ms */
@@ -432,7 +433,7 @@ int amdgpu_fence_driver_init(struct amdgpu_device *adev);
 void amdgpu_fence_driver_fini(struct amdgpu_device *adev);
 void amdgpu_fence_driver_force_completion(struct amdgpu_device *adev);
 
-void amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring);
+int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring);
 int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
 				   struct amdgpu_irq_src *irq_src,
 				   unsigned irq_type);
@@ -890,7 +891,7 @@ struct amdgpu_ring {
 	struct amdgpu_device		*adev;
 	const struct amdgpu_ring_funcs	*funcs;
 	struct amdgpu_fence_driver	fence_drv;
-	struct amd_gpu_scheduler 	*scheduler;
+	struct amd_gpu_scheduler 	sched;
 
 	spinlock_t              fence_lock;
 	struct mutex		*ring_lock;
@@ -1201,8 +1202,6 @@ struct amdgpu_gfx {
 	struct amdgpu_irq_src		priv_inst_irq;
 	/* gfx status */
 	uint32_t gfx_current_status;
-	/* sync signal for const engine */
-	unsigned ce_sync_offs;
 	/* ce ram size*/
 	unsigned ce_ram_size;
 };
@@ -1274,8 +1273,10 @@ struct amdgpu_job {
 	uint32_t		num_ibs;
 	struct mutex            job_lock;
 	struct amdgpu_user_fence uf;
-	int (*free_job)(struct amdgpu_job *sched_job);
+	int (*free_job)(struct amdgpu_job *job);
 };
+#define to_amdgpu_job(sched_job)		\
+		container_of((sched_job), struct amdgpu_job, base)
 
 static inline u32 amdgpu_get_ib_value(struct amdgpu_cs_parser *p, uint32_t ib_idx, int idx)
 {
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
index 496ed2192eba..84d68d658f8a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c
@@ -183,7 +183,7 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size,
 		return -ENOMEM;
 
 	r = amdgpu_bo_create(rdev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_GTT,
-			AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, &(*mem)->bo);
+			     AMDGPU_GEM_CREATE_CPU_GTT_USWC, NULL, NULL, &(*mem)->bo);
 	if (r) {
 		dev_err(rdev->dev,
 			"failed to allocate BO for amdkfd (%d)\n", r);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
index 98d59ee640ce..cd639c362df3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c
@@ -79,7 +79,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size,
 	int time;
 
 	n = AMDGPU_BENCHMARK_ITERATIONS;
-	r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, sdomain, 0, NULL, &sobj);
+	r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, sdomain, 0, NULL,
+			     NULL, &sobj);
 	if (r) {
 		goto out_cleanup;
 	}
@@ -91,7 +92,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size,
 	if (r) {
 		goto out_cleanup;
 	}
-	r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, ddomain, 0, NULL, &dobj);
+	r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, ddomain, 0, NULL,
+			     NULL, &dobj);
 	if (r) {
 		goto out_cleanup;
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
index 6b1243f9f86d..1c3fc99c5465 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c
@@ -86,7 +86,7 @@ static int amdgpu_cgs_gmap_kmem(void *cgs_device, void *kmem,
 
 	struct sg_table *sg = drm_prime_pages_to_sg(&kmem_page, npages);
 	ret = amdgpu_bo_create(adev, size, PAGE_SIZE, false,
-			       AMDGPU_GEM_DOMAIN_GTT, 0, sg, &bo);
+			       AMDGPU_GEM_DOMAIN_GTT, 0, sg, NULL, &bo);
 	if (ret)
 		return ret;
 	ret = amdgpu_bo_reserve(bo, false);
@@ -197,7 +197,8 @@ static int amdgpu_cgs_alloc_gpu_mem(void *cgs_device,
 
 	ret = amdgpu_bo_create_restricted(adev, size, PAGE_SIZE,
 					  true, domain, flags,
-					  NULL, &placement, &obj);
+					  NULL, &placement, NULL,
+					  &obj);
 	if (ret) {
 		DRM_ERROR("(%d) bo create failed\n", ret);
 		return ret;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
index 3b355aeb62fd..749420f1ea6f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c
@@ -154,42 +154,41 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 {
 	union drm_amdgpu_cs *cs = data;
 	uint64_t *chunk_array_user;
-	uint64_t *chunk_array = NULL;
+	uint64_t *chunk_array;
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	unsigned size, i;
-	int r = 0;
+	int ret;
 
-	if (!cs->in.num_chunks)
-		goto out;
+	if (cs->in.num_chunks == 0)
+		return 0;
+
+	chunk_array = kmalloc_array(cs->in.num_chunks, sizeof(uint64_t), GFP_KERNEL);
+	if (!chunk_array)
+		return -ENOMEM;
 
 	p->ctx = amdgpu_ctx_get(fpriv, cs->in.ctx_id);
 	if (!p->ctx) {
-		r = -EINVAL;
-		goto out;
+		ret = -EINVAL;
+		goto free_chunk;
 	}
+
 	p->bo_list = amdgpu_bo_list_get(fpriv, cs->in.bo_list_handle);
 
 	/* get chunks */
 	INIT_LIST_HEAD(&p->validated);
-	chunk_array = kmalloc_array(cs->in.num_chunks, sizeof(uint64_t), GFP_KERNEL);
-	if (chunk_array == NULL) {
-		r = -ENOMEM;
-		goto out;
-	}
-
 	chunk_array_user = (uint64_t __user *)(cs->in.chunks);
 	if (copy_from_user(chunk_array, chunk_array_user,
 			   sizeof(uint64_t)*cs->in.num_chunks)) {
-		r = -EFAULT;
-		goto out;
+		ret = -EFAULT;
+		goto put_bo_list;
 	}
 
 	p->nchunks = cs->in.num_chunks;
 	p->chunks = kmalloc_array(p->nchunks, sizeof(struct amdgpu_cs_chunk),
 			    GFP_KERNEL);
-	if (p->chunks == NULL) {
-		r = -ENOMEM;
-		goto out;
+	if (!p->chunks) {
+		ret = -ENOMEM;
+		goto put_bo_list;
 	}
 
 	for (i = 0; i < p->nchunks; i++) {
@@ -200,8 +199,9 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 		chunk_ptr = (void __user *)chunk_array[i];
 		if (copy_from_user(&user_chunk, chunk_ptr,
 				       sizeof(struct drm_amdgpu_cs_chunk))) {
-			r = -EFAULT;
-			goto out;
+			ret = -EFAULT;
+			i--;
+			goto free_partial_kdata;
 		}
 		p->chunks[i].chunk_id = user_chunk.chunk_id;
 		p->chunks[i].length_dw = user_chunk.length_dw;
@@ -212,13 +212,14 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 
 		p->chunks[i].kdata = drm_malloc_ab(size, sizeof(uint32_t));
 		if (p->chunks[i].kdata == NULL) {
-			r = -ENOMEM;
-			goto out;
+			ret = -ENOMEM;
+			i--;
+			goto free_partial_kdata;
 		}
 		size *= sizeof(uint32_t);
 		if (copy_from_user(p->chunks[i].kdata, cdata, size)) {
-			r = -EFAULT;
-			goto out;
+			ret = -EFAULT;
+			goto free_partial_kdata;
 		}
 
 		switch (p->chunks[i].chunk_id) {
@@ -238,15 +239,15 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 				gobj = drm_gem_object_lookup(p->adev->ddev,
 							     p->filp, handle);
 				if (gobj == NULL) {
-					r = -EINVAL;
-					goto out;
+					ret = -EINVAL;
+					goto free_partial_kdata;
 				}
 
 				p->uf.bo = gem_to_amdgpu_bo(gobj);
 				p->uf.offset = fence_data->offset;
 			} else {
-				r = -EINVAL;
-				goto out;
+				ret = -EINVAL;
+				goto free_partial_kdata;
 			}
 			break;
 
@@ -254,19 +255,35 @@ int amdgpu_cs_parser_init(struct amdgpu_cs_parser *p, void *data)
 			break;
 
 		default:
-			r = -EINVAL;
-			goto out;
+			ret = -EINVAL;
+			goto free_partial_kdata;
 		}
 	}
 
 
 	p->ibs = kcalloc(p->num_ibs, sizeof(struct amdgpu_ib), GFP_KERNEL);
-	if (!p->ibs)
-		r = -ENOMEM;
+	if (!p->ibs) {
+		ret = -ENOMEM;
+		goto free_all_kdata;
+	}
 
-out:
 	kfree(chunk_array);
-	return r;
+	return 0;
+
+free_all_kdata:
+	i = p->nchunks - 1;
+free_partial_kdata:
+	for (; i >= 0; i--)
+		drm_free_large(p->chunks[i].kdata);
+	kfree(p->chunks);
+put_bo_list:
+	if (p->bo_list)
+		amdgpu_bo_list_put(p->bo_list);
+	amdgpu_ctx_put(p->ctx);
+free_chunk:
+	kfree(chunk_array);
+
+	return ret;
 }
 
 /* Returns how many bytes TTM can move per IB.
@@ -321,25 +338,17 @@ static u64 amdgpu_cs_get_threshold_for_moves(struct amdgpu_device *adev)
 	return max(bytes_moved_threshold, 1024*1024ull);
 }
 
-int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p)
+int amdgpu_cs_list_validate(struct amdgpu_device *adev,
+			    struct amdgpu_vm *vm,
+			    struct list_head *validated)
 {
-	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
-	struct amdgpu_vm *vm = &fpriv->vm;
-	struct amdgpu_device *adev = p->adev;
 	struct amdgpu_bo_list_entry *lobj;
-	struct list_head duplicates;
 	struct amdgpu_bo *bo;
 	u64 bytes_moved = 0, initial_bytes_moved;
 	u64 bytes_moved_threshold = amdgpu_cs_get_threshold_for_moves(adev);
 	int r;
 
-	INIT_LIST_HEAD(&duplicates);
-	r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true, &duplicates);
-	if (unlikely(r != 0)) {
-		return r;
-	}
-
-	list_for_each_entry(lobj, &p->validated, tv.head) {
+	list_for_each_entry(lobj, validated, tv.head) {
 		bo = lobj->robj;
 		if (!bo->pin_count) {
 			u32 domain = lobj->prefered_domains;
@@ -373,7 +382,6 @@ int amdgpu_cs_list_validate(struct amdgpu_cs_parser *p)
 					domain = lobj->allowed_domains;
 					goto retry;
 				}
-				ttm_eu_backoff_reservation(&p->ticket, &p->validated);
 				return r;
 			}
 		}
@@ -386,6 +394,7 @@ static int amdgpu_cs_parser_relocs(struct amdgpu_cs_parser *p)
 {
 	struct amdgpu_fpriv *fpriv = p->filp->driver_priv;
 	struct amdgpu_cs_buckets buckets;
+	struct list_head duplicates;
 	bool need_mmap_lock = false;
 	int i, r;
 
@@ -405,8 +414,22 @@ static int amdgpu_cs_parser_relocs(struct amdgpu_cs_parser *p)
 	if (need_mmap_lock)
 		down_read(&current->mm->mmap_sem);
 
-	r = amdgpu_cs_list_validate(p);
+	INIT_LIST_HEAD(&duplicates);
+	r = ttm_eu_reserve_buffers(&p->ticket, &p->validated, true, &duplicates);
+	if (unlikely(r != 0))
+		goto error_reserve;
+
+	r = amdgpu_cs_list_validate(p->adev, &fpriv->vm, &p->validated);
+	if (r)
+		goto error_validate;
+
+	r = amdgpu_cs_list_validate(p->adev, &fpriv->vm, &duplicates);
+
+error_validate:
+	if (r)
+		ttm_eu_backoff_reservation(&p->ticket, &p->validated);
 
+error_reserve:
 	if (need_mmap_lock)
 		up_read(&current->mm->mmap_sem);
 
@@ -772,15 +795,15 @@ static int amdgpu_cs_dependencies(struct amdgpu_device *adev,
 	return 0;
 }
 
-static int amdgpu_cs_free_job(struct amdgpu_job *sched_job)
+static int amdgpu_cs_free_job(struct amdgpu_job *job)
 {
 	int i;
-	if (sched_job->ibs)
-		for (i = 0; i < sched_job->num_ibs; i++)
-			amdgpu_ib_free(sched_job->adev, &sched_job->ibs[i]);
-	kfree(sched_job->ibs);
-	if (sched_job->uf.bo)
-		drm_gem_object_unreference_unlocked(&sched_job->uf.bo->gem_base);
+	if (job->ibs)
+		for (i = 0; i < job->num_ibs; i++)
+			amdgpu_ib_free(job->adev, &job->ibs[i]);
+	kfree(job->ibs);
+	if (job->uf.bo)
+		drm_gem_object_unreference_unlocked(&job->uf.bo->gem_base);
 	return 0;
 }
 
@@ -804,7 +827,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	r = amdgpu_cs_parser_init(parser, data);
 	if (r) {
 		DRM_ERROR("Failed to initialize parser !\n");
-		amdgpu_cs_parser_fini(parser, r, false);
+		kfree(parser);
 		up_read(&adev->exclusive_lock);
 		r = amdgpu_cs_handle_lockup(adev, r);
 		return r;
@@ -842,7 +865,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		job = kzalloc(sizeof(struct amdgpu_job), GFP_KERNEL);
 		if (!job)
 			return -ENOMEM;
-		job->base.sched = ring->scheduler;
+		job->base.sched = &ring->sched;
 		job->base.s_entity = &parser->ctx->rings[ring->idx].entity;
 		job->adev = parser->adev;
 		job->ibs = parser->ibs;
@@ -857,7 +880,7 @@ int amdgpu_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 
 		job->free_job = amdgpu_cs_free_job;
 		mutex_lock(&job->job_lock);
-		r = amd_sched_entity_push_job((struct amd_sched_job *)job);
+		r = amd_sched_entity_push_job(&job->base);
 		if (r) {
 			mutex_unlock(&job->job_lock);
 			amdgpu_cs_free_job(job);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
index 20cbc4eb5a6f..e0b80ccdfe8a 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
@@ -43,10 +43,10 @@ int amdgpu_ctx_init(struct amdgpu_device *adev, bool kernel,
 		for (i = 0; i < adev->num_rings; i++) {
 			struct amd_sched_rq *rq;
 			if (kernel)
-				rq = &adev->rings[i]->scheduler->kernel_rq;
+				rq = &adev->rings[i]->sched.kernel_rq;
 			else
-				rq = &adev->rings[i]->scheduler->sched_rq;
-			r = amd_sched_entity_init(adev->rings[i]->scheduler,
+				rq = &adev->rings[i]->sched.sched_rq;
+			r = amd_sched_entity_init(&adev->rings[i]->sched,
 						  &ctx->rings[i].entity,
 						  rq, amdgpu_sched_jobs);
 			if (r)
@@ -55,7 +55,7 @@ int amdgpu_ctx_init(struct amdgpu_device *adev, bool kernel,
 
 		if (i < adev->num_rings) {
 			for (j = 0; j < i; j++)
-				amd_sched_entity_fini(adev->rings[j]->scheduler,
+				amd_sched_entity_fini(&adev->rings[j]->sched,
 						      &ctx->rings[j].entity);
 			kfree(ctx);
 			return r;
@@ -75,7 +75,7 @@ void amdgpu_ctx_fini(struct amdgpu_ctx *ctx)
 
 	if (amdgpu_enable_scheduler) {
 		for (i = 0; i < adev->num_rings; i++)
-			amd_sched_entity_fini(adev->rings[i]->scheduler,
+			amd_sched_entity_fini(&adev->rings[i]->sched,
 					      &ctx->rings[i].entity);
 	}
 }
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
index 6ff6ae945794..6068d8207d10 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c
@@ -246,7 +246,7 @@ static int amdgpu_vram_scratch_init(struct amdgpu_device *adev)
 		r = amdgpu_bo_create(adev, AMDGPU_GPU_PAGE_SIZE,
 				     PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,
 				     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-				     NULL, &adev->vram_scratch.robj);
+				     NULL, NULL, &adev->vram_scratch.robj);
 		if (r) {
 			return r;
 		}
@@ -449,7 +449,8 @@ static int amdgpu_wb_init(struct amdgpu_device *adev)
 
 	if (adev->wb.wb_obj == NULL) {
 		r = amdgpu_bo_create(adev, AMDGPU_MAX_WB * 4, PAGE_SIZE, true,
-				     AMDGPU_GEM_DOMAIN_GTT, 0,  NULL, &adev->wb.wb_obj);
+				     AMDGPU_GEM_DOMAIN_GTT, 0,  NULL, NULL,
+				     &adev->wb.wb_obj);
 		if (r) {
 			dev_warn(adev->dev, "(%d) create WB bo failed\n", r);
 			return r;
@@ -1650,9 +1651,11 @@ int amdgpu_suspend_kms(struct drm_device *dev, bool suspend, bool fbcon)
 	drm_kms_helper_poll_disable(dev);
 
 	/* turn off display hw */
+	drm_modeset_lock_all(dev);
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 		drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF);
 	}
+	drm_modeset_unlock_all(dev);
 
 	/* unpin the front buffers */
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
@@ -1747,9 +1750,11 @@ int amdgpu_resume_kms(struct drm_device *dev, bool resume, bool fbcon)
 	if (fbcon) {
 		drm_helper_resume_force_mode(dev);
 		/* turn on display hw */
+		drm_modeset_lock_all(dev);
 		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 			drm_helper_connector_dpms(connector, DRM_MODE_DPMS_ON);
 		}
+		drm_modeset_unlock_all(dev);
 	}
 
 	drm_kms_helper_poll_enable(dev);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
index 0fcc0bd1622c..adb48353f2e1 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c
@@ -79,6 +79,7 @@ int amdgpu_exp_hw_support = 0;
 int amdgpu_enable_scheduler = 0;
 int amdgpu_sched_jobs = 16;
 int amdgpu_sched_hw_submission = 2;
+int amdgpu_enable_semaphores = 1;
 
 MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes");
 module_param_named(vramlimit, amdgpu_vram_limit, int, 0600);
@@ -152,6 +153,9 @@ module_param_named(sched_jobs, amdgpu_sched_jobs, int, 0444);
 MODULE_PARM_DESC(sched_hw_submission, "the max number of HW submissions (default 2)");
 module_param_named(sched_hw_submission, amdgpu_sched_hw_submission, int, 0444);
 
+MODULE_PARM_DESC(enable_semaphores, "Enable semaphores (1 = enable (default), 0 = disable)");
+module_param_named(enable_semaphores, amdgpu_enable_semaphores, int, 0644);
+
 static struct pci_device_id pciidlist[] = {
 #ifdef CONFIG_DRM_AMDGPU_CIK
 	/* Kaveri */
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
index 1be2bd6d07ea..b3fc26c59787 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c
@@ -609,9 +609,9 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring,
  * Init the fence driver for the requested ring (all asics).
  * Helper function for amdgpu_fence_driver_init().
  */
-void amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring)
+int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring)
 {
-	int i;
+	int i, r;
 
 	ring->fence_drv.cpu_addr = NULL;
 	ring->fence_drv.gpu_addr = 0;
@@ -625,15 +625,19 @@ void amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring)
 			amdgpu_fence_check_lockup);
 	ring->fence_drv.ring = ring;
 
+	init_waitqueue_head(&ring->fence_drv.fence_queue);
+
 	if (amdgpu_enable_scheduler) {
-		ring->scheduler = amd_sched_create(&amdgpu_sched_ops,
-						   ring->idx,
-						   amdgpu_sched_hw_submission,
-						   (void *)ring->adev);
-		if (!ring->scheduler)
-			DRM_ERROR("Failed to create scheduler on ring %d.\n",
-				  ring->idx);
+		r = amd_sched_init(&ring->sched, &amdgpu_sched_ops,
+				   amdgpu_sched_hw_submission, ring->name);
+		if (r) {
+			DRM_ERROR("Failed to create scheduler on ring %s.\n",
+				  ring->name);
+			return r;
+		}
 	}
+
+	return 0;
 }
 
 /**
@@ -681,8 +685,7 @@ void amdgpu_fence_driver_fini(struct amdgpu_device *adev)
 		wake_up_all(&ring->fence_drv.fence_queue);
 		amdgpu_irq_put(adev, ring->fence_drv.irq_src,
 			       ring->fence_drv.irq_type);
-		if (ring->scheduler)
-			amd_sched_destroy(ring->scheduler);
+		amd_sched_fini(&ring->sched);
 		ring->fence_drv.initialized = false;
 	}
 	mutex_unlock(&adev->ring_lock);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
index cbd3a486c5c2..7312d729d300 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c
@@ -127,7 +127,7 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev)
 		r = amdgpu_bo_create(adev, adev->gart.table_size,
 				     PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM,
 				     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-				     NULL, &adev->gart.robj);
+				     NULL, NULL, &adev->gart.robj);
 		if (r) {
 			return r;
 		}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
index 5839fab374bf..7297ca3a0ba7 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c
@@ -69,7 +69,8 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size,
 		}
 	}
 retry:
-	r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain, flags, NULL, &robj);
+	r = amdgpu_bo_create(adev, size, alignment, kernel, initial_domain,
+			     flags, NULL, NULL, &robj);
 	if (r) {
 		if (r != -ERESTARTSYS) {
 			if (initial_domain == AMDGPU_GEM_DOMAIN_VRAM) {
@@ -426,6 +427,10 @@ int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 					   &args->data.data_size_bytes,
 					   &args->data.flags);
 	} else if (args->op == AMDGPU_GEM_METADATA_OP_SET_METADATA) {
+		if (args->data.data_size_bytes > sizeof(args->data.data)) {
+			r = -EINVAL;
+			goto unreserve;
+		}
 		r = amdgpu_bo_set_tiling_flags(robj, args->data.tiling_info);
 		if (!r)
 			r = amdgpu_bo_set_metadata(robj, args->data.data,
@@ -433,6 +438,7 @@ int amdgpu_gem_metadata_ioctl(struct drm_device *dev, void *data,
 						   args->data.flags);
 	}
 
+unreserve:
 	amdgpu_bo_unreserve(robj);
 out:
 	drm_gem_object_unreference_unlocked(gobj);
@@ -454,11 +460,12 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 	struct ttm_validate_buffer tv, *entry;
 	struct amdgpu_bo_list_entry *vm_bos;
 	struct ww_acquire_ctx ticket;
-	struct list_head list;
+	struct list_head list, duplicates;
 	unsigned domain;
 	int r;
 
 	INIT_LIST_HEAD(&list);
+	INIT_LIST_HEAD(&duplicates);
 
 	tv.bo = &bo_va->bo->tbo;
 	tv.shared = true;
@@ -468,7 +475,8 @@ static void amdgpu_gem_va_update_vm(struct amdgpu_device *adev,
 	if (!vm_bos)
 		return;
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
+	/* Provide duplicates to avoid -EALREADY */
+	r = ttm_eu_reserve_buffers(&ticket, &list, true, &duplicates);
 	if (r)
 		goto error_free;
 
@@ -651,7 +659,7 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv,
 	int r;
 
 	args->pitch = amdgpu_align_pitch(adev, args->width, args->bpp, 0) * ((args->bpp + 1) / 8);
-	args->size = args->pitch * args->height;
+	args->size = (u64)args->pitch * args->height;
 	args->size = ALIGN(args->size, PAGE_SIZE);
 
 	r = amdgpu_gem_object_create(adev, args->size, 0,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
index 5c8a803acedc..534fc04e80fd 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ih.c
@@ -43,7 +43,7 @@ static int amdgpu_ih_ring_alloc(struct amdgpu_device *adev)
 		r = amdgpu_bo_create(adev, adev->irq.ih.ring_size,
 				     PAGE_SIZE, true,
 				     AMDGPU_GEM_DOMAIN_GTT, 0,
-				     NULL, &adev->irq.ih.ring_obj);
+				     NULL, NULL, &adev->irq.ih.ring_obj);
 		if (r) {
 			DRM_ERROR("amdgpu: failed to create ih ring buffer (%d).\n", r);
 			return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
index 0aba8e9bc8a0..7c42ff670080 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_irq.c
@@ -140,7 +140,7 @@ void amdgpu_irq_preinstall(struct drm_device *dev)
  */
 int amdgpu_irq_postinstall(struct drm_device *dev)
 {
-	dev->max_vblank_count = 0x001fffff;
+	dev->max_vblank_count = 0x00ffffff;
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
index 22367939ebf1..8c735f544b66 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c
@@ -390,7 +390,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
 				    min((size_t)size, sizeof(vram_gtt))) ? -EFAULT : 0;
 	}
 	case AMDGPU_INFO_READ_MMR_REG: {
-		unsigned n, alloc_size = info->read_mmr_reg.count * 4;
+		unsigned n, alloc_size;
 		uint32_t *regs;
 		unsigned se_num = (info->read_mmr_reg.instance >>
 				   AMDGPU_INFO_MMR_SE_INDEX_SHIFT) &
@@ -406,9 +406,10 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file
 		if (sh_num == AMDGPU_INFO_MMR_SH_INDEX_MASK)
 			sh_num = 0xffffffff;
 
-		regs = kmalloc(alloc_size, GFP_KERNEL);
+		regs = kmalloc_array(info->read_mmr_reg.count, sizeof(*regs), GFP_KERNEL);
 		if (!regs)
 			return -ENOMEM;
+		alloc_size = info->read_mmr_reg.count * sizeof(*regs);
 
 		for (i = 0; i < info->read_mmr_reg.count; i++)
 			if (amdgpu_asic_read_register(adev, se_num, sh_num,
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 08b09d55b96f..1a7708f365f3 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -215,6 +215,7 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
 				bool kernel, u32 domain, u64 flags,
 				struct sg_table *sg,
 				struct ttm_placement *placement,
+				struct reservation_object *resv,
 				struct amdgpu_bo **bo_ptr)
 {
 	struct amdgpu_bo *bo;
@@ -261,7 +262,7 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
 	/* Kernel allocation are uninterruptible */
 	r = ttm_bo_init(&adev->mman.bdev, &bo->tbo, size, type,
 			&bo->placement, page_align, !kernel, NULL,
-			acc_size, sg, NULL, &amdgpu_ttm_bo_destroy);
+			acc_size, sg, resv, &amdgpu_ttm_bo_destroy);
 	if (unlikely(r != 0)) {
 		return r;
 	}
@@ -275,7 +276,9 @@ int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
 int amdgpu_bo_create(struct amdgpu_device *adev,
 		     unsigned long size, int byte_align,
 		     bool kernel, u32 domain, u64 flags,
-		     struct sg_table *sg, struct amdgpu_bo **bo_ptr)
+		     struct sg_table *sg,
+		     struct reservation_object *resv,
+		     struct amdgpu_bo **bo_ptr)
 {
 	struct ttm_placement placement = {0};
 	struct ttm_place placements[AMDGPU_GEM_DOMAIN_MAX + 1];
@@ -286,11 +289,9 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 	amdgpu_ttm_placement_init(adev, &placement,
 				  placements, domain, flags);
 
-	return amdgpu_bo_create_restricted(adev, size, byte_align,
-					   kernel, domain, flags,
-					   sg,
-					   &placement,
-					   bo_ptr);
+	return amdgpu_bo_create_restricted(adev, size, byte_align, kernel,
+					   domain, flags, sg, &placement,
+					   resv, bo_ptr);
 }
 
 int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr)
@@ -535,12 +536,10 @@ int amdgpu_bo_set_metadata (struct amdgpu_bo *bo, void *metadata,
 	if (metadata == NULL)
 		return -EINVAL;
 
-	buffer = kzalloc(metadata_size, GFP_KERNEL);
+	buffer = kmemdup(metadata, metadata_size, GFP_KERNEL);
 	if (buffer == NULL)
 		return -ENOMEM;
 
-	memcpy(buffer, metadata, metadata_size);
-
 	kfree(bo->metadata);
 	bo->metadata_flags = flags;
 	bo->metadata = buffer;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
index 6ea18dcec561..3c2ff4567798 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h
@@ -129,12 +129,14 @@ int amdgpu_bo_create(struct amdgpu_device *adev,
 			    unsigned long size, int byte_align,
 			    bool kernel, u32 domain, u64 flags,
 			    struct sg_table *sg,
+			    struct reservation_object *resv,
 			    struct amdgpu_bo **bo_ptr);
 int amdgpu_bo_create_restricted(struct amdgpu_device *adev,
 				unsigned long size, int byte_align,
 				bool kernel, u32 domain, u64 flags,
 				struct sg_table *sg,
 				struct ttm_placement *placement,
+			        struct reservation_object *resv,
 				struct amdgpu_bo **bo_ptr);
 int amdgpu_bo_kmap(struct amdgpu_bo *bo, void **ptr);
 void amdgpu_bo_kunmap(struct amdgpu_bo *bo);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
index d9652fe32d6a..59f735a933a9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c
@@ -61,12 +61,15 @@ struct drm_gem_object *amdgpu_gem_prime_import_sg_table(struct drm_device *dev,
 							struct dma_buf_attachment *attach,
 							struct sg_table *sg)
 {
+	struct reservation_object *resv = attach->dmabuf->resv;
 	struct amdgpu_device *adev = dev->dev_private;
 	struct amdgpu_bo *bo;
 	int ret;
 
+	ww_mutex_lock(&resv->lock, NULL);
 	ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, false,
-			       AMDGPU_GEM_DOMAIN_GTT, 0, sg, &bo);
+			       AMDGPU_GEM_DOMAIN_GTT, 0, sg, resv, &bo);
+	ww_mutex_unlock(&resv->lock);
 	if (ret)
 		return ERR_PTR(ret);
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
index 9bec91484c24..30dce235ddeb 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c
@@ -357,11 +357,11 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 		ring->adev = adev;
 		ring->idx = adev->num_rings++;
 		adev->rings[ring->idx] = ring;
-		amdgpu_fence_driver_init_ring(ring);
+		r = amdgpu_fence_driver_init_ring(ring);
+		if (r)
+			return r;
 	}
 
-	init_waitqueue_head(&ring->fence_drv.fence_queue);
-
 	r = amdgpu_wb_get(adev, &ring->rptr_offs);
 	if (r) {
 		dev_err(adev->dev, "(%d) ring rptr_offs wb alloc failed\n", r);
@@ -407,7 +407,7 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	if (ring->ring_obj == NULL) {
 		r = amdgpu_bo_create(adev, ring->ring_size, PAGE_SIZE, true,
 				     AMDGPU_GEM_DOMAIN_GTT, 0,
-				     NULL, &ring->ring_obj);
+				     NULL, NULL, &ring->ring_obj);
 		if (r) {
 			dev_err(adev->dev, "(%d) ring create failed\n", r);
 			return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
index 74dad270362c..e90712443fe9 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sa.c
@@ -64,8 +64,8 @@ int amdgpu_sa_bo_manager_init(struct amdgpu_device *adev,
 		INIT_LIST_HEAD(&sa_manager->flist[i]);
 	}
 
-	r = amdgpu_bo_create(adev, size, align, true,
-			     domain, 0, NULL, &sa_manager->bo);
+	r = amdgpu_bo_create(adev, size, align, true, domain,
+			     0, NULL, NULL, &sa_manager->bo);
 	if (r) {
 		dev_err(adev->dev, "(%d) failed to allocate bo for manager\n", r);
 		return r;
@@ -145,8 +145,13 @@ static uint32_t amdgpu_sa_get_ring_from_fence(struct fence *f)
 	struct amd_sched_fence *s_fence;
 
 	s_fence = to_amd_sched_fence(f);
-	if (s_fence)
-		return s_fence->scheduler->ring_id;
+	if (s_fence) {
+		struct amdgpu_ring *ring;
+
+		ring = container_of(s_fence->sched, struct amdgpu_ring, sched);
+		return ring->idx;
+	}
+
 	a_fence = to_amdgpu_fence(f);
 	if (a_fence)
 		return a_fence->ring->idx;
@@ -412,6 +417,26 @@ void amdgpu_sa_bo_free(struct amdgpu_device *adev, struct amdgpu_sa_bo **sa_bo,
 }
 
 #if defined(CONFIG_DEBUG_FS)
+
+static void amdgpu_sa_bo_dump_fence(struct fence *fence, struct seq_file *m)
+{
+	struct amdgpu_fence *a_fence = to_amdgpu_fence(fence);
+	struct amd_sched_fence *s_fence = to_amd_sched_fence(fence);
+
+	if (a_fence)
+		seq_printf(m, " protected by 0x%016llx on ring %d",
+			   a_fence->seq, a_fence->ring->idx);
+
+	if (s_fence) {
+		struct amdgpu_ring *ring;
+
+
+		ring = container_of(s_fence->sched, struct amdgpu_ring, sched);
+		seq_printf(m, " protected by 0x%016x on ring %d",
+			   s_fence->base.seqno, ring->idx);
+	}
+}
+
 void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager *sa_manager,
 				  struct seq_file *m)
 {
@@ -428,18 +453,8 @@ void amdgpu_sa_bo_dump_debug_info(struct amdgpu_sa_manager *sa_manager,
 		}
 		seq_printf(m, "[0x%010llx 0x%010llx] size %8lld",
 			   soffset, eoffset, eoffset - soffset);
-		if (i->fence) {
-			struct amdgpu_fence *a_fence = to_amdgpu_fence(i->fence);
-			struct amd_sched_fence *s_fence = to_amd_sched_fence(i->fence);
-			if (a_fence)
-				seq_printf(m, " protected by 0x%016llx on ring %d",
-					   a_fence->seq, a_fence->ring->idx);
-			if (s_fence)
-				seq_printf(m, " protected by 0x%016x on ring %d",
-					   s_fence->base.seqno,
-					   s_fence->scheduler->ring_id);
-
-		}
+		if (i->fence)
+			amdgpu_sa_bo_dump_fence(i->fence, m);
 		seq_printf(m, "\n");
 	}
 	spin_unlock(&sa_manager->wq.lock);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
index de98fbd2971e..2e946b2cad88 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c
@@ -27,63 +27,48 @@
 #include <drm/drmP.h>
 #include "amdgpu.h"
 
-static struct fence *amdgpu_sched_dependency(struct amd_sched_job *job)
+static struct fence *amdgpu_sched_dependency(struct amd_sched_job *sched_job)
 {
-	struct amdgpu_job *sched_job = (struct amdgpu_job *)job;
-	return amdgpu_sync_get_fence(&sched_job->ibs->sync);
+	struct amdgpu_job *job = to_amdgpu_job(sched_job);
+	return amdgpu_sync_get_fence(&job->ibs->sync);
 }
 
-static struct fence *amdgpu_sched_run_job(struct amd_sched_job *job)
+static struct fence *amdgpu_sched_run_job(struct amd_sched_job *sched_job)
 {
-	struct amdgpu_job *sched_job;
-	struct amdgpu_fence *fence;
+	struct amdgpu_fence *fence = NULL;
+	struct amdgpu_job *job;
 	int r;
 
-	if (!job) {
+	if (!sched_job) {
 		DRM_ERROR("job is null\n");
 		return NULL;
 	}
-	sched_job = (struct amdgpu_job *)job;
-	mutex_lock(&sched_job->job_lock);
-	r = amdgpu_ib_schedule(sched_job->adev,
-			       sched_job->num_ibs,
-			       sched_job->ibs,
-			       sched_job->base.owner);
-	if (r)
+	job = to_amdgpu_job(sched_job);
+	mutex_lock(&job->job_lock);
+	r = amdgpu_ib_schedule(job->adev,
+			       job->num_ibs,
+			       job->ibs,
+			       job->base.owner);
+	if (r) {
+		DRM_ERROR("Error scheduling IBs (%d)\n", r);
 		goto err;
-	fence = amdgpu_fence_ref(sched_job->ibs[sched_job->num_ibs - 1].fence);
-
-	if (sched_job->free_job)
-		sched_job->free_job(sched_job);
+	}
 
-	mutex_unlock(&sched_job->job_lock);
-	return &fence->base;
+	fence = amdgpu_fence_ref(job->ibs[job->num_ibs - 1].fence);
 
 err:
-	DRM_ERROR("Run job error\n");
-	mutex_unlock(&sched_job->job_lock);
-	job->sched->ops->process_job(job);
-	return NULL;
-}
+	if (job->free_job)
+		job->free_job(job);
 
-static void amdgpu_sched_process_job(struct amd_sched_job *job)
-{
-	struct amdgpu_job *sched_job;
-
-	if (!job) {
-		DRM_ERROR("job is null\n");
-		return;
-	}
-	sched_job = (struct amdgpu_job *)job;
-	/* after processing job, free memory */
-	fence_put(&sched_job->base.s_fence->base);
-	kfree(sched_job);
+	mutex_unlock(&job->job_lock);
+	fence_put(&job->base.s_fence->base);
+	kfree(job);
+	return fence ? &fence->base : NULL;
 }
 
 struct amd_sched_backend_ops amdgpu_sched_ops = {
 	.dependency = amdgpu_sched_dependency,
 	.run_job = amdgpu_sched_run_job,
-	.process_job = amdgpu_sched_process_job
 };
 
 int amdgpu_sched_ib_submit_kernel_helper(struct amdgpu_device *adev,
@@ -100,7 +85,7 @@ int amdgpu_sched_ib_submit_kernel_helper(struct amdgpu_device *adev,
 			kzalloc(sizeof(struct amdgpu_job), GFP_KERNEL);
 		if (!job)
 			return -ENOMEM;
-		job->base.sched = ring->scheduler;
+		job->base.sched = &ring->sched;
 		job->base.s_entity = &adev->kernel_ctx.rings[ring->idx].entity;
 		job->adev = adev;
 		job->ibs = ibs;
@@ -109,7 +94,7 @@ int amdgpu_sched_ib_submit_kernel_helper(struct amdgpu_device *adev,
 		mutex_init(&job->job_lock);
 		job->free_job = free_job;
 		mutex_lock(&job->job_lock);
-		r = amd_sched_entity_push_job((struct amd_sched_job *)job);
+		r = amd_sched_entity_push_job(&job->base);
 		if (r) {
 			mutex_unlock(&job->job_lock);
 			kfree(job);
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
index 068aeaff7183..4921de15b451 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c
@@ -65,8 +65,14 @@ static bool amdgpu_sync_same_dev(struct amdgpu_device *adev, struct fence *f)
 
 	if (a_fence)
 		return a_fence->ring->adev == adev;
-	if (s_fence)
-		return (struct amdgpu_device *)s_fence->scheduler->priv == adev;
+
+	if (s_fence) {
+		struct amdgpu_ring *ring;
+
+		ring = container_of(s_fence->sched, struct amdgpu_ring, sched);
+		return ring->adev == adev;
+	}
+
 	return false;
 }
 
@@ -251,6 +257,20 @@ int amdgpu_sync_wait(struct amdgpu_sync *sync)
 		fence_put(e->fence);
 		kfree(e);
 	}
+
+	if (amdgpu_enable_semaphores)
+		return 0;
+
+	for (i = 0; i < AMDGPU_MAX_RINGS; ++i) {
+		struct amdgpu_fence *fence = sync->sync_to[i];
+		if (!fence)
+			continue;
+
+		r = fence_wait(&fence->base, false);
+		if (r)
+			return r;
+	}
+
 	return 0;
 }
 
@@ -285,7 +305,8 @@ int amdgpu_sync_rings(struct amdgpu_sync *sync,
 			return -EINVAL;
 		}
 
-		if (amdgpu_enable_scheduler || (count >= AMDGPU_NUM_SYNCS)) {
+		if (amdgpu_enable_scheduler || !amdgpu_enable_semaphores ||
+		    (count >= AMDGPU_NUM_SYNCS)) {
 			/* not enough room, wait manually */
 			r = fence_wait(&fence->base, false);
 			if (r)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
index f80b1a43be8a..4865615e9c06 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c
@@ -59,8 +59,9 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
 		goto out_cleanup;
 	}
 
-	r = amdgpu_bo_create(adev, size, PAGE_SIZE, true, AMDGPU_GEM_DOMAIN_VRAM, 0,
-			     NULL, &vram_obj);
+	r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,
+			     AMDGPU_GEM_DOMAIN_VRAM, 0,
+			     NULL, NULL, &vram_obj);
 	if (r) {
 		DRM_ERROR("Failed to create VRAM object\n");
 		goto out_cleanup;
@@ -80,7 +81,8 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev)
 		struct fence *fence = NULL;
 
 		r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,
-				     AMDGPU_GEM_DOMAIN_GTT, 0, NULL, gtt_obj + i);
+				     AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+				     NULL, gtt_obj + i);
 		if (r) {
 			DRM_ERROR("Failed to create GTT object %d\n", i);
 			goto out_lclean;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
index b5abd5cde413..364cbe975332 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c
@@ -861,7 +861,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev)
 	r = amdgpu_bo_create(adev, 256 * 1024, PAGE_SIZE, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			     NULL, &adev->stollen_vga_memory);
+			     NULL, NULL, &adev->stollen_vga_memory);
 	if (r) {
 		return r;
 	}
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
index 482e66797ae6..5cc95f1a7dab 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c
@@ -247,7 +247,7 @@ int amdgpu_ucode_init_bo(struct amdgpu_device *adev)
 	const struct common_firmware_header *header = NULL;
 
 	err = amdgpu_bo_create(adev, adev->firmware.fw_size, PAGE_SIZE, true,
-			AMDGPU_GEM_DOMAIN_GTT, 0, NULL, bo);
+			AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL, bo);
 	if (err) {
 		dev_err(adev->dev, "(%d) Firmware buffer allocate failed\n", err);
 		err = -ENOMEM;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
index 2cf6c6b06e3b..d0312364d950 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c
@@ -156,7 +156,7 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 	r = amdgpu_bo_create(adev, bo_size, PAGE_SIZE, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			     NULL, &adev->uvd.vcpu_bo);
+			     NULL, NULL, &adev->uvd.vcpu_bo);
 	if (r) {
 		dev_err(adev->dev, "(%d) failed to allocate UVD bo\n", r);
 		return r;
@@ -543,46 +543,60 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx,
 		return -EINVAL;
 	}
 
-	if (msg_type == 1) {
+	switch (msg_type) {
+	case 0:
+		/* it's a create msg, calc image size (width * height) */
+		amdgpu_bo_kunmap(bo);
+
+		/* try to alloc a new handle */
+		for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
+			if (atomic_read(&adev->uvd.handles[i]) == handle) {
+				DRM_ERROR("Handle 0x%x already in use!\n", handle);
+				return -EINVAL;
+			}
+
+			if (!atomic_cmpxchg(&adev->uvd.handles[i], 0, handle)) {
+				adev->uvd.filp[i] = ctx->parser->filp;
+				return 0;
+			}
+		}
+
+		DRM_ERROR("No more free UVD handles!\n");
+		return -EINVAL;
+
+	case 1:
 		/* it's a decode msg, calc buffer sizes */
 		r = amdgpu_uvd_cs_msg_decode(msg, ctx->buf_sizes);
 		amdgpu_bo_kunmap(bo);
 		if (r)
 			return r;
 
-	} else if (msg_type == 2) {
+		/* validate the handle */
+		for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
+			if (atomic_read(&adev->uvd.handles[i]) == handle) {
+				if (adev->uvd.filp[i] != ctx->parser->filp) {
+					DRM_ERROR("UVD handle collision detected!\n");
+					return -EINVAL;
+				}
+				return 0;
+			}
+		}
+
+		DRM_ERROR("Invalid UVD handle 0x%x!\n", handle);
+		return -ENOENT;
+
+	case 2:
 		/* it's a destroy msg, free the handle */
 		for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i)
 			atomic_cmpxchg(&adev->uvd.handles[i], handle, 0);
 		amdgpu_bo_kunmap(bo);
 		return 0;
-	} else {
-		/* it's a create msg */
-		amdgpu_bo_kunmap(bo);
-
-		if (msg_type != 0) {
-			DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type);
-			return -EINVAL;
-		}
-
-		/* it's a create msg, no special handling needed */
-	}
-
-	/* create or decode, validate the handle */
-	for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
-		if (atomic_read(&adev->uvd.handles[i]) == handle)
-			return 0;
-	}
 
-	/* handle not found try to alloc a new one */
-	for (i = 0; i < AMDGPU_MAX_UVD_HANDLES; ++i) {
-		if (!atomic_cmpxchg(&adev->uvd.handles[i], 0, handle)) {
-			adev->uvd.filp[i] = ctx->parser->filp;
-			return 0;
-		}
+	default:
+		DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type);
+		return -EINVAL;
 	}
-
-	DRM_ERROR("No more free UVD handles!\n");
+	BUG();
 	return -EINVAL;
 }
 
@@ -805,10 +819,10 @@ int amdgpu_uvd_ring_parse_cs(struct amdgpu_cs_parser *parser, uint32_t ib_idx)
 }
 
 static int amdgpu_uvd_free_job(
-	struct amdgpu_job *sched_job)
+	struct amdgpu_job *job)
 {
-	amdgpu_ib_free(sched_job->adev, sched_job->ibs);
-	kfree(sched_job->ibs);
+	amdgpu_ib_free(job->adev, job->ibs);
+	kfree(job->ibs);
 	return 0;
 }
 
@@ -905,7 +919,7 @@ int amdgpu_uvd_get_create_msg(struct amdgpu_ring *ring, uint32_t handle,
 	r = amdgpu_bo_create(adev, 1024, PAGE_SIZE, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			     NULL, &bo);
+			     NULL, NULL, &bo);
 	if (r)
 		return r;
 
@@ -954,7 +968,7 @@ int amdgpu_uvd_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle,
 	r = amdgpu_bo_create(adev, 1024, PAGE_SIZE, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			     NULL, &bo);
+			     NULL, NULL, &bo);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
index 3cab96c42aa8..74f2038ac747 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c
@@ -143,7 +143,7 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
 	r = amdgpu_bo_create(adev, size, PAGE_SIZE, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			     NULL, &adev->vce.vcpu_bo);
+			     NULL, NULL, &adev->vce.vcpu_bo);
 	if (r) {
 		dev_err(adev->dev, "(%d) failed to allocate VCE bo\n", r);
 		return r;
@@ -342,10 +342,10 @@ void amdgpu_vce_free_handles(struct amdgpu_device *adev, struct drm_file *filp)
 }
 
 static int amdgpu_vce_free_job(
-	struct amdgpu_job *sched_job)
+	struct amdgpu_job *job)
 {
-	amdgpu_ib_free(sched_job->adev, sched_job->ibs);
-	kfree(sched_job->ibs);
+	amdgpu_ib_free(job->adev, job->ibs);
+	kfree(job->ibs);
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
index f68b7cdc370a..1e14531353e0 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
@@ -316,12 +316,12 @@ static void amdgpu_vm_update_pages(struct amdgpu_device *adev,
 	}
 }
 
-int amdgpu_vm_free_job(struct amdgpu_job *sched_job)
+int amdgpu_vm_free_job(struct amdgpu_job *job)
 {
 	int i;
-	for (i = 0; i < sched_job->num_ibs; i++)
-		amdgpu_ib_free(sched_job->adev, &sched_job->ibs[i]);
-	kfree(sched_job->ibs);
+	for (i = 0; i < job->num_ibs; i++)
+		amdgpu_ib_free(job->adev, &job->ibs[i]);
+	kfree(job->ibs);
 	return 0;
 }
 
@@ -686,31 +686,6 @@ static int amdgpu_vm_update_ptes(struct amdgpu_device *adev,
 }
 
 /**
- * amdgpu_vm_fence_pts - fence page tables after an update
- *
- * @vm: requested vm
- * @start: start of GPU address range
- * @end: end of GPU address range
- * @fence: fence to use
- *
- * Fence the page tables in the range @start - @end (cayman+).
- *
- * Global and local mutex must be locked!
- */
-static void amdgpu_vm_fence_pts(struct amdgpu_vm *vm,
-				uint64_t start, uint64_t end,
-				struct fence *fence)
-{
-	unsigned i;
-
-	start >>= amdgpu_vm_block_size;
-	end >>= amdgpu_vm_block_size;
-
-	for (i = start; i <= end; ++i)
-		amdgpu_bo_fence(vm->page_tables[i].bo, fence, true);
-}
-
-/**
  * amdgpu_vm_bo_update_mapping - update a mapping in the vm page table
  *
  * @adev: amdgpu_device pointer
@@ -813,8 +788,7 @@ static int amdgpu_vm_bo_update_mapping(struct amdgpu_device *adev,
 	if (r)
 		goto error_free;
 
-	amdgpu_vm_fence_pts(vm, mapping->it.start,
-			    mapping->it.last + 1, f);
+	amdgpu_bo_fence(vm->page_directory, f, true);
 	if (fence) {
 		fence_put(*fence);
 		*fence = fence_get(f);
@@ -855,7 +829,7 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev,
 	int r;
 
 	if (mem) {
-		addr = mem->start << PAGE_SHIFT;
+		addr = (u64)mem->start << PAGE_SHIFT;
 		if (mem->mem_type != TTM_PL_TT)
 			addr += adev->vm_manager.vram_base_offset;
 	} else {
@@ -1089,6 +1063,7 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 
 	/* walk over the address space and allocate the page tables */
 	for (pt_idx = saddr; pt_idx <= eaddr; ++pt_idx) {
+		struct reservation_object *resv = vm->page_directory->tbo.resv;
 		struct amdgpu_bo *pt;
 
 		if (vm->page_tables[pt_idx].bo)
@@ -1097,11 +1072,13 @@ int amdgpu_vm_bo_map(struct amdgpu_device *adev,
 		/* drop mutex to allocate and clear page table */
 		mutex_unlock(&vm->mutex);
 
+		ww_mutex_lock(&resv->lock, NULL);
 		r = amdgpu_bo_create(adev, AMDGPU_VM_PTE_COUNT * 8,
 				     AMDGPU_GPU_PAGE_SIZE, true,
 				     AMDGPU_GEM_DOMAIN_VRAM,
 				     AMDGPU_GEM_CREATE_NO_CPU_ACCESS,
-				     NULL, &pt);
+				     NULL, resv, &pt);
+		ww_mutex_unlock(&resv->lock);
 		if (r)
 			goto error_free;
 
@@ -1303,7 +1280,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 	r = amdgpu_bo_create(adev, pd_size, align, true,
 			     AMDGPU_GEM_DOMAIN_VRAM,
 			     AMDGPU_GEM_CREATE_NO_CPU_ACCESS,
-			     NULL, &vm->page_directory);
+			     NULL, NULL, &vm->page_directory);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/cz_smc.c b/drivers/gpu/drm/amd/amdgpu/cz_smc.c
index a72ffc7d6c26..e33180d3314a 100644
--- a/drivers/gpu/drm/amd/amdgpu/cz_smc.c
+++ b/drivers/gpu/drm/amd/amdgpu/cz_smc.c
@@ -814,7 +814,8 @@ int cz_smu_init(struct amdgpu_device *adev)
 	* 3. map kernel virtual address
 	*/
 	ret = amdgpu_bo_create(adev, priv->toc_buffer.data_size, PAGE_SIZE,
-				true, AMDGPU_GEM_DOMAIN_GTT, 0, NULL, toc_buf);
+			       true, AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL,
+			       toc_buf);
 
 	if (ret) {
 		dev_err(adev->dev, "(%d) SMC TOC buffer allocation failed\n", ret);
@@ -822,7 +823,8 @@ int cz_smu_init(struct amdgpu_device *adev)
 	}
 
 	ret = amdgpu_bo_create(adev, priv->smu_buffer.data_size, PAGE_SIZE,
-				true, AMDGPU_GEM_DOMAIN_GTT, 0, NULL, smu_buf);
+			       true, AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL,
+			       smu_buf);
 
 	if (ret) {
 		dev_err(adev->dev, "(%d) SMC Internal buffer allocation failed\n", ret);
diff --git a/drivers/gpu/drm/amd/amdgpu/fiji_smc.c b/drivers/gpu/drm/amd/amdgpu/fiji_smc.c
index 322edea65857..bda1249eb871 100644
--- a/drivers/gpu/drm/amd/amdgpu/fiji_smc.c
+++ b/drivers/gpu/drm/amd/amdgpu/fiji_smc.c
@@ -764,7 +764,7 @@ int fiji_smu_init(struct amdgpu_device *adev)
 	ret = amdgpu_bo_create(adev, image_size, PAGE_SIZE,
 			       true, AMDGPU_GEM_DOMAIN_VRAM,
 			       AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			       NULL, toc_buf);
+			       NULL, NULL, toc_buf);
 	if (ret) {
 		DRM_ERROR("Failed to allocate memory for TOC buffer\n");
 		return -ENOMEM;
@@ -774,7 +774,7 @@ int fiji_smu_init(struct amdgpu_device *adev)
 	ret = amdgpu_bo_create(adev, smu_internal_buffer_size, PAGE_SIZE,
 			       true, AMDGPU_GEM_DOMAIN_VRAM,
 			       AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			       NULL, smu_buf);
+			       NULL, NULL, smu_buf);
 	if (ret) {
 		DRM_ERROR("Failed to allocate memory for SMU internal buffer\n");
 		return -ENOMEM;
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
index 4bd1e5cf65ca..e992bf2ff66c 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c
@@ -3206,7 +3206,7 @@ static int gfx_v7_0_mec_init(struct amdgpu_device *adev)
 		r = amdgpu_bo_create(adev,
 				     adev->gfx.mec.num_mec *adev->gfx.mec.num_pipe * MEC_HPD_SIZE * 2,
 				     PAGE_SIZE, true,
-				     AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+				     AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL,
 				     &adev->gfx.mec.hpd_eop_obj);
 		if (r) {
 			dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
@@ -3373,7 +3373,7 @@ static int gfx_v7_0_cp_compute_resume(struct amdgpu_device *adev)
 			r = amdgpu_bo_create(adev,
 					     sizeof(struct bonaire_mqd),
 					     PAGE_SIZE, true,
-					     AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+					     AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL,
 					     &ring->mqd_obj);
 			if (r) {
 				dev_warn(adev->dev, "(%d) create MQD bo failed\n", r);
@@ -3610,41 +3610,6 @@ static int gfx_v7_0_cp_resume(struct amdgpu_device *adev)
 	return 0;
 }
 
-static void gfx_v7_0_ce_sync_me(struct amdgpu_ring *ring)
-{
-	struct amdgpu_device *adev = ring->adev;
-	u64 gpu_addr = adev->wb.gpu_addr + adev->gfx.ce_sync_offs * 4;
-
-	/* instruct DE to set a magic number */
-	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
-	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
-							 WRITE_DATA_DST_SEL(5)));
-	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, 1);
-
-	/* let CE wait till condition satisfied */
-	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
-	amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(0) | /* wait */
-							 WAIT_REG_MEM_MEM_SPACE(1) | /* memory */
-							 WAIT_REG_MEM_FUNCTION(3) |  /* == */
-							 WAIT_REG_MEM_ENGINE(2)));   /* ce */
-	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, 1);
-	amdgpu_ring_write(ring, 0xffffffff);
-	amdgpu_ring_write(ring, 4); /* poll interval */
-
-	/* instruct CE to reset wb of ce_sync to zero */
-	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
-	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
-							 WRITE_DATA_DST_SEL(5) |
-							 WR_CONFIRM));
-	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, 0);
-}
-
 /*
  * vm
  * VMID 0 is the physical GPU addresses as used by the kernel.
@@ -3663,6 +3628,13 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 					unsigned vm_id, uint64_t pd_addr)
 {
 	int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
+	if (usepfp) {
+		/* synce CE with ME to prevent CE fetch CEIB before context switch done */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+	}
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
@@ -3703,7 +3675,10 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 		amdgpu_ring_write(ring, 0x0);
 
 		/* synce CE with ME to prevent CE fetch CEIB before context switch done */
-		gfx_v7_0_ce_sync_me(ring);
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
 	}
 }
 
@@ -3788,7 +3763,8 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
 			r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true,
 					     AMDGPU_GEM_DOMAIN_VRAM,
 					     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-					     NULL, &adev->gfx.rlc.save_restore_obj);
+					     NULL, NULL,
+					     &adev->gfx.rlc.save_restore_obj);
 			if (r) {
 				dev_warn(adev->dev, "(%d) create RLC sr bo failed\n", r);
 				return r;
@@ -3831,7 +3807,8 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
 			r = amdgpu_bo_create(adev, dws * 4, PAGE_SIZE, true,
 					     AMDGPU_GEM_DOMAIN_VRAM,
 					     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-					     NULL, &adev->gfx.rlc.clear_state_obj);
+					     NULL, NULL,
+					     &adev->gfx.rlc.clear_state_obj);
 			if (r) {
 				dev_warn(adev->dev, "(%d) create RLC c bo failed\n", r);
 				gfx_v7_0_rlc_fini(adev);
@@ -3870,7 +3847,8 @@ static int gfx_v7_0_rlc_init(struct amdgpu_device *adev)
 			r = amdgpu_bo_create(adev, adev->gfx.rlc.cp_table_size, PAGE_SIZE, true,
 					     AMDGPU_GEM_DOMAIN_VRAM,
 					     AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-					     NULL, &adev->gfx.rlc.cp_table_obj);
+					     NULL, NULL,
+					     &adev->gfx.rlc.cp_table_obj);
 			if (r) {
 				dev_warn(adev->dev, "(%d) create RLC cp table bo failed\n", r);
 				gfx_v7_0_rlc_fini(adev);
@@ -4802,12 +4780,6 @@ static int gfx_v7_0_sw_init(void *handle)
 		return r;
 	}
 
-	r = amdgpu_wb_get(adev, &adev->gfx.ce_sync_offs);
-	if (r) {
-		DRM_ERROR("(%d) gfx.ce_sync_offs wb alloc failed\n", r);
-		return r;
-	}
-
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 		ring = &adev->gfx.gfx_ring[i];
 		ring->ring_obj = NULL;
@@ -4851,21 +4823,21 @@ static int gfx_v7_0_sw_init(void *handle)
 	r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size,
 			PAGE_SIZE, true,
 			AMDGPU_GEM_DOMAIN_GDS, 0,
-			NULL, &adev->gds.gds_gfx_bo);
+			NULL, NULL, &adev->gds.gds_gfx_bo);
 	if (r)
 		return r;
 
 	r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size,
 		PAGE_SIZE, true,
 		AMDGPU_GEM_DOMAIN_GWS, 0,
-		NULL, &adev->gds.gws_gfx_bo);
+		NULL, NULL, &adev->gds.gws_gfx_bo);
 	if (r)
 		return r;
 
 	r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size,
 			PAGE_SIZE, true,
 			AMDGPU_GEM_DOMAIN_OA, 0,
-			NULL, &adev->gds.oa_gfx_bo);
+			NULL, NULL, &adev->gds.oa_gfx_bo);
 	if (r)
 		return r;
 
@@ -4886,8 +4858,6 @@ static int gfx_v7_0_sw_fini(void *handle)
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_wb_free(adev, adev->gfx.ce_sync_offs);
-
 	gfx_v7_0_cp_compute_fini(adev);
 	gfx_v7_0_rlc_fini(adev);
 	gfx_v7_0_mec_fini(adev);
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
index 53f07439a512..cb4f68f53f24 100644
--- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c
@@ -868,7 +868,7 @@ static int gfx_v8_0_mec_init(struct amdgpu_device *adev)
 		r = amdgpu_bo_create(adev,
 				     adev->gfx.mec.num_mec *adev->gfx.mec.num_pipe * MEC_HPD_SIZE * 2,
 				     PAGE_SIZE, true,
-				     AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
+				     AMDGPU_GEM_DOMAIN_GTT, 0, NULL, NULL,
 				     &adev->gfx.mec.hpd_eop_obj);
 		if (r) {
 			dev_warn(adev->dev, "(%d) create HDP EOP bo failed\n", r);
@@ -940,12 +940,6 @@ static int gfx_v8_0_sw_init(void *handle)
 		return r;
 	}
 
-	r = amdgpu_wb_get(adev, &adev->gfx.ce_sync_offs);
-	if (r) {
-		DRM_ERROR("(%d) gfx.ce_sync_offs wb alloc failed\n", r);
-		return r;
-	}
-
 	/* set up the gfx ring */
 	for (i = 0; i < adev->gfx.num_gfx_rings; i++) {
 		ring = &adev->gfx.gfx_ring[i];
@@ -995,21 +989,21 @@ static int gfx_v8_0_sw_init(void *handle)
 	/* reserve GDS, GWS and OA resource for gfx */
 	r = amdgpu_bo_create(adev, adev->gds.mem.gfx_partition_size,
 			PAGE_SIZE, true,
-			AMDGPU_GEM_DOMAIN_GDS, 0,
+			AMDGPU_GEM_DOMAIN_GDS, 0, NULL,
 			NULL, &adev->gds.gds_gfx_bo);
 	if (r)
 		return r;
 
 	r = amdgpu_bo_create(adev, adev->gds.gws.gfx_partition_size,
 		PAGE_SIZE, true,
-		AMDGPU_GEM_DOMAIN_GWS, 0,
+		AMDGPU_GEM_DOMAIN_GWS, 0, NULL,
 		NULL, &adev->gds.gws_gfx_bo);
 	if (r)
 		return r;
 
 	r = amdgpu_bo_create(adev, adev->gds.oa.gfx_partition_size,
 			PAGE_SIZE, true,
-			AMDGPU_GEM_DOMAIN_OA, 0,
+			AMDGPU_GEM_DOMAIN_OA, 0, NULL,
 			NULL, &adev->gds.oa_gfx_bo);
 	if (r)
 		return r;
@@ -1033,8 +1027,6 @@ static int gfx_v8_0_sw_fini(void *handle)
 	for (i = 0; i < adev->gfx.num_compute_rings; i++)
 		amdgpu_ring_fini(&adev->gfx.compute_ring[i]);
 
-	amdgpu_wb_free(adev, adev->gfx.ce_sync_offs);
-
 	gfx_v8_0_mec_fini(adev);
 
 	return 0;
@@ -3106,7 +3098,7 @@ static int gfx_v8_0_cp_compute_resume(struct amdgpu_device *adev)
 					     sizeof(struct vi_mqd),
 					     PAGE_SIZE, true,
 					     AMDGPU_GEM_DOMAIN_GTT, 0, NULL,
-					     &ring->mqd_obj);
+					     NULL, &ring->mqd_obj);
 			if (r) {
 				dev_warn(adev->dev, "(%d) create MQD bo failed\n", r);
 				return r;
@@ -3965,6 +3957,7 @@ static void gfx_v8_0_ring_emit_fence_gfx(struct amdgpu_ring *ring, u64 addr,
 			  DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0));
 	amdgpu_ring_write(ring, lower_32_bits(seq));
 	amdgpu_ring_write(ring, upper_32_bits(seq));
+
 }
 
 /**
@@ -4005,49 +3998,34 @@ static bool gfx_v8_0_ring_emit_semaphore(struct amdgpu_ring *ring,
 	return true;
 }
 
-static void gfx_v8_0_ce_sync_me(struct amdgpu_ring *ring)
+static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
+					unsigned vm_id, uint64_t pd_addr)
 {
-	struct amdgpu_device *adev = ring->adev;
-	u64 gpu_addr = adev->wb.gpu_addr + adev->gfx.ce_sync_offs * 4;
-
-	/* instruct DE to set a magic number */
-	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
-	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) |
-							 WRITE_DATA_DST_SEL(5)));
-	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, 1);
+	int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
+	uint32_t seq = ring->fence_drv.sync_seq[ring->idx];
+	uint64_t addr = ring->fence_drv.gpu_addr;
 
-	/* let CE wait till condition satisfied */
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5));
-	amdgpu_ring_write(ring, (WAIT_REG_MEM_OPERATION(0) | /* wait */
-							 WAIT_REG_MEM_MEM_SPACE(1) | /* memory */
-							 WAIT_REG_MEM_FUNCTION(3) |  /* == */
-							 WAIT_REG_MEM_ENGINE(2)));   /* ce */
-	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, 1);
+	amdgpu_ring_write(ring, (WAIT_REG_MEM_MEM_SPACE(1) | /* memory */
+		 WAIT_REG_MEM_FUNCTION(3))); /* equal */
+	amdgpu_ring_write(ring, addr & 0xfffffffc);
+	amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff);
+	amdgpu_ring_write(ring, seq);
 	amdgpu_ring_write(ring, 0xffffffff);
 	amdgpu_ring_write(ring, 4); /* poll interval */
 
-	/* instruct CE to reset wb of ce_sync to zero */
-	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
-	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(2) |
-							 WRITE_DATA_DST_SEL(5) |
-							 WR_CONFIRM));
-	amdgpu_ring_write(ring, gpu_addr & 0xfffffffc);
-	amdgpu_ring_write(ring, upper_32_bits(gpu_addr) & 0xffffffff);
-	amdgpu_ring_write(ring, 0);
-}
-
-static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
-					unsigned vm_id, uint64_t pd_addr)
-{
-	int usepfp = (ring->type == AMDGPU_RING_TYPE_GFX);
+	if (usepfp) {
+		/* synce CE with ME to prevent CE fetch CEIB before context switch done */
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+	}
 
 	amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3));
 	amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) |
-				 WRITE_DATA_DST_SEL(0)));
+				 WRITE_DATA_DST_SEL(0)) |
+				 WR_CONFIRM);
 	if (vm_id < 8) {
 		amdgpu_ring_write(ring,
 				  (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vm_id));
@@ -4083,9 +4061,10 @@ static void gfx_v8_0_ring_emit_vm_flush(struct amdgpu_ring *ring,
 		/* sync PFP to ME, otherwise we might get invalid PFP reads */
 		amdgpu_ring_write(ring, PACKET3(PACKET3_PFP_SYNC_ME, 0));
 		amdgpu_ring_write(ring, 0x0);
-
-		/* synce CE with ME to prevent CE fetch CEIB before context switch done */
-		gfx_v8_0_ce_sync_me(ring);
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
+		amdgpu_ring_write(ring, PACKET3(PACKET3_SWITCH_BUFFER, 0));
+		amdgpu_ring_write(ring, 0);
 	}
 }
 
diff --git a/drivers/gpu/drm/amd/amdgpu/iceland_smc.c b/drivers/gpu/drm/amd/amdgpu/iceland_smc.c
index c900aa942ade..966d4b2ed9da 100644
--- a/drivers/gpu/drm/amd/amdgpu/iceland_smc.c
+++ b/drivers/gpu/drm/amd/amdgpu/iceland_smc.c
@@ -625,7 +625,7 @@ int iceland_smu_init(struct amdgpu_device *adev)
 	ret = amdgpu_bo_create(adev, image_size, PAGE_SIZE,
 			       true, AMDGPU_GEM_DOMAIN_VRAM,
 			       AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			       NULL, toc_buf);
+			       NULL, NULL, toc_buf);
 	if (ret) {
 		DRM_ERROR("Failed to allocate memory for TOC buffer\n");
 		return -ENOMEM;
diff --git a/drivers/gpu/drm/amd/amdgpu/tonga_smc.c b/drivers/gpu/drm/amd/amdgpu/tonga_smc.c
index 1f5ac941a610..5421309c1862 100644
--- a/drivers/gpu/drm/amd/amdgpu/tonga_smc.c
+++ b/drivers/gpu/drm/amd/amdgpu/tonga_smc.c
@@ -763,7 +763,7 @@ int tonga_smu_init(struct amdgpu_device *adev)
 	ret = amdgpu_bo_create(adev, image_size, PAGE_SIZE,
 			       true, AMDGPU_GEM_DOMAIN_VRAM,
 			       AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			       NULL, toc_buf);
+			       NULL, NULL, toc_buf);
 	if (ret) {
 		DRM_ERROR("Failed to allocate memory for TOC buffer\n");
 		return -ENOMEM;
@@ -773,7 +773,7 @@ int tonga_smu_init(struct amdgpu_device *adev)
 	ret = amdgpu_bo_create(adev, smu_internal_buffer_size, PAGE_SIZE,
 			       true, AMDGPU_GEM_DOMAIN_VRAM,
 			       AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED,
-			       NULL, smu_buf);
+			       NULL, NULL, smu_buf);
 	if (ret) {
 		DRM_ERROR("Failed to allocate memory for SMU internal buffer\n");
 		return -ENOMEM;
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
index 5fac5da694f0..ed50dd725788 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c
@@ -224,11 +224,11 @@ static int uvd_v4_2_suspend(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = uvd_v4_2_hw_fini(adev);
+	r = amdgpu_uvd_suspend(adev);
 	if (r)
 		return r;
 
-	r = amdgpu_uvd_suspend(adev);
+	r = uvd_v4_2_hw_fini(adev);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
index 2d5c59c318af..9ad8b9906c0b 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c
@@ -220,11 +220,11 @@ static int uvd_v5_0_suspend(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = uvd_v5_0_hw_fini(adev);
+	r = amdgpu_uvd_suspend(adev);
 	if (r)
 		return r;
 
-	r = amdgpu_uvd_suspend(adev);
+	r = uvd_v5_0_hw_fini(adev);
 	if (r)
 		return r;
 
diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
index d9f553fce531..7e9934fa4193 100644
--- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
+++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c
@@ -214,14 +214,16 @@ static int uvd_v6_0_suspend(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
+	/* Skip this for APU for now */
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_uvd_suspend(adev);
+		if (r)
+			return r;
+	}
 	r = uvd_v6_0_hw_fini(adev);
 	if (r)
 		return r;
 
-	r = amdgpu_uvd_suspend(adev);
-	if (r)
-		return r;
-
 	return r;
 }
 
@@ -230,10 +232,12 @@ static int uvd_v6_0_resume(void *handle)
 	int r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
-	r = amdgpu_uvd_resume(adev);
-	if (r)
-		return r;
-
+	/* Skip this for APU for now */
+	if (!(adev->flags & AMD_IS_APU)) {
+		r = amdgpu_uvd_resume(adev);
+		if (r)
+			return r;
+	}
 	r = uvd_v6_0_hw_init(adev);
 	if (r)
 		return r;
diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c
index 552d9e75ad1b..b55ceb14fdcd 100644
--- a/drivers/gpu/drm/amd/amdgpu/vi.c
+++ b/drivers/gpu/drm/amd/amdgpu/vi.c
@@ -1400,7 +1400,8 @@ static int vi_common_early_init(void *handle)
 	case CHIP_CARRIZO:
 		adev->has_uvd = true;
 		adev->cg_flags = 0;
-		adev->pg_flags = AMDGPU_PG_SUPPORT_UVD | AMDGPU_PG_SUPPORT_VCE;
+		/* Disable UVD pg */
+		adev->pg_flags = /* AMDGPU_PG_SUPPORT_UVD | */AMDGPU_PG_SUPPORT_VCE;
 		adev->external_rev_id = adev->rev_id + 0x1;
 		if (amdgpu_smc_load_fw && smc_enabled)
 			adev->firmware.smu_load = true;
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h b/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h
new file mode 100644
index 000000000000..144f50acc971
--- /dev/null
+++ b/drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h
@@ -0,0 +1,41 @@
+#if !defined(_GPU_SCHED_TRACE_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _GPU_SCHED_TRACE_H_
+
+#include <linux/stringify.h>
+#include <linux/types.h>
+#include <linux/tracepoint.h>
+
+#include <drm/drmP.h>
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM gpu_sched
+#define TRACE_INCLUDE_FILE gpu_sched_trace
+
+TRACE_EVENT(amd_sched_job,
+	    TP_PROTO(struct amd_sched_job *sched_job),
+	    TP_ARGS(sched_job),
+	    TP_STRUCT__entry(
+			     __field(struct amd_sched_entity *, entity)
+			     __field(const char *, name)
+			     __field(u32, job_count)
+			     __field(int, hw_job_count)
+			     ),
+
+	    TP_fast_assign(
+			   __entry->entity = sched_job->s_entity;
+			   __entry->name = sched_job->sched->name;
+			   __entry->job_count = kfifo_len(
+				   &sched_job->s_entity->job_queue) / sizeof(sched_job);
+			   __entry->hw_job_count = atomic_read(
+				   &sched_job->sched->hw_rq_count);
+			   ),
+	    TP_printk("entity=%p, ring=%s, job count:%u, hw job count:%d",
+		      __entry->entity, __entry->name, __entry->job_count,
+		      __entry->hw_job_count)
+);
+#endif
+
+/* This part must be outside protection */
+#undef TRACE_INCLUDE_PATH
+#define TRACE_INCLUDE_PATH .
+#include <trace/define_trace.h>
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
index 9259f1b6664c..3697eeeecf82 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.c
@@ -27,6 +27,9 @@
 #include <drm/drmP.h>
 #include "gpu_scheduler.h"
 
+#define CREATE_TRACE_POINTS
+#include "gpu_sched_trace.h"
+
 static struct amd_sched_job *
 amd_sched_entity_pop_job(struct amd_sched_entity *entity);
 static void amd_sched_wakeup(struct amd_gpu_scheduler *sched);
@@ -65,29 +68,29 @@ static struct amd_sched_job *
 amd_sched_rq_select_job(struct amd_sched_rq *rq)
 {
 	struct amd_sched_entity *entity;
-	struct amd_sched_job *job;
+	struct amd_sched_job *sched_job;
 
 	spin_lock(&rq->lock);
 
 	entity = rq->current_entity;
 	if (entity) {
 		list_for_each_entry_continue(entity, &rq->entities, list) {
-			job = amd_sched_entity_pop_job(entity);
-			if (job) {
+			sched_job = amd_sched_entity_pop_job(entity);
+			if (sched_job) {
 				rq->current_entity = entity;
 				spin_unlock(&rq->lock);
-				return job;
+				return sched_job;
 			}
 		}
 	}
 
 	list_for_each_entry(entity, &rq->entities, list) {
 
-		job = amd_sched_entity_pop_job(entity);
-		if (job) {
+		sched_job = amd_sched_entity_pop_job(entity);
+		if (sched_job) {
 			rq->current_entity = entity;
 			spin_unlock(&rq->lock);
-			return job;
+			return sched_job;
 		}
 
 		if (entity == rq->current_entity)
@@ -115,23 +118,27 @@ int amd_sched_entity_init(struct amd_gpu_scheduler *sched,
 			  struct amd_sched_rq *rq,
 			  uint32_t jobs)
 {
+	int r;
+
 	if (!(sched && entity && rq))
 		return -EINVAL;
 
 	memset(entity, 0, sizeof(struct amd_sched_entity));
-	entity->belongto_rq = rq;
-	entity->scheduler = sched;
-	entity->fence_context = fence_context_alloc(1);
-	if(kfifo_alloc(&entity->job_queue,
-		       jobs * sizeof(void *),
-		       GFP_KERNEL))
-		return -EINVAL;
+	INIT_LIST_HEAD(&entity->list);
+	entity->rq = rq;
+	entity->sched = sched;
 
 	spin_lock_init(&entity->queue_lock);
+	r = kfifo_alloc(&entity->job_queue, jobs * sizeof(void *), GFP_KERNEL);
+	if (r)
+		return r;
+
 	atomic_set(&entity->fence_seq, 0);
+	entity->fence_context = fence_context_alloc(1);
 
 	/* Add the entity to the run queue */
 	amd_sched_rq_add_entity(rq, entity);
+
 	return 0;
 }
 
@@ -146,8 +153,8 @@ int amd_sched_entity_init(struct amd_gpu_scheduler *sched,
 static bool amd_sched_entity_is_initialized(struct amd_gpu_scheduler *sched,
 					    struct amd_sched_entity *entity)
 {
-	return entity->scheduler == sched &&
-		entity->belongto_rq != NULL;
+	return entity->sched == sched &&
+		entity->rq != NULL;
 }
 
 /**
@@ -177,7 +184,7 @@ static bool amd_sched_entity_is_idle(struct amd_sched_entity *entity)
 void amd_sched_entity_fini(struct amd_gpu_scheduler *sched,
 			   struct amd_sched_entity *entity)
 {
-	struct amd_sched_rq *rq = entity->belongto_rq;
+	struct amd_sched_rq *rq = entity->rq;
 
 	if (!amd_sched_entity_is_initialized(sched, entity))
 		return;
@@ -198,22 +205,22 @@ static void amd_sched_entity_wakeup(struct fence *f, struct fence_cb *cb)
 		container_of(cb, struct amd_sched_entity, cb);
 	entity->dependency = NULL;
 	fence_put(f);
-	amd_sched_wakeup(entity->scheduler);
+	amd_sched_wakeup(entity->sched);
 }
 
 static struct amd_sched_job *
 amd_sched_entity_pop_job(struct amd_sched_entity *entity)
 {
-	struct amd_gpu_scheduler *sched = entity->scheduler;
-	struct amd_sched_job *job;
+	struct amd_gpu_scheduler *sched = entity->sched;
+	struct amd_sched_job *sched_job;
 
 	if (ACCESS_ONCE(entity->dependency))
 		return NULL;
 
-	if (!kfifo_out_peek(&entity->job_queue, &job, sizeof(job)))
+	if (!kfifo_out_peek(&entity->job_queue, &sched_job, sizeof(sched_job)))
 		return NULL;
 
-	while ((entity->dependency = sched->ops->dependency(job))) {
+	while ((entity->dependency = sched->ops->dependency(sched_job))) {
 
 		if (fence_add_callback(entity->dependency, &entity->cb,
 				       amd_sched_entity_wakeup))
@@ -222,32 +229,33 @@ amd_sched_entity_pop_job(struct amd_sched_entity *entity)
 			return NULL;
 	}
 
-	return job;
+	return sched_job;
 }
 
 /**
  * Helper to submit a job to the job queue
  *
- * @job		The pointer to job required to submit
+ * @sched_job		The pointer to job required to submit
  *
  * Returns true if we could submit the job.
  */
-static bool amd_sched_entity_in(struct amd_sched_job *job)
+static bool amd_sched_entity_in(struct amd_sched_job *sched_job)
 {
-	struct amd_sched_entity *entity = job->s_entity;
+	struct amd_sched_entity *entity = sched_job->s_entity;
 	bool added, first = false;
 
 	spin_lock(&entity->queue_lock);
-	added = kfifo_in(&entity->job_queue, &job, sizeof(job)) == sizeof(job);
+	added = kfifo_in(&entity->job_queue, &sched_job,
+			sizeof(sched_job)) == sizeof(sched_job);
 
-	if (added && kfifo_len(&entity->job_queue) == sizeof(job))
+	if (added && kfifo_len(&entity->job_queue) == sizeof(sched_job))
 		first = true;
 
 	spin_unlock(&entity->queue_lock);
 
 	/* first job wakes up scheduler */
 	if (first)
-		amd_sched_wakeup(job->sched);
+		amd_sched_wakeup(sched_job->sched);
 
 	return added;
 }
@@ -255,7 +263,7 @@ static bool amd_sched_entity_in(struct amd_sched_job *job)
 /**
  * Submit a job to the job queue
  *
- * @job		The pointer to job required to submit
+ * @sched_job		The pointer to job required to submit
  *
  * Returns 0 for success, negative error code otherwise.
  */
@@ -271,9 +279,9 @@ int amd_sched_entity_push_job(struct amd_sched_job *sched_job)
 	fence_get(&fence->base);
 	sched_job->s_fence = fence;
 
-	wait_event(entity->scheduler->job_scheduled,
+	wait_event(entity->sched->job_scheduled,
 		   amd_sched_entity_in(sched_job));
-
+	trace_amd_sched_job(sched_job);
 	return 0;
 }
 
@@ -301,30 +309,28 @@ static void amd_sched_wakeup(struct amd_gpu_scheduler *sched)
 static struct amd_sched_job *
 amd_sched_select_job(struct amd_gpu_scheduler *sched)
 {
-	struct amd_sched_job *job;
+	struct amd_sched_job *sched_job;
 
 	if (!amd_sched_ready(sched))
 		return NULL;
 
 	/* Kernel run queue has higher priority than normal run queue*/
-	job = amd_sched_rq_select_job(&sched->kernel_rq);
-	if (job == NULL)
-		job = amd_sched_rq_select_job(&sched->sched_rq);
+	sched_job = amd_sched_rq_select_job(&sched->kernel_rq);
+	if (sched_job == NULL)
+		sched_job = amd_sched_rq_select_job(&sched->sched_rq);
 
-	return job;
+	return sched_job;
 }
 
 static void amd_sched_process_job(struct fence *f, struct fence_cb *cb)
 {
-	struct amd_sched_job *sched_job =
-		container_of(cb, struct amd_sched_job, cb);
-	struct amd_gpu_scheduler *sched;
+	struct amd_sched_fence *s_fence =
+		container_of(cb, struct amd_sched_fence, cb);
+	struct amd_gpu_scheduler *sched = s_fence->sched;
 
-	sched = sched_job->sched;
-	amd_sched_fence_signal(sched_job->s_fence);
 	atomic_dec(&sched->hw_rq_count);
-	fence_put(&sched_job->s_fence->base);
-	sched->ops->process_job(sched_job);
+	amd_sched_fence_signal(s_fence);
+	fence_put(&s_fence->base);
 	wake_up_interruptible(&sched->wake_up_worker);
 }
 
@@ -338,87 +344,82 @@ static int amd_sched_main(void *param)
 
 	while (!kthread_should_stop()) {
 		struct amd_sched_entity *entity;
-		struct amd_sched_job *job;
+		struct amd_sched_fence *s_fence;
+		struct amd_sched_job *sched_job;
 		struct fence *fence;
 
 		wait_event_interruptible(sched->wake_up_worker,
 			kthread_should_stop() ||
-			(job = amd_sched_select_job(sched)));
+			(sched_job = amd_sched_select_job(sched)));
 
-		if (!job)
+		if (!sched_job)
 			continue;
 
-		entity = job->s_entity;
+		entity = sched_job->s_entity;
+		s_fence = sched_job->s_fence;
 		atomic_inc(&sched->hw_rq_count);
-		fence = sched->ops->run_job(job);
+		fence = sched->ops->run_job(sched_job);
 		if (fence) {
-			r = fence_add_callback(fence, &job->cb,
+			r = fence_add_callback(fence, &s_fence->cb,
 					       amd_sched_process_job);
 			if (r == -ENOENT)
-				amd_sched_process_job(fence, &job->cb);
+				amd_sched_process_job(fence, &s_fence->cb);
 			else if (r)
 				DRM_ERROR("fence add callback failed (%d)\n", r);
 			fence_put(fence);
+		} else {
+			DRM_ERROR("Failed to run job!\n");
+			amd_sched_process_job(NULL, &s_fence->cb);
 		}
 
-		count = kfifo_out(&entity->job_queue, &job, sizeof(job));
-		WARN_ON(count != sizeof(job));
+		count = kfifo_out(&entity->job_queue, &sched_job,
+				sizeof(sched_job));
+		WARN_ON(count != sizeof(sched_job));
 		wake_up(&sched->job_scheduled);
 	}
 	return 0;
 }
 
 /**
- * Create a gpu scheduler
+ * Init a gpu scheduler instance
  *
+ * @sched		The pointer to the scheduler
  * @ops			The backend operations for this scheduler.
- * @ring		The the ring id for the scheduler.
  * @hw_submissions	Number of hw submissions to do.
+ * @name		Name used for debugging
  *
- * Return the pointer to scheduler for success, otherwise return NULL
+ * Return 0 on success, otherwise error code.
 */
-struct amd_gpu_scheduler *amd_sched_create(struct amd_sched_backend_ops *ops,
-					   unsigned ring, unsigned hw_submission,
-					   void *priv)
+int amd_sched_init(struct amd_gpu_scheduler *sched,
+		   struct amd_sched_backend_ops *ops,
+		   unsigned hw_submission, const char *name)
 {
-	struct amd_gpu_scheduler *sched;
-
-	sched = kzalloc(sizeof(struct amd_gpu_scheduler), GFP_KERNEL);
-	if (!sched)
-		return NULL;
-
 	sched->ops = ops;
-	sched->ring_id = ring;
 	sched->hw_submission_limit = hw_submission;
-	sched->priv = priv;
-	snprintf(sched->name, sizeof(sched->name), "amdgpu[%d]", ring);
+	sched->name = name;
 	amd_sched_rq_init(&sched->sched_rq);
 	amd_sched_rq_init(&sched->kernel_rq);
 
 	init_waitqueue_head(&sched->wake_up_worker);
 	init_waitqueue_head(&sched->job_scheduled);
 	atomic_set(&sched->hw_rq_count, 0);
+
 	/* Each scheduler will run on a seperate kernel thread */
 	sched->thread = kthread_run(amd_sched_main, sched, sched->name);
 	if (IS_ERR(sched->thread)) {
-		DRM_ERROR("Failed to create scheduler for id %d.\n", ring);
-		kfree(sched);
-		return NULL;
+		DRM_ERROR("Failed to create scheduler for %s.\n", name);
+		return PTR_ERR(sched->thread);
 	}
 
-	return sched;
+	return 0;
 }
 
 /**
  * Destroy a gpu scheduler
  *
  * @sched	The pointer to the scheduler
- *
- * return 0 if succeed. -1 if failed.
  */
-int amd_sched_destroy(struct amd_gpu_scheduler *sched)
+void amd_sched_fini(struct amd_gpu_scheduler *sched)
 {
 	kthread_stop(sched->thread);
-	kfree(sched);
-	return  0;
 }
diff --git a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
index 2af0e4d4d817..80b64dc22214 100644
--- a/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
+++ b/drivers/gpu/drm/amd/scheduler/gpu_scheduler.h
@@ -38,13 +38,15 @@ struct amd_sched_rq;
 */
 struct amd_sched_entity {
 	struct list_head		list;
-	struct amd_sched_rq		*belongto_rq;
-	atomic_t			fence_seq;
-	/* the job_queue maintains the jobs submitted by clients */
-	struct kfifo                    job_queue;
+	struct amd_sched_rq		*rq;
+	struct amd_gpu_scheduler	*sched;
+
 	spinlock_t			queue_lock;
-	struct amd_gpu_scheduler	*scheduler;
+	struct kfifo                    job_queue;
+
+	atomic_t			fence_seq;
 	uint64_t                        fence_context;
+
 	struct fence			*dependency;
 	struct fence_cb			cb;
 };
@@ -62,13 +64,13 @@ struct amd_sched_rq {
 
 struct amd_sched_fence {
 	struct fence                    base;
-	struct amd_gpu_scheduler	*scheduler;
+	struct fence_cb                 cb;
+	struct amd_gpu_scheduler	*sched;
 	spinlock_t			lock;
 	void                            *owner;
 };
 
 struct amd_sched_job {
-	struct fence_cb                 cb;
 	struct amd_gpu_scheduler        *sched;
 	struct amd_sched_entity         *s_entity;
 	struct amd_sched_fence          *s_fence;
@@ -91,32 +93,29 @@ static inline struct amd_sched_fence *to_amd_sched_fence(struct fence *f)
  * these functions should be implemented in driver side
 */
 struct amd_sched_backend_ops {
-	struct fence *(*dependency)(struct amd_sched_job *job);
-	struct fence *(*run_job)(struct amd_sched_job *job);
-	void (*process_job)(struct amd_sched_job *job);
+	struct fence *(*dependency)(struct amd_sched_job *sched_job);
+	struct fence *(*run_job)(struct amd_sched_job *sched_job);
 };
 
 /**
  * One scheduler is implemented for each hardware ring
 */
 struct amd_gpu_scheduler {
-	struct task_struct		*thread;
+	struct amd_sched_backend_ops	*ops;
+	uint32_t			hw_submission_limit;
+	const char			*name;
 	struct amd_sched_rq		sched_rq;
 	struct amd_sched_rq		kernel_rq;
-	atomic_t			hw_rq_count;
-	struct amd_sched_backend_ops	*ops;
-	uint32_t			ring_id;
 	wait_queue_head_t		wake_up_worker;
 	wait_queue_head_t		job_scheduled;
-	uint32_t                        hw_submission_limit;
-	char                            name[20];
-	void                            *priv;
+	atomic_t			hw_rq_count;
+	struct task_struct		*thread;
 };
 
-struct amd_gpu_scheduler *
-amd_sched_create(struct amd_sched_backend_ops *ops,
-		 uint32_t ring, uint32_t hw_submission, void *priv);
-int amd_sched_destroy(struct amd_gpu_scheduler *sched);
+int amd_sched_init(struct amd_gpu_scheduler *sched,
+		   struct amd_sched_backend_ops *ops,
+		   uint32_t hw_submission, const char *name);
+void amd_sched_fini(struct amd_gpu_scheduler *sched);
 
 int amd_sched_entity_init(struct amd_gpu_scheduler *sched,
 			  struct amd_sched_entity *entity,
diff --git a/drivers/gpu/drm/amd/scheduler/sched_fence.c b/drivers/gpu/drm/amd/scheduler/sched_fence.c
index e62c37920e11..d802638094f4 100644
--- a/drivers/gpu/drm/amd/scheduler/sched_fence.c
+++ b/drivers/gpu/drm/amd/scheduler/sched_fence.c
@@ -36,7 +36,7 @@ struct amd_sched_fence *amd_sched_fence_create(struct amd_sched_entity *s_entity
 	if (fence == NULL)
 		return NULL;
 	fence->owner = owner;
-	fence->scheduler = s_entity->scheduler;
+	fence->sched = s_entity->sched;
 	spin_lock_init(&fence->lock);
 
 	seq = atomic_inc_return(&s_entity->fence_seq);
@@ -63,7 +63,7 @@ static const char *amd_sched_fence_get_driver_name(struct fence *fence)
 static const char *amd_sched_fence_get_timeline_name(struct fence *f)
 {
 	struct amd_sched_fence *fence = to_amd_sched_fence(f);
-	return (const char *)fence->scheduler->name;
+	return (const char *)fence->sched->name;
 }
 
 static bool amd_sched_fence_enable_signaling(struct fence *f)
diff --git a/drivers/gpu/drm/radeon/radeon_device.c b/drivers/gpu/drm/radeon/radeon_device.c
index d8319dae8358..f3f562f6d848 100644
--- a/drivers/gpu/drm/radeon/radeon_device.c
+++ b/drivers/gpu/drm/radeon/radeon_device.c
@@ -1573,10 +1573,12 @@ int radeon_suspend_kms(struct drm_device *dev, bool suspend, bool fbcon)
 
 	drm_kms_helper_poll_disable(dev);
 
+	drm_modeset_lock_all(dev);
 	/* turn off display hw */
 	list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 		drm_helper_connector_dpms(connector, DRM_MODE_DPMS_OFF);
 	}
+	drm_modeset_unlock_all(dev);
 
 	/* unpin the front buffers and cursors */
 	list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) {
@@ -1734,9 +1736,11 @@ int radeon_resume_kms(struct drm_device *dev, bool resume, bool fbcon)
 	if (fbcon) {
 		drm_helper_resume_force_mode(dev);
 		/* turn on display hw */
+		drm_modeset_lock_all(dev);
 		list_for_each_entry(connector, &dev->mode_config.connector_list, head) {
 			drm_helper_connector_dpms(connector, DRM_MODE_DPMS_ON);
 		}
+		drm_modeset_unlock_all(dev);
 	}
 
 	drm_kms_helper_poll_enable(dev);
diff --git a/drivers/gpu/drm/radeon/si_dpm.c b/drivers/gpu/drm/radeon/si_dpm.c
index 787cd8fd897f..e9115d3f67b0 100644
--- a/drivers/gpu/drm/radeon/si_dpm.c
+++ b/drivers/gpu/drm/radeon/si_dpm.c
@@ -2927,6 +2927,7 @@ static struct si_dpm_quirk si_dpm_quirk_list[] = {
 	{ PCI_VENDOR_ID_ATI, 0x6810, 0x1462, 0x3036, 0, 120000 },
 	{ PCI_VENDOR_ID_ATI, 0x6811, 0x174b, 0xe271, 0, 120000 },
 	{ PCI_VENDOR_ID_ATI, 0x6810, 0x174b, 0xe271, 85000, 90000 },
+	{ PCI_VENDOR_ID_ATI, 0x6811, 0x1762, 0x2015, 0, 120000 },
 	{ 0, 0, 0, 0 },
 };