drm: move amd_gpu_scheduler into common location

This moves and renames the AMDGPU scheduler to a common location in DRM
in order to facilitate re-use by other drivers. This is mostly a straight
forward rename with no code changes.

One notable exception is the function to_drm_sched_fence(), which is no
longer a inline header function to avoid the need to export the
drm_sched_fence_ops_scheduled and drm_sched_fence_ops_finished structures.

Reviewed-by: Chunming Zhou <david1.zhou@amd.com>
Tested-by: Dieter Nützel <Dieter@nuetzel-hh.de>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

3 files changed, 935 insertions, 0 deletions

diff --git a/drivers/gpu/drm/scheduler/Makefile b/drivers/gpu/drm/scheduler/Makefile
new file mode 100644
index 000000000000..ed877912d06d
--- /dev/null
+++ b/drivers/gpu/drm/scheduler/Makefile
@@ -0,0 +1,4 @@
+ccflags-y := -Iinclude/drm
+gpu-sched-y := gpu_scheduler.o sched_fence.o
+
+obj-$(CONFIG_DRM_SCHED) += gpu-sched.o
diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler.c b/drivers/gpu/drm/scheduler/gpu_scheduler.c
new file mode 100644
index 000000000000..2c18996d59c5
--- /dev/null
+++ b/drivers/gpu/drm/scheduler/gpu_scheduler.c
@@ -0,0 +1,744 @@
+/*
+ * Copyright 2015 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/kthread.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <uapi/linux/sched/types.h>
+#include <drm/drmP.h>
+#include <drm/gpu_scheduler.h>
+#include <drm/spsc_queue.h>
+
+#define CREATE_TRACE_POINTS
+#include <drm/gpu_scheduler_trace.h>
+
+#define to_drm_sched_job(sched_job)		\
+		container_of((sched_job), struct drm_sched_job, queue_node)
+
+static bool drm_sched_entity_is_ready(struct drm_sched_entity *entity);
+static void drm_sched_wakeup(struct drm_gpu_scheduler *sched);
+static void drm_sched_process_job(struct dma_fence *f, struct dma_fence_cb *cb);
+
+/* Initialize a given run queue struct */
+static void drm_sched_rq_init(struct drm_sched_rq *rq)
+{
+	spin_lock_init(&rq->lock);
+	INIT_LIST_HEAD(&rq->entities);
+	rq->current_entity = NULL;
+}
+
+static void drm_sched_rq_add_entity(struct drm_sched_rq *rq,
+				    struct drm_sched_entity *entity)
+{
+	if (!list_empty(&entity->list))
+		return;
+	spin_lock(&rq->lock);
+	list_add_tail(&entity->list, &rq->entities);
+	spin_unlock(&rq->lock);
+}
+
+static void drm_sched_rq_remove_entity(struct drm_sched_rq *rq,
+				       struct drm_sched_entity *entity)
+{
+	if (list_empty(&entity->list))
+		return;
+	spin_lock(&rq->lock);
+	list_del_init(&entity->list);
+	if (rq->current_entity == entity)
+		rq->current_entity = NULL;
+	spin_unlock(&rq->lock);
+}
+
+/**
+ * Select an entity which could provide a job to run
+ *
+ * @rq		The run queue to check.
+ *
+ * Try to find a ready entity, returns NULL if none found.
+ */
+static struct drm_sched_entity *
+drm_sched_rq_select_entity(struct drm_sched_rq *rq)
+{
+	struct drm_sched_entity *entity;
+
+	spin_lock(&rq->lock);
+
+	entity = rq->current_entity;
+	if (entity) {
+		list_for_each_entry_continue(entity, &rq->entities, list) {
+			if (drm_sched_entity_is_ready(entity)) {
+				rq->current_entity = entity;
+				spin_unlock(&rq->lock);
+				return entity;
+			}
+		}
+	}
+
+	list_for_each_entry(entity, &rq->entities, list) {
+
+		if (drm_sched_entity_is_ready(entity)) {
+			rq->current_entity = entity;
+			spin_unlock(&rq->lock);
+			return entity;
+		}
+
+		if (entity == rq->current_entity)
+			break;
+	}
+
+	spin_unlock(&rq->lock);
+
+	return NULL;
+}
+
+/**
+ * Init a context entity used by scheduler when submit to HW ring.
+ *
+ * @sched	The pointer to the scheduler
+ * @entity	The pointer to a valid drm_sched_entity
+ * @rq		The run queue this entity belongs
+ * @kernel	If this is an entity for the kernel
+ * @jobs	The max number of jobs in the job queue
+ *
+ * return 0 if succeed. negative error code on failure
+*/
+int drm_sched_entity_init(struct drm_gpu_scheduler *sched,
+			  struct drm_sched_entity *entity,
+			  struct drm_sched_rq *rq,
+			  uint32_t jobs, atomic_t *guilty)
+{
+	if (!(sched && entity && rq))
+		return -EINVAL;
+
+	memset(entity, 0, sizeof(struct drm_sched_entity));
+	INIT_LIST_HEAD(&entity->list);
+	entity->rq = rq;
+	entity->sched = sched;
+	entity->guilty = guilty;
+
+	spin_lock_init(&entity->rq_lock);
+	spin_lock_init(&entity->queue_lock);
+	spsc_queue_init(&entity->job_queue);
+
+	atomic_set(&entity->fence_seq, 0);
+	entity->fence_context = dma_fence_context_alloc(2);
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_sched_entity_init);
+
+/**
+ * Query if entity is initialized
+ *
+ * @sched       Pointer to scheduler instance
+ * @entity	The pointer to a valid scheduler entity
+ *
+ * return true if entity is initialized, false otherwise
+*/
+static bool drm_sched_entity_is_initialized(struct drm_gpu_scheduler *sched,
+					    struct drm_sched_entity *entity)
+{
+	return entity->sched == sched &&
+		entity->rq != NULL;
+}
+
+/**
+ * Check if entity is idle
+ *
+ * @entity	The pointer to a valid scheduler entity
+ *
+ * Return true if entity don't has any unscheduled jobs.
+ */
+static bool drm_sched_entity_is_idle(struct drm_sched_entity *entity)
+{
+	rmb();
+	if (spsc_queue_peek(&entity->job_queue) == NULL)
+		return true;
+
+	return false;
+}
+
+/**
+ * Check if entity is ready
+ *
+ * @entity	The pointer to a valid scheduler entity
+ *
+ * Return true if entity could provide a job.
+ */
+static bool drm_sched_entity_is_ready(struct drm_sched_entity *entity)
+{
+	if (spsc_queue_peek(&entity->job_queue) == NULL)
+		return false;
+
+	if (READ_ONCE(entity->dependency))
+		return false;
+
+	return true;
+}
+
+/**
+ * Destroy a context entity
+ *
+ * @sched       Pointer to scheduler instance
+ * @entity	The pointer to a valid scheduler entity
+ *
+ * Cleanup and free the allocated resources.
+ */
+void drm_sched_entity_fini(struct drm_gpu_scheduler *sched,
+			   struct drm_sched_entity *entity)
+{
+	int r;
+
+	if (!drm_sched_entity_is_initialized(sched, entity))
+		return;
+	/**
+	 * The client will not queue more IBs during this fini, consume existing
+	 * queued IBs or discard them on SIGKILL
+	*/
+	if ((current->flags & PF_SIGNALED) && current->exit_code == SIGKILL)
+		r = -ERESTARTSYS;
+	else
+		r = wait_event_killable(sched->job_scheduled,
+					drm_sched_entity_is_idle(entity));
+	drm_sched_entity_set_rq(entity, NULL);
+	if (r) {
+		struct drm_sched_job *job;
+
+		/* Park the kernel for a moment to make sure it isn't processing
+		 * our enity.
+		 */
+		kthread_park(sched->thread);
+		kthread_unpark(sched->thread);
+		if (entity->dependency) {
+			dma_fence_remove_callback(entity->dependency,
+						  &entity->cb);
+			dma_fence_put(entity->dependency);
+			entity->dependency = NULL;
+		}
+
+		while ((job = to_drm_sched_job(spsc_queue_pop(&entity->job_queue)))) {
+			struct drm_sched_fence *s_fence = job->s_fence;
+			drm_sched_fence_scheduled(s_fence);
+			dma_fence_set_error(&s_fence->finished, -ESRCH);
+			drm_sched_fence_finished(s_fence);
+			WARN_ON(s_fence->parent);
+			dma_fence_put(&s_fence->finished);
+			sched->ops->free_job(job);
+		}
+	}
+}
+EXPORT_SYMBOL(drm_sched_entity_fini);
+
+static void drm_sched_entity_wakeup(struct dma_fence *f, struct dma_fence_cb *cb)
+{
+	struct drm_sched_entity *entity =
+		container_of(cb, struct drm_sched_entity, cb);
+	entity->dependency = NULL;
+	dma_fence_put(f);
+	drm_sched_wakeup(entity->sched);
+}
+
+static void drm_sched_entity_clear_dep(struct dma_fence *f, struct dma_fence_cb *cb)
+{
+	struct drm_sched_entity *entity =
+		container_of(cb, struct drm_sched_entity, cb);
+	entity->dependency = NULL;
+	dma_fence_put(f);
+}
+
+void drm_sched_entity_set_rq(struct drm_sched_entity *entity,
+			     struct drm_sched_rq *rq)
+{
+	if (entity->rq == rq)
+		return;
+
+	spin_lock(&entity->rq_lock);
+
+	if (entity->rq)
+		drm_sched_rq_remove_entity(entity->rq, entity);
+
+	entity->rq = rq;
+	if (rq)
+		drm_sched_rq_add_entity(rq, entity);
+
+	spin_unlock(&entity->rq_lock);
+}
+EXPORT_SYMBOL(drm_sched_entity_set_rq);
+
+bool drm_sched_dependency_optimized(struct dma_fence* fence,
+				    struct drm_sched_entity *entity)
+{
+	struct drm_gpu_scheduler *sched = entity->sched;
+	struct drm_sched_fence *s_fence;
+
+	if (!fence || dma_fence_is_signaled(fence))
+		return false;
+	if (fence->context == entity->fence_context)
+		return true;
+	s_fence = to_drm_sched_fence(fence);
+	if (s_fence && s_fence->sched == sched)
+		return true;
+
+	return false;
+}
+EXPORT_SYMBOL(drm_sched_dependency_optimized);
+
+static bool drm_sched_entity_add_dependency_cb(struct drm_sched_entity *entity)
+{
+	struct drm_gpu_scheduler *sched = entity->sched;
+	struct dma_fence * fence = entity->dependency;
+	struct drm_sched_fence *s_fence;
+
+	if (fence->context == entity->fence_context) {
+		/* We can ignore fences from ourself */
+		dma_fence_put(entity->dependency);
+		return false;
+	}
+
+	s_fence = to_drm_sched_fence(fence);
+	if (s_fence && s_fence->sched == sched) {
+
+		/*
+		 * Fence is from the same scheduler, only need to wait for
+		 * it to be scheduled
+		 */
+		fence = dma_fence_get(&s_fence->scheduled);
+		dma_fence_put(entity->dependency);
+		entity->dependency = fence;
+		if (!dma_fence_add_callback(fence, &entity->cb,
+					    drm_sched_entity_clear_dep))
+			return true;
+
+		/* Ignore it when it is already scheduled */
+		dma_fence_put(fence);
+		return false;
+	}
+
+	if (!dma_fence_add_callback(entity->dependency, &entity->cb,
+				    drm_sched_entity_wakeup))
+		return true;
+
+	dma_fence_put(entity->dependency);
+	return false;
+}
+
+static struct drm_sched_job *
+drm_sched_entity_pop_job(struct drm_sched_entity *entity)
+{
+	struct drm_gpu_scheduler *sched = entity->sched;
+	struct drm_sched_job *sched_job = to_drm_sched_job(
+						spsc_queue_peek(&entity->job_queue));
+
+	if (!sched_job)
+		return NULL;
+
+	while ((entity->dependency = sched->ops->dependency(sched_job, entity)))
+		if (drm_sched_entity_add_dependency_cb(entity))
+			return NULL;
+
+	/* skip jobs from entity that marked guilty */
+	if (entity->guilty && atomic_read(entity->guilty))
+		dma_fence_set_error(&sched_job->s_fence->finished, -ECANCELED);
+
+	spsc_queue_pop(&entity->job_queue);
+	return sched_job;
+}
+
+/**
+ * Submit a job to the job queue
+ *
+ * @sched_job		The pointer to job required to submit
+ *
+ * Returns 0 for success, negative error code otherwise.
+ */
+void drm_sched_entity_push_job(struct drm_sched_job *sched_job,
+			       struct drm_sched_entity *entity)
+{
+	struct drm_gpu_scheduler *sched = sched_job->sched;
+	bool first = false;
+
+	trace_drm_sched_job(sched_job, entity);
+
+	spin_lock(&entity->queue_lock);
+	first = spsc_queue_push(&entity->job_queue, &sched_job->queue_node);
+
+	spin_unlock(&entity->queue_lock);
+
+	/* first job wakes up scheduler */
+	if (first) {
+		/* Add the entity to the run queue */
+		spin_lock(&entity->rq_lock);
+		drm_sched_rq_add_entity(entity->rq, entity);
+		spin_unlock(&entity->rq_lock);
+		drm_sched_wakeup(sched);
+	}
+}
+EXPORT_SYMBOL(drm_sched_entity_push_job);
+
+/* job_finish is called after hw fence signaled
+ */
+static void drm_sched_job_finish(struct work_struct *work)
+{
+	struct drm_sched_job *s_job = container_of(work, struct drm_sched_job,
+						   finish_work);
+	struct drm_gpu_scheduler *sched = s_job->sched;
+
+	/* remove job from ring_mirror_list */
+	spin_lock(&sched->job_list_lock);
+	list_del_init(&s_job->node);
+	if (sched->timeout != MAX_SCHEDULE_TIMEOUT) {
+		struct drm_sched_job *next;
+
+		spin_unlock(&sched->job_list_lock);
+		cancel_delayed_work_sync(&s_job->work_tdr);
+		spin_lock(&sched->job_list_lock);
+
+		/* queue TDR for next job */
+		next = list_first_entry_or_null(&sched->ring_mirror_list,
+						struct drm_sched_job, node);
+
+		if (next)
+			schedule_delayed_work(&next->work_tdr, sched->timeout);
+	}
+	spin_unlock(&sched->job_list_lock);
+	dma_fence_put(&s_job->s_fence->finished);
+	sched->ops->free_job(s_job);
+}
+
+static void drm_sched_job_finish_cb(struct dma_fence *f,
+				    struct dma_fence_cb *cb)
+{
+	struct drm_sched_job *job = container_of(cb, struct drm_sched_job,
+						 finish_cb);
+	schedule_work(&job->finish_work);
+}
+
+static void drm_sched_job_begin(struct drm_sched_job *s_job)
+{
+	struct drm_gpu_scheduler *sched = s_job->sched;
+
+	dma_fence_add_callback(&s_job->s_fence->finished, &s_job->finish_cb,
+			       drm_sched_job_finish_cb);
+
+	spin_lock(&sched->job_list_lock);
+	list_add_tail(&s_job->node, &sched->ring_mirror_list);
+	if (sched->timeout != MAX_SCHEDULE_TIMEOUT &&
+	    list_first_entry_or_null(&sched->ring_mirror_list,
+				     struct drm_sched_job, node) == s_job)
+		schedule_delayed_work(&s_job->work_tdr, sched->timeout);
+	spin_unlock(&sched->job_list_lock);
+}
+
+static void drm_sched_job_timedout(struct work_struct *work)
+{
+	struct drm_sched_job *job = container_of(work, struct drm_sched_job,
+						 work_tdr.work);
+
+	job->sched->ops->timedout_job(job);
+}
+
+void drm_sched_hw_job_reset(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad)
+{
+	struct drm_sched_job *s_job;
+	struct drm_sched_entity *entity, *tmp;
+	int i;;
+
+	spin_lock(&sched->job_list_lock);
+	list_for_each_entry_reverse(s_job, &sched->ring_mirror_list, node) {
+		if (s_job->s_fence->parent &&
+		    dma_fence_remove_callback(s_job->s_fence->parent,
+					      &s_job->s_fence->cb)) {
+			dma_fence_put(s_job->s_fence->parent);
+			s_job->s_fence->parent = NULL;
+			atomic_dec(&sched->hw_rq_count);
+		}
+	}
+	spin_unlock(&sched->job_list_lock);
+
+	if (bad && bad->s_priority != DRM_SCHED_PRIORITY_KERNEL) {
+		atomic_inc(&bad->karma);
+		/* don't increase @bad's karma if it's from KERNEL RQ,
+		 * becuase sometimes GPU hang would cause kernel jobs (like VM updating jobs)
+		 * corrupt but keep in mind that kernel jobs always considered good.
+		 */
+		for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_KERNEL; i++ ) {
+			struct drm_sched_rq *rq = &sched->sched_rq[i];
+
+			spin_lock(&rq->lock);
+			list_for_each_entry_safe(entity, tmp, &rq->entities, list) {
+				if (bad->s_fence->scheduled.context == entity->fence_context) {
+				    if (atomic_read(&bad->karma) > bad->sched->hang_limit)
+						if (entity->guilty)
+							atomic_set(entity->guilty, 1);
+					break;
+				}
+			}
+			spin_unlock(&rq->lock);
+			if (&entity->list != &rq->entities)
+				break;
+		}
+	}
+}
+EXPORT_SYMBOL(drm_sched_hw_job_reset);
+
+void drm_sched_job_recovery(struct drm_gpu_scheduler *sched)
+{
+	struct drm_sched_job *s_job, *tmp;
+	bool found_guilty = false;
+	int r;
+
+	spin_lock(&sched->job_list_lock);
+	s_job = list_first_entry_or_null(&sched->ring_mirror_list,
+					 struct drm_sched_job, node);
+	if (s_job && sched->timeout != MAX_SCHEDULE_TIMEOUT)
+		schedule_delayed_work(&s_job->work_tdr, sched->timeout);
+
+	list_for_each_entry_safe(s_job, tmp, &sched->ring_mirror_list, node) {
+		struct drm_sched_fence *s_fence = s_job->s_fence;
+		struct dma_fence *fence;
+		uint64_t guilty_context;
+
+		if (!found_guilty && atomic_read(&s_job->karma) > sched->hang_limit) {
+			found_guilty = true;
+			guilty_context = s_job->s_fence->scheduled.context;
+		}
+
+		if (found_guilty && s_job->s_fence->scheduled.context == guilty_context)
+			dma_fence_set_error(&s_fence->finished, -ECANCELED);
+
+		spin_unlock(&sched->job_list_lock);
+		fence = sched->ops->run_job(s_job);
+		atomic_inc(&sched->hw_rq_count);
+		if (fence) {
+			s_fence->parent = dma_fence_get(fence);
+			r = dma_fence_add_callback(fence, &s_fence->cb,
+						   drm_sched_process_job);
+			if (r == -ENOENT)
+				drm_sched_process_job(fence, &s_fence->cb);
+			else if (r)
+				DRM_ERROR("fence add callback failed (%d)\n",
+					  r);
+			dma_fence_put(fence);
+		} else {
+			drm_sched_process_job(NULL, &s_fence->cb);
+		}
+		spin_lock(&sched->job_list_lock);
+	}
+	spin_unlock(&sched->job_list_lock);
+}
+EXPORT_SYMBOL(drm_sched_job_recovery);
+
+/* init a sched_job with basic field */
+int drm_sched_job_init(struct drm_sched_job *job,
+		       struct drm_gpu_scheduler *sched,
+		       struct drm_sched_entity *entity,
+		       void *owner)
+{
+	job->sched = sched;
+	job->s_priority = entity->rq - sched->sched_rq;
+	job->s_fence = drm_sched_fence_create(entity, owner);
+	if (!job->s_fence)
+		return -ENOMEM;
+	job->id = atomic64_inc_return(&sched->job_id_count);
+
+	INIT_WORK(&job->finish_work, drm_sched_job_finish);
+	INIT_LIST_HEAD(&job->node);
+	INIT_DELAYED_WORK(&job->work_tdr, drm_sched_job_timedout);
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_sched_job_init);
+
+/**
+ * Return ture if we can push more jobs to the hw.
+ */
+static bool drm_sched_ready(struct drm_gpu_scheduler *sched)
+{
+	return atomic_read(&sched->hw_rq_count) <
+		sched->hw_submission_limit;
+}
+
+/**
+ * Wake up the scheduler when it is ready
+ */
+static void drm_sched_wakeup(struct drm_gpu_scheduler *sched)
+{
+	if (drm_sched_ready(sched))
+		wake_up_interruptible(&sched->wake_up_worker);
+}
+
+/**
+ * Select next entity to process
+*/
+static struct drm_sched_entity *
+drm_sched_select_entity(struct drm_gpu_scheduler *sched)
+{
+	struct drm_sched_entity *entity;
+	int i;
+
+	if (!drm_sched_ready(sched))
+		return NULL;
+
+	/* Kernel run queue has higher priority than normal run queue*/
+	for (i = DRM_SCHED_PRIORITY_MAX - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) {
+		entity = drm_sched_rq_select_entity(&sched->sched_rq[i]);
+		if (entity)
+			break;
+	}
+
+	return entity;
+}
+
+static void drm_sched_process_job(struct dma_fence *f, struct dma_fence_cb *cb)
+{
+	struct drm_sched_fence *s_fence =
+		container_of(cb, struct drm_sched_fence, cb);
+	struct drm_gpu_scheduler *sched = s_fence->sched;
+
+	dma_fence_get(&s_fence->finished);
+	atomic_dec(&sched->hw_rq_count);
+	drm_sched_fence_finished(s_fence);
+
+	trace_drm_sched_process_job(s_fence);
+	dma_fence_put(&s_fence->finished);
+	wake_up_interruptible(&sched->wake_up_worker);
+}
+
+static bool drm_sched_blocked(struct drm_gpu_scheduler *sched)
+{
+	if (kthread_should_park()) {
+		kthread_parkme();
+		return true;
+	}
+
+	return false;
+}
+
+static int drm_sched_main(void *param)
+{
+	struct sched_param sparam = {.sched_priority = 1};
+	struct drm_gpu_scheduler *sched = (struct drm_gpu_scheduler *)param;
+	int r;
+
+	sched_setscheduler(current, SCHED_FIFO, &sparam);
+
+	while (!kthread_should_stop()) {
+		struct drm_sched_entity *entity = NULL;
+		struct drm_sched_fence *s_fence;
+		struct drm_sched_job *sched_job;
+		struct dma_fence *fence;
+
+		wait_event_interruptible(sched->wake_up_worker,
+					 (!drm_sched_blocked(sched) &&
+					  (entity = drm_sched_select_entity(sched))) ||
+					 kthread_should_stop());
+
+		if (!entity)
+			continue;
+
+		sched_job = drm_sched_entity_pop_job(entity);
+		if (!sched_job)
+			continue;
+
+		s_fence = sched_job->s_fence;
+
+		atomic_inc(&sched->hw_rq_count);
+		drm_sched_job_begin(sched_job);
+
+		fence = sched->ops->run_job(sched_job);
+		drm_sched_fence_scheduled(s_fence);
+
+		if (fence) {
+			s_fence->parent = dma_fence_get(fence);
+			r = dma_fence_add_callback(fence, &s_fence->cb,
+						   drm_sched_process_job);
+			if (r == -ENOENT)
+				drm_sched_process_job(fence, &s_fence->cb);
+			else if (r)
+				DRM_ERROR("fence add callback failed (%d)\n",
+					  r);
+			dma_fence_put(fence);
+		} else {
+			drm_sched_process_job(NULL, &s_fence->cb);
+		}
+
+		wake_up(&sched->job_scheduled);
+	}
+	return 0;
+}
+
+/**
+ * Init a gpu scheduler instance
+ *
+ * @sched		The pointer to the scheduler
+ * @ops			The backend operations for this scheduler.
+ * @hw_submissions	Number of hw submissions to do.
+ * @name		Name used for debugging
+ *
+ * Return 0 on success, otherwise error code.
+*/
+int drm_sched_init(struct drm_gpu_scheduler *sched,
+		   const struct drm_sched_backend_ops *ops,
+		   unsigned hw_submission,
+		   unsigned hang_limit,
+		   long timeout,
+		   const char *name)
+{
+	int i;
+	sched->ops = ops;
+	sched->hw_submission_limit = hw_submission;
+	sched->name = name;
+	sched->timeout = timeout;
+	sched->hang_limit = hang_limit;
+	for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_MAX; i++)
+		drm_sched_rq_init(&sched->sched_rq[i]);
+
+	init_waitqueue_head(&sched->wake_up_worker);
+	init_waitqueue_head(&sched->job_scheduled);
+	INIT_LIST_HEAD(&sched->ring_mirror_list);
+	spin_lock_init(&sched->job_list_lock);
+	atomic_set(&sched->hw_rq_count, 0);
+	atomic64_set(&sched->job_id_count, 0);
+
+	/* Each scheduler will run on a seperate kernel thread */
+	sched->thread = kthread_run(drm_sched_main, sched, sched->name);
+	if (IS_ERR(sched->thread)) {
+		DRM_ERROR("Failed to create scheduler for %s.\n", name);
+		return PTR_ERR(sched->thread);
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL(drm_sched_init);
+
+/**
+ * Destroy a gpu scheduler
+ *
+ * @sched	The pointer to the scheduler
+ */
+void drm_sched_fini(struct drm_gpu_scheduler *sched)
+{
+	if (sched->thread)
+		kthread_stop(sched->thread);
+}
+EXPORT_SYMBOL(drm_sched_fini);
diff --git a/drivers/gpu/drm/scheduler/sched_fence.c b/drivers/gpu/drm/scheduler/sched_fence.c
new file mode 100644
index 000000000000..f6f2955890c4
--- /dev/null
+++ b/drivers/gpu/drm/scheduler/sched_fence.c
@@ -0,0 +1,187 @@
+/*
+ * Copyright 2015 Advanced Micro Devices, Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+
+#include <linux/kthread.h>
+#include <linux/wait.h>
+#include <linux/sched.h>
+#include <drm/drmP.h>
+#include <drm/gpu_scheduler.h>
+
+static struct kmem_cache *sched_fence_slab;
+
+int drm_sched_fence_slab_init(void)
+{
+	sched_fence_slab = kmem_cache_create(
+		"drm_sched_fence", sizeof(struct drm_sched_fence), 0,
+		SLAB_HWCACHE_ALIGN, NULL);
+	if (!sched_fence_slab)
+		return -ENOMEM;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(drm_sched_fence_slab_init);
+
+void drm_sched_fence_slab_fini(void)
+{
+	rcu_barrier();
+	kmem_cache_destroy(sched_fence_slab);
+}
+EXPORT_SYMBOL_GPL(drm_sched_fence_slab_fini);
+
+void drm_sched_fence_scheduled(struct drm_sched_fence *fence)
+{
+	int ret = dma_fence_signal(&fence->scheduled);
+
+	if (!ret)
+		DMA_FENCE_TRACE(&fence->scheduled,
+				"signaled from irq context\n");
+	else
+		DMA_FENCE_TRACE(&fence->scheduled,
+				"was already signaled\n");
+}
+
+void drm_sched_fence_finished(struct drm_sched_fence *fence)
+{
+	int ret = dma_fence_signal(&fence->finished);
+
+	if (!ret)
+		DMA_FENCE_TRACE(&fence->finished,
+				"signaled from irq context\n");
+	else
+		DMA_FENCE_TRACE(&fence->finished,
+				"was already signaled\n");
+}
+
+static const char *drm_sched_fence_get_driver_name(struct dma_fence *fence)
+{
+	return "drm_sched";
+}
+
+static const char *drm_sched_fence_get_timeline_name(struct dma_fence *f)
+{
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
+	return (const char *)fence->sched->name;
+}
+
+static bool drm_sched_fence_enable_signaling(struct dma_fence *f)
+{
+	return true;
+}
+
+/**
+ * amd_sched_fence_free - free up the fence memory
+ *
+ * @rcu: RCU callback head
+ *
+ * Free up the fence memory after the RCU grace period.
+ */
+static void drm_sched_fence_free(struct rcu_head *rcu)
+{
+	struct dma_fence *f = container_of(rcu, struct dma_fence, rcu);
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
+
+	dma_fence_put(fence->parent);
+	kmem_cache_free(sched_fence_slab, fence);
+}
+
+/**
+ * amd_sched_fence_release_scheduled - callback that fence can be freed
+ *
+ * @fence: fence
+ *
+ * This function is called when the reference count becomes zero.
+ * It just RCU schedules freeing up the fence.
+ */
+static void drm_sched_fence_release_scheduled(struct dma_fence *f)
+{
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
+
+	call_rcu(&fence->finished.rcu, drm_sched_fence_free);
+}
+
+/**
+ * amd_sched_fence_release_finished - drop extra reference
+ *
+ * @f: fence
+ *
+ * Drop the extra reference from the scheduled fence to the base fence.
+ */
+static void drm_sched_fence_release_finished(struct dma_fence *f)
+{
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
+
+	dma_fence_put(&fence->scheduled);
+}
+
+const struct dma_fence_ops drm_sched_fence_ops_scheduled = {
+	.get_driver_name = drm_sched_fence_get_driver_name,
+	.get_timeline_name = drm_sched_fence_get_timeline_name,
+	.enable_signaling = drm_sched_fence_enable_signaling,
+	.signaled = NULL,
+	.wait = dma_fence_default_wait,
+	.release = drm_sched_fence_release_scheduled,
+};
+
+const struct dma_fence_ops drm_sched_fence_ops_finished = {
+	.get_driver_name = drm_sched_fence_get_driver_name,
+	.get_timeline_name = drm_sched_fence_get_timeline_name,
+	.enable_signaling = drm_sched_fence_enable_signaling,
+	.signaled = NULL,
+	.wait = dma_fence_default_wait,
+	.release = drm_sched_fence_release_finished,
+};
+
+struct drm_sched_fence *to_drm_sched_fence(struct dma_fence *f)
+{
+	if (f->ops == &drm_sched_fence_ops_scheduled)
+		return container_of(f, struct drm_sched_fence, scheduled);
+
+	if (f->ops == &drm_sched_fence_ops_finished)
+		return container_of(f, struct drm_sched_fence, finished);
+
+	return NULL;
+}
+EXPORT_SYMBOL(to_drm_sched_fence);
+
+struct drm_sched_fence *drm_sched_fence_create(struct drm_sched_entity *entity,
+					       void *owner)
+{
+	struct drm_sched_fence *fence = NULL;
+	unsigned seq;
+
+	fence = kmem_cache_zalloc(sched_fence_slab, GFP_KERNEL);
+	if (fence == NULL)
+		return NULL;
+
+	fence->owner = owner;
+	fence->sched = entity->sched;
+	spin_lock_init(&fence->lock);
+
+	seq = atomic_inc_return(&entity->fence_seq);
+	dma_fence_init(&fence->scheduled, &drm_sched_fence_ops_scheduled,
+		       &fence->lock, entity->fence_context, seq);
+	dma_fence_init(&fence->finished, &drm_sched_fence_ops_finished,
+		       &fence->lock, entity->fence_context + 1, seq);
+
+	return fence;
+}