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>

drivers/gpu/drm/Kconfig

@@ -149,6 +149,10 @@ config DRM_VM
 	bool
 	depends on DRM && MMU
 
+config DRM_SCHED
+	tristate
+	depends on DRM
+
 source "drivers/gpu/drm/i2c/Kconfig"
 
 source "drivers/gpu/drm/arm/Kconfig"
@@ -178,6 +182,7 @@ config DRM_AMDGPU
 	depends on DRM && PCI && MMU
 	select FW_LOADER
         select DRM_KMS_HELPER
+	select DRM_SCHED
         select DRM_TTM
 	select POWER_SUPPLY
 	select HWMON

drivers/gpu/drm/Makefile

@@ -101,3 +101,4 @@ obj-$(CONFIG_DRM_MXSFB)	+= mxsfb/
 obj-$(CONFIG_DRM_TINYDRM) += tinydrm/
 obj-$(CONFIG_DRM_PL111) += pl111/
 obj-$(CONFIG_DRM_TVE200) += tve200/
+obj-$(CONFIG_DRM_SCHED)	+= scheduler/

drivers/gpu/drm/amd/amdgpu/Makefile

@@ -135,10 +135,7 @@ amdgpu-y += \
 amdgpu-y += amdgpu_cgs.o
 
 # GPU scheduler
-amdgpu-y += \
-	../scheduler/gpu_scheduler.o \
-	../scheduler/sched_fence.o \
-	amdgpu_job.o
+amdgpu-y += amdgpu_job.o
 
 # ACP componet
 ifneq ($(CONFIG_DRM_AMD_ACP),)

drivers/gpu/drm/amd/amdgpu/amdgpu.h

@@ -45,6 +45,7 @@
 #include <drm/drmP.h>
 #include <drm/drm_gem.h>
 #include <drm/amdgpu_drm.h>
+#include <drm/gpu_scheduler.h>
 
 #include <kgd_kfd_interface.h>
 #include "dm_pp_interface.h"
@@ -68,7 +69,6 @@
 #include "amdgpu_vcn.h"
 #include "amdgpu_mn.h"
 #include "amdgpu_dm.h"
-#include "gpu_scheduler.h"
 #include "amdgpu_virt.h"
 #include "amdgpu_gart.h"
 
@@ -689,7 +689,7 @@ struct amdgpu_ib {
 	uint32_t			flags;
 };
 
-extern const struct amd_sched_backend_ops amdgpu_sched_ops;
+extern const struct drm_sched_backend_ops amdgpu_sched_ops;
 
 int amdgpu_job_alloc(struct amdgpu_device *adev, unsigned num_ibs,
 		     struct amdgpu_job **job, struct amdgpu_vm *vm);
@@ -699,7 +699,7 @@ int amdgpu_job_alloc_with_ib(struct amdgpu_device *adev, unsigned size,
 void amdgpu_job_free_resources(struct amdgpu_job *job);
 void amdgpu_job_free(struct amdgpu_job *job);
 int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
-		      struct amd_sched_entity *entity, void *owner,
+		      struct drm_sched_entity *entity, void *owner,
 		      struct dma_fence **f);
 
 /*
@@ -732,7 +732,7 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev,
 struct amdgpu_ctx_ring {
 	uint64_t		sequence;
 	struct dma_fence	**fences;
-	struct amd_sched_entity	entity;
+	struct drm_sched_entity	entity;
 };
 
 struct amdgpu_ctx {
@@ -746,8 +746,8 @@ struct amdgpu_ctx {
 	struct dma_fence	**fences;
 	struct amdgpu_ctx_ring	rings[AMDGPU_MAX_RINGS];
 	bool			preamble_presented;
-	enum amd_sched_priority init_priority;
-	enum amd_sched_priority override_priority;
+	enum drm_sched_priority init_priority;
+	enum drm_sched_priority override_priority;
 	struct mutex            lock;
 	atomic_t	guilty;
 };
@@ -767,7 +767,7 @@ int amdgpu_ctx_add_fence(struct amdgpu_ctx *ctx, struct amdgpu_ring *ring,
 struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
 				   struct amdgpu_ring *ring, uint64_t seq);
 void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
-				  enum amd_sched_priority priority);
+				  enum drm_sched_priority priority);
 
 int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 		     struct drm_file *filp);
@@ -1116,7 +1116,7 @@ struct amdgpu_cs_parser {
 #define AMDGPU_HAVE_CTX_SWITCH              (1 << 2) /* bit set means context switch occured */
 
 struct amdgpu_job {
-	struct amd_sched_job    base;
+	struct drm_sched_job    base;
 	struct amdgpu_device	*adev;
 	struct amdgpu_vm	*vm;
 	struct amdgpu_ring	*ring;

drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c

@@ -1150,7 +1150,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 			    union drm_amdgpu_cs *cs)
 {
 	struct amdgpu_ring *ring = p->job->ring;
-	struct amd_sched_entity *entity = &p->ctx->rings[ring->idx].entity;
+	struct drm_sched_entity *entity = &p->ctx->rings[ring->idx].entity;
 	struct amdgpu_job *job;
 	unsigned i;
 	uint64_t seq;
@@ -1173,7 +1173,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	job = p->job;
 	p->job = NULL;
 
-	r = amd_sched_job_init(&job->base, &ring->sched, entity, p->filp);
+	r = drm_sched_job_init(&job->base, &ring->sched, entity, p->filp);
 	if (r) {
 		amdgpu_job_free(job);
 		amdgpu_mn_unlock(p->mn);
@@ -1202,7 +1202,7 @@ static int amdgpu_cs_submit(struct amdgpu_cs_parser *p,
 	amdgpu_ring_priority_get(job->ring, job->base.s_priority);
 
 	trace_amdgpu_cs_ioctl(job);
-	amd_sched_entity_push_job(&job->base, entity);
+	drm_sched_entity_push_job(&job->base, entity);
 
 	ttm_eu_fence_buffer_objects(&p->ticket, &p->validated, p->fence);
 	amdgpu_mn_unlock(p->mn);

drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c

@@ -28,10 +28,10 @@
 #include "amdgpu_sched.h"
 
 static int amdgpu_ctx_priority_permit(struct drm_file *filp,
-				      enum amd_sched_priority priority)
+				      enum drm_sched_priority priority)
 {
 	/* NORMAL and below are accessible by everyone */
-	if (priority <= AMD_SCHED_PRIORITY_NORMAL)
+	if (priority <= DRM_SCHED_PRIORITY_NORMAL)
 		return 0;
 
 	if (capable(CAP_SYS_NICE))
@@ -44,14 +44,14 @@ static int amdgpu_ctx_priority_permit(struct drm_file *filp,
 }
 
 static int amdgpu_ctx_init(struct amdgpu_device *adev,
-			   enum amd_sched_priority priority,
+			   enum drm_sched_priority priority,
 			   struct drm_file *filp,
 			   struct amdgpu_ctx *ctx)
 {
 	unsigned i, j;
 	int r;
 
-	if (priority < 0 || priority >= AMD_SCHED_PRIORITY_MAX)
+	if (priority < 0 || priority >= DRM_SCHED_PRIORITY_MAX)
 		return -EINVAL;
 
 	r = amdgpu_ctx_priority_permit(filp, priority);
@@ -78,19 +78,19 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 	ctx->reset_counter_query = ctx->reset_counter;
 	ctx->vram_lost_counter = atomic_read(&adev->vram_lost_counter);
 	ctx->init_priority = priority;
-	ctx->override_priority = AMD_SCHED_PRIORITY_UNSET;
+	ctx->override_priority = DRM_SCHED_PRIORITY_UNSET;
 
 	/* create context entity for each ring */
 	for (i = 0; i < adev->num_rings; i++) {
 		struct amdgpu_ring *ring = adev->rings[i];
-		struct amd_sched_rq *rq;
+		struct drm_sched_rq *rq;
 
 		rq = &ring->sched.sched_rq[priority];
 
 		if (ring == &adev->gfx.kiq.ring)
 			continue;
 
-		r = amd_sched_entity_init(&ring->sched, &ctx->rings[i].entity,
+		r = drm_sched_entity_init(&ring->sched, &ctx->rings[i].entity,
 					  rq, amdgpu_sched_jobs, &ctx->guilty);
 		if (r)
 			goto failed;
@@ -104,7 +104,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev,
 
 failed:
 	for (j = 0; j < i; j++)
-		amd_sched_entity_fini(&adev->rings[j]->sched,
+		drm_sched_entity_fini(&adev->rings[j]->sched,
 				      &ctx->rings[j].entity);
 	kfree(ctx->fences);
 	ctx->fences = NULL;
@@ -126,7 +126,7 @@ static void amdgpu_ctx_fini(struct amdgpu_ctx *ctx)
 	ctx->fences = NULL;
 
 	for (i = 0; i < adev->num_rings; i++)
-		amd_sched_entity_fini(&adev->rings[i]->sched,
+		drm_sched_entity_fini(&adev->rings[i]->sched,
 				      &ctx->rings[i].entity);
 
 	amdgpu_queue_mgr_fini(adev, &ctx->queue_mgr);
@@ -137,7 +137,7 @@ static void amdgpu_ctx_fini(struct amdgpu_ctx *ctx)
 static int amdgpu_ctx_alloc(struct amdgpu_device *adev,
 			    struct amdgpu_fpriv *fpriv,
 			    struct drm_file *filp,
-			    enum amd_sched_priority priority,
+			    enum drm_sched_priority priority,
 			    uint32_t *id)
 {
 	struct amdgpu_ctx_mgr *mgr = &fpriv->ctx_mgr;
@@ -266,7 +266,7 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 {
 	int r;
 	uint32_t id;
-	enum amd_sched_priority priority;
+	enum drm_sched_priority priority;
 
 	union drm_amdgpu_ctx *args = data;
 	struct amdgpu_device *adev = dev->dev_private;
@@ -278,8 +278,8 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
 
 	/* For backwards compatibility reasons, we need to accept
 	 * ioctls with garbage in the priority field */
-	if (priority == AMD_SCHED_PRIORITY_INVALID)
-		priority = AMD_SCHED_PRIORITY_NORMAL;
+	if (priority == DRM_SCHED_PRIORITY_INVALID)
+		priority = DRM_SCHED_PRIORITY_NORMAL;
 
 	switch (args->in.op) {
 	case AMDGPU_CTX_OP_ALLOC_CTX:
@@ -385,18 +385,18 @@ struct dma_fence *amdgpu_ctx_get_fence(struct amdgpu_ctx *ctx,
 }
 
 void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
-				  enum amd_sched_priority priority)
+				  enum drm_sched_priority priority)
 {
 	int i;
 	struct amdgpu_device *adev = ctx->adev;
-	struct amd_sched_rq *rq;
-	struct amd_sched_entity *entity;
+	struct drm_sched_rq *rq;
+	struct drm_sched_entity *entity;
 	struct amdgpu_ring *ring;
-	enum amd_sched_priority ctx_prio;
+	enum drm_sched_priority ctx_prio;
 
 	ctx->override_priority = priority;
 
-	ctx_prio = (ctx->override_priority == AMD_SCHED_PRIORITY_UNSET) ?
+	ctx_prio = (ctx->override_priority == DRM_SCHED_PRIORITY_UNSET) ?
 			ctx->init_priority : ctx->override_priority;
 
 	for (i = 0; i < adev->num_rings; i++) {
@@ -407,7 +407,7 @@ void amdgpu_ctx_priority_override(struct amdgpu_ctx *ctx,
 		if (ring->funcs->type == AMDGPU_RING_TYPE_KIQ)
 			continue;
 
-		amd_sched_entity_set_rq(entity, rq);
+		drm_sched_entity_set_rq(entity, rq);
 	}
 }
 

drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

@@ -3058,7 +3058,7 @@ int amdgpu_gpu_recover(struct amdgpu_device *adev, struct amdgpu_job *job)
 			continue;
 
 		kthread_park(ring->sched.thread);
-		amd_sched_hw_job_reset(&ring->sched, &job->base);
+		drm_sched_hw_job_reset(&ring->sched, &job->base);
 
 		/* after all hw jobs are reset, hw fence is meaningless, so force_completion */
 		amdgpu_fence_driver_force_completion(ring);
@@ -3111,7 +3111,7 @@ int amdgpu_gpu_recover(struct amdgpu_device *adev, struct amdgpu_job *job)
 			if (job && job->ring->idx != i)
 				continue;
 
-			amd_sched_job_recovery(&ring->sched);
+			drm_sched_job_recovery(&ring->sched);
 			kthread_unpark(ring->sched.thread);
 		}
 	} else {

drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c

@@ -912,7 +912,7 @@ static int __init amdgpu_init(void)
 	if (r)
 		goto error_fence;
 
-	r = amd_sched_fence_slab_init();
+	r = drm_sched_fence_slab_init();
 	if (r)
 		goto error_sched;
 
@@ -944,7 +944,7 @@ static void __exit amdgpu_exit(void)
 	pci_unregister_driver(pdriver);
 	amdgpu_unregister_atpx_handler();
 	amdgpu_sync_fini();
-	amd_sched_fence_slab_fini();
+	drm_sched_fence_slab_fini();
 	amdgpu_fence_slab_fini();
 }
 

drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c

@@ -445,7 +445,7 @@ int amdgpu_fence_driver_init_ring(struct amdgpu_ring *ring,
 			 */
 			timeout = MAX_SCHEDULE_TIMEOUT;
 		}
-		r = amd_sched_init(&ring->sched, &amdgpu_sched_ops,
+		r = drm_sched_init(&ring->sched, &amdgpu_sched_ops,
 				   num_hw_submission, amdgpu_job_hang_limit,
 				   timeout, ring->name);
 		if (r) {
@@ -503,7 +503,7 @@ void amdgpu_fence_driver_fini(struct amdgpu_device *adev)
 		}
 		amdgpu_irq_put(adev, ring->fence_drv.irq_src,
 			       ring->fence_drv.irq_type);
-		amd_sched_fini(&ring->sched);
+		drm_sched_fini(&ring->sched);
 		del_timer_sync(&ring->fence_drv.fallback_timer);
 		for (j = 0; j <= ring->fence_drv.num_fences_mask; ++j)
 			dma_fence_put(ring->fence_drv.fences[j]);

drivers/gpu/drm/amd/amdgpu/amdgpu_job.c

@@ -28,7 +28,7 @@
 #include "amdgpu.h"
 #include "amdgpu_trace.h"
 
-static void amdgpu_job_timedout(struct amd_sched_job *s_job)
+static void amdgpu_job_timedout(struct drm_sched_job *s_job)
 {
 	struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base);
 
@@ -96,7 +96,7 @@ void amdgpu_job_free_resources(struct amdgpu_job *job)
 		amdgpu_ib_free(job->adev, &job->ibs[i], f);
 }
 
-static void amdgpu_job_free_cb(struct amd_sched_job *s_job)
+static void amdgpu_job_free_cb(struct drm_sched_job *s_job)
 {
 	struct amdgpu_job *job = container_of(s_job, struct amdgpu_job, base);
 
@@ -118,7 +118,7 @@ void amdgpu_job_free(struct amdgpu_job *job)
 }
 
 int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
-		      struct amd_sched_entity *entity, void *owner,
+		      struct drm_sched_entity *entity, void *owner,
 		      struct dma_fence **f)
 {
 	int r;
@@ -127,7 +127,7 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
 	if (!f)
 		return -EINVAL;
 
-	r = amd_sched_job_init(&job->base, &ring->sched, entity, owner);
+	r = drm_sched_job_init(&job->base, &ring->sched, entity, owner);
 	if (r)
 		return r;
 
@@ -136,13 +136,13 @@ int amdgpu_job_submit(struct amdgpu_job *job, struct amdgpu_ring *ring,
 	*f = dma_fence_get(&job->base.s_fence->finished);
 	amdgpu_job_free_resources(job);
 	amdgpu_ring_priority_get(job->ring, job->base.s_priority);
-	amd_sched_entity_push_job(&job->base, entity);
+	drm_sched_entity_push_job(&job->base, entity);
 
 	return 0;
 }
 
-static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job,
-					       struct amd_sched_entity *s_entity)
+static struct dma_fence *amdgpu_job_dependency(struct drm_sched_job *sched_job,
+					       struct drm_sched_entity *s_entity)
 {
 	struct amdgpu_job *job = to_amdgpu_job(sched_job);
 	struct amdgpu_vm *vm = job->vm;
@@ -151,7 +151,7 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job,
 	struct dma_fence *fence = amdgpu_sync_get_fence(&job->sync, &explicit);
 
 	if (fence && explicit) {
-		if (amd_sched_dependency_optimized(fence, s_entity)) {
+		if (drm_sched_dependency_optimized(fence, s_entity)) {
 			r = amdgpu_sync_fence(job->adev, &job->sched_sync, fence, false);
 			if (r)
 				DRM_ERROR("Error adding fence to sync (%d)\n", r);
@@ -173,7 +173,7 @@ static struct dma_fence *amdgpu_job_dependency(struct amd_sched_job *sched_job,
 	return fence;
 }
 
-static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job)
+static struct dma_fence *amdgpu_job_run(struct drm_sched_job *sched_job)
 {
 	struct dma_fence *fence = NULL, *finished;
 	struct amdgpu_device *adev;
@@ -211,7 +211,7 @@ static struct dma_fence *amdgpu_job_run(struct amd_sched_job *sched_job)
 	return fence;
 }
 
-const struct amd_sched_backend_ops amdgpu_sched_ops = {
+const struct drm_sched_backend_ops amdgpu_sched_ops = {
 	.dependency = amdgpu_job_dependency,
 	.run_job = amdgpu_job_run,
 	.timedout_job = amdgpu_job_timedout,

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c

@@ -164,7 +164,7 @@ void amdgpu_ring_undo(struct amdgpu_ring *ring)
  * Release a request for executing at @priority
  */
 void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
-			      enum amd_sched_priority priority)
+			      enum drm_sched_priority priority)
 {
 	int i;
 
@@ -175,7 +175,7 @@ void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
 		return;
 
 	/* no need to restore if the job is already at the lowest priority */
-	if (priority == AMD_SCHED_PRIORITY_NORMAL)
+	if (priority == DRM_SCHED_PRIORITY_NORMAL)
 		return;
 
 	mutex_lock(&ring->priority_mutex);
@@ -184,8 +184,8 @@ void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
 		goto out_unlock;
 
 	/* decay priority to the next level with a job available */
-	for (i = priority; i >= AMD_SCHED_PRIORITY_MIN; i--) {
-		if (i == AMD_SCHED_PRIORITY_NORMAL
+	for (i = priority; i >= DRM_SCHED_PRIORITY_MIN; i--) {
+		if (i == DRM_SCHED_PRIORITY_NORMAL
 				|| atomic_read(&ring->num_jobs[i])) {
 			ring->priority = i;
 			ring->funcs->set_priority(ring, i);
@@ -206,7 +206,7 @@ void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
  * Request a ring's priority to be raised to @priority (refcounted).
  */
 void amdgpu_ring_priority_get(struct amdgpu_ring *ring,
-			      enum amd_sched_priority priority)
+			      enum drm_sched_priority priority)
 {
 	if (!ring->funcs->set_priority)
 		return;
@@ -317,12 +317,12 @@ int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 	}
 
 	ring->max_dw = max_dw;
-	ring->priority = AMD_SCHED_PRIORITY_NORMAL;
+	ring->priority = DRM_SCHED_PRIORITY_NORMAL;
 	mutex_init(&ring->priority_mutex);
 	INIT_LIST_HEAD(&ring->lru_list);
 	amdgpu_ring_lru_touch(adev, ring);
 
-	for (i = 0; i < AMD_SCHED_PRIORITY_MAX; ++i)
+	for (i = 0; i < DRM_SCHED_PRIORITY_MAX; ++i)
 		atomic_set(&ring->num_jobs[i], 0);
 
 	if (amdgpu_debugfs_ring_init(adev, ring)) {

drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h

@@ -25,7 +25,7 @@
 #define __AMDGPU_RING_H__
 
 #include <drm/amdgpu_drm.h>
-#include "gpu_scheduler.h"
+#include <drm/gpu_scheduler.h>
 
 /* max number of rings */
 #define AMDGPU_MAX_RINGS		18
@@ -154,14 +154,14 @@ struct amdgpu_ring_funcs {
 	void (*emit_tmz)(struct amdgpu_ring *ring, bool start);
 	/* priority functions */
 	void (*set_priority) (struct amdgpu_ring *ring,
-			      enum amd_sched_priority priority);
+			      enum drm_sched_priority priority);
 };
 
 struct amdgpu_ring {
 	struct amdgpu_device		*adev;
 	const struct amdgpu_ring_funcs	*funcs;
 	struct amdgpu_fence_driver	fence_drv;
-	struct amd_gpu_scheduler	sched;
+	struct drm_gpu_scheduler	sched;
 	struct list_head		lru_list;
 
 	struct amdgpu_bo	*ring_obj;
@@ -196,7 +196,7 @@ struct amdgpu_ring {
 	unsigned		vm_inv_eng;
 	bool			has_compute_vm_bug;
 
-	atomic_t		num_jobs[AMD_SCHED_PRIORITY_MAX];
+	atomic_t		num_jobs[DRM_SCHED_PRIORITY_MAX];
 	struct mutex		priority_mutex;
 	/* protected by priority_mutex */
 	int			priority;
@@ -212,9 +212,9 @@ void amdgpu_ring_generic_pad_ib(struct amdgpu_ring *ring, struct amdgpu_ib *ib);
 void amdgpu_ring_commit(struct amdgpu_ring *ring);
 void amdgpu_ring_undo(struct amdgpu_ring *ring);
 void amdgpu_ring_priority_get(struct amdgpu_ring *ring,
-			      enum amd_sched_priority priority);
+			      enum drm_sched_priority priority);
 void amdgpu_ring_priority_put(struct amdgpu_ring *ring,
-			      enum amd_sched_priority priority);
+			      enum drm_sched_priority priority);
 int amdgpu_ring_init(struct amdgpu_device *adev, struct amdgpu_ring *ring,
 		     unsigned ring_size, struct amdgpu_irq_src *irq_src,
 		     unsigned irq_type);

drivers/gpu/drm/amd/amdgpu/amdgpu_sched.c

@@ -29,29 +29,29 @@
 
 #include "amdgpu_vm.h"
 
-enum amd_sched_priority amdgpu_to_sched_priority(int amdgpu_priority)
+enum drm_sched_priority amdgpu_to_sched_priority(int amdgpu_priority)
 {
 	switch (amdgpu_priority) {
 	case AMDGPU_CTX_PRIORITY_VERY_HIGH:
-		return AMD_SCHED_PRIORITY_HIGH_HW;
+		return DRM_SCHED_PRIORITY_HIGH_HW;
 	case AMDGPU_CTX_PRIORITY_HIGH:
-		return AMD_SCHED_PRIORITY_HIGH_SW;
+		return DRM_SCHED_PRIORITY_HIGH_SW;
 	case AMDGPU_CTX_PRIORITY_NORMAL:
-		return AMD_SCHED_PRIORITY_NORMAL;
+		return DRM_SCHED_PRIORITY_NORMAL;
 	case AMDGPU_CTX_PRIORITY_LOW:
 	case AMDGPU_CTX_PRIORITY_VERY_LOW:
-		return AMD_SCHED_PRIORITY_LOW;
+		return DRM_SCHED_PRIORITY_LOW;
 	case AMDGPU_CTX_PRIORITY_UNSET:
-		return AMD_SCHED_PRIORITY_UNSET;
+		return DRM_SCHED_PRIORITY_UNSET;
 	default:
 		WARN(1, "Invalid context priority %d\n", amdgpu_priority);
-		return AMD_SCHED_PRIORITY_INVALID;
+		return DRM_SCHED_PRIORITY_INVALID;
 	}
 }
 
 static int amdgpu_sched_process_priority_override(struct amdgpu_device *adev,
 						  int fd,
-						  enum amd_sched_priority priority)
+						  enum drm_sched_priority priority)
 {
 	struct file *filp = fcheck(fd);
 	struct drm_file *file;
@@ -86,11 +86,11 @@ int amdgpu_sched_ioctl(struct drm_device *dev, void *data,
 {
 	union drm_amdgpu_sched *args = data;
 	struct amdgpu_device *adev = dev->dev_private;
-	enum amd_sched_priority priority;
+	enum drm_sched_priority priority;
 	int r;
 
 	priority = amdgpu_to_sched_priority(args->in.priority);
-	if (args->in.flags || priority == AMD_SCHED_PRIORITY_INVALID)
+	if (args->in.flags || priority == DRM_SCHED_PRIORITY_INVALID)
 		return -EINVAL;
 
 	switch (args->in.op) {

drivers/gpu/drm/amd/amdgpu/amdgpu_sched.h

@@ -27,7 +27,7 @@
 
 #include <drm/drmP.h>
 
-enum amd_sched_priority amdgpu_to_sched_priority(int amdgpu_priority);
+enum drm_sched_priority amdgpu_to_sched_priority(int amdgpu_priority);
 int amdgpu_sched_ioctl(struct drm_device *dev, void *data,
 		       struct drm_file *filp);
 

drivers/gpu/drm/amd/amdgpu/amdgpu_sync.c

@@ -64,7 +64,7 @@ void amdgpu_sync_create(struct amdgpu_sync *sync)
 static bool amdgpu_sync_same_dev(struct amdgpu_device *adev,
 				 struct dma_fence *f)
 {
-	struct amd_sched_fence *s_fence = to_amd_sched_fence(f);
+	struct drm_sched_fence *s_fence = to_drm_sched_fence(f);
 
 	if (s_fence) {
 		struct amdgpu_ring *ring;
@@ -85,7 +85,7 @@ static bool amdgpu_sync_same_dev(struct amdgpu_device *adev,
  */
 static void *amdgpu_sync_get_owner(struct dma_fence *f)
 {
-	struct amd_sched_fence *s_fence = to_amd_sched_fence(f);
+	struct drm_sched_fence *s_fence = to_drm_sched_fence(f);
 
 	if (s_fence)
 		return s_fence->owner;
@@ -248,7 +248,7 @@ struct dma_fence *amdgpu_sync_peek_fence(struct amdgpu_sync *sync,
 
 	hash_for_each_safe(sync->fences, i, tmp, e, node) {
 		struct dma_fence *f = e->fence;
-		struct amd_sched_fence *s_fence = to_amd_sched_fence(f);
+		struct drm_sched_fence *s_fence = to_drm_sched_fence(f);
 
 		if (dma_fence_is_signaled(f)) {
 			hash_del(&e->node);

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

@@ -76,7 +76,7 @@ static int amdgpu_ttm_global_init(struct amdgpu_device *adev)
 {
 	struct drm_global_reference *global_ref;
 	struct amdgpu_ring *ring;
-	struct amd_sched_rq *rq;
+	struct drm_sched_rq *rq;
 	int r;
 
 	adev->mman.mem_global_referenced = false;
@@ -108,8 +108,8 @@ static int amdgpu_ttm_global_init(struct amdgpu_device *adev)
 	mutex_init(&adev->mman.gtt_window_lock);
 
 	ring = adev->mman.buffer_funcs_ring;
-	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_KERNEL];
-	r = amd_sched_entity_init(&ring->sched, &adev->mman.entity,
+	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+	r = drm_sched_entity_init(&ring->sched, &adev->mman.entity,
 				  rq, amdgpu_sched_jobs, NULL);
 	if (r) {
 		DRM_ERROR("Failed setting up TTM BO move run queue.\n");
@@ -131,7 +131,7 @@ static int amdgpu_ttm_global_init(struct amdgpu_device *adev)
 static void amdgpu_ttm_global_fini(struct amdgpu_device *adev)
 {
 	if (adev->mman.mem_global_referenced) {
-		amd_sched_entity_fini(adev->mman.entity.sched,
+		drm_sched_entity_fini(adev->mman.entity.sched,
 				      &adev->mman.entity);
 		mutex_destroy(&adev->mman.gtt_window_lock);
 		drm_global_item_unref(&adev->mman.bo_global_ref.ref);

drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h

@@ -25,7 +25,7 @@
 #define __AMDGPU_TTM_H__
 
 #include "amdgpu.h"
-#include "gpu_scheduler.h"
+#include <drm/gpu_scheduler.h>
 
 #define AMDGPU_PL_GDS		(TTM_PL_PRIV + 0)
 #define AMDGPU_PL_GWS		(TTM_PL_PRIV + 1)
@@ -55,7 +55,7 @@ struct amdgpu_mman {
 
 	struct mutex				gtt_window_lock;
 	/* Scheduler entity for buffer moves */
-	struct amd_sched_entity			entity;
+	struct drm_sched_entity			entity;
 };
 
 struct amdgpu_copy_mem {

drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c

@@ -116,7 +116,7 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work);
 int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	struct amd_sched_rq *rq;
+	struct drm_sched_rq *rq;
 	unsigned long bo_size;
 	const char *fw_name;
 	const struct common_firmware_header *hdr;
@@ -230,8 +230,8 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev)
 	}
 
 	ring = &adev->uvd.ring;
-	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
-	r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity,
+	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
+	r = drm_sched_entity_init(&ring->sched, &adev->uvd.entity,
 				  rq, amdgpu_sched_jobs, NULL);
 	if (r != 0) {
 		DRM_ERROR("Failed setting up UVD run queue.\n");
@@ -272,7 +272,7 @@ int amdgpu_uvd_sw_fini(struct amdgpu_device *adev)
 	int i;
 	kfree(adev->uvd.saved_bo);
 
-	amd_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity);
+	drm_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity);
 
 	amdgpu_bo_free_kernel(&adev->uvd.vcpu_bo,
 			      &adev->uvd.gpu_addr,

drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h

@@ -51,8 +51,8 @@ struct amdgpu_uvd {
 	struct amdgpu_irq_src	irq;
 	bool			address_64_bit;
 	bool			use_ctx_buf;
-	struct amd_sched_entity entity;
-	struct amd_sched_entity entity_enc;
+	struct drm_sched_entity entity;
+	struct drm_sched_entity entity_enc;
 	uint32_t                srbm_soft_reset;
 	unsigned		num_enc_rings;
 };

drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c

@@ -85,7 +85,7 @@ static void amdgpu_vce_idle_work_handler(struct work_struct *work);
 int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
 {
 	struct amdgpu_ring *ring;
-	struct amd_sched_rq *rq;
+	struct drm_sched_rq *rq;
 	const char *fw_name;
 	const struct common_firmware_header *hdr;
 	unsigned ucode_version, version_major, version_minor, binary_id;
@@ -174,8 +174,8 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size)
 	}
 
 	ring = &adev->vce.ring[0];
-	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
-	r = amd_sched_entity_init(&ring->sched, &adev->vce.entity,
+	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
+	r = drm_sched_entity_init(&ring->sched, &adev->vce.entity,
 				  rq, amdgpu_sched_jobs, NULL);
 	if (r != 0) {
 		DRM_ERROR("Failed setting up VCE run queue.\n");
@@ -207,7 +207,7 @@ int amdgpu_vce_sw_fini(struct amdgpu_device *adev)
 	if (adev->vce.vcpu_bo == NULL)
 		return 0;
 
-	amd_sched_entity_fini(&adev->vce.ring[0].sched, &adev->vce.entity);
+	drm_sched_entity_fini(&adev->vce.ring[0].sched, &adev->vce.entity);
 
 	amdgpu_bo_free_kernel(&adev->vce.vcpu_bo, &adev->vce.gpu_addr,
 		(void **)&adev->vce.cpu_addr);

drivers/gpu/drm/amd/amdgpu/amdgpu_vce.h

@@ -46,7 +46,7 @@ struct amdgpu_vce {
 	struct amdgpu_ring	ring[AMDGPU_MAX_VCE_RINGS];
 	struct amdgpu_irq_src	irq;
 	unsigned		harvest_config;
-	struct amd_sched_entity	entity;
+	struct drm_sched_entity	entity;
 	uint32_t                srbm_soft_reset;
 	unsigned		num_rings;
 };

drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c

@@ -51,7 +51,7 @@ static void amdgpu_vcn_idle_work_handler(struct work_struct *work);
 int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 {
 	struct amdgpu_ring *ring;
-	struct amd_sched_rq *rq;
+	struct drm_sched_rq *rq;
 	unsigned long bo_size;
 	const char *fw_name;
 	const struct common_firmware_header *hdr;
@@ -104,8 +104,8 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 	}
 
 	ring = &adev->vcn.ring_dec;
-	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
-	r = amd_sched_entity_init(&ring->sched, &adev->vcn.entity_dec,
+	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
+	r = drm_sched_entity_init(&ring->sched, &adev->vcn.entity_dec,
 				  rq, amdgpu_sched_jobs, NULL);
 	if (r != 0) {
 		DRM_ERROR("Failed setting up VCN dec run queue.\n");
@@ -113,8 +113,8 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev)
 	}
 
 	ring = &adev->vcn.ring_enc[0];
-	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
-	r = amd_sched_entity_init(&ring->sched, &adev->vcn.entity_enc,
+	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
+	r = drm_sched_entity_init(&ring->sched, &adev->vcn.entity_enc,
 				  rq, amdgpu_sched_jobs, NULL);
 	if (r != 0) {
 		DRM_ERROR("Failed setting up VCN enc run queue.\n");
@@ -130,9 +130,9 @@ int amdgpu_vcn_sw_fini(struct amdgpu_device *adev)
 
 	kfree(adev->vcn.saved_bo);
 
-	amd_sched_entity_fini(&adev->vcn.ring_dec.sched, &adev->vcn.entity_dec);
+	drm_sched_entity_fini(&adev->vcn.ring_dec.sched, &adev->vcn.entity_dec);
 
-	amd_sched_entity_fini(&adev->vcn.ring_enc[0].sched, &adev->vcn.entity_enc);
+	drm_sched_entity_fini(&adev->vcn.ring_enc[0].sched, &adev->vcn.entity_enc);
 
 	amdgpu_bo_free_kernel(&adev->vcn.vcpu_bo,
 			      &adev->vcn.gpu_addr,

drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h

@@ -56,8 +56,8 @@ struct amdgpu_vcn {
 	struct amdgpu_ring	ring_dec;
 	struct amdgpu_ring	ring_enc[AMDGPU_VCN_MAX_ENC_RINGS];
 	struct amdgpu_irq_src	irq;
-	struct amd_sched_entity entity_dec;
-	struct amd_sched_entity entity_enc;
+	struct drm_sched_entity entity_dec;
+	struct drm_sched_entity entity_enc;
 	unsigned		num_enc_rings;
 };
 

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c

@@ -2643,7 +2643,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 		AMDGPU_VM_PTE_COUNT(adev) * 8);
 	unsigned ring_instance;
 	struct amdgpu_ring *ring;
-	struct amd_sched_rq *rq;
+	struct drm_sched_rq *rq;
 	int r, i;
 	u64 flags;
 	uint64_t init_pde_value = 0;
@@ -2663,8 +2663,8 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	ring_instance = atomic_inc_return(&adev->vm_manager.vm_pte_next_ring);
 	ring_instance %= adev->vm_manager.vm_pte_num_rings;
 	ring = adev->vm_manager.vm_pte_rings[ring_instance];
-	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_KERNEL];
-	r = amd_sched_entity_init(&ring->sched, &vm->entity,
+	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL];
+	r = drm_sched_entity_init(&ring->sched, &vm->entity,
 				  rq, amdgpu_sched_jobs, NULL);
 	if (r)
 		return r;
@@ -2744,7 +2744,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,
 	vm->root.base.bo = NULL;
 
 error_free_sched_entity:
-	amd_sched_entity_fini(&ring->sched, &vm->entity);
+	drm_sched_entity_fini(&ring->sched, &vm->entity);
 
 	return r;
 }
@@ -2803,7 +2803,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)
 		spin_unlock_irqrestore(&adev->vm_manager.pasid_lock, flags);
 	}
 
-	amd_sched_entity_fini(vm->entity.sched, &vm->entity);
+	drm_sched_entity_fini(vm->entity.sched, &vm->entity);
 
 	if (!RB_EMPTY_ROOT(&vm->va.rb_root)) {
 		dev_err(adev->dev, "still active bo inside vm\n");

drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h

@@ -24,10 +24,11 @@
 #ifndef __AMDGPU_VM_H__
 #define __AMDGPU_VM_H__
 
-#include <linux/rbtree.h>
 #include <linux/idr.h>
+#include <linux/kfifo.h>
+#include <linux/rbtree.h>
+#include <drm/gpu_scheduler.h>
 
-#include "gpu_scheduler.h"
 #include "amdgpu_sync.h"
 #include "amdgpu_ring.h"
 
@@ -175,7 +176,7 @@ struct amdgpu_vm {
 	spinlock_t		freed_lock;
 
 	/* Scheduler entity for page table updates */
-	struct amd_sched_entity	entity;
+	struct drm_sched_entity	entity;
 
 	/* client id and PASID (TODO: replace client_id with PASID) */
 	u64                     client_id;

drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c

@@ -6472,10 +6472,10 @@ static void gfx_v8_0_hqd_set_priority(struct amdgpu_device *adev,
 	mutex_unlock(&adev->srbm_mutex);
 }
 static void gfx_v8_0_ring_set_priority_compute(struct amdgpu_ring *ring,
-					       enum amd_sched_priority priority)
+					       enum drm_sched_priority priority)
 {
 	struct amdgpu_device *adev = ring->adev;
-	bool acquire = priority == AMD_SCHED_PRIORITY_HIGH_HW;
+	bool acquire = priority == DRM_SCHED_PRIORITY_HIGH_HW;
 
 	if (ring->funcs->type != AMDGPU_RING_TYPE_COMPUTE)
 		return;

drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c

@@ -412,10 +412,10 @@ static int uvd_v6_0_sw_init(void *handle)
 		return r;
 
 	if (uvd_v6_0_enc_support(adev)) {
-		struct amd_sched_rq *rq;
+		struct drm_sched_rq *rq;
 		ring = &adev->uvd.ring_enc[0];
-		rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
-		r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity_enc,
+		rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
+		r = drm_sched_entity_init(&ring->sched, &adev->uvd.entity_enc,
 					  rq, amdgpu_sched_jobs, NULL);
 		if (r) {
 			DRM_ERROR("Failed setting up UVD ENC run queue.\n");
@@ -456,7 +456,7 @@ static int uvd_v6_0_sw_fini(void *handle)
 		return r;
 
 	if (uvd_v6_0_enc_support(adev)) {
-		amd_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc);
+		drm_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc);
 
 		for (i = 0; i < adev->uvd.num_enc_rings; ++i)
 			amdgpu_ring_fini(&adev->uvd.ring_enc[i]);

drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c

@@ -385,7 +385,7 @@ static int uvd_v7_0_early_init(void *handle)
 static int uvd_v7_0_sw_init(void *handle)
 {
 	struct amdgpu_ring *ring;
-	struct amd_sched_rq *rq;
+	struct drm_sched_rq *rq;
 	int i, r;
 	struct amdgpu_device *adev = (struct amdgpu_device *)handle;
 
@@ -416,8 +416,8 @@ static int uvd_v7_0_sw_init(void *handle)
 	}
 
 	ring = &adev->uvd.ring_enc[0];
-	rq = &ring->sched.sched_rq[AMD_SCHED_PRIORITY_NORMAL];
-	r = amd_sched_entity_init(&ring->sched, &adev->uvd.entity_enc,
+	rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL];
+	r = drm_sched_entity_init(&ring->sched, &adev->uvd.entity_enc,
 				  rq, amdgpu_sched_jobs, NULL);
 	if (r) {
 		DRM_ERROR("Failed setting up UVD ENC run queue.\n");
@@ -472,7 +472,7 @@ static int uvd_v7_0_sw_fini(void *handle)
 	if (r)
 		return r;
 
-	amd_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc);
+	drm_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc);
 
 	for (i = 0; i < adev->uvd.num_enc_rings; ++i)
 		amdgpu_ring_fini(&adev->uvd.ring_enc[i]);

drivers/gpu/drm/scheduler/Makefile

+ccflags-y := -Iinclude/drm
+gpu-sched-y := gpu_scheduler.o sched_fence.o
+
+obj-$(CONFIG_DRM_SCHED) += gpu-sched.o

drivers/gpu/drm/amd/scheduler/sched_fence.c → drivers/gpu/drm/scheduler/sched_fence.c

@@ -19,57 +19,36 @@
  * 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 "gpu_scheduler.h"
+#include <drm/gpu_scheduler.h>
 
 static struct kmem_cache *sched_fence_slab;
 
-int amd_sched_fence_slab_init(void)
+int drm_sched_fence_slab_init(void)
 {
 	sched_fence_slab = kmem_cache_create(
-		"amd_sched_fence", sizeof(struct amd_sched_fence), 0,
+		"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 amd_sched_fence_slab_fini(void)
+void drm_sched_fence_slab_fini(void)
 {
 	rcu_barrier();
 	kmem_cache_destroy(sched_fence_slab);
 }
+EXPORT_SYMBOL_GPL(drm_sched_fence_slab_fini);
 
-struct amd_sched_fence *amd_sched_fence_create(struct amd_sched_entity *entity,
-					       void *owner)
-{
-	struct amd_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, &amd_sched_fence_ops_scheduled,
-		       &fence->lock, entity->fence_context, seq);
-	dma_fence_init(&fence->finished, &amd_sched_fence_ops_finished,
-		       &fence->lock, entity->fence_context + 1, seq);
-
-	return fence;
-}
-
-void amd_sched_fence_scheduled(struct amd_sched_fence *fence)
+void drm_sched_fence_scheduled(struct drm_sched_fence *fence)
 {
 	int ret = dma_fence_signal(&fence->scheduled);
 
@@ -81,7 +60,7 @@ void amd_sched_fence_scheduled(struct amd_sched_fence *fence)
 				"was already signaled\n");
 }
 
-void amd_sched_fence_finished(struct amd_sched_fence *fence)
+void drm_sched_fence_finished(struct drm_sched_fence *fence)
 {
 	int ret = dma_fence_signal(&fence->finished);
 
@@ -93,18 +72,18 @@ void amd_sched_fence_finished(struct amd_sched_fence *fence)
 				"was already signaled\n");
 }
 
-static const char *amd_sched_fence_get_driver_name(struct dma_fence *fence)
+static const char *drm_sched_fence_get_driver_name(struct dma_fence *fence)
 {
-	return "amd_sched";
+	return "drm_sched";
 }
 
-static const char *amd_sched_fence_get_timeline_name(struct dma_fence *f)
+static const char *drm_sched_fence_get_timeline_name(struct dma_fence *f)
 {
-	struct amd_sched_fence *fence = to_amd_sched_fence(f);
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
 	return (const char *)fence->sched->name;
 }
 
-static bool amd_sched_fence_enable_signaling(struct dma_fence *f)
+static bool drm_sched_fence_enable_signaling(struct dma_fence *f)
 {
 	return true;
 }
@@ -116,10 +95,10 @@ static bool amd_sched_fence_enable_signaling(struct dma_fence *f)
  *
  * Free up the fence memory after the RCU grace period.
  */
-static void amd_sched_fence_free(struct rcu_head *rcu)
+static void drm_sched_fence_free(struct rcu_head *rcu)
 {
 	struct dma_fence *f = container_of(rcu, struct dma_fence, rcu);
-	struct amd_sched_fence *fence = to_amd_sched_fence(f);
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
 
 	dma_fence_put(fence->parent);
 	kmem_cache_free(sched_fence_slab, fence);
@@ -133,11 +112,11 @@ static void amd_sched_fence_free(struct rcu_head *rcu)
  * This function is called when the reference count becomes zero.
  * It just RCU schedules freeing up the fence.
  */
-static void amd_sched_fence_release_scheduled(struct dma_fence *f)
+static void drm_sched_fence_release_scheduled(struct dma_fence *f)
 {
-	struct amd_sched_fence *fence = to_amd_sched_fence(f);
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
 
-	call_rcu(&fence->finished.rcu, amd_sched_fence_free);
+	call_rcu(&fence->finished.rcu, drm_sched_fence_free);
 }
 
 /**
@@ -147,27 +126,62 @@ static void amd_sched_fence_release_scheduled(struct dma_fence *f)
  *
  * Drop the extra reference from the scheduled fence to the base fence.
  */
-static void amd_sched_fence_release_finished(struct dma_fence *f)
+static void drm_sched_fence_release_finished(struct dma_fence *f)
 {
-	struct amd_sched_fence *fence = to_amd_sched_fence(f);
+	struct drm_sched_fence *fence = to_drm_sched_fence(f);
 
 	dma_fence_put(&fence->scheduled);
 }
 
-const struct dma_fence_ops amd_sched_fence_ops_scheduled = {
-	.get_driver_name = amd_sched_fence_get_driver_name,
-	.get_timeline_name = amd_sched_fence_get_timeline_name,
-	.enable_signaling = amd_sched_fence_enable_signaling,
+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 = amd_sched_fence_release_scheduled,
+	.release = drm_sched_fence_release_scheduled,
 };
 
-const struct dma_fence_ops amd_sched_fence_ops_finished = {
-	.get_driver_name = amd_sched_fence_get_driver_name,
-	.get_timeline_name = amd_sched_fence_get_timeline_name,
-	.enable_signaling = amd_sched_fence_enable_signaling,
+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 = amd_sched_fence_release_finished,
+	.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;
+}

drivers/gpu/drm/amd/scheduler/gpu_scheduler.h → include/drm/gpu_scheduler.h

@@ -21,28 +21,26 @@
  *
  */
 
-#ifndef _GPU_SCHEDULER_H_
-#define _GPU_SCHEDULER_H_
+#ifndef _DRM_GPU_SCHEDULER_H_
+#define _DRM_GPU_SCHEDULER_H_
 
-#include <linux/kfifo.h>
+#include <drm/spsc_queue.h>
 #include <linux/dma-fence.h>
-#include "spsc_queue.h"
-
-struct amd_gpu_scheduler;
-struct amd_sched_rq;
-
-enum amd_sched_priority {
-	AMD_SCHED_PRIORITY_MIN,
-	AMD_SCHED_PRIORITY_LOW = AMD_SCHED_PRIORITY_MIN,
-	AMD_SCHED_PRIORITY_NORMAL,
-	AMD_SCHED_PRIORITY_HIGH_SW,
-	AMD_SCHED_PRIORITY_HIGH_HW,
-	AMD_SCHED_PRIORITY_KERNEL,
-	AMD_SCHED_PRIORITY_MAX,
-	AMD_SCHED_PRIORITY_INVALID = -1,
-	AMD_SCHED_PRIORITY_UNSET = -2
-};
 
+struct drm_gpu_scheduler;
+struct drm_sched_rq;
+
+enum drm_sched_priority {
+	DRM_SCHED_PRIORITY_MIN,
+	DRM_SCHED_PRIORITY_LOW = DRM_SCHED_PRIORITY_MIN,
+	DRM_SCHED_PRIORITY_NORMAL,
+	DRM_SCHED_PRIORITY_HIGH_SW,
+	DRM_SCHED_PRIORITY_HIGH_HW,
+	DRM_SCHED_PRIORITY_KERNEL,
+	DRM_SCHED_PRIORITY_MAX,
+	DRM_SCHED_PRIORITY_INVALID = -1,
+	DRM_SCHED_PRIORITY_UNSET = -2
+};
 
 /**
  * A scheduler entity is a wrapper around a job queue or a group
@@ -50,21 +48,21 @@ enum amd_sched_priority {
  * job queues to corresponding hardware ring based on scheduling
  * policy.
 */
-struct amd_sched_entity {
+struct drm_sched_entity {
 	struct list_head		list;
-	struct amd_sched_rq		*rq;
+	struct drm_sched_rq		*rq;
 	spinlock_t			rq_lock;
-	struct amd_gpu_scheduler	*sched;
+	struct drm_gpu_scheduler	*sched;
 
 	spinlock_t			queue_lock;
-	struct spsc_queue	job_queue;
+	struct spsc_queue		job_queue;
 
 	atomic_t			fence_seq;
-	uint64_t                        fence_context;
+	uint64_t			fence_context;
 
 	struct dma_fence		*dependency;
 	struct dma_fence_cb		cb;
-	atomic_t	*guilty; /* points to ctx's guilty */
+	atomic_t			*guilty; /* points to ctx's guilty */
 };
 
 /**
@@ -72,49 +70,39 @@ struct amd_sched_entity {
  * one specific ring. It implements the scheduling policy that selects
  * the next entity to emit commands from.
 */
-struct amd_sched_rq {
-	spinlock_t		lock;
-	struct list_head	entities;
-	struct amd_sched_entity	*current_entity;
+struct drm_sched_rq {
+	spinlock_t			lock;
+	struct list_head		entities;
+	struct drm_sched_entity		*current_entity;
 };
 
-struct amd_sched_fence {
-	struct dma_fence                scheduled;
-	struct dma_fence                finished;
-	struct dma_fence_cb             cb;
-	struct dma_fence                *parent;
-	struct amd_gpu_scheduler	*sched;
+struct drm_sched_fence {
+	struct dma_fence		scheduled;
+	struct dma_fence		finished;
+	struct dma_fence_cb		cb;
+	struct dma_fence		*parent;
+	struct drm_gpu_scheduler	*sched;
 	spinlock_t			lock;
-	void                            *owner;
+	void				*owner;
 };
 
-struct amd_sched_job {
-	struct spsc_node queue_node;
-	struct amd_gpu_scheduler        *sched;
-	struct amd_sched_fence          *s_fence;
+struct drm_sched_fence *to_drm_sched_fence(struct dma_fence *f);
+
+struct drm_sched_job {
+	struct spsc_node		queue_node;
+	struct drm_gpu_scheduler	*sched;
+	struct drm_sched_fence		*s_fence;
 	struct dma_fence_cb		finish_cb;
 	struct work_struct		finish_work;
 	struct list_head		node;
 	struct delayed_work		work_tdr;
 	uint64_t			id;
-	atomic_t karma;
-	enum amd_sched_priority s_priority;
+	atomic_t			karma;
+	enum drm_sched_priority		s_priority;
 };
 
-extern const struct dma_fence_ops amd_sched_fence_ops_scheduled;
-extern const struct dma_fence_ops amd_sched_fence_ops_finished;
-static inline struct amd_sched_fence *to_amd_sched_fence(struct dma_fence *f)
-{
-	if (f->ops == &amd_sched_fence_ops_scheduled)
-		return container_of(f, struct amd_sched_fence, scheduled);
-
-	if (f->ops == &amd_sched_fence_ops_finished)
-		return container_of(f, struct amd_sched_fence, finished);
-
-	return NULL;
-}
-
-static inline bool amd_sched_invalidate_job(struct amd_sched_job *s_job, int threshold)
+static inline bool drm_sched_invalidate_job(struct drm_sched_job *s_job,
+					    int threshold)
 {
 	return (s_job && atomic_inc_return(&s_job->karma) > threshold);
 }
@@ -123,64 +111,66 @@ static inline bool amd_sched_invalidate_job(struct amd_sched_job *s_job, int thr
  * Define the backend operations called by the scheduler,
  * these functions should be implemented in driver side
 */
-struct amd_sched_backend_ops {
-	struct dma_fence *(*dependency)(struct amd_sched_job *sched_job,
-					struct amd_sched_entity *s_entity);
-	struct dma_fence *(*run_job)(struct amd_sched_job *sched_job);
-	void (*timedout_job)(struct amd_sched_job *sched_job);
-	void (*free_job)(struct amd_sched_job *sched_job);
+struct drm_sched_backend_ops {
+	struct dma_fence *(*dependency)(struct drm_sched_job *sched_job,
+					struct drm_sched_entity *s_entity);
+	struct dma_fence *(*run_job)(struct drm_sched_job *sched_job);
+	void (*timedout_job)(struct drm_sched_job *sched_job);
+	void (*free_job)(struct drm_sched_job *sched_job);
 };
 
 /**
  * One scheduler is implemented for each hardware ring
 */
-struct amd_gpu_scheduler {
-	const struct amd_sched_backend_ops	*ops;
+struct drm_gpu_scheduler {
+	const struct drm_sched_backend_ops	*ops;
 	uint32_t			hw_submission_limit;
 	long				timeout;
 	const char			*name;
-	struct amd_sched_rq		sched_rq[AMD_SCHED_PRIORITY_MAX];
+	struct drm_sched_rq		sched_rq[DRM_SCHED_PRIORITY_MAX];
 	wait_queue_head_t		wake_up_worker;
 	wait_queue_head_t		job_scheduled;
 	atomic_t			hw_rq_count;
 	atomic64_t			job_id_count;
 	struct task_struct		*thread;
-	struct list_head	ring_mirror_list;
+	struct list_head		ring_mirror_list;
 	spinlock_t			job_list_lock;
-	int hang_limit;
+	int				hang_limit;
 };
 
-int amd_sched_init(struct amd_gpu_scheduler *sched,
-		   const struct amd_sched_backend_ops *ops,
-		   uint32_t hw_submission, unsigned hang_limit, long timeout, 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,
-			  struct amd_sched_rq *rq,
-			  uint32_t jobs, atomic_t* guilty);
-void amd_sched_entity_fini(struct amd_gpu_scheduler *sched,
-			   struct amd_sched_entity *entity);
-void amd_sched_entity_push_job(struct amd_sched_job *sched_job,
-			       struct amd_sched_entity *entity);
-void amd_sched_entity_set_rq(struct amd_sched_entity *entity,
-			     struct amd_sched_rq *rq);
-
-int amd_sched_fence_slab_init(void);
-void amd_sched_fence_slab_fini(void);
-
-struct amd_sched_fence *amd_sched_fence_create(
-	struct amd_sched_entity *s_entity, void *owner);
-void amd_sched_fence_scheduled(struct amd_sched_fence *fence);
-void amd_sched_fence_finished(struct amd_sched_fence *fence);
-int amd_sched_job_init(struct amd_sched_job *job,
-		       struct amd_gpu_scheduler *sched,
-		       struct amd_sched_entity *entity,
+int drm_sched_init(struct drm_gpu_scheduler *sched,
+		   const struct drm_sched_backend_ops *ops,
+		   uint32_t hw_submission, unsigned hang_limit, long timeout,
+		   const char *name);
+void drm_sched_fini(struct drm_gpu_scheduler *sched);
+
+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);
+void drm_sched_entity_fini(struct drm_gpu_scheduler *sched,
+			   struct drm_sched_entity *entity);
+void drm_sched_entity_push_job(struct drm_sched_job *sched_job,
+			       struct drm_sched_entity *entity);
+void drm_sched_entity_set_rq(struct drm_sched_entity *entity,
+			     struct drm_sched_rq *rq);
+
+int drm_sched_fence_slab_init(void);
+void drm_sched_fence_slab_fini(void);
+
+struct drm_sched_fence *drm_sched_fence_create(
+	struct drm_sched_entity *s_entity, void *owner);
+void drm_sched_fence_scheduled(struct drm_sched_fence *fence);
+void drm_sched_fence_finished(struct drm_sched_fence *fence);
+int drm_sched_job_init(struct drm_sched_job *job,
+		       struct drm_gpu_scheduler *sched,
+		       struct drm_sched_entity *entity,
 		       void *owner);
-void amd_sched_hw_job_reset(struct amd_gpu_scheduler *sched, struct amd_sched_job *job);
-void amd_sched_job_recovery(struct amd_gpu_scheduler *sched);
-bool amd_sched_dependency_optimized(struct dma_fence* fence,
-				    struct amd_sched_entity *entity);
-void amd_sched_job_kickout(struct amd_sched_job *s_job);
+void drm_sched_hw_job_reset(struct drm_gpu_scheduler *sched,
+			    struct drm_sched_job *job);
+void drm_sched_job_recovery(struct drm_gpu_scheduler *sched);
+bool drm_sched_dependency_optimized(struct dma_fence* fence,
+				    struct drm_sched_entity *entity);
+void drm_sched_job_kickout(struct drm_sched_job *s_job);
 
 #endif

drivers/gpu/drm/amd/scheduler/gpu_sched_trace.h → include/drm/gpu_scheduler_trace.h

@@ -31,14 +31,14 @@
 #include <drm/drmP.h>
 
 #undef TRACE_SYSTEM
-#define TRACE_SYSTEM gpu_sched
-#define TRACE_INCLUDE_FILE gpu_sched_trace
+#define TRACE_SYSTEM gpu_scheduler
+#define TRACE_INCLUDE_FILE gpu_scheduler_trace
 
-TRACE_EVENT(amd_sched_job,
-	    TP_PROTO(struct amd_sched_job *sched_job, struct amd_sched_entity *entity),
+TRACE_EVENT(drm_sched_job,
+	    TP_PROTO(struct drm_sched_job *sched_job, struct drm_sched_entity *entity),
 	    TP_ARGS(sched_job, entity),
 	    TP_STRUCT__entry(
-			     __field(struct amd_sched_entity *, entity)
+			     __field(struct drm_sched_entity *, entity)
 			     __field(struct dma_fence *, fence)
 			     __field(const char *, name)
 			     __field(uint64_t, id)
@@ -61,8 +61,8 @@ TRACE_EVENT(amd_sched_job,
 		      __entry->job_count, __entry->hw_job_count)
 );
 
-TRACE_EVENT(amd_sched_process_job,
-	    TP_PROTO(struct amd_sched_fence *fence),
+TRACE_EVENT(drm_sched_process_job,
+	    TP_PROTO(struct drm_sched_fence *fence),
 	    TP_ARGS(fence),
 	    TP_STRUCT__entry(
 		    __field(struct dma_fence *, fence)

drivers/gpu/drm/amd/scheduler/spsc_queue.h → include/drm/spsc_queue.h

@@ -21,10 +21,11 @@
  *
  */
 
-#ifndef AMD_SCHEDULER_SPSC_QUEUE_H_
-#define AMD_SCHEDULER_SPSC_QUEUE_H_
+#ifndef DRM_SCHEDULER_SPSC_QUEUE_H_
+#define DRM_SCHEDULER_SPSC_QUEUE_H_
 
 #include <linux/atomic.h>
+#include <linux/preempt.h>
 
 /** SPSC lockless queue */
 
@@ -118,4 +119,4 @@ static inline struct spsc_node *spsc_queue_pop(struct spsc_queue *queue)
 
 
 
-#endif /* AMD_SCHEDULER_SPSC_QUEUE_H_ */
+#endif /* DRM_SCHEDULER_SPSC_QUEUE_H_ */