Switching more drivers to use drm_exec

Switching more drivers to use drm_exec

[Christian König]

Hi guys,

with the exception of VMWGFX this moves over all the drivers to using the new drm_exec object.

Please review and/or comment,
Christian.

[PATCH 1/7] drm/radeon: switch over to drm_exec

[Christian König]

Just a straightforward conversion without any optimization.

Signed-off-by: Christian König <christian.koenig@amd.com>
Cc: amd-gfx@lists.freedesktop.org
---
 drivers/gpu/drm/radeon/Kconfig         |  1 +
 drivers/gpu/drm/radeon/radeon.h        |  7 ++--
 drivers/gpu/drm/radeon/radeon_cs.c     | 45 +++++++++++++-------------
 drivers/gpu/drm/radeon/radeon_gem.c    | 39 ++++++++++++----------
 drivers/gpu/drm/radeon/radeon_object.c | 25 +++++++-------
 drivers/gpu/drm/radeon/radeon_object.h |  2 +-
 drivers/gpu/drm/radeon/radeon_vm.c     | 10 +++---
 7 files changed, 66 insertions(+), 63 deletions(-)

diff --git a/drivers/gpu/drm/radeon/Kconfig b/drivers/gpu/drm/radeon/Kconfig
index fe498c8af1bb..fdfede471a6d 100644
--- a/drivers/gpu/drm/radeon/Kconfig
+++ b/drivers/gpu/drm/radeon/Kconfig
@@ -12,6 +12,7 @@ config DRM_RADEON
         select DRM_TTM
 	select DRM_TTM_HELPER
 	select FB_IO_HELPERS if DRM_FBDEV_EMULATION
+	select DRM_EXEC
 	select SND_HDA_COMPONENT if SND_HDA_CORE
 	select POWER_SUPPLY
 	select HWMON
diff --git a/drivers/gpu/drm/radeon/radeon.h b/drivers/gpu/drm/radeon/radeon.h
index 8afb03bbce29..37a932a5195f 100644
--- a/drivers/gpu/drm/radeon/radeon.h
+++ b/drivers/gpu/drm/radeon/radeon.h
@@ -75,8 +75,8 @@
 
 #include <drm/ttm/ttm_bo.h>
 #include <drm/ttm/ttm_placement.h>
-#include <drm/ttm/ttm_execbuf_util.h>
 
+#include <drm/drm_exec.h>
 #include <drm/drm_gem.h>
 #include <drm/drm_audio_component.h>
 #include <drm/drm_suballoc.h>
@@ -458,7 +458,8 @@ struct radeon_mman {
 
 struct radeon_bo_list {
 	struct radeon_bo		*robj;
-	struct ttm_validate_buffer	tv;
+	struct list_head		list;
+	bool				shared;
 	uint64_t			gpu_offset;
 	unsigned			preferred_domains;
 	unsigned			allowed_domains;
@@ -1031,6 +1032,7 @@ struct radeon_cs_parser {
 	struct radeon_bo_list	*vm_bos;
 	struct list_head	validated;
 	unsigned		dma_reloc_idx;
+	struct drm_exec		exec;
 	/* indices of various chunks */
 	struct radeon_cs_chunk  *chunk_ib;
 	struct radeon_cs_chunk  *chunk_relocs;
@@ -1044,7 +1046,6 @@ struct radeon_cs_parser {
 	u32			cs_flags;
 	u32			ring;
 	s32			priority;
-	struct ww_acquire_ctx	ticket;
 };
 
 static inline u32 radeon_get_ib_value(struct radeon_cs_parser *p, int idx)
diff --git a/drivers/gpu/drm/radeon/radeon_cs.c b/drivers/gpu/drm/radeon/radeon_cs.c
index a6700d7278bf..ea0a1ed56bd8 100644
--- a/drivers/gpu/drm/radeon/radeon_cs.c
+++ b/drivers/gpu/drm/radeon/radeon_cs.c
@@ -182,11 +182,8 @@ static int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
 			}
 		}
 
-		p->relocs[i].tv.bo = &p->relocs[i].robj->tbo;
-		p->relocs[i].tv.num_shared = !r->write_domain;
-
-		radeon_cs_buckets_add(&buckets, &p->relocs[i].tv.head,
-				      priority);
+		p->relocs[i].shared = !r->write_domain;
+		radeon_cs_buckets_add(&buckets, &p->relocs[i].list, priority);
 	}
 
 	radeon_cs_buckets_get_list(&buckets, &p->validated);
@@ -197,7 +194,7 @@ static int radeon_cs_parser_relocs(struct radeon_cs_parser *p)
 	if (need_mmap_lock)
 		mmap_read_lock(current->mm);
 
-	r = radeon_bo_list_validate(p->rdev, &p->ticket, &p->validated, p->ring);
+	r = radeon_bo_list_validate(p->rdev, &p->exec, &p->validated, p->ring);
 
 	if (need_mmap_lock)
 		mmap_read_unlock(current->mm);
@@ -253,12 +250,11 @@ static int radeon_cs_sync_rings(struct radeon_cs_parser *p)
 	struct radeon_bo_list *reloc;
 	int r;
 
-	list_for_each_entry(reloc, &p->validated, tv.head) {
+	list_for_each_entry(reloc, &p->validated, list) {
 		struct dma_resv *resv;
 
 		resv = reloc->robj->tbo.base.resv;
-		r = radeon_sync_resv(p->rdev, &p->ib.sync, resv,
-				     reloc->tv.num_shared);
+		r = radeon_sync_resv(p->rdev, &p->ib.sync, resv, reloc->shared);
 		if (r)
 			return r;
 	}
@@ -276,6 +272,7 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
 	s32 priority = 0;
 
 	INIT_LIST_HEAD(&p->validated);
+	drm_exec_init(&p->exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
 
 	if (!cs->num_chunks) {
 		return 0;
@@ -397,8 +394,8 @@ int radeon_cs_parser_init(struct radeon_cs_parser *p, void *data)
 static int cmp_size_smaller_first(void *priv, const struct list_head *a,
 				  const struct list_head *b)
 {
-	struct radeon_bo_list *la = list_entry(a, struct radeon_bo_list, tv.head);
-	struct radeon_bo_list *lb = list_entry(b, struct radeon_bo_list, tv.head);
+	struct radeon_bo_list *la = list_entry(a, struct radeon_bo_list, list);
+	struct radeon_bo_list *lb = list_entry(b, struct radeon_bo_list, list);
 
 	/* Sort A before B if A is smaller. */
 	if (la->robj->tbo.base.size > lb->robj->tbo.base.size)
@@ -417,11 +414,13 @@ static int cmp_size_smaller_first(void *priv, const struct list_head *a,
  * If error is set than unvalidate buffer, otherwise just free memory
  * used by parsing context.
  **/
-static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error, bool backoff)
+static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error)
 {
 	unsigned i;
 
 	if (!error) {
+		struct radeon_bo_list *reloc;
+
 		/* Sort the buffer list from the smallest to largest buffer,
 		 * which affects the order of buffers in the LRU list.
 		 * This assures that the smallest buffers are added first
@@ -433,15 +432,17 @@ static void radeon_cs_parser_fini(struct radeon_cs_parser *parser, int error, bo
 		 * per frame under memory pressure.
 		 */
 		list_sort(NULL, &parser->validated, cmp_size_smaller_first);
-
-		ttm_eu_fence_buffer_objects(&parser->ticket,
-					    &parser->validated,
-					    &parser->ib.fence->base);
-	} else if (backoff) {
-		ttm_eu_backoff_reservation(&parser->ticket,
-					   &parser->validated);
+		list_for_each_entry(reloc, &parser->validated, list) {
+			dma_resv_add_fence(reloc->robj->tbo.base.resv,
+					   &parser->ib.fence->base,
+					   reloc->shared ?
+					   DMA_RESV_USAGE_READ :
+					   DMA_RESV_USAGE_WRITE);
+		}
 	}
 
+	drm_exec_fini(&parser->exec);
+
 	if (parser->relocs != NULL) {
 		for (i = 0; i < parser->nrelocs; i++) {
 			struct radeon_bo *bo = parser->relocs[i].robj;
@@ -693,7 +694,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	r = radeon_cs_parser_init(&parser, data);
 	if (r) {
 		DRM_ERROR("Failed to initialize parser !\n");
-		radeon_cs_parser_fini(&parser, r, false);
+		radeon_cs_parser_fini(&parser, r);
 		up_read(&rdev->exclusive_lock);
 		r = radeon_cs_handle_lockup(rdev, r);
 		return r;
@@ -707,7 +708,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 	}
 
 	if (r) {
-		radeon_cs_parser_fini(&parser, r, false);
+		radeon_cs_parser_fini(&parser, r);
 		up_read(&rdev->exclusive_lock);
 		r = radeon_cs_handle_lockup(rdev, r);
 		return r;
@@ -724,7 +725,7 @@ int radeon_cs_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)
 		goto out;
 	}
 out:
-	radeon_cs_parser_fini(&parser, r, true);
+	radeon_cs_parser_fini(&parser, r);
 	up_read(&rdev->exclusive_lock);
 	r = radeon_cs_handle_lockup(rdev, r);
 	return r;
diff --git a/drivers/gpu/drm/radeon/radeon_gem.c b/drivers/gpu/drm/radeon/radeon_gem.c
index d0119c5f7eb3..3af518895603 100644
--- a/drivers/gpu/drm/radeon/radeon_gem.c
+++ b/drivers/gpu/drm/radeon/radeon_gem.c
@@ -623,33 +623,40 @@ int radeon_gem_get_tiling_ioctl(struct drm_device *dev, void *data,
 static void radeon_gem_va_update_vm(struct radeon_device *rdev,
 				    struct radeon_bo_va *bo_va)
 {
-	struct ttm_validate_buffer tv, *entry;
-	struct radeon_bo_list *vm_bos;
-	struct ww_acquire_ctx ticket;
+	struct radeon_bo_list *vm_bos, *entry;
 	struct list_head list;
+	struct drm_exec exec;
 	unsigned domain;
 	int r;
 
 	INIT_LIST_HEAD(&list);
 
-	tv.bo = &bo_va->bo->tbo;
-	tv.num_shared = 1;
-	list_add(&tv.head, &list);
-
 	vm_bos = radeon_vm_get_bos(rdev, bo_va->vm, &list);
 	if (!vm_bos)
 		return;
 
-	r = ttm_eu_reserve_buffers(&ticket, &list, true, NULL);
-	if (r)
-		goto error_free;
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
+	drm_exec_until_all_locked(&exec) {
+		list_for_each_entry(entry, &list, list) {
+			r = drm_exec_prepare_obj(&exec, &entry->robj->tbo.base,
+						 1);
+			drm_exec_retry_on_contention(&exec);
+			if (unlikely(r))
+				goto error_cleanup;
+		}
 
-	list_for_each_entry(entry, &list, head) {
-		domain = radeon_mem_type_to_domain(entry->bo->resource->mem_type);
+		r = drm_exec_prepare_obj(&exec, &bo_va->bo->tbo.base, 1);
+		drm_exec_retry_on_contention(&exec);
+		if (unlikely(r))
+			goto error_cleanup;
+	}
+
+	list_for_each_entry(entry, &list, list) {
+		domain = radeon_mem_type_to_domain(entry->robj->tbo.resource->mem_type);
 		/* if anything is swapped out don't swap it in here,
 		   just abort and wait for the next CS */
 		if (domain == RADEON_GEM_DOMAIN_CPU)
-			goto error_unreserve;
+			goto error_cleanup;
 	}
 
 	mutex_lock(&bo_va->vm->mutex);
@@ -663,10 +670,8 @@ static void radeon_gem_va_update_vm(struct radeon_device *rdev,
 error_unlock:
 	mutex_unlock(&bo_va->vm->mutex);
 
-error_unreserve:
-	ttm_eu_backoff_reservation(&ticket, &list);
-
-error_free:
+error_cleanup:
+	drm_exec_fini(&exec);
 	kvfree(vm_bos);
 
 	if (r && r != -ERESTARTSYS)
diff --git a/drivers/gpu/drm/radeon/radeon_object.c b/drivers/gpu/drm/radeon/radeon_object.c
index 10c0fbd9d2b4..e47f04c001b8 100644
--- a/drivers/gpu/drm/radeon/radeon_object.c
+++ b/drivers/gpu/drm/radeon/radeon_object.c
@@ -468,23 +468,26 @@ static u64 radeon_bo_get_threshold_for_moves(struct radeon_device *rdev)
 }
 
 int radeon_bo_list_validate(struct radeon_device *rdev,
-			    struct ww_acquire_ctx *ticket,
+			    struct drm_exec *exec,
 			    struct list_head *head, int ring)
 {
 	struct ttm_operation_ctx ctx = { true, false };
 	struct radeon_bo_list *lobj;
-	struct list_head duplicates;
-	int r;
 	u64 bytes_moved = 0, initial_bytes_moved;
 	u64 bytes_moved_threshold = radeon_bo_get_threshold_for_moves(rdev);
+	int r;
 
-	INIT_LIST_HEAD(&duplicates);
-	r = ttm_eu_reserve_buffers(ticket, head, true, &duplicates);
-	if (unlikely(r != 0)) {
-		return r;
+	drm_exec_until_all_locked(exec) {
+		list_for_each_entry(lobj, head, list) {
+			r = drm_exec_prepare_obj(exec, &lobj->robj->tbo.base,
+						 1);
+			drm_exec_retry_on_contention(exec);
+			if (unlikely(r && r != -EALREADY))
+				return r;
+		}
 	}
 
-	list_for_each_entry(lobj, head, tv.head) {
+	list_for_each_entry(lobj, head, list) {
 		struct radeon_bo *bo = lobj->robj;
 		if (!bo->tbo.pin_count) {
 			u32 domain = lobj->preferred_domains;
@@ -523,7 +526,6 @@ int radeon_bo_list_validate(struct radeon_device *rdev,
 					domain = lobj->allowed_domains;
 					goto retry;
 				}
-				ttm_eu_backoff_reservation(ticket, head);
 				return r;
 			}
 		}
@@ -531,11 +533,6 @@ int radeon_bo_list_validate(struct radeon_device *rdev,
 		lobj->tiling_flags = bo->tiling_flags;
 	}
 
-	list_for_each_entry(lobj, &duplicates, tv.head) {
-		lobj->gpu_offset = radeon_bo_gpu_offset(lobj->robj);
-		lobj->tiling_flags = lobj->robj->tiling_flags;
-	}
-
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/radeon/radeon_object.h b/drivers/gpu/drm/radeon/radeon_object.h
index 39cc87a59a9a..d7bbb52db546 100644
--- a/drivers/gpu/drm/radeon/radeon_object.h
+++ b/drivers/gpu/drm/radeon/radeon_object.h
@@ -152,7 +152,7 @@ extern void radeon_bo_force_delete(struct radeon_device *rdev);
 extern int radeon_bo_init(struct radeon_device *rdev);
 extern void radeon_bo_fini(struct radeon_device *rdev);
 extern int radeon_bo_list_validate(struct radeon_device *rdev,
-				   struct ww_acquire_ctx *ticket,
+				   struct drm_exec *exec,
 				   struct list_head *head, int ring);
 extern int radeon_bo_set_tiling_flags(struct radeon_bo *bo,
 				u32 tiling_flags, u32 pitch);
diff --git a/drivers/gpu/drm/radeon/radeon_vm.c b/drivers/gpu/drm/radeon/radeon_vm.c
index 987cabbf1318..647c4a07d92a 100644
--- a/drivers/gpu/drm/radeon/radeon_vm.c
+++ b/drivers/gpu/drm/radeon/radeon_vm.c
@@ -142,10 +142,9 @@ struct radeon_bo_list *radeon_vm_get_bos(struct radeon_device *rdev,
 	list[0].robj = vm->page_directory;
 	list[0].preferred_domains = RADEON_GEM_DOMAIN_VRAM;
 	list[0].allowed_domains = RADEON_GEM_DOMAIN_VRAM;
-	list[0].tv.bo = &vm->page_directory->tbo;
-	list[0].tv.num_shared = 1;
+	list[0].shared = true;
 	list[0].tiling_flags = 0;
-	list_add(&list[0].tv.head, head);
+	list_add(&list[0].list, head);
 
 	for (i = 0, idx = 1; i <= vm->max_pde_used; i++) {
 		if (!vm->page_tables[i].bo)
@@ -154,10 +153,9 @@ struct radeon_bo_list *radeon_vm_get_bos(struct radeon_device *rdev,
 		list[idx].robj = vm->page_tables[i].bo;
 		list[idx].preferred_domains = RADEON_GEM_DOMAIN_VRAM;
 		list[idx].allowed_domains = RADEON_GEM_DOMAIN_VRAM;
-		list[idx].tv.bo = &list[idx].robj->tbo;
-		list[idx].tv.num_shared = 1;
+		list[idx].shared = true;
 		list[idx].tiling_flags = 0;
-		list_add(&list[idx++].tv.head, head);
+		list_add(&list[idx++].list, head);
 	}
 
 	return list;
-- 
2.34.1

[PATCH 2/7] drm/qxl: switch to using drm_exec

[Christian König]

Just a straightforward conversion without any optimization.

Only compile tested for now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: virtualization@lists.linux-foundation.org
Cc: spice-devel@lists.freedesktop.org
---
 drivers/gpu/drm/qxl/Kconfig       |  1 +
 drivers/gpu/drm/qxl/qxl_drv.h     |  7 ++--
 drivers/gpu/drm/qxl/qxl_release.c | 68 ++++++++++++++++---------------
 3 files changed, 40 insertions(+), 36 deletions(-)

diff --git a/drivers/gpu/drm/qxl/Kconfig b/drivers/gpu/drm/qxl/Kconfig
index ca3f51c2a8fe..9c8e433be33e 100644
--- a/drivers/gpu/drm/qxl/Kconfig
+++ b/drivers/gpu/drm/qxl/Kconfig
@@ -5,6 +5,7 @@ config DRM_QXL
 	select DRM_KMS_HELPER
 	select DRM_TTM
 	select DRM_TTM_HELPER
+	select DRM_EXEC
 	select CRC32
 	help
 	  QXL virtual GPU for Spice virtualization desktop integration.
diff --git a/drivers/gpu/drm/qxl/qxl_drv.h b/drivers/gpu/drm/qxl/qxl_drv.h
index ea993d7162e8..3e732648b332 100644
--- a/drivers/gpu/drm/qxl/qxl_drv.h
+++ b/drivers/gpu/drm/qxl/qxl_drv.h
@@ -38,12 +38,12 @@
 
 #include <drm/drm_crtc.h>
 #include <drm/drm_encoder.h>
+#include <drm/drm_exec.h>
 #include <drm/drm_gem_ttm_helper.h>
 #include <drm/drm_ioctl.h>
 #include <drm/drm_gem.h>
 #include <drm/qxl_drm.h>
 #include <drm/ttm/ttm_bo.h>
-#include <drm/ttm/ttm_execbuf_util.h>
 #include <drm/ttm/ttm_placement.h>
 
 #include "qxl_dev.h"
@@ -101,7 +101,8 @@ struct qxl_gem {
 };
 
 struct qxl_bo_list {
-	struct ttm_validate_buffer tv;
+	struct qxl_bo		*bo;
+	struct list_head	list;
 };
 
 struct qxl_crtc {
@@ -151,7 +152,7 @@ struct qxl_release {
 	struct qxl_bo *release_bo;
 	uint32_t release_offset;
 	uint32_t surface_release_id;
-	struct ww_acquire_ctx ticket;
+	struct drm_exec	exec;
 	struct list_head bos;
 };
 
diff --git a/drivers/gpu/drm/qxl/qxl_release.c b/drivers/gpu/drm/qxl/qxl_release.c
index 368d26da0d6a..5953e8ed5f47 100644
--- a/drivers/gpu/drm/qxl/qxl_release.c
+++ b/drivers/gpu/drm/qxl/qxl_release.c
@@ -121,13 +121,11 @@ qxl_release_free_list(struct qxl_release *release)
 {
 	while (!list_empty(&release->bos)) {
 		struct qxl_bo_list *entry;
-		struct qxl_bo *bo;
 
 		entry = container_of(release->bos.next,
-				     struct qxl_bo_list, tv.head);
-		bo = to_qxl_bo(entry->tv.bo);
-		qxl_bo_unref(&bo);
-		list_del(&entry->tv.head);
+				     struct qxl_bo_list, list);
+		qxl_bo_unref(&entry->bo);
+		list_del(&entry->list);
 		kfree(entry);
 	}
 	release->release_bo = NULL;
@@ -172,8 +170,8 @@ int qxl_release_list_add(struct qxl_release *release, struct qxl_bo *bo)
 {
 	struct qxl_bo_list *entry;
 
-	list_for_each_entry(entry, &release->bos, tv.head) {
-		if (entry->tv.bo == &bo->tbo)
+	list_for_each_entry(entry, &release->bos, list) {
+		if (entry->bo == bo)
 			return 0;
 	}
 
@@ -182,9 +180,8 @@ int qxl_release_list_add(struct qxl_release *release, struct qxl_bo *bo)
 		return -ENOMEM;
 
 	qxl_bo_ref(bo);
-	entry->tv.bo = &bo->tbo;
-	entry->tv.num_shared = 0;
-	list_add_tail(&entry->tv.head, &release->bos);
+	entry->bo = bo;
+	list_add_tail(&entry->list, &release->bos);
 	return 0;
 }
 
@@ -221,21 +218,28 @@ int qxl_release_reserve_list(struct qxl_release *release, bool no_intr)
 	if (list_is_singular(&release->bos))
 		return 0;
 
-	ret = ttm_eu_reserve_buffers(&release->ticket, &release->bos,
-				     !no_intr, NULL);
-	if (ret)
-		return ret;
-
-	list_for_each_entry(entry, &release->bos, tv.head) {
-		struct qxl_bo *bo = to_qxl_bo(entry->tv.bo);
-
-		ret = qxl_release_validate_bo(bo);
-		if (ret) {
-			ttm_eu_backoff_reservation(&release->ticket, &release->bos);
-			return ret;
+	drm_exec_init(&release->exec, no_intr ? 0 :
+		      DRM_EXEC_INTERRUPTIBLE_WAIT);
+	drm_exec_until_all_locked(&release->exec) {
+		list_for_each_entry(entry, &release->bos, list) {
+			ret = drm_exec_prepare_obj(&release->exec,
+						   &entry->bo->tbo.base,
+						   1);
+			drm_exec_retry_on_contention(&release->exec);
+			if (ret)
+				goto error;
 		}
 	}
+
+	list_for_each_entry(entry, &release->bos, list) {
+		ret = qxl_release_validate_bo(entry->bo);
+		if (ret)
+			goto error;
+	}
 	return 0;
+error:
+	drm_exec_fini(&release->exec);
+	return ret;
 }
 
 void qxl_release_backoff_reserve_list(struct qxl_release *release)
@@ -245,7 +249,7 @@ void qxl_release_backoff_reserve_list(struct qxl_release *release)
 	if (list_is_singular(&release->bos))
 		return;
 
-	ttm_eu_backoff_reservation(&release->ticket, &release->bos);
+	drm_exec_fini(&release->exec);
 }
 
 int qxl_alloc_surface_release_reserved(struct qxl_device *qdev,
@@ -404,18 +408,18 @@ void qxl_release_unmap(struct qxl_device *qdev,
 
 void qxl_release_fence_buffer_objects(struct qxl_release *release)
 {
-	struct ttm_buffer_object *bo;
 	struct ttm_device *bdev;
-	struct ttm_validate_buffer *entry;
+	struct qxl_bo_list *entry;
 	struct qxl_device *qdev;
+	struct qxl_bo *bo;
 
 	/* if only one object on the release its the release itself
 	   since these objects are pinned no need to reserve */
 	if (list_is_singular(&release->bos) || list_empty(&release->bos))
 		return;
 
-	bo = list_first_entry(&release->bos, struct ttm_validate_buffer, head)->bo;
-	bdev = bo->bdev;
+	bo = list_first_entry(&release->bos, struct qxl_bo_list, list)->bo;
+	bdev = bo->tbo.bdev;
 	qdev = container_of(bdev, struct qxl_device, mman.bdev);
 
 	/*
@@ -426,14 +430,12 @@ void qxl_release_fence_buffer_objects(struct qxl_release *release)
 		       release->id | 0xf0000000, release->base.seqno);
 	trace_dma_fence_emit(&release->base);
 
-	list_for_each_entry(entry, &release->bos, head) {
+	list_for_each_entry(entry, &release->bos, list) {
 		bo = entry->bo;
 
-		dma_resv_add_fence(bo->base.resv, &release->base,
+		dma_resv_add_fence(bo->tbo.base.resv, &release->base,
 				   DMA_RESV_USAGE_READ);
-		ttm_bo_move_to_lru_tail_unlocked(bo);
-		dma_resv_unlock(bo->base.resv);
+		ttm_bo_move_to_lru_tail_unlocked(&bo->tbo);
 	}
-	ww_acquire_fini(&release->ticket);
+	drm_exec_fini(&release->exec);
 }
-
-- 
2.34.1

[PATCH 3/7] drm/lima: switch to using drm_exec

[Christian König]

Just a straightforward conversion without any optimization.

Only compile tested for now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Cc: Qiang Yu <yuq825@gmail.com>
Cc: lima@lists.freedesktop.org
---
 drivers/gpu/drm/lima/Kconfig    |  1 +
 drivers/gpu/drm/lima/lima_gem.c | 17 +++++++++--------
 2 files changed, 10 insertions(+), 8 deletions(-)

diff --git a/drivers/gpu/drm/lima/Kconfig b/drivers/gpu/drm/lima/Kconfig
index fa1d4f5df31e..1d2871d9ddd2 100644
--- a/drivers/gpu/drm/lima/Kconfig
+++ b/drivers/gpu/drm/lima/Kconfig
@@ -9,6 +9,7 @@ config DRM_LIMA
        depends on COMMON_CLK
        depends on OF
        select DRM_SCHED
+       select DRM_EXEC
        select DRM_GEM_SHMEM_HELPER
        select PM_DEVFREQ
        select DEVFREQ_GOV_SIMPLE_ONDEMAND
diff --git a/drivers/gpu/drm/lima/lima_gem.c b/drivers/gpu/drm/lima/lima_gem.c
index 4f9736e5f929..d4688d19465b 100644
--- a/drivers/gpu/drm/lima/lima_gem.c
+++ b/drivers/gpu/drm/lima/lima_gem.c
@@ -8,6 +8,7 @@
 #include <linux/shmem_fs.h>
 #include <linux/dma-mapping.h>
 
+#include <drm/drm_exec.h>
 #include <drm/drm_file.h>
 #include <drm/drm_syncobj.h>
 #include <drm/drm_utils.h>
@@ -292,7 +293,7 @@ static int lima_gem_add_deps(struct drm_file *file, struct lima_submit *submit)
 int lima_gem_submit(struct drm_file *file, struct lima_submit *submit)
 {
 	int i, err = 0;
-	struct ww_acquire_ctx ctx;
+	struct drm_exec exec;
 	struct lima_drm_priv *priv = to_lima_drm_priv(file);
 	struct lima_vm *vm = priv->vm;
 	struct drm_syncobj *out_sync = NULL;
@@ -329,8 +330,11 @@ int lima_gem_submit(struct drm_file *file, struct lima_submit *submit)
 		bos[i] = bo;
 	}
 
-	err = drm_gem_lock_reservations((struct drm_gem_object **)bos,
-					submit->nr_bos, &ctx);
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
+	drm_exec_until_all_locked(&exec)
+		err = drm_exec_prepare_array(&exec,
+					     (struct drm_gem_object **)bos,
+					     submit->nr_bos, 0);
 	if (err)
 		goto err_out0;
 
@@ -360,9 +364,7 @@ int lima_gem_submit(struct drm_file *file, struct lima_submit *submit)
 				   submit->bos[i].flags & LIMA_SUBMIT_BO_WRITE ?
 				   DMA_RESV_USAGE_WRITE : DMA_RESV_USAGE_READ);
 	}
-
-	drm_gem_unlock_reservations((struct drm_gem_object **)bos,
-				    submit->nr_bos, &ctx);
+	drm_exec_fini(&exec);
 
 	for (i = 0; i < submit->nr_bos; i++)
 		drm_gem_object_put(&bos[i]->base.base);
@@ -379,8 +381,7 @@ int lima_gem_submit(struct drm_file *file, struct lima_submit *submit)
 err_out2:
 	lima_sched_task_fini(submit->task);
 err_out1:
-	drm_gem_unlock_reservations((struct drm_gem_object **)bos,
-				    submit->nr_bos, &ctx);
+	drm_exec_fini(&exec);
 err_out0:
 	for (i = 0; i < submit->nr_bos; i++) {
 		if (!bos[i])
-- 
2.34.1

[PATCH 4/7] drm/virtgpu: switch to using drm_exec

[Christian König]

Just a straightforward conversion without any optimization.

Only compile tested for now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Cc: David Airlie <airlied@redhat.com>
Cc: Gerd Hoffmann <kraxel@redhat.com>
Cc: Gurchetan Singh <gurchetansingh@chromium.org>
Cc: Chia-I Wu <olvaffe@gmail.com>
Cc: virtualization@lists.linux-foundation.org
---
 drivers/gpu/drm/virtio/Kconfig       |  1 +
 drivers/gpu/drm/virtio/virtgpu_drv.h |  3 ++-
 drivers/gpu/drm/virtio/virtgpu_gem.c | 28 +++++-----------------------
 3 files changed, 8 insertions(+), 24 deletions(-)

diff --git a/drivers/gpu/drm/virtio/Kconfig b/drivers/gpu/drm/virtio/Kconfig
index ea06ff2aa4b4..a24a1ce5e666 100644
--- a/drivers/gpu/drm/virtio/Kconfig
+++ b/drivers/gpu/drm/virtio/Kconfig
@@ -5,6 +5,7 @@ config DRM_VIRTIO_GPU
 	select VIRTIO
 	select DRM_KMS_HELPER
 	select DRM_GEM_SHMEM_HELPER
+	select DRM_EXEC
 	select VIRTIO_DMA_SHARED_BUFFER
 	help
 	   This is the virtual GPU driver for virtio.  It can be used with
diff --git a/drivers/gpu/drm/virtio/virtgpu_drv.h b/drivers/gpu/drm/virtio/virtgpu_drv.h
index 4126c384286b..cb5d9218a3ab 100644
--- a/drivers/gpu/drm/virtio/virtgpu_drv.h
+++ b/drivers/gpu/drm/virtio/virtgpu_drv.h
@@ -35,6 +35,7 @@
 #include <drm/drm_atomic.h>
 #include <drm/drm_drv.h>
 #include <drm/drm_encoder.h>
+#include <drm/drm_exec.h>
 #include <drm/drm_fourcc.h>
 #include <drm/drm_framebuffer.h>
 #include <drm/drm_gem.h>
@@ -116,7 +117,7 @@ struct virtio_gpu_object_vram {
 	container_of((virtio_gpu_object), struct virtio_gpu_object_vram, base)
 
 struct virtio_gpu_object_array {
-	struct ww_acquire_ctx ticket;
+	struct drm_exec exec;
 	struct list_head next;
 	u32 nents, total;
 	struct drm_gem_object *objs[];
diff --git a/drivers/gpu/drm/virtio/virtgpu_gem.c b/drivers/gpu/drm/virtio/virtgpu_gem.c
index 7db48d17ee3a..bb7928e7b3b6 100644
--- a/drivers/gpu/drm/virtio/virtgpu_gem.c
+++ b/drivers/gpu/drm/virtio/virtgpu_gem.c
@@ -171,6 +171,7 @@ struct virtio_gpu_object_array *virtio_gpu_array_alloc(u32 nents)
 
 	objs->nents = 0;
 	objs->total = nents;
+	drm_exec_init(&objs->exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
 	return objs;
 }
 
@@ -214,36 +215,17 @@ void virtio_gpu_array_add_obj(struct virtio_gpu_object_array *objs,
 
 int virtio_gpu_array_lock_resv(struct virtio_gpu_object_array *objs)
 {
-	unsigned int i;
 	int ret;
 
-	if (objs->nents == 1) {
-		ret = dma_resv_lock_interruptible(objs->objs[0]->resv, NULL);
-	} else {
-		ret = drm_gem_lock_reservations(objs->objs, objs->nents,
-						&objs->ticket);
-	}
-	if (ret)
-		return ret;
-
-	for (i = 0; i < objs->nents; ++i) {
-		ret = dma_resv_reserve_fences(objs->objs[i]->resv, 1);
-		if (ret) {
-			virtio_gpu_array_unlock_resv(objs);
-			return ret;
-		}
-	}
+	drm_exec_until_all_locked(&objs->exec)
+		ret = drm_exec_prepare_array(&objs->exec, objs->objs,
+					     objs->nents, 1);
 	return ret;
 }
 
 void virtio_gpu_array_unlock_resv(struct virtio_gpu_object_array *objs)
 {
-	if (objs->nents == 1) {
-		dma_resv_unlock(objs->objs[0]->resv);
-	} else {
-		drm_gem_unlock_reservations(objs->objs, objs->nents,
-					    &objs->ticket);
-	}
+	drm_exec_fini(&objs->exec);
 }
 
 void virtio_gpu_array_add_fence(struct virtio_gpu_object_array *objs,
-- 
2.34.1

[PATCH 5/7] drm/panfrost: switch to using drm_exec

[Christian König]

Just a straightforward conversion without any optimization.

Only compile tested for now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Cc: Rob Herring <robh@kernel.org>
Cc: Tomeu Vizoso <tomeu.vizoso@collabora.com>
Cc: Steven Price <steven.price@arm.com>
Cc: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
---
 drivers/gpu/drm/panfrost/Kconfig        |  1 +
 drivers/gpu/drm/panfrost/panfrost_job.c | 12 +++++++-----
 2 files changed, 8 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/panfrost/Kconfig b/drivers/gpu/drm/panfrost/Kconfig
index e6403a9d66ad..e86a1a2fd8e1 100644
--- a/drivers/gpu/drm/panfrost/Kconfig
+++ b/drivers/gpu/drm/panfrost/Kconfig
@@ -7,6 +7,7 @@ config DRM_PANFROST
 	depends on !GENERIC_ATOMIC64    # for IOMMU_IO_PGTABLE_LPAE
 	depends on MMU
 	select DRM_SCHED
+	select DRM_EXEC
 	select IOMMU_SUPPORT
 	select IOMMU_IO_PGTABLE_LPAE
 	select DRM_GEM_SHMEM_HELPER
diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
index dbc597ab46fb..8b9206e910b5 100644
--- a/drivers/gpu/drm/panfrost/panfrost_job.c
+++ b/drivers/gpu/drm/panfrost/panfrost_job.c
@@ -8,6 +8,7 @@
 #include <linux/platform_device.h>
 #include <linux/pm_runtime.h>
 #include <linux/dma-resv.h>
+#include <drm/drm_exec.h>
 #include <drm/gpu_scheduler.h>
 #include <drm/panfrost_drm.h>
 
@@ -275,13 +276,14 @@ static void panfrost_attach_object_fences(struct drm_gem_object **bos,
 int panfrost_job_push(struct panfrost_job *job)
 {
 	struct panfrost_device *pfdev = job->pfdev;
-	struct ww_acquire_ctx acquire_ctx;
+	struct drm_exec exec;
 	int ret = 0;
 
-	ret = drm_gem_lock_reservations(job->bos, job->bo_count,
-					    &acquire_ctx);
+	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
+	drm_exec_until_all_locked(&exec)
+		ret = drm_exec_prepare_array(&exec, job->bos, job->bo_count, 1);
 	if (ret)
-		return ret;
+		goto unlock;
 
 	mutex_lock(&pfdev->sched_lock);
 	drm_sched_job_arm(&job->base);
@@ -305,7 +307,7 @@ int panfrost_job_push(struct panfrost_job *job)
 				      job->render_done_fence);
 
 unlock:
-	drm_gem_unlock_reservations(job->bos, job->bo_count, &acquire_ctx);
+	drm_exec_fini(&exec);
 
 	return ret;
 }
-- 
2.34.1

[PATCH 6/7] drm/v3d: switch to using drm_exec

[Christian König]

Just a straightforward conversion without any optimization.

Only compile tested for now.

Signed-off-by: Christian König <christian.koenig@amd.com>
Cc: Emma Anholt <emma@anholt.net>
Cc: Melissa Wen <mwen@igalia.com>
---
 drivers/gpu/drm/v3d/v3d_gem.c | 44 ++++++++++++++++-------------------
 1 file changed, 20 insertions(+), 24 deletions(-)

diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
index 2e94ce788c71..190e2a9f64a4 100644
--- a/drivers/gpu/drm/v3d/v3d_gem.c
+++ b/drivers/gpu/drm/v3d/v3d_gem.c
@@ -10,6 +10,7 @@
 #include <linux/sched/signal.h>
 #include <linux/uaccess.h>
 
+#include <drm/drm_exec.h>
 #include <drm/drm_managed.h>
 #include <drm/drm_syncobj.h>
 #include <uapi/drm/v3d_drm.h>
@@ -249,20 +250,17 @@ v3d_invalidate_caches(struct v3d_dev *v3d)
  * to v3d, so we don't attach dma-buf fences to them.
  */
 static int
-v3d_lock_bo_reservations(struct v3d_job *job,
-			 struct ww_acquire_ctx *acquire_ctx)
+v3d_lock_bo_reservations(struct v3d_job *job, struct drm_exec *exec)
 {
 	int i, ret;
 
-	ret = drm_gem_lock_reservations(job->bo, job->bo_count, acquire_ctx);
+	drm_exec_init(exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
+	drm_exec_until_all_locked(exec)
+		ret = drm_exec_prepare_array(exec, job->bo, job->bo_count, 1);
 	if (ret)
-		return ret;
+		goto fail;
 
 	for (i = 0; i < job->bo_count; i++) {
-		ret = dma_resv_reserve_fences(job->bo[i]->resv, 1);
-		if (ret)
-			goto fail;
-
 		ret = drm_sched_job_add_implicit_dependencies(&job->base,
 							      job->bo[i], true);
 		if (ret)
@@ -272,7 +270,7 @@ v3d_lock_bo_reservations(struct v3d_job *job,
 	return 0;
 
 fail:
-	drm_gem_unlock_reservations(job->bo, job->bo_count, acquire_ctx);
+	drm_exec_fini(exec);
 	return ret;
 }
 
@@ -477,7 +475,7 @@ v3d_push_job(struct v3d_job *job)
 static void
 v3d_attach_fences_and_unlock_reservation(struct drm_file *file_priv,
 					 struct v3d_job *job,
-					 struct ww_acquire_ctx *acquire_ctx,
+					 struct drm_exec *exec,
 					 u32 out_sync,
 					 struct v3d_submit_ext *se,
 					 struct dma_fence *done_fence)
@@ -492,7 +490,7 @@ v3d_attach_fences_and_unlock_reservation(struct drm_file *file_priv,
 				   DMA_RESV_USAGE_WRITE);
 	}
 
-	drm_gem_unlock_reservations(job->bo, job->bo_count, acquire_ctx);
+	drm_exec_fini(exec);
 
 	/* Update the return sync object for the job */
 	/* If it only supports a single signal semaphore*/
@@ -669,7 +667,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
 	struct v3d_render_job *render = NULL;
 	struct v3d_job *clean_job = NULL;
 	struct v3d_job *last_job;
-	struct ww_acquire_ctx acquire_ctx;
+	struct drm_exec exec;
 	int ret = 0;
 
 	trace_v3d_submit_cl_ioctl(&v3d->drm, args->rcl_start, args->rcl_end);
@@ -731,7 +729,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
 	if (ret)
 		goto fail;
 
-	ret = v3d_lock_bo_reservations(last_job, &acquire_ctx);
+	ret = v3d_lock_bo_reservations(last_job, &exec);
 	if (ret)
 		goto fail;
 
@@ -775,7 +773,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
 
 	v3d_attach_fences_and_unlock_reservation(file_priv,
 						 last_job,
-						 &acquire_ctx,
+						 &exec,
 						 args->out_sync,
 						 &se,
 						 last_job->done_fence);
@@ -791,8 +789,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
 fail_unreserve:
 	mutex_unlock(&v3d->sched_lock);
 fail_perfmon:
-	drm_gem_unlock_reservations(last_job->bo,
-				    last_job->bo_count, &acquire_ctx);
+	drm_exec_fini(&exec);
 fail:
 	v3d_job_cleanup((void *)bin);
 	v3d_job_cleanup((void *)render);
@@ -819,7 +816,7 @@ v3d_submit_tfu_ioctl(struct drm_device *dev, void *data,
 	struct drm_v3d_submit_tfu *args = data;
 	struct v3d_submit_ext se = {0};
 	struct v3d_tfu_job *job = NULL;
-	struct ww_acquire_ctx acquire_ctx;
+	struct drm_exec exec;
 	int ret = 0;
 
 	trace_v3d_submit_tfu_ioctl(&v3d->drm, args->iia);
@@ -870,7 +867,7 @@ v3d_submit_tfu_ioctl(struct drm_device *dev, void *data,
 		job->base.bo[job->base.bo_count] = bo;
 	}
 
-	ret = v3d_lock_bo_reservations(&job->base, &acquire_ctx);
+	ret = v3d_lock_bo_reservations(&job->base, &exec);
 	if (ret)
 		goto fail;
 
@@ -879,7 +876,7 @@ v3d_submit_tfu_ioctl(struct drm_device *dev, void *data,
 	mutex_unlock(&v3d->sched_lock);
 
 	v3d_attach_fences_and_unlock_reservation(file_priv,
-						 &job->base, &acquire_ctx,
+						 &job->base, &exec,
 						 args->out_sync,
 						 &se,
 						 job->base.done_fence);
@@ -914,7 +911,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
 	struct v3d_submit_ext se = {0};
 	struct v3d_csd_job *job = NULL;
 	struct v3d_job *clean_job = NULL;
-	struct ww_acquire_ctx acquire_ctx;
+	struct drm_exec exec;
 	int ret;
 
 	trace_v3d_submit_csd_ioctl(&v3d->drm, args->cfg[5], args->cfg[6]);
@@ -957,7 +954,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
 	if (ret)
 		goto fail;
 
-	ret = v3d_lock_bo_reservations(clean_job, &acquire_ctx);
+	ret = v3d_lock_bo_reservations(clean_job, &exec);
 	if (ret)
 		goto fail;
 
@@ -983,7 +980,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
 
 	v3d_attach_fences_and_unlock_reservation(file_priv,
 						 clean_job,
-						 &acquire_ctx,
+						 &exec,
 						 args->out_sync,
 						 &se,
 						 clean_job->done_fence);
@@ -996,8 +993,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
 fail_unreserve:
 	mutex_unlock(&v3d->sched_lock);
 fail_perfmon:
-	drm_gem_unlock_reservations(clean_job->bo, clean_job->bo_count,
-				    &acquire_ctx);
+	drm_exec_fini(&exec);
 fail:
 	v3d_job_cleanup((void *)job);
 	v3d_job_cleanup(clean_job);
-- 
2.34.1

[PATCH 7/7] drm: remove drm_gem_(un)lock_reservations

[Christian König]

Not used any more.

Signed-off-by: Christian König <christian.koenig@amd.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
---
 drivers/gpu/drm/drm_gem.c              | 78 --------------------------
 drivers/gpu/drm/scheduler/sched_main.c |  5 +-
 include/drm/drm_gem.h                  |  4 --
 3 files changed, 2 insertions(+), 85 deletions(-)

diff --git a/drivers/gpu/drm/drm_gem.c b/drivers/gpu/drm/drm_gem.c
index c18686f434d4..2753926c74db 100644
--- a/drivers/gpu/drm/drm_gem.c
+++ b/drivers/gpu/drm/drm_gem.c
@@ -1214,84 +1214,6 @@ void drm_gem_vunmap_unlocked(struct drm_gem_object *obj, struct iosys_map *map)
 }
 EXPORT_SYMBOL(drm_gem_vunmap_unlocked);
 
-/**
- * drm_gem_lock_reservations - Sets up the ww context and acquires
- * the lock on an array of GEM objects.
- *
- * Once you've locked your reservations, you'll want to set up space
- * for your shared fences (if applicable), submit your job, then
- * drm_gem_unlock_reservations().
- *
- * @objs: drm_gem_objects to lock
- * @count: Number of objects in @objs
- * @acquire_ctx: struct ww_acquire_ctx that will be initialized as
- * part of tracking this set of locked reservations.
- */
-int
-drm_gem_lock_reservations(struct drm_gem_object **objs, int count,
-			  struct ww_acquire_ctx *acquire_ctx)
-{
-	int contended = -1;
-	int i, ret;
-
-	ww_acquire_init(acquire_ctx, &reservation_ww_class);
-
-retry:
-	if (contended != -1) {
-		struct drm_gem_object *obj = objs[contended];
-
-		ret = dma_resv_lock_slow_interruptible(obj->resv,
-								 acquire_ctx);
-		if (ret) {
-			ww_acquire_fini(acquire_ctx);
-			return ret;
-		}
-	}
-
-	for (i = 0; i < count; i++) {
-		if (i == contended)
-			continue;
-
-		ret = dma_resv_lock_interruptible(objs[i]->resv,
-							    acquire_ctx);
-		if (ret) {
-			int j;
-
-			for (j = 0; j < i; j++)
-				dma_resv_unlock(objs[j]->resv);
-
-			if (contended != -1 && contended >= i)
-				dma_resv_unlock(objs[contended]->resv);
-
-			if (ret == -EDEADLK) {
-				contended = i;
-				goto retry;
-			}
-
-			ww_acquire_fini(acquire_ctx);
-			return ret;
-		}
-	}
-
-	ww_acquire_done(acquire_ctx);
-
-	return 0;
-}
-EXPORT_SYMBOL(drm_gem_lock_reservations);
-
-void
-drm_gem_unlock_reservations(struct drm_gem_object **objs, int count,
-			    struct ww_acquire_ctx *acquire_ctx)
-{
-	int i;
-
-	for (i = 0; i < count; i++)
-		dma_resv_unlock(objs[i]->resv);
-
-	ww_acquire_fini(acquire_ctx);
-}
-EXPORT_SYMBOL(drm_gem_unlock_reservations);
-
 /**
  * drm_gem_lru_init - initialize a LRU
  *
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
index 7b2bfc10c1a5..335640369a0f 100644
--- a/drivers/gpu/drm/scheduler/sched_main.c
+++ b/drivers/gpu/drm/scheduler/sched_main.c
@@ -795,9 +795,8 @@ EXPORT_SYMBOL(drm_sched_job_add_resv_dependencies);
  * @write: whether the job might write the object (so we need to depend on
  * shared fences in the reservation object).
  *
- * This should be called after drm_gem_lock_reservations() on your array of
- * GEM objects used in the job but before updating the reservations with your
- * own fences.
+ * This should be called after locking your GEM objects used in the job but
+ * before updating the reservations with your own fences.
  *
  * Returns:
  * 0 on success, or an error on failing to expand the array.
diff --git a/include/drm/drm_gem.h b/include/drm/drm_gem.h
index bbc721870c13..35ffa5aa7b96 100644
--- a/include/drm/drm_gem.h
+++ b/include/drm/drm_gem.h
@@ -508,10 +508,6 @@ int drm_gem_objects_lookup(struct drm_file *filp, void __user *bo_handles,
 struct drm_gem_object *drm_gem_object_lookup(struct drm_file *filp, u32 handle);
 long drm_gem_dma_resv_wait(struct drm_file *filep, u32 handle,
 				    bool wait_all, unsigned long timeout);
-int drm_gem_lock_reservations(struct drm_gem_object **objs, int count,
-			      struct ww_acquire_ctx *acquire_ctx);
-void drm_gem_unlock_reservations(struct drm_gem_object **objs, int count,
-				 struct ww_acquire_ctx *acquire_ctx);
 int drm_gem_dumb_map_offset(struct drm_file *file, struct drm_device *dev,
 			    u32 handle, u64 *offset);
 
-- 
2.34.1

Re: [PATCH 5/7] drm/panfrost: switch to using drm_exec

[Steven Price]

On 12/07/2023 13:47, Christian König wrote:
> Just a straightforward conversion without any optimization.
> 
> Only compile tested for now.
> 
> Signed-off-by: Christian König <christian.koenig@amd.com>
> Cc: Rob Herring <robh@kernel.org>
> Cc: Tomeu Vizoso <tomeu.vizoso@collabora.com>
> Cc: Steven Price <steven.price@arm.com>
> Cc: Alyssa Rosenzweig <alyssa.rosenzweig@collabora.com>
> ---
>  drivers/gpu/drm/panfrost/Kconfig        |  1 +
>  drivers/gpu/drm/panfrost/panfrost_job.c | 12 +++++++-----
>  2 files changed, 8 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/panfrost/Kconfig b/drivers/gpu/drm/panfrost/Kconfig
> index e6403a9d66ad..e86a1a2fd8e1 100644
> --- a/drivers/gpu/drm/panfrost/Kconfig
> +++ b/drivers/gpu/drm/panfrost/Kconfig
> @@ -7,6 +7,7 @@ config DRM_PANFROST
>  	depends on !GENERIC_ATOMIC64    # for IOMMU_IO_PGTABLE_LPAE
>  	depends on MMU
>  	select DRM_SCHED
> +	select DRM_EXEC
>  	select IOMMU_SUPPORT
>  	select IOMMU_IO_PGTABLE_LPAE
>  	select DRM_GEM_SHMEM_HELPER
> diff --git a/drivers/gpu/drm/panfrost/panfrost_job.c b/drivers/gpu/drm/panfrost/panfrost_job.c
> index dbc597ab46fb..8b9206e910b5 100644
> --- a/drivers/gpu/drm/panfrost/panfrost_job.c
> +++ b/drivers/gpu/drm/panfrost/panfrost_job.c
> @@ -8,6 +8,7 @@
>  #include <linux/platform_device.h>
>  #include <linux/pm_runtime.h>
>  #include <linux/dma-resv.h>
> +#include <drm/drm_exec.h>
>  #include <drm/gpu_scheduler.h>
>  #include <drm/panfrost_drm.h>
>  
> @@ -275,13 +276,14 @@ static void panfrost_attach_object_fences(struct drm_gem_object **bos,
>  int panfrost_job_push(struct panfrost_job *job)
>  {
>  	struct panfrost_device *pfdev = job->pfdev;
> -	struct ww_acquire_ctx acquire_ctx;
> +	struct drm_exec exec;
>  	int ret = 0;
>  
> -	ret = drm_gem_lock_reservations(job->bos, job->bo_count,
> -					    &acquire_ctx);
> +	drm_exec_init(&exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
> +	drm_exec_until_all_locked(&exec)
> +		ret = drm_exec_prepare_array(&exec, job->bos, job->bo_count, 1);

This loop bothers me - the return value is ignored until
drm_exec_until_all_locked() decides we can exit. It also reads badly
without the braces and with the "if" unindented below. The documentation
("a typical usage pattern") suggests this should be something like:

	drm_exec_until_all_locked(&exec) {
		ret = drm_exec_prepare_array(...);
		drm_exec_retry_on_contention(&exec);
		if (ret)
			goto unlock;
	}

Although I'm not sure if that's actually different because if there's no
contention that drm_exec_until_all_locked() looks like it should exit.

Perhaps it's just the name drm_exec_until_all_locked() which perhaps
should be drm_exec_until_not_contended()...?

Anyway I gave it a quick spin on a Firefly-RK3288 and didn't see any
problems, but I don't think I've got any tests which would create
contention on the BOs.

Steve

>  	if (ret)
> -		return ret;
> +		goto unlock;
>  
>  	mutex_lock(&pfdev->sched_lock);
>  	drm_sched_job_arm(&job->base);
> @@ -305,7 +307,7 @@ int panfrost_job_push(struct panfrost_job *job)
>  				      job->render_done_fence);
>  
>  unlock:
> -	drm_gem_unlock_reservations(job->bos, job->bo_count, &acquire_ctx);
> +	drm_exec_fini(&exec);
>  
>  	return ret;
>  }

Re: [PATCH 6/7] drm/v3d: switch to using drm_exec

[Maira Canal]

Hi Christian,

I believe that `select DRM_EXEC` is missing on v3d's Kconfig file. If
we don't select it, we will get some compilation errors.

Apart from this problem, I ran some tests on the RPi 4 and didn't see
any problems.

Best Regards,
- Maíra

On 7/12/23 09:47, Christian König wrote:
> Just a straightforward conversion without any optimization.
> 
> Only compile tested for now.
> 
> Signed-off-by: Christian König <christian.koenig@amd.com>
> Cc: Emma Anholt <emma@anholt.net>
> Cc: Melissa Wen <mwen@igalia.com>
> ---
>   drivers/gpu/drm/v3d/v3d_gem.c | 44 ++++++++++++++++-------------------
>   1 file changed, 20 insertions(+), 24 deletions(-)
> 
> diff --git a/drivers/gpu/drm/v3d/v3d_gem.c b/drivers/gpu/drm/v3d/v3d_gem.c
> index 2e94ce788c71..190e2a9f64a4 100644
> --- a/drivers/gpu/drm/v3d/v3d_gem.c
> +++ b/drivers/gpu/drm/v3d/v3d_gem.c
> @@ -10,6 +10,7 @@
>   #include <linux/sched/signal.h>
>   #include <linux/uaccess.h>
>   
> +#include <drm/drm_exec.h>
>   #include <drm/drm_managed.h>
>   #include <drm/drm_syncobj.h>
>   #include <uapi/drm/v3d_drm.h>
> @@ -249,20 +250,17 @@ v3d_invalidate_caches(struct v3d_dev *v3d)
>    * to v3d, so we don't attach dma-buf fences to them.
>    */
>   static int
> -v3d_lock_bo_reservations(struct v3d_job *job,
> -			 struct ww_acquire_ctx *acquire_ctx)
> +v3d_lock_bo_reservations(struct v3d_job *job, struct drm_exec *exec)
>   {
>   	int i, ret;
>   
> -	ret = drm_gem_lock_reservations(job->bo, job->bo_count, acquire_ctx);
> +	drm_exec_init(exec, DRM_EXEC_INTERRUPTIBLE_WAIT);
> +	drm_exec_until_all_locked(exec)
> +		ret = drm_exec_prepare_array(exec, job->bo, job->bo_count, 1);
>   	if (ret)
> -		return ret;
> +		goto fail;
>   
>   	for (i = 0; i < job->bo_count; i++) {
> -		ret = dma_resv_reserve_fences(job->bo[i]->resv, 1);
> -		if (ret)
> -			goto fail;
> -
>   		ret = drm_sched_job_add_implicit_dependencies(&job->base,
>   							      job->bo[i], true);
>   		if (ret)
> @@ -272,7 +270,7 @@ v3d_lock_bo_reservations(struct v3d_job *job,
>   	return 0;
>   
>   fail:
> -	drm_gem_unlock_reservations(job->bo, job->bo_count, acquire_ctx);
> +	drm_exec_fini(exec);
>   	return ret;
>   }
>   
> @@ -477,7 +475,7 @@ v3d_push_job(struct v3d_job *job)
>   static void
>   v3d_attach_fences_and_unlock_reservation(struct drm_file *file_priv,
>   					 struct v3d_job *job,
> -					 struct ww_acquire_ctx *acquire_ctx,
> +					 struct drm_exec *exec,
>   					 u32 out_sync,
>   					 struct v3d_submit_ext *se,
>   					 struct dma_fence *done_fence)
> @@ -492,7 +490,7 @@ v3d_attach_fences_and_unlock_reservation(struct drm_file *file_priv,
>   				   DMA_RESV_USAGE_WRITE);
>   	}
>   
> -	drm_gem_unlock_reservations(job->bo, job->bo_count, acquire_ctx);
> +	drm_exec_fini(exec);
>   
>   	/* Update the return sync object for the job */
>   	/* If it only supports a single signal semaphore*/
> @@ -669,7 +667,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
>   	struct v3d_render_job *render = NULL;
>   	struct v3d_job *clean_job = NULL;
>   	struct v3d_job *last_job;
> -	struct ww_acquire_ctx acquire_ctx;
> +	struct drm_exec exec;
>   	int ret = 0;
>   
>   	trace_v3d_submit_cl_ioctl(&v3d->drm, args->rcl_start, args->rcl_end);
> @@ -731,7 +729,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
>   	if (ret)
>   		goto fail;
>   
> -	ret = v3d_lock_bo_reservations(last_job, &acquire_ctx);
> +	ret = v3d_lock_bo_reservations(last_job, &exec);
>   	if (ret)
>   		goto fail;
>   
> @@ -775,7 +773,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
>   
>   	v3d_attach_fences_and_unlock_reservation(file_priv,
>   						 last_job,
> -						 &acquire_ctx,
> +						 &exec,
>   						 args->out_sync,
>   						 &se,
>   						 last_job->done_fence);
> @@ -791,8 +789,7 @@ v3d_submit_cl_ioctl(struct drm_device *dev, void *data,
>   fail_unreserve:
>   	mutex_unlock(&v3d->sched_lock);
>   fail_perfmon:
> -	drm_gem_unlock_reservations(last_job->bo,
> -				    last_job->bo_count, &acquire_ctx);
> +	drm_exec_fini(&exec);
>   fail:
>   	v3d_job_cleanup((void *)bin);
>   	v3d_job_cleanup((void *)render);
> @@ -819,7 +816,7 @@ v3d_submit_tfu_ioctl(struct drm_device *dev, void *data,
>   	struct drm_v3d_submit_tfu *args = data;
>   	struct v3d_submit_ext se = {0};
>   	struct v3d_tfu_job *job = NULL;
> -	struct ww_acquire_ctx acquire_ctx;
> +	struct drm_exec exec;
>   	int ret = 0;
>   
>   	trace_v3d_submit_tfu_ioctl(&v3d->drm, args->iia);
> @@ -870,7 +867,7 @@ v3d_submit_tfu_ioctl(struct drm_device *dev, void *data,
>   		job->base.bo[job->base.bo_count] = bo;
>   	}
>   
> -	ret = v3d_lock_bo_reservations(&job->base, &acquire_ctx);
> +	ret = v3d_lock_bo_reservations(&job->base, &exec);
>   	if (ret)
>   		goto fail;
>   
> @@ -879,7 +876,7 @@ v3d_submit_tfu_ioctl(struct drm_device *dev, void *data,
>   	mutex_unlock(&v3d->sched_lock);
>   
>   	v3d_attach_fences_and_unlock_reservation(file_priv,
> -						 &job->base, &acquire_ctx,
> +						 &job->base, &exec,
>   						 args->out_sync,
>   						 &se,
>   						 job->base.done_fence);
> @@ -914,7 +911,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
>   	struct v3d_submit_ext se = {0};
>   	struct v3d_csd_job *job = NULL;
>   	struct v3d_job *clean_job = NULL;
> -	struct ww_acquire_ctx acquire_ctx;
> +	struct drm_exec exec;
>   	int ret;
>   
>   	trace_v3d_submit_csd_ioctl(&v3d->drm, args->cfg[5], args->cfg[6]);
> @@ -957,7 +954,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
>   	if (ret)
>   		goto fail;
>   
> -	ret = v3d_lock_bo_reservations(clean_job, &acquire_ctx);
> +	ret = v3d_lock_bo_reservations(clean_job, &exec);
>   	if (ret)
>   		goto fail;
>   
> @@ -983,7 +980,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
>   
>   	v3d_attach_fences_and_unlock_reservation(file_priv,
>   						 clean_job,
> -						 &acquire_ctx,
> +						 &exec,
>   						 args->out_sync,
>   						 &se,
>   						 clean_job->done_fence);
> @@ -996,8 +993,7 @@ v3d_submit_csd_ioctl(struct drm_device *dev, void *data,
>   fail_unreserve:
>   	mutex_unlock(&v3d->sched_lock);
>   fail_perfmon:
> -	drm_gem_unlock_reservations(clean_job->bo, clean_job->bo_count,
> -				    &acquire_ctx);
> +	drm_exec_fini(&exec);
>   fail:
>   	v3d_job_cleanup((void *)job);
>   	v3d_job_cleanup(clean_job);