Sync drm/panthor and drm/sched with 6.12-rc2

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

@@ -5399,7 +5399,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev,
 			if (!tmp_adev->asic_reset_res && !job_signaled)
 				drm_sched_resubmit_jobs(&ring->sched);
 
-			drm_sched_start(&ring->sched, !tmp_adev->asic_reset_res);
+			drm_sched_start(&ring->sched);
 		}
 
 		if (adev->enable_mes && adev->ip_versions[GC_HWIP][0] != IP_VERSION(11, 0, 3))
@@ -5854,7 +5854,7 @@ void amdgpu_pci_resume(struct pci_dev *pdev)
 
 
 		drm_sched_resubmit_jobs(&ring->sched);
-		drm_sched_start(&ring->sched, true);
+		drm_sched_start(&ring->sched);
 	}
 
 	amdgpu_device_unset_mp1_state(adev);

drivers/gpu/drm/etnaviv/etnaviv_sched.c

@@ -71,12 +71,12 @@ static enum drm_gpu_sched_stat etnaviv_sched_timedout_job(struct drm_sched_job
 
 	drm_sched_resubmit_jobs(&gpu->sched);
 
-	drm_sched_start(&gpu->sched, true);
+	drm_sched_start(&gpu->sched);
 	return DRM_GPU_SCHED_STAT_NOMINAL;
 
 out_no_timeout:
 	/* restart scheduler after GPU is usable again */
-	drm_sched_start(&gpu->sched, true);
+	drm_sched_start(&gpu->sched);
 	return DRM_GPU_SCHED_STAT_NOMINAL;
 }
 

drivers/gpu/drm/lima/lima_sched.c

@@ -430,7 +430,7 @@ static enum drm_gpu_sched_stat lima_sched_timedout_job(struct drm_sched_job *job
 	lima_pm_idle(ldev);
 
 	drm_sched_resubmit_jobs(&pipe->base);
-	drm_sched_start(&pipe->base, true);
+	drm_sched_start(&pipe->base);
 
 	return DRM_GPU_SCHED_STAT_NOMINAL;
 }

drivers/gpu/drm/panfrost/panfrost_job.c

@@ -696,7 +696,7 @@ panfrost_reset(struct panfrost_device *pfdev,
 
 	/* Restart the schedulers */
 	for (i = 0; i < NUM_JOB_SLOTS; i++)
-		drm_sched_start(&pfdev->js->queue[i].sched, true);
+		drm_sched_start(&pfdev->js->queue[i].sched);
 
 	/* Re-enable job interrupts now that everything has been restarted. */
 	job_write(pfdev, JOB_INT_MASK,

drivers/gpu/drm/panthor/panthor_mmu.c

@@ -575,6 +575,12 @@ static int mmu_hw_do_operation_locked(struct panthor_device *ptdev, int as_nr,
 	if (as_nr < 0)
 		return 0;
 
+	/*
+	 * If the AS number is greater than zero, then we can be sure
+	 * the device is up and running, so we don't need to explicitly
+	 * power it up
+	 */
+
 	if (op != AS_COMMAND_UNLOCK)
 		lock_region(ptdev, as_nr, iova, size);
 
@@ -826,7 +832,7 @@ static void panthor_vm_stop(struct panthor_vm *vm)
 
 static void panthor_vm_start(struct panthor_vm *vm)
 {
-	drm_sched_start(&vm->sched, true);
+	drm_sched_start(&vm->sched);
 }
 
 /**
@@ -873,10 +879,6 @@ static int panthor_vm_flush_range(struct panthor_vm *vm, u64 iova, u64 size)
 	if (!drm_dev_enter(&ptdev->base, &cookie))
 		return 0;
 
-	/*
-	 * If we made it this far, that means the device is awake, because
-	 * upon device suspension, all active VMs are given an AS id of -1
-	 */
 	ret = mmu_hw_do_operation(vm, iova, size, AS_COMMAND_FLUSH_PT);
 
 	drm_dev_exit(cookie);
@@ -1248,9 +1250,17 @@ static int panthor_vm_prepare_map_op_ctx(struct panthor_vm_op_ctx *op_ctx,
 		goto err_cleanup;
 	}
 
+	/* drm_gpuvm_bo_obtain_prealloc() will call drm_gpuvm_bo_put() on our
+	 * pre-allocated BO if the <BO,VM> association exists. Given we
+	 * only have one ref on preallocated_vm_bo, drm_gpuvm_bo_destroy() will
+	 * be called immediately, and we have to hold the VM resv lock when
+	 * calling this function.
+	 */
+	dma_resv_lock(panthor_vm_resv(vm), NULL);
 	mutex_lock(&bo->gpuva_list_lock);
 	op_ctx->map.vm_bo = drm_gpuvm_bo_obtain_prealloc(preallocated_vm_bo);
 	mutex_unlock(&bo->gpuva_list_lock);
+	dma_resv_unlock(panthor_vm_resv(vm));
 
 	/* If the a vm_bo for this <VM,BO> combination exists, it already
 	 * retains a pin ref, and we can release the one we took earlier.

drivers/gpu/drm/panthor/panthor_sched.c

@@ -1103,7 +1103,13 @@ cs_slot_sync_queue_state_locked(struct panthor_device *ptdev, u32 csg_id, u32 cs
 			list_move_tail(&group->wait_node,
 				       &group->ptdev->scheduler->groups.waiting);
 		}
-		group->blocked_queues |= BIT(cs_id);
+
+		/* The queue is only blocked if there's no deferred operation
+		 * pending, which can be checked through the scoreboard status.
+		 */
+		if (!cs_iface->output->status_scoreboards)
+			group->blocked_queues |= BIT(cs_id);
+
 		queue->syncwait.gpu_va = cs_iface->output->status_wait_sync_ptr;
 		queue->syncwait.ref = cs_iface->output->status_wait_sync_value;
 		status_wait_cond = cs_iface->output->status_wait & CS_STATUS_WAIT_SYNC_COND_MASK;
@@ -2046,6 +2052,7 @@ static void
 tick_ctx_cleanup(struct panthor_scheduler *sched,
 		 struct panthor_sched_tick_ctx *ctx)
 {
+	struct panthor_device *ptdev = sched->ptdev;
 	struct panthor_group *group, *tmp;
 	u32 i;
 
@@ -2054,7 +2061,7 @@ tick_ctx_cleanup(struct panthor_scheduler *sched,
 			/* If everything went fine, we should only have groups
 			 * to be terminated in the old_groups lists.
 			 */
-			drm_WARN_ON(&group->ptdev->base, !ctx->csg_upd_failed_mask &&
+			drm_WARN_ON(&ptdev->base, !ctx->csg_upd_failed_mask &&
 				    group_can_run(group));
 
 			if (!group_can_run(group)) {
@@ -2077,7 +2084,7 @@ tick_ctx_cleanup(struct panthor_scheduler *sched,
 		/* If everything went fine, the groups to schedule lists should
 		 * be empty.
 		 */
-		drm_WARN_ON(&group->ptdev->base,
+		drm_WARN_ON(&ptdev->base,
 			    !ctx->csg_upd_failed_mask && !list_empty(&ctx->groups[i]));
 
 		list_for_each_entry_safe(group, tmp, &ctx->groups[i], run_node) {
@@ -2538,7 +2545,7 @@ static void queue_start(struct panthor_queue *queue)
 	list_for_each_entry(job, &queue->scheduler.pending_list, base.list)
 		job->base.s_fence->parent = dma_fence_get(job->done_fence);
 
-	drm_sched_start(&queue->scheduler, true);
+	drm_sched_start(&queue->scheduler);
 }
 
 static void panthor_group_stop(struct panthor_group *group)
@@ -3242,6 +3249,18 @@ int panthor_group_destroy(struct panthor_file *pfile, u32 group_handle)
 	return 0;
 }
 
+static struct panthor_group *group_from_handle(struct panthor_group_pool *pool,
+					       u32 group_handle)
+{
+	struct panthor_group *group;
+
+	xa_lock(&pool->xa);
+	group = group_get(xa_load(&pool->xa, group_handle));
+	xa_unlock(&pool->xa);
+
+	return group;
+}
+
 int panthor_group_get_state(struct panthor_file *pfile,
 			    struct drm_panthor_group_get_state *get_state)
 {
@@ -3253,7 +3272,7 @@ int panthor_group_get_state(struct panthor_file *pfile,
 	if (get_state->pad)
 		return -EINVAL;
 
-	group = group_get(xa_load(&gpool->xa, get_state->group_handle));
+	group = group_from_handle(gpool, get_state->group_handle);
 	if (!group)
 		return -EINVAL;
 
@@ -3384,7 +3403,7 @@ panthor_job_create(struct panthor_file *pfile,
 	job->call_info.latest_flush = qsubmit->latest_flush;
 	INIT_LIST_HEAD(&job->node);
 
-	job->group = group_get(xa_load(&gpool->xa, group_handle));
+	job->group = group_from_handle(gpool, group_handle);
 	if (!job->group) {
 		ret = -EINVAL;
 		goto err_put_job;
@@ -3424,13 +3443,8 @@ void panthor_job_update_resvs(struct drm_exec *exec, struct drm_sched_job *sched
 {
 	struct panthor_job *job = container_of(sched_job, struct panthor_job, base);
 
-	/* Still not sure why we want USAGE_WRITE for external objects, since I
-	 * was assuming this would be handled through explicit syncs being imported
-	 * to external BOs with DMA_BUF_IOCTL_IMPORT_SYNC_FILE, but other drivers
-	 * seem to pass DMA_RESV_USAGE_WRITE, so there must be a good reason.
-	 */
 	panthor_vm_update_resvs(job->group->vm, exec, &sched_job->s_fence->finished,
-				DMA_RESV_USAGE_BOOKKEEP, DMA_RESV_USAGE_WRITE);
+				DMA_RESV_USAGE_BOOKKEEP, DMA_RESV_USAGE_BOOKKEEP);
 }
 
 void panthor_sched_unplug(struct panthor_device *ptdev)

drivers/gpu/drm/scheduler/sched_entity.c

@@ -71,13 +71,19 @@ int drm_sched_entity_init(struct drm_sched_entity *entity,
 	entity->guilty = guilty;
 	entity->num_sched_list = num_sched_list;
 	entity->priority = priority;
+	/*
+	 * It's perfectly valid to initialize an entity without having a valid
+	 * scheduler attached. It's just not valid to use the scheduler before it
+	 * is initialized itself.
+	 */
 	entity->sched_list = num_sched_list > 1 ? sched_list : NULL;
 	RCU_INIT_POINTER(entity->last_scheduled, NULL);
 	RB_CLEAR_NODE(&entity->rb_tree_node);
 
-	if (!sched_list[0]->sched_rq) {
-		/* Warn drivers not to do this and to fix their DRM
-		 * calling order.
+	if (num_sched_list && !sched_list[0]->sched_rq) {
+		/* Since every entry covered by num_sched_list
+		 * should be non-NULL and therefore we warn drivers
+		 * not to do this and to fix their DRM calling order.
 		 */
 		pr_warn("%s: called with uninitialized scheduler\n", __func__);
 	} else if (num_sched_list) {
@@ -127,8 +133,10 @@ void drm_sched_entity_modify_sched(struct drm_sched_entity *entity,
 {
 	WARN_ON(!num_sched_list || !sched_list);
 
+	spin_lock(&entity->rq_lock);
 	entity->sched_list = sched_list;
 	entity->num_sched_list = num_sched_list;
+	spin_unlock(&entity->rq_lock);
 }
 EXPORT_SYMBOL(drm_sched_entity_modify_sched);
 
@@ -374,7 +382,7 @@ static void drm_sched_entity_wakeup(struct dma_fence *f,
 		container_of(cb, struct drm_sched_entity, cb);
 
 	drm_sched_entity_clear_dep(f, cb);
-	drm_sched_wakeup(entity->rq->sched, entity);
+	drm_sched_wakeup(entity->rq->sched);
 }
 
 /**
@@ -591,6 +599,9 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
 
 	/* first job wakes up scheduler */
 	if (first) {
+		struct drm_gpu_scheduler *sched;
+		struct drm_sched_rq *rq;
+
 		/* Add the entity to the run queue */
 		spin_lock(&entity->rq_lock);
 		if (entity->stopped) {
@@ -600,13 +611,16 @@ void drm_sched_entity_push_job(struct drm_sched_job *sched_job)
 			return;
 		}
 
-		drm_sched_rq_add_entity(entity->rq, entity);
+		rq = entity->rq;
+		sched = rq->sched;
+
+		drm_sched_rq_add_entity(rq, entity);
 		spin_unlock(&entity->rq_lock);
 
 		if (drm_sched_policy == DRM_SCHED_POLICY_FIFO)
 			drm_sched_rq_update_fifo(entity, submit_ts);
 
-		drm_sched_wakeup(entity->rq->sched, entity);
+		drm_sched_wakeup(sched);
 	}
 }
 EXPORT_SYMBOL(drm_sched_entity_push_job);

drivers/gpu/drm/scheduler/sched_fence.c

@@ -33,9 +33,7 @@ static struct kmem_cache *sched_fence_slab;
 
 static int __init drm_sched_fence_slab_init(void)
 {
-	sched_fence_slab = kmem_cache_create(
-		"drm_sched_fence", sizeof(struct drm_sched_fence), 0,
-		SLAB_HWCACHE_ALIGN, NULL);
+	sched_fence_slab = KMEM_CACHE(drm_sched_fence, SLAB_HWCACHE_ALIGN);
 	if (!sched_fence_slab)
 		return -ENOMEM;