diff options
Diffstat (limited to 'drivers/gpu/drm/scheduler/sched_main.c')
| -rw-r--r-- | drivers/gpu/drm/scheduler/sched_main.c | 1217 |
1 files changed, 904 insertions, 313 deletions
diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c index a2a953693b45..1d4f1b822e7b 100644 --- a/drivers/gpu/drm/scheduler/sched_main.c +++ b/drivers/gpu/drm/scheduler/sched_main.c @@ -41,24 +41,135 @@ * 4. Entities themselves maintain a queue of jobs that will be scheduled on * the hardware. * - * The jobs in a entity are always scheduled in the order that they were pushed. + * The jobs in an entity are always scheduled in the order in which they were pushed. + * + * Note that once a job was taken from the entities queue and pushed to the + * hardware, i.e. the pending queue, the entity must not be referenced anymore + * through the jobs entity pointer. + */ + +/** + * DOC: Flow Control + * + * The DRM GPU scheduler provides a flow control mechanism to regulate the rate + * in which the jobs fetched from scheduler entities are executed. + * + * In this context the &drm_gpu_scheduler keeps track of a driver specified + * credit limit representing the capacity of this scheduler and a credit count; + * every &drm_sched_job carries a driver specified number of credits. + * + * Once a job is executed (but not yet finished), the job's credits contribute + * to the scheduler's credit count until the job is finished. If by executing + * one more job the scheduler's credit count would exceed the scheduler's + * credit limit, the job won't be executed. Instead, the scheduler will wait + * until the credit count has decreased enough to not overflow its credit limit. + * This implies waiting for previously executed jobs. */ -#include <linux/kthread.h> +#include <linux/export.h> #include <linux/wait.h> #include <linux/sched.h> #include <linux/completion.h> +#include <linux/dma-resv.h> #include <uapi/linux/sched/types.h> #include <drm/drm_print.h> +#include <drm/drm_gem.h> +#include <drm/drm_syncobj.h> #include <drm/gpu_scheduler.h> #include <drm/spsc_queue.h> +#include "sched_internal.h" + #define CREATE_TRACE_POINTS #include "gpu_scheduler_trace.h" -#define to_drm_sched_job(sched_job) \ - container_of((sched_job), struct drm_sched_job, queue_node) +int drm_sched_policy = DRM_SCHED_POLICY_FIFO; + +/** + * DOC: sched_policy (int) + * Used to override default entities scheduling policy in a run queue. + */ +MODULE_PARM_DESC(sched_policy, "Specify the scheduling policy for entities on a run-queue, " __stringify(DRM_SCHED_POLICY_RR) " = Round Robin, " __stringify(DRM_SCHED_POLICY_FIFO) " = FIFO (default)."); +module_param_named(sched_policy, drm_sched_policy, int, 0444); + +static u32 drm_sched_available_credits(struct drm_gpu_scheduler *sched) +{ + u32 credits; + + WARN_ON(check_sub_overflow(sched->credit_limit, + atomic_read(&sched->credit_count), + &credits)); + + return credits; +} + +/** + * drm_sched_can_queue -- Can we queue more to the hardware? + * @sched: scheduler instance + * @entity: the scheduler entity + * + * Return true if we can push at least one more job from @entity, false + * otherwise. + */ +static bool drm_sched_can_queue(struct drm_gpu_scheduler *sched, + struct drm_sched_entity *entity) +{ + struct drm_sched_job *s_job; + + s_job = drm_sched_entity_queue_peek(entity); + if (!s_job) + return false; + + /* If a job exceeds the credit limit, truncate it to the credit limit + * itself to guarantee forward progress. + */ + if (s_job->credits > sched->credit_limit) { + dev_WARN(sched->dev, + "Jobs may not exceed the credit limit, truncate.\n"); + s_job->credits = sched->credit_limit; + } + + return drm_sched_available_credits(sched) >= s_job->credits; +} + +static __always_inline bool drm_sched_entity_compare_before(struct rb_node *a, + const struct rb_node *b) +{ + struct drm_sched_entity *ent_a = rb_entry((a), struct drm_sched_entity, rb_tree_node); + struct drm_sched_entity *ent_b = rb_entry((b), struct drm_sched_entity, rb_tree_node); + + return ktime_before(ent_a->oldest_job_waiting, ent_b->oldest_job_waiting); +} + +static void drm_sched_rq_remove_fifo_locked(struct drm_sched_entity *entity, + struct drm_sched_rq *rq) +{ + if (!RB_EMPTY_NODE(&entity->rb_tree_node)) { + rb_erase_cached(&entity->rb_tree_node, &rq->rb_tree_root); + RB_CLEAR_NODE(&entity->rb_tree_node); + } +} + +void drm_sched_rq_update_fifo_locked(struct drm_sched_entity *entity, + struct drm_sched_rq *rq, + ktime_t ts) +{ + /* + * Both locks need to be grabbed, one to protect from entity->rq change + * for entity from within concurrent drm_sched_entity_select_rq and the + * other to update the rb tree structure. + */ + lockdep_assert_held(&entity->lock); + lockdep_assert_held(&rq->lock); + + drm_sched_rq_remove_fifo_locked(entity, rq); + + entity->oldest_job_waiting = ts; + + rb_add_cached(&entity->rb_tree_node, &rq->rb_tree_root, + drm_sched_entity_compare_before); +} /** * drm_sched_rq_init - initialize a given run queue struct @@ -73,6 +184,7 @@ static void drm_sched_rq_init(struct drm_gpu_scheduler *sched, { spin_lock_init(&rq->lock); INIT_LIST_HEAD(&rq->entities); + rq->rb_tree_root = RB_ROOT_CACHED; rq->current_entity = NULL; rq->sched = sched; } @@ -88,12 +200,14 @@ static void drm_sched_rq_init(struct drm_gpu_scheduler *sched, void drm_sched_rq_add_entity(struct drm_sched_rq *rq, struct drm_sched_entity *entity) { + lockdep_assert_held(&entity->lock); + lockdep_assert_held(&rq->lock); + if (!list_empty(&entity->list)) return; - spin_lock(&rq->lock); + atomic_inc(rq->sched->score); list_add_tail(&entity->list, &rq->entities); - spin_unlock(&rq->lock); } /** @@ -107,25 +221,40 @@ void drm_sched_rq_add_entity(struct drm_sched_rq *rq, void drm_sched_rq_remove_entity(struct drm_sched_rq *rq, struct drm_sched_entity *entity) { + lockdep_assert_held(&entity->lock); + if (list_empty(&entity->list)) return; + spin_lock(&rq->lock); + atomic_dec(rq->sched->score); list_del_init(&entity->list); + if (rq->current_entity == entity) rq->current_entity = NULL; + + if (drm_sched_policy == DRM_SCHED_POLICY_FIFO) + drm_sched_rq_remove_fifo_locked(entity, rq); + spin_unlock(&rq->lock); } /** - * drm_sched_rq_select_entity - Select an entity which could provide a job to run + * drm_sched_rq_select_entity_rr - Select an entity which could provide a job to run * + * @sched: the gpu scheduler * @rq: scheduler run queue to check. * - * Try to find a ready entity, returns NULL if none found. + * Try to find the next ready entity. + * + * Return an entity if one is found; return an error-pointer (!NULL) if an + * entity was ready, but the scheduler had insufficient credits to accommodate + * its job; return NULL, if no ready entity was found. */ static struct drm_sched_entity * -drm_sched_rq_select_entity(struct drm_sched_rq *rq) +drm_sched_rq_select_entity_rr(struct drm_gpu_scheduler *sched, + struct drm_sched_rq *rq) { struct drm_sched_entity *entity; @@ -134,23 +263,14 @@ drm_sched_rq_select_entity(struct drm_sched_rq *rq) entity = rq->current_entity; if (entity) { list_for_each_entry_continue(entity, &rq->entities, list) { - if (drm_sched_entity_is_ready(entity)) { - rq->current_entity = entity; - reinit_completion(&entity->entity_idle); - spin_unlock(&rq->lock); - return entity; - } + if (drm_sched_entity_is_ready(entity)) + goto found; } } list_for_each_entry(entity, &rq->entities, list) { - - if (drm_sched_entity_is_ready(entity)) { - rq->current_entity = entity; - reinit_completion(&entity->entity_idle); - spin_unlock(&rq->lock); - return entity; - } + if (drm_sched_entity_is_ready(entity)) + goto found; if (entity == rq->current_entity) break; @@ -159,28 +279,105 @@ drm_sched_rq_select_entity(struct drm_sched_rq *rq) spin_unlock(&rq->lock); return NULL; + +found: + if (!drm_sched_can_queue(sched, entity)) { + /* + * If scheduler cannot take more jobs signal the caller to not + * consider lower priority queues. + */ + entity = ERR_PTR(-ENOSPC); + } else { + rq->current_entity = entity; + reinit_completion(&entity->entity_idle); + } + + spin_unlock(&rq->lock); + + return entity; +} + +/** + * drm_sched_rq_select_entity_fifo - Select an entity which provides a job to run + * + * @sched: the gpu scheduler + * @rq: scheduler run queue to check. + * + * Find oldest waiting ready entity. + * + * Return an entity if one is found; return an error-pointer (!NULL) if an + * entity was ready, but the scheduler had insufficient credits to accommodate + * its job; return NULL, if no ready entity was found. + */ +static struct drm_sched_entity * +drm_sched_rq_select_entity_fifo(struct drm_gpu_scheduler *sched, + struct drm_sched_rq *rq) +{ + struct rb_node *rb; + + spin_lock(&rq->lock); + for (rb = rb_first_cached(&rq->rb_tree_root); rb; rb = rb_next(rb)) { + struct drm_sched_entity *entity; + + entity = rb_entry(rb, struct drm_sched_entity, rb_tree_node); + if (drm_sched_entity_is_ready(entity)) { + /* If we can't queue yet, preserve the current entity in + * terms of fairness. + */ + if (!drm_sched_can_queue(sched, entity)) { + spin_unlock(&rq->lock); + return ERR_PTR(-ENOSPC); + } + + reinit_completion(&entity->entity_idle); + break; + } + } + spin_unlock(&rq->lock); + + return rb ? rb_entry(rb, struct drm_sched_entity, rb_tree_node) : NULL; +} + +/** + * drm_sched_run_job_queue - enqueue run-job work + * @sched: scheduler instance + */ +static void drm_sched_run_job_queue(struct drm_gpu_scheduler *sched) +{ + if (!READ_ONCE(sched->pause_submit)) + queue_work(sched->submit_wq, &sched->work_run_job); +} + +/** + * drm_sched_run_free_queue - enqueue free-job work + * @sched: scheduler instance + */ +static void drm_sched_run_free_queue(struct drm_gpu_scheduler *sched) +{ + if (!READ_ONCE(sched->pause_submit)) + queue_work(sched->submit_wq, &sched->work_free_job); } /** * drm_sched_job_done - complete a job * @s_job: pointer to the job which is done * - * Finish the job's fence and wake up the worker thread. + * Finish the job's fence and resubmit the work items. */ -static void drm_sched_job_done(struct drm_sched_job *s_job) +static void drm_sched_job_done(struct drm_sched_job *s_job, int result) { struct drm_sched_fence *s_fence = s_job->s_fence; struct drm_gpu_scheduler *sched = s_fence->sched; - atomic_dec(&sched->hw_rq_count); + atomic_sub(s_job->credits, &sched->credit_count); atomic_dec(sched->score); - trace_drm_sched_process_job(s_fence); + trace_drm_sched_job_done(s_fence); dma_fence_get(&s_fence->finished); - drm_sched_fence_finished(s_fence); + drm_sched_fence_finished(s_fence, result); dma_fence_put(&s_fence->finished); - wake_up_interruptible(&sched->wake_up_worker); + drm_sched_run_free_queue(sched); } /** @@ -192,48 +389,47 @@ static void drm_sched_job_done_cb(struct dma_fence *f, struct dma_fence_cb *cb) { struct drm_sched_job *s_job = container_of(cb, struct drm_sched_job, cb); - drm_sched_job_done(s_job); + drm_sched_job_done(s_job, f->error); } /** - * drm_sched_dependency_optimized + * drm_sched_start_timeout - start timeout for reset worker * - * @fence: the dependency fence - * @entity: the entity which depends on the above fence + * @sched: scheduler instance to start the worker for * - * Returns true if the dependency can be optimized and false otherwise + * Start the timeout for the given scheduler. */ -bool drm_sched_dependency_optimized(struct dma_fence* fence, - struct drm_sched_entity *entity) +static void drm_sched_start_timeout(struct drm_gpu_scheduler *sched) { - struct drm_gpu_scheduler *sched = entity->rq->sched; - struct drm_sched_fence *s_fence; + lockdep_assert_held(&sched->job_list_lock); - if (!fence || dma_fence_is_signaled(fence)) - return false; - if (fence->context == entity->fence_context) - return true; - s_fence = to_drm_sched_fence(fence); - if (s_fence && s_fence->sched == sched) - return true; + if (sched->timeout != MAX_SCHEDULE_TIMEOUT && + !list_empty(&sched->pending_list)) + mod_delayed_work(sched->timeout_wq, &sched->work_tdr, sched->timeout); +} - return false; +static void drm_sched_start_timeout_unlocked(struct drm_gpu_scheduler *sched) +{ + spin_lock(&sched->job_list_lock); + drm_sched_start_timeout(sched); + spin_unlock(&sched->job_list_lock); } -EXPORT_SYMBOL(drm_sched_dependency_optimized); /** - * drm_sched_start_timeout - start timeout for reset worker + * drm_sched_tdr_queue_imm: - immediately start job timeout handler * - * @sched: scheduler instance to start the worker for + * @sched: scheduler for which the timeout handling should be started. * - * Start the timeout for the given scheduler. + * Start timeout handling immediately for the named scheduler. */ -static void drm_sched_start_timeout(struct drm_gpu_scheduler *sched) +void drm_sched_tdr_queue_imm(struct drm_gpu_scheduler *sched) { - if (sched->timeout != MAX_SCHEDULE_TIMEOUT && - !list_empty(&sched->pending_list)) - schedule_delayed_work(&sched->work_tdr, sched->timeout); + spin_lock(&sched->job_list_lock); + sched->timeout = 0; + drm_sched_start_timeout(sched); + spin_unlock(&sched->job_list_lock); } +EXPORT_SYMBOL(drm_sched_tdr_queue_imm); /** * drm_sched_fault - immediately start timeout handler @@ -244,7 +440,8 @@ static void drm_sched_start_timeout(struct drm_gpu_scheduler *sched) */ void drm_sched_fault(struct drm_gpu_scheduler *sched) { - mod_delayed_work(system_wq, &sched->work_tdr, 0); + if (sched->timeout_wq) + mod_delayed_work(sched->timeout_wq, &sched->work_tdr, 0); } EXPORT_SYMBOL(drm_sched_fault); @@ -270,7 +467,7 @@ unsigned long drm_sched_suspend_timeout(struct drm_gpu_scheduler *sched) * Modify the timeout to an arbitrarily large value. This also prevents * the timeout to be restarted when new submissions arrive */ - if (mod_delayed_work(system_wq, &sched->work_tdr, MAX_SCHEDULE_TIMEOUT) + if (mod_delayed_work(sched->timeout_wq, &sched->work_tdr, MAX_SCHEDULE_TIMEOUT) && time_after(sched_timeout, now)) return sched_timeout - now; else @@ -294,7 +491,7 @@ void drm_sched_resume_timeout(struct drm_gpu_scheduler *sched, if (list_empty(&sched->pending_list)) cancel_delayed_work(&sched->work_tdr); else - mod_delayed_work(system_wq, &sched->work_tdr, remaining); + mod_delayed_work(sched->timeout_wq, &sched->work_tdr, remaining); spin_unlock(&sched->job_list_lock); } @@ -310,24 +507,51 @@ static void drm_sched_job_begin(struct drm_sched_job *s_job) spin_unlock(&sched->job_list_lock); } +/** + * drm_sched_job_reinsert_on_false_timeout - reinsert the job on a false timeout + * @sched: scheduler instance + * @job: job to be reinserted on the pending list + * + * In the case of a "false timeout" - when a timeout occurs but the GPU isn't + * hung and is making progress, the scheduler must reinsert the job back into + * @sched->pending_list. Otherwise, the job and its resources won't be freed + * through the &struct drm_sched_backend_ops.free_job callback. + * + * This function must be used in "false timeout" cases only. + */ +static void drm_sched_job_reinsert_on_false_timeout(struct drm_gpu_scheduler *sched, + struct drm_sched_job *job) +{ + spin_lock(&sched->job_list_lock); + list_add(&job->list, &sched->pending_list); + + /* After reinserting the job, the scheduler enqueues the free-job work + * again if ready. Otherwise, a signaled job could be added to the + * pending list, but never freed. + */ + drm_sched_run_free_queue(sched); + spin_unlock(&sched->job_list_lock); +} + static void drm_sched_job_timedout(struct work_struct *work) { struct drm_gpu_scheduler *sched; struct drm_sched_job *job; - enum drm_gpu_sched_stat status = DRM_GPU_SCHED_STAT_NOMINAL; + enum drm_gpu_sched_stat status = DRM_GPU_SCHED_STAT_RESET; sched = container_of(work, struct drm_gpu_scheduler, work_tdr.work); - /* Protects against concurrent deletion in drm_sched_get_cleanup_job */ + /* Protects against concurrent deletion in drm_sched_get_finished_job */ spin_lock(&sched->job_list_lock); job = list_first_entry_or_null(&sched->pending_list, struct drm_sched_job, list); if (job) { /* - * Remove the bad job so it cannot be freed by concurrent - * drm_sched_cleanup_jobs. It will be reinserted back after sched->thread - * is parked at which point it's safe. + * Remove the bad job so it cannot be freed by a concurrent + * &struct drm_sched_backend_ops.free_job. It will be + * reinserted after the scheduler's work items have been + * cancelled, at which point it's safe. */ list_del_init(&job->list); spin_unlock(&sched->job_list_lock); @@ -342,37 +566,16 @@ static void drm_sched_job_timedout(struct work_struct *work) job->sched->ops->free_job(job); sched->free_guilty = false; } - } else { - spin_unlock(&sched->job_list_lock); - } - if (status != DRM_GPU_SCHED_STAT_ENODEV) { - spin_lock(&sched->job_list_lock); - drm_sched_start_timeout(sched); + if (status == DRM_GPU_SCHED_STAT_NO_HANG) + drm_sched_job_reinsert_on_false_timeout(sched, job); + } else { spin_unlock(&sched->job_list_lock); } -} - /** - * drm_sched_increase_karma - Update sched_entity guilty flag - * - * @bad: The job guilty of time out - * - * Increment on every hang caused by the 'bad' job. If this exceeds the hang - * limit of the scheduler then the respective sched entity is marked guilty and - * jobs from it will not be scheduled further - */ -void drm_sched_increase_karma(struct drm_sched_job *bad) -{ - drm_sched_increase_karma_ext(bad, 1); + if (status != DRM_GPU_SCHED_STAT_ENODEV) + drm_sched_start_timeout_unlocked(sched); } -EXPORT_SYMBOL(drm_sched_increase_karma); - -void drm_sched_reset_karma(struct drm_sched_job *bad) -{ - drm_sched_increase_karma_ext(bad, 0); -} -EXPORT_SYMBOL(drm_sched_reset_karma); /** * drm_sched_stop - stop the scheduler @@ -385,19 +588,26 @@ EXPORT_SYMBOL(drm_sched_reset_karma); * callers responsibility to release it manually if it's not part of the * pending list any more. * + * This function is typically used for reset recovery (see the docu of + * drm_sched_backend_ops.timedout_job() for details). Do not call it for + * scheduler teardown, i.e., before calling drm_sched_fini(). + * + * As it's only used for reset recovery, drivers must not call this function + * in their &struct drm_sched_backend_ops.timedout_job callback when they + * skip a reset using &enum drm_gpu_sched_stat.DRM_GPU_SCHED_STAT_NO_HANG. */ void drm_sched_stop(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad) { struct drm_sched_job *s_job, *tmp; - kthread_park(sched->thread); + drm_sched_wqueue_stop(sched); /* * Reinsert back the bad job here - now it's safe as - * drm_sched_get_cleanup_job cannot race against us and release the + * drm_sched_get_finished_job() cannot race against us and release the * bad job at this point - we parked (waited for) any in progress - * (earlier) cleanups and drm_sched_get_cleanup_job will not be called - * now until the scheduler thread is unparked. + * (earlier) cleanups and drm_sched_get_finished_job() will not be + * called now until the scheduler's work items are submitted again. */ if (bad && bad->sched == sched) /* @@ -410,14 +620,17 @@ void drm_sched_stop(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad) * Iterate the job list from later to earlier one and either deactive * their HW callbacks or remove them from pending list if they already * signaled. - * This iteration is thread safe as sched thread is stopped. + * This iteration is thread safe as the scheduler's work items have been + * cancelled. */ list_for_each_entry_safe_reverse(s_job, tmp, &sched->pending_list, list) { if (s_job->s_fence->parent && dma_fence_remove_callback(s_job->s_fence->parent, &s_job->cb)) { - atomic_dec(&sched->hw_rq_count); + dma_fence_put(s_job->s_fence->parent); + s_job->s_fence->parent = NULL; + atomic_sub(s_job->credits, &sched->credit_count); } else { /* * remove job from pending_list. @@ -455,89 +668,79 @@ void drm_sched_stop(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad) */ cancel_delayed_work(&sched->work_tdr); } - EXPORT_SYMBOL(drm_sched_stop); /** * drm_sched_start - recover jobs after a reset * * @sched: scheduler instance - * @full_recovery: proceed with complete sched restart + * @errno: error to set on the pending fences + * + * This function is typically used for reset recovery (see the docu of + * drm_sched_backend_ops.timedout_job() for details). Do not call it for + * scheduler startup. The scheduler itself is fully operational after + * drm_sched_init() succeeded. * + * As it's only used for reset recovery, drivers must not call this function + * in their &struct drm_sched_backend_ops.timedout_job callback when they + * skip a reset using &enum drm_gpu_sched_stat.DRM_GPU_SCHED_STAT_NO_HANG. */ -void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery) +void drm_sched_start(struct drm_gpu_scheduler *sched, int errno) { struct drm_sched_job *s_job, *tmp; - int r; /* - * Locking the list is not required here as the sched thread is parked - * so no new jobs are being inserted or removed. Also concurrent - * GPU recovers can't run in parallel. + * Locking the list is not required here as the scheduler's work items + * are currently not running, so no new jobs are being inserted or + * removed. Also concurrent GPU recovers can't run in parallel. */ list_for_each_entry_safe(s_job, tmp, &sched->pending_list, list) { struct dma_fence *fence = s_job->s_fence->parent; - atomic_inc(&sched->hw_rq_count); + atomic_add(s_job->credits, &sched->credit_count); - if (!full_recovery) + if (!fence) { + drm_sched_job_done(s_job, errno ?: -ECANCELED); continue; + } - if (fence) { - r = dma_fence_add_callback(fence, &s_job->cb, - drm_sched_job_done_cb); - if (r == -ENOENT) - drm_sched_job_done(s_job); - else if (r) - DRM_ERROR("fence add callback failed (%d)\n", - r); - } else - drm_sched_job_done(s_job); - } - - if (full_recovery) { - spin_lock(&sched->job_list_lock); - drm_sched_start_timeout(sched); - spin_unlock(&sched->job_list_lock); + if (dma_fence_add_callback(fence, &s_job->cb, + drm_sched_job_done_cb)) + drm_sched_job_done(s_job, fence->error ?: errno); } - kthread_unpark(sched->thread); + drm_sched_start_timeout_unlocked(sched); + drm_sched_wqueue_start(sched); } EXPORT_SYMBOL(drm_sched_start); /** - * drm_sched_resubmit_jobs - helper to relaunch jobs from the pending list + * drm_sched_resubmit_jobs - Deprecated, don't use in new code! * * @sched: scheduler instance * - */ -void drm_sched_resubmit_jobs(struct drm_gpu_scheduler *sched) -{ - drm_sched_resubmit_jobs_ext(sched, INT_MAX); -} -EXPORT_SYMBOL(drm_sched_resubmit_jobs); - -/** - * drm_sched_resubmit_jobs_ext - helper to relunch certain number of jobs from mirror ring list + * Re-submitting jobs was a concept AMD came up as cheap way to implement + * recovery after a job timeout. * - * @sched: scheduler instance - * @max: job numbers to relaunch + * This turned out to be not working very well. First of all there are many + * problem with the dma_fence implementation and requirements. Either the + * implementation is risking deadlocks with core memory management or violating + * documented implementation details of the dma_fence object. * + * Drivers can still save and restore their state for recovery operations, but + * we shouldn't make this a general scheduler feature around the dma_fence + * interface. */ -void drm_sched_resubmit_jobs_ext(struct drm_gpu_scheduler *sched, int max) +void drm_sched_resubmit_jobs(struct drm_gpu_scheduler *sched) { struct drm_sched_job *s_job, *tmp; uint64_t guilty_context; bool found_guilty = false; struct dma_fence *fence; - int i = 0; list_for_each_entry_safe(s_job, tmp, &sched->pending_list, list) { struct drm_sched_fence *s_fence = s_job->s_fence; - if (i >= max) - break; - if (!found_guilty && atomic_read(&s_job->karma) > sched->hang_limit) { found_guilty = true; guilty_context = s_job->s_fence->scheduled.context; @@ -546,9 +749,7 @@ void drm_sched_resubmit_jobs_ext(struct drm_gpu_scheduler *sched, int max) if (found_guilty && s_job->s_fence->scheduled.context == guilty_context) dma_fence_set_error(&s_fence->finished, -ECANCELED); - dma_fence_put(s_job->s_fence->parent); fence = sched->ops->run_job(s_job); - i++; if (IS_ERR_OR_NULL(fence)) { if (IS_ERR(fence)) @@ -556,85 +757,327 @@ void drm_sched_resubmit_jobs_ext(struct drm_gpu_scheduler *sched, int max) s_job->s_fence->parent = NULL; } else { - s_job->s_fence->parent = fence; + + s_job->s_fence->parent = dma_fence_get(fence); + + /* Drop for orignal kref_init */ + dma_fence_put(fence); } } } -EXPORT_SYMBOL(drm_sched_resubmit_jobs_ext); +EXPORT_SYMBOL(drm_sched_resubmit_jobs); /** * drm_sched_job_init - init a scheduler job - * * @job: scheduler job to init * @entity: scheduler entity to use + * @credits: the number of credits this job contributes to the schedulers + * credit limit * @owner: job owner for debugging + * @drm_client_id: &struct drm_file.client_id of the owner (used by trace + * events) * * Refer to drm_sched_entity_push_job() documentation * for locking considerations. * + * Drivers must make sure drm_sched_job_cleanup() if this function returns + * successfully, even when @job is aborted before drm_sched_job_arm() is called. + * + * Note that this function does not assign a valid value to each struct member + * of struct drm_sched_job. Take a look at that struct's documentation to see + * who sets which struct member with what lifetime. + * + * WARNING: amdgpu abuses &drm_sched.ready to signal when the hardware + * has died, which can mean that there's no valid runqueue for a @entity. + * This function returns -ENOENT in this case (which probably should be -EIO as + * a more meanigful return value). + * * Returns 0 for success, negative error code otherwise. */ int drm_sched_job_init(struct drm_sched_job *job, struct drm_sched_entity *entity, - void *owner) + u32 credits, void *owner, + uint64_t drm_client_id) { - struct drm_gpu_scheduler *sched; - - drm_sched_entity_select_rq(entity); - if (!entity->rq) + if (!entity->rq) { + /* This will most likely be followed by missing frames + * or worse--a blank screen--leave a trail in the + * logs, so this can be debugged easier. + */ + dev_err(job->sched->dev, "%s: entity has no rq!\n", __func__); return -ENOENT; + } - sched = entity->rq->sched; + if (unlikely(!credits)) { + pr_err("*ERROR* %s: credits cannot be 0!\n", __func__); + return -EINVAL; + } + + /* + * We don't know for sure how the user has allocated. Thus, zero the + * struct so that unallowed (i.e., too early) usage of pointers that + * this function does not set is guaranteed to lead to a NULL pointer + * exception instead of UB. + */ + memset(job, 0, sizeof(*job)); - job->sched = sched; job->entity = entity; - job->s_priority = entity->rq - sched->sched_rq; - job->s_fence = drm_sched_fence_create(entity, owner); + job->credits = credits; + job->s_fence = drm_sched_fence_alloc(entity, owner, drm_client_id); if (!job->s_fence) return -ENOMEM; - job->id = atomic64_inc_return(&sched->job_id_count); INIT_LIST_HEAD(&job->list); + xa_init_flags(&job->dependencies, XA_FLAGS_ALLOC); + return 0; } EXPORT_SYMBOL(drm_sched_job_init); /** - * drm_sched_job_cleanup - clean up scheduler job resources + * drm_sched_job_arm - arm a scheduler job for execution + * @job: scheduler job to arm * - * @job: scheduler job to clean up + * This arms a scheduler job for execution. Specifically it initializes the + * &drm_sched_job.s_fence of @job, so that it can be attached to struct dma_resv + * or other places that need to track the completion of this job. It also + * initializes sequence numbers, which are fundamental for fence ordering. + * + * Refer to drm_sched_entity_push_job() documentation for locking + * considerations. + * + * Once this function was called, you *must* submit @job with + * drm_sched_entity_push_job(). + * + * This can only be called if drm_sched_job_init() succeeded. */ -void drm_sched_job_cleanup(struct drm_sched_job *job) +void drm_sched_job_arm(struct drm_sched_job *job) { - dma_fence_put(&job->s_fence->finished); - job->s_fence = NULL; + struct drm_gpu_scheduler *sched; + struct drm_sched_entity *entity = job->entity; + + BUG_ON(!entity); + drm_sched_entity_select_rq(entity); + sched = entity->rq->sched; + + job->sched = sched; + job->s_priority = entity->priority; + + drm_sched_fence_init(job->s_fence, job->entity); } -EXPORT_SYMBOL(drm_sched_job_cleanup); +EXPORT_SYMBOL(drm_sched_job_arm); /** - * drm_sched_ready - is the scheduler ready + * drm_sched_job_add_dependency - adds the fence as a job dependency + * @job: scheduler job to add the dependencies to + * @fence: the dma_fence to add to the list of dependencies. * - * @sched: scheduler instance + * Note that @fence is consumed in both the success and error cases. + * + * Returns: + * 0 on success, or an error on failing to expand the array. + */ +int drm_sched_job_add_dependency(struct drm_sched_job *job, + struct dma_fence *fence) +{ + struct dma_fence *entry; + unsigned long index; + u32 id = 0; + int ret; + + if (!fence) + return 0; + + /* Deduplicate if we already depend on a fence from the same context. + * This lets the size of the array of deps scale with the number of + * engines involved, rather than the number of BOs. + */ + xa_for_each(&job->dependencies, index, entry) { + if (entry->context != fence->context) + continue; + + if (dma_fence_is_later(fence, entry)) { + dma_fence_put(entry); + xa_store(&job->dependencies, index, fence, GFP_KERNEL); + } else { + dma_fence_put(fence); + } + return 0; + } + + ret = xa_alloc(&job->dependencies, &id, fence, xa_limit_32b, GFP_KERNEL); + if (ret != 0) + dma_fence_put(fence); + + return ret; +} +EXPORT_SYMBOL(drm_sched_job_add_dependency); + +/** + * drm_sched_job_add_syncobj_dependency - adds a syncobj's fence as a job dependency + * @job: scheduler job to add the dependencies to + * @file: drm file private pointer + * @handle: syncobj handle to lookup + * @point: timeline point + * + * This adds the fence matching the given syncobj to @job. + * + * Returns: + * 0 on success, or an error on failing to expand the array. + */ +int drm_sched_job_add_syncobj_dependency(struct drm_sched_job *job, + struct drm_file *file, + u32 handle, + u32 point) +{ + struct dma_fence *fence; + int ret; + + ret = drm_syncobj_find_fence(file, handle, point, 0, &fence); + if (ret) + return ret; + + return drm_sched_job_add_dependency(job, fence); +} +EXPORT_SYMBOL(drm_sched_job_add_syncobj_dependency); + +/** + * drm_sched_job_add_resv_dependencies - add all fences from the resv to the job + * @job: scheduler job to add the dependencies to + * @resv: the dma_resv object to get the fences from + * @usage: the dma_resv_usage to use to filter the fences + * + * This adds all fences matching the given usage from @resv to @job. + * Must be called with the @resv lock held. + * + * Returns: + * 0 on success, or an error on failing to expand the array. + */ +int drm_sched_job_add_resv_dependencies(struct drm_sched_job *job, + struct dma_resv *resv, + enum dma_resv_usage usage) +{ + struct dma_resv_iter cursor; + struct dma_fence *fence; + int ret; + + dma_resv_assert_held(resv); + + dma_resv_for_each_fence(&cursor, resv, usage, fence) { + /* + * As drm_sched_job_add_dependency always consumes the fence + * reference (even when it fails), and dma_resv_for_each_fence + * is not obtaining one, we need to grab one before calling. + */ + ret = drm_sched_job_add_dependency(job, dma_fence_get(fence)); + if (ret) + return ret; + } + return 0; +} +EXPORT_SYMBOL(drm_sched_job_add_resv_dependencies); + +/** + * drm_sched_job_add_implicit_dependencies - adds implicit dependencies as job + * dependencies + * @job: scheduler job to add the dependencies to + * @obj: the gem object to add new dependencies from. + * @write: whether the job might write the object (so we need to depend on + * shared fences in the reservation object). * - * Return true if we can push more jobs to the hw, otherwise false. + * 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. + * + * Returns: + * 0 on success, or an error on failing to expand the array. */ -static bool drm_sched_ready(struct drm_gpu_scheduler *sched) +int drm_sched_job_add_implicit_dependencies(struct drm_sched_job *job, + struct drm_gem_object *obj, + bool write) { - return atomic_read(&sched->hw_rq_count) < - sched->hw_submission_limit; + return drm_sched_job_add_resv_dependencies(job, obj->resv, + dma_resv_usage_rw(write)); } +EXPORT_SYMBOL(drm_sched_job_add_implicit_dependencies); /** - * drm_sched_wakeup - Wake up the scheduler when it is ready + * drm_sched_job_has_dependency - check whether fence is the job's dependency + * @job: scheduler job to check + * @fence: fence to look for * + * Returns: + * True if @fence is found within the job's dependencies, or otherwise false. + */ +bool drm_sched_job_has_dependency(struct drm_sched_job *job, + struct dma_fence *fence) +{ + struct dma_fence *f; + unsigned long index; + + xa_for_each(&job->dependencies, index, f) { + if (f == fence) + return true; + } + + return false; +} +EXPORT_SYMBOL(drm_sched_job_has_dependency); + +/** + * drm_sched_job_cleanup - clean up scheduler job resources + * @job: scheduler job to clean up + * + * Cleans up the resources allocated with drm_sched_job_init(). + * + * Drivers should call this from their error unwind code if @job is aborted + * before drm_sched_job_arm() is called. + * + * drm_sched_job_arm() is a point of no return since it initializes the fences + * and their sequence number etc. Once that function has been called, you *must* + * submit it with drm_sched_entity_push_job() and cannot simply abort it by + * calling drm_sched_job_cleanup(). + * + * This function should be called in the &drm_sched_backend_ops.free_job callback. + */ +void drm_sched_job_cleanup(struct drm_sched_job *job) +{ + struct dma_fence *fence; + unsigned long index; + + if (kref_read(&job->s_fence->finished.refcount)) { + /* The job has been processed by the scheduler, i.e., + * drm_sched_job_arm() and drm_sched_entity_push_job() have + * been called. + */ + dma_fence_put(&job->s_fence->finished); + } else { + /* The job was aborted before it has been committed to be run; + * notably, drm_sched_job_arm() has not been called. + */ + drm_sched_fence_free(job->s_fence); + } + + job->s_fence = NULL; + + xa_for_each(&job->dependencies, index, fence) { + dma_fence_put(fence); + } + xa_destroy(&job->dependencies); + +} +EXPORT_SYMBOL(drm_sched_job_cleanup); + +/** + * drm_sched_wakeup - Wake up the scheduler if it is ready to queue * @sched: scheduler instance * + * Wake up the scheduler if we can queue jobs. */ void drm_sched_wakeup(struct drm_gpu_scheduler *sched) { - if (drm_sched_ready(sched)) - wake_up_interruptible(&sched->wake_up_worker); + drm_sched_run_job_queue(sched); } /** @@ -642,7 +1085,11 @@ void drm_sched_wakeup(struct drm_gpu_scheduler *sched) * * @sched: scheduler instance * - * Returns the entity to process or NULL if none are found. + * Return an entity to process or NULL if none are found. + * + * Note, that we break out of the for-loop when "entity" is non-null, which can + * also be an error-pointer--this assures we don't process lower priority + * run-queues. See comments in the respectively called functions. */ static struct drm_sched_entity * drm_sched_select_entity(struct drm_gpu_scheduler *sched) @@ -650,60 +1097,64 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched) struct drm_sched_entity *entity; int i; - if (!drm_sched_ready(sched)) - return NULL; - - /* Kernel run queue has higher priority than normal run queue*/ - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) { - entity = drm_sched_rq_select_entity(&sched->sched_rq[i]); + /* Start with the highest priority. + */ + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { + entity = drm_sched_policy == DRM_SCHED_POLICY_FIFO ? + drm_sched_rq_select_entity_fifo(sched, sched->sched_rq[i]) : + drm_sched_rq_select_entity_rr(sched, sched->sched_rq[i]); if (entity) break; } - return entity; + return IS_ERR(entity) ? NULL : entity; } /** - * drm_sched_get_cleanup_job - fetch the next finished job to be destroyed + * drm_sched_get_finished_job - fetch the next finished job to be destroyed * * @sched: scheduler instance + * @have_more: are there more finished jobs on the list + * + * Informs the caller through @have_more whether there are more finished jobs + * besides the returned one. * * Returns the next finished job from the pending list (if there is one) * ready for it to be destroyed. */ static struct drm_sched_job * -drm_sched_get_cleanup_job(struct drm_gpu_scheduler *sched) +drm_sched_get_finished_job(struct drm_gpu_scheduler *sched, bool *have_more) { struct drm_sched_job *job, *next; - /* - * Don't destroy jobs while the timeout worker is running OR thread - * is being parked and hence assumed to not touch pending_list - */ - if ((sched->timeout != MAX_SCHEDULE_TIMEOUT && - !cancel_delayed_work(&sched->work_tdr)) || - kthread_should_park()) - return NULL; - spin_lock(&sched->job_list_lock); job = list_first_entry_or_null(&sched->pending_list, struct drm_sched_job, list); - if (job && dma_fence_is_signaled(&job->s_fence->finished)) { /* remove job from pending_list */ list_del_init(&job->list); - /* make the scheduled timestamp more accurate */ + + /* cancel this job's TO timer */ + cancel_delayed_work(&sched->work_tdr); + + *have_more = false; next = list_first_entry_or_null(&sched->pending_list, typeof(*next), list); - if (next) - next->s_fence->scheduled.timestamp = - job->s_fence->finished.timestamp; + if (next) { + /* make the scheduled timestamp more accurate */ + if (test_bit(DMA_FENCE_FLAG_TIMESTAMP_BIT, + &next->s_fence->scheduled.flags)) + next->s_fence->scheduled.timestamp = + dma_fence_timestamp(&job->s_fence->finished); + *have_more = dma_fence_is_signaled(&next->s_fence->finished); + + /* start TO timer for next job */ + drm_sched_start_timeout(sched); + } } else { job = NULL; - /* queue timeout for next job */ - drm_sched_start_timeout(sched); } spin_unlock(&sched->job_list_lock); @@ -748,178 +1199,264 @@ drm_sched_pick_best(struct drm_gpu_scheduler **sched_list, EXPORT_SYMBOL(drm_sched_pick_best); /** - * drm_sched_blocked - check if the scheduler is blocked + * drm_sched_free_job_work - worker to call free_job * - * @sched: scheduler instance - * - * Returns true if blocked, otherwise false. + * @w: free job work */ -static bool drm_sched_blocked(struct drm_gpu_scheduler *sched) +static void drm_sched_free_job_work(struct work_struct *w) { - if (kthread_should_park()) { - kthread_parkme(); - return true; + struct drm_gpu_scheduler *sched = + container_of(w, struct drm_gpu_scheduler, work_free_job); + struct drm_sched_job *job; + bool have_more; + + job = drm_sched_get_finished_job(sched, &have_more); + if (job) { + sched->ops->free_job(job); + if (have_more) + drm_sched_run_free_queue(sched); } - return false; + drm_sched_run_job_queue(sched); } /** - * drm_sched_main - main scheduler thread - * - * @param: scheduler instance + * drm_sched_run_job_work - worker to call run_job * - * Returns 0. + * @w: run job work */ -static int drm_sched_main(void *param) +static void drm_sched_run_job_work(struct work_struct *w) { - struct drm_gpu_scheduler *sched = (struct drm_gpu_scheduler *)param; + struct drm_gpu_scheduler *sched = + container_of(w, struct drm_gpu_scheduler, work_run_job); + struct drm_sched_entity *entity; + struct dma_fence *fence; + struct drm_sched_fence *s_fence; + struct drm_sched_job *sched_job; int r; - sched_set_fifo_low(current); - - while (!kthread_should_stop()) { - struct drm_sched_entity *entity = NULL; - struct drm_sched_fence *s_fence; - struct drm_sched_job *sched_job; - struct dma_fence *fence; - struct drm_sched_job *cleanup_job = NULL; - - wait_event_interruptible(sched->wake_up_worker, - (cleanup_job = drm_sched_get_cleanup_job(sched)) || - (!drm_sched_blocked(sched) && - (entity = drm_sched_select_entity(sched))) || - kthread_should_stop()); - - if (cleanup_job) { - sched->ops->free_job(cleanup_job); - /* queue timeout for next job */ - drm_sched_start_timeout(sched); - } - - if (!entity) - continue; - - sched_job = drm_sched_entity_pop_job(entity); + /* Find entity with a ready job */ + entity = drm_sched_select_entity(sched); + if (!entity) { + /* + * Either no more work to do, or the next ready job needs more + * credits than the scheduler has currently available. + */ + return; + } - complete(&entity->entity_idle); + sched_job = drm_sched_entity_pop_job(entity); + if (!sched_job) { + complete_all(&entity->entity_idle); + drm_sched_run_job_queue(sched); + return; + } - if (!sched_job) - continue; + s_fence = sched_job->s_fence; - s_fence = sched_job->s_fence; + atomic_add(sched_job->credits, &sched->credit_count); + drm_sched_job_begin(sched_job); - atomic_inc(&sched->hw_rq_count); - drm_sched_job_begin(sched_job); + trace_drm_sched_job_run(sched_job, entity); + /* + * The run_job() callback must by definition return a fence whose + * refcount has been incremented for the scheduler already. + */ + fence = sched->ops->run_job(sched_job); + complete_all(&entity->entity_idle); + drm_sched_fence_scheduled(s_fence, fence); + + if (!IS_ERR_OR_NULL(fence)) { + r = dma_fence_add_callback(fence, &sched_job->cb, + drm_sched_job_done_cb); + if (r == -ENOENT) + drm_sched_job_done(sched_job, fence->error); + else if (r) + DRM_DEV_ERROR(sched->dev, "fence add callback failed (%d)\n", r); + + dma_fence_put(fence); + } else { + drm_sched_job_done(sched_job, IS_ERR(fence) ? + PTR_ERR(fence) : 0); + } - trace_drm_run_job(sched_job, entity); - fence = sched->ops->run_job(sched_job); - drm_sched_fence_scheduled(s_fence); + wake_up(&sched->job_scheduled); + drm_sched_run_job_queue(sched); +} - if (!IS_ERR_OR_NULL(fence)) { - s_fence->parent = dma_fence_get(fence); - r = dma_fence_add_callback(fence, &sched_job->cb, - drm_sched_job_done_cb); - if (r == -ENOENT) - drm_sched_job_done(sched_job); - else if (r) - DRM_ERROR("fence add callback failed (%d)\n", - r); - dma_fence_put(fence); - } else { - if (IS_ERR(fence)) - dma_fence_set_error(&s_fence->finished, PTR_ERR(fence)); +static struct workqueue_struct *drm_sched_alloc_wq(const char *name) +{ +#if (IS_ENABLED(CONFIG_LOCKDEP)) + static struct lockdep_map map = { + .name = "drm_sched_lockdep_map" + }; - drm_sched_job_done(sched_job); - } + /* + * Avoid leaking a lockdep map on each drm sched creation and + * destruction by using a single lockdep map for all drm sched + * allocated submit_wq. + */ - wake_up(&sched->job_scheduled); - } - return 0; + return alloc_ordered_workqueue_lockdep_map(name, WQ_MEM_RECLAIM, &map); +#else + return alloc_ordered_workqueue(name, WQ_MEM_RECLAIM); +#endif } /** * drm_sched_init - Init a gpu scheduler instance * * @sched: scheduler instance - * @ops: backend operations for this scheduler - * @hw_submission: number of hw submissions that can be in flight - * @hang_limit: number of times to allow a job to hang before dropping it - * @timeout: timeout value in jiffies for the scheduler - * @score: optional score atomic shared with other schedulers - * @name: name used for debugging + * @args: scheduler initialization arguments * * Return 0 on success, otherwise error code. */ -int drm_sched_init(struct drm_gpu_scheduler *sched, - const struct drm_sched_backend_ops *ops, - unsigned hw_submission, unsigned hang_limit, long timeout, - atomic_t *score, const char *name) -{ - int i, ret; - sched->ops = ops; - sched->hw_submission_limit = hw_submission; - sched->name = name; - sched->timeout = timeout; - sched->hang_limit = hang_limit; - sched->score = score ? score : &sched->_score; - for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_COUNT; i++) - drm_sched_rq_init(sched, &sched->sched_rq[i]); - - init_waitqueue_head(&sched->wake_up_worker); +int drm_sched_init(struct drm_gpu_scheduler *sched, const struct drm_sched_init_args *args) +{ + int i; + + sched->ops = args->ops; + sched->credit_limit = args->credit_limit; + sched->name = args->name; + sched->timeout = args->timeout; + sched->hang_limit = args->hang_limit; + sched->timeout_wq = args->timeout_wq ? args->timeout_wq : system_percpu_wq; + sched->score = args->score ? args->score : &sched->_score; + sched->dev = args->dev; + + if (args->num_rqs > DRM_SCHED_PRIORITY_COUNT) { + /* This is a gross violation--tell drivers what the problem is. + */ + dev_err(sched->dev, "%s: num_rqs cannot be greater than DRM_SCHED_PRIORITY_COUNT\n", + __func__); + return -EINVAL; + } else if (sched->sched_rq) { + /* Not an error, but warn anyway so drivers can + * fine-tune their DRM calling order, and return all + * is good. + */ + dev_warn(sched->dev, "%s: scheduler already initialized!\n", __func__); + return 0; + } + + if (args->submit_wq) { + sched->submit_wq = args->submit_wq; + sched->own_submit_wq = false; + } else { + sched->submit_wq = drm_sched_alloc_wq(args->name); + if (!sched->submit_wq) + return -ENOMEM; + + sched->own_submit_wq = true; + } + + sched->sched_rq = kmalloc_array(args->num_rqs, sizeof(*sched->sched_rq), + GFP_KERNEL | __GFP_ZERO); + if (!sched->sched_rq) + goto Out_check_own; + sched->num_rqs = args->num_rqs; + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { + sched->sched_rq[i] = kzalloc(sizeof(*sched->sched_rq[i]), GFP_KERNEL); + if (!sched->sched_rq[i]) + goto Out_unroll; + drm_sched_rq_init(sched, sched->sched_rq[i]); + } + init_waitqueue_head(&sched->job_scheduled); INIT_LIST_HEAD(&sched->pending_list); spin_lock_init(&sched->job_list_lock); - atomic_set(&sched->hw_rq_count, 0); + atomic_set(&sched->credit_count, 0); INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout); + INIT_WORK(&sched->work_run_job, drm_sched_run_job_work); + INIT_WORK(&sched->work_free_job, drm_sched_free_job_work); atomic_set(&sched->_score, 0); atomic64_set(&sched->job_id_count, 0); - - /* Each scheduler will run on a seperate kernel thread */ - sched->thread = kthread_run(drm_sched_main, sched, sched->name); - if (IS_ERR(sched->thread)) { - ret = PTR_ERR(sched->thread); - sched->thread = NULL; - DRM_ERROR("Failed to create scheduler for %s.\n", name); - return ret; - } + sched->pause_submit = false; sched->ready = true; return 0; +Out_unroll: + for (--i ; i >= DRM_SCHED_PRIORITY_KERNEL; i--) + kfree(sched->sched_rq[i]); + + kfree(sched->sched_rq); + sched->sched_rq = NULL; +Out_check_own: + if (sched->own_submit_wq) + destroy_workqueue(sched->submit_wq); + dev_err(sched->dev, "%s: Failed to setup GPU scheduler--out of memory\n", __func__); + return -ENOMEM; } EXPORT_SYMBOL(drm_sched_init); +static void drm_sched_cancel_remaining_jobs(struct drm_gpu_scheduler *sched) +{ + struct drm_sched_job *job, *tmp; + + /* All other accessors are stopped. No locking necessary. */ + list_for_each_entry_safe_reverse(job, tmp, &sched->pending_list, list) { + sched->ops->cancel_job(job); + list_del(&job->list); + sched->ops->free_job(job); + } +} + /** * drm_sched_fini - Destroy a gpu scheduler * * @sched: scheduler instance * * Tears down and cleans up the scheduler. + * + * This stops submission of new jobs to the hardware through &struct + * drm_sched_backend_ops.run_job. If &struct drm_sched_backend_ops.cancel_job + * is implemented, all jobs will be canceled through it and afterwards cleaned + * up through &struct drm_sched_backend_ops.free_job. If cancel_job is not + * implemented, memory could leak. */ void drm_sched_fini(struct drm_gpu_scheduler *sched) { struct drm_sched_entity *s_entity; int i; - if (sched->thread) - kthread_stop(sched->thread); - - for (i = DRM_SCHED_PRIORITY_COUNT - 1; i >= DRM_SCHED_PRIORITY_MIN; i--) { - struct drm_sched_rq *rq = &sched->sched_rq[i]; + drm_sched_wqueue_stop(sched); - if (!rq) - continue; + for (i = DRM_SCHED_PRIORITY_KERNEL; i < sched->num_rqs; i++) { + struct drm_sched_rq *rq = sched->sched_rq[i]; spin_lock(&rq->lock); - list_for_each_entry(s_entity, &rq->entities, list) + list_for_each_entry(s_entity, &rq->entities, list) { /* * Prevents reinsertion and marks job_queue as idle, - * it will removed from rq in drm_sched_entity_fini + * it will be removed from the rq in drm_sched_entity_fini() * eventually + * + * FIXME: + * This lacks the proper spin_lock(&s_entity->lock) and + * is, therefore, a race condition. Most notably, it + * can race with drm_sched_entity_push_job(). The lock + * cannot be taken here, however, because this would + * lead to lock inversion -> deadlock. + * + * The best solution probably is to enforce the life + * time rule of all entities having to be torn down + * before their scheduler. Then, however, locking could + * be dropped alltogether from this function. + * + * For now, this remains a potential race in all + * drivers that keep entities alive for longer than + * the scheduler. + * + * The READ_ONCE() is there to make the lockless read + * (warning about the lockless write below) slightly + * less broken... */ + if (!READ_ONCE(s_entity->stopped)) + dev_warn(sched->dev, "Tearing down scheduler with active entities!\n"); s_entity->stopped = true; + } spin_unlock(&rq->lock); - + kfree(sched->sched_rq[i]); } /* Wakeup everyone stuck in drm_sched_entity_flush for this scheduler */ @@ -928,18 +1465,31 @@ void drm_sched_fini(struct drm_gpu_scheduler *sched) /* Confirm no work left behind accessing device structures */ cancel_delayed_work_sync(&sched->work_tdr); + /* Avoid memory leaks if supported by the driver. */ + if (sched->ops->cancel_job) + drm_sched_cancel_remaining_jobs(sched); + + if (sched->own_submit_wq) + destroy_workqueue(sched->submit_wq); sched->ready = false; + kfree(sched->sched_rq); + sched->sched_rq = NULL; + + if (!list_empty(&sched->pending_list)) + dev_warn(sched->dev, "Tearing down scheduler while jobs are pending!\n"); } EXPORT_SYMBOL(drm_sched_fini); /** - * drm_sched_increase_karma_ext - Update sched_entity guilty flag + * drm_sched_increase_karma - Update sched_entity guilty flag * * @bad: The job guilty of time out - * @type: type for increase/reset karma * + * Increment on every hang caused by the 'bad' job. If this exceeds the hang + * limit of the scheduler then the respective sched entity is marked guilty and + * jobs from it will not be scheduled further */ -void drm_sched_increase_karma_ext(struct drm_sched_job *bad, int type) +void drm_sched_increase_karma(struct drm_sched_job *bad) { int i; struct drm_sched_entity *tmp; @@ -951,21 +1501,17 @@ void drm_sched_increase_karma_ext(struct drm_sched_job *bad, int type) * corrupt but keep in mind that kernel jobs always considered good. */ if (bad->s_priority != DRM_SCHED_PRIORITY_KERNEL) { - if (type == 0) - atomic_set(&bad->karma, 0); - else if (type == 1) - atomic_inc(&bad->karma); + atomic_inc(&bad->karma); - for (i = DRM_SCHED_PRIORITY_MIN; i < DRM_SCHED_PRIORITY_KERNEL; - i++) { - struct drm_sched_rq *rq = &sched->sched_rq[i]; + for (i = DRM_SCHED_PRIORITY_HIGH; i < sched->num_rqs; i++) { + struct drm_sched_rq *rq = sched->sched_rq[i]; spin_lock(&rq->lock); list_for_each_entry_safe(entity, tmp, &rq->entities, list) { if (bad->s_fence->scheduled.context == entity->fence_context) { if (entity->guilty) - atomic_set(entity->guilty, type); + atomic_set(entity->guilty, 1); break; } } @@ -975,4 +1521,49 @@ void drm_sched_increase_karma_ext(struct drm_sched_job *bad, int type) } } } -EXPORT_SYMBOL(drm_sched_increase_karma_ext); +EXPORT_SYMBOL(drm_sched_increase_karma); + +/** + * drm_sched_wqueue_ready - Is the scheduler ready for submission + * + * @sched: scheduler instance + * + * Returns true if submission is ready + */ +bool drm_sched_wqueue_ready(struct drm_gpu_scheduler *sched) +{ + return sched->ready; +} +EXPORT_SYMBOL(drm_sched_wqueue_ready); + +/** + * drm_sched_wqueue_stop - stop scheduler submission + * @sched: scheduler instance + * + * Stops the scheduler from pulling new jobs from entities. It also stops + * freeing jobs automatically through drm_sched_backend_ops.free_job(). + */ +void drm_sched_wqueue_stop(struct drm_gpu_scheduler *sched) +{ + WRITE_ONCE(sched->pause_submit, true); + cancel_work_sync(&sched->work_run_job); + cancel_work_sync(&sched->work_free_job); +} +EXPORT_SYMBOL(drm_sched_wqueue_stop); + +/** + * drm_sched_wqueue_start - start scheduler submission + * @sched: scheduler instance + * + * Restarts the scheduler after drm_sched_wqueue_stop() has stopped it. + * + * This function is not necessary for 'conventional' startup. The scheduler is + * fully operational after drm_sched_init() succeeded. + */ +void drm_sched_wqueue_start(struct drm_gpu_scheduler *sched) +{ + WRITE_ONCE(sched->pause_submit, false); + queue_work(sched->submit_wq, &sched->work_run_job); + queue_work(sched->submit_wq, &sched->work_free_job); +} +EXPORT_SYMBOL(drm_sched_wqueue_start); |