On 2020-12-04 3:16 a.m., Christian König wrote:
> Am 04.12.20 um 04:17 schrieb Luben Tuikov:
>> The drm_sched_job_done() callback now moves done
>> jobs from the pending list to a "done" list.
>>
>> In drm_sched_job_timeout, make use of the status
>> returned by a GPU driver job timeout handler to
>> decide whether to leave the oldest job in the
>> pending list, or to send it off to the done list.
>> If a driver's job timeout callback returns a
>> status that that job is done, it is added to the
>> done list and the done thread woken up. If that
>> job needs more time, it is left on the pending
>> list and the timeout timer restarted.
>>
>> The idea is that a GPU driver can check the IP to
>> which the passed-in job belongs to and determine
>> whether the IP is alive and well, or if it needs
>> more time to complete this job and perhaps others
>> also executing on it.
>>
>> In drm_sched_job_timeout(), the main scheduler
>> thread is now parked, before calling a driver's
>> timeout_job callback, so as to not compete pushing
>> jobs down to the GPU while the recovery method is
>> taking place.
>>
>> Eliminate the polling mechanism of picking out done
>> jobs from the pending list, i.e. eliminate
>> drm_sched_get_cleanup_job().
>>
>> This also eliminates the eldest job disappearing
>> from the pending list, while the driver timeout
>> handler is called.
>>
>> Various other optimizations to the GPU scheduler
>> and job recovery are possible with this format.
>>
>> Signed-off-by: Luben Tuikov <luben.tuikov@xxxxxxx>
>>
>> Cc: Alexander Deucher <Alexander.Deucher@xxxxxxx>
>> Cc: Andrey Grodzovsky <Andrey.Grodzovsky@xxxxxxx>
>> Cc: Christian König <christian.koenig@xxxxxxx>
>> Cc: Daniel Vetter <daniel.vetter@xxxxxxxx>
>> Cc: Lucas Stach <l.stach@xxxxxxxxxxxxxx>
>> Cc: Russell King <linux+etnaviv@xxxxxxxxxxxxxxx>
>> Cc: Christian Gmeiner <christian.gmeiner@xxxxxxxxx>
>> Cc: Qiang Yu <yuq825@xxxxxxxxx>
>> Cc: Rob Herring <robh@xxxxxxxxxx>
>> Cc: Tomeu Vizoso <tomeu.vizoso@xxxxxxxxxxxxx>
>> Cc: Steven Price <steven.price@xxxxxxx>
>> Cc: Alyssa Rosenzweig <alyssa.rosenzweig@xxxxxxxxxxxxx>
>> Cc: Eric Anholt <eric@xxxxxxxxxx>
>>
>> v2: Dispell using a done thread, so as to keep
>> the cache hot on the same processor.
>> ---
>> drivers/gpu/drm/scheduler/sched_main.c | 247 +++++++++++++------------
>> include/drm/gpu_scheduler.h | 4 +
>> 2 files changed, 134 insertions(+), 117 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/scheduler/sched_main.c b/drivers/gpu/drm/scheduler/sched_main.c
>> index b9876cad94f2..d77180b44998 100644
>> --- a/drivers/gpu/drm/scheduler/sched_main.c
>> +++ b/drivers/gpu/drm/scheduler/sched_main.c
>> @@ -164,7 +164,9 @@ drm_sched_rq_select_entity(struct drm_sched_rq *rq)
>> * 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.
>> + * Move the completed task to the done list,
>> + * signal the its fence to mark it finished,
>> + * and wake up the worker thread.
>> */
>> static void drm_sched_job_done(struct drm_sched_job *s_job)
>> {
>> @@ -176,9 +178,14 @@ static void drm_sched_job_done(struct drm_sched_job *s_job)
>>
>> trace_drm_sched_process_job(s_fence);
>>
>> + spin_lock(&sched->job_list_lock);
>> + list_move(&s_job->list, &sched->done_list);
>> + spin_unlock(&sched->job_list_lock);
>> +
>
> That is racy, as soon as the spinlock is dropped the job and with it the
> s_fence might haven been destroyed.
Yeah, I had it the other way around, (the correct way),
and changed it--not sure why. I revert it back. Thanks
for catching this.
Regards,
Luben
>
>> dma_fence_get(&s_fence->finished);
>> drm_sched_fence_finished(s_fence);
>> dma_fence_put(&s_fence->finished);
>
> In other words this here needs to come first.
>
> Regards,
> Christian.
>
>> +
>> wake_up_interruptible(&sched->wake_up_worker);
>> }
>>
>> @@ -309,6 +316,37 @@ static void drm_sched_job_begin(struct drm_sched_job *s_job)
>> spin_unlock(&sched->job_list_lock);
>> }
>>
>> +/** drm_sched_job_timeout -- a timer timeout occurred
>> + * @work: pointer to work_struct
>> + *
>> + * First, park the scheduler thread whose IP timed out,
>> + * so that we don't race with the scheduler thread pushing
>> + * jobs down the IP as we try to investigate what
>> + * happened and give drivers a chance to recover.
>> + *
>> + * Second, take the fist job in the pending list
>> + * (oldest), leave it in the pending list and call the
>> + * driver's timer timeout callback to find out what
>> + * happened, passing this job as the suspect one.
>> + *
>> + * The driver may return DRM_TASK_STATUS COMPLETE,
>> + * which means the task is not in the IP(*) and we move
>> + * it to the done list to free it.
>> + *
>> + * (*) A reason for this would be, say, that the job
>> + * completed in due time, or the driver has aborted
>> + * this job using driver specific methods in the
>> + * timedout_job callback and has now removed it from
>> + * the hardware.
>> + *
>> + * Or, the driver may return DRM_TASK_STATUS_ALIVE, to
>> + * indicate that it had inquired about this job, and it
>> + * has verified that this job is alive and well, and
>> + * that the DRM layer should give this task more time
>> + * to complete. In this case, we restart the timeout timer.
>> + *
>> + * Lastly, we unpark the scheduler thread.
>> + */
>> static void drm_sched_job_timedout(struct work_struct *work)
>> {
>> struct drm_gpu_scheduler *sched;
>> @@ -316,37 +354,32 @@ static void drm_sched_job_timedout(struct work_struct *work)
>>
>> sched = container_of(work, struct drm_gpu_scheduler, work_tdr.work);
>>
>> - /* Protects against concurrent deletion in drm_sched_get_cleanup_job */
>> + kthread_park(sched->thread);
>> +
>> spin_lock(&sched->job_list_lock);
>> job = list_first_entry_or_null(&sched->pending_list,
>> struct drm_sched_job, list);
>> + spin_unlock(&sched->job_list_lock);
>>
>> 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.
>> - */
>> - list_del_init(&job->list);
>> - spin_unlock(&sched->job_list_lock);
>> -
>> - job->sched->ops->timedout_job(job);
>> + int res;
>>
>> - /*
>> - * Guilty job did complete and hence needs to be manually removed
>> - * See drm_sched_stop doc.
>> - */
>> - if (sched->free_guilty) {
>> - job->sched->ops->free_job(job);
>> - sched->free_guilty = false;
>> + res = job->sched->ops->timedout_job(job);
>> + if (res == DRM_TASK_STATUS_COMPLETE) {
>> + /* The job is out of the device.
>> + */
>> + spin_lock(&sched->job_list_lock);
>> + list_move(&job->list, &sched->done_list);
>> + spin_unlock(&sched->job_list_lock);
>> + wake_up_interruptible(&sched->wake_up_worker);
>> + } else {
>> + /* The job needs more time.
>> + */
>> + drm_sched_start_timeout(sched);
>> }
>> - } else {
>> - spin_unlock(&sched->job_list_lock);
>> }
>>
>> - spin_lock(&sched->job_list_lock);
>> - drm_sched_start_timeout(sched);
>> - spin_unlock(&sched->job_list_lock);
>> + kthread_unpark(sched->thread);
>> }
>>
>> /**
>> @@ -413,24 +446,13 @@ void drm_sched_stop(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad)
>> kthread_park(sched->thread);
>>
>> /*
>> - * Reinsert back the bad job here - now it's safe as
>> - * drm_sched_get_cleanup_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.
>> - */
>> - if (bad && bad->sched == sched)
>> - /*
>> - * Add at the head of the queue to reflect it was the earliest
>> - * job extracted.
>> - */
>> - list_add(&bad->list, &sched->pending_list);
>> -
>> - /*
>> - * 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.
>> + * Iterate the pending list in reverse order,
>> + * from most recently submitted to oldest
>> + * tasks. Tasks which haven't completed, leave
>> + * them in the pending list, but decrement
>> + * their hardware run queue count.
>> + * Else, the fence must've signalled, and the job
>> + * is in the done list.
>> */
>> list_for_each_entry_safe_reverse(s_job, tmp, &sched->pending_list,
>> list) {
>> @@ -439,36 +461,52 @@ void drm_sched_stop(struct drm_gpu_scheduler *sched, struct drm_sched_job *bad)
>> &s_job->cb)) {
>> atomic_dec(&sched->hw_rq_count);
>> } else {
>> - /*
>> - * remove job from pending_list.
>> - * Locking here is for concurrent resume timeout
>> - */
>> - spin_lock(&sched->job_list_lock);
>> - list_del_init(&s_job->list);
>> - spin_unlock(&sched->job_list_lock);
>> -
>> - /*
>> - * Wait for job's HW fence callback to finish using s_job
>> - * before releasing it.
>> - *
>> - * Job is still alive so fence refcount at least 1
>> - */
>> - dma_fence_wait(&s_job->s_fence->finished, false);
>> -
>> - /*
>> - * We must keep bad job alive for later use during
>> - * recovery by some of the drivers but leave a hint
>> - * that the guilty job must be released.
>> - */
>> - if (bad != s_job)
>> - sched->ops->free_job(s_job);
>> - else
>> - sched->free_guilty = true;
>> + if (bad == s_job) {
>> + /* This is the oldest job on the pending list
>> + * whose IP timed out. The
>> + * drm_sched_job_timeout() function calls the
>> + * driver's timedout_job callback passing @bad,
>> + * who then calls this function here--as such
>> + * we shouldn't move @bad or free it. This will
>> + * be decided by drm_sched_job_timeout() when
>> + * this function here returns back to the caller
>> + * (the driver) and the driver's timedout_job
>> + * callback returns a result to
>> + * drm_sched_job_timeout().
>> + */
>> + ;
>> + } else {
>> + int res;
>> +
>> + /* This job is not the @bad job passed above.
>> + * Note that perhaps it was *this* job which
>> + * timed out. The wait below is suspect. Since,
>> + * it waits with maximum timeout and "intr" set
>> + * to false, it will either return 0 indicating
>> + * that the fence has signalled, or negative on
>> + * error. What if, the whole IP is stuck and
>> + * this ends up waiting forever?
>> + *
>> + * Wait for job's HW fence callback to finish
>> + * using s_job before releasing it.
>> + *
>> + * Job is still alive so fence
>> + * refcount at least 1
>> + */
>> + res = dma_fence_wait(&s_job->s_fence->finished,
>> + false);
>> +
>> + if (res == 0)
>> + sched->ops->free_job(s_job);
>> + else
>> + pr_err_once("%s: dma_fence_wait: %d\n",
>> + sched->name, res);
>> + }
>> }
>> }
>>
>> /*
>> - * Stop pending timer in flight as we rearm it in drm_sched_start. This
>> + * Stop pending timer in flight as we rearm it in drm_sched_start. This
>> * avoids the pending timeout work in progress to fire right away after
>> * this TDR finished and before the newly restarted jobs had a
>> * chance to complete.
>> @@ -511,8 +549,9 @@ void drm_sched_start(struct drm_gpu_scheduler *sched, bool full_recovery)
>> else if (r)
>> DRM_ERROR("fence add callback failed (%d)\n",
>> r);
>> - } else
>> + } else {
>> drm_sched_job_done(s_job);
>> + }
>> }
>>
>> if (full_recovery) {
>> @@ -665,47 +704,6 @@ drm_sched_select_entity(struct drm_gpu_scheduler *sched)
>> return entity;
>> }
>>
>> -/**
>> - * drm_sched_get_cleanup_job - fetch the next finished job to be destroyed
>> - *
>> - * @sched: scheduler instance
>> - *
>> - * 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)
>> -{
>> - struct drm_sched_job *job;
>> -
>> - /*
>> - * 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);
>> - } else {
>> - job = NULL;
>> - /* queue timeout for next job */
>> - drm_sched_start_timeout(sched);
>> - }
>> -
>> - spin_unlock(&sched->job_list_lock);
>> -
>> - return job;
>> -}
>> -
>> /**
>> * drm_sched_pick_best - Get a drm sched from a sched_list with the least load
>> * @sched_list: list of drm_gpu_schedulers
>> @@ -759,6 +757,25 @@ static bool drm_sched_blocked(struct drm_gpu_scheduler *sched)
>> return false;
>> }
>>
>> +static void drm_sched_free_done(struct drm_gpu_scheduler *sched)
>> +{
>> + LIST_HEAD(done_q);
>> +
>> + spin_lock(&sched->job_list_lock);
>> + list_splice_init(&sched->done_list, &done_q);
>> + spin_unlock(&sched->job_list_lock);
>> +
>> + while (!list_empty(&done_q)) {
>> + struct drm_sched_job *job;
>> +
>> + job = list_first_entry(&done_q,
>> + struct drm_sched_job,
>> + list);
>> + list_del_init(&job->list);
>> + sched->ops->free_job(job);
>> + }
>> +}
>> +
>> /**
>> * drm_sched_main - main scheduler thread
>> *
>> @@ -768,7 +785,7 @@ static bool drm_sched_blocked(struct drm_gpu_scheduler *sched)
>> */
>> static int drm_sched_main(void *param)
>> {
>> - struct drm_gpu_scheduler *sched = (struct drm_gpu_scheduler *)param;
>> + struct drm_gpu_scheduler *sched = param;
>> int r;
>>
>> sched_set_fifo_low(current);
>> @@ -778,19 +795,14 @@ static int drm_sched_main(void *param)
>> 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)) ||
>> + (!list_empty(&sched->done_list)) ||
>> (!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);
>> - }
>> + drm_sched_free_done(sched);
>>
>> if (!entity)
>> continue;
>> @@ -864,6 +876,7 @@ int drm_sched_init(struct drm_gpu_scheduler *sched,
>> init_waitqueue_head(&sched->wake_up_worker);
>> init_waitqueue_head(&sched->job_scheduled);
>> INIT_LIST_HEAD(&sched->pending_list);
>> + INIT_LIST_HEAD(&sched->done_list);
>> spin_lock_init(&sched->job_list_lock);
>> atomic_set(&sched->hw_rq_count, 0);
>> INIT_DELAYED_WORK(&sched->work_tdr, drm_sched_job_timedout);
>> diff --git a/include/drm/gpu_scheduler.h b/include/drm/gpu_scheduler.h
>> index cedfc5394e52..11278695fed0 100644
>> --- a/include/drm/gpu_scheduler.h
>> +++ b/include/drm/gpu_scheduler.h
>> @@ -289,6 +289,7 @@ struct drm_gpu_scheduler {
>> uint32_t hw_submission_limit;
>> long timeout;
>> const char *name;
>> +
>> struct drm_sched_rq sched_rq[DRM_SCHED_PRIORITY_COUNT];
>> wait_queue_head_t wake_up_worker;
>> wait_queue_head_t job_scheduled;
>> @@ -296,8 +297,11 @@ struct drm_gpu_scheduler {
>> atomic64_t job_id_count;
>> struct delayed_work work_tdr;
>> struct task_struct *thread;
>> +
>> struct list_head pending_list;
>> + struct list_head done_list;
>> spinlock_t job_list_lock;
>> +
>> int hang_limit;
>> atomic_t score;
>> bool ready;
>
_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
