[Intel-xe] [PATCH v3 5/8] drm/xe: Long running job update
Matthew Brost
matthew.brost at intel.com
Thu Jun 15 03:22:48 UTC 2023
On Wed, Jun 14, 2023 at 01:40:31PM +0200, Thomas Hellström wrote:
>
> On 6/7/23 18:03, Matthew Brost wrote:
> Some commit message nits:
> > For long running (LR) jobs with the DRM scheduler we must return NULL in
> > run_job which results in signaling the job's finished fence immediately.
> > This prevents LR jobs from creating infinite dma-fences.
> >
> > Signaling job's finished fence immediately breaks flow controling ring
> s/controling/controlling/
> > with the DRM scheduler, to work around this the ring is flow controlled
> > and written in the exec IOCTL.
> Imperative: To work around this, flow control...
> > Signaling job's finished fence
> > immediately also breaks the TDR which is used in reset / cleanup entity
> > paths so we write a new path for LR entities.
> so write
Will fix all of this.
> >
> > v2: Better commit, white space, remove rmb(), better comment next to
> > emit_job()
> >
> > Signed-off-by: Matthew Brost <matthew.brost at intel.com>
> > ---
> > drivers/gpu/drm/xe/xe_engine.c | 32 ++++++++
> > drivers/gpu/drm/xe/xe_engine.h | 4 +
> > drivers/gpu/drm/xe/xe_exec.c | 8 ++
> > drivers/gpu/drm/xe/xe_guc_engine_types.h | 2 +
> > drivers/gpu/drm/xe/xe_guc_submit.c | 93 +++++++++++++++++++++---
> > drivers/gpu/drm/xe/xe_trace.h | 5 ++
> > 6 files changed, 135 insertions(+), 9 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/xe/xe_engine.c b/drivers/gpu/drm/xe/xe_engine.c
> > index b3036c4a8ec3..eb5f66f07da8 100644
> > --- a/drivers/gpu/drm/xe/xe_engine.c
> > +++ b/drivers/gpu/drm/xe/xe_engine.c
> > @@ -18,6 +18,7 @@
> > #include "xe_macros.h"
> > #include "xe_migrate.h"
> > #include "xe_pm.h"
> > +#include "xe_ring_ops_types.h"
> > #include "xe_trace.h"
> > #include "xe_vm.h"
> > @@ -679,6 +680,37 @@ static void engine_kill_compute(struct xe_engine *e)
> > up_write(&e->vm->lock);
> > }
> > +/**
> > + * xe_engine_is_lr() - Whether an engine is long-running
> > + * @e: The engine
> > + *
> > + * Return: True if the engine is long-running, false otherwise.
> > + */
> > +bool xe_engine_is_lr(struct xe_engine *e)
> > +{
> > + return e->vm && xe_vm_no_dma_fences(e->vm) &&
> > + !(e->flags & ENGINE_FLAG_VM);
> > +}
> > +
> > +static s32 xe_engine_num_job_inflight(struct xe_engine *e)
> > +{
> > + return e->lrc->fence_ctx.next_seqno - xe_lrc_seqno(e->lrc) - 1;
> > +}
> > +
> > +/**
> > + * xe_engine_ring_full() - Whether an engine's ring is full
> > + * @e: The engine
> > + *
> > + * Return: True if the engine's ring is full, false otherwise.
> > + */
> > +bool xe_engine_ring_full(struct xe_engine *e)
> > +{
> > + struct xe_lrc *lrc = e->lrc;
> > + s32 max_job = lrc->ring.size / MAX_JOB_SIZE_BYTES;
> > +
> > + return xe_engine_num_job_inflight(e) >= max_job;
> > +}
> > +
> > /**
> > * xe_engine_is_idle() - Whether an engine is idle.
> > * @engine: The engine
> > diff --git a/drivers/gpu/drm/xe/xe_engine.h b/drivers/gpu/drm/xe/xe_engine.h
> > index a49cf2ab405e..2e60f6d90226 100644
> > --- a/drivers/gpu/drm/xe/xe_engine.h
> > +++ b/drivers/gpu/drm/xe/xe_engine.h
> > @@ -42,6 +42,10 @@ static inline bool xe_engine_is_parallel(struct xe_engine *engine)
> > return engine->width > 1;
> > }
> > +bool xe_engine_is_lr(struct xe_engine *e);
> > +
> > +bool xe_engine_ring_full(struct xe_engine *e);
> > +
> > bool xe_engine_is_idle(struct xe_engine *engine);
> > void xe_engine_kill(struct xe_engine *e);
> > diff --git a/drivers/gpu/drm/xe/xe_exec.c b/drivers/gpu/drm/xe/xe_exec.c
> > index e44076ee2e11..9e9523939de8 100644
> > --- a/drivers/gpu/drm/xe/xe_exec.c
> > +++ b/drivers/gpu/drm/xe/xe_exec.c
> > @@ -14,6 +14,7 @@
> > #include "xe_device.h"
> > #include "xe_engine.h"
> > #include "xe_macros.h"
> > +#include "xe_ring_ops_types.h"
> > #include "xe_sched_job.h"
> > #include "xe_sync.h"
> > #include "xe_vm.h"
> > @@ -300,6 +301,11 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
> > goto err_engine_end;
> > }
> > + if (xe_engine_is_lr(engine) && xe_engine_ring_full(engine)) {
> > + err = -EWOULDBLOCK;
> > + goto err_engine_end;
> > + }
> > +
> > job = xe_sched_job_create(engine, xe_engine_is_parallel(engine) ?
> > addresses : &args->address);
> > if (IS_ERR(job)) {
> > @@ -386,6 +392,8 @@ int xe_exec_ioctl(struct drm_device *dev, void *data, struct drm_file *file)
> > xe_sync_entry_signal(&syncs[i], job,
> > &job->drm.s_fence->finished);
> > + if (xe_engine_is_lr(engine))
> > + engine->ring_ops->emit_job(job);
> > xe_sched_job_push(job);
> > xe_vm_reactivate_rebind(vm);
> > diff --git a/drivers/gpu/drm/xe/xe_guc_engine_types.h b/drivers/gpu/drm/xe/xe_guc_engine_types.h
> > index cbfb13026ec1..5d83132034a6 100644
> > --- a/drivers/gpu/drm/xe/xe_guc_engine_types.h
> > +++ b/drivers/gpu/drm/xe/xe_guc_engine_types.h
> > @@ -31,6 +31,8 @@ struct xe_guc_engine {
> > */
> > #define MAX_STATIC_MSG_TYPE 3
> > struct drm_sched_msg static_msgs[MAX_STATIC_MSG_TYPE];
> > + /** @lr_tdr: long running TDR worker */
> > + struct work_struct lr_tdr;
> > /** @fini_async: do final fini async from this worker */
> > struct work_struct fini_async;
> > /** @resume_time: time of last resume */
> > diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> > index a7a0d9f806bd..7bde687fd188 100644
> > --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> > +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> > @@ -482,6 +482,14 @@ static void register_engine(struct xe_engine *e)
> > parallel_write(xe, map, wq_desc.wq_status, WQ_STATUS_ACTIVE);
> > }
> > + /*
> > + * We must keep a reference for LR engines if engine is registered with
> > + * the GuC as jobs signal immediately and can't destroy an engine if the
> > + * GuC has a reference to it.
> > + */
> > + if (xe_engine_is_lr(e))
> > + xe_engine_get(e);
> > +
> > set_engine_registered(e);
> > trace_xe_engine_register(e);
> > if (xe_engine_is_parallel(e))
> > @@ -644,6 +652,7 @@ guc_engine_run_job(struct drm_sched_job *drm_job)
> > {
> > struct xe_sched_job *job = to_xe_sched_job(drm_job);
> > struct xe_engine *e = job->engine;
> > + bool lr = xe_engine_is_lr(e);
> > XE_BUG_ON((engine_destroyed(e) || engine_pending_disable(e)) &&
> > !engine_banned(e) && !engine_suspended(e));
> > @@ -653,14 +662,19 @@ guc_engine_run_job(struct drm_sched_job *drm_job)
> > if (!engine_killed_or_banned(e) && !xe_sched_job_is_error(job)) {
> > if (!engine_registered(e))
> > register_engine(e);
> > - e->ring_ops->emit_job(job);
> > + if (!lr) /* LR jobs are emitted in the exec IOCTL */
> > + e->ring_ops->emit_job(job);
> > submit_engine(e);
> > }
> > - if (test_and_set_bit(JOB_FLAG_SUBMIT, &job->fence->flags))
> > + if (lr) {
> > + xe_sched_job_set_error(job, -EOPNOTSUPP);
> > + return NULL;
> > + } else if (test_and_set_bit(JOB_FLAG_SUBMIT, &job->fence->flags)) {
> > return job->fence;
> > - else
> > + } else {
> > return dma_fence_get(job->fence);
> > + }
> > }
> > static void guc_engine_free_job(struct drm_sched_job *drm_job)
> > @@ -764,6 +778,56 @@ static void simple_error_capture(struct xe_engine *e)
> > }
> > #endif
> > +static void xe_guc_engine_trigger_cleanup(struct xe_engine *e)
> > +{
> > + struct xe_guc *guc = engine_to_guc(e);
> > +
> > + if (xe_engine_is_lr(e))
> > + queue_work(guc_to_gt(guc)->ordered_wq, &e->guc->lr_tdr);
> > + else
> > + drm_sched_set_timeout(&e->guc->sched, MIN_SCHED_TIMEOUT);
> > +}
> > +
> > +static void xe_guc_engine_lr_cleanup(struct work_struct *w)
> > +{
> > + struct xe_guc_engine *ge =
> > + container_of(w, struct xe_guc_engine, lr_tdr);
> > + struct xe_engine *e = ge->engine;
> > + struct drm_gpu_scheduler *sched = &ge->sched;
> > +
> > + XE_WARN_ON(!xe_engine_is_lr(e));
> > + trace_xe_engine_lr_cleanup(e);
> > +
> > + /* Kill the run_job / process_msg entry points */
> > + drm_sched_run_wq_stop(sched);
> > +
> > + /* Engine state now stable, disable scheduling / deregister if needed */
> > + if (engine_registered(e)) {
> > + struct xe_guc *guc = engine_to_guc(e);
> > + int ret;
> > +
> > + set_engine_banned(e);
> > + xe_engine_get(e);
> > + disable_scheduling_deregister(guc, e);
> > +
> > + /*
> > + * Must wait for scheduling to be disabled before signalling
> > + * any fences, if GT broken the GT reset code should signal us.
> > + */
> > + ret = wait_event_timeout(guc->ct.wq,
> > + !engine_pending_disable(e) ||
> > + guc_read_stopped(guc), HZ * 5);
> > + if (!ret) {
> > + XE_WARN_ON("Schedule disable failed to respond");
> > + drm_sched_run_wq_start(sched);
> > + xe_gt_reset_async(e->gt);
> > + return;
> > + }
> > + }
> > +
> > + drm_sched_run_wq_start(sched);
> > +}
> > +
> > static enum drm_gpu_sched_stat
> > guc_engine_timedout_job(struct drm_sched_job *drm_job)
> > {
> > @@ -815,7 +879,7 @@ guc_engine_timedout_job(struct drm_sched_job *drm_job)
> > err = -EIO;
> > set_engine_banned(e);
> > xe_engine_get(e);
> > - disable_scheduling_deregister(engine_to_guc(e), e);
> > + disable_scheduling_deregister(guc, e);
> > /*
> > * Must wait for scheduling to be disabled before signalling
> > @@ -848,7 +912,7 @@ guc_engine_timedout_job(struct drm_sched_job *drm_job)
> > */
> > list_add(&drm_job->list, &sched->pending_list);
> > drm_sched_run_wq_start(sched);
> > - drm_sched_set_timeout(&e->guc->sched, MIN_SCHED_TIMEOUT);
> > + xe_guc_engine_trigger_cleanup(e);
> > /* Mark all outstanding jobs as bad, thus completing them */
> > spin_lock(&sched->job_list_lock);
> > @@ -872,6 +936,8 @@ static void __guc_engine_fini_async(struct work_struct *w)
> > trace_xe_engine_destroy(e);
> > + if (xe_engine_is_lr(e))
> > + cancel_work_sync(&ge->lr_tdr);
> > if (e->flags & ENGINE_FLAG_PERSISTENT)
> > xe_device_remove_persistent_engines(gt_to_xe(e->gt), e);
> > release_guc_id(guc, e);
> > @@ -889,7 +955,7 @@ static void guc_engine_fini_async(struct xe_engine *e)
> > bool kernel = e->flags & ENGINE_FLAG_KERNEL;
> > INIT_WORK(&e->guc->fini_async, __guc_engine_fini_async);
> > - queue_work(system_unbound_wq, &e->guc->fini_async);
> > + queue_work(system_wq, &e->guc->fini_async);
> > /* We must block on kernel engines so slabs are empty on driver unload */
> > if (kernel) {
> > @@ -1078,6 +1144,9 @@ static int guc_engine_init(struct xe_engine *e)
> > if (err)
> > goto err_sched;
> > + if (xe_engine_is_lr(e))
> > + INIT_WORK(&e->guc->lr_tdr, xe_guc_engine_lr_cleanup);
> > +
> > mutex_lock(&guc->submission_state.lock);
> > err = alloc_guc_id(guc, e);
> > @@ -1129,7 +1198,7 @@ static void guc_engine_kill(struct xe_engine *e)
> > {
> > trace_xe_engine_kill(e);
> > set_engine_killed(e);
> > - drm_sched_set_timeout(&e->guc->sched, MIN_SCHED_TIMEOUT);
> > + xe_guc_engine_trigger_cleanup(e);
> > }
> > static void guc_engine_add_msg(struct xe_engine *e, struct drm_sched_msg *msg,
> > @@ -1279,6 +1348,9 @@ static void guc_engine_stop(struct xe_guc *guc, struct xe_engine *e)
> > /* Stop scheduling + flush any DRM scheduler operations */
> > drm_sched_run_wq_stop(sched);
> > + if (engine_registered(e) && xe_engine_is_lr(e))
> > + xe_engine_put(e);
> > +
> > /* Clean up lost G2H + reset engine state */
> > if (engine_destroyed(e) && engine_registered(e)) {
> > if (engine_banned(e))
> > @@ -1503,6 +1575,9 @@ int xe_guc_deregister_done_handler(struct xe_guc *guc, u32 *msg, u32 len)
> > trace_xe_engine_deregister_done(e);
> > clear_engine_registered(e);
> > + if (xe_engine_is_lr(e))
> > + xe_engine_put(e);
> > +
>
> This looks suspicious. We're putting the reference of "e", but continue to
> dereference "e" afterwards. If this is intentional, could you add a comment
> as to which pointer's reference we put? Typically wen you put the reference
> of a pointer you don't use it afterwards unless it is obvious that you're
> only caching the value from the pointer that carries the reference?
>
This is goofy and does indeed work as if a engine is LR here is will
also always be banned. Let me change the reference counting here so the
code can change to:
if (xe_engine_is_lr(e) || engine_banned(e))
xe_engine_put(e);
else
...
This will affect a few other code paths too.
Matt
>
> > if (engine_banned(e))
> > xe_engine_put(e);
> > else
> > @@ -1540,7 +1615,7 @@ int xe_guc_engine_reset_handler(struct xe_guc *guc, u32 *msg, u32 len)
> > */
> > set_engine_reset(e);
> > if (!engine_banned(e))
> > - drm_sched_set_timeout(&e->guc->sched, MIN_SCHED_TIMEOUT);
> > + xe_guc_engine_trigger_cleanup(e);
> > return 0;
> > }
> > @@ -1567,7 +1642,7 @@ int xe_guc_engine_memory_cat_error_handler(struct xe_guc *guc, u32 *msg,
> > /* Treat the same as engine reset */
> > set_engine_reset(e);
> > if (!engine_banned(e))
> > - drm_sched_set_timeout(&e->guc->sched, MIN_SCHED_TIMEOUT);
> > + xe_guc_engine_trigger_cleanup(e);
> > return 0;
> > }
> > diff --git a/drivers/gpu/drm/xe/xe_trace.h b/drivers/gpu/drm/xe/xe_trace.h
> > index 2f8eb7ebe9a7..02861c26e145 100644
> > --- a/drivers/gpu/drm/xe/xe_trace.h
> > +++ b/drivers/gpu/drm/xe/xe_trace.h
> > @@ -219,6 +219,11 @@ DEFINE_EVENT(xe_engine, xe_engine_resubmit,
> > TP_ARGS(e)
> > );
> > +DEFINE_EVENT(xe_engine, xe_engine_lr_cleanup,
> > + TP_PROTO(struct xe_engine *e),
> > + TP_ARGS(e)
> > +);
> > +
> > DECLARE_EVENT_CLASS(xe_sched_job,
> > TP_PROTO(struct xe_sched_job *job),
> > TP_ARGS(job),
More information about the Intel-xe
mailing list