| Message ID | 20240614081657.408397-2-pierre-eric.pelloux-prayer@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Improve gpu_scheduler trace events + uAPI | expand |
Am 14.06.24 um 10:16 schrieb Pierre-Eric Pelloux-Prayer: > Until the switch from kthread to workqueue, a userspace application could > determine the source device from the pid of the thread sending the event. > > With workqueues this is not possible anymore, so the event needs to contain > the dev_name() to identify the device. > > Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com> > --- > drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > index c75302ca3427..1f9c5a884487 100644 > --- a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > +++ b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > @@ -42,6 +42,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, > __field(uint64_t, id) > __field(u32, job_count) > __field(int, hw_job_count) > + __string(dev, dev_name(sched_job->sched->dev)) > ), > > TP_fast_assign( > @@ -52,6 +53,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, > __entry->job_count = spsc_queue_count(&entity->job_queue); > __entry->hw_job_count = atomic_read( > &sched_job->sched->credit_count); > + __assign_str(dev); > ), > TP_printk("entity=%p, id=%llu, fence=%p, ring=%s, job count:%u, hw job count:%d", > __entry->entity, __entry->id, > @@ -64,9 +66,13 @@ DEFINE_EVENT(drm_sched_job, drm_sched_job, > TP_ARGS(sched_job, entity) > ); > > -DEFINE_EVENT(drm_sched_job, drm_run_job, > +DEFINE_EVENT_PRINT(drm_sched_job, drm_run_job, > TP_PROTO(struct drm_sched_job *sched_job, struct drm_sched_entity *entity), > - TP_ARGS(sched_job, entity) > + TP_ARGS(sched_job, entity), > + TP_printk("dev=%s, entity=%p id=%llu, fence=%p, ring=%s, job count:%u, hw job count:%d", > + __get_str(dev), __entry->entity, __entry->id, > + __entry->fence, __get_str(name), > + __entry->job_count, __entry->hw_job_count) Why not doing that in the TP_printk() of the drm_sched_job event class? The device should now be available for all trace events of that class. Regards, Christian. > ); > > TRACE_EVENT(drm_sched_process_job,
Hi, Le 14/06/2024 à 11:08, Christian König a écrit : > > > Am 14.06.24 um 10:16 schrieb Pierre-Eric Pelloux-Prayer: >> Until the switch from kthread to workqueue, a userspace application could >> determine the source device from the pid of the thread sending the event. >> >> With workqueues this is not possible anymore, so the event needs to >> contain >> the dev_name() to identify the device. >> >> Signed-off-by: Pierre-Eric Pelloux-Prayer >> <pierre-eric.pelloux-prayer@amd.com> >> --- >> drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 10 ++++++++-- >> 1 file changed, 8 insertions(+), 2 deletions(-) >> >> diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >> b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >> index c75302ca3427..1f9c5a884487 100644 >> --- a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >> +++ b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >> @@ -42,6 +42,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, >> __field(uint64_t, id) >> __field(u32, job_count) >> __field(int, hw_job_count) >> + __string(dev, dev_name(sched_job->sched->dev)) >> ), >> TP_fast_assign( >> @@ -52,6 +53,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, >> __entry->job_count = >> spsc_queue_count(&entity->job_queue); >> __entry->hw_job_count = atomic_read( >> &sched_job->sched->credit_count); >> + __assign_str(dev); >> ), >> TP_printk("entity=%p, id=%llu, fence=%p, ring=%s, job >> count:%u, hw job count:%d", >> __entry->entity, __entry->id, >> @@ -64,9 +66,13 @@ DEFINE_EVENT(drm_sched_job, drm_sched_job, >> TP_ARGS(sched_job, entity) >> ); >> -DEFINE_EVENT(drm_sched_job, drm_run_job, >> +DEFINE_EVENT_PRINT(drm_sched_job, drm_run_job, >> TP_PROTO(struct drm_sched_job *sched_job, struct >> drm_sched_entity *entity), >> - TP_ARGS(sched_job, entity) >> + TP_ARGS(sched_job, entity), >> + TP_printk("dev=%s, entity=%p id=%llu, fence=%p, ring=%s, job >> count:%u, hw job count:%d", >> + __get_str(dev), __entry->entity, __entry->id, >> + __entry->fence, __get_str(name), >> + __entry->job_count, __entry->hw_job_count) > > Why not doing that in the TP_printk() of the drm_sched_job event class? > I didn't because drm_sched_job + drm_run_job usually work in pair so the ring+device information will be available anyway. But maybe you're right and it's better to add it in the event the job never gets scheduled. For the other events it's not necessary IMO: the first event will always be drm_sched_job, so it's possible to figure out which the device/ring for a given job. Regards, Pierre-eric > The device should now be available for all trace events of that class. > > Regards, > Christian. > >> ); >> TRACE_EVENT(drm_sched_process_job,
Am 17.06.24 um 14:54 schrieb Pierre-Eric Pelloux-Prayer: > Hi, > > Le 14/06/2024 à 11:08, Christian König a écrit : >> >> >> Am 14.06.24 um 10:16 schrieb Pierre-Eric Pelloux-Prayer: >>> Until the switch from kthread to workqueue, a userspace application >>> could >>> determine the source device from the pid of the thread sending the >>> event. >>> >>> With workqueues this is not possible anymore, so the event needs to >>> contain >>> the dev_name() to identify the device. >>> >>> Signed-off-by: Pierre-Eric Pelloux-Prayer >>> <pierre-eric.pelloux-prayer@amd.com> >>> --- >>> drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 10 ++++++++-- >>> 1 file changed, 8 insertions(+), 2 deletions(-) >>> >>> diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >>> b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >>> index c75302ca3427..1f9c5a884487 100644 >>> --- a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >>> +++ b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h >>> @@ -42,6 +42,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, >>> __field(uint64_t, id) >>> __field(u32, job_count) >>> __field(int, hw_job_count) >>> + __string(dev, dev_name(sched_job->sched->dev)) >>> ), >>> TP_fast_assign( >>> @@ -52,6 +53,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, >>> __entry->job_count = >>> spsc_queue_count(&entity->job_queue); >>> __entry->hw_job_count = atomic_read( >>> &sched_job->sched->credit_count); >>> + __assign_str(dev); >>> ), >>> TP_printk("entity=%p, id=%llu, fence=%p, ring=%s, job >>> count:%u, hw job count:%d", >>> __entry->entity, __entry->id, >>> @@ -64,9 +66,13 @@ DEFINE_EVENT(drm_sched_job, drm_sched_job, >>> TP_ARGS(sched_job, entity) >>> ); >>> -DEFINE_EVENT(drm_sched_job, drm_run_job, >>> +DEFINE_EVENT_PRINT(drm_sched_job, drm_run_job, >>> TP_PROTO(struct drm_sched_job *sched_job, struct >>> drm_sched_entity *entity), >>> - TP_ARGS(sched_job, entity) >>> + TP_ARGS(sched_job, entity), >>> + TP_printk("dev=%s, entity=%p id=%llu, fence=%p, ring=%s, >>> job count:%u, hw job count:%d", >>> + __get_str(dev), __entry->entity, __entry->id, >>> + __entry->fence, __get_str(name), >>> + __entry->job_count, __entry->hw_job_count) >> >> Why not doing that in the TP_printk() of the drm_sched_job event class? >> > > I didn't because drm_sched_job + drm_run_job usually work in pair so > the ring+device information will be available anyway. > > But maybe you're right and it's better to add it in the event the job > never gets scheduled. > > For the other events it's not necessary IMO: the first event will > always be drm_sched_job, so it's possible to figure out which the > device/ring for a given job. Yeah, but we have that strace in the trace event even when we don't print it. I would just go ahead and always print it even when it is available somehow else. Regards, Christian. > > Regards, > Pierre-eric > >> The device should now be available for all trace events of that class. >> >> Regards, >> Christian. >> >>> ); >>> TRACE_EVENT(drm_sched_process_job,
diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h index c75302ca3427..1f9c5a884487 100644 --- a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h +++ b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h @@ -42,6 +42,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, __field(uint64_t, id) __field(u32, job_count) __field(int, hw_job_count) + __string(dev, dev_name(sched_job->sched->dev)) ), TP_fast_assign( @@ -52,6 +53,7 @@ DECLARE_EVENT_CLASS(drm_sched_job, __entry->job_count = spsc_queue_count(&entity->job_queue); __entry->hw_job_count = atomic_read( &sched_job->sched->credit_count); + __assign_str(dev); ), TP_printk("entity=%p, id=%llu, fence=%p, ring=%s, job count:%u, hw job count:%d", __entry->entity, __entry->id, @@ -64,9 +66,13 @@ DEFINE_EVENT(drm_sched_job, drm_sched_job, TP_ARGS(sched_job, entity) ); -DEFINE_EVENT(drm_sched_job, drm_run_job, +DEFINE_EVENT_PRINT(drm_sched_job, drm_run_job, TP_PROTO(struct drm_sched_job *sched_job, struct drm_sched_entity *entity), - TP_ARGS(sched_job, entity) + TP_ARGS(sched_job, entity), + TP_printk("dev=%s, entity=%p id=%llu, fence=%p, ring=%s, job count:%u, hw job count:%d", + __get_str(dev), __entry->entity, __entry->id, + __entry->fence, __get_str(name), + __entry->job_count, __entry->hw_job_count) ); TRACE_EVENT(drm_sched_process_job,
Until the switch from kthread to workqueue, a userspace application could determine the source device from the pid of the thread sending the event. With workqueues this is not possible anymore, so the event needs to contain the dev_name() to identify the device. Signed-off-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com> --- drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)