| Message ID | 1533049014-6934-1-git-send-email-andrey.grodzovsky@amd.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [1/2] drm/scheduler: Add job dependency trace. | expand |
Am 31.07.2018 um 16:56 schrieb Andrey Grodzovsky: > During debug sessions I encountered a need to trace > back a job dependecy a few steps back to the first failing > job. This trace helpped me a lot. > > Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com> Reviewed-by: Christian König <christian.koenig@amd.com> for the series. > --- > drivers/gpu/drm/scheduler/gpu_scheduler.c | 8 ++++++-- > drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 24 ++++++++++++++++++++++++ > 2 files changed, 30 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler.c b/drivers/gpu/drm/scheduler/gpu_scheduler.c > index 3f2fc5e..45703b9 100644 > --- a/drivers/gpu/drm/scheduler/gpu_scheduler.c > +++ b/drivers/gpu/drm/scheduler/gpu_scheduler.c > @@ -506,9 +506,13 @@ drm_sched_entity_pop_job(struct drm_sched_entity *entity) > if (!sched_job) > return NULL; > > - while ((entity->dependency = sched->ops->dependency(sched_job, entity))) > - if (drm_sched_entity_add_dependency_cb(entity)) > + while ((entity->dependency = sched->ops->dependency(sched_job, entity))) { > + if (drm_sched_entity_add_dependency_cb(entity)) { > + > + trace_drm_sched_job_wait_dep(sched_job, entity->dependency); > return NULL; > + } > + } > > /* skip jobs from entity that marked guilty */ > if (entity->guilty && atomic_read(entity->guilty)) > diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > index 4998ad9..1626f39 100644 > --- a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > +++ b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > @@ -74,6 +74,30 @@ TRACE_EVENT(drm_sched_process_job, > TP_printk("fence=%p signaled", __entry->fence) > ); > > +TRACE_EVENT(drm_sched_job_wait_dep, > + TP_PROTO(struct drm_sched_job *sched_job, struct dma_fence *fence), > + TP_ARGS(sched_job, fence), > + TP_STRUCT__entry( > + __field(const char *,name) > + __field(uint64_t, id) > + __field(struct dma_fence *, fence) > + __field(uint64_t, ctx) > + __field(unsigned, seqno) > + ), > + > + TP_fast_assign( > + __entry->name = sched_job->sched->name; > + __entry->id = sched_job->id; > + __entry->fence = fence; > + __entry->ctx = fence->context; > + __entry->seqno = fence->seqno; > + ), > + TP_printk("job ring=%s, id=%llu, depends fence=%p, context=%llu, seq=%u", > + __entry->name, __entry->id, > + __entry->fence, __entry->ctx, > + __entry->seqno) > +); > + > #endif > > /* This part must be outside protection */
On Tue, Jul 31, 2018 at 10:56:53AM -0400, Andrey Grodzovsky wrote: > During debug sessions I encountered a need to trace > back a job dependecy a few steps back to the first failing > job. This trace helpped me a lot. > > Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com> Series are Reviewed-by: Huang Rui <ray.huang@amd.com> > --- > drivers/gpu/drm/scheduler/gpu_scheduler.c | 8 ++++++-- > drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 24 ++++++++++++++++++++++++ > 2 files changed, 30 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler.c b/drivers/gpu/drm/scheduler/gpu_scheduler.c > index 3f2fc5e..45703b9 100644 > --- a/drivers/gpu/drm/scheduler/gpu_scheduler.c > +++ b/drivers/gpu/drm/scheduler/gpu_scheduler.c > @@ -506,9 +506,13 @@ drm_sched_entity_pop_job(struct drm_sched_entity *entity) > if (!sched_job) > return NULL; > > - while ((entity->dependency = sched->ops->dependency(sched_job, entity))) > - if (drm_sched_entity_add_dependency_cb(entity)) > + while ((entity->dependency = sched->ops->dependency(sched_job, entity))) { > + if (drm_sched_entity_add_dependency_cb(entity)) { > + > + trace_drm_sched_job_wait_dep(sched_job, entity->dependency); > return NULL; > + } > + } > > /* skip jobs from entity that marked guilty */ > if (entity->guilty && atomic_read(entity->guilty)) > diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > index 4998ad9..1626f39 100644 > --- a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > +++ b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h > @@ -74,6 +74,30 @@ TRACE_EVENT(drm_sched_process_job, > TP_printk("fence=%p signaled", __entry->fence) > ); > > +TRACE_EVENT(drm_sched_job_wait_dep, > + TP_PROTO(struct drm_sched_job *sched_job, struct dma_fence *fence), > + TP_ARGS(sched_job, fence), > + TP_STRUCT__entry( > + __field(const char *,name) > + __field(uint64_t, id) > + __field(struct dma_fence *, fence) > + __field(uint64_t, ctx) > + __field(unsigned, seqno) > + ), > + > + TP_fast_assign( > + __entry->name = sched_job->sched->name; > + __entry->id = sched_job->id; > + __entry->fence = fence; > + __entry->ctx = fence->context; > + __entry->seqno = fence->seqno; > + ), > + TP_printk("job ring=%s, id=%llu, depends fence=%p, context=%llu, seq=%u", > + __entry->name, __entry->id, > + __entry->fence, __entry->ctx, > + __entry->seqno) > +); > + > #endif > > /* This part must be outside protection */ > -- > 2.7.4 > > _______________________________________________ > amd-gfx mailing list > amd-gfx@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/amd-gfx
diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler.c b/drivers/gpu/drm/scheduler/gpu_scheduler.c index 3f2fc5e..45703b9 100644 --- a/drivers/gpu/drm/scheduler/gpu_scheduler.c +++ b/drivers/gpu/drm/scheduler/gpu_scheduler.c @@ -506,9 +506,13 @@ drm_sched_entity_pop_job(struct drm_sched_entity *entity) if (!sched_job) return NULL; - while ((entity->dependency = sched->ops->dependency(sched_job, entity))) - if (drm_sched_entity_add_dependency_cb(entity)) + while ((entity->dependency = sched->ops->dependency(sched_job, entity))) { + if (drm_sched_entity_add_dependency_cb(entity)) { + + trace_drm_sched_job_wait_dep(sched_job, entity->dependency); return NULL; + } + } /* skip jobs from entity that marked guilty */ if (entity->guilty && atomic_read(entity->guilty)) diff --git a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h index 4998ad9..1626f39 100644 --- a/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h +++ b/drivers/gpu/drm/scheduler/gpu_scheduler_trace.h @@ -74,6 +74,30 @@ TRACE_EVENT(drm_sched_process_job, TP_printk("fence=%p signaled", __entry->fence) ); +TRACE_EVENT(drm_sched_job_wait_dep, + TP_PROTO(struct drm_sched_job *sched_job, struct dma_fence *fence), + TP_ARGS(sched_job, fence), + TP_STRUCT__entry( + __field(const char *,name) + __field(uint64_t, id) + __field(struct dma_fence *, fence) + __field(uint64_t, ctx) + __field(unsigned, seqno) + ), + + TP_fast_assign( + __entry->name = sched_job->sched->name; + __entry->id = sched_job->id; + __entry->fence = fence; + __entry->ctx = fence->context; + __entry->seqno = fence->seqno; + ), + TP_printk("job ring=%s, id=%llu, depends fence=%p, context=%llu, seq=%u", + __entry->name, __entry->id, + __entry->fence, __entry->ctx, + __entry->seqno) +); + #endif /* This part must be outside protection */
During debug sessions I encountered a need to trace back a job dependecy a few steps back to the first failing job. This trace helpped me a lot. Signed-off-by: Andrey Grodzovsky <andrey.grodzovsky@amd.com> --- drivers/gpu/drm/scheduler/gpu_scheduler.c | 8 ++++++-- drivers/gpu/drm/scheduler/gpu_scheduler_trace.h | 24 ++++++++++++++++++++++++ 2 files changed, 30 insertions(+), 2 deletions(-)