[Intel-gfx] [PATCH v3] drm/i915: store HW tagging information into tracepoints

[Lionel Landwerlin]

In Gpuvis [1] we added matching of the OA report tags against i915
tracepoints fields to figure what workload was submitted by what
process. It doesn't matter much whether HW tags get reused for a
single request, as it gets preempted for example. All we need is link
between the OA report and seqno/engine over a short period of time.
That is enough to find the relationship between the different elements
on a timeline.

The tags got removed from the tracepoints in 2935ed5339c4 ("drm/i915:
Remove logical HW ID"). I was fine with the idea of reuse and dropping
them from most tracepoints, but we still need it at execlist port
insertion.

This change brings the hw_id back just for the i915_request_in
tracepoint.

TODO: someone figures what Guc's putting in the upper 32bits of the
      execlist port

v2: s/hw_tag/hw_id/ to keep the old gpuvis code going

v3: Don't store anything on i915_request

[1] : https://github.com/mikesart/gpuvis/wiki/TechDocs-Intel#gpu-generated-countersevents

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
Fixes: 2935ed5339c4 ("drm/i915: Remove logical HW ID")
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio at intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin at intel.com>
Cc: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/gt/intel_lrc.c               | 6 +++++-
 drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c | 3 ++-
 drivers/gpu/drm/i915/i915_trace.h                 | 8 +++++---
 3 files changed, 12 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 6fbad5e2343f..58970099db25 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1278,7 +1278,6 @@ execlists_schedule_in(struct i915_request *rq, int idx)
 	struct intel_engine_cs *old;
 
 	GEM_BUG_ON(!intel_engine_pm_is_awake(rq->engine));
-	trace_i915_request_in(rq, idx);
 
 	old = READ_ONCE(ce->inflight);
 	do {
@@ -1288,6 +1287,11 @@ execlists_schedule_in(struct i915_request *rq, int idx)
 		}
 	} while (!try_cmpxchg(&ce->inflight, &old, ptr_inc(old)));
 
+	/*
+	 * Emit the tracepoint once the ce->lrc_desc has been updated.
+	 */
+	trace_i915_request_in(rq, idx, ce->lrc_desc >> 32);
+
 	GEM_BUG_ON(intel_context_inflight(ce) != rq->engine);
 	return i915_request_get(rq);
 }
diff --git a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
index fe7778c28d2d..fd3d63d720d9 100644
--- a/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
+++ b/drivers/gpu/drm/i915/gt/uc/intel_guc_submission.c
@@ -263,7 +263,8 @@ static inline int rq_prio(const struct i915_request *rq)
 
 static struct i915_request *schedule_in(struct i915_request *rq, int idx)
 {
-	trace_i915_request_in(rq, idx);
+	trace_i915_request_in(rq, idx,
+			      0 /* TODO: figure out what value GuC chooses */);
 
 	/*
 	 * Currently we are not tracking the rq->context being inflight
diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index bc854ad60954..3a36adb8094b 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -790,7 +790,7 @@ DEFINE_EVENT(i915_request, i915_request_execute,
 );
 
 TRACE_EVENT(i915_request_in,
-	    TP_PROTO(struct i915_request *rq, unsigned int port),
+	    TP_PROTO(struct i915_request *rq, unsigned int port, u32 hw_id),
 	    TP_ARGS(rq, port),
 
 	    TP_STRUCT__entry(
@@ -801,6 +801,7 @@ TRACE_EVENT(i915_request_in,
 			     __field(u32, seqno)
 			     __field(u32, port)
 			     __field(s32, prio)
+			     __field(u32, hw_id)
 			    ),
 
 	    TP_fast_assign(
@@ -811,12 +812,13 @@ TRACE_EVENT(i915_request_in,
 			   __entry->seqno = rq->fence.seqno;
 			   __entry->prio = rq->sched.attr.priority;
 			   __entry->port = port;
+			   __entry->hw_id = hw_id;
 			   ),
 
-	    TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, prio=%d, port=%u",
+	    TP_printk("dev=%u, engine=%u:%u, ctx=%llu, seqno=%u, prio=%d, port=%u hw_id=%u",
 		      __entry->dev, __entry->class, __entry->instance,
 		      __entry->ctx, __entry->seqno,
-		      __entry->prio, __entry->port)
+		      __entry->prio, __entry->port, __entry->hw_id)
 );
 
 TRACE_EVENT(i915_request_out,
-- 
2.26.1