[PATCH 2/2 v3] drm/i915: tracepoints for semaphores

[Ben Widawsky <ben@bwidawsk.net>]

The tracepoints give enough info to figure the updates and compares
(document terminology for signals and waits) and dependencies therein of
the semaphore mailboxes.

Here are arguments to perf to get interesting info (mostly copied from
Chris):
record -f -g -c 1 -e i915:intel_ringbuffer_add_request -e i915:intel_ringbuffer_consume_semaphore -a

Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Ben Widawsky <ben@bwidawsk.net>
---
 drivers/gpu/drm/i915/i915_trace.h       |   61 +++++++++++++++++++++++++++++++
 drivers/gpu/drm/i915/intel_ringbuffer.c |    4 ++
 2 files changed, 65 insertions(+), 0 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_trace.h b/drivers/gpu/drm/i915/i915_trace.h
index d623fef..493a0d3 100644
--- a/drivers/gpu/drm/i915/i915_trace.h
+++ b/drivers/gpu/drm/i915/i915_trace.h
@@ -410,6 +410,67 @@ TRACE_EVENT(i915_reg_rw,
 		     (u32)(__entry->val >> 32))
 );
 
+#define SEMAPHORE_REG_NAME(x) (\
+	(x == RENDER_RING_BASE + 0x40) ? ("RVSYNC") : \
+	((x == RENDER_RING_BASE + 0x44) ? ("RBSYNC") : \
+	((x == GEN6_BSD_RING_BASE + 0x40) ? ("VRSYNC") : \
+	((x == GEN6_BSD_RING_BASE + 0x44) ? ("VBSYNC") : \
+	((x == BLT_RING_BASE + 0x40) ? ("BVSYNC") : \
+	((x == BLT_RING_BASE + 0x44) ? ("BRSYNC") : \
+	("UNKNOWN")))))))
+
+#define RING_NAME(x) (\
+	(x == 1) ? ("rcs") : \
+	((x == 2) ? ("vcs") : \
+	((x == 4) ? ("bcs") : \
+	("unk"))))
+
+TRACE_EVENT(intel_ringbuffer_add_request,
+	    TP_PROTO(struct intel_ring_buffer *ring,
+		     u32 seqno,
+		     u32 reg1,
+		     u32 reg2),
+	    TP_ARGS(ring, seqno, reg1, reg2),
+	    TP_STRUCT__entry(
+			     __field(int, id)
+			     __field(u32, seqno)
+			     __field(u32, reg1)
+			     __field(u32, reg2)
+			    ),
+	    TP_fast_assign(
+			   __entry->id = ring->id;
+			   __entry->seqno = seqno;
+			   __entry->reg1 = reg1;
+			   __entry->reg2 = reg2;
+			  ),
+	    TP_printk("ring = %s, seqno = %u, signal mbox 1 = %s, signal mbox 2 = %s",
+		      RING_NAME(__entry->id), __entry->seqno,
+		      SEMAPHORE_REG_NAME(__entry->reg1),
+		      SEMAPHORE_REG_NAME(__entry->reg2))
+);
+
+TRACE_EVENT(intel_ringbuffer_consume_semaphore,
+	    TP_PROTO(struct intel_ring_buffer *updater,
+		     struct intel_ring_buffer *comparer,
+		     u32 seqno),
+	    TP_ARGS(updater, comparer, seqno),
+	    TP_STRUCT__entry(
+			     __field(int, src)
+			     __field(int, dest)
+			     __field(u32, seqno)
+			    ),
+	    TP_fast_assign(
+			   __entry->src = updater->id;
+			   __entry->dest = comparer->id;
+			   __entry->seqno = seqno;
+			  ),
+	    /* It's not easy to macro the consuming mbox register */
+	    TP_printk("signaller = %s, waiter = %s, seqno = %u",
+		      RING_NAME(__entry->src),
+		      RING_NAME(__entry->dest),
+		      __entry->seqno)
+);
+
 #endif /* _I915_TRACE_H_ */
 
 /* This part must be outside protection */
diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
index d25c275..accd520b 100644
--- a/drivers/gpu/drm/i915/intel_ringbuffer.c
+++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
@@ -360,6 +360,8 @@ gen6_add_request(struct intel_ring_buffer *ring,
 	intel_ring_emit(ring, MI_USER_INTERRUPT);
 	intel_ring_advance(ring);
 
+	trace_intel_ringbuffer_add_request(ring, seqno, mbox1_reg, mbox2_reg);
+
 	return 0;
 }
 
@@ -420,6 +422,8 @@ intel_ring_sync(struct intel_ring_buffer *waiter,
 	intel_ring_emit(waiter, MI_NOOP);
 	intel_ring_advance(waiter);
 
+	trace_intel_ringbuffer_consume_semaphore(signaller, waiter, seqno);
+
 	return 0;
 }
 
-- 
1.7.6.1