[Intel-gfx] [PATCH 2/3] drm/i915/perf: reintroduce wait on OA configuration completion

[Lionel Landwerlin]

We still need to wait for the initial OA configuration to happen
before we enable OA report writes to the OA buffer.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin at intel.com>
Fixes: 15d0ace1f876 ("drm/i915/perf: execute OA configuration from command stream")
Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_perf.c       | 49 +++++++++++++++++++++++---
 drivers/gpu/drm/i915/i915_perf_types.h |  8 +++++
 2 files changed, 52 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index b474a0240280..541a818ff3a8 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1378,6 +1378,23 @@ free_noa_wait(struct i915_perf_stream *stream)
 	i915_vma_unpin_and_release(&stream->noa_wait, 0);
 }
 
+static int
+wait_and_put_configure_request(struct i915_perf_stream *stream)
+{
+	struct i915_request *rq = stream->configure_request;
+	int ret = 0;
+
+	stream->configure_request = NULL;
+	GEM_BUG_ON(rq == NULL);
+
+	if (i915_request_wait(rq, 0, MAX_SCHEDULE_TIMEOUT) < 0)
+		ret = -ETIME;
+
+	i915_request_put(rq);
+
+	return ret;
+}
+
 static void i915_oa_stream_destroy(struct i915_perf_stream *stream)
 {
 	struct i915_perf *perf = stream->perf;
@@ -1392,6 +1409,7 @@ static void i915_oa_stream_destroy(struct i915_perf_stream *stream)
 	 */
 	WRITE_ONCE(perf->exclusive_stream, NULL);
 	perf->ops.disable_metric_set(stream);
+	GEM_BUG_ON(stream->configure_request != NULL);
 
 	free_oa_buffer(stream);
 
@@ -1956,7 +1974,8 @@ get_oa_vma(struct i915_perf_stream *stream, struct i915_oa_config *oa_config)
 
 static int emit_oa_config(struct i915_perf_stream *stream,
 			  struct i915_oa_config *oa_config,
-			  struct intel_context *ce)
+			  struct intel_context *ce,
+			  bool store_on_stream)
 {
 	struct i915_request *rq;
 	struct i915_vma *vma;
@@ -1989,6 +2008,12 @@ static int emit_oa_config(struct i915_perf_stream *stream,
 	err = rq->engine->emit_bb_start(rq,
 					vma->node.start, 0,
 					I915_DISPATCH_SECURE);
+
+	if (err == 0 && store_on_stream) {
+		GEM_BUG_ON(stream->configure_request != NULL);
+		stream->configure_request = i915_request_get(rq);
+	}
+
 err_add_request:
 	i915_request_add(rq);
 err_vma_unpin:
@@ -2022,7 +2047,9 @@ static int hsw_enable_metric_set(struct i915_perf_stream *stream)
 	intel_uncore_rmw(uncore, GEN6_UCGCTL1,
 			 0, GEN6_CSUNIT_CLOCK_GATE_DISABLE);
 
-	return emit_oa_config(stream, stream->oa_config, oa_context(stream));
+	return emit_oa_config(stream, stream->oa_config,
+			      oa_context(stream),
+			      true /* store_on_stream */);
 }
 
 static void hsw_disable_metric_set(struct i915_perf_stream *stream)
@@ -2452,7 +2479,9 @@ static int gen8_enable_metric_set(struct i915_perf_stream *stream)
 	if (ret)
 		return ret;
 
-	return emit_oa_config(stream, oa_config, oa_context(stream));
+	return emit_oa_config(stream, oa_config,
+			      oa_context(stream),
+			      true /* store_on_stream */);
 }
 
 static u32 oag_report_ctx_switches(const struct i915_perf_stream *stream)
@@ -2506,7 +2535,9 @@ static int gen12_enable_metric_set(struct i915_perf_stream *stream)
 			return ret;
 	}
 
-	return emit_oa_config(stream, oa_config, oa_context(stream));
+	return emit_oa_config(stream, oa_config,
+			      oa_context(stream),
+			      true /* store_on_stream */);
 }
 
 static void gen8_disable_metric_set(struct i915_perf_stream *stream)
@@ -2841,6 +2872,12 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 		goto err_enable;
 	}
 
+	ret = wait_and_put_configure_request(stream);
+	if (ret) {
+		DRM_DEBUG("Wait on OA config request timed out\n");
+		goto err_enable;
+	}
+
 	DRM_DEBUG("opening stream oa config uuid=%s\n",
 		  stream->oa_config->uuid);
 
@@ -2855,6 +2892,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 err_enable:
 	WRITE_ONCE(perf->exclusive_stream, NULL);
 	perf->ops.disable_metric_set(stream);
+	GEM_BUG_ON(stream->configure_request != NULL);
 
 	free_oa_buffer(stream);
 
@@ -3163,7 +3201,8 @@ static long i915_perf_config_locked(struct i915_perf_stream *stream,
 		 * When set globally, we use a low priority kernel context,
 		 * so it will effectively take effect when idle.
 		 */
-		err = emit_oa_config(stream, config, oa_context(stream));
+		err = emit_oa_config(stream, config, oa_context(stream),
+				     false /* store_on_stream */);
 		if (err == 0)
 			config = xchg(&stream->oa_config, config);
 		else
diff --git a/drivers/gpu/drm/i915/i915_perf_types.h b/drivers/gpu/drm/i915/i915_perf_types.h
index d994fa6a1c5f..92fd3cf5afcc 100644
--- a/drivers/gpu/drm/i915/i915_perf_types.h
+++ b/drivers/gpu/drm/i915/i915_perf_types.h
@@ -309,6 +309,14 @@ struct i915_perf_stream {
 	 * reprogrammed.
 	 */
 	struct i915_vma *noa_wait;
+
+	/**
+	 * @configure_request: Request on which to wait for HW to complete its
+	 * initial configuration. This is required for applications caring
+	 * about system wide monitoring. We want all the data they can get
+	 * through the OA buffer to be valid.
+	 */
+	struct i915_request *configure_request;
 };
 
 /**
-- 
2.25.1