drm/i915/perf: Group engines into respective OA groups

Now that we may have multiple OA units in a single GT as well as on
separate GTs, create an engine group that maps to a single OA unit.

v2: (Jani)
- Drop warning on ENOMEM
- Reorder patch in the series

v3: (Ashutosh)
- Remove unused members from perf structs
- Update comments
- Update engine_supports_oa check
- Just return 1 in num_perf_groups_per_gt for now
- Set engine->oa_group to NULL to begin with

v4: Use engine_supports_oa() check in oa_init_reg_state (Ashutosh)
v5: Rebase after dropping engine_supports_oa helper

Signed-off-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa@intel.com>
Reviewed-by: Ashutosh Dixit <ashutosh.dixit@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230323225901.3743681-5-umesh.nerlige.ramappa@intel.com

1 files changed, 86 insertions, 7 deletions

diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index ebd76b5c9712..cf4ea5e389a5 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1570,12 +1570,18 @@ free_noa_wait(struct i915_perf_stream *stream)
 	i915_vma_unpin_and_release(&stream->noa_wait, 0);
 }
 
+static bool engine_supports_oa(const struct intel_engine_cs *engine)
+{
+	return engine->oa_group;
+}
+
 static void i915_oa_stream_destroy(struct i915_perf_stream *stream)
 {
 	struct i915_perf *perf = stream->perf;
 	struct intel_gt *gt = stream->engine->gt;
+	struct i915_perf_group *g = stream->engine->oa_group;
 
-	if (WARN_ON(stream != gt->perf.exclusive_stream))
+	if (WARN_ON(stream != g->exclusive_stream))
 		return;
 
 	/*
@@ -1584,7 +1590,7 @@ static void i915_oa_stream_destroy(struct i915_perf_stream *stream)
 	 *
 	 * See i915_oa_init_reg_state() and lrc_configure_all_contexts()
 	 */
-	WRITE_ONCE(gt->perf.exclusive_stream, NULL);
+	WRITE_ONCE(g->exclusive_stream, NULL);
 	perf->ops.disable_metric_set(stream);
 
 	free_oa_buffer(stream);
@@ -3182,6 +3188,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 {
 	struct drm_i915_private *i915 = stream->perf->i915;
 	struct i915_perf *perf = stream->perf;
+	struct i915_perf_group *g;
 	struct intel_gt *gt;
 	int ret;
 
@@ -3191,6 +3198,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 		return -EINVAL;
 	}
 	gt = props->engine->gt;
+	g = props->engine->oa_group;
 
 	/*
 	 * If the sysfs metrics/ directory wasn't registered for some
@@ -3221,7 +3229,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 	 * counter reports and marshal to the appropriate client
 	 * we currently only allow exclusive access
 	 */
-	if (gt->perf.exclusive_stream) {
+	if (g->exclusive_stream) {
 		drm_dbg(&stream->perf->i915->drm,
 			"OA unit already in use\n");
 		return -EBUSY;
@@ -3316,7 +3324,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 	stream->ops = &i915_oa_stream_ops;
 
 	stream->engine->gt->perf.sseu = props->sseu;
-	WRITE_ONCE(gt->perf.exclusive_stream, stream);
+	WRITE_ONCE(g->exclusive_stream, stream);
 
 	ret = i915_perf_stream_enable_sync(stream);
 	if (ret) {
@@ -3339,7 +3347,7 @@ static int i915_oa_stream_init(struct i915_perf_stream *stream,
 	return 0;
 
 err_enable:
-	WRITE_ONCE(gt->perf.exclusive_stream, NULL);
+	WRITE_ONCE(g->exclusive_stream, NULL);
 	perf->ops.disable_metric_set(stream);
 
 	free_oa_buffer(stream);
@@ -3373,7 +3381,7 @@ void i915_oa_init_reg_state(const struct intel_context *ce,
 		return;
 
 	/* perf.exclusive_stream serialised by lrc_configure_all_contexts() */
-	stream = READ_ONCE(engine->gt->perf.exclusive_stream);
+	stream = READ_ONCE(engine->oa_group->exclusive_stream);
 	if (stream && GRAPHICS_VER(stream->perf->i915) < 12)
 		gen8_update_reg_state_unlocked(ce, stream);
 }
@@ -3965,6 +3973,13 @@ static int read_properties_unlocked(struct i915_perf *perf,
 		return -EINVAL;
 	}
 
+	if (!engine_supports_oa(props->engine)) {
+		drm_dbg(&perf->i915->drm,
+			"Engine not supported by OA %d:%d\n",
+			I915_ENGINE_CLASS_RENDER, 0);
+		return -EINVAL;
+	}
+
 	/* Considering that ID = 0 is reserved and assuming that we don't
 	 * (currently) expect any configurations to ever specify duplicate
 	 * values for a particular property ID then the last _PROP_MAX value is
@@ -4743,6 +4758,60 @@ static struct ctl_table oa_table[] = {
 	{}
 };
 
+static u32 num_perf_groups_per_gt(struct intel_gt *gt)
+{
+	return 1;
+}
+
+static u32 __oa_engine_group(struct intel_engine_cs *engine)
+{
+	if (engine->class == RENDER_CLASS)
+		return PERF_GROUP_OAG;
+	else
+		return PERF_GROUP_INVALID;
+}
+
+static int oa_init_gt(struct intel_gt *gt)
+{
+	u32 num_groups = num_perf_groups_per_gt(gt);
+	struct intel_engine_cs *engine;
+	struct i915_perf_group *g;
+	intel_engine_mask_t tmp;
+
+	g = kcalloc(num_groups, sizeof(*g), GFP_KERNEL);
+	if (!g)
+		return -ENOMEM;
+
+	for_each_engine_masked(engine, gt, ALL_ENGINES, tmp) {
+		u32 index = __oa_engine_group(engine);
+
+		engine->oa_group = NULL;
+		if (index < num_groups) {
+			g[index].num_engines++;
+			engine->oa_group = &g[index];
+		}
+	}
+
+	gt->perf.num_perf_groups = num_groups;
+	gt->perf.group = g;
+
+	return 0;
+}
+
+static int oa_init_engine_groups(struct i915_perf *perf)
+{
+	struct intel_gt *gt;
+	int i, ret;
+
+	for_each_gt(gt, perf->i915, i) {
+		ret = oa_init_gt(gt);
+		if (ret)
+			return ret;
+	}
+
+	return 0;
+}
+
 static void oa_init_supported_formats(struct i915_perf *perf)
 {
 	struct drm_i915_private *i915 = perf->i915;
@@ -4909,7 +4978,7 @@ void i915_perf_init(struct drm_i915_private *i915)
 
 	if (perf->ops.enable_metric_set) {
 		struct intel_gt *gt;
-		int i;
+		int i, ret;
 
 		for_each_gt(gt, i915, i)
 			mutex_init(&gt->perf.lock);
@@ -4948,6 +5017,11 @@ void i915_perf_init(struct drm_i915_private *i915)
 
 		perf->i915 = i915;
 
+		ret = oa_init_engine_groups(perf);
+		if (ret)
+			drm_err(&i915->drm,
+				"OA initialization failed %d\n", ret);
+
 		oa_init_supported_formats(perf);
 	}
 }
@@ -4976,10 +5050,15 @@ void i915_perf_sysctl_unregister(void)
 void i915_perf_fini(struct drm_i915_private *i915)
 {
 	struct i915_perf *perf = &i915->perf;
+	struct intel_gt *gt;
+	int i;
 
 	if (!perf->i915)
 		return;
 
+	for_each_gt(gt, perf->i915, i)
+		kfree(gt->perf.group);
+
 	idr_for_each(&perf->metrics_idr, destroy_config, perf);
 	idr_destroy(&perf->metrics_idr);