drm/i915/pmu: Only enumerate available counters in sysfs

Switch over to dynamically creating device attributes, which are in turn
used by the perf core to expose available counters in sysfs.

This way we do not expose counters which are not avaiable on the current
platform, and are so more consistent between what we reply to open
attempts via the perf_event_open(2), and what is discoverable in sysfs.

v2:
 * Simplify attribute pointer freeing loop.
 * Changed attr init from macro to function.
 * More common error unwind. (Chris Wilson)
 * Rename some locals. (Chris Wilson)

v3:
 * Fixed double semi-colon. (Chris Wilson)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Link: https://patchwork.freedesktop.org/patch/msgid/20180111083525.32394-1-tvrtko.ursulin@linux.intel.com

drivers/gpu/drm/i915/i915_pmu.c

@@ -290,23 +290,44 @@ static void i915_pmu_event_destroy(struct perf_event *event)
 	WARN_ON(event->parent);
 }
 
-static int engine_event_init(struct perf_event *event)
+static int
+engine_event_status(struct intel_engine_cs *engine,
+		    enum drm_i915_pmu_engine_sample sample)
 {
-	struct drm_i915_private *i915 =
-		container_of(event->pmu, typeof(*i915), pmu.base);
-
-	if (!intel_engine_lookup_user(i915, engine_event_class(event),
-				      engine_event_instance(event)))
-		return -ENODEV;
-
-	switch (engine_event_sample(event)) {
+	switch (sample) {
 	case I915_SAMPLE_BUSY:
 	case I915_SAMPLE_WAIT:
 		break;
 	case I915_SAMPLE_SEMA:
+		if (INTEL_GEN(engine->i915) < 6)
+			return -ENODEV;
+		break;
+	default:
+		return -ENOENT;
+	}
+
+	return 0;
+}
+
+static int
+config_status(struct drm_i915_private *i915, u64 config)
+{
+	switch (config) {
+	case I915_PMU_ACTUAL_FREQUENCY:
+		if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915))
+			/* Requires a mutex for sampling! */
+			return -ENODEV;
+		/* Fall-through. */
+	case I915_PMU_REQUESTED_FREQUENCY:
 		if (INTEL_GEN(i915) < 6)
 			return -ENODEV;
 		break;
+	case I915_PMU_INTERRUPTS:
+		break;
+	case I915_PMU_RC6_RESIDENCY:
+		if (!HAS_RC6(i915))
+			return -ENODEV;
+		break;
 	default:
 		return -ENOENT;
 	}
@@ -314,6 +335,20 @@ static int engine_event_init(struct perf_event *event)
 	return 0;
 }
 
+static int engine_event_init(struct perf_event *event)
+{
+	struct drm_i915_private *i915 =
+		container_of(event->pmu, typeof(*i915), pmu.base);
+	struct intel_engine_cs *engine;
+
+	engine = intel_engine_lookup_user(i915, engine_event_class(event),
+					  engine_event_instance(event));
+	if (!engine)
+		return -ENODEV;
+
+	return engine_event_status(engine, engine_event_sample(event));
+}
+
 static int i915_pmu_event_init(struct perf_event *event)
 {
 	struct drm_i915_private *i915 =
@@ -337,30 +372,10 @@ static int i915_pmu_event_init(struct perf_event *event)
 	if (!cpumask_test_cpu(event->cpu, &i915_pmu_cpumask))
 		return -EINVAL;
 
-	if (is_engine_event(event)) {
+	if (is_engine_event(event))
 		ret = engine_event_init(event);
-	} else {
-		ret = 0;
-		switch (event->attr.config) {
-		case I915_PMU_ACTUAL_FREQUENCY:
-			if (IS_VALLEYVIEW(i915) || IS_CHERRYVIEW(i915))
-				 /* Requires a mutex for sampling! */
-				ret = -ENODEV;
-		case I915_PMU_REQUESTED_FREQUENCY:
-			if (INTEL_GEN(i915) < 6)
-				ret = -ENODEV;
-			break;
-		case I915_PMU_INTERRUPTS:
-			break;
-		case I915_PMU_RC6_RESIDENCY:
-			if (!HAS_RC6(i915))
-				ret = -ENODEV;
-			break;
-		default:
-			ret = -ENOENT;
-			break;
-		}
-	}
+	else
+		ret = config_status(i915, event->attr.config);
 	if (ret)
 		return ret;
 
@@ -657,52 +672,9 @@ static ssize_t i915_pmu_event_show(struct device *dev,
 	return sprintf(buf, "config=0x%lx\n", eattr->val);
 }
 
-#define I915_EVENT_ATTR(_name, _config) \
-	(&((struct i915_ext_attribute[]) { \
-		{ .attr = __ATTR(_name, 0444, i915_pmu_event_show, NULL), \
-		  .val = _config, } \
-	})[0].attr.attr)
-
-#define I915_EVENT_STR(_name, _str) \
-	(&((struct perf_pmu_events_attr[]) { \
-		{ .attr	     = __ATTR(_name, 0444, perf_event_sysfs_show, NULL), \
-		  .id	     = 0, \
-		  .event_str = _str, } \
-	})[0].attr.attr)
-
-#define I915_EVENT(_name, _config, _unit) \
-	I915_EVENT_ATTR(_name, _config), \
-	I915_EVENT_STR(_name.unit, _unit)
-
-#define I915_ENGINE_EVENT(_name, _class, _instance, _sample) \
-	I915_EVENT_ATTR(_name, __I915_PMU_ENGINE(_class, _instance, _sample)), \
-	I915_EVENT_STR(_name.unit, "ns")
-
-#define I915_ENGINE_EVENTS(_name, _class, _instance) \
-	I915_ENGINE_EVENT(_name##_instance-busy, _class, _instance, I915_SAMPLE_BUSY), \
-	I915_ENGINE_EVENT(_name##_instance-sema, _class, _instance, I915_SAMPLE_SEMA), \
-	I915_ENGINE_EVENT(_name##_instance-wait, _class, _instance, I915_SAMPLE_WAIT)
-
-static struct attribute *i915_pmu_events_attrs[] = {
-	I915_ENGINE_EVENTS(rcs, I915_ENGINE_CLASS_RENDER, 0),
-	I915_ENGINE_EVENTS(bcs, I915_ENGINE_CLASS_COPY, 0),
-	I915_ENGINE_EVENTS(vcs, I915_ENGINE_CLASS_VIDEO, 0),
-	I915_ENGINE_EVENTS(vcs, I915_ENGINE_CLASS_VIDEO, 1),
-	I915_ENGINE_EVENTS(vecs, I915_ENGINE_CLASS_VIDEO_ENHANCE, 0),
-
-	I915_EVENT(actual-frequency,    I915_PMU_ACTUAL_FREQUENCY,    "MHz"),
-	I915_EVENT(requested-frequency, I915_PMU_REQUESTED_FREQUENCY, "MHz"),
-
-	I915_EVENT_ATTR(interrupts, I915_PMU_INTERRUPTS),
-
-	I915_EVENT(rc6-residency,   I915_PMU_RC6_RESIDENCY,   "ns"),
-
-	NULL,
-};
-
-static const struct attribute_group i915_pmu_events_attr_group = {
+static struct attribute_group i915_pmu_events_attr_group = {
 	.name = "events",
-	.attrs = i915_pmu_events_attrs,
+	/* Patch in attrs at runtime. */
 };
 
 static ssize_t
@@ -720,7 +692,7 @@ static struct attribute *i915_cpumask_attrs[] = {
 	NULL,
 };
 
-static struct attribute_group i915_pmu_cpumask_attr_group = {
+static const struct attribute_group i915_pmu_cpumask_attr_group = {
 	.attrs = i915_cpumask_attrs,
 };
 
@@ -731,6 +703,191 @@ static const struct attribute_group *i915_pmu_attr_groups[] = {
 	NULL
 };
 
+#define __event(__config, __name, __unit) \
+{ \
+	.config = (__config), \
+	.name = (__name), \
+	.unit = (__unit), \
+}
+
+#define __engine_event(__sample, __name) \
+{ \
+	.sample = (__sample), \
+	.name = (__name), \
+}
+
+static struct i915_ext_attribute *
+add_i915_attr(struct i915_ext_attribute *attr, const char *name, u64 config)
+{
+	attr->attr.attr.name = name;
+	attr->attr.attr.mode = 0444;
+	attr->attr.show = i915_pmu_event_show;
+	attr->val = config;
+
+	return ++attr;
+}
+
+static struct perf_pmu_events_attr *
+add_pmu_attr(struct perf_pmu_events_attr *attr, const char *name,
+	     const char *str)
+{
+	attr->attr.attr.name = name;
+	attr->attr.attr.mode = 0444;
+	attr->attr.show = perf_event_sysfs_show;
+	attr->event_str = str;
+
+	return ++attr;
+}
+
+static struct attribute **
+create_event_attributes(struct drm_i915_private *i915)
+{
+	static const struct {
+		u64 config;
+		const char *name;
+		const char *unit;
+	} events[] = {
+		__event(I915_PMU_ACTUAL_FREQUENCY, "actual-frequency", "MHz"),
+		__event(I915_PMU_REQUESTED_FREQUENCY, "requested-frequency", "MHz"),
+		__event(I915_PMU_INTERRUPTS, "interrupts", NULL),
+		__event(I915_PMU_RC6_RESIDENCY, "rc6-residency", "ns"),
+	};
+	static const struct {
+		enum drm_i915_pmu_engine_sample sample;
+		char *name;
+	} engine_events[] = {
+		__engine_event(I915_SAMPLE_BUSY, "busy"),
+		__engine_event(I915_SAMPLE_SEMA, "sema"),
+		__engine_event(I915_SAMPLE_WAIT, "wait"),
+	};
+	unsigned int count = 0;
+	struct perf_pmu_events_attr *pmu_attr = NULL, *pmu_iter;
+	struct i915_ext_attribute *i915_attr = NULL, *i915_iter;
+	struct attribute **attr = NULL, **attr_iter;
+	struct intel_engine_cs *engine;
+	enum intel_engine_id id;
+	unsigned int i;
+
+	/* Count how many counters we will be exposing. */
+	for (i = 0; i < ARRAY_SIZE(events); i++) {
+		if (!config_status(i915, events[i].config))
+			count++;
+	}
+
+	for_each_engine(engine, i915, id) {
+		for (i = 0; i < ARRAY_SIZE(engine_events); i++) {
+			if (!engine_event_status(engine,
+						 engine_events[i].sample))
+				count++;
+		}
+	}
+
+	/* Allocate attribute objects and table. */
+	i915_attr = kzalloc(count * sizeof(*i915_attr), GFP_KERNEL);
+	if (!i915_attr)
+		goto err_alloc;
+
+	pmu_attr = kzalloc(count * sizeof(*pmu_attr), GFP_KERNEL);
+	if (!pmu_attr)
+		goto err_alloc;
+
+	/* Max one pointer of each attribute type plus a termination entry. */
+	attr = kzalloc((count * 2 + 1) * sizeof(attr), GFP_KERNEL);
+	if (!attr)
+		goto err_alloc;
+
+	i915_iter = i915_attr;
+	pmu_iter = pmu_attr;
+	attr_iter = attr;
+
+	/* Initialize supported non-engine counters. */
+	for (i = 0; i < ARRAY_SIZE(events); i++) {
+		char *str;
+
+		if (config_status(i915, events[i].config))
+			continue;
+
+		str = kstrdup(events[i].name, GFP_KERNEL);
+		if (!str)
+			goto err;
+
+		*attr_iter++ = &i915_iter->attr.attr;
+		i915_iter = add_i915_attr(i915_iter, str, events[i].config);
+
+		if (events[i].unit) {
+			str = kasprintf(GFP_KERNEL, "%s.unit", events[i].name);
+			if (!str)
+				goto err;
+
+			*attr_iter++ = &pmu_iter->attr.attr;
+			pmu_iter = add_pmu_attr(pmu_iter, str, events[i].unit);
+		}
+	}
+
+	/* Initialize supported engine counters. */
+	for_each_engine(engine, i915, id) {
+		for (i = 0; i < ARRAY_SIZE(engine_events); i++) {
+			char *str;
+
+			if (engine_event_status(engine,
+						engine_events[i].sample))
+				continue;
+
+			str = kasprintf(GFP_KERNEL, "%s-%s",
+					engine->name, engine_events[i].name);
+			if (!str)
+				goto err;
+
+			*attr_iter++ = &i915_iter->attr.attr;
+			i915_iter =
+				add_i915_attr(i915_iter, str,
+					      __I915_PMU_ENGINE(engine->class,
+								engine->instance,
+								engine_events[i].sample));
+
+			str = kasprintf(GFP_KERNEL, "%s-%s.unit",
+					engine->name, engine_events[i].name);
+			if (!str)
+				goto err;
+
+			*attr_iter++ = &pmu_iter->attr.attr;
+			pmu_iter = add_pmu_attr(pmu_iter, str, "ns");
+		}
+	}
+
+	i915->pmu.i915_attr = i915_attr;
+	i915->pmu.pmu_attr = pmu_attr;
+
+	return attr;
+
+err:;
+	for (attr_iter = attr; *attr_iter; attr_iter++)
+		kfree((*attr_iter)->name);
+
+err_alloc:
+	kfree(attr);
+	kfree(i915_attr);
+	kfree(pmu_attr);
+
+	return NULL;
+}
+
+static void free_event_attributes(struct drm_i915_private *i915)
+{
+	struct attribute **attr_iter = i915_pmu_events_attr_group.attrs;
+
+	for (; *attr_iter; attr_iter++)
+		kfree((*attr_iter)->name);
+
+	kfree(i915_pmu_events_attr_group.attrs);
+	kfree(i915->pmu.i915_attr);
+	kfree(i915->pmu.pmu_attr);
+
+	i915_pmu_events_attr_group.attrs = NULL;
+	i915->pmu.i915_attr = NULL;
+	i915->pmu.pmu_attr = NULL;
+}
+
 static int i915_pmu_cpu_online(unsigned int cpu, struct hlist_node *node)
 {
 	struct i915_pmu *pmu = hlist_entry_safe(node, typeof(*pmu), node);
@@ -806,6 +963,12 @@ void i915_pmu_register(struct drm_i915_private *i915)
 		return;
 	}
 
+	i915_pmu_events_attr_group.attrs = create_event_attributes(i915);
+	if (!i915_pmu_events_attr_group.attrs) {
+		ret = -ENOMEM;
+		goto err;
+	}
+
 	i915->pmu.base.attr_groups	= i915_pmu_attr_groups;
 	i915->pmu.base.task_ctx_nr	= perf_invalid_context;
 	i915->pmu.base.event_init	= i915_pmu_event_init;
@@ -838,6 +1001,7 @@ void i915_pmu_register(struct drm_i915_private *i915)
 	perf_pmu_unregister(&i915->pmu.base);
 err:
 	i915->pmu.base.event_init = NULL;
+	free_event_attributes(i915);
 	DRM_NOTE("Failed to register PMU! (err=%d)\n", ret);
 }
 
@@ -862,4 +1026,5 @@ void i915_pmu_unregister(struct drm_i915_private *i915)
 
 	perf_pmu_unregister(&i915->pmu.base);
 	i915->pmu.base.event_init = NULL;
+	free_event_attributes(i915);
 }

drivers/gpu/drm/i915/i915_pmu.h

@@ -94,6 +94,14 @@ struct i915_pmu {
 	 * struct intel_engine_cs.
 	 */
 	struct i915_pmu_sample sample[__I915_NUM_PMU_SAMPLERS];
+	/**
+	 * @i915_attr: Memory block holding device attributes.
+	 */
+	void *i915_attr;
+	/**
+	 * @pmu_attr: Memory block holding device attributes.
+	 */
+	void *pmu_attr;
 };
 
 #ifdef CONFIG_PERF_EVENTS