X-Git-Url: http://git.monstr.eu/?a=blobdiff_plain;f=kernel%2Fevents%2Fcore.c;h=479c9e672ec4371219d89cf35d16cb9470a9bc4c;hb=f8c7e4ede46fe63ff10000669652648aab09d112;hp=fc18664f49b08743883b4ac0758ccdc58c7e7e08;hpb=1d1df41c5a33359a00e919d54eaebfb789711fdc;p=linux-2.6-microblaze.git

diff --git a/kernel/events/core.c b/kernel/events/core.c
index fc18664f49b0..479c9e672ec4 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -674,6 +674,23 @@ perf_event_set_state(struct perf_event *event, enum perf_event_state state)
 	WRITE_ONCE(event->state, state);
 }
 
+/*
+ * UP store-release, load-acquire
+ */
+
+#define __store_release(ptr, val)					\
+do {									\
+	barrier();							\
+	WRITE_ONCE(*(ptr), (val));					\
+} while (0)
+
+#define __load_acquire(ptr)						\
+({									\
+	__unqual_scalar_typeof(*(ptr)) ___p = READ_ONCE(*(ptr));	\
+	barrier();							\
+	___p;								\
+})
+
 #ifdef CONFIG_CGROUP_PERF
 
 static inline bool
@@ -719,34 +736,51 @@ static inline u64 perf_cgroup_event_time(struct perf_event *event)
 	return t->time;
 }
 
-static inline void __update_cgrp_time(struct perf_cgroup *cgrp)
+static inline u64 perf_cgroup_event_time_now(struct perf_event *event, u64 now)
 {
-	struct perf_cgroup_info *info;
-	u64 now;
-
-	now = perf_clock();
+	struct perf_cgroup_info *t;
 
-	info = this_cpu_ptr(cgrp->info);
+	t = per_cpu_ptr(event->cgrp->info, event->cpu);
+	if (!__load_acquire(&t->active))
+		return t->time;
+	now += READ_ONCE(t->timeoffset);
+	return now;
+}
 
-	info->time += now - info->timestamp;
+static inline void __update_cgrp_time(struct perf_cgroup_info *info, u64 now, bool adv)
+{
+	if (adv)
+		info->time += now - info->timestamp;
 	info->timestamp = now;
+	/*
+	 * see update_context_time()
+	 */
+	WRITE_ONCE(info->timeoffset, info->time - info->timestamp);
 }
 
-static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
+static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx, bool final)
 {
 	struct perf_cgroup *cgrp = cpuctx->cgrp;
 	struct cgroup_subsys_state *css;
+	struct perf_cgroup_info *info;
 
 	if (cgrp) {
+		u64 now = perf_clock();
+
 		for (css = &cgrp->css; css; css = css->parent) {
 			cgrp = container_of(css, struct perf_cgroup, css);
-			__update_cgrp_time(cgrp);
+			info = this_cpu_ptr(cgrp->info);
+
+			__update_cgrp_time(info, now, true);
+			if (final)
+				__store_release(&info->active, 0);
 		}
 	}
 }
 
 static inline void update_cgrp_time_from_event(struct perf_event *event)
 {
+	struct perf_cgroup_info *info;
 	struct perf_cgroup *cgrp;
 
 	/*
@@ -760,8 +794,10 @@ static inline void update_cgrp_time_from_event(struct perf_event *event)
 	/*
 	 * Do not update time when cgroup is not active
 	 */
-	if (cgroup_is_descendant(cgrp->css.cgroup, event->cgrp->css.cgroup))
-		__update_cgrp_time(event->cgrp);
+	if (cgroup_is_descendant(cgrp->css.cgroup, event->cgrp->css.cgroup)) {
+		info = this_cpu_ptr(event->cgrp->info);
+		__update_cgrp_time(info, perf_clock(), true);
+	}
 }
 
 static inline void
@@ -785,7 +821,8 @@ perf_cgroup_set_timestamp(struct task_struct *task,
 	for (css = &cgrp->css; css; css = css->parent) {
 		cgrp = container_of(css, struct perf_cgroup, css);
 		info = this_cpu_ptr(cgrp->info);
-		info->timestamp = ctx->timestamp;
+		__update_cgrp_time(info, ctx->timestamp, false);
+		__store_release(&info->active, 1);
 	}
 }
 
@@ -981,14 +1018,6 @@ out:
 	return ret;
 }
 
-static inline void
-perf_cgroup_set_shadow_time(struct perf_event *event, u64 now)
-{
-	struct perf_cgroup_info *t;
-	t = per_cpu_ptr(event->cgrp->info, event->cpu);
-	event->shadow_ctx_time = now - t->timestamp;
-}
-
 static inline void
 perf_cgroup_event_enable(struct perf_event *event, struct perf_event_context *ctx)
 {
@@ -1066,7 +1095,8 @@ static inline void update_cgrp_time_from_event(struct perf_event *event)
 {
 }
 
-static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
+static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx,
+						bool final)
 {
 }
 
@@ -1098,12 +1128,12 @@ perf_cgroup_switch(struct task_struct *task, struct task_struct *next)
 {
 }
 
-static inline void
-perf_cgroup_set_shadow_time(struct perf_event *event, u64 now)
+static inline u64 perf_cgroup_event_time(struct perf_event *event)
 {
+	return 0;
 }
 
-static inline u64 perf_cgroup_event_time(struct perf_event *event)
+static inline u64 perf_cgroup_event_time_now(struct perf_event *event, u64 now)
 {
 	return 0;
 }
@@ -1525,22 +1555,59 @@ static void perf_unpin_context(struct perf_event_context *ctx)
 /*
  * Update the record of the current time in a context.
  */
-static void update_context_time(struct perf_event_context *ctx)
+static void __update_context_time(struct perf_event_context *ctx, bool adv)
 {
 	u64 now = perf_clock();
 
-	ctx->time += now - ctx->timestamp;
+	if (adv)
+		ctx->time += now - ctx->timestamp;
 	ctx->timestamp = now;
+
+	/*
+	 * The above: time' = time + (now - timestamp), can be re-arranged
+	 * into: time` = now + (time - timestamp), which gives a single value
+	 * offset to compute future time without locks on.
+	 *
+	 * See perf_event_time_now(), which can be used from NMI context where
+	 * it's (obviously) not possible to acquire ctx->lock in order to read
+	 * both the above values in a consistent manner.
+	 */
+	WRITE_ONCE(ctx->timeoffset, ctx->time - ctx->timestamp);
+}
+
+static void update_context_time(struct perf_event_context *ctx)
+{
+	__update_context_time(ctx, true);
 }
 
 static u64 perf_event_time(struct perf_event *event)
 {
 	struct perf_event_context *ctx = event->ctx;
 
+	if (unlikely(!ctx))
+		return 0;
+
 	if (is_cgroup_event(event))
 		return perf_cgroup_event_time(event);
 
-	return ctx ? ctx->time : 0;
+	return ctx->time;
+}
+
+static u64 perf_event_time_now(struct perf_event *event, u64 now)
+{
+	struct perf_event_context *ctx = event->ctx;
+
+	if (unlikely(!ctx))
+		return 0;
+
+	if (is_cgroup_event(event))
+		return perf_cgroup_event_time_now(event, now);
+
+	if (!(__load_acquire(&ctx->is_active) & EVENT_TIME))
+		return ctx->time;
+
+	now += READ_ONCE(ctx->timeoffset);
+	return now;
 }
 
 static enum event_type_t get_event_type(struct perf_event *event)
@@ -2350,7 +2417,7 @@ __perf_remove_from_context(struct perf_event *event,
 
 	if (ctx->is_active & EVENT_TIME) {
 		update_context_time(ctx);
-		update_cgrp_time_from_cpuctx(cpuctx);
+		update_cgrp_time_from_cpuctx(cpuctx, false);
 	}
 
 	event_sched_out(event, cpuctx, ctx);
@@ -2361,6 +2428,9 @@ __perf_remove_from_context(struct perf_event *event,
 	list_del_event(event, ctx);
 
 	if (!ctx->nr_events && ctx->is_active) {
+		if (ctx == &cpuctx->ctx)
+			update_cgrp_time_from_cpuctx(cpuctx, true);
+
 		ctx->is_active = 0;
 		ctx->rotate_necessary = 0;
 		if (ctx->task) {
@@ -2482,40 +2552,6 @@ void perf_event_disable_inatomic(struct perf_event *event)
 	irq_work_queue(&event->pending);
 }
 
-static void perf_set_shadow_time(struct perf_event *event,
-				 struct perf_event_context *ctx)
-{
-	/*
-	 * use the correct time source for the time snapshot
-	 *
-	 * We could get by without this by leveraging the
-	 * fact that to get to this function, the caller
-	 * has most likely already called update_context_time()
-	 * and update_cgrp_time_xx() and thus both timestamp
-	 * are identical (or very close). Given that tstamp is,
-	 * already adjusted for cgroup, we could say that:
-	 *    tstamp - ctx->timestamp
-	 * is equivalent to
-	 *    tstamp - cgrp->timestamp.
-	 *
-	 * Then, in perf_output_read(), the calculation would
-	 * work with no changes because:
-	 * - event is guaranteed scheduled in
-	 * - no scheduled out in between
-	 * - thus the timestamp would be the same
-	 *
-	 * But this is a bit hairy.
-	 *
-	 * So instead, we have an explicit cgroup call to remain
-	 * within the time source all along. We believe it
-	 * is cleaner and simpler to understand.
-	 */
-	if (is_cgroup_event(event))
-		perf_cgroup_set_shadow_time(event, event->tstamp);
-	else
-		event->shadow_ctx_time = event->tstamp - ctx->timestamp;
-}
-
 #define MAX_INTERRUPTS (~0ULL)
 
 static void perf_log_throttle(struct perf_event *event, int enable);
@@ -2556,8 +2592,6 @@ event_sched_in(struct perf_event *event,
 
 	perf_pmu_disable(event->pmu);
 
-	perf_set_shadow_time(event, ctx);
-
 	perf_log_itrace_start(event);
 
 	if (event->pmu->add(event, PERF_EF_START)) {
@@ -3251,16 +3285,6 @@ static void ctx_sched_out(struct perf_event_context *ctx,
 		return;
 	}
 
-	ctx->is_active &= ~event_type;
-	if (!(ctx->is_active & EVENT_ALL))
-		ctx->is_active = 0;
-
-	if (ctx->task) {
-		WARN_ON_ONCE(cpuctx->task_ctx != ctx);
-		if (!ctx->is_active)
-			cpuctx->task_ctx = NULL;
-	}
-
 	/*
 	 * Always update time if it was set; not only when it changes.
 	 * Otherwise we can 'forget' to update time for any but the last
@@ -3274,7 +3298,22 @@ static void ctx_sched_out(struct perf_event_context *ctx,
 	if (is_active & EVENT_TIME) {
 		/* update (and stop) ctx time */
 		update_context_time(ctx);
-		update_cgrp_time_from_cpuctx(cpuctx);
+		update_cgrp_time_from_cpuctx(cpuctx, ctx == &cpuctx->ctx);
+		/*
+		 * CPU-release for the below ->is_active store,
+		 * see __load_acquire() in perf_event_time_now()
+		 */
+		barrier();
+	}
+
+	ctx->is_active &= ~event_type;
+	if (!(ctx->is_active & EVENT_ALL))
+		ctx->is_active = 0;
+
+	if (ctx->task) {
+		WARN_ON_ONCE(cpuctx->task_ctx != ctx);
+		if (!ctx->is_active)
+			cpuctx->task_ctx = NULL;
 	}
 
 	is_active ^= ctx->is_active; /* changed bits */
@@ -3711,13 +3750,19 @@ static noinline int visit_groups_merge(struct perf_cpu_context *cpuctx,
 	return 0;
 }
 
+/*
+ * Because the userpage is strictly per-event (there is no concept of context,
+ * so there cannot be a context indirection), every userpage must be updated
+ * when context time starts :-(
+ *
+ * IOW, we must not miss EVENT_TIME edges.
+ */
 static inline bool event_update_userpage(struct perf_event *event)
 {
 	if (likely(!atomic_read(&event->mmap_count)))
 		return false;
 
 	perf_event_update_time(event);
-	perf_set_shadow_time(event, event->ctx);
 	perf_event_update_userpage(event);
 
 	return true;
@@ -3801,13 +3846,23 @@ ctx_sched_in(struct perf_event_context *ctx,
 	     struct task_struct *task)
 {
 	int is_active = ctx->is_active;
-	u64 now;
 
 	lockdep_assert_held(&ctx->lock);
 
 	if (likely(!ctx->nr_events))
 		return;
 
+	if (is_active ^ EVENT_TIME) {
+		/* start ctx time */
+		__update_context_time(ctx, false);
+		perf_cgroup_set_timestamp(task, ctx);
+		/*
+		 * CPU-release for the below ->is_active store,
+		 * see __load_acquire() in perf_event_time_now()
+		 */
+		barrier();
+	}
+
 	ctx->is_active |= (event_type | EVENT_TIME);
 	if (ctx->task) {
 		if (!is_active)
@@ -3818,13 +3873,6 @@ ctx_sched_in(struct perf_event_context *ctx,
 
 	is_active ^= ctx->is_active; /* changed bits */
 
-	if (is_active & EVENT_TIME) {
-		/* start ctx time */
-		now = perf_clock();
-		ctx->timestamp = now;
-		perf_cgroup_set_timestamp(task, ctx);
-	}
-
 	/*
 	 * First go through the list and put on any pinned groups
 	 * in order to give them the best chance of going on.
@@ -4418,6 +4466,18 @@ static inline u64 perf_event_count(struct perf_event *event)
 	return local64_read(&event->count) + atomic64_read(&event->child_count);
 }
 
+static void calc_timer_values(struct perf_event *event,
+				u64 *now,
+				u64 *enabled,
+				u64 *running)
+{
+	u64 ctx_time;
+
+	*now = perf_clock();
+	ctx_time = perf_event_time_now(event, *now);
+	__perf_update_times(event, ctx_time, enabled, running);
+}
+
 /*
  * NMI-safe method to read a local event, that is an event that
  * is:
@@ -4477,10 +4537,9 @@ int perf_event_read_local(struct perf_event *event, u64 *value,
 
 	*value = local64_read(&event->count);
 	if (enabled || running) {
-		u64 now = event->shadow_ctx_time + perf_clock();
-		u64 __enabled, __running;
+		u64 __enabled, __running, __now;;
 
-		__perf_update_times(event, now, &__enabled, &__running);
+		calc_timer_values(event, &__now, &__enabled, &__running);
 		if (enabled)
 			*enabled = __enabled;
 		if (running)
@@ -5802,18 +5861,6 @@ static int perf_event_index(struct perf_event *event)
 	return event->pmu->event_idx(event);
 }
 
-static void calc_timer_values(struct perf_event *event,
-				u64 *now,
-				u64 *enabled,
-				u64 *running)
-{
-	u64 ctx_time;
-
-	*now = perf_clock();
-	ctx_time = event->shadow_ctx_time + *now;
-	__perf_update_times(event, ctx_time, enabled, running);
-}
-
 static void perf_event_init_userpage(struct perf_event *event)
 {
 	struct perf_event_mmap_page *userpg;
@@ -6353,7 +6400,6 @@ accounting:
 		ring_buffer_attach(event, rb);
 
 		perf_event_update_time(event);
-		perf_set_shadow_time(event, event->ctx);
 		perf_event_init_userpage(event);
 		perf_event_update_userpage(event);
 	} else {