From: Ian Rogers irogers@google.com
[ Upstream commit fd7d55172d1e2e501e6da0a5c1de25f06612dc2e ]
Currently perf_rotate_context assumes that if the context's nr_events != nr_active a rotation is necessary for perf event multiplexing. With cgroups, nr_events is the total count of events for all cgroups and nr_active will not include events in a cgroup other than the current task's. This makes rotation appear necessary for cgroups when it is not.
Add a perf_event_context flag that is set when rotation is necessary. Clear the flag during sched_out and set it when a flexible sched_in fails due to resources.
Signed-off-by: Ian Rogers irogers@google.com Signed-off-by: Peter Zijlstra (Intel) peterz@infradead.org Cc: Alexander Shishkin alexander.shishkin@linux.intel.com Cc: Arnaldo Carvalho de Melo acme@kernel.org Cc: Arnaldo Carvalho de Melo acme@redhat.com Cc: Borislav Petkov bp@alien8.de Cc: Jiri Olsa jolsa@redhat.com Cc: Kan Liang kan.liang@linux.intel.com Cc: Linus Torvalds torvalds@linux-foundation.org Cc: Namhyung Kim namhyung@kernel.org Cc: Peter Zijlstra peterz@infradead.org Cc: Stephane Eranian eranian@google.com Cc: Thomas Gleixner tglx@linutronix.de Cc: Vince Weaver vincent.weaver@maine.edu Link: https://lkml.kernel.org/r/20190601082722.44543-1-irogers@google.com Signed-off-by: Ingo Molnar mingo@kernel.org Cc: stable@vger.kernel.org # 4.19+ Signed-off-by: Wen Yang wenyang@linux.alibaba.com --- include/linux/perf_event.h | 5 +++++ kernel/events/core.c | 42 ++++++++++++++++++++++-------------------- 2 files changed, 27 insertions(+), 20 deletions(-)
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h index 0d35c4d..99fc355 100644 --- a/include/linux/perf_event.h +++ b/include/linux/perf_event.h @@ -747,6 +747,11 @@ struct perf_event_context { int nr_stat; int nr_freq; int rotate_disable; + /* + * Set when nr_events != nr_active, except tolerant to events not + * necessary to be active due to scheduling constraints, such as cgroups. + */ + int rotate_necessary; refcount_t refcount; struct task_struct *task;
diff --git a/kernel/events/core.c b/kernel/events/core.c index 28fa3e7..e7b7e27 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -2947,6 +2947,12 @@ static void ctx_sched_out(struct perf_event_context *ctx, if (!ctx->nr_active || !(is_active & EVENT_ALL)) return;
+ /* + * If we had been multiplexing, no rotations are necessary, now no events + * are active. + */ + ctx->rotate_necessary = 0; + perf_pmu_disable(ctx->pmu); if (is_active & EVENT_PINNED) { list_for_each_entry_safe(event, tmp, &ctx->pinned_active, active_list) @@ -3314,10 +3320,13 @@ static int flexible_sched_in(struct perf_event *event, void *data) return 0;
if (group_can_go_on(event, sid->cpuctx, sid->can_add_hw)) { - if (!group_sched_in(event, sid->cpuctx, sid->ctx)) - list_add_tail(&event->active_list, &sid->ctx->flexible_active); - else + int ret = group_sched_in(event, sid->cpuctx, sid->ctx); + if (ret) { sid->can_add_hw = 0; + sid->ctx->rotate_necessary = 1; + return 0; + } + list_add_tail(&event->active_list, &sid->ctx->flexible_active); }
return 0; @@ -3685,24 +3694,17 @@ static void rotate_ctx(struct perf_event_context *ctx, struct perf_event *event) static bool perf_rotate_context(struct perf_cpu_context *cpuctx) { struct perf_event *cpu_event = NULL, *task_event = NULL; - bool cpu_rotate = false, task_rotate = false; - struct perf_event_context *ctx = NULL; + struct perf_event_context *task_ctx = NULL; + int cpu_rotate, task_rotate;
/* * Since we run this from IRQ context, nobody can install new * events, thus the event count values are stable. */
- if (cpuctx->ctx.nr_events) { - if (cpuctx->ctx.nr_events != cpuctx->ctx.nr_active) - cpu_rotate = true; - } - - ctx = cpuctx->task_ctx; - if (ctx && ctx->nr_events) { - if (ctx->nr_events != ctx->nr_active) - task_rotate = true; - } + cpu_rotate = cpuctx->ctx.rotate_necessary; + task_ctx = cpuctx->task_ctx; + task_rotate = task_ctx ? task_ctx->rotate_necessary : 0;
if (!(cpu_rotate || task_rotate)) return false; @@ -3711,7 +3713,7 @@ static bool perf_rotate_context(struct perf_cpu_context *cpuctx) perf_pmu_disable(cpuctx->ctx.pmu);
if (task_rotate) - task_event = ctx_first_active(ctx); + task_event = ctx_first_active(task_ctx); if (cpu_rotate) cpu_event = ctx_first_active(&cpuctx->ctx);
@@ -3719,17 +3721,17 @@ static bool perf_rotate_context(struct perf_cpu_context *cpuctx) * As per the order given at ctx_resched() first 'pop' task flexible * and then, if needed CPU flexible. */ - if (task_event || (ctx && cpu_event)) - ctx_sched_out(ctx, cpuctx, EVENT_FLEXIBLE); + if (task_event || (task_ctx && cpu_event)) + ctx_sched_out(task_ctx, cpuctx, EVENT_FLEXIBLE); if (cpu_event) cpu_ctx_sched_out(cpuctx, EVENT_FLEXIBLE);
if (task_event) - rotate_ctx(ctx, task_event); + rotate_ctx(task_ctx, task_event); if (cpu_event) rotate_ctx(&cpuctx->ctx, cpu_event);
- perf_event_sched_in(cpuctx, ctx, current); + perf_event_sched_in(cpuctx, task_ctx, current);
perf_pmu_enable(cpuctx->ctx.pmu); perf_ctx_unlock(cpuctx, cpuctx->task_ctx);
From: Song Liu songliubraving@fb.com
[ Upstream commit 7fa343b7fdc4f351de4e3f28d5c285937dd1f42f ]
In perf_rotate_context(), when the first cpu flexible event fail to schedule, cpu_rotate is 1, while cpu_event is NULL. Since cpu_event is NULL, perf_rotate_context will _NOT_ call cpu_ctx_sched_out(), thus cpuctx->ctx.is_active will have EVENT_FLEXIBLE set. Then, the next perf_event_sched_in() will skip all cpu flexible events because of the EVENT_FLEXIBLE bit.
In the next call of perf_rotate_context(), cpu_rotate stays 1, and cpu_event stays NULL, so this process repeats. The end result is, flexible events on this cpu will not be scheduled (until another event being added to the cpuctx).
Here is an easy repro of this issue. On Intel CPUs, where ref-cycles could only use one counter, run one pinned event for ref-cycles, one flexible event for ref-cycles, and one flexible event for cycles. The flexible ref-cycles is never scheduled, which is expected. However, because of this issue, the cycles event is never scheduled either.
$ perf stat -e ref-cycles:D,ref-cycles,cycles -C 5 -I 1000
time counts unit events 1.000152973 15,412,480 ref-cycles:D 1.000152973 <not counted> ref-cycles (0.00%) 1.000152973 <not counted> cycles (0.00%) 2.000486957 18,263,120 ref-cycles:D 2.000486957 <not counted> ref-cycles (0.00%) 2.000486957 <not counted> cycles (0.00%)
To fix this, when the flexible_active list is empty, try rotate the first event in the flexible_groups. Also, rename ctx_first_active() to ctx_event_to_rotate(), which is more accurate.
Signed-off-by: Song Liu songliubraving@fb.com Signed-off-by: Peter Zijlstra (Intel) peterz@infradead.org Cc: kernel-team@fb.com Cc: Arnaldo Carvalho de Melo acme@kernel.org Cc: Linus Torvalds torvalds@linux-foundation.org Cc: Peter Zijlstra peterz@infradead.org Cc: Sasha Levin sashal@kernel.org Cc: Thomas Gleixner tglx@linutronix.de Fixes: 8d5bce0c37fa ("perf/core: Optimize perf_rotate_context() event scheduling") Link: https://lkml.kernel.org/r/20191008165949.920548-1-songliubraving@fb.com Signed-off-by: Ingo Molnar mingo@kernel.org Cc: stable@vger.kernel.org # 4.19+ Signed-off-by: Wen Yang wenyang@linux.alibaba.com --- kernel/events/core.c | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-)
diff --git a/kernel/events/core.c b/kernel/events/core.c index e7b7e27..d8c4f84 100644 --- a/kernel/events/core.c +++ b/kernel/events/core.c @@ -3684,11 +3684,23 @@ static void rotate_ctx(struct perf_event_context *ctx, struct perf_event *event) perf_event_groups_insert(&ctx->flexible_groups, event); }
+/* pick an event from the flexible_groups to rotate */ static inline struct perf_event * -ctx_first_active(struct perf_event_context *ctx) +ctx_event_to_rotate(struct perf_event_context *ctx) { - return list_first_entry_or_null(&ctx->flexible_active, - struct perf_event, active_list); + struct perf_event *event; + + /* pick the first active flexible event */ + event = list_first_entry_or_null(&ctx->flexible_active, + struct perf_event, active_list); + + /* if no active flexible event, pick the first event */ + if (!event) { + event = rb_entry_safe(rb_first(&ctx->flexible_groups.tree), + typeof(*event), group_node); + } + + return event; }
static bool perf_rotate_context(struct perf_cpu_context *cpuctx) @@ -3713,9 +3725,9 @@ static bool perf_rotate_context(struct perf_cpu_context *cpuctx) perf_pmu_disable(cpuctx->ctx.pmu);
if (task_rotate) - task_event = ctx_first_active(task_ctx); + task_event = ctx_event_to_rotate(task_ctx); if (cpu_rotate) - cpu_event = ctx_first_active(&cpuctx->ctx); + cpu_event = ctx_event_to_rotate(&cpuctx->ctx);
/* * As per the order given at ctx_resched() first 'pop' task flexible
linux-stable-mirror@lists.linaro.org
-
Wen Yang