* [PATCH v2] perf: update perf_cgroup time for ancestor cgroup(s)
@ 2018-03-12 16:59 Song Liu
2018-03-14 21:30 ` Song Liu
` (2 more replies)
0 siblings, 3 replies; 4+ messages in thread
From: Song Liu @ 2018-03-12 16:59 UTC (permalink / raw)
To: linux-kernel, peterz, jolsa; +Cc: kernel-team, ephiepark, Song Liu
When a perf_event is attached to parent cgroup, it should count events
for all children cgroups:
parent_group <---- perf_event
\
- child_group <---- process(es)
However, in our tests, we found this perf_event cannot report reliable
results. Here is an example case:
# create cgroups
mkdir -p /sys/fs/cgroup/p/c
# start perf for parent group
perf stat -e instructions -G "p"
# on another console, run test process in child cgroup:
stressapptest -s 2 -M 1000 & echo $! > /sys/fs/cgroup/p/c/cgroup.procs
# after the test process is done, stop perf in the first console shows
<not counted> instructions p
The instruction should not be "not counted" as the process runs in the
child cgroup.
We found this is because perf_event->cgrp and cpuctx->cgrp are not
identical, thus perf_event->cgrp are not updated properly.
This patch fixes this by updating perf_cgroup properly for ancestor
cgroup(s).
Signed-off-by: Song Liu <songliubraving@fb.com>
Reported-by: Ephraim Park <ephiepark@fb.com>
---
kernel/events/core.c | 20 +++++++++++++++-----
1 file changed, 15 insertions(+), 5 deletions(-)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 5789810..6f015ff 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -724,9 +724,14 @@ static inline void __update_cgrp_time(struct perf_cgroup *cgrp)
static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
{
- struct perf_cgroup *cgrp_out = cpuctx->cgrp;
- if (cgrp_out)
- __update_cgrp_time(cgrp_out);
+ struct perf_cgroup *cgrp = cpuctx->cgrp;
+ struct cgroup_subsys_state *css;
+
+ if (cgrp)
+ for (css = &cgrp->css; css; css = css->parent) {
+ cgrp = container_of(css, struct perf_cgroup, css);
+ __update_cgrp_time(cgrp);
+ }
}
static inline void update_cgrp_time_from_event(struct perf_event *event)
@@ -754,6 +759,7 @@ perf_cgroup_set_timestamp(struct task_struct *task,
{
struct perf_cgroup *cgrp;
struct perf_cgroup_info *info;
+ struct cgroup_subsys_state *css;
/*
* ctx->lock held by caller
@@ -764,8 +770,12 @@ perf_cgroup_set_timestamp(struct task_struct *task,
return;
cgrp = perf_cgroup_from_task(task, ctx);
- info = this_cpu_ptr(cgrp->info);
- info->timestamp = ctx->timestamp;
+
+ 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;
+ }
}
static DEFINE_PER_CPU(struct list_head, cgrp_cpuctx_list);
--
2.9.5
^ permalink raw reply related [flat|nested] 4+ messages in thread
* Re: [PATCH v2] perf: update perf_cgroup time for ancestor cgroup(s)
2018-03-12 16:59 [PATCH v2] perf: update perf_cgroup time for ancestor cgroup(s) Song Liu
@ 2018-03-14 21:30 ` Song Liu
2018-03-16 14:24 ` Peter Zijlstra
2018-03-20 11:18 ` [tip:perf/urgent] perf/cgroup: Fix child event counting bug tip-bot for Song Liu
2 siblings, 0 replies; 4+ messages in thread
From: Song Liu @ 2018-03-14 21:30 UTC (permalink / raw)
To: LKML, peterz, jolsa; +Cc: Kernel Team, Ephraim Park
Dear Peter,
Could you please share your comments/suggestions on this patch? We would
like to fix this issue in our kernel, as we are using perf events with
nested cgroups.
Thanks,
Song
> On Mar 12, 2018, at 9:59 AM, Song Liu <songliubraving@fb.com> wrote:
>
> When a perf_event is attached to parent cgroup, it should count events
> for all children cgroups:
>
> parent_group <---- perf_event
> \
> - child_group <---- process(es)
>
> However, in our tests, we found this perf_event cannot report reliable
> results. Here is an example case:
>
> # create cgroups
> mkdir -p /sys/fs/cgroup/p/c
> # start perf for parent group
> perf stat -e instructions -G "p"
>
> # on another console, run test process in child cgroup:
> stressapptest -s 2 -M 1000 & echo $! > /sys/fs/cgroup/p/c/cgroup.procs
>
> # after the test process is done, stop perf in the first console shows
>
> <not counted> instructions p
>
> The instruction should not be "not counted" as the process runs in the
> child cgroup.
>
> We found this is because perf_event->cgrp and cpuctx->cgrp are not
> identical, thus perf_event->cgrp are not updated properly.
>
> This patch fixes this by updating perf_cgroup properly for ancestor
> cgroup(s).
>
> Signed-off-by: Song Liu <songliubraving@fb.com>
> Reported-by: Ephraim Park <ephiepark@fb.com>
> ---
> kernel/events/core.c | 20 +++++++++++++++-----
> 1 file changed, 15 insertions(+), 5 deletions(-)
>
> diff --git a/kernel/events/core.c b/kernel/events/core.c
> index 5789810..6f015ff 100644
> --- a/kernel/events/core.c
> +++ b/kernel/events/core.c
> @@ -724,9 +724,14 @@ static inline void __update_cgrp_time(struct perf_cgroup *cgrp)
>
> static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
> {
> - struct perf_cgroup *cgrp_out = cpuctx->cgrp;
> - if (cgrp_out)
> - __update_cgrp_time(cgrp_out);
> + struct perf_cgroup *cgrp = cpuctx->cgrp;
> + struct cgroup_subsys_state *css;
> +
> + if (cgrp)
> + for (css = &cgrp->css; css; css = css->parent) {
> + cgrp = container_of(css, struct perf_cgroup, css);
> + __update_cgrp_time(cgrp);
> + }
> }
>
> static inline void update_cgrp_time_from_event(struct perf_event *event)
> @@ -754,6 +759,7 @@ perf_cgroup_set_timestamp(struct task_struct *task,
> {
> struct perf_cgroup *cgrp;
> struct perf_cgroup_info *info;
> + struct cgroup_subsys_state *css;
>
> /*
> * ctx->lock held by caller
> @@ -764,8 +770,12 @@ perf_cgroup_set_timestamp(struct task_struct *task,
> return;
>
> cgrp = perf_cgroup_from_task(task, ctx);
> - info = this_cpu_ptr(cgrp->info);
> - info->timestamp = ctx->timestamp;
> +
> + 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;
> + }
> }
>
> static DEFINE_PER_CPU(struct list_head, cgrp_cpuctx_list);
> --
> 2.9.5
>
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [PATCH v2] perf: update perf_cgroup time for ancestor cgroup(s)
2018-03-12 16:59 [PATCH v2] perf: update perf_cgroup time for ancestor cgroup(s) Song Liu
2018-03-14 21:30 ` Song Liu
@ 2018-03-16 14:24 ` Peter Zijlstra
2018-03-20 11:18 ` [tip:perf/urgent] perf/cgroup: Fix child event counting bug tip-bot for Song Liu
2 siblings, 0 replies; 4+ messages in thread
From: Peter Zijlstra @ 2018-03-16 14:24 UTC (permalink / raw)
To: Song Liu; +Cc: linux-kernel, jolsa, kernel-team, ephiepark
On Mon, Mar 12, 2018 at 09:59:43AM -0700, Song Liu wrote:
> When a perf_event is attached to parent cgroup, it should count events
> for all children cgroups:
>
> parent_group <---- perf_event
> \
> - child_group <---- process(es)
>
> However, in our tests, we found this perf_event cannot report reliable
> results. Here is an example case:
>
> # create cgroups
> mkdir -p /sys/fs/cgroup/p/c
> # start perf for parent group
> perf stat -e instructions -G "p"
>
> # on another console, run test process in child cgroup:
> stressapptest -s 2 -M 1000 & echo $! > /sys/fs/cgroup/p/c/cgroup.procs
>
> # after the test process is done, stop perf in the first console shows
>
> <not counted> instructions p
>
> The instruction should not be "not counted" as the process runs in the
> child cgroup.
>
> We found this is because perf_event->cgrp and cpuctx->cgrp are not
> identical, thus perf_event->cgrp are not updated properly.
>
> This patch fixes this by updating perf_cgroup properly for ancestor
> cgroup(s).
>
> Signed-off-by: Song Liu <songliubraving@fb.com>
> Reported-by: Ephraim Park <ephiepark@fb.com>
Yeah, that looks about right, Thanks!
^ permalink raw reply [flat|nested] 4+ messages in thread
* [tip:perf/urgent] perf/cgroup: Fix child event counting bug
2018-03-12 16:59 [PATCH v2] perf: update perf_cgroup time for ancestor cgroup(s) Song Liu
2018-03-14 21:30 ` Song Liu
2018-03-16 14:24 ` Peter Zijlstra
@ 2018-03-20 11:18 ` tip-bot for Song Liu
2 siblings, 0 replies; 4+ messages in thread
From: tip-bot for Song Liu @ 2018-03-20 11:18 UTC (permalink / raw)
To: linux-tip-commits
Cc: hpa, peterz, mingo, vincent.weaver, songliubraving, acme,
torvalds, linux-kernel, tglx, kernel-team, eranian,
alexander.shishkin, jolsa, ephiepark
Commit-ID: c917e0f259908e75bd2a65877e25f9d90c22c848
Gitweb: https://git.kernel.org/tip/c917e0f259908e75bd2a65877e25f9d90c22c848
Author: Song Liu <songliubraving@fb.com>
AuthorDate: Mon, 12 Mar 2018 09:59:43 -0700
Committer: Ingo Molnar <mingo@kernel.org>
CommitDate: Tue, 20 Mar 2018 08:58:47 +0100
perf/cgroup: Fix child event counting bug
When a perf_event is attached to parent cgroup, it should count events
for all children cgroups:
parent_group <---- perf_event
\
- child_group <---- process(es)
However, in our tests, we found this perf_event cannot report reliable
results. Here is an example case:
# create cgroups
mkdir -p /sys/fs/cgroup/p/c
# start perf for parent group
perf stat -e instructions -G "p"
# on another console, run test process in child cgroup:
stressapptest -s 2 -M 1000 & echo $! > /sys/fs/cgroup/p/c/cgroup.procs
# after the test process is done, stop perf in the first console shows
<not counted> instructions p
The instruction should not be "not counted" as the process runs in the
child cgroup.
We found this is because perf_event->cgrp and cpuctx->cgrp are not
identical, thus perf_event->cgrp are not updated properly.
This patch fixes this by updating perf_cgroup properly for ancestor
cgroup(s).
Reported-by: Ephraim Park <ephiepark@fb.com>
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <jolsa@redhat.com>
Cc: <kernel-team@fb.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.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: http://lkml.kernel.org/r/20180312165943.1057894-1-songliubraving@fb.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
kernel/events/core.c | 21 ++++++++++++++++-----
1 file changed, 16 insertions(+), 5 deletions(-)
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 4b838470fac4..709a55b9ad97 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -724,9 +724,15 @@ static inline void __update_cgrp_time(struct perf_cgroup *cgrp)
static inline void update_cgrp_time_from_cpuctx(struct perf_cpu_context *cpuctx)
{
- struct perf_cgroup *cgrp_out = cpuctx->cgrp;
- if (cgrp_out)
- __update_cgrp_time(cgrp_out);
+ struct perf_cgroup *cgrp = cpuctx->cgrp;
+ struct cgroup_subsys_state *css;
+
+ if (cgrp) {
+ for (css = &cgrp->css; css; css = css->parent) {
+ cgrp = container_of(css, struct perf_cgroup, css);
+ __update_cgrp_time(cgrp);
+ }
+ }
}
static inline void update_cgrp_time_from_event(struct perf_event *event)
@@ -754,6 +760,7 @@ perf_cgroup_set_timestamp(struct task_struct *task,
{
struct perf_cgroup *cgrp;
struct perf_cgroup_info *info;
+ struct cgroup_subsys_state *css;
/*
* ctx->lock held by caller
@@ -764,8 +771,12 @@ perf_cgroup_set_timestamp(struct task_struct *task,
return;
cgrp = perf_cgroup_from_task(task, ctx);
- info = this_cpu_ptr(cgrp->info);
- info->timestamp = ctx->timestamp;
+
+ 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;
+ }
}
static DEFINE_PER_CPU(struct list_head, cgrp_cpuctx_list);
^ permalink raw reply related [flat|nested] 4+ messages in thread
end of thread, other threads:[~2018-03-20 11:18 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-03-12 16:59 [PATCH v2] perf: update perf_cgroup time for ancestor cgroup(s) Song Liu
2018-03-14 21:30 ` Song Liu
2018-03-16 14:24 ` Peter Zijlstra
2018-03-20 11:18 ` [tip:perf/urgent] perf/cgroup: Fix child event counting bug tip-bot for Song Liu
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).