[PATCH] perf cgroups: Fix perf_cgroup_switch schedule in warning

From: Jiri Olsa
Date: Tue Oct 02 2012 - 07:19:19 EST


hi,
I'm getting following warning when running basic cgroup perf stuff:
(perf stat -a -e faults -G krava -- sleep 10)

WARNING: at kernel/events/core.c:397 perf_cgroup_switch+0x1c6/0x1e0()
Hardware name: Montevina platform
Modules linked in:
Pid: 1173, comm: bash Not tainted 3.6.0+ #129
Call Trace:
[<ffffffff8103736f>] warn_slowpath_common+0x7f/0xc0
[<ffffffff810dc025>] ? perf_ctx_lock+0x15/0x30
[<ffffffff810373ca>] warn_slowpath_null+0x1a/0x20
[<ffffffff810e0b96>] perf_cgroup_switch+0x1c6/0x1e0
[<ffffffff810e09d0>] ? perf_event_context_sched_in+0xc0/0xc0
[<ffffffff810e0fa9>] __perf_event_task_sched_in+0xa9/0x200
[<ffffffff81089c67>] ? lock_release_holdtime.part.3+0xc7/0x160
[<ffffffff810689f8>] finish_task_switch+0xb8/0xf0
[<ffffffff814fab2b>] __schedule+0x2eb/0x930
[<ffffffff814fca10>] ? _raw_spin_unlock_irq+0x30/0x60
[<ffffffff8108fd75>] ? trace_hardirqs_on_caller+0x105/0x190
[<ffffffff8108fe0d>] ? trace_hardirqs_on+0xd/0x10
[<ffffffff814fca1b>] ? _raw_spin_unlock_irq+0x3b/0x60
[<ffffffff81055238>] ? start_flush_work+0x108/0x180
[<ffffffff814fb289>] schedule+0x29/0x70
[<ffffffff814f8725>] schedule_timeout+0x1c5/0x210
[<ffffffff8105cecd>] ? add_wait_queue+0x4d/0x60
[<ffffffff814fc9c5>] ? _raw_spin_unlock_irqrestore+0x65/0x80
[<ffffffff8108fd75>] ? trace_hardirqs_on_caller+0x105/0x190
[<ffffffff8108fe0d>] ? trace_hardirqs_on+0xd/0x10
[<ffffffff814fc9a2>] ? _raw_spin_unlock_irqrestore+0x42/0x80
[<ffffffff812b5151>] n_tty_read+0x461/0x8b0
[<ffffffff8106d040>] ? try_to_wake_up+0x310/0x310
[<ffffffff812af2b9>] tty_read+0x99/0xf0
[<ffffffff81128fff>] vfs_read+0xaf/0x180
[<ffffffff8112911d>] sys_read+0x4d/0x90
[<ffffffff81504412>] system_call_fastpath+0x16/0x1b


Looking at the code the cpuctx->cgrp condition seems legal,
and need just some adjustment.

However, I don't fully understand that code and could be
wrong.. just want to get rid of probably wrong warning.. ;)

thanks,
jirka


---
Changing WARN_ON_ONCE condition for PERF_CGROUP_SWIN leg.

It's legal to have cpuctx->cgrp already defined, but it's
not legal not to have events active at the same time.

Cc: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
Cc: Frederic Weisbecker <fweisbec@xxxxxxxxx>
Cc: Ingo Molnar <mingo@xxxxxxx>
Cc: Paul Mackerras <paulus@xxxxxxxxx>
Cc: Peter Zijlstra <a.p.zijlstra@xxxxxxxxx>
CC: Arnaldo Carvalho de Melo <acme@xxxxxxxxxx>
Signed-off-by: Jiri Olsa <jolsa@xxxxxxxxxx>
---
kernel/events/core.c | 3 ++-
1 files changed, 2 insertions(+), 1 deletions(-)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index 7b9df35..733f794 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -394,7 +394,8 @@ void perf_cgroup_switch(struct task_struct *task, int mode)
}

if (mode & PERF_CGROUP_SWIN) {
- WARN_ON_ONCE(cpuctx->cgrp);
+ WARN_ON_ONCE(cpuctx->cgrp && !cpuctx->ctx.is_active);
+
/* set cgrp before ctxsw in to
* allow event_filter_match() to not
* have to pass task around
--
1.7.7.6

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/