From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: rostedt <rostedt@goodmis.org>,
"Paul E. McKenney" <paulmck@linux.vnet.ibm.com>,
linux-kernel <linux-kernel@vger.kernel.org>
Subject: Re: Perf: event wakeup discards sched_waking events
Date: Mon, 14 Jan 2019 16:36:59 -0500 (EST) [thread overview]
Message-ID: <615183542.671.1547501819850.JavaMail.zimbra@efficios.com> (raw)
In-Reply-To: <20190114130957.GA10486@hirez.programming.kicks-ass.net>
----- On Jan 14, 2019, at 8:09 AM, Peter Zijlstra peterz@infradead.org wrote:
> On Thu, Jan 10, 2019 at 12:45:27PM -0500, Mathieu Desnoyers wrote:
>
>> static void perf_pending_event(struct irq_work *entry)
>> {
>> struct perf_event *event = container_of(entry,
>> struct perf_event, pending);
>> int rctx;
>>
>> rctx = perf_swevent_get_recursion_context();
>> /*
>> * If we 'fail' here, that's OK, it means recursion is already disabled
>> * and we won't recurse 'further'.
>> */
>>
>> if (event->pending_disable) {
>> event->pending_disable = 0;
>> perf_event_disable_local(event);
>> }
>>
>> if (event->pending_wakeup) {
>> event->pending_wakeup = 0;
>> perf_event_wakeup(event);
>> }
>>
>> if (rctx >= 0)
>> perf_swevent_put_recursion_context(rctx);
>> }
>>
>> One side-effect of perf_event_wakeup() is to generate a sched_waking
>> event. But I suspect it won't be traced by perf because it is invoked before
>> putting the recursion context.
>>
>> Is there a reason why the wakeup is done before putting the recursion
>> context ?
>
> d525211f9d1b ("perf: Fix irq_work 'tail' recursion")
>
> If we were to allow perf_event_wakeup() to generate its tracepoint, we'd
> be back to square #1, no?
Considering that perf tracing code has side-effects that generate additional
events, it's indeed best not to trace them, otherwise you indeed end up with
tail recursion.
Can ftrace end up in the same situation through rb_wake_up_waiters() ? I suspect
the tail recursion would be hard to trigger if the wakeup only happens once per
page though, unless the events generated end up filling up a page.
FWIW, LTTng avoids this entire issue by using a timer-based polling mechanism
to ensure the tracing code does not call into the scheduler wakeup.
Thanks,
Mathieu
--
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
next prev parent reply other threads:[~2019-01-14 21:37 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-01-10 1:38 Possible use of RCU while in extended QS: idle vs RCU read-side in interrupt vs rcu_eqs_exit Mathieu Desnoyers
2019-01-10 4:13 ` Paul E. McKenney
2019-01-10 6:30 ` Mathieu Desnoyers
2019-01-10 14:19 ` Paul E. McKenney
2019-01-10 16:08 ` Steven Rostedt
2019-01-10 16:44 ` Mathieu Desnoyers
2019-01-10 17:11 ` Mathieu Desnoyers
2019-01-10 17:23 ` Steven Rostedt
2019-01-10 17:25 ` Mathieu Desnoyers
2019-01-10 17:45 ` Perf: event wakeup discards sched_waking events Mathieu Desnoyers
2019-01-14 13:09 ` Peter Zijlstra
2019-01-14 21:36 ` Mathieu Desnoyers [this message]
2019-01-14 22:04 ` Steven Rostedt
2019-01-14 22:31 ` Mathieu Desnoyers
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=615183542.671.1547501819850.JavaMail.zimbra@efficios.com \
--to=mathieu.desnoyers@efficios.com \
--cc=linux-kernel@vger.kernel.org \
--cc=paulmck@linux.vnet.ibm.com \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).