linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH 0/3] tracing/perf: perf_trace_buf/perf_xxx hacks.
@ 2013-06-18 19:21 Oleg Nesterov
  2013-06-18 19:22 ` [PATCH 1/3] tracing/perf: expand TRACE_EVENT(sched_stat_runtime) Oleg Nesterov
                   ` (4 more replies)
  0 siblings, 5 replies; 21+ messages in thread
From: Oleg Nesterov @ 2013-06-18 19:21 UTC (permalink / raw)
  To: Steven Rostedt, Peter Zijlstra
  Cc: Frederic Weisbecker, Ingo Molnar, Masami Hiramatsu,
	Srikar Dronamraju, zhangwei(Jovi),
	linux-kernel

Hello.

On top of "PATCH 0/3] tracing: more list_empty(perf_events) checks"
series I sent yesterday.

Compile tested only, not for inclusion yet.

But I'll appreciate if you can take a look. I'll try to test this
tomorrow somehow and let you know. Right now I am looking at asm code,
looks correct...

I also compiled the kernel with the additional patch below, everything
compiles except sched/core.o as expected.

Oleg.

--- a/include/trace/ftrace.h
+++ b/include/trace/ftrace.h
@@ -664,6 +664,8 @@ perf_trace_##call(void *__data, proto)					\
 									\
 	__data_size = ftrace_get_offsets_##call(&__data_offsets, args); \
 									\
+	BUILD_BUG_ON(!(__builtin_constant_p(!__task) && !__task));	\
+									\
 	head = this_cpu_ptr(event_call->perf_events);			\
 	if (__builtin_constant_p(!__task) && !__task &&			\
 				hlist_empty(head))			\


^ permalink raw reply	[flat|nested] 21+ messages in thread
* [PATCH RESEND 0/3] Teach perf_trace_##call() to check hlist_empty(perf_events)
@ 2013-07-18 18:30 Oleg Nesterov
  2013-07-18 18:30 ` [PATCH 2/3] tracing/perf: reimplement TP_perf_assign() logic Oleg Nesterov
  0 siblings, 1 reply; 21+ messages in thread
From: Oleg Nesterov @ 2013-07-18 18:30 UTC (permalink / raw)
  To: Frederic Weisbecker, Peter Zijlstra, Steven Rostedt
  Cc: David Ahern, Ingo Molnar, Masami Hiramatsu, zhangwei(Jovi), linux-kernel

Hello.

The patches are the same, I only tried to update the changelogs a bit.
I am also quoting my old email below, to explain what this hack tries
to do.

	Say, "perf record -e sched:sched_switch -p1".

	Every task except /sbin/init will do perf_trace_sched_switch() and
	perf_trace_buf_prepare() + perf_trace_buf_submit for no reason(),
	it doesn't have a counter.

	So it makes sense to add the fast-path check at the start of
	perf_trace_##call(),

		if (hlist_empty(event_call->perf_events))
			return;

	The problem is, we should not do this if __task != NULL (iow, if
	DECLARE_EVENT_CLASS() uses __perf_task()), perf_tp_event() has the
	additional code for this case.

	So we should do

		if (!__task && hlist_empty(event_call->perf_events))
			return;

	But __task is changed by "{ assign; }" block right before
	perf_trace_buf_submit(). Too late for the fast-path check,
	we already called perf_trace_buf_prepare/fetch_regs.

	So. After 2/3 __perf_task() (and __perf_count/addr) is called
	when ftrace_get_offsets_##call(args) evaluates the arguments,
	and we can check !__task && hlist_empty() right after that.

Oleg.


^ permalink raw reply	[flat|nested] 21+ messages in thread
* [PATCH 0/3] Teach perf_trace_##call() to check hlist_empty(perf_events)
@ 2013-08-05 16:50 Oleg Nesterov
  2013-08-05 16:50 ` [PATCH 2/3] tracing/perf: Reimplement TP_perf_assign() logic Oleg Nesterov
  0 siblings, 1 reply; 21+ messages in thread
From: Oleg Nesterov @ 2013-08-05 16:50 UTC (permalink / raw)
  To: Ingo Molnar
  Cc: Steven Rostedt, Frederic Weisbecker, Peter Zijlstra, David Ahern,
	Masami Hiramatsu, zhangwei(Jovi),
	linux-kernel

Sorry for double post, forgot to cc lkml...

On 07/19, Ingo Molnar wrote:
>
> * Oleg Nesterov <oleg@redhat.com> wrote:
>
> > Hello.
> >
> > The patches are the same, I only tried to update the changelogs a bit.
> > I am also quoting my old email below, to explain what this hack tries
> > to do.
> >
> > 	Say, "perf record -e sched:sched_switch -p1".
> >
> > 	Every task except /sbin/init will do perf_trace_sched_switch() and
> > 	perf_trace_buf_prepare() + perf_trace_buf_submit for no reason(),
> > 	it doesn't have a counter.
> >
> > 	So it makes sense to add the fast-path check at the start of
> > 	perf_trace_##call(),
> >
> > 		if (hlist_empty(event_call->perf_events))
> > 			return;
> >
> > 	The problem is, we should not do this if __task != NULL (iow, if
> > 	DECLARE_EVENT_CLASS() uses __perf_task()), perf_tp_event() has the
> > 	additional code for this case.
> >
> > 	So we should do
> >
> > 		if (!__task && hlist_empty(event_call->perf_events))
> > 			return;
> >
> > 	But __task is changed by "{ assign; }" block right before
> > 	perf_trace_buf_submit(). Too late for the fast-path check,
> > 	we already called perf_trace_buf_prepare/fetch_regs.
> >
> > 	So. After 2/3 __perf_task() (and __perf_count/addr) is called
> > 	when ftrace_get_offsets_##call(args) evaluates the arguments,
> > 	and we can check !__task && hlist_empty() right after that.
> >
> > Oleg.
>
> Nice improvement.
>
> Peter, Steve, any objections?

Ingo,

It seems that everybody agree with this hack but it was forgotten,
let me resend it again.

The only change is that I added the following tags:

	Tested-by: David Ahern <dsahern@gmail.com>
	Reviewed-and-Acked-by: Steven Rostedt <rostedt@goodmis.org>

Oleg.

 include/trace/events/sched.h |   22 ++++++++--------------
 include/trace/ftrace.h       |   33 ++++++++++++++++++++-------------
 2 files changed, 28 insertions(+), 27 deletions(-)


^ permalink raw reply	[flat|nested] 21+ messages in thread
end of thread, other threads:[~2013-08-05 16:56 UTC | newest]

Thread overview: 21+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2013-06-18 19:21 [PATCH 0/3] tracing/perf: perf_trace_buf/perf_xxx hacks Oleg Nesterov
2013-06-18 19:22 ` [PATCH 1/3] tracing/perf: expand TRACE_EVENT(sched_stat_runtime) Oleg Nesterov
2013-06-18 19:22 ` [PATCH 2/3] tracing/perf: reimplement TP_perf_assign() logic Oleg Nesterov
2013-06-18 19:22 ` [PATCH 3/3] tracing/perf: Avoid perf_trace_buf_*() in perf_trace_##call() when possible Oleg Nesterov
2013-06-18 20:02   ` Steven Rostedt
2013-06-19 18:12     ` Oleg Nesterov
2013-06-19 18:24       ` Steven Rostedt
2013-06-19 12:10 ` [PATCH 0/3] tracing/perf: perf_trace_buf/perf_xxx hacks Peter Zijlstra
2013-06-19 15:28   ` Oleg Nesterov
2013-06-19 17:51     ` Oleg Nesterov
2013-06-19 18:50       ` David Ahern
2013-06-19 19:58         ` Oleg Nesterov
2013-06-20 18:23           ` Steven Rostedt
2013-06-20 18:35             ` David Ahern
2013-06-20 18:47               ` Steven Rostedt
2013-06-20 18:53                 ` David Ahern
2013-06-20 18:53                 ` Steven Rostedt
2013-06-20 22:20                   ` Steven Rostedt
2013-07-18  3:06 ` Steven Rostedt
2013-07-18 18:30 [PATCH RESEND 0/3] Teach perf_trace_##call() to check hlist_empty(perf_events) Oleg Nesterov
2013-07-18 18:30 ` [PATCH 2/3] tracing/perf: reimplement TP_perf_assign() logic Oleg Nesterov
2013-08-05 16:50 [PATCH 0/3] Teach perf_trace_##call() to check hlist_empty(perf_events) Oleg Nesterov
2013-08-05 16:50 ` [PATCH 2/3] tracing/perf: Reimplement TP_perf_assign() logic Oleg Nesterov

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).