Message-ID: <20110111013625.GE2570@nowhere>
Date:	Tue, 11 Jan 2011 02:36:28 +0100
From:	Frederic Weisbecker <fweisbec@...il.com>
To:	Thomas Renninger <trenn@...e.de>
Cc:	Arnaldo Carvalho de Melo <acme@...hat.com>,
	Ingo Molnar <mingo@...e.hu>, linux-perf-users@...r.kernel.org,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: perf timechart broken

On Fri, Jan 07, 2011 at 11:04:37AM +0100, Thomas Renninger wrote:
> Hi,
> 
> Latest x86 tip has another perf timechart issue introduce
> by latest commits.
> 
> ./perf timechart
> gives me:
> "no trace data in the file"
> 
> I reverted  the latest changes and things seem to break
> between d854861c4292a4e675a5d3bfd862c5f7421c81e8
> and
> 69aad6f1ee69546dea8535ab8f3da9f445d57328
> (linux-2.6-x86 tree ids)
> 
> Be aware that ./perf timechart is currently broken
> and segfaults on idle events (in 2.6.36 and 2.6.37).

You mean .36 perf tools faults on .37 kernel? or the opposite?
Is it because power_idle events weren't present on old tools?

Do you have a pointer to those patches?

> I submitted a patch to stable@ but it did not get accepted,
> becaust it's not mainline yet.
> I concentrated on my latest patches instead of making sure
> the tiny fix gets into 2.6.37 first.
> 
> Ingo: It's a bit late for 2.6.37, if there is any chance
> to still get it in, the patch below is against linus master...
> otherwise I'll submit it for stable 36+37 asap.

It's too late for .37, but it's fine, we just need to add
a "Cc: stable@...nel.org" tag in the patch for it to be
backported.

> 
> Anyway to test above regression this mail is about, first
> apply below patch otherwise you get a segfault if idle
> (power_start/end) events got logged.
> 
> Thanks,
> 
>    Thomas
> 
> ---
> perf timechart: Fix segfault on cpu idle (power_start/end) events
> 
> power_end event does not have type and other attributes and thus does
> not match struct power_event.
> Another struct could be created, but data.cpu is fine for fixing the segfault
> and will work as long as C-states got initiated on the same CPU the idle state
> takes place which is the case for all recent HW.
> 
> The power_start/end events get deprecated anyway, thus this is an easy,
> riskless and sufficient solution for the segfault problem.
> 
> Signed-off-by: Thomas Renninger <trenn@...e.de>
> CC: mingo@...e.hu
> CC: arjan@...ux.intel.com
> 
> ---
>  tools/perf/builtin-timechart.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c
> index 9bcc38f..b3028eb 100644
> --- a/tools/perf/builtin-timechart.c
> +++ b/tools/perf/builtin-timechart.c
> @@ -502,7 +502,7 @@ static int process_sample_event(event_t *event, struct perf_session *session)
>  			c_state_start(pe->cpu_id, data.time, pe->value);
>  
>  		if (strcmp(event_str, "power:power_end") == 0)
> -			c_state_end(pe->cpu_id, data.time);
> +			c_state_end(data.cpu, data.time);

On which tree is this based of?

What I have by looking at tip/master is:

		else if (strcmp(event_str, "power:power_end") == 0)
			c_state_end(sample->cpu, sample->time);

>  
>  		if (strcmp(event_str, "power:power_frequency") == 0)
>  			p_state_change(pe->cpu_id, data.time, pe->value);
> --
> To unsubscribe from this list: send the line "unsubscribe linux-perf-users" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists