Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1303936 > unrolled thread

Re: [RFC] perf record: missing buildid for callstack modules

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [RFC] perf record: missing buildid for callstack modules Andi Kleen <ak@linux.intel.com> - 2016-01-07 23:00 +0100
    Re: [RFC] perf record: missing buildid for callstack modules Stephane Eranian <eranian@google.com> - 2016-01-07 23:10 +0100

#1303936 — Re: [RFC] perf record: missing buildid for callstack modules

On Thu, Jan 07, 2016 at 01:56:14PM -0800, Stephane Eranian wrote:
> Hi,
> 
> Whenever you do:
> 
>     $ perf record -g -a sleep 10
> 
> Perf will collect the callstack for each sample. At the end of the
> run, perf record
> adds the buildid for all dso with at least one sample. But when it does this, it
> only looks at the sampled IP and ignore the modules traversed by the callstack.
> That means that, it is not possible to uniquely identify the modules executed,
> unless they had at least one IP sample captured. But this is not
> always the case.
> 
> How about providing an option to perf record to force collecting
> buildid for all IPs
> captured in the callstack? I understand that would cost more at the end of the
> collection, but this would be beneficial to several monitoring scenarios.

If you do that I would rather collect buildid for everything that is synthesized
(everything in /proc)

-Andi

[toc] | [next] | [standalone]

#1303941

On Thu, Jan 7, 2016 at 1:57 PM, Andi Kleen <ak@linux.intel.com> wrote:
> On Thu, Jan 07, 2016 at 01:56:14PM -0800, Stephane Eranian wrote:
>> Hi,
>>
>> Whenever you do:
>>
>>     $ perf record -g -a sleep 10
>>
>> Perf will collect the callstack for each sample. At the end of the
>> run, perf record
>> adds the buildid for all dso with at least one sample. But when it does this, it
>> only looks at the sampled IP and ignore the modules traversed by the callstack.
>> That means that, it is not possible to uniquely identify the modules executed,
>> unless they had at least one IP sample captured. But this is not
>> always the case.
>>
>> How about providing an option to perf record to force collecting
>> buildid for all IPs
>> captured in the callstack? I understand that would cost more at the end of the
>> collection, but this would be beneficial to several monitoring scenarios.
>
> If you do that I would rather collect buildid for everything that is synthesized
> (everything in /proc)
>
But that would be much more than whatever would be captured with hits in
callstacks.

A valid scenario here is sampling on tracepoints (which are kernel only events)
but you want to know where the syscall originated from or which user code was
interrupted.

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web