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

Groups > linux.kernel > #1664393 > unrolled thread

[PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode

Started byDavid Carrillo-Cisneros <davidcc@google.com>
First post2017-06-13 06:40 +0200
Last post2017-06-14 20:40 +0200
Articles 6 on this page of 26 — 2 participants

Back to article view | Back to linux.kernel

Contents

  [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 16/16] perf header: add event desc to pipe-mode header David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 03/16] perf header: fail on write_padded error David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 13/16] perf header: change FEAT_OP* macros David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 02/16] perf header: add PROCESS_STR_FUN macro David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 10/16] perf header: use struct feat_fd in read header records David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 07/16] perf header: use struct feat_fd for print David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 01/16] perf header: encapsulate read and swap David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 04/16] perf util: add const modifier to buf in "writen" function David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 11/16] perf header: make write_pmu_mappings pipe-mode friendly David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 09/16] perf header: don't pass struct perf_file_section to process_##_feat David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
      Re: [PATCH v4 09/16] perf header: don't pass struct  perf_file_section to process_##_feat Jiri Olsa <jolsa@redhat.com> - 2017-06-13 21:00 +0200
        Re: [PATCH v4 09/16] perf header: don't pass struct perf_file_section  to process_##_feat David Carrillo-Cisneros <davidcc@google.com> - 2017-06-14 04:00 +0200
    [PATCH v4 12/16] perf header: add a buffer to struct feat_fd David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
      Re: [PATCH v4 12/16] perf header: add a buffer to struct feat_fd Jiri Olsa <jolsa@redhat.com> - 2017-06-13 21:00 +0200
        Re: [PATCH v4 12/16] perf header: add a buffer to struct feat_fd David Carrillo-Cisneros <davidcc@google.com> - 2017-06-14 04:00 +0200
    [PATCH v4 14/16] perf tool: add show_feature_header to perf_tool David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
      Re: [PATCH v4 14/16] perf tool: add show_feature_header to perf_tool Jiri Olsa <jolsa@redhat.com> - 2017-06-13 21:00 +0200
    [PATCH v4 06/16] perf header: add struct feat_fd for write David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    [PATCH v4 08/16] perf header: use struct feat_fd to process header records David Carrillo-Cisneros <davidcc@google.com> - 2017-06-13 06:40 +0200
    Re: [PATCH v4 00/16] perf tool: add meta-data header support for  pipe-mode Jiri Olsa <jolsa@redhat.com> - 2017-06-13 21:00 +0200
      Re: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode David Carrillo-Cisneros <davidcc@google.com> - 2017-06-14 08:50 +0200
    Re: [PATCH v4 00/16] perf tool: add meta-data header support for  pipe-mode Jiri Olsa <jolsa@redhat.com> - 2017-06-13 21:00 +0200
      Re: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode David Carrillo-Cisneros <davidcc@google.com> - 2017-06-14 09:00 +0200
        Re: [PATCH v4 00/16] perf tool: add meta-data header support for  pipe-mode Jiri Olsa <jolsa@redhat.com> - 2017-06-14 09:30 +0200
          Re: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode David Carrillo-Cisneros <davidcc@google.com> - 2017-06-14 20:40 +0200

Page 2 of 2 — ← Prev page 1 [2]

#1665106 — Re: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode

FromJiri Olsa <jolsa@redhat.com>
Date2017-06-13 21:00 +0200
SubjectRe: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode
Message-ID<tRZcJ-380-1@gated-at.bofh.it>
In reply to#1664393
On Mon, Jun 12, 2017 at 09:29:16PM -0700, David Carrillo-Cisneros wrote:
> v4: - Limit write_* functions to page_size.
>     - Fixed bugs spotted by Jiri.
>     - Add information about pipe-mode to some error messages. 

hum, looks like we have pipe report broken.. on arnaldo's perf/core branch:

[jolsa@krava perf]$ ./perf record ls | ./perf report 
# To display the perf.data header info, please use --header/--header-only options.
#

...

[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.000 MB (null) ]
#
# Total Lost Samples: 0
#
# Samples: 10  of event 'cycles:uhH'
# Event count (approx.): 1810081
#
# Overhead  Trace output
# ........  ............
#
   100.00%  N/A


thanks,
jirka

[toc] | [prev] | [next] | [standalone]

#1665485

FromDavid Carrillo-Cisneros <davidcc@google.com>
Date2017-06-14 08:50 +0200
Message-ID<tSahQ-1IN-5@gated-at.bofh.it>
In reply to#1665106
On Tue, Jun 13, 2017 at 11:54 AM, Jiri Olsa <jolsa@redhat.com> wrote:
> On Mon, Jun 12, 2017 at 09:29:16PM -0700, David Carrillo-Cisneros wrote:
>> v4: - Limit write_* functions to page_size.
>>     - Fixed bugs spotted by Jiri.
>>     - Add information about pipe-mode to some error messages.
>
> hum, looks like we have pipe report broken.. on arnaldo's perf/core branch:
>
> [jolsa@krava perf]$ ./perf record ls | ./perf report
> # To display the perf.data header info, please use --header/--header-only options.
> #
>
> ...
>
> [ perf record: Woken up 1 times to write data ]
> [ perf record: Captured and wrote 0.000 MB (null) ]
> #
> # Total Lost Samples: 0
> #
> # Samples: 10  of event 'cycles:uhH'
> # Event count (approx.): 1810081
> #
> # Overhead  Trace output
> # ........  ............
> #
>    100.00%  N/A

I saw this bug before, that time I bisected it to commit d49dadea7862
("perf tools: Make 'trace' or 'trace_fields' sort key default for
tracepoint events") from Dec 2015 but didn't have time to look further
into it ...

David

[toc] | [prev] | [next] | [standalone]

#1665116 — Re: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode

FromJiri Olsa <jolsa@redhat.com>
Date2017-06-13 21:00 +0200
SubjectRe: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode
Message-ID<tRZcK-380-23@gated-at.bofh.it>
In reply to#1664393
On Mon, Jun 12, 2017 at 09:29:16PM -0700, David Carrillo-Cisneros wrote:
> v4: - Limit write_* functions to page_size.
>     - Fixed bugs spotted by Jiri.
>     - Add information about pipe-mode to some error messages. 

looks like this version breaks the normal report

[jolsa@krava perf]$ ./perf report --header-only -I
[jolsa@krava perf]$ 

jirka

[toc] | [prev] | [next] | [standalone]

#1665487

FromDavid Carrillo-Cisneros <davidcc@google.com>
Date2017-06-14 09:00 +0200
Message-ID<tSarw-1Mo-1@gated-at.bofh.it>
In reply to#1665116
On Tue, Jun 13, 2017 at 11:55 AM, Jiri Olsa <jolsa@redhat.com> wrote:
> On Mon, Jun 12, 2017 at 09:29:16PM -0700, David Carrillo-Cisneros wrote:
>> v4: - Limit write_* functions to page_size.
>>     - Fixed bugs spotted by Jiri.
>>     - Add information about pipe-mode to some error messages.
>
> looks like this version breaks the normal report
>
> [jolsa@krava perf]$ ./perf report --header-only -I
> [jolsa@krava perf]$

Is the command simply terminating without any output? I cannot
reproduce it. In my machines it works:

$ ./perf report --header-only -I
# ========
# captured on: Tue Jun 13 23:49:13 2017
# hostname : lphh20
# os release : 4.3.5-smp-801.43.0.0
# perf version : 4.12.rc4.gf2978e
# arch : x86_64
# nrcpus online : 72
# nrcpus avail : 72
# cpudesc : Intel(R) Xeon(R) CPU E5-2696 v3 @ 2.30GHz
# cpuid : GenuineIntel,6,63,2
# total memory : 264134144 kB
# cmdline : /root/perf record ls
# event : name = cycles:pp, , size = 112, { sample_period, sample_freq
} = 4000, sample_type = IP|TID|TIME|PERIOD, disa
# sibling cores   : 0-17,36-53
# sibling cores   : 18-35,54-71
[SNIP]

Is your perf.data from the same version?

>
> jirka

[toc] | [prev] | [next] | [standalone]

#1665505 — Re: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode

FromJiri Olsa <jolsa@redhat.com>
Date2017-06-14 09:30 +0200
SubjectRe: [PATCH v4 00/16] perf tool: add meta-data header support for pipe-mode
Message-ID<tSaUy-2dU-13@gated-at.bofh.it>
In reply to#1665487
On Tue, Jun 13, 2017 at 11:51:43PM -0700, David Carrillo-Cisneros wrote:
> On Tue, Jun 13, 2017 at 11:55 AM, Jiri Olsa <jolsa@redhat.com> wrote:
> > On Mon, Jun 12, 2017 at 09:29:16PM -0700, David Carrillo-Cisneros wrote:
> >> v4: - Limit write_* functions to page_size.
> >>     - Fixed bugs spotted by Jiri.
> >>     - Add information about pipe-mode to some error messages.
> >
> > looks like this version breaks the normal report
> >
> > [jolsa@krava perf]$ ./perf report --header-only -I
> > [jolsa@krava perf]$
> 
> Is the command simply terminating without any output? I cannot
> reproduce it. In my machines it works:
> 
> $ ./perf report --header-only -I
> # ========
> # captured on: Tue Jun 13 23:49:13 2017
> # hostname : lphh20
> # os release : 4.3.5-smp-801.43.0.0
> # perf version : 4.12.rc4.gf2978e
> # arch : x86_64
> # nrcpus online : 72
> # nrcpus avail : 72
> # cpudesc : Intel(R) Xeon(R) CPU E5-2696 v3 @ 2.30GHz
> # cpuid : GenuineIntel,6,63,2
> # total memory : 264134144 kB
> # cmdline : /root/perf record ls
> # event : name = cycles:pp, , size = 112, { sample_period, sample_freq
> } = 4000, sample_type = IP|TID|TIME|PERIOD, disa
> # sibling cores   : 0-17,36-53
> # sibling cores   : 18-35,54-71
> [SNIP]
> 
> Is your perf.data from the same version?

yes:

[jolsa@krava perf]$ ./perf record kill
kill: not enough arguments
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.016 MB perf.data (7 samples) ]
[jolsa@krava perf]$ ./perf report --header-only -I
[jolsa@krava perf]$ 

but I'm actualy using branch I found in your tree: ccdavid/toup_june12_pipeheaders_00,
maybe it's not the latest

[jolsa@krava perf]$ git log --oneline | head -1
a16f0acca1d2 perf header: add event desc to pipe-mode header


jirka

[toc] | [prev] | [next] | [standalone]

#1666141

FromDavid Carrillo-Cisneros <davidcc@google.com>
Date2017-06-14 20:40 +0200
Message-ID<tSlmV-fF-1@gated-at.bofh.it>
In reply to#1665505
On Wed, Jun 14, 2017 at 12:20 AM, Jiri Olsa <jolsa@redhat.com> wrote:
> On Tue, Jun 13, 2017 at 11:51:43PM -0700, David Carrillo-Cisneros wrote:
>> On Tue, Jun 13, 2017 at 11:55 AM, Jiri Olsa <jolsa@redhat.com> wrote:
>> > On Mon, Jun 12, 2017 at 09:29:16PM -0700, David Carrillo-Cisneros wrote:
>> >> v4: - Limit write_* functions to page_size.
>> >>     - Fixed bugs spotted by Jiri.
>> >>     - Add information about pipe-mode to some error messages.
>> >
>> > looks like this version breaks the normal report
>> >
>> > [jolsa@krava perf]$ ./perf report --header-only -I
>> > [jolsa@krava perf]$
>>
>> Is the command simply terminating without any output? I cannot
>> reproduce it. In my machines it works:
>>
>> $ ./perf report --header-only -I
>> # ========
>> # captured on: Tue Jun 13 23:49:13 2017
>> # hostname : lphh20
>> # os release : 4.3.5-smp-801.43.0.0
>> # perf version : 4.12.rc4.gf2978e
>> # arch : x86_64
>> # nrcpus online : 72
>> # nrcpus avail : 72
>> # cpudesc : Intel(R) Xeon(R) CPU E5-2696 v3 @ 2.30GHz
>> # cpuid : GenuineIntel,6,63,2
>> # total memory : 264134144 kB
>> # cmdline : /root/perf record ls
>> # event : name = cycles:pp, , size = 112, { sample_period, sample_freq
>> } = 4000, sample_type = IP|TID|TIME|PERIOD, disa
>> # sibling cores   : 0-17,36-53
>> # sibling cores   : 18-35,54-71
>> [SNIP]
>>
>> Is your perf.data from the same version?
>
> yes:
>
> [jolsa@krava perf]$ ./perf record kill
> kill: not enough arguments
> [ perf record: Woken up 1 times to write data ]
> [ perf record: Captured and wrote 0.016 MB perf.data (7 samples) ]
> [jolsa@krava perf]$ ./perf report --header-only -I
> [jolsa@krava perf]$
>
> but I'm actualy using branch I found in your tree: ccdavid/toup_june12_pipeheaders_00,
> maybe it's not the latest
>
> [jolsa@krava perf]$ git log --oneline | head -1
> a16f0acca1d2 perf header: add event desc to pipe-mode header

Ok, found the problem. Output is hidden by the ui in certain perf
configurations. Will fix for next version. Thanks for looking into it.

David

[toc] | [prev] | [standalone]

Page 2 of 2 — ← Prev page 1 [2]

Back to top | Article view | linux.kernel

csiph-web