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


Groups > linux.kernel > #1470062 > unrolled thread

perf-4.8.0-rc3: Segmentation fault

Started byAlexey Dobriyan <adobriyan@gmail.com>
First post2016-08-25 13:00 +0200
Last post2016-08-30 12:10 +0200
Articles 4 — 2 participants

Back to article view | Back to linux.kernel


Contents

  perf-4.8.0-rc3: Segmentation fault Alexey Dobriyan <adobriyan@gmail.com> - 2016-08-25 13:00 +0200
    Re: perf-4.8.0-rc3: Segmentation fault Arnaldo Carvalho de Melo <acme@kernel.org> - 2016-08-29 16:50 +0200
      Re: perf-4.8.0-rc3: Segmentation fault Arnaldo Carvalho de Melo <acme@kernel.org> - 2016-08-29 22:10 +0200
        Re: perf-4.8.0-rc3: Segmentation fault Alexey Dobriyan <adobriyan@gmail.com> - 2016-08-30 12:10 +0200

#1470062 — perf-4.8.0-rc3: Segmentation fault

FromAlexey Dobriyan <adobriyan@gmail.com>
Date2016-08-25 13:00 +0200
Subjectperf-4.8.0-rc3: Segmentation fault
Message-ID<sa0y5-8ll-17@gated-at.bofh.it>
I profiled kernel allmodconfig build (don't ask) and perf tui reliably
segfaults when trying to enter annotation mode for any of the
individual functions. It segfaults immediately before even showing
menu with choices.

perf: Segmentation fault
-------- backtrace --------
perf[0x5451fb]
/lib64/libc.so.6(+0x33290)[0x7fcec0a80290]
perf(__map__is_kernel+0x26)[0x4bbcd6]
perf[0x54287c]
perf(perf_evlist__tui_browse_hists+0x94)[0x543e84]
perf(cmd_report+0x1a34)[0x434884]
perf[0x478471]
perf(main+0x6c6)[0x4237e6]
/lib64/libc.so.6(__libc_start_main+0xf0)[0x7fcec0a6d630]
perf(_start+0x29)[0x4238e9]

Steps to reproduce:
    rm -rf ../xxx
    mkdir ../xxx
    make O=../xxx ARCH=x86_64 allmodconfig
    perf record -e cycles make O=../xxx -k -j5
    perf report

perf.data is 1.23 GB long , so be prepared.

[toc] | [next] | [standalone]


#1471890

FromArnaldo Carvalho de Melo <acme@kernel.org>
Date2016-08-29 16:50 +0200
Message-ID<sbw2R-62-11@gated-at.bofh.it>
In reply to#1470062
Em Thu, Aug 25, 2016 at 01:49:41PM +0300, Alexey Dobriyan escreveu:
> I profiled kernel allmodconfig build (don't ask) and perf tui reliably
> segfaults when trying to enter annotation mode for any of the
> individual functions. It segfaults immediately before even showing
> menu with choices.
> 
> perf: Segmentation fault
> -------- backtrace --------
> perf[0x5451fb]
> /lib64/libc.so.6(+0x33290)[0x7fcec0a80290]
> perf(__map__is_kernel+0x26)[0x4bbcd6]
> perf[0x54287c]
> perf(perf_evlist__tui_browse_hists+0x94)[0x543e84]
> perf(cmd_report+0x1a34)[0x434884]
> perf[0x478471]
> perf(main+0x6c6)[0x4237e6]
> /lib64/libc.so.6(__libc_start_main+0xf0)[0x7fcec0a6d630]
> perf(_start+0x29)[0x4238e9]
> 
> Steps to reproduce:
>     rm -rf ../xxx
>     mkdir ../xxx
>     make O=../xxx ARCH=x86_64 allmodconfig
>     perf record -e cycles make O=../xxx -k -j5
>     perf report
> 
> perf.data is 1.23 GB long , so be prepared.

Thanks for the report, trying to reproduce this.

- Arnaldo

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


#1472091

FromArnaldo Carvalho de Melo <acme@kernel.org>
Date2016-08-29 22:10 +0200
Message-ID<sbB2y-3pN-15@gated-at.bofh.it>
In reply to#1471890
Em Mon, Aug 29, 2016 at 11:44:50AM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Thu, Aug 25, 2016 at 01:49:41PM +0300, Alexey Dobriyan escreveu:
> > I profiled kernel allmodconfig build (don't ask) and perf tui reliably
> > segfaults when trying to enter annotation mode for any of the
> > individual functions. It segfaults immediately before even showing
> > menu with choices.
> > 
> > perf: Segmentation fault
> > -------- backtrace --------
> > perf[0x5451fb]
> > /lib64/libc.so.6(+0x33290)[0x7fcec0a80290]
> > perf(__map__is_kernel+0x26)[0x4bbcd6]
> > perf[0x54287c]
> > perf(perf_evlist__tui_browse_hists+0x94)[0x543e84]
> > perf(cmd_report+0x1a34)[0x434884]
> > perf[0x478471]
> > perf(main+0x6c6)[0x4237e6]
> > /lib64/libc.so.6(__libc_start_main+0xf0)[0x7fcec0a6d630]
> > perf(_start+0x29)[0x4238e9]
> > 
> > Steps to reproduce:
> >     rm -rf ../xxx
> >     mkdir ../xxx
> >     make O=../xxx ARCH=x86_64 allmodconfig
> >     perf record -e cycles make O=../xxx -k -j5
> >     perf report
> > 
> > perf.data is 1.23 GB long , so be prepared.
> 
> Thanks for the report, trying to reproduce this.

So, I built the kernel with my normal .config, using your command line and got:

[acme@jouet linux]$ perf report --header-only perf.data
# ========
# captured on: Mon Aug 29 15:48:10 2016
# hostname : jouet
# os release : 4.8.0-rc3+
# perf version : 4.8.rc2.g21a116
# arch : x86_64
# nrcpus online : 4
# nrcpus avail : 4
# cpudesc : Intel(R) Core(TM) i7-5600U CPU @ 2.60GHz
# cpuid : GenuineIntel,6,61,4
# total memory : 7869032 kB
# cmdline : /home/acme/bin/perf record -e cycles make O=../build/v4.8.0-rc4 -j -j5 
# event : name = cycles:u, , size = 112, { sample_period, sample_freq } = 4000, sample_type = IP|TID|TIME|PERIOD, disabled = 1, inherit = 1, exclude_kernel = 
# HEADER_CPU_TOPOLOGY info available, use -I to display
# HEADER_NUMA_TOPOLOGY info available, use -I to display
# pmu mappings: intel_pt = 7, intel_bts = 6, uncore_arb = 12, cstate_pkg = 15, breakpoint = 5, uncore_cbox_1 = 11, power = 13, cpu = 4, software = 1, uncore_i
# HEADER_CACHE info available, use -I to display
# ========
#

But it didn't crash:

[acme@jouet linux]$ ls -lah perf.data
-rw-------. 1 acme acme 1.8G Aug 29 13:00 perf.data
[acme@jouet linux]$ 

[acme@jouet linux]$ perf --version
perf version 4.8.rc4.g3eab88

And this was with:

[acme@jouet linux]$ git log --oneline | head -5
3eab887a5542 Linux 4.8-rc4
25d0d91af770 Merge tag 'drm-fixes-for-4.8-rc4' of git://people.freedesktop.org/~airlied/linux
add1fa751012 drm/atomic: Don't potentially reset color_mgmt_changed on successive property updates.
908e373f1c81 Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
5d84ee796448 Merge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
[acme@jouet linux]$ 

I.e. 4.8-rc4 from git.

- Arnaldo

- Arnaldo

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


#1472387

FromAlexey Dobriyan <adobriyan@gmail.com>
Date2016-08-30 12:10 +0200
Message-ID<sbO9r-3ob-17@gated-at.bofh.it>
In reply to#1472091
On Mon, Aug 29, 2016 at 11:02 PM, Arnaldo Carvalho de Melo
<arnaldo.melo@gmail.com> wrote:
> Em Mon, Aug 29, 2016 at 11:44:50AM -0300, Arnaldo Carvalho de Melo escreveu:
>> Em Thu, Aug 25, 2016 at 01:49:41PM +0300, Alexey Dobriyan escreveu:
>> > I profiled kernel allmodconfig build (don't ask) and perf tui reliably
>> > segfaults when trying to enter annotation mode for any of the
>> > individual functions. It segfaults immediately before even showing
>> > menu with choices.
>> >
>> > perf: Segmentation fault
>> > -------- backtrace --------
>> > perf[0x5451fb]
>> > /lib64/libc.so.6(+0x33290)[0x7fcec0a80290]
>> > perf(__map__is_kernel+0x26)[0x4bbcd6]
>> > perf[0x54287c]
>> > perf(perf_evlist__tui_browse_hists+0x94)[0x543e84]
>> > perf(cmd_report+0x1a34)[0x434884]
>> > perf[0x478471]
>> > perf(main+0x6c6)[0x4237e6]
>> > /lib64/libc.so.6(__libc_start_main+0xf0)[0x7fcec0a6d630]
>> > perf(_start+0x29)[0x4238e9]
>> >
>> > Steps to reproduce:
>> >     rm -rf ../xxx
>> >     mkdir ../xxx
>> >     make O=../xxx ARCH=x86_64 allmodconfig
>> >     perf record -e cycles make O=../xxx -k -j5
>> >     perf report
>> >
>> > perf.data is 1.23 GB long , so be prepared.
>>
>> Thanks for the report, trying to reproduce this.
>
> So, I built the kernel with my normal .config, using your command line and got:

I've lost vmlinux and can't recreate it anymore. Oh well.

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web