Re: [GIT PULL 00/16] perf/core improvements and fixes

[Namhyung Kim <namhyung.kim@lge.com>]

2012-02-14 12:07 PM, Namhyung Kim wrote:
> 2012-02-14 오전 11:50, Namhyung Kim 쓴 글:
>> Hello,
>>
>> 2012-02-14 10:52 AM, Arnaldo Carvalho de Melo wrote:
>>> The following changes since commit c98fdeaa92731308ed80386261fa2589addefa47:
>>>
>>> x86/sched/perf/AMD: Set sched_clock_stable (2012-02-07 13:12:08 +0100)
>>>
>>> are available in the git repository at:
>>>
>>> git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux perf-core-for-mingo
>>>
>>> for you to fetch changes up to 7e1ccd3804281fc0755eb726b654469c40a96d89:
>>>
>>> perf tools: cleanup initialization of attr->size (2012-02-13 23:35:04 -0200)
>>>
>>> ----------------------------------------------------------------
>>> perf/core fixes and improvements.
>>>
>>> Signed-off-by: Arnaldo Carvalho de Melo<acme@redhat.com>
>>>
>>
>> I've got a following error during a simple test:
>>
>> $ ./perf record -- sleep 1
>>
>> Error: sys_perf_event_open() syscall returned with 22 (Invalid argument).
>> /bin/dmesg may provide additional information.
>>
>> Fatal: No CONFIG_PERF_EVENTS=y kernel support configured?
>>
>> sleep: Terminated
>>
>>
>> Any thoughts?
>>
>> Thanks,
>> Namhyung
>>
>
> git bisect told me:
>
> c4a7dca92bbb9881a5d678720f1d0c2153499749 is the first bad commit
> commit c4a7dca92bbb9881a5d678720f1d0c2153499749
> Author: Joerg Roedel<joerg.roedel@amd.com>
> Date:   Fri Feb 10 18:05:05 2012 +0100
>
>      perf tools: Change perf_guest default back to false
>
>      Setting perf_guest to true by default makes no sense because the perf
>      subcommands can not setup guest symbol information and thus not process
>      and guest samples. The only exception is perf-kvm which changes the
>      perf_guest value on its own.  So change the default for perf_guest back
>      to false.
>
>      Cc: David Ahern<dsahern@gmail.com>
>      Cc: Ingo Molnar<mingo@elte.hu>
>      Cc: Jason Wang<jasowang@redhat.com>
>      Cc: Paul Mackerras<paulus@samba.org>
>      Cc: Peter Zijlstra<a.p.zijlstra@chello.nl>
>      Link: http://lkml.kernel.org/r/1328893505-4115-3-git-send-email-joerg.roedel@amd.com
>      Signed-off-by: Joerg Roedel<joerg.roedel@amd.com>
>      Signed-off-by: Arnaldo Carvalho de Melo<acme@redhat.com>
>
> :040000 040000 64d65944c9081dfd9ace65dbc19fca655cbdc708 4f0e86ab85314ba68f2295ae1e599f3bb5ab9484 M      tools
>
>
> Thanks,
> Namhyung
>

It happened because I used old kernel version that isn't aware of the 
exclude_guest bit. So the check on attr->__reserved_1 in 
core.c::perf_copy_attr() failed and then ended up seeing EINVAL at 
perf_evsel__open().

So I think it should be handled in the same way of handling sample_id_all bit 
in perf record and top if we want to set perf_guest to false by default. I'll 
send patches for them soon.

Thanks,
Namhyung