Re: perf_guest default bug

[Arnaldo Carvalho de Melo <acme@ghostprotocols.net>]

Em Sat, Mar 03, 2012 at 11:04:05AM -0300, Arnaldo Carvalho de Melo escreveu:
> Em Fri, Mar 02, 2012 at 12:00:01PM -0700, David Ahern escreveu:
> > Similar to my last email, this patch is not in urgent trees or
> > Linus' yet the problem exists in Linus' tree:
> > https://lkml.org/lkml/2012/2/17/82
> > 
> > $ /tmp/pbuild/perf record -a  -- sleep 1
> > [ perf record: Woken up 1 times to write data ]
> > [ perf record: Captured and wrote 0.319 MB perf.data (~13935 samples) ]
> > Segmentation fault
> > 
> > Root cause is the perf_guest setting.
> > 
> > $ /tmp/pbuild/perf --version
> > perf version 3.3.rc5.60.g203738
> 
> This one was tricky, IIRC, weren't follow up patches needed to plug some
> extra problems?

Yeah, just trying cherry picking it and trying it on an RHEL6.2 kernel
and it didn't work because older kernels don't have
attr.exclude_{guest,host}, so I had to also cherrypick:

[acme@sandy linux]$ git cherry-pick
0c9781280fb672ca09c997df3f14ba506bbdb977
Finished one cherry-pick.
[perf/urgent b5f81a4] perf tools: Handle kernels that don't support attr.exclude_{guest,host}

- Arnaldo