Re: [PATCH] perf: do not flush maps on COMM for perf report

[David Ahern]

On 8/22/12 12:16 PM, Arnaldo Carvalho de Melo wrote:
> Em Wed, Aug 22, 2012 at 10:56:13AM -0600, David Ahern escreveu:
> > On 8/22/12 10:29 AM, Arnaldo Carvalho de Melo wrote:
> > > Adding the EXEC event, ditto. And I agree that while adding it we want
> > > to do 1/2 as pre-requisite.
> >
> > maps should not be flushed on a COMM event, so that was a mistake.
> > Given that what new information does an EXEC event provide? Same
> > process id. A COMM event is generated on an exec, so the name change
> > happens. Mappings are dropped prior to that - and there is no unmap
> > event. That seems to be a missing piece. Maps are added which is
> > handled by MMAP events. After that why is an exec event relevant?
>
> Please read the original discussion about it:
>
> https://lkml.org/lkml/2012/2/13/545

I do recall that discussion (and re-read just now). If maps are not 
flushed on a rename (and they should not be), then I still do not 
understand what information is conveyed by an exec event? It's not a new 
process or new map and the name change is already handled by a comm event.

David

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/