Re: [PATCH 1/2] perf,top: fix events overflow in top command

[Jiri Olsa <jolsa@redhat.com>]

On Wed, Mar 09, 2011 at 11:06:10AM -0300, Arnaldo Carvalho de Melo wrote:
> Em Wed, Mar 09, 2011 at 10:59:25AM -0300, Arnaldo Carvalho de Melo escreveu:
> > Em Mon, Mar 07, 2011 at 09:13:40PM +0100, Jiri Olsa escreveu:
> > > The snprintf function returns number of printed characters even
> > > if it cross the size parameter. So passing enough events via '-e'
> > > parameter will cause segmentation fault.
> > > 
> > > It's reproduced by following command:
> > > 
> > > perf top -e `perf list | grep Tracepoint | awk -F'[' '\
> > > {gsub(/[[:space:]]+/,"",$1);array[FNR]=$1}END{outputs=array[1];\
> > > for (i=2;i<=FNR;i++){ outputs=outputs "," array[i];};print outputs}'`
> > > 
> > > Attached patch is adding SNPRINTF macro that provides the
> > > overflow check and returns actuall number of printed characters.
> > 
> > Good catch, applying to perf/urgent.
> 
> Tried but it doesn't apply nor can I reproduce the overflow (albeit
> granted the long line of events is annoying).

patches were based on tip tree

hm, it crashes for me on tip tree,
but I haven't checked the perf/core tree..

not sure if it's any help, but while debugging I could see the sigsegv happened
when the pointer was far beyond the 160 chars buffer size,
maybe the buffer got bigger in the perf tree..

jirka

> 
> So I'm applying both patches to perf/core instead.
> 
> - Arnaldo