Re: [PATCH] perf buildid: Avoid copy of uninitialized memory

[Arnaldo Carvalho de Melo <acme@kernel.org>]

Em Thu, Jan 19, 2023 at 08:57:21AM -0800, Ian Rogers escreveu:
> On Thu, Jan 19, 2023 at 8:39 AM Arnaldo Carvalho de Melo
> <acme@kernel.org> wrote:
> >
> > Em Thu, Jan 19, 2023 at 01:29:30PM -0300, Arnaldo Carvalho de Melo escreveu:
> > > Em Thu, Jan 19, 2023 at 08:08:13AM -0800, Ian Rogers escreveu:
> > > > On Fri, Jan 13, 2023 at 10:57 AM Ian Rogers <irogers@google.com> wrote:
> > > > >
> > > > > build_id__init only copies the buildid data up to size leaving the
> > > > > rest of the data array uninitialized. Copying the full array during
> > > > > synthesis means the written event contains uninitialized memory.  This
> > > > > was detected by the Clang/LLVM memory sanitizer.
> > > > >
> > > > > Signed-off-by: Ian Rogers <irogers@google.com>
> > > > > ---
> > > > >  tools/perf/util/synthetic-events.c | 2 +-
> > > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/tools/perf/util/synthetic-events.c b/tools/perf/util/synthetic-events.c
> > > > > index 3ab6a92b1a6d..348d05e4ec03 100644
> > > > > --- a/tools/perf/util/synthetic-events.c
> > > > > +++ b/tools/perf/util/synthetic-events.c
> > > > > @@ -2219,7 +2219,7 @@ int perf_event__synthesize_build_id(struct perf_tool *tool, struct dso *pos, u16
> > > > >
> > > > >         len = pos->long_name_len + 1;
> > > > >         len = PERF_ALIGN(len, NAME_ALIGN);
> > > > > -       memcpy(&ev.build_id.build_id, pos->bid.data, sizeof(pos->bid.data));
> > > > > +       memcpy(&ev.build_id.build_id, pos->bid.data, pos->bid.size);
> > > >
> > > > Ping. Should be an uncontroversial change to fix a copy of
> > > > uninitialized memory into the perf.data file during synthesis.
> > >
> > > Indeed, applied.
> >
> > Humm, don't we better do it as:
> >
> > +       memcpy(&ev.build_id.build_id, pos->bid.data, min(pos->bid.size, sizeof(pos->bid.data));
> >
> > Lemme check what is setting that pos->bid.size....
> >
> > Things like sysfs__read_build_id() that does such checks, but perhaps we
> > should be defensive and do it in this function as well?
> 
> Defensive is good, another option would be an assert but they can be
> compiled out. Do you want me to repost?

Please.

- Arnaldo