Re: [PATCH v2 10/10] perf symbol: Fix ENOENT case for filename__read_build_id

[Ian Rogers]

On Fri, Dec 5, 2025 at 1:18 PM Namhyung Kim <namhyung@xxxxxxxxxx> wrote:
>
> Hi Ian,
>
> On Fri, Dec 05, 2025 at 11:40:12AM -0800, Ian Rogers wrote:
> > On Mon, Dec 1, 2025 at 12:55 PM Ian Rogers <irogers@xxxxxxxxxx> wrote:
> > >
> > > Some callers of filename__read_build_id assume the error value must be
> > > -1, fix by making them handle all < 0 values.
> > >
> > > If is_regular_file fails in filename__read_build_id then it could be
> > > the file is missing (ENOENT) and it would be wrong to return
> > > -EWOULDBLOCK in that case. Fix the logic so -EWOULDBLOCK is only
> > > reported if other errors with stat haven't occurred.
> > >
> > > Fixes: 834ebb5678d7 ("perf tools: Don't read build-ids from non-regular files")
> >
> > We might want to prioritize this fix.
>
> Sure.
>
> >
> > > Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>
> > > ---
> > >  tools/perf/builtin-buildid-cache.c | 6 ++++--
> > >  tools/perf/util/symbol.c           | 3 ++-
> > >  2 files changed, 6 insertions(+), 3 deletions(-)
> > >
> > > diff --git a/tools/perf/builtin-buildid-cache.c b/tools/perf/builtin-buildid-cache.c
> > > index c98104481c8a..539e779e3268 100644
> > > --- a/tools/perf/builtin-buildid-cache.c
> > > +++ b/tools/perf/builtin-buildid-cache.c
> > > @@ -276,12 +276,14 @@ static bool dso__missing_buildid_cache(struct dso *dso, int parm __maybe_unused)
> > >  {
> > >         char filename[PATH_MAX];
> > >         struct build_id bid = { .size = 0, };
> > > +       int err;
> > >
> > >         if (!dso__build_id_filename(dso, filename, sizeof(filename), false))
> > >                 return true;
> > >
> > > -       if (filename__read_build_id(filename, &bid) == -1) {
> >
> > This check here is clearly wrong when -EWOULDBLOCK is returned from
> > James' change.
> >
> > > -               if (errno == ENOENT)
> > > +       err = filename__read_build_id(filename, &bid);
> > > +       if (err < 0) {
> > > +               if (err == -ENOENT)
> > >                         return false;
> > >
> > >                 pr_warning("Problems with %s file, consider removing it from the cache\n",
> > > diff --git a/tools/perf/util/symbol.c b/tools/perf/util/symbol.c
> > > index 76dc5b70350a..f43e30019e21 100644
> > > --- a/tools/perf/util/symbol.c
> > > +++ b/tools/perf/util/symbol.c
> > > @@ -2008,8 +2008,9 @@ int filename__read_build_id(const char *filename, struct build_id *bid)
> > >         if (!filename)
> > >                 return -EFAULT;
> > >
> > > +       errno = 0;
> > >         if (!is_regular_file(filename))
> > > -               return -EWOULDBLOCK;
> > > +               return errno == 0 ? -EWOULDBLOCK : -errno;
> >
> > I've made the fix after the other changes as it is simpler to fix in
> > one filename__read_build_id rather than all the libbfd, .. variants.
> > If we don't want the series in the short-term perhaps we still want to
> > carry some parts of this fix.
>
> Yeah, I hesitate to merge a big series at this moment unless it's a
> critical and obvious fix.  Can you please separate this change and send
> it out?

Sent in:
https://lore.kernel.org/lkml/20251207022345.909535-1-irogers@xxxxxxxxxx/

Thanks,
Ian


> Thanks,
> Namhyung
>
>
> >
> > >
> > >         err = kmod_path__parse(&m, filename);
> > >         if (err)
> > > --
> > > 2.52.0.158.g65b55ccf14-goog
> > >