Re: [PATCH v2 1/2] tools/resolve_btfids: Add --fatal-warnings option

[Andrii Nakryiko]

On Tue, Dec 3, 2024 at 10:19 PM Thomas Weißschuh <linux@xxxxxxxxxxxxxx> wrote:
>
> On 2024-12-03 18:06:26-0800, Andrii Nakryiko wrote:
> > On Tue, Dec 3, 2024 at 3:09 PM Thomas Weißschuh <linux@xxxxxxxxxxxxxx> wrote:
> > >
> > > On 2024-12-03 14:31:01-0800, Andrii Nakryiko wrote:
> > > > On Tue, Nov 26, 2024 at 1:17 PM Thomas Weißschuh <linux@xxxxxxxxxxxxxx> wrote:
> > > > >
> > > > > Currently warnings emitted by resolve_btfids are buried in the build log
> > > > > and are slipping into mainline frequently.
> > > > > Add an option to elevate warnings to hard errors so the CI bots can
> > > > > catch any new warnings.
> > > > >
> > > > > Signed-off-by: Thomas Weißschuh <linux@xxxxxxxxxxxxxx>
> > > > > Acked-by: Jiri Olsa <jolsa@xxxxxxxxxx>
> > > > > ---
> > > > >  tools/bpf/resolve_btfids/main.c | 12 ++++++++++--
> > > > >  1 file changed, 10 insertions(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/tools/bpf/resolve_btfids/main.c b/tools/bpf/resolve_btfids/main.c
> > > > > index bd9f960bce3d5b74dc34159b35af1e0b33524d2d..571d29d2da97fea75e5f9c544a95b9ac65f9e579 100644
> > > > > --- a/tools/bpf/resolve_btfids/main.c
> > > > > +++ b/tools/bpf/resolve_btfids/main.c
> > > > > @@ -141,6 +141,7 @@ struct object {
> > > > >  };
> > > > >
> > > > >  static int verbose;
> > > > > +static int warnings;
> > > > >
> > > > >  static int eprintf(int level, int var, const char *fmt, ...)
> > > > >  {
> > > > > @@ -604,6 +605,7 @@ static int symbols_resolve(struct object *obj)
> > > > >                         if (id->id) {
> > > > >                                 pr_info("WARN: multiple IDs found for '%s': %d, %d - using %d\n",
> > > > >                                         str, id->id, type_id, id->id);
> > > > > +                               warnings++;
> > > > >                         } else {
> > > > >                                 id->id = type_id;
> > > > >                                 (*nr)--;
> > > > > @@ -625,8 +627,10 @@ static int id_patch(struct object *obj, struct btf_id *id)
> > > > >         int i;
> > > > >
> > > > >         /* For set, set8, id->id may be 0 */
> > > > > -       if (!id->id && !id->is_set && !id->is_set8)
> > > > > +       if (!id->id && !id->is_set && !id->is_set8) {
> > > > >                 pr_err("WARN: resolve_btfids: unresolved symbol %s\n", id->name);
> > > > > +               warnings++;
> > > > > +       }
> > > > >
> > > > >         for (i = 0; i < id->addr_cnt; i++) {
> > > > >                 unsigned long addr = id->addr[i];
> > > > > @@ -782,6 +786,7 @@ int main(int argc, const char **argv)
> > > > >                 .funcs    = RB_ROOT,
> > > > >                 .sets     = RB_ROOT,
> > > > >         };
> > > > > +       bool fatal_warnings = false;
> > > > >         struct option btfid_options[] = {
> > > > >                 OPT_INCR('v', "verbose", &verbose,
> > > > >                          "be more verbose (show errors, etc)"),
> > > > > @@ -789,6 +794,8 @@ int main(int argc, const char **argv)
> > > > >                            "BTF data"),
> > > > >                 OPT_STRING('b', "btf_base", &obj.base_btf_path, "file",
> > > > >                            "path of file providing base BTF"),
> > > > > +               OPT_BOOLEAN(0, "fatal-warnings", &fatal_warnings,
> > > > > +                           "turn warnings into errors"),
> > > >
> > > > We are mixing naming styles here: we have "btf_base" with underscore
> > > > separator, and you are adding "fatal-warnings" with dash separator. I
> > > > personally like dashes, but whichever way we should stay consistent.
> > > > So let's fix it, otherwise it looks a bit sloppy.
> > >
> > > Ack.
> > >
> > > >
> > > > Please also use [PATCH bpf-next v3] subject prefix to make it explicit
> > > > that this should go through bpf-next tree.
> > >
> > > Ack.
> > >
> > > >
> > > > pw-bot: cr
> > > >
> > > > >                 OPT_END()
> > > > >         };
> > > > >         int err = -1;
> > > > > @@ -823,7 +830,8 @@ int main(int argc, const char **argv)
> > > > >         if (symbols_patch(&obj))
> > > > >                 goto out;
> > > > >
> > > > > -       err = 0;
> > > > > +       if (!(fatal_warnings && warnings))
> > > > > +               err = 0;
> > > >
> > > > nit: just
> > > >
> > > > if (!fatal_warnings)
> > > >     err = 0;
> > > >
> > > > ?
> > >
> > > This seems wrong. Now the actual warning counter is never evaluated.
> > > And --fatal_warnings will always lead to an error exit code.
> >
> > Ah, I missed that you are using default -1 value here. I wonder if we
> > should make it a bit more explicit?
> >
> > if (fatal_warnings)
> >     err = warnings ? -1 : 0;
> > else
> >     err = 0;
> >
> > Something like that?
>
> The existing code was the same. Also the rest of the function
> relies on this. IMO the pattern is clear when looking at the resulting
> code and not the diff.
> But if you prefer I can change it of course.

That new condition breaks my brain, but luckily I don't have to look
at it often, so I don't care all that much. Feel free to leave it as
is.

>
> > > > >  out:
> > > > >         if (obj.efile.elf) {
> > > > >                 elf_end(obj.efile.elf);
> > > > >
> > > > > --
> > > > > 2.47.1
> > > > >