On Wed, Jul 1, 2020 at 9:36 AM Alexei Starovoitov
<alexei.starovoitov@xxxxxxxxx> wrote:
>
> On Wed, Jul 01, 2020 at 09:08:45AM -0700, Andrii Nakryiko wrote:
> > On Wed, Jul 1, 2020 at 8:02 AM Alexei Starovoitov
> > <alexei.starovoitov@xxxxxxxxx> wrote:
> > >
> > > On Tue, Jun 30, 2020 at 11:46 PM Andrii Nakryiko <andriin@xxxxxx> wrote:
> > > >
> > > > Fix bpftool logic of stripping away const/volatile modifiers for all global
> > > > variables during BPF skeleton generation. See patch #1 for details on when
> > > > existing logic breaks and why it's important. Support special .strip_mods=true
> > > > mode in btf_dump. Add selftests validating that everything works as expected.
> > >
> > > Why bother with the flag?
> >
> > You mean btf_dump should do this always? That's a bit too invasive a
> > change, I don't like it.
> >
> > > It looks like bugfix to me.
> >
> > It can be considered a bug fix for bpftool's skeleton generation, but
> > it depends on non-trivial changes in libbpf, which are not bug fix per
> > se, so should probably better go through bpf-next.
>
> I'm not following.
> Without tweaking opts and introducing new flag the actual fix is only
> two hunks in patch 1:
Right, but from the btf_dump point of view this is not a bug fix, its
current behavior is correct and precise. So this change is a change in
behavior and not universally correct for all the possible use cases.
So I can't just make it always strip modifiers, it's changing
generated output. It has to be an optional feature.
>
> @@ -1045,6 +1050,10 @@ static void btf_dump_emit_type_decl(struct btf_dump *d, __u32 id,
>
> stack_start = d->decl_stack_cnt;
> for (;;) {
> + t = btf__type_by_id(d->btf, id);
> + if (btf_is_mod(t))
> + goto skip_mod;
> +
> err = btf_dump_push_decl_stack_id(d, id);
> if (err < 0) {
> /*
> @@ -1056,12 +1065,11 @@ static void btf_dump_emit_type_decl(struct btf_dump *d, __u32 id,
> d->decl_stack_cnt = stack_start;
> return;
> }
> -
> +skip_mod:
> /* VOID */
> if (id == 0)
> break;
>
> - t = btf__type_by_id(d->btf, id);
>
