On Wed, Mar 08, 2023 at 10:53:51PM -0300, Arnaldo Carvalho de Melo wrote:
> Em Mon, Feb 13, 2023 at 10:09:21PM -0800, Alexei Starovoitov escreveu:
> > On Mon, Feb 13, 2023 at 7:12 PM Alexei Starovoitov <alexei.starovoitov@xxxxxxxxx> wrote:
> > > On Thu, Feb 9, 2023 at 5:29 AM Alan Maguire <alan.maguire@xxxxxxxxxx> wrote:
> > > > +++ b/scripts/pahole-flags.sh
> > > > @@ -23,5 +23,8 @@ if [ "${pahole_ver}" -ge "124" ]; then
> > > > # see PAHOLE_HAS_LANG_EXCLUDE
> > > > extra_paholeopt="${extra_paholeopt} --lang_exclude=rust"
> > > > fi
> > > > +if [ "${pahole_ver}" -ge "125" ]; then
> > > > + extra_paholeopt="${extra_paholeopt} --skip_encoding_btf_inconsistent_proto --btf_gen_optimized"
> > > > +fi
> > >
> > > We landed this too soon.
> > > #229 tracing_struct:FAIL
> > > is failing now.
> > > since bpf_testmod.ko is missing a bunch of functions though they're global.
> > >
> > > I've tried a bunch of different flags and attributes, but none of them
> > > helped.
> > > The only thing that works is:
> > > diff --git a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> > > b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> > > index 46500636d8cd..5fd0f75d5d20 100644
> > > --- a/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> > > +++ b/tools/testing/selftests/bpf/bpf_testmod/bpf_testmod.c
> > > @@ -28,6 +28,7 @@ struct bpf_testmod_struct_arg_2 {
> > > long b;
> > > };
> > >
> > > +__attribute__((optimize("-O0")))
> > > noinline int
> > > bpf_testmod_test_struct_arg_1(struct bpf_testmod_struct_arg_2 a, int
> > > b, int c) {
> > >
> > > We cannot do:
> > > --- a/tools/testing/selftests/bpf/bpf_testmod/Makefile
> > > +++ b/tools/testing/selftests/bpf/bpf_testmod/Makefile
> > > @@ -10,7 +10,7 @@ endif
> > > MODULES = bpf_testmod.ko
> > >
> > > obj-m += bpf_testmod.o
> > > -CFLAGS_bpf_testmod.o = -I$(src)
> > > +CFLAGS_bpf_testmod.o = -I$(src) -O0
> > >
> > > The build fails due to asm stuff.
> > >
> > > Maybe we should make scripts/pahole-flags.sh selective
> > > and don't apply skip_encoding_btf_inconsiste to bpf_testmod ?
> > >
> > > Thoughts?
> >
> > It's even worse with clang compiled kernel:
>
> I tested what is now in the master branch with both gcc and clang, on
> fedora:37, Alan also tested it, Jiri, it would be great if you could
> check if reverting the revert works for you as well.
ok, will check your master branch
jirka
>
> Thanks,
>
> - Arnaldo
>
> > WARN: resolve_btfids: unresolved symbol tcp_reno_cong_avoid
> > WARN: resolve_btfids: unresolved symbol dctcp_update_alpha
> > WARN: resolve_btfids: unresolved symbol cubictcp_cong_avoid
> > WARN: resolve_btfids: unresolved symbol bpf_xdp_metadata_rx_timestamp
> > WARN: resolve_btfids: unresolved symbol bpf_xdp_metadata_rx_hash
> > WARN: resolve_btfids: unresolved symbol bpf_task_kptr_get
> > WARN: resolve_btfids: unresolved symbol bpf_task_acquire_not_zero
> > WARN: resolve_btfids: unresolved symbol bpf_rdonly_cast
> > WARN: resolve_btfids: unresolved symbol
> > bpf_kfunc_call_test_static_unused_arg
> > WARN: resolve_btfids: unresolved symbol bpf_kfunc_call_test_ref
> >
> > so I reverted this commit for now.
> > Looks like pahole with skip_encoding_btf_inconsistent_proto needs
> > to be more accurate.
> > It's way too aggressive removing valid functions.
