Re: [RFC bpf-next 5/6] bpf: selftests: test_tunnel: Disable CO-RE relocations

[Andrii Nakryiko]

On Sun, Oct 29, 2023 at 4:22 PM Daniel Xu <dxu@xxxxxxxxx> wrote:
>
> On Fri, Oct 27, 2023 at 01:33:09PM -0700, Andrii Nakryiko wrote:
> > On Fri, Oct 27, 2023 at 11:46 AM Daniel Xu <dxu@xxxxxxxxx> wrote:
> > >
> > > Switching to vmlinux.h definitions seems to make the verifier very
> > > unhappy with bitfield accesses. The error is:
> > >
> > >     ; md.u.md2.dir = direction;
> > >     33: (69) r1 = *(u16 *)(r2 +11)
> > >     misaligned stack access off (0x0; 0x0)+-64+11 size 2
> > >
> > > It looks like disabling CO-RE relocations seem to make the error go
> > > away.
> > >
> >
> > for accessing bitfields libbpf provides
> > BPF_CORE_READ_BITFIELD_PROBED() and BPF_CORE_READ_BITFIELD() macros
>
> In this case the code in question is:
>
>         __u8 direction = 0;
>         md.u.md2.dir = direction;
>
> IOW the problem is assigning to bitfields, not reading from them.
>
> Is that something that libbpf needs to support as well?

Ah, I missed that this is a write into a struct. I think we can
support BPF_CORE_WRITE_BITFIELD() (not the PROBED version, though)
using all the same CO-RE relocations. It's probably a very niche case,
but BPF_CORE_READ_BITFIELD() is niche as well (though an absolute
necessity when the need does come up).

>
> Thanks,
> Daniel