Re: [PATCH bpf-next v2 2/2] selftests/bpf: Cover 4-byte load from remote_port in bpf_sk_lookup

[Ilya Leoshkevich <iii@linux.ibm.com>]

On Wed, 2022-02-16 at 13:44 -0800, Andrii Nakryiko wrote:
> On Wed, Feb 9, 2022 at 10:43 AM Jakub Sitnicki <jakub@cloudflare.com>
> wrote:
> > 
> > Extend the context access tests for sk_lookup prog to cover the
> > surprising
> > case of a 4-byte load from the remote_port field, where the
> > expected value
> > is actually shifted by 16 bits.
> > 
> > Acked-by: Yonghong Song <yhs@fb.com>
> > Signed-off-by: Jakub Sitnicki <jakub@cloudflare.com>
> > ---
> >  tools/include/uapi/linux/bpf.h                     | 3 ++-
> >  tools/testing/selftests/bpf/progs/test_sk_lookup.c | 6 ++++++
> >  2 files changed, 8 insertions(+), 1 deletion(-)
> > 
> > diff --git a/tools/include/uapi/linux/bpf.h
> > b/tools/include/uapi/linux/bpf.h
> > index a7f0ddedac1f..afe3d0d7f5f2 100644
> > --- a/tools/include/uapi/linux/bpf.h
> > +++ b/tools/include/uapi/linux/bpf.h
> > @@ -6453,7 +6453,8 @@ struct bpf_sk_lookup {
> >         __u32 protocol;         /* IP protocol (IPPROTO_TCP,
> > IPPROTO_UDP) */
> >         __u32 remote_ip4;       /* Network byte order */
> >         __u32 remote_ip6[4];    /* Network byte order */
> > -       __u32 remote_port;      /* Network byte order */
> > +       __be16 remote_port;     /* Network byte order */
> > +       __u16 :16;              /* Zero padding */
> >         __u32 local_ip4;        /* Network byte order */
> >         __u32 local_ip6[4];     /* Network byte order */
> >         __u32 local_port;       /* Host byte order */
> > diff --git a/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> > b/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> > index 83b0aaa52ef7..bf5b7caefdd0 100644
> > --- a/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> > +++ b/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> > @@ -392,6 +392,7 @@ int ctx_narrow_access(struct bpf_sk_lookup
> > *ctx)
> >  {
> >         struct bpf_sock *sk;
> >         int err, family;
> > +       __u32 val_u32;
> >         bool v4;
> > 
> >         v4 = (ctx->family == AF_INET);
> > @@ -418,6 +419,11 @@ int ctx_narrow_access(struct bpf_sk_lookup
> > *ctx)
> >         if (LSW(ctx->remote_port, 0) != SRC_PORT)
> >                 return SK_DROP;
> > 
> > +       /* Load from remote_port field with zero padding (backward
> > compatibility) */
> > +       val_u32 = *(__u32 *)&ctx->remote_port;
> > +       if (val_u32 != bpf_htonl(bpf_ntohs(SRC_PORT) << 16))
> > +               return SK_DROP;
> > +
> 
> Jakub, can you please double check that your patch set doesn't break
> big-endian architectures? I've noticed that our s390x test runner is
> now failing in the sk_lookup selftest. See [0]. Also CC'ing Ilya.

I agree that this looks like an endianness issue. The new check seems
to make little sense on big-endian to me, so I would just #ifdef it
out.

> 
>   [0]
> https://github.com/libbpf/libbpf/runs/5220996832?check_suite_focus=true
> 
> >         /* Narrow loads from local_port field. Expect DST_PORT. */
> >         if (LSB(ctx->local_port, 0) != ((DST_PORT >> 0) & 0xff) ||
> >             LSB(ctx->local_port, 1) != ((DST_PORT >> 8) & 0xff) ||
> > --
> > 2.31.1
> >