On Mon, 2022-02-21 at 19:03 +0100, Jakub Sitnicki wrote:
> Shifting 16-bit type by 16 bits is implementation-defined for BPF
> programs.
> Don't rely on it in case it is causing the test failures we are
> seeing on
> s390x z15 target.
>
> Fixes: 2ed0dc5937d3 ("selftests/bpf: Cover 4-byte load from
> remote_port in bpf_sk_lookup")
> Reported-by: Andrii Nakryiko <andrii.nakryiko@xxxxxxxxx>
> Signed-off-by: Jakub Sitnicki <jakub@xxxxxxxxxxxxxx>
> ---
>
> I don't have a dev env for s390x/z15 set up yet, so can't definitely
> confirm the fix.
> That said, it seems worth fixing either way.
>
> tools/testing/selftests/bpf/progs/test_sk_lookup.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> b/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> index bf5b7caefdd0..7d47276a8964 100644
> --- a/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> +++ b/tools/testing/selftests/bpf/progs/test_sk_lookup.c
> @@ -65,6 +65,7 @@ static const __u32 KEY_SERVER_A = SERVER_A;
> static const __u32 KEY_SERVER_B = SERVER_B;
>
> static const __u16 SRC_PORT = bpf_htons(8008);
> +static const __u32 SRC_PORT_U32 = bpf_htonl(8008U << 16);
> static const __u32 SRC_IP4 = IP4(127, 0, 0, 2);
> static const __u32 SRC_IP6[] = IP6(0xfd000000, 0x0, 0x0,
> 0x00000002);
>
> @@ -421,7 +422,7 @@ int ctx_narrow_access(struct bpf_sk_lookup *ctx)
>
> /* 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))
> + if (val_u32 != SRC_PORT_U32)
> return SK_DROP;
>
> /* Narrow loads from local_port field. Expect DST_PORT. */
Unfortunately this doesn't help with the s390 problem.
I'll try to debug this.
