Re: [PATCH bpf-next] bpf: pack struct bpf_fib_lookup

["Arnd Bergmann" <arnd@arndb.de>]

On Wed, Apr 3, 2024, at 14:33, Anton Protopopov wrote:
> The struct bpf_fib_lookup is supposed to be of size 64. A recent commit
> 59b418c7063d ("bpf: Add a check for struct bpf_fib_lookup size") added
> a static assertion to check this property so that future changes to the
> structure will not accidentally break this assumption.
>
> As it immediately turned out, on some 32-bit arm systems, when AEABI=n,
> the total size of the structure was equal to 68, see [1]. This happened
> because the bpf_fib_lookup structure contains a union of two 16-bit
> fields:
>
>     union {
>             __u16 tot_len;
>             __u16 mtu_result;
>     };

This union was introduced three years ago, so fixing it now means
another incompatible ABI change. While you clearly change it back
to what it should have been the entire time, it's not obvious that
this is the correct thing to do after three years.

I do wonder to what degree we still want to care about OABI
at all. As I mentioned, I stopped testing it myself because it
is no longer relevant to most people, and there are probably
a number of other ABI issues.

> which was supposed to compile to a 16-bit-aligned 16-bit field. On the
> aforementioned setups it was instead both aligned and padded to 32-bits.
>
> Declare this inner union as __attribute__((packed, aligned(2))) such
> that it always is of size 2 and is aligned to 16 bits.

I think you probably want 32-bit alignment for the structure,
to keep the ABI unchanged on all other architectures.

     Arnd