Re: [PATCH bpf-next v2 4/9] bpf: Introduce load-acquire and store-release instructions

[Peilin Ye <yepeilin@google.com>]

Hi Ilya,

On Fri, Feb 07, 2025 at 12:28:43PM +0100, Ilya Leoshkevich wrote:
> Acked-by: Ilya Leoshkevich <iii@linux.ibm.com>

Thanks!

> s390x has a strong memory model, and the regular load and store
> instructions are atomic as long as operand addresses are aligned.

I see.

> IIUC the verifier already enforces this unless BPF_F_ANY_ALIGNMENT
> is set, in which case whoever loaded the program is responsible for the
> consequences: memory accesses that happen to be unaligned would
> not trigger an exception, but they would not be atomic either.

The verifier rejects misaligned BPF_ATOMIC instructions since commit
ca36960211eb ("bpf: allow xadd only on aligned memory"), even if
BPF_F_ANY_ALIGNMENT is set - so this patch makes the verifier reject
misaligned load-acquires and store-releases, too, to keep the behaviour
consistent:

Specifically, check_atomic_load() calls check_load_mem() (and
check_atomic_store() calls check_store_reg()) with the
@strict_alignment_once argument equals true.  See also selftests
load_acquire_misaligned() and store_release_misaligned() in PATCH 8/9.

> So I can implement the new instructions as normal loads/stores after
> this series is merged.

That would be great!

Thanks,
Peilin Ye