Re: [PATCH bpf] bpf: verifier: support BPF_LOAD_ACQ in insn_def_regno()

[Peilin Ye <yepeilin@google.com>]

On Wed, May 21, 2025 at 01:04:47PM -0700, Alexei Starovoitov wrote:
> On Wed, May 21, 2025 at 12:13 PM Eduard Zingerman <eddyz87@gmail.com> wrote:
> > I'm confused, is_atomic_load_insn() is defined as:
> >
> >           return BPF_CLASS(insn->code) == BPF_STX &&
> >                  BPF_MODE(insn->code) == BPF_ATOMIC &&
> >                  insn->imm == BPF_LOAD_ACQ;
> >
> > And insn_def_regno() has the following case:
> >
> >           case BPF_STX:
> >                   if (BPF_MODE(insn->code) == BPF_ATOMIC ||
> >                       BPF_MODE(insn->code) == BPF_PROBE_ATOMIC) {
> >                           if (insn->imm == BPF_CMPXCHG)
> >                                   return BPF_REG_0;
> >                           else if (insn->imm == BPF_LOAD_ACQ)
> >                                   return insn->dst_reg;
> >                           else if (insn->imm & BPF_FETCH)
> >                                   return insn->src_reg;
> >                   }
> >                   return -1;
> >
> > Why is it not triggering?
> >
> > Also, can this be tested with a BPF_F_TEST_RND_HI32 flag?
> > E.g. see verifier_scalar_ids.c:linked_regs_and_subreg_def() test case.
> 
> I suspect it was already fixed by commit
> fce7bd8e385a ("bpf/verifier: Handle BPF_LOAD_ACQ instructions in
> insn_def_regno()")

Ah, right; I did it when adding support for riscv64 (which needs_zext).
I targeted bpf-next because at that time only x86_64 and arm64 (both
!needs_zext) supported BPF_LOAD_ACQ, and didn't realize this could
affect arm32 (needs_zext).

It should've targeted bpf with a Fixes: tag instead.  Sorry for any
confusion.

Thanks,
Peilin Ye