Re: [REGRESSION][SECURITY] bpf, s390: verifier/JIT mismatch allows unprivileged kernel memory access

Min-gyu Kim <[email protected]> Fri, 24 Jul 2026 15:53:41 +0900
Newsgroups dev.linux.lists.regressions,org.kernel.vger.bpf,org.kernel.vger.linux-kernel,org.kernel.vger.linux-s390
Message-ID <CAGKGUvk6sOo4kMV0z20Vaq-eAAJn2t79Hgfhuc_sYLoC+rvqTw@mail.gmail.com>
Hi Greg, Eduard,

Thank you.

Eduard has reproduced the issue and is already working on a fix that extend=
s
compute_live_registers() to track upper 32-bit subregister liveness and rem=
oves
the dependency on reg->subreg_def.

Once the patch is posted, I can test it against my reproducer in
the s390x QEMU environment, run the relevant BPF selftests, and report the
results. I can also help prepare or test the regression test if that would =
be
useful.

I can additionally test older kernels to help determine whether the issue i=
s
triggerable before v6.18, as Eduard suggested.

Best regards,
Min-gyu Kim

2026=EB=85=84 7=EC=9B=94 24=EC=9D=BC (=EA=B8=88) =EC=98=A4=ED=9B=84 3:37, G=
reg KH <[email protected]>=EB=8B=98=EC=9D=B4 =EC=9E=91=EC=84=B1:
>
> On Thu, Jul 23, 2026 at 04:55:12PM +0900, Min-gyu Kim wrote:
> > I have not yet tested a patch.
>
> As you have a reproducer, please create a proposed patch that solves the
> issue for you for us to work off of, as our security documentation asks
> for.
>
> Also, as you cc:ed public mailing lists, there's no need to get
> [email protected] involved here.
>
> thanks,
>
> greg k-h