Re: [PATCH bpf v2] bpf: tcp: Fix use-after-free in bpf_iter_tcp_established_batch()

On Tue, Jul 21, 2026 at 09:24:05PM +0200, Kumar Kartikeya Dwivedi wrote:
> On Fri Jul 17, 2026 at 4:52 AM CEST, sashiko-bot wrote:
> > [Severity: Low]
> > Does this multi-line comment follow the BPF subsystem comment style?
> > The BPF subsystem guide expects the opening block to be on its own line:

On the comment-style note, checkpatch accepts both block comment forms
and the one used here matches the surrounding code in
net/ipv4/tcp_ipv4.c, so I don't think the bot's flag applies. Either way
is fine by me, whatever the maintainers prefer.

> > Pre-existing issues:
> > - [High] Double-put and wild pointer dereference of batch sockets on the
> > first bpf_iter_tcp_realloc_batch failure path.
>
> Hi Kuniyuki,
> Thoughts on the fix and Sashiko's concern? (The latter should be probably be
> fixed separately, if needed).

The double-put is pre-existing and already fixed separately by
980a81345275 ("bpf: tcp: fix double sock release on batch realloc"),
now in the bpf tree. The hunks are disjoint, so this series applies
cleanly on top of it.

Thanks,
Jose