Re: [PATCH net-next] net: do not write to msg_get_inq in caller

[Willem de Bruijn <willemdebruijn.kernel@gmail.com>]

Jens Axboe wrote:
> On 1/5/26 10:42 AM, Willem de Bruijn wrote:
> > Eric Dumazet wrote:
> >> On Mon, Jan 5, 2026 at 5:33?PM Willem de Bruijn
> >> <willemdebruijn.kernel@gmail.com> wrote:
> >>>
> >>> From: Willem de Bruijn <willemb@google.com>
> >>>
> >>> msg_get_inq is an input field from caller to callee. Don't set it in
> >>> the callee, as the caller may not clear it on struct reuse.
> >>>
> >>> This is a kernel-internal variant of msghdr only, and the only user
> >>> does reinitialize the field. So this is not critical.
> >>>
> >>> But it is more robust to avoid the write, and slightly simpler code.
> >>>
> >>> Callers set msg_get_inq to request the input queue length to be
> >>> returned in msg_inq. This is equivalent to but independent from the
> >>> SO_INQ request to return that same info as a cmsg (tp->recvmsg_inq).
> >>> To reduce branching in the hot path the second also sets the msg_inq.
> >>> That is WAI.
> >>>
> >>> This is a small follow-on to commit 4d1442979e4a ("af_unix: don't
> >>> post cmsg for SO_INQ unless explicitly asked for"), which fixed the
> >>> inverse.
> >>>
> >>> Also collapse two branches using a bitwise or.
> >>>
> >>> Link: https://lore.kernel.org/netdev/willemdebruijn.kernel.24d8030f7a3de@gmail.com/
> >>> Signed-off-by: Willem de Bruijn <willemb@google.com>
> >>> ---
> >>
> >> Patch looks sane to me, but the title is a bit confusing, I guess you meant
> >>
> >> "net: do not write to msg_get_inq in callee" ?
> > 
> > Indeed, thanks. Will fix.
> > 
> >>
> >> Also, unix_stream_read_generic() is currently potentially adding a NULL deref
> >> if u->recvmsg_inq is non zero, but msg is NULL ?
> >>
> >> If this is the case  we need a Fixes: tag.
> > 
> > Oh good point. state->msg can be NULL as of commit 2b514574f7e8 ("net:
> > af_unix: implement splice for stream af_unix sockets"). That commit
> > mentions "we mostly have to deal with a non-existing struct msghdr
> > argument".
> 
> Worth noting that this is currently not possible, as io_uring should
> be the only one setting ->recvmsg_inq and it would not do that via
> splice. Should still be fixed of course.

recvmsg_inq is written from setsockopt SO_INQ. Do you mean
msg_get_inq?

I think this is reachable with a setsockopt + splice:

        do_cmsg = READ_ONCE(u->recvmsg_inq);
        if (do_cmsg)
                msg->msg_get_inq = 1;