]> git.ipfire.org Git - thirdparty/kernel/stable.git/commitdiff
io_uring/net: fix sendzc double notif flush
authorPavel Begunkov <asml.silence@gmail.com>
Sat, 22 Mar 2025 11:47:27 +0000 (11:47 +0000)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 28 Mar 2025 21:04:58 +0000 (22:04 +0100)
commit 67c007d6c12da3e456c005083696c20d4498ae72 upstream.

refcount_t: underflow; use-after-free.
WARNING: CPU: 0 PID: 5823 at lib/refcount.c:28 refcount_warn_saturate+0x15a/0x1d0 lib/refcount.c:28
RIP: 0010:refcount_warn_saturate+0x15a/0x1d0 lib/refcount.c:28
Call Trace:
 <TASK>
 io_notif_flush io_uring/notif.h:40 [inline]
 io_send_zc_cleanup+0x121/0x170 io_uring/net.c:1222
 io_clean_op+0x58c/0x9a0 io_uring/io_uring.c:406
 io_free_batch_list io_uring/io_uring.c:1429 [inline]
 __io_submit_flush_completions+0xc16/0xd20 io_uring/io_uring.c:1470
 io_submit_flush_completions io_uring/io_uring.h:159 [inline]

Before the blamed commit, sendzc relied on io_req_msg_cleanup() to clear
REQ_F_NEED_CLEANUP, so after the following snippet the request will
never hit the core io_uring cleanup path.

io_notif_flush();
io_req_msg_cleanup();

The easiest fix is to null the notification. io_send_zc_cleanup() can
still be called after, but it's tolerated.

Reported-by: syzbot+cf285a028ffba71b2ef5@syzkaller.appspotmail.com
Tested-by: syzbot+cf285a028ffba71b2ef5@syzkaller.appspotmail.com
Fixes: cc34d8330e036 ("io_uring/net: don't clear REQ_F_NEED_CLEANUP unconditionally")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/e1306007458b8891c88c4f20c966a17595f766b0.1742643795.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
io_uring/net.c

index 28688b6a3a50025ecde69f0d006137e8090e8131..f4c626029ba7b47c6108f65106c6632df1928af9 100644 (file)
@@ -1439,6 +1439,7 @@ int io_send_zc(struct io_kiocb *req, unsigned int issue_flags)
         */
        if (!(issue_flags & IO_URING_F_UNLOCKED)) {
                io_notif_flush(zc->notif);
+               zc->notif = NULL;
                io_req_msg_cleanup(req, 0);
        }
        io_req_set_res(req, ret, IORING_CQE_F_MORE);
@@ -1499,6 +1500,7 @@ int io_sendmsg_zc(struct io_kiocb *req, unsigned int issue_flags)
         */
        if (!(issue_flags & IO_URING_F_UNLOCKED)) {
                io_notif_flush(sr->notif);
+               sr->notif = NULL;
                io_req_msg_cleanup(req, 0);
        }
        io_req_set_res(req, ret, IORING_CQE_F_MORE);