]> git.baikalelectronics.ru Git - kernel.git/commit
io_uring/poll: fix double poll req->flags races
authorPavel Begunkov <asml.silence@gmail.com>
Fri, 11 Nov 2022 16:51:29 +0000 (16:51 +0000)
committerJens Axboe <axboe@kernel.dk>
Fri, 11 Nov 2022 16:59:27 +0000 (09:59 -0700)
commit30a33669fa21cd3dc7d92a00ba736358059014b7
tree6cd63d9678ff7392667e8d2fa3df741dc7cecebf
parent3851d25c75ed03117268a8feb34adca5a843a126
io_uring/poll: fix double poll req->flags races

io_poll_double_prepare()            | io_poll_wake()
                                    | poll->head = NULL
smp_load(&poll->head); /* NULL */   |
flags = req->flags;                 |
                                    | req->flags &= ~SINGLE_POLL;
req->flags = flags | DOUBLE_POLL    |

The idea behind io_poll_double_prepare() is to serialise with the
first poll entry by taking the wq lock. However, it's not safe to assume
that io_poll_wake() is not running when we can't grab the lock and so we
may race modifying req->flags.

Skip double poll setup if that happens. It's ok because the first poll
entry will only be removed when it's definitely completing, e.g.
pollfree or oneshot with a valid mask.

Fixes: 49f1c68e048f1 ("io_uring: optimise submission side poll_refs")
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/b7fab2d502f6121a7d7b199fe4d914a43ca9cdfd.1668184658.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
io_uring/poll.c