Pavel Begunkov 4aa84f2ffa io_uring: dont kill fasync under completion_lock
CPU0                    CPU1
       ----                    ----
  lock(&new->fa_lock);
                               local_irq_disable();
                               lock(&ctx->completion_lock);
                               lock(&new->fa_lock);
  <Interrupt>
    lock(&ctx->completion_lock);

 *** DEADLOCK ***

Move kill_fasync() out of io_commit_cqring() to io_cqring_ev_posted(),
so it doesn't hold completion_lock while doing it. That saves from the
reported deadlock, and it's just nice to shorten the locking time and
untangle nested locks (compl_lock -> wq_head::lock).

Reported-by: syzbot+91ca3f25bd7f795f019c@syzkaller.appspotmail.com
Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
2021-01-07 07:48:09 -07:00
..
2020-12-21 10:28:02 -08:00
2020-12-16 12:57:51 -08:00
2020-12-17 11:18:00 -08:00
2020-12-17 11:18:00 -08:00
2020-12-10 15:33:14 +01:00
2020-12-17 12:15:03 -08:00
\n
2020-12-17 10:56:27 -08:00
\n
2020-12-17 10:56:27 -08:00
2020-12-14 15:26:14 +01:00
2020-12-16 11:30:10 -08:00
2020-12-17 13:22:17 -08:00
\n
2020-12-17 11:00:37 -08:00
2020-09-10 14:03:31 -07:00
2020-12-18 12:50:18 -08:00
2020-12-22 08:43:06 -07:00
2020-12-16 12:57:51 -08:00
2020-12-15 22:46:19 -08:00
\n
2020-10-15 15:03:10 -07:00
2020-12-19 11:18:38 -08:00
2020-08-04 21:02:38 -04:00
2020-12-16 12:57:51 -08:00
2020-07-31 08:16:01 +02:00
2020-08-07 11:33:24 -07:00
2020-12-17 11:18:00 -08:00
2020-10-23 11:33:41 -07:00
2020-08-19 14:14:19 +02:00
2020-12-01 14:53:39 -07:00
2020-10-23 11:33:41 -07:00
2020-07-31 08:16:00 +02:00
2020-10-24 12:40:18 -07:00
2020-09-26 22:55:05 -04:00
2020-12-01 14:53:39 -07:00
2020-12-01 14:53:39 -07:00
2020-06-09 15:40:50 -07:00
2020-07-31 08:16:01 +02:00