nbd: clear disconnected on reconnect
authorJosef Bacik <jbacik@fb.com>
Tue, 25 Jul 2017 17:31:19 +0000 (13:31 -0400)
committerJens Axboe <axboe@kernel.dk>
Tue, 25 Jul 2017 19:58:34 +0000 (13:58 -0600)
If our device loses its connection for longer than the dead timeout we
will set NBD_DISCONNECTED in order to quickly fail any pending IO's that
flood in after the IO's that were waiting during the dead timer.
However if we re-connect at some point in the future we'll still see
this DISCONNECTED flag set if we then lose our connection again after
that, which means we won't get notifications for our newly lost
connections.  Fix this by just clearing the DISCONNECTED flag on
reconnect in order to make sure everything works as it's supposed to.

Reported-by: Dan Melnic <dmm@fb.com>
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
drivers/block/nbd.c

index 64b19b10b73975a6eb7b6799b39962da4f8c2aa5..5bdf923294a5d610429581c83e558529f7d40b62 100644 (file)
@@ -923,6 +923,8 @@ static int nbd_reconnect_socket(struct nbd_device *nbd, unsigned long arg)
                mutex_unlock(&nsock->tx_lock);
                sockfd_put(old);
 
+               clear_bit(NBD_DISCONNECTED, &config->runtime_flags);
+
                /* We take the tx_mutex in an error path in the recv_work, so we
                 * need to queue_work outside of the tx_mutex.
                 */