nbd: fix shutdown and recv work deadlock v2

This fixes a regression added with:

commit e9e006f
Author: Mike Christie <mchristi@redhat.com>
Date:   Sun Aug 4 14:10:06 2019 -0500

    nbd: fix max number of supported devs

where we can deadlock during device shutdown. The problem occurs if
the recv_work's nbd_config_put occurs after nbd_start_device_ioctl has
returned and the userspace app has droppped its reference via closing
the device and running nbd_release. The recv_work nbd_config_put call
would then drop the refcount to zero and try to destroy the config which
would try to do destroy_workqueue from the recv work.

This patch just has nbd_start_device_ioctl do a flush_workqueue when it
wakes so we know after the ioctl returns running works have exited. This
also fixes a possible race where we could try to reuse the device while
old recv_works are still running.

Cc: stable@vger.kernel.org
Fixes: e9e006f ("nbd: fix max number of supported devs")
Signed-off-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

drivers/block/nbd.c

@@ -1296,10 +1296,10 @@ static int nbd_start_device_ioctl(struct nbd_device *nbd, struct block_device *b
 	mutex_unlock(&nbd->config_lock);
 	ret = wait_event_interruptible(config->recv_wq,
 					 atomic_read(&config->recv_threads) == 0);
-	if (ret) {
+	if (ret)
 		sock_shutdown(nbd);
-		flush_workqueue(nbd->recv_workq);
-	}
+	flush_workqueue(nbd->recv_workq);
+
 	mutex_lock(&nbd->config_lock);
 	nbd_bdev_reset(bdev);
 	/* user requested, ignore socket errors */