drbd: add missing kref_get in handle_write_conflicts

[ Upstream commit 00c9c9628b49e368d140cfa61d7df9b8922ec2a8 ]

With `two-primaries` enabled, DRBD tries to detect "concurrent" writes
and handle write conflicts, so that even if you write to the same sector
simultaneously on both nodes, they end up with the identical data once
the writes are completed.

In handling "superseeded" writes, we forgot a kref_get,
resulting in a premature drbd_destroy_device and use after free,
and further to kernel crashes with symptoms.

Relevance: No one should use DRBD as a random data generator, and apparently
all users of "two-primaries" handle concurrent writes correctly on layer up.
That is cluster file systems use some distributed lock manager,
and live migration in virtualization environments stops writes on one node
before starting writes on the other node.

Which means that other than for "test cases",
this code path is never taken in real life.

FYI, in DRBD 9, things are handled differently nowadays.  We still detect
"write conflicts", but no longer try to be smart about them.
We decided to disconnect hard instead: upper layers must not submit concurrent
writes. If they do, that's their fault.

Signed-off-by: Sarah Newman <[email protected]>
Signed-off-by: Lars Ellenberg <[email protected]>
Signed-off-by: Christoph Böhmwalder <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Signed-off-by: Jens Axboe <[email protected]>
Signed-off-by: Sasha Levin <[email protected]>
(cherry picked from commit 7d483ad300fc0a06f69b019dda8f74970714baf8)
Signed-off-by: Jack Vogel <[email protected]>

Author: Sarah Newman

drivers/block/drbd/drbd_receiver.c

@@ -2500,7 +2500,11 @@ static int handle_write_conflicts(struct drbd_device *device,
 			peer_req->w.cb = superseded ? e_send_superseded :
 						   e_send_retry_write;
 			list_add_tail(&peer_req->w.list, &device->done_ee);
-			queue_work(connection->ack_sender, &peer_req->peer_device->send_acks_work);
+			/* put is in drbd_send_acks_wf() */
+			kref_get(&device->kref);
+			if (!queue_work(connection->ack_sender,
+					&peer_req->peer_device->send_acks_work))
+				kref_put(&device->kref, drbd_destroy_device);
 
 			err = -ENOENT;
 			goto out;