pipe: wakeup wr_wait after setting max_usage

Commit c73be61 ("pipe: Add general notification queue support") a
regression was introduced that would lock up resized pipes under certain
conditions. See the reproducer in [1].

The commit resizing the pipe ring size was moved to a different
function, doing that moved the wakeup for pipe->wr_wait before actually
raising pipe->max_usage. If a pipe was full before the resize occured it
would result in the wakeup never actually triggering pipe_write.

Set @max_usage and @nr_accounted before waking writers if this isn't a
watch queue.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=212295 [1]
Link: https://lore.kernel.org/r/20231201-orchideen-modewelt-e009de4562c6@brauner
Fixes: c73be61 ("pipe: Add general notification queue support")
Reviewed-by: David Howells <dhowells@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Lukas Schauer <lukas@schauer.dev>
[Christian Brauner <brauner@kernel.org>: rewrite to account for watch queues]
Signed-off-by: Christian Brauner <brauner@kernel.org>

fs/pipe.c

@@ -1324,6 +1324,11 @@ int pipe_resize_ring(struct pipe_inode_info *pipe, unsigned int nr_slots)
 	pipe->tail = tail;
 	pipe->head = head;
 
+	if (!pipe_has_watch_queue(pipe)) {
+		pipe->max_usage = nr_slots;
+		pipe->nr_accounted = nr_slots;
+	}
+
 	spin_unlock_irq(&pipe->rd_wait.lock);
 
 	/* This might have made more room for writers */
@@ -1375,8 +1380,6 @@ static long pipe_set_size(struct pipe_inode_info *pipe, unsigned int arg)
 	if (ret < 0)
 		goto out_revert_acct;
 
-	pipe->max_usage = nr_slots;
-	pipe->nr_accounted = nr_slots;
 	return pipe->max_usage * PAGE_SIZE;
 
 out_revert_acct: