io_uring: fix wrong arm_poll error handling

[ upstream commmit 9d2ad29 ]

Leaving ip.error set when a request was punted to task_work execution is
problematic, don't forget to clear it.

Fixes: aa43477 ("io_uring: poll rework")
Signed-off-by: Pavel Begunkov <[email protected]>
Link: https://lore.kernel.org/r/a6c84ef4182c6962380aebe11b35bdcb25b0ccfb.1655852245.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <[email protected]>
[pavel: backport]
Signed-off-by: Pavel Begunkov <[email protected]>
Signed-off-by: Greg Kroah-Hartman <[email protected]>

Author: isilence

fs/io_uring.c

@@ -5627,8 +5627,10 @@ static int __io_arm_poll_handler(struct io_kiocb *req,
 
 	if (mask) {
 		/* can't multishot if failed, just queue the event we've got */
-		if (unlikely(ipt->error || !ipt->nr_entries))
+		if (unlikely(ipt->error || !ipt->nr_entries)) {
 			poll->events |= EPOLLONESHOT;
+			ipt->error = 0;
+		}
 		__io_poll_execute(req, mask);
 		return 0;
 	}