io_uring: only assign io_uring_enter() SQPOLL error in actual error case

[ Upstream commit 21f9652 ]

If an SQPOLL based ring is newly created and an application issues an
io_uring_enter(2) system call on it, then we can return a spurious
-EOWNERDEAD error. This happens because there's nothing to submit, and
if the caller doesn't specify any other action, the initial error
assignment of -EOWNERDEAD never gets overwritten. This causes us to
return it directly, even if it isn't valid.

Move the error assignment into the actual failure case instead.

Cc: stable@vger.kernel.org
Fixes: d9d0521 ("io_uring: stop SQPOLL submit on creator's death")
Reported-by: Sherlock Holo sherlockya@gmail.com
Link: axboe/liburing#413
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

fs/io_uring.c

@@ -9362,9 +9362,10 @@ SYSCALL_DEFINE6(io_uring_enter, unsigned int, fd, u32, to_submit,
 	if (ctx->flags & IORING_SETUP_SQPOLL) {
 		io_cqring_overflow_flush(ctx, false);
 
-		ret = -EOWNERDEAD;
-		if (unlikely(ctx->sq_data->thread == NULL))
+		if (unlikely(ctx->sq_data->thread == NULL)) {
+			ret = -EOWNERDEAD;
 			goto out;
+		}
 		if (flags & IORING_ENTER_SQ_WAKEUP)
 			wake_up(&ctx->sq_data->wait);
 		if (flags & IORING_ENTER_SQ_WAIT) {