aio: Fix incorrect usage of eventfd_signal_allowed()
authorXie Yongji <xieyongji@bytedance.com>
Mon, 13 Sep 2021 11:19:28 +0000 (19:19 +0800)
committerEric Biggers <ebiggers@google.com>
Thu, 9 Dec 2021 18:52:55 +0000 (10:52 -0800)
We should defer eventfd_signal() to the workqueue when
eventfd_signal_allowed() return false rather than return
true.

Fixes: b542e383d8c0 ("eventfd: Make signal recursion protection a task bit")
Signed-off-by: Xie Yongji <xieyongji@bytedance.com>
Link: https://lore.kernel.org/r/20210913111928.98-1-xieyongji@bytedance.com
Reviewed-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
fs/aio.c

index c9bb0d3..f6f1cbf 100644 (file)
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -1760,7 +1760,7 @@ static int aio_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
                list_del_init(&req->wait.entry);
                list_del(&iocb->ki_list);
                iocb->ki_res.res = mangle_poll(mask);
-               if (iocb->ki_eventfd && eventfd_signal_allowed()) {
+               if (iocb->ki_eventfd && !eventfd_signal_allowed()) {
                        iocb = NULL;
                        INIT_WORK(&req->work, aio_poll_put_work);
                        schedule_work(&req->work);