[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-Id: <20170324101452.19832-2-emilio@crisal.io>
Date: Fri, 24 Mar 2017 11:14:52 +0100
From: Emilio Cobos Álvarez <emilio@...sal.io>
To: linux-kernel@...r.kernel.org, Benjamin LaHaise <bcrl@...ck.org>,
linux-aio@...ck.org
Cc: Emilio Cobos Álvarez <emilio@...sal.io>
Subject: [PATCH 2/2] fs/aio.c: use READ_ONCE instead of ACCESS_ONCE for kiocb->ki_cancel
This is suggested by checkpath, but thought that wasn't as trivial as the
previous patch and deserved a bit more thought.
AIUI, this should prevent the compiler reordering the assignment in the same
way, so let's just use it.
Signed-off-by: Emilio Cobos Álvarez <emilio@...sal.io>
---
fs/aio.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/fs/aio.c b/fs/aio.c
index be0ca6d9a6b3..dce62ed78c5c 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -578,7 +578,7 @@ static int kiocb_cancel(struct aio_kiocb *kiocb)
* actually has a cancel function, hence the cmpxchg()
*/
- cancel = ACCESS_ONCE(kiocb->ki_cancel);
+ cancel = READ_ONCE(kiocb->ki_cancel);
do {
if (!cancel || cancel == KIOCB_CANCELLED)
return -EINVAL;
--
2.12.1
Powered by blists - more mailing lists