aio: do not return ERESTARTSYS as a result of AIO
authorJan Kara <jack@suse.cz>
Wed, 22 Sep 2010 20:05:03 +0000 (13:05 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 23 Sep 2010 00:22:39 +0000 (17:22 -0700)
OCFS2 can return ERESTARTSYS from its write function when the process is
signalled while waiting for a cluster lock (and the filesystem is mounted
with intr mount option).  Generally, it seems reasonable to allow
filesystems to return this error code from its IO functions.  As we must
not leak ERESTARTSYS (and similar error codes) to userspace as a result of
an AIO operation, we have to properly convert it to EINTR inside AIO code
(restarting the syscall isn't really an option because other AIO could
have been already submitted by the same io_submit syscall).

Signed-off-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/aio.c

index 1320b2a..250b0a7 100644 (file)
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -712,8 +712,16 @@ static ssize_t aio_run_iocb(struct kiocb *iocb)
         */
        ret = retry(iocb);
 
-       if (ret != -EIOCBRETRY && ret != -EIOCBQUEUED)
+       if (ret != -EIOCBRETRY && ret != -EIOCBQUEUED) {
+               /*
+                * There's no easy way to restart the syscall since other AIO's
+                * may be already running. Just fail this IO with EINTR.
+                */
+               if (unlikely(ret == -ERESTARTSYS || ret == -ERESTARTNOINTR ||
+                            ret == -ERESTARTNOHAND || ret == -ERESTART_RESTARTBLOCK))
+                       ret = -EINTR;
                aio_complete(iocb, ret, 0);
+       }
 out:
        spin_lock_irq(&ctx->ctx_lock);