[PATCH] dio: remove duplicate bio wait code
Now that we have a single refcount and waiting path we can reuse it in the
async 'should_wait' path. It continues to rely on the fragile link between
the conditional in dio_complete_aio() which decides to complete the AIO and
the conditional in direct_io_worker() which decides to wait and free.
By waiting before dropping the reference we stop dio_bio_end_aio() from
calling dio_complete_aio() which used to wake up the waiter after seeing the
reference count drop to 0. We hoist this wake up into dio_bio_end_aio() which
now notices when it's left a single remaining reference that is held by the
waiter.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Suparna Bhattacharya <suparna@in.ibm.com>
Acked-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/fs/direct-io.c b/fs/direct-io.c
index bc1cbf9..f11f05d 100644
--- a/fs/direct-io.c
+++ b/fs/direct-io.c
@@ -257,7 +257,6 @@
*/
static void dio_complete_aio(struct dio *dio)
{
- unsigned long flags;
int ret;
ret = dio_complete(dio, dio->iocb->ki_pos, 0);
@@ -267,14 +266,6 @@
((dio->rw == READ) && dio->result)) {
aio_complete(dio->iocb, ret, 0);
kfree(dio);
- } else {
- /*
- * Falling back to buffered
- */
- spin_lock_irqsave(&dio->bio_lock, flags);
- if (dio->waiter)
- wake_up_process(dio->waiter);
- spin_unlock_irqrestore(&dio->bio_lock, flags);
}
}
@@ -285,6 +276,8 @@
static int dio_bio_end_aio(struct bio *bio, unsigned int bytes_done, int error)
{
struct dio *dio = bio->bi_private;
+ int waiter_holds_ref = 0;
+ int remaining;
if (bio->bi_size)
return 1;
@@ -292,7 +285,12 @@
/* cleanup the bio */
dio_bio_complete(dio, bio);
- if (atomic_dec_and_test(&dio->refcount))
+ waiter_holds_ref = !!dio->waiter;
+ remaining = atomic_sub_return(1, (&dio->refcount));
+ if (remaining == 1 && waiter_holds_ref)
+ wake_up_process(dio->waiter);
+
+ if (remaining == 0)
dio_complete_aio(dio);
return 0;
@@ -1097,30 +1095,15 @@
if (ret == 0)
ret = dio->result;
+ if (should_wait)
+ dio_await_completion(dio);
+
/* this can free the dio */
if (atomic_dec_and_test(&dio->refcount))
dio_complete_aio(dio);
- if (should_wait) {
- unsigned long flags;
- /*
- * Wait for already issued I/O to drain out and
- * release its references to user-space pages
- * before returning to fallback on buffered I/O
- */
-
- spin_lock_irqsave(&dio->bio_lock, flags);
- set_current_state(TASK_UNINTERRUPTIBLE);
- while (atomic_read(&dio->refcount)) {
- spin_unlock_irqrestore(&dio->bio_lock, flags);
- io_schedule();
- spin_lock_irqsave(&dio->bio_lock, flags);
- set_current_state(TASK_UNINTERRUPTIBLE);
- }
- spin_unlock_irqrestore(&dio->bio_lock, flags);
- set_current_state(TASK_RUNNING);
+ if (should_wait)
kfree(dio);
- }
} else {
dio_await_completion(dio);