From: Jeff Moyer <jmoyer@redhat.com>
Date: Mon, 20 Aug 2007 13:17:44 -0400
Subject: [fs] aio: account for I/O wait properly
Message-id: x49odh2rupz.fsf@segfault.boston.devel.redhat.com
O-Subject: [rhel5 patch] aio: account for I/O wait properly
Bugzilla: 253337

Hi,

Currently, Asynchronous I/O does not show up as io_wait in tools such
as vmstat.  The problem is that the AIO code doesn't call
io_schedule().  The attached patch fixes this problem.  I tested it by
running aio-stress to raw disk and to a file system to ensure that the
i/o wait time was properly accounted for.

This addresses Bugzilla Bug 253337: AIO not shown by iostat, top, etc.

The patch was posted for inclusion upstream last week, and was
accepted into -mm.  Below is a copy of the email I sent to linux-aio
and Andrew.

Cheers,

Jeff

Some months back I proposed changing the schedule() call in
read_events to an io_schedule():
	http://osdir.com/ml/linux.kernel.aio.general/2006-10/msg00024.html
This was rejected as there are AIO operations that do not initiate
disk I/O.  I've had another look at the problem, and the only AIO
operation that will not initiate disk I/O is IOCB_CMD_NOOP.  However,
this command isn't even wired up!

Given that it doesn't work, and hasn't for *years*, I'm going to
suggest again that we do proper I/O accounting when using AIO.  Please
consider the attached patch.

Thanks!

Acked-by: Jarod Wilson <jwilson@redhat.com>
Acked-by: Peter Staubach <staubach@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
---
 fs/aio.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/aio.c b/fs/aio.c
index a140041..a590865 100644
--- a/fs/aio.c
+++ b/fs/aio.c
@@ -305,7 +305,7 @@ static void wait_for_all_aios(struct kioctx *ctx)
 	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
 	while (ctx->reqs_active) {
 		spin_unlock_irq(&ctx->ctx_lock);
-		schedule();
+		io_schedule();
 		set_task_state(tsk, TASK_UNINTERRUPTIBLE);
 		spin_lock_irq(&ctx->ctx_lock);
 	}
@@ -325,7 +325,7 @@ ssize_t fastcall wait_on_sync_kiocb(struct kiocb *iocb)
 		set_current_state(TASK_UNINTERRUPTIBLE);
 		if (!iocb->ki_users)
 			break;
-		schedule();
+		io_schedule();
 	}
 	__set_current_state(TASK_RUNNING);
 	return iocb->ki_user_data;
@@ -1176,7 +1176,7 @@ retry:
 			ret = 0;
 			if (to.timed_out)	/* Only check after read evt */
 				break;
-			schedule();
+			io_schedule();
 			if (signal_pending(tsk)) {
 				ret = -EINTR;
 				break;
-- 
1.5.3.5.645.gbb47