2010-01-07 02:58:29

by Divyesh Shah

[permalink] [raw]
Subject: [PATCH] cfq-iosched: Respect ioprio_class when preempting

In cfq_should_preempt(), we currently allow some cases where a non-RT request
can preempt an ongoing RT cfqq timeslice. This should not happen.
Examples include:

o A sync_noidle wl type non-RT request pre-empting a sync_noidle wl type cfqq
on which we are idling.
o Once we have per-cgroup async queues, a non-RT sync request pre-empting a RT
async cfqq.

Signed-off-by: Divyesh Shah<[email protected]>
---

block/cfq-iosched.c | 6 ++++++
1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 7d6cef9..e9e6610 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -3104,6 +3104,12 @@ cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
return true;

/*
+ * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
+ */
+ if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
+ return false;
+
+ /*
* if the new request is sync, but the currently running queue is
* not, let the sync request have priority.
*/


2010-01-11 15:15:43

by Jens Axboe

[permalink] [raw]
Subject: Re: [PATCH] cfq-iosched: Respect ioprio_class when preempting

On Wed, Jan 06 2010, Divyesh Shah wrote:
> In cfq_should_preempt(), we currently allow some cases where a non-RT request
> can preempt an ongoing RT cfqq timeslice. This should not happen.
> Examples include:
>
> o A sync_noidle wl type non-RT request pre-empting a sync_noidle wl type cfqq
> on which we are idling.
> o Once we have per-cgroup async queues, a non-RT sync request pre-empting a RT
> async cfqq.
>
> Signed-off-by: Divyesh Shah<[email protected]>
> ---
>
> block/cfq-iosched.c | 6 ++++++
> 1 files changed, 6 insertions(+), 0 deletions(-)
>
> diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
> index 7d6cef9..e9e6610 100644
> --- a/block/cfq-iosched.c
> +++ b/block/cfq-iosched.c
> @@ -3104,6 +3104,12 @@ cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
> return true;
>
> /*
> + * Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice.
> + */
> + if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq))
> + return false;

It would be nice to just be able to do

if (new_cfqq->ioprio_class > cfqq->ioprio_class)
return false;

but that doesn't quite work given IOPRIO_CLASS_NONE < _RT. So I think
your patch is fine, I've added it. Thanks.

--
Jens Axboe