[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20090811062243.GP12579@kernel.dk>
Date: Tue, 11 Aug 2009 08:22:43 +0200
From: Jens Axboe <jens.axboe@...cle.com>
To: Gui Jianfeng <guijianfeng@...fujitsu.com>
Cc: linux-kernel@...r.kernel.org
Subject: Re: [PATCH] CFQ: mark must_alloc flag when a cfqq doesn't allocate
any request
On Fri, Aug 07 2009, Gui Jianfeng wrote:
> Hi Jens,
>
> CFQ judges whether must_alloc flag is set, but never marks this flag.
> This patch set must_alloc flag when there is no request allocated by
> a cfqq. This change ensures cfqq will get at least one request even
> if the queue is full.
> Hope it helps.
The flag became redundant with the fix to not fail in set_request(). I
don't think your below patch is a very good idea, it would basically
mark any queue that has previously had a request allocated as a must
alloc queue. It would make more sense to mark an idling queue as must
alloc, since it's critical that this queue is allowed to allocate a
request once we idle a queue on its behalf. But we already check for the
wait_request flag in ->may_queue, so that part is already covered. So
I'd suggest we just delete any reference to must_alloc instead.
>
> Signed-off-by: Gui Jianfeng <guijianfeng@...fujitsu.com>
> ---
> block/cfq-iosched.c | 3 +++
> 1 files changed, 3 insertions(+), 0 deletions(-)
>
> diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
> index fd7080e..531a404 100644
> --- a/block/cfq-iosched.c
> +++ b/block/cfq-iosched.c
> @@ -2279,6 +2279,9 @@ static void cfq_put_request(struct request *rq)
> BUG_ON(!cfqq->allocated[rw]);
> cfqq->allocated[rw]--;
>
> + if (!(cfqq->allocated[0] + cfqq->allocated[1]))
> + cfq_mark_cfqq_must_alloc(cfqq);
> +
> put_io_context(RQ_CIC(rq)->ioc);
>
> rq->elevator_private = NULL;
> --
> 1.5.4.rc3
>
>
--
Jens Axboe
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists