From: Vivek Goyal <vgoyal@redhat.com>
To: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
Cc: Jens Axboe <jens.axboe@oracle.com>,
linux kernel mailing list <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] cfq: Take whether cfq group is changed into account when choosing service tree
Date: Fri, 11 Dec 2009 10:07:27 -0500 [thread overview]
Message-ID: <20091211150727.GB2756@redhat.com> (raw)
In-Reply-To: <4B21D252.1060902@cn.fujitsu.com>
On Fri, Dec 11, 2009 at 01:02:10PM +0800, Gui Jianfeng wrote:
> Currently, with IO Controller introduced, CFQ chooses cfq group
> at the top, and then choose service tree. So we need to take
> whether cfq group is changed into account to decide whether we
> should choose service tree start from scratch.
>
I am not able to understand the need/purpose of this patch. Once we
switched the group during scheduling, why should we reset the order
of workload with-in group. In fact we don't want to do that and we
want to start from where we left so that no workload with-in group is
starved.
When a group is resumed, choose_service_tree() always checks if any RT
queue got backlogged or not. If yes, it does choose that first.
Can you give more details why do you want to do this change?
Thanks
Vivek
> Signed-off-by: Gui Jianfeng <guijianfeng@cn.fujitsu.com>
> ---
> block/cfq-iosched.c | 27 ++++++++++++++++++---------
> 1 files changed, 18 insertions(+), 9 deletions(-)
>
> diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
> index f3f6239..16084ca 100644
> --- a/block/cfq-iosched.c
> +++ b/block/cfq-iosched.c
> @@ -1964,7 +1964,7 @@ static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq)
>
> static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd,
> struct cfq_group *cfqg, enum wl_prio_t prio,
> - bool prio_changed)
> + bool cfqg_changed, bool prio_changed)
> {
> struct cfq_queue *queue;
> int i;
> @@ -1972,7 +1972,7 @@ static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd,
> unsigned long lowest_key = 0;
> enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD;
>
> - if (prio_changed) {
> + if (cfqg_changed || prio_changed) {
> /*
> * When priorities switched, we prefer starting
> * from SYNC_NOIDLE (first choice), or just SYNC
> @@ -2001,7 +2001,8 @@ static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd,
> return cur_best;
> }
>
> -static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg)
> +static void choose_service_tree(struct cfq_data *cfqd,
> + struct cfq_group *cfqg, bool cfqg_changed)
> {
> enum wl_prio_t previous_prio = cfqd->serving_prio;
> bool prio_changed;
> @@ -2033,21 +2034,22 @@ static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg)
> * expiration time
> */
> prio_changed = (cfqd->serving_prio != previous_prio);
> - st = service_tree_for(cfqg, cfqd->serving_prio, cfqd->serving_type,
> - cfqd);
> + st = service_tree_for(cfqg, cfqd->serving_prio, cfqd->serving_type,
> + cfqd);
> count = st->count;
>
> /*
> * If priority didn't change, check workload expiration,
> * and that we still have other queues ready
> */
> - if (!prio_changed && count &&
> + if (!cfqg_changed && !prio_changed && count &&
> !time_after(jiffies, cfqd->workload_expires))
> return;
>
> /* otherwise select new workload type */
> cfqd->serving_type =
> - cfq_choose_wl(cfqd, cfqg, cfqd->serving_prio, prio_changed);
> + cfq_choose_wl(cfqd, cfqg, cfqd->serving_prio,
> + cfqg_changed, prio_changed);
> st = service_tree_for(cfqg, cfqd->serving_prio, cfqd->serving_type,
> cfqd);
> count = st->count;
> @@ -2104,9 +2106,16 @@ static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd)
>
> static void cfq_choose_cfqg(struct cfq_data *cfqd)
> {
> + bool cfqg_changed = false;
> +
> + struct cfq_group *orig_cfqg = cfqd->serving_group;
> +
> struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd);
>
> - cfqd->serving_group = cfqg;
> + if (orig_cfqg != cfqg) {
> + cfqg_changed = 1;
> + cfqd->serving_group = cfqg;
> + }
>
> /* Restore the workload type data */
> if (cfqg->saved_workload_slice) {
> @@ -2114,7 +2123,7 @@ static void cfq_choose_cfqg(struct cfq_data *cfqd)
> cfqd->serving_type = cfqg->saved_workload;
> cfqd->serving_prio = cfqg->saved_serving_prio;
> }
> - choose_service_tree(cfqd, cfqg);
> + choose_service_tree(cfqd, cfqg, cfqg_changed);
> }
>
> /*
> --
> 1.5.4.rc3
>
next prev parent reply other threads:[~2009-12-11 15:07 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-12-11 5:02 [PATCH] cfq: Take whether cfq group is changed into account when choosing service tree Gui Jianfeng
2009-12-11 15:07 ` Vivek Goyal [this message]
2009-12-11 18:01 ` Corrado Zoccolo
2009-12-11 18:46 ` Vivek Goyal
2009-12-14 2:37 ` Gui Jianfeng
2009-12-14 8:39 ` Corrado Zoccolo
2009-12-14 9:54 ` Gui Jianfeng
2009-12-14 11:01 ` Corrado Zoccolo
2009-12-15 0:52 ` Vivek Goyal
2009-12-15 1:06 ` Gui Jianfeng
2009-12-15 15:23 ` Vivek Goyal
2009-12-15 16:04 ` Corrado Zoccolo
2009-12-15 16:22 ` Vivek Goyal
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20091211150727.GB2756@redhat.com \
--to=vgoyal@redhat.com \
--cc=guijianfeng@cn.fujitsu.com \
--cc=jens.axboe@oracle.com \
--cc=linux-kernel@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox