2010-07-15 02:01:34

by Gui, Jianfeng/归 剑峰

Subject: [PATCH] CFQ: delay getting cfq group at the very point we really need it

Delay searching or allocating a cfq group at the very point that
we really need it. This change would save some cycles every time
we're getting a cfq queue.

Signed-off-by: Gui Jianfeng <[email protected]>
---
block/cfq-iosched.c | 5 +++--
1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index eb4086f..9b0cf13 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -2820,10 +2820,8 @@ cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync,
{
struct cfq_queue *cfqq, *new_cfqq = NULL;
struct cfq_io_context *cic;
- struct cfq_group *cfqg;

retry:
- cfqg = cfq_get_cfqg(cfqd, 1);
cic = cfq_cic_lookup(cfqd, ioc);
/* cic always exists here */
cfqq = cic_to_cfqq(cic, is_sync);
@@ -2852,6 +2850,9 @@ retry:
}

if (cfqq) {
+ struct cfq_group *cfqg;
+
+ cfqg = cfq_get_cfqg(cfqd, 1);
cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
cfq_init_prio_data(cfqq, ioc);
cfq_link_cfqq_cfqg(cfqq, cfqg);
-- 1.5.4.rc3

2010-07-16 13:52:28

by Vivek Goyal

Subject: Re: [PATCH] CFQ: delay getting cfq group at the very point we really need it

On Thu, Jul 15, 2010 at 09:59:20AM +0800, Gui Jianfeng wrote:
> Delay searching or allocating a cfq group at the very point that
> we really need it. This change would save some cycles every time
> we're getting a cfq queue.
>
> Signed-off-by: Gui Jianfeng <[email protected]>
> ---
> block/cfq-iosched.c | 5 +++--
> 1 files changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
> index eb4086f..9b0cf13 100644
> --- a/block/cfq-iosched.c
> +++ b/block/cfq-iosched.c
> @@ -2820,10 +2820,8 @@ cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync,
> {
> struct cfq_queue *cfqq, *new_cfqq = NULL;
> struct cfq_io_context *cic;
> - struct cfq_group *cfqg;
>
> retry:
> - cfqg = cfq_get_cfqg(cfqd, 1);
> cic = cfq_cic_lookup(cfqd, ioc);
> /* cic always exists here */
> cfqq = cic_to_cfqq(cic, is_sync);
> @@ -2852,6 +2850,9 @@ retry:
> }
>
> if (cfqq) {
> + struct cfq_group *cfqg;
> +
> + cfqg = cfq_get_cfqg(cfqd, 1);
> cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);

Gui, I noticed that once we have setup a cfqq and cfqg, for subequent
requests we should not come into this function at all because in
cfq_set_request(), cic will have pointer to cfqq.

So I am not sure in what call path there is substantial saving but anyway,
it should not harm moving cfq_get_cfqg() call down and call it after
cfqq has been allocated.

Acked-by: Vivek Goyal <[email protected]>

Thanks
Vivek