On Thu, 28 Jul 2011 16:13:09 +0800
Shaohua Li <shaohua.li@xxxxxxxxx> wrote:
> cleanup kswapd_try_to_sleep() a little bit. Sometimes kswapd doesn't
> really sleep. In such case, don't call prepare_to_wait/finish_wait.
> It just wastes CPU.
>
> Signed-off-by: Shaohua Li <shaohua.li@xxxxxxxxx>
> ---
> mm/vmscan.c | 7 +++----
> 1 file changed, 3 insertions(+), 4 deletions(-)
>
> Index: linux/mm/vmscan.c
> ===================================================================
> --- linux.orig/mm/vmscan.c 2011-07-28 15:52:35.000000000 +0800
> +++ linux/mm/vmscan.c 2011-07-28 15:55:56.000000000 +0800
> @@ -2709,13 +2709,11 @@ static void kswapd_try_to_sleep(pg_data_
> if (freezing(current) || kthread_should_stop())
> return;
>
> - prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
> -
> /* Try to sleep for a short interval */
> if (!sleeping_prematurely(pgdat, order, remaining, classzone_idx)) {
> + prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
> remaining = schedule_timeout(HZ/10);
> finish_wait(&pgdat->kswapd_wait, &wait);
> - prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
> }
>
> /*
> @@ -2734,7 +2732,9 @@ static void kswapd_try_to_sleep(pg_data_
> * them before going back to sleep.
> */
> set_pgdat_percpu_threshold(pgdat, calculate_normal_threshold);
> + prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
> schedule();
> + finish_wait(&pgdat->kswapd_wait, &wait);
> set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
> } else {
> if (remaining)
> @@ -2742,7 +2742,6 @@ static void kswapd_try_to_sleep(pg_data_
> else
> count_vm_event(KSWAPD_HIGH_WMARK_HIT_QUICKLY);
> }
> - finish_wait(&pgdat->kswapd_wait, &wait);
> }
Well. Here's some correct waiting code:
prepare_to_wait(...);
if (condition)
schedule();
finish_wait();
And here's come incorrect waiting code:
if (condition) {
<-- if `condition' becomese false here we can
sleep incorrectly and even miss a wakeup.
prepare_to_wait(...);
schedule();
finish_wait();
}
Your patch converts balance_pgdat() from the correct pattern to the
incorrect pattern. This may be OK given the overall sloppiness of the
vmscan synchronisation. But I think we need to convince ourselves that
we aren't adding rarely-occurring bugs or inefficiencies, and that this
change won't cause us to accidentally introduce rarely-occurring bugs
or inefficiencies as the code eveolves.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>