Re: [PATCH] cleanup kswapd()

Previous thread: [PATCH/RFC] netfilter: nf_conntrack_sip: Handle quirky Cisco phones by Kevin Cernekee on Sunday, November 14, 2010 - 1:32 am. (9 messages)

Next thread: [GIT PULL] Important ocfs2 fix for 2.6.37-rc1 by Joel Becker on Sunday, November 14, 2010 - 2:31 am. (1 message)
From: KOSAKI Motohiro
Date: Sunday, November 14, 2010 - 2:05 am

Currently, kswapd() function has deeper nest and it slightly harder to
read. cleanup it.

Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
---
 mm/vmscan.c |   71 +++++++++++++++++++++++++++++++---------------------------
 1 files changed, 38 insertions(+), 33 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 8cc90d5..82ffe5f 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2364,6 +2364,42 @@ out:
 	return sc.nr_reclaimed;
 }
 
+void kswapd_try_to_sleep(pg_data_t *pgdat, int order)
+{
+	long remaining = 0;
+	DEFINE_WAIT(wait);
+
+	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)) {
+		remaining = schedule_timeout(HZ/10);
+		finish_wait(&pgdat->kswapd_wait, &wait);
+		prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
+	}
+
+	/*
+	 * After a short sleep, check if it was a
+	 * premature sleep. If not, then go fully
+	 * to sleep until explicitly woken up
+	 */
+	if (!sleeping_prematurely(pgdat, order, remaining)) {
+		trace_mm_vmscan_kswapd_sleep(pgdat->node_id);
+		set_pgdat_percpu_threshold(pgdat, calculate_normal_threshold);
+		schedule();
+		set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
+	} else {
+		if (remaining)
+			count_vm_event(KSWAPD_LOW_WMARK_HIT_QUICKLY);
+		else
+			count_vm_event(KSWAPD_HIGH_WMARK_HIT_QUICKLY);
+	}
+	finish_wait(&pgdat->kswapd_wait, &wait);
+}
+
 /*
  * The background pageout daemon, started as a kernel thread
  * from the init process.
@@ -2382,7 +2418,7 @@ static int kswapd(void *p)
 	unsigned long order;
 	pg_data_t *pgdat = (pg_data_t*)p;
 	struct task_struct *tsk = current;
-	DEFINE_WAIT(wait);
+
 	struct reclaim_state reclaim_state = {
 		.reclaimed_slab = 0,
 	};
@@ -2414,7 +2450,6 @@ static int kswapd(void *p)
 		unsigned long new_order;
 		int ...
From: Jesper Juhl
Date: Sunday, November 14, 2010 - 4:03 am

Shouldn't this be

  static void kswapd_try_to_sleep(pg_data_t *pgdat, int order)

??


-- 
Jesper Juhl <jj@chaosbits.net>            http://www.chaosbits.net/
Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html
Plain text mails only, please.

--

From: KOSAKI Motohiro
Date: Sunday, November 14, 2010 - 5:27 pm

Right. thank you.
I'll respin.




--

From: KOSAKI Motohiro
Date: Sunday, November 14, 2010 - 6:37 pm

Done.



From a29f0f5b780170fc26eb9210df03ed974aad8362 Mon Sep 17 00:00:00 2001
From: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Date: Fri, 3 Dec 2010 10:48:41 +0900
Subject: [PATCH] factor out kswapd sleeping logic from kswapd()

Currently, kswapd() function has deeper nest and it slightly harder to
read. cleanup it.

Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
---
 mm/vmscan.c |   71 +++++++++++++++++++++++++++++++---------------------------
 1 files changed, 38 insertions(+), 33 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 8cc90d5..3ee33a8 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2364,6 +2364,42 @@ out:
 	return sc.nr_reclaimed;
 }
 
+static void kswapd_try_to_sleep(pg_data_t *pgdat, int order)
+{
+	long remaining = 0;
+	DEFINE_WAIT(wait);
+
+	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)) {
+		remaining = schedule_timeout(HZ/10);
+		finish_wait(&pgdat->kswapd_wait, &wait);
+		prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
+	}
+
+	/*
+	 * After a short sleep, check if it was a
+	 * premature sleep. If not, then go fully
+	 * to sleep until explicitly woken up
+	 */
+	if (!sleeping_prematurely(pgdat, order, remaining)) {
+		trace_mm_vmscan_kswapd_sleep(pgdat->node_id);
+		set_pgdat_percpu_threshold(pgdat, calculate_normal_threshold);
+		schedule();
+		set_pgdat_percpu_threshold(pgdat, calculate_pressure_threshold);
+	} else {
+		if (remaining)
+			count_vm_event(KSWAPD_LOW_WMARK_HIT_QUICKLY);
+		else
+			count_vm_event(KSWAPD_HIGH_WMARK_HIT_QUICKLY);
+	}
+	finish_wait(&pgdat->kswapd_wait, &wait);
+}
+
 /*
  * The background pageout daemon, started as a kernel thread
  * from the init process.
@@ -2382,7 +2418,7 @@ static int kswapd(void *p)
 	unsigned long order;
 ...
From: Mel Gorman
Date: Monday, November 15, 2010 - 2:42 am

I posted a patch adding a comment on why set_pgdat_percpu_threshold() is
called. I do not believe it has been picked up by Andrew but it if is,
the patches will conflict. The resolution will be obvious but you may
need to respin this patch if the comment patch gets picked up in mmotm.


-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
--

From: KOSAKI Motohiro
Date: Monday, November 15, 2010 - 11:07 pm

OK, I've rebased the patch on top your comment patch. 



From 1bd232713d55f033676f80cc7451ff83d4483884 Mon Sep 17 00:00:00 2001
From: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Date: Mon, 6 Dec 2010 20:44:27 +0900
Subject: [PATCH] factor out kswapd sleeping logic from kswapd()

Currently, kswapd() function has deeper nest and it slightly harder to
read. cleanup it.

Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
---
 mm/vmscan.c |   92 +++++++++++++++++++++++++++++-----------------------------
 1 files changed, 46 insertions(+), 46 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 33994b7..cd07b97 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2364,6 +2364,50 @@ out:
 	return sc.nr_reclaimed;
 }
 
+static void kswapd_try_to_sleep(pg_data_t *pgdat, int order)
+{
+	long remaining = 0;
+	DEFINE_WAIT(wait);
+
+	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)) {
+		remaining = schedule_timeout(HZ/10);
+		finish_wait(&pgdat->kswapd_wait, &wait);
+		prepare_to_wait(&pgdat->kswapd_wait, &wait, TASK_INTERRUPTIBLE);
+	}
+
+	/*
+	 * After a short sleep, check if it was a premature sleep. If not, then
+	 * go fully to sleep until explicitly woken up.
+	 */
+	if (!sleeping_prematurely(pgdat, order, remaining)) {
+		trace_mm_vmscan_kswapd_sleep(pgdat->node_id);
+
+		/*
+		 * vmstat counters are not perfectly accurate and the estimated
+		 * value for counters such as NR_FREE_PAGES can deviate from the
+		 * true value by nr_online_cpus * threshold. To avoid the zone
+		 * watermarks being breached while under pressure, we reduce the
+		 * per-cpu vmstat threshold while kswapd is awake and restore
+		 * them before going back to sleep.
+		 */
+		set_pgdat_percpu_threshold(pgdat, ...
From: Mel Gorman
Date: Thursday, November 18, 2010 - 10:27 am

Acked-by: Mel Gorman <mel@csn.ul.ie>

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
--

Previous thread: [PATCH/RFC] netfilter: nf_conntrack_sip: Handle quirky Cisco phones by Kevin Cernekee on Sunday, November 14, 2010 - 1:32 am. (9 messages)

Next thread: [GIT PULL] Important ocfs2 fix for 2.6.37-rc1 by Joel Becker on Sunday, November 14, 2010 - 2:31 am. (1 message)