Re: [PATCH 4/4] mm: vmscan: If kswapd has been running too long,allow it to sleep

[KOSAKI Motohiro]

> > > > > > While it appears unlikely, there are bad conditions which can result
> > > >
> > > > in cond_resched() being avoided.
> >
> > Every reclaim priority decreasing or every shrink_zone() calling makes more
> > fine grained preemption. I think.
>
> It could be.
> But in direct reclaim case, I have a concern about losing pages
> reclaimed to other tasks by preemption.

Nope, I proposed to add cond_resched() into balance_pgdat().

> Hmm,, anyway, we also needs test.
> Hmm,, how long should we bother them(Colins and James)?
> First of all, Let's fix one just between us and ask test to them and
> send the last patch to akpm.
>
> 1. shrink_slab
> 2. right after balance_pgdat
> 3. shrink_zone
> 4. reclaim priority decreasing routine.
>
> Now, I vote 1) and 2).
>
> Mel, KOSAKI?

I think following patch makes enough preemption.

Thanks.



From e7d88be1916184ea7c93a6f2746b15c7a32d1973 Mon Sep 17 00:00:00 2001
From: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
Date: Wed, 18 May 2011 15:00:39 +0900
Subject: [PATCH] vmscan: balance_pgdat() call cond_resched() unconditionally

Under constant allocation pressure, kswapd can be in the situation where
sleeping_prematurely() will always return true even if kswapd has been
running a long time. Check if kswapd needs to be scheduled.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx>
Cc: Mel Gorman <mgorman@xxxxxxx>
Cc: James Bottomley <James.Bottomley@xxxxxxxxxxxxxxxxxxxxx>
Cc: Colin King <colin.king@xxxxxxxxxxxxx>
Cc: Minchan Kim <minchan.kim@xxxxxxxxx>
---
 mm/vmscan.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 19e179b..87c88fd 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -2449,6 +2449,7 @@ loop_again:
 			sc.nr_reclaimed += reclaim_state->reclaimed_slab;
 			total_scanned += sc.nr_scanned;

+			cond_resched();
 			if (zone->all_unreclaimable)
 				continue;
 			if (nr_slab == 0 &&
@@ -2518,8 +2519,6 @@ out:
 	 *             for the node to be balanced
 	 */
 	if (!(all_zones_ok || (order && pgdat_balanced(pgdat, balanced, *classzone_idx)))) {
-		cond_resched();
-
 		try_to_freeze();

 		/*
--
1.7.3.1




--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/