[PATCH] Use min of two prio settings in calculating distress forreclaim

From: Martin Bligh
Date: Mon Oct 16 2006 - 21:31:32 EST


Another bug is that if try_to_free_pages / balance_pgdat are called
with a gfp_mask specifying GFP_IO and/or GFP_FS, they may reclaim
the requisite number of pages, and reset prev_priority to DEF_PRIORITY.

However, another reclaimer without those gfp_mask flags set may still
be struggling to reclaim pages. The easy fix for this is to key the
distress calculation not off zone->prev_priority, but also take into
account the local caller's priority by using:
min(zone->prev_priority, sc->priority)

Signed-off-by: Martin J. Bligh <mbligh@xxxxxxxxxx> diff -aurpN -X /home/mbligh/.diff.exclude linux-2.6.18/mm/vmscan.c 2.6.18-min_prio/mm/vmscan.c
--- linux-2.6.18/mm/vmscan.c 2006-09-20 12:24:42.000000000 -0700
+++ 2.6.18-min_prio/mm/vmscan.c 2006-10-16 18:16:39.000000000 -0700
@@ -713,7 +713,7 @@ done:
* But we had to alter page->flags anyway.
*/
static void shrink_active_list(unsigned long nr_pages, struct zone *zone,
- struct scan_control *sc)
+ struct scan_control *sc, int priority)
{
unsigned long pgmoved;
int pgdeactivate = 0;
@@ -734,7 +734,7 @@ static void shrink_active_list(unsigned
* `distress' is a measure of how much trouble we're having
* reclaiming pages. 0 -> no problems. 100 -> great trouble.
*/
- distress = 100 >> zone->prev_priority;
+ distress = 100 >> min(zone->prev_priority, priority);

/*
* The point of this algorithm is to decide when to start
@@ -885,7 +885,7 @@ static unsigned long shrink_zone(int pri
nr_to_scan = min(nr_active,
(unsigned long)sc->swap_cluster_max);
nr_active -= nr_to_scan;
- shrink_active_list(nr_to_scan, zone, sc);
+ shrink_active_list(nr_to_scan, zone, sc, priority);
}

if (nr_inactive) {
@@ -1315,7 +1315,7 @@ static unsigned long shrink_all_zones(un
if (zone->nr_scan_active >= nr_pages || pass > 3) {
zone->nr_scan_active = 0;
nr_to_scan = min(nr_pages, zone->nr_active);
- shrink_active_list(nr_to_scan, zone, sc);
+ shrink_active_list(nr_to_scan, zone, sc, prio);
}
}