linux-kernel - [PATCH 2/2] mm: fix endless looping around false-positive too_many

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:	Wed, 07 Dec 2011 19:16:47 +0300
From:	Konstantin Khlebnikov <khlebnikov@...nvz.org>
To:	linux-mm@...ck.org, Andrew Morton <akpm@...ux-foundation.org>
Cc:	linux-kernel@...r.kernel.org, Mel Gorman <mgorman@...e.de>
Subject: [PATCH 2/2] mm: fix endless looping around false-positive
 too_many_isolated()

Due to vmstat counters percpu drift result of too_many_isolated() check
can be false-positive. Unfortunately it can be stable false-positive:
for example if zone at the one moment hasn't active/inactive pages at all
(for small zones like "DMA" this is very likely) but its atomic part of
isolated-pages counter is non-zero. In this sutuation shrink_inactive_list()
and isolate_migratepages() will loop forever around too_many_isolated().

After this patch too_many_isolated() will sum percpu fractions of
isolated pages counter if atomic part above watermark, but not higher than
watermark plus possible percpu drift.

We can ignore drift for active/inactive pages counters, because sooner or later
isolate pages counter drops to zero.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@...nvz.org>
---
 mm/compaction.c |   11 +++++++++--
 mm/vmscan.c     |    5 +++++
 2 files changed, 14 insertions(+), 2 deletions(-)

diff --git a/mm/compaction.c b/mm/compaction.c
index 899d956..2d6fced 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -231,7 +231,7 @@ static void acct_isolated(struct zone *zone, struct compact_control *cc)
 /* Similar to reclaim, but different enough that they don't share logic */
 static bool too_many_isolated(struct zone *zone)
 {
-	unsigned long active, inactive, isolated;
+	unsigned long active, inactive, isolated, watermark;
 
 	inactive = zone_page_state(zone, NR_INACTIVE_FILE) +
 					zone_page_state(zone, NR_INACTIVE_ANON);
@@ -240,7 +240,14 @@ static bool too_many_isolated(struct zone *zone)
 	isolated = zone_page_state(zone, NR_ISOLATED_FILE) +
 					zone_page_state(zone, NR_ISOLATED_ANON);
 
-	return isolated > (inactive + active) / 2;
+	watermark = (inactive + active) / 2;
+
+	if (isolated > watermark &&
+	    isolated - watermark <= zone->percpu_drift * 2)
+		isolated = zone_page_state_snapshot(zone, NR_ISOLATED_FILE) +
+			   zone_page_state_snapshot(zone, NR_ISOLATED_ANON);
+
+	return isolated > watermark;
 }
 
 /* possible outcome of isolate_migratepages */
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 393ebce..3918c5f 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1320,6 +1320,11 @@ static int too_many_isolated(struct zone *zone, int file,
 		isolated = zone_page_state(zone, NR_ISOLATED_ANON);
 	}
 
+	if (isolated > inactive &&
+	    isolated - inactive <= zone->percpu_drift)
+		isolated = zone_page_state_snapshot(zone,
+				file ? NR_ISOLATED_FILE : NR_ISOLATED_ANON);
+
 	return isolated > inactive;
 }
 

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