In __zone_watermark_ok, free and min are signed long type
while z->lowmem_reserve[classzone_idx] is unsigned long type.
So comparision of them could be wrong due to type conversion
to unsigned although free_pages is minus value.
It could return true instead of false in case of order-0 check
so that kswapd could sleep forever. It means livelock because
direct reclaimer loops forever until kswapd set
zone->all_unreclaimable.
Aaditya reported this problem when he test my hotplug patch.
Reported-off-by: Aaditya Kumar <aaditya.ku...@ap.sony.com>
Tested-by: Aaditya Kumar <aaditya.ku...@ap.sony.com>
Signed-off-by: Aaditya Kumar <aaditya.ku...@ap.sony.com>
Signed-off-by: Minchan Kim <minc...@kernel.org>
---
This patch isn't dependent with this series.
It seems to be candidate for -stable but I'm not sure because of this part.
So, pass the decision to akpm.
" - It must fix a real bug that bothers people (not a, "This could be a
problem..." type thing)."
mm/page_alloc.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index f17e6e4..627653c 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1594,6 +1594,7 @@ static bool __zone_watermark_ok(struct zone *z, int
order, unsigned long mark,
{
/* free_pages my go negative - that's OK */
long min = mark;
+ long lowmem_reserve = z->lowmem_reserve[classzone_idx];
int o;
free_pages -= (1 << order) - 1;
@@ -1602,7 +1603,7 @@ static bool __zone_watermark_ok(struct zone *z, int
order, unsigned long mark,
if (alloc_flags & ALLOC_HARDER)
min -= min / 4;
- if (free_pages <= min + z->lowmem_reserve[classzone_idx])
+ if (free_pages <= min + lowmem_reserve)
return false;
for (o = 0; o < order; o++) {
/* At the next order, this order's pages become unavailable */
--
1.7.9.5
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
