[PATCH] mm/page_alloc: Fix race conditions on getting migratetype in buffered_rmqueue

[Hui Zhu]

From: Hui Zhu <zhuhui@xxxxxxxxxx>

To test the patch [1], I use KGTP and a script [2] to show NR_FREE_CMA_PAGES
and gross of cma_nr_free.  The values are always not same.
I check the code of pages alloc and free and found that race conditions
on getting migratetype in buffered_rmqueue.
Then I add move the code of getting migratetype inside the zone->lock
protection part.

Because this issue will affect system even if the Linux kernel does't
have [1].  So I post this patch separately.

This patchset is based on fc7f0dd381720ea5ee5818645f7d0e9dece41cb0.

[1] https://lkml.org/lkml/2015/1/18/28
[2] https://github.com/teawater/kgtp/blob/dev/add-ons/cma_free.py

Signed-off-by: Hui Zhu <zhuhui@xxxxxxxxxx>
---
 mm/page_alloc.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 7633c50..f3d6922 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1694,11 +1694,12 @@ again:
 		}
 		spin_lock_irqsave(&zone->lock, flags);
 		page = __rmqueue(zone, order, migratetype);
+		if (page)
+			migratetype = get_pageblock_migratetype(page);
+		else
+			goto failed_unlock;
 		spin_unlock(&zone->lock);
-		if (!page)
-			goto failed;
-		__mod_zone_freepage_state(zone, -(1 << order),
-					  get_freepage_migratetype(page));
+		__mod_zone_freepage_state(zone, -(1 << order), migratetype);
 	}
 
 	__mod_zone_page_state(zone, NR_ALLOC_BATCH, -(1 << order));
@@ -1715,6 +1716,8 @@ again:
 		goto again;
 	return page;
 
+failed_unlock:
+	spin_unlock(&zone->lock);
 failed:
 	local_irq_restore(flags);
 	return NULL;
-- 
1.9.3

--
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/