[PATCH v2] mm,ksm: add __GFP_HIGH to the allocation in alloc_stable_node()

[PATCH v2] mm,ksm: add __GFP_HIGH to the allocation in alloc_stable_node()

[zhongjiang]

From: zhong jiang <zhongjiang@huawei.com>

According to HUgh's suggestion, alloc_stable_node() with GFP_KERNEL
will cause the hungtask, despite less possiblity.

At present, if alloc_stable_node allocate fails, two break_cow may
want to allocate a couple of pages, and the issue will come up when
free memory is under pressure.

we fix it by adding the __GFP_HIGH to GFP. because it grant access to
some of meory reserves. it will make progress to make it allocation
successful at the utmost.

Acked-by: Hugh Dickins <hughd@google.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
---
 mm/ksm.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/mm/ksm.c b/mm/ksm.c
index 5048083..5e98c0b 100644
--- a/mm/ksm.c
+++ b/mm/ksm.c
@@ -299,7 +299,14 @@ static inline void free_rmap_item(struct rmap_item *rmap_item)
 
 static inline struct stable_node *alloc_stable_node(void)
 {
-	return kmem_cache_alloc(stable_node_cache, GFP_KERNEL);
+	/*
+	 * The caller can take too long time with GFP_KERNEL when memory
+	 * is under pressure, it may be lead to the hung task. Therefore,
+	 * Adding the __GFP_HIGH to this. it grant access to some of
+	 * memory reserves. and it will make progress to make it allocation
+	 * successful at the utmost.
+	 */
+	return kmem_cache_alloc(stable_node_cache, GFP_KERNEL | __GFP_HIGH);
 }
 
 static inline void free_stable_node(struct stable_node *stable_node)
-- 
1.8.3.1

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>