mm/memory-failure: me_huge_page() does nothing for thp

memory_failure() is supposed not to handle thp itself, but to split it.
But if something were wrong and page_action() were called on thp,
me_huge_page() (action routine for hugepages) should be better to take
no action, rather than to take wrong action prepared for hugetlb (which
triggers BUG_ON().)

This change is for potential problems, but makes sense to me because thp
is an actively developing feature and this code path can be open in the
future.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/mm/memory-failure.c b/mm/memory-failure.c
index 62ebb1b..c72f41b 100644 (file)
--- a/mm/memory-failure.c
+++ b/mm/memory-failure.c
@@ -777,6 +777,10 @@ static int me_huge_page(struct page *p, unsigned long pfn)
 {
        int res = 0;
        struct page *hpage = compound_head(p);
+
+       if (!PageHuge(hpage))
+               return MF_DELAYED;
+
        /*
         * We can safely recover from error on free or reserved (i.e.
         * not in-use) hugepage by dequeuing it from freelist.