mm: fix BUG in __split_huge_page_pmd

[Hugh Dickins]

Occasionally we hit the BUG_ON(pmd_trans_huge(*pmd)) at the end of
__split_huge_page_pmd(): seen when doing madvise(,,MADV_DONTNEED).

It's invalid: we don't always have down_write of mmap_sem there:
a racing do_huge_pmd_wp_page() might have copied-on-write to another
huge page before our split_huge_page() got the anon_vma lock.

Forget the BUG_ON, just go back and try again if this happens.
    
Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
Cc: stable@xxxxxxxxxxxxxxx
---

 mm/huge_memory.c |   10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)

--- 3.12-rc5/mm/huge_memory.c	2013-09-16 17:37:56.811072270 -0700
+++ linux/mm/huge_memory.c	2013-10-15 03:40:02.044138488 -0700
@@ -2697,6 +2697,7 @@ void __split_huge_page_pmd(struct vm_are
 
 	mmun_start = haddr;
 	mmun_end   = haddr + HPAGE_PMD_SIZE;
+again:
 	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
 	spin_lock(&mm->page_table_lock);
 	if (unlikely(!pmd_trans_huge(*pmd))) {
@@ -2719,7 +2720,14 @@ void __split_huge_page_pmd(struct vm_are
 	split_huge_page(page);
 
 	put_page(page);
-	BUG_ON(pmd_trans_huge(*pmd));
+
+	/*
+	 * We don't always have down_write of mmap_sem here: a racing
+	 * do_huge_pmd_wp_page() might have copied-on-write to another
+	 * huge page before our split_huge_page() got the anon_vma lock.
+	 */
+	if (unlikely(pmd_trans_huge(*pmd)))
+		goto again;
 }
 
 void split_huge_page_pmd_mm(struct mm_struct *mm, unsigned long address,
--
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/