rmap: recompute pgoff for unmapping huge page
authorHillf Danton <dhillf@gmail.com>
Mon, 29 Apr 2013 22:06:14 +0000 (15:06 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Mon, 29 Apr 2013 22:54:28 +0000 (15:54 -0700)
We have to recompute pgoff if the given page is huge, since result based
on HPAGE_SIZE is not approapriate for scanning the vma interval tree, as
shown by commit 36e4f20af833 ("hugetlb: do not use vma_hugecache_offset()
for vma_prio_tree_foreach").

Signed-off-by: Hillf Danton <dhillf@gmail.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Michel Lespinasse <walken@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/rmap.c

index 807c96bf0dc6d08cdc167ce432a68999b1e21692..6280da86b5d6761ed8a245c3fadb6e42015216c5 100644 (file)
--- a/mm/rmap.c
+++ b/mm/rmap.c
@@ -1513,6 +1513,9 @@ static int try_to_unmap_file(struct page *page, enum ttu_flags flags)
        unsigned long max_nl_size = 0;
        unsigned int mapcount;
 
+       if (PageHuge(page))
+               pgoff = page->index << compound_order(page);
+
        mutex_lock(&mapping->i_mmap_mutex);
        vma_interval_tree_foreach(vma, &mapping->i_mmap, pgoff, pgoff) {
                unsigned long address = vma_address(page, vma);