thp: use is_zero_pfn() only after pte_present() check

Use is_zero_pfn() on pteval only after pte_present() check on pteval
(It might be better idea to introduce is_zero_pte() which checks
pte_present() first).

Otherwise when working on a swap or migration entry and if pte_pfn's
result is equal to zero_pfn by chance, we lose user's data in
__collapse_huge_page_copy().  So if you're unlucky, the application
segfaults and finally you could see below message on exit:

BUG: Bad rss-counter state mm:ffff88007f099300 idx:2 val:3

Fixes: ca0984caa823 ("mm: incorporate zero pages into transparent huge pages")
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org>	[4.1+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

diff --git a/mm/huge_memory.c b/mm/huge_memory.c
index 4b06b8db9df23c8f33406586507bbaecf7f5444c..bbac913f96bc1662ac3d7ef37c82080d9c17562f 100644 (file)
--- a/mm/huge_memory.c
+++ b/mm/huge_memory.c
@@ -2206,7 +2206,8 @@ static int __collapse_huge_page_isolate(struct vm_area_struct *vma,
        for (_pte = pte; _pte < pte+HPAGE_PMD_NR;
             _pte++, address += PAGE_SIZE) {
                pte_t pteval = *_pte;
-               if (pte_none(pteval) || is_zero_pfn(pte_pfn(pteval))) {
+               if (pte_none(pteval) || (pte_present(pteval) &&
+                               is_zero_pfn(pte_pfn(pteval)))) {
                        if (!userfaultfd_armed(vma) &&
                            ++none_or_zero <= khugepaged_max_ptes_none)
                                continue;