queue-4.14/mm-swap-fix-swapoff-with-ksm-pages.patch

   1 From 7af7a8e19f0c5425ff639b0f0d2d244c2a647724 Mon Sep 17 00:00:00 2001
   2 From: Huang Ying <ying.huang@intel.com>
   3 Date: Fri, 28 Dec 2018 00:39:53 -0800
   4 Subject: mm, swap: fix swapoff with KSM pages
   5
   6 From: Huang Ying <ying.huang@intel.com>
   7
   8 commit 7af7a8e19f0c5425ff639b0f0d2d244c2a647724 upstream.
   9
  10 KSM pages may be mapped to the multiple VMAs that cannot be reached from
  11 one anon_vma.  So during swapin, a new copy of the page need to be
  12 generated if a different anon_vma is needed, please refer to comments of
  13 ksm_might_need_to_copy() for details.
  14
  15 During swapoff, unuse_vma() uses anon_vma (if available) to locate VMA and
  16 virtual address mapped to the page, so not all mappings to a swapped out
  17 KSM page could be found.  So in try_to_unuse(), even if the swap count of
  18 a swap entry isn't zero, the page needs to be deleted from swap cache, so
  19 that, in the next round a new page could be allocated and swapin for the
  20 other mappings of the swapped out KSM page.
  21
  22 But this contradicts with the THP swap support.  Where the THP could be
  23 deleted from swap cache only after the swap count of every swap entry in
  24 the huge swap cluster backing the THP has reach 0.  So try_to_unuse() is
  25 changed in commit e07098294adf ("mm, THP, swap: support to reclaim swap
  26 space for THP swapped out") to check that before delete a page from swap
  27 cache, but this has broken KSM swapoff too.
  28
  29 Fortunately, KSM is for the normal pages only, so the original behavior
  30 for KSM pages could be restored easily via checking PageTransCompound().
  31 That is how this patch works.
  32
  33 The bug is introduced by e07098294adf ("mm, THP, swap: support to reclaim
  34 swap space for THP swapped out"), which is merged by v4.14-rc1.  So I
  35 think we should backport the fix to from 4.14 on.  But Hugh thinks it may
  36 be rare for the KSM pages being in the swap device when swapoff, so nobody
  37 reports the bug so far.
  38
  39 Link: http://lkml.kernel.org/r/20181226051522.28442-1-ying.huang@intel.com
  40 Fixes: e07098294adf ("mm, THP, swap: support to reclaim swap space for THP swapped out")
  41 Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
  42 Reported-by: Hugh Dickins <hughd@google.com>
  43 Tested-by: Hugh Dickins <hughd@google.com>
  44 Acked-by: Hugh Dickins <hughd@google.com>
  45 Cc: Rik van Riel <riel@redhat.com>
  46 Cc: Johannes Weiner <hannes@cmpxchg.org>
  47 Cc: Minchan Kim <minchan@kernel.org>
  48 Cc: Shaohua Li <shli@kernel.org>
  49 Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
  50 Cc: <stable@vger.kernel.org>
  51 Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
  52 Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
  53 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  54
  55 ---
  56  mm/swapfile.c |    3 ++-
  57  1 file changed, 2 insertions(+), 1 deletion(-)
  58
  59 --- a/mm/swapfile.c
  60 +++ b/mm/swapfile.c
  61 @@ -2218,7 +2218,8 @@ int try_to_unuse(unsigned int type, bool
  62                  */
  63                 if (PageSwapCache(page) &&
  64                     likely(page_private(page) == entry.val) &&
  65 -                   !page_swapped(page))
  66 +                   (!PageTransCompound(page) ||
  67 +                    !swap_page_trans_huge_swapped(si, entry)))
  68                         delete_from_swap_cache(compound_head(page));
  69
  70                 /*