mm: add !pte_present() check on existing hugetlb_entry callbacks
authorNaoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Fri, 6 Jun 2014 14:00:01 +0000 (10:00 -0400)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 6 Jun 2014 20:21:16 +0000 (13:21 -0700)
commitd4c54919ed86302094c0ca7d48a8cbd4ee753e92
tree0e2f79515abf8d056d6c32912194e0c69212fad5
parentd54d14bfb49f0b61aed9f20cb84cb692566cf83b
mm: add !pte_present() check on existing hugetlb_entry callbacks

The age table walker doesn't check non-present hugetlb entry in common
path, so hugetlb_entry() callbacks must check it.  The reason for this
behavior is that some callers want to handle it in its own way.

[ I think that reason is bogus, btw - it should just do what the regular
  code does, which is to call the "pte_hole()" function for such hugetlb
  entries  - Linus]

However, some callers don't check it now, which causes unpredictable
result, for example when we have a race between migrating hugepage and
reading /proc/pid/numa_maps.  This patch fixes it by adding !pte_present
checks on buggy callbacks.

This bug exists for years and got visible by introducing hugepage
migration.

ChangeLog v2:
- fix if condition (check !pte_present() instead of pte_present())

Reported-by: Sasha Levin <sasha.levin@oracle.com>
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org> [3.12+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Backported to 3.15.  Signed-off-by: Josh Boyer <jwboyer@fedoraproject.org> ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/proc/task_mmu.c
mm/mempolicy.c