mm: make compound_head() robust
authorKirill A. Shutemov <kirill.shutemov@linux.intel.com>
Sat, 7 Nov 2015 00:29:54 +0000 (16:29 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 7 Nov 2015 01:50:42 +0000 (17:50 -0800)
commit1d798ca3f16437c71ff63e36597ff07f9c12e4d6
tree4b70d32439fb18ef699175413e4b82c4af206f81
parentf1e61557f0230d51a3df8d825f2c156e75563bff
mm: make compound_head() robust

Hugh has pointed that compound_head() call can be unsafe in some
context. There's one example:

CPU0 CPU1

isolate_migratepages_block()
  page_count()
    compound_head()
      !!PageTail() == true
put_page()
  tail->first_page = NULL
      head = tail->first_page
alloc_pages(__GFP_COMP)
   prep_compound_page()
     tail->first_page = head
     __SetPageTail(p);
      !!PageTail() == true
    <head == NULL dereferencing>

The race is pure theoretical. I don't it's possible to trigger it in
practice. But who knows.

We can fix the race by changing how encode PageTail() and compound_head()
within struct page to be able to update them in one shot.

The patch introduces page->compound_head into third double word block in
front of compound_dtor and compound_order. Bit 0 encodes PageTail() and
the rest bits are pointer to head page if bit zero is set.

The patch moves page->pmd_huge_pte out of word, just in case if an
architecture defines pgtable_t into something what can have the bit 0
set.

hugetlb_cgroup uses page->lru.next in the second tail page to store
pointer struct hugetlb_cgroup. The patch switch it to use page->private
in the second tail page instead. The space is free since ->first_page is
removed from the union.

The patch also opens possibility to remove HUGETLB_CGROUP_MIN_ORDER
limitation, since there's now space in first tail page to store struct
hugetlb_cgroup pointer. But that's out of scope of the patch.

That means page->compound_head shares storage space with:

 - page->lru.next;
 - page->next;
 - page->rcu_head.next;

That's too long list to be absolutely sure, but looks like nobody uses
bit 0 of the word.

page->rcu_head.next guaranteed[1] to have bit 0 clean as long as we use
call_rcu(), call_rcu_bh(), call_rcu_sched(), or call_srcu(). But future
call_rcu_lazy() is not allowed as it makes use of the bit and we can
get false positive PageTail().

[1] http://lkml.kernel.org/g/20150827163634.GD4029@linux.vnet.ibm.com

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
15 files changed:
Documentation/vm/split_page_table_lock
arch/xtensa/configs/iss_defconfig
include/linux/hugetlb_cgroup.h
include/linux/mm.h
include/linux/mm_types.h
include/linux/page-flags.h
mm/Kconfig
mm/debug.c
mm/huge_memory.c
mm/hugetlb.c
mm/hugetlb_cgroup.c
mm/internal.h
mm/memory-failure.c
mm/page_alloc.c
mm/swap.c