mm/memory.c: use entry = ACCESS_ONCE(*pte) in handle_pte_fault()
authorHugh Dickins <hughd@google.com>
Wed, 6 Aug 2014 23:05:08 +0000 (16:05 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Thu, 7 Aug 2014 01:01:15 +0000 (18:01 -0700)
commitc0d73261f5c1355a35b8b40e871d31578ce0c044
tree9612726cd85eab1bb6f89dcf5c8dd49c532cc572
parent474750aba88817c53f39424e5567b8e4acc4b39b
mm/memory.c: use entry = ACCESS_ONCE(*pte) in handle_pte_fault()

Use ACCESS_ONCE() in handle_pte_fault() when getting the entry or
orig_pte upon which all subsequent decisions and pte_same() tests will
be made.

I have no evidence that its lack is responsible for the mm/filemap.c:202
BUG_ON(page_mapped(page)) in __delete_from_page_cache() found by
trinity, and I am not optimistic that it will fix it.  But I have found
no other explanation, and ACCESS_ONCE() here will surely not hurt.

If gcc does re-access the pte before passing it down, then that would be
disastrous for correct page fault handling, and certainly could explain
the page_mapped() BUGs seen (concurrent fault causing page to be mapped
in a second time on top of itself: mapcount 2 for a single pte).

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Sasha Levin <sasha.levin@oracle.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Konstantin Khlebnikov <koct9i@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/memory.c