mm: hwpoison: drop lru_add_drain_all() in __soft_offline_page()
authorNaoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Thu, 12 Feb 2015 23:00:25 +0000 (15:00 -0800)
committerSasha Levin <sasha.levin@oracle.com>
Tue, 28 Apr 2015 14:48:53 +0000 (10:48 -0400)
commitad931555388824fb12bc8f555e6fb6ee57ad4352
tree51b4722b890ac57a0848b0de85a0ee02488c212e
parent3e16c7f2f05592c9dc7c337623786d875035dcdf
mm: hwpoison: drop lru_add_drain_all() in __soft_offline_page()

[ Upstream commit 9ab3b598d2dfbdb0153ffa7e4b1456bbff59a25d ]

A race condition starts to be visible in recent mmotm, where a PG_hwpoison
flag is set on a migration source page *before* it's back in buddy page
poo= l.

This is problematic because no page flag is supposed to be set when
freeing (see __free_one_page().) So the user-visible effect of this race
is that it could trigger the BUG_ON() when soft-offlining is called.

The root cause is that we call lru_add_drain_all() to make sure that the
page is in buddy, but that doesn't work because this function just
schedule= s a work item and doesn't wait its completion.
drain_all_pages() does drainin= g directly, so simply dropping
lru_add_drain_all() solves this problem.

Fixes: f15bdfa802bf ("mm/memory-failure.c: fix memory leak in successful soft offlining")
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Chen Gong <gong.chen@linux.intel.com>
Cc: <stable@vger.kernel.org> [3.11+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
mm/memory-failure.c