mm: Clear page active before releasing pages

Active pages should not be freed to the page allocator as it triggers a
bad page state warning.  Fengguang Wu reported the following bug and
bisected it to the patch "mm: remove lru parameter from __lru_cache_add
and lru_cache_add_lru" which is currently in mmotm as
mm-remove-lru-parameter-from-__lru_cache_add-and-lru_cache_add_lru.patch

[   84.212960] BUG: Bad page state in process rm  pfn:0b0c9
[   84.214682] page:ffff88000d646240 count:0 mapcount:0 mapping:          (null) index:0x0
[   84.216883] page flags: 0x20000000004c(referenced|uptodate|active)
[   84.218697] CPU: 1 PID: 283 Comm: rm Not tainted 3.10.0-rc4-04361-geeb9bfc #49
[   84.220729]  ffff88000d646240 ffff88000d179bb8 ffffffff82562956 ffff88000d179bd8
[   84.223242]  ffffffff811333f1 000020000000004c ffff88000d646240 ffff88000d179c28
[   84.225387]  ffffffff811346a4 ffff880000270000 0000000000000000 0000000000000006
[   84.227294] Call Trace:
[   84.227867]  [<ffffffff82562956>] dump_stack+0x27/0x30
[   84.229045]  [<ffffffff811333f1>] bad_page+0x130/0x158
[   84.230261]  [<ffffffff811346a4>] free_pages_prepare+0x8b/0x1e3
[   84.231765]  [<ffffffff8113542a>] free_hot_cold_page+0x28/0x1cf
[   84.233171]  [<ffffffff82585830>] ? _raw_spin_unlock_irqrestore+0x6b/0xc6
[   84.234822]  [<ffffffff81135b59>] free_hot_cold_page_list+0x30/0x5a
[   84.236311]  [<ffffffff8113a4ed>] release_pages+0x251/0x267
[   84.237653]  [<ffffffff8112a88d>] ? delete_from_page_cache+0x48/0x9e
[   84.239142]  [<ffffffff8113ad93>] __pagevec_release+0x2b/0x3d
[   84.240473]  [<ffffffff8113b45a>] truncate_inode_pages_range+0x1b0/0x7ce
[   84.242032]  [<ffffffff810e76ab>] ? put_lock_stats.isra.20+0x1c/0x53
[   84.243480]  [<ffffffff810e77f5>] ? lock_release_holdtime+0x113/0x11f
[   84.244935]  [<ffffffff8113ba8c>] truncate_inode_pages+0x14/0x1d
[   84.246337]  [<ffffffff8119b3ef>] evict+0x11f/0x232
[   84.247501]  [<ffffffff8119c527>] iput+0x1a5/0x218
[   84.248607]  [<ffffffff8118f015>] do_unlinkat+0x19b/0x25a
[   84.249828]  [<ffffffff810ea993>] ? trace_hardirqs_on_caller+0x210/0x2ce
[   84.251382]  [<ffffffff8144372e>] ? trace_hardirqs_on_thunk+0x3a/0x3f
[   84.252879]  [<ffffffff8118f10d>] SyS_unlinkat+0x39/0x4c
[   84.254174]  [<ffffffff825874d6>] system_call_fastpath+0x1a/0x1f
[   84.255596] Disabling lock debugging due to kernel taint

The problem was that a page marked for activation was released via
pagevec. This patch clears the active bit before freeing in this case.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Reported-and-Tested-by: Fengguang Wu <fengguang.wu@intel.com>
Tested-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

diff --git a/mm/swap.c b/mm/swap.c
index ac23602b79aacf570ae5cb14ac35021db7e2f7f3..4a1d0d2c52fa4ab4ca66ff4a2025e093521556c9 100644 (file)
--- a/mm/swap.c
+++ b/mm/swap.c
@@ -739,6 +739,9 @@ void release_pages(struct page **pages, int nr, int cold)
                        del_page_from_lru_list(page, lruvec, page_off_lru(page));
                }
 
+               /* Clear Active bit in case of parallel mark_page_accessed */
+               ClearPageActive(page);
+
                list_add(&page->lru, &pages_to_free);
        }
        if (zone)