[Devel] [PATCH RHEL7 COMMIT] mm/zswap: fix potential deadlock in zswap_frontswap_store()

[Konstantin Khorenko]

The commit is pushed to "branch-rh7-3.10.0-514.10.2.vz7.29.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-514.10.2.vz7.29.7
------>
commit d89e689bd0e81aad2145d26f6003c0aece875f63
Author: Andrey Ryabinin <aryabinin at virtuozzo.com>
Date:   Wed Mar 29 16:34:22 2017 +0400

    mm/zswap: fix potential deadlock in zswap_frontswap_store()
    
    zswap_frontswap_store() is called during memory reclaim, therefore
    it can't use __GFP_FS, otherwise we may renter into fs code and deadlock.
    zswap_frontswap_store() also shouldn't do __GFP_IO, to avoid reentering
    into zswap code itself.
    
     WARNING: at mm/slub.c:1250 slab_pre_alloc_hook.isra.42.part.43+0x15/0x17()
    ...
     Call Trace:
     [<ffffffff8163413f>] dump_stack+0x19/0x1b
     [<ffffffff8107b620>] warn_slowpath_common+0x70/0xb0
     [<ffffffff8107b76a>] warn_slowpath_null+0x1a/0x20
     [<ffffffff816311c8>] slab_pre_alloc_hook.isra.42.part.43+0x15/0x17
     [<ffffffff811d7235>] kmem_cache_alloc+0x55/0x220
     [<ffffffff811c4a09>] zswap_frontswap_store+0xe9/0x320
     [<ffffffff811c3d5b>] __frontswap_store+0x7b/0x110
     [<ffffffff811bf6a3>] swap_writepage+0x23/0x80
     [<ffffffff8118fff2>] shrink_page_list+0x4b2/0xa80
     [<ffffffff81190c1b>] shrink_inactive_list+0x1fb/0x6c0
     [<ffffffff811918b5>] shrink_lruvec+0x395/0x800
     [<ffffffff81191e0f>] shrink_zone+0xef/0x2d0
     [<ffffffff81192390>] do_try_to_free_pages+0x170/0x530
     [<ffffffff81192825>] try_to_free_pages+0xd5/0x160
     [<ffffffff81185aa5>] __alloc_pages_nodemask+0x8a5/0xc10
     [<ffffffff811cb1da>] alloc_pages_current+0xaa/0x170
     [<ffffffff811803de>] __get_free_pages+0xe/0x50
     [<ffffffffa029a5aa>] jbd2_alloc+0x8a/0x90 [jbd2]
     [<ffffffffa028fab4>] do_get_write_access+0x1d4/0x4d0 [jbd2]
     [<ffffffffa028fdd7>] jbd2_journal_get_write_access+0x27/0x40 [jbd2]
     [<ffffffffa02e8e0b>] __ext4_journal_get_write_access+0x3b/0x80 [ext4]
     [<ffffffffa02efccd>] ext4_mb_mark_diskspace_used+0x7d/0x4f0 [ext4]
     [<ffffffffa02f156d>] ext4_mb_new_blocks+0x36d/0x620 [ext4]
     [<ffffffffa02e601d>] ext4_ext_map_blocks+0x49d/0xed0 [ext4]
     [<ffffffffa02b2689>] ext4_map_blocks+0x179/0x590 [ext4]
     [<ffffffffa02b5bc2>] ext4_writepages+0x692/0xd60 [ext4]
     [<ffffffff8118828e>] do_writepages+0x1e/0x40
     [<ffffffff812249da>] __writeback_single_inode+0x7a/0x410
     [<ffffffff812257d7>] writeback_sb_inodes+0x2c7/0x540
     [<ffffffff81225aef>] __writeback_inodes_wb+0x9f/0xd0
     [<ffffffff81225d83>] wb_writeback+0x263/0x320
     [<ffffffff8122821d>] bdi_writeback_workfn+0x11d/0x420
     [<ffffffff8109ef4b>] process_one_work+0x17b/0x470
     [<ffffffff8109fd1b>] worker_thread+0x11b/0x400
     [<ffffffff810a779f>] kthread+0xcf/0xe0
     [<ffffffff81644918>] ret_from_fork+0x58/0x90
     ---[ end trace 892c68052e9c3d3f ]---
    
    Also add __GFP_NOWARN as failures here is fine.
    
    https://jira.sw.ru/browse/PSBM-58279
    
    Signed-off-by: Andrey Ryabinin <aryabinin at virtuozzo.com>
---
 mm/zswap.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/zswap.c b/mm/zswap.c
index e2f8c7e..c582024 100644
--- a/mm/zswap.c
+++ b/mm/zswap.c
@@ -665,7 +665,7 @@ static int zswap_frontswap_store(unsigned type, pgoff_t offset,
 	}
 
 	/* allocate entry */
-	entry = zswap_entry_cache_alloc(GFP_KERNEL);
+	entry = zswap_entry_cache_alloc(GFP_NOIO | __GFP_NOWARN);
 	if (!entry) {
 		zswap_reject_kmemcache_fail++;
 		ret = -ENOMEM;