[PATCH 3.13.y-ckt 22/57] mm, vmscan: Do not wait for page writeback for GFP_NOFS allocations

From	Kamal Mostafa <kamal@canonical.com>
Newsgroups	linux.kernel
Subject	[PATCH 3.13.y-ckt 22/57] mm, vmscan: Do not wait for page writeback for GFP_NOFS allocations
Date	2015-09-22 21:20 +0200
Message-ID	<qbBgD-665-39@gated-at.bofh.it> (permalink)
References	<qbzRv-3UF-5@gated-at.bofh.it>
Organization	linux.* mail to news gateway

Show all headers | View raw

3.13.11-ckt27 -stable review patch.  If anyone has any objections, please let me know.

------------------

From: Michal Hocko <mhocko@suse.cz>

commit ecf5fc6e9654cd7a268c782a523f072b2f1959f9 upstream.

Nikolay has reported a hang when a memcg reclaim got stuck with the
following backtrace:

PID: 18308  TASK: ffff883d7c9b0a30  CPU: 1   COMMAND: "rsync"
  #0 __schedule at ffffffff815ab152
  #1 schedule at ffffffff815ab76e
  #2 schedule_timeout at ffffffff815ae5e5
  #3 io_schedule_timeout at ffffffff815aad6a
  #4 bit_wait_io at ffffffff815abfc6
  #5 __wait_on_bit at ffffffff815abda5
  #6 wait_on_page_bit at ffffffff8111fd4f
  #7 shrink_page_list at ffffffff81135445
  #8 shrink_inactive_list at ffffffff81135845
  #9 shrink_lruvec at ffffffff81135ead
 #10 shrink_zone at ffffffff811360c3
 #11 shrink_zones at ffffffff81136eff
 #12 do_try_to_free_pages at ffffffff8113712f
 #13 try_to_free_mem_cgroup_pages at ffffffff811372be
 #14 try_charge at ffffffff81189423
 #15 mem_cgroup_try_charge at ffffffff8118c6f5
 #16 __add_to_page_cache_locked at ffffffff8112137d
 #17 add_to_page_cache_lru at ffffffff81121618
 #18 pagecache_get_page at ffffffff8112170b
 #19 grow_dev_page at ffffffff811c8297
 #20 __getblk_slow at ffffffff811c91d6
 #21 __getblk_gfp at ffffffff811c92c1
 #22 ext4_ext_grow_indepth at ffffffff8124565c
 #23 ext4_ext_create_new_leaf at ffffffff81246ca8
 #24 ext4_ext_insert_extent at ffffffff81246f09
 #25 ext4_ext_map_blocks at ffffffff8124a848
 #26 ext4_map_blocks at ffffffff8121a5b7
 #27 mpage_map_one_extent at ffffffff8121b1fa
 #28 mpage_map_and_submit_extent at ffffffff8121f07b
 #29 ext4_writepages at ffffffff8121f6d5
 #30 do_writepages at ffffffff8112c490
 #31 __filemap_fdatawrite_range at ffffffff81120199
 #32 filemap_flush at ffffffff8112041c
 #33 ext4_alloc_da_blocks at ffffffff81219da1
 #34 ext4_rename at ffffffff81229b91
 #35 ext4_rename2 at ffffffff81229e32
 #36 vfs_rename at ffffffff811a08a5
 #37 SYSC_renameat2 at ffffffff811a3ffc
 #38 sys_renameat2 at ffffffff811a408e
 #39 sys_rename at ffffffff8119e51e
 #40 system_call_fastpath at ffffffff815afa89

Dave Chinner has properly pointed out that this is a deadlock in the
reclaim code because ext4 doesn't submit pages which are marked by
PG_writeback right away.

The heuristic was introduced by commit e62e384e9da8 ("memcg: prevent OOM
with too many dirty pages") and it was applied only when may_enter_fs
was specified.  The code has been changed by c3b94f44fcb0 ("memcg:
further prevent OOM with too many dirty pages") which has removed the
__GFP_FS restriction with a reasoning that we do not get into the fs
code.  But this is not sufficient apparently because the fs doesn't
necessarily submit pages marked PG_writeback for IO right away.

ext4_bio_write_page calls io_submit_add_bh but that doesn't necessarily
submit the bio.  Instead it tries to map more pages into the bio and
mpage_map_one_extent might trigger memcg charge which might end up
waiting on a page which is marked PG_writeback but hasn't been submitted
yet so we would end up waiting for something that never finishes.

Fix this issue by replacing __GFP_IO by may_enter_fs check (for case 2)
before we go to wait on the writeback.  The page fault path, which is
the only path that triggers memcg oom killer since 3.12, shouldn't
require GFP_NOFS and so we shouldn't reintroduce the premature OOM
killer issue which was originally addressed by the heuristic.

As per David Chinner the xfs is doing similar thing since 2.6.15 already
so ext4 is not the only affected filesystem.  Moreover he notes:

: For example: IO completion might require unwritten extent conversion
: which executes filesystem transactions and GFP_NOFS allocations. The
: writeback flag on the pages can not be cleared until unwritten
: extent conversion completes. Hence memory reclaim cannot wait on
: page writeback to complete in GFP_NOFS context because it is not
: safe to do so, memcg reclaim or otherwise.

[tytso@mit.edu: corrected the control flow]
Fixes: c3b94f44fcb0 ("memcg: further prevent OOM with too many dirty pages")
Reported-by: Nikolay Borisov <kernel@kyup.com>
Signed-off-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
[ luis: backported to 3.16: used Hugh's backport for 4.1 ]
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
---
 mm/vmscan.c | 14 +++++---------
 1 file changed, 5 insertions(+), 9 deletions(-)

diff --git a/mm/vmscan.c b/mm/vmscan.c
index 9e2f264..e45f42b 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -852,21 +852,17 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 		 *
 		 * 2) Global reclaim encounters a page, memcg encounters a
 		 *    page that is not marked for immediate reclaim or
-		 *    the caller does not have __GFP_IO. In this case mark
+		 *    the caller does not have __GFP_FS (or __GFP_IO if it's
+		 *    simply going to swap, not to fs). In this case mark
 		 *    the page for immediate reclaim and continue scanning.
 		 *
-		 *    __GFP_IO is checked  because a loop driver thread might
+		 *    Require may_enter_fs because we would wait on fs, which
+		 *    may not have submitted IO yet. And the loop driver might
 		 *    enter reclaim, and deadlock if it waits on a page for
 		 *    which it is needed to do the write (loop masks off
 		 *    __GFP_IO|__GFP_FS for this reason); but more thought
 		 *    would probably show more reasons.
 		 *
-		 *    Don't require __GFP_FS, since we're not going into the
-		 *    FS, just waiting on its writeback completion. Worryingly,
-		 *    ext4 gfs2 and xfs allocate pages with
-		 *    grab_cache_page_write_begin(,,AOP_FLAG_NOFS), so testing
-		 *    may_enter_fs here is liable to OOM on them.
-		 *
 		 * 3) memcg encounters a page that is not already marked
 		 *    PageReclaim. memcg does not have any dirty pages
 		 *    throttling so we could easily OOM just because too many
@@ -883,7 +879,7 @@ static unsigned long shrink_page_list(struct list_head *page_list,
 
 			/* Case 2 above */
 			} else if (global_reclaim(sc) ||
-			    !PageReclaim(page) || !(sc->gfp_mask & __GFP_IO)) {
+			    !PageReclaim(page) || !may_enter_fs) {
 				/*
 				 * This is slightly racy - end_page_writeback()
 				 * might have just cleared PageReclaim, then
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Thread

[3.13.y-ckt stable] Linux 3.13.11-ckt27 stable review Kamal Mostafa <kamal@canonical.com> - 2015-09-22 19:50 +0200
  [PATCH 3.13.y-ckt 15/57] target/iscsi: Fix double free of a TUR followed by a solicited NOPOUT Kamal Mostafa <kamal@canonical.com> - 2015-09-22 19:50 +0200
  [PATCH 3.13.y-ckt 12/57] xen-blkback: replace work_pending with work_busy in purge_persistent_gnt() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 19:50 +0200
  [PATCH 3.13.y-ckt 16/57] rbd: fix copyup completion race Kamal Mostafa <kamal@canonical.com> - 2015-09-22 19:50 +0200
  [PATCH 3.13.y-ckt 57/57] net: gso: use feature flag argument in all protocol gso handlers Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 50/57] ipv6: lock socket in ip6_datagram_connect() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 47/57] rds: fix an integer overflow test in rds_info_getsockopt() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 53/57] net: pktgen: fix race between pktgen_thread_worker() and kthread_stop() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 54/57] net: call rcu_read_lock early in process_backlog Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 52/57] net/tipc: initialize security state for new connection socket Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 40/57] ipc/sem.c: update/correct memory barriers Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 21/57] perf: Fix fasync handling on inherited events Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 51/57] rtnetlink: verify IFLA_VF_INFO attributes before passing them to driver Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 56/57] netlink: don't hold mutex in rcu callback when releasing mmapd ring Kamal Mostafa <kamal@canonical.com> - 2015-09-22 20:00 +0200
  [PATCH 3.13.y-ckt 48/57] udp: fix dst races with multicast early demux Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:00 +0200
  [PATCH 3.13.y-ckt 49/57] sparc64: Fix userspace FPU register corruptions. Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:00 +0200
  [PATCH 3.13.y-ckt 44/57] ip6_gre: release cached dst on tunnel removal Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:00 +0200
  [PATCH 3.13.y-ckt 46/57] net: Fix skb_set_peeked use-after-free bug Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:00 +0200
  [PATCH 3.13.y-ckt 39/57] ipc/sem.c: change memory barrier in sem_lock() to smp_rmb() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 43/57] batman-adv: protect tt_local_entry from concurrent delete events Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 36/57] crypto: caam - fix memory corruption in ahash_final_ctx Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 32/57] libiscsi: Fix host busy blocking during connection teardown Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 24/57] ipc: modify message queue accounting to not take kernel data structures into account Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 42/57] arm64: KVM: Fix host crash when injecting a fault into a 32bit guest Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 35/57] EDAC, ppc4xx: Access mci->csrows array elements properly Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 33/57] libfc: Fix fc_exch_recv_req() error path Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 38/57] ipc,sem: fix use after free on IPC_RMID after a task using same semaphore set exits Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 37/57] mm/hwpoison: fix page refcount of unknown non LRU page Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 25/57] ocfs2: fix BUG in ocfs2_downconvert_thread_do_work() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 17/57] md/raid1: extend spinlock to protect raid1_end_read_request against inconsistencies Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 45/57] net: Fix RCU splat in af_key Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 34/57] libfc: Fix fc_fcp_cleanup_each_cmd() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 41/57] Add factory recertified Crucial M500s to blacklist Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:10 +0200
  [PATCH 3.13.y-ckt 04/57] virtio-net: drop NETIF_F_FRAGLIST Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 14/57] regmap: regcache-rbtree: Clean new present bits on present bitmap resize Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 27/57] KVM: x86: Use adjustment in guest cycles when handling MSR_IA32_TSC_ADJUST Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 05/57] RDS: verify the underlying transport exists before creating a connection Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 07/57] xen/gntdevt: Fix race condition in gntdev_release() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 13/57] USB: sierra: add 1199:68AB device ID Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 31/57] drm/radeon: add new OLAND pci id Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 10/57] crypto: ixp4xx - Remove bogus BUG_ON on scattered dst buffer Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 29/57] dm thin metadata: delete btrees when releasing metadata snapshot Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 20/57] xhci: fix off by one error in TRB DMA address boundary check Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 30/57] dm btree: add ref counting ops for the leaves of top level btrees Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 18/57] target: REPORT LUNS should return LUN 0 even for dynamic ACLs Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 02/57] md: flush ->event_work before stopping array. Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 03/57] ipv6: addrconf: validate new MTU before applying it Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 22/57] mm, vmscan: Do not wait for page writeback for GFP_NOFS allocations Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 19/57] MIPS: Fix sched_getaffinity with MT FPAFF enabled Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 28/57] localmodconfig: Use Kbuild files too Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 09/57] nfsd: Drop BUG_ON and ignore SECLABEL on absent filesystem Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 11/57] xen-blkfront: don't add indirect pages to list when !feature_persistent Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 08/57] PCI: Restore PCI_MSIX_FLAGS_BIRMASK definition Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 23/57] MIPS: Make set_pte() SMP safe. Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200
  [PATCH 3.13.y-ckt 26/57] fsnotify: fix oops in fsnotify_clear_marks_by_group_flags() Kamal Mostafa <kamal@canonical.com> - 2015-09-22 21:20 +0200

csiph-web