Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1315387

[PATCH 3.13.y-ckt 057/108] mm, vmstat: allow WQ concurrency to discover memory reclaim doesn't make any progress

From Kamal Mostafa <kamal@canonical.com>
Newsgroups linux.kernel
Subject [PATCH 3.13.y-ckt 057/108] mm, vmstat: allow WQ concurrency to discover memory reclaim doesn't make any progress
Date 2016-01-23 00:50 +0100
Message-ID <qTTCN-2K6-1@gated-at.bofh.it> (permalink)
References <qTTt7-2G5-3@gated-at.bofh.it>
Organization linux.* mail to news gateway

Show all headers | View raw

3.13.11-ckt33 -stable review patch.  If anyone has any objections, please let me know.

---8<------------------------------------------------------------

From: Michal Hocko <mhocko@suse.com>

commit 373ccbe5927034b55bdc80b0f8b54d6e13fe8d12 upstream.

Tetsuo Handa has reported that the system might basically livelock in
OOM condition without triggering the OOM killer.

The issue is caused by internal dependency of the direct reclaim on
vmstat counter updates (via zone_reclaimable) which are performed from
the workqueue context.  If all the current workers get assigned to an
allocation request, though, they will be looping inside the allocator
trying to reclaim memory but zone_reclaimable can see stalled numbers so
it will consider a zone reclaimable even though it has been scanned way
too much.  WQ concurrency logic will not consider this situation as a
congested workqueue because it relies that worker would have to sleep in
such a situation.  This also means that it doesn't try to spawn new
workers or invoke the rescuer thread if the one is assigned to the
queue.

In order to fix this issue we need to do two things.  First we have to
let wq concurrency code know that we are in trouble so we have to do a
short sleep.  In order to prevent from issues handled by 0e093d99763e
("writeback: do not sleep on the congestion queue if there are no
congested BDIs or if significant congestion is not being encountered in
the current zone") we limit the sleep only to worker threads which are
the ones of the interest anyway.

The second thing to do is to create a dedicated workqueue for vmstat and
mark it WQ_MEM_RECLAIM to note it participates in the reclaim and to
have a spare worker thread for it.

Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Tejun Heo <tj@kernel.org>
Cc: Cristopher Lameter <clameter@sgi.com>
Cc: Joonsoo Kim <js1304@gmail.com>
Cc: Arkadiusz Miskiewicz <arekm@maven.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ben Hutchings <ben@decadent.org.uk>
[ luis: backported to 3.16, based on Ben's backport to 3.2:
  - use schedule_delayed_work instead of queue_delayed_work_on function
    vmstat_update()
  - change start_cpu_timer() instead of vmstat_shepherd()
  - adjusted context ]
Signed-off-by: Luis Henriques <luis.henriques@canonical.com>
[ kamal: backport to 3.13-stable: context ]
Signed-off-by: Kamal Mostafa <kamal@canonical.com>
---
 mm/backing-dev.c | 19 ++++++++++++++++---
 mm/vmstat.c      |  6 ++++--
 2 files changed, 20 insertions(+), 5 deletions(-)

diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 09d9591..d8b4973 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -611,8 +611,9 @@ EXPORT_SYMBOL(congestion_wait);
  * jiffies for either a BDI to exit congestion of the given @sync queue
  * or a write to complete.
  *
- * In the absence of zone congestion, cond_resched() is called to yield
- * the processor if necessary but otherwise does not sleep.
+ * In the absence of zone congestion, a short sleep or a cond_resched is
+ * performed to yield the processor and to allow other subsystems to make
+ * a forward progress.
  *
  * The return value is 0 if the sleep is for the full timeout. Otherwise,
  * it is the number of jiffies that were still remaining when the function
@@ -632,7 +633,19 @@ long wait_iff_congested(struct zone *zone, int sync, long timeout)
 	 */
 	if (atomic_read(&nr_bdi_congested[sync]) == 0 ||
 			!zone_is_reclaim_congested(zone)) {
-		cond_resched();
+
+		/*
+		 * Memory allocation/reclaim might be called from a WQ
+		 * context and the current implementation of the WQ
+		 * concurrency control doesn't recognize that a particular
+		 * WQ is congested if the worker thread is looping without
+		 * ever sleeping. Therefore we have to do a short sleep
+		 * here rather than calling cond_resched().
+		 */
+		if (current->flags & PF_WQ_WORKER)
+			schedule_timeout(1);
+		else
+			cond_resched();
 
 		/* In case we scheduled, work out time remaining */
 		ret = timeout - (jiffies - start);
diff --git a/mm/vmstat.c b/mm/vmstat.c
index 7249614..674de36 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -1212,13 +1212,14 @@ static const struct file_operations proc_vmstat_file_operations = {
 #endif /* CONFIG_PROC_FS */
 
 #ifdef CONFIG_SMP
+static struct workqueue_struct *vmstat_wq;
 static DEFINE_PER_CPU(struct delayed_work, vmstat_work);
 int sysctl_stat_interval __read_mostly = HZ;
 
 static void vmstat_update(struct work_struct *w)
 {
 	refresh_cpu_vm_stats();
-	schedule_delayed_work(&__get_cpu_var(vmstat_work),
+	queue_delayed_work(vmstat_wq, &__get_cpu_var(vmstat_work),
 		round_jiffies_relative(sysctl_stat_interval));
 }
 
@@ -1227,7 +1228,7 @@ static void start_cpu_timer(int cpu)
 	struct delayed_work *work = &per_cpu(vmstat_work, cpu);
 
 	INIT_DEFERRABLE_WORK(work, vmstat_update);
-	schedule_delayed_work_on(cpu, work, __round_jiffies_relative(HZ, cpu));
+	queue_delayed_work_on(cpu, vmstat_wq, work, __round_jiffies_relative(HZ, cpu));
 }
 
 static void vmstat_cpu_dead(int node)
@@ -1298,6 +1299,7 @@ static int __init setup_vmstat(void)
 		node_set_state(cpu_to_node(cpu), N_CPU);
 	}
 	put_online_cpus();
+	vmstat_wq = alloc_workqueue("vmstat", WQ_FREEZABLE|WQ_MEM_RECLAIM, 0);
 #endif
 #ifdef CONFIG_PROC_FS
 	proc_create("buddyinfo", S_IRUGO, NULL, &fragmentation_file_operations);
-- 
1.9.1

Back to linux.kernel | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

[3.13.y-ckt stable] Linux 3.13.11-ckt33 stable review Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 086/108] x86/mce: Ensure offline CPUs don't participate in rendezvous process Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 053/108] xhci: fix usb2 resume timing and races. Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 009/108] sh_eth: fix kernel oops in skb_put() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 093/108] async_tx: use GFP_NOWAIT rather than GFP_IO Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 102/108] pinctrl: bcm2835: Fix initial value for direction_output Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 105/108] qlcnic: fix a timeout loop Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 095/108] ftrace/module: Call clean up function when module init fails early Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 088/108] ALSA: hda/realtek - Fix silent headphone output on MacPro 4,1 (v2) Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 107/108] include/linux/mmdebug.h: should include linux/bug.h Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 090/108] mm/memory_hotplug.c: check for missing sections in test_pages_in_a_zone() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 075/108] xen-netback: use RING_COPY_REQUEST() throughout Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 096/108] vmstat: allocate vmstat_wq before it is used Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 103/108] mISDN: fix a loop count Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 092/108] tracing: Fix setting of start_index in find_next() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 099/108] ipv6/addrlabel: fix ip6addrlbl_get() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 091/108] ftrace/scripts: Fix incorrect use of sprintf in recordmcount Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 077/108] xen-blkback: read from indirect descriptors only once Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:40 +0100
  [PATCH 3.13.y-ckt 057/108] mm, vmstat: allow WQ concurrency to discover memory reclaim doesn't make any progress Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 076/108] xen-blkback: only read request operation from shared ring once Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 050/108] ses: Fix problems with simple enclosures Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 059/108] drivers/base/memory.c: prohibit offlining of memory blocks with missing sections Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 049/108] dm btree: fix bufio buffer leaks in dm_btree_del() error path Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 052/108] ses: fix additional element traversal bug Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 073/108] xen: Add RING_COPY_REQUEST() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 072/108] s390/dis: Fix handling of format specifiers Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 070/108] ARC: dw2 unwind: Ignore CIE version !=1 gracefully instead of bailing Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 045/108] dm space map metadata: fix ref counting bug when bootstrapping a new space map Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 071/108] ALSA: hda - Set SKL+ hda controller power at freeze() and thaw() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 068/108] ftrace/scripts: Have recordmcount copy the object file Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 066/108] spi: fix parent-device reference leak Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 064/108] ALSA: usb-audio: Add a more accurate volume quirk for AudioQuest DragonFly Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 058/108] mm: hugetlb: call huge_pte_alloc() only if ptep is null Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 062/108] tty: Fix GPF in flush_to_ldisc() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 082/108] xen/pciback: For XEN_PCI_OP_disable_msi[|x] only disable if device has MSI(X) enabled. Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 081/108] xen/pciback: Do not install an IRQ handler for MSI interrupts. Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 054/108] USB: add quirk for devices with broken LPM Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 083/108] xen/pciback: Don't allow MSI-X ops if PCI_COMMAND_MEMORY is not set. Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 051/108] vgaarb: fix signal handling in vga_get() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 046/108] ipmi: move timer init to before irq is setup Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 069/108] ARC: dw2 unwind: Reinstante unwinding out of modules Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 074/108] xen-netback: don't use last request to determine minimum Tx credit Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 063/108] genirq: Prevent chip buslock deadlock Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 060/108] sh64: fix __NR_fgetxattr Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 048/108] rfkill: copy the name into the rfkill struct Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 078/108] xen/pciback: Save xen_pci_op commands before processing it Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 079/108] xen/pciback: Return error on XEN_PCI_OP_enable_msi when device has MSI or MSI-X enabled Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 061/108] n_tty: Fix poll() after buffer-limited eof push read Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 056/108] parisc iommu: fix panic due to trying to allocate too large region Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 055/108] powercap / RAPL: fix BIOS lock check Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 080/108] xen/pciback: Return error on XEN_PCI_OP_enable_msix when device has MSI or MSI-X enabled Kamal Mostafa <kamal@canonical.com> - 2016-01-23 00:50 +0100
  [PATCH 3.13.y-ckt 022/108] USB: cdc_acm: Ignore Infineon Flash Loader utility Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 015/108] tools: Add a "make all" rule Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 020/108] iio: fix some warning messages Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 031/108] irqchip/versatile-fpga: Fix PCI IRQ mapping on Versatile PB Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 014/108] KVM: x86: Reload pit counters for all channels when restoring state Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 026/108] drm/ttm: Fixed a read/write lock imbalance Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 023/108] USB: serial: Another Infineon flash loader USB ID Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 042/108] radeon/cik: Fix GFX IB test on Big-Endian Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 025/108] jbd2: Fix unreclaimed pages after truncate in data=journal mode Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 021/108] USB: cp210x: Remove CP2110 ID from compatibility list Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 038/108] ALSA: rme96: Fix unexpected volume reset after rate changes Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 017/108] net: ipmr: fix static mfc/dev leaks on table destruction Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 036/108] SCSI: Fix NULL pointer dereference in runtime PM Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 007/108] ipv6: sctp: clone options to avoid use after free Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 018/108] fuse: break infinite loop in fuse_fill_write_pages() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 041/108] 9p: ->evict_inode() should kick out ->i_data, not ->i_mapping Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 029/108] sata_sil: disable trim Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 040/108] virtio: fix memory leak of virtio ida cache layers Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 010/108] pptp: verify sockaddr_len in pptp_bind() and pptp_connect() Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 043/108] crypto: skcipher - Copy iv from desc even for 0-len walks Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 019/108] usb: gadget: pxa27x: fix suspend callback Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 030/108] staging: lustre: echo_copy.._lsm() dereferences userland pointers directly Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 037/108] usb: xhci: fix config fail of FS hub behind a HS hub with MTT Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 034/108] usb: Use the USB_SS_MULT() macro to decode burst multiplier for log message Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 027/108] i2c: mv64xxx: The n clockdiv factor is 0 based on sunxi SoCs Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 035/108] dm btree: fix leak of bufio-backed block in btree_split_sibling error path Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 033/108] USB: whci-hcd: add check for dma mapping error Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 016/108] efi: Disable interrupts around EFI calls, not in the epilog/prolog calls Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 008/108] net: add validation for the socket syscall protocol argument Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 028/108] AHCI: Fix softreset failed issue of Port Multiplier Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 044/108] dm thin metadata: fix bug when taking a metadata snapshot Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 011/108] bluetooth: Validate socket address length in sco_sock_bind(). Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 012/108] af_unix: Revert 'lock_interruptible' in stream receive code Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:00 +0100
  [PATCH 3.13.y-ckt 005/108] sctp: use the same clock as if sock source timestamps were on Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:10 +0100
  [PATCH 3.13.y-ckt 003/108] gre6: allow to update all parameters via rtnl Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:10 +0100
  [PATCH 3.13.y-ckt 004/108] atl1c: Improve driver not to do order 4 GFP_ATOMIC allocation Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:10 +0100
  [PATCH 3.13.y-ckt 001/108] ARC: Fix silly typo in MAINTAINERS file Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:10 +0100
  [PATCH 3.13.y-ckt 006/108] sctp: update the netstamp_needed counter when copying sockets Kamal Mostafa <kamal@canonical.com> - 2016-01-23 01:10 +0100

csiph-web