Page allocation failure on Pogoplug E02

Hello,

I am running Arch Linux on a Pogoplug E02 with 2 pendrives (one for root, one for swap & /var) and one 3TB HDD connected via USB. The pogoplug is used as a NAS and as a Bittorrent downloader (transmission, 5-10 ongoing torrents at a time). From time to time I see below messages in the log. Any ideas what could be the reason? I have a 682mb swap file but set vm.swappines to 1 to avoid using it if possible to prevent failure of the pendrive due to frequent writes (otoh I put /var and swap intentionally on a separate drive which I can simply replace if it fails). I am running a few services besides transmission (nginx, collectd, samba etc).

[477090.803782] swapper: page allocation failure: order:0, mode:0x20
[477090.803795] CPU: 0 PID: 0 Comm: swapper Tainted: G O 3.11.6-1-ARCH #1
[477090.803816] [<c000e124>] (unwind_backtrace+0x0/0xf0) from [<c000ba4c>] (show_stack+0x10/0x14)
[477090.803834] [<c000ba4c>] (show_stack+0x10/0x14) from [<c00a4b00>] (warn_alloc_failed+0xd4/0x118)
[477090.803851] [<c00a4b00>] (warn_alloc_failed+0xd4/0x118) from [<c00a7808>] (__alloc_pages_nodemask+0x608/0x828)
[477090.803868] [<c00a7808>] (__alloc_pages_nodemask+0x608/0x828) from [<c0429780>] (__netdev_alloc_frag+0x98/0x134)
[477090.803883] [<c0429780>] (__netdev_alloc_frag+0x98/0x134) from [<c042ad7c>] (__netdev_alloc_skb+0x70/0xd0)
[477090.803898] [<c042ad7c>] (__netdev_alloc_skb+0x70/0xd0) from [<c039202c>] (rxq_refill+0xf4/0x194)
[477090.803912] [<c039202c>] (rxq_refill+0xf4/0x194) from [<c03923c4>] (mv643xx_eth_poll+0xd0/0x6e0)
[477090.803926] [<c03923c4>] (mv643xx_eth_poll+0xd0/0x6e0) from [<c04357a0>] (net_rx_action+0x11c/0x27c)
[477090.803942] [<c04357a0>] (net_rx_action+0x11c/0x27c) from [<c001d204>] (__do_softirq+0xd4/0x288)
[477090.803957] [<c001d204>] (__do_softirq+0xd4/0x288) from [<c001d470>] (do_softirq+0x5c/0x64)
[477090.803970] [<c001d470>] (do_softirq+0x5c/0x64) from [<c001d6f0>] (irq_exit+0x90/0xcc)
[477090.803984] [<c001d6f0>] (irq_exit+0x90/0xcc) from [<c0009b64>] (handle_IRQ+0x34/0x84)
[477090.804000] [<c0009b64>] (handle_IRQ+0x34/0x84) from [<c04d50f4>] (__irq_svc+0x34/0x98)
[477090.804019] [<c04d50f4>] (__irq_svc+0x34/0x98) from [<c03e5b1c>] (cpuidle_enter_state+0x60/0xf4)
[477090.804037] [<c03e5b1c>] (cpuidle_enter_state+0x60/0xf4) from [<c03e5c54>] (cpuidle_idle_call+0xa4/0x23c)
[477090.804052] [<c03e5c54>] (cpuidle_idle_call+0xa4/0x23c) from [<c0009cdc>] (arch_cpu_idle+0x8/0x44)
[477090.804068] [<c0009cdc>] (arch_cpu_idle+0x8/0x44) from [<c004d934>] (cpu_startup_entry+0x64/0x220)
[477090.804083] [<c004d934>] (cpu_startup_entry+0x64/0x220) from [<c068e9e8>] (start_kernel+0x2cc/0x324)
[477090.804090] Mem-info:
[477090.804095] Normal per-cpu:
[477090.804103] CPU 0: hi: 90, btch: 15 usd: 41
[477090.804119] active_anon:10697 inactive_anon:11674 isolated_anon:0
active_file:12742 inactive_file:14979 isolated_file:0
unevictable:0 dirty:6635 writeback:1 unstable:0
free:188 slab_reclaimable:5106 slab_unreclaimable:3055
mapped:1449 shmem:6163 pagetables:395 bounce:0
free_cma:0
[477090.804156] Normal free:752kB min:2004kB low:2504kB high:3004kB active_anon:42788kB inactive_anon:46696kB active_file:50968kB inactive_file:59916kB unevictable:0kB isolated(anon):0kB isolated(file):0kB present:262144kB managed:251988kB mlocked:0kB dirty:26540kB writeback:4kB mapped:5796kB shmem:24652kB slab_reclaimable:20424kB slab_unreclaimable:12220kB kernel_stack:824kB pagetables:1580kB unstable:0kB bounce:0kB free_cma:0kB writeback_tmp:0kB pages_scanned:1114 all_unreclaimable? no
[477090.804166] lowmem_reserve[]: 0 0
[477090.804175] Normal: 174*4kB (UEM) 3*8kB (E) 2*16kB (E) 0*32kB 0*64kB 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 752kB
[477090.804210] 35390 total pagecache pages
[477090.804217] 1506 pages in swap cache
[477090.804224] Swap cache stats: add 69438, delete 67932, find 1399311/1402897
[477090.804230] Free swap = 668204kB
[477090.804235] Total swap = 682724kB

I guess setting swappiness to 1 was a quite bad idea. Now you seem to run out of RAM, it seems and the swapping doesn't kick in fast enough. You could try the solutions suggested here:
http://serverfault.com/questions/236170/page-allocation-failure-am-i-running-out-of-memory
Just the result of a little quick googling.

See this also:
http://forum.doozan.com/read.php?2,12096,12450#msg-12450

-bodhi
===========================
Forum Wiki
bodhi's corner (buy bodhi a beer)

Interesting!
When setting up my Goflex some days ago I noticed some strange behaviour, too.
You could try setting the priority of the swap, explicitly, to something like 1 or 10. Per default I think it gets assigned a priority of -1.
When experimenting with swap I found that the swap with no explicit priority setting just didn't get utilized. Set the swap priority to some explicit number and it worked.
Could be worth a try and sounds like some kind of a bug.
BTW, kernel 3.12 on my tests, too.



Edited 1 time(s). Last edit at 11/20/2013 06:21PM by ingmar_k.