linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Minchan Kim <minchan@kernel.org>
To: Luigi Semenzato <semenzato@google.com>
Cc: linux-mm@kvack.org
Subject: Re: another allocation livelock with zram
Date: Wed, 21 Nov 2012 10:27:26 +0900	[thread overview]
Message-ID: <20121121012726.GA5121@bbox> (raw)
In-Reply-To: <CAA25o9T8cBhuFnesnxHDsv3PmV8tiHKoLz0dGQeUSCvtpBBv3A@mail.gmail.com>

Hi Luigi,

Question.
Is it a 3.4.0 vanilla kernel?
Otherwise, some hacky patches(ex, min_filelist_kbytes) are applied?

On Tue, Nov 20, 2012 at 03:46:34PM -0800, Luigi Semenzato wrote:
> Greetings MM folks,
> 
> and thanks again for fixing my previous hang-with-zram problem.  I am
> now running into a similar problem and I hope I will not take
> advantage of your kindness by asking for further advice.
> 
> By running a few dozen memory-hungry processes on an ARM cpu with 2 Gb
> RAM, with zram enabled, I can easily get into a situation where all
> processes are either:
> 
> 1. blocked in a futex
> 2. trying unsuccessfully to allocate memory
> 
> This happens when there should still be plenty of memory: the zram
> swap device is about 1/3 full. (The output of SysRq-M is at the end.)
> Yet the SI and SO fields of vmstat stay at 0, and CPU utilization is
> 100% system.
> 
> procs -----------memory---------- ---swap-- -----io---- -system-- ----cpu----
>  r  b   swpd   free   buff  cache   si   so    bi    bo   in   cs us sy id wa
> 46  0 1076432  13636   2844 216648    0    0     0     0  621  229  0 100  0  0
> 44  0 1076432  13636   2844 216648    0    0     0     0  618  204  0 100  0  0
> 
> I added counters in various places in the page allocator to see which
> paths were being taken and noticed the following facts:
> 
> - alloc_page_slowpath is looping, apparently trying to rebalance.  It
> calls alloc_pages_direct_reclaim at a rate of about 155 times/second,
> and gets one page about once every 500 calls.  Did_some_progress is
> always set to true.  Then should_alloc_retry returns true (because
> order < PAGE_ALLOC_COSTLY_ORDER).
> 
> - kswapd is asleep and is not woken up because alloc_page_slowpath
> never goes to the "restart" label.
> 
> My questions are:
> 
> 1. is it obvious to any of you what is going wrong?
> 1.1 is the allocation failing because nobody is waking up kswapd?  And
> if so, why not?
> 
> 2. if it's not obvious, what are the next things to look into?
> 
> 3. is there a better way of debugging this?
> 
> Thanks!
> Luigi
> 
> [    0.000000] Linux version 3.4.0
> (semenzato@luigi.mtv.corp.google.com) (gcc version 4.6.x-google
> 20120301 (prerelease) (gcc-4.6.3_cos_gg_2a32ae6) ) #26 SMP Tue Nov 20
> 14:27:15 PST 2012
> [    0.000000] CPU: ARMv7 Processor [410fc0f4] revision 4 (ARMv7), cr=10c5387d
> [    0.000000] CPU: PIPT / VIPT nonaliasing data cache, PIPT instruction cache
> [    0.000000] Machine: SAMSUNG EXYNOS5 (Flattened Device Tree),
> model: Google Snow
> ...
> [  198.564328] SysRq : Show Memory
> [  198.564347] Mem-info:
> [  198.564355] Normal per-cpu:
> [  198.564364] CPU    0: hi:  186, btch:  31 usd:   0
> [  198.564373] CPU    1: hi:  186, btch:  31 usd:   0
> [  198.564381] HighMem per-cpu:
> [  198.564389] CPU    0: hi:   90, btch:  15 usd:   0
> [  198.564397] CPU    1: hi:   90, btch:  15 usd:   0
> [  198.564411] active_anon:196868 inactive_anon:66835 isolated_anon:47
> [  198.564415]  active_file:13931 inactive_file:11043 isolated_file:0
> [  198.564419]  unevictable:0 dirty:4 writeback:1 unstable:0
> [  198.564423]  free:3409 slab_reclaimable:2583 slab_unreclaimable:3337
> [  198.564427]  mapped:137910 shmem:29899 pagetables:3972 bounce:0
> [  198.564449] Normal free:13384kB min:5380kB low:6724kB high:8068kB
> active_anon:782052kB inactive_anon:261808kB active_file:25020kB
> inactive_file:24900kB unevictable:0kB isolated(anon):16kB
> isolated(file):0kB present:1811520kB mlocked:0kB dirty:12kB
> writeback:0kB mapped:461296kB shmem:115892kB slab_reclaimable:10332kB
> slab_unreclaimable:13348kB kernel_stack:3008kB pagetables:15888kB
> unstable:0kB bounce:0kB writeback_tmp:0kB pages_scanned:107282320
> all_unreclaimable? no
> [  198.564474] lowmem_reserve[]: 0 2095 2095
> [  198.564499] HighMem free:252kB min:260kB low:456kB high:656kB
> active_anon:5420kB inactive_anon:5532kB active_file:30704kB
> inactive_file:19272kB unevictable:0kB isolated(anon):172kB
> isolated(file):0kB present:268224kB mlocked:0kB dirty:4kB
> writeback:4kB mapped:90344kB shmem:3704kB slab_reclaimable:0kB
> slab_unreclaimable:0kB kernel_stack:0kB pagetables:0kB unstable:0kB
> bounce:0kB writeback_tmp:0kB pages_scanned:7081406 all_unreclaimable?
> no
> [  198.564523] lowmem_reserve[]: 0 0 0
> [  198.564536] Normal: 1570*4kB 6*8kB 1*16kB 0*32kB 0*64kB 1*128kB
> 1*256kB 1*512kB 0*1024kB 1*2048kB 1*4096kB = 13384kB
> [  198.564574] HighMem: 59*4kB 2*8kB 0*16kB 0*32kB 0*64kB 0*128kB
> 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 252kB
> [  198.564610] 123112 total pagecache pages
> [  198.564616] 68239 pages in swap cache
> [  198.564622] Swap cache stats: add 466115, delete 397876, find 31817/56350
> [  198.564630] Free swap  = 1952336kB
> [  198.564635] Total swap = 3028768kB
> [  198.564640] xxcount_nr_reclaimed 358488
> [  198.564646] xxcount_nr_reclaims 6201
> [  198.564651] xxcount_aborted_reclaim 0
> [  198.564656] xxcount_more_to_do 5137
> [  198.564662] xxcount_direct_reclaims 17065
> [  198.564667] xxcount_failed_direct_reclaims 10708
> [  198.564673] xxcount_no_progress 5696
> [  198.564678] xxcount_restarts 5696
> [  198.564683] xxcount_should_alloc_retry 5008
> [  198.564688] xxcount_direct_compact 1
> [  198.564693] xxcount_alloc_failed 115
> [  198.564699] xxcount_gfp_nofail 0
> [  198.564704] xxcount_costly_order 5009
> [  198.564709] xxcount_repeat 0
> [  198.564714] xxcount_kswapd_nap 2210
> [  198.564719] xxcount_kswapd_sleep 17
> [  198.564724] xxcount_kswapd_loop 2211
> [  198.564729] xxcount_kswapd_try_to_sleep 2210
> [  198.575349] 524288 pages of RAM
> [  198.575358] 4420 free pages
> [  198.575365] 7122 reserved pages
> [  198.575371] 4091 slab pages
> [  198.575378] 302549 pages shared
> [  198.575384] 68239 pages swap cached
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@kvack.org.  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

-- 
Kind regards,
Minchan Kim

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  reply	other threads:[~2012-11-21  1:27 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2012-11-20 23:46 another allocation livelock with zram Luigi Semenzato
2012-11-21  1:27 ` Minchan Kim [this message]
2012-11-21  1:47   ` Luigi Semenzato
2012-11-21 13:59     ` Minchan Kim
2012-11-21 18:21       ` Luigi Semenzato
2012-11-23  5:44         ` Minchan Kim
2012-11-27  1:00           ` Luigi Semenzato
2012-11-28  6:12             ` Minchan Kim
2013-02-17  3:33 ` Jaegeuk Hanse
2013-02-17  4:55   ` Luigi Semenzato

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20121121012726.GA5121@bbox \
    --to=minchan@kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=semenzato@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).