- 1. 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Thu, 21 Apr 2011 18:57:16 -0700 (PDT)
- Hi, after the block layer regression[0] seemed to be fixed, the machine appeared to be running fine. But after putting some disk I/O to the system (PowerBook G4) it became unresponsive, I/O wait went
- /archives/xfs/2011-04/msg00315.html (8,563 bytes)
- 2. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Minchan Kim <minchan.kim@xxxxxxxxx>
- Date: Fri, 22 Apr 2011 11:58:34 +0900
- You would try to allocate a page from DMA as you don't have a normal zone. Although free pages in DMA zone is about 3M, free pages of zone is below min of DMA zone. So zone_watermark_ok would be fail
- /archives/xfs/2011-04/msg00316.html (11,360 bytes)
- 3. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Thu, 21 Apr 2011 20:03:27 -0700 (PDT)
- No, I forgot to add 2.6.38 (and before) did not show that behaviour. I would think so. I'll try to bisect, but with one compilation taking ~20min on this box, this might take a while. When it happens
- /archives/xfs/2011-04/msg00317.html (8,404 bytes)
- 4. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Fri, 22 Apr 2011 10:41:26 -0700 (PDT)
- It happened again, and again it's du(1) invoking the OOM killer (I'm running du(1) accross a few directories every 6hrs). This time I got /proc/slabinfo as well: http://nerdbynature.de/bits/2.6.39-rc
- /archives/xfs/2011-04/msg00326.html (9,714 bytes)
- 5. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Fri, 22 Apr 2011 11:46:57 -0700 (PDT)
- Btw, xfs* showed up again in the first backtrace. This might be just because du(1) came across an XFS filesystem (~1TB) or it's something else: [20546.774480] du invoked oom-killer: gfp_mask=0x842d0,
- /archives/xfs/2011-04/msg00327.html (12,545 bytes)
- 6. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Minchan Kim <minchan.kim@xxxxxxxxx>
- Date: Sat, 23 Apr 2011 07:47:10 +0900
- xfs_inode slab is using about 200M. It seems rather huge. Let's Cced XFS guys. -- Kind regards, Minchan Kim
- /archives/xfs/2011-04/msg00329.html (14,800 bytes)
- 7. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Dave Chinner <david@xxxxxxxxxxxxx>
- Date: Mon, 25 Apr 2011 09:46:55 +1000
- You memory is full of xfs inodes, and it doesn't appear that memory reclaim has kicked in at all to free any - the numbers just keep growing at 1-2000 inodes/s. I'd say they are not being reclaimmmed
- /archives/xfs/2011-04/msg00338.html (10,419 bytes)
- 8. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Sun, 24 Apr 2011 22:51:17 -0700 (PDT)
- Will do that. /dev/mapper/wdc1 /mnt/backup xfs rw,nosuid,nodev,noexec,relatime,attr2,delaylog,noquota 0 0 /dev/mapper/wdc1 /data xfs ro,relatime,attr2,delaylog,noquota 0 0 The "/dev/mapper/wdc1" is a
- /archives/xfs/2011-04/msg00340.html (9,801 bytes)
- 9. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Mon, 25 Apr 2011 00:19:24 -0700 (PDT)
- Please see http://nerdbynature.de/bits/2.6.39-rc4/oom/ - slabinfo-4.txt.bz2, contains /proc/sys/fs/{dentry,inode}-state and /proc/fs/xfs/stat (is this interesting at all?) and a few other /proc outpu
- /archives/xfs/2011-04/msg00341.html (9,774 bytes)
- 10. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Mon, 25 Apr 2011 01:02:33 -0700 (PDT)
- When mounted with noatime, running du(1) accross the XFS volume still triggered the OOM killer, but the backtrace looked somewhat different this time: date invoked oom-killer: gfp_mask=0x84d0, order=
- /archives/xfs/2011-04/msg00342.html (9,978 bytes)
- 11. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Mon, 25 Apr 2011 02:50:12 -0700 (PDT)
- http://nerdbynature.de/bits/2.6.39-rc4/oom/oom-6.JPG http://nerdbynature.de/bits/2.6.39-rc4/oom/slabinfo-6.txt.bz2 Christian. -- BOFH excuse #272: Netscape has crashed
- /archives/xfs/2011-04/msg00343.html (9,786 bytes)
- 12. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Tue, 26 Apr 2011 08:14:46 -0700 (PDT)
- Dave, did you have time to look at those outputs yet? Do you need some more information? Right now 2.6.39-rc4 is not an option for me, 2.6.38 works just fine.... Thanks, Christian. -- BOFH excuse #17
- /archives/xfs/2011-04/msg00346.html (9,757 bytes)
- 13. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Dave Chinner <david@xxxxxxxxxxxxx>
- Date: Wed, 27 Apr 2011 12:26:55 +1000
- Ok, so looking at slabinfo-5.txt.bz2, this is the pattern of dentry, VFS inode and XFS inode use patterns: http://userweb.kernel.org/~dgc/slab-usage.png What this shows is that VFS inode cache memory
- /archives/xfs/2011-04/msg00358.html (10,747 bytes)
- 14. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Wed, 27 Apr 2011 00:46:51 -0700 (PDT)
- Do you remember any XFS changes past 2.6.38 that could be related to something like this? Bisecting is pretty slow on this machine. Could I somehow try to run 2.6.39-rc4 but w/o the XFS changes merge
- /archives/xfs/2011-04/msg00362.html (10,714 bytes)
- 15. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Dave Chinner <david@xxxxxxxxxxxxx>
- Date: Wed, 27 Apr 2011 20:28:24 +1000
- There's plenty of changes that coul dbe the cause - we've changed the inode reclaim to run in the background out of a workqueue as well as via the shrinker, so it could even be workqueue starvation c
- /archives/xfs/2011-04/msg00365.html (11,283 bytes)
- 16. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Minchan Kim <minchan.kim@xxxxxxxxx>
- Date: Thu, 28 Apr 2011 08:16:29 +0900
- RCU free starvation is another possibility? https://lkml.org/lkml/2011/4/25/124 -- Kind regards, Minchan Kim
- /archives/xfs/2011-04/msg00373.html (11,964 bytes)
- 17. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Dave Chinner <david@xxxxxxxxxxxxx>
- Date: Thu, 28 Apr 2011 09:56:58 +1000
- You know, I've been waching that thread with interest, but it didn't seem to be related. However, now that I go look at the config file provided, I see: CONFIG_TINY_RCU=y CONFIG_PREEMPT_NONE=y which
- /archives/xfs/2011-04/msg00374.html (11,936 bytes)
- 18. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Thu, 28 Apr 2011 10:30:47 -0700 (PDT)
- Hi David, sorry for the late response: No, I did not change anything underneath /proc/sys/fs/xfs. Yeah, I noticed. I wanted to do something like this: $ for c in `git log --pretty=oneline --no-merges
- /archives/xfs/2011-04/msg00391.html (11,922 bytes)
- 19. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Dave Chinner <david@xxxxxxxxxxxxx>
- Date: Fri, 29 Apr 2011 09:37:51 +1000
- Did you read the email that pointed to CONFIG_TINY_RCU as the ptotential source of the problem? Can you change to CONFIG_TREE_RCU and see if the problem goes away? Cheers, Dave. -- Dave Chinner david
- /archives/xfs/2011-04/msg00394.html (11,574 bytes)
- 20. Re: 2.6.39-rc4+: oom-killer busy killing tasks (score: 1)
- Author: Christian Kujau <lists@xxxxxxxxxxxxxxx>
- Date: Fri, 29 Apr 2011 10:32:16 -0700 (PDT)
- For that I had to select SMP on this UP-PowerPC G4 machine. But compiling -rc5 (as the bug is strill present in -rc5) fails with: /usr/local/src/linux-2.6-git/arch/powerpc/kernel/crash.c: In function
- /archives/xfs/2011-04/msg00412.html (12,019 bytes)
This search system is powered by
Namazu