Koozali.org: home of the SME Server

Obsolete Releases => SME Server 7.x => Topic started by: dadou on April 22, 2012, 08:07:43 PM

Title: kernel: Eeek! page_mapcount(page) went negative!
Post by: dadou on April 22, 2012, 08:07:43 PM
Hello,

Last night I got a bad crash on our system.

After log-rotate in crontab it just crashed :


Quote
Apr 22 01:12:01 fmcp esmith::event[7374]: S75purge-old-logs=action|Event|logrotate|Action|S75purge-old-logs|Start|1335049921 555497|End|1335049921 640327|Elapsed|0.08483
Apr 22 01:12:01 fmcp esmith::event[7374]: Running event handler: /etc/e-smith/events/actions/adjust-services
Apr 22 01:12:01 fmcp esmith::event[7374]: adjusting supervised httpd-e-smith (sigusr1)
Apr 22 01:12:01 fmcp esmith::event[7374]: adjusting supervised httpd-e-smith (up)
Apr 22 01:12:01 fmcp esmith::event[7374]: adjusting non-supervised syslog (reload)
Apr 22 01:12:01 fmcp kernel: Eeek! page_mapcount(page) went negative! (-1)
Apr 22 01:12:01 fmcp kernel:   page->flags = 4002016c
Apr 22 01:12:01 fmcp kernel:   page->count = 1
Apr 22 01:12:01 fmcp kernel:   page->mapping = d6c8dcf0
Apr 22 01:12:01 fmcp kernel: ------------[ cut here ]------------
Apr 22 01:12:01 fmcp kernel: kernel BUG at mm/rmap.c:508!
Apr 22 01:12:01 fmcp kernel: invalid operand: 0000 [#1]
Apr 22 01:12:01 fmcp kernel: SMP
Apr 22 01:12:01 fmcp kernel: Modules linked in: appletalk(U) r8169 bonding(U) ipt_ULOG ipt_REJECT ipt_MASQUERADE ipt_state ip_nat_ftp ip_conntrack_ftp iptable_mangle iptable_nat ip_conntrack iptable_filter ip_tables loop button battery ac hw_random mii sr_mod dm_snapshot dm_zero dm_mirror ext3 jbd raid1 dm_mod usb_storage uhci_hcd ohci_hcd ehci_hcd ata_piix libata 3w_9xxx sd_mod scsi_mod
Apr 22 01:12:01 fmcp kernel: CPU:    0
Apr 22 01:12:01 fmcp kernel: EIP:    0060:[page_remove_rmap+111/258]    Not tainted VLI
Apr 22 01:12:01 fmcp kernel: EIP:    0060:[<c0155197>]    Not tainted VLI
Apr 22 01:12:01 fmcp kernel: EFLAGS: 00010246   (2.6.9-89.31.1.ELsmp)
Apr 22 01:12:01 fmcp kernel: EIP is at page_remove_rmap+0x6f/0x102
Apr 22 01:12:01 fmcp kernel: eax: 0000001e   ebx: c95fa860   ecx: f3a7eeac   edx: c02f2938
Apr 22 01:12:01 fmcp kernel: esi: 00000001   edi: e0c85000   ebp: fffcf8d8   esp: f3a7eea8
Apr 22 01:12:01 fmcp kernel: ds: 007b   es: 007b   ss: 0068
Apr 22 01:12:01 fmcp kernel: Process httpd (pid: 2154, threadinfo=f3a7e000 task=f6ac4d70)
Apr 22 01:12:01 fmcp kernel: Stack: c02f2938 d6c8dcf0 c02f2921 00000001 085fa860 00000001 c014eb8e 2fd43425
Apr 22 01:12:01 fmcp kernel:        00000004 c95fa860 0000a000 00111000 c969b180 c0120f01 f3a7eee0 00111000
Apr 22 01:12:01 fmcp kernel:        00191000 e0c85008 c969b180 c014ecb0 00080000 00000000 00111000 f6bda548

I'm on contribs 7.5.1 and it seems to be the first time it happens.
When I google the problems it seems to be a material RAM problem. What do you think ?

Thanks by advance,

David
Title: Re: kernel: Eeek! page_mapcount(page) went negative!
Post by: cactus on April 22, 2012, 09:43:12 PM
Quote from: dadou on April 22, 2012, 08:07:43 PM
When I google the problems it seems to be a material RAM problem. What do you think ?
Most of the times Google is very accurate on this. I suggest you run a memory test like memtest86+, to confirm this.
Title: Re: kernel: Eeek! page_mapcount(page) went negative!
Post by: dadou on May 09, 2012, 09:30:22 AM
Hello Thanks for the answer,

I'll test that.

I also see something I never took care of. At startup system give this message :


kernel: ********************************************************
kernel: * This system has more than 16 Gigabyte of memory.     *
kernel: * It is recommended that you read the release notes    *
kernel: * that accompany your copy of CentOS distribution      *
kernel: * about the recommended kernel for such configurations *
kernel: ********************************************************

I have 4x4096Mo Kingston DDR3 (so it's 16Go ?)

It seems that kernel has to be updated for better configuration about this. Is it necessary ?
If yes how do you suggest to update it ?

It seems to need kernel-hugemem does someone already installed it ?

Thanks by advance,

David
Title: Re: kernel: Eeek! page_mapcount(page) went negative!
Post by: CharlieBrady on May 09, 2012, 02:35:19 PM
You don't need kernel-hugemem. I'm pretty sure that you don't need 16GB of RAM either.

You do need to test your memory.
Title: Re: kernel: Eeek! page_mapcount(page) went negative!
Post by: jameswilson on May 10, 2012, 09:35:46 AM
Re 16 gig of ram
I was planning a powerful sme and use virtual box. Would I be better with a modest sme and a seperate xen box for the vm's

I know performance would be better on xen but I love the stability and ease of use of sme

James
Title: Re: kernel: Eeek! page_mapcount(page) went negative!
Post by: Stefano on May 10, 2012, 09:38:39 AM
Quote from: jameswilson on May 10, 2012, 09:35:46 AM
Re 16 gig of ram
I was planning a powerful sme and use virtual box. Would I be better with a modest sme and a seperate xen box for the vm's

I know performance would be better on xen but I love the stability and ease of use of sme

James

on such an hw, I wolud install proxmox ve and virtualize all my servers (SME too)
Title: Re: kernel: Eeek! page_mapcount(page) went negative!
Post by: jameswilson on May 10, 2012, 01:36:57 PM
Ok ta
Title: Re: kernel: Eeek! page_mapcount(page) went negative!
Post by: CharlieBrady on May 10, 2012, 03:07:46 PM
Quote from: jameswilson on May 10, 2012, 09:35:46 AM
I was planning a powerful sme and use virtual box.

If you proceed with that, you should use SME8.