2.4.3 Kernel Freeze with highmem BUG at highmem.c:155 - CRASH

2.4.3 Kernel Freeze with highmem BUG at highmem.c:155 - CRASH

[Marcin Kowalski]

Hi All

After two weeks of quasi stability the server has crashed.. again... with the 
following message... (partial)
------------------
kernel BUG at highmem.c:155

Invalid Operand : 0000
CPU : 1
EIP : 0010:[<c012fcb>]
EFLAGS: 00010286
eax:0000001d ebx: 00000000 esi:c2147ec0
edi: 00000000 ebp: 00001000 esp: df587ebc

Process sshd (pid : 6230, stackpage : df587000)
......
-----------------
This is a standard 2.4.3 kernel with four patches, to correct the 
dcache inode non-clearance (ever growing inode and directory cache) as well 
as the patch to apply vm pressure to lower these caches (Ed Tomlinsons, I 
beleive...).., next patches are for aacraid.. including jens axboes' SCSI 
patch as well as the aacraid patch.

The symptons were an ever more sluggish machine over time, memory usage 
looked pretty standard with the majority of memory assigned to cache... what 
would happen is that at terminal it would go into semi-freeze states of about 
5-10 seconds (increasing with time), where no user interaction was possible. 
By terminal I mean through a ssh remote terminal.... The load would also 
occasionally just increase for no apparent reason to values of 7,8,9...

The server is a dual 933mhz Xeon (PIII) on a ServerWorks Motherboard (HP 
NetServer) with 1.2 gig or ram, 6 SCSI III 18.6 GIG drives in HP Netraid and 
1 9 gig SCSI as the root FS... Running suse 7.0 and reiserfs...

ANy and ALL advice/patches would be greatly appreciated.... I will probably 
be moving back to 2.2 kernel series as a result of my stability problems..

Thanks 
MarCin

-- 
-----------------------------
     Marcin Kowalski
     Linux/Perl Developer
-----------------------------

Re: 2.4.3 Kernel Freeze with highmem BUG at highmem.c:155 - CRASH

[Rico Tudor]

> The symptons were an ever more sluggish machine over time, memory usage 
> looked pretty standard with the majority of memory assigned to cache... what 
> would happen is that at terminal it would go into semi-freeze states of about 
> 5-10 seconds (increasing with time), where no user interaction was possible. 
> By terminal I mean through a ssh remote terminal.... The load would also 
> occasionally just increase for no apparent reason to values of 7,8,9...

I reported the same sluggishness problem on Feb 25.  Capsule summary
is 64GB option does not work.  I was easily able to reproduce the
sluggishness in 2.4.2, but need to test again for 2.4.4.  See if your
problem goes away with the 4GB option.

Re: 2.4.3 Kernel Freeze with highmem BUG at highmem.c:155 - CRASH

[Marcin Kowalski]

> I reported the same sluggishness problem on Feb 25.  Capsule summary
> is 64GB option does not work.  I was easily able to reproduce the
> sluggishness in 2.4.2, but need to test again for 2.4.4.  See if your
> problem goes away with the 4GB option.

Hi

Thanks for the advice, I am however using the 4GB setting already. I find it 
difficult to reproduce the results as the freeze happens at various times 
when I am wokring on the console. I do however suspect that it is related to 
AACRAID array write activity....

Regards
MarCin

-- 
-----------------------------
     Marcin Kowalski
     Linux/Perl Developer
     Datrix Solutions
     Cel. 082-400-7603
      ***Open Source Kicks Ass***
-----------------------------