Re: [coLinux-devel] initrd crashed with mem=128

[Henry N. <Henry.Ne@Arcor.de>]

Hello George,

George P Boutwell wrote:
> On 5/17/06, Henry Nestler wrote:
>> Command line (basics):
>> colinux-daemon.exe kernel=vmlinux initrd=initrd.gz cobd0=%IMG% cofs0=d:\
>> root=/dev/cobd0 eth0=slirp eth1=tuntap mem=128
>>
>> http://home.arcor.de/henryn/colinux/screenshoots/colinux064pre2-128mb-initrd.png 
>>
>>
>>
>> All testings with 0.6.4-pre2
>> - mem=32, no initrd  --> OK, many times
>> - mem=32, initrd=initrd.gz  --> OK, many times
>> - mem=128, no initrd --> OK, many times
>> - mem=128, initrd=initrd.gz --> Reboots host
>> - mem=128, no initrd --> OK, many times
>> - mem=120, no initrd --> OK, many times
>> - mem=120, initrd=initrd.gz  --> OK, tested 2 times
>> - mem=128, initrd=initrd.gz --> Holds (see attachment and sceenshoot),
>> colinux-daemon have closed simple with cross on right corner of screen.
>> - mem=128, initrd=initrd.gz --> Reboots host (nothing changed between
>> this and the start before!)
>> - mem=128, initrd=initrd.gz --> Holds on same place. Close the console
>> - mem=120, initrd=initrd.gz --> OK, tested 3 times
>> - mem=128, initrd=initrd.gz --> 2 times OK, than...
>> - mem=128, initrd=initrd.gz --> have a Kernel oops (attached) ;-)
>>
>>
>> About the system:
>> AMD 1.9GHz, 512MB real memory, XP with SP2 (DEP changes not need),
>> Debian woody in 600MB image file.
>>
>> It crashed or holded all times on the same step.  After the line
>> "RAMDISK: Compressed image found...".  That the state, the ramdisk
>> should start.
>> This is long before the initrd try to mount some. So it is not the old
>> "sync;umount" error.
>>
> 
> We shoudl do these same tests on the 0.7.x series, including the
> current basically neglected 0.7.x and your hn series to see if this
> bug/problem exists in them.  In the mean-time, I'll try to isolate
> what has changed (hopefully combined with your back trace) between
> 0.6.3-rc4 and 0.6.3 release to see if I can identify what the problem
> code is, and get a fix going.  I might be wrong, but it seems to be a
> problem on low memory machines...  ie mine that I see the problem on
> are 384MB, yours is 512MB, I think the others who report the problem
> have similar 512MB memory...  I don't think any reports have been from
>> 512MB (eg 768, or 1g, or the like)
> 

The problem comes not with the new 0.6.4.
I have the same problem also on 0.7.1-hn* series on this PC.

It comes sometimes olso on my other PC with coLinux mem=256, real 
mem=512MB, coLinux 0.6.3 and 0.7.1.

The problem exist typical with coLinux memory 128MB and multiples of 
this size.  Only, if I used initrd.

The rootfs image is a Debian 3.0, or a Fedora core 4 partition, or a 
SuSE 9.0 partition.

It never gave such bugtrace.  Mostly it's simple reboots the host 
system, and after the next host booting, it runs mostly without problems.

The problem exist, if the host memory is enougth free.  I have not 
starts other programs before (fresh booted PC).

Machines with exact 128/256/512MB memory have some times specific 
problems inside linux kernel/mm (no coLinux specific).  A machine with 
512MB real memory is not typical in real world of linux.  For sample 
640KB low DOS memory is mostly not usable under linux.  If you have 
512MB installed in a machine, Linux can only use 512MB-640KB.   So, the 
Linux kernel was not full tested for exact 512MB and the other sizes 
(128MB,256MB,384MB).  I known, it was also an old fixed bug in older kernel.

-- 
Henry Nestler