Syslinux - Dec 2003 - Asus A7V600 (3C940 1gb) + Oopsing kernel with PXELINUX

Hi!

We (an astronomical center) are building a small cluster for our 
number-crunching. The problem is that we have 12 mobos A7V600 by Asus (for 
athlons) - 1 server + 11 diskless nodes. And here comes the trouble...

hardware:
mobo: a7v600
net: 1 gigabit by 3com - 3c940, integrated in mobo

the problem is:

dhcp - works fine
tftp - works fine
pxelinux.0 - downloads fine (I hope)
linux kernel - downloads and...

and here begins the lottery. sometimes:
Uncompressing vmlinux....... crc error, system halted

sometimes the node uncompresses the kernel and advances in startup, but 
the kernel makes Oops and produces the stack info during services startup. 
mostly non-recoverable.

but sometimes everything seemes to be fine - even the login prompt ;-)
but the system is corrupted - kernel oopses after some time. for sure will 
oops when shutting the system down.

it seems as some part of bios/mba randomly overwrites some part of the 
kernel or sth alike. this is only my guess.

the same kernel works fine when loaded from a "hard" media.

I have tried various versions of pxelinux: 2.05, 2.07, 1.76. prhaps I have 
missed sth? kernels 2.6.0-test11 and 2.4.23 as well (but it is not a 
kernel issue).

at the moment I can not remember the version of MBA (I am at home now) - 
will update tomorrow.

I however have no idea how to solve this.
1. is it a common behaviour with some hardware configurations? I have 
digged the archives and google but found nothing of that type...
2. is there a chance to use pxelinux anyway? I am not sure if the mobo 
provides an option to update the mba...
3. does anyone use asus a7v600 with pxelinux???
4. any alternative booting way? those diskless nodes do not even have 
floppies... usb pen-drive?

what are the alternatives to pxelinux (bootp?) that could work?

thanx a lot - I am really in need for help...

regards
michal frackowiak

Michal Frackowiak wrote:
> Hi!
> 
> We (an astronomical center) are building a small cluster for our 
> number-crunching. The problem is that we have 12 mobos A7V600 by Asus (for 
> athlons) - 1 server + 11 diskless nodes. And here comes the trouble...
> 
> hardware:
> mobo: a7v600
> net: 1 gigabit by 3com - 3c940, integrated in mobo
> 
> the problem is:
> 
> dhcp - works fine
> tftp - works fine
> pxelinux.0 - downloads fine (I hope)
> linux kernel - downloads and...
> 
> and here begins the lottery. sometimes:
> Uncompressing vmlinux....... crc error, system halted
> 
> sometimes the node uncompresses the kernel and advances in startup, but 
> the kernel makes Oops and produces the stack info during services startup. 
> mostly non-recoverable.
> 
> but sometimes everything seemes to be fine - even the login prompt ;-)
> but the system is corrupted - kernel oopses after some time. for sure will 
> oops when shutting the system down.
> 
> it seems as some part of bios/mba randomly overwrites some part of the 
> kernel or sth alike. this is only my guess.
> 
Once the kernel is started, there should be no BIOS or MBA still running.

The 3Com MBA is definitely one of the most problematic PXE stacks out there.  I
would try to find
the latest version available for your mobo and see if that helps.

	-hpa

Try just attaching a cd drive and running some live cd distro like
knoppix.

Apart from that, are you sure about your memory? Try running memtest86,
you can start it directly as a kernel from syslinux.

I've had some weird memory problems myself, that mostly went away by
downclocking my memory to 266MHz.

Good luck!

Wout.

Yesterday at 23:55 evil squirrels forced Michal Frackowiak to write:
> Hi!
>
> We (an astronomical center) are building a small cluster for our
> number-crunching. The problem is that we have 12 mobos A7V600 by Asus (for
> athlons) - 1 server + 11 diskless nodes. And here comes the trouble...
>
> hardware:
> mobo: a7v600
> net: 1 gigabit by 3com - 3c940, integrated in mobo
>
> the problem is:
>
> dhcp - works fine
> tftp - works fine
> pxelinux.0 - downloads fine (I hope)
> linux kernel - downloads and...
>
> and here begins the lottery. sometimes:
> Uncompressing vmlinux....... crc error, system halted
>
> sometimes the node uncompresses the kernel and advances in startup, but
> the kernel makes Oops and produces the stack info during services startup.
> mostly non-recoverable.
>
> but sometimes everything seemes to be fine - even the login prompt ;-)
> but the system is corrupted - kernel oopses after some time. for sure will
> oops when shutting the system down.
>
> it seems as some part of bios/mba randomly overwrites some part of the
> kernel or sth alike. this is only my guess.
>
> the same kernel works fine when loaded from a "hard" media.
>
> I have tried various versions of pxelinux: 2.05, 2.07, 1.76. prhaps I have
> missed sth? kernels 2.6.0-test11 and 2.4.23 as well (but it is not a
> kernel issue).
>
> at the moment I can not remember the version of MBA (I am at home now) -
> will update tomorrow.
>
> I however have no idea how to solve this.
> 1. is it a common behaviour with some hardware configurations? I have
> digged the archives and google but found nothing of that type...
> 2. is there a chance to use pxelinux anyway? I am not sure if the mobo
> provides an option to update the mba...
> 3. does anyone use asus a7v600 with pxelinux???
> 4. any alternative booting way? those diskless nodes do not even have
> floppies... usb pen-drive?
>
> what are the alternatives to pxelinux (bootp?) that could work?
>
> thanx a lot - I am really in need for help...
>
> regards
> michal frackowiak
>
> _______________________________________________
> SYSLINUX mailing list
> Submissions to SYSLINUX at zytor.com
> Unsubscribe or set options at:
> http://www.zytor.com/mailman/listinfo/syslinux
> Please do not send private replies to mailing list traffic.
>