[Soekris] Combination of EXT2 & JFFS2 (as RootFS) don't boot Soekris net4521
bill at stsx.org
Fri Sep 28 05:58:50 UTC 2007
On Fri, 2007-09-28 at 10:08 +0530, Bhagyashri Katole wrote:
> Hi ,
> >So the problem seems to lie within the kernel - after all it's the
> kernel that reboots, not Lilo (AFAIK > Lilo can't do that).
> >Kernels have bugs too.
> But , I tried same kernel image to boot my computer, it worked
> fine ...
"Bug" here doesn't necessarily mean a programming error. It could be
some deeply nested anomaly in the communication between kernel, chipset
and/or CPU, leading to errors that are very hard to trace, and the
"blame" might lie with the hardware and not the kernel.
The kernel boot options seem to be passed properly, and you don't get
the "OK, booting.." message. If you dig into the docs, you'll probably
find that things go wrong while the kernel tries to relocate itself and
tries to execute the unpacking routines, (I haven't been keeping myself
up to data with the Linux kernel and its boot process, but that's at
least how it used to happen - interesting stuff anyway).
The case I referred to involved a 2.4* kernel running fine on anything
except one particular machine (as: motherboard/CPU combination). Another
[minor] kernel version ran without a problem. The unlucky kernel was
compiled by the RedHat company, by the way. In such a case trying a
different kernel version is a lot quicker and easier than trying to
debug the current one. I personally wouldn't even be bothered to file a
So there's the whole story - hope you'll agree that trying to boot from
another kernel is definitely a worthwhile, and relatively easy, option.
If something runs on 1 machine, that's no guarantee that it will run on
another. The proof is in the pudding, or the Soekris box in this case.
Just for clarity and disambiguation: when I wrote about "configuring the
system for serial output", I meant setting the appropriate entry
# Run agetty on COM1/ttyS0
s0:2345:respawn:/sbin/agetty -L -f /etc/issueserial 9600 ttyS0 vt100
That's just an example of course.
[Mmm.. just discovered that my Ubuntu 6.10 doesn't have an inittab anymore.
Yet another well-known standard config file chopped up by those "smart"
Linux maintainers. Now lives in /etc/event.d/, as a collection of chunks.
Another 10.000 docs or so outdated. Those guys/gals are crazy!;]
> All jffs2 modules got inserted properly .... Not even that I am using
> same kernel image now...
> >You could try fdisk-ing the CF again with the real CHS values the
> Soekris combios sees as args to > fdisk. Still, Lilo would usually
> display as screenful of 0x0N's if it fails to load the kernel
> I will definitely try this option today ...
> Hope I will get an error this time to figure out where I am going
> wrong ..
> Thanks ...
> Have a Gr8 day ..
> Bhagyashri Katole
> Project Engineer
> NISG, C-DAC
More information about the Soekris-tech