Kernel compilation on eBox33xx

[Daniel Berenguer]

In order to enable the watchdog timer in the kernel, I'm recompiling
the kernel INTO my eBox 3310s with Debian Lenny (kernel 2.6.30)
running on it. Everything goes fine except for the final test. After
the installation of the compiled deb files, rebooting the system ends
by a long "Loading, please wait..." and then "ALERT! /dev/hdc1 does
not exist. Dropping to a shell!!"

In my last try, I've even recompiled the new kernel (2.6.30.4) with
the current working config file (2.6.30) without modifying any option
in the kernel, but no lack, the same error appears after rebooting the
PC. For the rest of tries, I've followed Kelvin's tutorial:
http://kelvinism.com/tech-blog/compiling-kernel-vortex86dx/ with the
difference that my compilations are done into the eBox itself.

Should the kernel compilation work given that I'm keeping the current
working options? Should I try maybe with a kernel with version 2.6.30,
the same than my current config file?

Thanks for your feedback.

Daniel.

[Ben Tullis]

Hi Daniel,
I haven't followed those instructions myself, but I've run Lenny on my
Microclient Jr without issue.
It's currently running Ubuntu 10.4, so I can't check against it now.

My thoughts are, in no particular order:

If you're dropping to a Busybox shell, then that suggests that the
initrd is present and unpacks successfully.
It could be that either udev isn't starting, or that your compact flash
card isn't showing up as /dev/hdc

I would have expected it to be /dev/hda anyway. Try doing an "ls -l
/dev" from your shell, to see what is in /dev
If there is very little, then it suggests that udev hasn't started
properly. Try running "udevadm trigger" and see if this detects anything
else. If that fixes it, then check the contents of
/etc/initramfs-tools/scripts/ on the CF card, or wherever you compiled
the kernel.

If there's a /dev/hda and not a /dev/hdc then you might be able to fix
it by editing the grub config interactively, during the boot process.
Press "e" to edit a line and change the root device to /dev/sda1, then
press "b" to boot. This doesn't update the config file, but you'll be
able to do that after a successful boot.

It's worth checking whether there are any /dev/sd* devices, because it
might show up as a SCSI (libATA I think) device now. Again, this should
be fixable with a bit of grub editing. If you're using lilo, then you'll
probably have to edit the config file on another machine, then rerun
lilo on it.

I hope that helps a bit.

Regards,

Ben

[Daniel Berenguer]

Thanks Ben but there is no hd* nor sd* device in /dev.

I've just tried with a new compilation, this time with a 2.6.30
kernel, the same version that I installed, with no extra modifications
in the kernel, but I get the same error after rebooting.

Daniel.

[Ben Tullis]

As I say, try running "udevadm trigger" from your busybox shell, because
it sounds like udev isn't running.
Are you able to list what does appear in /dev ?

Ben

[dberenguer]

Thanks Ben,

ls /dev does list a lot of devices: mainly pty and tty and a bunch of
numeric N:NNN devices. Running "udevadm trigger" does not seem to
change anything so I guess that udev was already running.

I have really no idea about what to do next. I'm even thinking in
letting go the idea of activating the watchdog...

Thanks again.

Daniel.

> >>>http://kelvinism.com/tech-blog/compiling-kernel-vortex86dx/with the