Oh my God, someone ate my entire pie!

Wednesday, 11 March 2009

Yesterday was my birthday, and it was full of cookies, pies, magical pixie dust and things made entirely of sugar and spice!

The remainder of the day was spent re-installing Debian Lenny upon my EEE PC - Somehow I managed to completely screw the system.

Because the EEE PC is one of those ultra-portable machines I mostly used it when I was travelling, or outdoors. That mean I was generally receiving poor connectivity and the system packages weren't up to date.

While I was in bed I figured I'd dist-upgrade it to the recently released Lenny. Unfortunately I started the dist-upgrade inside X.org, once I realised this I figured I'd cancel the operation via Ctrl-c.

Bad news everbody: I think I was unlucky enough to interrupt an upgrade of libc, or something equally critical. Every single application gave segfaults afterward.

I had two open root terminals and I could navigate around via cd .., and "echo *", but all other commands such as sudo, dpkg, strace just gave segfaults. (Even static commands gave errors - so it might have been the dynamic loader that was borked, I admit I didn't look too closely.)

I figured reinstalling would be a good solution since the machine has a 4Gb root partition and /home was stored on a separate 16Gb volume. Unfortunately I managed to misjudge the installer's partitioning step and nuke the partition table on the external volume so I ended up losing the whole system.

Happily reinstallation was a breeze as my home network is setup to allow installation via PXE network booting (at some point I should document NFS-root PXE-booting). It took me longer to fiddle with the BIOS on the EEE PC to allow network booting than it did to complete a minimal install. Which I guess is good.

I still need to restore my backup of /home/, but that can wait a few days. Right now I'm loathe to touch the machine at all - although I did distract myself by getting KVM to PXE boot:

# create 4gb disk image
dd if=/dev/zero of=/tmp/img.img bs=1024 count=4096k

# launch KVM
sudo kvm -no-acpi
  -boot n -tftp /var/lib/tftpboot/ -bootp /pxelinux.0
  -hda /tmp/img.img
  -net nic,macaddr=00:0E:35:be:de:ad  -net user

It seems that KVM wants to have access to the local TFTP root directory so I just pointed it at that. Since my desktop machine is also my TFTP + DHCP host that works out nicely. (A quick scan of the manual suggests that QEMU/KVM has funky built-in TFTP code, so it doesn't actually forward TFTP requests over the network.)

DHCP requests were certainly passed around as expected though and were answered via my local dnsmasq installation. I did see errors at every DHCP request in syslog, but they seemed harmleess enough:

gold dnsmasq[29241]: no address range available for DHCP request via qemu0

*shrugs*

ObFilm: Never Been Kissed.

| 1 comment.

 

Comments On This Entry

[gravitar] Matt Simmons

Submitted at 01:58:58 on 12 march 2009

Hey, Happy Birthday! Sorry about the problems...

 

Comments are closed on posts which are more than ten days old.

Recent Posts

Recent Tags