Hi,
I finally had some time to tinker with my HURD box again.
Justus Winter wrote:
If you are running a debug kernel (and you should), you can break into
the kernel debugger using ctrl-alt-d and do:
Sadly I am not. I have only one kernel in my /boot dir and it doesn't
respond to ctrl-alt-d. apt-get installs apparently only the non-debug
kernel without more hints. I bet I had a debug kernel installed too, but
it is gone or iI hallucinate.
To recover from this, use a Hurd live CD (I guess the Debian installer
CD will do), or even a random Debian/Linux live CD. Run fsck on your
root filesystem. If that doesn't help, some important files might
have been corrupted. You could then try reinstalling the hurd,
hurd-libs, and libc0.3 packages. This is doable even from Linux
(mount your rootfs using mount -t ext2 /dev/your/root /mnt, use
dpkg-deb --extract hurd.deb /mnt).
I tried a slightly different procedure: I am able to boot the last HURD
cdrom I had, I mount the target, chroot to it and I had in the cache
three different hurd versions with respective libraries, I used "dpkg
-i" to install the previous one, kernel + lib, (which most probably did
work, since I was able to reboot and install the latter one) but still
my box freezes.
Is my procedure sound as yours? I suppose so. perhaps what's missing is
a libc downgrade then.
To prevent this in the future, regularly run touch /forcefsck to force
a fsck on your filesystems. There is a known problem in ext2fs
affecting root filesystem translators, deleted inodes with zero dtime
are accumulating until no free inodes are left.
:) I'll remember that, however it would be a nice fix to see in the kernel.
Riccardo
