19:59 Sunday, September 09 2006

initrd fiasco

I built 2.6.17.13 today on a system that had been running 2.6.16.18 for
the past 100+ days (of uptime, no less). I used the same kernel config to build 2.6.17.13 as I had used for 2.6.16.18. I built it with "make oldconfig && make all && make modules_install && make install". All attempts to boot the 2.6.17.13 kernel kept failing with a kernel panic when attempting to mount the / filesystem complaining that "could not mount /dev/root".
Since I have no clue what /dev/root was, I googled a bit, and determined
that its part of the initrd. I have an initrd for 2.6.17.13 just as I have one for 2.6.16.18. I like to build my kernels with support for the / filesystem, so in reality, I didn't really need an initrd (no SCSI disks here). I commented out the initrd entry (which was valid) in lilo.conf (yes, this is my last box using LILO, i've been converted to grub everywhere else, old habits die hard), ran /sbin/lilo, and rebooted to 2.6.17.13.
Guess what?! It boots fine without the initrd. I'm quite stumped on what is going on here. I'm glad that it boots now, but i don't get how the initrd can be busticated like this when the kernel build process created it. Anyone? Beuller?