Hello.

To be safe I bought an new identical 80gb sata harddrive to install Fedora Core 6 on. The old one has Fedora Core 5 installed on it. I installed fc6 on the new drive with the old drive unplugged. I then plugged the old drive back in. After configuring /boot/grub/grub.conf, I was able to see and switch to both systems on their own drive from the fc5 initial boot screen. Everything was working until I decided to enter the boot menu and boot into fc6 on the new disc directly. I get this message from dmesg on fc5:

hfs: unable to find HFS+ superblock
VFS: Can't find ext3 filesystem on dev sdb.

Using the rescue disc I chroot /dev/sysimage -> nano /boot/grub/grub.conf. But grub.conf is empty. I was able to do this on initial install to get the grub.conf entry info to add to the fc5 master grub.conf.

I can can still boot into fc5, but when I choose the fc6 entry on the fc5 initial boot screen, I get a lock up and kernel panic.

How can I correct this? Please help.

Pre thankyou,

-ron-

Sounds like the booting kernel is aimed at the wrong partition or maybe something kernel. What's the exact panic? Is it, "Unable to mount Root FS on blah blah blah?"

Cheers,

Finegan

Hello,

Selecting the fc6 entry on the initial Grub boot screen from fc5 on the old sata drive.

Switchroot: Mount Failed: No Such File or Directory.
Kernel Panic Not Syncing: Attempted to Kill Init.

-ron-

It looks like its initrd is aimed at a partition that's not even there. This could get mungly. Normally its just a matter of booting using the kernel argument of root=/dev/hda4 (or whatever is your root), but FC has always used initrds, so the kernel actually boots to the initrd, there's a tiny /init script in there that basically just loads a bunch of kernel modules (also in the initrd file, maybe does some other fun and games, and then exec's the init on the real root filesystem.

The only way I know of to fix that is to unpack the initrd, usually named initrd.gz, loopback mount it, hand edit the script so its aimed at the right partition, unmount the loop and repack it.

Something weird happened though if you've got FC5 on one partition and FC6 on another? Do you have /boot shared between the two? If I remember right (going back a bit, haven't messed with FC since 3ish), initrd.gz is a symplink to whichever one matches the kernel, like initrd-2.6.14.blah.tar.gz

This is why initrds suck.

Regardless, if you want to go wacky in depth on this:

Find the initrd that the FC6 kernel is using from your grub.conf
gunzip initrd.gz
mount -o loop initrd /mnt/somewhere
nano /mnt/somewhere/init (it might be /mnt/somewhere/sbin/init)
Look for the switchroot call, its probably going to be a variable, track up in the script to where that partition is declared, fix it, umount, re-gunzip, etc...

If you want to just forgo using an initrd, you kinda can't. FC modularized ext3 support and its one of the modules that the initrd loads before switch rooting. So, the kernel aimed directly at the rootfs can't read the rootfs. This is wacky fun.

If you want to compile your own kernel with all the ext3 support in it and just aim that right at the FC6 partition, its probably the best way to go, depending on your skill level. Heck 3/4s of this whole missive may seem pedantic.

Then again, I may be on crack and you can just pass the kernel root=/dev/hda4 or whatever and the initrd will take the kernel argument and run with it...

Cheers,

Finegan