Slackware 10.1 hates me.

March 8, 2005 1:38 PM

Help needed with linux (re)install. Managed to somehow kill my slackware 10.1 box, so I reinstalled - now it's not networking because it can't properly reference the hardware auto-detect features. Any help would be appreciated...

Basically, eth0 isn't detected on startup due to a modprobe dependency failure.

Looks like it's trying to reference /lib/modules/2.4.26/modules.dep (which doesn't exist). /lib/modules/2.4.29 does exist though.

So... might be a dumb question, but can I just rename the 2.4.26 directory, or will this break things severely? Any relatively easy way to change the references to the incorrect directory across the board? I'm assuming the numbers refer to the kernel in use; I didn't change kernels (at least i don't think I ever did) so not sure why the numbers don't match. The only recent changes made to the box were to upgrade from 10 to 10.1 using slackupdate (successful) and installing X (unsuccessful due to glibc issues, which prompted the reinstall). I used the same default kernel from the boot disk for the initial install as well as for the reinstall.

All I want is for my little server to be up and running again...

posted by caution live frogs to Computers & Internet (9 answers total)

Best answer: Renaming the directory won't work, as the symbols in the modules and the kernel won't match. The directory that modprobe looks in is based on the kernel currently running, so its not as easy as changing the kernel across the board.

Been ages since I've used Slackware, so I've no idea what kind of package management you're using. But I imagine you could fix things by recompiling your kernel and the associated modules.

Also, did your bootloader (LILO, GRUB, etc.) config get overlooked in the confusion? I know Gentoo and some other distros depend on the initrd file (initial ramdisk) that is specified in the bootloader config to be available so they can do hardware autodetection.
posted by kableh at 1:54 PM on March 8, 2005

Err, changing the references across the board (first para). Sorry.
posted by kableh at 2:00 PM on March 8, 2005

Try either installing the kernel package for linux 2.4.29, or the kernel modules package (I assume they're separate on slackware) for 2.4.26.
posted by fvw at 2:08 PM on March 8, 2005

Response by poster: No, I configured LILO but I think somehow the kernel must have been updated when using slackupdate, and then taking the (older) kernel used on the boot disk when reinstalling it didn't end up matching the module symbols.

Biting the bullet and recompiling the kernel now... have never done that before, so this should be a hoot.
posted by caution live frogs at 2:16 PM on March 8, 2005

Definitely don't rename it. Listen to others if you want to do it the right way, but if you want to do it the quick-and-dirty-this-probably-won't-break-things-way, do this:

ln -s /lib/modules/2.4.29 /lib/modules/2.4.26

which creates a symbolic link to 2.4.29 named 2.4.26.
posted by thebabelfish at 2:20 PM on March 8, 2005

Response by poster: Ah, fudge. Now I'm out of disk space - that will teach me to use the smallest drive available when building a server.

Not sure how I managed to get the system this screwed up... trying to delete files to make more room but it doesn't seem to be working, df still gives me 100% used and the package tool won't load so I can't uninstall anything.

Crap.
posted by caution live frogs at 2:35 PM on March 8, 2005

Response by poster: kableh: I think rebuilding will fix it if I can get some free space to do so (/tmp rm * seemed to do it, although df is still pegged at 100% - weird).

Apparently the kernel for 2.4.29 is sitting in /boot but isn't being pulled up by LILO. I'll see if I can get LILO to use that kernel (since it's already there) before trying anything esle drastic...
posted by caution live frogs at 2:42 PM on March 8, 2005

You'd have to edit your lilo.conf, of course, and that'd certainly be easier than recompiling your kernel.

Were you in the middle of building when you ran out of disk? Try going into your kernel source tree and running 'make clean'.

- M
posted by kableh at 2:50 PM on March 8, 2005

Response by poster: Editing lilo.conf then resetting LILO using lilo -v to reference the new kernel seems to have done the trick. Server is up and I can ssh into it again. I am still low on disk space but I didn't want this server to be used for large file transfers anyway. Thanks for the help, MeFites (and lurkers).
posted by caution live frogs at 4:26 PM on March 8, 2005