minimal root file system for testing dangerous kernels

[hakbeest]

the linux driver for my ide chipset seems to have a lot of trouble with dma. If I boot it up without ide=nodma it completely corrupts the root partition. Still for debugging purposes I need to boot in this way, so I've created an auxiliary partition on my disk for being the root fs for such kernels which corrupt it. Unfortunately since it is still empty I cannot boot any kernels with it.

What I woukd like to know is a fast way (my disk is doing ~2MB/s reads) of populating this filesystem with just enough files to boot a kernel and be able to scp a copy of dmesg to another box. I may have to do this a lot of times, so it has to be simple and fast.

[mbar]

Why don't you just use a stage3 tarball?

[hakbeest]

Yes, good idea. I extracted a stage 3 tarball to my victim partition and rebooted with a hungry kernel. Unfortunately while booting it corrupts my filesystem so the init scripts can't run and I have to enter the root passwd (which I believe is autoscrambled) or press ctrl-D to reboot

hmm, I guess I need to construct an initial rootdisk to load into ram.

[hakbeest]

I've tried using an initrd from the iso of minimax linux. For some reason though, dmesg output output is incomplete. The first half or so just isn't there. I've not been able to find out how this can happen.

I'm thinking about building a custom livecd with catalyst.