Kevin P. Fleming wrote:
Not a big deal, this an old machine... but the kernel startup dies right after the Seagate SCSI driver loads "seagate: driver for ST-0x/TMC-8xx". No error messages or anything, just a complete lock up.
I've been successful making a modified live CD using the 2.6.11-rc4 kernel and turning off nearly all the SCSI drivers (otherwise no changes, I just replaced the kernel binary and the modules with my new ones). This boots fine on this box, so I think that the whole "hotplug-based module loading" deal may be more valuable than I first thought.
Quite. Currently we have to include *all* SCSI we can as built-ins to catch whatever cd drive the cd might be in. :/ If we could get this working well, it would help a great deal.
I think the size problem you mentioned before could be somewhat taken care of. IIUC, this is what we'd have:
1) Initramfs is filled with a custom init, hotplug, modprobe, udev?, and (all?)modules.
2) All that is compressed and built into the kernel, so now you've got the entire kernel, including the compressed initramfs image in memory
3) The kernel creates a ramdisk, mounts it as '/' and dumps the contents of the initramfs image there (so now, do we have the contents of the initramfs in memory twice?)
4) Inside the init, after the modules are loaded, conceivably, we could have init delete all that stuff from the ramdisk (modules, modprobe, hotplug, udev, etc) because it will then mount the cd and symlink to the appropriate dirs which will have copies of all that stuff on the cd.
Does that make sense? Btw, this is just me thinking out loud...
-- Jeremy H. -- http://linuxfromscratch.org/mailman/listinfo/livecd FAQ: http://www.linuxfromscratch.org/faq/ Unsubscribe: See the above information page
