As noted in bug #283193, d-i's current handling of failure to detect a hard disk is suboptimal. The installer fails with an error message in partman, and the user is left without a way to continue unless they know how to modprobe modules in the shell. I've got a patch that does away with hw-detect-full (bonus; never liked that name), and replaces it with disk-detect, which is very similar to ethdetect: It runs hw-detect to try to find a disk, if this fails it prompts the user with a (currently long and uncommented) list of disk driver modules, lets them load a module, and tries again. The list of modules also lets the user choose "continue with no disk drive", or "none of the above", which lets the user load third party disk drivers from a floppy. The tricky bit is, of course, detecting whether a partitionable disk has been found. The technique I've used is to run parted_devices from partman if it's available, otherwise do some silly find in /dev/discs. The find is essentially how the partitioner udeb does it, so should keep things working on the few arches that still use it (which I really hope will all switch to partman soon). Using parted_devices perhaps bends the partman model a bit, since partman addons could conceivably provide a non-disk device which is still a valid partman device. I think this is probably ok, it only means that hypothetical diskless NFS installs would need to choose "continue with no disk drive" in disk-detect. Wanted to make sure this was ok before committing it. -- see shy jo
Attachment:
signature.asc
Description: Digital signature