Hello, I have a mostly working partition implementation for ubd
This is a patch against 2.4.9-8um

I have booted off of the second partition on the second device but don't
know how you would like to handle the problem of naming -- currently
root=/dev/ubd0 (whole disk 0)
root=/deb/ubd1 (disk 0 part 1)
root=/dev/ubd16 (whole disk 1)
It could have multiple entries for "ubda","ubdb" etc. like ide does in the
device name table. "ubd0" "ubd1" would also cause confusion ubd0 vs ubd00 ?

I am unhappy with simple_strtoul() if the base is set it will not accept hex
or octal values so root=/dev/ubd0x12 does not work nor root=ubd0x12
or even root=0x6212
but root=/dev/0x6212 and root=25106 do I think
it is also possible to use the same shift as the ide drives and with the
fake major have the partitions work as expected with the default
device nodes. hda1... hdb1...

currently it just uses a 4 bit shift
if the shift is set to 0 the current behaviour is possible (devfs will
create the disc/disc0 entries though)

What about those "<< 9" in ubd.c I think most are intended as hardsect_size

Note that to loop mount the partitioned filesystem
mount -o loop -o offset=$((512 * start_sector)) part_fs /mnt
hardsect_size = 512 bytes/sector usually but is set to 1K in ubd.c?

The ubd.c code currently does not support revalidate is this a design
decision or just something that has not been gotten to?

Get your FREE download of MSN Explorer at http://explorer.msn.com