for encrypted root. That can be extended for the entire disk if you boot from
CDROM or USB using the entire disk /dev/sda for instance.
This is not very useful, as I'll end up with a bunch of /dev/loop* devices then. Those devices would have partition tables in them, and in particular, they contain RAID arrays (partition type RAID autodetect). It is difficult to keep track of which HDD is which, but if the kernel can autodetect that, it'll be a lot easier. That is why I want the kernel to read them as HDDs, rather than just provide me some block devices.
The above is only my solution to the problem. There probably are others (that still do things automatically!), but this was the first thing that I thought about.
Example:
1. Setup cryptoloops
/dev/sda = /dev/loop0
/dev/sdb = /dev/loop1
/dev/sdc = /dev/loop2
2. Let the kernel detect them as HDDs
/dev/loop0 = HDD => kernel detects partitions
/dev/loop0-partition1 = RAID-5 md0 disk 2
/dev/loop0-partition2 = RAID-1 md1 disk 0
/dev/loop1 = HDD => kernel detects partitions
/dev/loop1-partition1 = RAID-1 md1 disk 1
/dev/loop1-partition2 = RAID-5 md0 disk 1
/dev/loop2-partition2 = RAID-5 md0 disk 0
3. The kernel found RAID partitions, so it automatically assembles them:
/dev/md0 = my RAID-5
/dev/md1 = my RAID-1
4. I can easily mount /dev/md1 as root, etc.
If the kernel cannot detect them that way, I have serious trouble trying to figure out where exactly is each partition, which array it belongs to, etc.
