Date:        Wed, 19 Sep 2001 18:32:55 -0600
    From:        Greg Oster <oster@cs.usask.ca>
    Message-ID:  <200109200032.SAA08834@cs.usask.ca>

  | No, it shouldn't.  I'm guessing that 'booted_device' is getting changed
  | on the sparc port after RAIDframe changes it to be raid0a.... but I've not
  | tracked this down yet...

My assumption is that Manuel must have "root on sd0a" (or similar) in his
kernel config - otherwise the root device would be defaulting to the
place where /netbsd came from, which isn't a root filesys from what was
said.

Given that, if RAIDframe is altering it, then I submit it shouldn't be.
If I expressly config the root device in my kernel, then that's the one
I want to be used, no matter what some smart code in the kernel thinks.
If I wanted "root on raid0a" that's what I'd configure instead of "root
on sd0a".

The only time RAIDframe should be playing is when there's just "root on ?"
in which case if the place (drive/filesys) that would be selected as the
root is a part of a raidset, then using the raid as the root device rather
than one if its components makes lots of sense.

The reason this is probably more noticeable on sparc than i386, is that
the sparc boot proms pass through the actual partition that was booted
from, so NetBSD will know exactly where root should be - so a "root on"
is going to be needed if you want /netbsd from one partition and the
root from another.   i386's just pass the device, and I believe that
NetBSD then assumes that the a partition in the label found on that device
is the root, regardless of where the /netbsd actually came from - so there
things tend to work without the "root on" (even before the 386 boot blocks
learned to deal with booting from raid, if that happened yet).

kre