bootblk problems with raidframe RAID1 root

[Anders Lindgren <ali%df.lth.se@localhost>]

Dear port-sparc64,

  I have tried to set up a RAID1 root on one of my Ultra1 boxes, following 
step by step (literally) the sparc64 instructions of the excellent 
raidFRAME chapter in the NetBSD Guide.

  I have a pair of 36GB SCSI disks (sd0 and sd1) of identical make, model, 
and capacity (as read by disklabel(8)). Both have identical disklabels, as 
created when the first raid component was set up according to the guide, 
and both drives have had /usr/mdec/bootblk installboot(8)'d. They're set 
up with a single root fs on raid0a, swap on raid0b, and a dump partition 
on the physical sd[01]b overlapping the raid0b swap(s), exactly as 
outlined in the guide.

  Booting from the original raid component ("disk1" in OFW, or sd1) works 
fine; the system boots, and raidctl -v -s raid0 is happy about the state, 
listing both sd0a and sd1a as in "optimal" state.

  All _looks_ well, but when trying to verify that it can indeed boot off 
of the other component however, it barfs immediately: Upon typing "boot 
disk0" in OFW the bootblock spews out (immediately after printing its 
header) the error message:

.Inode not directory
Fast Data Access MMU miss

.. and I'm back at the OFW "ok" prompt. Does this mean the bootblock 
thinks the root directory inode of the disk0 component is not a dir?

  I'm confused as to what I've done wrong. I've double checked that both 
disks have the same bootblocks, identical disklabels, and RAIDframe thinks 
the mirror is just fine. What else is there to check?

Many thanks in advance,

/ali -- RAIDframe virgin :)