Hi All,
Michael (macallan18) and I have been working on identifying a problem
with the kernel not properly recognizing the boot device on SCSI hard
drives. The particular instance is where there are two valid NetBSD
installs on two separate hard drives. If the hard drive with the higher
SCSI ID is used to boot, the kernel will not recognize this and insist
on using the lower SCSI ID drive for the file system.

It is also possible that the problem only occurs when the
stock hard drive is used as one of the two hard drives. However, the
problem may not be limited to this particular aspect, as this could
impact single installs as well, if the booting drive does not match the
Open Firmware node @0. This would most likely become apparent at booting
when the kernel requests the boot device to be manually entered, even
if there is only one hard drive installed and fstab matches sd0.

We believe that the problem is in how the kernel identifies the node. In
some Open Firmware implementations, the node will be identified as @0
when OF routines are used, even if the node is @5 or @6, which are
common SCSI IDs for second hard drives.

In an attempt to determine how widespread this is, I am posting to the
list. What we're looking for is people that can reproduce this, so that
if we have a patch we can test it quickly. This problem does occur in
-current. 

thanks,
tim