I've had similar problems when trying to boot a kernel with a '.' in its
name. Try changing the kernel name to netbsd-2hi, and see if that helps.

On Sat, 30 May 1998, Frederick Bruckman wrote:

> 
> The Booter displays this messge:
> 
> 	Booting...File "netbsd.2hi" not found
> 	Could not open kernel "(0,?)netbsd.2hi"
> 
> on my system whenever the inode number of the kernel >= 47500. 
> I know this because there happen to be three in a row with inode numbers
> 47499, 47500, 47501. One of these used to boot, and I got the other to
> work by deleting some files and "cp"'ing it before "mv", thus taking the
> identical file to a lower inode number. Same thing with Booter 1.11.1 and
> 1.11.2.
> 
> A strange thing recently occurred with a new kernel. It booted at first,
> but at some point, there were all kinds of wierd errors in single user
> mode--"..executable is in wrong format"; also "sh" complained that it
> couldn't parse simple commands like ``mount -u /''. I had thought at the
> time, that my filesystem was corrupted. The problems uncovered by fsck:
> one directory named ",", and "summary information bad," don't really
> explain this problem. Perhaps these two oddities are related.
> 
> There are other problems with this file-system. It was originally made
> with ``newfs -O ...'', and then later upgraded with ``fsck -c 1 ...''. 
> ``fsck -c 3'' dumps core, and then I had to use ``fsck -b 2'' to clean up. 
> It now won't take ``newfs -n ...'' without complaining
> 
> 	newfs: /dev/rsd0a: bad rotational layout count
> 
> Therefore, I really couldn't say if it's a bug in the Booter, or if the
> problem is dependant on this particular file-system.
> 
> It's reproducible as described. All the kernels made now come in with high
> inode numbers, but of course that could easily change. I won't be able to
> back up and rebuild this filesystem anytime soon, and intend to keep at
> least one kernel around for testing. I'm willing to work with whoever is
> currently working on the Booter.
>