I can pretty reliably reproduce this bug by doing "MAKEDEV all" (or a
sufficiently large subset thereof) if the dev directory is empty and
initially 512 bytes long.  If I manually
	MAKEDEV std
	MAKEDEV	fd
	etc.
then the panic doesn't occur.  Once the directory has been filled,
MAKEDEV all works fine.

After the panic, fsck complains that the last directory to be extended
has an oddball size.

This problem does not seem to occur if all the filesystems are built
with 4K blocks and 1K fragments.  The panic reliably occurs if I use
8K blocks and 1K fragments.

Looks like some sort of wierd filesystem bug to me.

On Dec 1,  5:39pm, Don Lewis wrote:
} Subject: Re: problem installing NetBSD-1.1 on hp425t
} On Dec 1,  7:01am, Don Lewis wrote:
} } Subject: problem installing NetBSD-1.1 on hp425t
} } I've run into a problem installing 1.1 on an hp425t.  The installation
} } script gets most of the way through, but blows up making the devices.
} } 
} } Making devices...vm_fault(105000, 4ef80000, 1, 0) -> 1
} }   type 8 code[mmu,,ssw]:505
} } trap type 8, code: 505, v=4ef80400
} } kernel: MMU fault trap
} } pid=462, pc=0002FE06, s=2004, sfc=0001, dfc=0001
} } 
} } followed by a register dump and a kernel stack dump.
} 
} More information (please excuse any typos):
} 
} _Debugger(...) + 6
} _panic(...) + 34
} _trap(8, 505, 4ef8040) + 1be
} _addrerr(?)
} _vgone(1c4b300, 5, 2180, 1c4b300, 1c4b31c) + de
} _ufs_inactive(797dc8, 1c4b300, 1c1e500, a8520, 1c4b300) + 18e
} _vrele(1c4b300, 2f99a, a85a4, 1c4b300, 797e3c) + 52
} _vput(1c4b300, 1c4b480, 1c1ac00, c, 0) + 2c
} _ufs_makeinode(2180, 1c4b480, 797eb0, 797ec4) + 248
} _ufs_mknode(797e8c, e8, e, fa056, 0) + 32
} _sys_mknode(1c51d00, 797f88, 797f80) + 1d8
} _syscall(e) + 10a
} _trap0() + e
}-- End of excerpt from Don Lewis