I suspect, there's something else that is wrong with your machine. 
Might very well be broken RAM or the like. I've seens random core dumps 
once I had bad RAM -- in fact the RAM was good, but it was not seated 
correctly.

Anyhow I have a 604ev based machine (ANS 700) that has been running a 
1.6.2 kernel since april 2004:

     NetBSD 1.6.2_STABLE (ANS700MLX-$Revision: 1.6 $) #2: Thu Apr 22 
14:43:20 UTC 2004
     total memory = 416 MB
     avail memory = 376 MB
     using 2048 buffers containing 21400 KB of memory
     mainbus0 (root)
     cpu0 at mainbus0: 604ev (Revision 2.2), ID 0 (primary)
     cpu0: HID0 8090c084<EMCP,DOZE,DPM,ICE,DCE,SGE,BHT>

Hope this helps.

/Daniel


On Feb 24, 2005, at 1:31 PM, John Klos wrote:

> Hi,
>
> 1.6.2 appears to be broken, at least for 604 processors. On a 603e 
> system, I was able to run build.sh and install an entire new system 
> and run it, but at least since last summer, I cannot do this on a 
> 604ev system.
>
> When the 604 system gets to the point where it'd install 
> libexec/ld.elf_so, it starts dumping core everywhere:
>
> STRIP=/usr/tools/bin/powerpc--netbsd-strip /usr/tools/bin/nbinstall  
> -c -p -r -o root -g wheel -m 444  ld.aout_so.cat1 
> //usr/share/man/cat1/ld.aout_so.0
> install ===> libexec/ld.elf_so
> STRIP=/usr/tools/bin/powerpc--netbsd-strip /usr/tools/bin/nbinstall  
> -c -p -r -o root -g wheel -m 555   ld.elf_so //usr/libexec/ld.elf_so
> Illegal instruction - core dumped
>
> If I move the new ld.elf_so from the 603e system to the chroot, the 
> system becomes happy again.
>
> So - the question is this - what's broken that would cause or allow 
> ld.elf_so to be compiled differently between a 603 and a 604 system?
>
> Both are 1.6.2, both are running a new kernel compiled this month, 
> and both have new src trees (-rnetbsd-1-6). No CPU optimizations are 
> used.
>
> Ideas?
>
> John Klos