On Sun, Jul 14, 2002 at 11:43:35PM -0700, Matt Thomas wrote:
> >
> >I think it's likely that when DEC increased the firmware from 256KB in
> >4k60 to 512KB in 4k90 they just had a lot of space they didn't need.
> >Those last 4 bytes could well be the checksum, but I have no idea what
> >method they use to derive it.
> 
> Where's a valid image I can peruse?  I can probably figure it out.
> My guess is either a CRC-32 or a simple SUM32.
> 
The images I've collected so far are avialable at:
	http://www.oxide.org/vaxproms/

I've also spent some time running small programs and a custom kernel
through Chuck Cranor's machine, but nothing I do seems to restore that
obviously errant nul byte.

I did a little reading about FEPROM though, and I get the impression
that some may allow bits to be cleared at any time, but set to 1 only
in complete blocks.

If the Cougar works this way then a fix will likely require blanking
and rewriting the whole part, but without specific knowledge of the
specifics (such as timing) of this operation I'm hesitant to write a
driver to do it. If someone who actually owns one of these machines
wants to check if the proms are socketed, and has access to a burner
for restoring when things go wrong, then they will have a safety net
for debugging a mopable prom rewriter.

On the other hard, DEC apparently did have software to do this in the
field, and locating their software would be the most robust solution.

I'm still interested in receiving more prom images, to confirm once
and for all that dz's misprobing was at fault, and to spot any further
corruption in non obvious places. I'll continue to provide the images
should anyone with more familiarity with FEPROM want to work on the
problem, and it's still possible that some magic write / read / write
combo remains to be discovered.

Regards,

/Hugh

> 
> -- 
> Matt Thomas               Internet:   matt@3am-software.com
> 3am Software Foundry      WWW URL:    http://www.3am-software.com/bio/matt/
> Cupertino, CA             Disclaimer: I avow all knowledge of this message
>