"David G. Andersen" <danderse@cs.utah.edu> writes:
>   I decided to take our shark cluster up to the 1.4-ALPHA snapshot
> recently, and found that 1.4 won't netboot on dnards with the August
> 1997 firmware, but will netboot with anything past February 1998.  Is
> this a known bug (e.g., "Just go upgrade the bloody firmware"), or is
> it something someone feels like fixing?  I'm not all that familiar
> with the CS89x0 driver, or I'd peek in to it a bit more deeply.

It was an issue with the old firmware.


>   My most likely guess is that the old firmware is probably leaving
> the device in some weird state, but that's a guess.  The patched
> NetBSD 1.3.1 from DEC worked like a charm on sharks with this release
> of the firmware, so it seems like a driver issue.

The old firmware did not specify the correct properties (interrupt,
drq, I/O addresses, etc.) for the devices.

The hacked version of NetBSD 1.3.1 released by Digital coped with that
by simply ignoring the values provided by the firmware, and using
values known to be correct.

That's not the right solution for a generic OFW device attachment, for
many reasons.  You can't even kludge around it, because, for all you
know, _it might be right_...  (it's not like the FW isn't supplying
values, it's supplying the _wrong_ values.)


At one point, i went through the entire firmware properties list and
sent in an update to the firmware folks, and subsequents releases of
the firmware have the correct information.  I think i might have
missed one or two things, but they weren't important w.r.t. having
NetBSD "do the right thing."



cgd
-- 
Chris Demetriou - cgd@netbsd.org - http://www.netbsd.org/People/Pages/cgd.html
Disclaimer: Not speaking for NetBSD, just expressing my own opinion.