Re: HP9000 K-Series support

Making, drinking tea and reading an opus magnum from Hellmuth Michaelis:
> Michael Shalayeff wrote:
> 
> > > Now i need a driver for the built in Fast-Wide-Diff SCSI bus .. ;-)
> > 
> > i must be blind and do not see fwd in your dmesg?
> 
> Its not in the dmesg. I have no idea yet why, but i going to find out
> why its not there (and a couple of other cards which are in the K but
> do not display).

it's possible that it's behind the runway.
i beleive runway is the ccio driver although
only exists on 64bit machines as told.
seems like not only. fortunately it has some
dumb mode to work transparently w/ minimal
code size and thus should be quick-hackable.
my guess.

> Debian probes (and uses) the built-in FWD SCSI as
> 
> 4. KittyHawk GSY Core FW-SCSI (4) at 0xf1800000 [10/0], versions 0x15, 0x1, 0x89
> 
> so its really there .. ;-)
> 
> I have seen that you seem to have experimented with a siop driver,
> do you have something which might help me with the 53c720 ?

i have not tried siop(4) for 720 yet. there is a tiny bit of
code to be written to make it usable. dunno about siop(4)
usability for 720 at all though.

> > did you reverse the device probe in your changes for some particular reason ?
> 
> The PDC was scanned from max index to 0 and i have changed it to go from
> 0 to PDC_SYSTEM_MAP return code equal -5. This completes (for me) in 6
> calls to PDC as opposed to 64-6 calls and conforms to the probe algorithm
> as described in PDC doc (1.0, Aug 2001, p4-79).

that makes sense although looks ugly (;
i think an approach maybe to scan for hpa's forward
and then attach devices off the array backwards
just to avoid recursion that would be more appropriate
although stack is not big enough for that.

> The cpu and the mem device do not appear at the beginning because they are
> not returned by probing via the PDC. It just happened that i called the
> new probe routine below the others so they show up at the bottom of the
> list.

oh. i guess you are writing code for that funky third scan method
we do not support yet. i think it only exists on 32bit servers
and 64bit machines thus i did not write it (;

> > also why does it list same device multiple times?
> 
> Because the code needs (much) more work. I'm really glad that i have it
> running and working on the K220 now (on an old 2G SE SCSI drive) and
> i'm able to work directly on the target instead of the dog slow 715/64.

cool. i bet it's fast (w/ those caches)!
can you show me please whatever diffs you have now --
just to grasp the idea of what is happenning.
i'd know how to collaborate that w/ the stuff
i'm working on to make it all more usable.

10x
cu

-- 
    paranoic mickey       (my employers have changed but, the name has remained)