Subject: Re: Can someone test out a GENERIC kernel on a 3100?
To: Jonathan Stone <jonathan@NetBSD.ORG>
From: Michael L. Hitch <mhitch@lightning.oscs.montana.edu>
List: port-pmax
Date: 05/16/1997 21:50:59
On May 16,  2:25pm, Jonathan Stone wrote:
> 
> mhitch@lightning.oscs.montana.edu (Michael L. Hitch) writes:
> 
> >  One interesting thing I noticed:  the sii driver does not negotiate
> >synchronous transfers, but does accept synchronous transfers if the disk
> >requests them.  The HP disk and RZ58 drives did not request synchronous
> >transfer, but the RZ56 drives do.
> 
> Sync vs. async. Interesting. Yet how come 1.2.1 works, and 1.2D doesn't?

  It gets even more interesting:  for the heck of it, I modified the
response to the sync transfer request to set the period and offset to 0.
Now the tar extract gets farther before it hangs.  And more interesting yet -
I found that when it hangs, the command is a write of 64K bytes.  [Setting
sii_debug to 2 or 3 makes for lots of console output :-), and it's not a
good idea to have syslogd running! :-) :-)]

  I'm not sure why 1.2.1 would work and later versions don't.  One
possibility I can think of is a timing problem of some kind.  All the kernel
changes between 1.2.1 and 1.2D almost certainly affects timing in subtle
ways.  Forcing async transfers could also possibly affect timing as well.

Michael

-- 
Michael L. Hitch			mhitch@montana.edu
Computer Consultant
Information Technology Center
Montana State University	Bozeman, MT	USA