On Tue, 4 Nov 1997 13:49:45 +0100 
 Marc Baudoin <Marc.Baudoin@solsoft.com> wrote:

 > >  	GO_WINDOW(1);
 > >  	switch (i)  {
 > >  	case EP_LARGEWIN_PROBE:
 > > +	case EP_LARGEWIN_3C589_BUG
 > >  	case (EP_LARGEWIN_PROBE & EP_LARGEWIN_MASK):
 > >  		sc->ep_pktlenshift = 0;
 > >  
 > > 
 > > That should fix your problem. I still don't understand why 
 > > it doens't show up with jtk's patches, though.
 > > 
 > > If that works, please send a PR (so the database knows what model of
 > > 3c589 this affects ), cc: me, and I'll see about committing the fix to
 > > the tree and the 1.3 branch.
 > 
 > Well, for the moment, I can't compile a 1.3A kernel because I
 > have no machine available to tamper with.  I have a friend who
 > can do that but he's on holidays so I'll have to wait until he
 > returns.

Except THAT'S NOT THE CORRECT FIX!  The problem is probably an address
lines problem...

Basically, some laptops have the pcic wired with 10 address bits.  This
means that addresses 0x400 and higher won't work.  You want to use
0x300 -> 0x3ff on such laptops.

Unfortunately, low 0x300-range addresses make some laptops lose, as
well.  Notably, the NEC Versa 6030X (such as the one on my desk over
here...)

I think I'm going to experiment with some high 0x300 range addresses
and see if I can find a set that works more universally (since the
bus width probe is evidently broken).

Jason R. Thorpe                                       thorpej@nas.nasa.gov
NASA Ames Research Center                            Home: +1 408 866 1912
NAS: M/S 258-6                                       Work: +1 650 604 0935
Moffett Field, CA 94035                             Pager: +1 415 428 6939