> > I wanted to ask when the change went into -current, but I shall do it
> > now: why do you remove the DATA32 capability in this case?
> 
> Just being extra conservative I guess. It seemed to be one of the required
> changes to fix the problem...

I've had a bit more time to look at the "problem".  On my rev 11 OPTi
chipset, the only "problem" is that you cannot use a pci (or isa?) card
that takes irq 10, AFAICT.  As soon as I removed the card I had in
there (older Intel 182557 EtherExpress 100B clone), no more warnings
were issued.  I haven't seen it documented anywhere, but irq 10 seems
to be reserved on this machine (IBM PC750, type 6883).  [Could there be
such a thing that bus-master support itself requires a certain irq?]

Anyway, I'm in the process of shuffling cards around.  I've traded the
one fixed at irq 10 for one fixed at irq 6, but I've got to disable the
floppy drive to use it.  It's going to take a few days to get two nics
in there that don't conflict with the bus-master stuff or themselves.

Anyway, in my case at least, the patch that was sent to the list was
incorrect and should not be applied IMHO.  The problem is not that the
bus-master support is not there or not working, but that for some reason
the irq tables from the BIOS are bogus and need to be adjusted.  The
1.5.1 kernel seems to be able to do that as long as you don't stick a
card in there using irq 10.

henry nelson