Nate Williams wrote:
> > "Matthew N. Dodd" wrote:
> > > On Sun, 5 Sep 1999, Warner Losh wrote:
> > > > Might be a good time have a sys/dev/sio and have pccard, cardbus, pci
> > > > and isa attachments there. Yes, I did say cardbus, since I have seen
> > > > cardbus PCI modems that are NOT winmodems.
> > >
> > > And MCA and EISA attachments.
> >
> > Well, it seems Bruce objects to this.. I don't know why though. If he's
> > concerned about loosing the tightly integrated sio<->isa stuff then I guess
> > there could be an "osio" (old sio) or "isasio" or something driver that
> > remains isa-specific. I could well imagine this could be important for
> > older/slower machines.
[full quote restored for bde's benefit, he may not be reading this thread
about adding support for this in -hackers]
> My guess is that he's worried that we'll end up with lots of additional
> 'indirection' through the system, thus slowing down the ability to
> service interrupts in a quick manner.
If that's what he's worried about then I wish he'd say so. Nothing that
we want to do requires breaking this.
> Also, don't fast interrupts depend on the ISA bus? Fast interrupts are
> a requirement for *any* machine to run at a reasonable speed, old/slow
> or new/fast, it doesn't make any difference.
Yes.. Which is part of the reason for moving the interrupt *setup* to
bus-specific code. The handler would be the same regardless and there
wouldn't need to be any indirections or performance impairments as a
result.
One of the things I was going to fix was the hack where com_addr() is using
device_get_softc(). This is not fast interrupt safe and needs to go back
to something along the lines of what it was before. The end result would
be that there would be *less* new-bus code in the core sio.c as it'd be
moved aside.
> Nate
Cheers,
-Peter
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-hackers" in the body of the message
