Re: midi problem, an isa device on a pci card

In message: <20030408042202.GS17533_at_cnd.mcgill.ca>
            Mathew Kanner <mat_at_cnd.mcgill.ca> writes:
: On Apr 07, M. Warner Losh wrote:
: > [... trimmed ...]
: > : >  You have to make them be full pci devices.
: > : >  Hweover, this
: > : > is relatively easy to do.  You hacks will just not work.
: > : 
: > : 	But it's an ISA device on on a pci card.
: > 
: > that's not possible.  It is *NOT* an ISA device unless it is on a card
: > that has ISA fingers on it.  Otherwise it *IS* a PCI device.  Does the
: > pci card have a PCI to ISA bridge on it?  If not, then you have to
: > deal with it as a PCI bus.
: 
: 	I learned my lesson in math class not to argue from instinct
: alone, so I accept that what you are saying is the truth and try to
: make it hang off the pci device.

Yes.  I'm sorry if I'm sounding grumpy about it :-).

: 	Does your gut say I can bus_bus_alloc_resource in the pci driver and do
: port io given that I could from the isa driver?  Will it catch overlap
: from the pci and isa busses?  (suppose I really had an isa pnp mpu
: card and this pci card both wanting ports 0x300-0x302)

Yes.  You can do that.  In fact, most of the drivers in the tree have
common 'resource allocation' routines.

: 	I had a look at the netbsd code and they can say stuff like
: this is a config file:
: mpu at isa?
: mpu at somesoundcard?
: 	I hope my mpu code doesn't need to know about particular
: soundcards, like
: DRIVER_MODULE(mpu, snd_cmi,...)
: DRIVER_MODULE(mpu, snd_visamstrcrd,...)

While you can do that, chances you'll want to do a mii-like bus to
make this easier.

There are a couple of different ways to deal with doing these sorts of
things.  I think the easiest follows after the mii model.

Network drivers with mii hardware attach an miibus child to them when
they detect that their NIC has an mii attachment.  What mii is doesn't
matter too much, but it is basically similar.  The sound cards would
attach, in their attach routine, a sbbus (sbbus == soundblaster bus)
child.  The sbbus would then have different children: mpu, joystick
and maybe others that are common to all sbbus.  I think that the sbbus
may need to have the sound card driver code enumerate these devices
and assign their resources so that they can later attach.

This is, in some ways, similar to what you were doing with the 'reach
around' attachment to the ISA bus.

I'd be happy to help you out a little with writing some sample code
for you so that you can concentrate on the mpu devices and such if my
description isn't clear or detailed enough to do this.

: 	Incidentally, what about vga and joysticks for that matter?
: Not isa, but isa drivers.

vga is a very special case.  joysticks also are poorly handled in
FreeBSD right now.

: > [...more snips...]
: > You likely need to attach the mpu device to the pci soundcard in some
: > way, or if the pci device has its own function, directly to the pci
: > device.  Chances are excellent that's what the interrupt problem you
: > are encountering are.
: 
: 	Not an interrupt problem.  Looking at code for other OSes, it
: doesn't seem like anybody has interrupts for tx ready for mpu401s.

OK.

: 	I dug out my other soundcard and the older midi driver that I
: did a few months ago and it now has the exact same problem of blasting
: the stuff too fast.  I can only assume something has changed in the
: synth timer code (unlikely), or the general freebsd timer code in last
: few months.  

FreeBSD timer could would be my guess.  Maybe the defualt 'HZ' has
changed?

Warner