Re: Question about Interrupt Pin PCI Config. Register.

[michael.bender@Eng.Sun.COM (Michael Bender)]

> From H_John_McGrath@ccm.jf.intel.com Fri Oct 11 17:44:21 1996
> 
> If it is a bridge to 16-bit PCMCIA cards (known now as PC Card-16 cards) 
> the interrupts it is generating are probably legacy IRQs as side-band 
> signals rather than PCI INTs. Hence the '0' return.
>
> If the bridge supported PCMCIA CardBus cards, as well as PC Card-16, it 
> would provide support for PCI INTs and the register wouldn't be fixed at 
> '0'.

All of the PCI-PCMCIA bridges that I've worked with support both PCI-style
interrupts and ISA side-band interrupts (the latter is a crock, IMHO).

The PCI-R2 bridges such as the Cirrus 6729 and 6730 and the Intel 82092 can
be configured in software as to which type of interrupt to generate. Some
old versions of the 6729 indicated in PCI config space that they generated
NO interrupt, while later versions indicate one IRQ required. Our software
winds up routing adapter status change and card IO interrupts to a single
PCI pin in those systems where ISA side-band IRQs are not available or not
desired.

All of these bridges can be configured so as to route any of the possible
interrupt sources (adapter status change, PC Card IO IRQ) to any of the four
PCI INTx lines, but whether or not the bridge is actually wired into the
system such that any or all of the adapter's PCI INTx outputs are connected
to the PCI bus INTx inputs is another matter. It seems that laptop
manufacturers have taken very creative liberties as to how they wire their
PCI-PCMCIA bridge interrupts into their system. Unfortunately, most of them
seem to route the adapter's INTx lines to ISA side-band IRQs, and not
provide a mechanism to route the adapter's INTx lines to the PCI bus INTx
lines. This is yet again the philosophy that "All The World Is DOS Legacy
Devices" and too bad about real OSes that don't need or want ISA side-band
IRQ routing.

I don't know what the "right" value is to place in the Interrupt Line
Register is for a PCI-R2 bridge is; I suppose a non-zero value is better
than a zero value, so that at least the BIOS on x86 systems will assign the
bridge a PCI INTx-to-ISA-IRQ routing.

I also don't think that it makes a difference whether the bridge supports
only R2 cards or R2/CardBus cards (like the TI-1130 for example) - both
cases still potentially require PCI INTx line resources.

mike
vH7