<tm_wanka@earthling.net> (Thomas Michael Wanka) writes:

> On 16 Nov 2000, at 9:44, Nathan J. Williams wrote:
> 
> > PCI devices and
> > their drivers *must* be able to share interrupts. 
> 
> Hi,
> 
> could it be that IRQ sharing was introduced with PCI 2.x spec?

I suppose it's possible. PCI 2.0 is pretty old, though. I don't have a
timeline of PCI in general, or a sense of what, if anything, used PCI
1.x. 

> In fact most hardware manufacturers will ask to assign an own IRQ 
> to their cards if called for help when there are problems with their 
> products. 

It is still possible for hardware manufacturers to be bozos.

As Manuel pointed out, a board/chip manufacturer can make life
annoying by making it very hard or impossible to tell if it catually
generated an interrupt, which does mean that it can only really work
well with its own IRQ. Such a device is broken - it won't work in
perfectly valid PCI arrangements - but that never seems to stop PC
hardware makers. The fact that most PCs have ~4 IRQs for plug-in PCI
cards means that such devices will work on many machines.

But they're still broken.

        - Nathan