Re: Legacy Addresses

[havard@brooktree.com (havard scott)]

> From pierce@hogranch.com Fri Jan 17 14:01:46 1997
> 
> > 3C0h to 3DFh. For other classes, such as a Generic 8237 DMA
> > controller, the PCI specification doesn't specify which addresses
> > are reserved, but I've always assumed that it reserves the
> > standard addresses of that part. Am I wrong?
> 
> Actually, I don't believe the 8237 DMA controller IS a PCI device... its behind
> the ISA bridge, so is a legacy ISA device. Its existance (and fixed resources)
> is assumed by any PC bios and operating system. Ditto the floppy, keyboard,
> RTC, COM ports, etc. They are all ISA devices and don't even EXIST in PCI
> config space.
> 
> Far as I know, the only legacy PCI devices in common use are VGA and IDE ports.
>  I'm probably forgetting something.
> 
> -jrp
> 

These devices might not typically be put on a PCI bus, but I thought that
the idea of the class type field in PCI config space was to allow any 
PCI device to look like one of these devices and grab their address space. 
There's no inherent reason why you wouldn't want to put an ISA Legacy
device on the PCI bus. If you take a look at the PCI 2.1 specification,
you'll see lots of Legacy device classes defined.

For example, we have put legacy audio on the PCI bus, even though, to
this day almost everybody else still puts it in on the ISA bus. Unfortunately
for us, there wasn't a PCI class for it. This meant that we had to
be careful about having another Legacy audio (i.e. SoundBlaster 
compatible) device in the same system. It also meant that we couldn't
operate behind a bridge.

As for the other standard Legacy devices such as communication controllers
or DMA controllers, I'm not sure why you would put this functionality
on the PCI bus, but it seems to me that PCI allows you to look like one
of these devices and grab their IO/Memory space. Whether or not the original
device was on the ISA bus seems irrelevant. Maybe I'm missing something.
øàÐ