Re: small note to GENERIC for isa dependency

On Tue, 20 Apr 2004, Bruce M Simpson wrote:

> On Tue, Apr 20, 2004 at 06:03:27PM +0400, Roman Kurakin wrote:
> > >The problem is that ISA and the i386 architecture are still intimately
> > >involved. It's probably hidden away as something called LPC on yours. :-)
> >
> > I only want to say that this is normal wish to remove device "isa" as a
> > "hardware missing in my modern box", cause I open it and see nothing
> > except PCI slots.
>
> Appearances can be deceiving. I can understand the source of your confusion.
> Let me attempt to clarify.
>
> ISA is there in some form. If your machine has legacy hardware of any kind,
> such as an AT-PIC programmable interrupt controller, a PS/2 mouse port, or
> even a keyboard controller, you need ISA support, even if the devices are
> probed via ACPI.
>
> Many machines have PCI slots and not ISA slots, but ISA is still present;
> many VIA and Intel based motherboards have a SuperIO chip connected to
> the south bridge via what is known as the LPC (Low Pin Count) bus, which is
> covered by Intel specifications; this is essentially an ISA bus with less
> pins.

E.g., on a new system (A7N8X-E) with no isa slots, pciconf -lv says:

% isab0_at_pci0:1:0:	class=0x060100 card=0x80ad1043 chip=0x006010de rev=0xa4 hdr=0x00
%     vendor   = 'NVIDIA Corporation'
%     device   = 'nForce MCP2 ISA Bridge'
%     class    = bridge
%     subclass = PCI-ISA

> Only a truly legacy-free machine (and most machines which are i386 and
> claim to be legacy-free are not) could live without the isa bus support
> in FreeBSD, and to the best of my knowledge, these don't exist yet.

If isa were properly optioned, then omitting it would omit support for all
the hardware that depends on it, not just 8087 through 80387 compatibility
mode in npx.  This system wouldn't work, starting with interrupt handling
becauses its APIC doesn't work.

Bruce