> In message <Pine.LNX.4.05.9811252223010.11212-100000@tahallah.demon.co.uk>,
> Ale
> x Buell writes:
> +-----
> | On Wed, 25 Nov 1998, Brandon S. Allbery KF8NH wrote:
> | > Linuxers, marred only by small issues such as (IIRC) ATI and S3 video
> | > cards vs. COM4. And it's right smack in the middle of the list of
> |
> | Ah yeah, the infamous COM4 bug with S3 chipsets. Seems some fuckwit used
> | the usual COM4 I/O port as one of its I/O registers. Ahh, that particular
> | chipset designer should have been taken out and made to use Microsoft
> | products in eternity (with a Windows 95 Beta!).
> +--->8
>
> Not exactly.
>
> The ATI and S3 chipsets started out as enhancements of IBM's 8514/A video
> card. But the 8514/A was designed for Microchannel, which addresses the
> entire I/O space. So the 8514/A used several registers with addresses
> 0xX2e8, for several values of X.
>
> The ISA bus only has 10 I/O address lines, so I/O addresses above 0x03FF
> effectively wrap around. The 8514/A ports ended up aliased to 0x02e8....
Even the first ISA bus had 20 address lines (although x86 I/O instructions
use only 16). Probably the reason to decode only 10 bits was rather that
the address decoder could be implemented in a simple 16 or 20 pin DIL chip
(PAL or whatever) on what was then an expansion board, because at the time
anything with more than 20 pins was huge and expensive.
[I have just checked the schematics, in the IBM Technical Reference,
Copyright 1981, Revised Januray 1983 that I've found: all address
pins are demultiplexed simultaneously from the 8088 and sent directly
to the ISA bus. All 20 bits are simultaneously valid.]
Gabriel.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/