> Some applications need to be able to access the ISA I/O and ISA memory spaces.
> On Intel, this is easy because you know where those spaces are (at 0 :-).
>
> On other architectures that have memory mapped I/O, things are more complex
> because the addresses are not fixed. They may even be different for different
> machines within the same architecture. E.g. on Linux/PPC, this is the case for
> the PReP and CHRP platforms. So the application has to be informed of the base
> addresses to mmap.
>
> One solution would be to add 2 lines to /proc/cpuinfo, like
>
> isa_io_base: 0xf8000000
> isa_mem_base: 0xf7000000
>
> (for CHRP LongTrail).
>
> Another solution (IMHO better) would be to add /proc/bus/isa/io and
> /proc/bus/isa/mem. On machines with memory mapped I/O, you can mmap() these, on
> ordinary PCs you have to use read/write. The advantage of a /proc/bus/isa is
> that you can add ISA PnP (yuk) device identifation later (cfr. the other
> /proc/bus/ drivers).
>
> Or make them real memory devices (char major 1): /dev/isa_io and /dev/isa_mem?
>
> What do you think?
It would be better to solve these issues for all buses at once. I plan to
include such thing in my resource manager code I'll work on in those golden and
long-awaited 2.3 days. I'd also like to add some basic I/O support to glibc or
create a I/O library, so that processes needing raw I/O will be portable at
least across different Linux architectures.
Have a nice fortnight
-- Martin `MJ' Mares <mj@ucw.cz> http://atrey.karlin.mff.cuni.cz/~mj/ Faculty of Math and Physics, Charles University, Prague, Czech Rep., Earth "DOS never says "EXCELLENT command or filename"..."- 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/