Geert Uytterhoeven wrote:
> On a related issue: using request_mem_region() to request the A0000 region
> works on PC only. On other machines ISA memory space doesn't start at CPU
> physical address 0. On my CHRP LongTrail /proc/iomem says:
[...]
> Hence you need an isa_request_mem_region() function to fix this.
>
> A small patch (untested in real life!) that takes into account the offset for
> ia32 and PPC is below.
Patch looks good, I agree with the general trend that isa I/O will need
its own functions separate from PCI I/O for situations like these.. As
ISA ages and PCI advantages, distinction between the two will become
increasingly more useful.
> Another related issue: on non-PC boxes with memory mapped I/O, it's quite
> difficult to access ISA I/O (and memory) space from userspace since you need to
> know at which offset it resides, which of course depends upon the machine type.
> This is a problem for XFree86 (the hardware banging user app :-) on at least
> PPC and MIPS.
>
> I see two possible solutions:
>
> - Create /dev/isaio and /dev/isamem so you can mmap() ISA I/O and memory
> space.
>
> - Create /proc/bus/isa/map which contains e.g.
>
> | IO f8000000 f8ffffff
> | MEM f7000000 f7ffffff
>
> so user space can mmap() those ranges using /dev/mem, or fall back to
> I/O instructions (inb() and friends on ia32. Note that inb() on PPC needs
> the base pointer isa_io_base, cfr. include/asm-ppc/io.h) for the I/O part
> if it doesn't exist..
I favor the second one since it puts more work in userspace, but there
may be a need for special properties which only a /dev/isaio or
/dev/isamem can provide. I defer to any Upper Penguins on this one :)
Regards,
Jeff
-- Jeff Garzik | Liberty is always dangerous, but Building 1024 | it is the safest thing we have. MandrakeSoft, Inc. | -- Harry Emerson Fosdick- 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/
This archive was generated by hypermail 2b29 : Thu Jun 15 2000 - 21:00:33 EST