tony <spamtrap@[EMAIL PROTECTED]
> wrote:
>
>> You don't use memory mapped I/O to talk to an IDE controller.
>
>Say if a malware aims to change the default beahviour of OS, can it
>map the registers of IDE controller to main memory and hook page fault
>handler to subvert the return results, or it is not possible at all?
Not possible. I/O space and memory space are very different, and you
can't
reroute one to the other.
>Sorry another related query, (though it is not concerned about x86),
>processors like ARM,PowerPC,Intel Xscale that employ memory mapped IO,
>do they have a certain fixed memory map for IO and all devices therein
>(just like standard IO Address Map of x86 PC) or it is not
>standerdized?
Remember that the "standardized" memory map is not x86 -- it's from the PC
architecture. There's nothing inherent in the x86 that requires it.
Similarly, there's nothing in the ARM or PowerPC that demands a particular
memory map. It's all up to the specific platform.
Also remember that, because of PCI, the memory map isn't nearly as fixed
as
it used to be.
--
Tim Roberts, timr@[EMAIL PROTECTED]
& Boekelheide, Inc.
|
