I would like a user-mode program to have access to my RAM in the area
from 4Megs to 20Megs, while linux itself is _not_ using (or even
acknowledging) that area (due to "mem=3900k" in the lilo command line).
I know this is possible somehow, as XFree86 uses linear mapping of some
graphic cards, but am absolutely clueless as to the HOW of it.
Any hints? Other than getting the source for XFree ;-) ? Or maybe even some
simple sample code?
The background of this question: I got my hands on 16 Megs of RAM from a
junked graphics computer. Unfortunately, those are 32 ZIP chips, soldered to
a PCB, and are also 80ns RAMs. I solved the first problem by stringing a
bunch of wires from that PCB to SIP-to-SIMM-adaptors, and (after adding some
wait states here and there in the BIOS), the boot-up memtest is happy as
well as a DOS based mem tester.
Linux, however, dislikes this memory tremendously and gets lots of "Oops"
messages from the kernel. So rather than either junking the whole project or
trusting anything DOS based, I'd like to throw some test patterns at it
myself, and see where the errors occur and of what kind they are. Only ---
linux won't let me (I have hacked around in the kernel for the last few
hours, trying to get those pages mapped so a kernel space routine I added
into an unused driver could do the test for me, but all I get is "Ooops"
over and over :-( )
Thanks for any help,