Messages in this thread | | | Date | Tue, 26 Jan 1999 12:59:50 -0800 (PST) | From | Linus Torvalds <> | Subject | Re: Big Fix for 2.2.1 |
| |
On Tue, 26 Jan 1999, Leonard N. Zubkoff wrote: > > If the default setup is only correct for PAGE_OFFSET == 0xC0000000, then why > did you permit the comment in page.h telling precisely how to change it to a > value that violated this assumption?
I can happily remove the comment if that is your problem.
That comment was written in the 2.1.x series, _long_ before anybody worried about stability. At that point I might have accepted patches to various drivers, trying to get the issue worked out.
As it is, the only working values for PAGE_OFFSET are of the form (2^32 - 2^n), where n goes from 31 down:
0x80000000 0xC0000000 0xE0000000 0xF0000000 ...
Of these, obviously only 0xC0000000 and 0x80000000 are actually all that useful (the others give slightly more user space, but not enough to matter, and they obviously seriously cut down into the amount of physical memory you can have - exactly the opposite of what you want).
> I had assumed that the legacy ISA region was also mapped at PAGE_OFFSET, in > which case the above readb would still work properly. I should have noticed > that the change to traps.c invalidated this assumption.
The legacy ISA region actually does happen to be mapped at PAGE_OFFSET.
But that just happens to be the case, and it happens to be the case only for an x86 implementation. It is NOT the case for any other Linux PCI implementation I know of.
Linus
- 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/
| |