>>>>> On Sun, 17 Dec 1995 15:08:04 +0100 (MET), Thomas Bogendoerfer <tsbogend@bigbug.franken.de> said:
Thomas> Another question: I have some problems with the NCR on my
Thomas> Noname. Under heavy disk-load I get memory checks with the
Thomas> reason: Reason: access to non-existent memory (long frame):
Hmmh, could you provide all the information that's printed with the
machine check? Also, the kernel-version would be important. I'd also
be a good idea to build the kernel with -g so it's easy to lookup the
code that was hit by the fault.
Thomas> There are two exc_addr, where this happens. Both are in the
Thomas> NCR driver, but I couldn't find a reason, why this happens.
Actually, I saw this a long time ago on my Noname. Are the two
addresses by chance somewhere close to inb() or some such call? Are
you booting via SRM or MILO? If memory serves, my problem went away
once I disabled parity on the PCI bus. The SRM has an option to
disable PCI parity. I'm not sure what MILO does but I never saw this
problem since I switched to MILO, so I assume it turns off parity by
default. The problem seems to be that many (older?) PCI devices do
not correctly implement parity generation.
--david