On Tue, Apr 20, 1999 at 11:09:10AM -0700, Michael Graff wrote:
> Jukka Marin <jmarin@pyy.jmp.fi> writes:
> 
> > What does NetBSD do in that situation?  Dump core and reboot?
> 
> My pentium machine with 128M ram seems to log it.
> 
> 	NMI port 61 a0, port 70 ff
> 
> It would be nice to have it do the right thing, that is, not only log
> it but try to correct it if it can be corrected (ECC) or die if it
> cannot (parity).

You should only get the NMI if it *couldn't* correct the error.  The
operating system doesn't get to "try to correct it" -- if the chipset's
ECC could correct it, it would have.  If it's telling you it couldn't,
you Just Lose... and you should find a new memory vendor, or stop rubbing
your shoes on the carpet while you build machines. ;-)