[ On , April 21, 1999 at 11:10:27 (-0700), Wolfgang Rupprecht wrote: ]
> Subject: Re: 256 MB RAM?
>
> Its been a while, I can't recall what they did, but it might be a
> smart think to do if some modern OS took up that ball and ran with it.

I think the only smart thing to do with hard ECC errors (beyond not
panicing, that is) would be keep stats on the errors/page (perhaps even
just 2 or 3 bit counters per page), and only map out a page if it rolls
the counter over.  It would also be a good idea to have a hook that lets
the administrator add the page back into the pool, perhaps after putting
the cover back on the case, to see if it still suffers errors, as well
as a hook to list the non-zero counters (even though you'd probably
still have all the syslog entries as well).

If there's a soft-ECC (i.e. corrected error) interrupt it might even be
nifty to try and co-ordinate previous soft errors with hard errors,
though for this to be of much pracical use you'd have to do it on a
per-component (SIMM/DIMM) level, because then you'd maybe know better
when a problem is becoming critical and parts need to be replaced.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>