[ On Sat, November 29, 1997 at 20:26:23 (CST), Chuck Cranor wrote: ]
> Subject: Re: so, does arch/sun3/dev/memerr.c do anything or not????
>
> At one point I had some code that would talk to sun3/sun4 200 ECC memory
> boards.   however, i seem to have lost it somewhere.   but... i think
> i sent Jason a copy of it at some point, so it may be possible to fill
> in this part of the kernel.

That would be good if Jason still has it.

Now that I look closer at this issue I note that there is a tiny bit of
code in sun3/dev/memerr.c that says:

                /*
                 * XXX:  Some day, figure out how to decode
                 * correctable errors and set ME_ECC_CE_ENA
                 * here so we can log them...
                 */
                mer->me_csr = ME_CSR_IENA; /* | ME_ECC_CE_ENA */

This of course suggests the comment I quoted previously from
sun3/sun3/machdep.c was misleading since this enabling is actually done
(or planned) in memerr.c

Currently there is no code in memerr.c to decode the board/chip from the
address, though this shouldn't be too difficult, and from what I can see
it gets a whole lot easier if you're allowed to read the SunOS
<sun3/memerr.h> and <sun3/eccreg.h> headers.

I think I'm going to turn that CE_ENA flag on in any case as at least
then I'll have them logged even if it doesn't tell me which exact chip
had the error....

I may even peek in the SunOS headers to see how to detect which board
the error was on so that I can log at least that info too.

-- 
							Greg A. Woods

+1 416 443-1734      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>