Re: Linux and memory ECC on i386

Ralf Baechle (ralf@julia.de)
Fri, 6 Sep 1996 21:01:15 +0200 (MET DST)


> (Semi-off topic, but I would appreciate any help...)
>
> What does Linux do with memory parity and ECC errors on
> the i386 architectures? I find no references to same
> in the 2.0.x kernel sources. (There are references for
> some of the non-i386 architectures.)

PCs generate an NMI when a memory parity error is detected.
The only think that Linux/i386 currently does is to print:

Uhhuh. NMI received. Dazed and confused, but trying to continue
You probably have a hardware problem with your RAM chips or a
power saving mode enabled.

and trying to continue. With the little information that is available about
the source of the problem on PC board this is about what can be done.

> The hardware in question is an Intel PentiumPro with the
> Natoma (82440FX) chipset (Intel VS440FX motherboard).
> The technical specs say that ECC is supported if parity
> memory is used.
>
> Some of these hardware docs are also ambiguous.
> Would someone more knowledgeable than I be willing to go
> off-mailing-list to help me clear this up?

ECC is mostly a hardware topic for designers of l1/l2 caches (At least MIPS
caches do use ECC), system buses and memory. During normal operation no
software interaction is needed and thus Linux doesn't need to support it.

Only in the case that the ECC hardware detects an uncorrectable error
system software gets involved. And at that point it's usually to late
anyway ...

Ralf

-- 
A weird imagination is most useful to gain full advantage
of all the features - manpage of amd(8).