> On Mon, 24 Nov 1997, David S. Miller wrote:
>
> > The point is that in this case the kernel did output a legitimate
> > address and it was not shown in the bug reports.
>
> But what can we do for the future so that we do get a valid address?
> (Somehow I don't think there's going to be a widely used release of gcc
> that fixes this problem anytime in the near future.)
klogd used to translate the oops addresses automagically. Unfortunately
this feature broke with 2.1 and noone has it fixed yet. And it doesn't really
help for panics that halt the box. But I don't think that the lookup should
be put into kernel space. Other unixes solve the problem with a kernel crash
dump to the swap partition, but we _don't_ want that in Linux, really.
(ever watched when a SCO box dumped its kernel over your /home partition?)
So overall the current situation isn't that bad.
-Andi