Re: 3.0 wishlist Was: Overview of 2.2.x goals?
Dan Hollis (goemon@sasami.anime.net)
Tue, 20 Jan 1998 10:13:09 -0800 (PST)
On Tue, 20 Jan 1998, Geert Uytterhoeven wrote:
> On Mon, 19 Jan 1998, Jordan Mendelson wrote:
> > On Mon, 19 Jan 1998, MOLNAR Ingo wrote:
> > > an OOPS 100% tells you where and why the oops happened. Now, to determine
> > > what the real reason for the oops was is quite complex, and in many cases
> > > a crashdump doesnt help much. I would say the existance of a crashdump
> > > hinders development more than the nonexistance of it ...
> > My problem with OOPs's is that usually, when one happens.. it is > 25
> > lines, and scrolls off the screen before the machine is at a dead halt. Of
> > course, when this happens, that oops sure isn't going to be written to the
> > logs.
> Linux/m68k on Amiga has a solution for this: it allocates a piece of Chip RAM
> and stores all kernel messages in it. After a reboot, you can use a special
> program to find them.
This actually presents a fascinating possibility : for systems with enough
memory, it might be possible to save a 'crashdump' into some part of ram
(of course with special consistency checks like a CRC32) which is then
picked up after a reboot and saved to disk for postmortem analyis.
This completely nullifies any arguments about "peripherals might not be in
a sane state to do a kernel dump". Of course, it may not be applicable to
machines with limited memory. But its better than nothing.
-Dan