Crash dumps (was: Re: 3.0 wishlist Was: Overview of 2.2.x goals?)

Werner Almesberger (
Tue, 20 Jan 1998 19:38:30 +0100 (MET)

Benjamin C.R. LaHaise wrote:
> If people really care about doing post-mortem analysis, modify the
> boot-loader to capture an image of RAM after reboot.

I did this once with LILO and found to my great disappointment that the
BIOS would reliably garble the memory on every reset. I've tested this
on three very different PCs (AMI, Award, and IBM(?) BIOS), so it seems
unlikely that there are many machines around which leave useful
information through a reset. (I used the reset button for those tests.)

> using the old 0x1234 warm reboot trick should avoid wiping the memory.

I've tried now with reboot=w,b and executing the reboot command,
and the (AMI) BIOS seems to ignore this completely and still happily
clears all my memory. Did anybody get the BIOS to _not_ interfere ?

> IMHO oops reports tend to be enough to get one looking in the
> right direction.

I think the main use for crash dumps is for remote support. Asking for
a single big file is likely to be easier than asking for a few dozen
items from /proc and such, which may also have changed since. Also, a
crash dump will contain the last few precious kernel messages.

- Werner

 / Werner Almesberger, DI-ICA,EPFL,CH /