Re: crashme crashes 2.0.20 with guarantee :-(

Michael J. Micek (mmicek@muddcs.cs.hmc.edu)
Fri, 27 Sep 1996 19:06:43 -0700 (PDT)



Ages (like over a week) ago Linus wrote:

> Now, it would be interesting to have others running crashme too, and see if
> there is some pattern to failures (specific configurations and/or CPU's or
> whatever). For example, crashme _always_ succeeds to bring a machine down on
> old i386 machines (the popad bug is fatal in protected mode) regardless of
> OS.

Well I tried it a few days ago (after searching all over for
the program; finally found it in ftp.std.com:/pub/gjc/) and
it locked up my console. Specifically, I tried the example
in the man page:

crashme +2000 666 100 1:00:00

and it got to about 684 (after less than a minute) and froze
up. No response to keyboard or mouse activity was evident;
however, (I surmise) the update daemon woke up about 30
seconds later (there was a small amount of disk activity),
and so there was no damage to the file systems after the
required reset. So I wouldn't say it "locked solid". Also,
I don't have a way to telnet or a terminal to attach, so I
don't know if that would have worked.

This is with kernel 2.0.21 built for i386, on an Nx586 Pf110
motherboard with an AHA2940UW. I haven't had any trouble
with Linux, but I don't stress my system very hard. (I even
shut it down most nights. How uncouth!) I don't give any
guarantees that my system is configured properly, either...
For instance, the news that pty's should be major number 2
not 4 took me by surprise. I apparently have an obsolete
version (1.1) of sysklogd. (There was no info in the logs.)
Basically it's Slackware 3.0 + linux/Documentation/Changes.
(Hm... should update binutils, too.)

What I really want to know is: Does the Nx586 have the
"popad bug", or should I keep trying to narrow down what the
problem is? (I've only run crashme twice; once in X and
once from the VGA console with identical results.)

-- 
Michael J. Micek, peripatetic philosopher. (currently) mmicek@muddcs.cs.hmc.edu