Re: [OOPS] repeatable 2.4.8-ac7, 2.4.7-ac6 [I] just run xdos

[ebiederm@xmission.com (Eric W. Biederman)]

Paul <set@pobox.com> writes:

> "Eric W. Biederman" <ebiederm@xmission.com>, on Mon Aug 20, 2001 [12:09:27 AM]
> said:
> 
> 
> > If you can rule out X stracing dosemu might be of some help.   The
> > challenge now is to track down what dosemu is doing that is triggering
> > the problem. 
> > 
> > As an interrupt handler is where the oops is occuring.  Finding an
> > immediate cause and effect could be tricky.
> > 
> > Eric
> 
> 	Dear Eric;
> 
> 	Ok. I oopsed/locked the machine running 'dos' in a vt,
> without X in single user mode. Then I did it again, stracing the
> session.  Unfortunately, the fs was left in such a state, that
> fsck completely chucked the logfile out. Then I booted 2.2.18, and
> tried.  I could not make it oops.
> 	I need to setup a test machine to persue this farther, as
> locking and fscking my main box is no fun:) Ill try to get that
> strace...

O.k.  That rules out all kinds of things.  What is interesting at first
glance is that a) Every oops has been in an interrupt handler. and
b) It is never remotely at the same location.

I'm beginning to suspect there is some kind of hardware problem.  But it
is very weird.  I wonder if dos 6.2 somehow tickles a bug in the media GX
processor.  Though my top canidate is probably the lazy state switching
introduced in 2.4.  I know there were some problems with the ldt that were
fixed, and there might be another case out there.  But I'm just guessing
in the dark.

If you can reproduce this on a second machine that would definentily help.

Eric