Crash a RISC machine from user-mode code:

[Bruce Worden]

In article <481 at demott.COM> kdq at demott.COM (Kevin D. Quitt) writes:
>In article <1990Aug13.053147.11714 at laguna.ccsf.caltech.edu> bruce at seismo.gps.caltech.edu (Bruce Worden) writes:
>>In article <477 at demott.COM> kdq at demott.COM (Kevin D. Quitt) writes:
>...
>>
>    Please read what I said.  I did *not* claim they were only useful for
>single user mode, just that it was one condition where it could be used.
>Another situation is where you have users of (well debugged) applications,
>which appears to be your situation.

You may assume that to be our situation, but it is not.  I never said
anything about running "well debugged" applications.  Applications are 
developed here all the time, and, as I said before, not by professional
programmers.

Now YOU read what YOU wrote:

>>>                                              ....  But when you try to
>>>stress the system, it's going to fail.

This bug existed on Suns (for example) for at least a couple of years (it 
was only fixed a few months ago with release 4.1 of the OS.)  Surely in that 
time enough people stressed them to make it apparent that they were "broken 
systems."  You would think the word would have leaked out somewhere that
they were "failing" (like maybe one of the dozens of users groups or the
various news groups that deal in some capacity with Sun?)  Don't you think
this bad press would hurt Sun's sales?  Don't you think we would have 
heard about it?  Don't you think Sun would have fixed it sooner?  I do. 
Maybe it is just not happening.  (The point of my original posting, which you 
chose to ignore.)  So either very few people are stressing their Suns (not 
likely), or you are just plain wrong (more likely.)

The only legitimate issue of discussion that the crashme program brings 
up, and lord, the one I wish it was limited to, is that of system
security.  If the crashme program crashes your system, talk to an expert,
it represents a potential security problem.
 
>If it is a multi-level privelege system, and the hardware can fail to
>provide the necessary protection to the operating system, then the hardware
>is broken. Period.  

How interesting.  Good thing it's not a hardware bug (at least on Sparc).

--------------------------------------------------------------------------
Bruce Worden                            bruce at seismo.gps.caltech.edu
252-21 Seismological Laboratory, Caltech, Pasadena, CA 91125
--------------------------------------------------------------------------