[2724] in Athena Bugs
problems with the VS2 kernel on movies (a vax parallax)
daemon@ATHENA.MIT.EDU (John D. Kubiatowicz)
Mon Aug 7 18:03:08 1989
To: bugs@ATHENA.MIT.EDU
Cc: probe@ATHENA.MIT.EDU, raeburn@ATHENA.MIT.EDU
Date: Mon, 07 Aug 89 18:02:34 EDT
From: John D. Kubiatowicz <kubitron@ATHENA.MIT.EDU>
As you know, the 6.3B kernel on movies has been panicking rather regularly
with a "reserved operand" error. Since we haven't seen this behavior
on other VS2 machines, we assumed that the problem was with the new
parallax driver code with was distributed with 6.3B.
It appears to be something else. Namely, zonker, which is also
a vax parallax, has NOT crashed in this way. Further, I seem to
have fixed movie's problem by BACKING OUT THE RECENT CHANGES TO
LOCORE.S AND EMULATE.S, which effect ONLY the VS2 kernel. It is
quite possible that movies is stressed in way which other 6.3B VS2
kernels are not (it always has 2+ people logged into it).
While I would expect the above changes to cause "illegal instruction"
problems (if anything), there may be some complicated interaction.
These changes HAVE come between the 6.2 and 6.3 releases.
We should probably wait a day or two more to make sure that movies
is cured, but it seems to be at this time...(Has been up for 2.5 days).
--KUBI--
p.s. I would attempt to isolate the problem further, but movies is a
regularly used machine, and I hate to bother them...
p.p.s I don't understand what was happening...
p.p.p.s If you end up recompiling the 6.3B kernel, I have a change to
plxconstty.c which should permit the kernel to dump core (has to do
with console use when the processor is not mapping).
