[13698] in Athena Bugs
Re: sun4 7.7R: sparc classic
daemon@ATHENA.MIT.EDU (John Hawkinson)
Fri Jul 28 15:26:01 1995
Date: Fri, 28 Jul 1995 15:25:54 -0400
To: Brett David Rosen <bdrosen@MIT.EDU>
Cc: bugs@MIT.EDU
In-Reply-To: "[13697] in Athena Bugs"
From: John Hawkinson <jhawk@MIT.EDU>
> What were you trying to do?
> Drop into rom mode, run probe-scsi, then hit
> go to return back to X.
This is not guaranteed to work.
In fact, it is never guaranteed that the machine will be in
any sane state if you L1-A and then ``go'', but it's _usually_ safe.
Executing operations in the OpenBOOT PROM that perform IO or
change memory are particularly prone to problems.
It is possible that the machine might be more receptive to this if
you had a quiescent filesystem, and ran sync a few times before
hitting L1-A.
> What's wrong:
> The machine froze. I have reproduced this 4 times.
> However when I used a Sparc 20 or a Sparc 5 instead of
> a Sparc Classic, it did not freeze.
The Classic has a different version of the OpenBOOT prom, and it is
probably more susceptible to this problem than the 5. It's also possible
that that version of OpenBOOT is trashing some registers or memory when
this happenss.
If you really want this to work, I would suggest two courses of action:
(1) Do this while the machine is not running X, and figure out
why it actually hangs (if it continues to do so). You might
find the output of the ``ctrace'' command in the prom useful
here, along with adb if you can make the kernel panic.
(2) Try and see how the sparc 5 and 20 are different. Use
``see probe-scsi'' to examine the FORTH code used and compare
them. You might also compare the values returned by the
``.registers'' command before and after you run probe-scsi.
It might be interesting to see if the same problem
affects probe-scsi-all, as well.
There is some poor Sun documentation on the OpenBOOT PROM available
in E40 (Sun OpenBOOT Programmers' Manual or somesuch).
--jhawk
