Subject: Speed hacks....
To: None <email@example.com>
From: Dave Burgess <firstname.lastname@example.org>
Date: 12/24/1994 12:50:52
I have been playing around with -current a little bit and I have found
an interested problem with my hardware and the fast noop option.
generic two port serial card with VT-102 terminal hanging from it.
When I enable the fast no-op option in the kernel, at some point in the
future, something will happen that will cause my serial terminal (on
tty01) to hang and lock. The port is wedged tight in some kind of serious
hardware way. There is absolutley nothing I can do to recover from the
lock except power off and power back on.
While it is obviously a problem with my com port, I have noticed that
disabling the fast no-op options keeps it from happening. I haven't
really looked into it closely, but I expect that if I did, I would find
something in there that relies on no-ops taking more time than they do
with the fast no-op option.
I used to have a problem like this that was resoved by improving the wd
driver interrupt latency. That fixed most of the problems, but it still
happens from time to time.
The exact description is that I was running 'top' on the terminal, and
had been for about 2 days. During a period of high activity (several
com overflows, some disk activity, etc.) the terminal just froze in the
middle of an update. I killed the process as soon as I got home (on
Wednesday). It is now Saturday, and here are the processes:
147 01- IEs 0:01.33 (tcsh)
420 01- IE+ 27:27.88 top
I have tried every trick on the book (from ^Q through powering down and
up, to ioctls on the port trying to revive it) and have had no luck
getting it back.
Power cycle always fixes it ;-)