Subject: Re: hangs with macppc in -current?
To: Erik Bertelsen <erik@mediator.uni-c.dk>
From: Erik Bertelsen <erik@mediator.uni-c.dk>
List: port-macppc
Date: 01/30/2000 22:21:29
On Sat, Jan 22, 2000 at 09:32:49AM +0100, Erik Bertelsen wrote:
> On Fri, Jan 21, 2000 at 09:30:42PM -0500, Dan Winship wrote:
> > For a long time now, my macppc box has been wedging solid every now
> > and then. I chalked it up to "port-macppc flakiness". After a while I
...
> > -- Dan
...
> I've seen similar events on my G3 233MHz Desktop (first generation or so G3),
> 64MB RAM, running MacOS on the IDE disk and NetBSD on an internally mounted
> SCSI disk.
>
> The difference in my case is that when the machine dies, it is always about to
> open a new window in X (typically emacs or xterm, the X app may be running
> locally or via ssh on another machine). This typically
> happens after a few hours or days of uptime and some significant system activities.
> In my way of using the machine, this is often after updating the NetBSD sources
> when I want to take a look on what was updated since last time. A few times,
> it dies at the first xterm after system start, which is somewhat inconvenient
> because a restart with fsck does take its time :-( I've never seen the problems
> with non-X applications, e.g. it has not yet died when I was rsync'ing my
> NetBSD repository from another system.
>
> When the problem started several months ago, the machine just hang and had to be
> reset, but when I removed support for ddb, it just reboots. Nothing is logged
> about the cause in /var/log/messages or other places where I've looked.
>
> - Erik Bertelsen
>
> ps: maybe I should really find the cables and hook up a serial console and use
> that with ddb re-enabled.......
In follow-up to my own message:
My problem was not related to the kern_proc lossage that was fixed by Jason.
I have now hooked up my mac68k system as a console for my powermac and by setting
the openfirmware output device to ttya, the openfirmware output is displayed on the
other machine while it still uses the powermac's own keyboard. This setup allows
me to still use X on the powermac. Actually if ddb is activated, then input will
be taken from the serial console...
Without options DDB in the kernel, it will eventually die with
panic: trap
syncing disks... 14 14 4 done
dumpsys: TBD
rebooting
Then I added options DDB,DEBUG and after a while, it died again:
Stopped in Xmacppc at 0x1f1518: lwz r0, r1, 0x14,
db> t
at ?+ffffffff
at ?+ffffffff
saved LR(0x6db6db69) is invalid.
db>
db> ps
PID PPID PGRP UID S FLAGS COMMAND WAIT
2351 2350 2351 269 2 0x4006 emacs
2350 2349 2350 269 3 0x4086 bash wait
2349 1 196 269 3 0x4184 xterm select
20312 20310 20312 269 3 0x4086 bash ttyin
20310 1 196 269 3 0x4184 xterm select
220 219 220 269 3 0x4086 bash ttyin
219 1 196 269 3 0x4184 xterm select
211 1 196 269 3 0x4084 oclock select
208 1 196 269 3 0x4084 smproxy select
206 1 196 269 3 0x4084 twm select
201 200 201 269 3 0x84 ssh-agent1 select
200 196 196 269 3 0x4084 xsm select
196 174 196 269 3 0x4084 sh wait
187 1 185 0 2 0x84 xconsole
This happened once more, with the same address, but a different set processes
running.
I'm now current at Friday's sup, but the problem has persisted for several months
now. It always happens when opening a new window in X (client either local on another
machine -- except once, when it happened while resizing a window. If I don't
create new windows, it'll normally keep up for days, while rebuilding NetBSD or
doing a complete make World of X, just to keep it working...
Any ideas of the cause of the death or some ddb exercises that I could perform?
- Erik