Subject: 2 new problems with 1.3.3 on SPARC
To: None <port-sparc@netbsd.org>
From: Greg Earle <earle@isolar.Tujunga.CA.US>
List: port-sparc
Date: 01/06/1999 05:57:51
Well, I finally made the big leap to 1.3.3 (from 1.2.1) on my work SS20/71.
I hit a snag during the upgrade because my /usr didn't have enough free
space in it. (Had an AFS cache in /usr/vice taking up lotsa room.) Suggest
that the install scripts check for sufficient disk space before doing the
un-tar'ing of the .tgz files, if that's possible.
I've hit a couple of problems in 1.3.3 that weren't in 1.2.1 so far.
(1) If I exit out of the TriTeal TED CDE I'm running, the machine crashes
with a Watchdog Reset. Oi! This is with the GENERIC_SCSI3 kernel, btw.
I haven't recompiled a new custom 1.3.3 kernel yet. The only major
difference between my 1.2.1 setup and now is that the 1.3.3 setup uses
the X11R6.3 server et al. that's provided in our x*.tgz files for 1.3.3.
Since it Watchdog Resets, there is no core dump for "savecore".
I'll try to build a DDB-enabled kernel when I get a chance, but in the
meantime, I'm amazed that quitting a user-level process could get the
machine to reliably cause a Watchdog Reset. (It's happened 3 times now,
each time I tried to log out of TED/CDE.)
(2) I'm trying to do something like the following:
foreach i ( machine1 machine2 machine3 )
? echo '<root password>' | rsh $i "sudo sh /etc/init.d/syslog start"
? end
to restart some failed "syslogd" daemons on some Solaris 2.6 hosts.
This has always worked fine for me in the past. But now, it completes
the first iteration, and then it stalls and I find the remote commands
have turned into unreaped Zombies:
netbsd4me# ps -auxwtp2
USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND
earle 2816 0.0 0.0 0 0 p2 Z+ 5:03AM 0:00.00 (rsh)
earle 2817 0.0 0.0 0 0 p2 Z+ 5:03AM 0:00.00 (rcmd)
earle 389 0.0 0.3 604 208 p2 Is 7:59PM 0:00.84 -usr/local/bin/tcsh
earle 2813 0.0 0.4 132 220 p2 I+ 4:50AM 0:00.06 rsh machine1 sudo sh -x /etc/init.d/syslog start
earle 2815 0.0 0.3 356 160 p2 I+ 4:50AM 0:00.03 rcmd machine1 sudo sh -x /etc/init.d/syslog start
The child processes have not gotten reaped for some reason. Most commands
run this way exit normally, so it seems to be something endemic to
running "sh /etc/init.d/syslog start" on the remote side. But this didn't
used to happen with 1.2.1 as far as I know, and if I do a "sh -x" instead,
it shows that the remote shell command runs "syslogd" and then does an
"exit 0", as expected. Yet the commands don't return, and I have to hit
^C (twice!) to get the shell to skip to the next entry in the loop.
Any ideas on either problem?
(On a completely different note, does anyone have a cheat sheet on how to set
up a dial-in/dial-out modem arrangement with "pppd" under NetBSD/SPARC 1.3.3?)
Thanks in advance,
- Greg