1. The 3com driver is still giving error messages. I had mistakenly
written in a previous message that they had stopped. They
continue, but with reduced frequency:
Mar 8 11:40:11 borg kernel: eth0: Missed interrupt, status then 2011 now 2011
Mar 8 15:58:40 borg kernel: eth0: Missed interrupt, status then 2011 now 2011
Mar 8 16:10:59 borg kernel: eth0: Missed interrupt, status then 2011 now 2011
Mar 8 21:08:17 borg kernel: eth0: transmit timed out, tx_status 00 status 2000.
Mar 8 21:29:31 borg kernel: eth0: transmit timed out, tx_status 00 status 2000.
Mar 8 21:56:25 borg kernel: eth0: transmit timed out, tx_status 00 status 2000.
Each of these messages occurred under periods of extremely heavy NFS
usage.
The only machine I haven't had these kind of messages is the one with
an EtherExpress16, all the machines with 3c509's have gotten them. My
server just got upgraded tonight, and it has an SMC WD8013 as well as
3c509, no error messages so far, but the uptime is yet short.
2. There are a number of incompatibilities that make running binaries
built against 1.2 include files chancy. I've identified two
programs that are definitely broken by an upgrade:
XEmacs 19.13 and Procmail 3.11pre4.
I don't know what's happening with XEmacs, but a simple recompile and
relink seems to do the trick.
Procmail has problems with kernel file locking. The solution there is
to use .lock files instead. If you are using procmail as a local
mailer, definitely make sure that you aren't using kernel locking at
present. A recompile and relink is not sufficient to get it working.
Emacs 19.30 is probably broken too, but I haven't tested it in the
same fashion yet.
3. Sound configuration.
I still can't figure out how to configure for a SoundBlaster Pro or
SoundBlaster 16. I was able to get working sound on an SB16 by
running ``make config'' in the drivers/sound directory, and then
manually inserting the defines for SBPRO and SB16. This method is
not completely clean. Where formerly ``playmidi file.mid'' did
something relatively harmless, it now causes a system crash (the
process is unkillable, and an endless stream of error messages get
written to syslog. Playmidi -f works great and does seem to sound
better with the new sound driver (playmidi-2.3).
4a. Bootstrapping Kerneld.
The latest modules package will not compile completely with
/usr/include/linux pointing to a 1.2.13 kernel source tree. A
satisfactory solution seems to be to change the link before compiling.
I've modified the standard slackware setup here to have an extra set
of symbolic links. /usr/include/linux -> /var/linux, and /var/linux
points to the specific kernel source tree. This allows a
configuration where /usr may be mounted NFS, and allows multiple
kernel versions on the various client machines. It's worked for me.
4b. depmod -a gives an error message at boot.
I've placed the lines of code in the kerneld mini-howto into rc.S and
the call to /sbin/depmod -a always gives a ``Can't find .../modules.dep''
error message, even when the file exists. It doesn't seem to affect
how kerneld and modprobe work though. (Kerneld is a wonderful
invention).
The following are nits.
5. dnsdomainname is called during kernel building.
And gives an error message which I assume can be ignored if one is not
running a DNS locally. It hasn't seemed to hurt anything. Since all
of the hosts have FQDN as hostnames anyway, it doesn't do a whole lot
in the setup here.
6. The Slackware /etc/rc.d/rc.cdrom is broken.
It gives a whole stream of error messages when a cdrom isn't found,
mostly from kerneld. I haven't figured out yet what this file buys in
the first place.
As chaotic as the configuration is right now, it's still better than
some O/S upgrades I've made in the past, and it is definitely
remarkable considering the free nature of the software and the fact
that stable Linux is less than two years old.
Regards,
-- steve@miranova.com baur Unsolicited commercial e-mail will be proofread for $250/hour. Andrea Seastrand: For your vote on the Telecom bill, I will vote for anyone except you in November.