Greg's diary--October 2001

Monday, 1 October 2001	Echunga
		Top of page
		next day
		last day

Just in time, last night, we discovered that today was a public holiday (Labour Day). That didn't help much; spent most of the day trying to catch up on my mail. Why is it that when I'm away, so much mail worth reading carefully comes along?

Spent some time tidying up the Mike Smith Memorial Room. Much more needs to be done.

Tuesday, 2 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

The Telstra Internet bill arrived today. $384.75. Normally it's a little over $100; the rest is accounted for by Code Red and Nimda. High time to find a better solution.

More work in catch-up mode.

Wednesday, 3 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Still more catchup work. Some weeks I seem to get nowhere.

In the evening to band practice again. This is rather frustrating. It's been 5 weeks since I joined, and I still don't have all the music. We have a concert on Sunday, and assuming I get the music by then, I'll have to sight-read it.

Thursday, 4 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Finally got to do some real work, mainly thoughts and discussions about EVMS. There's a lot of stuff to think about there.

Decided to finally move the lemis.com class C network off Telstra, since Nimda was costing too much. That was interesting because the old IP address I had from Telstra was no longer available, but we got it done pretty quickly. As soon as things settle, I'll move it to the satellite link.

Diane Saunders in in the afternoon. Her computer had died a CMOS battery death, and so she used sydney, my Dell laptop, to write her newsletter. Again she had to “put up” with Emacs, and again she had no problems, though it did take me 5 minutes to tidy up the text and format it with groff. It's really interesting that beginners can use so supposedly hard to use editors without any help.

In the evening, we were supposed to go to Warrawong sanctuary, but we left it so late that we would have missed the beginning, so I didn't go. I'm sure there will be a next time.

Friday, 5 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

This was one of those weeks which seem to be nothing but work, but at the end of it there's nothing to show for it. I was too busy to go riding with the others this morning, but I didn't really get much done. On reflection, I suppose it's the higher mail load: in the five days Monday to Friday I received a total of 8928 mail messages.

One of those was from Jerry Dunham, saying that his domain (which I DNS host) had dropped off the Internet. On examination, it seemed that lemis.com had done so as well: the information on freefall.FreeBSD.org was nearly two weeks old, and I couldn't get a DNS query through. Ethereal showed the request come in and the reply go out again, but it never got back to freefall. Called up Telstra and logged a fault. Spent much time following the fault and found:

A number of secondary name servers were trying to load the zone from freebie.lemis.com, the secondary address for echunga.lemis.com which went away yesterday, so they were marginally out of date.
freefall (more correctly, hub.FreeBSD.org) was loading the zone from sat-gw.lemis.com, which was not loading from echunga, since I was planning to migrate it to primary name server as soon as I got my satellite routing sorted out, so sat-gw was out of date.
Jerry had not updated his name server records for years, and they were pointing to allegro (which went away 2 years ago or so) and freebie (now gone).
Jerry's ISP, Insync, had somehow managed to mess up their mail configuration so that they wouldn't relay for Jerry.
Somebody at Yahoo! was blocking UDP reply packets to freefall and hub, which is the problem we had been following.

What a mess!

Saturday, 6 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Boy, am I tired! Up relatively early today to say goodbye to our guests, and found that sat-gw, the current name for the new machine I bought in August, had frozen up and needed the Big Red Button. That's the same machine that gave me all the grief on 17 September.

Apart from that, didn't do much all day.

Sunday, 7 October 2001	Echunga	Images for 7 October 2001
		Top of page
		previous day
		next day
		last day

Slept in a long time today; I'm just not shaking this tiredness. sat-gw froze again this morning. I'm going to have to accept that there's something wrong with it.

In the afternoon to The Cedars to perform the concert. Despite the lack of preparation, things went quite well, and Yvonne was quite happy with the sound. She took a lot of photos of the event.

Monday, 8 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Today I was finally ready to change the routing of my class C network from Telstra to IHUG. Things did not go smoothly. To remove the route from Telstra I had to go to their web page, which acknowledged the change, but didn't perform it. I had to log a fault before it actually got done.

On the other side it also took a while to change over the IP addresses. I got a phone call from a rather confused network operator in New Zealand asking why he should route a German class C to Australia, and I had to explain the background of how I brought it with me when I returned from Germany. Finally they switched over the addresses, but the routing didn't follow. More phone calls. “Epoch must have been confused by the existing Telstra route. It should be right in half an hour”. Still nothing until late evening, “but it'll definitely be there tomorrow morning”.

While that was happening, got a phone call from Alan Modra. His cable service died, apparently with a large proportion of Telstra's cable infrastructure. Not a good day for networking. At least we still have one IP address (gregl1.lnk.telstra.net).

Tuesday, 9 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Up early this morning to see whether we were back on the net. No. Tried to call IHUG, but their annoying answering service wasn't in the office yet, though I knew the person I was looking for was. Finally called their network people in New Zealand, who said that they had reminded Epoch of the problem. Finally got on to my contact person in Sydney (forwarded from New Zealand), who told me that they had logged a fault with Epoch, and it would definitely be fixed by 10 am EST.

Round 10:45 I decided that we had spent enough time waiting, and that we should switch back. Reinstated my Telstra route and asked to have the old /29 reinstated. As I was watching the traffic, I found traffic for my class C come in... for about 3 minutes, until it was replaced by traffic for the /29. What timing! Had everything undone again, not without the help of Peter Kingston of Telstra: deleting routes just doesn't seem to work.

So finally I have a properly functioning satellite connection, along with one “low-latency” PPP connection. It's interesting to see the difference. A ping to just about anywhere in the world takes 500 ms: one way or the other, it has to go to America. There's more to it than just timing, though: the routes are very different from each end. Here's what we see tracing to the AUUG Web site:

$ traceroute www.auug.org.au
traceroute to www.auug.org.au (150.101.248.57), 30 hops max, 40 byte packets
 1  echunga (192.109.197.82)  0.283 ms  0.236 ms  0.197 ms
 2  FastEthernet0.way3.Adelaide.telstra.net (139.130.237.247)  2600.215 ms  458.823 ms  448.331 ms
 3  GigabitEthernet3-0.way-core4.Adelaide.telstra.net (139.130.237.235)  994.723 ms  1028.790 ms  470.535 ms
 4  Pos5-0.exi-core1.Melbourne.telstra.net (203.50.6.161)  461.331 ms  447.109 ms  456.040 ms
 5  Pos6-0.chw-core2.Sydney.telstra.net (203.50.6.17)  442.765 ms  458.835 ms  485.509 ms
 6  FastEthernet1-0-0.ken4.Sydney.telstra.net (203.50.19.14)  485.119 ms  452.143 ms  447.313 ms
 7  ozemail9.lnk.telstra.net (139.130.4.58)  459.989 ms  456.253 ms  510.727 ms
 8  431.at-6-0-0.XR2.SYD2.Alter.Net (210.80.3.121)  459.201 ms  474.900 ms  438.278 ms
 9  0.SO-5-0-0.XR2.SYD4.ALTER.NET (210.80.33.242)  578.468 ms  513.263 ms  507.201 ms
10  0.SO-0-0-0.XR2.ADL1.ALTER.NET (210.80.33.14)  517.760 ms  518.056 ms  510.188 ms
11  412.ATM8-0-0.GW1.ADL1.ALTER.NET (210.80.32.158)  575.643 ms  528.677 ms  518.572 ms
12  internode-adl-gw.customer.alter.net (203.166.43.98)  501.597 ms  512.511 ms  499.830 ms
13  fa1-0.ka6.internode.on.net (203.16.212.155)  523.867 ms  528.250 ms  514.994 ms
14  www.auug.org.au (150.101.248.57)  524.050 ms  534.704 ms  512.959 ms

In other words, the messages going there go via Melbourne and Sydney, change to a different network in Sydney, and come back again. With the exception of the timing, it's pretty much the same for either the single PPP IP or the class C. Looking back from the other direction is very different, though. To the PPP link, we get:

$ traceroute gregl1.lnk.telstra.net
traceroute to gregl1.lnk.telstra.net (139.130.136.138), 30 hops max, 40 byte packets
 1  fa0-0-557.ka6.internode.on.net (150.101.248.58)  0.764 ms  0.715 ms  0.581 ms
 2  fa0-0.boomerang.internode.on.net (203.16.212.153)  1.048 ms  0.842 ms  0.684 ms
 3  fddi0-1-0.way14.Adelaide.telstra.net (139.130.136.25)  26.693 ms  30.098 ms  25.784 ms
 4  FastEthernet0.way3.Adelaide.telstra.net (139.130.237.247)  26.230 ms  26.387 ms  31.418 ms
 5  gregl1.lnk.telstra.net (139.130.136.138)  160.258 ms  137.733 ms  129.431 ms

On the other hand, tracing to the same machine via the class C address gives:

$ traceroute echunga.lemis.com
traceroute to echunga.lemis.com (192.109.197.82), 30 hops max, 40 byte packets
 1  fa0-0-557.ka6.internode.on.net (150.101.248.58)  0.703 ms  0.656 ms  0.515 ms
 2  fa0-0.zephyrus.internode.on.net (203.16.212.157)  1.105 ms  1.568 ms  1.315 ms
 3  Serial0-0-0.GW1.ADL1.customer.ALTER.NET (203.166.43.97)  3.967 ms  2.736 ms  4.678 ms
 4  421.AT-6-0-0.XR2.ADL1.ALTER.NET (210.80.32.157)  4.984 ms  3.073 ms  2.736 ms
 5  0.SO-1-0-0.XR2.SYD4.ALTER.NET (210.80.33.13)  22.958 ms  22.437 ms  23.493 ms
 6  0.SO-6-0-0.TR1.SYD4.ALTER.NET (210.80.51.77)  23.852 ms  33.446 ms  23.236 ms
 7  209.ATM4-0.IR1.LAX12.Alter.Net (210.80.48.29)  191.280 ms  192.603 ms  190.197 ms
 8  137.39.31.222 (137.39.31.222)  190.067 ms  191.860 ms  191.096 ms
 9  122.at-6-1-0.TR1.LAX9.ALTER.NET (152.63.10.218)  189.432 ms  192.105 ms  191.009 ms
10  0.so-0-0-0.XR1.LAX9.ALTER.NET (152.63.15.117)  189.889 ms  199.343 ms  190.292 ms
11  0.so-2-0-0.XL1.LAX9.ALTER.NET (152.63.115.158)  196.940 ms  194.999 ms  194.013 ms
12  0.so-1-0-0.TL1.LAX9.ALTER.NET (152.63.115.142)  190.199 ms  189.571 ms  194.113 ms
13  0.so-5-1-0.TL1.DFW9.ALTER.NET (152.63.0.54)  226.740 ms  227.422 ms  226.002 ms
14  0.so-7-0-0.XL1.DFW9.ALTER.NET (152.63.0.194)  225.858 ms  226.392 ms  227.076 ms
15  0.so-7-0-0.XR1.DFW9.ALTER.NET (152.63.101.250)  225.366 ms  224.319 ms  224.324 ms
16  185.ATM5-0.BR3.DFW9.ALTER.NET (152.63.100.161)  224.192 ms  224.315 ms  226.541 ms
17  204.255.174.38 (204.255.174.38)  248.618 ms  238.014 ms  237.027 ms
18  pos10-0-0.lax-c100.gw.epoch.net (155.229.123.125)  228.825 ms  230.922 ms  243.022 ms
19  fast5-7.lad-c000.gw.epoch.net (155.229.57.150)  229.751 ms  253.802 ms  231.612 ms
20  pos8-0.pao-c000.gw.epoch.net (155.229.57.201)  236.713 ms  243.848 ms  242.454 ms
21  pos5-0.npa-m000.gw.epoch.net (155.229.120.58)  236.429 ms  232.717 ms  235.555 ms
22  fast12-1-0.npa-m100.gw.epoch.net (155.229.123.29)  255.176 ms  234.811 ms  246.918 ms
23  205-214-35-74.npa-m100.cust.gw.epoch.net (205.214.35.74)  248.842 ms  273.752 ms  324.766 ms
24  * * *
25  sat-gw.lemis.com (192.109.197.1)  755.842 ms  655.609 ms  591.090 ms
26  echunga.lemis.com (192.109.197.82)  598.692 ms  581.522 ms  609.125 ms

I found an unpleasant surprise in my mail. I suppose it's bad enough to get spam, but it's hundreds of times worse when it purports to come from me. I suppose I've annoyed one spammer once too often. I wish I knew what I could do about this one.

Return-path: <groggyhimself@lemis.com>
Received: from vip by mail.vcr.istar.ca with local (Exim 1.92 #2)
        for dan@nwave.com
        id 15qnRn-0000qg-00; Mon, 8 Oct 2001 23:20:47 -0400
Received: from ns.losangeles.cl ([164.77.210.122] helo=server.losangeles.cl)
        by mail.vcr.istar.ca with esmtp (Exim 1.92 #2)
        for grodzuik@wimsey.com
        id 15qnRk-0000qd-00; Mon, 8 Oct 2001 23:20:44 -0400
Received: from plain (localhost [127.0.0.1])
        by server.losangeles.cl (8.11.2/8.8.7) with SMTP id f931cd405044;
        Tue, 2 Oct 2001 21:38:45 -0400
Message-Id: <200110030138.f931cd405044@server.losangeles.cl>
From: groggyhimself@lemis.com
Reply-To: lance1075@digicom.ch
To: groggyhimself@lemis.com
Subject: Finally A REAL way to make $$$$ Dont miss this one! Time:9:38:21 PM
Date: Tue, 2 Oct 2001 21:38:21
Mime-Version: 1.0
Content-Type: text/plain; charset=""
X-BadReturnPath: lance1075@digicom.ch rewritten as groggyhimself@lemis.com
  using "From" header

Dear, Mr.  grugg,

Hello Fellow Entrepreneurs,

WHY are you getting this? Simple...  Because I make TONS of money from
sending out a few of these Emails every day.  That's why - PERIOD! You want
Honesty, you just got it.  You want proof.  You just got that too.  That's
right, it's the email you're reading right now! Why would you be reading
it if I wasn't making money? You wouldn't.  But it reached you and this is
all I do ALL day EVERY day.

(etc)

Wednesday, 10 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

This evening I had promised a demonstration of installing FreeBSD at the AUUG SA Chapter meeting. To my annoyance, I couldn't find a machine to test it on. The Duron box kept freezing up, and the old Pentium/133 that I use for odds and ends had problems that proved to be related to the new disk drives that had been donated to me: two of the four wouldn't power up properly. Grr. That took up an inordinate amount of time, and I didn't get much done.

To the meeting, taking Yana to another Mount Barker Concert performance, this time at the Hyatt. The presentation went OK, and Yana didn't get finished until after me. Turned out they had been invited to dinner at the hotel's expense. If I had known that, I wouldn't have had anything to eat during the AUUG meeting: it looked pretty good.

Thursday, 11 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Now things are more or less back to normal, it occurs to me how much time little things take up. Managed to get back to setting up some machines for storage system test and porting, which took up a fair amount of time. The RS/6000 now has ten disks on it. Unfortunately the Linux system disk is the ninth, so just about any change in disk configuration will change its name.

Having the Internet feed come in via satellite means that I have to set up a Linux firewall as well. Finally found the documentation for iptables, and spent some time reading that. An interesting approach, one that Michael Neuling talked about at the AOSS-2.

Friday, 12 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

More work on the test machines today. This Duron box is a real problem, and the fact that I no longer have a usable disk doesn't help much either. Spent most of the day trying to update it to recent -CURRENT. Interestingly, it didn't crash. Also more work on Linux firewalling. Things are looking better now.

Saturday, 13 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Another quiet day. If it weren't raining so much, we could go riding, but in any case tried to do less work, with some success.

In the evening, a long power outage brought home to me that since the network redeployment we haven't paid much attention to the issue. The satellite receiver machine (sat-gw) is without a UPS, and reacted particularly sensitively to the voltage fluctuations when the water pump cut in, so I had multiple reboots. air-gw, the machine which runs the wireless networks, was in an even worse situation: not only no UPS, the power circuit it is on is not connected to the emergency generator, so we had to connect an extension cable to run it at all. sigh It'll be a while before we have things the way I want them.

Sunday, 14 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Spent some time looking at Vinum today. My kernel debugging setup leaves something to be desired; I'll need to look at that in more detail. Fixed one silly bug, but there are plenty more waiting for me.

In the afternoon, out riding for the first time in weeks. It has been too wet recently. Down to the southernmost part of Kuitpo Forest, past where we went on 28 July. It's quite nice there, and we kept up a good pace for some time.

Monday, 15 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

A quiet day. Spent most of my spare time doing administrative stuff such as expenses. On the other hand, it was a nice day, so also went out riding with Yvonne and her friend Essey, who is agisting her horses at Hazlemere. I would have thought that Darah would have been tired after yesterday, but she was raring to go. I've never seen her so wet.

Tuesday, 16 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Finally got a bit of work done, punctuated by a long phone conference in the afternoon. We need to do more about the quality of the telecommunications equipment; it was very difficult to understand.

Got sphinx up to the latest version of Linux and EVMS, which, to my surprise, went without any particular problems, though I still don't understand how I can track a release which isn't kept under any form of source control.

Wednesday, 17 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

More work on the PPC Linux today. There's a lot of work needed to keep things in sync. JFS supports the 2.4.7 kernel and earlier, and EVMS uses the 2.4.12 kernel. It required a lot of patching, which makes me wonder how easy it's going to be to track the respective CVS trees. It really makes me appreciate the “single source tree” approach of FreeBSD. Anyway, apart from logistical things, everything worked well, and sphinx now produces the following messages on boot:

EVMS v0.2.1 initializing ....  info level(5).
Exporting EVMS Volume(63,1) from "/dev/evms/aix/hd5".
Exporting EVMS Volume(63,2) from "/dev/evms/aix/hd6".
Exporting EVMS Volume(63,3) from "/dev/evms/aix/hd8".
Exporting EVMS Volume(63,4) from "/dev/evms/aix/hd4".
Exporting EVMS Volume(63,5) from "/dev/evms/aix/hd2".
Exporting EVMS Volume(63,6) from "/dev/evms/aix/hd9var".
Exporting EVMS Volume(63,7) from "/dev/evms/aix/hd3".
Exporting EVMS Volume(63,8) from "/dev/evms/aix/hd1".
Exporting EVMS Volume(63,9) from "/dev/evms/aix/hd10opt".

Looks like we're one step closer to reading AIX file systems.

Thursday, 18 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

More work on the EVMS stuff today. There's something funny about creating the devices I noted yesterday: there aren't any, even after a reboot. I'll have to take a bit of time to get a kernel debugger installed and do a bit of snooping around. In the meantime, though, I created the device nodes and found:

# dd if=/dev/evms/aix/hd5  count=2 | od -c
2+0 records in
2+0 records out
0000000   A   I   X       L   V   C   B  \0  \0   b   o   o   t  \0  \0
0000020  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0
0000040  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0   0   0   0   b   7   0
0000060   0   f   0   0   0   0   4   c   0   0   0   0   0   0   0   0
0000100   e   7  \0   h   d   5  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0
0000120  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0
*
0000200  \0  \0  \0   S   u   n       J   u   n       1   7       1   1
0000220   :   4   4   :   1   3       2   0   0   1  \n  \0  \0  \0  \0
0000240  \0   S   u   n       J   u   n       1   7       1   1   :   4
0000260   4   :   1   3       2   0   0   1  \n  \0  \0  \0  \0  \0   B
0000300   7   0   0   F   4   C   0   0  \0   n   m   e  \0   y  \0
0000320  \0 001  \0 001   N   o   n   e  \0  \0  \0  \0  \0  \0  \0  \0
0000340  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0  \0
*

That's encouraging.

In the afternoon, Robert Bickle brought along components for the new wantadilla: an Athlon XP 1700+ (1.47 GHz) and an Epox EP-8HKA+ motherboard, along with 512 MB of DDR-RAM. Both are hot off the press: the new Athlon series was announced only 8 days ago, and the Epox board was also first shipped earlier this month. Put the machine together and tried a make buildworld time. The old wantadilla (750 MHz Athlon) did that in about 75 minutes. The new machine took 135 minutes. Grrr. There's obviously something to look at there, but I didn't have time.

Robert also brought a replacement for the flaky motherboard for the Duron which he delivered in August. This one was even worse: it seems to have severe BIOS problems. Didn't even get as far as booting. Grrr again.

Friday, 19 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Up early this morning and built a new kernel for the new machine, without the WITNESS debugging. Bingo! It built the world in about 35 minutes. WITNESS seems to take up about 75% of the CPU when it's enabled; definitely not something that should be done in any kind of normal environment. Carried on installing software; this time I won't cut over to it until I know everything is functional.

Also played around with the new Duron motherboard. That was a bad idea. I got it to boot, but it severely corrupted the disk when I tried to turn it off, even though the OS had halted. It also wrote all over a partition (NetBSD) which I hadn't used. Further investigation found that the board handled the power switch correctly if I installed the Athlon, which might point to problems with the processor, or just that the obvious breakage is CPU sensitive. Even with the Athlon, it was unable to install FreeBSD: the disk accesses kept timing out. sigh This is costing me more in time than the components are worth.

Saturday, 20 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

More work on the new wantadilla today. For some reason, lots of programs try to link with libc.so.4, although they should be using libc.so.5. To make matters worse, there is an undefined reference there. I suspect that somebody has been lax in bumping the revision number. Finally worked around it by linking libc.so.4 to libc.so.5.

Out riding in the afternoon. Unfortunately Darah lost a shoe, so we didn't go as far as we wanted.

Sunday, 21 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Quiet day today. Yana was away at a friend's place, and we didn't do much. Out in the afternoon with the horses, but there was some cycling competition going on, and we had to drive some time before we found a place to ride, to the East of the Kuitpo Headquarters. That's a bit of a record for me, going riding 4 times in 8 days.

Monday, 22 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

More work on the Linux kgdb implementation today. It's taking me into an area I haven't been before, Linux kernel configuration. Confusingly, there seem to be at least three different kernel debuggers, all different, and the one I chose doesn't have support for PowerPC—yet.

In the evening, Rasmus Lerdorf discovered an auction of cheap wireless network cards, finishing tomorrow, and so many that they would go for a guaranteed reserve price. Started a bulk purchase.

Tuesday, 23 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

More work on the kgdb implementation today. This stuff is much more complicated than I expected, and the version I downloaded from Sourceforge seems to be particularly Intel-dependent. Looked at the PPC implementation in more detail and discovered it was particularly Macintosh-dependent, and the serial port driver for Mac is really different from the generic driver (which is what the RS/6000 uses). Went on until my eyes went funny.

Wednesday, 24 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Another teleconference this morning. We have a proper teleconferencing setup now, so it wasn't as painful as last week, but it's still very difficult to understand when most people are in one room. A pity, because these discussions seem worthwhile.

Somehow I couldn't face the kgdb stuff today, so found some excuses and updated the satellite IP program to include statistics. The only way to get these statistics seems to be to read /proc/net/dev, which is a text file. There must be a tidier solution. I had been concerned that I hadn't been getting my full 400 kb/s, but discovered that the transfer rate maxed out at 52301 bytes/second. Given rounding errors, that's close enough to 400 kb/s.

In the process of testing, downloaded the new Emacs 21.1, which built out of the box. It certainly looks new (what? Emacs with a toolbar? Thankfully you can turn it off), but on the whole I think it's an improvement.

Thursday, 25 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

I'm still facing lack of motivation for doing the kgdb stuff, so spent some time tidying up my office, which sorely needed it. Moved the table at the entrance until it touched my desk, so now I have a three-sided desk setup, enabling me to put yet another monitor on the desk.

That was a bad idea, of course, because it meant that I had to go and move my X installation to the new machine and put a total of four display cards in it. That just plain Didn't Work. In fact, I wasn't even able to duplicate the configuration of the old machine: with both Matrox Millennium cards in the box, any attempt to start X froze the machine. Spent an inordinate amount of time trying to work out what was wrong.

Friday, 26 October 2001	Echunga	Images for 26 October 2001
		Top of page
		previous day
		next day
		last day

sigh X is really a pain. I keep forgetting that every time I do a configuration change, but a relatively simple modification (well, what's 5 screens?) took nearly 2 days to sort out.

I had to interrupt the work to go to the ADUUG lunch, which was very well attended today, a total of 16 people. I don't think I've ever seen that many. After that to a gate place to look at automatic gates, which was well worthwhile. Came away with some additional ideas for how to do the installation; maybe we won't have to dig a trench at all if we use solar power and wireless for the intercom.

Back home and finally put the old display configuration back in wantadilla, and put an Elsa Vanta card in monorchid (must change that name), along with the ATI Mach 64 PCI card. The Elsa card looked like it had a better spec than the Matrox (350 MHz pixel clock instead of 250, 16 MB memory instead of 4), but I couldn't get it to work. Finally put in the M200, but not before noting that the el cheapo AGP card I bought last week (AUD 25) had the same amount of memory as the Matrox, and actually a slightly higher clock frequency (256 MHz). It was another TNT card, though, and I had had enough.

So, now I have 5 screens, currently arranged from left to right:

A Matrox M200 on monorchid:0.1, running at 1920x1440x24. This replaces echunga:0, which had the same specifications.
A Matrox Millennium on wantadilla:0.0, running at 1600x1200x16.
A Matrox G400 on wantadilla:0.1, running at 2048x1536x24.
A Matrox Millennium on wantadilla:0.2, running at 1600x1200x8.
The ATI Mach 64 on monorchid:0.0, running at 1280x1024x8.

That gives me a total of 11,061,248 pixels. That's a little less than I got on 7 August, but then I needed a laptop as well. Still, I have the display in echunga.0, which is another 1920x1440x24, but I don't have a monitor to connect it to. I suppose I should drag out the old Eizo 9500, which could handle the the ATI, and use another monitor to get up to a grand total of 13,826,048 pixels.

Saturday, 27 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

A quiet day, or at least an attempt at one, but I found enough things to do to keep me busy, and despite the good weather, when Yvonne asked me if I wanted to go riding, I declined. I did get some reading done.

There was an interesting oboe for sale on eBay, and I put in a bid for it, but I was outbid immediately. Spent some time examining other offers on the web, however. There were some interesting leads in the UK; maybe I'll go and look at some while I'm there.

Sunday, 28 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

The clocks went forward today. I don't know what it is about that, but it seems to be strangely disorienting.

Didn't do much work, and didn't go riding again. I did spend some time investigating how to make Bulgogi on a raclette grill. Bulgogi is a Korean dish traditionally barbequed at the table. In Korea, they do this with gas grills. Raclette is an originally Swiss cheese dish made at a fireplace; the Germans have modified this tradition by making it under an electric grill, which also has a Teflon top, on which we tried grilling the Bulgogi. The results? Not too bad, though I need to work more on the side dishes, and the house still stinks of grilled meat.

Monday, 29 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

Gradually the BSDCon Europe is coming closer, and I still need to do the materials for the tutorial I'm giving. Started upgrading my laptops for that, more than slightly hampered by problems upgrading to a recent -CURRENT. It seems that something has gone wrong with library versioning in libc.so, and some older programs failed with an undefined reference to __stdoutp. I've seen this one before, but it's still a nuisance. The solution (well, workaround) is to replace /usr/lib/libc.so.4 with a link to /usr/lib/libc.so.5.

In addition, somebody decided to “improve” the system by adding a call to panic where previously a warning had been issued. The result was that after building the new kernel, it paniced on boot:

Pentium Pro MTRR support enabled
WARNING: Driver mistake: destroy_dev on 154/0
panic: don't do that
Debugger("panic")
Stopped at      Debugger+0x44:  pushl   %ebx
db> t
Debugger(c03acf5b) at Debugger+0x44
panic(c03aa690,c03aa6a0,9a,0,c16af400) at panic+0x70
destroy_dev(c16af400,9a,0,c03e1160,c0536d98) at destroy_dev+0x31
asr_drvinit(0,533c00,533000,0,c012b6dc) at asr_drvinit+0xa9
mi_startup() at mi_startup+0x90
begin() at begin+0x43
db>

The old kernel wouldn't boot without further attention, either. First, I had to disable acpi, because the old kernel didn't handle it. Grrr. It's times like this that I really love -CURRENT.

Tuesday, 30 October 2001	Echunga	Images for 30 October 2001
		Top of page
		previous day
		next day
		last day

More work on the tutorial today. In the process, tried to merge the debug macros which will end up in /usr/src/sys/scripts/ when I get my act together. That act was somewhat delayed by the fact that we now have kernel threads, and the proc structure has changed almost beyond recognition. Neither my ps macro nor Kirk's xps macro work any more, and I'm not even sure what information I need to present any more.

Also spoke with Greg Rodgers, who gave me some useful input on JFS 1. It looks like it'll be relatively simple to write a Linux file system which can read the stuff. Spent some time puzzling over the header files. JFS 1 seems to be nothing like JFS 2.

Wednesday, 31 October 2001	Echunga
		Top of page
		previous day
		next day
		last day

More chasing things around today. Sometimes it seems that there isn't even enough time to handle the routine things of life, let alone achieve something new. Today it wasn't helped by the spurious panics I was getting on startup, which I had somehow managed to rebuild into the system. Ah well, it'll be a good example for my tutorial next week.