|
| Greg's diary |
| December 2000 |
|
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Friday, 1 December 2000 | Echunga –> Canberra | Images for 1 December 2000 |
| Top of page | ||
| next day | ||
| last day |
Ugh, I hate getting up early. Still, I woke up just before the alarm went off, but I was left with the feeling that it was the middle of the night. Up and down to the airport with no breakfast, and today the Qantas club had nothing useful, so ended up eating toast and Vegemite.
In Canberra, uneventfully but yawningly to the hotel and then to the office. Yvonne quickly retired to catch up on her sleep, while I avoided reading my mail in a fruitless attempt to get my Aviators working under FreeBSD. At least Chris confirmed that they are working. The driver seems to have some resource allocation problem: the first time round it never finds the card, but the second time round it recognizes them, then trashes the callout list. sigh
In the evening the Christmas party. Good food, good company and unfortunately the photos weren't spectacular.
On 29 December 2015 I reprocessed these images with DxO Optics “Pro”, producing images that were gaudy but better.
Got one of Paulus, Sue and Tridge which should be a good candidate for a caption contest:
We later gave this one captions on IRC:
dsifry: Paulus: “So, wife swapping tonight, right?”
Rasmus: Paulus: “Wanna dance disco boy?”
grog: Sue: “What, me, a Microsoft spy?”
| Saturday, 2 December 2000 | Canberra –> Brogo | Images for 2 December 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spending the night over P.J. O'Reilly's was not a good idea. As we discovered, they shut at 5 am, and they make a lot of noise until then. Next time we come we must ensure that we stay at the other end of the hotel.
Up for breakfast and then to pick up our car from “we try harder” Avis. I think I'll recommend that they concentrate more on achievement than trying; it took three people an average time of 20 minutes per customer to process picking up reserved cars. They didn't have the tiny Barina we reserved, of course, and we ended up with a Commodore. But it took us an hour, the tank was only half full, and the boot was full of newspapers. Ugh.
|
|||||||||||||||||||||||
Down to Brogo by the Princes Highway, which proved less interesting than we had hoped. Spent a relaxing afternoon talking and looking at horses. Too relaxing: I could barely keep my eyes open. Must be the lack of pressure.
| Sunday, 3 December 2000 | Brogo | Images for 3 December 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Another relaxing day. I was really able to keep away from my laptop, except for copying photos, but I did lapse enough to take a look in Christiane's “Networking Handbook”, published by McGraw-Hill. I don't think I have ever seen a worse book on the subject. Years ago, I used to think McGraw-Hill was a really good publisher, but this book made it clear how long ago it must have been. My favourite quote: “IP is used for routing”. Ugh.
| Monday, 4 December 2000 | Brogo –> Canberra –> Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Up to a slow start this morning, took some more photos of horses, and off to Canberra again. It's nice to have time to do nothing again; I had almost forgotten what it's like. Took a little look around Canberra, then back home. The couple of days we have been away have made a big difference to the appearance of the place: it must have been very hot and dry while we were away, and it looks very summery now.
| Tuesday, 5 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to work. Ah, the joy of having 3,000 mail messages waiting. I'm beginning to think that I should spend some of my copious spare time writing a better MUA.
The Yoyodyne box was still happily running. I think we'll be able to call this one a success. Instead, spent some time trying to bring zaphod up to speed. The last buildworld worked fine, but today it kept hanging. We need to spend more time looking at reliability.
| Wednesday, 6 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Finally there seems to be less to do, and I can get back to more neglected stuff. Started the day with a meeting in San Francisco about the BBC; it looks as if we're finally getting down to something.
Apart from that, continued work on getting zaphod up to date. We seem to have significant stability problems.
| Thursday, 7 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent most of the day catching up on mail and trying to get my test machine up to speed. CURRENT is not in very good shape at the moment.
| Friday, 8 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
I could almost use yesterday's entry for today as well. I'm getting rather concerned about the SMPng project's attention to reliability.
| Saturday, 9 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Still more problems with zaphod. It seems that at some point we exchanged problems with dual processor Celerons for problems with four processor Xeons. Sent out a message to the mailing list suggesting that we try to get more stability into the system rather than moving on with optimizations. I don't know how we're going to be able to test the lightweight threads like this. The message was not well received.
To get my mind off the problem, decided to look at the Aviator driver instead, and discovered much more about newbus structure than I really wanted to. It's beginning to look as if the big problem on my laptop is that the driver wants 48 kB of I/O memory, and it's just not available. Also tried it on zaphod, where it was recognized, but it killed both the IDE driver and the Ethernet card, making it rather hard to use.
Miss Teak is still very fat. Yvonne's still worried that she might founder, so I rode her, not once, but twice. The first time we rode around the North paddock, since La Tigre still needs to be taken easy, and we can't float Miss Teak. Miss Teak was not happy. In the evening rode her again, and she was much better.
| Sunday, 10 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
More work on the Aviator cards today. I spent a lot of time trying to work out why the card is not getting any memory. There seem to be 64 kB of I/O memory free at 0xc0000, and I couldn't find a way to confirm that, so spent much of the day writing some debug macros which would dump the device tree. That took a while, and I didn't finish everything, but I was surprised to find out how many “devices” there were on a simple laptop.
In the afternoon, Daniel O'Connor and Kris Kennaway came in to pick up a laptop. Daniel had just finished emptying a 5 litre can of oil onto the floor of his car, and spent some time wiping it up again. Had a bit of fun with the laptop, an ancient Dell Latitude LX with a 486 processor, destined for Jonathan Michaels. The disk we had planned to use proved to be dead, and the installation wasn't able to fire up the (PCMCIA) Ethernet card to do a net installation, so we put another disk in zaphod and installed there. I had forgotten how slow these old 2½" drives are.
| Monday, 11 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
This Aviator driver is getting on my nerves. Spent the whole day tracing through the kernel, finding much more about the bus structure than I really wanted to know, and finally came to the conclusions:
The annoying thing was that I had this card working on this machine 6 months ago, but that version of the driver no longer works. I'm beginning to suspect there's something fishy with the information about memory allocation at 0xe0000.
In the evening, after it had cooled down a little, Yvonne tried to saddle Miss Teak for me. She (Miss Teak) pulled back, broke the (newly repaired) buckle on the lead rope, and fell over on her back, fortunately without the saddle. Yvonne was more than a little upset. Went out and saddled her at liberty. She wasn't perfect, but a whole lot better. It's strange how Yvonne and Miss Teak don't get on.
| Tuesday, 12 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
After all this frustrating wireless stuff, decided to do something completely different today and tidy up the office, which sorely needs it. I actually got quite a bit tidied up, but finally the wireless Ethernet stuff got the better of me, and I was back looking for the problems. Finally I found it, in sys/pccard/pccard_nbk.c:
static struct resource *
pccard_alloc_resource(device_t bus, device_t child, int type, int *rid,
u_long start, u_long end, u_long count, u_int flags)
{
/*
* Consider adding a resource definition. We allow rid 0 for
* irq, 0-3 for memory and 0-1 for ports
*/
int passthrough = (device_get_parent(child) != bus);
int isdefault;
struct pccard_devinfo *devi = device_get_ivars(child);
struct resource_list *rl = &devi->resources;
struct resource_list_entry *rle;
struct resource *res;
if (start == 0 && end == ~0 && type == SYS_RES_MEMORY && count != 1) {
start = 0xd0000;
end = 0xdffff;
}
The code was simply limiting the range of allocatable memory to 0xd0000 to 0xdffff. The locations I had at 0xe0000 were just not being used. I changed end to 0xfffff, built a new kernel with that, and bingo! it recognized the card.
The joy was short-lived. When I ran ifconfig -a, I got a panic:
#1 0xc01c065c in panic (fmt=0xc0339518 "ray%d: %s(%d) spun too long\n") at ../../kern/kern_shutdown.c:568
#2 0xc01655e9 in ray_com_ecf (sc=0xc0f88800, com=0xc0fa4280) at ../../dev/ray/if_ray.c:3045
#3 0xc0164eeb in ray_repparams (sc=0xc0f88800, com=0xc0fa4280) at ../../dev/ray/if_ray.c:2647
#4 0xc0165535 in ray_com_runq (sc=0xc0f88800) at ../../dev/ray/if_ray.c:2982
#5 0xc016545f in ray_com_runq_add (sc=0xc0f88800, com=0xc8b16ce8, ncom=0x1, wmesg=0xc03393b5 "rayrparm")
at ../../dev/ray/if_ray.c:2932
#6 0xc0164df5 in ray_repparams_user (sc=0xc0f88800, pr=0xc8b16d2c) at ../../dev/ray/if_ray.c:2616
#7 0xc01621fe in ray_ioctl (ifp=0xc0f88804, command=0xc020693a, data=0xc8b16ebc "ray0") at ../../dev/ray/if_ray.c:680
#8 0xc020d8fa in ifioctl (so=0xc7fc9e40, cmd=0xc020693a, data=0xc8b16ebc "ray0", p=0xc7b77200) at ../../net/if.c:926
#9 0xc01d6b3e in soo_ioctl (fp=0xc0f85000, cmd=0xc020693a, data=0xc8b16ebc "ray0", p=0xc7b77200)
at ../../kern/sys_socket.c:141
#10 0xc01d3e7a in ioctl (p=0xc7b77200, uap=0xc8b16f80) at ../../sys/file.h:172
#11 0xc0312505 in syscall2 (frame={tf_fs = 0x2f, tf_es = 0x2f, tf_ds = 0x2f, tf_edi = 0x0, tf_esi = 0xbfbfefe4,
tf_ebp = 0xbfbfeff8, tf_isp = 0xc8b16fd4, tf_ebx = 0x3, tf_edx = 0x0, tf_ecx = 0x0, tf_eax = 0x36,
tf_trapno = 0xc, tf_err = 0x2, tf_eip = 0x804c1cc, tf_cs = 0x1f, tf_eflags = 0x283, tf_esp = 0xbfbfefac,
tf_ss = 0x2f}) at ../../i386/i386/trap.c:1141
It seems that the memory wasn't being mapped correctly. sigh.
| Wednesday, 13 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Still no joy with the Aviator cards. I spent more time installing the same driver on other machines. On Yana's Latitude it installed without problems, and seemed to come up, but the effect was rather like the sound of one hand clapping. On monorchid it came up, and I could run ifconfig without a panic, but the call hung. Recompiled the modules with all debugging code and then ran them again. The debug output is significantly different, so that will give me something to look at tomorrow.
| Thursday, 14 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Still more work with the Aviators. It's becoming clear that the real issue is memory mapping; after a lot of time in the kernel debugger, I realised what I should have realised a couple of days ago: I can extend the I/O map area, but I need to tell the system elsewhere as well. It turns out that there are a number of implicit assumptions that the area is only 64 kB long.
I can fix this of course, but at the moment Mike Smith and Warner Losh are rewriting all this code. This might just be a problem that would go away by itself.
In the evening to the inaugural meeting of SAGE-AU's South Australian Chapter. Phil Kernick jumped the gun on AUUG and pronounced a chapter. We discussed the matter, and Phil, Sarah and myself came to the conclusion that it would be a Good Thing to resurrect the AUUG Chapter as well and do joint meetings, at least in the beginning. Now to get the AUUG Executive committee to agree on that one.
| Friday, 15 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Tried to get my mind off the Aviator cards today, and actually got a lot of work done in the morning. Real Soon Now I'll even get my next expense report put in, a thing I don't normally do until I notice a lack of cash.
In the afternoon it got the better of me, though, and after a lot of reading through the kernel code found that we do in fact map the entire lower 1 MB of physical memory into the kernel at the kernel base address, currently 0xc0000000 on Intel. So much for yesterday's assumptions that I was out of range.
So I wondered if I could fake it by mapping both the attribute memory and the shared memory sequentially at the same location. Poked around in the kernel debugger and set the attribute memory to 0xd0000 and... it worked! Looking more carefully, however, I discovered that my intended fix didn't work after all: I still had attribute memory at 0xe0000. But the card worked, and without further ado I was able to set up a link to Yana's machine. I'm too tired to try to work out what's going on, but there's still monorchid to get going.
| Saturday, 16 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
More work on the Aviators today, of course. The problem with monorchid was a trivial interrupt conflict: my BIOS had enabled VGA interrupts on IRQ 9, the one I had chosen for the Aviator card. More interesting wer the issues of network partitioning and range.
Did some testing with Yana's laptop. It works acceptably from her office, but an ftp of a large file brought transfer rates of about 55 kB/s, closer to ADSL speeds than to Ethernet. When I brought it into my office, the rate went up to about 170 kB/s, or about 1.3 Mb/s payload data. That should be higher. ping times were interesting, too: with a good (local) connection, I get about 9 ms. At 1 Mb/s (fallback speed) you'd expect a transfer of a total of 64 bytes to take 500 µs in each direction, or 1 ms total. When the distance increased, so did the ping time; in extreme cases it reached 200 ms. Obviously the radio circuitry is doing a lot of things here.
With monorchid as the hub, things improved marginally. I was able to use sydney in the lounge room, but connectivity was sporadic, and I found it easier to route through firefly. It looks as if I'll need a gateway machine in Yana's office, especially when I get an 802.11b network up as well.
The other issue is simply how to handle changes in configuration. When sydney is in my office, I want it on the (much faster) wire. When I take it out onto the verandah, I don't want to have to reboot. But I have nfs file systems mounted, and nfs attaches the mounts to IP addresses, so if I just pop the Ethernet card and walk out to the porch, all my nfs file systems will hang. I haven't come to a good solution for that one yet.
| Sunday, 17 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Spent most of today reshaping my network. I thought it was a problem when I found I needed two networks in the house (10 Mb/s and 100 Mb/s; there's no way I know to combine 10/100 Mbs UTP and RG58 cabling in a single network), but now I have four, each /26s. The other two are the current Aviator 802.11 FHSS stuff, and a theoretical 802.11b DSSS network, currently with only one WaveLAN card, rather like the sound of one hand clapping.
Found an old 486 DX/2-66 lying around in the Mike Smith Memorial Room. On examination, it proved to be Yana's old machine. Put both PCMCIA/ISA adapters in it and installed FreeBSD 5.0 on it. Fairly straightforward, but getting the machine set up properly still took as much time as I was prepared to work today. Now I have:
=== root@fhss-gw (/dev/ttyp1) ~ 8 -> ifconfig -aed0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500 inet6 fe80::200:b4ff:fe33:6da2%ed0 prefixlen 64 scopeid 0x1 inet 192.109.197.187 netmask 0xffffffc0 broadcast 192.109.197.191 ether 00:00:b4:33:6d:a2 lp0: flags=8810<POINTOPOINT,SIMPLEX,MULTICAST> mtu 1500 lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> mtu 16384 inet6 fe80::1%lo0 prefixlen 64 scopeid 0x7 inet6 ::1 prefixlen 128 inet 127.0.0.1 netmask 0xff000000 ppp0: flags=8010<POINTOPOINT,MULTICAST> mtu 1500 ray0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500 inet 192.109.197.245 netmask 0xffffffc0 broadcast 192.109.197.255 inet6 fe80::200:f1ff:fe10:f99f%ray0 prefixlen 64 scopeid 0xa ether 00:00:f1:10:f9:9f vlan: 0 parent interface: <none> wi0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> mtu 1500 inet 192.109.197.62 netmask 0xffffffc0 broadcast 192.109.197.63 inet6 fe80::202:2dff:fe04:93a%wi0 prefixlen 64 scopeid 0xb ether 00:02:2d:04:09:3a=== root@fhss-gw (/dev/ttyp1) ~ 9 -> netstat -rRouting tables Internet: Destination Gateway Flags Refs Use Netif Expire default freebie UGSc 6 3178 ed0 localhost localhost UH 2 36 lo0 dsss-net/26 link#11 UC 0 0 wi0 => 10mb-enet/26 link#1 UC 0 0 ed0 => freebie 0:80:ad:b7:c9:c7 UHLW 12 1142921 ed0 1199 yana link#1 UHLW 0 4 ed0 => fhss-gw 0:0:b4:33:6d:a2 UHLW 0 1 lo0 192.109.197.191 ff:ff:ff:ff:ff:ff UHLWb 1 360 ed0 fhss-net/26 link#10 UC 0 0 ray0 => sydney 0:0:f1:10:fc:c2 UHLW 2 191774 ray0 391 firefly 0:0:f1:10:fc:c1 UHLW 1 26444 ray0 1100 broadcast ff:ff:ff:ff:ff:ff UHLWb 1 872 ray0
Installed the machine in Yana's study, currently on her table. When I get the Hi-Fi closet working, I'll put it in there. Now monorchid can go back to being deviant again.
| Monday, 18 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Today was something like a lull after the storm. No problems with the wireless cards, so caught up on my mail and tidied up a little more. It's amazing how much area you can free up by moving away the junk of a few months.
On IRC with Chris Yeoh and Anton Blanchard talking about the range of the Aviator cards. Chris thought I should have been able to get the whole length of the house, and that brickwork shouldn't affect them. Wandered around the property with sydney and confirmed that I could (barely) get a connection from the entrance to the house if I pointed the machine in the right direction. As Chris pointed out, this is probably a good thing, considering that the cards don't offer encryption. On the other side of the property I was able to get about 200 metres before the connection dropped, though it was very dependent on height and alignment.
| Tuesday, 19 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
| Topic: technology | Link here |
Spent some time working on the Bootable Business Card for FreeBSD today. Since this was supposed to be released, I tried it with 4-STABLE, which proved not to be a good idea. It looks as if our PicoBSD build tools have changed a lot since then, and I was no longer able to build crunch. Finally gave up and reverted to 5-CURRENT, but by that time the day was over.
|
LinuxSA Christmas dinner
|
Topic: technology | Link here |
Into town to the LinuxSA Christmas dinner, at the Kublai Khan Mongolian Barbecue. Pleasant time was had by all. Mark Newton told me that Leigh Hart [look at those domain names] had bought an Alpha, a RAID controller and 20 GB of disk, all for AUD 500 (about $US 270). Why does this always happen to other people? As if to rub salt into the wound, Leigh called me on my mobile to ask some questions about the isp SCSI driver.
|
Lilac catches a rabbit
|
Topic: animals | Link here |
When I got back home, went to the fridge to get a drink and found Lilac in her basket on top of the fridge—not alone. She had a baby rabbit in there with her. Interestingly, it was completely unharmed. We're left wondering whether the rabbit's body language was sufficiently similar to cats that she couldn't harm it, or if she had brought the thing in for companionship. Yvonne wasn't convinced, and we let it go again. Only later did it occur to us that we should at least have taken a photo.
| Wednesday, 20 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Back to working on zaphod and SMPng today, in preparation for developing the BBC on that platform. Noted with pleasure that I can now make world -j4 without the thing hanging, and took the opportunity to do some long-overdue updates to Vinum.
In the afternoon, back to the BBC. Had the same problems that I had with -STABLE. Grrr.
| Thursday, 21 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Chris Cravens and I have been discussing some problems with Netscape, and he thought that they might be related to the BSD version of Netscape, and they may not happen under Linux. Well, there's an easy way to check that, right? Wrong. I would never have thought that installing Netscape on Debian Linux would be such a problem. dselect wanted to replace the C library, which seemed a bit excessive and likely to break other things, but the official Netscape install script had library problems as well.
Call me a wimp, but it seems that there's something wrong if you need different versions of the C library to run programs. In the end, I took the rather more radical step of removing Debian Linux (to tape) and installing RedHat 7.0 (isn't that the one with the 3 C compilers? Oh well). Anyway, after some hours, got Netscape running on the RedHat box and confirmed that the breakage was identical.
In the meantime, to keep from getting bored, noted some strange packets coming in from the Internet:
09:20:06.731350 pc5.tokyosoc-unet.ocn.ne.jp.netbios-ns > dialup89.lemis.com.netbios-ns:
>>> NBT UDP PACKET(137): QUERY; REQUEST; BROADCAST
According to John Edwards, they're from the network.vbs worm. Since they're Microsoft-centric, they can't do any harm, but it reminded me that it was time to finally get my firewall up to date. More fun. It's amazing how an incorrect firewall configuration can reduce a network to a notwork. More fun for tomorrow.
| Friday, 22 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
As expected, more fun with the firewall code today. This time I did things right and RTFMed before installing it. Also built a kernel which defaults to open, so if anything goes wrong I can blow away the firewall config and work normally. Starting in mid-session somewhat confused things, since many of the rules apply to connection setup. As a result it took me quite a while to realise that I wasn't getting anything through. Looked at the rules, and they all looked good to me. Finally I posted a message on IRC:
* groggy kicks his firewall.
> Grrr.
> 01100 0 0 allow tcp from any to 192.197.109.0/24 25
> 02700 1 44 deny log logamount 100 tcp from any to any in recv ppp0 setup
> Dec 22 15:22:42 echunga /kernel: ipfw: 2700 Deny TCP 208.50.99.198:7418 192.109.197.82:25 in via ppp0
> What the hell's going on here?
<kkenn> groggy: you have a dyslexism in your netmask
<wca> umm groggy, i think you got your ip masks wrong
* groggy crawls into a corner and dies.
<wca> yer allowing 192.197.109.0/24 which denies 192.109.197.82
<wca> :-)
After a bit more work, things looked OK. I note that we need a rule to allow passive mode FTP, which assumes it can access ports 49152 to 65535:
${fwcmd} add allow tcp from any 49152-65535 to ftp.lemis.com
Now at least I'm not getting those annoying netbios-ns messages:
03700 1530 119340 deny udp from any to any 137 in recv ppp0
These fields are:
So in the course of about 18 hours I've rejected over 1500 of these packets.
Spent a bit of time thinking about performance implications, and did some ftp transfers between the 10 Mb/s and the 100 Mb/s networks. Interrupt time on echunga went up to 70% of the CPU, but I couldn't really see any difference, either in CPU load or throughput, with or without the firewall. I suppose I should try with a kernel with no firewall enabled. Those interrupt times look very high.
| Saturday, 23 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
A quiet day today. Spent some time thinking about the performance of the Aviator cards. They run at 2 Mb/s, with a fallback to 1 Mb/s when propagation conditions are poor. Under those circumstances you'd expect ideal ftp performance of about 220 kB/s and a ping time of about 1 ms. In fact, Chris Yeoh tells me that a friend of a friend explained to him why you can't get more than 160 kB/s. Chris couldn't remember the reasoning, though he found it sound at the time.
Anyway, the figures I was getting on the 16th looked OK: they tend to disprove Chris' FOAF, since we got up to about 170 kB/s. The real question is why the ping times are so long (I've never had one under 9 ms). Chris showed me some results with his laptop running Linux. Not surprisingly, the ftp throughput was no better, but the ping times were in the order of 2 to 3 ms. That's a puzzling combination. Others also reported that other wireless network cards under FreeBSD also give a much better ping time.
Reading the code didn't help, so I built a profiling kernel. Ah, the joys of running -CURRENT. Something has changed in the meantime, and ps wasn't working, so I tried a make world and watched the system panic on a known bug on nfs timeouts. Spent the rest of the day chasing that.
| Sunday, 24 December 2000 | Echunga | Images for 24 December 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Christmas is coming! Today I had invited some of the FreeBSD crowd to a barbecue lunch. Kris Kennaway (FreeBSD Security officer, currently studying in Los Angeles) and Benno Rice (working on the PPC port, currently living in Melbourne) were both back for Christmas. Kris arrived first, and we decided to make another attempt at putting together a machine for him to use while he's here. We took a whole lot of photos; the whole story's here
| Monday, 25 December 2000 | Echunga | Images for 25 December 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Christmas day has come too fast. I really didn't feel like Christmas today. It's also still, after four years, a funny feeling to spend it with relatives, after all the time in Europe where there were no relatives in reach. Down to my cousin Karen Besant and had Christmas dinner there with her family, mother Audrey Schaedel and my other cousin, Sandy Semmens and her husband Dean Semmens. Took a number of forgettable photos.
| Tuesday, 26 December 2000 | Echunga | Images for 26 December 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
The second day of Christmas today. I had toyed with the idea of riding Miss Teak, which I hadn't done for some time. First I went out to take a look at a horse which had been lying in the same place for a surprisingly long time. It was Miss Teak: she had broken off her left hind leg, just above the fetlock. No more riding her. sigh.
Some change in -CURRENT has broken my Aviator kludges. Set down to rewrite the attribute memory allocation, but didn't get it finished.
| Wednesday, 27 December 2000 | Echunga | Images for 27 December 2000 |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
First thing this morning John Bailey's son came in to bury Miss Teak, leaving all the gates open as he came in. I went up to take a look, but there wasn't much to see. An hour later, all we could see of her was a slight mound of earth. sigh again.
In the afternoon, went into Adelaide, encouraged by Yana. She had received the manual for our old video recorder for Christmas, claims she has now RTFM'd from cover to cover, and was waiting for us to buy a new one. Unfortunately, we didn't find anything, and came home with little achieved. I decided to check the web for video recorders, but it seems that Yahoo! now only lists people they like, and I'm not sure where to find the el cheapo fly-by-night vendors I knew last year.
More work on the Aviators. I wish I understood newbus better.
| Thursday, 28 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
More work on the Aviators. I wish I hadn't started this stuff, but now I've started, I have to continue. Got them working again, this time in a slightly less filthy way, by declaring both memory regions sequentially shareable. It turns out you need to lie and say that both regions are the same size.
Spent a bit of time preparing for installing the fire water tank. Steve was supposed to come this evening to fit the fittings, but he put it off until tomorrow.
| Friday, 29 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
More work on the Aviator performance issues. Added ktr calls to the driver to track the progress, and then discovered that one of John Baldwin's recent commits effectively decreased the time resolution to more than 100 ms. Fixed that, except for the KTR_LOCK case, and finally got some results, but at a snail's pace over the 9600 bps serial link. I must get tdump(8) working. Anyway, it seems that we're getting two interrupts from the card for each packet received. I wonder why.
In the afternoon, finally got the fire water tank in place. I suppose we need a fire hose now. It looks like we're in for a high-risk summer.
| Saturday, 30 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
Not much work today. Decided that we had got far enough with the fire water tank that I should finish it off. To my surprise, got the fire pump going on second pull (for the record: full throttle, no choke). Down to Mount Barker to buy some fire hose, and was rather surprised to find that the standard diameter is ¾". Bought 100 metres of that, along with some fittings, and back home, where Yana helped me unravel it. As I feared, 100 metres of hose is not enough to reach once round the house, so cut it into a longer and a shorter piece; the longer one reaches just over halfway round, and the pump is on the more endangered side of the house.
| Sunday, 31 December 2000 | Echunga | |
| Top of page | ||
| previous day | ||
| next day | ||
| last day |
The last day of the year already! This year seems to have gone at such a pace that I've scarcely had time to catch my breath.
Finally got round to taking another look at the O'Reilly gmat software that I last looked at on 2 October. I wonder what I was smoking then; I didn't need sgmls, I needed sp, which was there. Got past that and put together a port, textproc/gmat, coming soon to a ports collection near you. Everything seemed to install alright, but then the instructions went on:
Once you have completed the installation, you should try to format the sample chapters in the test/ directory that is in the same directory as this README file.
A slight problem: they don't explain how. No Makefile, no man pages, 38 undocumented programs in the bin directory. Ah well, I suppose they'll tell me if I ask.
In the afternoon, riding Darah again, still without shoes, to see if she's lame. She is not. She was raring to go, and I had to hold her back, a far cry from Miss Teak. sigh again.
We hadn't intended to stay up till midnight, but somehow we did anyway. I think I'm getting too old to enjoy seeing yet another year roll in.
Do you have a comment about something I have written? This is a diary, not a “blog”, and there is deliberately no provision for directly adding comments. It's also not a vehicle for third-party content. But I welcome feedback and try to reply to all messages I receive. See the diary overview for more details. If you do send me a message relating to something I have written, please indicate whether you'd prefer me not to mention your name. Otherwise I'll assume that it's OK to do so.
| Top of page | Previous month | Greg's home page | Today's diary entry | Next month | Greg's photos | Copyright information |