Greg
Greg's diary
January 2001
Translate this page
Select day in January 2001:
Su Mo Tu We Th Fr Sa
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31
Select month:
2000 Sep Oct Nov Dec
2001 Jan Feb Mar Apr
2001 May Jun Jul Aug
Today's diary entry
Diary index
About this diary
Previous month
Next month
Greg's home page
Greg's photos
Network link stats
Greg's other links
Copyright information
    
Groogle

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

Up correspondingly early today, but didn't do much. Finished the port of gmat, and confirmed that yes, indeed, I had absolutely no information about how to run it.

Peter Wemm was boasting on IRC yesterday about how he could build the entire system on his dual processor Xeon machine in just over 30 minutes. I thought I'd go for the opposite and see how slowly I could build the world. I stopped short of reviving a 16 MHz 386SX, but started building on tweedledum instead. After 14 hours, and still way short of the end, it died with an “out of memory” message, so rebooted with four times as much swap and started again.


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

Spent most of today catching up with stuff I've left lying. Yana finally got round to looking for multistandard VCRs, and located a Toshiba one in Adelaide and a JVC on special order. Prices looked reasonable, but there was no way to get reliable specs. Looked at the Toshiba Australia web site, but they seem to be unaware that they sell consumer electronics. Finally into town to take a look at it, and forgot my SECAM tape. It looked OK, and I bought it, but when we got home I discovered it doesn't do real SECAM after all, only MESECAM. It also doesn't autodetect, so to play a MESECAM tape I need to go into a menu and select it. I wonder when I will ever see a VCR which doesn't have silly design decisions.


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

More catching up today. The last three months have left an enormous backlog. To make things worse, a number of people have started claiming on the FreeBSD-STABLE list that they're all having the same problem with Vinum. None can give adequate descriptions, but it's clear that there are at least two problems, and people don't seem to notice. Oh for a good, clear problem report.


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

More catching up work today, along with a surprising amount of activity on the lists about Vinum. If this carries on for any length of time, we should be able to iron out the remaining bugs fairly quickly.

More work on gmat; looks like I'll be able to use it soon.

Down below two pages of incoming mail (168 messages) for the first time since September.


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

Yet another day split between Vinum, catching up on things left behind, and gmat. Made progress on all three fronts: only 120 unanswered mail messages, plenty of input on Vinum, particularly from Tor Egge, and I was finally able to format the sample chapter supplied with gmat (well, sort of: the footnotes ended up one word per line). Amusingly, found my name in the sample chapter: it's from Programming with GNU software. I didn't know that they referred to Porting UNIX Software.


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

It's time to get a life, I think. Vinum is still keeping me busy, but somehow I get the feeling that I'm getting on top of the work. More playing around with gmat, but not much more progress. Lenny Muellner has been very helpful, but there's still something that I'm missing.

In the afternoon with Orlando and Darah to the Razorback section of Kuitpo Forest. When we left it was about 22°, but it seemed a whole lot warmer in the forest. The horses were certainly warm, anyway, and we had difficulty keeping them from cantering all the time. Yvonne was worried they might harm themselves, but I suspect they would have slowed down if we had gone longer. Got back to the car in record time.


Sunday, 7 January 2001 Echunga
Top of page
previous day
next day
last day

We were going to go riding again today, but Yvonne was still a little exhausted from the previous day's ride. I considered going by myself, but I reasoned that if Yvonne were exhausted from just having sat on the horses, they might be a little tired too.

Messed around the house a bit and otherwise did non-computer things. Took Yana out in the arena again with the Nimbus to get her used to starting and stopping the car.


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

I finally did my expenses! This must be the first time in years that I've done that without being prodded, either by somebody who wanted me to get them done, or because of financial problems.

More tidying up. It doesn't make for good copy, but I'm now down to a single page of mail messages. I will have time to spend the rest of the week looking at Vinum problems.


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

Finally some time to work on Vinum again. Tor Egge has sent in quite a few fixes and suggestions, and I finally had time to work on it. As I suspected, the range locking code had some interesting race conditions: I was expanding the lock table without quiescing the system (for obvious performance reasons), but this almost certainly meant that I was writing the unlock data to the wrong place. 3 words at a time; sounds suspiciously like the 6 word hole syndrome we have been seeing in the buffer headers. Decided that quiescing the system would be a bad idea, but that there's probably never a reason to have more than 256 locks on a plex, so set to rewriting the lock code.


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

More work on Vinum, and got the locking code working in -CURRENT. Still a way to go in -STABLE.


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

A lot of time on the phone today. Looks like more changes are under way. Also spent a lot of time working on documentation.


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

Finally got some documentation in DocBook. It's more of a pain than writing C code which compiles without warnings.

Rumours are going around that we're going to merge with Turbo Linux. Never a dull moment.


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

The weather has been ridiculously hot lately. Today it hit 39.8°C, and they're predicting even hotter weather tomorrow. Looks like good bushfire weather.

Didn't do too much during the day. Following IRC on the Turbolinux/Linuxcare merger makes interesting reading. It's amusing to read things like at the bottom of Linuxgram's report, which states:

Linuxcare, on the other hand, has a bunch of highly regarded guru-type kernel hackers whose presence on the payroll brings a certain cachet and authority with customers.

I wonder what people really think of us.


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

The 8 am news confirmed my fears this morning: “The greatest danger of bushfires since the Ash Wednesday catastrophe”. The weather forecast promised winds from the North and a maximum temperature of 42° C, so after breakfast we dragged out the fire pump and connected it up, showing Yvonne how to use it in the process.

What a waste of effort. The temperature only made it to 37.7°, and there was almost no wind. Still, I can't say I was disappointed.

More work on Vinum. Peter Wemm had found another problem associated with configuration changes on degraded arrays. I still don't understand why it happened, but at least we catch it now. I didn't catch the fact that I broke make world in the process. For the moment, I have the pointy hat.


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

I can't make up my mind what's up with the weather forecast. Last night they said that the temperature would be only 32° today, this morning they said 29°. In fact, we hit a maximum of 21.7°. Anyway, it was a nice relief from the heat of the last few days.

Lots of bits and pieces today. Somehow I end always up in a spin before leaving on a journey. Got some more fixes in for Vinum.


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

John Baldwin has made some fixes to SMPng which appear to make it less likely to hang. A good thing too. It took me three reboots to make world.

More teleconferences. We don't know anything about the proposed merger with TurboLinux, but it doesn't seem to stop us talking about it.


Wednesday, 17 January 2001 Echunga → Sydney
Top of page
previous day
next day
last day

For a change, the flight to Sydney was at a relatively civilized time, supposedly 12:20, but we were delayed and didn't take off until about 13:00. Qantas are getting even sillier about their regulations about the use of laptops, and now expect you to wait for (or guess the time) 20 minutes after takeoff. In view of the fact that I strongly doubt any interference in the first place, I judge their technical understanding to be on a par with people who say you shouldn't use mobile phones in petrol stations.

To the hotel with Alan Modra. It turns out that we have to share rooms, not a thing I intend to do again. To the venue, where we set up a mobile network and messed around a bit. In the evening to a Thai restaurant, and relatively early to bed.


Thursday, 18 January 2001 Sydney Images for 18 January 2001
Top of page
previous day
next day
last day

Into the first day of the conference. I'm beginning to wonder whether multi-track conferences aren't overdoing things; you end up wanting to go to competing presentations. One way or another, there's not enough time to sit around and discuss things with people. The first keynote was with Alan Cox, who described the new 2.4 kernel. Judging by the title of the talk (“World Domination: Classified Progress Report and Briefing”), I suspect that Alan hadn't been expecting it to have been released either.

Then to Richard Gooch's devfs talk. He seems to have cleverly avoided the bikesheds which have held up FreeBSD devfs for so long, but traded it for surprisingly long pathnames. Something to follow up about.

In the afternoon, Daniel Phillip's talk about Tux2, a rearrangement of ext2fs which guarantees consistency. An interesting concept, but like all of these approaches, it's a tradeoff between performance and reliability. I'm still not convinced that soft updates are the way to go, but I'm left feeling that they're a better approach than this one.

Juan Quintela talked about his VM system test programs. Very much a work in progress, but it's good to see that somebody is doing it.

Rik van Riel spoke about the Linux VM system. It's interesting to see how much has been borrowed from FreeBSD, but the topics about which he spoke (very lucidly) don't seem to be the same topics that cause the heated discussions in FreeBSD-developers.

In the evening, first to a “networking” reception, then to dinner at the Red Hat Chinese restaurant, chosen because of the name, though the food was good. Rusty Russell tried to buy one of their teeshirts, and was finally successful.

Back to the hotel in pouring rain. How could Paulus have claimed that the weather here is like Adelaide's weather a day or two later? The Great Divide makes its presence felt.


Friday, 19 January 2001 Sydney
Top of page
previous day
next day
last day

Started off this morning with David Miller's talk about zero copy file transfer for Linux. It was particularly interesting in that it seemed to be an adaptation of David Greenman's sendfile concept to Linux, somewhat complicated by the fact that Linux doesn't have mbufs. The details were above my head in the sense that, despite Jay Schulist's tutorial at the AOSS, I still don't understand Linux networking.

Then to Stephane Eranian's talk on IA64, which I personally found the most rewarding talk to date. Unfortunately got called out by Liz Carroll, and didn't get back until the talk was nearly over. I rather like some of the ideas of the IA64.

Next was Hugh Blemings' talk about reverse engineering. He had told me before that it was his first ever presentation of the kind, but he did an excellent job, demonstrating the setup he used to decode the Nokia serial protocol. Excellent. Excellent. Did I mention that Hugh is my boss?

Had lunch with Richard Sharpe. We only ever see each other outside Adelaide. We're thinking about targeting the education process with a book about computers for schools.

Missed the early afternoon talks, and Sarah Bolderoff and I spoke to Maddog Hall about what AUUG needs to do. It seems that he's on the USENIX board of directors, and he came up with a lot of good ideas. Now to see if we can convince people to follow up on them. Noted also that Kirk McKusick has asked USENIX to host the next BSDCon, and they apparently have agreed. Funny that the FreeBSD core team hasn't heard of this.

Then Rusty Russell's talk about how kernel hackers get the girls. Another enthusiastic talk about how hacking is so much fun. Rusty's a very different person from Hugh, of course, but it was interesting how both talks conveyed the enthusiasm that comes out of Canberra.

In the evening to the conference dinner. A pleasant, nay merry time was had by all. Late to bed.


Saturday, 20 January 2001 Sydney
Top of page
previous day
next day
last day

Up a little later than usual today, and found to my surprise that just about everybody else had survived the night unscathed. To Tridge's keynote, about the Tivo hacking. I suppose it was deliberate that Tridge, who doesn't drink, was put on first talk after the conference dinner. Still, the auditorium was packed, and of course he produced yet another enthusiastic talk from Canberra about the joys of hacking.

Tried to quickly grab my mail after that, and by the time I looked up again I found that I had missed the next session. In to hear Alan Au talking about TLB sharing in IA64. I'm not convinced of the approach.

After that, Neil Brown talking about his improvements to RAID-5 in the Linux md driver. His figures looked very impressive, but they were designed to optimize the sequential I/O case, and I can't see any direct application to Vinum. For example, he went to a lot of trouble to grab entire bands when writing, so that he didn't need a read before write phase. To do that he needed small stripe sizes; the largest he even tried were 32 kB in size, and he concentrated on stripes of 4 or 8 kB. The throughput was impressive, up to 50 MB/s as lied^H^H^H^Hreported by bonnie, but under BSD they would incur a much higher read and write I/O load for the more interesting multi-process random I/O case. I've recently been told that Linux still writes a maximum of 4kB at a time, though I thought I recalled hearing that they were doing much larger writes, up to a megabyte at a time. Once that happens, this hack will be worthless.

I came away with a number of interesting ideas to think about for Vinum, but gradually they faded as I realised they relied on these small stripes. It's interesting, though, that they allow multiple concurrent access to a stripe which is being written, depending on the state (i.e. progress) of the transfer. I wonder if this makes any real difference in random access situations. It obviously will if you're coalescing data for full band writes.

In the afternoon, tried to catch up on my mail, but spent a fair amount of time talking. Conference roundup was a speaker panel, slightly too Linux-specific to keep me interested.

In the evening down Anzac Parade yet again and had a talk with Daniel Phillips about the Tux2 file system and other things he's been doing. Seems he lives in Berlin, and used to live in Siegen. He was surprised that I knew where Siegen is.

I later put up a web page with photos taken at the conference.


Sunday, 21 January 2001 Sydney → Echunga Images for 21 January 2001
Top of page
previous day
next day
last day

Up a little later this morning, finally, and had a leisurely breakfast discussing kernel debugging with Rusty and Tridge. I'm still baffled that Linux doesn't have any facility for dumping core after a crash. It does display a stack backtrace on the console, but it suffers from the same problems with X as BSD does.

They kicked us out of the motel at 10 am for some obscure reason, so down to Coogee with Tridge to pick up Alan and Telsa Cox and into town to show them Yet Another Big City. Parked between two monorail stations, went to one and bought tickets at $3.50 each to go the long way round to the other station we could have gone to. Planned to go up the Centrepoint tower, but changed out minds when we found it cost $19.80 each.

To Andrew van der Stock's place, conveniently losing the address but finding it anyway. Had a pleasant time and took a number of further photos.

Back home, where I discovered that, despite the best of intentions, I had brought Anton's Aviator card and given him one of mine, thus thoroughly confusing DHCP and fhss-gw. Yana had somehow also got her machine catatonic, so I ended up rebooting both machines. echunga had had another panic, and I noted that I had removed the flash card from sydney this afternoon without umounting it, so I suppose there's another crash there just waiting to happen. sigh.


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

The first day back after a journey always seems to achieve nothing. I suppose the mail load, currently round 1300 a day, makes a big dent, especially as I returned with another 1000 awaiting a reply. As if things weren't bad enough, echunga paniced and couldn't get through the reboot. Investigation showed that the CPU temperature, normally round 35° C, was at 59° C. Previous experience has shown that this is the temperature where the CPU starts to misbehave.

Took the machine apart and noted a loose power connection to the CPU fan. After fixing that, the temperature was back to normal. Wondering whether a Linux person would have found this problem based on the backtrace, rebooted and... got another panic before the end of the reboot. Finally rebooted into single user mode and took a look at the dump, which pointed at the firewall code. Removed the firewall start, which allowed me to finish the reboot, and after that I was able to start the firewall. I suspected a race condition with named, but the interesting thing was that all three dumps, including the two which occurred during normal operations, showed the same fingerprint. Spent the rest of the day building new worlds for echunga and firefly, Yana's laptop. Just before it finished, went in to find that Yana had rebooted firefly into Microsoft to run the scanner. High time to install sane.


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

Still in catchup mode. While trying to rebuild Yana's world again, echunga crashed again, this time running the new -STABLE build. Sighed and bit the bullet to analyse the dumps.

The dump turned out to be interesting. It was building some information about ICMP packets, and it had decided, incorrectly, that the packet was an IPV6 ICMP packet. This particular packet was aligned exactly at the end of a page, and since the function thought it was IPV6, it tried to write to an offset beyond the end of the structure, falling off the edge of the world in the process. It seems that this function has been completely broken all the time: it set an integer to the IP version number, 4 or 6, and executed different code on the basis of this value, doing nothing if the version had one of the 4294967294 other possible values. The problem was, the variable wasn't initialized. It's fascinating to think that such a bug would occur at all: literally a one-in-a-billion occurrence, and then only when the packet was aligned at the end of a page which had no successor.

All this was in -STABLE, of course, but we can't commit fixes to -STABLE until they've been running in -CURRENT for at least a few days. Went to patch -CURRENT and... the fix was already there, and had been for months, hidden with non-committal messages:

Revision 1.18
date: 2000/10/26 12:33:42; author: pointyhat; state: Exp; lines: +86 -58
fix conflicts from rcsids
----------------------------
revision 1.17
date: 2000/08/13 04:31:06; author: pointyhat; state: Exp; lines: +248 -50
resolve conflicts
----------------------------
revision 1.16
date: 2000/07/19 14:02:09; author: pointyhat; state: Exp; lines: +22 -4
fix conflicts
----------------------------
revision 1.15
date: 2000/05/24 19:38:17; author: pointyhat; state: Exp; lines: +1 -4
fix duplicate rcsid's
----------------------------
revision 1.14
date: 2000/05/24 04:09:13; author: pointyhat; state: Exp; lines: +700 -241
fix conflicts

Look at those line counts! I wonder how many other bugs have been fixed in -CURRENT and not in -STABLE.

I was reminded of my discussion with Rusty and Tridge on Sunday. Rusty derived his IPCHAINS work from exactly this code. I wonder if he would have found the bug without a dump. It's possible, of course, but not knowing what the data really was makes it a whole lot more difficult to analyse.


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

Another day catching up, for the most part. In the late morning, I heard a noise from one of my test machines, looked over and saw some rather spectacular flashes from within the power supply. Turned off the power as quickly as I could, but it seemed already to have done what it wanted to do.

Called up Robert Bickle and organized a couple of second-hand AT power supplies for $10 a throw. Yvonne brought them back, and I tried to bring the machine back to life, but to no avail. A bit of swapping around with the machine I had lent to Kris Kennaway a month ago confirmed that the processor was OK, but the memory chip was dead.

That was the machine which housed the SCSI drives connected to zaphod. I didn't get round to checking whether the disks are OK, but I noticed that when trying to boot without any drives, the host adaptor (AHA 2940) went crazy. I hope it didn't blow it as well.


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

Another relatively inactive day. I had to go into Adelaide to do some personal business, so I didn't really get started until about 2 pm. A surprising number of similar complaints are coming in about Vinum, and I can't reproduce them. Fortunately it looks as if my disk cabinet will be arriving on Monday, and that should help a lot.

In the afternoon took another look at these problems with the O'Reilly gmat package (in /usr/ports/textproc/gmat). Lenny Muellner had replied to my message earlier this month saying that it must be something wrong with the macros, but I was very sure that there was nothing wrong. Finally found the problem: I had erroneously slipped a blank into a pathname in such a way that the diffs didn't find it. A nice debug story, though.


Friday, 26 January 2001 Echunga
Top of page
previous day
next day
last day

Australia Day and a public holiday, for all the difference that made.

For some months now I have been blocking spammers en masse: if a spam message comes from a domain I don't know, and the headers haven't been forged, I send a message and then give the ISP a day to respond (a thing I don't mention in the messages), then I block the domain. Today I got one of the most interesting responses I've seen in a long time (the last was a porn star doing her own manual spamming, who apologized for her behaviour). It's interesting to see how many spam messages are coming from Argentina nowadays.

Building NetBSD 1.5 on tweedledum is becoming a problem; apart from a power failure yesterday, I've been having NFS hangs which can only be solved by a reboot. During one of these, I accidentally rebooted the wrong machine, wantadilla. Took advantage of the mistake to insert another Matrox Millennium display card I found lying around. wantadilla now has three Matrox cards: a G400 AGP card and two PCI Millennium cards. The total display environment includes, from left to right,

All these are joined together with x2x, making me frustratingly short of 10 mexapixels: I only have 9994800 pixels. I suppose I could connect up the second output of the G400, which is limited to 1280x1024, but I don't know where I would put the monitor.

The configuration for wantadilla is here.


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

A quiet day. I don't seem to be able to keep up with my mail any more.

The interesting news of the day was that USENIX have accepted my paper on SMPng this year. From the comments on the proposal, it looks as if they're expecting the audience to be a lot more technical than I was expecting.


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

More build problems. Decided to upgrade checkpoint from NetBSD 1.4.2 to 1.5, but the build went badly wrong. I'm still trying to get firefly back to a consistent usable state; the non-functional Aviator driver in -CURRENT meant I had to back out to a previous state, but that no longer builds.

In the afternoon decided to finally take the plunge and start the port of Vinum to Linux. Maybe it won't be as bad as I feared. Obviously I'm going to have to rewrite the driver proper, since the Linux interfaces are so different from UNIX, but at least I understand the logic. RAID-5 will be the most challenging part, again.


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

Got my renewal notice for lemis.com from Network Solutions today. They're learning, but not much: the invoices now have the name of the country on them, and they're sent from Australia. Last year's invoice came from Denmark by surface mail, and had only “AU” for the country. It arrived after the last renewal date.

This year it arrived a few days before; it was dated 10 January, but had a postmark 24 January. So I had to renew by credit card. I was astounded, though, when a little later I got a mail message:

From: MAILER-DAEMON@lemis.com (Mail Delivery System) To: postmaster@lemis.com
(Postmaster) Subject: Postfix SMTP server: errors from
mailhost.networksolutions.com[216.168.233.68]

Transcript of session follows.

 Out: 220 wantadilla.lemis.com ESMTP Postfix In: EHLO opsmail.prod.netsol.com
 Out: 250-wantadilla.lemis.com Out: 250-PIPELINING Out: 250-SIZE 10240000 Out:
 250-ETRN Out: 250 8BITMIME In: MAIL From:<billing@networksolutions.com>
 SIZE=1468 Out: 250 Ok In: RCPT To:<GROG@LEMIS.COM> Out: 450
 <opsmail.prod.netsol.com>: Helo command rejected: Host not found In: RSET
 Out: 250 Ok In: QUIT Out: 221 Bye

In other words, Network Solutions have difficulties with DNS. In fact, the whole subdomain prod.netsol.com is not known to DNS. Sheesh. If I had known that, I would have looked harder for a domain registrar with a clue.


Tuesday, 30 January 2001 Echunga
Top of page
previous day
next day
last day

I was planning to go to Canberra tomorrow, but it got called off at the last minute. That's fine, since I have enough to do, but it rather got my plans to upgrade to Linux kernel 2.4.0 off track. Instead, downloaded the source to capellorosso and tried to compile it.

Well, compiling wasn't the issue, but installation was. It seems that every Linux user has his own incantation to install a kernel. I used “make all” and “make install”, which worked as I expected, until I tried to reboot and found I was still running the old kernel. Even when I explicitly overwrote the old kernel with the new one, the boot process somehow found the old 2.2.16 kernel, dumbfounding not only myself but also the Linux hackers on IRC.

Finally wrote a Makefile target (everything) which did the whole lot. This time the boot found the correct kernel, which hung the system while trying to uncompress it. sigh.

Network Solutions are not letting up. Not content with having a messed up DNS configuration, they're now showing that their internal mail setup is also broken:

From: System Administrator <postmaster@netsol.com> To: grog@lemis.com
Subject: Undeliverable: Date: Mon, 29 Jan 2001 19:19:57 -0500 X-Mailer: Internet
Mail Service (5.5.2653.19) X-MS-Embedded-Report: Content-Type: multipart/mixed;
boundary="----_=_NextPart_000_01C08A52.603D8610" Status: RO Content-Length: 3675
Lines: 111

[-- Attachment #1 --] [-- Type: text/plain, Encoding: 7bit, Size: 0.5K --]
Content-Type: text/plain; charset="iso-8859-1"

Your message

  To: billing@networksolutions.com; daemon@opsmail.prod.netsol.com;
nobody@www3.networksolutions.com Cc: abuse@networksolutions.com;
hostmaster@networksolutions.com Sent: Mon, 29 Jan 2001 19:20:01 -0500

did not reach the following recipient(s):

ccurtis@netsol.com on Mon, 29 Jan 2001 19:20:39 -0500 The recipient name is not
    recognized The MTS-ID of the original message is: c=us;a=
    ;p=netsol;l=IMC0101300020DC09YBPX
    MSEXCH:IMS:Netsol:US-Herndon-NIC:NETSOL-NIC-EX04 0 (000C05A6) Unknown
    Recipient

Note the headers: they're running Microsoft, the message was in multipart/mixed format, and they've managed to expose the name ccurtis, presumably an ex-employee. What a load of amateurs.


Wednesday, 31 January 2001 Echunga
Top of page
previous day

After pondering the death of capellorosso, I recalled that RedHat 7.0 was noted for having three different compilers, depending on what you were doing, and one of the alternatives was for compiling kernels. Rather than unravel that can of worms, I decided to try a new machine with Debian Potato.

Ah, the joys of installing Debian! I had enough space, so I just said “install everything and don't ask any questions”. In Debian-speak, that translated to the message level “critical”. And I got questions about paper size and similar trivia.

Finally got it installed and found that it had installed almost nothing. After going nearly crazy with dselect (which, to add insult to injury, insisted on reverse video), I NFS mounted the FreeBSD ports tree and installed my software from there. That worked relatively well, as did the subsequent kernel build and install:

Linux yoyodyne 2.4.0 #2 SMP Thu Feb 1 02:42:27 CST 2001 i686 unknown

Interesting, that date. It seems that the CMOS clock had been keeping local time instead of UTC, and for some reason the date(1) command in Linux doesn't reset the hardware clock. Neither do ntpdate or xntpd. And of course the man page was missing on all three machines to which I had access. For reference, the command is hwclock --utc --systohc.

I still don't know why people need so many incantations to install Linux. I added this to the top-level Makefile, and it seems to work fine:

everything: dep
    ${MAKE} bzImage
    if [ -f /vmlinuz ]; then mv /vmlinuz /vmlinuz.old; fi
    if [ -f /boot/vmlinuz-${KERNELRELEASE} ]; then \
    mv /boot/vmlinuz-${KERNELRELEASE} /boot/vmlinuz-${KERNELRELEASE}.old; \
    fi
        cp arch/${ARCH}/boot/bzImage /boot/vmlinuz-${KERNELRELEASE}
    ln -s /boot/vmlinuz-${KERNELRELEASE} /vmlinuz
    ${MAKE} modules modules_install
    lilo

Next step is to get remote kernel debugging running.


Top of page Previous month Greg's home page Today's diary entry Next month Greg's photos Copyright information

Valid XHTML 1.0!

$Id: diary-jan2001.php,v 1.20 2010/04/28 02:03:49 grog Exp $