Older blog entries for mike750 (starting at number 12)

Made the receive buffer changes in olympic and some of the simple fix-ups suggested by Jeff Garzik.

Other than that its been a quiet day - I've actually had to do some of the work I get paid for - eeek !!!

Merged in the ppc changes to ibmtr.c Nothing too drastic, just sorting out the byte ordering on 16 bit reads. Tested out fine on ix86 so we don't appear to have broken anything.

Just wait for some feedback from the ppc testers and we'll release it to the world.

Back to olympic now and the receive buffer copying issue.

It's actually managed to get above 80 degrees in Chicago. Everybody up here is complaining like mad, not me, this is my type of weather. Living in MD for 3.5 years kinda conditions you to the heat and humidity.

Bashed the ibmtr driver to death and it stood up to all the abuse. Finalized the fixes and send them off to a couple of testers to give them the once over

Put the patch all together to be send off the Alan to find out that Alan has released -ac9 with more updates to ibmtr.c. So had to go and merge in these changes with the driver. Got it all sorted out and sent the patch off for inclusion

Finished the new Terry Pratchett yesterday, pretty good, not his best but not his worse but a long stretch, and their is a hilarious reference to salad cream / mayonaise that's you'd have to be a Brit living in America to fully appreciate.

The Clangers videos have arrived, now to get a vcr in the US that will play UK tapes ;)

Fixed up the problems with ibmtr, a swift ioremap here and an ioremap there and everything is working ok.
Now I've got access to my test machines I am going to beat the hell out of the driver this weekend before I release it again.

Found out why ssh has died, turns out the debian package upgrade put back the file /etc/ssh/sshd_not_to_be_run, so, of course, after the reboot sshd didn't start up. I think I need to find the ssh package maintainer and slap them with a wet fish.

Douglas Adams died. A minutes silence please.

Got to the bottom of the problems with olympic. Turns out to be the pci hardware setup on the iSeries machines, not the driver at all :)

Working on the problems with ibmtr.c, shouldn't take too much to fix it up. I think we'll add the ppc support with this patch as well.

Life is good !!

Just heard that the one user having problems with cardbus olympic is now up and running after upgrading to 2.4.4 and gcc 2.95.3.This had been worrying me for a while.

Life is definately good.

Lots of fun today !! Trying to track down the problem with ibmtr.c and the isa adapters I decided to update the kernel and driver on my machine at home 680 miles away. So a ssh'ing we go into the box, successfully upgraded the kernel, rebooted the machine, logged back in again.

Life is good, so we take the plunge and insmod the driver which immediately panics the kernel but luckily doesn't lock the box solid so we managed to reboot remotely again.

Then confidence got ahead of ourselves and decided to apt-get update/upgrade the box, which then proceeds to download 291 packages (140mb). No trouble so far. So we re-compile the kernel with kdb support, reboot the machine and try to log back in again.

Then we scream very loudly (and cause great mirth and merriment on #linux) as we realize that the update upgraded ssh and changed hosts.allow|deny and now we can't log into the machine. Tried various different methods to get back into the machine, not happening, no way, no how. So it will have to wait until I get back home tomorrow and can physically access machine.

I wonder if I can pipe kdb over a ssh connection though, hmmm.

Not much development work today.

For some perverse reason I decided to get a Lotus Domino server up and running on the laptop. The Linux install is pretty painless, but you've got to run Domino as a web service to be able to do the initial configuration, and the browser requires java support, so not being quite brave enough to attempt using Netscape 6 I had to fire up the other os in vmware to run the browser to configure the server.

Only had to run it twice :) to get the configuration sorted out and then fought with the Notes Administration client for a while getting it to utilize my company id file on the local database. Next task is to get Administrator to run under Wine so we don't need to run vmware at all.
But why I hear the crowds call out. Well it turns out that corporate is seriously looking to use Linux with Apache as a web interface and they want to be able to get Apache to redirect to the Domino http server if the file ends in .nsf. No trouble, just a quick tweak of httpd.conf and we should be there, and I'll be able to get this up and running a lot faster than the corporate MIS weanies. Then it'll be time to make a major push to move them over from NT to Linux. :)

Updated the web site with the latest patch for the Madge drivers that now allows pcmcia/cardbus to work with the 2.4.x kernels.

Updated the ibmtr patch for the 2.4.4 kernels and mailed it off for inclusion in the kernel (I'm going to get beaten for this one, the patch is over 130K !!!)

We've found a really obscure condition in the olympic driver. Until certain conditions, the transmit ring will fill up and not empty itself, stalling the driver. There is a subtle race somewhere in the driver and we're working on tracking it down.

I'm still pondering the ppc architecture, I think it may be time to get hold of a ppc machine to play with. I did suggest using the wife's iMac but she wasn't too happy at the suggestion :)

Finished off watching the last Cowboy Bebop dvd, shame there isn't any more of them, the series was excellent. Now to crack on through the Neon Genesis series.

Spent the weekend doing house stuff. Putting up more bookcases, fixing up the garden (managed to bust a blade on the lawn tractor when I went over a hidden stump though, not good, especially when you've got 2 acres of grass to cut and then have to get the old walk behind lawn mower out to finish the job, luckily only a small piece of grass was left.) Took the motorbike out Sunday morning for a long run, nothing quite like it for blowing the cobwebs away and giving you that nice adrenaline rush.

Discovered that Red Hat have included the latest ibmtr patch from the website in their 7.1 release. Hmmmm, I haven't finished fully testing the patch yet with the standard isa adapters. (Which don't work with 7.1 anyway as they've compiled in isapnp support in their standard kernel which goes and reconfigures the adapters so they don't get detected.)

The obsolete ibmtr.h has been deleted from the latest kernels :) Now to get the rest of the patch into shape and submitted.

Got up a 4am this morning to catch the plane to Chicago and spent the time on the plane trip looking at the differences between ppc and i386 for write/read[w|l], ntoh [s|l]. It would be really nice if the byte swapping that goes on was consistent, i.e. ntohs(readw(xxx)) gives you different ordering on the two architectures.

Started on the heinous task of getting the ibmtr / ibmtr_cs changes applied to the kernel. This is going to be fun !! The entire patch is over 160k, posting that on lkml is not going to make me very popular. Time to break the patch down into smaller chunks. The most difficult one is going to be ibmtr.c, not a lot of actual code was changed but the driver was completely re-formatted and made a *lot* more readable. (The patch for ibmtr.c alone is 3230 lines, whereas the original file is only 1963 lines long.)

Got bored so played with CML, found an omission in rules.cml for the token ring device so worked up the corrected file and sent it of to Eric Raymond for inclusion.

So, it turns out that McDonalds is using beef extracts in their fries when they claimed back in 1990 that the fries were vegetarian. Another place off the list, its bad enough already being restricted to 2 or 3 choices from the menus in most restaurants. (At least back home the foot and mouth is turning a lot more people veggie.;)

Started lurking on #linux, maybe one day I'll be brave enough to actually join in the conversations :)