Older blog entries for kroah (starting at number 32)

work

I left WireX a week ago, and am now working for IBM. What a contrast in the two companies, the internal commitment to Linux that IBM has is amazing. I would have never guessed, and this should be a fun job (basically getting paid to do the things I was doing on my own anyway...)

USB

Lots of little things need to be cleaned up. The flaws in the usb-serial drivers keep getting pointed out (the SMP locking is all b0rked) so I will spend some time this week sorting it all out.

I wrote a small usb-skeleton.c driver and got lots of good review by a few people on how I messed up the SMP locking on it (any common thread here?...) The usb-skeleton.c driver was written as an example for others, and a good excuse to keep putting off a LJ article that I finally finished about Linux USB drivers. Hope it turned out ok, it seemed pretty messy.

Hotplug PCI

Compaq supprised me with a machine to test out the hotplug PCI patches with. It's a very impressive piece of hardware. I finally got it all set up, and now have quite the room heaters :) I need to work on the drivers a bunch this week also, luckily my day job allows this.

survived that, now onward to new stuff...

tomorrow is not going to be a fun day...

Wee, got the Hotplug PCI driver working on my box.
Now to attack it with a big stick...

OLS

Finished my paper late last night. Glad to see it done, hopefully some people find it interesting. Now to postpone writing my presentation till the last minute also :)

work

It really sucks to see a good, talented, person let go for no good reason. And of course management explained it all very badly. I hope they realize what they did when all of their Windows machines need something and there's no one to help them. Of course I feel extra bad, as I had pointed the job out to him in the first place a year ago :( Now who do I get to bother when the server's disc is full?

USB

Received a new USB 2.0 card. Looks real nice, hopefully I'll get some time now that my paper is done to hack on it. Also have to get this hotplug PCI driver working soon too, I hate cleaning up NTisms in Linux drivers.

Should be an "interesting" month...

gotta get working on my paper for OLS...time is running out...

Now I remember why I like writing code instead of documents :)

Lots of things done, lots of places gone to, and yet...

Finally made my room reservations for OLS and now I have to finish this paper by the end of the month.

Got a kick butt machine from VA Linux to do some HotPlug PCI development work, trying to tie it into the Linux Hotplug core.

Now to go figure out where I am going to get some spare time in the upcomming months...

long time no post....

USB on the Tyan Tiger 133 motherboard

I have been working for a while with Randy Dunlap on getting the onboard USB controller (a VIA chip) on this motherboard to work in SMP mode. All that we could get was in noapic mode, which isn't the best. Anyway, after loads of emails back and forth, I drug the machine to the local "Advanced Topic Workshop" put on by our local LUG (actually the monitor was the pain, the machine was easy to carry) and we sat in the corner of the room, and I let Randy loose on the box. After a few hours, and me learning lots more about SMP interrupts and APIC mode than I knew, he gave up.

Seems that even if the machine would work in APIC mode with the USB controller, so many pins are not connected to the APIC controller that there wouldn't be even much benifit to run in that mode.

So in summary, if you have this motherboard, use a 2.4.x kernel, and add "noapic" to the boot line, and things should be fine. And don't buy this board if you have a choice :)

But Linux does better than Win2000 in this regard, they don't support this configuration at all!

Linux-Hotplug

Also, I'm having fun working on the scripts for the Linux-Hotplug project. I'm going to be giving a talk about linux-hotplug at the Ottawa Linux Symposium this year, so I need to get working on that paper...

Bitkeeper

Been using Bitkeeper a bunch lately in my work to port SubDomain to the 2.4.x kernel. Very impressed with it. Like it lots better than cvs, but then that might not be a fair comparison :)

Finally released 12 security advisaries at once. Of course I made a few typos, and lots of grammer errors. That's what I get to try to rush the announcement to meet a self imposed deadline...

Hopefully the code updates don't have any such problems.

I have to say again, auditing code isn't fun at all. And it's amazing the kind of response that you get when you point out a potential security problem in people's code to them. The responses were grouped:

• no response at all. I'm amazed that the apache group fell in this category.
• rudeness that there could be anything wrong in their code. Now in defense of the author of links, he was right, there is nothing wrong with links, but the other people that really did have problems had a hard time accepting it.
• very helpful in fixing the problem, and learning about how to prevent things like this from happening again. The author of mgetty is in this category. Great response. Very nice to see.

Unfortunatly, the rudeness category had the most people :(

Off to the "simple" world of the kernel, that's enough security stuff for a few days...

Got a very generous donation of a SMP motherboard and 2 processors to try to get the USB code on Linux to work properly. Right now it only works if APIC is turned off, not exactly the best way to run a SMP machine :)

Others are also having the same problem, and the maintainer of the pci code is working with us to get this figured out (see here and here for more info.)

Things were very busy at work over the holidays. All I can say now is we understand why the OpenBSD people are so grumpy all the time. Auditing other peoples code sucks at times :(

off to update the USB backport for 2.2.19...