Muli Ben-Yehuda's journal

I had a great weekend, doing various hackety things. I worked on the Introduction to Linux Device Drivers talk, including writing the driver I will use for demonstration – an in-kernel Game of Life implementation. More on this as the talk draws near. I also started reviewing a SCSI driver on linux-scsi, and refreshed my memory on the finer details of hardware management and PCI via the excellent Linux Device Drivers.

Other activities include sleeping, almost finishing the last Hyperion book, Endymion, having a great time with ladypine and either snuggling or wrestling with the cat. Oh, and studying. A well rounded weekend!

Last year, I guided a group of students from Eddie A.‘s Linux Development Workshop who wanted to work on syscalltrack. Together, we ported syscalltrack’s hijack module to kernel 2.6[0], and while it took a lot of effort from all involved, I believe we all enjoyed the experience. They got a pretty good grade for their work, too, 95/100.

On Friday afternoon, Lior, one of the guys, came by and gave me a small present from all of them – a bottle of Jameson’s Irish Whiskey. This is the first time I get a gift for working on free software[1], and I am deeply touched.

Lior, Ori, David, if you read this – thank you! As I kept saying, the pleasure was all mine… Cheers!

On a side note, I met Eddie at the last Telux meeting. I agreed to give a talk at this year’s Linux Development Workshop on kernel development, and told Eddie that if any student should be interested in doing a Linux kernel project, I’ll be delighted to guide them.

[0] the code is in a branch in CVS. I need to finish it and merge it into HEAD, and make a new release.

[1] I’ve had a link to the amazon.com wishlist for ages on mulix.org, and no one I didn’t know in real life ever bought me anything from there. Maybe I should replace it with a link to the local alcoholic beverages purveyor…

Unlike my usual custom, I am actually writing a design document well ahead of when it’s due. What a refreshing feeling!

There’s a discussion going on right now on lkml about the evils (read: inherent raciness) of the sleep_on() interface. Looking at trident.c, I see two uses of interruptible_sleep_on_timeout(). I have to confess I can’t immediately see the race there, but moving to the wait_event() interface is still probably in order. Unfortunately, Ranny borrowed tea, the laptop with the trident card in it, for his flight to the US and won’t return it until he comes back in a week and a half. Not being able to test your driver changes sucks.