OpenBSD's presence at BSDCan 2006

While BSDCan was rather largely dominated by FreeBSD (and its users), OpenBSD had a respectable showing. There were two talks by OpenBSD developers, one by David Gwynne (Marco Peereboom contributed but was not able to attend) and the other by Reyk Floeter. Both talks were held in the same room, at the same time slot both days of the conference (conference schedule). Bob Beck and Mathieu Sauve-Frankel also attended (and heckled).

Once a table was aquired on the second (and slightly less rainy) day of the conference, the various shirts and CD's were offered for sale. From what I could tell it was rather well recieved, and a CD set amongst a few shirts were given away by the organizers at the conference wrap up.

bio And sensors in OpenBSD:
This was the first of the two talks and was given by an exhausted David Gynne. Marco Peereboom was supposed to be there as well but he had 'green card' issues and was unable to attend although there definitely were a few questions that David had to refer to Marco (specifics about the ami driver). As the title implies David talked about his work on the bio driver and sensor framework. The sensors framework will (for the currently supported devices) spit out readings from the available devices via the sysctl tree. The drivers in the kernel do most of the work and the sysctl command in userland queries the interface which looks deceptively easy to implement (see slides 14-16). David made is abundantly clear that the sensor daemon needs a lot of work and is getting better daily but is not quite perfect. While all architectures have the sensors framework enabled by default, the i386 and amd64 ports contain most of the available drivers.

The rest of his talk was regarding bio and the RAID managment framework. This grew out of a need to not only minimally manage a working raid set online but also provide a standard framework that could be applied to managing RAID cards from all kinds of vendors. Currently the ami driver (ie. MegaRAID) is the only supported adapter though there are plans to add more adapters as they become available (check out mpt,mfi,ciss) as well as eventually supporting the S.M.A.R.T. drive interface. As a start, the reason why there is only minimal knobs to tweak with bioctl is that at the bare minimum one only needs a few inquery functions which are provided through the sensors framework (like state of the array/s and hot spare/s) as well as a few discovery options (like blinking leds in enclosures) and finally extremely important recovery functions are provided by bioctl (such as adding and replacing hot spares and rebuidling there with). To expand the adapter support for the RAID/bio framework we need to convince vendors to provide developers with relevant documentation and if possible sample adapters. This most valuable thing I took from this talk was to find out that when requesting documentation from vendors it is important to ask for the register set documentation as well as the device control structures and formats. While I'm sure many people would love to help convince vendors to give up the documentation we now know exactly what to ask for (as opposed to the seemingly rather useless documentation Sun provides for it's newer products).

The talk ended with a few questions and an explanation about David's unimpressive experience entering into Canada which was for some reason quite common amongst the speakers. He played the strangest and most confusing video I've ever seen which he credits as being what got him across the border. I don't have a direct like to the video but lets just say it involved salami nunchunks and bacon belts.

Proactive Wireless Networks With OpenBSD:
Reyk's talk on his wireless ethernet work took place on the second day of the conference. He started his talk giving us a brief synopsys of the supported wireless chipsets (of which the rtw and (u)ral are the best supported). Reyk also was very emphatic over the OpenBSD's team's stance on binary blobs and the work done with the ath driver to make it blob free.

He then went to talk about his work with hostapd and it's use during the WhatTheHack conference. Since I've not looked at the hostapd man page and have not tried setting up a custom access point, a good portion of the talk was well over my head. One interesting thing I learned was instead of flooding a particular channel with noise to get rid of people using "your" channel you can use hostapd to send deauth frames and effectively (and more cleverly) do the same thing. Reyk made mention of this when he had problems when other people in his building were all trying to use "his" channel. His gave an example using hostapd which would send deauth frames to access points which were not "our own." The hostapd configuration file is very similar to the PF configuration which definately helped me figure things out when I got lost in jargon (frequently). Reyk has also been doing a lot of work adding W/IDS infrastructure as well as some rudimentry AP roaming to hostapd.

To show that the roaming (via trunk) worked Reyk gave a little demo where he was playing an MP3 (a hummpa track with a really long play time) over a little network and proceeded to switch between his wireless and a wired connection and the track didn't skip a note.

The talk finished up with a few mentions on wireless security options. The first being setting up an ad-hoc VPN tunnel through OpenSSH 4.3. I was particularly interested in this idea considering the overhead setting up a VPN for a client (especially mobile ones) but it was only briefly mentioned. The other alternative is using IPSec which will be made far simpler with ipsecctl in 3.9 which is largly a work in progress. Reyk mentions that there is and he is work on implementing WPA support but he made sure to mention that it was an ugly and over complicated protocol.

On the whole both talks were extremely interesting and I learned something new and personally valuable in both. This is the second time I've been to BSDCan and my only wish would be for more OpenBSD based talks even if just from users. Despite the rather small group of 'OpenBSD people' it was nice to be able to put a face to names I've only seen flame others on the misc list.

Media:
On the off chance that someone would find it useful I've decided to put up my audio recordings of the talk. I would like to thank Michael Knudsen (aka 'mk/reverse') for donating the bandwidth. The audio is not great but it is passable. I am told that the BSDCan organizers have professionally recorded each talk though there was no indication as to when and/or if that media would be put online.

bio And sensors in OpenBSD	Slides	[MP3 15MB][OGG 21MB]	43m 39s
Proactive Wireless Networks With OpenBSD	Slides	[MP3 20MB][OGG 25MB]	58m 00s