[Freedombox-discuss] FreedomBox Unstable Image 2012.0819 Available

Wed Aug 22 14:50:13 UTC 2012

On Wed, 22 Aug 2012, Nick Daly wrote:

> Are you building your own image or using the published ones?  In
> either case, only images built from the "shiny" branch will have
> wireless enabled, and then only when you plug an ethernet cord into
> the eth1 slot during the first boot.  The /etc/init.d/first-boot file
> controls this process and it removes itself after configuring the
> system on the first boot.

In this case I used the published image. The first time around I didn't 
have ethernet plugged in at boot so I re-fetched the first-boot init.d 
script and ran it by hand (including the wireless fetch). Later I started 
over with the fresh 2012.0819 image and made sure I had ethernet 
connectivity to fetch everything (and saw the wget from spinifex succeed).

> Reading this, it seems like either I built the images from the master
> branch (entirely possible, that branch doesn't have wireless), or you
> didn't have an ethernet cord plugged into the eth1 slot during the
> first boot.  If you did, you would've downloaded spinfex's drivers
> automatically.  I'll make sure to call this out in the notes for this
> week's image.

I think the image I got from you is a shiny one: the "wifi-ap-setup" line 
in init.d/first-run was uncommented (it is commented out in master), and 
the spinfex *firmware* was downloaded and placed in /lib/firmware/mrvl 
(after which the first-run file was deleted). Still no uap8xxx *kernel 
module* though.

Looking more carefully through freedom-maker history, it looks like it 
compiled a custom kernel at some point (based on work by Bauermann; it 
presumably included the uap8xx module?), but this was removed because 
"hallelujah! the Dreamplug patches were back-ported already!" and "no 
longer need local kernel content since all we need will be in Debian 
wheezy":

https://github.com/NickDaly/freedom-maker/commit/2bd110374a5b39e8bd2c7c4583a4b8dbf8d132a9
https://github.com/NickDaly/freedom-maker/commit/fae93dbf3f2f03f606bb81d09431f50b32321f6d

I assume this has to do with device tree information or other bits 
required to get the kernel to run on a dreamplug.

However, while I do see libtertas_sdio.ko and "usb8xxx.ko" in the wheezy 
kernel filelist, I don't see uap8xxx.ko:
http://packages.debian.org/wheezy/armel/linux-image-3.2.0-3-kirkwood/filelist

or in experimental's 3.4 image:
http://packages.debian.org/experimental/armel/linux-image-3.4-trunk-kirkwood/filelist

or in mainline:
https://github.com/mirrors/linux/tree/master/drivers/net/wireless/libertas
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux.git;a=tree;f=drivers/net/wireless;h=6690fa09b446d59773f022e7d9ec72516368dbcb;hb=d9875690d9b89a866022ff49e3fcea892345ad92

What I ended up doing last night was to grab Bauermann's 3.4 kernel 
dreamplug patch set [0], which was based on a 3.2 kernel patches (wheezy 
kernel is 3.2)  and extracted just the uap8xxx module code (which lives in 
a "libertas_uap" directory) and then followed Dan Gilmore's very old 
directions [1] to compile[2] this on my dreamplug (running the 2012.0819 
freedom box image, which already had the mrvl .bin firmware), install it, 
and blacklist the other libertas drivers. After this I did a 
power-down-reboot (to ensure that any .bin firmware was flushed from 
chipset memory?) and during the next power up uap0 finally existed, was 
automatically configured, and came up as a protected "freedombox" SSID, 
which I was able to connect to and SSH over from my laptop. So it seems 
all the wireless configuration works just fine as long as the uap8xxx.ko 
kernel module exists and is loaded.

Anyways, I've gone way out on a limb here, and the question remains: does 
the wireless access point work for other people (especially with older 
DreamPlugs) using the weekly image? Perhaps the uap8xxx driver gets snuck 
in somewhere that I missed, or perhaps there were improvements to the 
libertas_sdio driver in the debian wheezy kernel and I simply have a 
configuration or hardware problem.

--bryan

[0] https://github.com/bauermann/dreamplug
[1] http://lists.debian.org/debian-arm/2010/05/msg00081.html
[2] had to kludge out line 74 of uap_debug.c; will post soon