WLAN: towards a solution for the roaming freeze

Werner Almesberger werner at openmoko.org
Thu Jan 22 23:59:57 CET 2009


Andy Green wrote:
> So... it's a firmware problem we can't do anything about?

We'll see. I haven't heard back from Atheros yet.

Since the module duly reports when it has crashed, we could also
catch that indication (there's already a notification mechanism
in place that can also convey this kind of errors) and restore
sanity by resetting the module.

And if everything else fails, there's always Boomerang ;-)

But yes, closed source firmware sucks, particularly if it's rather
complex and thus has obscure failure modes :-(

> Beacon packets contain the channel they're getting transmitted on to
> avoid problems from bleedthrough.

Yup, so it seems that there's no up to date scan information available
when the module tries to re-associate. In most cases, it notices the
problem within seconds, so this might just be normal behaviour. (*)

Sometimes it takes longer, though. Steve has friends at Intelligraphics
who have suggested that changing the background scan interval may
improve this detection period. So this is what tonight's test run will
explore.

(*) The AP could send a channel change announcement that should
    expedite the process and avoid missing the channel. I still have
    to find out if this is happening.

- Werner



More information about the devel mailing list