802.11 status and futures

[Sam Leffler]

We're getting closer to 6.0 and the next STABLE branch.  This will 
define another ABI freeze which will have significant impact on my 
importing 802.11 work so I figure it's time to review where we are and 
describe what's going on in terms of future development.

6.0 will be the first release that has WPA support included in the base 
system.  While the support has been working (for some definition of 
"working") for >18 months it's only been available to CURRENT users so 
this is a major milestone.  The ath driver is well-tested in this 
regard; a few other drivers have minimal WPA support (e.g. no h/w crypto 
is yet supported).

I have fixes pending in other depots to address miscellaneous bugs and 
resolve compliance issues (the linux version of the code has passed many 
standard compliance suites and some of the changes from that code base 
have not been back-merged).  These changes will be rolled into HEAD as 
time permits and all will be MFC'd to RELENG_6 if possible.

As soon as it's ok to have HEAD diverge I want to bring in an entirely 
new framework for doing scanning.  This supports things like background 
scanning (scanning for ap's while associated), roaming, and enables 
station mode power save operation.  These changes affect all drivers so 
committing them won't happen until I get help in updating and testing 
other drivers.  Also these changes cannot be MFC'd because they would 
break existing API's.

Past the scanning-related changes I have an entirely new architecture 
for the 802.11 layer that I described at BSDCan 2005.  Slides are 
available at:

http://www.freebsd.org/~sam/BSDCan2005.pdf

These changes enable things like support for multiple ssid and bssid on 
a single device (e.g. so you can have a virtual ap using WPA operating 
in parallel with an ap that uses open authentication), WDS repeater 
operation, and mesh networks (this was designed with 802.11s in mind). 
The virtual ap (or vap) code has been in development for nearly a year 
and is about to be released for linux (and has gone through extensive 
testing).  Getting this code in the tree will again require help from 
folks to update and test drivers.  And again this cannot be MFC'd 
because of API changes.

While not directly related to the above work I will soon commit an 
update to the Atheros hal that will add support for some of their most 
recent parts that people are encountering in retail products.  This will 
be MFC'd to REELNG_6; I haven't decided yet whether to backport it to 
other branches (by the time I can consider RELENG_5 it may be irrelevant).

The above may change depending on my time and work commitments.

	Sam