Re: Anybody working on 802.11 generic framework?

On 01-Jul-2002 Brooks Davis wrote:
> On Mon, Jul 01, 2002 at 02:01:28PM -0600, M. Warner Losh wrote:
>> Greetings,
>> 
>> I've been told that newer chipsets are more "winmodem"-like for the
>> 802.11{a,b,g,ti's22mbs} cards.  These cards generally just give you
>> beacon packets and nothing else (maybe ack too).  This, coupled with
>> HostAP mode, means it might make sense to have a generic 802.11 stack
>> in the kernel to deal with the MAC layer issues that future drivers
>> will require.

We won't be able to do real "MAC" (aka DLC) processing in s/w. The
turn around times are too small, at work we ended up using two ARMs to
implement the 802.11a MAC. One handling the real time stuff and the
other dealing with higher layer functions - e.g. beacon decoding for
power save paging (this for the ICs we develop).

However, if you mean handling the different packet types (e.g
management, control etc) that can be done and is done in the infamous
Raylink driver. The Raylink driver has state machines for finding a
network, association and authentication. There is some decoding of
other stuff - but that was mainly for interest. Some of the state
machines have bad error handling. There is certainly code missing
for handling roaming and re-association.

The other area that the Raylink driver needed help with was MAC header
furtling - but this may just have been to support Ethernet2 and LLC/SNAP
encapsulations.

The raylink driver does work, notwithstanding the comments at the
head of if_ray.c that were there when I felt the code needed tidying.
Given that only a few people use these cards and the driver is okay,
the suggestions are "turd polishing".

>> Has anybody else thought of this?  Anybody working on it?
> 
> I thought Julian did this in netgraph for this firmware based AP stuff
> so he might have some thoughts.  I think there was also a suggestion on
> the FreeBSD lists (maybe -net) at some point of adding this stuff as a
> layer on the level of ethernet, but I don't remember who suggested that.

It was probably me whilst writing the Raylink driver - I think I
suggested some generic code to handle encapsulation and signal level
caches etc.

> -- Brooks
> 
> -- 
> Any statement of the form "X is the one, true Y" is FALSE.
> PGP fingerprint 655D 519C 26A7 82E7 2529  9BF0 5D8E 8BE9 F238 1AD4

-- 
________________________________________________________________________
Duncan Barclay  | God smiles upon the little children,
[email protected]   | the alcoholics, and the permanently stoned.
[email protected]| Steven King
--
*bsd wireless list, a bawug thing <http://www.bawug.org/>
[un]subscribe: http://lists.bawug.org/mailman/listinfo/bsd-wireless/