Re: DTIM (Power Save Mode) w/ HostAP

On Fri, Mar 07, 2003 at 10:56:03AM -0600, David Young wrote:

> Maybe somebody out there knows off the top of their head

Isn't IEEE Std 802.11 like mandatory reading for everyone? ;-)

> 1 What are the timing requirements for answering PS-Polls? Can we
>   meet the requirements if the HostAP driver will not queue at least
>   one PS packet per STA and hold it until until that STA's PS-Poll?

PS Poll frames can be replied either directly with the buffered data
frame or with ACK. This reply has to happen immediately (within SIFS),
so there is not really time for it to go through the driver. Prism2
station firmware (in Host AP mode) does indeed reply with an ACK to PS
Poll frames, so the driver does not need to do it. After this ACK, there
are no strict time requirements, but the station needs to stay awake all
the time, so it would be better to send the buffered frame as soon as
possible.

> 2 Will the HostAP driver actually queue and hold one PS packet?

No, the AP needs to buffer more than one frame. To be more exact, it
would need to buffer all unicast frames (apart from few exception) to
the station in PS mode. Of course there could be some kind of limit for
the buffer size and aging function to expire old frames.


On Fri, Mar 07, 2003 at 02:16:33PM -0800, Sam Leffler wrote:

> I believe Jouni Malinen's Linux hostap code supports PS-polls; not
> sure.  I
> know he's done a bunch of similar "estorica" like
> fragmentation/reassembly.
> It might be a good place to look.

Yes, Linux Host AP driver supports PS buffering (and has supported since
2001 ;-). I haven't implemented 802.11 fragmentation, but the driver
does indeed has defragmentation that is used with host-based WEP
decryption.

-- 
Jouni Malinen                                            PGP id EFC895FA
--
*bsd wireless list, a bawug thing <http://www.bawug.org/>
[un]subscribe: http://lists.bawug.org/mailman/listinfo/bsd-wireless/