Re: [PATCH 06/17] cw1200: Mini-AP implementation

On Sun, Dec 23, 2012 at 04:49:05PM +0100, Bartosz Markowski wrote:

Are you using this bss loss evt subscription? It was under STE_EXTENSIONS
if I good remember. This is part of STE CQM design. But answering to the
question FW should not spam with loss/regained events like that and if I
good remember (I do not work with this for more than 6 months now) this
behave quite well those days. I do not know lmac internals however , Ive
used to work with the driver only.

This "spamming" isn't a problem using the GPL driver thanks to its 
hysterisis and delays.

From my own investigations into the LMAC sources, that event is 
triggered by an internal LMAC timer if it hasn't received a beacon at 
all in that time.  Disabling powersave operation makes the problem go 
away entirely, so I suspect what's happening is that the LMAC is 
consistently waking up a little too late (or going back to sleep too 
early) to hear the beacons.

My guess is that the LMAC is pushing its powersaving margins too close.

Btw. how old is the fw bin you have?

I'm using WSM395 currently, dated June 18 2012.  (It fixed a problem 
that a client of ours observed with WSM392 and WSM385; the LMAC would 
lock up on noisy environments when block acks were in use)

 - Solomon
-- 
Solomon Peachy        		       pizza at shaftnet dot org	 
Melbourne, FL                          ^^ (mail/jabber/gtalk) ^^
Quidquid latine dictum sit, altum viditur.

Solomon Peachy [off-list ref] writes:

From my own investigations into the LMAC sources, that event is 
triggered by an internal LMAC timer if it hasn't received a beacon at 
all in that time.  Disabling powersave operation makes the problem go 
away entirely, so I suspect what's happening is that the LMAC is 
consistently waking up a little too late (or going back to sleep too 
early) to hear the beacons.

My guess is that the LMAC is pushing its powersaving margins too close.

Or synchronisation to TBTT is busted for some reason, like not
receiving a beacon before enabling power save etc. (just guessing here).

-- 
Kalle Valo

On Thu, Dec 27, 2012 at 10:48:24AM +0200, Kalle Valo wrote:

quoted

away entirely, so I suspect what's happening is that the LMAC is 
consistently waking up a little too late (or going back to sleep too 
early) to hear the beacons.

My guess is that the LMAC is pushing its powersaving margins too close.

Or synchronisation to TBTT is busted for some reason, like not
receiving a beacon before enabling power save etc. (just guessing here).

Unlikely; the "join" command to the LMAC is blocking and won't return 
until it hears a beacon.

 - Solomon
-- 
Solomon Peachy        		       pizza at shaftnet dot org	 
Melbourne, FL                          ^^ (mail/jabber/gtalk) ^^
Quidquid latine dictum sit, altum viditur.