Excesive RTT if cell flooded with AP to STA traffic + management frames in f/w&driver queues

[Sorin PAUN]

Hello all,

The large RTT values was (evidently) because of the kernel queue size of 
100. Smaller values give smaller RTTs.

Also, what I am investigating is the lost of management frames (actually 
  Auth/Ass) in such lareg traffic situations. I understand that the 
driver actually does not prioritize this management frames

My questions are:

Which management frames go in the driver and which ones rest in the f/w 
? I would say that Auth/Ass only are going to the driver and Probe 
Req/Resp and Beacons are analyzed by the f/w ?

Furhermore, I see that the driver has a queue which length is set by 
PRISM2_TXFID_LEN. What queue are in the f/w, of what size and does it 
prioritize the management frames?




Sorin PAUN wrote:
> Hi all,
> 
> I'm looking at packet loss in condition of large traffic in the cell. 
> I've found this interesting fact about the RTTs in the cell in some 
> condition
> 
> AP, STA1 and STA2 associated.
> 
> Cell is flooded with UDP traffic, 1500B packet size, from AP to STA1.
> 
> STA2 pings the AP :
> 
> 100 packets transmitted, 99 received, 1% packet loss, time 99966ms
> rtt min/avg/max/mdev = 267.894/280.973/324.914/7.935 ms
> 
> Whenever, when the cell is flooded with traffic from STA1 to AP, the RTT 
> get normal values:
> 
> 100 packets transmitted, 100 received, +2 duplicates, 0% packet loss, 
> time 99990ms
> rtt min/avg/max/mdev = 2.949/8.593/31.640/5.129 ms
> 
> What you think about this 300ms RTT?
> 
> Thank youu
> 
> I'm using hostap 0.2.0 with LinkSys WPC11 cards.