Sensitivity setting problem

[Denis Vlasenko]

On 24 June 2003 04:55, Jouni Malinen wrote:
> On Tue, Jun 24, 2003 at 12:04:21AM +0300, insecure wrote:
> 
> > I tried iwconfig wlan0 sens 1,2,3 but to no avail.
> > I presume that foreign signal exceeds threshold at any of these
> > settings.
> > 
> > I've read the code and it seems hostap does not accept any other
> > sens values.
> 
> Current version is mapping sensitivity configuration into Prism2
> cnfSystemScale with those three options. However, this does not really
> control what you would like to set with sensitivity threshold. Some of
> the BBP control registers might be more helpful, but I have not
> experimented with that kind of configuration.

I'd be glad to experiment with. I still fail to find Intersil
docs on prism, anybody feel free to mail me any and all non-NDA
pdfs you have, because today I am limited to guessing off the
hostap source.
 
> > 22:49:11 kernel: wlan0: dropped received packet from 00:02:2d:51:3e:ab with no ToDS flag (type=0x02, subtype=0x00)
> > 22:49:11 kernel: wlan0: RX signal=164 noise=156 rate=20 len=50 jiffies=14576029
> > 22:49:11 kernel:    FC=0x0008 (type=2:0) dur=0x0000 seq=0x00f0
> > 22:49:11 kernel:    A1=01:02:2d:61:49:cd A2=00:02:2d:51:3e:ab A3=00:00:00:00:00:00 A4=aa:aa:03:00:20:f6
> 
> That looks like a multicast frame from an (pseudo?) adhoc network..
> Since the destination address is multicast, it will be delivered to the
> driver in Host AP mode. That kernel message should probably be just
> removed and those frames could then be silently dropped..

I can shut up the kernel, yes,
but it would not eliminate the hw/fw side of the problem.
AP tries to receive 'em and this interferes with 'good' packets:

	       Signal levels
time	foreign sta	'my' sta	AP action
1	*		*
2	*		*
3	***		*		tries to pick foreign packet
4       ****		*
5       ***		*
6       ****		*
7       ****		*
8       ***		***********	KABOOM! collision,
9       ****		************	  no ack sent to 'my' sta,
10      ****		************	  'my' sta will need to re-xmit,
11      ***		************	  maybe at lower rate...
12      ****		***********	  

In the log attached to my first email you can see ~30 foreign
multicast packets/sec. God knows how many unicast packets
are there...

I think I really need to make that signal be treated as noise.

Folks, gimme the docs! ;)
--
vda