Re: [PATCH] nl80211: allow ad-hoc to set WMM parameters from outside

[Simon Wunderlich <simon.wunderlich@s2003.tu-chemnitz.de>]

[-- Attachment #1: Type: text/plain, Size: 3095 bytes --]

On Wed, Jan 02, 2013 at 03:41:28PM +0100, Johannes Berg wrote:
> On Wed, 2013-01-02 at 14:36 +0100, Christian Lamparter wrote:
> 
> > I can't find anything useful in the 802.11-2012 spec either. There's
> > something vague in 4.7 "Difference between ESS and IBSS LANs" on page 80:
> > "The services that apply to an IBSS are the SSs [Station ServiceS]. [...].
> > The parameters that  control differentiation of the delivery of MSDUs with
> > different priority using EDCA are >FIXED<. [...]".
> > 
> > But what does this mean? Does it translate into: "Every peer of an IBSS
> > truly independent of each other when it comes to the WMM parameters.
> > Each peer can advertise (in beacons, probes, ...) and have its own set
> > of fixed parameters and no one needs to sync those from anyone else"?
> > 
> > Or does "fixed" just mean that everyone has to go with the default WMM
> > parameters from 8-105 [i.e.: no one can change them at all]?
> 
> Yeah it's a little vague, but I think it means *fixed* as you describe.
> The WMM spec describes them as fixed and also doesn't include the WMM
> Parameter IE in beacons/probe responses, so all of this would be an
> extension to the spec and we have to be careful not to break the spec
> (default) case (which is correct right now, afaict, with default
> parameters and no parameters IE)

It appears that WMM IEs never carry EDCA parameters, and this is also not
specified anywhere that IBSS-stations should do that. All I can find is to use
the default EDCA parameters, and this is done right now in Linux - external WMM
IEs are ignored.

For the original patch, I'd still opt to have it integrated:
 * if no WMM EDCA parameters are provided from userspace, the default behaviour
   is still to use the default EDCA parameters - 99% of the people won't change
   that
 * changed parameter sets don't have to be exposed to other nodes - the BSS
   won't break if nodes will have other parameters, it might just behave
   unexpected in terms of priority. Normal users won't tamper with EDCA parameters
   and shouldn't notice that at all.
 * Exposing EDCA parameters in the WMM IEs is technically possible, but will
   probably give us more headache (syncing them, most clients won't support that,
   etc) then solve problems.

With the current patch, one can just set the parameters locally, using the
already existing nl80211-commands for AP. One thing we still need to consider
is the ibss-merge case (mac80211 will reset to EDCA defaults then, we want
to keep the custom values). I can provide a patch for that, of course.

> 
> Interoperating with FreeBSD would be useful though.

Yeah I agree. Don't know what the state in FreeBSD is now, but I'd still suggest
to not use EDCA-parameters in WMM IEs for reasons above. Even if Linux and FreeBSD
supports it in their open source drivers, there will probably tons of other drivers
not supporting these parameters and breaking the mechanism - this is not standard
behaviour, after all.

Adrian, others, what do you think?

Cheers,
	Simon

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]