Re: mac80211 QoS/aggregation questions, thoughts

[Ivo van Doorn <ivdoorn@gmail.com>]

On Friday 01 February 2008, Johannes Berg wrote:
> 
> > Like I mentioned earlier, ring/queue handling has received an overhaul,
> > so not much of the rt2x00 part of this patch will apply after the update.
> > However the change should be easier afterwards. ;)
> 
> :)
> Keep the patches coming. No need to test them in rt2x00.git first,
> wireless-2.6 git is pretty broken anyway right now. Due to me,
> unfortunately.

Ok. I'll try to implement LED classes this weekend. I already have the initial
patches for that ready, but those were not really correct. Ones those are in
I'll submit all patches as the 2.1.0 release.
If I release them all this weekend they will be compile tested only, and I would
like to have at least 1 run test to make sure no obvious NULL pointers or segmentation
faults are there during init.

> > Personally I think these queue identifiers should be completely seperated from the ieee80211_queue
> > enumeration, and not be used inside a variable mixed.
> > But that will be implementation details for later, and something I should probably take care of. ;)
> 
> Yeah, but this seemed easiest for now. We never use the beacon queue in
> mac80211 and I believe that we shouldn't because not all hardware uses a
> queue for beacons.

I understand the change, and my comment was more a first idea for teh rt2x00 implementation. ;)
I think I'll try to remove the IEEE80211_TX_QUEUE_BEACON tomorrow already and setup the correct
implementation. That value is set as identifier by rt2x00 manually anyways (it never relies on
mac80211 passing that value, so moving it to a different identifier can easily be done seperately from
your patch. Saves you some time updating the patch anyway. :)


> > Other then that, the patch is fine with me, but not this patch probably needs to be respinned after
> > my rt2x00 queue and virtual interfaces overhaul.
> 
> Right, no worries, I'm not entirely sure about this patch anyway.

Ivo