Blogs

Access Point Supporting PCF - (Contention Free Parameter)

By gstefanick posted Apr 22, 2015 02:00 PM

  

A VIP user was complaining about WiFi connectivity issues. It appeared to be local to this user and no one else. Users in the immediate area weren’t reporting a problem. I do about 60 minutes of captures and the user doesn't have a problem the entire time. Typical, “motor is knocking bring it to the shop and it doesn’t knock.”   

 

During the capture something caught my eye. Something rather interesting. The examination of a  neighboring beacon from a Cisco-Linksys access point I seen the “Contention Free Parameter Set”. At first I thought it was a CRC. But after closer examination and reviewing other beacons it was indeed legit.  It would appear at some level the access point might be supporting PCF (Point Coordination Function). 

 

PCF is this mythical creature. Legend has it, few if any vendors supported it. Interestingly enough, here it is. It’s not a CRC. Not sure to what level this is implemented. 

contention.free.period.png

 

I was curious to see if any clients were actually supporting PCF and associated, but I didn't see any. I filtered the traffic and a CFE (Contention Free End) frame could be seen from time to time. Again, not a CRC.

 

contention.free.period.2.jpg

 

 

Reference: http://community.arubanetworks.com/t5/Airheads-Dictionary/PCF-Point-Coordination-Function/ta-p/221738

 

Im curious has anyone else seen this or PCF in the wild ?

 

 

8 comments
0 views

Comments

Jon Linton
May 11, 2015 09:03 PM

I know this is a bit old now, however, I may have an answer.  While no devices actually implemented PCF (Point Coordination Function), the CFP does in fact exist.  The CFP is a variable length duration prior to the Contention Period.  So, if you send out a CF-End you can prematurely end the CFP and get on with your DCF goodness.  It simply means that the CFP (when PCF would occur) is complete and the Contention Period (when DCF occurs) can now begin.  As far as PCF showing up in the beacon, that does not provide any performance enhancement.  However, it would notify the BSS to look for the CF-End.  I've heard this question a number of times and it baffles just about everyone.  It is in the same spririt of chip vendors using high duration values to "reserve" the medium for as long as possible.  Not breaking the standard but still usually looks like errors. Nice catch. Sorry it took so long to reply, I got this in a linked in email.

gstefanick
May 05, 2015 09:33 PM

While I agree with the CF-End .. What do you guys thing about the CFP in the beacon ?

Z273
May 01, 2015 04:31 PM

Hi George,

Sure next time I go to Houston to visit my uncle, we should meet!

 

About CF-end comments from eljay is in line with my traces, I look at some old intel5300 HT traces, I do see intel does CF-end right after BlockAck, also I do see Intel does RTS with Duration=4108 microseconds before txing the AMPDU of 31 subframes at HT40 MCS23 SGI to the AP.

 

gstefanick
Apr 30, 2015 06:53 PM

Come on down to Houston and we can jump through some ceilings! Lol 

Z273
Apr 30, 2015 08:40 AM

Hi George,

AP in rental office... how about we tag team like Mission Impossible(old school) and you get closer to that AP 8-)

Me too my CF-end frames has CRC error even if they look legit, I used latest wireshark, I will try to "look" at those frames in omnipeek.

Cheers.

gstefanick
Apr 24, 2015 11:08 AM

I love the conversation guy. Ive seen the Intel CFE as well. But this is a bit different, this CFP is in the beacon which I find intersting. Maybe I never looked hard enough before, but this is a first for me. Normally you see CF elements in CRCs. This AP was in a rental office so I dont have access to the AP. 

 

 

eljay
Apr 24, 2015 10:58 AM

I've seen CFE packets in a similar situation to the one Chi-Thanh references, but with an Intel chipset on a laptop. Basically it set the duration field to 4000 or so so it could happily send a bunch of packets without getting interrupted. They're easy to spot with Omnipeek as it picks these up as a "duration attack".

Z273
Apr 23, 2015 08:21 PM

Hi George,

A few years ago there were some discussion about CF-end among other weird stuff with packet duration on Andrew VonNagy ( @revolutionwifi ) blog http://revolutionwifi.blogspot.ca/2012/05/are-apple-iphones-misbehaving-on-wi-fi.html

Not sure if it is related to your PCF as it is about Intel client that does CF-end.

Can you tell which Linksys device is doing your PCF "in the wild"?

 

Chi-Thanh