Bugzilla – Bug 2201
Simulation fails when active probing is used in 802.11n/ac with channel bonding enabled
Last modified: 2015-11-22 16:55:31 UTC
Created attachment 2164 [details] proposed fix to solve this bug Initial discussion can be found at https://groups.google.com/forum/#!topic/ns-3-users/2jFF8jYrceI If the channel width used is larger than 20 MHz in 802.11n/ac, simulation fails with the following assert when we enable active probing : assert failed. cond="duration >= MicroSeconds (0)", file=../src/wifi/model/mac-low.cc, line=2280 In the considered case, the probe response gets transmitted at 13.5 Mbit/s (MCS = 0, channel width = 40). When it need to compute the duration of the ACK response, it selects the highest non-HT rate smaller or equal to 13.5 Mbit/s, i.e. 12 Mbit/s. At the receiver side, the lowest non-HT rate of 6 Mbit/s is picked for the ACK response since the station is not yet associated. Consequently, the duration field is set to a too small value in the probe response, resulting in a negative duration at the end (which thus triggers the assert). I propose to fix this by preventing the transmitter to select a control answer mode with a constellation size bigger than the one used for the transmission of the received frame (this should indeed be avoided anyhow).
Created attachment 2168 [details] new proposed fix to solve this bug Created a new patch. The reason is that another bug was observed (https://groups.google.com/forum/#!topic/ns-3-users/2jFF8jYrceI): AddSupportedMode was never called when disabling beacons and enabling active probing. IMO this should be added once a probe response was received (and if previous check about supported rates successfully passed).
Does someone have any comments about the latest proposed fix? If not, I will deliver it in few days.
Fixed in changeset 11771:ce31a4af11d6