Bug 2605 – A HT/VHT station transmitting to a legacy access point results in a null throughput

Bug 2605 - A HT/VHT station transmitting to a legacy access point results in a null throughput


Summary:	A HT/VHT station transmitting to a legacy access point results in a null thro...

Status:	RESOLVED FIXED

Product:	ns-3
Classification:	Unclassified
Component:	wifi
Version:	ns-3.26
Hardware:	All All

Importance:	P2 major
Assigned To:	sebastien.deronne

URL:
Keywords:

Depends on:
Blocks:	2606 2607
	Show dependency tree / graph

Reported:	2017-01-02 17:30 UTC by sebastien.deronne
Modified:	2017-01-11 14:39 UTC (History)
CC List:	2 users (show)

See Also:

Attachments
test script to reproduce (7.80 KB, text/x-csrc) 2017-01-02 17:30 UTC, sebastien.deronne	Details
patch to fix (7.80 KB, patch) 2017-01-02 17:37 UTC, sebastien.deronne	Details \| Diff
patch to fix (15.51 KB, patch) 2017-01-07 09:53 UTC, sebastien.deronne	Details \| Diff
patch to fix (23.34 KB, patch) 2017-01-07 15:06 UTC, sebastien.deronne	Details \| Diff
Show Obsolete (2) Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.

Description sebastien.deronne 2017-01-02 17:30:56 UTC

Created attachment 2726 [details]
test script to reproduce

When a HT/VHT station is transmitting data to a legacy access point (a/b/g), I obtain a throughput of 0 Mbit/s. 
From traces, it points to the fact the station attempts to associate using a HT rate, so it cannot be understood by the AP.
See test script to reproduce the issue.

Comment 1 sebastien.deronne 2017-01-02 17:37:09 UTC

Created attachment 2727 [details]
patch to fix

There are several issues that are fixed by the attached match:
- Any HT/VHT stations assume MCS 0 is widely supported. Based on beacon capabilities, they do not tell the station managers to discard all MCS.
- HT/VHT stations always enable HT/VHT support without checking HT/VHT capabilities are valid.
- ADDBA is sent to the AP regardless whether AP does support QoS or not.

Comment 2 sebastien.deronne 2017-01-05 03:34:23 UTC

This blocks 2 other bugs and the fix is quite obvious, I suggest to push to mainstream this weekend.

Comment 3 sebastien.deronne 2017-01-07 09:53:07 UTC

Created attachment 2739 [details]
patch to fix

Adapted previous fix because m_htSupported is local and is not part of serialization process. Instead, we check whether it supports at least default mcs 0.

Comment 4 Tom Henderson 2017-01-07 12:20:33 UTC

I agree with the patch; it would be nice to also commit the example and instrument our test runner to try various combinations and to check that sane output is obtained (i.e. more than just checking for zero).

We have a bug open on this topic but discussion died out:
https://www.nsnam.org/bugzilla/﷒0﷓

Comment 5 sebastien.deronne 2017-01-07 15:06:46 UTC

Created attachment 2740 [details]
patch to fix

Previous fix was not handling VHT case.

Comment 6 sebastien.deronne 2017-01-07 15:07:50 UTC

(In reply to Tom Henderson from comment #4)
> I agree with the patch; it would be nice to also commit the example and
> instrument our test runner to try various combinations and to check that
> sane output is obtained (i.e. more than just checking for zero).
> 
> We have a bug open on this topic but discussion died out:
> https://www.nsnam.org/bugzilla/﷒0﷓

I will add this example to the regression. The only reason I compare with 0 is that there are still issues with HT Minstrel and Ideal. Once fixed we could change this.

Comment 7 sebastien.deronne 2017-01-11 14:39:44 UTC

Fixed in changeset 12525:e0b63ca01399