test on latest hostapd with acs

Wed Sep 18 08:17:23 PDT 2013

On 18 September 2013 09:49, Sarah Rumpel <Sarah.Rumpel at barco.com> wrote:
> Hi,
>

Hi Sarah,

> I did some tests on the latest hosapd with ACS (commit
> 50f4f2a066e60be7cc03e291635e9a0f953922bd 31 Aug 2013).
> I ran it without and with additionally generated continuous interference on
> channel 11, but I achieve for both cases more or less the same results.
> Does someone achieve similar results or have a solution?
> I attached my config file, perhaps there is a mistake in my configuration.
>
> Without interference:
>
> ACS: Survey analysis for channel 7 (2442 MHz)
> ACS: 1: min_nf=-99 interference_factor=0.447154 nf=-98 time=123 busy=55
> rx=38
> ACS: 2: min_nf=-99 interference_factor=0.633333 nf=-99 time=123 busy=79
> rx=62
> ACS: 3: min_nf=-99 interference_factor=0.585366 nf=-98 time=123 busy=72
> rx=64
> ACS: 4: min_nf=-99 interference_factor=0.634146 nf=-99 time=123 busy=78
> rx=70
> ACS: 5: min_nf=-99 interference_factor=0.672131 nf=-99 time=123 busy=83
> rx=69
> ACS:  * interference factor average: 0.594426
> ACS: Survey analysis for channel 8 (2447 MHz)
> ACS: 1: min_nf=-101 interference_factor=0.396694 nf=-98 time=123 busy=50
> rx=12
> ACS: 2: min_nf=-101 interference_factor=0.390244 nf=-98 time=123 busy=48
> rx=33
> ACS: 3: min_nf=-101 interference_factor=0.642202 nf=-99 time=123 busy=84
> rx=52
> ACS: 4: min_nf=-101 interference_factor=0.831776 nf=-101 time=123 busy=105
> rx=46
> ACS: 5: min_nf=-101 interference_factor=0.565574 nf=-98 time=123 busy=70
> rx=54
> ACS:  * interference factor average: 0.565298
> ACS: Survey analysis for channel 9 (2452 MHz)
> ACS: 1: min_nf=-99 interference_factor=0.528926 nf=-98 time=123 busy=66
> rx=14
> ACS: 2: min_nf=-99 interference_factor=0.57377 nf=-99 time=123 busy=71 rx=38
> ACS: 3: min_nf=-99 interference_factor=0.565574 nf=-98 time=123 busy=70
> rx=34
> ACS: 4: min_nf=-99 interference_factor=0.694444 nf=-98 time=123 busy=90
> rx=51
> ACS: 5: min_nf=-99 interference_factor=0.455285 nf=-98 time=123 busy=56
> rx=41
> ACS:  * interference factor average: 0.5636
> ACS: Survey analysis for channel 10 (2457 MHz)
> ACS: 1: min_nf=-104 interference_factor=0.256198 nf=-98 time=123 busy=33
> rx=27
> ACS: 2: min_nf=-104 interference_factor=0.349593 nf=-104 time=123 busy=43
> rx=36
> ACS: 3: min_nf=-104 interference_factor=0.577236 nf=-98 time=123 busy=71
> rx=65
> ACS: 4: min_nf=-104 interference_factor=0.38843 nf=-98 time=123 busy=49
> rx=41
> ACS: 5: min_nf=-104 interference_factor=0.393443 nf=-99 time=123 busy=49
> rx=44
> ACS:  * interference factor average: 0.39298
> ACS: Survey analysis for channel 11 (2462 MHz)
> ACS: 1: min_nf=-99 interference_factor=0.22622 nf=-98 time=37320 busy=9019
> rx=7361
> ACS: 2: min_nf=-99 interference_factor=0.226773 nf=-99 time=37470 busy=9074
> rx=7407
> ACS: 3: min_nf=-99 interference_factor=0.22749 nf=-98 time=37620 busy=9136
> rx=7462
> ACS: 4: min_nf=-99 interference_factor=0.227893 nf=-98 time=37770 busy=9192
> rx=7504
> ACS: 5: min_nf=-99 interference_factor=0.228157 nf=-99 time=37920 busy=9236
> rx=7541

This is an interesting bit. See the disparity between time spent on
channel 10 and 11?

If you calculate relative values you get the following for channel 11:

time=150 busy=55
time=150 busy=62
time=150 busy=56
time=150 busy=44

This seems to be close to channel 10, as expected.

> ACS:  * interference factor average: 0.227306
> ACS: Survey analysis for selected bandwidth 20 MHz
> ACS:  * channel 1: total interference = 1.6204
> ACS:  * channel 2: total interference = 2.09924
> ACS:  * channel 3: total interference = 2.57981
> ACS:  * channel 4: total interference = 2.71252
> ACS:  * channel 5: total interference = 2.80184
> ACS:  * channel 6: total interference = 2.82541
> ACS:  * channel 7: total interference = 2.91017
> ACS:  * channel 8: total interference = 2.82258
> ACS:  * channel 9: total interference = 2.34361
> ACS:  * channel 10: total interference = 1.74918
> ACS:  * channel 11: total interference = 1.18389
> ACS: Ideal channel is 11 (2462 MHz) with total interference factor of
> 1.18389
>
> With interference:
>
> ACS: Survey analysis for channel 11 (2462 MHz)
> ACS: 1: min_nf=-99 interference_factor=0.228277 nf=-99 time=38102 busy=9282
> rx=7578
> ACS: 2: min_nf=-99 interference_factor=0.228057 nf=-95 time=38252 busy=9308
> rx=7601
> ACS: 3: min_nf=-99 interference_factor=0.228001 nf=-95 time=38406 busy=9341
> rx=7631
> ACS: 4: min_nf=-99 interference_factor=0.227652 nf=-95 time=38556 busy=9362
> rx=7649
> ACS: 5: min_nf=-99 interference_factor=0.227331 nf=-95 time=38706 busy=9384
> rx=7668

Relative numbers for this would be:

time=150 busy=26
time=154 busy=33
time=150 busy=21
time=150 busy=22

Actually this reports less interference on channel 11.

> ACS:  * interference factor average: 0.227864
> ACS: Survey analysis for selected bandwidth 20 MHz
> ACS:  * channel 1: total interference = 1.59097
> ACS:  * channel 2: total interference = 2.10033
> ACS:  * channel 3: total interference = 2.71575
> ACS:  * channel 4: total interference = 2.743
> ACS:  * channel 5: total interference = 3.01982
> ACS:  * channel 6: total interference = 3.42506
> ACS:  * channel 7: total interference = 3.5531
> ACS:  * channel 8: total interference = 3.33931
> ACS:  * channel 9: total interference = 3.08912

It seems that channels 5/6/7/8/9 have increased interference here (in
comparison to the interference report for "without interference").

Either way it seems the driver/device doesn't reset the channel survey
information.

+ath9k-devel
Can someone from ath9k-devel confirm if ath9k doesn't reset survey
data after disassociation, etc? It looks to be the case here.

@Sarah: Can you reload the driver before you try each ACS run and see
if the issue persists, please?

Micha?.

Hi Michal,

first thank you for your fast answer.
I figured out that I always have this issue with the large numbers on the channel in use. I can prove this and my results by using iw dev wlan0 survey dump. So it seems like these results of acs are correct, but the driver does not reset the survey data.
The version I made the tests with is 3.0. I did an update to the latest version (v3.10.4), but there I have the same issues.

I tried, what you suggested (with modprobe) with the old version of the ath9k driver.
If I start hostapd too fast after the driver reload, I get one of the two following error message after the last survey. Also hostapd does always, if one of these errors occur, 7 instead of 5 surveys. It jumps to survey 2 again after survey 3.

the errors are either:
ACS: Trying survey-based ACS
ACS: Survey is missing noise floor
ACS: Channel 7 has insufficient survey data
ACS: Surveys have insufficient data
ACS: All study options have failed
ACS: Failed to start

or this one after the calculation of ACS:
wlan0: Flushing old station entries
nl80211: flush -> DEL_STATION wlan0 (all)
nl80211: Station flush failed: ret=-22 (Invalid argument)
wlan0: Could not connect to kernel driver
wlan0: Deauthenticate all stations
nl80211: No monitor socket available for wpa_driver_nl80211_send_mntr

If I just wait a moment with the start of hostapd after reloading the driver than I don't get an error.

With the latest ath9k driver, hostapd always does 5 surveys, but I still can get the error that acs fails to start.

Best regards,
Sarah

This message is subject to the following terms and conditions: http://www.barco.com/en/maildisclaimer