Wireshark · Wireshark-dev: [Wireshark-dev] Computing average wireless signal strength off with tshark

["Luis R. Rodriguez" <mcgrof@xxxxxxxxx>]

Using tshark from svn (today) with ath5k from wireless-testing (today)
I'm noticing some reporting of a signal value which is completely off
when using the statistics feature (io,stat). At first I though it was
ath5k so I went to verify we don't report huge signal values and we
don't. Right before we hand of the skb to mac80211 I added checked the
signal value and in all cases it was negative and with in reasonable
range -- my awk script says -55.219, for instance.

Now if I apply the same awk scrip to the output of 'tshark -V -i moni0
> dump' and then grep for "SSI Signal" I get a reasonable average.
This at least allows me to rule out a driver issue.

Instead of running awk scripts I'd like to just use tshark to compute this.

tshark -i moni0 -q -z
"io,stat,1,AVG(radiotap.dbm_antsignal)radiotap.dbm_antsignal,AVG(radiotap.datarate)radiotap.datarate"

This prints the average signal and rate every second, if you use it
you'll see the first column is gross:

Capturing on moni0
^C1355 packets captured

===================================================================
IO Statistics
Interval: 1.000 secs
Column #0: AVG(radiotap.dbm_antsignal)radiotap.dbm_antsignal
Column #1: AVG(radiotap.datarate)radiotap.datarate
                |   Column #0    |   Column #1
Time            |            AVG |            AVG
000.000-001.000          14035780                4
001.000-002.000          16085960                4
002.000-003.000          16976095                5
003.000-004.000          19088682                5
004.000-005.000          19792417                5
005.000-006.000          49367380                5


If I decrease the interval to 0.001 I get more reasonable values:

tshark -i moni0 -q -z
"io,stat,0.001,AVG(radiotap.dbm_antsignal)radiotap.dbm_antsignal,AVG(radiotap.datarate)radiotap.datarate"

5 packets captured

===================================================================
IO Statistics
Interval: 0.001 secs
Column #0: AVG(radiotap.dbm_antsignal)radiotap.dbm_antsignal
Column #1: AVG(radiotap.datarate)radiotap.datarate
                |   Column #0    |   Column #1
Time            |            AVG |            AVG
000.000-000.001               -40                2
000.001-000.002               -41                2
000.002-000.003                 0                0
000.003-000.004          2147483602                3
000.004-000.005                 0                0
000.005-000.006                 0                0
000.006-000.007                 0                0
000.007-000.008                 0                0
000.008-000.009                 0                0
000.009-000.010                 0                0
000.010-000.011                 0                0
000.011-000.012                 0                0
000.012-000.013                 0                0
000.013-000.014                 0                0
000.014-000.015                 0                0
000.015-000.016                 0                0
000.016-000.017                 0                0
000.017-000.018                 0                0
000.018-000.019                 0                0
000.019-000.020                 0                0
000.020-000.021                 0                0
000.021-000.022                 0                0
000.022-000.023                 0                0
000.023-000.024                 0                0
000.024-000.025                 0                0
000.025-000.026                 0                0
000.026-000.027                 0                0
000.027-000.028               -64                2
===================================================================

But notice that big fat 2147483602. And this is just 5 packets, if you
let it sit for a few seconds you'll see a lot of these spread out and
that ruins the average computation.

I'd send a patch but I'm not yet sure where to poke. If you are aware
where to look at let me know and I'll check it out. I'd like to fix
this or see this fixed.

  Luis