Bad host-vased AP performace -- and qouple of questions about
FreeBSD WiFi stack tuning
Bernhard Schmidt
bschmidt at freebsd.org
Tue Aug 9 06:29:56 UTC 2011
On Monday, August 08, 2011 20:07:20 Lev Serebryakov wrote:
> Hello, Freebsd-wireless.
>
> I have host-based AP, which is running FreeBSD 8.2-STABLE (i386) and
> equipped with Senao EMP 8602 Plus-S Mini-PCI WiFi card. It is Atheros
> 5413 based 600mW card:
>
> ath0: <Atheros 5413> mem 0xa0060000-0xa006ffff irq 15 at device 17.0 on pci0
> ath0: [ITHREAD]
> ath0: AR5413 mac 10.4 RF5413 phy 6.1
>
> I'm using it with 10dBi omnidirectional antenna in 802.11g (2.4Ghz)
> standard.
>
> It worked well some time ago, but now its performance is horrible.
>
> Main clients are Intel-based notebooks with Intel 3945ABG cards,
> Windows-based.
>
> Typical "Real" (file copy) bandwidth with only one client active is
> 500-600KiB/s. Windows shows "Excellent" signal level and 54Mbit.
> InSSIDer (Windows-based WiFi monitoring software) shows RSSI between
> -50 and -40 (dBm)m but I can not trust it, as it shows some very
> distant APs with RSSI 100-120 (yes, +100 - +120), and it is nonsense.
>
> I've bring additional notebook with Linux and Kismet installed.
> Kismet shows, that my AP oscillate between -80 and -40 dBm RSSI
> permanently, even if here is connected (and doenloading!) client
> connected. Signal level jumps up and down, and often Kismet receive
> frames with broken SSID (one-two chars are replaced by other).
>
> "ifconfig wlan0 list sta" on AP shows something like this (running in
> cycle with "sleep 1", one client copy big file, 600-700KiB/s):
>
> 00:18:de:08:e8:1d 1 9 24M 14.5 0 23084 35488 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 18M 11.0 0 23612 39952 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 13.5 0 24116 44320 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 13.0 0 24688 49104 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 13.0 0 25341 54720 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 13.0 0 25813 58640 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 54M 11.0 0 26092 61104 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 54M 12.0 0 26782 1424 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 54M 14.0 0 27241 5312 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 14.0 0 27615 8496 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 12.5 0 27872 10768 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 14.0 0 28302 14432 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 13.5 0 28912 19600 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 18M 14.0 0 29501 24624 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 13.5 0 30305 31440 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 24M 13.5 0 31971 45584 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 36M 13.0 0 32429 49472 EPS AQE RSN WME
> 00:18:de:08:e8:1d 1 9 11M 12.5 0 32758 52288 EPS AQE RSN WME
>
> Also, tcpdump shows, that AP announce only speeds up to 18Mbit (tag
> number 1 in beacon frame, list of supported speeds -- 1, 2, 5.5, 11,
> 6, 9, 12, 18).
>
> Here is my setting of AP:
>
> ifconfig_wlan0="inet 192.168.135.1 netmask 255.255.255.0 mode 11g channel 9 ssid home.serebryakov.spb.ru country ru regdomain row txpower 30"
>
> So, Questions are:
>
> (1) Hopping signal strength and total terrible WiFi network
> performance -- is it broken hardware or something strange in
> FreeBSD code?
Either the hardware really has some issues or the signal strength is
just too high? I've seen that.. can you try setting a lower txpower
for a test?
> (2) Why card announce only 18Mbit, but not more?
Seems like you're looking at the "basic rates", if you examine the
beacon you should find another IE which contains all rates.
> (3) What "txpower" setting means? dBm is relative unit. My 600mW card
> lists "30" as max, and simple notebook Intel card lists "30" as
> max too. Is it relative to max power of card (100mW for standard
> cards, 400-600-1000mW for high-power "procider-grade" cards from
> Ubiquiti and other such wendors, like Senao) or 1W or 100mW or what?
> Manual pages didn't give answer.
It probably isn't as absolute as you might think. net80211 currently
has no way for the driver to tell it what the max is (except setting
the maxtxpower for each channel). Depending on which value you look
at it might even in half-dbm. The iwn(4) cards for example have a
txpowerlimit of 15 dBm which is 30 in half-dBm.
The txpower command for ifconfig is ment to pass over that number
to the driver unmodified, the driver applies whatever limit it has.
For ath(4) afaik that means you get min(txpower, cardlimit).
> (4) What units are uses for RSSI output in "ifconfig list sta"?
RSSI is something absolutely unspecified by design. The field itself
is defined to be 8 bit width, but that's it, no unit whatsoever.
ath(4) for examples reports (64?) unique values which are mapped
through a table to dBm. For other cards we try to mimic that
behaviour.
Basically, if any of the ifconfig commands show something negative
it's dBm, if it's possitive it should be ratio between something.
For RSSI case it should be signal to noise (SNR), haven't explicitly
checked that, so I might be wrong.
--
Bernhard
More information about the freebsd-wireless
mailing list