Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 23 Jul 2005 19:44:03 -0700
From:      Sam Leffler <sam@errno.com>
To:        Michal Mertl <mime@traveller.cz>
Cc:        freebsd-current@freebsd.org
Subject:   Re: Recent fragility with if_wi, 802.11 adhoc/wep, and Tiger
Message-ID:  <42E30073.1080202@errno.com>
In-Reply-To: <1121987685.61017.31.camel@genius1.i.cz>
References:  <20050719094905.F15510@fledge.watson.org>	<1121771151.764.42.camel@genius1.i.cz> <42DDD710.4030503@errno.com>	<1121855884.796.8.camel@genius1.i.cz> <42DE648B.1060402@errno.com>	<1121881805.929.38.camel@genius1.i.cz> <1121987685.61017.31.camel@genius1.i.cz>

next in thread | previous in thread | raw e-mail | index | archive | help
Michal Mertl wrote:
> Michal Mertl wrote:
> 
>>Sam Leffler wrote:
>>
>>>Michal Mertl wrote:
>>>
>>>>I run an ath card in hostap mode and several wireless clients connect to
>>>>it and are on the same IP network. The ping from one client to another
>>>>is slow yet both ping the AP fine. I think that in this situation the
>>>>bridging is done by ath (in HAL?) and configured by 'ifconfig apbridge'.
>>>
>>>The bridging is done in the net80211 layer, not "in the card".  I will 
>>>test, thank you.
>>
>>Thanks for the correction.
>>
>>And thank you for looking into it.
> 
> 
> Some new findings:
> 
> I can confirm that the atheros based client really connects much worse
> than an IPW based one. When I restart the atheros AP, the ipw card
> connects immediately after the AP is back but the atheros client either
> never connects or it will take long time (I didn't wait long enough). It
> connects immediately after I issue ifconfig down/up on it.

Once a station has failed to associate with an ap it marks it "bad" and 
won't try again for a while.  I need to bring in some improvements from 
another tree that improve this area of the code; we wait too long to 
re-enable going back to an ap.  Marking the interface down-then-up 
clears this state so you can immediately re-associate.  The ipw firmware 
does not use this code so it behaves differently.

> 
> More interesting finding that I have is about the bridging issue. I
> wasn't able to find which debug setting (via dev.ath.0.debug or
> net.wlan.0.debug) will show me any usefull information. Anyways it now
> seems to me that I was wrong saying that it works at all. The AP bridges
> the packets only when there is another IP communication between the AP
> host and one of the clients. It seems to me that the bridged packets are
> queued somewhere and sent only when there are some non bridged.

I don't recall what "the bridging issue" was but use 80211debug and 
athdebug to manipulate these sysctls using mnemonics--80211debug -? will 
list the controls.  Code is in tools/tools/ath.  At some point these 
tools probably should be integrated with another program and not stay 
hidden in the tools directory.

> 
> Test conditions - I have 192.168.1.1 on the AP, .2 on the IPW notebook
> and .3 on the atheros client. The settings of ath0 on AP are: "mode 11b
> mediaopt hostap channel 1 ssid test_ap_xx". The settings on clients are
> almost the same except there I don't issue any mediaopt. I hope I'm not
> doing anything extra stupid :-). The nodes are just several centimeters
> apart from each other and I only have tiny antennas. When the only IP
> communication is the ping from 192.168.1.2 to .3 (between the clients) I
> don't get any answer. When I ping at the same time from between any of
> the clients and the AP it works. When I let the first ping run for
> several seconds and then start the second one I get all the answers at
> the same time.

I'm guessing this was the problem where frames bridged internal to the 
net80211 layer were not dispatched immediately.  If so we both know this 
was fixed.

> 
> This is probably complete nonsense but I think that what I'm
> experiencing looks as if the bridged packets weren't generating
> interrupts or something.

The frames were queued for transmit but the driver was never told to 
send them; hence the random delay.

> 
> I've got serial consoles hooked to the machines and am able to sprinkle
> some debug prints somewhere if required.
> 
> 
> Michal
> 
> 
> _______________________________________________
> freebsd-current@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-current
> To unsubscribe, send any mail to "freebsd-current-unsubscribe@freebsd.org"
> 




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?42E30073.1080202>