[maemo-developers] WLAN Horrible Roaming Performance (N800, OS2008), Software or Hardware Problem ?

[From: Kalle Valo]

"ext Benno Senoner" <benno.senoner at googlemail.com> writes:

>> An easy way to confirm, if this is a TCP problem or something else,
>> is to run ping in the background and see how long timeouts (ie. how
>> much you see packet loss) you get with it. ICMP ping does not have
>> any of problems that TCP has.
>
> Yes I did and indeed ping works much better. i tried ping -s 1024
> mylocalgatewayip (1024 byte long packets) and indeed it works ok.
> sometimes it loses 1-4 packets while roaming but afterwards the number
> of packet loss is usually small. when roaming around the wlan it is
> about 3-4% but the nokia wlan lockup remains, see below.

Ok, 3-4% packet loss with ping implies that roaming per se is working
quite well. Though I don't know how well TCP handles 3-4% packet loss,
but my guess is that it's not a problem.

But the lock ups you are describing is a different issue. Is that the
real problem here? Not the roaming performance but the fact that WLAN
connection gets stuck very easily and no data is transfered after
that.

I would like to get more information about the lockup situation so
that I can understand what's really happening:

o You earlier mentioned that disabling WLAN Power Save Mode does not
  help, so we can rule that out. That's good.

o Does the lockup happen with only one AP as well? I'm just wondering if the
  lockup is related to roaming.

o Does the lockup happen with other APs (different make and model)?

o What does UI show? Any errors?

>> As for the network lock ups we are experiencing (you have to
>> > disable,enable wlan on the n800 to make it work again)
>>
>> Does ping work when that happens? For example, you could try to ping
>> your gateway. Also can you send me the output from dmesg after you
>> experience this, please?
>
> when it locks up even ping stops responding, I get a message
> connection lost (from maemo).

I'm sorry, but I want to be clear here. Do you mean that N800 UI
displays "connection lost" banner? Please be very specific, because
all minor details count here.

But if you really see "connection lost" in UI, that's not a lockup but
an indication that the N800 lost the connection to the AP. And a
properly ported maemo application would handle that situation just
fine (ie. it would receive disconnected signal from libconic, would
close the TCP connections and the UI would not block etc.)

> I just tested it today with an xterm open running a constant ping
> and walking around the area. after 10min it stopped working (I was
> out of coverage area and returning) I did unfortunately not look at
> the dmesg output, I will try again and report my findings.

dmesg output is very important here. If you can get one, please send
it here. Also please describe, in detail, what UI shows you at the
time when dmesg output was saved.

>> other users reported it too as Luca Olivetti pointed out. and it
>> > seems like the problem and fix is described here:
>> > http://internettablettalk.com/forums/showpost.php?p=134914&postcount=15
>> > at least for the 770 the fix seems to exist,
>>
>> What I read from the link, someone had written a workaround to try
>> again whenever the chip is responding. That would good a feature,
>> but I would like to get more information about what's happening in
>> this case.
>
> i will investigate this issue and will look at the sources of the
> patch and see if it can be adapted to the n800 wlan driver (I can ask
> for help on the project's mailinglist) i hope it will solve the issue
> as this problem occurs several times per day in my wlan environment.

If you create a patch, please send it to cx3110x-devel. I would like
to see all patches (even if it's just for testing or just a quick
hack) on that mailing list so that others, including me, can easily
find them.

But I would like to remind here that even if other people see similar
symptoms somewhere else, it does not automatically mean that the
problem is the same. Lots of different WLAN problems have the symptom,
that is data transfer gets stuck. So a fix helping someone else does
not necessarily mean that it will fix your problem. 

This is the reason why I need some much information so that I can
really analyse the problem and I can be sure that I'm fixing the
correct problem.

> PS: meanwhile in order to automate WLAN lockup recovey, how could I
> write a script which basically does the same like clickingon the
> connectivity icon, basically bring down wlan and then bring it back
> up. I tried with ifdown wlan0 ; ifup wlan0 but it does not work it
> says no such device. any hints ? some commands to send to wcond ?

'ifdown wlan0' doesn't work because wlan0 is not specified in
/etc/network/interfaces. You need to run 'ifconfig wlan0 down', but
that doesn't work because wlancond does not follow wlan0 interface
status.

I think the easiest would be to send disconnected signal to wlancond
using D-BUS, for example with dbus-send commandline tool.
Unfortunately I don't have any examples at hand, but osso-wlan sources
might help you.

-- 
Kalle Valo