castor_fou wrote: 
> ... any effect in the number of connectivity losses per day...
> I don't follow completely your explanation under Log File Analysis. ...
> Ping 2s-1q6f Events, Fails[0..7] 8154s :    Qr:0 Fr:27   Wr:0 Wc:0  [
> 2299 10 0 0 0 0 27 0 ]
> ...Are you confident we could get an actual solution packaged in
> community firmware fixing all our disconnection issues?
The number of connectivity losses may go up if the quick reset effects
don't last very long, but 'hopefully', those brief outages they will not
result in music interruptions. You are experiencing 12 resets per hour,
that's high, especially considering your very good signal level of -54
dBm.

Too bad the script is so complicated as it is. The modest mitigation
portion is just detection, a full "brute force" reset, one of many
possible 'quick' resets, and making sure that the os software stack is
running (e.g. 'wpa_cli' occasionally disappears, and we relaunch it).
Most of the script is data capture, which takes a lot of processing
time. Some of the data is for validating the script, but most of it is
for evaluating the environment and the radio's response to the outages.
It is not practical to do other than the most basic analysis on the
radio using a script. So we have to do it ourselves, perhaps by
processing the captured data with some (e.g., your) desktop analysis
code. 

Explaining the data in an understandable and concise manner is not easy,
and there are so many different cases to explain, especially as the
software evolves into new areas of investigation. For example, we are
not logging actual music interruptions for correlation to outages, but
the most important measure is music interruptions, and is critical to
evaluating different mitigation methods.

"Confident" would be going too far. The attitude is "Hopeful." The real
solution would be for Qualcomm/Atheros to investigate and patch their
(by necessity closed) chip firmware, if not to prevent loss of receiver
sensitivity then at least to monitor it and offer some much quicker
method for resetting or adjusting the systems on the chip so that they
do not repeatedly fail to respond to packets. What we can do is build on
the expressly open yet unavailable source os driver stack to implement
detection and mitigation that can work far more quickly and efficiently
than the current mitigation script, so that the usual buffering would
make very brief degradations or interruptions unnoticeable. Whether the
latter mitigation can be adequately justified, tested, and validated and
thus be fit for inclusion into an official software release is for
others to decide.


------------------------------------------------------------------------
POMdev's Profile: http://forums.slimdevices.com/member.php?userid=70558
View this thread: http://forums.slimdevices.com/showthread.php?t=111663

_______________________________________________
Radio mailing list
[email protected]
http://lists.slimdevices.com/mailman/listinfo/radio

Reply via email to