status report layer1 and layer23
Andreas.Eversberg at versatel.de
Fri May 20 10:03:54 UTC 2011
i just fixed some locking issues the last days. fix will follow. it took
a bit longer, because there were some race conditions. it took up to
about one hour until it crashed. my way to detect the area where the
crash happened, was to turn on buzzer before that area, and turn it off
after that area. after many hours of approximation, i finally found out
that the major crash happend during _talloc_zero. (first it looks for a
free memory chunk, then it allocates it.) since it can be called from
all contexts (main, irq, fiq), it need to be locked against any
interrupt, otherwise the memory chunk can be assigned multiple times.
(the process of _talloc_free is "atomic" and requires no locking.)
because it seems pretty stable, i think it is time to merge some
branches into the master. (i made a 6 hours call yesterday. and no crash
after bugfix ever since.) i will do that together with sylvain, if we
find the time this weekend.
currently i use the jolly/voice together with the sylvain/traffic
branch. i am able to use an isdn phone togehter with linux-call-router
and make/receive calls. audio is passed both ways. i think this is a
stage where it actually become "usable". (if not moving arround.)
one of my major work for the next weeks/months will be the neighbour
cell measurement, cell re-selection, and handover. this is essential
when moving with the phone.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the baseband-devel