I have checked in a fix for the cc2420x acknowledgement loss issue. Please update to the svn head from google code. I have tested it with apps/test/cc2420/TestAck, compiled for the "telosb cc2420x" target. Let me know if it works for you, and thanks for reporting this issue.
Janos On Sun, Jul 17, 2011 at 8:40 PM, Xiaohui Liu <[email protected]> wrote: > The application compiles correctly with cc2420x stack, but after it runs for > a while on nodes, nodes die and the application collapses. Before its > collapse, for the same sender and receiver, very few acks are received using > cc2420x stack (i.e., < 10%) but most acks (i.e., > 90%) are received with > cc2420 stack. > > I added Packet.clear() and the result does not change. > On Sun, Jul 17, 2011 at 8:33 PM, Janos Sallai <[email protected]> > wrote: >> >> Hi, >> >> What do you mean by "The application crashes if it is using the >> cc2420x stack"? Does it compile? Does it run on a mote, but no acks >> are received? >> >> Do you call Packet.clear() before reusing a message_t buffer? The >> proper way of using non-default metadata settings (acks, tx power, >> etc.) in the rfxlink stack is: >> - calling Packet.clear() on the reused buffer >> - filling in the payload >> - setting parameters in the metadata (acks, tx power, etc.) >> - calling AMSend.send() >> >> >> > 1) use cc2420x stack: it's necessary to get all functionality of CTP, >> > i.e., PacketAcknowledgements and LinkPacketMetadata and possibly others, >> > to work.under the new stack >> You can live without LinkPacketMetadata: modify CtpP such that it >> wires DummyActiveMessageP instead of CC2420ActiveMessageC if the >> cc2420x stack is used. >> >> > 2) use default cc2420 stack: fix its timestamping (both packet >> > timestamping >> > and packet-level time sync) by porting the timestamping from cc2420x to >> > cc2420 >> > If you have any suggestion on which venue to go and how to implement it, >> > please let me know. Thanks, again. >> There's a TEP on the cc2420 stack, there's the cc2420 data sheet, and >> there's the source code. You might get some answers from the three or >> four people familiar with that code from the mailing list as well. >> >> Janos > > > > -- > -Xiaohui Liu > _______________________________________________ Tinyos-help mailing list [email protected] https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
