On Tue, 2006-01-17 at 17:20 +0800, fiasco wrote:
> Maybe I did not express clearly, since English is not my native
> language. :)
>
> Of cource we can ensure there is no conflict in application level,
> such as
> we do not call SendMsg.send() until we get sendDone() .
>
> But I am not sure whether it is enough to guarantee there is no
> conflict
> in system level.
>
> For instance, back to the previous code,
>
> if(!call SendSensorsData.send(TOS_BCAST_ADDR,sizeof
> (SensorsMsg ),&gMsgBuffer))
> {
> call Leds.set(7);
> }
>
> In a surge-like application, when a mote receives a bcast message, it
> has two messages to
> send out. One is the bcast message, since it is required to
> rebroadcast it. The other is the
> sensoring-data, which will be sent using up-code.
>
> Is it possible that these two send() conflict each other, which cause
> sending failed?
>
> If so, how can we aovid it, these two actions are in different level,
> one is in application level, the other is in library level.
Two ways: used QueuedSendC and make sure you don't fill the queue, or
use the sendDone() event of GenericComm:
configuration GenericComm
{
provides {
interface StdControl as Control;
// The interface are as parameterised by the active message id
interface SendMsg[uint8_t id];
interface ReceiveMsg[uint8_t id];
// How many packets were received in the past second
command uint16_t activity();
}
uses {
// signaled after every send completion for components which wish to
// retry failed sends
event result_t sendDone();
}
}
E.g.:
result_t tryToSend() {
if (state == SENDING) {
return FAIL;
}
if (call SendMsg.send(...) == FAIL) {
state = WAITING_TO_SEND;
}
else {
state = SENDING;
}
}
event result_t SendMsg.sendDone(...) {
state = IDLE;
}
event result_t sendDone() {
if (state == WAITING_TO_SEND) {
tryToSend();
}
}
Phil
_______________________________________________
Tinyos-help mailing list
[email protected]
https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
