Method reassignFailure

Control/L3LogicalChannel.cpp:220–245

This occurs on the channel being assigned from. We need to release this channel and nextChannel. Release of nextChan also occurs if a channel drops out of the main service loop and the nextChan state is still ReassignTarget

// We need to release this channel and nextChannel.
// Release of nextChan also occurs if a channel drops out of the main service loop and the nextChan state is still ReassignTarget
void L3LogicalChannel::reassignFailure()
{
	ScopedLock lock(gMMLock,__FILE__,__LINE__);
	LOG(DEBUG) << this;
	// Clean up the next chan we were trying to reassign to.
	if (mNextChan) {
		devassert(mNextChan->isTCHF());
		mNextChan->chanSetState(L3LogicalChannel::chRequestRelease);	// This will probably block the chan for 30 seconds.
		// If we never received the ESTABLISH primitive on nextChan, then the service loop is not runnning,
		// so it will never detect the change of state on nextChan, so we must free the channel here.
		// The service loops check dcch->chanRunning(), so they will not do anything that might try to use the Context we are freeing..
		mNextChan->chanFreeContext();
		mNextChan = NULL;
	} else {
		LOG(ERR) << "reassignment failure but no nextChan? "<<this;
	}


	// channel()->l3sendm(L3ReleaseComplete(l3ti,l3cause));	// Release the transaction identifier.
	// We might as well drop the whole channel.
	// Old code had cause 4, but that does not seem right:
	// RR Cause 0x04 -- "abnormal release, no activity on the radio path"
	// Caller does this now.
	//l3sendm(GSM::L3ChannelRelease(L3RRCause::NoActivityOnTheRadio));
	// The MS is supposed to release the channel, which will send a RELEASE primitive up to Layer3 to close the channel.
}

// Both old and new L3LogicalChannel point to the same MMContext.
// The message arrives on the new channel, but we run this function on the old channel