AW: AW: first patch: mISDN handling

ok, next time i will split the patches for easier understanding. if you require splitting of my first patch, let me know. for better understanding and review, i will explain the usleep replacement patch again. only commit it, if you fully understand the changes. the usleep is replaced by a timer, one for each time slot (signalling only ts). 1. whenever data is queued into the tx-queue, a tx-delay timer is checked: - the timer is not running: mISDN write event is triggered with BSC_FD_WRITE flag on file descriptor. - the timer is running: nothing is triggered, because we still have to wait until the tx-timer timed out 2. whenever a time out occurrs: - the BSC_FD_WRITE flag is set to trigger write event, even if tx-queue is empty. 3. whenever a write event is triggered: if it is ok to write (select() indicated that writing is possible), the tx-queue is checked: - if queue is empty: nothing is done, delay timer remains disabled, no BSC_FD_WRITE flag is set. - if queue is not empty: the first message is dequeued and sent to mISDN. the tx-delay timer is started and set for next write event. special cases: - if tx-queue is not empty, delay timer is not running and BSC_FD_WRITE flag is set (e.g. right after timeout event), subsequent data is queued and BSC_FD_WRITE flag remains set. - if tx-timer times out, but there is no data in the tx-queue, the BSC_FD_WRITE flag will be set. the write event handler always checks for data in the tx-queue. if it is empty, the delay timer will not be started and the BSC_FD_WRITE flag will be cleared. - the timers is not cleared on release of libbsc data structures. i expect all times to be removed on release of libbsc also. -----Ursprüngliche Nachricht----- Von: openbsc-bounces(a)lists.gnumonks.org [mailto:openbsc-bounces@lists.gnumonks.org] Im Auftrag von Holger Freyther Gesendet: Donnerstag, 23. April 2009 00:11 An: openbsc(a)lists.gnumonks.org Betreff: Re: AW: first patch: mISDN handling On Wednesday 22 April 2009 15:44:44 Holger Freyther wrote: Okay, I managed to do this today. It would be cool if you could split out your improved log messages (__func__) and extended error checking (making sure we have 30 bchan's) into a separate patch and I would be happy to merge this then, afterwards the killing of the usleep should be done. z.