Jan Kiszka wrote: > matrix_df hotmail wrote: >>> ... >>>> ########################################################### >>>> >>>> see attached ethereal file for tdma, tdmac packets >>> On which box did you capture, master or slave? Or on an independent >>> sniffer box? Only the sync messages are included in the dump, no RTcfg >>> frame which should pop up at least once per second. Either you skipped >>> them accidentally (the dump includes only 20 ms), or the RTcfg service >>> has some real problem here. >> I just shortend the capturing, because the sync is the only message. >> Messages were captured on master, but also verified by a third computer >> (sniffer). >> There are no "RTcfg frames" !! >> >> Who should send the RTcfg frames (invitation), the master or? >>
I'm writing this mail while running a latest CVS vulcano. I started the standard rtnet script for a master, cross-connected to another box, and launched a packet sniffer on the other side. I can confirm, that there is *not a single* RTcfg frame between the TDMA sync packets. Instead, I'm also facing "RTcfg: error -105 while sending stage 1 frame" after a while. [a bit later] Fine, I tracked it down: it's a combination of a bug in the RTAI RTDM layer together with an useless return value check in RTnet. That check is harmless over Xenomai as it effectively cannot trigger, but RTAI is returning mistakenly an error code in case rtdm_task_sleep_until is called with an already passed date. It's a bit funny as it is clearly stated in rtdm_task_sleep_until only a few lines above the bug what this service is supposed to return. I would accept the additional error EIDRM in case some crazy person stopped the system timer (though this will not be catched by RTnet anymore). But returning ETIMEDOUT is wrong here, specifically when comparing this service with its big POSIX brother clock_nanosleep(..., TIMER_ABSTIME, ...). I will post a patch for RTDM on the RTAI list later/tomorrow. Meanwhile, I committed a fix for RTnet that ignores the return value and should make it run over RTAI now - at least the logs contain all required frames, but I'm too tired to start up a full test scenario. Jan
signature.asc
Description: OpenPGP digital signature
