From: Jonathan W. <jw...@ph...> - 2008-10-16 22:45:57
|
Hi Patrick > > Can you provide a log where jack runs in realtime mode? > > Here you go. > Meant to send this to the list too. > > Can you tell from the logs what layer is failing? Is it ffado or something > in libraw1394? Well, for starters I see cannot lock down memory for jackd (Cannot allocate memory) JACK: unable to mlock() port buffers: Cannot allocate memory JACK: unable to mlock() port buffers: Cannot allocate memory This usually never helps, and in my experience it can sometimes cause other issues within jackd. Therefore I think this needs to be addressed before getting too concerned about other things. The mlock limit is set in the same place as the RT scheduling limit - often in PAM's limits.conf file. Jackd needs quite a bit of mlocked memory - I think on my systems I allow about 300 MB. Running down the log I see: (ISOXMT) Create RT thread 0x8074fa0 with priority 15 (ISORCV) Create RT thread 0x80750f8 with priority 8 Switching IsoManager to (rt=1, prio=14) and so forth. These are *very* low priority numbers and are almost certainly right in the middle of the priorities used by other things on the system. This won't be helping things. Could you supply your jackd command line? Are you using the "-P" option in addition to "-R"? If not you might like to try adding "-P 60" (what I use) or "-P 70" (what others use) and see if that makes any difference. Somewhat later in the log we get Execute: re-init dll due to too large tick diff: 1550484378.000000 >> 98304000.000000 and this is possibly what's causing the MOTU to get upset - it's rather picky regarding its incoming timestamps, and recovery from errors like this is something we're still working on. However, at a guess I'd say that the root cause of this error is at least related to the two issues previously noted (lack of mlock, low priorities), with the low priorities probably being the most significant. Therefore I think it's definitely worth addressing those first since there's a good chance that once they're fixed the other issues will go away. As an aside, I note in your logs that there's a device state error: prepare: Incorrect state, expected E_Initialized, got 4 so the ffado stream re-init mechanism may still have some gremlins in it. Regards jonathan |