Re: isochronous cycle inconsistent
Brought to you by:
aeb,
bencollins
From: Stefan R. <st...@s5...> - 2011-12-17 17:22:11
|
On Dec 16 Carl Karsten wrote: > On Fri, Dec 16, 2011 at 2:53 AM, Clemens Ladisch <cl...@la...> wrote: > > Carl Karsten wrote: > >> [ 112.817782] firewire_ohci: isochronous cycle inconsistent > > > > FireWire has isochronous transfers, which rely on a common bus time. > > (AV devices and all host controllers are capable of iso transfers.) > > To synchronize the time on the bus, one of the iso-capable nodes, the > > cycle master, broadcasts its current time every 125 µs. > > > > As long as the two controllers are not connected, each one is cycle > > master. When you connect them, one loses its status as cycle master and > > receives its time from the other one. The resulting change in time is > > reported as "isochronous cycle inconsistent". This is not a bug when > > it doesn't happen during normal operation, and the kernel probably > > shouldn't report it when it is expected. firewire-ohci's irq_handler() provides a rather awkward context if we wanted to determine whether this is expected or not. We could make the message conditional on whether there is any IT or IR DMA context in use. (Better: IT or IR DMA with cycleMatch control flag enabled, but that would require more code to find out.) See comment in irq_handler() and OHCI clauses 9.5.1 and 10.5.1. Better yet, we just write a cycleInconsistent handler. > >> [ 143.392293] firewire_core: giving up on config rom for node id ffc1 > >> [ 165.404068] firewire_core: giving up on config rom for node id ffc0 > > > > Is your problem that the cards don't eventually see each other? > > > > > > correct. > > I also had problems with 2 cams hooked up. I used the loop to remove > the cams as a source of problem. I'll hook up the cams and post.. Please # echo 3 > /sys/module/firewire_ohci/parameters/debug before you plug the cameras in or before you plug the two cards together. -- Stefan Richter -=====-==-== ==-- =---= http://arcgraph.de/sr/ |