Hi,
This patchset is to update my previous one: [alsa-devel] [RFC][PATCH] ALSA: firewire-lib: support Linux tracing to dump a part of packet data http://mailman.alsa-project.org/pipermail/alsa-devel/2016-March/106146.html
Changes: - Drop condition statement enabled only on tracing mode. - Compute cycle count correctly and report. - Add more parameters to report. - Node IDs of source/destination. - The value of isochronous channel to use. - Add parse script written by Python3.
The motivation of this patchset is to check timestamp processing of ALSA firewire stack, as writing in this message. http://mailman.alsa-project.org/pipermail/alsa-devel/2016-March/105608.html
As a result, I confirm it compliant to IEC 61883-1/6 in these points: - Transfer 8,000 packets per second of presentation timestamp. - Transfer the same number of data blocks as sampling transfer frequency per second of presentation timestamp. - The gap of sequence of calculated presentation timestamp correctly describes the elapsed time of data blocks in a packet. - The offset between cycle start and presentation timestamp is proper, including transfer delay.
Therefore, I believe that issues on dice-based models come from specific quirks, not due to bugs of ALSA firewire stack. I guess that the presentation timestamps for the models should be somewhat calculated with cycle count or the other parameters. Current implementation calculates it ignoring the cycle count.
Further investigation is required to solve it.
Regards
Takashi Sakamoto (3): firewire-lib: compute the value of second field in cycle count for IT context firewire-lib: compute the value of second field in cycle count for IR context firewire-lib: support Linux tracing to dump a part of packet data
sound/firewire/Makefile | 3 ++ sound/firewire/amdtp-stream-trace.h | 98 +++++++++++++++++++++++++++++++++++++ sound/firewire/amdtp-stream.c | 75 +++++++++++++++++++++------- 3 files changed, 159 insertions(+), 17 deletions(-) create mode 100644 sound/firewire/amdtp-stream-trace.h