The following patches fix opening perf.data files that have timestamps (ordered data), aren't recorded with --per-thread, and that have discontinuous data in a single aux trace buffer.
I have some open questions: * Can cs_etm__update_queues() be removed from cs_etm__flush_events()? * Why does the second commit start making some files process correctly? * Is it ok to wait for the flush to start processing? Previously processing happened when the first aux record was delivered to cs_etm__process_event(). * Do the aux records need to be saved into a new buffer or can they be pulled from elsewhere?
I also have some further changes to make to make per-thread mode work where the cpu field of the sample is set to -1. And when there are no timestamps cs_etm__process_timeless_queues() is used, which is a completely different code path.
Thanks James
James Clark (5): perf cs-etm: Split up etm queue setup function perf cs-etm: Only search timestamp in current sample's queue. perf cs-etm: Save aux records in each etm queue perf cs-etm: don't process queues until cs_etm__flush_events perf cs-etm: split decode by aux records.
tools/perf/util/cs-etm.c | 200 +++++++++++++++++++++++---------------- 1 file changed, 121 insertions(+), 79 deletions(-)