Hi Peter, Markus reported that PT is losing a lot of its data in cpu-wide mode. There are two problems in play here, both are amplified by the fact that cpu-wide events don't normally get re-scheduled, so if we don't disable them when their buffers are full, they get forever stuck. Hence two patches. I propose to put these through perf/urgent and also stable trees so that people with older kernels still get to collect humongous amounts of PT traces in cpu-wide mode.
Alexander Shishkin (2): perf/x86/intel/pt: Generate PMI in the STOP region as well perf: Disable the event on a truncated AUX record arch/x86/events/intel/pt.c | 2 ++ kernel/events/ring_buffer.c | 10 +++++++++- 2 files changed, 11 insertions(+), 1 deletion(-) -- 2.8.1
