Hi Tom, I noticed that you created an error_log file in every instance, but they all show the same errors. These three patches make it so that the errors appear in the instance directory that they happened in. If you write a bad error to /sys/kernel/tracing/instance/foo/events/sched/sched_switch/hist It appears only in /sys/kernel/tracing/instances/foo/error_log Which I think is the proper approach, especially instances should not affect the top directory or other instances. For those errors that do not have an associated instance (creating a kprobe/uprobe event or perf), a NULL passed to tracing_log_err() will result in the error message in the top level error message. Do you (or Masami) have any issues with this patch set? If not, please add a "reviewed-by" or "acked-by" and I'll add it to your patch series and push them to for-next (after more testing). If this isn't obvious, this patch series is on top of: http://lkml.kernel.org/r/cover.1554072478.git.tom.zanussi@xxxxxxxxxxxxxxx Actually, I added it right after patch 5 of that series (before the selftests and documentation). Thanks! -- Steve Steven Rostedt (VMware) (3): tracing: Add trace_array parameter to create_event_filter() tracing: Have histogram code pass around trace_array for error handling tracing: Have the error logs show up in the proper instances ---- kernel/trace/trace.c | 55 +++++++++----- kernel/trace/trace.h | 8 +- kernel/trace/trace_events_filter.c | 25 ++++--- kernel/trace/trace_events_hist.c | 145 ++++++++++++++++++++---------------- kernel/trace/trace_events_trigger.c | 3 +- kernel/trace/trace_probe.c | 2 +- 6 files changed, 142 insertions(+), 96 deletions(-)
