A couple of months ago I was playing with getting recursion
fixed with the function tracer. There's a lot of places that could
use improvements, and I also found a few cases where the recursion
protection would have false positives causing missed function tracing.

I implemented a per context bit checking (Normal, SoftIRQ, IRQ and NMI).
Using 4 bits in the task_struct trace_recursion field, I was
able to fix these issues as well as improve the speed of the function
tracer and ring buffer.

Before the patches I had:

 # echo function > /debug/tracing/current_tracer
 # for i in `seq 10`; do ./hackbench 50; done
Time: 12.028
Time: 11.945
Time: 11.925
Time: 11.964
Time: 12.002
Time: 11.910
Time: 11.944
Time: 11.929
Time: 11.941
Time: 11.924

(average: 11.9512)

Now we have:

 # echo function > /debug/tracing/current_tracer
 # for i in `seq 10`; do ./hackbench 50; done
Time: 9.712
Time: 9.824
Time: 9.861
Time: 9.827
Time: 9.962
Time: 9.905
Time: 9.886
Time: 10.088
Time: 9.861
Time: 9.834

(average: 9.876)
    
This is a total of a 17.4% improvement in function tracing!


Steven Rostedt (8):
      ftrace: Fix global function tracers that are not recursion safe
      ftrace: Fix function tracing recursion self test
      ftrace: Optimize the function tracer list loop
      ftrace: Add context level recursion bit checking
      tracing: Make the trace recursion bits into enums
      tracing: Avoid unnecessary multiple recursion checks
      ftrace: Use only the preempt version of function tracing
      ring-buffer: User context bit recursion checking

----
 kernel/trace/ftrace.c          |   82 ++++++++++++++++---------
 kernel/trace/ring_buffer.c     |   85 +++++++++++++++++--------
 kernel/trace/trace.h           |  133 ++++++++++++++++++++++++++++++++++++----
 kernel/trace/trace_functions.c |   61 +++++-------------
 kernel/trace/trace_selftest.c  |    3 +-
 5 files changed, 248 insertions(+), 116 deletions(-)
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Reply via email to