Use ftrace

perf ftrace is a simple wrapper for kernel’s ftrace functionality, and only supports single thread tracing now. Check following example: 

  1. # perf ftrace ./lock
  2.    2)               |  switch_mm_irqs_off() {
  3.    2)   1.792 us    |    load_new_mm_cr3();
  4.    2)   9.157 us    |  }
  5.  ------------------------------------------
  6.    2)    <idle>-0    =>   <...>-5785
  7.  ------------------------------------------
  9.    2)               |  finish_task_switch() {
  10.    2)               |    _raw_spin_unlock_irq() {
  11.    2)   ==========> |
  12.    2)               |      smp_irq_work_interrupt() {
  13.    2)               |        irq_enter() {
  14.    2)   0.612 us    |          rcu_irq_enter();
  15.    2)   0.631 us    |          irqtime_account_irq();
  16.    2)   0.457 us    |          preempt_count_add();
  17.    2)   3.818 us    |        }
  18.    2)               |        __wake_up() {
  19. ......

-T option can be used to specify only tracing interested functions: 

  1. # perf ftrace -T __kmalloc ./lock
  2.   20) + 17.698 us   |  __kmalloc();
  3.   20)   3.167 us    |  __kmalloc();
  4.   20)   0.952 us    |  __kmalloc();