ftrace: Initialize the ftrace profiler for each possible cpu
authorMiao Xie <miaox@cn.fujitsu.com>
Mon, 16 Dec 2013 07:20:01 +0000 (15:20 +0800)
committerSteven Rostedt <rostedt@goodmis.org>
Mon, 16 Dec 2013 15:53:46 +0000 (10:53 -0500)
Ftrace currently initializes only the online CPUs. This implementation has
two problems:
- If we online a CPU after we enable the function profile, and then run the
  test, we will lose the trace information on that CPU.
  Steps to reproduce:
  # echo 0 > /sys/devices/system/cpu/cpu1/online
  # cd <debugfs>/tracing/
  # echo <some function name> >> set_ftrace_filter
  # echo 1 > function_profile_enabled
  # echo 1 > /sys/devices/system/cpu/cpu1/online
  # run test
- If we offline a CPU before we enable the function profile, we will not clear
  the trace information when we enable the function profile. It will trouble
  the users.
  Steps to reproduce:
  # cd <debugfs>/tracing/
  # echo <some function name> >> set_ftrace_filter
  # echo 1 > function_profile_enabled
  # run test
  # cat trace_stat/function*
  # echo 0 > /sys/devices/system/cpu/cpu1/online
  # echo 0 > function_profile_enabled
  # echo 1 > function_profile_enabled
  # cat trace_stat/function*
  # run test
  # cat trace_stat/function*

So it is better that we initialize the ftrace profiler for each possible cpu
every time we enable the function profile instead of just the online ones.

Link: http://lkml.kernel.org/r/1387178401-10619-1-git-send-email-miaox@cn.fujitsu.com
Cc: stable@vger.kernel.org # 2.6.31+
Signed-off-by: Miao Xie <miaox@cn.fujitsu.com>
Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
kernel/trace/ftrace.c

index 0e9f9ea..72a0f81 100644 (file)
@@ -775,7 +775,7 @@ static int ftrace_profile_init(void)
        int cpu;
        int ret = 0;
 
        int cpu;
        int ret = 0;
 
-       for_each_online_cpu(cpu) {
+       for_each_possible_cpu(cpu) {
                ret = ftrace_profile_init_cpu(cpu);
                if (ret)
                        break;
                ret = ftrace_profile_init_cpu(cpu);
                if (ret)
                        break;