Deadlock while using stack tracer(ftrace)

Issue

• When writing debugfs/tracing/stack_max_size while reading debugfs/tracing/stack_trace, we got deadlock. Because the 2 actions can be racy, and result in holding the same lock twice.

• Step to Reproduce:

1. echo 1 > /proc/sys/kernel/stack_tracer_enabled
2. mount -t debugfs xxx /sys/kernel/debug
3. for ((; ;)) ; do echo 1 > /sys/kernel/debug/tracing/stack_max_size; done
4. for ((; ;)) ; do cat /sys/kernel/debug/tracing/stack_trace > /dev/null; done
5. Then, wait for a while (until transparent hugepage is used.)
6. move the task to /cgroup/A/
7. rmdir /cgroup/A/B

• Actual Results:
  • lockup
• Expected Results:
  • Never lockup