[PATCH 2/3] powerpc/ftrace: Remove mod_return_to_handler

[Steven Rostedt]

On Wed, 17 Sep 2014 17:07:03 +1000
Anton Blanchard <anton at samba.org> wrote:

> mod_return_to_handler is the same as return_to_handler, except
> it handles the change of the TOC (r2). Add this into
> return_to_handler and remove mod_return_to_handler.

Adding this patch actually gave me some more output. Funny that?

electra login: INFO: rcu_sched self-detected stall on CPU { 1}  (t=5250 jiffies g=3579 c=3578 q=7)
Task dump for CPU 1:
trace-cmd       R  running task        0  3553   3550 0x00008014
Call Trace:
[c0000000054a6ce0] [c000000000012b34] .show_stack+0x104/0x260 (unreliable)
[c0000000054a6dc0] [c0000000000c2270] .sched_show_task+0xd0/0x150
[c0000000054a6e40] [c0000000000eedb0] .rcu_dump_cpu_stacks+0xe0/0x150
[c0000000054a6ee0] [c0000000000f2a48] .rcu_check_callbacks+0x4f8/0x8d0
[c0000000054a7020] [c0000000000f8118] .update_process_times+0x48/0xa0
[c0000000054a70b0] [c00000000010cfe8] .tick_sched_timer+0x88/0xd0
[c0000000054a7150] [c0000000000f8b1c] .__run_hrtimer+0xcc/0x2c0
[c0000000054a7200] [c0000000000f9a98] .hrtimer_interrupt+0x158/0x330
[c0000000054a7310] [c00000000001a6b8] .__timer_interrupt+0xa8/0x280
[c0000000054a73c0] [c00000000001a920] .timer_interrupt+0x90/0x100
[c0000000054a7440] [c000000000002260] decrementer_common+0x160/0x180
--- interrupt: 901 at .trace_buffer_lock_reserve+0x2c/0x90
    LR = .trace_function+0x54/0xe0
[c0000000054a77c0] [c0000000001449cc] .function_trace_call+0x7c/0x120
[c0000000054a7840] [c00000000012d750] .ftrace_ops_no_ops+0xf0/0x170
[c0000000054a78e0] [c000000000009d8c] ftrace_call+0x4/0x8
[c0000000054a7950] [c00000000072e128] .mutex_unlock+0x18/0x70
[c0000000054a79d0] [c000000000137534] .tracing_buffers_splice_read+0x424/0x4c0
[c0000000054a7c80] [c00000000021bfe8] .do_splice_to+0xa8/0xe0
[c0000000054a7d20] [c00000000021eae4] .SyS_splice+0x694/0x6b0
[c0000000054a7e30] [c000000000009224] syscall_exit+0x0/0x98


Note, the stress test is basically this:

perf record -o perf-test.dat -a -- trace-cmd record -e all -p function hackbench 2

It actually dies as it finishes the hackbench run and starts stopping
the tracing.

-- Steve