Re: dynamic ftrace - graph

[Steven Rostedt <rostedt@goodmis.org>]

Hi Michal,

Sorry for such a late reply, you sent this email to me at the worst
time. My wife was having shoulder surgery and I heading for a
conference. I fell very far behind in email and I only now just noticed
this.


I'll answer these ever though you may already figured it out.

On Mon, 2009-11-09 at 17:52 +0100, Michal Simek wrote:
> Hi Steve and others,
> 
> I have working dynamic ftrace function. :-)
> 
> I look at __ftrace_modify_code function and I have one question about.
> Below are function which are called when I enable function_graph.
> 
> I personally think that make more sense to call 
> ftrace_enable_ftrace_graph_caller before ftrace_update_ftrace_func.
> The reason for it is that graph_caller enable calling graph tracing 
> which should be setup before ftrace_update_ftrace_func which enable 
> whole tracing function.

Have you seen any difference in output with this?

I never thought about this order too much.

-- Steve

> 
> 
> 
> # cd ; mkdir /debug; mount -t debugfs none /debug; cat 
> /debug/tracing/available_tracers; echo function_graph > 
> /debug/tracing/current_tracer;echo 0 > /debug/tra
> cing/tracing_enabled;cat /debug/tracing/trace | head -n 10
> function_graph function sched_switch nop
> 
> ftrace_update_ftrace_func 0xc0009100 0xc00091c4, 0xb000c000, 0x32809100
> ftrace_enable_ftrace_graph_caller
> # tracer: function_graph
> 
> 
> For disabling graph trace is sequence ok I think.
> 
> # echo function > /debug/tracing/current_tracer
> ftrace_disable_ftrace_graph_caller
> ftrace_update_ftrace_func 0xc006afe4 0xc00091c4, 0xb000c006, 0x3280afe4
> 
> 
> What do you think?
> 
> Thanks,
> Michal
> 
>