Re: [PATCH v2 2/4] perf ftrace: Add ftrace.tracer config option

[Arnaldo Carvalho de Melo <acme@kernel.org>]

Em Tue, Jan 31, 2017 at 08:38:29PM +0900, Taeung Song escreveu:
> Currently perf ftrace command will select 'function_graph' or 'function'.
> So add ftrace.tracer config option to select tracer

The above is confusing, I changed it to:

    Currently 'perf ftrace' command allows selecting 'function_graph' or
    'function', defaulting to 'function_graph'.
    
    Add the ftrace.tracer config option to select the default tracer:
 
>     # cat ~/.perfconfig
>     [ftrace]
>         tracer = function
> 
>     # perf ftrace usleep 123456 | head -10
>       <...>-14450 [002] d... 10089.284231: finish_task_switch <-__schedule
>       <...>-14450 [002] .... 10089.284232: finish_wait <-pipe_wait
>       <...>-14450 [002] .... 10089.284232: mutex_lock <-pipe_wait
>       <...>-14450 [002] .... 10089.284232: _cond_resched <-mutex_lock
> 
> Cc: Jiri Olsa <jolsa@kernel.org>
> Cc: Namhyung Kim <namhyung@kernel.org>
> Signed-off-by: Taeung Song <treeze.taeung@gmail.com>
> ---
>  tools/perf/builtin-ftrace.c | 25 +++++++++++++++++++++++++
>  1 file changed, 25 insertions(+)
> 
> diff --git a/tools/perf/builtin-ftrace.c b/tools/perf/builtin-ftrace.c
> index 414444d..00e228f 100644
> --- a/tools/perf/builtin-ftrace.c
> +++ b/tools/perf/builtin-ftrace.c
> @@ -17,6 +17,7 @@
>  #include "evlist.h"
>  #include "target.h"
>  #include "thread_map.h"
> +#include "util/config.h"
>  
>  
>  #define DEFAULT_TRACER  "function_graph"
> @@ -198,6 +199,26 @@ static int __cmd_ftrace(struct perf_ftrace *ftrace, int argc, const char **argv)
>  	return done ? 0 : -1;
>  }
>  
> +static int perf_ftrace_config(const char *var, const char *value, void *cb)
> +{
> +	struct perf_ftrace *ftrace = cb;
> +
> +	if (!strcmp(var, "ftrace.tracer")) {
> +		if (!strcmp(value, "function_graph"))
> +			ftrace->tracer = DEFAULT_TRACER;
> +		else if (!strcmp(value, "function"))
> +			ftrace->tracer = "function";
> +		else {
> +			pr_err("Please select function_graph(default)"
> +			       "or function to use tracer.\n");
> +			return -1;
> +		}

If you use {} in the else case, use it in the other branch, also I
simplified it to be:

+	if (!strcmp(var, "ftrace.tracer")) {
+		if (!strcmp(value, "function_graph") ||
+		    !strcmp(value, "function"))
+			ftrace->tracer = value;
+		else {
+			pr_err("Please select function_graph(default)"
+			       "or function to use tracer.\n");
+			return -1;
+		}

As we know item->value will not go away (its a perf_config_item,
allocated and kept in that perf_config_set structure).

Also it doesn't make sense comparing against "function_graph" to set it
to DEFAULT_TRACER, as if we decide to change the DEFAULT_TRACER this
simply breaks.

Also you forgot to test by setting an invalid tracer, I did and noticed
that you forgot to add a space between "(default)" and "or ", please be
more careful with testing.

Also you are silently ignoring any unknown variable in this section, so
if someone has this:

cat ~/.perfconfig

[ftrace]

	trace = function


I.e. forgets the 'r', it will keep using the current default,
"function_graph" till the user checks the config and adds that missing
'r'...

I haven't changed this because I think this is so common that we should
instead have a different error return for those config callbacks that
will tell perf_config() to say something like:

  "unknown variable %s.%s\n", section_name, variable_name

- Arnaldo


> +		return 0;
> +	}
> +
> +	return 0;
> +}
> +
>  int cmd_ftrace(int argc, const char **argv, const char *prefix __maybe_unused)
>  {
>  	int ret;
> @@ -218,6 +239,10 @@ int cmd_ftrace(int argc, const char **argv, const char *prefix __maybe_unused)
>  	OPT_END()
>  	};
>  
> +	ret = perf_config(perf_ftrace_config, &ftrace);
> +	if (ret < 0)
> +		return -1;
> +
>  	argc = parse_options(argc, argv, ftrace_options, ftrace_usage,
>  			    PARSE_OPT_STOP_AT_NON_OPTION);
>  	if (!argc)
> -- 
> 2.7.4