Re: [PATCH] tracing/fgraph: remove unnecessary unlikely()

[Steven Rostedt <rostedt@goodmis.org>]

On Sat,  3 Nov 2018 22:35:37 -0400
Yangtao Li <tiny.windzz@gmail.com> wrote:

> WARN_ON() already contains an unlikely(), so it's not necessary to use
> unlikely.

NACK... see below.

> 
> Signed-off-by: Yangtao Li <tiny.windzz@gmail.com>
> ---
>  kernel/trace/trace_functions_graph.c | 6 ++----
>  1 file changed, 2 insertions(+), 4 deletions(-)
> 
> diff --git a/kernel/trace/trace_functions_graph.c b/kernel/trace/trace_functions_graph.c
> index 169b3c44ee97..f8c2a08e4985 100644
> --- a/kernel/trace/trace_functions_graph.c
> +++ b/kernel/trace/trace_functions_graph.c
> @@ -201,9 +201,8 @@ ftrace_pop_return_trace(struct ftrace_graph_ret *trace, unsigned long *ret,
>  	if (index < 0)
>  		index += FTRACE_NOTRACE_DEPTH;
>  
> -	if (unlikely(index < 0 || index >= FTRACE_RETFUNC_DEPTH)) {
> +	if (WARN_ON(index < 0 || index >= FTRACE_RETFUNC_DEPTH)) {

When function graph is enabled, the printk from the warn on will likely
be traced as well and trigger the same WARN_ON, which will again cause
a recursion loop and crash and reboot the box with no output at all.

>  		ftrace_graph_stop();

Notice that I call ftrace_graph_stop() *before* the WARN_ON(). This
disables the ftrace graph tracer and prevents the recursion loop from
happening.

There's a reason that the WARN_ON() is placed where it is.

But thanks for the report, it shows that I need to add a comment here
so that someone else doesn't send a similar patch in the future.

-- Steve


> -		WARN_ON(1);
>  		/* Might as well panic, otherwise we have no where to go */
>  		*ret = (unsigned long)panic;
>  		return;
> @@ -274,9 +273,8 @@ unsigned long ftrace_return_to_handler(unsigned long frame_pointer)
>  	 */
>  	ftrace_graph_return(&trace);
>  
> -	if (unlikely(!ret)) {
> +	if (WARN_ON(!ret)) {
>  		ftrace_graph_stop();
> -		WARN_ON(1);
>  		/* Might as well panic. What else to do? */
>  		ret = (unsigned long)panic;
>  	}