Re: [PATCH] ftrace: avoid potential division by zero

[Steven Rostedt]

On Wed,  1 Jan 2020 17:32:19 +0800
Wen Yang <wenyang@xxxxxxxxxxxxxxxxx> wrote:

> The ftrace_profile->counter is unsigned long and
> do_div truncates it to 32 bits, which means it can test
> non-zero and be truncated to zero for division.
> Fix this issue by using div64_ul() instead.

Thanks, but since we are using div64_ul() which has different semantics
than do_div() let's clean up the code that was written to deal with the
strange do_div() semantics.

> 
> Signed-off-by: Wen Yang <wenyang@xxxxxxxxxxxxxxxxx>
> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> ---
>  kernel/trace/ftrace.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
> index ac99a35..a490ba5 100644
> --- a/kernel/trace/ftrace.c
> +++ b/kernel/trace/ftrace.c
> @@ -527,7 +527,7 @@ static int function_stat_show(struct seq_file *m, void *v)
>  
>  #ifdef CONFIG_FUNCTION_GRAPH_TRACER
>  	avg = rec->time;
> -	do_div(avg, rec->counter);
> +	avg = div64_ul(avg, rec->counter);

The above should be:

	avg = div64_ul(rec->time, rec->counter);

and get rid of the pre-assigning of avg.


>  	if (tracing_thresh && (avg < tracing_thresh))
>  		goto out;
>  #endif
> @@ -553,7 +553,8 @@ static int function_stat_show(struct seq_file *m, void *v)
>  		 * Divide only 1000 for ns^2 -> us^2 conversion.
>  		 * trace_print_graph_duration will divide 1000 again.
>  		 */
> -		do_div(stddev, rec->counter * (rec->counter - 1) * 1000);
> +		stddev = div64_ul(stddev, 
> +				  rec->counter * (rec->counter - 1) * 1000);

This can stay as is, because of the complex dividend in the equation.

Thanks,

-- Steve


>  	}
>  
>  	trace_seq_init(&s);