Re: [PATCH 8/8][RFC] perf: be paranoid about child times?

[Ingo Molnar <mingo@elte.hu>]

* Peter Zijlstra <a.p.zijlstra@chello.nl> wrote:

> Normally we flatten the inherited hierarchy by attaching all childs to
> the top parent, therefore a child's child_total_time_* should never
> get incremented, add it anyway?
> 
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
> ---
>  kernel/perf_event.c |    6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> Index: linux-2.6/kernel/perf_event.c
> ===================================================================
> --- linux-2.6.orig/kernel/perf_event.c
> +++ linux-2.6/kernel/perf_event.c
> @@ -1780,8 +1780,10 @@ u64 perf_event_read_value(struct perf_ev
>  
>  	list_for_each_entry(child, &event->child_list, child_list) {
>  		total += perf_event_read(child);
> -		*enabled += child->total_time_enabled;
> -		*running += child->total_time_running;
> +		*enabled += child->total_time_enabled +
> +			atomic64_read(&child->child_total_time_enabled);
> +		*running += child->total_time_running +
> +			atomic64_read(&child->child_total_time_running);

Stick in a WARN_ON_ONCE() instead?

	Ingo