Re: [PATCH v3 05/10] sched/fair: Hoist idle_stamp up from idle_balance

[Steven Sistare <steven.sistare@oracle.com>]

On 11/9/2018 2:07 PM, Valentin Schneider wrote:
> Hi Steve,
> 
> On 09/11/2018 12:50, Steve Sistare wrote:
>> Move the update of idle_stamp from idle_balance to the call site in
>> pick_next_task_fair, to prepare for a future patch that adds work to
>> pick_next_task_fair which must be included in the idle_stamp interval.
>> No functional change.
>>
>> Signed-off-by: Steve Sistare <steven.sistare@oracle.com>
>> ---
>>  kernel/sched/fair.c | 23 ++++++++++++++---------
>>  1 file changed, 14 insertions(+), 9 deletions(-)
>>
>> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
>> index 9031d39..da368ed 100644
>> --- a/kernel/sched/fair.c
>> +++ b/kernel/sched/fair.c
>> @@ -3725,6 +3725,8 @@ static inline void update_misfit_status(struct task_struct *p, struct rq *rq)
>>  	rq->misfit_task_load = task_h_load(p);
>>  }
>>  
>> +#define IF_SMP(statement)	statement
>> +
> 
> I'm not too hot on those IF_SMP() macros. Since you're not introducing
> any other user for them, what about an inline function for rq->idle_stamp
> setting ? When it's mapped to an empty statement (!CONFIG_SMP) GCC is
> smart enough to remove the rq_clock() that would be passed to it on
> CONFIG_SMP:

That may be true now, but I worry that rq_clock or its subroutines may gain
side effects in the future that prevent the compiler from removing it.  However,
I could push rq_clock into the inline function:

  static inline void rq_idle_stamp_set(rq)    { rq->idle_stamp = rq_clock(rq); }
  static inline void rq_idle_stamp_clear(rq)  { rq->idle_stamp = 0; }

I like that better, do you?

- Steve
 
> ----->8-----
> 
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index c11adf3..34d9864 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -3725,7 +3725,10 @@ static inline void update_misfit_status(struct task_struct *p, struct rq *rq)
>         rq->misfit_task_load = task_h_load(p);
>  }
>  
> -#define IF_SMP(statement)      statement
> +static inline void set_rq_idle_stamp(struct rq *rq, u64 value)
> +{
> +       rq->idle_stamp = value;
> +}
>  
>  static void overload_clear(struct rq *rq)
>  {
> @@ -3772,7 +3775,7 @@ static inline int idle_balance(struct rq *rq, struct rq_flags *rf)
>         return 0;
>  }
>  
> -#define IF_SMP(statement)      /* empty */
> +static inline void set_rq_idle_stamp(struct rq *rq, u64 value) {}
>  
>  static inline void overload_clear(struct rq *rq) {}
>  static inline void overload_set(struct rq *rq) {}
> @@ -6773,12 +6776,12 @@ done: __maybe_unused;
>          * We must set idle_stamp _before_ calling idle_balance(), such that we
>          * measure the duration of idle_balance() as idle time.
>          */
> -       IF_SMP(rq->idle_stamp = rq_clock(rq);)
> +       set_rq_idle_stamp(rq, rq_clock(rq));
>  
>         new_tasks = idle_balance(rq, rf);
>  
>         if (new_tasks)
> -               IF_SMP(rq->idle_stamp = 0;)
> +               set_rq_idle_stamp(rq, 0);
>  
>         /*
>          * Because idle_balance() releases (and re-acquires) rq->lock, it is
>