On 28/04/20 06:02, Scott Wood wrote:
> Thus, newidle_balance() is entered with interrupts enabled, which allows
> (in the next patch) enabling interrupts when the lock is dropped.
>
> Signed-off-by: Scott Wood <swood@xxxxxxxxxx>
> ---
> kernel/sched/core.c | 7 ++++---
> kernel/sched/fair.c | 45 ++++++++++++++++----------------------------
> kernel/sched/sched.h | 6 ++----
> 3 files changed, 22 insertions(+), 36 deletions(-)
>
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 9a2fbf98fd6f..0294beb8d16c 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -3241,6 +3241,10 @@ static struct rq *finish_task_switch(struct task_struct *prev)
> }
>
> tick_nohz_task_switch();
> +
> + if (is_idle_task(current))
> + newidle_balance();
> +
This means we must go through a switch_to(idle) before figuring out we
could've switched to a CFS task, and do it then. I'm curious to see the
performance impact of that.
> return rq;
> }
>
> @@ -10425,14 +10408,23 @@ static inline void nohz_newidle_balance(struct rq *this_rq) { }
> * 0 - failed, no new tasks
> * > 0 - success, new (fair) tasks present
> */
> -int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
> +int newidle_balance(void)
> {
> unsigned long next_balance = jiffies + HZ;
> - int this_cpu = this_rq->cpu;
> + int this_cpu;
> struct sched_domain *sd;
> + struct rq *this_rq;
> int pulled_task = 0;
> u64 curr_cost = 0;
>
> + preempt_disable();
> + this_rq = this_rq();
> + this_cpu = this_rq->cpu;
> + local_bh_disable();
> + raw_spin_lock_irq(&this_rq->lock);
> +
> + update_rq_clock(this_rq);
> +
> update_misfit_status(NULL, this_rq);
I'm thinking this should be moved to where newidle_balance() used to be,
otherwise we have a window where the rq is flagged as having a misfit
task despite not having any runnable CFS tasks.
> /*
> * We must set idle_stamp _before_ calling idle_balance(), such that we
