Received-SPF: pass (google.com: domain of linux-kernel+bounces-92183-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
From: Valentin Schneider <vschneid@redhat.com>
To: Ingo Molnar <mingo@kernel.org>, linux-kernel@vger.kernel.org
Cc: Shrikanth Hegde <sshegde@linux.ibm.com>, Peter Zijlstra
 <peterz@infradead.org>, Vincent Guittot <vincent.guittot@linaro.org>,
 Dietmar Eggemann <dietmar.eggemann@arm.com>, Linus Torvalds
 <torvalds@linux-foundation.org>
Subject: Re: [PATCH 6/9] sched/balancing: Update run_rebalance_domains()
 comments
In-Reply-To: <20240304094831.3639338-7-mingo@kernel.org>
References: <20240304094831.3639338-1-mingo@kernel.org>
 <20240304094831.3639338-7-mingo@kernel.org>
Date: Tue, 05 Mar 2024 11:50:51 +0100
Message-ID: <xhsmhbk7tdkac.mognet@vschneid-thinkpadt14sgen2i.remote.csb>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain

On 04/03/24 10:48, Ingo Molnar wrote:
> The first sentence of the comment explaining run_rebalance_domains()
> is historic and not true anymore:
>
>     * run_rebalance_domains is triggered when needed from the scheduler tick.
>
> ... contradicted/modified by the second sentence:
>
>     * Also triggered for NOHZ idle balancing (with NOHZ_BALANCE_KICK set).
>
> Avoid that kind of confusion straight away and explain from what
> places sched_balance_softirq() is triggered.
>
> Signed-off-by: Ingo Molnar <mingo@kernel.org>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Vincent Guittot <vincent.guittot@linaro.org>
> Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
> Cc: Linus Torvalds <torvalds@linux-foundation.org>
> Cc: Valentin Schneider <vschneid@redhat.com>
> ---
>  kernel/sched/fair.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index 4c46bffb6a7a..18b7d2999cff 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -12409,8 +12409,13 @@ static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
>  }
>
>  /*
> - * run_rebalance_domains is triggered when needed from the scheduler tick.
> - * Also triggered for NOHZ idle balancing (with NOHZ_BALANCE_KICK set).
> + * The run_rebalance_domains() softirq handler is triggered via SCHED_SOFTIRQ
> + * from two places:
> + *
> + *  - the scheduler_tick(),
> + *
> + *  - from the SMP cross-call function nohz_csd_func(),
> + *    used by NOHZ idle balancing (with NOHZ_BALANCE_KICK set).

Bit of a nit but the CSD is also triggered via the scheduler_tick():

  scheduler_tick()
  `\
    trigger_load_balance()
    `\
      raise_softirq(SCHED_SOFTIRQ)

  scheduler_tick()
  `\
    trigger_load_balance()
    `\
      nohz_balance_kick()
      `\
        kick_ilb()
        `\
          smp_call_function_single_async(ilb_cpu, &cpu_rq(ilb_cpu)->nohz_csd);

I got to the below which is still somewhat confusing, thoughts?

"""
The run_rebalance_domains() softirq handler is triggered via SCHED_SOFTIRQ
from two places:

- directly from trigger_load_balance() in scheduler_tick(), for periodic
  load balance

- indirectly from kick_ilb() (invoked down the scheduler_tick() too), which
  issues an SMP cross-call to nohz_csd_func() which will itself raise the
  softirq, for NOHZ idle balancing.
"""

>   */
>  static __latent_entropy void run_rebalance_domains(struct softirq_action *h)
>  {
> --
> 2.40.1