Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Barry Song <21cnbao@gmail.com>
To:     aubrey.li@linux.intel.com
Cc:     linux-kernel@vger.kernel.org, mgorman@techsingularity.net,
        mingo@kernel.org, peterz@infradead.org, song.bao.hua@hisilicon.com,
        valentin.schneider@arm.com, vincent.guittot@linaro.org,
        yangyicong@huawei.com
Subject: Re: [PATCH 8/9] sched/fair: select idle cpu from idle cpumask for task wakeup
Date:   Fri, 17 Sep 2021 11:44:00 +0800
Message-Id: <20210917034400.7620-1-21cnbao@gmail.com>
In-Reply-To: <125eb98a-241b-078f-1844-b0521425ed1e@linux.intel.com>
References: <125eb98a-241b-078f-1844-b0521425ed1e@linux.intel.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

[Aubrey wrote]
> The idle core information is already in idle cpu mask, do we have a quick
> and cheap way to extract out?

seems hard.

since we need an intermediate cpumask memory to save the result of cpumask
bitmap operations, the only way i can get the idle core and make sure "cpus"
isn't broken is:

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 361927803e31..abd844bcfb86 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -6187,8 +6187,14 @@ void __update_idle_core(struct rq *rq)
  */
 static int select_idle_core(struct task_struct *p, int core, struct cpumask *cpus, int *idle_cpu)
 {
-       bool idle = true;
-       int cpu;
+       bool idle = true, possible_idle;
+       int cpu, wb, wa;
+
+       wb = cpumask_weight(cpus);
+       cpumask_andnot(cpus, cpus, cpu_smt_mask(core));
+       wa = cpumask_weight(cpus);
+
+       possible_idle = (wa - wb == cpumask_weight(cpu_smt_mask(core)));
 
        if (!static_branch_likely(&sched_smt_present))
                return __select_idle_cpu(core, p);
@@ -6212,7 +6218,6 @@ static int select_idle_core(struct task_struct *p, int core, struct cpumask *cpu
        if (idle)
                return core;
 
-       cpumask_andnot(cpus, cpus, cpu_smt_mask(core));
        return -1;
 }

but the questions are:
1. is it really cheap?
2. how to use possible_idle? right now, select_idle_core() is also selecting idle cpu. we shouldn't simply return when possible_idle is false.

Thanks
Barry