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:   Srikar Dronamraju <srikar@linux.vnet.ibm.com>
To:     Ingo Molnar <mingo@kernel.org>,
        Peter Zijlstra <peterz@infradead.org>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Mel Gorman <mgorman@techsingularity.net>,
        Rik van Riel <riel@surriel.com>,
        Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Valentin Schneider <valentin.schneider@arm.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Gautham R Shenoy <ego@linux.vnet.ibm.com>,
        Parth Shah <parth@linux.ibm.com>
Subject: [PATCH v2 8/8] sched/fair: Dont iterate if no idle CPUs
Date:   Thu,  6 May 2021 22:15:43 +0530
Message-Id: <20210506164543.90688-9-srikar@linux.vnet.ibm.com>
In-Reply-To: <20210506164543.90688-1-srikar@linux.vnet.ibm.com>
References: <20210506164543.90688-1-srikar@linux.vnet.ibm.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

Now that the nr_busy_cpus for a LLC are updated in idle callbacks,
scheduler can detect if all threads of a LLC are busy. In such cases, it
can avoid searching for idle CPUs in the LLC that can run the wakee
thread.

Cc: LKML <linux-kernel@vger.kernel.org>
Cc: Gautham R Shenoy <ego@linux.vnet.ibm.com>
Cc: Parth Shah <parth@linux.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Cc: Dietmar Eggemann <dietmar.eggemann@arm.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
---
 kernel/sched/fair.c | 23 +++++++++++++++--------
 1 file changed, 15 insertions(+), 8 deletions(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c70f0889258f..83104d3bd0f9 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -715,7 +715,7 @@ static u64 sched_vslice(struct cfs_rq *cfs_rq, struct sched_entity *se)
 #include "pelt.h"
 #ifdef CONFIG_SMP
 
-static int select_idle_sibling(struct task_struct *p, int prev_cpu, int cpu);
+static int select_idle_sibling(struct task_struct *p, int prev_cpu, int cpu, bool idle);
 static unsigned long task_h_load(struct task_struct *p);
 static unsigned long capacity_of(int cpu);
 
@@ -5870,7 +5870,8 @@ wake_affine_weight(struct sched_domain *sd, struct task_struct *p,
 
 static inline bool test_reset_idle_core(struct sched_domain_shared *sds, int val);
 
-static int wake_affine_idler_llc(struct task_struct *p, int pref_cpu, int try_cpu, int sync)
+static int wake_affine_idler_llc(struct task_struct *p, int pref_cpu, int try_cpu,
+				int sync, bool *idle)
 {
 	int tnr_busy, tllc_size, pnr_busy, pllc_size;
 	struct sched_domain_shared *pref_sds, *try_sds;
@@ -5905,8 +5906,10 @@ static int wake_affine_idler_llc(struct task_struct *p, int pref_cpu, int try_cp
 	pllc_size = per_cpu(sd_llc_size, pref_cpu);
 	tllc_size = per_cpu(sd_llc_size, try_cpu);
 
-	if (tnr_busy == tllc_size && pnr_busy == pllc_size)
+	if (tnr_busy == tllc_size && pnr_busy == pllc_size) {
+		*idle = false;
 		return nr_cpumask_bits;
+	}
 
 	diff = tnr_busy * pllc_size - pnr_busy * tllc_size;
 	if (diff > 0)
@@ -5918,7 +5921,7 @@ static int wake_affine_idler_llc(struct task_struct *p, int pref_cpu, int try_cp
 }
 
 static int wake_affine(struct sched_domain *sd, struct task_struct *p,
-		       int this_cpu, int prev_cpu, int sync)
+		       int this_cpu, int prev_cpu, int sync, bool *idle)
 {
 	bool share_caches = cpus_share_cache(prev_cpu, this_cpu);
 	int target = nr_cpumask_bits;
@@ -5927,7 +5930,7 @@ static int wake_affine(struct sched_domain *sd, struct task_struct *p,
 		target = wake_affine_idle(this_cpu, prev_cpu);
 
 	else if (sched_feat(WA_IDLER_LLC) && !share_caches)
-		target = wake_affine_idler_llc(p, this_cpu, prev_cpu, sync);
+		target = wake_affine_idler_llc(p, this_cpu, prev_cpu, sync, idle);
 
 	if (sched_feat(WA_WEIGHT) && target == nr_cpumask_bits)
 		target = wake_affine_weight(sd, p, this_cpu, prev_cpu, sync);
@@ -6343,7 +6346,7 @@ static inline bool asym_fits_capacity(int task_util, int cpu)
 /*
  * Try and locate an idle core/thread in the LLC cache domain.
  */
-static int select_idle_sibling(struct task_struct *p, int prev, int target)
+static int select_idle_sibling(struct task_struct *p, int prev, int target, bool idle)
 {
 	struct sched_domain *sd;
 	unsigned long task_util;
@@ -6420,6 +6423,9 @@ static int select_idle_sibling(struct task_struct *p, int prev, int target)
 		}
 	}
 
+	if (!idle)
+		return target;
+
 	sd = rcu_dereference(per_cpu(sd_llc, target));
 	if (!sd)
 		return target;
@@ -6828,6 +6834,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags)
 	int want_affine = 0;
 	/* SD_flags and WF_flags share the first nibble */
 	int sd_flag = wake_flags & 0xF;
+	bool idle = true;
 
 	if (wake_flags & WF_TTWU) {
 		record_wakee(p);
@@ -6851,7 +6858,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags)
 		if (want_affine && (tmp->flags & SD_WAKE_AFFINE) &&
 		    cpumask_test_cpu(prev_cpu, sched_domain_span(tmp))) {
 			if (cpu != prev_cpu)
-				new_cpu = wake_affine(tmp, p, cpu, prev_cpu, sync);
+				new_cpu = wake_affine(tmp, p, cpu, prev_cpu, sync, &idle);
 
 			sd = NULL; /* Prefer wake_affine over balance flags */
 			break;
@@ -6868,7 +6875,7 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int wake_flags)
 		new_cpu = find_idlest_cpu(sd, p, cpu, prev_cpu, sd_flag);
 	} else if (wake_flags & WF_TTWU) { /* XXX always ? */
 		/* Fast path */
-		new_cpu = select_idle_sibling(p, prev_cpu, new_cpu);
+		new_cpu = select_idle_sibling(p, prev_cpu, new_cpu, idle);
 
 		if (want_affine)
 			current->recent_used_cpu = cpu;
-- 
2.18.2