From e2cabe48c20efb174ce0c01190f8b9c5f3ea1d13 Mon Sep 17 00:00:00 2001
From: Viresh Kumar <viresh.kumar@linaro.org>
Date: Thu, 10 Aug 2017 09:50:55 +0530
Subject: [PATCH] cpufreq: schedutil: Don't restrict kthread to related_cpus
 unnecessarily

Utilization update callbacks are now processed remotely, even on the
CPUs that don't share cpufreq policy with the target CPU (if
dvfs_possible_from_any_cpu flag is set).

But in non-fast switch paths, the frequency is changed only from one of
policy->related_cpus. This happens because the kthread which does the
actual update is bound to a subset of CPUs (i.e. related_cpus).

Allow frequency to be remotely updated as well (i.e. call
__cpufreq_driver_target()) if dvfs_possible_from_any_cpu flag is set.

Reported-by: Pavan Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
---
 kernel/sched/cpufreq_schedutil.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
index 2ba04bb3182a5..69571ee6a175a 100644
--- a/kernel/sched/cpufreq_schedutil.c
+++ b/kernel/sched/cpufreq_schedutil.c
@@ -487,7 +487,11 @@ static int sugov_kthread_create(struct sugov_policy *sg_policy)
 	}
 
 	sg_policy->thread = thread;
-	kthread_bind_mask(thread, policy->related_cpus);
+
+	/* Kthread is bound to all CPUs by default */
+	if (!policy->dvfs_possible_from_any_cpu)
+		kthread_bind_mask(thread, policy->related_cpus);
+
 	init_irq_work(&sg_policy->irq_work, sugov_irq_work);
 	mutex_init(&sg_policy->work_lock);