From: Chen Liang <cl@rock-chips.com>
Date: Fri, 8 Jul 2016 02:12:43 +0000 (+0800)
Subject: ARM64: sched: fix bug: avoid infinite loop
X-Git-Tag: firefly_0821_release~2181
X-Git-Url: http://plrg.eecs.uci.edu/git/?p=firefly-linux-kernel-4.4.55.git;a=commitdiff_plain;h=d4773e1407b6d59a202bd4ce8acc83deaceafa29

ARM64: sched: fix bug: avoid infinite loop

All of the sched domains will be destroied and then rebuilded when a cpu
online/offline, if a softirq comes after the shced domains are destroied,
this cpu will be stucked in the infinite loop in sched_group_energy(),
because of it can not get the shced donmain in for_each_domain(cpu, sd).

Change-Id: I154cf560e4e1af4a7a2547154ad321e936196ce3
Signed-off-by: Chen Liang <cl@rock-chips.com>
---

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 6d84df9366ed..67629fa15d58 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -4866,6 +4866,7 @@ static int sched_group_energy(struct energy_env *eenv)
 		struct sched_group *sg_shared_cap = NULL;
 
 		cpu = cpumask_first(&visit_cpus);
+		cpumask_clear_cpu(cpu, &visit_cpus);
 
 		/*
 		 * Is the group utilization affected by cpus outside this
@@ -4928,8 +4929,12 @@ static int sched_group_energy(struct energy_env *eenv)
 
 				total_energy += sg_busy_energy + sg_idle_energy;
 
-				if (!sd->child)
-					cpumask_xor(&visit_cpus, &visit_cpus, sched_group_cpus(sg));
+				if (!sd->child) {
+					int i;
+
+					for_each_cpu(i, sched_group_cpus(sg))
+						cpumask_clear_cpu(i, &visit_cpus);
+				}
 
 				if (cpumask_equal(sched_group_cpus(sg), sched_group_cpus(eenv->sg_top)))
 					goto next_cpu;