sched/fair: Fix sched_group_energy() to support per-cpu capacity states

sched_group_energy() was supposed to support per-cpu capacity states
(DVFS), however, while fixing a hotplug issue this was broken as we bail
out if there is no SD_SHARE_CAP_STATES flag set.

This patch implements the hotplug race check differently and should
therefore reinstate support for per-cpu capacity states.

Change-Id: I5b865666c9ce833dcfa6514c574580d75aa0a195
Signed-off-by: Morten Rasmussen <morten.rasmussen@arm.com>

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index fc4e2529fbd29cba598721b64696b9673926e6bd..cbd6d12e91fe4cb5155f87163340828b9c241fc5 100644 (file)
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -5442,15 +5442,7 @@ static int sched_group_energy(struct energy_env *eenv)
                 */
                sd = rcu_dereference(per_cpu(sd_scs, cpu));
 
-               if (!sd)
-                       /*
-                        * We most probably raced with hotplug; returning a
-                        * wrong energy estimation is better than entering an
-                        * infinite loop.
-                        */
-                       return -EINVAL;
-
-               if (sd->parent)
+               if (sd && sd->parent)
                        sg_shared_cap = sd->parent->groups;
 
                for_each_domain(cpu, sd) {
@@ -5505,6 +5497,14 @@ static int sched_group_energy(struct energy_env *eenv)
 
                        } while (sg = sg->next, sg != sd->groups);
                }
+
+               /*
+                * If we raced with hotplug and got an sd NULL-pointer;
+                * returning a wrong energy estimation is better than
+                * entering an infinite loop.
+                */
+               if (cpumask_test_cpu(cpu, &visit_cpus))
+                       return -EINVAL;
 next_cpu:
                cpumask_clear_cpu(cpu, &visit_cpus);
                continue;