sched: Break out cpu_power from the sched_group structure

In order to prepare for non-unique sched_groups per domain, we need to
carry the cpu_power elsewhere, so put a level of indirection in.

Reported-and-tested-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/n/tip-qkho2byuhe4482fuknss40ad@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>

2 files changed, 49 insertions, 29 deletions

diff --git a/kernel/sched.c b/kernel/sched.c
index 3dc716f6d8ad..36c10d25d4cd 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -6557,7 +6557,7 @@ static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
                         break;
                 }
 
-                if (!group->cpu_power) {
+                if (!group->sgp->power) {
                         printk(KERN_CONT "\n");
                         printk(KERN_ERR "ERROR: domain->cpu_power not "
                                         "set\n");
@@ -6581,9 +6581,9 @@ static int sched_domain_debug_one(struct sched_domain *sd, int cpu, int level,
                 cpulist_scnprintf(str, sizeof(str), sched_group_cpus(group));
 
                 printk(KERN_CONT " %s", str);
-                if (group->cpu_power != SCHED_POWER_SCALE) {
+                if (group->sgp->power != SCHED_POWER_SCALE) {
                         printk(KERN_CONT " (cpu_power = %d)",
-                                group->cpu_power);
+                                group->sgp->power);
                 }
 
                 group = group->next;
@@ -6777,8 +6777,10 @@ static struct root_domain *alloc_rootdomain(void)
 static void free_sched_domain(struct rcu_head *rcu)
 {
         struct sched_domain *sd = container_of(rcu, struct sched_domain, rcu);
-        if (atomic_dec_and_test(&sd->groups->ref))
+        if (atomic_dec_and_test(&sd->groups->ref)) {
+                kfree(sd->groups->sgp);
                 kfree(sd->groups);
+        }
         kfree(sd);
 }
 
@@ -6945,6 +6947,7 @@ int sched_smt_power_savings = 0, sched_mc_power_savings = 0;
 struct sd_data {
         struct sched_domain **__percpu sd;
         struct sched_group **__percpu sg;
+        struct sched_group_power **__percpu sgp;
 };
 
 struct s_data {
@@ -6981,8 +6984,10 @@ static int get_group(int cpu, struct sd_data *sdd, struct sched_group **sg)
         if (child)
                 cpu = cpumask_first(sched_domain_span(child));
 
-        if (sg)
+        if (sg) {
                 *sg = *per_cpu_ptr(sdd->sg, cpu);
+                (*sg)->sgp = *per_cpu_ptr(sdd->sgp, cpu);
+        }
 
         return cpu;
 }
@@ -7020,7 +7025,7 @@ build_sched_groups(struct sched_domain *sd)
                         continue;
 
                 cpumask_clear(sched_group_cpus(sg));
-                sg->cpu_power = 0;
+                sg->sgp->power = 0;
 
                 for_each_cpu(j, span) {
                         if (get_group(j, sdd, NULL) != group)
@@ -7185,6 +7190,7 @@ static void claim_allocations(int cpu, struct sched_domain *sd)
         if (cpu == cpumask_first(sched_group_cpus(sg))) {
                 WARN_ON_ONCE(*per_cpu_ptr(sdd->sg, cpu) != sg);
                 *per_cpu_ptr(sdd->sg, cpu) = NULL;
+                *per_cpu_ptr(sdd->sgp, cpu) = NULL;
         }
 }
 
@@ -7234,9 +7240,14 @@ static int __sdt_alloc(const struct cpumask *cpu_map)
                 if (!sdd->sg)
                         return -ENOMEM;
 
+                sdd->sgp = alloc_percpu(struct sched_group_power *);
+                if (!sdd->sgp)
+                        return -ENOMEM;
+
                 for_each_cpu(j, cpu_map) {
                         struct sched_domain *sd;
                         struct sched_group *sg;
+                        struct sched_group_power *sgp;
 
                         sd = kzalloc_node(sizeof(struct sched_domain) + cpumask_size(),
                                         GFP_KERNEL, cpu_to_node(j));
@@ -7251,6 +7262,13 @@ static int __sdt_alloc(const struct cpumask *cpu_map)
                                 return -ENOMEM;
 
                         *per_cpu_ptr(sdd->sg, j) = sg;
+
+                        sgp = kzalloc_node(sizeof(struct sched_group_power),
+                                        GFP_KERNEL, cpu_to_node(j));
+                        if (!sgp)
+                                return -ENOMEM;
+
+                        *per_cpu_ptr(sdd->sgp, j) = sgp;
                 }
         }
 
@@ -7268,9 +7286,11 @@ static void __sdt_free(const struct cpumask *cpu_map)
                 for_each_cpu(j, cpu_map) {
                         kfree(*per_cpu_ptr(sdd->sd, j));
                         kfree(*per_cpu_ptr(sdd->sg, j));
+                        kfree(*per_cpu_ptr(sdd->sgp, j));
                 }
                 free_percpu(sdd->sd);
                 free_percpu(sdd->sg);
+                free_percpu(sdd->sgp);
         }
 }
 
diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 433491c2dc8f..c768588e180b 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -1585,7 +1585,7 @@ find_idlest_group(struct sched_domain *sd, struct task_struct *p,
                 }
 
                 /* Adjust by relative CPU power of the group */
-                avg_load = (avg_load * SCHED_POWER_SCALE) / group->cpu_power;
+                avg_load = (avg_load * SCHED_POWER_SCALE) / group->sgp->power;
 
                 if (local_group) {
                         this_load = avg_load;
@@ -2631,7 +2631,7 @@ static void update_cpu_power(struct sched_domain *sd, int cpu)
                 power >>= SCHED_POWER_SHIFT;
         }
 
-        sdg->cpu_power_orig = power;
+        sdg->sgp->power_orig = power;
 
         if (sched_feat(ARCH_POWER))
                 power *= arch_scale_freq_power(sd, cpu);
@@ -2647,7 +2647,7 @@ static void update_cpu_power(struct sched_domain *sd, int cpu)
                 power = 1;
 
         cpu_rq(cpu)->cpu_power = power;
-        sdg->cpu_power = power;
+        sdg->sgp->power = power;
 }
 
 static void update_group_power(struct sched_domain *sd, int cpu)
@@ -2665,11 +2665,11 @@ static void update_group_power(struct sched_domain *sd, int cpu)
 
         group = child->groups;
         do {
-                power += group->cpu_power;
+                power += group->sgp->power;
                 group = group->next;
         } while (group != child->groups);
 
-        sdg->cpu_power = power;
+        sdg->sgp->power = power;
 }
 
 /*
@@ -2691,7 +2691,7 @@ fix_small_capacity(struct sched_domain *sd, struct sched_group *group)
         /*
          * If ~90% of the cpu_power is still there, we're good.
          */
-        if (group->cpu_power * 32 > group->cpu_power_orig * 29)
+        if (group->sgp->power * 32 > group->sgp->power_orig * 29)
                 return 1;
 
         return 0;
@@ -2771,7 +2771,7 @@ static inline void update_sg_lb_stats(struct sched_domain *sd,
         }
 
         /* Adjust by relative CPU power of the group */
-        sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->cpu_power;
+        sgs->avg_load = (sgs->group_load*SCHED_POWER_SCALE) / group->sgp->power;
 
         /*
          * Consider the group unbalanced when the imbalance is larger
@@ -2788,7 +2788,7 @@ static inline void update_sg_lb_stats(struct sched_domain *sd,
         if ((max_cpu_load - min_cpu_load) >= avg_load_per_task && max_nr_running > 1)
                 sgs->group_imb = 1;
 
-        sgs->group_capacity = DIV_ROUND_CLOSEST(group->cpu_power,
+        sgs->group_capacity = DIV_ROUND_CLOSEST(group->sgp->power,
                                                 SCHED_POWER_SCALE);
         if (!sgs->group_capacity)
                 sgs->group_capacity = fix_small_capacity(sd, group);
@@ -2877,7 +2877,7 @@ static inline void update_sd_lb_stats(struct sched_domain *sd, int this_cpu,
                         return;
 
                 sds->total_load += sgs.group_load;
-                sds->total_pwr += sg->cpu_power;
+                sds->total_pwr += sg->sgp->power;
 
                 /*
                  * In case the child domain prefers tasks go to siblings
@@ -2962,7 +2962,7 @@ static int check_asym_packing(struct sched_domain *sd,
         if (this_cpu > busiest_cpu)
                 return 0;
 
-        *imbalance = DIV_ROUND_CLOSEST(sds->max_load * sds->busiest->cpu_power,
+        *imbalance = DIV_ROUND_CLOSEST(sds->max_load * sds->busiest->sgp->power,
                                        SCHED_POWER_SCALE);
         return 1;
 }
@@ -2993,7 +2993,7 @@ static inline void fix_small_imbalance(struct sd_lb_stats *sds,
 
         scaled_busy_load_per_task = sds->busiest_load_per_task
                                          * SCHED_POWER_SCALE;
-        scaled_busy_load_per_task /= sds->busiest->cpu_power;
+        scaled_busy_load_per_task /= sds->busiest->sgp->power;
 
         if (sds->max_load - sds->this_load + scaled_busy_load_per_task >=
                         (scaled_busy_load_per_task * imbn)) {
@@ -3007,28 +3007,28 @@ static inline void fix_small_imbalance(struct sd_lb_stats *sds,
          * moving them.
          */
 
-        pwr_now += sds->busiest->cpu_power *
+        pwr_now += sds->busiest->sgp->power *
                         min(sds->busiest_load_per_task, sds->max_load);
-        pwr_now += sds->this->cpu_power *
+        pwr_now += sds->this->sgp->power *
                         min(sds->this_load_per_task, sds->this_load);
         pwr_now /= SCHED_POWER_SCALE;
 
         /* Amount of load we'd subtract */
         tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
-                sds->busiest->cpu_power;
+                sds->busiest->sgp->power;
         if (sds->max_load > tmp)
-                pwr_move += sds->busiest->cpu_power *
+                pwr_move += sds->busiest->sgp->power *
                         min(sds->busiest_load_per_task, sds->max_load - tmp);
 
         /* Amount of load we'd add */
-        if (sds->max_load * sds->busiest->cpu_power <
+        if (sds->max_load * sds->busiest->sgp->power <
                 sds->busiest_load_per_task * SCHED_POWER_SCALE)
-                tmp = (sds->max_load * sds->busiest->cpu_power) /
-                        sds->this->cpu_power;
+                tmp = (sds->max_load * sds->busiest->sgp->power) /
+                        sds->this->sgp->power;
         else
                 tmp = (sds->busiest_load_per_task * SCHED_POWER_SCALE) /
-                        sds->this->cpu_power;
-        pwr_move += sds->this->cpu_power *
+                        sds->this->sgp->power;
+        pwr_move += sds->this->sgp->power *
                         min(sds->this_load_per_task, sds->this_load + tmp);
         pwr_move /= SCHED_POWER_SCALE;
 
@@ -3074,7 +3074,7 @@ static inline void calculate_imbalance(struct sd_lb_stats *sds, int this_cpu,
 
                 load_above_capacity *= (SCHED_LOAD_SCALE * SCHED_POWER_SCALE);
 
-                load_above_capacity /= sds->busiest->cpu_power;
+                load_above_capacity /= sds->busiest->sgp->power;
         }
 
         /*
@@ -3090,8 +3090,8 @@ static inline void calculate_imbalance(struct sd_lb_stats *sds, int this_cpu,
         max_pull = min(sds->max_load - sds->avg_load, load_above_capacity);
 
         /* How much load to actually move to equalise the imbalance */
-        *imbalance = min(max_pull * sds->busiest->cpu_power,
-                (sds->avg_load - sds->this_load) * sds->this->cpu_power)
+        *imbalance = min(max_pull * sds->busiest->sgp->power,
+                (sds->avg_load - sds->this_load) * sds->this->sgp->power)
                         / SCHED_POWER_SCALE;
 
         /*