sched: Rename load-balancing fields
authorPeter Zijlstra <a.p.zijlstra@chello.nl>
Wed, 22 Feb 2012 18:27:40 +0000 (19:27 +0100)
committerIngo Molnar <mingo@elte.hu>
Thu, 1 Mar 2012 09:51:23 +0000 (10:51 +0100)
 s/env->this_/env->dst_/g
 s/env->busiest_/env->src_/g
 s/pull_task/move_task/g

Makes everything clearer.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: pjt@google.com
Link: http://lkml.kernel.org/n/tip-0yvgms8t8x962drpvl0fu0kk@git.kernel.org
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/sched/fair.c

index 55b1f11..233d051 100644 (file)
@@ -2918,7 +2918,7 @@ static void check_preempt_wakeup(struct rq *rq, struct task_struct *p, int wake_
                return;
 
        /*
-        * This is possible from callers such as pull_task(), in which we
+        * This is possible from callers such as move_task(), in which we
         * unconditionally check_prempt_curr() after an enqueue (which may have
         * lead to a throttle).  This both saves work and prevents false
         * next-buddy nomination below.
@@ -3084,17 +3084,37 @@ static bool yield_to_task_fair(struct rq *rq, struct task_struct *p, bool preemp
 
 static unsigned long __read_mostly max_load_balance_interval = HZ/10;
 
+#define LBF_ALL_PINNED 0x01
+#define LBF_NEED_BREAK 0x02    /* clears into HAD_BREAK */
+#define LBF_HAD_BREAK  0x04
+#define LBF_HAD_BREAKS 0x0C    /* count HAD_BREAKs overflows into ABORT */
+#define LBF_ABORT      0x10
+
+struct lb_env {
+       struct sched_domain     *sd;
+
+       int                     src_cpu;
+       struct rq               *src_rq;
+       struct cfs_rq           *src_cfs_rq;
+
+       int                     dst_cpu;
+       struct rq               *dst_rq;
+
+       enum cpu_idle_type      idle;
+       unsigned long           max_load_move;
+       unsigned int            flags;
+};
+
 /*
- * pull_task - move a task from a remote runqueue to the local runqueue.
+ * move_task - move a task from one runqueue to another runqueue.
  * Both runqueues must be locked.
  */
-static void pull_task(struct rq *src_rq, struct task_struct *p,
-                     struct rq *this_rq, int this_cpu)
+static void move_task(struct task_struct *p, struct lb_env *env)
 {
-       deactivate_task(src_rq, p, 0);
-       set_task_cpu(p, this_cpu);
-       activate_task(this_rq, p, 0);
-       check_preempt_curr(this_rq, p, 0);
+       deactivate_task(env->src_rq, p, 0);
+       set_task_cpu(p, env->dst_cpu);
+       activate_task(env->dst_rq, p, 0);
+       check_preempt_curr(env->dst_rq, p, 0);
 }
 
 /*
@@ -3129,26 +3149,6 @@ task_hot(struct task_struct *p, u64 now, struct sched_domain *sd)
        return delta < (s64)sysctl_sched_migration_cost;
 }
 
-#define LBF_ALL_PINNED 0x01
-#define LBF_NEED_BREAK 0x02    /* clears into HAD_BREAK */
-#define LBF_HAD_BREAK  0x04
-#define LBF_HAD_BREAKS 0x0C    /* count HAD_BREAKs overflows into ABORT */
-#define LBF_ABORT      0x10
-
-struct lb_env {
-       struct sched_domain     *sd;
-
-       int                     this_cpu;
-       struct rq               *this_rq;
-
-       struct rq               *busiest_rq;
-       struct cfs_rq           *busiest_cfs_rq;
-
-       enum cpu_idle_type      idle;
-       unsigned long           max_load_move;
-       unsigned int            flags;
-};
-
 /*
  * can_migrate_task - may task p from runqueue rq be migrated to this_cpu?
  */
@@ -3162,13 +3162,13 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
         * 2) cannot be migrated to this CPU due to cpus_allowed, or
         * 3) are cache-hot on their current CPU.
         */
-       if (!cpumask_test_cpu(env->this_cpu, tsk_cpus_allowed(p))) {
+       if (!cpumask_test_cpu(env->dst_cpu, tsk_cpus_allowed(p))) {
                schedstat_inc(p, se.statistics.nr_failed_migrations_affine);
                return 0;
        }
        env->flags &= ~LBF_ALL_PINNED;
 
-       if (task_running(env->busiest_rq, p)) {
+       if (task_running(env->src_rq, p)) {
                schedstat_inc(p, se.statistics.nr_failed_migrations_running);
                return 0;
        }
@@ -3179,7 +3179,7 @@ int can_migrate_task(struct task_struct *p, struct lb_env *env)
         * 2) too many balance attempts have failed.
         */
 
-       tsk_cache_hot = task_hot(p, env->busiest_rq->clock_task, env->sd);
+       tsk_cache_hot = task_hot(p, env->src_rq->clock_task, env->sd);
        if (!tsk_cache_hot ||
                env->sd->nr_balance_failed > env->sd->cache_nice_tries) {
 #ifdef CONFIG_SCHEDSTATS
@@ -3210,20 +3210,20 @@ static int move_one_task(struct lb_env *env)
        struct task_struct *p, *n;
        struct cfs_rq *cfs_rq;
 
-       for_each_leaf_cfs_rq(env->busiest_rq, cfs_rq) {
+       for_each_leaf_cfs_rq(env->src_rq, cfs_rq) {
                list_for_each_entry_safe(p, n, &cfs_rq->tasks, se.group_node) {
                        if (throttled_lb_pair(task_group(p),
-                                             env->busiest_rq->cpu, env->this_cpu))
+                                             env->src_cpu, env->dst_cpu))
                                break;
 
                        if (!can_migrate_task(p, env))
                                continue;
 
-                       pull_task(env->busiest_rq, p, env->this_rq, env->this_cpu);
+                       move_task(p, env);
                        /*
-                        * Right now, this is only the second place pull_task()
-                        * is called, so we can safely collect pull_task()
-                        * stats here rather than inside pull_task().
+                        * Right now, this is only the second place move_task()
+                        * is called, so we can safely collect move_task()
+                        * stats here rather than inside move_task().
                         */
                        schedstat_inc(env->sd, lb_gained[env->idle]);
                        return 1;
@@ -3242,7 +3242,7 @@ static unsigned long balance_tasks(struct lb_env *env)
        if (env->max_load_move == 0)
                goto out;
 
-       list_for_each_entry_safe(p, n, &env->busiest_cfs_rq->tasks, se.group_node) {
+       list_for_each_entry_safe(p, n, &env->src_cfs_rq->tasks, se.group_node) {
                if (loops++ > sysctl_sched_nr_migrate) {
                        env->flags |= LBF_NEED_BREAK;
                        break;
@@ -3252,7 +3252,7 @@ static unsigned long balance_tasks(struct lb_env *env)
                    !can_migrate_task(p, env))
                        continue;
 
-               pull_task(env->busiest_rq, p, env->this_rq, env->this_cpu);
+               move_task(p, env);
                pulled++;
                rem_load_move -= p->se.load.weight;
 
@@ -3277,9 +3277,9 @@ static unsigned long balance_tasks(struct lb_env *env)
        }
 out:
        /*
-        * Right now, this is one of only two places pull_task() is called,
-        * so we can safely collect pull_task() stats here rather than
-        * inside pull_task().
+        * Right now, this is one of only two places move_task() is called,
+        * so we can safely collect move_task() stats here rather than
+        * inside move_task().
         */
        schedstat_add(env->sd, lb_gained[env->idle], pulled);
 
@@ -3372,11 +3372,11 @@ static unsigned long load_balance_fair(struct lb_env *env)
        long rem_load_move = env->max_load_move;
 
        rcu_read_lock();
-       update_h_load(cpu_of(env->busiest_rq));
+       update_h_load(cpu_of(env->src_rq));
 
-       for_each_leaf_cfs_rq(env->busiest_rq, env->busiest_cfs_rq) {
-               unsigned long busiest_h_load = env->busiest_cfs_rq->h_load;
-               unsigned long busiest_weight = env->busiest_cfs_rq->load.weight;
+       for_each_leaf_cfs_rq(env->src_rq, env->src_cfs_rq) {
+               unsigned long busiest_h_load = env->src_cfs_rq->h_load;
+               unsigned long busiest_weight = env->src_cfs_rq->load.weight;
                u64 rem_load, moved_load;
 
                if (env->flags & (LBF_NEED_BREAK|LBF_ABORT))
@@ -3385,12 +3385,12 @@ static unsigned long load_balance_fair(struct lb_env *env)
                /*
                 * empty group or part of a throttled hierarchy
                 */
-               if (!env->busiest_cfs_rq->task_weight)
+               if (!env->src_cfs_rq->task_weight)
                        continue;
 
-               if (throttled_lb_pair(env->busiest_cfs_rq->tg,
-                                     cpu_of(env->busiest_rq),
-                                     env->this_cpu))
+               if (throttled_lb_pair(env->src_cfs_rq->tg,
+                                     cpu_of(env->src_rq),
+                                     env->dst_cpu))
                        continue;
 
                rem_load = (u64)rem_load_move * busiest_weight;
@@ -3420,7 +3420,7 @@ static inline void update_shares(int cpu)
 
 static unsigned long load_balance_fair(struct lb_env *env)
 {
-       env->busiest_cfs_rq = &env->busiest_rq->cfs;
+       env->src_cfs_rq = &env->src_rq->cfs;
        return balance_tasks(env);
 }
 #endif
@@ -3451,7 +3451,7 @@ static int move_tasks(struct lb_env *env)
                 * kernels will stop after the first task is pulled to minimize
                 * the critical section.
                 */
-               if (env->idle == CPU_NEWLY_IDLE && env->this_rq->nr_running) {
+               if (env->idle == CPU_NEWLY_IDLE && env->dst_rq->nr_running) {
                        env->flags |= LBF_ABORT;
                        break;
                }
@@ -4461,8 +4461,8 @@ static int load_balance(int this_cpu, struct rq *this_rq,
 
        struct lb_env env = {
                .sd             = sd,
-               .this_cpu       = this_cpu,
-               .this_rq        = this_rq,
+               .dst_cpu        = this_cpu,
+               .dst_rq         = this_rq,
                .idle           = idle,
        };
 
@@ -4502,7 +4502,8 @@ redo:
                 */
                env.flags |= LBF_ALL_PINNED;
                env.max_load_move = imbalance;
-               env.busiest_rq = busiest;
+               env.src_cpu = busiest->cpu;
+               env.src_rq = busiest;
 
                local_irq_save(flags);
                double_rq_lock(this_rq, busiest);
@@ -4722,9 +4723,10 @@ static int active_load_balance_cpu_stop(void *data)
        if (likely(sd)) {
                struct lb_env env = {
                        .sd             = sd,
-                       .this_cpu       = target_cpu,
-                       .this_rq        = target_rq,
-                       .busiest_rq     = busiest_rq,
+                       .dst_cpu        = target_cpu,
+                       .dst_rq         = target_rq,
+                       .src_cpu        = busiest_rq->cpu,
+                       .src_rq         = busiest_rq,
                        .idle           = CPU_IDLE,
                };