rdb-5.13-mfs.patch

diff --git a/init/Kconfig b/init/Kconfig
index 089e3bfe5dbc..7c46008e5a7b 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -844,6 +844,40 @@ config CACULE_SCHED
 
 	  If unsure, say Y here.
 
+config CACULE_RDB
+	bool "RDB (Response Driven Balancer)"
+	default y
+	depends on CACULE_SCHED
+	help
+	  This is an experimental load balancer for CacULE. It is a lightweight
+	  load balancer which is a replacement of CFS load balancer. It migrates
+	  tasks based on their interactivity scores.
+
+	  If unsure, say Y here.
+
+config RDB_INTERVAL
+	int "RDB load balancer interval"
+	default 19
+	depends on CACULE_RDB
+	help
+	  This is an interval to control load balance time period.
+	  The trigger_load_balance runs in every tick. For High HZ values, the
+	  load balance could be overwhelming. RDB load balance includes rq locking
+	  which can reduce the performance. The balance interval can help to avoid
+	  running load balance on every tick. For example, RDB_INTERVAL=3 will
+	  only run load balance every 3ms. Setting RDB_INTERVAL depends on HZ.
+	  If you want load balancer run every 2ms while HZ=500 then it is not
+	  needed and better to set RDB_INTERVAL=0 since 500HZ already (1000ms
+	  / 500HZ = 2ms). However, if you have 1000HZ and want to avoid load
+	  balancer from running every 1ms, you could set RDB_INTERVAL=4ms for
+	  example to make load balancer run every 4ms. Less RDB_INTERVAL values
+	  (or 0 to disable) could make sure tasks are balanced ASAP, but with
+	  the cost of locking/blocking time. High RDB_INTERVAL values can relax
+	  balancing locking but with the cost of imbalanced workload for that
+	  period of time (i.e. if RDB_INTERVAL=100ms) there will be no balancing
+	  for 100ms (except for newidle_balance which is not effected by RDB_INTERVAL).
+
+	  If in doubt, use the default value.
 
 #
 # For architectures that want to enable the support for NUMA-affine scheduler
@@ -962,6 +996,7 @@ config CGROUP_WRITEBACK
 
 menuconfig CGROUP_SCHED
 	bool "CPU controller"
+	depends on !CACULE_RDB
 	default n
 	help
 	  This feature lets CPU scheduler recognize task groups and control CPU
@@ -1239,10 +1274,10 @@ config CHECKPOINT_RESTORE
 
 config SCHED_AUTOGROUP
 	bool "Automatic process group scheduling"
+	depends on !CACULE_RDB
 	select CGROUPS
 	select CGROUP_SCHED
 	select FAIR_GROUP_SCHED
-	default y
 	help
 	  This option optimizes the scheduler for common desktop workloads by
 	  automatically creating and populating task groups.  This separation
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index bdedde199504..18d2b5d41b36 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -8129,8 +8129,11 @@ void __init sched_init(void)
 	BUG_ON(&dl_sched_class + 1 != &stop_sched_class);
 #endif
 
-#ifdef CONFIG_CACULE_SCHED
-	printk(KERN_INFO "CacULE CPU scheduler v5.13-r2 by Hamad Al Marri.");
+#ifdef CONFIG_CACULE_RDB
+	/*
+	 * This patch is on top cacule-5.13.patch
+	 */
+	printk(KERN_INFO "CacULE CPU scheduler (RDB) v5.13-r2 by Hamad Al Marri.");
 #endif
 
 	wait_bit_init();
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index c64337fd121d..e64aa617b3ed 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -841,6 +841,10 @@ static void __dequeue_entity(struct cfs_rq *cfs_rq, struct sched_entity *_se)
 		cfs_rq->head = NULL;
 		cfs_rq->tail = NULL;
 
+#ifdef CONFIG_CACULE_RDB
+	WRITE_ONCE(cfs_rq->IS_head, ~0);
+#endif
+
 	} else if (se == cfs_rq->head) {
 		// if it is the head
 		cfs_rq->head		= cfs_rq->head->next;
@@ -4562,7 +4566,9 @@ place_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int initial)
 }
 #endif /* CONFIG_CACULE_SCHED */
 
+#if !defined(CONFIG_CACULE_RDB)
 static void check_enqueue_throttle(struct cfs_rq *cfs_rq);
+#endif
 
 static inline void check_schedstat_required(void)
 {
@@ -4671,6 +4677,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 		__enqueue_entity(cfs_rq, se);
 	se->on_rq = 1;
 
+#if !defined(CONFIG_CACULE_RDB)
 	/*
 	 * When bandwidth control is enabled, cfs might have been removed
 	 * because of a parent been throttled but cfs->nr_running > 1. Try to
@@ -4681,6 +4688,7 @@ enqueue_entity(struct cfs_rq *cfs_rq, struct sched_entity *se, int flags)
 
 	if (cfs_rq->nr_running == 1)
 		check_enqueue_throttle(cfs_rq);
+#endif
 }
 
 #if !defined(CONFIG_CACULE_SCHED)
@@ -5223,6 +5231,9 @@ static int tg_throttle_down(struct task_group *tg, void *data)
 
 static bool throttle_cfs_rq(struct cfs_rq *cfs_rq)
 {
+#ifdef CONFIG_CACULE_RDB
+	return false;
+#else
 	struct rq *rq = rq_of(cfs_rq);
 	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
 	struct sched_entity *se;
@@ -5300,10 +5311,12 @@ static bool throttle_cfs_rq(struct cfs_rq *cfs_rq)
 	cfs_rq->throttled = 1;
 	cfs_rq->throttled_clock = rq_clock(rq);
 	return true;
+#endif
 }
 
 void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
 {
+#if !defined(CONFIG_CACULE_RDB)
 	struct rq *rq = rq_of(cfs_rq);
 	struct cfs_bandwidth *cfs_b = tg_cfs_bandwidth(cfs_rq->tg);
 	struct sched_entity *se;
@@ -5385,6 +5398,7 @@ void unthrottle_cfs_rq(struct cfs_rq *cfs_rq)
 	/* Determine whether we need to wake up potentially idle CPU: */
 	if (rq->curr == rq->idle && rq->cfs.nr_running)
 		resched_curr(rq);
+#endif
 }
 
 static void distribute_cfs_runtime(struct cfs_bandwidth *cfs_b)
@@ -5837,7 +5851,11 @@ static inline bool cfs_bandwidth_used(void)
 
 static void account_cfs_rq_runtime(struct cfs_rq *cfs_rq, u64 delta_exec) {}
 static bool check_cfs_rq_runtime(struct cfs_rq *cfs_rq) { return false; }
+
+#if !defined(CONFIG_CACULE_RDB)
 static void check_enqueue_throttle(struct cfs_rq *cfs_rq) {}
+#endif
+
 static inline void sync_throttle(struct task_group *tg, int cpu) {}
 static __always_inline void return_cfs_rq_runtime(struct cfs_rq *cfs_rq) {}
 
@@ -6081,7 +6099,9 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
 	struct sched_entity *se = &p->se;
 	int task_sleep = flags & DEQUEUE_SLEEP;
 	int idle_h_nr_running = task_has_idle_policy(p);
+#if !defined(CONFIG_CACULE_RDB)
 	bool was_sched_idle = sched_idle_rq(rq);
+#endif
 
 	util_est_dequeue(&rq->cfs, p);
 
@@ -6132,9 +6152,11 @@ static void dequeue_task_fair(struct rq *rq, struct task_struct *p, int flags)
 	/* At this point se is NULL and we are at root level*/
 	sub_nr_running(rq, 1);
 
+#if !defined(CONFIG_CACULE_RDB)
 	/* balance early to pull high priority tasks */
 	if (unlikely(!was_sched_idle && sched_idle_rq(rq)))
 		rq->next_balance = jiffies;
+#endif
 
 dequeue_throttle:
 	util_est_update(&rq->cfs, p, task_sleep);
@@ -7659,11 +7681,23 @@ pick_next_task_fair(struct rq *rq, struct task_struct *prev, struct rq_flags *rf
 	if (prev)
 		put_prev_task(rq, prev);
 
+#ifdef CONFIG_CACULE_RDB
+	se = pick_next_entity(cfs_rq, NULL);
+	set_next_entity(cfs_rq, se);
+
+	if (cfs_rq->head) {
+		unsigned int IS_head = calc_interactivity(sched_clock(), cfs_rq->head);
+		WRITE_ONCE(cfs_rq->IS_head, IS_head);
+	} else {
+		WRITE_ONCE(cfs_rq->IS_head, ~0);
+	}
+#else
 	do {
 		se = pick_next_entity(cfs_rq, NULL);
 		set_next_entity(cfs_rq, se);
 		cfs_rq = group_cfs_rq(se);
 	} while (cfs_rq);
+#endif
 
 	p = task_of(se);
 
@@ -7685,6 +7719,10 @@ done: __maybe_unused;
 	return p;
 
 idle:
+#ifdef CONFIG_CACULE_RDB
+	WRITE_ONCE(cfs_rq->IS_head, ~0);
+#endif
+
 	if (!rf)
 		return NULL;
 
@@ -7991,6 +8029,7 @@ struct lb_env {
 	struct list_head	tasks;
 };
 
+#if !defined(CONFIG_CACULE_RDB)
 /*
  * Is this task likely cache-hot:
  */
@@ -8412,6 +8451,7 @@ static void attach_tasks(struct lb_env *env)
 
 	rq_unlock(env->dst_rq, &rf);
 }
+#endif
 
 #ifdef CONFIG_NO_HZ_COMMON
 static inline bool cfs_rq_has_blocked(struct cfs_rq *cfs_rq)
@@ -8461,6 +8501,7 @@ static inline void update_blocked_load_tick(struct rq *rq) {}
 static inline void update_blocked_load_status(struct rq *rq, bool has_blocked) {}
 #endif
 
+#if !defined(CONFIG_CACULE_RDB)
 static bool __update_blocked_others(struct rq *rq, bool *done)
 {
 	const struct sched_class *curr_class;
@@ -8486,6 +8527,7 @@ static bool __update_blocked_others(struct rq *rq, bool *done)
 
 	return decayed;
 }
+#endif
 
 #ifdef CONFIG_FAIR_GROUP_SCHED
 
@@ -8576,6 +8618,7 @@ static unsigned long task_h_load(struct task_struct *p)
 			cfs_rq_load_avg(cfs_rq) + 1);
 }
 #else
+#if !defined(CONFIG_CACULE_RDB)
 static bool __update_blocked_fair(struct rq *rq, bool *done)
 {
 	struct cfs_rq *cfs_rq = &rq->cfs;
@@ -8587,6 +8630,7 @@ static bool __update_blocked_fair(struct rq *rq, bool *done)
 
 	return decayed;
 }
+#endif
 
 static unsigned long task_h_load(struct task_struct *p)
 {
@@ -8594,6 +8638,7 @@ static unsigned long task_h_load(struct task_struct *p)
 }
 #endif
 
+#if !defined(CONFIG_CACULE_RDB)
 static void update_blocked_averages(int cpu)
 {
 	bool decayed = false, done = true;
@@ -8612,6 +8657,7 @@ static void update_blocked_averages(int cpu)
 		cpufreq_update_util(rq, 0);
 	rq_unlock_irqrestore(rq, &rf);
 }
+#endif
 
 /********** Helpers for find_busiest_group ************************/
 
@@ -9715,6 +9761,7 @@ static inline void calculate_imbalance(struct lb_env *env, struct sd_lb_stats *s
  *            different in groups.
  */
 
+#if !defined(CONFIG_CACULE_RDB)
 /**
  * find_busiest_group - Returns the busiest group within the sched_domain
  * if there is an imbalance.
@@ -9983,6 +10030,7 @@ static struct rq *find_busiest_queue(struct lb_env *env,
 
 	return busiest;
 }
+#endif
 
 /*
  * Max backoff if we encounter pinned tasks. Pretty arbitrary value, but
@@ -10019,6 +10067,7 @@ imbalanced_active_balance(struct lb_env *env)
 	return 0;
 }
 
+#if !defined(CONFIG_CACULE_RDB)
 static int need_active_balance(struct lb_env *env)
 {
 	struct sched_domain *sd = env->sd;
@@ -10351,6 +10400,7 @@ static int load_balance(int this_cpu, struct rq *this_rq,
 out:
 	return ld_moved;
 }
+#endif
 
 static inline unsigned long
 get_sd_balance_interval(struct sched_domain *sd, int cpu_busy)
@@ -10389,6 +10439,7 @@ update_next_balance(struct sched_domain *sd, unsigned long *next_balance)
 		*next_balance = next;
 }
 
+#if !defined(CONFIG_CACULE_RDB)
 /*
  * active_load_balance_cpu_stop is run by the CPU stopper. It pushes
  * running tasks off the busiest CPU onto idle CPUs. It requires at
@@ -10474,6 +10525,7 @@ static int active_load_balance_cpu_stop(void *data)
 }
 
 static DEFINE_SPINLOCK(balancing);
+#endif
 
 /*
  * Scale the max load_balance interval with the number of CPUs in the system.
@@ -10484,6 +10536,7 @@ void update_max_interval(void)
 	max_load_balance_interval = HZ*num_online_cpus()/10;
 }
 
+#if !defined(CONFIG_CACULE_RDB)
 /*
  * It checks each scheduling domain to see if it is due to be balanced,
  * and initiates a balancing operation if so.
@@ -10576,6 +10629,7 @@ static void rebalance_domains(struct rq *rq, enum cpu_idle_type idle)
 		rq->next_balance = next_balance;
 
 }
+#endif
 
 static inline int on_null_domain(struct rq *rq)
 {
@@ -10609,6 +10663,7 @@ static inline int find_new_ilb(void)
 	return nr_cpu_ids;
 }
 
+#if !defined(CONFIG_CACULE_RDB)
 /*
  * Kick a CPU to do the nohz balancing, if it is time for it. We pick any
  * idle CPU in the HK_FLAG_MISC housekeeping set (if there is one).
@@ -10759,6 +10814,7 @@ static void nohz_balancer_kick(struct rq *rq)
 	if (flags)
 		kick_ilb(flags);
 }
+#endif /* CONFIG_CACULE_RDB */
 
 static void set_cpu_sd_state_busy(int cpu)
 {
@@ -10879,11 +10935,17 @@ static bool update_nohz_stats(struct rq *rq)
 	if (!time_after(jiffies, READ_ONCE(rq->last_blocked_load_update_tick)))
 		return true;
 
+#if !defined(CONFIG_CACULE_RDB)
 	update_blocked_averages(cpu);
+#endif
 
 	return rq->has_blocked_load;
 }
 
+#ifdef CONFIG_CACULE_RDB
+static int idle_try_pull_any(struct cfs_rq *cfs_rq);
+#endif
+
 /*
  * Internal function that runs load balance for all idle cpus. The load balance
  * can be a simple update of blocked load or a complete load balance with
@@ -10953,7 +11015,11 @@ static void _nohz_idle_balance(struct rq *this_rq, unsigned int flags,
 			rq_unlock_irqrestore(rq, &rf);
 
 			if (flags & NOHZ_BALANCE_KICK)
+#if !defined(CONFIG_CACULE_RDB)
 				rebalance_domains(rq, CPU_IDLE);
+#else
+				idle_try_pull_any(&rq->cfs);
+#endif
 		}
 
 		if (time_after(next_balance, rq->next_balance)) {
@@ -10979,6 +11045,7 @@ static void _nohz_idle_balance(struct rq *this_rq, unsigned int flags,
 		WRITE_ONCE(nohz.has_blocked, 1);
 }
 
+#if !defined(CONFIG_CACULE_RDB)
 /*
  * In CONFIG_NO_HZ_COMMON case, the idle balance kickee will do the
  * rebalancing for all the cpus for whom scheduler ticks are stopped.
@@ -10999,6 +11066,7 @@ static bool nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle)
 
 	return true;
 }
+#endif
 
 /*
  * Check if we need to run the ILB for updating blocked load before entering
@@ -11048,6 +11116,7 @@ static void nohz_newidle_balance(struct rq *this_rq)
 }
 
 #else /* !CONFIG_NO_HZ_COMMON */
+#if !defined(CONFIG_CACULE_RDB)
 static inline void nohz_balancer_kick(struct rq *rq) { }
 
 static inline bool nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle)
@@ -11056,8 +11125,130 @@ static inline bool nohz_idle_balance(struct rq *this_rq, enum cpu_idle_type idle
 }
 
 static inline void nohz_newidle_balance(struct rq *this_rq) { }
+#endif
+
 #endif /* CONFIG_NO_HZ_COMMON */
 
+#ifdef CONFIG_CACULE_RDB
+static int
+can_migrate_task(struct task_struct *p, int dst_cpu, struct rq *src_rq)
+{
+	if (task_running(src_rq, p))
+		return 0;
+
+	if (!cpumask_test_cpu(dst_cpu, p->cpus_ptr))
+		return 0;
+
+	if (p->se.exec_start == 0)
+		return 0;
+
+	return 1;
+}
+
+static void push_to_unlock(struct rq *this_rq,
+			   struct rq *dst_rq,
+			   struct task_struct *p,
+			   int dst_cpu)
+{
+	struct rq_flags rf;
+
+	// detach task
+	deactivate_task(this_rq, p, DEQUEUE_NOCLOCK);
+	set_task_cpu(p, dst_cpu);
+
+	// unlock this rq
+	raw_spin_unlock(&this_rq->lock);
+
+	/* push to */
+	rq_lock_irqsave(dst_rq, &rf);
+	update_rq_clock(dst_rq);
+
+	activate_task(dst_rq, p, ENQUEUE_NOCLOCK);
+	check_preempt_curr(dst_rq, p, 0);
+
+	// unlock src rq
+	rq_unlock(dst_rq, &rf);
+	local_irq_restore(rf.flags);
+}
+
+static void pull_from_unlock(struct rq *this_rq,
+			     struct rq *src_rq,
+			     struct rq_flags *rf,
+			     struct task_struct *p,
+			     int dst_cpu)
+{
+	// detach task
+	deactivate_task(src_rq, p, DEQUEUE_NOCLOCK);
+	set_task_cpu(p, dst_cpu);
+
+	// unlock src rq
+	rq_unlock(src_rq, rf);
+	local_irq_restore(rf->flags);
+
+	// lock this rq
+	raw_spin_lock(&this_rq->lock);
+	update_rq_clock(this_rq);
+
+	activate_task(this_rq, p, ENQUEUE_NOCLOCK);
+	check_preempt_curr(this_rq, p, 0);
+
+	// unlock this rq
+	raw_spin_unlock(&this_rq->lock);
+}
+
+static inline struct rq *
+find_max_IS_rq(struct cfs_rq *cfs_rq, int dst_cpu)
+{
+	struct rq *tmp_rq, *max_rq = NULL;
+	int cpu;
+	unsigned int max_IS = cfs_rq->IS_head;
+	unsigned int local_IS;
+
+	// find max hrrn
+	for_each_online_cpu(cpu) {
+		if (cpu == dst_cpu)
+			continue;
+
+		tmp_rq = cpu_rq(cpu);
+
+		if (tmp_rq->cfs.nr_running < 2 || !tmp_rq->cfs.head)
+			continue;
+
+		local_IS = READ_ONCE(tmp_rq->cfs.IS_head);
+
+		if (local_IS < max_IS) {
+			max_IS = local_IS;
+			max_rq = tmp_rq;
+		}
+	}
+
+	return max_rq;
+}
+
+static int try_pull_from(struct rq *src_rq, struct rq *this_rq)
+{
+	struct rq_flags rf;
+	int dst_cpu = cpu_of(this_rq);
+	struct task_struct *p;
+
+	rq_lock_irqsave(src_rq, &rf);
+	update_rq_clock(src_rq);
+
+	if (src_rq->cfs.head && src_rq->cfs.nr_running > 1) {
+		p = task_of(se_of(src_rq->cfs.head));
+
+		if (can_migrate_task(p, dst_cpu, src_rq)) {
+			pull_from_unlock(this_rq, src_rq, &rf, p, dst_cpu);
+			return 1;
+		}
+	}
+
+	rq_unlock(src_rq, &rf);
+	local_irq_restore(rf.flags);
+
+	return 0;
+}
+
 /*
  * newidle_balance is called by schedule() if this_cpu is about to become
  * idle. Attempts to pull tasks from other CPUs.
@@ -11068,6 +11259,107 @@ static inline void nohz_newidle_balance(struct rq *this_rq) { }
  *   > 0 - success, new (fair) tasks present
  */
 static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
+{
+	int this_cpu = this_rq->cpu;
+	struct task_struct *p = NULL;
+	struct rq *src_rq;
+	int src_cpu;
+	struct rq_flags src_rf;
+	int pulled_task = 0;
+	int cores_round = 1;
+
+	update_misfit_status(NULL, this_rq);
+	/*
+	 * We must set idle_stamp _before_ calling idle_balance(), such that we
+	 * measure the duration of idle_balance() as idle time.
+	 */
+	this_rq->idle_stamp = rq_clock(this_rq);
+
+	/*
+	 * Do not pull tasks towards !active CPUs...
+	 */
+	if (!cpu_active(this_cpu))
+		return 0;
+
+	/*
+	 * This is OK, because current is on_cpu, which avoids it being picked
+	 * for load-balance and preemption/IRQs are still disabled avoiding
+	 * further scheduler activity on it and we're being very careful to
+	 * re-start the picking loop.
+	 */
+	rq_unpin_lock(this_rq, rf);
+	raw_spin_unlock(&this_rq->lock);
+
+again:
+	for_each_online_cpu(src_cpu) {
+
+		if (src_cpu == this_cpu)
+			continue;
+
+		if (cores_round && !cpus_share_cache(src_cpu, this_cpu))
+			continue;
+
+		src_rq = cpu_rq(src_cpu);
+
+		rq_lock_irqsave(src_rq, &src_rf);
+		update_rq_clock(src_rq);
+
+		if (src_rq->cfs.nr_running < 2 || !(src_rq->cfs.head))
+			goto next;
+
+		p = task_of(se_of(src_rq->cfs.head));
+
+		if (can_migrate_task(p, this_cpu, src_rq)) {
+			pull_from_unlock(this_rq, src_rq, &src_rf, p, this_cpu);
+
+			pulled_task = 1;
+			goto out;
+		}
+
+next:
+		rq_unlock(src_rq, &src_rf);
+		local_irq_restore(src_rf.flags);
+
+		/*
+		 * Stop searching for tasks to pull if there are
+		 * now runnable tasks on this rq.
+		 */
+		if (pulled_task || this_rq->nr_running > 0)
+			goto out;
+	}
+
+	if (cores_round) {
+		// now search for all cpus
+		cores_round = 0;
+		goto again;
+	}
+
+out:
+	raw_spin_lock(&this_rq->lock);
+
+	/*
+	 * While browsing the domains, we released the rq lock, a task could
+	 * have been enqueued in the meantime. Since we're not going idle,
+	 * pretend we pulled a task.
+	 */
+	if (this_rq->cfs.h_nr_running && !pulled_task)
+		pulled_task = 1;
+
+	/* Is there a task of a high priority class? */
+	if (this_rq->nr_running != this_rq->cfs.h_nr_running)
+		pulled_task = -1;
+
+	if (pulled_task)
+		this_rq->idle_stamp = 0;
+	else
+		nohz_newidle_balance(this_rq);
+
+	rq_repin_lock(this_rq, rf);
+
+	return pulled_task;
+}
+#else
+static int newidle_balance(struct rq *this_rq, struct rq_flags *rf)
 {
 	unsigned long next_balance = jiffies + HZ;
 	int this_cpu = this_rq->cpu;
@@ -11224,6 +11516,214 @@ void trigger_load_balance(struct rq *rq)
 
 	nohz_balancer_kick(rq);
 }
+#endif
+
+#ifdef CONFIG_CACULE_RDB
+static int
+idle_try_pull_any(struct cfs_rq *cfs_rq)
+{
+	struct task_struct *p = NULL;
+	struct rq *this_rq = rq_of(cfs_rq), *src_rq;
+	int dst_cpu = cpu_of(this_rq);
+	int src_cpu;
+	struct rq_flags rf;
+	int pulled = 0;
+	int cores_round = 1;
+
+again:
+	for_each_online_cpu(src_cpu) {
+
+		if (src_cpu == dst_cpu)
+			continue;
+
+		if (cores_round && !cpus_share_cache(src_cpu, dst_cpu))
+			continue;
+
+		src_rq = cpu_rq(src_cpu);
+
+		rq_lock_irqsave(src_rq, &rf);
+		update_rq_clock(src_rq);
+
+		if (src_rq->cfs.nr_running < 2 || !(src_rq->cfs.head))
+			goto next;
+
+		p = task_of(se_of(src_rq->cfs.head));
+
+		if (can_migrate_task(p, dst_cpu, src_rq)) {
+			pull_from_unlock(this_rq, src_rq, &rf, p, dst_cpu);
+			pulled = 1;
+			goto out;
+		}
+
+next:
+		rq_unlock(src_rq, &rf);
+		local_irq_restore(rf.flags);
+	}
+
+	if (cores_round) {
+		// now search for all cpus
+		cores_round = 0;
+		goto again;
+	}
+
+out:
+	return pulled;
+}
+
+
+static int
+try_pull_higher_IS(struct cfs_rq *cfs_rq)
+{
+	struct rq *this_rq = rq_of(cfs_rq), *max_rq;
+	int dst_cpu = cpu_of(this_rq);
+
+	max_rq = find_max_IS_rq(cfs_rq, dst_cpu);
+
+	if (!max_rq)
+		return 0;
+
+	if (try_pull_from(max_rq, this_rq))
+		return 1;
+
+	return 0;
+}
+
+static void try_push_any(struct rq *this_rq)
+{
+	struct task_struct *p = NULL;
+	struct rq *dst_rq;
+	int dst_cpu;
+	int src_cpu = cpu_of(this_rq);
+	int cores_round = 1;
+
+again:
+	for_each_online_cpu(dst_cpu) {
+
+		if (dst_cpu == src_cpu)
+			continue;
+
+		if (cores_round && !cpus_share_cache(src_cpu, dst_cpu))
+			continue;
+
+		dst_rq = cpu_rq(dst_cpu);
+
+		if (dst_rq->cfs.nr_running >= this_rq->cfs.nr_running - 1)
+			continue;
+
+		// lock this rq
+		raw_spin_lock(&this_rq->lock);
+		update_rq_clock(this_rq);
+
+		if (!this_rq->cfs.head) {
+			// unlock this rq
+			raw_spin_unlock(&this_rq->lock);
+			return;
+		}
+
+		p = task_of(se_of(this_rq->cfs.head));
+
+		if (can_migrate_task(p, dst_cpu, this_rq)) {
+			push_to_unlock(this_rq, dst_rq, p, dst_cpu);
+			return;
+		}
+
+		// unlock this rq
+		raw_spin_unlock(&this_rq->lock);
+	}
+
+	if (cores_round) {
+		// now search for all cpus
+		cores_round = 0;
+		goto again;
+	}
+}
+
+static void try_pull_any(struct rq *this_rq)
+{
+	struct task_struct *p = NULL;
+	struct rq *src_rq;
+	int dst_cpu = cpu_of(this_rq);
+	int src_cpu;
+	struct rq_flags src_rf;
+	int cores_round = 1;
+	unsigned int this_head = this_rq->cfs.IS_head;
+
+again:
+	for_each_online_cpu(src_cpu) {
+
+		if (src_cpu == dst_cpu)
+			continue;
+
+		if (cores_round && !cpus_share_cache(src_cpu, dst_cpu))
+			continue;
+
+		src_rq = cpu_rq(src_cpu);
+
+		if (src_rq->cfs.nr_running < 2 || !(src_rq->cfs.head)
+                    || READ_ONCE(src_rq->cfs.IS_head) >= this_head)
+			continue;
+
+		rq_lock_irqsave(src_rq, &src_rf);
+		update_rq_clock(src_rq);
+
+		if (src_rq->cfs.nr_running < 2 || !(src_rq->cfs.head)
+                    || src_rq->cfs.IS_head >= this_head)
+			goto next;
+
+		p = task_of(se_of(src_rq->cfs.head));
+
+		if (can_migrate_task(p, dst_cpu, src_rq)) {
+			pull_from_unlock(this_rq, src_rq, &src_rf, p, dst_cpu);
+			return;
+		}
+
+next:
+		rq_unlock(src_rq, &src_rf);
+		local_irq_restore(src_rf.flags);
+	}
+
+	if (cores_round) {
+		// now search for all cpus
+		cores_round = 0;
+		goto again;
+	}
+}
+
+static inline void
+active_balance(struct rq *rq)
+{
+	struct cfs_rq *cfs_rq = &rq->cfs;
+
+	if (!cfs_rq->head || cfs_rq->nr_running < 2)
+		try_pull_higher_IS(&rq->cfs);
+	else {
+		try_push_any(rq);
+		try_pull_any(rq);
+	}
+}
+
+void trigger_load_balance(struct rq *rq)
+{
+	unsigned long interval;
+
+#ifdef CONFIG_RDB_INTERVAL
+	if (time_before(jiffies, rq->next_balance))
+		return;
+#endif
+
+	if (rq->idle_balance)
+		idle_try_pull_any(&rq->cfs);
+	else {
+		active_balance(rq);
+
+#ifdef CONFIG_RDB_INTERVAL
+		/* scale ms to jiffies */
+		interval = msecs_to_jiffies(CONFIG_RDB_INTERVAL);
+		rq->next_balance = jiffies + interval;
+#endif
+	}
+}
+#endif
 
 static void rq_online_fair(struct rq *rq)
 {
@@ -11864,7 +12364,9 @@ void show_numa_stats(struct task_struct *p, struct seq_file *m)
 __init void init_sched_fair_class(void)
 {
 #ifdef CONFIG_SMP
+#if !defined(CONFIG_CACULE_RDB)
 	open_softirq(SCHED_SOFTIRQ, run_rebalance_domains);
+#endif
 
 #ifdef CONFIG_NO_HZ_COMMON
 	nohz.next_balance = jiffies;
diff --git a/kernel/sched/idle.c b/kernel/sched/idle.c
index 7ca3d3d86c2a..a7422dea8a9f 100644
--- a/kernel/sched/idle.c
+++ b/kernel/sched/idle.c
@@ -262,10 +262,12 @@ static void do_idle(void)
 {
 	int cpu = smp_processor_id();
 
+#if !defined(CONFIG_CACULE_RDB)
 	/*
 	 * Check if we need to update blocked load
 	 */
 	nohz_run_idle_balance(cpu);
+#endif
 
 	/*
 	 * If the arch has a polling bit, we maintain an invariant:
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 09a8290fc883..4d46ef7190bf 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -544,6 +544,10 @@ struct cfs_rq {
 	struct cacule_node	*head;
 	struct cacule_node	*tail;
 
+#ifdef CONFIG_CACULE_RDB
+	unsigned int		IS_head;
+#endif
+
 #else
 	struct sched_entity	*next;
 	struct sched_entity	*last;