6 files changed, 402 insertions, 77 deletions

diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index f4d02201f424..5800b0623ff3 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -701,6 +701,7 @@ static void update_rq_clock_task(struct rq *rq, s64 delta)
 
 	rq->prev_irq_time += irq_delta;
 	delta -= irq_delta;
+	psi_account_irqtime(rq->curr, irq_delta);
 #endif
 #ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
 	if (static_key_false((&paravirt_steal_rq_enabled))) {
@@ -4389,6 +4390,17 @@ void set_numabalancing_state(bool enabled)
 }
 
 #ifdef CONFIG_PROC_SYSCTL
+static void reset_memory_tiering(void)
+{
+	struct pglist_data *pgdat;
+
+	for_each_online_pgdat(pgdat) {
+		pgdat->nbp_threshold = 0;
+		pgdat->nbp_th_nr_cand = node_page_state(pgdat, PGPROMOTE_CANDIDATE);
+		pgdat->nbp_th_start = jiffies_to_msecs(jiffies);
+	}
+}
+
 int sysctl_numa_balancing(struct ctl_table *table, int write,
 			  void *buffer, size_t *lenp, loff_t *ppos)
 {
@@ -4405,6 +4417,9 @@ int sysctl_numa_balancing(struct ctl_table *table, int write,
 	if (err < 0)
 		return err;
 	if (write) {
+		if (!(sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING) &&
+		    (state & NUMA_BALANCING_MEMORY_TIERING))
+			reset_memory_tiering();
 		sysctl_numa_balancing_mode = state;
 		__set_numabalancing_state(state);
 	}
@@ -5159,6 +5174,7 @@ context_switch(struct rq *rq, struct task_struct *prev,
 		 * finish_task_switch()'s mmdrop().
 		 */
 		switch_mm_irqs_off(prev->active_mm, next->mm, next);
+		lru_gen_use_mm(next->mm);
 
 		if (!prev->mm) {                        // from kernel
 			/* will mmdrop() in finish_task_switch(). */
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 667876da8382..1637b65ba07a 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -333,6 +333,7 @@ static __init int sched_init_debug(void)
 	debugfs_create_u32("scan_period_min_ms", 0644, numa, &sysctl_numa_balancing_scan_period_min);
 	debugfs_create_u32("scan_period_max_ms", 0644, numa, &sysctl_numa_balancing_scan_period_max);
 	debugfs_create_u32("scan_size_mb", 0644, numa, &sysctl_numa_balancing_scan_size);
+	debugfs_create_u32("hot_threshold_ms", 0644, numa, &sysctl_numa_balancing_hot_threshold);
 #endif
 
 	debugfs_create_file("debug", 0444, debugfs_sched, NULL, &sched_debug_fops);
diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 5ffec4370602..e4a0b8bd941c 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -40,6 +40,7 @@
 
 #include <linux/cpuidle.h>
 #include <linux/interrupt.h>
+#include <linux/memory-tiers.h>
 #include <linux/mempolicy.h>
 #include <linux/mutex_api.h>
 #include <linux/profile.h>
@@ -1090,6 +1091,12 @@ unsigned int sysctl_numa_balancing_scan_size = 256;
 /* Scan @scan_size MB every @scan_period after an initial @scan_delay in ms */
 unsigned int sysctl_numa_balancing_scan_delay = 1000;
 
+/* The page with hint page fault latency < threshold in ms is considered hot */
+unsigned int sysctl_numa_balancing_hot_threshold = MSEC_PER_SEC;
+
+/* Restrict the NUMA promotion throughput (MB/s) for each target node. */
+unsigned int sysctl_numa_balancing_promote_rate_limit = 65536;
+
 struct numa_group {
 	refcount_t refcount;
 
@@ -1432,6 +1439,120 @@ static inline unsigned long group_weight(struct task_struct *p, int nid,
 	return 1000 * faults / total_faults;
 }
 
+/*
+ * If memory tiering mode is enabled, cpupid of slow memory page is
+ * used to record scan time instead of CPU and PID.  When tiering mode
+ * is disabled at run time, the scan time (in cpupid) will be
+ * interpreted as CPU and PID.  So CPU needs to be checked to avoid to
+ * access out of array bound.
+ */
+static inline bool cpupid_valid(int cpupid)
+{
+	return cpupid_to_cpu(cpupid) < nr_cpu_ids;
+}
+
+/*
+ * For memory tiering mode, if there are enough free pages (more than
+ * enough watermark defined here) in fast memory node, to take full
+ * advantage of fast memory capacity, all recently accessed slow
+ * memory pages will be migrated to fast memory node without
+ * considering hot threshold.
+ */
+static bool pgdat_free_space_enough(struct pglist_data *pgdat)
+{
+	int z;
+	unsigned long enough_wmark;
+
+	enough_wmark = max(1UL * 1024 * 1024 * 1024 >> PAGE_SHIFT,
+			   pgdat->node_present_pages >> 4);
+	for (z = pgdat->nr_zones - 1; z >= 0; z--) {
+		struct zone *zone = pgdat->node_zones + z;
+
+		if (!populated_zone(zone))
+			continue;
+
+		if (zone_watermark_ok(zone, 0,
+				      wmark_pages(zone, WMARK_PROMO) + enough_wmark,
+				      ZONE_MOVABLE, 0))
+			return true;
+	}
+	return false;
+}
+
+/*
+ * For memory tiering mode, when page tables are scanned, the scan
+ * time will be recorded in struct page in addition to make page
+ * PROT_NONE for slow memory page.  So when the page is accessed, in
+ * hint page fault handler, the hint page fault latency is calculated
+ * via,
+ *
+ *	hint page fault latency = hint page fault time - scan time
+ *
+ * The smaller the hint page fault latency, the higher the possibility
+ * for the page to be hot.
+ */
+static int numa_hint_fault_latency(struct page *page)
+{
+	int last_time, time;
+
+	time = jiffies_to_msecs(jiffies);
+	last_time = xchg_page_access_time(page, time);
+
+	return (time - last_time) & PAGE_ACCESS_TIME_MASK;
+}
+
+/*
+ * For memory tiering mode, too high promotion/demotion throughput may
+ * hurt application latency.  So we provide a mechanism to rate limit
+ * the number of pages that are tried to be promoted.
+ */
+static bool numa_promotion_rate_limit(struct pglist_data *pgdat,
+				      unsigned long rate_limit, int nr)
+{
+	unsigned long nr_cand;
+	unsigned int now, start;
+
+	now = jiffies_to_msecs(jiffies);
+	mod_node_page_state(pgdat, PGPROMOTE_CANDIDATE, nr);
+	nr_cand = node_page_state(pgdat, PGPROMOTE_CANDIDATE);
+	start = pgdat->nbp_rl_start;
+	if (now - start > MSEC_PER_SEC &&
+	    cmpxchg(&pgdat->nbp_rl_start, start, now) == start)
+		pgdat->nbp_rl_nr_cand = nr_cand;
+	if (nr_cand - pgdat->nbp_rl_nr_cand >= rate_limit)
+		return true;
+	return false;
+}
+
+#define NUMA_MIGRATION_ADJUST_STEPS	16
+
+static void numa_promotion_adjust_threshold(struct pglist_data *pgdat,
+					    unsigned long rate_limit,
+					    unsigned int ref_th)
+{
+	unsigned int now, start, th_period, unit_th, th;
+	unsigned long nr_cand, ref_cand, diff_cand;
+
+	now = jiffies_to_msecs(jiffies);
+	th_period = sysctl_numa_balancing_scan_period_max;
+	start = pgdat->nbp_th_start;
+	if (now - start > th_period &&
+	    cmpxchg(&pgdat->nbp_th_start, start, now) == start) {
+		ref_cand = rate_limit *
+			sysctl_numa_balancing_scan_period_max / MSEC_PER_SEC;
+		nr_cand = node_page_state(pgdat, PGPROMOTE_CANDIDATE);
+		diff_cand = nr_cand - pgdat->nbp_th_nr_cand;
+		unit_th = ref_th * 2 / NUMA_MIGRATION_ADJUST_STEPS;
+		th = pgdat->nbp_threshold ? : ref_th;
+		if (diff_cand > ref_cand * 11 / 10)
+			th = max(th - unit_th, unit_th);
+		else if (diff_cand < ref_cand * 9 / 10)
+			th = min(th + unit_th, ref_th * 2);
+		pgdat->nbp_th_nr_cand = nr_cand;
+		pgdat->nbp_threshold = th;
+	}
+}
+
 bool should_numa_migrate_memory(struct task_struct *p, struct page * page,
 				int src_nid, int dst_cpu)
 {
@@ -1439,9 +1560,44 @@ bool should_numa_migrate_memory(struct task_struct *p, struct page * page,
 	int dst_nid = cpu_to_node(dst_cpu);
 	int last_cpupid, this_cpupid;
 
+	/*
+	 * The pages in slow memory node should be migrated according
+	 * to hot/cold instead of private/shared.
+	 */
+	if (sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING &&
+	    !node_is_toptier(src_nid)) {
+		struct pglist_data *pgdat;
+		unsigned long rate_limit;
+		unsigned int latency, th, def_th;
+
+		pgdat = NODE_DATA(dst_nid);
+		if (pgdat_free_space_enough(pgdat)) {
+			/* workload changed, reset hot threshold */
+			pgdat->nbp_threshold = 0;
+			return true;
+		}
+
+		def_th = sysctl_numa_balancing_hot_threshold;
+		rate_limit = sysctl_numa_balancing_promote_rate_limit << \
+			(20 - PAGE_SHIFT);
+		numa_promotion_adjust_threshold(pgdat, rate_limit, def_th);
+
+		th = pgdat->nbp_threshold ? : def_th;
+		latency = numa_hint_fault_latency(page);
+		if (latency >= th)
+			return false;
+
+		return !numa_promotion_rate_limit(pgdat, rate_limit,
+						  thp_nr_pages(page));
+	}
+
 	this_cpupid = cpu_pid_to_cpupid(dst_cpu, current->pid);
 	last_cpupid = page_cpupid_xchg_last(page, this_cpupid);
 
+	if (!(sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING) &&
+	    !node_is_toptier(src_nid) && !cpupid_valid(last_cpupid))
+		return false;
+
 	/*
 	 * Allow first faults or private faults to migrate immediately early in
 	 * the lifetime of a task. The magic number 4 is based on waiting for
@@ -2681,6 +2837,15 @@ void task_numa_fault(int last_cpupid, int mem_node, int pages, int flags)
 	if (!p->mm)
 		return;
 
+	/*
+	 * NUMA faults statistics are unnecessary for the slow memory
+	 * node for memory tiering mode.
+	 */
+	if (!node_is_toptier(mem_node) &&
+	    (sysctl_numa_balancing_mode & NUMA_BALANCING_MEMORY_TIERING ||
+	     !cpupid_valid(last_cpupid)))
+		return;
+
 	/* Allocate buffer to track faults on a per-node basis */
 	if (unlikely(!p->numa_faults)) {
 		int size = sizeof(*p->numa_faults) *
@@ -2761,6 +2926,7 @@ static void task_numa_work(struct callback_head *work)
 	struct task_struct *p = current;
 	struct mm_struct *mm = p->mm;
 	u64 runtime = p->se.sum_exec_runtime;
+	MA_STATE(mas, &mm->mm_mt, 0, 0);
 	struct vm_area_struct *vma;
 	unsigned long start, end;
 	unsigned long nr_pte_updates = 0;
@@ -2817,13 +2983,16 @@ static void task_numa_work(struct callback_head *work)
 
 	if (!mmap_read_trylock(mm))
 		return;
-	vma = find_vma(mm, start);
+	mas_set(&mas, start);
+	vma = mas_find(&mas, ULONG_MAX);
 	if (!vma) {
 		reset_ptenuma_scan(p);
 		start = 0;
-		vma = mm->mmap;
+		mas_set(&mas, start);
+		vma = mas_find(&mas, ULONG_MAX);
 	}
-	for (; vma; vma = vma->vm_next) {
+
+	for (; vma; vma = mas_find(&mas, ULONG_MAX)) {
 		if (!vma_migratable(vma) || !vma_policy_mof(vma) ||
 			is_vm_hugetlb_page(vma) || (vma->vm_flags & VM_MIXEDMAP)) {
 			continue;
diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
index 7f6030091aee..ee2ecc081422 100644
--- a/kernel/sched/psi.c
+++ b/kernel/sched/psi.c
@@ -181,6 +181,7 @@ static void group_init(struct psi_group *group)
 {
 	int cpu;
 
+	group->enabled = true;
 	for_each_possible_cpu(cpu)
 		seqcount_init(&per_cpu_ptr(group->pcpu, cpu)->seq);
 	group->avg_last_update = sched_clock();
@@ -201,6 +202,7 @@ void __init psi_init(void)
 {
 	if (!psi_enable) {
 		static_branch_enable(&psi_disabled);
+		static_branch_disable(&psi_cgroups_enabled);
 		return;
 	}
 
@@ -211,7 +213,7 @@ void __init psi_init(void)
 	group_init(&psi_system);
 }
 
-static bool test_state(unsigned int *tasks, enum psi_states state)
+static bool test_state(unsigned int *tasks, enum psi_states state, bool oncpu)
 {
 	switch (state) {
 	case PSI_IO_SOME:
@@ -224,9 +226,9 @@ static bool test_state(unsigned int *tasks, enum psi_states state)
 		return unlikely(tasks[NR_MEMSTALL] &&
 			tasks[NR_RUNNING] == tasks[NR_MEMSTALL_RUNNING]);
 	case PSI_CPU_SOME:
-		return unlikely(tasks[NR_RUNNING] > tasks[NR_ONCPU]);
+		return unlikely(tasks[NR_RUNNING] > oncpu);
 	case PSI_CPU_FULL:
-		return unlikely(tasks[NR_RUNNING] && !tasks[NR_ONCPU]);
+		return unlikely(tasks[NR_RUNNING] && !oncpu);
 	case PSI_NONIDLE:
 		return tasks[NR_IOWAIT] || tasks[NR_MEMSTALL] ||
 			tasks[NR_RUNNING];
@@ -688,35 +690,53 @@ static void psi_group_change(struct psi_group *group, int cpu,
 			     bool wake_clock)
 {
 	struct psi_group_cpu *groupc;
-	u32 state_mask = 0;
 	unsigned int t, m;
 	enum psi_states s;
+	u32 state_mask;
 
 	groupc = per_cpu_ptr(group->pcpu, cpu);
 
 	/*
-	 * First we assess the aggregate resource states this CPU's
-	 * tasks have been in since the last change, and account any
-	 * SOME and FULL time these may have resulted in.
-	 *
-	 * Then we update the task counts according to the state
+	 * First we update the task counts according to the state
 	 * change requested through the @clear and @set bits.
+	 *
+	 * Then if the cgroup PSI stats accounting enabled, we
+	 * assess the aggregate resource states this CPU's tasks
+	 * have been in since the last change, and account any
+	 * SOME and FULL time these may have resulted in.
 	 */
 	write_seqcount_begin(&groupc->seq);
 
-	record_times(groupc, now);
+	/*
+	 * Start with TSK_ONCPU, which doesn't have a corresponding
+	 * task count - it's just a boolean flag directly encoded in
+	 * the state mask. Clear, set, or carry the current state if
+	 * no changes are requested.
+	 */
+	if (unlikely(clear & TSK_ONCPU)) {
+		state_mask = 0;
+		clear &= ~TSK_ONCPU;
+	} else if (unlikely(set & TSK_ONCPU)) {
+		state_mask = PSI_ONCPU;
+		set &= ~TSK_ONCPU;
+	} else {
+		state_mask = groupc->state_mask & PSI_ONCPU;
+	}
 
+	/*
+	 * The rest of the state mask is calculated based on the task
+	 * counts. Update those first, then construct the mask.
+	 */
 	for (t = 0, m = clear; m; m &= ~(1 << t), t++) {
 		if (!(m & (1 << t)))
 			continue;
 		if (groupc->tasks[t]) {
 			groupc->tasks[t]--;
 		} else if (!psi_bug) {
-			printk_deferred(KERN_ERR "psi: task underflow! cpu=%d t=%d tasks=[%u %u %u %u %u] clear=%x set=%x\n",
+			printk_deferred(KERN_ERR "psi: task underflow! cpu=%d t=%d tasks=[%u %u %u %u] clear=%x set=%x\n",
 					cpu, t, groupc->tasks[0],
 					groupc->tasks[1], groupc->tasks[2],
-					groupc->tasks[3], groupc->tasks[4],
-					clear, set);
+					groupc->tasks[3], clear, set);
 			psi_bug = 1;
 		}
 	}
@@ -725,9 +745,25 @@ static void psi_group_change(struct psi_group *group, int cpu,
 		if (set & (1 << t))
 			groupc->tasks[t]++;
 
-	/* Calculate state mask representing active states */
+	if (!group->enabled) {
+		/*
+		 * On the first group change after disabling PSI, conclude
+		 * the current state and flush its time. This is unlikely
+		 * to matter to the user, but aggregation (get_recent_times)
+		 * may have already incorporated the live state into times_prev;
+		 * avoid a delta sample underflow when PSI is later re-enabled.
+		 */
+		if (unlikely(groupc->state_mask & (1 << PSI_NONIDLE)))
+			record_times(groupc, now);
+
+		groupc->state_mask = state_mask;
+
+		write_seqcount_end(&groupc->seq);
+		return;
+	}
+
 	for (s = 0; s < NR_PSI_STATES; s++) {
-		if (test_state(groupc->tasks, s))
+		if (test_state(groupc->tasks, s, state_mask & PSI_ONCPU))
 			state_mask |= (1 << s);
 	}
 
@@ -739,9 +775,11 @@ static void psi_group_change(struct psi_group *group, int cpu,
 	 * task in a cgroup is in_memstall, the corresponding groupc
 	 * on that cpu is in PSI_MEM_FULL state.
 	 */
-	if (unlikely(groupc->tasks[NR_ONCPU] && cpu_curr(cpu)->in_memstall))
+	if (unlikely((state_mask & PSI_ONCPU) && cpu_curr(cpu)->in_memstall))
 		state_mask |= (1 << PSI_MEM_FULL);
 
+	record_times(groupc, now);
+
 	groupc->state_mask = state_mask;
 
 	write_seqcount_end(&groupc->seq);
@@ -753,27 +791,12 @@ static void psi_group_change(struct psi_group *group, int cpu,
 		schedule_delayed_work(&group->avgs_work, PSI_FREQ);
 }
 
-static struct psi_group *iterate_groups(struct task_struct *task, void **iter)
+static inline struct psi_group *task_psi_group(struct task_struct *task)
 {
-	if (*iter == &psi_system)
-		return NULL;
-
 #ifdef CONFIG_CGROUPS
-	if (static_branch_likely(&psi_cgroups_enabled)) {
-		struct cgroup *cgroup = NULL;
-
-		if (!*iter)
-			cgroup = task->cgroups->dfl_cgrp;
-		else
-			cgroup = cgroup_parent(*iter);
-
-		if (cgroup && cgroup_parent(cgroup)) {
-			*iter = cgroup;
-			return cgroup_psi(cgroup);
-		}
-	}
+	if (static_branch_likely(&psi_cgroups_enabled))
+		return cgroup_psi(task_dfl_cgroup(task));
 #endif
-	*iter = &psi_system;
 	return &psi_system;
 }
 
@@ -796,8 +819,6 @@ void psi_task_change(struct task_struct *task, int clear, int set)
 {
 	int cpu = task_cpu(task);
 	struct psi_group *group;
-	bool wake_clock = true;
-	void *iter = NULL;
 	u64 now;
 
 	if (!task->pid)
@@ -806,19 +827,11 @@ void psi_task_change(struct task_struct *task, int clear, int set)
 	psi_flags_change(task, clear, set);
 
 	now = cpu_clock(cpu);
-	/*
-	 * Periodic aggregation shuts off if there is a period of no
-	 * task changes, so we wake it back up if necessary. However,
-	 * don't do this if the task change is the aggregation worker
-	 * itself going to sleep, or we'll ping-pong forever.
-	 */
-	if (unlikely((clear & TSK_RUNNING) &&
-		     (task->flags & PF_WQ_WORKER) &&
-		     wq_worker_last_func(task) == psi_avgs_work))
-		wake_clock = false;
 
-	while ((group = iterate_groups(task, &iter)))
-		psi_group_change(group, cpu, clear, set, now, wake_clock);
+	group = task_psi_group(task);
+	do {
+		psi_group_change(group, cpu, clear, set, now, true);
+	} while ((group = group->parent));
 }
 
 void psi_task_switch(struct task_struct *prev, struct task_struct *next,
@@ -826,34 +839,30 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next,
 {
 	struct psi_group *group, *common = NULL;
 	int cpu = task_cpu(prev);
-	void *iter;
 	u64 now = cpu_clock(cpu);
 
 	if (next->pid) {
-		bool identical_state;
-
 		psi_flags_change(next, 0, TSK_ONCPU);
 		/*
-		 * When switching between tasks that have an identical
-		 * runtime state, the cgroup that contains both tasks
-		 * we reach the first common ancestor. Iterate @next's
-		 * ancestors only until we encounter @prev's ONCPU.
+		 * Set TSK_ONCPU on @next's cgroups. If @next shares any
+		 * ancestors with @prev, those will already have @prev's
+		 * TSK_ONCPU bit set, and we can stop the iteration there.
 		 */
-		identical_state = prev->psi_flags == next->psi_flags;
-		iter = NULL;
-		while ((group = iterate_groups(next, &iter))) {
-			if (identical_state &&
-			    per_cpu_ptr(group->pcpu, cpu)->tasks[NR_ONCPU]) {
+		group = task_psi_group(next);
+		do {
+			if (per_cpu_ptr(group->pcpu, cpu)->state_mask &
+			    PSI_ONCPU) {
 				common = group;
 				break;
 			}
 
 			psi_group_change(group, cpu, 0, TSK_ONCPU, now, true);
-		}
+		} while ((group = group->parent));
 	}
 
 	if (prev->pid) {
 		int clear = TSK_ONCPU, set = 0;
+		bool wake_clock = true;
 
 		/*
 		 * When we're going to sleep, psi_dequeue() lets us
@@ -867,26 +876,74 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next,
 				clear |= TSK_MEMSTALL_RUNNING;
 			if (prev->in_iowait)
 				set |= TSK_IOWAIT;
+
+			/*
+			 * Periodic aggregation shuts off if there is a period of no
+			 * task changes, so we wake it back up if necessary. However,
+			 * don't do this if the task change is the aggregation worker
+			 * itself going to sleep, or we'll ping-pong forever.
+			 */
+			if (unlikely((prev->flags & PF_WQ_WORKER) &&
+				     wq_worker_last_func(prev) == psi_avgs_work))
+				wake_clock = false;
 		}
 
 		psi_flags_change(prev, clear, set);
 
-		iter = NULL;
-		while ((group = iterate_groups(prev, &iter)) && group != common)
-			psi_group_change(group, cpu, clear, set, now, true);
+		group = task_psi_group(prev);
+		do {
+			if (group == common)
+				break;
+			psi_group_change(group, cpu, clear, set, now, wake_clock);
+		} while ((group = group->parent));
 
 		/*
-		 * TSK_ONCPU is handled up to the common ancestor. If we're tasked
-		 * with dequeuing too, finish that for the rest of the hierarchy.
+		 * TSK_ONCPU is handled up to the common ancestor. If there are
+		 * any other differences between the two tasks (e.g. prev goes
+		 * to sleep, or only one task is memstall), finish propagating
+		 * those differences all the way up to the root.
 		 */
-		if (sleep) {
+		if ((prev->psi_flags ^ next->psi_flags) & ~TSK_ONCPU) {
 			clear &= ~TSK_ONCPU;
-			for (; group; group = iterate_groups(prev, &iter))
-				psi_group_change(group, cpu, clear, set, now, true);
+			for (; group; group = group->parent)
+				psi_group_change(group, cpu, clear, set, now, wake_clock);
 		}
 	}
 }
 
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+void psi_account_irqtime(struct task_struct *task, u32 delta)
+{
+	int cpu = task_cpu(task);
+	struct psi_group *group;
+	struct psi_group_cpu *groupc;
+	u64 now;
+
+	if (!task->pid)
+		return;
+
+	now = cpu_clock(cpu);
+
+	group = task_psi_group(task);
+	do {
+		if (!group->enabled)
+			continue;
+
+		groupc = per_cpu_ptr(group->pcpu, cpu);
+
+		write_seqcount_begin(&groupc->seq);
+
+		record_times(groupc, now);
+		groupc->times[PSI_IRQ_FULL] += delta;
+
+		write_seqcount_end(&groupc->seq);
+
+		if (group->poll_states & (1 << PSI_IRQ_FULL))
+			psi_schedule_poll_work(group, 1);
+	} while ((group = group->parent));
+}
+#endif
+
 /**
  * psi_memstall_enter - mark the beginning of a memory stall section
  * @flags: flags to handle nested sections
@@ -952,7 +1009,7 @@ EXPORT_SYMBOL_GPL(psi_memstall_leave);
 #ifdef CONFIG_CGROUPS
 int psi_cgroup_alloc(struct cgroup *cgroup)
 {
-	if (static_branch_likely(&psi_disabled))
+	if (!static_branch_likely(&psi_cgroups_enabled))
 		return 0;
 
 	cgroup->psi = kzalloc(sizeof(struct psi_group), GFP_KERNEL);
@@ -965,12 +1022,13 @@ int psi_cgroup_alloc(struct cgroup *cgroup)
 		return -ENOMEM;
 	}
 	group_init(cgroup->psi);
+	cgroup->psi->parent = cgroup_psi(cgroup_parent(cgroup));
 	return 0;
 }
 
 void psi_cgroup_free(struct cgroup *cgroup)
 {
-	if (static_branch_likely(&psi_disabled))
+	if (!static_branch_likely(&psi_cgroups_enabled))
 		return;
 
 	cancel_delayed_work_sync(&cgroup->psi->avgs_work);
@@ -998,7 +1056,7 @@ void cgroup_move_task(struct task_struct *task, struct css_set *to)
 	struct rq_flags rf;
 	struct rq *rq;
 
-	if (static_branch_likely(&psi_disabled)) {
+	if (!static_branch_likely(&psi_cgroups_enabled)) {
 		/*
 		 * Lame to do this here, but the scheduler cannot be locked
 		 * from the outside, so we move cgroups from inside sched/.
@@ -1046,10 +1104,45 @@ void cgroup_move_task(struct task_struct *task, struct css_set *to)
 
 	task_rq_unlock(rq, task, &rf);
 }
+
+void psi_cgroup_restart(struct psi_group *group)
+{
+	int cpu;
+
+	/*
+	 * After we disable psi_group->enabled, we don't actually
+	 * stop percpu tasks accounting in each psi_group_cpu,
+	 * instead only stop test_state() loop, record_times()
+	 * and averaging worker, see psi_group_change() for details.
+	 *
+	 * When disable cgroup PSI, this function has nothing to sync
+	 * since cgroup pressure files are hidden and percpu psi_group_cpu
+	 * would see !psi_group->enabled and only do task accounting.
+	 *
+	 * When re-enable cgroup PSI, this function use psi_group_change()
+	 * to get correct state mask from test_state() loop on tasks[],
+	 * and restart groupc->state_start from now, use .clear = .set = 0
+	 * here since no task status really changed.
+	 */
+	if (!group->enabled)
+		return;
+
+	for_each_possible_cpu(cpu) {
+		struct rq *rq = cpu_rq(cpu);
+		struct rq_flags rf;
+		u64 now;
+
+		rq_lock_irq(rq, &rf);
+		now = cpu_clock(cpu);
+		psi_group_change(group, cpu, 0, 0, now, true);
+		rq_unlock_irq(rq, &rf);
+	}
+}
 #endif /* CONFIG_CGROUPS */
 
 int psi_show(struct seq_file *m, struct psi_group *group, enum psi_res res)
 {
+	bool only_full = false;
 	int full;
 	u64 now;
 
@@ -1064,7 +1157,11 @@ int psi_show(struct seq_file *m, struct psi_group *group, enum psi_res res)
 		group->avg_next_update = update_averages(group, now);
 	mutex_unlock(&group->avgs_lock);
 
-	for (full = 0; full < 2; full++) {
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+	only_full = res == PSI_IRQ;
+#endif
+
+	for (full = 0; full < 2 - only_full; full++) {
 		unsigned long avg[3] = { 0, };
 		u64 total = 0;
 		int w;
@@ -1078,7 +1175,7 @@ int psi_show(struct seq_file *m, struct psi_group *group, enum psi_res res)
 		}
 
 		seq_printf(m, "%s avg10=%lu.%02lu avg60=%lu.%02lu avg300=%lu.%02lu total=%llu\n",
-			   full ? "full" : "some",
+			   full || only_full ? "full" : "some",
 			   LOAD_INT(avg[0]), LOAD_FRAC(avg[0]),
 			   LOAD_INT(avg[1]), LOAD_FRAC(avg[1]),
 			   LOAD_INT(avg[2]), LOAD_FRAC(avg[2]),
@@ -1106,6 +1203,11 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
 	else
 		return ERR_PTR(-EINVAL);
 
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+	if (res == PSI_IRQ && --state != PSI_IRQ_FULL)
+		return ERR_PTR(-EINVAL);
+#endif
+
 	if (state >= PSI_NONIDLE)
 		return ERR_PTR(-EINVAL);
 
@@ -1390,6 +1492,33 @@ static const struct proc_ops psi_cpu_proc_ops = {
 	.proc_release	= psi_fop_release,
 };
 
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+static int psi_irq_show(struct seq_file *m, void *v)
+{
+	return psi_show(m, &psi_system, PSI_IRQ);
+}
+
+static int psi_irq_open(struct inode *inode, struct file *file)
+{
+	return psi_open(file, psi_irq_show);
+}
+
+static ssize_t psi_irq_write(struct file *file, const char __user *user_buf,
+			     size_t nbytes, loff_t *ppos)
+{
+	return psi_write(file, user_buf, nbytes, PSI_IRQ);
+}
+
+static const struct proc_ops psi_irq_proc_ops = {
+	.proc_open	= psi_irq_open,
+	.proc_read	= seq_read,
+	.proc_lseek	= seq_lseek,
+	.proc_write	= psi_irq_write,
+	.proc_poll	= psi_fop_poll,
+	.proc_release	= psi_fop_release,
+};
+#endif
+
 static int __init psi_proc_init(void)
 {
 	if (psi_enable) {
@@ -1397,6 +1526,9 @@ static int __init psi_proc_init(void)
 		proc_create("pressure/io", 0666, NULL, &psi_io_proc_ops);
 		proc_create("pressure/memory", 0666, NULL, &psi_memory_proc_ops);
 		proc_create("pressure/cpu", 0666, NULL, &psi_cpu_proc_ops);
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+		proc_create("pressure/irq", 0666, NULL, &psi_irq_proc_ops);
+#endif
 	}
 	return 0;
 }
diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
index 1fc198be1ffd..1644242ecd11 100644
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -2446,6 +2446,7 @@ extern unsigned int sysctl_numa_balancing_scan_delay;
 extern unsigned int sysctl_numa_balancing_scan_period_min;
 extern unsigned int sysctl_numa_balancing_scan_period_max;
 extern unsigned int sysctl_numa_balancing_scan_size;
+extern unsigned int sysctl_numa_balancing_hot_threshold;
 #endif
 
 #ifdef CONFIG_SCHED_HRTICK
diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index baa839c1ba96..84a188913cc9 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -107,6 +107,11 @@ __schedstats_from_se(struct sched_entity *se)
 }
 
 #ifdef CONFIG_PSI
+void psi_task_change(struct task_struct *task, int clear, int set);
+void psi_task_switch(struct task_struct *prev, struct task_struct *next,
+		     bool sleep);
+void psi_account_irqtime(struct task_struct *task, u32 delta);
+
 /*
  * PSI tracks state that persists across sleeps, such as iowaits and
  * memory stalls. As a result, it has to distinguish between sleeps,
@@ -201,6 +206,7 @@ static inline void psi_ttwu_dequeue(struct task_struct *p) {}
 static inline void psi_sched_switch(struct task_struct *prev,
 				    struct task_struct *next,
 				    bool sleep) {}
+static inline void psi_account_irqtime(struct task_struct *task, u32 delta) {}
 #endif /* CONFIG_PSI */
 
 #ifdef CONFIG_SCHED_INFO