diff --git a/arch/ia64/include/asm/acpi.h b/arch/ia64/include/asm/acpi.h
index b66ba907019c..87927eb824cc 100644
--- a/arch/ia64/include/asm/acpi.h
+++ b/arch/ia64/include/asm/acpi.h
@@ -74,8 +74,6 @@ static inline void arch_acpi_set_pdc_bits(u32 *buf)
 	buf[2] |= ACPI_PDC_EST_CAPABILITY_SMP;
 }
 
-#define acpi_unlazy_tlb(x)
-
 #ifdef CONFIG_ACPI_NUMA
 extern cpumask_t early_cpu_possible_map;
 #define for_each_possible_early_cpu(cpu)  \
diff --git a/arch/x86/include/asm/acpi.h b/arch/x86/include/asm/acpi.h
index ca0976456a6b..6d2df1ee427b 100644
--- a/arch/x86/include/asm/acpi.h
+++ b/arch/x86/include/asm/acpi.h
@@ -159,8 +159,6 @@ static inline u64 x86_default_get_root_pointer(void)
 extern int x86_acpi_numa_init(void);
 #endif /* CONFIG_ACPI_NUMA */
 
-#define acpi_unlazy_tlb(x)	leave_mm(x)
-
 #ifdef CONFIG_ACPI_APEI
 static inline pgprot_t arch_apei_get_mem_attribute(phys_addr_t addr)
 {
diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
index 71a30b0d0f05..7ecb90e90afd 100644
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -161,18 +161,10 @@ static void lapic_timer_propagate_broadcast(struct acpi_processor *pr)
 }
 
 /* Power(C) State timer broadcast control */
-static void lapic_timer_state_broadcast(struct acpi_processor *pr,
-				       struct acpi_processor_cx *cx,
-				       int broadcast)
+static bool lapic_timer_needs_broadcast(struct acpi_processor *pr,
+					struct acpi_processor_cx *cx)
 {
-	int state = cx - pr->power.states;
-
-	if (state >= pr->power.timer_broadcast_on_state) {
-		if (broadcast)
-			tick_broadcast_enter();
-		else
-			tick_broadcast_exit();
-	}
+	return cx - pr->power.states >= pr->power.timer_broadcast_on_state;
 }
 
 #else
@@ -180,9 +172,9 @@ static void lapic_timer_state_broadcast(struct acpi_processor *pr,
 static void lapic_timer_check_state(int state, struct acpi_processor *pr,
 				   struct acpi_processor_cx *cstate) { }
 static void lapic_timer_propagate_broadcast(struct acpi_processor *pr) { }
-static void lapic_timer_state_broadcast(struct acpi_processor *pr,
-				       struct acpi_processor_cx *cx,
-				       int broadcast)
+
+static bool lapic_timer_needs_broadcast(struct acpi_processor *pr,
+					struct acpi_processor_cx *cx)
 {
 }
 
@@ -566,32 +558,43 @@ static DEFINE_RAW_SPINLOCK(c3_lock);
 
 /**
  * acpi_idle_enter_bm - enters C3 with proper BM handling
+ * @drv: cpuidle driver
  * @pr: Target processor
  * @cx: Target state context
- * @timer_bc: Whether or not to change timer mode to broadcast
+ * @index: index of target state
  */
-static void acpi_idle_enter_bm(struct acpi_processor *pr,
-			       struct acpi_processor_cx *cx, bool timer_bc)
+static int acpi_idle_enter_bm(struct cpuidle_driver *drv,
+			       struct acpi_processor *pr,
+			       struct acpi_processor_cx *cx,
+			       int index)
 {
-	acpi_unlazy_tlb(smp_processor_id());
-
-	/*
-	 * Must be done before busmaster disable as we might need to
-	 * access HPET !
-	 */
-	if (timer_bc)
-		lapic_timer_state_broadcast(pr, cx, 1);
+	static struct acpi_processor_cx safe_cx = {
+		.entry_method = ACPI_CSTATE_HALT,
+	};
 
 	/*
 	 * disable bus master
 	 * bm_check implies we need ARB_DIS
 	 * bm_control implies whether we can do ARB_DIS
 	 *
-	 * That leaves a case where bm_check is set and bm_control is
-	 * not set. In that case we cannot do much, we enter C3
-	 * without doing anything.
+	 * That leaves a case where bm_check is set and bm_control is not set.
+	 * In that case we cannot do much, we enter C3 without doing anything.
 	 */
-	if (pr->flags.bm_control) {
+	bool dis_bm = pr->flags.bm_control;
+
+	/* If we can skip BM, demote to a safe state. */
+	if (!cx->bm_sts_skip && acpi_idle_bm_check()) {
+		dis_bm = false;
+		index = drv->safe_state_index;
+		if (index >= 0) {
+			cx = this_cpu_read(acpi_cstate[index]);
+		} else {
+			cx = &safe_cx;
+			index = -EBUSY;
+		}
+	}
+
+	if (dis_bm) {
 		raw_spin_lock(&c3_lock);
 		c3_cpu_count++;
 		/* Disable bus master arbitration when all CPUs are in C3 */
@@ -600,18 +603,21 @@ static void acpi_idle_enter_bm(struct acpi_processor *pr,
 		raw_spin_unlock(&c3_lock);
 	}
 
+	rcu_idle_enter();
+
 	acpi_idle_do_entry(cx);
 
+	rcu_idle_exit();
+
 	/* Re-enable bus master arbitration */
-	if (pr->flags.bm_control) {
+	if (dis_bm) {
 		raw_spin_lock(&c3_lock);
 		acpi_write_bit_register(ACPI_BITREG_ARB_DISABLE, 0);
 		c3_cpu_count--;
 		raw_spin_unlock(&c3_lock);
 	}
 
-	if (timer_bc)
-		lapic_timer_state_broadcast(pr, cx, 0);
+	return index;
 }
 
 static int acpi_idle_enter(struct cpuidle_device *dev,
@@ -625,32 +631,21 @@ static int acpi_idle_enter(struct cpuidle_device *dev,
 		return -EINVAL;
 
 	if (cx->type != ACPI_STATE_C1) {
+		if (cx->type == ACPI_STATE_C3 && pr->flags.bm_check)
+			return acpi_idle_enter_bm(drv, pr, cx, index);
+
+		/* C2 to C1 demotion. */
 		if (acpi_idle_fallback_to_c1(pr) && num_online_cpus() > 1) {
 			index = ACPI_IDLE_STATE_START;
 			cx = per_cpu(acpi_cstate[index], dev->cpu);
-		} else if (cx->type == ACPI_STATE_C3 && pr->flags.bm_check) {
-			if (cx->bm_sts_skip || !acpi_idle_bm_check()) {
-				acpi_idle_enter_bm(pr, cx, true);
-				return index;
-			} else if (drv->safe_state_index >= 0) {
-				index = drv->safe_state_index;
-				cx = per_cpu(acpi_cstate[index], dev->cpu);
-			} else {
-				acpi_safe_halt();
-				return -EBUSY;
-			}
 		}
 	}
 
-	lapic_timer_state_broadcast(pr, cx, 1);
-
 	if (cx->type == ACPI_STATE_C3)
 		ACPI_FLUSH_CPU_CACHE();
 
 	acpi_idle_do_entry(cx);
 
-	lapic_timer_state_broadcast(pr, cx, 0);
-
 	return index;
 }
 
@@ -666,7 +661,13 @@ static int acpi_idle_enter_s2idle(struct cpuidle_device *dev,
 			return 0;
 
 		if (pr->flags.bm_check) {
-			acpi_idle_enter_bm(pr, cx, false);
+			u8 bm_sts_skip = cx->bm_sts_skip;
+
+			/* Don't check BM_STS, do an unconditional ARB_DIS for S2IDLE */
+			cx->bm_sts_skip = 1;
+			acpi_idle_enter_bm(drv, pr, cx, index);
+			cx->bm_sts_skip = bm_sts_skip;
+
 			return 0;
 		} else {
 			ACPI_FLUSH_CPU_CACHE();
@@ -682,11 +683,13 @@ static int acpi_processor_setup_cpuidle_cx(struct acpi_processor *pr,
 {
 	int i, count = ACPI_IDLE_STATE_START;
 	struct acpi_processor_cx *cx;
+	struct cpuidle_state *state;
 
 	if (max_cstate == 0)
 		max_cstate = 1;
 
 	for (i = 1; i < ACPI_PROCESSOR_MAX_POWER && i <= max_cstate; i++) {
+		state = &acpi_idle_driver.states[count];
 		cx = &pr->power.states[i];
 
 		if (!cx->valid)
@@ -694,6 +697,15 @@ static int acpi_processor_setup_cpuidle_cx(struct acpi_processor *pr,
 
 		per_cpu(acpi_cstate[count], dev->cpu) = cx;
 
+		if (lapic_timer_needs_broadcast(pr, cx))
+			state->flags |= CPUIDLE_FLAG_TIMER_STOP;
+
+		if (cx->type == ACPI_STATE_C3) {
+			state->flags |= CPUIDLE_FLAG_TLB_FLUSHED;
+			if (pr->flags.bm_check)
+				state->flags |= CPUIDLE_FLAG_RCU_IDLE;
+		}
+
 		count++;
 		if (count == CPUIDLE_STATE_MAX)
 			break;
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
index 04becd70cc41..6c7e5621cf9a 100644
--- a/drivers/cpuidle/cpuidle.c
+++ b/drivers/cpuidle/cpuidle.c
@@ -138,6 +138,7 @@ static void enter_s2idle_proper(struct cpuidle_driver *drv,
 				struct cpuidle_device *dev, int index)
 {
 	ktime_t time_start, time_end;
+	struct cpuidle_state *target_state = &drv->states[index];
 
 	time_start = ns_to_ktime(local_clock());
 
@@ -153,8 +154,9 @@ static void enter_s2idle_proper(struct cpuidle_driver *drv,
 	 * suspended is generally unsafe.
 	 */
 	stop_critical_timings();
-	rcu_idle_enter();
-	drv->states[index].enter_s2idle(dev, drv, index);
+	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
+		rcu_idle_enter();
+	target_state->enter_s2idle(dev, drv, index);
 	if (WARN_ON_ONCE(!irqs_disabled()))
 		local_irq_disable();
 	/*
@@ -162,7 +164,8 @@ static void enter_s2idle_proper(struct cpuidle_driver *drv,
 	 * first CPU executing it calls functions containing RCU read-side
 	 * critical sections, so tell RCU about that.
 	 */
-	rcu_idle_exit();
+	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
+		rcu_idle_exit();
 	tick_unfreeze();
 	start_critical_timings();
 
@@ -239,9 +242,11 @@ int cpuidle_enter_state(struct cpuidle_device *dev, struct cpuidle_driver *drv,
 	time_start = ns_to_ktime(local_clock());
 
 	stop_critical_timings();
-	rcu_idle_enter();
+	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
+		rcu_idle_enter();
 	entered_state = target_state->enter(dev, drv, index);
-	rcu_idle_exit();
+	if (!(target_state->flags & CPUIDLE_FLAG_RCU_IDLE))
+		rcu_idle_exit();
 	start_critical_timings();
 
 	sched_clock_idle_wakeup_event();
diff --git a/include/linux/cpuidle.h b/include/linux/cpuidle.h
index 75895e6363b8..6175c77bf25e 100644
--- a/include/linux/cpuidle.h
+++ b/include/linux/cpuidle.h
@@ -82,6 +82,7 @@ struct cpuidle_state {
 #define CPUIDLE_FLAG_UNUSABLE		BIT(3) /* avoid using this state */
 #define CPUIDLE_FLAG_OFF		BIT(4) /* disable this state by default */
 #define CPUIDLE_FLAG_TLB_FLUSHED	BIT(5) /* idle-state flushes TLBs */
+#define CPUIDLE_FLAG_RCU_IDLE		BIT(6) /* idle-state takes care of RCU */
 
 struct cpuidle_device_kobj;
 struct cpuidle_state_kobj;