cpuidle: fix HP nx6125 regression

Fix for http://bugzilla.kernel.org/show_bug.cgi?id=9355

cpuidle always used to fallback to C2 if there is some bm activity while
entering C3. But, presence of C2 is not always guaranteed. Change cpuidle
algorithm to detect a safe_state to fallback in case of bm_activity and
use that state instead of C2.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

drivers/acpi/processor_idle.c

@@ -197,6 +197,19 @@ static inline u32 ticks_elapsed_in_us(u32 t1, u32 t2)
 		return PM_TIMER_TICKS_TO_US((0xFFFFFFFF - t1) + t2);
 }
 
+static void acpi_safe_halt(void)
+{
+	current_thread_info()->status &= ~TS_POLLING;
+	/*
+	 * TS_POLLING-cleared state must be visible before we
+	 * test NEED_RESCHED:
+	 */
+	smp_mb();
+	if (!need_resched())
+		safe_halt();
+	current_thread_info()->status |= TS_POLLING;
+}
+
 #ifndef CONFIG_CPU_IDLE
 
 static void
@@ -239,19 +252,6 @@ acpi_processor_power_activate(struct acpi_processor *pr,
 	return;
 }
 
-static void acpi_safe_halt(void)
-{
-	current_thread_info()->status &= ~TS_POLLING;
-	/*
-	 * TS_POLLING-cleared state must be visible before we
-	 * test NEED_RESCHED:
-	 */
-	smp_mb();
-	if (!need_resched())
-		safe_halt();
-	current_thread_info()->status |= TS_POLLING;
-}
-
 static atomic_t c3_cpu_count;
 
 /* Common C-state entry for C2, C3, .. */
@@ -1385,15 +1385,7 @@ static int acpi_idle_enter_c1(struct cpuidle_device *dev,
 	if (pr->flags.bm_check)
 		acpi_idle_update_bm_rld(pr, cx);
 
-	current_thread_info()->status &= ~TS_POLLING;
+	acpi_safe_halt();
-	/*
-	 * TS_POLLING-cleared state must be visible before we test
-	 * NEED_RESCHED:
-	 */
-	smp_mb();
-	if (!need_resched())
-		safe_halt();
-	current_thread_info()->status |= TS_POLLING;
 
 	cx->usage++;
 
@@ -1493,6 +1485,15 @@ static int acpi_idle_enter_bm(struct cpuidle_device *dev,
 	if (acpi_idle_suspend)
 		return(acpi_idle_enter_c1(dev, state));
 
+	if (acpi_idle_bm_check()) {
+		if (dev->safe_state) {
+			return dev->safe_state->enter(dev, dev->safe_state);
+		} else {
+			acpi_safe_halt();
+			return 0;
+		}
+	}
+
 	local_irq_disable();
 	current_thread_info()->status &= ~TS_POLLING;
 	/*
@@ -1515,49 +1516,39 @@ static int acpi_idle_enter_bm(struct cpuidle_device *dev,
 	 */
 	acpi_state_timer_broadcast(pr, cx, 1);
 
-	if (acpi_idle_bm_check()) {
+	acpi_idle_update_bm_rld(pr, cx);
-		cx = pr->power.bm_state;
 
-		acpi_idle_update_bm_rld(pr, cx);
+	/*
+	 * disable bus master
+	 * bm_check implies we need ARB_DIS
+	 * !bm_check implies we need cache flush
+	 * 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.
+	 */
+	if (pr->flags.bm_check && pr->flags.bm_control) {
+		spin_lock(&c3_lock);
+		c3_cpu_count++;
+		/* Disable bus master arbitration when all CPUs are in C3 */
+		if (c3_cpu_count == num_online_cpus())
+			acpi_set_register(ACPI_BITREG_ARB_DISABLE, 1);
+		spin_unlock(&c3_lock);
+	} else if (!pr->flags.bm_check) {
+		ACPI_FLUSH_CPU_CACHE();
+	}
 
-		t1 = inl(acpi_gbl_FADT.xpm_timer_block.address);
+	t1 = inl(acpi_gbl_FADT.xpm_timer_block.address);
-		acpi_idle_do_entry(cx);
+	acpi_idle_do_entry(cx);
-		t2 = inl(acpi_gbl_FADT.xpm_timer_block.address);
+	t2 = inl(acpi_gbl_FADT.xpm_timer_block.address);
-	} else {
-		acpi_idle_update_bm_rld(pr, cx);
 
-		/*
+	/* Re-enable bus master arbitration */
-		 * disable bus master
+	if (pr->flags.bm_check && pr->flags.bm_control) {
-		 * bm_check implies we need ARB_DIS
+		spin_lock(&c3_lock);
-		 * !bm_check implies we need cache flush
+		acpi_set_register(ACPI_BITREG_ARB_DISABLE, 0);
-		 * bm_control implies whether we can do ARB_DIS
+		c3_cpu_count--;
-		 *
+		spin_unlock(&c3_lock);
-		 * 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_check && pr->flags.bm_control) {
-			spin_lock(&c3_lock);
-			c3_cpu_count++;
-			/* Disable bus master arbitration when all CPUs are in C3 */
-			if (c3_cpu_count == num_online_cpus())
-				acpi_set_register(ACPI_BITREG_ARB_DISABLE, 1);
-			spin_unlock(&c3_lock);
-		} else if (!pr->flags.bm_check) {
-			ACPI_FLUSH_CPU_CACHE();
-		}
-
-		t1 = inl(acpi_gbl_FADT.xpm_timer_block.address);
-		acpi_idle_do_entry(cx);
-		t2 = inl(acpi_gbl_FADT.xpm_timer_block.address);
-
-		/* Re-enable bus master arbitration */
-		if (pr->flags.bm_check && pr->flags.bm_control) {
-			spin_lock(&c3_lock);
-			acpi_set_register(ACPI_BITREG_ARB_DISABLE, 0);
-			c3_cpu_count--;
-			spin_unlock(&c3_lock);
-		}
 	}
 
 #if defined (CONFIG_GENERIC_TIME) && defined (CONFIG_X86_TSC)
@@ -1626,12 +1617,14 @@ static int acpi_processor_setup_cpuidle(struct acpi_processor *pr)
 			case ACPI_STATE_C1:
 			state->flags |= CPUIDLE_FLAG_SHALLOW;
 			state->enter = acpi_idle_enter_c1;
+			dev->safe_state = state;
 			break;
 
 			case ACPI_STATE_C2:
 			state->flags |= CPUIDLE_FLAG_BALANCED;
 			state->flags |= CPUIDLE_FLAG_TIME_VALID;
 			state->enter = acpi_idle_enter_simple;
+			dev->safe_state = state;
 			break;
 
 			case ACPI_STATE_C3:
@@ -1652,14 +1645,6 @@ static int acpi_processor_setup_cpuidle(struct acpi_processor *pr)
 	if (!count)
 		return -EINVAL;
 
-	/* find the deepest state that can handle active BM */
-	if (pr->flags.bm_check) {
-		for (i = 1; i < ACPI_PROCESSOR_MAX_POWER && i <= max_cstate; i++)
-			if (pr->power.states[i].type == ACPI_STATE_C3)
-				break;
-		pr->power.bm_state = &pr->power.states[i-1];
-	}
-
 	return 0;
 }
 

include/acpi/processor.h

@@ -78,7 +78,6 @@ struct acpi_processor_cx {
 struct acpi_processor_power {
 	struct cpuidle_device dev;
 	struct acpi_processor_cx *state;
-	struct acpi_processor_cx *bm_state;
 	unsigned long bm_check_timestamp;
 	u32 default_state;
 	u32 bm_activity;

include/linux/cpuidle.h

@@ -92,6 +92,7 @@ struct cpuidle_device {
 	struct kobject		kobj;
 	struct completion	kobj_unregister;
 	void			*governor_data;
+	struct cpuidle_state	*safe_state;
 };
 
 DECLARE_PER_CPU(struct cpuidle_device *, cpuidle_devices);