Merge branch 'pm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm

* 'pm-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (63 commits)
  PM / Clocks: Remove redundant NULL checks before kfree()
  PM / Documentation: Update docs about suspend and CPU hotplug
  ACPI / PM: Add Sony VGN-FW21E to nonvs blacklist.
  ARM: mach-shmobile: sh7372 A4R support (v4)
  ARM: mach-shmobile: sh7372 A3SP support (v4)
  PM / Sleep: Mark devices involved in wakeup signaling during suspend
  PM / Hibernate: Improve performance of LZO/plain hibernation, checksum image
  PM / Hibernate: Do not initialize static and extern variables to 0
  PM / Freezer: Make fake_signal_wake_up() wake TASK_KILLABLE tasks too
  PM / Hibernate: Add resumedelay kernel param in addition to resumewait
  MAINTAINERS: Update linux-pm list address
  PM / ACPI: Blacklist Vaio VGN-FW520F machine known to require acpi_sleep=nonvs
  PM / ACPI: Blacklist Sony Vaio known to require acpi_sleep=nonvs
  PM / Hibernate: Add resumewait param to support MMC-like devices as resume file
  PM / Hibernate: Fix typo in a kerneldoc comment
  PM / Hibernate: Freeze kernel threads after preallocating memory
  PM: Update the policy on default wakeup settings
  PM / VT: Cleanup #if defined uglyness and fix compile error
  PM / Suspend: Off by one in pm_suspend()
  PM / Hibernate: Include storage keys in hibernation image on s390
  ...

10 files changed, 931 insertions, 270 deletions

diff --git a/drivers/base/power/Makefile b/drivers/base/power/Makefile
index 2639ae79a37..81676dd1790 100644
--- a/drivers/base/power/Makefile
+++ b/drivers/base/power/Makefile
@@ -1,4 +1,4 @@
-obj-$(CONFIG_PM)	+= sysfs.o generic_ops.o
+obj-$(CONFIG_PM)	+= sysfs.o generic_ops.o common.o qos.o
 obj-$(CONFIG_PM_SLEEP)	+= main.o wakeup.o
 obj-$(CONFIG_PM_RUNTIME)	+= runtime.o
 obj-$(CONFIG_PM_TRACE_RTC)	+= trace.o
@@ -6,4 +6,4 @@ obj-$(CONFIG_PM_OPP)	+= opp.o
 obj-$(CONFIG_PM_GENERIC_DOMAINS)	+=  domain.o
 obj-$(CONFIG_HAVE_CLK)	+= clock_ops.o
 
-ccflags-$(CONFIG_DEBUG_DRIVER) := -DDEBUG
\ No newline at end of file
+ccflags-$(CONFIG_DEBUG_DRIVER) := -DDEBUG
diff --git a/drivers/base/power/clock_ops.c b/drivers/base/power/clock_ops.c
index b97294e2d95..5f0f85d5c57 100644
--- a/drivers/base/power/clock_ops.c
+++ b/drivers/base/power/clock_ops.c
@@ -10,18 +10,13 @@
 #include <linux/kernel.h>
 #include <linux/io.h>
 #include <linux/pm.h>
-#include <linux/pm_runtime.h>
+#include <linux/pm_clock.h>
 #include <linux/clk.h>
 #include <linux/slab.h>
 #include <linux/err.h>
 
 #ifdef CONFIG_PM
 
-struct pm_clk_data {
-	struct list_head clock_list;
-	spinlock_t lock;
-};
-
 enum pce_status {
 	PCE_STATUS_NONE = 0,
 	PCE_STATUS_ACQUIRED,
@@ -36,11 +31,6 @@ struct pm_clock_entry {
 	enum pce_status status;
 };
 
-static struct pm_clk_data *__to_pcd(struct device *dev)
-{
-	return dev ? dev->power.subsys_data : NULL;
-}
-
 /**
  * pm_clk_acquire - Acquire a device clock.
  * @dev: Device whose clock is to be acquired.
@@ -67,10 +57,10 @@ static void pm_clk_acquire(struct device *dev, struct pm_clock_entry *ce)
  */
 int pm_clk_add(struct device *dev, const char *con_id)
 {
-	struct pm_clk_data *pcd = __to_pcd(dev);
+	struct pm_subsys_data *psd = dev_to_psd(dev);
 	struct pm_clock_entry *ce;
 
-	if (!pcd)
+	if (!psd)
 		return -EINVAL;
 
 	ce = kzalloc(sizeof(*ce), GFP_KERNEL);
@@ -91,9 +81,9 @@ int pm_clk_add(struct device *dev, const char *con_id)
 
 	pm_clk_acquire(dev, ce);
 
-	spin_lock_irq(&pcd->lock);
-	list_add_tail(&ce->node, &pcd->clock_list);
-	spin_unlock_irq(&pcd->lock);
+	spin_lock_irq(&psd->lock);
+	list_add_tail(&ce->node, &psd->clock_list);
+	spin_unlock_irq(&psd->lock);
 	return 0;
 }
 
@@ -114,9 +104,7 @@ static void __pm_clk_remove(struct pm_clock_entry *ce)
 			clk_put(ce->clk);
 	}
 
-	if (ce->con_id)
-		kfree(ce->con_id);
-
+	kfree(ce->con_id);
 	kfree(ce);
 }
 
@@ -130,15 +118,15 @@ static void __pm_clk_remove(struct pm_clock_entry *ce)
  */
 void pm_clk_remove(struct device *dev, const char *con_id)
 {
-	struct pm_clk_data *pcd = __to_pcd(dev);
+	struct pm_subsys_data *psd = dev_to_psd(dev);
 	struct pm_clock_entry *ce;
 
-	if (!pcd)
+	if (!psd)
 		return;
 
-	spin_lock_irq(&pcd->lock);
+	spin_lock_irq(&psd->lock);
 
-	list_for_each_entry(ce, &pcd->clock_list, node) {
+	list_for_each_entry(ce, &psd->clock_list, node) {
 		if (!con_id && !ce->con_id)
 			goto remove;
 		else if (!con_id || !ce->con_id)
@@ -147,12 +135,12 @@ void pm_clk_remove(struct device *dev, const char *con_id)
 			goto remove;
 	}
 
-	spin_unlock_irq(&pcd->lock);
+	spin_unlock_irq(&psd->lock);
 	return;
 
  remove:
 	list_del(&ce->node);
-	spin_unlock_irq(&pcd->lock);
+	spin_unlock_irq(&psd->lock);
 
 	__pm_clk_remove(ce);
 }
@@ -161,23 +149,27 @@ void pm_clk_remove(struct device *dev, const char *con_id)
  * pm_clk_init - Initialize a device's list of power management clocks.
  * @dev: Device to initialize the list of PM clocks for.
  *
- * Allocate a struct pm_clk_data object, initialize its lock member and
- * make the @dev's power.subsys_data field point to it.
+ * Initialize the lock and clock_list members of the device's pm_subsys_data
+ * object.
  */
-int pm_clk_init(struct device *dev)
+void pm_clk_init(struct device *dev)
 {
-	struct pm_clk_data *pcd;
-
-	pcd = kzalloc(sizeof(*pcd), GFP_KERNEL);
-	if (!pcd) {
-		dev_err(dev, "Not enough memory for PM clock data.\n");
-		return -ENOMEM;
-	}
+	struct pm_subsys_data *psd = dev_to_psd(dev);
+	if (psd)
+		INIT_LIST_HEAD(&psd->clock_list);
+}
 
-	INIT_LIST_HEAD(&pcd->clock_list);
-	spin_lock_init(&pcd->lock);
-	dev->power.subsys_data = pcd;
-	return 0;
+/**
+ * pm_clk_create - Create and initialize a device's list of PM clocks.
+ * @dev: Device to create and initialize the list of PM clocks for.
+ *
+ * Allocate a struct pm_subsys_data object, initialize its lock and clock_list
+ * members and make the @dev's power.subsys_data field point to it.
+ */
+int pm_clk_create(struct device *dev)
+{
+	int ret = dev_pm_get_subsys_data(dev);
+	return ret < 0 ? ret : 0;
 }
 
 /**
@@ -185,29 +177,28 @@ int pm_clk_init(struct device *dev)
  * @dev: Device to destroy the list of PM clocks for.
  *
  * Clear the @dev's power.subsys_data field, remove the list of clock entries
- * from the struct pm_clk_data object pointed to by it before and free
+ * from the struct pm_subsys_data object pointed to by it before and free
  * that object.
  */
 void pm_clk_destroy(struct device *dev)
 {
-	struct pm_clk_data *pcd = __to_pcd(dev);
+	struct pm_subsys_data *psd = dev_to_psd(dev);
 	struct pm_clock_entry *ce, *c;
 	struct list_head list;
 
-	if (!pcd)
+	if (!psd)
 		return;
 
-	dev->power.subsys_data = NULL;
 	INIT_LIST_HEAD(&list);
 
-	spin_lock_irq(&pcd->lock);
+	spin_lock_irq(&psd->lock);
 
-	list_for_each_entry_safe_reverse(ce, c, &pcd->clock_list, node)
+	list_for_each_entry_safe_reverse(ce, c, &psd->clock_list, node)
 		list_move(&ce->node, &list);
 
-	spin_unlock_irq(&pcd->lock);
+	spin_unlock_irq(&psd->lock);
 
-	kfree(pcd);
+	dev_pm_put_subsys_data(dev);
 
 	list_for_each_entry_safe_reverse(ce, c, &list, node) {
 		list_del(&ce->node);
@@ -225,25 +216,25 @@ void pm_clk_destroy(struct device *dev)
  */
 int pm_clk_suspend(struct device *dev)
 {
-	struct pm_clk_data *pcd = __to_pcd(dev);
+	struct pm_subsys_data *psd = dev_to_psd(dev);
 	struct pm_clock_entry *ce;
 	unsigned long flags;
 
 	dev_dbg(dev, "%s()\n", __func__);
 
-	if (!pcd)
+	if (!psd)
 		return 0;
 
-	spin_lock_irqsave(&pcd->lock, flags);
+	spin_lock_irqsave(&psd->lock, flags);
 
-	list_for_each_entry_reverse(ce, &pcd->clock_list, node) {
+	list_for_each_entry_reverse(ce, &psd->clock_list, node) {
 		if (ce->status < PCE_STATUS_ERROR) {
 			clk_disable(ce->clk);
 			ce->status = PCE_STATUS_ACQUIRED;
 		}
 	}
 
-	spin_unlock_irqrestore(&pcd->lock, flags);
+	spin_unlock_irqrestore(&psd->lock, flags);
 
 	return 0;
 }
@@ -254,25 +245,25 @@ int pm_clk_suspend(struct device *dev)
  */
 int pm_clk_resume(struct device *dev)
 {
-	struct pm_clk_data *pcd = __to_pcd(dev);
+	struct pm_subsys_data *psd = dev_to_psd(dev);
 	struct pm_clock_entry *ce;
 	unsigned long flags;
 
 	dev_dbg(dev, "%s()\n", __func__);
 
-	if (!pcd)
+	if (!psd)
 		return 0;
 
-	spin_lock_irqsave(&pcd->lock, flags);
+	spin_lock_irqsave(&psd->lock, flags);
 
-	list_for_each_entry(ce, &pcd->clock_list, node) {
+	list_for_each_entry(ce, &psd->clock_list, node) {
 		if (ce->status < PCE_STATUS_ERROR) {
 			clk_enable(ce->clk);
 			ce->status = PCE_STATUS_ENABLED;
 		}
 	}
 
-	spin_unlock_irqrestore(&pcd->lock, flags);
+	spin_unlock_irqrestore(&psd->lock, flags);
 
 	return 0;
 }
@@ -310,7 +301,7 @@ static int pm_clk_notify(struct notifier_block *nb,
 		if (dev->pm_domain)
 			break;
 
-		error = pm_clk_init(dev);
+		error = pm_clk_create(dev);
 		if (error)
 			break;
 
@@ -345,22 +336,22 @@ static int pm_clk_notify(struct notifier_block *nb,
  */
 int pm_clk_suspend(struct device *dev)
 {
-	struct pm_clk_data *pcd = __to_pcd(dev);
+	struct pm_subsys_data *psd = dev_to_psd(dev);
 	struct pm_clock_entry *ce;
 	unsigned long flags;
 
 	dev_dbg(dev, "%s()\n", __func__);
 
 	/* If there is no driver, the clocks are already disabled. */
-	if (!pcd || !dev->driver)
+	if (!psd || !dev->driver)
 		return 0;
 
-	spin_lock_irqsave(&pcd->lock, flags);
+	spin_lock_irqsave(&psd->lock, flags);
 
-	list_for_each_entry_reverse(ce, &pcd->clock_list, node)
+	list_for_each_entry_reverse(ce, &psd->clock_list, node)
 		clk_disable(ce->clk);
 
-	spin_unlock_irqrestore(&pcd->lock, flags);
+	spin_unlock_irqrestore(&psd->lock, flags);
 
 	return 0;
 }
@@ -371,22 +362,22 @@ int pm_clk_suspend(struct device *dev)
  */
 int pm_clk_resume(struct device *dev)
 {
-	struct pm_clk_data *pcd = __to_pcd(dev);
+	struct pm_subsys_data *psd = dev_to_psd(dev);
 	struct pm_clock_entry *ce;
 	unsigned long flags;
 
 	dev_dbg(dev, "%s()\n", __func__);
 
 	/* If there is no driver, the clocks should remain disabled. */
-	if (!pcd || !dev->driver)
+	if (!psd || !dev->driver)
 		return 0;
 
-	spin_lock_irqsave(&pcd->lock, flags);
+	spin_lock_irqsave(&psd->lock, flags);
 
-	list_for_each_entry(ce, &pcd->clock_list, node)
+	list_for_each_entry(ce, &psd->clock_list, node)
 		clk_enable(ce->clk);
 
-	spin_unlock_irqrestore(&pcd->lock, flags);
+	spin_unlock_irqrestore(&psd->lock, flags);
 
 	return 0;
 }
diff --git a/drivers/base/power/common.c b/drivers/base/power/common.c
new file mode 100644
index 00000000000..29820c39618
--- /dev/null
+++ b/drivers/base/power/common.c
@@ -0,0 +1,86 @@
+/*
+ * drivers/base/power/common.c - Common device power management code.
+ *
+ * Copyright (C) 2011 Rafael J. Wysocki <rjw@sisk.pl>, Renesas Electronics Corp.
+ *
+ * This file is released under the GPLv2.
+ */
+
+#include <linux/init.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/pm_clock.h>
+
+/**
+ * dev_pm_get_subsys_data - Create or refcount power.subsys_data for device.
+ * @dev: Device to handle.
+ *
+ * If power.subsys_data is NULL, point it to a new object, otherwise increment
+ * its reference counter.  Return 1 if a new object has been created, otherwise
+ * return 0 or error code.
+ */
+int dev_pm_get_subsys_data(struct device *dev)
+{
+	struct pm_subsys_data *psd;
+	int ret = 0;
+
+	psd = kzalloc(sizeof(*psd), GFP_KERNEL);
+	if (!psd)
+		return -ENOMEM;
+
+	spin_lock_irq(&dev->power.lock);
+
+	if (dev->power.subsys_data) {
+		dev->power.subsys_data->refcount++;
+	} else {
+		spin_lock_init(&psd->lock);
+		psd->refcount = 1;
+		dev->power.subsys_data = psd;
+		pm_clk_init(dev);
+		psd = NULL;
+		ret = 1;
+	}
+
+	spin_unlock_irq(&dev->power.lock);
+
+	/* kfree() verifies that its argument is nonzero. */
+	kfree(psd);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dev_pm_get_subsys_data);
+
+/**
+ * dev_pm_put_subsys_data - Drop reference to power.subsys_data.
+ * @dev: Device to handle.
+ *
+ * If the reference counter of power.subsys_data is zero after dropping the
+ * reference, power.subsys_data is removed.  Return 1 if that happens or 0
+ * otherwise.
+ */
+int dev_pm_put_subsys_data(struct device *dev)
+{
+	struct pm_subsys_data *psd;
+	int ret = 0;
+
+	spin_lock_irq(&dev->power.lock);
+
+	psd = dev_to_psd(dev);
+	if (!psd) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if (--psd->refcount == 0) {
+		dev->power.subsys_data = NULL;
+		kfree(psd);
+		ret = 1;
+	}
+
+ out:
+	spin_unlock_irq(&dev->power.lock);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dev_pm_put_subsys_data);
diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
index 1c374579407..6790cf7eba5 100644
--- a/drivers/base/power/domain.c
+++ b/drivers/base/power/domain.c
@@ -29,10 +29,20 @@ static struct generic_pm_domain *dev_to_genpd(struct device *dev)
 	return pd_to_genpd(dev->pm_domain);
 }
 
-static void genpd_sd_counter_dec(struct generic_pm_domain *genpd)
+static bool genpd_sd_counter_dec(struct generic_pm_domain *genpd)
 {
-	if (!WARN_ON(genpd->sd_count == 0))
-			genpd->sd_count--;
+	bool ret = false;
+
+	if (!WARN_ON(atomic_read(&genpd->sd_count) == 0))
+		ret = !!atomic_dec_and_test(&genpd->sd_count);
+
+	return ret;
+}
+
+static void genpd_sd_counter_inc(struct generic_pm_domain *genpd)
+{
+	atomic_inc(&genpd->sd_count);
+	smp_mb__after_atomic_inc();
 }
 
 static void genpd_acquire_lock(struct generic_pm_domain *genpd)
@@ -71,81 +81,119 @@ static void genpd_set_active(struct generic_pm_domain *genpd)
 }
 
 /**
- * pm_genpd_poweron - Restore power to a given PM domain and its parents.
+ * __pm_genpd_poweron - Restore power to a given PM domain and its masters.
  * @genpd: PM domain to power up.
  *
- * Restore power to @genpd and all of its parents so that it is possible to
+ * Restore power to @genpd and all of its masters so that it is possible to
  * resume a device belonging to it.
  */
-int pm_genpd_poweron(struct generic_pm_domain *genpd)
+int __pm_genpd_poweron(struct generic_pm_domain *genpd)
+	__releases(&genpd->lock) __acquires(&genpd->lock)
 {
-	struct generic_pm_domain *parent = genpd->parent;
+	struct gpd_link *link;
+	DEFINE_WAIT(wait);
 	int ret = 0;
 
- start:
-	if (parent) {
-		genpd_acquire_lock(parent);
-		mutex_lock_nested(&genpd->lock, SINGLE_DEPTH_NESTING);
-	} else {
+	/* If the domain's master is being waited for, we have to wait too. */
+	for (;;) {
+		prepare_to_wait(&genpd->status_wait_queue, &wait,
+				TASK_UNINTERRUPTIBLE);
+		if (genpd->status != GPD_STATE_WAIT_MASTER)
+			break;
+		mutex_unlock(&genpd->lock);
+
+		schedule();
+
 		mutex_lock(&genpd->lock);
 	}
+	finish_wait(&genpd->status_wait_queue, &wait);
 
 	if (genpd->status == GPD_STATE_ACTIVE
 	    || (genpd->prepared_count > 0 && genpd->suspend_power_off))
-		goto out;
+		return 0;
 
 	if (genpd->status != GPD_STATE_POWER_OFF) {
 		genpd_set_active(genpd);
-		goto out;
+		return 0;
 	}
 
-	if (parent && parent->status != GPD_STATE_ACTIVE) {
+	/*
+	 * The list is guaranteed not to change while the loop below is being
+	 * executed, unless one of the masters' .power_on() callbacks fiddles
+	 * with it.
+	 */
+	list_for_each_entry(link, &genpd->slave_links, slave_node) {
+		genpd_sd_counter_inc(link->master);
+		genpd->status = GPD_STATE_WAIT_MASTER;
+
 		mutex_unlock(&genpd->lock);
-		genpd_release_lock(parent);
 
-		ret = pm_genpd_poweron(parent);
-		if (ret)
-			return ret;
+		ret = pm_genpd_poweron(link->master);
 
-		goto start;
+		mutex_lock(&genpd->lock);
+
+		/*
+		 * The "wait for parent" status is guaranteed not to change
+		 * while the master is powering on.
+		 */
+		genpd->status = GPD_STATE_POWER_OFF;
+		wake_up_all(&genpd->status_wait_queue);
+		if (ret) {
+			genpd_sd_counter_dec(link->master);
+			goto err;
+		}
 	}
 
 	if (genpd->power_on) {
 		ret = genpd->power_on(genpd);
 		if (ret)
-			goto out;
+			goto err;
 	}
 
 	genpd_set_active(genpd);
-	if (parent)
-		parent->sd_count++;
 
- out:
-	mutex_unlock(&genpd->lock);
-	if (parent)
-		genpd_release_lock(parent);
+	return 0;
+
+ err:
+	list_for_each_entry_continue_reverse(link, &genpd->slave_links, slave_node)
+		genpd_sd_counter_dec(link->master);
 
 	return ret;
 }
 
+/**
+ * pm_genpd_poweron - Restore power to a given PM domain and its masters.
+ * @genpd: PM domain to power up.
+ */
+int pm_genpd_poweron(struct generic_pm_domain *genpd)
+{
+	int ret;
+
+	mutex_lock(&genpd->lock);
+	ret = __pm_genpd_poweron(genpd);
+	mutex_unlock(&genpd->lock);
+	return ret;
+}
+
 #endif /* CONFIG_PM */
 
 #ifdef CONFIG_PM_RUNTIME
 
 /**
  * __pm_genpd_save_device - Save the pre-suspend state of a device.
- * @dle: Device list entry of the device to save the state of.
+ * @pdd: Domain data of the device to save the state of.
  * @genpd: PM domain the device belongs to.
  */
-static int __pm_genpd_save_device(struct dev_list_entry *dle,
+static int __pm_genpd_save_device(struct pm_domain_data *pdd,
 				  struct generic_pm_domain *genpd)
 	__releases(&genpd->lock) __acquires(&genpd->lock)
 {
-	struct device *dev = dle->dev;
+	struct generic_pm_domain_data *gpd_data = to_gpd_data(pdd);
+	struct device *dev = pdd->dev;
 	struct device_driver *drv = dev->driver;
 	int ret = 0;
 
-	if (dle->need_restore)
+	if (gpd_data->need_restore)
 		return 0;
 
 	mutex_unlock(&genpd->lock);
@@ -163,24 +211,25 @@ static int __pm_genpd_save_device(struct dev_list_entry *dle,
 	mutex_lock(&genpd->lock);
 
 	if (!ret)
-		dle->need_restore = true;
+		gpd_data->need_restore = true;
 
 	return ret;
 }
 
 /**
  * __pm_genpd_restore_device - Restore the pre-suspend state of a device.
- * @dle: Device list entry of the device to restore the state of.
+ * @pdd: Domain data of the device to restore the state of.
  * @genpd: PM domain the device belongs to.
  */
-static void __pm_genpd_restore_device(struct dev_list_entry *dle,
+static void __pm_genpd_restore_device(struct pm_domain_data *pdd,
 				      struct generic_pm_domain *genpd)
 	__releases(&genpd->lock) __acquires(&genpd->lock)
 {
-	struct device *dev = dle->dev;
+	struct generic_pm_domain_data *gpd_data = to_gpd_data(pdd);
+	struct device *dev = pdd->dev;
 	struct device_driver *drv = dev->driver;
 
-	if (!dle->need_restore)
+	if (!gpd_data->need_restore)
 		return;
 
 	mutex_unlock(&genpd->lock);
@@ -197,7 +246,7 @@ static void __pm_genpd_restore_device(struct dev_list_entry *dle,
 
 	mutex_lock(&genpd->lock);
 
-	dle->need_restore = false;
+	gpd_data->need_restore = false;
 }
 
 /**
@@ -211,7 +260,8 @@ static void __pm_genpd_restore_device(struct dev_list_entry *dle,
  */
 static bool genpd_abort_poweroff(struct generic_pm_domain *genpd)
 {
-	return genpd->status == GPD_STATE_ACTIVE || genpd->resume_count > 0;
+	return genpd->status == GPD_STATE_WAIT_MASTER
+		|| genpd->status == GPD_STATE_ACTIVE || genpd->resume_count > 0;
 }
 
 /**
@@ -238,8 +288,8 @@ void genpd_queue_power_off_work(struct generic_pm_domain *genpd)
 static int pm_genpd_poweroff(struct generic_pm_domain *genpd)
 	__releases(&genpd->lock) __acquires(&genpd->lock)
 {
-	struct generic_pm_domain *parent;
-	struct dev_list_entry *dle;
+	struct pm_domain_data *pdd;
+	struct gpd_link *link;
 	unsigned int not_suspended;
 	int ret = 0;
 
@@ -247,19 +297,22 @@ static int pm_genpd_poweroff(struct generic_pm_domain *genpd)
 	/*
 	 * Do not try to power off the domain in the following situations:
 	 * (1) The domain is already in the "power off" state.
-	 * (2) System suspend is in progress.
+	 * (2) The domain is waiting for its master to power up.
 	 * (3) One of the domain's devices is being resumed right now.
+	 * (4) System suspend is in progress.
 	 */
-	if (genpd->status == GPD_STATE_POWER_OFF || genpd->prepared_count > 0
-	    || genpd->resume_count > 0)
+	if (genpd->status == GPD_STATE_POWER_OFF
+	    || genpd->status == GPD_STATE_WAIT_MASTER
+	    || genpd->resume_count > 0 || genpd->prepared_count > 0)
 		return 0;
 
-	if (genpd->sd_count > 0)
+	if (atomic_read(&genpd->sd_count) > 0)
 		return -EBUSY;
 
 	not_suspended = 0;
-	list_for_each_entry(dle, &genpd->dev_list, node)
-		if (dle->dev->driver && !pm_runtime_suspended(dle->dev))
+	list_for_each_entry(pdd, &genpd->dev_list, list_node)
+		if (pdd->dev->driver && (!pm_runtime_suspended(pdd->dev)
+		    || pdd->dev->power.irq_safe))
 			not_suspended++;
 
 	if (not_suspended > genpd->in_progress)
@@ -282,54 +335,50 @@ static int pm_genpd_poweroff(struct generic_pm_domain *genpd)
 	genpd->status = GPD_STATE_BUSY;
 	genpd->poweroff_task = current;
 
-	list_for_each_entry_reverse(dle, &genpd->dev_list, node) {
-		ret = __pm_genpd_save_device(dle, genpd);
+	list_for_each_entry_reverse(pdd, &genpd->dev_list, list_node) {
+		ret = atomic_read(&genpd->sd_count) == 0 ?
+			__pm_genpd_save_device(pdd, genpd) : -EBUSY;
+
+		if (genpd_abort_poweroff(genpd))
+			goto out;
+
 		if (ret) {
 			genpd_set_active(genpd);
 			goto out;
 		}
 
-		if (genpd_abort_poweroff(genpd))
-			goto out;
-
 		if (genpd->status == GPD_STATE_REPEAT) {
 			genpd->poweroff_task = NULL;
 			goto start;
 		}
 	}
 
-	parent = genpd->parent;
-	if (parent) {
-		mutex_unlock(&genpd->lock);
-
-		genpd_acquire_lock(parent);
-		mutex_lock_nested(&genpd->lock, SINGLE_DEPTH_NESTING);
-
-		if (genpd_abort_poweroff(genpd)) {
-			genpd_release_lock(parent);
+	if (genpd->power_off) {
+		if (atomic_read(&genpd->sd_count) > 0) {
+			ret = -EBUSY;
 			goto out;
 		}
-	}
 
-	if (genpd->power_off) {
+		/*
+		 * If sd_count > 0 at this point, one of the subdomains hasn't
+		 * managed to call pm_genpd_poweron() for the master yet after
+		 * incrementing it.  In that case pm_genpd_poweron() will wait
+		 * for us to drop the lock, so we can call .power_off() and let
+		 * the pm_genpd_poweron() restore power for us (this shouldn't
+		 * happen very often).
+		 */
 		ret = genpd->power_off(genpd);
 		if (ret == -EBUSY) {
 			genpd_set_active(genpd);
-			if (parent)
-				genpd_release_lock(parent);
-
 			goto out;
 		}
 	}
 
 	genpd->status = GPD_STATE_POWER_OFF;
 
-	if (parent) {
-		genpd_sd_counter_dec(parent);
-		if (parent->sd_count == 0)
-			genpd_queue_power_off_work(parent);
-
-		genpd_release_lock(parent);
+	list_for_each_entry(link, &genpd->slave_links, slave_node) {
+		genpd_sd_counter_dec(link->master);
+		genpd_queue_power_off_work(link->master);
 	}
 
  out:
@@ -371,12 +420,21 @@ static int pm_genpd_runtime_suspend(struct device *dev)
 	if (IS_ERR(genpd))
 		return -EINVAL;
 
+	might_sleep_if(!genpd->dev_irq_safe);
+
 	if (genpd->stop_device) {
 		int ret = genpd->stop_device(dev);
 		if (ret)
 			return ret;
 	}
 
+	/*
+	 * If power.irq_safe is set, this routine will be run with interrupts
+	 * off, so it can't use mutexes.
+	 */
+	if (dev->power.irq_safe)
+		return 0;
+
 	mutex_lock(&genpd->lock);
 	genpd->in_progress++;
 	pm_genpd_poweroff(genpd);
@@ -387,24 +445,6 @@ static int pm_genpd_runtime_suspend(struct device *dev)
 }
 
 /**
- * __pm_genpd_runtime_resume - Resume a device belonging to I/O PM domain.
- * @dev: Device to resume.
- * @genpd: PM domain the device belongs to.
- */
-static void __pm_genpd_runtime_resume(struct device *dev,
-				      struct generic_pm_domain *genpd)
-{
-	struct dev_list_entry *dle;
-
-	list_for_each_entry(dle, &genpd->dev_list, node) {
-		if (dle->dev == dev) {
-			__pm_genpd_restore_device(dle, genpd);
-			break;
-		}
-	}
-}
-
-/**
  * pm_genpd_runtime_resume - Resume a device belonging to I/O PM domain.
  * @dev: Device to resume.
  *
@@ -424,11 +464,18 @@ static int pm_genpd_runtime_resume(struct device *dev)
 	if (IS_ERR(genpd))
 		return -EINVAL;
 
-	ret = pm_genpd_poweron(genpd);
-	if (ret)
-		return ret;
+	might_sleep_if(!genpd->dev_irq_safe);
+
+	/* If power.irq_safe, the PM domain is never powered off. */
+	if (dev->power.irq_safe)
+		goto out;
 
 	mutex_lock(&genpd->lock);
+	ret = __pm_genpd_poweron(genpd);
+	if (ret) {
+		mutex_unlock(&genpd->lock);
+		return ret;
+	}
 	genpd->status = GPD_STATE_BUSY;
 	genpd->resume_count++;
 	for (;;) {
@@ -448,12 +495,13 @@ static int pm_genpd_runtime_resume(struct device *dev)
 		mutex_lock(&genpd->lock);
 	}
 	finish_wait(&genpd->status_wait_queue, &wait);
-	__pm_genpd_runtime_resume(dev, genpd);
+	__pm_genpd_restore_device(dev->power.subsys_data->domain_data, genpd);
 	genpd->resume_count--;
 	genpd_set_active(genpd);
 	wake_up_all(&genpd->status_wait_queue);
 	mutex_unlock(&genpd->lock);
 
+ out:
 	if (genpd->start_device)
 		genpd->start_device(dev);
 
@@ -478,8 +526,6 @@ void pm_genpd_poweroff_unused(void)
 #else
 
 static inline void genpd_power_off_work_fn(struct work_struct *work) {}
-static inline void __pm_genpd_runtime_resume(struct device *dev,
-					     struct generic_pm_domain *genpd) {}
 
 #define pm_genpd_runtime_suspend	NULL
 #define pm_genpd_runtime_resume		NULL
@@ -489,11 +535,11 @@ static inline void __pm_genpd_runtime_resume(struct device *dev,
 #ifdef CONFIG_PM_SLEEP
 
 /**
- * pm_genpd_sync_poweroff - Synchronously power off a PM domain and its parents.
+ * pm_genpd_sync_poweroff - Synchronously power off a PM domain and its masters.
  * @genpd: PM domain to power off, if possible.
  *
  * Check if the given PM domain can be powered off (during system suspend or
- * hibernation) and do that if so.  Also, in that case propagate to its parent.
+ * hibernation) and do that if so.  Also, in that case propagate to its masters.
  *
  * This function is only called in "noirq" stages of system power transitions,
  * so it need not acquire locks (all of the "noirq" callbacks are executed
@@ -501,21 +547,23 @@ static inline void __pm_genpd_runtime_resume(struct device *dev,
  */
 static void pm_genpd_sync_poweroff(struct generic_pm_domain *genpd)
 {
-	struct generic_pm_domain *parent = genpd->parent;
+	struct gpd_link *link;
 
 	if (genpd->status == GPD_STATE_POWER_OFF)
 		return;
 
-	if (genpd->suspended_count != genpd->device_count || genpd->sd_count > 0)
+	if (genpd->suspended_count != genpd->device_count
+	    || atomic_read(&genpd->sd_count) > 0)
 		return;
 
 	if (genpd->power_off)
 		genpd->power_off(genpd);
 
 	genpd->status = GPD_STATE_POWER_OFF;
-	if (parent) {
-		genpd_sd_counter_dec(parent);
-		pm_genpd_sync_poweroff(parent);
+
+	list_for_each_entry(link, &genpd->slave_links, slave_node) {
+		genpd_sd_counter_dec(link->master);
+		pm_genpd_sync_poweroff(link->master);
 	}
 }
 
@@ -666,7 +714,7 @@ static int pm_genpd_suspend_noirq(struct device *dev)
 	if (ret)
 		return ret;
 
-	if (device_may_wakeup(dev)
+	if (dev->power.wakeup_path
 	    && genpd->active_wakeup && genpd->active_wakeup(dev))
 		return 0;
 
@@ -890,7 +938,7 @@ static int pm_genpd_dev_poweroff_noirq(struct device *dev)
 	if (ret)
 		return ret;
 
-	if (device_may_wakeup(dev)
+	if (dev->power.wakeup_path
 	    && genpd->active_wakeup && genpd->active_wakeup(dev))
 		return 0;
 
@@ -1034,7 +1082,8 @@ static void pm_genpd_complete(struct device *dev)
  */
 int pm_genpd_add_device(struct generic_pm_domain *genpd, struct device *dev)
 {
-	struct dev_list_entry *dle;
+	struct generic_pm_domain_data *gpd_data;
+	struct pm_domain_data *pdd;
 	int ret = 0;
 
 	dev_dbg(dev, "%s()\n", __func__);
@@ -1054,26 +1103,26 @@ int pm_genpd_add_device(struct generic_pm_domain *genpd, struct device *dev)
 		goto out;
 	}
 
-	list_for_each_entry(dle, &genpd->dev_list, node)
-		if (dle->dev == dev) {
+	list_for_each_entry(pdd, &genpd->dev_list, list_node)
+		if (pdd->dev == dev) {
 			ret = -EINVAL;
 			goto out;
 		}
 
-	dle = kzalloc(sizeof(*dle), GFP_KERNEL);
-	if (!dle) {
+	gpd_data = kzalloc(sizeof(*gpd_data), GFP_KERNEL);
+	if (!gpd_data) {
 		ret = -ENOMEM;
 		goto out;
 	}
 
-	dle->dev = dev;
-	dle->need_restore = false;
-	list_add_tail(&dle->node, &genpd->dev_list);
 	genpd->device_count++;
 
-	spin_lock_irq(&dev->power.lock);
 	dev->pm_domain = &genpd->domain;
-	spin_unlock_irq(&dev->power.lock);
+	dev_pm_get_subsys_data(dev);
+	dev->power.subsys_data->domain_data = &gpd_data->base;
+	gpd_data->base.dev = dev;
+	gpd_data->need_restore = false;
+	list_add_tail(&gpd_data->base.list_node, &genpd->dev_list);
 
  out:
 	genpd_release_lock(genpd);
@@ -1089,7 +1138,7 @@ int pm_genpd_add_device(struct generic_pm_domain *genpd, struct device *dev)
 int pm_genpd_remove_device(struct generic_pm_domain *genpd,
 			   struct device *dev)
 {
-	struct dev_list_entry *dle;
+	struct pm_domain_data *pdd;
 	int ret = -EINVAL;
 
 	dev_dbg(dev, "%s()\n", __func__);
@@ -1104,17 +1153,17 @@ int pm_genpd_remove_device(struct generic_pm_domain *genpd,
 		goto out;
 	}
 
-	list_for_each_entry(dle, &genpd->dev_list, node) {
-		if (dle->dev != dev)
+	list_for_each_entry(pdd, &genpd->dev_list, list_node) {
+		if (pdd->dev != dev)
 			continue;
 
-		spin_lock_irq(&dev->power.lock);
+		list_del_init(&pdd->list_node);
+		pdd->dev = NULL;
+		dev_pm_put_subsys_data(dev);
 		dev->pm_domain = NULL;
-		spin_unlock_irq(&dev->power.lock);
+		kfree(to_gpd_data(pdd));
 
 		genpd->device_count--;
-		list_del(&dle->node);
-		kfree(dle);
 
 		ret = 0;
 		break;
@@ -1129,48 +1178,55 @@ int pm_genpd_remove_device(struct generic_pm_domain *genpd,
 /**
  * pm_genpd_add_subdomain - Add a subdomain to an I/O PM domain.
  * @genpd: Master PM domain to add the subdomain to.
- * @new_subdomain: Subdomain to be added.
+ * @subdomain: Subdomain to be added.
  */
 int pm_genpd_add_subdomain(struct generic_pm_domain *genpd,
-			   struct generic_pm_domain *new_subdomain)
+			   struct generic_pm_domain *subdomain)
 {
-	struct generic_pm_domain *subdomain;
+	struct gpd_link *link;
 	int ret = 0;
 
-	if (IS_ERR_OR_NULL(genpd) || IS_ERR_OR_NULL(new_subdomain))
+	if (IS_ERR_OR_NULL(genpd) || IS_ERR_OR_NULL(subdomain))
 		return -EINVAL;
 
  start:
 	genpd_acquire_lock(genpd);
-	mutex_lock_nested(&new_subdomain->lock, SINGLE_DEPTH_NESTING);
+	mutex_lock_nested(&subdomain->lock, SINGLE_DEPTH_NESTING);
 
-	if (new_subdomain->status != GPD_STATE_POWER_OFF
-	    && new_subdomain->status != GPD_STATE_ACTIVE) {
-		mutex_unlock(&new_subdomain->lock);
+	if (subdomain->status != GPD_STATE_POWER_OFF
+	    && subdomain->status != GPD_STATE_ACTIVE) {
+		mutex_unlock(&subdomain->lock);
 		genpd_release_lock(genpd);
 		goto start;
 	}
 
 	if (genpd->status == GPD_STATE_POWER_OFF
-	    &&  new_subdomain->status != GPD_STATE_POWER_OFF) {
+	    &&  subdomain->status != GPD_STATE_POWER_OFF) {
 		ret = -EINVAL;
 		goto out;
 	}
 
-	list_for_each_entry(subdomain, &genpd->sd_list, sd_node) {
-		if (subdomain == new_subdomain) {
+	list_for_each_entry(link, &genpd->slave_links, slave_node) {
+		if (link->slave == subdomain && link->master == genpd) {
 			ret = -EINVAL;
 			goto out;
 		}
 	}
 
-	list_add_tail(&new_subdomain->sd_node, &genpd->sd_list);
-	new_subdomain->parent = genpd;
+	link = kzalloc(sizeof(*link), GFP_KERNEL);
+	if (!link) {
+		ret = -ENOMEM;
+		goto out;
+	}
+	link->master = genpd;
+	list_add_tail(&link->master_node, &genpd->master_links);
+	link->slave = subdomain;
+	list_add_tail(&link->slave_node, &subdomain->slave_links);
 	if (subdomain->status != GPD_STATE_POWER_OFF)
-		genpd->sd_count++;
+		genpd_sd_counter_inc(genpd);
 
  out:
-	mutex_unlock(&new_subdomain->lock);
+	mutex_unlock(&subdomain->lock);
 	genpd_release_lock(genpd);
 
 	return ret;
@@ -1179,22 +1235,22 @@ int pm_genpd_add_subdomain(struct generic_pm_domain *genpd,
 /**
  * pm_genpd_remove_subdomain - Remove a subdomain from an I/O PM domain.
  * @genpd: Master PM domain to remove the subdomain from.
- * @target: Subdomain to be removed.
+ * @subdomain: Subdomain to be removed.
  */
 int pm_genpd_remove_subdomain(struct generic_pm_domain *genpd,
-			      struct generic_pm_domain *target)
+			      struct generic_pm_domain *subdomain)
 {
-	struct generic_pm_domain *subdomain;
+	struct gpd_link *link;
 	int ret = -EINVAL;
 
-	if (IS_ERR_OR_NULL(genpd) || IS_ERR_OR_NULL(target))
+	if (IS_ERR_OR_NULL(genpd) || IS_ERR_OR_NULL(subdomain))
 		return -EINVAL;
 
  start:
 	genpd_acquire_lock(genpd);
 
-	list_for_each_entry(subdomain, &genpd->sd_list, sd_node) {
-		if (subdomain != target)
+	list_for_each_entry(link, &genpd->master_links, master_node) {
+		if (link->slave != subdomain)
 			continue;
 
 		mutex_lock_nested(&subdomain->lock, SINGLE_DEPTH_NESTING);
@@ -1206,8 +1262,9 @@ int pm_genpd_remove_subdomain(struct generic_pm_domain *genpd,
 			goto start;
 		}
 
-		list_del(&subdomain->sd_node);
-		subdomain->parent = NULL;
+		list_del(&link->master_node);
+		list_del(&link->slave_node);
+		kfree(link);
 		if (subdomain->status != GPD_STATE_POWER_OFF)
 			genpd_sd_counter_dec(genpd);
 
@@ -1234,15 +1291,14 @@ void pm_genpd_init(struct generic_pm_domain *genpd,
 	if (IS_ERR_OR_NULL(genpd))
 		return;
 
-	INIT_LIST_HEAD(&genpd->sd_node);
-	genpd->parent = NULL;
+	INIT_LIST_HEAD(&genpd->master_links);
+	INIT_LIST_HEAD(&genpd->slave_links);
 	INIT_LIST_HEAD(&genpd->dev_list);
-	INIT_LIST_HEAD(&genpd->sd_list);
 	mutex_init(&genpd->lock);
 	genpd->gov = gov;
 	INIT_WORK(&genpd->power_off_work, genpd_power_off_work_fn);
 	genpd->in_progress = 0;
-	genpd->sd_count = 0;
+	atomic_set(&genpd->sd_count, 0);
 	genpd->status = is_off ? GPD_STATE_POWER_OFF : GPD_STATE_ACTIVE;
 	init_waitqueue_head(&genpd->status_wait_queue);
 	genpd->poweroff_task = NULL;
diff --git a/drivers/base/power/main.c b/drivers/base/power/main.c
index a85459126bc..59f8ab23548 100644
--- a/drivers/base/power/main.c
+++ b/drivers/base/power/main.c
@@ -46,6 +46,7 @@ LIST_HEAD(dpm_prepared_list);
 LIST_HEAD(dpm_suspended_list);
 LIST_HEAD(dpm_noirq_list);
 
+struct suspend_stats suspend_stats;
 static DEFINE_MUTEX(dpm_list_mtx);
 static pm_message_t pm_transition;
 
@@ -65,6 +66,7 @@ void device_pm_init(struct device *dev)
 	spin_lock_init(&dev->power.lock);
 	pm_runtime_init(dev);
 	INIT_LIST_HEAD(&dev->power.entry);
+	dev->power.power_state = PMSG_INVALID;
 }
 
 /**
@@ -96,6 +98,7 @@ void device_pm_add(struct device *dev)
 		dev_warn(dev, "parent %s should not be sleeping\n",
 			dev_name(dev->parent));
 	list_add_tail(&dev->power.entry, &dpm_list);
+	dev_pm_qos_constraints_init(dev);
 	mutex_unlock(&dpm_list_mtx);
 }
 
@@ -109,6 +112,7 @@ void device_pm_remove(struct device *dev)
 		 dev->bus ? dev->bus->name : "No Bus", dev_name(dev));
 	complete_all(&dev->power.completion);
 	mutex_lock(&dpm_list_mtx);
+	dev_pm_qos_constraints_destroy(dev);
 	list_del_init(&dev->power.entry);
 	mutex_unlock(&dpm_list_mtx);
 	device_wakeup_disable(dev);
@@ -464,8 +468,12 @@ void dpm_resume_noirq(pm_message_t state)
 		mutex_unlock(&dpm_list_mtx);
 
 		error = device_resume_noirq(dev, state);
-		if (error)
+		if (error) {
+			suspend_stats.failed_resume_noirq++;
+			dpm_save_failed_step(SUSPEND_RESUME_NOIRQ);
+			dpm_save_failed_dev(dev_name(dev));
 			pm_dev_err(dev, state, " early", error);
+		}
 
 		mutex_lock(&dpm_list_mtx);
 		put_device(dev);
@@ -626,8 +634,12 @@ void dpm_resume(pm_message_t state)
 			mutex_unlock(&dpm_list_mtx);
 
 			error = device_resume(dev, state, false);
-			if (error)
+			if (error) {
+				suspend_stats.failed_resume++;
+				dpm_save_failed_step(SUSPEND_RESUME);
+				dpm_save_failed_dev(dev_name(dev));
 				pm_dev_err(dev, state, "", error);
+			}
 
 			mutex_lock(&dpm_list_mtx);
 		}
@@ -802,6 +814,9 @@ int dpm_suspend_noirq(pm_message_t state)
 		mutex_lock(&dpm_list_mtx);
 		if (error) {
 			pm_dev_err(dev, state, " late", error);
+			suspend_stats.failed_suspend_noirq++;
+			dpm_save_failed_step(SUSPEND_SUSPEND_NOIRQ);
+			dpm_save_failed_dev(dev_name(dev));
 			put_device(dev);
 			break;
 		}
@@ -902,7 +917,11 @@ static int __device_suspend(struct device *dev, pm_message_t state, bool async)
 	}
 
  End:
-	dev->power.is_suspended = !error;
+	if (!error) {
+		dev->power.is_suspended = true;
+		if (dev->power.wakeup_path && dev->parent)
+			dev->parent->power.wakeup_path = true;
+	}
 
 	device_unlock(dev);
 	complete_all(&dev->power.completion);
@@ -923,8 +942,10 @@ static void async_suspend(void *data, async_cookie_t cookie)
 	int error;
 
 	error = __device_suspend(dev, pm_transition, true);
-	if (error)
+	if (error) {
+		dpm_save_failed_dev(dev_name(dev));
 		pm_dev_err(dev, pm_transition, " async", error);
+	}
 
 	put_device(dev);
 }
@@ -967,6 +988,7 @@ int dpm_suspend(pm_message_t state)
 		mutex_lock(&dpm_list_mtx);
 		if (error) {
 			pm_dev_err(dev, state, "", error);
+			dpm_save_failed_dev(dev_name(dev));
 			put_device(dev);
 			break;
 		}
@@ -980,7 +1002,10 @@ int dpm_suspend(pm_message_t state)
 	async_synchronize_full();
 	if (!error)
 		error = async_error;
-	if (!error)
+	if (error) {
+		suspend_stats.failed_suspend++;
+		dpm_save_failed_step(SUSPEND_SUSPEND);
+	} else
 		dpm_show_time(starttime, state, NULL);
 	return error;
 }
@@ -999,6 +1024,8 @@ static int device_prepare(struct device *dev, pm_message_t state)
 
 	device_lock(dev);
 
+	dev->power.wakeup_path = device_may_wakeup(dev);
+
 	if (dev->pm_domain) {
 		pm_dev_dbg(dev, state, "preparing power domain ");
 		if (dev->pm_domain->ops.prepare)
@@ -1088,7 +1115,10 @@ int dpm_suspend_start(pm_message_t state)
 	int error;
 
 	error = dpm_prepare(state);
-	if (!error)
+	if (error) {
+		suspend_stats.failed_prepare++;
+		dpm_save_failed_step(SUSPEND_PREPARE);
+	} else
 		error = dpm_suspend(state);
 	return error;
 }
diff --git a/drivers/base/power/opp.c b/drivers/base/power/opp.c
index b23de185cb0..434a6c01167 100644
--- a/drivers/base/power/opp.c
+++ b/drivers/base/power/opp.c
@@ -73,6 +73,7 @@ struct opp {
  *		RCU usage: nodes are not modified in the list of device_opp,
  *		however addition is possible and is secured by dev_opp_list_lock
  * @dev:	device pointer
+ * @head:	notifier head to notify the OPP availability changes.
  * @opp_list:	list of opps
  *
  * This is an internal data structure maintaining the link to opps attached to
@@ -83,6 +84,7 @@ struct device_opp {
 	struct list_head node;
 
 	struct device *dev;
+	struct srcu_notifier_head head;
 	struct list_head opp_list;
 };
 
@@ -404,6 +406,7 @@ int opp_add(struct device *dev, unsigned long freq, unsigned long u_volt)
 		}
 
 		dev_opp->dev = dev;
+		srcu_init_notifier_head(&dev_opp->head);
 		INIT_LIST_HEAD(&dev_opp->opp_list);
 
 		/* Secure the device list modification */
@@ -428,6 +431,11 @@ int opp_add(struct device *dev, unsigned long freq, unsigned long u_volt)
 	list_add_rcu(&new_opp->node, head);
 	mutex_unlock(&dev_opp_list_lock);
 
+	/*
+	 * Notify the changes in the availability of the operable
+	 * frequency/voltage list.
+	 */
+	srcu_notifier_call_chain(&dev_opp->head, OPP_EVENT_ADD, new_opp);
 	return 0;
 }
 
@@ -504,6 +512,14 @@ static int opp_set_availability(struct device *dev, unsigned long freq,
 	mutex_unlock(&dev_opp_list_lock);
 	synchronize_rcu();
 
+	/* Notify the change of the OPP availability */
+	if (availability_req)
+		srcu_notifier_call_chain(&dev_opp->head, OPP_EVENT_ENABLE,
+					 new_opp);
+	else
+		srcu_notifier_call_chain(&dev_opp->head, OPP_EVENT_DISABLE,
+					 new_opp);
+
 	/* clean up old opp */
 	new_opp = opp;
 	goto out;
@@ -643,3 +659,17 @@ void opp_free_cpufreq_table(struct device *dev,
 	*table = NULL;
 }
 #endif		/* CONFIG_CPU_FREQ */
+
+/**
+ * opp_get_notifier() - find notifier_head of the device with opp
+ * @dev:	device pointer used to lookup device OPPs.
+ */
+struct srcu_notifier_head *opp_get_notifier(struct device *dev)
+{
+	struct device_opp *dev_opp = find_device_opp(dev);
+
+	if (IS_ERR(dev_opp))
+		return ERR_PTR(PTR_ERR(dev_opp)); /* matching type */
+
+	return &dev_opp->head;
+}
diff --git a/drivers/base/power/power.h b/drivers/base/power/power.h
index f2a25f18fde..9bf62323aaf 100644
--- a/drivers/base/power/power.h
+++ b/drivers/base/power/power.h
@@ -1,3 +1,5 @@
+#include <linux/pm_qos.h>
+
 #ifdef CONFIG_PM_RUNTIME
 
 extern void pm_runtime_init(struct device *dev);
@@ -35,15 +37,21 @@ extern void device_pm_move_last(struct device *);
 static inline void device_pm_init(struct device *dev)
 {
 	spin_lock_init(&dev->power.lock);
+	dev->power.power_state = PMSG_INVALID;
 	pm_runtime_init(dev);
 }
 
+static inline void device_pm_add(struct device *dev)
+{
+	dev_pm_qos_constraints_init(dev);
+}
+
 static inline void device_pm_remove(struct device *dev)
 {
+	dev_pm_qos_constraints_destroy(dev);
 	pm_runtime_remove(dev);
 }
 
-static inline void device_pm_add(struct device *dev) {}
 static inline void device_pm_move_before(struct device *deva,
 					 struct device *devb) {}
 static inline void device_pm_move_after(struct device *deva,
diff --git a/drivers/base/power/qos.c b/drivers/base/power/qos.c
new file mode 100644
index 00000000000..91e06141738
--- /dev/null
+++ b/drivers/base/power/qos.c
@@ -0,0 +1,419 @@
+/*
+ * Devices PM QoS constraints management
+ *
+ * Copyright (C) 2011 Texas Instruments, Inc.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ *
+ * This module exposes the interface to kernel space for specifying
+ * per-device PM QoS dependencies. It provides infrastructure for registration
+ * of:
+ *
+ * Dependents on a QoS value : register requests
+ * Watchers of QoS value : get notified when target QoS value changes
+ *
+ * This QoS design is best effort based. Dependents register their QoS needs.
+ * Watchers register to keep track of the current QoS needs of the system.
+ * Watchers can register different types of notification callbacks:
+ *  . a per-device notification callback using the dev_pm_qos_*_notifier API.
+ *    The notification chain data is stored in the per-device constraint
+ *    data struct.
+ *  . a system-wide notification callback using the dev_pm_qos_*_global_notifier
+ *    API. The notification chain data is stored in a static variable.
+ *
+ * Note about the per-device constraint data struct allocation:
+ * . The per-device constraints data struct ptr is tored into the device
+ *    dev_pm_info.
+ * . To minimize the data usage by the per-device constraints, the data struct
+ *   is only allocated at the first call to dev_pm_qos_add_request.
+ * . The data is later free'd when the device is removed from the system.
+ *  . A global mutex protects the constraints users from the data being
+ *     allocated and free'd.
+ */
+
+#include <linux/pm_qos.h>
+#include <linux/spinlock.h>
+#include <linux/slab.h>
+#include <linux/device.h>
+#include <linux/mutex.h>
+
+
+static DEFINE_MUTEX(dev_pm_qos_mtx);
+
+static BLOCKING_NOTIFIER_HEAD(dev_pm_notifiers);
+
+/**
+ * dev_pm_qos_read_value - Get PM QoS constraint for a given device.
+ * @dev: Device to get the PM QoS constraint value for.
+ */
+s32 dev_pm_qos_read_value(struct device *dev)
+{
+	struct pm_qos_constraints *c;
+	unsigned long flags;
+	s32 ret = 0;
+
+	spin_lock_irqsave(&dev->power.lock, flags);
+
+	c = dev->power.constraints;
+	if (c)
+		ret = pm_qos_read_value(c);
+
+	spin_unlock_irqrestore(&dev->power.lock, flags);
+
+	return ret;
+}
+
+/*
+ * apply_constraint
+ * @req: constraint request to apply
+ * @action: action to perform add/update/remove, of type enum pm_qos_req_action
+ * @value: defines the qos request
+ *
+ * Internal function to update the constraints list using the PM QoS core
+ * code and if needed call the per-device and the global notification
+ * callbacks
+ */
+static int apply_constraint(struct dev_pm_qos_request *req,
+			    enum pm_qos_req_action action, int value)
+{
+	int ret, curr_value;
+
+	ret = pm_qos_update_target(req->dev->power.constraints,
+				   &req->node, action, value);
+
+	if (ret) {
+		/* Call the global callbacks if needed */
+		curr_value = pm_qos_read_value(req->dev->power.constraints);
+		blocking_notifier_call_chain(&dev_pm_notifiers,
+					     (unsigned long)curr_value,
+					     req);
+	}
+
+	return ret;
+}
+
+/*
+ * dev_pm_qos_constraints_allocate
+ * @dev: device to allocate data for
+ *
+ * Called at the first call to add_request, for constraint data allocation
+ * Must be called with the dev_pm_qos_mtx mutex held
+ */
+static int dev_pm_qos_constraints_allocate(struct device *dev)
+{
+	struct pm_qos_constraints *c;
+	struct blocking_notifier_head *n;
+
+	c = kzalloc(sizeof(*c), GFP_KERNEL);
+	if (!c)
+		return -ENOMEM;
+
+	n = kzalloc(sizeof(*n), GFP_KERNEL);
+	if (!n) {
+		kfree(c);
+		return -ENOMEM;
+	}
+	BLOCKING_INIT_NOTIFIER_HEAD(n);
+
+	plist_head_init(&c->list);
+	c->target_value = PM_QOS_DEV_LAT_DEFAULT_VALUE;
+	c->default_value = PM_QOS_DEV_LAT_DEFAULT_VALUE;
+	c->type = PM_QOS_MIN;
+	c->notifiers = n;
+
+	spin_lock_irq(&dev->power.lock);
+	dev->power.constraints = c;
+	spin_unlock_irq(&dev->power.lock);
+
+	return 0;
+}
+
+/**
+ * dev_pm_qos_constraints_init - Initalize device's PM QoS constraints pointer.
+ * @dev: target device
+ *
+ * Called from the device PM subsystem during device insertion under
+ * device_pm_lock().
+ */
+void dev_pm_qos_constraints_init(struct device *dev)
+{
+	mutex_lock(&dev_pm_qos_mtx);
+	dev->power.constraints = NULL;
+	dev->power.power_state = PMSG_ON;
+	mutex_unlock(&dev_pm_qos_mtx);
+}
+
+/**
+ * dev_pm_qos_constraints_destroy
+ * @dev: target device
+ *
+ * Called from the device PM subsystem on device removal under device_pm_lock().
+ */
+void dev_pm_qos_constraints_destroy(struct device *dev)
+{
+	struct dev_pm_qos_request *req, *tmp;
+	struct pm_qos_constraints *c;
+
+	mutex_lock(&dev_pm_qos_mtx);
+
+	dev->power.power_state = PMSG_INVALID;
+	c = dev->power.constraints;
+	if (!c)
+		goto out;
+
+	/* Flush the constraints list for the device */
+	plist_for_each_entry_safe(req, tmp, &c->list, node) {
+		/*
+		 * Update constraints list and call the notification
+		 * callbacks if needed
+		 */
+		apply_constraint(req, PM_QOS_REMOVE_REQ, PM_QOS_DEFAULT_VALUE);
+		memset(req, 0, sizeof(*req));
+	}
+
+	spin_lock_irq(&dev->power.lock);
+	dev->power.constraints = NULL;
+	spin_unlock_irq(&dev->power.lock);
+
+	kfree(c->notifiers);
+	kfree(c);
+
+ out:
+	mutex_unlock(&dev_pm_qos_mtx);
+}
+
+/**
+ * dev_pm_qos_add_request - inserts new qos request into the list
+ * @dev: target device for the constraint
+ * @req: pointer to a preallocated handle
+ * @value: defines the qos request
+ *
+ * This function inserts a new entry in the device constraints list of
+ * requested qos performance characteristics. It recomputes the aggregate
+ * QoS expectations of parameters and initializes the dev_pm_qos_request
+ * handle.  Caller needs to save this handle for later use in updates and
+ * removal.
+ *
+ * Returns 1 if the aggregated constraint value has changed,
+ * 0 if the aggregated constraint value has not changed,
+ * -EINVAL in case of wrong parameters, -ENOMEM if there's not enough memory
+ * to allocate for data structures, -ENODEV if the device has just been removed
+ * from the system.
+ */
+int dev_pm_qos_add_request(struct device *dev, struct dev_pm_qos_request *req,
+			   s32 value)
+{
+	int ret = 0;
+
+	if (!dev || !req) /*guard against callers passing in null */
+		return -EINVAL;
+
+	if (dev_pm_qos_request_active(req)) {
+		WARN(1, KERN_ERR "dev_pm_qos_add_request() called for already "
+			"added request\n");
+		return -EINVAL;
+	}
+
+	req->dev = dev;
+
+	mutex_lock(&dev_pm_qos_mtx);
+
+	if (!dev->power.constraints) {
+		if (dev->power.power_state.event == PM_EVENT_INVALID) {
+			/* The device has been removed from the system. */
+			req->dev = NULL;
+			ret = -ENODEV;
+			goto out;
+		} else {
+			/*
+			 * Allocate the constraints data on the first call to
+			 * add_request, i.e. only if the data is not already
+			 * allocated and if the device has not been removed.
+			 */
+			ret = dev_pm_qos_constraints_allocate(dev);
+		}
+	}
+
+	if (!ret)
+		ret = apply_constraint(req, PM_QOS_ADD_REQ, value);
+
+ out:
+	mutex_unlock(&dev_pm_qos_mtx);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dev_pm_qos_add_request);
+
+/**
+ * dev_pm_qos_update_request - modifies an existing qos request
+ * @req : handle to list element holding a dev_pm_qos request to use
+ * @new_value: defines the qos request
+ *
+ * Updates an existing dev PM qos request along with updating the
+ * target value.
+ *
+ * Attempts are made to make this code callable on hot code paths.
+ *
+ * Returns 1 if the aggregated constraint value has changed,
+ * 0 if the aggregated constraint value has not changed,
+ * -EINVAL in case of wrong parameters, -ENODEV if the device has been
+ * removed from the system
+ */
+int dev_pm_qos_update_request(struct dev_pm_qos_request *req,
+			      s32 new_value)
+{
+	int ret = 0;
+
+	if (!req) /*guard against callers passing in null */
+		return -EINVAL;
+
+	if (!dev_pm_qos_request_active(req)) {
+		WARN(1, KERN_ERR "dev_pm_qos_update_request() called for "
+			"unknown object\n");
+		return -EINVAL;
+	}
+
+	mutex_lock(&dev_pm_qos_mtx);
+
+	if (req->dev->power.constraints) {
+		if (new_value != req->node.prio)
+			ret = apply_constraint(req, PM_QOS_UPDATE_REQ,
+					       new_value);
+	} else {
+		/* Return if the device has been removed */
+		ret = -ENODEV;
+	}
+
+	mutex_unlock(&dev_pm_qos_mtx);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dev_pm_qos_update_request);
+
+/**
+ * dev_pm_qos_remove_request - modifies an existing qos request
+ * @req: handle to request list element
+ *
+ * Will remove pm qos request from the list of constraints and
+ * recompute the current target value. Call this on slow code paths.
+ *
+ * Returns 1 if the aggregated constraint value has changed,
+ * 0 if the aggregated constraint value has not changed,
+ * -EINVAL in case of wrong parameters, -ENODEV if the device has been
+ * removed from the system
+ */
+int dev_pm_qos_remove_request(struct dev_pm_qos_request *req)
+{
+	int ret = 0;
+
+	if (!req) /*guard against callers passing in null */
+		return -EINVAL;
+
+	if (!dev_pm_qos_request_active(req)) {
+		WARN(1, KERN_ERR "dev_pm_qos_remove_request() called for "
+			"unknown object\n");
+		return -EINVAL;
+	}
+
+	mutex_lock(&dev_pm_qos_mtx);
+
+	if (req->dev->power.constraints) {
+		ret = apply_constraint(req, PM_QOS_REMOVE_REQ,
+				       PM_QOS_DEFAULT_VALUE);
+		memset(req, 0, sizeof(*req));
+	} else {
+		/* Return if the device has been removed */
+		ret = -ENODEV;
+	}
+
+	mutex_unlock(&dev_pm_qos_mtx);
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dev_pm_qos_remove_request);
+
+/**
+ * dev_pm_qos_add_notifier - sets notification entry for changes to target value
+ * of per-device PM QoS constraints
+ *
+ * @dev: target device for the constraint
+ * @notifier: notifier block managed by caller.
+ *
+ * Will register the notifier into a notification chain that gets called
+ * upon changes to the target value for the device.
+ */
+int dev_pm_qos_add_notifier(struct device *dev, struct notifier_block *notifier)
+{
+	int retval = 0;
+
+	mutex_lock(&dev_pm_qos_mtx);
+
+	/* Silently return if the constraints object is not present. */
+	if (dev->power.constraints)
+		retval = blocking_notifier_chain_register(
+				dev->power.constraints->notifiers,
+				notifier);
+
+	mutex_unlock(&dev_pm_qos_mtx);
+	return retval;
+}
+EXPORT_SYMBOL_GPL(dev_pm_qos_add_notifier);
+
+/**
+ * dev_pm_qos_remove_notifier - deletes notification for changes to target value
+ * of per-device PM QoS constraints
+ *
+ * @dev: target device for the constraint
+ * @notifier: notifier block to be removed.
+ *
+ * Will remove the notifier from the notification chain that gets called
+ * upon changes to the target value.
+ */
+int dev_pm_qos_remove_notifier(struct device *dev,
+			       struct notifier_block *notifier)
+{
+	int retval = 0;
+
+	mutex_lock(&dev_pm_qos_mtx);
+
+	/* Silently return if the constraints object is not present. */
+	if (dev->power.constraints)
+		retval = blocking_notifier_chain_unregister(
+				dev->power.constraints->notifiers,
+				notifier);
+
+	mutex_unlock(&dev_pm_qos_mtx);
+	return retval;
+}
+EXPORT_SYMBOL_GPL(dev_pm_qos_remove_notifier);
+
+/**
+ * dev_pm_qos_add_global_notifier - sets notification entry for changes to
+ * target value of the PM QoS constraints for any device
+ *
+ * @notifier: notifier block managed by caller.
+ *
+ * Will register the notifier into a notification chain that gets called
+ * upon changes to the target value for any device.
+ */
+int dev_pm_qos_add_global_notifier(struct notifier_block *notifier)
+{
+	return blocking_notifier_chain_register(&dev_pm_notifiers, notifier);
+}
+EXPORT_SYMBOL_GPL(dev_pm_qos_add_global_notifier);
+
+/**
+ * dev_pm_qos_remove_global_notifier - deletes notification for changes to
+ * target value of PM QoS constraints for any device
+ *
+ * @notifier: notifier block to be removed.
+ *
+ * Will remove the notifier from the notification chain that gets called
+ * upon changes to the target value for any device.
+ */
+int dev_pm_qos_remove_global_notifier(struct notifier_block *notifier)
+{
+	return blocking_notifier_chain_unregister(&dev_pm_notifiers, notifier);
+}
+EXPORT_SYMBOL_GPL(dev_pm_qos_remove_global_notifier);
diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
index acb3f83b807..6bb3aafa85e 100644
--- a/drivers/base/power/runtime.c
+++ b/drivers/base/power/runtime.c
@@ -9,6 +9,7 @@
 
 #include <linux/sched.h>
 #include <linux/pm_runtime.h>
+#include <trace/events/rpm.h>
 #include "power.h"
 
 static int rpm_resume(struct device *dev, int rpmflags);
@@ -155,6 +156,31 @@ static int rpm_check_suspend_allowed(struct device *dev)
 }
 
 /**
+ * __rpm_callback - Run a given runtime PM callback for a given device.
+ * @cb: Runtime PM callback to run.
+ * @dev: Device to run the callback for.
+ */
+static int __rpm_callback(int (*cb)(struct device *), struct device *dev)
+	__releases(&dev->power.lock) __acquires(&dev->power.lock)
+{
+	int retval;
+
+	if (dev->power.irq_safe)
+		spin_unlock(&dev->power.lock);
+	else
+		spin_unlock_irq(&dev->power.lock);
+
+	retval = cb(dev);
+
+	if (dev->power.irq_safe)
+		spin_lock(&dev->power.lock);
+	else
+		spin_lock_irq(&dev->power.lock);
+
+	return retval;
+}
+
+/**
  * rpm_idle - Notify device bus type if the device can be suspended.
  * @dev: Device to notify the bus type about.
  * @rpmflags: Flag bits.
@@ -171,6 +197,7 @@ static int rpm_idle(struct device *dev, int rpmflags)
 	int (*callback)(struct device *);
 	int retval;
 
+	trace_rpm_idle(dev, rpmflags);
 	retval = rpm_check_suspend_allowed(dev);
 	if (retval < 0)
 		;	/* Conditions are wrong. */
@@ -225,24 +252,14 @@ static int rpm_idle(struct device *dev, int rpmflags)
 	else
 		callback = NULL;
 
-	if (callback) {
-		if (dev->power.irq_safe)
-			spin_unlock(&dev->power.lock);
-		else
-			spin_unlock_irq(&dev->power.lock);
-
-		callback(dev);
-
-		if (dev->power.irq_safe)
-			spin_lock(&dev->power.lock);
-		else
-			spin_lock_irq(&dev->power.lock);
-	}
+	if (callback)
+		__rpm_callback(callback, dev);
 
 	dev->power.idle_notification = false;
 	wake_up_all(&dev->power.wait_queue);
 
  out:
+	trace_rpm_return_int(dev, _THIS_IP_, retval);
 	return retval;
 }
 
@@ -252,22 +269,14 @@ static int rpm_idle(struct device *dev, int rpmflags)
  * @dev: Device to run the callback for.
  */
 static int rpm_callback(int (*cb)(struct device *), struct device *dev)
-	__releases(&dev->power.lock) __acquires(&dev->power.lock)
 {
 	int retval;
 
 	if (!cb)
 		return -ENOSYS;
 
-	if (dev->power.irq_safe) {
-		retval = cb(dev);
-	} else {
-		spin_unlock_irq(&dev->power.lock);
-
-		retval = cb(dev);
+	retval = __rpm_callback(cb, dev);
 
-		spin_lock_irq(&dev->power.lock);
-	}
 	dev->power.runtime_error = retval;
 	return retval != -EACCES ? retval : -EIO;
 }
@@ -277,14 +286,16 @@ static int rpm_callback(int (*cb)(struct device *), struct device *dev)
  * @dev: Device to suspend.
  * @rpmflags: Flag bits.
  *
- * Check if the device's runtime PM status allows it to be suspended.  If
- * another suspend has been started earlier, either return immediately or wait
- * for it to finish, depending on the RPM_NOWAIT and RPM_ASYNC flags.  Cancel a
- * pending idle notification.  If the RPM_ASYNC flag is set then queue a
- * suspend request; otherwise run the ->runtime_suspend() callback directly.
- * If a deferred resume was requested while the callback was running then carry
- * it out; otherwise send an idle notification for the device (if the suspend
- * failed) or for its parent (if the suspend succeeded).
+ * Check if the device's runtime PM status allows it to be suspended.
+ * Cancel a pending idle notification, autosuspend or suspend. If
+ * another suspend has been started earlier, either return immediately
+ * or wait for it to finish, depending on the RPM_NOWAIT and RPM_ASYNC
+ * flags. If the RPM_ASYNC flag is set then queue a suspend request;
+ * otherwise run the ->runtime_suspend() callback directly. When
+ * ->runtime_suspend succeeded, if a deferred resume was requested while
+ * the callback was running then carry it out, otherwise send an idle
+ * notification for its parent (if the suspend succeeded and both
+ * ignore_children of parent->power and irq_safe of dev->power are not set).
  *
  * This function must be called under dev->power.lock with interrupts disabled.
  */
@@ -295,7 +306,7 @@ static int rpm_suspend(struct device *dev, int rpmflags)
 	struct device *parent = NULL;
 	int retval;
 
-	dev_dbg(dev, "%s flags 0x%x\n", __func__, rpmflags);
+	trace_rpm_suspend(dev, rpmflags);
 
  repeat:
 	retval = rpm_check_suspend_allowed(dev);
@@ -347,6 +358,15 @@ static int rpm_suspend(struct device *dev, int rpmflags)
 			goto out;
 		}
 
+		if (dev->power.irq_safe) {
+			spin_unlock(&dev->power.lock);
+
+			cpu_relax();
+
+			spin_lock(&dev->power.lock);
+			goto repeat;
+		}
+
 		/* Wait for the other suspend running in parallel with us. */
 		for (;;) {
 			prepare_to_wait(&dev->power.wait_queue, &wait,
@@ -400,15 +420,16 @@ static int rpm_suspend(struct device *dev, int rpmflags)
 			dev->power.runtime_error = 0;
 		else
 			pm_runtime_cancel_pending(dev);
-	} else {
+		wake_up_all(&dev->power.wait_queue);
+		goto out;
+	}
  no_callback:
-		__update_runtime_status(dev, RPM_SUSPENDED);
-		pm_runtime_deactivate_timer(dev);
+	__update_runtime_status(dev, RPM_SUSPENDED);
+	pm_runtime_deactivate_timer(dev);
 
-		if (dev->parent) {
-			parent = dev->parent;
-			atomic_add_unless(&parent->power.child_count, -1, 0);
-		}
+	if (dev->parent) {
+		parent = dev->parent;
+		atomic_add_unless(&parent->power.child_count, -1, 0);
 	}
 	wake_up_all(&dev->power.wait_queue);
 
@@ -430,7 +451,7 @@ static int rpm_suspend(struct device *dev, int rpmflags)
 	}
 
  out:
-	dev_dbg(dev, "%s returns %d\n", __func__, retval);
+	trace_rpm_return_int(dev, _THIS_IP_, retval);
 
 	return retval;
 }
@@ -459,7 +480,7 @@ static int rpm_resume(struct device *dev, int rpmflags)
 	struct device *parent = NULL;
 	int retval = 0;
 
-	dev_dbg(dev, "%s flags 0x%x\n", __func__, rpmflags);
+	trace_rpm_resume(dev, rpmflags);
 
  repeat:
 	if (dev->power.runtime_error)
@@ -496,6 +517,15 @@ static int rpm_resume(struct device *dev, int rpmflags)
 			goto out;
 		}
 
+		if (dev->power.irq_safe) {
+			spin_unlock(&dev->power.lock);
+
+			cpu_relax();
+
+			spin_lock(&dev->power.lock);
+			goto repeat;
+		}
+
 		/* Wait for the operation carried out in parallel with us. */
 		for (;;) {
 			prepare_to_wait(&dev->power.wait_queue, &wait,
@@ -615,7 +645,7 @@ static int rpm_resume(struct device *dev, int rpmflags)
 		spin_lock_irq(&dev->power.lock);
 	}
 
-	dev_dbg(dev, "%s returns %d\n", __func__, retval);
+	trace_rpm_return_int(dev, _THIS_IP_, retval);
 
 	return retval;
 }
@@ -732,13 +762,16 @@ EXPORT_SYMBOL_GPL(pm_schedule_suspend);
  * return immediately if it is larger than zero.  Then carry out an idle
  * notification, either synchronous or asynchronous.
  *
- * This routine may be called in atomic context if the RPM_ASYNC flag is set.
+ * This routine may be called in atomic context if the RPM_ASYNC flag is set,
+ * or if pm_runtime_irq_safe() has been called.
  */
 int __pm_runtime_idle(struct device *dev, int rpmflags)
 {
 	unsigned long flags;
 	int retval;
 
+	might_sleep_if(!(rpmflags & RPM_ASYNC) && !dev->power.irq_safe);
+
 	if (rpmflags & RPM_GET_PUT) {
 		if (!atomic_dec_and_test(&dev->power.usage_count))
 			return 0;
@@ -761,13 +794,16 @@ EXPORT_SYMBOL_GPL(__pm_runtime_idle);
  * return immediately if it is larger than zero.  Then carry out a suspend,
  * either synchronous or asynchronous.
  *
- * This routine may be called in atomic context if the RPM_ASYNC flag is set.
+ * This routine may be called in atomic context if the RPM_ASYNC flag is set,
+ * or if pm_runtime_irq_safe() has been called.
  */
 int __pm_runtime_suspend(struct device *dev, int rpmflags)
 {
 	unsigned long flags;
 	int retval;
 
+	might_sleep_if(!(rpmflags & RPM_ASYNC) && !dev->power.irq_safe);
+
 	if (rpmflags & RPM_GET_PUT) {
 		if (!atomic_dec_and_test(&dev->power.usage_count))
 			return 0;
@@ -789,13 +825,16 @@ EXPORT_SYMBOL_GPL(__pm_runtime_suspend);
  * If the RPM_GET_PUT flag is set, increment the device's usage count.  Then
  * carry out a resume, either synchronous or asynchronous.
  *
- * This routine may be called in atomic context if the RPM_ASYNC flag is set.
+ * This routine may be called in atomic context if the RPM_ASYNC flag is set,
+ * or if pm_runtime_irq_safe() has been called.
  */
 int __pm_runtime_resume(struct device *dev, int rpmflags)
 {
 	unsigned long flags;
 	int retval;
 
+	might_sleep_if(!(rpmflags & RPM_ASYNC) && !dev->power.irq_safe);
+
 	if (rpmflags & RPM_GET_PUT)
 		atomic_inc(&dev->power.usage_count);
 
diff --git a/drivers/base/power/wakeup.c b/drivers/base/power/wakeup.c
index 84f7c7d5a09..14ee07e9cc4 100644
--- a/drivers/base/power/wakeup.c
+++ b/drivers/base/power/wakeup.c
@@ -276,7 +276,9 @@ EXPORT_SYMBOL_GPL(device_set_wakeup_capable);
  *
  * By default, most devices should leave wakeup disabled.  The exceptions are
  * devices that everyone expects to be wakeup sources: keyboards, power buttons,
- * possibly network interfaces, etc.
+ * possibly network interfaces, etc.  Also, devices that don't generate their
+ * own wakeup requests but merely forward requests from one bus to another
+ * (like PCI bridges) should have wakeup enabled by default.
  */
 int device_init_wakeup(struct device *dev, bool enable)
 {