4 files changed, 347 insertions, 212 deletions

diff --git a/drivers/base/power/Makefile b/drivers/base/power/Makefile
index 2639ae79a372..6488ce12f586 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
 obj-$(CONFIG_PM_SLEEP)	+= main.o wakeup.o
 obj-$(CONFIG_PM_RUNTIME)	+= runtime.o
 obj-$(CONFIG_PM_TRACE_RTC)	+= trace.o
diff --git a/drivers/base/power/clock_ops.c b/drivers/base/power/clock_ops.c
index b97294e2d95b..b876e60a53ef 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;
 }
 
@@ -130,15 +120,15 @@ 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 +137,12 @@ 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 +151,27 @@ 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 +179,28 @@ 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 +218,25 @@ 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 +247,25 @@ 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 +303,7 @@ 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 +338,22 @@ 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 +364,22 @@ 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 000000000000..29820c396182
--- /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 1c374579407c..22fe029ca212 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);
 	}
 }
 
@@ -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;