[PATCH 0/9] PM: Asynchronous suspend of devices

[PATCH 0/9] PM: Asynchronous suspend of devices

[Rafael J. Wysocki]

Hi,

The following series of patches implements asynchronous suspend and resume of
devices.

All of the patches have already been discussed, please refer to the changelogs
for details.

The patches are targeted at 2.6.33, so if there are no major objections, I'll
put them into the linux-next branch of the suspend-2.6 tree after .31-rc1 is
out.

Thanks,
Rafael

[PATCH 1/9] PM: Introduce PM links framework

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Introduce a framework for representing off-tree PM dependencies
between devices.

There are PM dependencies between devices that are not reflected by
the structure of the device tree.  In other words, as far as PM is
concerned, a device may depend on some other devices which are not
its children and none of which is its parent.

Every such dependency involves two devices, one of which is a
"master" and the other of which is a "slave", meaning that the
"slave" have to be suspended before the "master" and cannot be
woken up before it.  Thus every device can be given two lists of
"dependency objects", one for the dependencies where the device is
the "master" and the other for the dependencies where the device is
the "slave".  Then, each "dependency object" can be represented as

struct pm_link {
    struct device *master;
    struct list_head master_hook;
    struct device *slave;
    struct list_head slave_hook;
};

Add some synchronization, helpers for adding and removing
'struct pm_link' objects.

In addition to checking a device's parent, walk the list of its
"masters", in addition to walking the list of a device's children,
walk the list of its "slaves".

'struct pm_link' objects are created automatically for ACPI devices
and "regular" devices associated with them.  In the other cases such
objects will have to be added directly by platforms / bus types /
drivers etc.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/acpi/glue.c          |    3 
 drivers/base/core.c          |    4 
 drivers/base/power/Makefile  |    2 
 drivers/base/power/common.c  |  280 +++++++++++++++++++++++++++++++++++++++++++
 drivers/base/power/main.c    |   27 ----
 drivers/base/power/power.h   |   31 ++--
 drivers/base/power/runtime.c |    4 
 include/linux/pm.h           |    4 
 include/linux/pm_link.h      |   30 ++++
 9 files changed, 343 insertions(+), 42 deletions(-)

Index: linux-2.6/include/linux/pm.h
===================================================================
--- linux-2.6.orig/include/linux/pm.h
+++ linux-2.6/include/linux/pm.h
@@ -408,6 +408,9 @@ enum rpm_request {
 };
 
 struct dev_pm_info {
+	spinlock_t		lock;
+	struct list_head	master_links;
+	struct list_head	slave_links;
 	pm_message_t		power_state;
 	unsigned int		can_wakeup:1;
 	unsigned int		should_wakeup:1;
@@ -420,7 +423,6 @@ struct dev_pm_info {
 	unsigned long		timer_expires;
 	struct work_struct	work;
 	wait_queue_head_t	wait_queue;
-	spinlock_t		lock;
 	atomic_t		usage_count;
 	atomic_t		child_count;
 	unsigned int		disable_depth:3;
Index: linux-2.6/drivers/base/power/power.h
===================================================================
--- linux-2.6.orig/drivers/base/power/power.h
+++ linux-2.6/drivers/base/power/power.h
@@ -1,3 +1,7 @@
+extern void device_pm_init(struct device *dev);
+extern void device_pm_add(struct device *dev);
+extern void device_pm_remove(struct device *dev);
+
 #ifdef CONFIG_PM_RUNTIME
 
 extern void pm_runtime_init(struct device *dev);
@@ -23,7 +27,8 @@ static inline struct device *to_device(s
 	return container_of(entry, struct device, power.entry);
 }
 
-extern void device_pm_init(struct device *dev);
+extern void device_pm_list_add(struct device *dev);
+extern void device_pm_list_remove(struct device *dev);
 extern void device_pm_add(struct device *);
 extern void device_pm_remove(struct device *);
 extern void device_pm_move_before(struct device *, struct device *);
@@ -32,17 +37,8 @@ extern void device_pm_move_last(struct d
 
 #else /* !CONFIG_PM_SLEEP */
 
-static inline void device_pm_init(struct device *dev)
-{
-	pm_runtime_init(dev);
-}
-
-static inline void device_pm_remove(struct device *dev)
-{
-	pm_runtime_remove(dev);
-}
-
-static inline void device_pm_add(struct device *dev) {}
+static inline void device_pm_list_add(struct device *dev) {}
+static inline void device_pm_list_remove(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,
@@ -60,7 +56,11 @@ static inline void device_pm_move_last(s
 extern int dpm_sysfs_add(struct device *);
 extern void dpm_sysfs_remove(struct device *);
 
-#else /* CONFIG_PM */
+/* drivers/base/power/link.c */
+extern int pm_link_init(void);
+extern void pm_link_remove_all(struct device *dev);
+
+#else /* !CONFIG_PM */
 
 static inline int dpm_sysfs_add(struct device *dev)
 {
@@ -71,4 +71,7 @@ static inline void dpm_sysfs_remove(stru
 {
 }
 
-#endif
+static inline int pm_link_init(void) { return 0; }
+static inline void pm_link_remove_all(struct device *dev) {}
+
+#endif  /* !CONFIG_PM */
Index: linux-2.6/drivers/base/core.c
===================================================================
--- linux-2.6.orig/drivers/base/core.c
+++ linux-2.6/drivers/base/core.c
@@ -1252,9 +1252,13 @@ int __init devices_init(void)
 	sysfs_dev_char_kobj = kobject_create_and_add("char", dev_kobj);
 	if (!sysfs_dev_char_kobj)
 		goto char_kobj_err;
+	if (pm_link_init())
+		goto pm_link_err;
 
 	return 0;
 
+ pm_link_err:
+	kobject_put(sysfs_dev_char_kobj);
  char_kobj_err:
 	kobject_put(sysfs_dev_block_kobj);
  block_kobj_err:
Index: linux-2.6/include/linux/pm_link.h
===================================================================
--- /dev/null
+++ linux-2.6/include/linux/pm_link.h
@@ -0,0 +1,30 @@
+/*
+ * include/linux/pm_link.h - PM links manipulation core.
+ *
+ * Copyright (c) 2009 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc.
+ *
+ * This file is released under the GPLv2.
+ */
+
+#ifndef _LINUX_PM_LINK_H
+#define _LINUX_PM_LINK_H
+
+#include <linux/list.h>
+
+struct device;
+
+struct pm_link {
+	struct device *master;
+	struct list_head master_hook;
+	struct device *slave;
+	struct list_head slave_hook;
+};
+
+extern int pm_link_add(struct device *slave, struct device *master);
+extern void pm_link_remove(struct device *dev, struct device *master);
+extern int device_for_each_master(struct device *slave, void *data,
+			   int (*fn)(struct device *dev, void *data));
+extern int device_for_each_slave(struct device *master, void *data,
+			  int (*fn)(struct device *dev, void *data));
+
+#endif
Index: linux-2.6/drivers/base/power/Makefile
===================================================================
--- linux-2.6.orig/drivers/base/power/Makefile
+++ linux-2.6/drivers/base/power/Makefile
@@ -1,4 +1,4 @@
-obj-$(CONFIG_PM)	+= sysfs.o
+obj-$(CONFIG_PM)	+= sysfs.o common.o
 obj-$(CONFIG_PM_SLEEP)	+= main.o
 obj-$(CONFIG_PM_RUNTIME)	+= runtime.o
 obj-$(CONFIG_PM_TRACE_RTC)	+= trace.o
Index: linux-2.6/drivers/base/power/runtime.c
===================================================================
--- linux-2.6.orig/drivers/base/power/runtime.c
+++ linux-2.6/drivers/base/power/runtime.c
@@ -972,8 +972,6 @@ EXPORT_SYMBOL_GPL(pm_runtime_enable);
  */
 void pm_runtime_init(struct device *dev)
 {
-	spin_lock_init(&dev->power.lock);
-
 	dev->power.runtime_status = RPM_SUSPENDED;
 	dev->power.idle_notification = false;
 
@@ -993,8 +991,6 @@ void pm_runtime_init(struct device *dev)
 	dev->power.timer_expires = 0;
 	setup_timer(&dev->power.suspend_timer, pm_suspend_timer_fn,
 			(unsigned long)dev);
-
-	init_waitqueue_head(&dev->power.wait_queue);
 }
 
 /**
Index: linux-2.6/drivers/base/power/main.c
===================================================================
--- linux-2.6.orig/drivers/base/power/main.c
+++ linux-2.6/drivers/base/power/main.c
@@ -21,6 +21,7 @@
 #include <linux/kallsyms.h>
 #include <linux/mutex.h>
 #include <linux/pm.h>
+#include <linux/pm_link.h>
 #include <linux/pm_runtime.h>
 #include <linux/resume-trace.h>
 #include <linux/rwsem.h>
@@ -50,16 +51,6 @@ static DEFINE_MUTEX(dpm_list_mtx);
 static bool transition_started;
 
 /**
- * device_pm_init - Initialize the PM-related part of a device object.
- * @dev: Device object being initialized.
- */
-void device_pm_init(struct device *dev)
-{
-	dev->power.status = DPM_ON;
-	pm_runtime_init(dev);
-}
-
-/**
  * device_pm_lock - Lock the list of active devices used by the PM core.
  */
 void device_pm_lock(void)
@@ -76,14 +67,11 @@ void device_pm_unlock(void)
 }
 
 /**
- * device_pm_add - Add a device to the PM core's list of active devices.
+ * device_pm_list_add - Add a device to the PM core's list of active devices.
  * @dev: Device to add to the list.
  */
-void device_pm_add(struct device *dev)
+void device_pm_list_add(struct device *dev)
 {
-	pr_debug("PM: Adding info for %s:%s\n",
-		 dev->bus ? dev->bus->name : "No Bus",
-		 kobject_name(&dev->kobj));
 	mutex_lock(&dpm_list_mtx);
 	if (dev->parent) {
 		if (dev->parent->power.status >= DPM_SUSPENDING)
@@ -97,24 +85,19 @@ void device_pm_add(struct device *dev)
 		 */
 		dev_WARN(dev, "Parentless device registered during a PM transaction\n");
 	}
-
 	list_add_tail(&dev->power.entry, &dpm_list);
 	mutex_unlock(&dpm_list_mtx);
 }
 
 /**
- * device_pm_remove - Remove a device from the PM core's list of active devices.
+ * device_pm_list_remove - Remove a device from the PM core's list of devices.
  * @dev: Device to be removed from the list.
  */
-void device_pm_remove(struct device *dev)
+void device_pm_list_remove(struct device *dev)
 {
-	pr_debug("PM: Removing info for %s:%s\n",
-		 dev->bus ? dev->bus->name : "No Bus",
-		 kobject_name(&dev->kobj));
 	mutex_lock(&dpm_list_mtx);
 	list_del_init(&dev->power.entry);
 	mutex_unlock(&dpm_list_mtx);
-	pm_runtime_remove(dev);
 }
 
 /**
Index: linux-2.6/drivers/base/power/common.c
===================================================================
--- /dev/null
+++ linux-2.6/drivers/base/power/common.c
@@ -0,0 +1,280 @@
+/*
+ * drivers/base/power/common.c - device PM common functions.
+ *
+ * Copyright (c) 2009 Rafael J. Wysocki <rjw@sisk.pl>, Novell Inc.
+ *
+ * This file is released under the GPLv2.
+ */
+
+#include <linux/rculist.h>
+#include <linux/device.h>
+#include <linux/srcu.h>
+#include <linux/pm_link.h>
+
+#include "power.h"
+
+/**
+ * device_pm_init - Initialize the PM part of a device object.
+ * @dev: Device object being initialized.
+ */
+void device_pm_init(struct device *dev)
+{
+	dev->power.status = DPM_ON;
+	spin_lock_init(&dev->power.lock);
+	INIT_LIST_HEAD(&dev->power.master_links);
+	INIT_LIST_HEAD(&dev->power.slave_links);
+	pm_runtime_init(dev);
+}
+
+/**
+ * device_pm_add - Handle the PM part of a device added to device tree.
+ * @dev: Device object being added to device tree.
+ */
+void device_pm_add(struct device *dev)
+{
+	pr_debug("PM: Adding info for %s:%s\n",
+		 dev->bus ? dev->bus->name : "No Bus",
+		 kobject_name(&dev->kobj));
+	device_pm_list_add(dev);
+}
+
+/**
+ * device_pm_remove - Handle the PM part of a device removed from device tree.
+ * @dev: Device object being removed from device tree.
+ */
+void device_pm_remove(struct device *dev)
+{
+	pr_debug("PM: Removing info for %s:%s\n",
+		 dev->bus ? dev->bus->name : "No Bus",
+		 kobject_name(&dev->kobj));
+	device_pm_list_remove(dev);
+	pm_runtime_remove(dev);
+	pm_link_remove_all(dev);
+}
+
+/*
+ * PM links framework.
+ *
+ * There are PM dependencies between devices that are not reflected by the
+ * structure of the device tree.  In other words, as far as PM is concerned, a
+ * device may depend on some other devices which are not its children and none
+ * of which is its parent.
+ *
+ * Every such dependency involves two devices, one of which is a "master" and
+ * the other of which is a "slave", meaning that the "slave" have to be
+ * suspended before the "master" and cannot be woken up before it.  Thus every
+ * device can be given two lists of "dependency objects", one for the
+ * dependencies where the device is the "master" and the other for the
+ * dependencies where the device is the "slave".  Then, each "dependency object"
+ * can be represented as 'struct pm_link' as defined in include/linux/pm_link.h.
+ *
+ * The PM links of a device can help decide when the device should be suspended
+ * or resumed.  Namely, In addition to checking the device's parent, the PM core
+ * can walk the list of its "masters" and check their PM status.  Similarly, in
+ * addition to walking the list of a device's children, the PM core can walk the
+ * list of its "slaves".
+ */
+
+static struct srcu_struct pm_link_ss;
+static DEFINE_MUTEX(pm_link_mtx);
+
+/**
+ * pm_link_add - Create a PM link object connecting two devices.
+ * @slave: Device to be the slave in this link.
+ * @master: Device to be the master in this link.
+ */
+int pm_link_add(struct device *slave, struct device *master)
+{
+	struct pm_link *link;
+	int error = -ENODEV;
+
+	if (!get_device(master))
+		return error;
+
+	if (!get_device(slave))
+		goto err_slave;
+
+	link = kzalloc(sizeof(*link), GFP_KERNEL);
+	if (!link)
+		goto err_link;
+
+	dev_dbg(slave, "PM: Creating PM link to (master) %s %s\n",
+		dev_driver_string(master), dev_name(master));
+
+	link->master = master;
+	INIT_LIST_HEAD(&link->master_hook);
+	link->slave = slave;
+	INIT_LIST_HEAD(&link->slave_hook);
+
+	spin_lock_irq(&master->power.lock);
+	list_add_tail_rcu(&link->master_hook, &master->power.master_links);
+	spin_unlock_irq(&master->power.lock);
+
+	spin_lock_irq(&slave->power.lock);
+	list_add_tail_rcu(&link->slave_hook, &slave->power.slave_links);
+	spin_unlock_irq(&slave->power.lock);
+
+	return 0;
+
+ err_link:
+	error = -ENOMEM;
+	put_device(slave);
+
+ err_slave:
+	put_device(master);
+
+	return error;
+}
+EXPORT_SYMBOL_GPL(pm_link_add);
+
+/**
+ * __pm_link_remove - Remove a PM link object.
+ * @link: PM link object to remove
+ */
+static void __pm_link_remove(struct pm_link *link)
+{
+	struct device *master = link->master;
+	struct device *slave = link->slave;
+
+	dev_dbg(slave, "PM: Removing PM link to (master) %s %s\n",
+		dev_driver_string(master), dev_name(master));
+
+	spin_lock_irq(&master->power.lock);
+	list_del_rcu(&link->master_hook);
+	spin_unlock_irq(&master->power.lock);
+
+	spin_lock_irq(&slave->power.lock);
+	list_del_rcu(&link->slave_hook);
+	spin_unlock_irq(&slave->power.lock);
+
+	synchronize_srcu(&pm_link_ss);
+
+	kfree(link);
+
+	put_device(master);
+	put_device(slave);
+}
+
+/**
+ * pm_link_remove_all - Remove all PM link objects for given device.
+ * @dev: Device to handle.
+ */
+void pm_link_remove_all(struct device *dev)
+{
+	struct pm_link *link, *n;
+
+	mutex_lock(&pm_link_mtx);
+
+	list_for_each_entry_safe(link, n, &dev->power.master_links, master_hook)
+		__pm_link_remove(link);
+
+	list_for_each_entry_safe(link, n, &dev->power.slave_links, slave_hook)
+		__pm_link_remove(link);
+
+	mutex_unlock(&pm_link_mtx);
+}
+
+/**
+ * pm_link_remove - Remove a PM link object connecting two devices.
+ * @dev: Slave device of the PM link to remove.
+ * @master: Master device of the PM link to remove.
+ */
+void pm_link_remove(struct device *dev, struct device *master)
+{
+	struct pm_link *link, *n;
+
+	mutex_lock(&pm_link_mtx);
+
+	list_for_each_entry_safe(link, n, &dev->power.slave_links, slave_hook) {
+		if (link->master != master)
+			continue;
+
+		__pm_link_remove(link);
+		break;
+	}
+
+	mutex_unlock(&pm_link_mtx);
+}
+EXPORT_SYMBOL_GPL(pm_link_remove);
+
+/**
+ * device_for_each_master - Execute given function for each master of a device.
+ * @slave: Device whose masters to execute the function for.
+ * @data: Data pointer to pass to the function.
+ * @fn: Function to execute for each master of @slave.
+ *
+ * The function is executed for the parent of the device, if there is one, and
+ * for each device connected to it via a pm_link object where @slave is the
+ * "slave".
+ */
+int device_for_each_master(struct device *slave, void *data,
+			   int (*fn)(struct device *dev, void *data))
+{
+	struct pm_link *link;
+	int idx;
+	int error = 0;
+
+	if (slave->parent) {
+		error = fn(slave->parent, data);
+		if (error)
+			return error;
+	}
+
+	idx = srcu_read_lock(&pm_link_ss);
+
+	list_for_each_entry_rcu(link, &slave->power.slave_links, slave_hook) {
+		struct device *master = link->master;
+
+		error = fn(master, data);
+		if (error)
+			break;
+	}
+
+	srcu_read_unlock(&pm_link_ss, idx);
+
+	return error;
+}
+EXPORT_SYMBOL_GPL(device_for_each_master);
+
+/**
+ * device_for_each_slave - Execute given function for each slave of a device.
+ * @master: Device whose slaves to execute the function for.
+ * @data: Data pointer to pass to the function.
+ * @fn: Function to execute for each slave of @master.
+ *
+ * The function is executed for all children of the device, if there are any,
+ * and for each device connected to it via a pm_link object where @master is the
+ * "master".
+ */
+int device_for_each_slave(struct device *master, void *data,
+			  int (*fn)(struct device *dev, void *data))
+{
+	struct pm_link *link;
+	int idx;
+	int error;
+
+	error = device_for_each_child(master, data, fn);
+	if (error)
+		return error;
+
+	idx = srcu_read_lock(&pm_link_ss);
+
+	list_for_each_entry_rcu(link, &master->power.master_links,
+				master_hook) {
+		struct device *slave = link->slave;
+
+		error = fn(slave, data);
+		if (error)
+			break;
+	}
+
+	srcu_read_unlock(&pm_link_ss, idx);
+
+	return error;
+}
+EXPORT_SYMBOL_GPL(device_for_each_slave);
+
+int __init pm_link_init(void)
+{
+	return init_srcu_struct(&pm_link_ss);
+}
Index: linux-2.6/drivers/acpi/glue.c
===================================================================
--- linux-2.6.orig/drivers/acpi/glue.c
+++ linux-2.6/drivers/acpi/glue.c
@@ -11,6 +11,7 @@
 #include <linux/device.h>
 #include <linux/rwsem.h>
 #include <linux/acpi.h>
+#include <linux/pm_link.h>
 
 #define ACPI_GLUE_DEBUG	0
 #if ACPI_GLUE_DEBUG
@@ -170,6 +171,7 @@ static int acpi_bind_one(struct device *
 			device_set_wakeup_enable(dev,
 						acpi_dev->wakeup.state.enabled);
 		}
+		pm_link_add(dev, &acpi_dev->dev);
 	}
 
 	return 0;
@@ -189,6 +191,7 @@ static int acpi_unbind_one(struct device
 					&acpi_dev)) {
 			sysfs_remove_link(&dev->kobj, "firmware_node");
 			sysfs_remove_link(&acpi_dev->dev.kobj, "physical_node");
+			pm_link_remove(dev, &acpi_dev->dev);
 		}
 
 		acpi_detach_data(dev->archdata.acpi_handle,

[PATCH 2/9] PM: Asynchronous resume of devices

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Theoretically, the total time of system sleep transitions (suspend
to RAM, hibernation) can be reduced by running suspend and resume
callbacks of device drivers in parallel with each other.  However,
there are dependencies between devices such that, for example, we may
not be allowed to put one device into a low power state before
anohter one has been suspended (e.g. we cannot suspend a bridge
before suspending all devices behind it).  In particular, we're not
allowed to suspend the parent of a device before suspending the
device itself.  Analogously, we're not allowed to resume a device
before resuming its parent.

Thus, to make it possible to execute suspend and resume callbacks
provided by device drivers in parallel with each other, we need to
provide a synchronization mechanism preventing the dependencies
between devices from being violated.

The patch below allows some devices to be resumed asynchronously,
with the following rules:
(1) There is a wait queue head, dev->power.wait_queue, and an
    "operation complete" flag, dev->power.op_complete for each device
    object.
(2) All of the power.op_complete flags are reset before suspend as
    well as after each resume stage (dpm_resume_noirq(),
    dpm_resume()).
(3) If power.async_suspend is set for dev or for one of devices it
    depends on, the PM core waits for the "master" device's
    power.op_complete flag to be set before attempting to run the
    resume callbacks, appropriate for this particular stage of
    resume, for dev.
(4) dev->power.op_complete is set for each device after running its
    resume callbacks at each stage of resume (dpm_resume_noirq(),
    dpm_resume()) and all threads waiting in the devices wait
    queue are woken up.

With this mechanism in place, the drivers wanting their resume
callbacks to be executed asynchronously can set
dev->power.async_suspend for them, with the help of
device_enable_async_suspend().  In addition to that, the PM off-tree
dependencies between devices have to be represented by
'struct pm_link' objects introduced by the previous patch.

In this version of the patch the async threads started to execute
the resume callbacks of specific device don't exit immediately having
done that, but search dpm_list for devices whose PM dependencies have
already been satisfied and execute their callbacks without waiting.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/base/power/common.c |    5 
 drivers/base/power/main.c   |  327 ++++++++++++++++++++++++++++++++++++++++++--
 include/linux/device.h      |    6 
 include/linux/pm.h          |    6 
 4 files changed, 329 insertions(+), 15 deletions(-)

Index: linux-2.6/include/linux/pm.h
===================================================================
--- linux-2.6.orig/include/linux/pm.h
+++ linux-2.6/include/linux/pm.h
@@ -26,6 +26,7 @@
 #include <linux/spinlock.h>
 #include <linux/wait.h>
 #include <linux/timer.h>
+#include <linux/completion.h>
 
 /*
  * Callbacks for platform drivers to implement.
@@ -409,20 +410,23 @@ enum rpm_request {
 
 struct dev_pm_info {
 	spinlock_t		lock;
+	wait_queue_head_t	wait_queue;
 	struct list_head	master_links;
 	struct list_head	slave_links;
 	pm_message_t		power_state;
 	unsigned int		can_wakeup:1;
 	unsigned int		should_wakeup:1;
+	unsigned int		async_suspend:1;
 	enum dpm_state		status;		/* Owned by the PM core */
 #ifdef CONFIG_PM_SLEEP
 	struct list_head	entry;
+	unsigned int		op_started:1;
+	unsigned int		op_complete:1;
 #endif
 #ifdef CONFIG_PM_RUNTIME
 	struct timer_list	suspend_timer;
 	unsigned long		timer_expires;
 	struct work_struct	work;
-	wait_queue_head_t	wait_queue;
 	atomic_t		usage_count;
 	atomic_t		child_count;
 	unsigned int		disable_depth:3;
Index: linux-2.6/include/linux/device.h
===================================================================
--- linux-2.6.orig/include/linux/device.h
+++ linux-2.6/include/linux/device.h
@@ -472,6 +472,12 @@ static inline int device_is_registered(s
 	return dev->kobj.state_in_sysfs;
 }
 
+static inline void device_enable_async_suspend(struct device *dev, bool enable)
+{
+	if (dev->power.status == DPM_ON)
+		dev->power.async_suspend = enable;
+}
+
 void driver_init(void);
 
 /*
Index: linux-2.6/drivers/base/power/main.c
===================================================================
--- linux-2.6.orig/drivers/base/power/main.c
+++ linux-2.6/drivers/base/power/main.c
@@ -26,6 +26,8 @@
 #include <linux/resume-trace.h>
 #include <linux/rwsem.h>
 #include <linux/interrupt.h>
+#include <linux/async.h>
+#include <linux/completion.h>
 
 #include "../base.h"
 #include "power.h"
@@ -43,6 +45,7 @@
 LIST_HEAD(dpm_list);
 
 static DEFINE_MUTEX(dpm_list_mtx);
+static pm_message_t pm_transition;
 
 /*
  * Set once the preparation of devices for a PM transition has started, reset
@@ -145,6 +148,131 @@ void device_pm_move_last(struct device *
 }
 
 /**
+ * dpm_reset - Clear power.op_started and power.op_complete for given device.
+ * @dev: Device to handle.
+ */
+static void dpm_reset(struct device *dev)
+{
+	dev->power.op_started = false;
+	dev->power.op_complete = false;
+}
+
+/**
+ * dpm_reset_all - Call dpm_reset() for all devices.
+ */
+static void dpm_reset_all(void)
+{
+	struct device *dev;
+
+	list_for_each_entry(dev, &dpm_list, power.entry)
+		dpm_reset(dev);
+}
+
+/**
+ * dpm_synchronize_noirq - Wait for "late" or "early" PM callbacks to complete.
+ *
+ * Wait for the "late" or "early" suspend/resume callbacks of all devices to
+ * complete and clear power.op_started and power.op_complete for all devices.
+ */
+static void dpm_synchronize_noirq(void)
+{
+	async_synchronize_full();
+	dpm_reset_all();
+}
+
+/**
+ * dpm_synchronize_noirq - Wait for PM callbacks to complete.
+ *
+ * Wait for the "regular" suspend/resume callbacks of all devices to complete
+ * and clear power.op_started and power.op_complete for all devices.
+ */
+static void dpm_synchronize(void)
+{
+	async_synchronize_full();
+	mutex_lock(&dpm_list_mtx);
+	dpm_reset_all();
+	mutex_unlock(&dpm_list_mtx);
+}
+
+/**
+ * device_pm_wait - Wait for a PM operation to complete.
+ * @sub: "Slave" device.
+ * @dev: Device to wait for.
+ *
+ * Wait for a PM operation carried out for @dev to complete, unless both @sub
+ * and @dev have to be handled synchronously (in such a case they are going to
+ * be handled in the right order anyway thanks to the pm_list ordering).
+ */
+static void device_pm_wait(struct device *sub, struct device *dev)
+{
+	if (!dev)
+		return;
+
+	if (!(sub->power.async_suspend || dev->power.async_suspend))
+		return;
+
+	if (!dev->power.op_complete) {
+		dev_dbg(sub, "PM: Waiting for %s %s\n", dev_driver_string(dev),
+			dev_name(dev));
+		wait_event(dev->power.wait_queue, !!dev->power.op_complete);
+	}
+}
+
+/**
+ * device_pm_wait_fn - Wrapper for device_pm_wait().
+ * @dev: Device to wait for.
+ * @data: Pointer to the "slave" device object.
+ */
+static int device_pm_wait_fn(struct device *dev, void *data)
+{
+	device_pm_wait((struct device *)data, dev);
+	return 0;
+}
+
+/**
+ * device_pm_wait_for_masters - Wait for all masters of given device.
+ * @slave: Device to wait for the masters of.
+ */
+static void device_pm_wait_for_masters(struct device *slave)
+{
+	if (!pm_trace_enabled)
+		device_for_each_master(slave, slave, device_pm_wait_fn);
+}
+
+/**
+ * device_pm_check - Check the power.op_complete flag of given device.
+ * @dev: Device to check.
+ */
+static bool device_pm_check(struct device *dev)
+{
+	int ret = 0;
+
+	if (dev)
+		ret = !dev->power.op_complete;
+
+	return ret;
+}
+
+/**
+ * device_pm_check_fn - Wrapper for device_pm_check().
+ * @dev: Device to check.
+ * @data: Ignored.
+ */
+static int device_pm_check_fn(struct device *dev, void *data)
+{
+	return device_pm_check(dev);
+}
+
+/**
+ * device_pm_check_masters - Check power.op_complete for masters of a device.
+ * @slave: Device to check the masters of.
+ */
+static int device_pm_check_masters(struct device *slave)
+{
+	return device_for_each_master(slave, NULL, device_pm_check_fn);
+}
+
+/**
  * pm_op - Execute the PM operation appropriate for given PM event.
  * @dev: Device to handle.
  * @ops: PM operations to choose from.
@@ -269,6 +397,24 @@ static int pm_noirq_op(struct device *de
 	return error;
 }
 
+/**
+ * pm_op_started - Mark the beginning of a PM operation for given device.
+ * @dev: Device to handle.
+ */
+static bool pm_op_started(struct device *dev)
+{
+	bool ret = false;
+
+	spin_lock_irq(&dev->power.lock);
+	if (dev->power.op_started)
+		ret = true;
+	else
+		dev->power.op_started = true;
+	spin_unlock_irq(&dev->power.lock);
+
+	return ret;
+}
+
 static char *pm_verb(int event)
 {
 	switch (event) {
@@ -310,33 +456,102 @@ static void pm_dev_err(struct device *de
 /*------------------------- Resume routines -------------------------*/
 
 /**
- * device_resume_noirq - Execute an "early resume" callback for given device.
+ * __device_resume_noirq - Execute an "early resume" callback for given device.
  * @dev: Device to handle.
  * @state: PM transition of the system being carried out.
  *
  * The driver of @dev will not receive interrupts while this function is being
  * executed.
  */
-static int device_resume_noirq(struct device *dev, pm_message_t state)
+static int __device_resume_noirq(struct device *dev, pm_message_t state)
 {
 	int error = 0;
 
 	TRACE_DEVICE(dev);
 	TRACE_RESUME(0);
 
-	if (!dev->bus)
-		goto End;
-
-	if (dev->bus->pm) {
+	if (dev->bus && dev->bus->pm) {
 		pm_dev_dbg(dev, state, "EARLY ");
 		error = pm_noirq_op(dev, dev->bus->pm, state);
 	}
- End:
+
+	dev->power.op_complete = true;
+	wake_up_all(&dev->power.wait_queue);
+
 	TRACE_RESUME(error);
 	return error;
 }
 
 /**
+ * async_device_resume_noirq - Wrapper of __device_resume_noirq().
+ * @dev: Device to resume.
+ */
+static void async_device_resume_noirq(struct device *dev)
+{
+	int error;
+
+	pm_dev_dbg(dev, pm_transition, "async EARLY ");
+	error = __device_resume_noirq(dev, pm_transition);
+	if (error)
+		pm_dev_err(dev, pm_transition, " async EARLY", error);
+}
+
+/**
+ * async_resume_noirq - Execute "early" resume callbacks asynchronously.
+ * @data: Pointer to the first device to resume.
+ * @cookie: Ignored.
+ *
+ * The execution of this function is scheduled with async_schedule(), so it runs
+ * in its own kernel thread.  It first calls the "early" resume callback for the
+ * device passed to it as @data.  Next, it walks dpm_list looking for devices
+ * that can be resumed without waiting for their "masters".  If such a device is
+ * found, its "early" resume callback is run.
+ */
+static void async_resume_noirq(void *data, async_cookie_t cookie)
+{
+	struct device *dev = (struct device *)data;
+
+	device_pm_wait_for_masters(dev);
+	async_device_resume_noirq(dev);
+
+	list_for_each_entry_continue(dev, &dpm_list, power.entry) {
+		if (!dev->power.async_suspend || dev->power.status <= DPM_OFF)
+			continue;
+
+		if (device_pm_check_masters(dev))
+			continue;
+
+		if (pm_op_started(dev))
+			continue;
+
+		pm_dev_dbg(dev, pm_transition, "out of order EARLY ");
+		async_device_resume_noirq(dev);
+	}
+}
+
+/**
+ * device_resume_noirq - Execute or schedule "early" resume callback.
+ * @dev: Device to resume.
+ *
+ * If @dev can be resumed asynchronously, schedule the execution of
+ * async_resume_noirq() for it.  Otherwise, execute its "early" resume callback
+ * directly.
+ */
+static int device_resume_noirq(struct device *dev)
+{
+	if (pm_op_started(dev))
+		return 0;
+
+	if (dev->power.async_suspend && !pm_trace_enabled) {
+		async_schedule(async_resume_noirq, dev);
+		return 0;
+	}
+
+	device_pm_wait_for_masters(dev);
+	return __device_resume_noirq(dev, pm_transition);
+}
+
+/**
  * dpm_resume_noirq - Execute "early resume" callbacks for non-sysdev devices.
  * @state: PM transition of the system being carried out.
  *
@@ -349,26 +564,28 @@ void dpm_resume_noirq(pm_message_t state
 
 	mutex_lock(&dpm_list_mtx);
 	transition_started = false;
+	pm_transition = state;
 	list_for_each_entry(dev, &dpm_list, power.entry)
 		if (dev->power.status > DPM_OFF) {
 			int error;
 
 			dev->power.status = DPM_OFF;
-			error = device_resume_noirq(dev, state);
+			error = device_resume_noirq(dev);
 			if (error)
-				pm_dev_err(dev, state, " early", error);
+				pm_dev_err(dev, state, " EARLY", error);
 		}
+	dpm_synchronize_noirq();
 	mutex_unlock(&dpm_list_mtx);
 	resume_device_irqs();
 }
 EXPORT_SYMBOL_GPL(dpm_resume_noirq);
 
 /**
- * device_resume - Execute "resume" callbacks for given device.
+ * __device_resume - Execute "resume" callbacks for given device.
  * @dev: Device to handle.
  * @state: PM transition of the system being carried out.
  */
-static int device_resume(struct device *dev, pm_message_t state)
+static int __device_resume(struct device *dev, pm_message_t state)
 {
 	int error = 0;
 
@@ -409,12 +626,92 @@ static int device_resume(struct device *
 	}
  End:
 	up(&dev->sem);
+	dev->power.op_complete = true;
+	wake_up_all(&dev->power.wait_queue);
 
 	TRACE_RESUME(error);
 	return error;
 }
 
 /**
+ * async_device_resume - Wrapper of __device_resume().
+ * @dev: Device to resume.
+ */
+static void async_device_resume(struct device *dev)
+{
+	int error;
+
+	pm_dev_dbg(dev, pm_transition, "async ");
+	error = __device_resume(dev, pm_transition);
+	if (error)
+		pm_dev_err(dev, pm_transition, " async", error);
+}
+
+/**
+ * async_resume - Execute resume callbacks asynchronously.
+ * @data: Pointer to the first device to resume.
+ * @cookie: Ignored.
+ *
+ * The execution of this function is scheduled with async_schedule(), so it runs
+ * in its own kernel thread.  It first calls the resume callbacks for the device
+ * passed to it as @data.  Next, it walks dpm_list looking for devices that can
+ * be resumed without waiting for their "masters".  If such a device is found,
+ * its resume callbacks are run.
+ */
+static void async_resume(void *data, async_cookie_t cookie)
+{
+	struct device *dev = (struct device *)data;
+
+	device_pm_wait_for_masters(dev);
+
+ repeat:
+	async_device_resume(dev);
+	put_device(dev);
+
+	mutex_lock(&dpm_list_mtx);
+	if (dev->power.status < DPM_OFF)
+		dev = to_device(dpm_list.next);
+	list_for_each_entry_continue(dev, &dpm_list, power.entry) {
+		if (!dev->power.async_suspend || dev->power.status < DPM_OFF)
+			continue;
+
+		if (device_pm_check_masters(dev))
+			continue;
+
+		if (pm_op_started(dev))
+			continue;
+
+		get_device(dev);
+		mutex_unlock(&dpm_list_mtx);
+		pm_dev_dbg(dev, pm_transition, "out of order ");
+		goto repeat;
+	}
+	mutex_unlock(&dpm_list_mtx);
+}
+
+/**
+ * device_resume - Execute or schedule resume callbacks for given device.
+ * @dev: Device to resume.
+ *
+ * If @dev can be resumed asynchronously, schedule the execution of
+ * async_resume() for it.  Otherwise, execute its resume callbacks directly.
+ */
+static int device_resume(struct device *dev)
+{
+	if (pm_op_started(dev))
+		return 0;
+
+	if (dev->power.async_suspend && !pm_trace_enabled) {
+		get_device(dev);
+		async_schedule(async_resume, dev);
+		return 0;
+	}
+
+	device_pm_wait_for_masters(dev);
+	return __device_resume(dev, pm_transition);
+}
+
+/**
  * dpm_resume - Execute "resume" callbacks for non-sysdev devices.
  * @state: PM transition of the system being carried out.
  *
@@ -427,6 +724,7 @@ static void dpm_resume(pm_message_t stat
 
 	INIT_LIST_HEAD(&list);
 	mutex_lock(&dpm_list_mtx);
+	pm_transition = state;
 	while (!list_empty(&dpm_list)) {
 		struct device *dev = to_device(dpm_list.next);
 
@@ -437,7 +735,7 @@ static void dpm_resume(pm_message_t stat
 			dev->power.status = DPM_RESUMING;
 			mutex_unlock(&dpm_list_mtx);
 
-			error = device_resume(dev, state);
+			error = device_resume(dev);
 
 			mutex_lock(&dpm_list_mtx);
 			if (error)
@@ -452,6 +750,7 @@ static void dpm_resume(pm_message_t stat
 	}
 	list_splice(&list, &dpm_list);
 	mutex_unlock(&dpm_list_mtx);
+	dpm_synchronize();
 }
 
 /**
@@ -775,8 +1074,10 @@ static int dpm_prepare(pm_message_t stat
 			break;
 		}
 		dev->power.status = DPM_SUSPENDING;
-		if (!list_empty(&dev->power.entry))
+		if (!list_empty(&dev->power.entry)) {
 			list_move_tail(&dev->power.entry, &list);
+			dpm_reset(dev);
+		}
 		put_device(dev);
 	}
 	list_splice(&list, &dpm_list);
Index: linux-2.6/drivers/base/power/common.c
===================================================================
--- linux-2.6.orig/drivers/base/power/common.c
+++ linux-2.6/drivers/base/power/common.c
@@ -19,10 +19,11 @@
  */
 void device_pm_init(struct device *dev)
 {
-	dev->power.status = DPM_ON;
 	spin_lock_init(&dev->power.lock);
+	init_waitqueue_head(&dev->power.wait_queue);
 	INIT_LIST_HEAD(&dev->power.master_links);
 	INIT_LIST_HEAD(&dev->power.slave_links);
+	dev->power.status = DPM_ON;
 	pm_runtime_init(dev);
 }
 
@@ -117,6 +118,8 @@ int pm_link_add(struct device *slave, st
 	return 0;
 
  err_link:
+	master->power.async_suspend = false;
+	slave->power.async_suspend = false;
 	error = -ENOMEM;
 	put_device(slave);
 

[PATCH 3/9] PM: Asynchronous suspend of devices

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Extend the approach used in the previous patch to the suspend part of
the PM core.

Asynchronous suspend is slightly more complicated, because if any of
the suspend callbacks executed asynchronously returns error code, the
entire suspend has to be terminated and rolled back.  Apart from
this, it's completely analogous to the asynchronous resume.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/base/power/main.c |  225 +++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 214 insertions(+), 11 deletions(-)

Index: linux-2.6/drivers/base/power/main.c
===================================================================
--- linux-2.6.orig/drivers/base/power/main.c
+++ linux-2.6/drivers/base/power/main.c
@@ -240,6 +240,15 @@ static void device_pm_wait_for_masters(s
 }
 
 /**
+ * device_pm_wait_for_slaves - Wait for all slaves of given device.
+ * @slave: Device to wait for the slaves of.
+ */
+static void device_pm_wait_for_slaves(struct device *master)
+{
+	device_for_each_slave(master, master, device_pm_wait_fn);
+}
+
+/**
  * device_pm_check - Check the power.op_complete flag of given device.
  * @dev: Device to check.
  */
@@ -273,6 +282,15 @@ static int device_pm_check_masters(struc
 }
 
 /**
+ * device_pm_check_slaves - Check power.op_complete for slaves of a device.
+ * @slave: Device to check the slaves of.
+ */
+static int device_pm_check_slaves(struct device *master)
+{
+	return device_for_each_slave(master, NULL, device_pm_check_fn);
+}
+
+/**
  * pm_op - Execute the PM operation appropriate for given PM event.
  * @dev: Device to handle.
  * @ops: PM operations to choose from.
@@ -832,6 +850,8 @@ EXPORT_SYMBOL_GPL(dpm_resume_end);
 
 /*------------------------- Suspend routines -------------------------*/
 
+static atomic_t async_error;
+
 /**
  * resume_event - Return a "resume" message for given "suspend" sleep state.
  * @sleep_state: PM message representing a sleep state.
@@ -861,21 +881,100 @@ static pm_message_t resume_event(pm_mess
  * The driver of @dev will not receive interrupts while this function is being
  * executed.
  */
-static int device_suspend_noirq(struct device *dev, pm_message_t state)
+static int __device_suspend_noirq(struct device *dev, pm_message_t state)
 {
 	int error = 0;
 
-	if (!dev->bus)
-		return 0;
-
-	if (dev->bus->pm) {
+	if (dev->bus && dev->bus->pm) {
 		pm_dev_dbg(dev, state, "LATE ");
 		error = pm_noirq_op(dev, dev->bus->pm, state);
 	}
+
+	dev->power.op_complete = true;
+	wake_up_all(&dev->power.wait_queue);
+
 	return error;
 }
 
 /**
+ * async_device_suspend_noirq - Wrapper of __device_suspend_noirq().
+ * @dev: Device to suspend.
+ *
+ * Execute __device_suspend_noirq() for given device unless async_error is
+ * set and if that returns error code, copy it to async_error and change the
+ * PM status of @dev to DPM_OFF.
+ */
+static void async_device_suspend_noirq(struct device *dev)
+{
+	int error = atomic_read(&async_error);
+
+	if (error)
+		return;
+	pm_dev_dbg(dev, pm_transition, "async LATE ");
+	error = __device_suspend_noirq(dev, pm_transition);
+	if (!error)
+		return;
+	pm_dev_err(dev, pm_transition, " async LATE", error);
+	dev->power.status = DPM_OFF;
+	atomic_set(&async_error, error);
+}
+
+/**
+ * async_suspend_noirq - Execute "late" suspend callbacks asynchronously.
+ * @data: Pointer to the first device to suspend.
+ * @cookie: Ignored.
+ *
+ * The execution of this function is scheduled with async_schedule(), so it runs
+ * in its own kernel thread.  It first calls the "late" suspend callback for the
+ * device passed to it as @data.  Next, it walks dpm_list looking for devices
+ * that can be suspended without waiting for their "slaves".  If such a device
+ * is found, its "late" suspend callback is run.
+ */
+static void async_suspend_noirq(void *data, async_cookie_t cookie)
+{
+	struct device *dev = (struct device *)data;
+
+	device_pm_wait_for_slaves(dev);
+	async_device_suspend_noirq(dev);
+
+	list_for_each_entry_continue_reverse(dev, &dpm_list, power.entry) {
+		if (!dev->power.async_suspend)
+			continue;
+
+		if (device_pm_check_slaves(dev))
+			continue;
+
+		if (pm_op_started(dev))
+			continue;
+
+		pm_dev_dbg(dev, pm_transition, "out of order LATE ");
+		async_device_suspend_noirq(dev);
+	}
+}
+
+/**
+ * device_suspend_noirq - Execute or schedule "late" suspend callback.
+ * @dev: Device to suspend.
+ *
+ * If @dev can be resumed asynchronously, schedule the execution of
+ * async_suspend_noirq() for it.  Otherwise, execute its "late" suspend callback
+ * directly.
+ */
+static int device_suspend_noirq(struct device *dev)
+{
+	if (pm_op_started(dev))
+		return 0;
+
+	if (dev->power.async_suspend) {
+		async_schedule(async_suspend_noirq, dev);
+		return 0;
+	}
+
+	device_pm_wait_for_slaves(dev);
+	return __device_suspend_noirq(dev, pm_transition);
+}
+
+/**
  * dpm_suspend_noirq - Execute "late suspend" callbacks for non-sysdev devices.
  * @state: PM transition of the system being carried out.
  *
@@ -889,14 +988,20 @@ int dpm_suspend_noirq(pm_message_t state
 
 	suspend_device_irqs();
 	mutex_lock(&dpm_list_mtx);
+	pm_transition = state;
 	list_for_each_entry_reverse(dev, &dpm_list, power.entry) {
-		error = device_suspend_noirq(dev, state);
+		dev->power.status = DPM_OFF_IRQ;
+		error = device_suspend_noirq(dev);
 		if (error) {
-			pm_dev_err(dev, state, " late", error);
+			pm_dev_err(dev, state, " LATE", error);
+			dev->power.status = DPM_OFF;
 			break;
 		}
-		dev->power.status = DPM_OFF_IRQ;
+		error = atomic_read(&async_error);
+		if (error)
+			break;
 	}
+	dpm_synchronize_noirq();
 	mutex_unlock(&dpm_list_mtx);
 	if (error)
 		dpm_resume_noirq(resume_event(state));
@@ -909,7 +1014,7 @@ EXPORT_SYMBOL_GPL(dpm_suspend_noirq);
  * @dev: Device to handle.
  * @state: PM transition of the system being carried out.
  */
-static int device_suspend(struct device *dev, pm_message_t state)
+static int __device_suspend(struct device *dev, pm_message_t state)
 {
 	int error = 0;
 
@@ -949,11 +1054,102 @@ static int device_suspend(struct device 
 	}
  End:
 	up(&dev->sem);
+	dev->power.op_complete = true;
+	wake_up_all(&dev->power.wait_queue);
 
 	return error;
 }
 
 /**
+ * async_device_suspend - Wrapper of __device_suspend().
+ * @dev: Device to suspend.
+ *
+ * Execute __device_suspend() for given device unless async_error is set and if
+ * that returns error code, copy it to async_error and change the PM status of
+ * @dev to DPM_SUSPENDING.
+ */
+static void async_device_suspend(struct device *dev)
+{
+	int error = atomic_read(&async_error);
+
+	if (error)
+		return;
+	pm_dev_dbg(dev, pm_transition, "async ");
+	error = __device_suspend(dev, pm_transition);
+	if (!error)
+		return;
+	pm_dev_err(dev, pm_transition, " async", error);
+	mutex_lock(&dpm_list_mtx);
+	dev->power.status = DPM_SUSPENDING;
+	atomic_set(&async_error, error);
+	mutex_unlock(&dpm_list_mtx);
+}
+
+/**
+ * async_suspend - Execute suspend callbacks asynchronously.
+ * @data: Pointer to the first device to resume.
+ * @cookie: Ignored.
+ *
+ * The execution of this function is scheduled with async_schedule(), so it runs
+ * in its own kernel thread.  It first calls the suspend callbacks for the
+ * device passed to it as @data.  Next, it walks dpm_list looking for devices
+ * that can be suspended without waiting for their "slaves".  If such a device
+ * is found, its suspend callbacks are run.
+ */
+static void async_suspend(void *data, async_cookie_t cookie)
+{
+	struct device *dev = (struct device *)data;
+
+	device_pm_wait_for_slaves(dev);
+
+ repeat:
+	async_device_suspend(dev);
+	put_device(dev);
+
+	mutex_lock(&dpm_list_mtx);
+	if (dev->power.status >= DPM_OFF)
+		dev = to_device(dpm_list.prev);
+	list_for_each_entry_continue_reverse(dev, &dpm_list, power.entry) {
+		if (!dev->power.async_suspend)
+			continue;
+
+		if (device_pm_check_slaves(dev))
+			continue;
+
+		if (pm_op_started(dev))
+			continue;
+
+		get_device(dev);
+		mutex_unlock(&dpm_list_mtx);
+		pm_dev_dbg(dev, pm_transition, "out of order ");
+		goto repeat;
+	}
+	mutex_unlock(&dpm_list_mtx);
+}
+
+/**
+ * device_suspend - Execute or schedule suspend callbacks for given device.
+ * @dev: Device to suspend.
+ *
+ * If @dev can be suspended asynchronously, schedule the execution of
+ * async_suspend() for it.  Otherwise, execute its suspend callbacks directly.
+ */
+static int device_suspend(struct device *dev)
+{
+	if (pm_op_started(dev))
+		return 0;
+
+	if (dev->power.async_suspend) {
+		get_device(dev);
+		async_schedule(async_suspend, dev);
+		return 0;
+	}
+
+	device_pm_wait_for_slaves(dev);
+	return __device_suspend(dev, pm_transition);
+}
+
+/**
  * dpm_suspend - Execute "suspend" callbacks for all non-sysdev devices.
  * @state: PM transition of the system being carried out.
  */
@@ -964,27 +1160,33 @@ static int dpm_suspend(pm_message_t stat
 
 	INIT_LIST_HEAD(&list);
 	mutex_lock(&dpm_list_mtx);
+	pm_transition = state;
 	while (!list_empty(&dpm_list)) {
 		struct device *dev = to_device(dpm_list.prev);
 
 		get_device(dev);
+		dev->power.status = DPM_OFF;
 		mutex_unlock(&dpm_list_mtx);
 
-		error = device_suspend(dev, state);
+		error = device_suspend(dev);
 
 		mutex_lock(&dpm_list_mtx);
 		if (error) {
 			pm_dev_err(dev, state, "", error);
+			dev->power.status = DPM_SUSPENDING;
 			put_device(dev);
 			break;
 		}
-		dev->power.status = DPM_OFF;
 		if (!list_empty(&dev->power.entry))
 			list_move(&dev->power.entry, &list);
 		put_device(dev);
+		error = atomic_read(&async_error);
+		if (error)
+			break;
 	}
 	list_splice(&list, dpm_list.prev);
 	mutex_unlock(&dpm_list_mtx);
+	dpm_synchronize();
 	return error;
 }
 
@@ -1043,6 +1245,7 @@ static int dpm_prepare(pm_message_t stat
 	INIT_LIST_HEAD(&list);
 	mutex_lock(&dpm_list_mtx);
 	transition_started = true;
+	atomic_set(&async_error, 0);
 	while (!list_empty(&dpm_list)) {
 		struct device *dev = to_device(dpm_list.next);
 

[PATCH 4/9] PM: Allow PCI devices to suspend/resume asynchronously

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Set async_suspend for all PCI devices and PCIe port services, so that
they can be suspended and resumed asynchronously with other devices
they don't depend on in a known way (i.e. devices which are not their
parents or children and to which they are not connected via struct
pm_link objects).

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/pci/pci.c               |    1 +
 drivers/pci/pcie/portdrv_core.c |    1 +
 2 files changed, 2 insertions(+)

Index: linux-2.6/drivers/pci/pci.c
===================================================================
--- linux-2.6.orig/drivers/pci/pci.c
+++ linux-2.6/drivers/pci/pci.c
@@ -1373,6 +1373,7 @@ void pci_pm_init(struct pci_dev *dev)
 	int pm;
 	u16 pmc;
 
+	device_enable_async_suspend(&dev->dev, true);
 	dev->wakeup_prepared = false;
 	dev->pm_cap = 0;
 
Index: linux-2.6/drivers/pci/pcie/portdrv_core.c
===================================================================
--- linux-2.6.orig/drivers/pci/pcie/portdrv_core.c
+++ linux-2.6/drivers/pci/pcie/portdrv_core.c
@@ -280,6 +280,7 @@ static void pcie_device_init(struct pci_
 	dev_set_name(device, "%s:pcie%02x",
 		 pci_name(parent), get_descriptor_id(port_type, service_type));
 	device->parent = &parent->dev;
+	device_enable_async_suspend(device, true);
 }
 
 /**

[PATCH 5/9] PM: Allow ACPI devices to suspend/resume asynchronously

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Set async_suspend for all ACPI devices, so that they can be suspended
and resumed asynchronously with other devices they don't depend on in
a known way (i.e. devices which are not their parents or children and
to which they are not connected via struct pm_link objects).

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/acpi/scan.c |    1 +
 1 file changed, 1 insertion(+)

Index: linux-2.6/drivers/acpi/scan.c
===================================================================
--- linux-2.6.orig/drivers/acpi/scan.c
+++ linux-2.6/drivers/acpi/scan.c
@@ -539,6 +539,7 @@ static int acpi_device_register(struct a
 		       dev_name(&device->dev));
 
 	device->removal_type = ACPI_BUS_REMOVAL_NORMAL;
+	device_enable_async_suspend(&device->dev, true);
 	return 0;
 end:
 	mutex_lock(&acpi_device_lock);

[PATCH 6/9] PM: Add a switch for disabling/enabling asynchronous suspend/resume

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Add sysfs attribute kernel/power/pm_async allowing the user space to
disable and enable async suspend/resume of devices.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/base/power/main.c  |   13 +++++++------
 drivers/base/power/power.h |    6 +++---
 kernel/power/main.c        |   31 ++++++++++++++++++++++++++++++-
 3 files changed, 40 insertions(+), 10 deletions(-)

Index: linux-2.6/kernel/power/main.c
===================================================================
--- linux-2.6.orig/kernel/power/main.c
+++ linux-2.6/kernel/power/main.c
@@ -44,6 +44,32 @@ int pm_notifier_call_chain(unsigned long
 			== NOTIFY_BAD) ? -EINVAL : 0;
 }
 
+/* If set, devices may be suspended and resumed asynchronously. */
+int pm_async_enabled = 1;
+
+static ssize_t pm_async_show(struct kobject *kobj, struct kobj_attribute *attr,
+			     char *buf)
+{
+	return sprintf(buf, "%d\n", pm_async_enabled);
+}
+
+static ssize_t pm_async_store(struct kobject *kobj, struct kobj_attribute *attr,
+			      const char *buf, size_t n)
+{
+	unsigned long val;
+
+	if (strict_strtoul(buf, 10, &val))
+		return -EINVAL;
+
+	if (val > 1)
+		return -EINVAL;
+
+	pm_async_enabled = val;
+	return n;
+}
+
+power_attr(pm_async);
+
 #ifdef CONFIG_PM_DEBUG
 int pm_test_level = TEST_NONE;
 
@@ -208,9 +234,12 @@ static struct attribute * g[] = {
 #ifdef CONFIG_PM_TRACE
 	&pm_trace_attr.attr,
 #endif
-#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_PM_DEBUG)
+#ifdef CONFIG_PM_SLEEP
+	&pm_async_attr.attr,
+#ifdef CONFIG_PM_DEBUG
 	&pm_test_attr.attr,
 #endif
+#endif
 	NULL,
 };
 
Index: linux-2.6/drivers/base/power/main.c
===================================================================
--- linux-2.6.orig/drivers/base/power/main.c
+++ linux-2.6/drivers/base/power/main.c
@@ -235,7 +235,7 @@ static int device_pm_wait_fn(struct devi
  */
 static void device_pm_wait_for_masters(struct device *slave)
 {
-	if (!pm_trace_enabled)
+	if (pm_async_enabled && !pm_trace_enabled)
 		device_for_each_master(slave, slave, device_pm_wait_fn);
 }
 
@@ -245,7 +245,8 @@ static void device_pm_wait_for_masters(s
  */
 static void device_pm_wait_for_slaves(struct device *master)
 {
-	device_for_each_slave(master, master, device_pm_wait_fn);
+	if (pm_async_enabled)
+		device_for_each_slave(master, master, device_pm_wait_fn);
 }
 
 /**
@@ -560,7 +561,7 @@ static int device_resume_noirq(struct de
 	if (pm_op_started(dev))
 		return 0;
 
-	if (dev->power.async_suspend && !pm_trace_enabled) {
+	if (pm_async_enabled && !pm_trace_enabled && dev->power.async_suspend) {
 		async_schedule(async_resume_noirq, dev);
 		return 0;
 	}
@@ -719,7 +720,7 @@ static int device_resume(struct device *
 	if (pm_op_started(dev))
 		return 0;
 
-	if (dev->power.async_suspend && !pm_trace_enabled) {
+	if (pm_async_enabled && !pm_trace_enabled && dev->power.async_suspend) {
 		get_device(dev);
 		async_schedule(async_resume, dev);
 		return 0;
@@ -965,7 +966,7 @@ static int device_suspend_noirq(struct d
 	if (pm_op_started(dev))
 		return 0;
 
-	if (dev->power.async_suspend) {
+	if (pm_async_enabled && dev->power.async_suspend) {
 		async_schedule(async_suspend_noirq, dev);
 		return 0;
 	}
@@ -1139,7 +1140,7 @@ static int device_suspend(struct device 
 	if (pm_op_started(dev))
 		return 0;
 
-	if (dev->power.async_suspend) {
+	if (pm_async_enabled && dev->power.async_suspend) {
 		get_device(dev);
 		async_schedule(async_suspend, dev);
 		return 0;
Index: linux-2.6/drivers/base/power/power.h
===================================================================
--- linux-2.6.orig/drivers/base/power/power.h
+++ linux-2.6/drivers/base/power/power.h
@@ -16,10 +16,10 @@ static inline void pm_runtime_remove(str
 
 #ifdef CONFIG_PM_SLEEP
 
-/*
- * main.c
- */
+/* kernel/power/main.c */
+extern int pm_async_enabled;
 
+/* drivers/base/power/main.c */
 extern struct list_head dpm_list;	/* The active device list */
 
 static inline struct device *to_device(struct list_head *entry)

Re: [PATCH 6/9] PM: Add a switch for disabling/enabling asynchronous suspend/resume

[Pavel Machek]

> From: Rafael J. Wysocki <rjw@sisk.pl>
> 
> Add sysfs attribute kernel/power/pm_async allowing the user space to
> disable and enable async suspend/resume of devices.

for bisect reasons, should this go before 5?

> 
> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
> ---
>  drivers/base/power/main.c  |   13 +++++++------
>  drivers/base/power/power.h |    6 +++---
>  kernel/power/main.c        |   31 ++++++++++++++++++++++++++++++-
>  3 files changed, 40 insertions(+), 10 deletions(-)
> 
> Index: linux-2.6/kernel/power/main.c
> ===================================================================
> --- linux-2.6.orig/kernel/power/main.c
> +++ linux-2.6/kernel/power/main.c
> @@ -44,6 +44,32 @@ int pm_notifier_call_chain(unsigned long
>  			== NOTIFY_BAD) ? -EINVAL : 0;
>  }
>  
> +/* If set, devices may be suspended and resumed asynchronously. */
> +int pm_async_enabled = 1;
> +
> +static ssize_t pm_async_show(struct kobject *kobj, struct kobj_attribute *attr,
> +			     char *buf)
> +{
> +	return sprintf(buf, "%d\n", pm_async_enabled);
> +}
> +
> +static ssize_t pm_async_store(struct kobject *kobj, struct kobj_attribute *attr,
> +			      const char *buf, size_t n)
> +{
> +	unsigned long val;
> +
> +	if (strict_strtoul(buf, 10, &val))
> +		return -EINVAL;
> +
> +	if (val > 1)
> +		return -EINVAL;
> +
> +	pm_async_enabled = val;
> +	return n;
> +}
> +
> +power_attr(pm_async);
> +
>  #ifdef CONFIG_PM_DEBUG
>  int pm_test_level = TEST_NONE;
>  
> @@ -208,9 +234,12 @@ static struct attribute * g[] = {
>  #ifdef CONFIG_PM_TRACE
>  	&pm_trace_attr.attr,
>  #endif
> -#if defined(CONFIG_PM_SLEEP) && defined(CONFIG_PM_DEBUG)
> +#ifdef CONFIG_PM_SLEEP
> +	&pm_async_attr.attr,
> +#ifdef CONFIG_PM_DEBUG
>  	&pm_test_attr.attr,
>  #endif
> +#endif
>  	NULL,
>  };
>  
> Index: linux-2.6/drivers/base/power/main.c
> ===================================================================
> --- linux-2.6.orig/drivers/base/power/main.c
> +++ linux-2.6/drivers/base/power/main.c
> @@ -235,7 +235,7 @@ static int device_pm_wait_fn(struct devi
>   */
>  static void device_pm_wait_for_masters(struct device *slave)
>  {
> -	if (!pm_trace_enabled)
> +	if (pm_async_enabled && !pm_trace_enabled)
>  		device_for_each_master(slave, slave, device_pm_wait_fn);
>  }
>  
> @@ -245,7 +245,8 @@ static void device_pm_wait_for_masters(s
>   */
>  static void device_pm_wait_for_slaves(struct device *master)
>  {
> -	device_for_each_slave(master, master, device_pm_wait_fn);
> +	if (pm_async_enabled)
> +		device_for_each_slave(master, master, device_pm_wait_fn);
>  }
>  
>  /**
> @@ -560,7 +561,7 @@ static int device_resume_noirq(struct de
>  	if (pm_op_started(dev))
>  		return 0;
>  
> -	if (dev->power.async_suspend && !pm_trace_enabled) {
> +	if (pm_async_enabled && !pm_trace_enabled && dev->power.async_suspend) {
>  		async_schedule(async_resume_noirq, dev);
>  		return 0;
>  	}
> @@ -719,7 +720,7 @@ static int device_resume(struct device *
>  	if (pm_op_started(dev))
>  		return 0;
>  
> -	if (dev->power.async_suspend && !pm_trace_enabled) {
> +	if (pm_async_enabled && !pm_trace_enabled && dev->power.async_suspend) {
>  		get_device(dev);
>  		async_schedule(async_resume, dev);
>  		return 0;
> @@ -965,7 +966,7 @@ static int device_suspend_noirq(struct d
>  	if (pm_op_started(dev))
>  		return 0;
>  
> -	if (dev->power.async_suspend) {
> +	if (pm_async_enabled && dev->power.async_suspend) {
>  		async_schedule(async_suspend_noirq, dev);
>  		return 0;
>  	}
> @@ -1139,7 +1140,7 @@ static int device_suspend(struct device 
>  	if (pm_op_started(dev))
>  		return 0;
>  
> -	if (dev->power.async_suspend) {
> +	if (pm_async_enabled && dev->power.async_suspend) {
>  		get_device(dev);
>  		async_schedule(async_suspend, dev);
>  		return 0;
> Index: linux-2.6/drivers/base/power/power.h
> ===================================================================
> --- linux-2.6.orig/drivers/base/power/power.h
> +++ linux-2.6/drivers/base/power/power.h
> @@ -16,10 +16,10 @@ static inline void pm_runtime_remove(str
>  
>  #ifdef CONFIG_PM_SLEEP
>  
> -/*
> - * main.c
> - */
> +/* kernel/power/main.c */
> +extern int pm_async_enabled;
>  
> +/* drivers/base/power/main.c */
>  extern struct list_head dpm_list;	/* The active device list */
>  
>  static inline struct device *to_device(struct list_head *entry)

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [PATCH 6/9] PM: Add a switch for disabling/enabling asynchronous suspend/resume

[Rafael J. Wysocki]

On Thursday 10 September 2009, Pavel Machek wrote:
> > From: Rafael J. Wysocki <rjw@sisk.pl>
> > 
> > Add sysfs attribute kernel/power/pm_async allowing the user space to
> > disable and enable async suspend/resume of devices.
> 
> for bisect reasons, should this go before 5?

It is set to "enabled" by default, so it shouldn't change bisection results.

Thanks,
Rafael

Re: [PATCH 6/9] PM: Add a switch for disabling/enabling asynchronous suspend/resume

[Pavel Machek]

On Thu 2009-09-10 20:56:39, Rafael J. Wysocki wrote:
> On Thursday 10 September 2009, Pavel Machek wrote:
> > > From: Rafael J. Wysocki <rjw@sisk.pl>
> > > 
> > > Add sysfs attribute kernel/power/pm_async allowing the user space to
> > > disable and enable async suspend/resume of devices.
> > 
> > for bisect reasons, should this go before 5?
> 
> It is set to "enabled" by default, so it shouldn't change bisection results.

Well, 5 enables it, 6 makes it configurable. So if you need async
disabled for a bisect, hitting commit between 5 and 6 will enable it
for you...

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [PATCH 6/9] PM: Add a switch for disabling/enabling asynchronous suspend/resume

[Rafael J. Wysocki]

On Thursday 10 September 2009, Pavel Machek wrote:
> On Thu 2009-09-10 20:56:39, Rafael J. Wysocki wrote:
> > On Thursday 10 September 2009, Pavel Machek wrote:
> > > > From: Rafael J. Wysocki <rjw@sisk.pl>
> > > > 
> > > > Add sysfs attribute kernel/power/pm_async allowing the user space to
> > > > disable and enable async suspend/resume of devices.
> > > 
> > > for bisect reasons, should this go before 5?
> > 
> > It is set to "enabled" by default, so it shouldn't change bisection results.
> 
> Well, 5 enables it, 6 makes it configurable. So if you need async
> disabled for a bisect, hitting commit between 5 and 6 will enable it
> for you...

OK, I'll move the patch before the [4/9].

Thanks,
Rafael

[PATCH 7/9] PM: Measure device suspend and resume times

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Measure and print the time of suspending and resuming all devices.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/base/power/main.c |   56 +++++++++++++++++++++++++++++++++++++++++++++-
 include/linux/pm.h        |    3 ++
 kernel/power/swsusp.c     |    6 ----
 3 files changed, 59 insertions(+), 6 deletions(-)

Index: linux-2.6/kernel/power/swsusp.c
===================================================================
--- linux-2.6.orig/kernel/power/swsusp.c
+++ linux-2.6/kernel/power/swsusp.c
@@ -169,14 +169,10 @@ int swsusp_swap_in_use(void)
 void swsusp_show_speed(struct timeval *start, struct timeval *stop,
 			unsigned nr_pages, char *msg)
 {
-	s64 elapsed_centisecs64;
-	int centisecs;
+	int centisecs = pm_time_elapsed(start, stop);
 	int k;
 	int kps;
 
-	elapsed_centisecs64 = timeval_to_ns(stop) - timeval_to_ns(start);
-	do_div(elapsed_centisecs64, NSEC_PER_SEC / 100);
-	centisecs = elapsed_centisecs64;
 	if (centisecs == 0)
 		centisecs = 1;	/* avoid div-by-zero */
 	k = nr_pages * (PAGE_SIZE / 1024);
Index: linux-2.6/drivers/base/power/main.c
===================================================================
--- linux-2.6.orig/drivers/base/power/main.c
+++ linux-2.6/drivers/base/power/main.c
@@ -28,6 +28,7 @@
 #include <linux/interrupt.h>
 #include <linux/async.h>
 #include <linux/completion.h>
+#include <linux/time.h>
 
 #include "../base.h"
 #include "power.h"
@@ -434,6 +435,20 @@ static bool pm_op_started(struct device 
 	return ret;
 }
 
+/**
+ * pm_time_elapsed - Compute time elapsed between two timestamps.
+ * @start: First timestamp.
+ * @stop: Second timestamp.
+ */
+int pm_time_elapsed(struct timeval *start, struct timeval *stop)
+{
+	s64 elapsed_centisecs64;
+
+	elapsed_centisecs64 = timeval_to_ns(stop) - timeval_to_ns(start);
+	do_div(elapsed_centisecs64, NSEC_PER_SEC / 100);
+	return elapsed_centisecs64;
+}
+
 static char *pm_verb(int event)
 {
 	switch (event) {
@@ -458,6 +473,16 @@ static char *pm_verb(int event)
 	}
 }
 
+static void dpm_show_time(struct timeval *start, struct timeval *stop,
+			  pm_message_t state, const char *info)
+{
+	int centisecs = pm_time_elapsed(start, stop);
+
+	printk(KERN_INFO "PM: %s%s%s of devices complete in %d.%02d seconds\n",
+		info ? info : "", info ? " " : "", pm_verb(state.event),
+		centisecs / 100, centisecs % 100);
+}
+
 static void pm_dev_dbg(struct device *dev, pm_message_t state, char *info)
 {
 	dev_dbg(dev, "%s%s%s\n", info, pm_verb(state.event),
@@ -580,6 +605,9 @@ static int device_resume_noirq(struct de
 void dpm_resume_noirq(pm_message_t state)
 {
 	struct device *dev;
+	struct timeval start, stop;
+
+	do_gettimeofday(&start);
 
 	mutex_lock(&dpm_list_mtx);
 	transition_started = false;
@@ -595,6 +623,10 @@ void dpm_resume_noirq(pm_message_t state
 		}
 	dpm_synchronize_noirq();
 	mutex_unlock(&dpm_list_mtx);
+
+	do_gettimeofday(&stop);
+	dpm_show_time(&start, &stop, state, "EARLY");
+
 	resume_device_irqs();
 }
 EXPORT_SYMBOL_GPL(dpm_resume_noirq);
@@ -740,6 +772,9 @@ static int device_resume(struct device *
 static void dpm_resume(pm_message_t state)
 {
 	struct list_head list;
+	struct timeval start, stop;
+
+	do_gettimeofday(&start);
 
 	INIT_LIST_HEAD(&list);
 	mutex_lock(&dpm_list_mtx);
@@ -770,6 +805,9 @@ static void dpm_resume(pm_message_t stat
 	list_splice(&list, &dpm_list);
 	mutex_unlock(&dpm_list_mtx);
 	dpm_synchronize();
+
+	do_gettimeofday(&stop);
+	dpm_show_time(&start, &stop, state, NULL);
 }
 
 /**
@@ -985,8 +1023,11 @@ static int device_suspend_noirq(struct d
 int dpm_suspend_noirq(pm_message_t state)
 {
 	struct device *dev;
+	struct timeval start, stop;
 	int error = 0;
 
+	do_gettimeofday(&start);
+
 	suspend_device_irqs();
 	mutex_lock(&dpm_list_mtx);
 	pm_transition = state;
@@ -1004,8 +1045,12 @@ int dpm_suspend_noirq(pm_message_t state
 	}
 	dpm_synchronize_noirq();
 	mutex_unlock(&dpm_list_mtx);
-	if (error)
+	if (error) {
 		dpm_resume_noirq(resume_event(state));
+	} else {
+		do_gettimeofday(&stop);
+		dpm_show_time(&start, &stop, state, "LATE");
+	}
 	return error;
 }
 EXPORT_SYMBOL_GPL(dpm_suspend_noirq);
@@ -1157,8 +1202,11 @@ static int device_suspend(struct device 
 static int dpm_suspend(pm_message_t state)
 {
 	struct list_head list;
+	struct timeval start, stop;
 	int error = 0;
 
+	do_gettimeofday(&start);
+
 	INIT_LIST_HEAD(&list);
 	mutex_lock(&dpm_list_mtx);
 	pm_transition = state;
@@ -1188,6 +1236,12 @@ static int dpm_suspend(pm_message_t stat
 	list_splice(&list, dpm_list.prev);
 	mutex_unlock(&dpm_list_mtx);
 	dpm_synchronize();
+
+	if (!error) {
+		do_gettimeofday(&stop);
+		dpm_show_time(&start, &stop, state, NULL);
+	}
+
 	return error;
 }
 
Index: linux-2.6/include/linux/pm.h
===================================================================
--- linux-2.6.orig/include/linux/pm.h
+++ linux-2.6/include/linux/pm.h
@@ -505,6 +505,9 @@ extern int sysdev_suspend(pm_message_t s
 extern int dpm_suspend_noirq(pm_message_t state);
 extern int dpm_suspend_start(pm_message_t state);
 
+struct timeval;
+extern int pm_time_elapsed(struct timeval *start, struct timeval *stop);
+
 extern void __suspend_report_result(const char *function, void *fn, int ret);
 
 #define suspend_report_result(fn, ret)					\

[PATCH 8/9] PM: Add facility for advanced testing of async suspend/resume

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

Add configuration switch CONFIG_PM_ADVANCED_DEBUG for compiling in
extra PM debugging/testing code allowing one to access some
PM-related attributes of devices from the user space via sysfs.

If CONFIG_PM_ADVANCED_DEBUG is set, add sysfs attribute power/async
for every device allowing the user space to access the device's
power.async_suspend flag and modify it, if desired.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/base/power/sysfs.c |   47 +++++++++++++++++++++++++++++++++++++++++++++
 include/linux/device.h     |    5 ++++
 kernel/power/Kconfig       |   14 +++++++++++++
 3 files changed, 66 insertions(+)

Index: linux-2.6/drivers/base/power/sysfs.c
===================================================================
--- linux-2.6.orig/drivers/base/power/sysfs.c
+++ linux-2.6/drivers/base/power/sysfs.c
@@ -38,6 +38,22 @@
  *	wakeup events internally (unless they are disabled), keeping
  *	their hardware in low power modes whenever they're unused.  This
  *	saves runtime power, without requiring system-wide sleep states.
+ *
+ *	async - Report/change current async suspend setting for the device
+ *
+ *	If set, the PM core will attempt to suspend and resume the device during
+ *	system power transitions (e.g. suspend to RAM, hibernation) in parallel
+ *	with other devices it doesn't appear to depend on (to the PM core's
+ *	knowledge).
+ *
+ *	 + "enabled\n" to permit the asynchronous suspend/resume of the device
+ *	 + "disabled\n" to forbid it
+ *
+ *	NOTE: It generally is unsafe to permit the asynchronous suspend/resume
+ *	of a device unless it is certain that all of the PM dependencies of the
+ *	device are known to the PM core.  However, for some devices this
+ *	attribute is set to "enabled" by the kernel and in that cases it should
+ *	be safe to leave the default value.
  */
 
 static const char enabled[] = "enabled";
@@ -77,9 +93,40 @@ wake_store(struct device * dev, struct d
 
 static DEVICE_ATTR(wakeup, 0644, wake_show, wake_store);
 
+#ifdef CONFIG_PM_SLEEP_ADVANCED_DEBUG
+static ssize_t async_show(struct device *dev, struct device_attribute *attr,
+			  char *buf)
+{
+	return sprintf(buf, "%s\n",
+			device_async_suspend_enabled(dev) ? enabled : disabled);
+}
+
+static ssize_t async_store(struct device *dev, struct device_attribute *attr,
+			   const char *buf, size_t n)
+{
+	char *cp;
+	int len = n;
+
+	cp = memchr(buf, '\n', n);
+	if (cp)
+		len = cp - buf;
+	if (len == sizeof enabled - 1 && strncmp(buf, enabled, len) == 0)
+		device_enable_async_suspend(dev, true);
+	else if (len == sizeof disabled - 1 && strncmp(buf, disabled, len) == 0)
+		device_enable_async_suspend(dev, false);
+	else
+		return -EINVAL;
+	return n;
+}
+
+static DEVICE_ATTR(async, 0644, async_show, async_store);
+#endif /* CONFIG_PM_SLEEP_ADVANCED_DEBUG */
 
 static struct attribute * power_attrs[] = {
 	&dev_attr_wakeup.attr,
+#ifdef CONFIG_PM_SLEEP_ADVANCED_DEBUG
+	&dev_attr_async.attr,
+#endif
 	NULL,
 };
 static struct attribute_group pm_attr_group = {
Index: linux-2.6/include/linux/device.h
===================================================================
--- linux-2.6.orig/include/linux/device.h
+++ linux-2.6/include/linux/device.h
@@ -478,6 +478,11 @@ static inline void device_enable_async_s
 		dev->power.async_suspend = enable;
 }
 
+static inline bool device_async_suspend_enabled(struct device *dev)
+{
+	return !!dev->power.async_suspend;
+}
+
 void driver_init(void);
 
 /*
Index: linux-2.6/kernel/power/Kconfig
===================================================================
--- linux-2.6.orig/kernel/power/Kconfig
+++ linux-2.6/kernel/power/Kconfig
@@ -27,6 +27,15 @@ config PM_DEBUG
 	code. This is helpful when debugging and reporting PM bugs, like
 	suspend support.
 
+config PM_ADVANCED_DEBUG
+	bool "Extra PM attributes in sysfs for low-level debugging/testing"
+	depends on PM_DEBUG
+	default n
+	---help---
+	Add extra sysfs attributes allowing one to access some Power Management
+	fields of device objects from user space.  If you are not a kernel
+	developer interested in debugging/testing Power Management, say "no".
+
 config PM_VERBOSE
 	bool "Verbose Power Management debugging"
 	depends on PM_DEBUG
@@ -85,6 +94,11 @@ config PM_SLEEP
 	depends on SUSPEND || HIBERNATION || XEN_SAVE_RESTORE
 	default y
 
+config PM_SLEEP_ADVANCED_DEBUG
+	bool
+	depends on PM_ADVANCED_DEBUG
+	default n
+
 config SUSPEND
 	bool "Suspend to RAM and standby"
 	depends on PM && ARCH_SUSPEND_POSSIBLE

[PATCH 9/9] PM: Measure suspend and resume times for individual devices

[Rafael J. Wysocki]

From: Rafael J. Wysocki <rjw@sisk.pl>

If verbose PM debugging is enabled, measure and print the time of
suspending and resuming of individual devices.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
---
 drivers/base/power/main.c |   50 +++++++++++++++++++++++++++++++++++++++++-----
 kernel/power/swsusp.c     |    2 -
 2 files changed, 46 insertions(+), 6 deletions(-)

Index: linux-2.6/drivers/base/power/main.c
===================================================================
--- linux-2.6.orig/drivers/base/power/main.c
+++ linux-2.6/drivers/base/power/main.c
@@ -442,11 +442,11 @@ static bool pm_op_started(struct device 
  */
 int pm_time_elapsed(struct timeval *start, struct timeval *stop)
 {
-	s64 elapsed_centisecs64;
+	s64 elapsed_msecs64;
 
-	elapsed_centisecs64 = timeval_to_ns(stop) - timeval_to_ns(start);
-	do_div(elapsed_centisecs64, NSEC_PER_SEC / 100);
-	return elapsed_centisecs64;
+	elapsed_msecs64 = timeval_to_ns(stop) - timeval_to_ns(start);
+	do_div(elapsed_msecs64, NSEC_PER_SEC / MSEC_PER_SEC);
+	return elapsed_msecs64;
 }
 
 static char *pm_verb(int event)
@@ -476,7 +476,7 @@ static char *pm_verb(int event)
 static void dpm_show_time(struct timeval *start, struct timeval *stop,
 			  pm_message_t state, const char *info)
 {
-	int centisecs = pm_time_elapsed(start, stop);
+	int centisecs = pm_time_elapsed(start, stop) / 10;
 
 	printk(KERN_INFO "PM: %s%s%s of devices complete in %d.%02d seconds\n",
 		info ? info : "", info ? " " : "", pm_verb(state.event),
@@ -497,6 +497,32 @@ static void pm_dev_err(struct device *de
 		kobject_name(&dev->kobj), pm_verb(state.event), info, error);
 }
 
+#ifdef DEBUG
+static inline void dbg_get_time(struct timeval *start)
+{
+	do_gettimeofday(start);
+}
+
+static void dbg_show_time(struct timeval *start, struct device *dev,
+			  pm_message_t state, char *info)
+{
+	struct timeval stop;
+	int msecs;
+
+	do_gettimeofday(&stop);
+	msecs = pm_time_elapsed(start, &stop);
+	dev_dbg(dev, "PID %d: %s%s%s complete in %d.%03d seconds\n",
+		task_pid_nr(current), info ? info : "", info ? " " : "",
+		pm_verb(state.event), msecs / 1000, msecs % 1000);
+}
+
+#else /* !DEBUG */
+static void dbg_get_time(struct timeval *start) {}
+static void dbg_show_time(struct timeval *start, struct device *dev,
+			  pm_message_t state, char *info) {}
+
+#endif /* !DEBUG */
+
 /*------------------------- Resume routines -------------------------*/
 
 /**
@@ -510,7 +536,9 @@ static void pm_dev_err(struct device *de
 static int __device_resume_noirq(struct device *dev, pm_message_t state)
 {
 	int error = 0;
+	struct timeval start;
 
+	dbg_get_time(&start);
 	TRACE_DEVICE(dev);
 	TRACE_RESUME(0);
 
@@ -523,6 +551,7 @@ static int __device_resume_noirq(struct 
 	wake_up_all(&dev->power.wait_queue);
 
 	TRACE_RESUME(error);
+	dbg_show_time(&start, dev, state, "EARLY");
 	return error;
 }
 
@@ -639,7 +668,9 @@ EXPORT_SYMBOL_GPL(dpm_resume_noirq);
 static int __device_resume(struct device *dev, pm_message_t state)
 {
 	int error = 0;
+	struct timeval start;
 
+	dbg_get_time(&start);
 	TRACE_DEVICE(dev);
 	TRACE_RESUME(0);
 
@@ -681,6 +712,7 @@ static int __device_resume(struct device
 	wake_up_all(&dev->power.wait_queue);
 
 	TRACE_RESUME(error);
+	dbg_show_time(&start, dev, state, NULL);
 	return error;
 }
 
@@ -923,6 +955,9 @@ static pm_message_t resume_event(pm_mess
 static int __device_suspend_noirq(struct device *dev, pm_message_t state)
 {
 	int error = 0;
+	struct timeval start;
+
+	dbg_get_time(&start);
 
 	if (dev->bus && dev->bus->pm) {
 		pm_dev_dbg(dev, state, "LATE ");
@@ -932,6 +967,7 @@ static int __device_suspend_noirq(struct
 	dev->power.op_complete = true;
 	wake_up_all(&dev->power.wait_queue);
 
+	dbg_show_time(&start, dev, state, "LATE");
 	return error;
 }
 
@@ -1063,6 +1099,9 @@ EXPORT_SYMBOL_GPL(dpm_suspend_noirq);
 static int __device_suspend(struct device *dev, pm_message_t state)
 {
 	int error = 0;
+	struct timeval start;
+
+	dbg_get_time(&start);
 
 	down(&dev->sem);
 
@@ -1103,6 +1142,7 @@ static int __device_suspend(struct devic
 	dev->power.op_complete = true;
 	wake_up_all(&dev->power.wait_queue);
 
+	dbg_show_time(&start, dev, state, NULL);
 	return error;
 }
 
Index: linux-2.6/kernel/power/swsusp.c
===================================================================
--- linux-2.6.orig/kernel/power/swsusp.c
+++ linux-2.6/kernel/power/swsusp.c
@@ -169,7 +169,7 @@ int swsusp_swap_in_use(void)
 void swsusp_show_speed(struct timeval *start, struct timeval *stop,
 			unsigned nr_pages, char *msg)
 {
-	int centisecs = pm_time_elapsed(start, stop);
+	int centisecs = pm_time_elapsed(start, stop) / 10;
 	int k;
 	int kps;
 

Re: [PATCH 0/9] PM: Asynchronous suspend of devices

[Pavel Machek]

Hi!
> 
> The following series of patches implements asynchronous suspend and resume of
> devices.
> 
> All of the patches have already been discussed, please refer to the changelogs
> for details.
> 
> The patches are targeted at 2.6.33, so if there are no major objections, I'll
> put them into the linux-next branch of the suspend-2.6 tree after .31-rc1 is
> out.

Patches 1-6 look mostly ok to me, I guess you can add my Acked-by. The
rest is fine, too, but I'm not sure if we really need so much of
debugging infrastructure...
									Pavel

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [PATCH 0/9] PM: Asynchronous suspend of devices

[Rafael J. Wysocki]

On Thursday 10 September 2009, Pavel Machek wrote:
> Hi!
> > 
> > The following series of patches implements asynchronous suspend and resume of
> > devices.
> > 
> > All of the patches have already been discussed, please refer to the changelogs
> > for details.
> > 
> > The patches are targeted at 2.6.33, so if there are no major objections, I'll
> > put them into the linux-next branch of the suspend-2.6 tree after .31-rc1 is
> > out.
> 
> Patches 1-6 look mostly ok to me, I guess you can add my Acked-by.

Thanks!

> The rest is fine, too, but I'm not sure if we really need so much of
> debugging infrastructure...

I believe it will be useful.

Thanks,
Rafael