Skip to contentSkip to conversation

PM: Remove unused and unnecessary features from suspend and resume core

PATCH In review Rafael J. Wysocki 8 participants 33 messages Jun 13, 2007 – Jun 17, 2007

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

cover PM: Remove unused and unnecessary features from suspend and resume core

Hi,

The following series of patches removes some unused and unnecessary features from the suspend and resume core code.

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#1 lore raw

PATCH 1/7 PM: Remove pm_parent from struct dev_pm_info

From: Rafael J. Wysocki <[email protected]>

The pm_parent member of struct dev_pm_info (defined in include/linux/pm.h) is only used to check if the device's parent is in the right state while the device is being suspended or resumed. However, this can be done just as well with the help of the parent pointer in struct device, so pm_parent can be removed along with some code that handles it.

---
 drivers/base/power/main.c    |   30 ++++--------------------------
 drivers/base/power/resume.c  |    7 +++----
 drivers/base/power/suspend.c |    7 +++----
 include/linux/pm.h           |    3 ---
 4 files changed, 10 insertions(+), 37 deletions(-)
drivers/base/power/main.c +4 -26
Index: linux-2.6.22-rc4-mm2/drivers/base/power/main.c
===================================================================
@@ -33,28 +33,7 @@ DEFINE_MUTEX(dpm_list_mtx);
 
 int (*platform_enable_wakeup)(struct device *dev, int is_on);
 
-
-/**
- *	device_pm_set_parent - Specify power dependency.
- *	@dev:		Device who needs power.
- *	@parent:	Device that supplies power.
- *
- *	This function is used to manually describe a power-dependency
- *	relationship. It may be used to specify a transversal relationship
- *	(where the power supplier is not the physical (or electrical)
- *	ancestor of a specific device.
- *	The effect of this is that the supplier will not be powered down
- *	before the power dependent.
- */
-
-void device_pm_set_parent(struct device * dev, struct device * parent)
-{
-	put_device(dev->power.pm_parent);
-	dev->power.pm_parent = get_device(parent);
-}
-EXPORT_SYMBOL_GPL(device_pm_set_parent);
-
-int device_pm_add(struct device * dev)
+int device_pm_add(struct device *dev)
 {
 	int error;
 
@@ -63,21 +42,20 @@ int device_pm_add(struct device * dev)
 		 kobject_name(&dev->kobj));
 	mutex_lock(&dpm_list_mtx);
 	list_add_tail(&dev->power.entry, &dpm_active);
-	device_pm_set_parent(dev, dev->parent);
-	if ((error = dpm_sysfs_add(dev)))
+	error = dpm_sysfs_add(dev);
+	if (error)
 		list_del(&dev->power.entry);
 	mutex_unlock(&dpm_list_mtx);
 	return error;
 }
 
-void device_pm_remove(struct device * dev)
+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));
 	mutex_lock(&dpm_list_mtx);
 	dpm_sysfs_remove(dev);
-	put_device(dev->power.pm_parent);
 	list_del_init(&dev->power.entry);
 	mutex_unlock(&dpm_list_mtx);
 }
drivers/base/power/resume.c +3 -4
Index: linux-2.6.22-rc4-mm2/drivers/base/power/resume.c
===================================================================
@@ -29,12 +29,11 @@ int resume_device(struct device * dev)
 
 	down(&dev->sem);
 
-	if (dev->power.pm_parent
-			&& dev->power.pm_parent->power.power_state.event) {
+	if (dev->parent && dev->parent->power.power_state.event) {
 		dev_err(dev, "PM: resume from %d, parent %s still %d\n",
 			dev->power.power_state.event,
-			dev->power.pm_parent->bus_id,
-			dev->power.pm_parent->power.power_state.event);
+			dev->parent->bus_id,
+			dev->parent->power.power_state.event);
 	}
 
 	if (dev->bus && dev->bus->resume) {
drivers/base/power/suspend.c +3 -4
Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================
@@ -55,13 +55,12 @@ int suspend_device(struct device * dev, 
 		dev_dbg(dev, "PM: suspend %d-->%d\n",
 			dev->power.power_state.event, state.event);
 	}
-	if (dev->power.pm_parent
-			&& dev->power.pm_parent->power.power_state.event) {
+	if (dev->parent && dev->parent->power.power_state.event) {
 		dev_err(dev,
 			"PM: suspend %d->%d, parent %s already %d\n",
 			dev->power.power_state.event, state.event,
-			dev->power.pm_parent->bus_id,
-			dev->power.pm_parent->power.power_state.event);
+			dev->parent->bus_id,
+			dev->parent->power.power_state.event);
 	}
 
 	dev->power.prev_state = dev->power.power_state;
include/linux/pm.h +0 -3
Index: linux-2.6.22-rc4-mm2/include/linux/pm.h
===================================================================
@@ -237,13 +237,10 @@ struct dev_pm_info {
 	unsigned		should_wakeup:1;
 	pm_message_t		prev_state;
 	void			* saved_state;
-	struct device		* pm_parent;
 	struct list_head	entry;
 #endif
 };
 
-extern void device_pm_set_parent(struct device * dev, struct device * parent);
-
 extern int device_power_down(pm_message_t state);
 extern void device_power_up(void);
 extern void device_resume(void);
#2 lore raw

PATCH 2/7 PM: Remove saved_state from struct dev_pm_info

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The saved_state member of struct dev_pm_info, defined in include/linux/pm.h, is not used anywhere, so it can be removed.

---
 include/linux/pm.h |    1 -
 1 file changed, 1 deletion(-)
include/linux/pm.h +0 -1
Index: linux-2.6.22-rc4-mm2/include/linux/pm.h
===================================================================
@@ -236,7 +236,6 @@ struct dev_pm_info {
 #ifdef	CONFIG_PM
 	unsigned		should_wakeup:1;
 	pm_message_t		prev_state;
-	void			* saved_state;
 	struct list_head	entry;
 #endif
 };
#3 lore raw

PATCH 3/7 PM: Simplify suspend_device

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

Reduce code duplication in drivers/base/suspend.c by introducing a separate function for printing diagnostic messages.

---
 drivers/base/power/suspend.c |   49 +++++++++++++++----------------------------
 1 file changed, 18 insertions(+), 31 deletions(-)
drivers/base/power/suspend.c +18 -31
Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================
@@ -40,6 +40,14 @@ static inline char *suspend_verb(u32 eve
 }
 
 
+static void
+suspend_device_dbg(struct device * dev, pm_message_t state, char *info)
+{
+	dev_dbg(dev, "%s%s%s\n", info, suspend_verb(state.event),
+		((state.event == PM_EVENT_SUSPEND) && device_may_wakeup(dev)) ?
+		", may wakeup" : "");
+}
+
 /**
  *	suspend_device - Save state of one device.
  *	@dev:	Device.
@@ -66,37 +74,21 @@ int suspend_device(struct device * dev, 
 	dev->power.prev_state = dev->power.power_state;
 
 	if (dev->class && dev->class->suspend && !dev->power.power_state.event) {
-		dev_dbg(dev, "class %s%s\n",
-			suspend_verb(state.event),
-			((state.event == PM_EVENT_SUSPEND)
-					&& device_may_wakeup(dev))
-				? ", may wakeup"
-				: ""
-			);
+		suspend_device_dbg(dev, state, "class ");
 		error = dev->class->suspend(dev, state);
 		suspend_report_result(dev->class->suspend, error);
 	}
 
-	if (!error && dev->type && dev->type->suspend && !dev->power.power_state.event) {
-		dev_dbg(dev, "%s%s\n",
-			suspend_verb(state.event),
-			((state.event == PM_EVENT_SUSPEND)
-					&& device_may_wakeup(dev))
-				? ", may wakeup"
-				: ""
-			);
+	if (!error && dev->type && dev->type->suspend
+	    && !dev->power.power_state.event) {
+		suspend_device_dbg(dev, state, "type ");
 		error = dev->type->suspend(dev, state);
 		suspend_report_result(dev->type->suspend, error);
 	}
 
-	if (!error && dev->bus && dev->bus->suspend && !dev->power.power_state.event) {
-		dev_dbg(dev, "%s%s\n",
-			suspend_verb(state.event),
-			((state.event == PM_EVENT_SUSPEND)
-					&& device_may_wakeup(dev))
-				? ", may wakeup"
-				: ""
-			);
+	if (!error && dev->bus && dev->bus->suspend
+	    && !dev->power.power_state.event) {
+		suspend_device_dbg(dev, state, "");
 		error = dev->bus->suspend(dev, state);
 		suspend_report_result(dev->bus->suspend, error);
 	}
@@ -114,14 +106,9 @@ static int suspend_device_late(struct de
 {
 	int error = 0;
 
-	if (dev->bus && dev->bus->suspend_late && !dev->power.power_state.event) {
-		dev_dbg(dev, "LATE %s%s\n",
-			suspend_verb(state.event),
-			((state.event == PM_EVENT_SUSPEND)
-					&& device_may_wakeup(dev))
-				? ", may wakeup"
-				: ""
-			);
+	if (dev->bus && dev->bus->suspend_late
+	    && !dev->power.power_state.event) {
+		suspend_device_dbg(dev, state, "LATE ");
 		error = dev->bus->suspend_late(dev, state);
 		suspend_report_result(dev->bus->suspend_late, error);
 	}
#4 lore raw

PATCH 4/7 PM: Remove suspend and resume support from struct device_type

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend and resume support in struct device_type (include/linux/device.h) is not used anywhere. It is also undocumented, so it can be removed.

---
 drivers/base/power/resume.c  |    5 -----
 drivers/base/power/suspend.c |    7 -------
 include/linux/device.h       |    2 --
 3 files changed, 14 deletions(-)
drivers/base/power/resume.c +0 -5
Index: linux-2.6.22-rc4-mm2/drivers/base/power/resume.c
===================================================================
@@ -41,11 +41,6 @@ int resume_device(struct device * dev)
 		error = dev->bus->resume(dev);
 	}
 
-	if (!error && dev->type && dev->type->resume) {
-		dev_dbg(dev,"resuming\n");
-		error = dev->type->resume(dev);
-	}
-
 	if (!error && dev->class && dev->class->resume) {
 		dev_dbg(dev,"class resume\n");
 		error = dev->class->resume(dev);
drivers/base/power/suspend.c +0 -7
Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================
@@ -79,13 +79,6 @@ int suspend_device(struct device * dev, 
 		suspend_report_result(dev->class->suspend, error);
 	}
 
-	if (!error && dev->type && dev->type->suspend
-	    && !dev->power.power_state.event) {
-		suspend_device_dbg(dev, state, "type ");
-		error = dev->type->suspend(dev, state);
-		suspend_report_result(dev->type->suspend, error);
-	}
-
 	if (!error && dev->bus && dev->bus->suspend
 	    && !dev->power.power_state.event) {
 		suspend_device_dbg(dev, state, "");
include/linux/device.h +0 -2
Index: linux-2.6.22-rc4-mm2/include/linux/device.h
===================================================================
@@ -340,8 +340,6 @@ struct device_type {
 	int (*uevent)(struct device *dev, char **envp, int num_envp,
 		      char *buffer, int buffer_size);
 	void (*release)(struct device *dev);
-	int (*suspend)(struct device * dev, pm_message_t state);
-	int (*resume)(struct device * dev);
 };
 
 /* interface for exporting device attributes */
#5 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Thanks,
Kay

Quoted 66 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend and resume support in struct device_type (include/linux/device.h) is not used anywhere. It is also undocumented, so it can be removed.

---
 drivers/base/power/resume.c  |    5 -----
 drivers/base/power/suspend.c |    7 -------
 include/linux/device.h       |    2 --
 3 files changed, 14 deletions(-)

Index: linux-2.6.22-rc4-mm2/drivers/base/power/resume.c
===================================================================

drivers/base/power/resume.c +0 -5
@@ -41,11 +41,6 @@ int resume_device(struct device * dev)
                error = dev->bus->resume(dev);
        }
 
-       if (!error && dev->type && dev->type->resume) {
-               dev_dbg(dev,"resuming\n");
-               error = dev->type->resume(dev);
-       }
-
        if (!error && dev->class && dev->class->resume) {
                dev_dbg(dev,"class resume\n");
                error = dev->class->resume(dev);

Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================

drivers/base/power/suspend.c +0 -7
@@ -79,13 +79,6 @@ int suspend_device(struct device * dev,
                suspend_report_result(dev->class->suspend, error);
        }
 
-       if (!error && dev->type && dev->type->suspend
-           && !dev->power.power_state.event) {
-               suspend_device_dbg(dev, state, "type ");
-               error = dev->type->suspend(dev, state);
-               suspend_report_result(dev->type->suspend, error);
-       }
-
        if (!error && dev->bus && dev->bus->suspend
            && !dev->power.power_state.event) {
                suspend_device_dbg(dev, state, "");

Index: linux-2.6.22-rc4-mm2/include/linux/device.h
===================================================================

include/linux/device.h +0 -3
@@ -340,8 +340,6 @@ struct device_type {
        int (*uevent)(struct device *dev, char **envp, int num_envp,
                      char *buffer, int buffer_size);
        void (*release)(struct device *dev);
-       int (*suspend)(struct device * dev, pm_message_t state);
-       int (*resume)(struct device * dev);
 };
 
 /* interface for exporting device attributes */
 
-

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#6 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore LED state and repeat rate for keyboards).

Signature · 1 line
Dmitry
#7 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 7 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore LED state and repeat rate for keyboards).

Would that be a big problem to reintroduce it along with the user?

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#8 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 11 lines · Rafael J. Wysocki
Quoted 7 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore LED state and repeat rate for keyboards).

Would that be a big problem to reintroduce it along with the user?

Yes because that part is in Greg's domain so I am trying to set infrastructure up before I can commit my patches into my tree so that Andrew can safely pull from me into -mm. Greg normally adds such stuff during merge window so that means if we remove it now we'd need ~2 releases to get it and the user back in.

Signature · 1 line
Dmitry
#9 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 11 lines · Dmitry Torokhov
Quoted 10 lines · Rafael J. Wysocki
Quoted 6 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore

LED

Quoted 1 line

state and repeat rate for keyboards).

Quoted 8 lines
Quoted 2 lines

Would that be a big problem to reintroduce it along with the user?

Yes because that part is in Greg's domain so I am trying to set infrastructure up before I can commit my patches into my tree so that Andrew can safely pull from me into -mm. Greg normally adds such stuff during merge window so that means if we remove it now we'd need ~2 releases to get it and the user back in.

Yes, don't worry, I'm not going to remove this as I know you rely on it.

Sorry for the delay in getting to these patches, I'm in meetings down in CA for the rest of the week...

thanks,

greg k-h

#10 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 15 lines · Greg KH
Quoted 14 lines · Dmitry Torokhov
Quoted 12 lines · Rafael J. Wysocki
Quoted 6 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore

LED

Quoted 1 line

state and repeat rate for keyboards).

Would that be a big problem to reintroduce it along with the user?

Quoted 7 lines
Quoted 5 lines

Yes because that part is in Greg's domain so I am trying to set infrastructure up before I can commit my patches into my tree so that Andrew can safely pull from me into -mm. Greg normally adds such stuff during merge window so that means if we remove it now we'd need ~2 releases to get it and the user back in.

Yes, don't worry, I'm not going to remove this as I know you rely on it.

Hm, in that case I'd have to rework the patches 5-7. Perhaps I should resend the entire series once again, without the $subject patch?

Quoted 2 lines

Sorry for the delay in getting to these patches, I'm in meetings down in CA for the rest of the week...

No big deal. :-)

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#11 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 22 lines · Rafael J. Wysocki
Quoted 21 lines · Greg KH
Quoted 19 lines · Dmitry Torokhov
Quoted 6 lines · Rafael J. Wysocki
Quoted 5 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Quoted 6 lines
Quoted 1 line

Yes, I will need it to implement input device resume (mainly to restore

LED

Quoted 1 line

state and repeat rate for keyboards).

Would that be a big problem to reintroduce it along with the user?

Yes because that part is in Greg's domain so I am trying to set infrastructure up before I can commit my patches into my tree so that Andrew can safely pull from me into -mm. Greg normally adds such stuff during merge window so that means if we remove it now we'd need ~2 releases to get it and the user back in.

Quoted 4 lines
Quoted 1 line

Yes, don't worry, I'm not going to remove this as I know you rely on it.

Hm, in that case I'd have to rework the patches 5-7. Perhaps I should resend the entire series once again, without the $subject patch?

You do? Conflicts are that bad?

Let me try and apply them and let you know what I've accepted in my tree...

thanks,

greg k-h

#12 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

In reply to Rafael J. Wysocki
Quoted 22 lines · Rafael J. Wysocki
Quoted 21 lines · Greg KH
Quoted 19 lines · Dmitry Torokhov
Quoted 6 lines · Rafael J. Wysocki
Quoted 5 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Quoted 6 lines
Quoted 1 line

Yes, I will need it to implement input device resume (mainly to restore

LED

Quoted 1 line

state and repeat rate for keyboards).

Would that be a big problem to reintroduce it along with the user?

Yes because that part is in Greg's domain so I am trying to set infrastructure up before I can commit my patches into my tree so that Andrew can safely pull from me into -mm. Greg normally adds such stuff during merge window so that means if we remove it now we'd need ~2 releases to get it and the user back in.

Quoted 4 lines
Quoted 1 line

Yes, don't worry, I'm not going to remove this as I know you rely on it.

Hm, in that case I'd have to rework the patches 5-7. Perhaps I should resend the entire series once again, without the $subject patch?

Ok, yeah, can you rework those three patches and resend them? I took patches 1-3 already.

thanks,

greg k-h

#13 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 23 lines · Greg KH
Quoted 22 lines · Rafael J. Wysocki
Quoted 21 lines · Greg KH
Quoted 19 lines · Dmitry Torokhov
Quoted 6 lines · Rafael J. Wysocki
Quoted 5 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Quoted 6 lines
Quoted 1 line

Yes, I will need it to implement input device resume (mainly to restore

LED

Quoted 1 line

state and repeat rate for keyboards).

Would that be a big problem to reintroduce it along with the user?

Yes because that part is in Greg's domain so I am trying to set infrastructure up before I can commit my patches into my tree so that Andrew can safely pull from me into -mm. Greg normally adds such stuff during merge window so that means if we remove it now we'd need ~2 releases to get it and the user back in.

Quoted 6 lines
Quoted 4 lines
Quoted 1 line

Yes, don't worry, I'm not going to remove this as I know you rely on it.

Hm, in that case I'd have to rework the patches 5-7. Perhaps I should resend the entire series once again, without the $subject patch?

Ok, yeah, can you rework those three patches and resend them?

Sure, I will.

Quoted 1 line

I took patches 1-3 already.

Thanks!

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#14 lore raw

cover PM: Remove unused and unnecessary features from core suspend code (continued)

In reply to Greg KH

Hi,

The next messages contain the remaining three patches from the previous series "PM: Remove unused and unnecessary features from suspend and resume core" that I have promised to resend.

The patches apply to 2.6.22-rc4-mm2, which is particularly important for the first one, as it is on top of some recent USB changes that right now are mm-only. If there are any problems with that, please let me know.

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#15 lore raw

PATCH 1/3 PM: Remove prev_state from struct dev_pm_info

From: Rafael J. Wysocki <[email protected]>

The prev_state member of struct dev_pm_info (defined in include/linux/pm.h) is only used during a resume to check if the device's state before the suspend was 'off', in which case the device is not resumed. However, in such cases the decision whether or not to resume the device should be made on the driver level and the resume callbacks from the device's bus and class should be executed anyway (the may be needed for some things other than just powering on the device).

---
 drivers/base/power/resume.c  |    3 +--
 drivers/base/power/suspend.c |    2 --
 drivers/usb/core/hub.c       |    5 -----
 drivers/usb/host/ohci-pci.c  |    2 --
 include/linux/pm.h           |    1 -
 5 files changed, 1 insertion(+), 12 deletions(-)
drivers/base/power/resume.c +1 -2
Index: linux-2.6.22-rc4-mm2/drivers/base/power/resume.c
===================================================================
@@ -88,8 +88,7 @@ void dpm_resume(void)
 		list_move_tail(entry, &dpm_active);
 
 		mutex_unlock(&dpm_list_mtx);
-		if (!dev->power.prev_state.event)
-			resume_device(dev);
+		resume_device(dev);
 		mutex_lock(&dpm_list_mtx);
 		put_device(dev);
 	}
drivers/base/power/suspend.c +0 -2
Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================
@@ -71,8 +71,6 @@ int suspend_device(struct device * dev, 
 			dev->parent->power.power_state.event);
 	}
 
-	dev->power.prev_state = dev->power.power_state;
-
 	if (dev->class && dev->class->suspend && !dev->power.power_state.event) {
 		suspend_device_dbg(dev, state, "class ");
 		error = dev->class->suspend(dev, state);
drivers/usb/core/hub.c +0 -5
Index: linux-2.6.22-rc4-mm2/drivers/usb/core/hub.c
===================================================================
@@ -623,7 +623,6 @@ static void mark_children_for_reset_resu
 
 		if (child) {
 			child->reset_resume = 1;
-			child->dev.power.prev_state.event = PM_EVENT_ON;
 			clear_port_feature(hdev, port1,
 					USB_PORT_FEAT_C_CONNECTION);
 		}
@@ -1190,10 +1189,6 @@ void usb_root_hub_lost_power(struct usb_
 	dev_warn(&rhdev->dev, "root hub lost power or was reset\n");
 	rhdev->reset_resume = 1;
 
-	/* Force the root hub to be resumed, even if it was suspended,
-	 * so that no wakeup events will get lost.
-	 */
-	rhdev->dev.power.prev_state.event = PM_EVENT_ON;
 }
 EXPORT_SYMBOL_GPL(usb_root_hub_lost_power);
 
include/linux/pm.h +0 -1
Index: linux-2.6.22-rc4-mm2/include/linux/pm.h
===================================================================
@@ -235,7 +235,6 @@ struct dev_pm_info {
 	unsigned		can_wakeup:1;
 #ifdef	CONFIG_PM
 	unsigned		should_wakeup:1;
-	pm_message_t		prev_state;
 	struct list_head	entry;
 #endif
 };
drivers/usb/host/ohci-pci.c +0 -2
Index: linux-2.6.22-rc4-mm2/drivers/usb/host/ohci-pci.c
===================================================================
@@ -281,8 +281,6 @@ static int ohci_pci_resume (struct usb_h
 	/* FIXME: we should try to detect loss of VBUS power here */
 	prepare_for_handover(hcd);
 
-	/* Force the PM core to resume the root hub */
-	hcd_to_bus(hcd)->root_hub->dev.power.prev_state.event = PM_EVENT_ON;
 	return 0;
 }
 
#16 lore raw

PATCH 2/3 PM: Remove power_state.event checks from suspend core code

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend routines should be called for every device during a system sleep transition, regardless of the device's state, so that drivers can regard these method calls as notifications that the system is about to go to sleep, rather than as directives to put their devices into the 'off' state.

This is documented in Documentation/power/devices.txt and is already done in the core resume code, so it seems reasonable to make the core suspend code behave accordingly.

---
 drivers/base/power/suspend.c |   11 ++++-------
 1 file changed, 4 insertions(+), 7 deletions(-)
drivers/base/power/suspend.c +4 -7
Index: linux-2.6.22-rc4/drivers/base/power/suspend.c
===================================================================
@@ -71,21 +71,19 @@ int suspend_device(struct device * dev, 
 			dev->parent->power.power_state.event);
 	}
 
-	if (dev->class && dev->class->suspend && !dev->power.power_state.event) {
+	if (dev->class && dev->class->suspend) {
 		suspend_device_dbg(dev, state, "class ");
 		error = dev->class->suspend(dev, state);
 		suspend_report_result(dev->class->suspend, error);
 	}
 
-	if (!error && dev->type && dev->type->suspend
-	    && !dev->power.power_state.event) {
+	if (!error && dev->type && dev->type->suspend) {
 		suspend_device_dbg(dev, state, "type ");
 		error = dev->type->suspend(dev, state);
 		suspend_report_result(dev->type->suspend, error);
 	}
 
-	if (!error && dev->bus && dev->bus->suspend
-	    && !dev->power.power_state.event) {
+	if (!error && dev->bus && dev->bus->suspend) {
 		suspend_device_dbg(dev, state, "");
 		error = dev->bus->suspend(dev, state);
 		suspend_report_result(dev->bus->suspend, error);
@@ -104,8 +102,7 @@ static int suspend_device_late(struct de
 {
 	int error = 0;
 
-	if (dev->bus && dev->bus->suspend_late
-	    && !dev->power.power_state.event) {
+	if (dev->bus && dev->bus->suspend_late) {
 		suspend_device_dbg(dev, state, "LATE ");
 		error = dev->bus->suspend_late(dev, state);
 		suspend_report_result(dev->bus->suspend_late, error);
#17 lore raw

PATCH 3/3 PM: Do not check parent state in suspend and resume core code

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The checks if the device's parent is in the right state done in drivers/base/power/suspend.c and drivers/base/power/resume.c serve no particular purpose, since if the parent is in a wrong power state, the device's suspend or resume callbacks are supposed to return an error anyway. Moreover, they are also useless from the sanity checking point of view, because they rely on the code being checked to set dev->parent->power.power_state.event appropriately, which need not happen if that code is buggy. For these reasons they can be removed.

---
 drivers/base/power/resume.c  |    7 -------
 drivers/base/power/suspend.c |    7 -------
 2 files changed, 14 deletions(-)
drivers/base/power/resume.c +0 -7
Index: linux-2.6.22-rc4/drivers/base/power/resume.c
===================================================================
@@ -29,13 +29,6 @@ int resume_device(struct device * dev)
 
 	down(&dev->sem);
 
-	if (dev->parent && dev->parent->power.power_state.event) {
-		dev_err(dev, "PM: resume from %d, parent %s still %d\n",
-			dev->power.power_state.event,
-			dev->parent->bus_id,
-			dev->parent->power.power_state.event);
-	}
-
 	if (dev->bus && dev->bus->resume) {
 		dev_dbg(dev,"resuming\n");
 		error = dev->bus->resume(dev);
drivers/base/power/suspend.c +0 -7
Index: linux-2.6.22-rc4/drivers/base/power/suspend.c
===================================================================
@@ -63,13 +63,6 @@ int suspend_device(struct device * dev, 
 		dev_dbg(dev, "PM: suspend %d-->%d\n",
 			dev->power.power_state.event, state.event);
 	}
-	if (dev->parent && dev->parent->power.power_state.event) {
-		dev_err(dev,
-			"PM: suspend %d->%d, parent %s already %d\n",
-			dev->power.power_state.event, state.event,
-			dev->parent->bus_id,
-			dev->parent->power.power_state.event);
-	}
 
 	if (dev->class && dev->class->suspend) {
 		suspend_device_dbg(dev, state, "class ");
#18 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

In reply to Dmitry Torokhov
Quoted 7 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore LED state and repeat rate for keyboards).

Why wouldn't that use the class device suspend/resume mechanism?

#19 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 11 lines · David Brownell
Quoted 7 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore LED state and repeat rate for keyboards).

Why wouldn't that use the class device suspend/resume mechanism?

Because they are not class devices anymore.

Signature · 1 line
Dmitry
#20 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 7 lines · Dmitry Torokhov
Quoted 6 lines · David Brownell
Quoted 5 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Quoted 7 lines
Quoted 5 lines
Quoted 2 lines

Yes, I will need it to implement input device resume (mainly to restore LED state and repeat rate for keyboards).

Why wouldn't that use the class device suspend/resume mechanism?

Because they are not class devices anymore.

Perhaps you mis-understood me: the class level device suspend/resume hooks don't rely on class_device. I observe that /sys/class/input still exists (2.6.22-rc4-git), so if all you mean is that it's no longer using class_device, that's good. One part of the reason to stop using class_device is specifically to let framework code (like input, network, or rtc) receive class level suspend/resume calls.

ISTR the RTC framework was the first to do that, but there's some work afoot to teach the network stack to use that mechanism too.

Or are you referring to some input patches which I've not yet seen, which cause /sys/class/input to vanish?

- Dave

#21 lore raw

Re: [PATCH -mm 4/7] PM: Remove suspend and resume support from struct device_type

Quoted 14 lines · David Brownell
Quoted 13 lines · Dmitry Torokhov
Quoted 11 lines · David Brownell
Quoted 7 lines · Dmitry Torokhov
Quoted 3 lines · Kay Sievers

Dmitry, you added this recently, is this used in any code you plan to merge soon?

Yes, I will need it to implement input device resume (mainly to restore LED state and repeat rate for keyboards).

Why wouldn't that use the class device suspend/resume mechanism?

Quoted 6 lines
Quoted 1 line

Because they are not class devices anymore.

Perhaps you mis-understood me: the class level device suspend/resume hooks don't rely on class_device. I observe that /sys/class/input still exists (2.6.22-rc4-git), so if all you mean is that it's no longer using class_device, that's good.

Yes, I did misunderstood you.

The reason I can't use class-wide resume support is that since we can't have class hierarchy we have mess of different objects dumped into single pool (input_dev, evdev. mousedev, joydev, tsdev all relate to /sys/class/input). And so instead of having class define standard interface for all objects in it we have to rely on something else (struct device_type - not visible fom userspace) to separate them and add tweaks to interface. Can you see that I am not a fan of the design? ;)

As far as input concerned - I need to resume input_dev but not evdev, mousedev, tsdev, etc. as they are all just different representations of the same object.

Quoted 3 lines

One part of the reason to
stop using class_device is specifically to let framework code (like input, network, or rtc) receive class level suspend/resume calls.

It could be done within the class_device infrastructure as well.

Quoted 6 lines

ISTR the RTC framework was the first to do that, but there's some work afoot to teach the network stack to use that mechanism too.

Or are you referring to some input patches which I've not yet seen, which cause /sys/class/input to vanish?

No, it will stay.

Signature · 1 line
Dmitry
#22 lore raw

PATCH 5/7 PM: Remove prev_state from struct dev_pm_info

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The prev_state member of struct dev_pm_info (defined in include/linux/pm.h) is only used during a resume to check if the device's state before the suspend was 'off', in which case the device is not resumed. However, in such cases the decision whether or not to resume the device should be made on the driver level and the resume callbacks from the device's bus and class should be executed anyway (the may be needed for some things other than just powering on the device).

---
 drivers/base/power/resume.c  |    3 +--
 drivers/base/power/suspend.c |    2 --
 drivers/usb/core/hub.c       |    5 -----
 drivers/usb/host/ohci-pci.c  |    2 --
 include/linux/pm.h           |    1 -
 5 files changed, 1 insertion(+), 12 deletions(-)
drivers/base/power/resume.c +1 -2
Index: linux-2.6.22-rc4-mm2/drivers/base/power/resume.c
===================================================================
@@ -83,8 +83,7 @@ void dpm_resume(void)
 		list_move_tail(entry, &dpm_active);
 
 		mutex_unlock(&dpm_list_mtx);
-		if (!dev->power.prev_state.event)
-			resume_device(dev);
+		resume_device(dev);
 		mutex_lock(&dpm_list_mtx);
 		put_device(dev);
 	}
drivers/base/power/suspend.c +0 -2
Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================
@@ -71,8 +71,6 @@ int suspend_device(struct device * dev, 
 			dev->parent->power.power_state.event);
 	}
 
-	dev->power.prev_state = dev->power.power_state;
-
 	if (dev->class && dev->class->suspend && !dev->power.power_state.event) {
 		suspend_device_dbg(dev, state, "class ");
 		error = dev->class->suspend(dev, state);
drivers/usb/core/hub.c +0 -5
Index: linux-2.6.22-rc4-mm2/drivers/usb/core/hub.c
===================================================================
@@ -623,7 +623,6 @@ static void mark_children_for_reset_resu
 
 		if (child) {
 			child->reset_resume = 1;
-			child->dev.power.prev_state.event = PM_EVENT_ON;
 			clear_port_feature(hdev, port1,
 					USB_PORT_FEAT_C_CONNECTION);
 		}
@@ -1190,10 +1189,6 @@ void usb_root_hub_lost_power(struct usb_
 	dev_warn(&rhdev->dev, "root hub lost power or was reset\n");
 	rhdev->reset_resume = 1;
 
-	/* Force the root hub to be resumed, even if it was suspended,
-	 * so that no wakeup events will get lost.
-	 */
-	rhdev->dev.power.prev_state.event = PM_EVENT_ON;
 }
 EXPORT_SYMBOL_GPL(usb_root_hub_lost_power);
 
include/linux/pm.h +0 -1
Index: linux-2.6.22-rc4-mm2/include/linux/pm.h
===================================================================
@@ -235,7 +235,6 @@ struct dev_pm_info {
 	unsigned		can_wakeup:1;
 #ifdef	CONFIG_PM
 	unsigned		should_wakeup:1;
-	pm_message_t		prev_state;
 	struct list_head	entry;
 #endif
 };
drivers/usb/host/ohci-pci.c +0 -2
Index: linux-2.6.22-rc4-mm2/drivers/usb/host/ohci-pci.c
===================================================================
@@ -281,8 +281,6 @@ static int ohci_pci_resume (struct usb_h
 	/* FIXME: we should try to detect loss of VBUS power here */
 	prepare_for_handover(hcd);
 
-	/* Force the PM core to resume the root hub */
-	hcd_to_bus(hcd)->root_hub->dev.power.prev_state.event = PM_EVENT_ON;
 	return 0;
 }
 
#23 lore raw

PATCH 6/7 PM: Remove power_state.event checks from suspend core code

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend routines should be called for every device during a system sleep transition, regardless of the device's state, so that drivers can regard these method calls as notifications that the system is about to go to sleep, rather than as directives to put their devices into the 'off' state.

This is documented in Documentation/power/devices.txt and is already done in the core resume code, so it seems reasonable to make the core suspend code behave accordingly.

---
 drivers/base/power/suspend.c |    8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)
drivers/base/power/suspend.c +3 -5
Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================
@@ -71,14 +71,13 @@ int suspend_device(struct device * dev, 
 			dev->parent->power.power_state.event);
 	}
 
-	if (dev->class && dev->class->suspend && !dev->power.power_state.event) {
+	if (dev->class && dev->class->suspend) {
 		suspend_device_dbg(dev, state, "class ");
 		error = dev->class->suspend(dev, state);
 		suspend_report_result(dev->class->suspend, error);
 	}
 
-	if (!error && dev->bus && dev->bus->suspend
-	    && !dev->power.power_state.event) {
+	if (!error && dev->bus && dev->bus->suspend) {
 		suspend_device_dbg(dev, state, "");
 		error = dev->bus->suspend(dev, state);
 		suspend_report_result(dev->bus->suspend, error);
@@ -97,8 +96,7 @@ static int suspend_device_late(struct de
 {
 	int error = 0;
 
-	if (dev->bus && dev->bus->suspend_late
-	    && !dev->power.power_state.event) {
+	if (dev->bus && dev->bus->suspend_late) {
 		suspend_device_dbg(dev, state, "LATE ");
 		error = dev->bus->suspend_late(dev, state);
 		suspend_report_result(dev->bus->suspend_late, error);
#24 lore raw

Re: [PATCH -mm 6/7] PM: Remove power_state.event checks from suspend core code

Quoted 6 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend routines should be called for every device during a system sleep transition, regardless of the device's state, so that drivers can regard these method calls as notifications that the system is about to go to sleep, rather than as directives to put their devices into the 'off' state.

Did you audit all the drivers to make sure this won't break things? Like for example through inappropriate pci_save_state() calls?

I'd really expect this patch would break things...

#25 lore raw

Re: [PATCH -mm 6/7] PM: Remove power_state.event checks from suspend core code

Quoted 10 lines · David Brownell
Quoted 6 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend routines should be called for every device during a system sleep transition, regardless of the device's state, so that drivers can regard these method calls as notifications that the system is about to go to sleep, rather than as directives to put their devices into the 'off' state.

Did you audit all the drivers to make sure this won't break things? Like for example through inappropriate pci_save_state() calls?

I did, but not very carefully.

Quoted 1 line

I'd really expect this patch would break things...

Well, in that case I'll have a closer look at them.

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
maintainer
#26 lore raw

Re: [PATCH -mm 6/7] PM: Remove power_state.event checks from suspend core code

Quoted 9 lines · Rafael J. Wysocki
Quoted 8 lines · David Brownell
Quoted 6 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend routines should be called for every device during a system sleep transition, regardless of the device's state, so that drivers can regard these method calls as notifications that the system is about to go to sleep, rather than as directives to put their devices into the 'off' state.

Quoted 8 lines
Quoted 2 lines

Did you audit all the drivers to make sure this won't break things? Like for example through inappropriate pci_save_state() calls?

I did, but not very carefully.

Quoted 1 line

I'd really expect this patch would break things...

Well, in that case I'll have a closer look at them.

It might not be all that bad. One would expect problems to occur only in cases where devices were already suspended at the time of a system sleep transition. Since relatively few drivers currently implement runtime PM -- and those that do are likely to be more careful about not blindly making state changes -- there might not be too much trouble.

Alan Stern

#27 lore raw

Re: [PATCH -mm 6/7] PM: Remove power_state.event checks from suspend core code

Quoted 18 lines · Alan Stern
Quoted 16 lines · Rafael J. Wysocki
Quoted 10 lines · David Brownell
Quoted 6 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend routines should be called for every device during a system sleep transition, regardless of the device's state, so that drivers can regard these method calls as notifications that the system is about to go to sleep, rather than as directives to put their devices into the 'off' state.

Did you audit all the drivers to make sure this won't break things? Like for example through inappropriate pci_save_state() calls?

I did, but not very carefully.

Quoted 1 line

I'd really expect this patch would break things...

Quoted 8 lines
Quoted 1 line

Well, in that case I'll have a closer look at them.

It might not be all that bad. One would expect problems to occur only in cases where devices were already suspended at the time of a system sleep transition. Since relatively few drivers currently implement runtime PM -- and those that do are likely to be more careful about not blindly making state changes -- there might not be too much trouble.

Yes, in fact I've had no problems related to that so far (tested the patch on four different machines).

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#28 lore raw

Re: [PATCH -mm 6/7] PM: Remove power_state.event checks from suspend core code

Quoted 28 lines · Rafael J. Wysocki
Quoted 18 lines · Alan Stern
Quoted 16 lines · Rafael J. Wysocki
Quoted 10 lines · David Brownell
Quoted 6 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The suspend routines should be called for every device during a system sleep transition, regardless of the device's state, so that drivers can regard these method calls as notifications that the system is about to go to sleep, rather than as directives to put their devices into the 'off' state.

Did you audit all the drivers to make sure this won't break things? Like for example through inappropriate pci_save_state() calls?

I did, but not very carefully.

Quoted 1 line

I'd really expect this patch would break things...

Quoted 8 lines
Quoted 1 line

Well, in that case I'll have a closer look at them.

It might not be all that bad. One would expect problems to occur only in cases where devices were already suspended at the time of a system sleep transition. Since relatively few drivers currently implement runtime PM -- and those that do are likely to be more careful about not blindly making state changes -- there might not be too much trouble.

Quoted 2 lines

Yes, in fact I've had no problems related to that so far (tested the patch on four different machines).

It seems the drivers for which that could be relevant do the checks as needed.

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#29 lore raw

PATCH 7/7 PM: Do not check parent state in suspend and resume core code

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

The checks if the device's parent is in the right state done in drivers/base/power/suspend.c and drivers/base/power/resume.c serve no particular purpose, since if the parent is in a wrong power state, the device's suspend or resume callbacks are supposed to return an error anyway. Moreover, they are also useless from the sanity checking point of view, because they rely on the code being checked to set dev->parent->power.power_state.event appropriately, which need not happen if that code is buggy. For these reasons they can be removed.

---
 drivers/base/power/resume.c  |    7 -------
 drivers/base/power/suspend.c |    7 -------
 2 files changed, 14 deletions(-)
drivers/base/power/resume.c +0 -7
Index: linux-2.6.22-rc4-mm2/drivers/base/power/resume.c
===================================================================
@@ -29,13 +29,6 @@ int resume_device(struct device * dev)
 
 	down(&dev->sem);
 
-	if (dev->parent && dev->parent->power.power_state.event) {
-		dev_err(dev, "PM: resume from %d, parent %s still %d\n",
-			dev->power.power_state.event,
-			dev->parent->bus_id,
-			dev->parent->power.power_state.event);
-	}
-
 	if (dev->bus && dev->bus->resume) {
 		dev_dbg(dev,"resuming\n");
 		error = dev->bus->resume(dev);
drivers/base/power/suspend.c +0 -7
Index: linux-2.6.22-rc4-mm2/drivers/base/power/suspend.c
===================================================================
@@ -63,13 +63,6 @@ int suspend_device(struct device * dev, 
 		dev_dbg(dev, "PM: suspend %d-->%d\n",
 			dev->power.power_state.event, state.event);
 	}
-	if (dev->parent && dev->parent->power.power_state.event) {
-		dev_err(dev,
-			"PM: suspend %d->%d, parent %s already %d\n",
-			dev->power.power_state.event, state.event,
-			dev->parent->bus_id,
-			dev->parent->power.power_state.event);
-	}
 
 	if (dev->class && dev->class->suspend) {
 		suspend_device_dbg(dev, state, "class ");
#30 lore raw
In reply to Rafael J. Wysocki
Quoted 4 lines · Rafael J. Wysocki

Hi,

The following series of patches removes some unused and unnecessary features from the suspend and resume core code.

These are the same as you sent out the other day, right?

If so, I'll put them in my tree, trying to catch up on my queue right now.

thanks,

greg k-h

#31 lore raw
Quoted 7 lines · Greg KH
Quoted 4 lines · Rafael J. Wysocki

Hi,

The following series of patches removes some unused and unnecessary features from the suspend and resume core code.

These are the same as you sent out the other day, right?

Almost.

The "PM: Remove pm_parent from struct dev_pm_info" (1/7) has been updated.

The "PM: Remove prev_state from struct dev_pm_info" patch (5/7) has been rediffed against 2.6.22-rc4-mm2 (to synchronize with the USB changes in there).

The others are the same.

Quoted 2 lines

If so, I'll put them in my tree, trying to catch up on my queue right now.

OK

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#32 lore raw
In reply to Rafael J. Wysocki

Hi!

Quoted 2 lines

The following series of patches removes some unused and unnecessary features from the suspend and resume core code.

Hmm, patches 1-4 are pretty much obvious cleanups, I guess they can go in.

Patches 5+ change semantics of suspend/resume callbacks in subtle way, I guess they need to stay in -mm for a while...
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html