[RFC PATCH V1 0/2] cpuidle: global registration of idle states with per-cpu statistics

[RFC PATCH V1 0/2] cpuidle: global registration of idle states with per-cpu statistics

[Trinabh Gupta]

This patch series is an early RFC to discuss the feasibility of
avoiding registering of idle states from each cpu.

The core change is to split the cpuidle_device structure into parts
that can be global and parts that has to remain per-cpu.  The per-cpu
pieces are mostly generic statistics that can be independent of
current running driver.

Motivation:
* Simplify the cpuidle subsystem framework and have
  registration/unregistration done by single cpu.

* Minimise the data structure that needs to be maintained for multiple
  cpuidle drivers

* Reference: https://lkml.org/lkml/2011/2/10/37

Advantages:
* Make the cpuidle framework simple for most use cases where C-States
  are symmetric.  In case there are asymmetric C-States detected,
  fallback mechanism should be incorporated to maintain the system
  functional
  https://lkml.org/lkml/2011/2/10/257
  https://lkml.org/lkml/2011/2/10/37

* Non x86 archs that does not have asymmetric C-States like POWER, may
  not need the fallback mechanism and hence the framework will be
  simple for most use cases.

Disadvantages:
* Asymmetric C-States are part of x86 ACPI specification.  Incorrect
  handling may functionally affect the system

* Incorporating per-cpu masks for each state to allow/dis-allow global
  states on subset of CPUs may result in an implementation that is
  not better than current solution of having per-cpu states.

This patch series applies on top of the pm_idle cleanup patch
https://lkml.org/lkml/2011/3/22/150 (cpuidle: Cleanup pm_idle and 
include driver/cpuidle.c in-kernel)

This patch series is tested on x86 Nehalem system with multiple ACPI
C-States.

This patch series has limitations of not handling multiple driver
registration and switching between drivers on all CPUs mainly due to
incomplete handling of per-cpu enable/disable and driver_data.

Please let us know your comments and suggest possible approaches to
the problem.

---

Trinabh Gupta (2):
      cpuidle: API changes in callers using new cpuidle_state_stats
      cpuidle: Data structure changes for global cpuidle device


 drivers/acpi/processor_driver.c    |   19 +---
 drivers/acpi/processor_idle.c      |   45 ++++++----
 drivers/cpuidle/cpuidle.c          |  169 ++++++++++++++++++++++--------------
 drivers/cpuidle/cpuidle.h          |    4 -
 drivers/cpuidle/driver.c           |   24 -----
 drivers/cpuidle/governor.c         |   10 +-
 drivers/cpuidle/governors/Makefile |    2
 drivers/cpuidle/governors/menu.c   |   16 ++-
 drivers/cpuidle/sysfs.c            |   62 +++++++------
 drivers/idle/default_driver.c      |   39 +++-----
 include/linux/cpuidle.h            |   70 ++++++++++-----
 11 files changed, 245 insertions(+), 215 deletions(-)

--
-Trinabh

[RFC PATCH V1 1/2] cpuidle: Data structure changes for global cpuidle device

[Trinabh Gupta]

Currently cpuidle subsystem has the following data structures:

* Global cpuidle_driver which is maintained in a list for each
  registered driver

* cpuidle_device is per-cpu per-driver structure that contains the
  array of cpuidle_state defined for that CPU

* cpuidle_state contains the definition of the idle routine and place
  holders for statistics that are used by the governor to make the
  selection.

Proposed new data structures:

* Basically the cpuidle_state structure along with the cpuidle_device
  is made global and per-driver. The cpuidle_state contains the idle state
  definitions.

* The statistics part of the cpuidle_state that is needed by the
  governor on a per-cpu bases is split into a new per-cpu structure
  called cpuidle_state_stats.  This structure is generic and
  independent of the cpuidle_driver/device that is currently in force.


<<- Global--->>                 <<per-cpu>>

cpuidle_driver                  cpuidle_device_stats
  |                                    -> state_stats[]
  \
    -> cpuidle_device
        -> states[]


Open Issues:

* Handle per-cpu, per-driver's driver_data appropriately within the
  cpuidle_driver structure.  For this RFC, it is pushed into
  cpuidle_device_stats structure.

* Handle double initialization of kobj structs on multiple driver
  registration.

Signed-off-by: Trinabh Gupta <trinabh@linux.vnet.ibm.com>
---

 include/linux/cpuidle.h |   75 +++++++++++++++++++++++++++++++++--------------
 1 files changed, 52 insertions(+), 23 deletions(-)

diff --git a/include/linux/cpuidle.h b/include/linux/cpuidle.h
index a5f5fd0..16423df 100644
--- a/include/linux/cpuidle.h
+++ b/include/linux/cpuidle.h
@@ -22,6 +22,7 @@
 #define CPUIDLE_DESC_LEN	32
 
 struct cpuidle_device;
+struct cpuidle_state_stats;
 
 
 /****************************
@@ -31,18 +32,21 @@ struct cpuidle_device;
 struct cpuidle_state {
 	char		name[CPUIDLE_NAME_LEN];
 	char		desc[CPUIDLE_DESC_LEN];
-	void		*driver_data;
 
 	unsigned int	flags;
 	unsigned int	exit_latency; /* in US */
 	unsigned int	power_usage; /* in mW */
 	unsigned int	target_residency; /* in US */
 
-	unsigned long long	usage;
-	unsigned long long	time; /* in US */
-
 	int (*enter)	(struct cpuidle_device *dev,
-			 struct cpuidle_state *state);
+			 struct cpuidle_state *state,
+			 struct cpuidle_state_stats *state_stats);
+};
+
+struct cpuidle_state_stats {
+	void		*driver_data;
+	unsigned long long usage;
+	unsigned long long time; /* in US */
 };
 
 /* Idle State Flags */
@@ -55,9 +59,9 @@ struct cpuidle_state {
  * cpuidle_get_statedata - retrieves private driver state data
  * @state: the state
  */
-static inline void * cpuidle_get_statedata(struct cpuidle_state *state)
+static inline void *cpuidle_get_statedata(struct cpuidle_state_stats *stats)
 {
-	return state->driver_data;
+	return stats->driver_data;
 }
 
 /**
@@ -66,9 +70,9 @@ static inline void * cpuidle_get_statedata(struct cpuidle_state *state)
  * @data: the private data
  */
 static inline void
-cpuidle_set_statedata(struct cpuidle_state *state, void *data)
+cpuidle_set_statedata(struct cpuidle_state_stats *state_stats, void *data)
 {
-	state->driver_data = data;
+	state_stats->driver_data = data;
 }
 
 struct cpuidle_state_kobj {
@@ -77,29 +81,42 @@ struct cpuidle_state_kobj {
 	struct kobject kobj;
 };
 
+struct cpuidle_device_stats {
+	unsigned int			enabled:1;
+	unsigned int			cpu;
+	int				last_residency;
+	struct cpuidle_state		*last_state;
+	struct cpuidle_state_stats	state_stats[CPUIDLE_STATE_MAX];
+
+	struct cpuidle_state_kobj	*kobjs[CPUIDLE_STATE_MAX];
+	struct kobject			kobj;
+	struct completion		kobj_unregister;
+
+	struct cpuidle_driver		*drv;
+};
+
 struct cpuidle_device {
 	unsigned int		registered:1;
 	unsigned int		enabled:1;
 	unsigned int		power_specified:1;
-	unsigned int		cpu;
 
-	int			last_residency;
 	int			state_count;
 	struct cpuidle_state	states[CPUIDLE_STATE_MAX];
-	struct cpuidle_state_kobj *kobjs[CPUIDLE_STATE_MAX];
-	struct cpuidle_state	*last_state;
 
-	struct list_head 	device_list;
 	struct cpuidle_driver	*drv;
-	struct kobject		kobj;
-	struct completion	kobj_unregister;
 	void			*governor_data;
 	struct cpuidle_state	*safe_state;
 
-	int (*prepare)		(struct cpuidle_device *dev);
+	int (*prepare)		(struct cpuidle_device_stats *dev_stats);
 };
 
-DECLARE_PER_CPU(struct cpuidle_device *, cpuidle_devices);
+
+DECLARE_PER_CPU(struct cpuidle_device_stats, cpuidle_dev_stats);
+static inline struct cpuidle_state_stats *get_cpuidle_state_stats(int cpu,
+						int count)
+{
+	return &per_cpu(cpuidle_dev_stats, cpu).state_stats[count];
+}
 
 /**
  * cpuidle_get_last_residency - retrieves the last state's residency time
@@ -107,7 +124,7 @@ DECLARE_PER_CPU(struct cpuidle_device *, cpuidle_devices);
  *
  * NOTE: this value is invalid if CPUIDLE_FLAG_TIME_VALID isn't set
  */
-static inline int cpuidle_get_last_residency(struct cpuidle_device *dev)
+static inline int cpuidle_get_last_residency(struct cpuidle_device_stats *dev)
 {
 	return dev->last_residency;
 }
@@ -122,6 +139,7 @@ struct cpuidle_driver {
 	struct module 		*owner;
 	unsigned int		priority;
 	struct list_head	driver_list;
+	struct cpuidle_device	*dev;
 };
 
 #ifdef CONFIG_CPU_IDLE
@@ -145,12 +163,22 @@ static inline struct cpuidle_driver *cpuidle_get_driver(void) {return NULL; }
 static inline void cpuidle_unregister_driver(struct cpuidle_driver *drv) { }
 static inline int cpuidle_register_device(struct cpuidle_device *dev)
 {return -ENODEV; }
+static inline int cpuidle_register_device_stats(
+				struct cpuidle_device_stats *dev_stats)
+{return -ENODEV; }
+static inline void cpuidle_unregister_device_stats(
+			struct cpuidle_device_stats *dev_stats) { };
 static inline void cpuidle_unregister_device(struct cpuidle_device *dev) { }
 
 static inline void cpuidle_pause_and_lock(void) { }
 static inline void cpuidle_resume_and_unlock(void) { }
 static inline int cpuidle_enable_device(struct cpuidle_device *dev)
 {return -ENODEV; }
+static inline int cpuidle_enable_device_stats(
+			struct cpuidle_device_stats *dev_stats)
+{return -ENODEV; }
+static inline void cpuidle_disable_device_stats(
+				struct cpuidle_device_stats *dev_stats) { }
 static inline void cpuidle_disable_device(struct cpuidle_device *dev) { }
 
 #endif
@@ -164,11 +192,12 @@ struct cpuidle_governor {
 	struct list_head 	governor_list;
 	unsigned int		rating;
 
-	int  (*enable)		(struct cpuidle_device *dev);
-	void (*disable)		(struct cpuidle_device *dev);
+	int  (*enable)		(struct cpuidle_device_stats *dev_stats);
+	void (*disable)		(struct cpuidle_device_stats *dev_stats);
 
-	int  (*select)		(struct cpuidle_device *dev);
-	void (*reflect)		(struct cpuidle_device *dev);
+	int  (*select)		(struct cpuidle_device *dev,
+					struct cpuidle_device_stats *dev_stats);
+	void (*reflect)		(struct cpuidle_device_stats *dev_stats);
 
 	struct module 		*owner;
 };

Re: [RFC PATCH V1 1/2] cpuidle: Data structure changes for global cpuidle device

[Len Brown]

I agree it is silly to allocate a cpuidle_device
for every cpu in the system as we do today.

Yes, splitting the counters out of cpuidle_device
is a necessary part of fixing that.

However, cpuidle_device.cpuidle_state[] is currently not per-driver,
it is per-cpu, and it is writable.

In particular, the cpuidle_device->prepare() mechanism
causes updates to the cpuidle_state[].flags,
setting and clearing CPUIDLE_FLAG_IGNORE to
tell the governor not to chose a state
on a per-cpu basis at run-time.

I don't like that mechanism.
I'd like to see it replaced, and when replaced,
cpuidle_state[] can be per system-wide driver.

I think the real problem that prepare() was trying to solve
is that the driver today does not have the ability to over-rule
the choice made by the governor.  The driver may discover
in the course of trying to satisfy the request of the governor
that it needs to demote to a shallower state; or it may
do its best to satisfy the governor's request, and the hardware
may demote its request to a shallower state.

Unfortunately, when this happens, the driver dutifully
returns the time spent in the state to cpuidle_idle_call(),
who then updates the wrong last_residency, time, and usage counters.

Sure is ironic for the driver to allocate the data structures and
then hand the timer to the uppper layer, just to have the upper layer
update the wrong data structures...

Surely the driver enter routine should update the counters
that the driver was obligated to allocate, and it should return
the state actually entered (for tracing), rather than the time spent
there.

The generic cpuidle code should simply handle where the counters live
in the sysfs namespace, not updating the counters.

This needs to be addressed before cpuidle_device.cpuidle_state[]
can be made one/system.

cheers,
Len Brown, Intel Open Source Technology Center

Re: [RFC PATCH V1 1/2] cpuidle: Data structure changes for global cpuidle device

[Vaidyanathan Srinivasan]

* Len Brown <lenb@kernel.org> [2011-03-25 04:12:03]:

> I agree it is silly to allocate a cpuidle_device
> for every cpu in the system as we do today.
> 
> Yes, splitting the counters out of cpuidle_device
> is a necessary part of fixing that.
> 
> However, cpuidle_device.cpuidle_state[] is currently not per-driver,
> it is per-cpu, and it is writable.
> 
> In particular, the cpuidle_device->prepare() mechanism
> causes updates to the cpuidle_state[].flags,
> setting and clearing CPUIDLE_FLAG_IGNORE to
> tell the governor not to chose a state
> on a per-cpu basis at run-time.
> 
> I don't like that mechanism.
> I'd like to see it replaced, and when replaced,
> cpuidle_state[] can be per system-wide driver.

Thanks for the detailed review.  I agree that we should rework
handling of the cpuidle_state[].flags.  However, is the prepare()
mechanism used at all?  Can we remove the option completely?

> I think the real problem that prepare() was trying to solve
> is that the driver today does not have the ability to over-rule
> the choice made by the governor.  The driver may discover
> in the course of trying to satisfy the request of the governor
> that it needs to demote to a shallower state; or it may
> do its best to satisfy the governor's request, and the hardware
> may demote its request to a shallower state.
> 
> Unfortunately, when this happens, the driver dutifully
> returns the time spent in the state to cpuidle_idle_call(),
> who then updates the wrong last_residency, time, and usage counters.

I did not get this scenario.  Are you saying 

target_state->enter(dev, target_state) can enter a different state
than the one suggested by target_state?  

I understand the hardware demotion part, but can we really detect the
target 'demoted' state in that case?  I guess not.

> Sure is ironic for the driver to allocate the data structures and
> then hand the timer to the uppper layer, just to have the upper layer
> update the wrong data structures...
> 
> Surely the driver enter routine should update the counters
> that the driver was obligated to allocate, and it should return
> the state actually entered (for tracing), rather than the time spent
> there.

Can we do something like this:

last_state = target_state->enter(dev, target_state)

dev->last_state and dev->last_residency are updated inside
target_state->enter() 

The returned last_state is just for tracing, actual data is already
updated in the cpuidle_dev structure and used for sysfs display.

> The generic cpuidle code should simply handle where the counters live
> in the sysfs namespace, not updating the counters.
> This needs to be addressed before cpuidle_device.cpuidle_state[]
> can be made one/system.

Agreed.

Thanks again for the recommendations.

--Vaidy

Re: [RFC PATCH V1 1/2] cpuidle: Data structure changes for global cpuidle device

[Len Brown]

> > I think the real problem that prepare() was trying to solve
> > is that the driver today does not have the ability to over-rule
> > the choice made by the governor.  The driver may discover
> > in the course of trying to satisfy the request of the governor
> > that it needs to demote to a shallower state; or it may
> > do its best to satisfy the governor's request, and the hardware
> > may demote its request to a shallower state.
> > 
> > Unfortunately, when this happens, the driver dutifully
> > returns the time spent in the state to cpuidle_idle_call(),
> > who then updates the wrong last_residency, time, and usage counters.
> 
> I did not get this scenario.  Are you saying 
> 
> target_state->enter(dev, target_state) can enter a different state
> than the one suggested by target_state?  

Yes.

> I understand the hardware demotion part, but can we really detect the
> target 'demoted' state in that case?  I guess not.

In addition to HW demotion, SW may also notice at the last minute
that some criteria for entering a state is not met, and need
to demote in SW.

Effectively it is the same as HW demotion, except SW can see
that it is going to happen ahead of time.

The example is MRST S0i3, where a number of dependencies on
SOC device state must be met to enter the state, and those
dependencies can change state at any time.

> > Sure is ironic for the driver to allocate the data structures and
> > then hand the timer to the uppper layer, just to have the upper layer
> > update the wrong data structures...
> > 
> > Surely the driver enter routine should update the counters
> > that the driver was obligated to allocate, and it should return
> > the state actually entered (for tracing), rather than the time spent
> > there.
> 
> Can we do something like this:
> 
> last_state = target_state->enter(dev, target_state)
> 
> dev->last_state and dev->last_residency are updated inside
> target_state->enter() 
> 
> The returned last_state is just for tracing, actual data is already
> updated in the cpuidle_dev structure and used for sysfs display.

dev->last_state and
dev->last_residency
seem like part of hacky support for run-time changing of c-states.

I think if we move the couter incrementing into the driver,
then the upper layer shouldn't need to care about it.
The driver will always have a better idea of what happened
than cpuidle_idle_call().

> > The generic cpuidle code should simply handle where the counters live
> > in the sysfs namespace, not updating the counters.
> > This needs to be addressed before cpuidle_device.cpuidle_state[]
> > can be made one/system.
> 
> Agreed.

thanks,
Len Brown, Intel Open Source Technolgy Center

[RFC PATCH V1 2/2] cpuidle: API changes in callers using new cpuidle_state_stats

[Trinabh Gupta]

cpuidle subsystem users will need to use struct cpuidle_state *state
and struct cpuidle_state_stats *state_stats.

Driver's per-cpu context is stored in cpuidle_state_stats for this
RFC. This need to be moved into cpuidle_driver.

Signed-off-by: Trinabh Gupta <trinabh@linux.vnet.ibm.com>
---

 drivers/acpi/processor_driver.c    |   19 +---
 drivers/acpi/processor_idle.c      |   49 +++++++---
 drivers/cpuidle/cpuidle.c          |  170 +++++++++++++++++++++++-------------
 drivers/cpuidle/cpuidle.h          |    4 -
 drivers/cpuidle/driver.c           |   24 -----
 drivers/cpuidle/governor.c         |   10 +-
 drivers/cpuidle/governors/Makefile |    2 
 drivers/cpuidle/governors/menu.c   |   19 ++--
 drivers/cpuidle/sysfs.c            |   63 +++++++------
 drivers/idle/default_driver.c      |   40 +++-----
 10 files changed, 210 insertions(+), 190 deletions(-)

diff --git a/drivers/acpi/processor_driver.c b/drivers/acpi/processor_driver.c
index 360a74e..4837861 100644
--- a/drivers/acpi/processor_driver.c
+++ b/drivers/acpi/processor_driver.c
@@ -503,9 +503,7 @@ static int __cpuinit acpi_processor_add(struct acpi_device *device)
 	acpi_processor_get_throttling_info(pr);
 	acpi_processor_get_limit_info(pr);
 
-
-	if (cpuidle_get_driver() == &acpi_idle_driver)
-		acpi_processor_power_init(pr, device);
+	acpi_processor_power_init(pr, device);
 
 	pr->cdev = thermal_cooling_device_register("Processor", device,
 						&processor_cooling_ops);
@@ -800,17 +798,9 @@ static int __init acpi_processor_init(void)
 
 	memset(&errata, 0, sizeof(errata));
 
-	if (!cpuidle_register_driver(&acpi_idle_driver)) {
-		printk(KERN_DEBUG "ACPI: %s registered with cpuidle\n",
-			acpi_idle_driver.name);
-	} else {
-		printk(KERN_DEBUG "ACPI: acpi_idle yielding to %s\n",
-			cpuidle_get_driver()->name);
-	}
-
 	result = acpi_bus_register_driver(&acpi_processor_driver);
 	if (result < 0)
-		goto out_cpuidle;
+		goto out;
 
 	acpi_processor_install_hotplug_notify();
 
@@ -822,8 +812,7 @@ static int __init acpi_processor_init(void)
 
 	return 0;
 
-out_cpuidle:
-	cpuidle_unregister_driver(&acpi_idle_driver);
+out:
 
 	return result;
 }
@@ -841,8 +830,6 @@ static void __exit acpi_processor_exit(void)
 
 	acpi_bus_unregister_driver(&acpi_processor_driver);
 
-	cpuidle_unregister_driver(&acpi_idle_driver);
-
 	return;
 }
 
diff --git a/drivers/acpi/processor_idle.c b/drivers/acpi/processor_idle.c
index 65dde00..407acd7 100644
--- a/drivers/acpi/processor_idle.c
+++ b/drivers/acpi/processor_idle.c
@@ -746,12 +746,13 @@ static inline void acpi_idle_do_entry(struct acpi_processor_cx *cx)
  * This is equivalent to the HALT instruction.
  */
 static int acpi_idle_enter_c1(struct cpuidle_device *dev,
-			      struct cpuidle_state *state)
+			      struct cpuidle_state *state,
+				struct cpuidle_state_stats *state_stats)
 {
 	ktime_t  kt1, kt2;
 	s64 idle_time;
 	struct acpi_processor *pr;
-	struct acpi_processor_cx *cx = cpuidle_get_statedata(state);
+	struct acpi_processor_cx *cx = cpuidle_get_statedata(state_stats);
 
 	pr = __this_cpu_read(processors);
 
@@ -786,13 +787,14 @@ static int acpi_idle_enter_c1(struct cpuidle_device *dev,
  * @state: the state data
  */
 static int acpi_idle_enter_simple(struct cpuidle_device *dev,
-				  struct cpuidle_state *state)
+				  struct cpuidle_state *state,
+				struct cpuidle_state_stats *state_stats)
 {
 	struct acpi_processor *pr;
-	struct acpi_processor_cx *cx = cpuidle_get_statedata(state);
 	ktime_t  kt1, kt2;
 	s64 idle_time_ns;
 	s64 idle_time;
+	struct acpi_processor_cx *cx = cpuidle_get_statedata(state_stats);
 
 	pr = __this_cpu_read(processors);
 
@@ -800,7 +802,7 @@ static int acpi_idle_enter_simple(struct cpuidle_device *dev,
 		return 0;
 
 	if (acpi_idle_suspend)
-		return(acpi_idle_enter_c1(dev, state));
+		return acpi_idle_enter_c1(dev, state, state_stats);
 
 	local_irq_disable();
 
@@ -862,14 +864,15 @@ static DEFINE_SPINLOCK(c3_lock);
  * If BM is detected, the deepest non-C3 idle state is entered instead.
  */
 static int acpi_idle_enter_bm(struct cpuidle_device *dev,
-			      struct cpuidle_state *state)
+			      struct cpuidle_state *state,
+				struct cpuidle_state_stats *state_stats)
 {
 	struct acpi_processor *pr;
-	struct acpi_processor_cx *cx = cpuidle_get_statedata(state);
 	ktime_t  kt1, kt2;
 	s64 idle_time_ns;
 	s64 idle_time;
-
+	struct cpuidle_device_stats *dev_stats;
+	struct acpi_processor_cx *cx = cpuidle_get_statedata(state_stats);
 
 	pr = __this_cpu_read(processors);
 
@@ -877,12 +880,15 @@ static int acpi_idle_enter_bm(struct cpuidle_device *dev,
 		return 0;
 
 	if (acpi_idle_suspend)
-		return(acpi_idle_enter_c1(dev, state));
+		return acpi_idle_enter_c1(dev, state, state_stats);
 
 	if (!cx->bm_sts_skip && acpi_idle_bm_check()) {
 		if (dev->safe_state) {
-			dev->last_state = dev->safe_state;
-			return dev->safe_state->enter(dev, dev->safe_state);
+			dev_stats = &per_cpu(cpuidle_dev_stats,
+						smp_processor_id());
+			dev_stats->last_state = dev->safe_state;
+			return dev->safe_state->enter(dev, dev->safe_state,
+							state_stats);
 		} else {
 			local_irq_disable();
 			acpi_safe_halt();
@@ -983,6 +989,7 @@ static int acpi_processor_setup_cpuidle(struct acpi_processor *pr)
 	int i, count = CPUIDLE_DRIVER_STATE_START;
 	struct acpi_processor_cx *cx;
 	struct cpuidle_state *state;
+	struct cpuidle_state_stats *state_stats;
 	struct cpuidle_device *dev = &pr->power.dev;
 
 	if (!pr->flags.power_setup_done)
@@ -992,7 +999,6 @@ static int acpi_processor_setup_cpuidle(struct acpi_processor *pr)
 		return -EINVAL;
 	}
 
-	dev->cpu = pr->id;
 	dev->drv = &acpi_idle_driver;
 	for (i = 0; i < CPUIDLE_STATE_MAX; i++) {
 		dev->states[i].name[0] = '\0';
@@ -1005,6 +1011,12 @@ static int acpi_processor_setup_cpuidle(struct acpi_processor *pr)
 	for (i = 1; i < ACPI_PROCESSOR_MAX_POWER && i <= max_cstate; i++) {
 		cx = &pr->power.states[i];
 		state = &dev->states[count];
+		/*
+		 * To Do: driver_data would have to be made per
+		 * cpuidle driver. It works as there is just one
+		 * driver using this.
+		 * */
+		state_stats = get_cpuidle_state_stats(pr->id, count);
 
 		if (!cx->valid)
 			continue;
@@ -1015,7 +1027,7 @@ static int acpi_processor_setup_cpuidle(struct acpi_processor *pr)
 		    !(acpi_gbl_FADT.flags & ACPI_FADT_C2_MP_SUPPORTED))
 			continue;
 #endif
-		cpuidle_set_statedata(state, cx);
+		cpuidle_set_statedata(state_stats, cx);
 
 		snprintf(state->name, CPUIDLE_NAME_LEN, "C%d", i);
 		strncpy(state->desc, cx->desc, CPUIDLE_DESC_LEN);
@@ -1127,10 +1139,15 @@ int __cpuinit acpi_processor_power_init(struct acpi_processor *pr,
 	 * Note that we use previously set idle handler will be used on
 	 * platforms that only support C1.
 	 */
-	if (pr->flags.power) {
+	if (pr->flags.power)
 		acpi_processor_setup_cpuidle(pr);
-		if (cpuidle_register_device(&pr->power.dev))
-			return -EIO;
+	/*
+	 * Driver registration has to be done by last cpu.
+	 * To Do: Remove hard coded cpu number.
+	 */
+	if (pr->flags.power && pr->id == (num_online_cpus() - 1)) {
+		acpi_idle_driver.dev = &pr->power.dev;
+		cpuidle_register_driver(&acpi_idle_driver);
 	}
 	return 0;
 }
diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
index 5335598..a232845 100644
--- a/drivers/cpuidle/cpuidle.c
+++ b/drivers/cpuidle/cpuidle.c
@@ -21,10 +21,10 @@
 
 #include "cpuidle.h"
 
-DEFINE_PER_CPU(struct cpuidle_device *, cpuidle_devices);
+DEFINE_PER_CPU(struct cpuidle_device_stats, cpuidle_dev_stats);
+EXPORT_PER_CPU_SYMBOL(cpuidle_dev_stats);
 
 DEFINE_MUTEX(cpuidle_lock);
-LIST_HEAD(cpuidle_detected_devices);
 static void (*pm_idle_old)(void);
 
 static int enabled_devices;
@@ -41,6 +41,8 @@ static void cpuidle_kick_cpus(void) {}
 #endif
 
 static int __cpuidle_register_device(struct cpuidle_device *dev);
+static int __cpuidle_register_device_stats(
+		struct cpuidle_device_stats *dev_stats);
 
 /**
  * cpuidle_idle_call - the main idle loop
@@ -49,12 +51,16 @@ static int __cpuidle_register_device(struct cpuidle_device *dev);
  */
 void cpuidle_idle_call(void)
 {
-	struct cpuidle_device *dev = __this_cpu_read(cpuidle_devices);
+	struct cpuidle_device_stats *dev_stats = &per_cpu(cpuidle_dev_stats,
+							smp_processor_id());
+	struct cpuidle_device *dev;
 	struct cpuidle_state *target_state;
+	struct cpuidle_state_stats *target_state_stats;
+	struct cpuidle_driver *drv = cpuidle_get_driver();
 	int next_state;
 
 	/* check if the device is ready */
-	if (!dev || !dev->enabled) {
+	if (!drv || !drv->dev || !drv->dev->enabled) {
 		if (pm_idle_old)
 			pm_idle_old();
 		else
@@ -65,6 +71,7 @@ void cpuidle_idle_call(void)
 #endif
 		return;
 	}
+	dev = drv->dev;
 
 #if 0
 	/* shows regressions, re-enable for 2.6.29 */
@@ -83,37 +90,40 @@ void cpuidle_idle_call(void)
 	 * state availability, latencies, residencies, etc.
 	 */
 	if (dev->prepare)
-		dev->prepare(dev);
+		dev->prepare(dev_stats);
 
 	/* ask the governor for the next state */
-	next_state = cpuidle_curr_governor->select(dev);
+	next_state = cpuidle_curr_governor->select(dev, dev_stats);
 	if (need_resched()) {
 		local_irq_enable();
 		return;
 	}
 
 	target_state = &dev->states[next_state];
+	target_state_stats = &dev_stats->state_stats[next_state];
 
 	/* enter the state and update stats */
-	dev->last_state = target_state;
+	dev_stats->last_state = target_state;
 
-	trace_power_start(POWER_CSTATE, next_state, dev->cpu);
-	trace_cpu_idle(next_state, dev->cpu);
+	trace_power_start(POWER_CSTATE, next_state, dev_stats->cpu);
+	trace_cpu_idle(next_state, dev_stats->cpu);
 
-	dev->last_residency = target_state->enter(dev, target_state);
+	dev_stats->last_residency = target_state->enter(dev, target_state,
+							target_state_stats);
 
-	trace_power_end(dev->cpu);
-	trace_cpu_idle(PWR_EVENT_EXIT, dev->cpu);
+	trace_power_end(dev_stats->cpu);
+	trace_cpu_idle(PWR_EVENT_EXIT, dev_stats->cpu);
 
-	if (dev->last_state)
-		target_state = dev->last_state;
+	if (dev_stats->last_state)
+		target_state = dev_stats->last_state;
 
-	target_state->time += (unsigned long long)dev->last_residency;
-	target_state->usage++;
+	target_state_stats->time +=
+			(unsigned long long)dev_stats->last_residency;
+	target_state_stats->usage++;
 
 	/* give the governor an opportunity to reflect on the outcome */
 	if (cpuidle_curr_governor->reflect)
-		cpuidle_curr_governor->reflect(dev);
+		cpuidle_curr_governor->reflect(dev_stats);
 }
 
 #ifdef CONFIG_ARCH_USES_PMIDLE
@@ -167,7 +177,8 @@ void cpuidle_resume_and_unlock(void)
 EXPORT_SYMBOL_GPL(cpuidle_resume_and_unlock);
 
 #ifdef CONFIG_ARCH_HAS_CPU_RELAX
-static int poll_idle(struct cpuidle_device *dev, struct cpuidle_state *st)
+static int poll_idle(struct cpuidle_device *dev, struct cpuidle_state *st,
+			struct cpuidle_state_stats *st_stats)
 {
 	ktime_t	t1, t2;
 	s64 diff;
@@ -191,8 +202,6 @@ static void poll_idle_init(struct cpuidle_device *dev)
 {
 	struct cpuidle_state *state = &dev->states[0];
 
-	cpuidle_set_statedata(state, NULL);
-
 	snprintf(state->name, CPUIDLE_NAME_LEN, "POLL");
 	snprintf(state->desc, CPUIDLE_DESC_LEN, "CPUIDLE CORE POLL IDLE");
 	state->exit_latency = 0;
@@ -205,6 +214,28 @@ static void poll_idle_init(struct cpuidle_device *dev)
 static void poll_idle_init(struct cpuidle_device *dev) {}
 #endif /* CONFIG_ARCH_HAS_CPU_RELAX */
 
+static int cpuidle_enable_device_stats(struct cpuidle_device_stats *dev_stats)
+{
+	int ret, i;
+
+	if ((ret = cpuidle_add_state_sysfs(dev_stats)))
+		return ret;
+
+	if (cpuidle_curr_governor->enable &&
+		(ret = cpuidle_curr_governor->enable(dev_stats)))
+			return ret;
+
+	for (i = 0; i < dev_stats->drv->dev->state_count; i++) {
+		dev_stats->state_stats[i].usage = 0;
+		dev_stats->state_stats[i].time = 0;
+	}
+	dev_stats->last_residency = 0;
+	dev_stats->last_state = NULL;
+
+	smp_wmb();
+	return 0;
+}
+
 /**
  * cpuidle_enable_device - enables idle PM for a CPU
  * @dev: the CPU
@@ -231,35 +262,29 @@ int cpuidle_enable_device(struct cpuidle_device *dev)
 
 	poll_idle_init(dev);
 
-	if ((ret = cpuidle_add_state_sysfs(dev)))
-		return ret;
-
-	if (cpuidle_curr_governor->enable &&
-	    (ret = cpuidle_curr_governor->enable(dev)))
-		goto fail_sysfs;
-
-	for (i = 0; i < dev->state_count; i++) {
-		dev->states[i].usage = 0;
-		dev->states[i].time = 0;
+	for_each_online_cpu(i) {
+		cpuidle_enable_device_stats(&per_cpu(cpuidle_dev_stats, i));
 	}
-	dev->last_residency = 0;
-	dev->last_state = NULL;
 
 	smp_wmb();
 
 	dev->enabled = 1;
 
-	enabled_devices++;
 	return 0;
 
-fail_sysfs:
-	cpuidle_remove_state_sysfs(dev);
-
-	return ret;
 }
 
 EXPORT_SYMBOL_GPL(cpuidle_enable_device);
 
+static void cpuidle_disable_device_stats(struct cpuidle_device_stats *dev_stats)
+{
+	cpuidle_remove_state_sysfs(dev_stats);
+
+	if (cpuidle_curr_governor->disable)
+		cpuidle_curr_governor->disable(dev_stats);
+
+}
+
 /**
  * cpuidle_disable_device - disables idle PM for a CPU
  * @dev: the CPU
@@ -269,6 +294,8 @@ EXPORT_SYMBOL_GPL(cpuidle_enable_device);
  */
 void cpuidle_disable_device(struct cpuidle_device *dev)
 {
+	int i;
+
 	if (!dev->enabled)
 		return;
 	if (!cpuidle_get_driver() || !cpuidle_curr_governor)
@@ -276,15 +303,31 @@ void cpuidle_disable_device(struct cpuidle_device *dev)
 
 	dev->enabled = 0;
 
-	if (cpuidle_curr_governor->disable)
-		cpuidle_curr_governor->disable(dev);
+	for_each_online_cpu(i) {
+		cpuidle_disable_device_stats(&per_cpu(cpuidle_dev_stats, i));
+	}
 
-	cpuidle_remove_state_sysfs(dev);
-	enabled_devices--;
 }
 
 EXPORT_SYMBOL_GPL(cpuidle_disable_device);
 
+static int cpuidle_register_device_stats(struct cpuidle_device_stats *dev_stats)
+{
+	int ret;
+	struct sys_device *sys_dev =
+			get_cpu_sysdev((unsigned long)dev_stats->cpu);
+
+	if (!sys_dev)
+		return -EINVAL;
+
+	init_completion(&dev_stats->kobj_unregister);
+
+	if ((ret = cpuidle_add_sysfs(sys_dev)))
+		return ret;
+
+	return 0;
+}
+
 /**
  * __cpuidle_register_device - internal register function called before register
  * and enable routines
@@ -294,14 +337,10 @@ EXPORT_SYMBOL_GPL(cpuidle_disable_device);
  */
 static int __cpuidle_register_device(struct cpuidle_device *dev)
 {
-	int ret;
-	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
 	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
 
 	if (dev->registered)
 		return 0;
-	if (!sys_dev)
-		return -EINVAL;
 	/*
 	 * This will maintain compatibility with old cpuidle_device
 	 * structure where driver pointer is not set.
@@ -315,8 +354,6 @@ static int __cpuidle_register_device(struct cpuidle_device *dev)
 	if (!try_module_get(dev->drv->owner))
 		return -EINVAL;
 
-	init_completion(&dev->kobj_unregister);
-
 	/*
 	 * cpuidle driver should set the dev->power_specified bit
 	 * before registering the device if the driver provides
@@ -337,14 +374,6 @@ static int __cpuidle_register_device(struct cpuidle_device *dev)
 			dev->states[i].power_usage = -1 - i;
 	}
 
-	if (cpuidle_driver == dev->drv)
-		per_cpu(cpuidle_devices, dev->cpu) = dev;
-	list_add(&dev->device_list, &cpuidle_detected_devices);
-	if ((ret = cpuidle_add_sysfs(sys_dev))) {
-		module_put(dev->drv->owner);
-		return ret;
-	}
-
 	dev->registered = 1;
 	return 0;
 }
@@ -355,7 +384,8 @@ static int __cpuidle_register_device(struct cpuidle_device *dev)
  */
 int cpuidle_register_device(struct cpuidle_device *dev)
 {
-	int ret;
+	int ret, i;
+	struct cpuidle_device_stats *dev_stats;
 
 	mutex_lock(&cpuidle_lock);
 
@@ -364,6 +394,13 @@ int cpuidle_register_device(struct cpuidle_device *dev)
 		return ret;
 	}
 
+	for_each_online_cpu(i) {
+		dev_stats = &per_cpu(cpuidle_dev_stats, i);
+		dev_stats->cpu = i;
+		dev_stats->drv = cpuidle_get_driver();
+		cpuidle_register_device_stats(dev_stats);
+	}
+
 	cpuidle_enable_device(dev);
 	cpuidle_install_idle_handler();
 
@@ -375,14 +412,23 @@ int cpuidle_register_device(struct cpuidle_device *dev)
 
 EXPORT_SYMBOL_GPL(cpuidle_register_device);
 
+static void cpuidle_unregister_device_stats(
+				struct cpuidle_device_stats *dev_stats)
+{
+	struct sys_device *sys_dev =
+			get_cpu_sysdev((unsigned long)dev_stats->cpu);
+	cpuidle_remove_sysfs(sys_dev);
+	wait_for_completion(&dev_stats->kobj_unregister);
+}
+
+
 /**
  * cpuidle_unregister_device - unregisters a CPU's idle PM feature
  * @dev: the cpu
  */
 void cpuidle_unregister_device(struct cpuidle_device *dev)
 {
-	struct sys_device *sys_dev = get_cpu_sysdev((unsigned long)dev->cpu);
-	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
+	int i;
 
 	if (dev->registered == 0)
 		return;
@@ -390,11 +436,9 @@ void cpuidle_unregister_device(struct cpuidle_device *dev)
 	cpuidle_pause_and_lock();
 
 	cpuidle_disable_device(dev);
-
-	cpuidle_remove_sysfs(sys_dev);
-	wait_for_completion(&dev->kobj_unregister);
-	if (cpuidle_driver == dev->drv)
-		per_cpu(cpuidle_devices, dev->cpu) = NULL;
+	for_each_online_cpu(i) {
+		cpuidle_unregister_device_stats(&per_cpu(cpuidle_dev_stats, i));
+	}
 
 	cpuidle_resume_and_unlock();
 
diff --git a/drivers/cpuidle/cpuidle.h b/drivers/cpuidle/cpuidle.h
index 33e50d5..ea48dd8 100644
--- a/drivers/cpuidle/cpuidle.h
+++ b/drivers/cpuidle/cpuidle.h
@@ -24,8 +24,8 @@ extern int cpuidle_switch_governor(struct cpuidle_governor *gov);
 /* sysfs */
 extern int cpuidle_add_class_sysfs(struct sysdev_class *cls);
 extern void cpuidle_remove_class_sysfs(struct sysdev_class *cls);
-extern int cpuidle_add_state_sysfs(struct cpuidle_device *device);
-extern void cpuidle_remove_state_sysfs(struct cpuidle_device *device);
+extern int cpuidle_add_state_sysfs(struct cpuidle_device_stats *dev_stats);
+extern void cpuidle_remove_state_sysfs(struct cpuidle_device_stats *dev_stats);
 extern int cpuidle_add_sysfs(struct sys_device *sysdev);
 extern void cpuidle_remove_sysfs(struct sys_device *sysdev);
 
diff --git a/drivers/cpuidle/driver.c b/drivers/cpuidle/driver.c
index c235286..19b2000 100644
--- a/drivers/cpuidle/driver.c
+++ b/drivers/cpuidle/driver.c
@@ -38,18 +38,13 @@ static struct cpuidle_driver *select_cpuidle_driver(void)
 
 static void set_current_cpuidle_driver(struct cpuidle_driver *drv)
 {
-	struct cpuidle_device *item = NULL;
 	if (drv == cpuidle_curr_driver)
 		return;
 
 	/* Unregister the previous drivers devices */
 	/* Do we need to take cpuidle_lock ? {un}register_device takes lock*/
 	if (cpuidle_curr_driver) {
-		list_for_each_entry(item, &cpuidle_detected_devices,
-			device_list) {
-			if (item->drv == cpuidle_curr_driver)
-				cpuidle_unregister_device(item);
-		}
+		cpuidle_unregister_device(cpuidle_curr_driver->dev);
 	}
 
 	cpuidle_curr_driver = drv;
@@ -57,12 +52,7 @@ static void set_current_cpuidle_driver(struct cpuidle_driver *drv)
 	if (drv == NULL)
 		return;
 	else {
-		/* Register the new driver devices */
-		list_for_each_entry(item, &cpuidle_detected_devices,
-			device_list) {
-			if (item->drv == drv)
-				cpuidle_register_device(item);
-		}
+		cpuidle_register_device(drv->dev);
 	}
 }
 
@@ -143,8 +133,6 @@ EXPORT_SYMBOL_GPL(cpuidle_get_driver);
  */
 void cpuidle_unregister_driver(struct cpuidle_driver *drv)
 {
-	struct cpuidle_device *item = NULL;
-
 	if (arch_allow_unregister(drv))
 		return;
 
@@ -154,14 +142,6 @@ void cpuidle_unregister_driver(struct cpuidle_driver *drv)
 	list_del(&drv->driver_list);
 	set_current_cpuidle_driver(select_cpuidle_driver());
 
-	/* Delete all devices corresponding to this driver */
-	mutex_lock(&cpuidle_lock);
-	list_for_each_entry(item, &cpuidle_detected_devices, device_list) {
-		if (item->drv == drv)
-			list_del(&item->device_list);
-	}
-	mutex_unlock(&cpuidle_lock);
-
 	spin_unlock(&cpuidle_driver_lock);
 }
 
diff --git a/drivers/cpuidle/governor.c b/drivers/cpuidle/governor.c
index 724c164..c4bf5f0 100644
--- a/drivers/cpuidle/governor.c
+++ b/drivers/cpuidle/governor.c
@@ -43,7 +43,7 @@ static struct cpuidle_governor * __cpuidle_find_governor(const char *str)
  */
 int cpuidle_switch_governor(struct cpuidle_governor *gov)
 {
-	struct cpuidle_device *dev;
+	struct cpuidle_driver *cpuidle_driver = cpuidle_get_driver();
 
 	if (gov == cpuidle_curr_governor)
 		return 0;
@@ -51,8 +51,8 @@ int cpuidle_switch_governor(struct cpuidle_governor *gov)
 	cpuidle_uninstall_idle_handler();
 
 	if (cpuidle_curr_governor) {
-		list_for_each_entry(dev, &cpuidle_detected_devices, device_list)
-			cpuidle_disable_device(dev);
+		if (cpuidle_driver)
+			cpuidle_disable_device(cpuidle_driver->dev);
 		module_put(cpuidle_curr_governor->owner);
 	}
 
@@ -61,8 +61,8 @@ int cpuidle_switch_governor(struct cpuidle_governor *gov)
 	if (gov) {
 		if (!try_module_get(cpuidle_curr_governor->owner))
 			return -EINVAL;
-		list_for_each_entry(dev, &cpuidle_detected_devices, device_list)
-			cpuidle_enable_device(dev);
+		if (cpuidle_driver)
+			cpuidle_enable_device(cpuidle_driver->dev);
 		cpuidle_install_idle_handler();
 		printk(KERN_INFO "cpuidle: using governor %s\n", gov->name);
 	}
diff --git a/drivers/cpuidle/governors/Makefile b/drivers/cpuidle/governors/Makefile
index 1b51272..a2afe56 100644
--- a/drivers/cpuidle/governors/Makefile
+++ b/drivers/cpuidle/governors/Makefile
@@ -2,5 +2,5 @@
 # Makefile for cpuidle governors.
 #
 
-obj-$(CONFIG_CPU_IDLE_GOV_LADDER) += ladder.o
+obj-n += ladder.o
 obj-$(CONFIG_CPU_IDLE_GOV_MENU) += menu.o
diff --git a/drivers/cpuidle/governors/menu.c b/drivers/cpuidle/governors/menu.c
index f508690..8de7b76 100644
--- a/drivers/cpuidle/governors/menu.c
+++ b/drivers/cpuidle/governors/menu.c
@@ -182,7 +182,8 @@ static inline int performance_multiplier(void)
 
 static DEFINE_PER_CPU(struct menu_device, menu_devices);
 
-static void menu_update(struct cpuidle_device *dev);
+static void menu_update(struct cpuidle_device *dev,
+			struct cpuidle_device_stats *dev_stats);
 
 /* This implements DIV_ROUND_CLOSEST but avoids 64 bit division */
 static u64 div_round64(u64 dividend, u32 divisor)
@@ -230,7 +231,8 @@ static void detect_repeating_patterns(struct menu_device *data)
  * menu_select - selects the next idle state to enter
  * @dev: the CPU
  */
-static int menu_select(struct cpuidle_device *dev)
+static int menu_select(struct cpuidle_device *dev,
+			struct cpuidle_device_stats *dev_stats)
 {
 	struct menu_device *data = &__get_cpu_var(menu_devices);
 	int latency_req = pm_qos_request(PM_QOS_CPU_DMA_LATENCY);
@@ -239,7 +241,7 @@ static int menu_select(struct cpuidle_device *dev)
 	int multiplier;
 
 	if (data->needs_update) {
-		menu_update(dev);
+		menu_update(dev, dev_stats);
 		data->needs_update = 0;
 	}
 
@@ -312,7 +314,7 @@ static int menu_select(struct cpuidle_device *dev)
  * NOTE: it's important to be fast here because this operation will add to
  *       the overall exit latency.
  */
-static void menu_reflect(struct cpuidle_device *dev)
+static void menu_reflect(struct cpuidle_device_stats *dev_stats)
 {
 	struct menu_device *data = &__get_cpu_var(menu_devices);
 	data->needs_update = 1;
@@ -322,11 +324,12 @@ static void menu_reflect(struct cpuidle_device *dev)
  * menu_update - attempts to guess what happened after entry
  * @dev: the CPU
  */
-static void menu_update(struct cpuidle_device *dev)
+static void menu_update(struct cpuidle_device *dev,
+			struct cpuidle_device_stats *dev_stats)
 {
 	struct menu_device *data = &__get_cpu_var(menu_devices);
 	int last_idx = data->last_state_idx;
-	unsigned int last_idle_us = cpuidle_get_last_residency(dev);
+	unsigned int last_idle_us = cpuidle_get_last_residency(dev_stats);
 	struct cpuidle_state *target = &dev->states[last_idx];
 	unsigned int measured_us;
 	u64 new_factor;
@@ -383,9 +386,9 @@ static void menu_update(struct cpuidle_device *dev)
  * menu_enable_device - scans a CPU's states and does setup
  * @dev: the CPU
  */
-static int menu_enable_device(struct cpuidle_device *dev)
+static int menu_enable_device(struct cpuidle_device_stats *dev_stats)
 {
-	struct menu_device *data = &per_cpu(menu_devices, dev->cpu);
+	struct menu_device *data = &per_cpu(menu_devices, dev_stats->cpu);
 
 	memset(data, 0, sizeof(struct menu_device));
 
diff --git a/drivers/cpuidle/sysfs.c b/drivers/cpuidle/sysfs.c
index 0310ffa..0737fea 100644
--- a/drivers/cpuidle/sysfs.c
+++ b/drivers/cpuidle/sysfs.c
@@ -157,8 +157,9 @@ void cpuidle_remove_class_sysfs(struct sysdev_class *cls)
 
 struct cpuidle_attr {
 	struct attribute attr;
-	ssize_t (*show)(struct cpuidle_device *, char *);
-	ssize_t (*store)(struct cpuidle_device *, const char *, size_t count);
+	ssize_t (*show)(struct cpuidle_device_stats *, char *);
+	ssize_t (*store)(struct cpuidle_device_stats *, const char *,
+							size_t count);
 };
 
 #define define_one_ro(_name, show) \
@@ -166,12 +167,12 @@ struct cpuidle_attr {
 #define define_one_rw(_name, show, store) \
 	static struct cpuidle_attr attr_##_name = __ATTR(_name, 0644, show, store)
 
-#define kobj_to_cpuidledev(k) container_of(k, struct cpuidle_device, kobj)
+#define kobj_to_cpuidledev(k) container_of(k, struct cpuidle_device_stats, kobj)
 #define attr_to_cpuidleattr(a) container_of(a, struct cpuidle_attr, attr)
 static ssize_t cpuidle_show(struct kobject * kobj, struct attribute * attr ,char * buf)
 {
 	int ret = -EIO;
-	struct cpuidle_device *dev = kobj_to_cpuidledev(kobj);
+	struct cpuidle_device_stats *dev = kobj_to_cpuidledev(kobj);
 	struct cpuidle_attr * cattr = attr_to_cpuidleattr(attr);
 
 	if (cattr->show) {
@@ -186,7 +187,7 @@ static ssize_t cpuidle_store(struct kobject * kobj, struct attribute * attr,
 		     const char * buf, size_t count)
 {
 	int ret = -EIO;
-	struct cpuidle_device *dev = kobj_to_cpuidledev(kobj);
+	struct cpuidle_device_stats *dev = kobj_to_cpuidledev(kobj);
 	struct cpuidle_attr * cattr = attr_to_cpuidleattr(attr);
 
 	if (cattr->store) {
@@ -204,7 +205,7 @@ static const struct sysfs_ops cpuidle_sysfs_ops = {
 
 static void cpuidle_sysfs_release(struct kobject *kobj)
 {
-	struct cpuidle_device *dev = kobj_to_cpuidledev(kobj);
+	struct cpuidle_device_stats *dev = kobj_to_cpuidledev(kobj);
 
 	complete(&dev->kobj_unregister);
 }
@@ -245,8 +246,11 @@ static ssize_t show_state_##_name(struct cpuidle_state *state, char *buf) \
 
 define_show_state_function(exit_latency)
 define_show_state_function(power_usage)
-define_show_state_ull_function(usage)
-define_show_state_ull_function(time)
+/*
+ * To Do: Show usage and time as well in
+ * sysfs; which are currently removed because
+ * of cpuidle_state structure split.
+ */
 define_show_state_str_function(name)
 define_show_state_str_function(desc)
 
@@ -254,16 +258,12 @@ define_one_state_ro(name, show_state_name);
 define_one_state_ro(desc, show_state_desc);
 define_one_state_ro(latency, show_state_exit_latency);
 define_one_state_ro(power, show_state_power_usage);
-define_one_state_ro(usage, show_state_usage);
-define_one_state_ro(time, show_state_time);
 
 static struct attribute *cpuidle_state_default_attrs[] = {
 	&attr_name.attr,
 	&attr_desc.attr,
 	&attr_latency.attr,
 	&attr_power.attr,
-	&attr_usage.attr,
-	&attr_time.attr,
 	NULL
 };
 
@@ -300,7 +300,8 @@ static struct kobj_type ktype_state_cpuidle = {
 	.release = cpuidle_state_sysfs_release,
 };
 
-static void inline cpuidle_free_state_kobj(struct cpuidle_device *device, int i)
+static inline void cpuidle_free_state_kobj(struct cpuidle_device_stats *device,
+								int i)
 {
 	kobject_put(&device->kobjs[i]->kobj);
 	wait_for_completion(&device->kobjs[i]->kobj_unregister);
@@ -312,34 +313,35 @@ static void inline cpuidle_free_state_kobj(struct cpuidle_device *device, int i)
  * cpuidle_add_driver_sysfs - adds driver-specific sysfs attributes
  * @device: the target device
  */
-int cpuidle_add_state_sysfs(struct cpuidle_device *device)
+int cpuidle_add_state_sysfs(struct cpuidle_device_stats *dev_stats)
 {
 	int i, ret = -ENOMEM;
 	struct cpuidle_state_kobj *kobj;
 
 	/* state statistics */
-	for (i = 0; i < device->state_count; i++) {
+	for (i = 0; i < dev_stats->drv->dev->state_count; i++) {
 		kobj = kzalloc(sizeof(struct cpuidle_state_kobj), GFP_KERNEL);
 		if (!kobj)
 			goto error_state;
-		kobj->state = &device->states[i];
+		kobj->state = &dev_stats->drv->dev->states[i];
 		init_completion(&kobj->kobj_unregister);
 
-		ret = kobject_init_and_add(&kobj->kobj, &ktype_state_cpuidle, &device->kobj,
-					   "state%d", i);
+		ret = kobject_init_and_add(&kobj->kobj, &ktype_state_cpuidle,
+						&dev_stats->kobj,
+						"state%d", i);
 		if (ret) {
 			kfree(kobj);
 			goto error_state;
 		}
 		kobject_uevent(&kobj->kobj, KOBJ_ADD);
-		device->kobjs[i] = kobj;
+		dev_stats->kobjs[i] = kobj;
 	}
 
 	return 0;
 
 error_state:
 	for (i = i - 1; i >= 0; i--)
-		cpuidle_free_state_kobj(device, i);
+		cpuidle_free_state_kobj(dev_stats, i);
 	return ret;
 }
 
@@ -347,12 +349,12 @@ error_state:
  * cpuidle_remove_driver_sysfs - removes driver-specific sysfs attributes
  * @device: the target device
  */
-void cpuidle_remove_state_sysfs(struct cpuidle_device *device)
+void cpuidle_remove_state_sysfs(struct cpuidle_device_stats *dev_stats)
 {
 	int i;
 
-	for (i = 0; i < device->state_count; i++)
-		cpuidle_free_state_kobj(device, i);
+	for (i = 0; i < dev_stats->drv->dev->state_count; i++)
+		cpuidle_free_state_kobj(dev_stats, i);
 }
 
 /**
@@ -362,14 +364,15 @@ void cpuidle_remove_state_sysfs(struct cpuidle_device *device)
 int cpuidle_add_sysfs(struct sys_device *sysdev)
 {
 	int cpu = sysdev->id;
-	struct cpuidle_device *dev;
+	struct cpuidle_device_stats *dev_stats;
 	int error;
 
-	dev = per_cpu(cpuidle_devices, cpu);
-	error = kobject_init_and_add(&dev->kobj, &ktype_cpuidle, &sysdev->kobj,
+	dev_stats = &per_cpu(cpuidle_dev_stats, cpu);
+	error = kobject_init_and_add(&dev_stats->kobj, &ktype_cpuidle,
+					&sysdev->kobj,
 				     "cpuidle");
 	if (!error)
-		kobject_uevent(&dev->kobj, KOBJ_ADD);
+		kobject_uevent(&dev_stats->kobj, KOBJ_ADD);
 	return error;
 }
 
@@ -380,8 +383,8 @@ int cpuidle_add_sysfs(struct sys_device *sysdev)
 void cpuidle_remove_sysfs(struct sys_device *sysdev)
 {
 	int cpu = sysdev->id;
-	struct cpuidle_device *dev;
+	struct cpuidle_device_stats *dev_stats;
 
-	dev = per_cpu(cpuidle_devices, cpu);
-	kobject_put(&dev->kobj);
+	dev_stats = &per_cpu(cpuidle_dev_stats, cpu);
+	kobject_put(&dev_stats->kobj);
 }
diff --git a/drivers/idle/default_driver.c b/drivers/idle/default_driver.c
index c8702b5..41ad63d 100644
--- a/drivers/idle/default_driver.c
+++ b/drivers/idle/default_driver.c
@@ -275,28 +275,32 @@ static void c1e_idle(void)
 }
 
 static int poll_idle_wrapper(struct cpuidle_device *dev,
-		struct cpuidle_state *state)
+		struct cpuidle_state *state,
+		struct cpuidle_state_stats *state_stats)
 {
 	poll_idle();
 	return 0;
 }
 
 static int mwait_idle_wrapper(struct cpuidle_device *dev,
-		struct cpuidle_state *state)
+		struct cpuidle_state *state,
+		struct cpuidle_state_stats *state_stats)
 {
 	mwait_idle();
 	return 0;
 }
 
 static int c1e_idle_wrapper(struct cpuidle_device *dev,
-		struct cpuidle_state *state)
+		struct cpuidle_state *state,
+		struct cpuidle_state_stats *state_stats)
 {
 	c1e_idle();
 	return 0;
 }
 
 int default_idle_wrapper(struct cpuidle_device *dev,
-		struct cpuidle_state *state)
+		struct cpuidle_state *state,
+		struct cpuidle_state_stats *state_stats)
 {
 	default_idle();
 	return 0;
@@ -305,7 +309,6 @@ int default_idle_wrapper(struct cpuidle_device *dev,
 static struct cpuidle_state state_poll = {
 		.name = "POLL",
 		.desc = "POLL",
-		.driver_data = (void *) 0x00,
 		.flags = CPUIDLE_FLAG_TIME_VALID,
 		.exit_latency = 1,
 		.target_residency = 1,
@@ -315,7 +318,6 @@ static struct cpuidle_state state_poll = {
 static struct cpuidle_state state_mwait = {
 		.name = "C1",
 		.desc = "MWAIT No Hints",
-		.driver_data = (void *) 0x01,
 		.flags = CPUIDLE_FLAG_TIME_VALID,
 		.exit_latency = 1,
 		.target_residency = 1,
@@ -325,7 +327,6 @@ static struct cpuidle_state state_mwait = {
 static struct cpuidle_state state_c1e = {
 		.name = "C1E",
 		.desc = "C1E",
-		.driver_data = (void *) 0x02,
 		.flags = CPUIDLE_FLAG_TIME_VALID,
 		.exit_latency = 1,
 		.target_residency = 1,
@@ -335,7 +336,6 @@ static struct cpuidle_state state_c1e = {
 struct cpuidle_state state_default_idle = {
 		.name = "DEFAULT-IDLE",
 		.desc = "Default idle routine",
-		.driver_data = (void *) 0x03,
 		.flags = CPUIDLE_FLAG_TIME_VALID,
 		.exit_latency = 1,
 		.target_residency = 1,
@@ -416,34 +416,20 @@ static struct cpuidle_driver default_idle_driver = {
 	.priority = 100,
 };
 
-static int setup_cpuidle(int cpu)
+static int __init default_idle_init(void)
 {
-	struct cpuidle_device *dev = kzalloc(sizeof(struct cpuidle_device),
-					GFP_KERNEL);
+	int retval;
 	int count = CPUIDLE_DRIVER_STATE_START;
-	dev->cpu = cpu;
+	struct cpuidle_device *dev = kzalloc(sizeof(struct cpuidle_device),
+						GFP_KERNEL);
 	dev->drv = &default_idle_driver;
-
-	BUG_ON(opt_state == NULL);
 	dev->states[count] = *opt_state;
 	count++;
-
 	dev->state_count = count;
 
-	if (cpuidle_register_device(dev))
-		return -EIO;
-	return 0;
-}
-
-static int __init default_idle_init(void)
-{
-	int retval, i;
+	default_idle_driver.dev = dev;
 	retval = cpuidle_register_driver(&default_idle_driver);
 
-	for_each_online_cpu(i) {
-		setup_cpuidle(i);
-	}
-
 	return 0;
 }
 

Re: [RFC PATCH V1 0/2] cpuidle: global registration of idle states with per-cpu statistics

[Len Brown]

On Tue, 22 Mar 2011, Trinabh Gupta wrote:

> This patch series is an early RFC to discuss the feasibility of
> avoiding registering of idle states from each cpu.
> 
> The core change is to split the cpuidle_device structure into parts
> that can be global and parts that has to remain per-cpu.  The per-cpu
> pieces are mostly generic statistics that can be independent of
> current running driver.
> 
> Motivation:
> * Simplify the cpuidle subsystem framework and have
>   registration/unregistration done by single cpu.
> 
> * Minimise the data structure that needs to be maintained for multiple
>   cpuidle drivers
> 
> * Reference: https://lkml.org/lkml/2011/2/10/37
> 
> Advantages:
> * Make the cpuidle framework simple for most use cases where C-States
>   are symmetric.  In case there are asymmetric C-States detected,
>   fallback mechanism should be incorporated to maintain the system
>   functional
>   https://lkml.org/lkml/2011/2/10/257
>   https://lkml.org/lkml/2011/2/10/37
> 
> * Non x86 archs that does not have asymmetric C-States like POWER, may
>   not need the fallback mechanism and hence the framework will be
>   simple for most use cases.
> 
> Disadvantages:
> * Asymmetric C-States are part of x86 ACPI specification.  Incorrect
>   handling may functionally affect the system

I think this is a non-issue.

> * Incorporating per-cpu masks for each state to allow/dis-allow global
>   states on subset of CPUs may result in an implementation that is
>   not better than current solution of having per-cpu states.

I don't think this is needed.

> This patch series applies on top of the pm_idle cleanup patch
> https://lkml.org/lkml/2011/3/22/150 (cpuidle: Cleanup pm_idle and 
> include driver/cpuidle.c in-kernel)
> 
> This patch series is tested on x86 Nehalem system with multiple ACPI
> C-States.
> 
> This patch series has limitations of not handling multiple driver
> registration and switching between drivers on all CPUs mainly due to
> incomplete handling of per-cpu enable/disable and driver_data.

I think this series is more important than the feature of multiple
driver/system support, and thus should come first.

thanks,
Len Brown, Intel Open Source Technology Center

Re: [RFC PATCH V1 0/2] cpuidle: global registration of idle states with per-cpu statistics

[Vaidyanathan Srinivasan]

* Len Brown <lenb@kernel.org> [2011-03-25 03:28:59]:

> On Tue, 22 Mar 2011, Trinabh Gupta wrote:
> 
> > This patch series is an early RFC to discuss the feasibility of
> > avoiding registering of idle states from each cpu.
> > 
> > The core change is to split the cpuidle_device structure into parts
> > that can be global and parts that has to remain per-cpu.  The per-cpu
> > pieces are mostly generic statistics that can be independent of
> > current running driver.
> > 
> > Motivation:
> > * Simplify the cpuidle subsystem framework and have
> >   registration/unregistration done by single cpu.
> > 
> > * Minimise the data structure that needs to be maintained for multiple
> >   cpuidle drivers
> > 
> > * Reference: https://lkml.org/lkml/2011/2/10/37
> > 
> > Advantages:
> > * Make the cpuidle framework simple for most use cases where C-States
> >   are symmetric.  In case there are asymmetric C-States detected,
> >   fallback mechanism should be incorporated to maintain the system
> >   functional
> >   https://lkml.org/lkml/2011/2/10/257
> >   https://lkml.org/lkml/2011/2/10/37
> > 
> > * Non x86 archs that does not have asymmetric C-States like POWER, may
> >   not need the fallback mechanism and hence the framework will be
> >   simple for most use cases.
> > 
> > Disadvantages:
> > * Asymmetric C-States are part of x86 ACPI specification.  Incorrect
> >   handling may functionally affect the system
> 
> I think this is a non-issue.

Hi Len,

Thanks for the confirmation.  This gives us the right direction to
proceed with the cleanup.

> > * Incorporating per-cpu masks for each state to allow/dis-allow global
> >   states on subset of CPUs may result in an implementation that is
> >   not better than current solution of having per-cpu states.
> 
> I don't think this is needed.
> 
> > This patch series applies on top of the pm_idle cleanup patch
> > https://lkml.org/lkml/2011/3/22/150 (cpuidle: Cleanup pm_idle and 
> > include driver/cpuidle.c in-kernel)
> > 
> > This patch series is tested on x86 Nehalem system with multiple ACPI
> > C-States.
> > 
> > This patch series has limitations of not handling multiple driver
> > registration and switching between drivers on all CPUs mainly due to
> > incomplete handling of per-cpu enable/disable and driver_data.
> 
> I think this series is more important than the feature of multiple
> driver/system support, and thus should come first.

Removing pm_idle() gets us to the situation where we have to support
multiple registration to allow default_idle driver to register and do
mwait.  Your cleanup patch to x86 idle will make the transition much
simple.  Lets us have removal of pm_idle() and removal of per-cpu
registration in the same patch series as the next step.

--Vaidy