diff mbox series

[1/4] PM/Domains: Add support for retrieving genpd performance states information

Message ID 1565398727-23090-2-git-send-email-thara.gopinath@linaro.org
State New
Headers show
Series qcom: Model RPMH power domains as thermal cooling devices | expand

Commit Message

Thara Gopinath Aug. 10, 2019, 12:58 a.m. UTC
Add two new APIs in the genpd framework,
dev_pm_genpd_get_performance_state to return the current performance
state of a power domain and dev_pm_genpd_performance_state_count to
return the total number of performance states supported by a
power domain. Since the genpd framework does not maintain
a count of number of performance states supported by a power domain,
introduce a new callback(.get_performance_state_count) that can be used
to retrieve this information from power domain drivers.

Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org>

---
 drivers/base/power/domain.c | 38 ++++++++++++++++++++++++++++++++++++++
 include/linux/pm_domain.h   | 18 ++++++++++++++++++
 2 files changed, 56 insertions(+)

-- 
2.1.4

Comments

Ulf Hansson Aug. 22, 2019, 3:03 p.m. UTC | #1
On Sat, 10 Aug 2019 at 02:58, Thara Gopinath <thara.gopinath@linaro.org> wrote:
>

> Add two new APIs in the genpd framework,

> dev_pm_genpd_get_performance_state to return the current performance

> state of a power domain and dev_pm_genpd_performance_state_count to

> return the total number of performance states supported by a

> power domain. Since the genpd framework does not maintain

> a count of number of performance states supported by a power domain,

> introduce a new callback(.get_performance_state_count) that can be used

> to retrieve this information from power domain drivers.


I think some brief background to *why* this is useful needs to be
squeezed into the changelog. Or at least state that following changes
makes use of it, somehow.

>

> Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org>

> ---

>  drivers/base/power/domain.c | 38 ++++++++++++++++++++++++++++++++++++++

>  include/linux/pm_domain.h   | 18 ++++++++++++++++++

>  2 files changed, 56 insertions(+)

>

> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c

> index b063bc4..17e0375 100644

> --- a/drivers/base/power/domain.c

> +++ b/drivers/base/power/domain.c

> @@ -413,6 +413,44 @@ int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state)

>  }

>  EXPORT_SYMBOL_GPL(dev_pm_genpd_set_performance_state);

>

> +int dev_pm_genpd_get_performance_state(struct device *dev,

> +                                      unsigned int *state)

> +{

> +       struct generic_pm_domain *genpd;

> +

> +       genpd = dev_to_genpd(dev);


We need to verify that the there is a genpd attached before doing this
cast. Let me post a patch in a day or so, it will give you a helper
function that covers this.

> +       if (IS_ERR(genpd))

> +               return -ENODEV;

> +

> +       genpd_lock(genpd);

> +       *state = genpd->performance_state;


Why not return the state, rather than assigning an out-parameter?

> +       genpd_unlock(genpd);

> +

> +       return 0;

> +}

> +EXPORT_SYMBOL_GPL(dev_pm_genpd_get_performance_state);

> +

> +int dev_pm_genpd_performance_state_count(struct device *dev,

> +                                        unsigned int *count)

> +{

> +       struct generic_pm_domain *genpd;

> +       int ret;

> +

> +       genpd = dev_to_genpd(dev);

> +       if (IS_ERR(genpd))

> +               return -ENODEV;

> +

> +       if (unlikely(!genpd->get_performance_state_count))

> +               return -EINVAL;

> +

> +       genpd_lock(genpd);

> +       ret = genpd->get_performance_state_count(genpd, count);


Why not having the callback to return the state, rather than using an
out-parameter?

> +       genpd_unlock(genpd);

> +

> +       return ret;

> +}

> +EXPORT_SYMBOL_GPL(dev_pm_genpd_performance_state_count);

> +

>  static int _genpd_power_on(struct generic_pm_domain *genpd, bool timed)

>  {

>         unsigned int state_idx = genpd->state_idx;

> diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h

> index 91d9bf4..0e5f502 100644

> --- a/include/linux/pm_domain.h

> +++ b/include/linux/pm_domain.h

> @@ -117,6 +117,8 @@ struct generic_pm_domain {

>                                                  struct dev_pm_opp *opp);

>         int (*set_performance_state)(struct generic_pm_domain *genpd,

>                                      unsigned int state);

> +       int (*get_performance_state_count)(struct generic_pm_domain *genpd,

> +                                          unsigned int *count);

>         struct gpd_dev_ops dev_ops;

>         s64 max_off_time_ns;    /* Maximum allowed "suspended" time. */

>         bool max_off_time_changed;

> @@ -204,6 +206,10 @@ int pm_genpd_init(struct generic_pm_domain *genpd,

>                   struct dev_power_governor *gov, bool is_off);

>  int pm_genpd_remove(struct generic_pm_domain *genpd);

>  int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state);

> +int dev_pm_genpd_get_performance_state(struct device *dev,

> +                                      unsigned int *state);

> +int dev_pm_genpd_performance_state_count(struct device *dev,

> +                                        unsigned int *count);

>

>  extern struct dev_power_governor simple_qos_governor;

>  extern struct dev_power_governor pm_domain_always_on_gov;

> @@ -251,6 +257,18 @@ static inline int dev_pm_genpd_set_performance_state(struct device *dev,

>         return -ENOTSUPP;

>  }

>

> +static inline int dev_pm_genpd_get_performance_state(struct device *dev,

> +                                                    unsigned int *state);

> +{

> +       return -ENOTSUPP;

> +}

> +

> +static inline int dev_pm_genpd_performance_state_count(struct device *dev,

> +                                                      unsigned int *count);

> +{

> +       return -ENOTSUPP;

> +}

> +

>  #define simple_qos_governor            (*(struct dev_power_governor *)(NULL))

>  #define pm_domain_always_on_gov                (*(struct dev_power_governor *)(NULL))

>  #endif

> --

> 2.1.4

>


Kind regards
Uffe
Thara Gopinath Aug. 23, 2019, 5:39 p.m. UTC | #2
Hi Ulf,

Thanks for the review.

On 08/22/2019 11:03 AM, Ulf Hansson wrote:
> On Sat, 10 Aug 2019 at 02:58, Thara Gopinath <thara.gopinath@linaro.org> wrote:

>>

>> Add two new APIs in the genpd framework,

>> dev_pm_genpd_get_performance_state to return the current performance

>> state of a power domain and dev_pm_genpd_performance_state_count to

>> return the total number of performance states supported by a

>> power domain. Since the genpd framework does not maintain

>> a count of number of performance states supported by a power domain,

>> introduce a new callback(.get_performance_state_count) that can be used

>> to retrieve this information from power domain drivers.

> 

> I think some brief background to *why* this is useful needs to be

> squeezed into the changelog. Or at least state that following changes

> makes use of it, somehow.


Will do.
> 

>>

>> Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org>

>> ---

>>  drivers/base/power/domain.c | 38 ++++++++++++++++++++++++++++++++++++++

>>  include/linux/pm_domain.h   | 18 ++++++++++++++++++

>>  2 files changed, 56 insertions(+)

>>

>> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c

>> index b063bc4..17e0375 100644

>> --- a/drivers/base/power/domain.c

>> +++ b/drivers/base/power/domain.c

>> @@ -413,6 +413,44 @@ int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state)

>>  }

>>  EXPORT_SYMBOL_GPL(dev_pm_genpd_set_performance_state);

>>

>> +int dev_pm_genpd_get_performance_state(struct device *dev,

>> +                                      unsigned int *state)

>> +{

>> +       struct generic_pm_domain *genpd;

>> +

>> +       genpd = dev_to_genpd(dev);

> 

> We need to verify that the there is a genpd attached before doing this

> cast. Let me post a patch in a day or so, it will give you a helper

> function that covers this.


Sounds good.. Thanks.. I will wait for it.

> 

>> +       if (IS_ERR(genpd))

>> +               return -ENODEV;

>> +

>> +       genpd_lock(genpd);

>> +       *state = genpd->performance_state;

> 

> Why not return the state, rather than assigning an out-parameter?

We can. I will change it for this and dev_pm_genpd_performance_state_count.

Regards
Thara

> 

>> +       genpd_unlock(genpd);

>> +

>> +       return 0;

>> +}

>> +EXPORT_SYMBOL_GPL(dev_pm_genpd_get_performance_state);

>> +

>> +int dev_pm_genpd_performance_state_count(struct device *dev,

>> +                                        unsigned int *count)

>> +{

>> +       struct generic_pm_domain *genpd;

>> +       int ret;

>> +

>> +       genpd = dev_to_genpd(dev);

>> +       if (IS_ERR(genpd))

>> +               return -ENODEV;

>> +

>> +       if (unlikely(!genpd->get_performance_state_count))

>> +               return -EINVAL;

>> +

>> +       genpd_lock(genpd);

>> +       ret = genpd->get_performance_state_count(genpd, count);

> 

> Why not having the callback to return the state, rather than using an

> out-parameter?

> 

>> +       genpd_unlock(genpd);

>> +

>> +       return ret;

>> +}

>> +EXPORT_SYMBOL_GPL(dev_pm_genpd_performance_state_count);

>> +

>>  static int _genpd_power_on(struct generic_pm_domain *genpd, bool timed)

>>  {

>>         unsigned int state_idx = genpd->state_idx;

>> diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h

>> index 91d9bf4..0e5f502 100644

>> --- a/include/linux/pm_domain.h

>> +++ b/include/linux/pm_domain.h

>> @@ -117,6 +117,8 @@ struct generic_pm_domain {

>>                                                  struct dev_pm_opp *opp);

>>         int (*set_performance_state)(struct generic_pm_domain *genpd,

>>                                      unsigned int state);

>> +       int (*get_performance_state_count)(struct generic_pm_domain *genpd,

>> +                                          unsigned int *count);

>>         struct gpd_dev_ops dev_ops;

>>         s64 max_off_time_ns;    /* Maximum allowed "suspended" time. */

>>         bool max_off_time_changed;

>> @@ -204,6 +206,10 @@ int pm_genpd_init(struct generic_pm_domain *genpd,

>>                   struct dev_power_governor *gov, bool is_off);

>>  int pm_genpd_remove(struct generic_pm_domain *genpd);

>>  int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state);

>> +int dev_pm_genpd_get_performance_state(struct device *dev,

>> +                                      unsigned int *state);

>> +int dev_pm_genpd_performance_state_count(struct device *dev,

>> +                                        unsigned int *count);

>>

>>  extern struct dev_power_governor simple_qos_governor;

>>  extern struct dev_power_governor pm_domain_always_on_gov;

>> @@ -251,6 +257,18 @@ static inline int dev_pm_genpd_set_performance_state(struct device *dev,

>>         return -ENOTSUPP;

>>  }

>>

>> +static inline int dev_pm_genpd_get_performance_state(struct device *dev,

>> +                                                    unsigned int *state);

>> +{

>> +       return -ENOTSUPP;

>> +}

>> +

>> +static inline int dev_pm_genpd_performance_state_count(struct device *dev,

>> +                                                      unsigned int *count);

>> +{

>> +       return -ENOTSUPP;

>> +}

>> +

>>  #define simple_qos_governor            (*(struct dev_power_governor *)(NULL))

>>  #define pm_domain_always_on_gov                (*(struct dev_power_governor *)(NULL))

>>  #endif

>> --

>> 2.1.4

>>

> 

> Kind regards

> Uffe

> 



-- 
Regards
Thara
Thara Gopinath Sept. 6, 2019, 10:24 p.m. UTC | #3
On 08/22/2019 11:03 AM, Ulf Hansson wrote:
> On Sat, 10 Aug 2019 at 02:58, Thara Gopinath <thara.gopinath@linaro.org> wrote:

>>

>> Add two new APIs in the genpd framework,

>> dev_pm_genpd_get_performance_state to return the current performance

>> state of a power domain and dev_pm_genpd_performance_state_count to

>> return the total number of performance states supported by a

>> power domain. Since the genpd framework does not maintain

>> a count of number of performance states supported by a power domain,

>> introduce a new callback(.get_performance_state_count) that can be used

>> to retrieve this information from power domain drivers.

> 

> I think some brief background to *why* this is useful needs to be

> squeezed into the changelog. Or at least state that following changes

> makes use of it, somehow.

> 

>>

>> Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org>

>> ---

>>  drivers/base/power/domain.c | 38 ++++++++++++++++++++++++++++++++++++++

>>  include/linux/pm_domain.h   | 18 ++++++++++++++++++

>>  2 files changed, 56 insertions(+)

>>

>> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c

>> index b063bc4..17e0375 100644

>> --- a/drivers/base/power/domain.c

>> +++ b/drivers/base/power/domain.c

>> @@ -413,6 +413,44 @@ int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state)

>>  }

>>  EXPORT_SYMBOL_GPL(dev_pm_genpd_set_performance_state);

>>

>> +int dev_pm_genpd_get_performance_state(struct device *dev,

>> +                                      unsigned int *state)

>> +{

>> +       struct generic_pm_domain *genpd;

>> +

>> +       genpd = dev_to_genpd(dev);

> 

> We need to verify that the there is a genpd attached before doing this

> cast. Let me post a patch in a day or so, it will give you a helper

> function that covers this.

> 

>> +       if (IS_ERR(genpd))

>> +               return -ENODEV;

>> +

>> +       genpd_lock(genpd);

>> +       *state = genpd->performance_state;

> 

> Why not return the state, rather than assigning an out-parameter?

> 

>> +       genpd_unlock(genpd);

>> +

>> +       return 0;

>> +}

>> +EXPORT_SYMBOL_GPL(dev_pm_genpd_get_performance_state);

>> +

>> +int dev_pm_genpd_performance_state_count(struct device *dev,

>> +                                        unsigned int *count)

>> +{

>> +       struct generic_pm_domain *genpd;

>> +       int ret;

>> +

>> +       genpd = dev_to_genpd(dev);

>> +       if (IS_ERR(genpd))

>> +               return -ENODEV;

>> +

>> +       if (unlikely(!genpd->get_performance_state_count))

>> +               return -EINVAL;

>> +

>> +       genpd_lock(genpd);

>> +       ret = genpd->get_performance_state_count(genpd, count);

> 

> Why not having the callback to return the state, rather than using an

> out-parameter?

Hi Ulf,
I just realized that returning the state instead of using a parameter
will prevent me from access under lock. Is that okay ?

Regards
Thara
> 

>> +       genpd_unlock(genpd);

>> +

>> +       return ret;

>> +}

>> +EXPORT_SYMBOL_GPL(dev_pm_genpd_performance_state_count);

>> +

>>  static int _genpd_power_on(struct generic_pm_domain *genpd, bool timed)

>>  {

>>         unsigned int state_idx = genpd->state_idx;

>> diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h

>> index 91d9bf4..0e5f502 100644

>> --- a/include/linux/pm_domain.h

>> +++ b/include/linux/pm_domain.h

>> @@ -117,6 +117,8 @@ struct generic_pm_domain {

>>                                                  struct dev_pm_opp *opp);

>>         int (*set_performance_state)(struct generic_pm_domain *genpd,

>>                                      unsigned int state);

>> +       int (*get_performance_state_count)(struct generic_pm_domain *genpd,

>> +                                          unsigned int *count);

>>         struct gpd_dev_ops dev_ops;

>>         s64 max_off_time_ns;    /* Maximum allowed "suspended" time. */

>>         bool max_off_time_changed;

>> @@ -204,6 +206,10 @@ int pm_genpd_init(struct generic_pm_domain *genpd,

>>                   struct dev_power_governor *gov, bool is_off);

>>  int pm_genpd_remove(struct generic_pm_domain *genpd);

>>  int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state);

>> +int dev_pm_genpd_get_performance_state(struct device *dev,

>> +                                      unsigned int *state);

>> +int dev_pm_genpd_performance_state_count(struct device *dev,

>> +                                        unsigned int *count);

>>

>>  extern struct dev_power_governor simple_qos_governor;

>>  extern struct dev_power_governor pm_domain_always_on_gov;

>> @@ -251,6 +257,18 @@ static inline int dev_pm_genpd_set_performance_state(struct device *dev,

>>         return -ENOTSUPP;

>>  }

>>

>> +static inline int dev_pm_genpd_get_performance_state(struct device *dev,

>> +                                                    unsigned int *state);

>> +{

>> +       return -ENOTSUPP;

>> +}

>> +

>> +static inline int dev_pm_genpd_performance_state_count(struct device *dev,

>> +                                                      unsigned int *count);

>> +{

>> +       return -ENOTSUPP;

>> +}

>> +

>>  #define simple_qos_governor            (*(struct dev_power_governor *)(NULL))

>>  #define pm_domain_always_on_gov                (*(struct dev_power_governor *)(NULL))

>>  #endif

>> --

>> 2.1.4

>>

> 

> Kind regards

> Uffe

> 



-- 
Regards
Thara
Ulf Hansson Sept. 9, 2019, 9:40 a.m. UTC | #4
On Sat, 7 Sep 2019 at 00:24, Thara Gopinath <thara.gopinath@linaro.org> wrote:
>

> On 08/22/2019 11:03 AM, Ulf Hansson wrote:

> > On Sat, 10 Aug 2019 at 02:58, Thara Gopinath <thara.gopinath@linaro.org> wrote:

> >>

> >> Add two new APIs in the genpd framework,

> >> dev_pm_genpd_get_performance_state to return the current performance

> >> state of a power domain and dev_pm_genpd_performance_state_count to

> >> return the total number of performance states supported by a

> >> power domain. Since the genpd framework does not maintain

> >> a count of number of performance states supported by a power domain,

> >> introduce a new callback(.get_performance_state_count) that can be used

> >> to retrieve this information from power domain drivers.

> >

> > I think some brief background to *why* this is useful needs to be

> > squeezed into the changelog. Or at least state that following changes

> > makes use of it, somehow.

> >

> >>

> >> Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org>

> >> ---

> >>  drivers/base/power/domain.c | 38 ++++++++++++++++++++++++++++++++++++++

> >>  include/linux/pm_domain.h   | 18 ++++++++++++++++++

> >>  2 files changed, 56 insertions(+)

> >>

> >> diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c

> >> index b063bc4..17e0375 100644

> >> --- a/drivers/base/power/domain.c

> >> +++ b/drivers/base/power/domain.c

> >> @@ -413,6 +413,44 @@ int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state)

> >>  }

> >>  EXPORT_SYMBOL_GPL(dev_pm_genpd_set_performance_state);

> >>

> >> +int dev_pm_genpd_get_performance_state(struct device *dev,

> >> +                                      unsigned int *state)

> >> +{

> >> +       struct generic_pm_domain *genpd;

> >> +

> >> +       genpd = dev_to_genpd(dev);

> >

> > We need to verify that the there is a genpd attached before doing this

> > cast. Let me post a patch in a day or so, it will give you a helper

> > function that covers this.

> >

> >> +       if (IS_ERR(genpd))

> >> +               return -ENODEV;

> >> +

> >> +       genpd_lock(genpd);

> >> +       *state = genpd->performance_state;

> >

> > Why not return the state, rather than assigning an out-parameter?

> >

> >> +       genpd_unlock(genpd);

> >> +

> >> +       return 0;

> >> +}

> >> +EXPORT_SYMBOL_GPL(dev_pm_genpd_get_performance_state);

> >> +

> >> +int dev_pm_genpd_performance_state_count(struct device *dev,

> >> +                                        unsigned int *count)

> >> +{

> >> +       struct generic_pm_domain *genpd;

> >> +       int ret;

> >> +

> >> +       genpd = dev_to_genpd(dev);

> >> +       if (IS_ERR(genpd))

> >> +               return -ENODEV;

> >> +

> >> +       if (unlikely(!genpd->get_performance_state_count))

> >> +               return -EINVAL;

> >> +

> >> +       genpd_lock(genpd);

> >> +       ret = genpd->get_performance_state_count(genpd, count);

> >

> > Why not having the callback to return the state, rather than using an

> > out-parameter?

> Hi Ulf,

> I just realized that returning the state instead of using a parameter

> will prevent me from access under lock. Is that okay ?


Not sure I understand. Why can't you just assign a local variable and
return that?

Like this:

genpd_lock();
count = genpd->get_performance_state_count();
genpd_unlock();

return count;

[...]

Kind regards
Uffe
diff mbox series

Patch

diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c
index b063bc4..17e0375 100644
--- a/drivers/base/power/domain.c
+++ b/drivers/base/power/domain.c
@@ -413,6 +413,44 @@  int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state)
 }
 EXPORT_SYMBOL_GPL(dev_pm_genpd_set_performance_state);
 
+int dev_pm_genpd_get_performance_state(struct device *dev,
+				       unsigned int *state)
+{
+	struct generic_pm_domain *genpd;
+
+	genpd = dev_to_genpd(dev);
+	if (IS_ERR(genpd))
+		return -ENODEV;
+
+	genpd_lock(genpd);
+	*state = genpd->performance_state;
+	genpd_unlock(genpd);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dev_pm_genpd_get_performance_state);
+
+int dev_pm_genpd_performance_state_count(struct device *dev,
+					 unsigned int *count)
+{
+	struct generic_pm_domain *genpd;
+	int ret;
+
+	genpd = dev_to_genpd(dev);
+	if (IS_ERR(genpd))
+		return -ENODEV;
+
+	if (unlikely(!genpd->get_performance_state_count))
+		return -EINVAL;
+
+	genpd_lock(genpd);
+	ret = genpd->get_performance_state_count(genpd, count);
+	genpd_unlock(genpd);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(dev_pm_genpd_performance_state_count);
+
 static int _genpd_power_on(struct generic_pm_domain *genpd, bool timed)
 {
 	unsigned int state_idx = genpd->state_idx;
diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h
index 91d9bf4..0e5f502 100644
--- a/include/linux/pm_domain.h
+++ b/include/linux/pm_domain.h
@@ -117,6 +117,8 @@  struct generic_pm_domain {
 						 struct dev_pm_opp *opp);
 	int (*set_performance_state)(struct generic_pm_domain *genpd,
 				     unsigned int state);
+	int (*get_performance_state_count)(struct generic_pm_domain *genpd,
+					   unsigned int *count);
 	struct gpd_dev_ops dev_ops;
 	s64 max_off_time_ns;	/* Maximum allowed "suspended" time. */
 	bool max_off_time_changed;
@@ -204,6 +206,10 @@  int pm_genpd_init(struct generic_pm_domain *genpd,
 		  struct dev_power_governor *gov, bool is_off);
 int pm_genpd_remove(struct generic_pm_domain *genpd);
 int dev_pm_genpd_set_performance_state(struct device *dev, unsigned int state);
+int dev_pm_genpd_get_performance_state(struct device *dev,
+				       unsigned int *state);
+int dev_pm_genpd_performance_state_count(struct device *dev,
+					 unsigned int *count);
 
 extern struct dev_power_governor simple_qos_governor;
 extern struct dev_power_governor pm_domain_always_on_gov;
@@ -251,6 +257,18 @@  static inline int dev_pm_genpd_set_performance_state(struct device *dev,
 	return -ENOTSUPP;
 }
 
+static inline int dev_pm_genpd_get_performance_state(struct device *dev,
+						     unsigned int *state);
+{
+	return -ENOTSUPP;
+}
+
+static inline int dev_pm_genpd_performance_state_count(struct device *dev,
+						       unsigned int *count);
+{
+	return -ENOTSUPP;
+}
+
 #define simple_qos_governor		(*(struct dev_power_governor *)(NULL))
 #define pm_domain_always_on_gov		(*(struct dev_power_governor *)(NULL))
 #endif