Jean Pihet | 7 Feb 09:34

[PATCH 0/2] PM / QoS: unconditionally build the feature

The PM QoS features depend on CONFIG_PM which depends on
PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
not having these CONFIGs.

Those patches allow the features in all cases.

Based on 3.3-rc1 from mainline.
Tested on OMAP3 Beagleboard, with CONFIG_PM.
Compile tested using the i386 defconfig, with and without CONFIG_PM.

Jean Pihet (2):
  PM / QoS: unconditionally build the feature
  PM / QoS: unconditionally build the per-device constraints feature

 drivers/base/power/Makefile |    3 +-
 include/linux/pm_qos.h      |   66 -------------------------------------------
 kernel/power/Makefile       |    3 +-
 3 files changed, 4 insertions(+), 68 deletions(-)

--

-- 
1.7.5.4

Jean Pihet | 7 Feb 09:34

[PATCH 1/2] PM / QoS: unconditionally build the feature

The PM QoS feature depends on CONFIG_PM which depends on
PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
not having these CONFIGs.

This patch allows the feature in all cases.

Reported-by: Venkatesh Pallipadi <venki <at> google.com>
Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
Cc: Mark Gross <markgross <at> thegnar.org>
---
 include/linux/pm_qos.h |   29 +----------------------------
 kernel/power/Makefile  |    3 ++-
 2 files changed, 3 insertions(+), 29 deletions(-)

diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h
index e5bbcba..105be69 100644
--- a/include/linux/pm_qos.h
+++ b/include/linux/pm_qos.h
@@ -63,7 +63,6 @@ static inline int dev_pm_qos_request_active(struct dev_pm_qos_request *req)
 	return req->dev != 0;
 }

-#ifdef CONFIG_PM
 int pm_qos_update_target(struct pm_qos_constraints *c, struct plist_node *node,
 			 enum pm_qos_req_action action, int value);
 void pm_qos_add_request(struct pm_qos_request *req, int pm_qos_class,
@@ -78,6 +77,7 @@ int pm_qos_remove_notifier(int pm_qos_class, struct notifier_block *notifier);
 int pm_qos_request_active(struct pm_qos_request *req);
 s32 pm_qos_read_value(struct pm_qos_constraints *c);
(Continue reading)

mark gross | 12 Feb 03:06

Re: [PATCH 1/2] PM / QoS: unconditionally build the feature

acked-by <markgross <at> thegnar.org>

On Tue, Feb 07, 2012 at 09:34:05AM +0100, Jean Pihet wrote:
> The PM QoS feature depends on CONFIG_PM which depends on
> PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> not having these CONFIGs.
> 
> This patch allows the feature in all cases.
> 
> Reported-by: Venkatesh Pallipadi <venki <at> google.com>
> Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> Cc: Mark Gross <markgross <at> thegnar.org>
> ---
>  include/linux/pm_qos.h |   29 +----------------------------
>  kernel/power/Makefile  |    3 ++-
>  2 files changed, 3 insertions(+), 29 deletions(-)
> 
> diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h
> index e5bbcba..105be69 100644
> --- a/include/linux/pm_qos.h
> +++ b/include/linux/pm_qos.h
> @@ -63,7 +63,6 @@ static inline int dev_pm_qos_request_active(struct dev_pm_qos_request *req)
>  	return req->dev != 0;
>  }
>  
> -#ifdef CONFIG_PM
>  int pm_qos_update_target(struct pm_qos_constraints *c, struct plist_node *node,
>  			 enum pm_qos_req_action action, int value);
>  void pm_qos_add_request(struct pm_qos_request *req, int pm_qos_class,
(Continue reading)

Rafael J. Wysocki | 12 Feb 22:37
Picon
Gravatar

Re: [PATCH 1/2] PM / QoS: unconditionally build the feature

On Tuesday, February 07, 2012, Jean Pihet wrote:
> The PM QoS feature depends on CONFIG_PM which depends on
> PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> not having these CONFIGs.
> 
> This patch allows the feature in all cases.

Since the Venki's patch has been merged in the meantime, can you please
rebase this on top of the current Linus' tree?

Rafael

> Reported-by: Venkatesh Pallipadi <venki <at> google.com>
> Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> Cc: Mark Gross <markgross <at> thegnar.org>
> ---
>  include/linux/pm_qos.h |   29 +----------------------------
>  kernel/power/Makefile  |    3 ++-
>  2 files changed, 3 insertions(+), 29 deletions(-)
> 
> diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h
> index e5bbcba..105be69 100644
> --- a/include/linux/pm_qos.h
> +++ b/include/linux/pm_qos.h
> @@ -63,7 +63,6 @@ static inline int dev_pm_qos_request_active(struct dev_pm_qos_request *req)
>  	return req->dev != 0;
>  }
>  
> -#ifdef CONFIG_PM
(Continue reading)

Jean Pihet | 13 Feb 14:50

[PATCH] PM / QoS: unconditionally build the feature

The PM QoS feature depends on CONFIG_PM which depends on
PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
not having these CONFIGs.

This patch allows the feature in all cases.

Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
Reported-by: Venkatesh Pallipadi <venki <at> google.com>
Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
Cc: Mark Gross <markgross <at> thegnar.org>
---
 include/linux/pm_qos.h |   41 +----------------------------------------
 kernel/power/Makefile  |    3 ++-
 2 files changed, 3 insertions(+), 41 deletions(-)

diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h
index 4d99e4e..105be69 100644
--- a/include/linux/pm_qos.h
+++ b/include/linux/pm_qos.h
@@ -63,7 +63,6 @@ static inline int dev_pm_qos_request_active(struct dev_pm_qos_request *req)
 	return req->dev != 0;
 }

-#ifdef CONFIG_PM
 int pm_qos_update_target(struct pm_qos_constraints *c, struct plist_node *node,
 			 enum pm_qos_req_action action, int value);
 void pm_qos_add_request(struct pm_qos_request *req, int pm_qos_class,
@@ -78,6 +77,7 @@ int pm_qos_remove_notifier(int pm_qos_class, struct notifier_block *notifier);
 int pm_qos_request_active(struct pm_qos_request *req);
 s32 pm_qos_read_value(struct pm_qos_constraints *c);
(Continue reading)

Rafael J. Wysocki | 13 Feb 16:41
Picon
Gravatar

Re: [PATCH] PM / QoS: unconditionally build the feature

On Monday, February 13, 2012, Jean Pihet wrote:
> The PM QoS feature depends on CONFIG_PM which depends on
> PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> not having these CONFIGs.
> 
> This patch allows the feature in all cases.
> 
> Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> Reported-by: Venkatesh Pallipadi <venki <at> google.com>
> Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> Cc: Mark Gross <markgross <at> thegnar.org>

Applied, but I modified the changelog (please have a look at the
linux-pm/pm-qos branch for details).

Thanks,
Rafael

> ---
>  include/linux/pm_qos.h |   41 +----------------------------------------
>  kernel/power/Makefile  |    3 ++-
>  2 files changed, 3 insertions(+), 41 deletions(-)
> 
> diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h
> index 4d99e4e..105be69 100644
> --- a/include/linux/pm_qos.h
> +++ b/include/linux/pm_qos.h
> @@ -63,7 +63,6 @@ static inline int dev_pm_qos_request_active(struct dev_pm_qos_request *req)
>  	return req->dev != 0;
>  }
(Continue reading)

Jean Pihet | 13 Feb 16:40

Re: [PATCH] PM / QoS: unconditionally build the feature

Hi Rafael,

2012/2/13 Rafael J. Wysocki <rjw <at> sisk.pl>:
> On Monday, February 13, 2012, Jean Pihet wrote:
>> The PM QoS feature depends on CONFIG_PM which depends on
>> PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
>> not having these CONFIGs.
>>
>> This patch allows the feature in all cases.
>>
>> Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
>> Reported-by: Venkatesh Pallipadi <venki <at> google.com>
>> Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
>> Cc: Mark Gross <markgross <at> thegnar.org>
>
> Applied, but I modified the changelog (please have a look at the
> linux-pm/pm-qos branch for details).
Ok thanks! Will look at it as soon as the changes are propagated to linux-pm.

>
> Thanks,
> Rafael
>

Jean

>
>> ---
>>  include/linux/pm_qos.h |   41 +----------------------------------------
>>  kernel/power/Makefile  |    3 ++-
(Continue reading)

Pavel Machek | 17 Feb 20:27
Picon

Re: [PATCH] PM / QoS: unconditionally build the feature

On Mon 2012-02-13 16:41:10, Rafael J. Wysocki wrote:
> On Monday, February 13, 2012, Jean Pihet wrote:
> > The PM QoS feature depends on CONFIG_PM which depends on
> > PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> > not having these CONFIGs.
> > 
> > This patch allows the feature in all cases.
> > 
> > Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> > Reported-by: Venkatesh Pallipadi <venki <at> google.com>
> > Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> > Cc: Mark Gross <markgross <at> thegnar.org>
> 
> Applied, but I modified the changelog (please have a look at the
> linux-pm/pm-qos branch for details).

Is it good idea? For servers, power management / QoS is not
neccessary... What about fixing kconfig, instead?
									Pavel
--

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
Rafael J. Wysocki | 17 Feb 21:48
Picon
Gravatar

Re: [PATCH] PM / QoS: unconditionally build the feature

On Friday, February 17, 2012, Pavel Machek wrote:
> On Mon 2012-02-13 16:41:10, Rafael J. Wysocki wrote:
> > On Monday, February 13, 2012, Jean Pihet wrote:
> > > The PM QoS feature depends on CONFIG_PM which depends on
> > > PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> > > not having these CONFIGs.
> > > 
> > > This patch allows the feature in all cases.
> > > 
> > > Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> > > Reported-by: Venkatesh Pallipadi <venki <at> google.com>
> > > Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> > > Cc: Mark Gross <markgross <at> thegnar.org>
> > 
> > Applied, but I modified the changelog (please have a look at the
> > linux-pm/pm-qos branch for details).
> 
> Is it good idea? For servers, power management / QoS is not
> neccessary... 

Yes, it is.  They surely use CPUidle.

Thanks,
Rafael
Jean Pihet | 7 Feb 09:34

[PATCH 2/2] PM / QoS: unconditionally build the per-device constraints feature

The per-device PM QoS feature depends on CONFIG_PM which depends
on PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
not having these CONFIGs.

This patch allows the feature in all cases.

Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
Cc: Mark Gross <markgross <at> thegnar.org>
---
 drivers/base/power/Makefile |    3 ++-
 include/linux/pm_qos.h      |   39 ---------------------------------------
 2 files changed, 2 insertions(+), 40 deletions(-)

diff --git a/drivers/base/power/Makefile b/drivers/base/power/Makefile
index 2e58ebb..312eb65 100644
--- a/drivers/base/power/Makefile
+++ b/drivers/base/power/Makefile
@@ -1,4 +1,5 @@
-obj-$(CONFIG_PM)	+= sysfs.o generic_ops.o common.o qos.o
+obj-y			+= qos.o
+obj-$(CONFIG_PM)	+= sysfs.o generic_ops.o common.o
 obj-$(CONFIG_PM_SLEEP)	+= main.o wakeup.o
 obj-$(CONFIG_PM_RUNTIME)	+= runtime.o
 obj-$(CONFIG_PM_TRACE_RTC)	+= trace.o
diff --git a/include/linux/pm_qos.h b/include/linux/pm_qos.h
index 105be69..37b017a 100644
--- a/include/linux/pm_qos.h
+++ b/include/linux/pm_qos.h
@@ -77,7 +77,6 @@ int pm_qos_remove_notifier(int pm_qos_class, struct notifier_block *notifier);
(Continue reading)

mark gross | 12 Feb 03:07

Re: [PATCH 2/2] PM / QoS: unconditionally build the per-device constraints feature

why are these two patches?

I think you could squash them into one and it would be just fine.

--mark

On Tue, Feb 07, 2012 at 09:34:06AM +0100, Jean Pihet wrote:
> The per-device PM QoS feature depends on CONFIG_PM which depends
> on PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> not having these CONFIGs.
> 
> This patch allows the feature in all cases.
> 
> Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> Cc: Mark Gross <markgross <at> thegnar.org>
> ---
>  drivers/base/power/Makefile |    3 ++-
>  include/linux/pm_qos.h      |   39 ---------------------------------------
>  2 files changed, 2 insertions(+), 40 deletions(-)
> 
> diff --git a/drivers/base/power/Makefile b/drivers/base/power/Makefile
> index 2e58ebb..312eb65 100644
> --- a/drivers/base/power/Makefile
> +++ b/drivers/base/power/Makefile
> @@ -1,4 +1,5 @@
> -obj-$(CONFIG_PM)	+= sysfs.o generic_ops.o common.o qos.o
> +obj-y			+= qos.o
> +obj-$(CONFIG_PM)	+= sysfs.o generic_ops.o common.o
>  obj-$(CONFIG_PM_SLEEP)	+= main.o wakeup.o
(Continue reading)

Rafael J. Wysocki | 12 Feb 22:27
Picon
Gravatar

Re: [PATCH 2/2] PM / QoS: unconditionally build the per-device constraints feature

On Sunday, February 12, 2012, mark gross wrote:
> why are these two patches?
> 
> I think you could squash them into one and it would be just fine.

Well, what sense does _device_ PM QoS make without CONFIG_PM?

Rafael

> On Tue, Feb 07, 2012 at 09:34:06AM +0100, Jean Pihet wrote:
> > The per-device PM QoS feature depends on CONFIG_PM which depends
> > on PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> > not having these CONFIGs.
> > 
> > This patch allows the feature in all cases.
> > 
> > Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> > Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> > Cc: Mark Gross <markgross <at> thegnar.org>
> > ---
> >  drivers/base/power/Makefile |    3 ++-
> >  include/linux/pm_qos.h      |   39 ---------------------------------------
> >  2 files changed, 2 insertions(+), 40 deletions(-)
> > 
> > diff --git a/drivers/base/power/Makefile b/drivers/base/power/Makefile
> > index 2e58ebb..312eb65 100644
> > --- a/drivers/base/power/Makefile
> > +++ b/drivers/base/power/Makefile
> > @@ -1,4 +1,5 @@
> > -obj-$(CONFIG_PM)	+= sysfs.o generic_ops.o common.o qos.o
(Continue reading)

mark gross | 15 Feb 16:33

Re: [PATCH 2/2] PM / QoS: unconditionally build the per-device constraints feature

On Sun, Feb 12, 2012 at 10:27:39PM +0100, Rafael J. Wysocki wrote:
> On Sunday, February 12, 2012, mark gross wrote:
> > why are these two patches?
> > 
> > I think you could squash them into one and it would be just fine.
> 
> Well, what sense does _device_ PM QoS make without CONFIG_PM?

I forgot about device pm qos.  

your right.

--mark

> Rafael
> 
> 
> > On Tue, Feb 07, 2012 at 09:34:06AM +0100, Jean Pihet wrote:
> > > The per-device PM QoS feature depends on CONFIG_PM which depends
> > > on PM_SLEEP || PM_RUNTIME. This breaks CPU C-states with kernels
> > > not having these CONFIGs.
> > > 
> > > This patch allows the feature in all cases.
> > > 
> > > Signed-off-by: Jean Pihet <j-pihet <at> ti.com>
> > > Cc: Rafael J. Wysocki <rjw <at> sisk.pl>
> > > Cc: Mark Gross <markgross <at> thegnar.org>
> > > ---
> > >  drivers/base/power/Makefile |    3 ++-
> > >  include/linux/pm_qos.h      |   39 ---------------------------------------
(Continue reading)


Gmane