All of lore.kernel.org
 help / color / mirror / Atom feed
From: Kevin Hilman <khilman@linaro.org>
To: Rajagopal Venkat <rajagopal.venkat@linaro.org>
Cc: myungjoo.ham@samsung.com, rjw@sisk.pl,
	linaro-kernel@lists.linaro.org, linux-pm@vger.kernel.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 2/2] PM / devfreq: tie suspend/resume to runtime-pm
Date: Fri, 22 Mar 2013 10:34:34 -0700	[thread overview]
Message-ID: <87fvzncp91.fsf@linaro.org> (raw)
In-Reply-To: <1363872506-3630-3-git-send-email-rajagopal.venkat@linaro.org> (Rajagopal Venkat's message of "Thu, 21 Mar 2013 18:58:26 +0530")

Hi Rajagopal,

Rajagopal Venkat <rajagopal.venkat@linaro.org> writes:

> Allow device devfreq to be suspend/resume automatically with
> runtime pm suspend/resume. The devfreq drivers should be least
> cared when to suspend/resume the devfreq.

That is the "what", but the changelog should also answer the question
"why?".

> pm_runtime_suspend(dev) will first suspend device devfreq(if available)
> before device is suspended from runtime pm core.

And here you should describe what you mentioned in your response to Alan
to clarify that suspending devfreq is suspending/resuming the load
monitoring, not any actual device hardware.  You should also describe
the rationale behind the ordering of devfreq susepnd/resume and device
suspend/resume.

> pm_runtime_resume(dev) will resume device devfreq(if available) after
> device is resumed from runtime pm core.
>
> Signed-off-by: Rajagopal Venkat <rajagopal.venkat@linaro.org>
> ---
>  drivers/base/power/runtime.c |   18 ++++++++++++++-
>  drivers/devfreq/devfreq.c    |   51 ++++++++++++++++++++++++++++++++++++++----
>  include/linux/devfreq.h      |   18 ++++-----------
>  include/linux/pm.h           |    2 ++
>  4 files changed, 70 insertions(+), 19 deletions(-)
>
> diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
> index 3148b10..951bf92 100644
> --- a/drivers/base/power/runtime.c
> +++ b/drivers/base/power/runtime.c
> @@ -11,6 +11,7 @@
>  #include <linux/export.h>
>  #include <linux/pm_runtime.h>
>  #include <trace/events/rpm.h>
> +#include <linux/devfreq.h>
>  #include "power.h"
>  
>  static int rpm_resume(struct device *dev, int rpmflags);
> @@ -406,6 +407,10 @@ static int rpm_suspend(struct device *dev, int rpmflags)
>  
>  	__update_runtime_status(dev, RPM_SUSPENDING);
>  
> +	if (dev->power.devfreq &&
> +			dev->power.devfreq->runtime_suspend(dev))
> +		goto fail;

It's not obvious to me why you want to abort the entire runtime suspend
path if the devfreq governor fails to runtime suspend.  That should also
be described someplace.

Same comment on skipping the callbacks in rpm_resume below.

>  	if (dev->pm_domain)
>  		callback = dev->pm_domain->ops.runtime_suspend;
>  	else if (dev->type && dev->type->pm)
> @@ -421,8 +426,11 @@ static int rpm_suspend(struct device *dev, int rpmflags)
>  		callback = dev->driver->pm->runtime_suspend;
>  
>  	retval = rpm_callback(callback, dev);
> -	if (retval)
> +	if (retval) {
> +		if (dev->power.devfreq)
> +			dev->power.devfreq->runtime_resume(dev);

What if ->runtime_resume is NULL?

>  		goto fail;
> +	}
>  
>   no_callback:
>  	__update_runtime_status(dev, RPM_SUSPENDED);
> @@ -638,6 +646,11 @@ static int rpm_resume(struct device *dev, int rpmflags)
>  
>  	__update_runtime_status(dev, RPM_RESUMING);
>  
> +	if (dev->power.devfreq)
> +		retval = dev->power.devfreq->runtime_resume(dev);

ditto.

I see that the devfreq core always sets these function pointers in this
patch, but that's not an assumption the PM core should make, IMO.

> +		if (retval)
> +			goto skip_callback;
> +
>  	if (dev->pm_domain)
>  		callback = dev->pm_domain->ops.runtime_resume;
>  	else if (dev->type && dev->type->pm)
> @@ -653,9 +666,12 @@ static int rpm_resume(struct device *dev, int rpmflags)
>  		callback = dev->driver->pm->runtime_resume;
>  
>  	retval = rpm_callback(callback, dev);
> +skip_callback:
>  	if (retval) {
>  		__update_runtime_status(dev, RPM_SUSPENDED);
>  		pm_runtime_cancel_pending(dev);
> +		if (dev->power.devfreq)
> +			dev->power.devfreq->runtime_suspend(dev);
>  	} else {
>   no_callback:
>  		__update_runtime_status(dev, RPM_ACTIVE);
> diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
> index 44c4079..ed6cb88 100644
> --- a/drivers/devfreq/devfreq.c
> +++ b/drivers/devfreq/devfreq.c
> @@ -25,6 +25,7 @@
>  #include <linux/list.h>
>  #include <linux/printk.h>
>  #include <linux/hrtimer.h>
> +#include <linux/pm_runtime.h>
>  #include "governor.h"
>  
>  static struct class *devfreq_class;
> @@ -42,6 +43,9 @@ static LIST_HEAD(devfreq_governor_list);
>  static LIST_HEAD(devfreq_list);
>  static DEFINE_MUTEX(devfreq_list_lock);
>  
> +static int devfreq_runtime_suspend(struct device *dev);
> +static int devfreq_runtime_resume(struct device *dev);
> +
>  /**
>   * find_device_devfreq() - find devfreq struct using device pointer
>   * @dev:	device pointer used to lookup device devfreq.
> @@ -476,6 +480,9 @@ struct devfreq *devfreq_add_device(struct device *dev,
>  	devfreq->previous_freq = profile->initial_freq;
>  	devfreq->data = data;
>  	devfreq->nb.notifier_call = devfreq_notifier_call;
> +	devfreq->runtime_suspend = devfreq_runtime_suspend;
> +	devfreq->runtime_resume = devfreq_runtime_resume;
> +	dev->power.devfreq = devfreq;
>  
>  	devfreq->trans_table =	devm_kzalloc(dev, sizeof(unsigned int) *
>  						devfreq->profile->max_state *
> @@ -549,7 +556,7 @@ EXPORT_SYMBOL(devfreq_remove_device);
>   * (e.g., runtime_suspend, suspend) of the device driver that
>   * holds the devfreq.
>   */
> -int devfreq_suspend_device(struct devfreq *devfreq)
> +static int devfreq_suspend_device(struct devfreq *devfreq)
>  {
>  	if (!devfreq)
>  		return -EINVAL;
> @@ -560,7 +567,6 @@ int devfreq_suspend_device(struct devfreq *devfreq)
>  	return devfreq->governor->event_handler(devfreq,
>  				DEVFREQ_GOV_SUSPEND, NULL);
>  }
> -EXPORT_SYMBOL(devfreq_suspend_device);

>  /**
>   * devfreq_resume_device() - Resume devfreq of a device.
> @@ -570,7 +576,7 @@ EXPORT_SYMBOL(devfreq_suspend_device);
>   * (e.g., runtime_resume, resume) of the device driver that
>   * holds the devfreq.
>   */
> -int devfreq_resume_device(struct devfreq *devfreq)
> +static int devfreq_resume_device(struct devfreq *devfreq)
>  {
>  	if (!devfreq)
>  		return -EINVAL;
> @@ -581,7 +587,44 @@ int devfreq_resume_device(struct devfreq *devfreq)
>  	return devfreq->governor->event_handler(devfreq,
>  				DEVFREQ_GOV_RESUME, NULL);
>  }
> -EXPORT_SYMBOL(devfreq_resume_device);
> +
> +/**
> + * devfreq_runtime_suspend() - Suspend devfreq of a device.
> + * @dev: the device associated with devfreq
> + *
> + * This function is intended to be called by the runtime-pm
> + * core when the device associated with devfreq is
> + * runtime suspended.
> + */
> +static int devfreq_runtime_suspend(struct device *dev)
> +{
> +	struct devfreq *devfreq;
> +
> +	mutex_lock(&devfreq_list_lock);
> +	devfreq = find_device_devfreq(dev);
> +	mutex_unlock(&devfreq_list_lock);
> +
> +	return devfreq_suspend_device(devfreq);
> +}
> +
> +/**
> + * devfreq_runtime_resume() - Resume devfreq of a device.
> + * @dev: the device associated with devfreq
> + *
> + * This function is intended to be called by the runtime-pm
> + * core when the device associated with devfreq is
> + * runtime resumed.
> + */
> +static int devfreq_runtime_resume(struct device *dev)
> +{
> +	struct devfreq *devfreq;
> +
> +	mutex_lock(&devfreq_list_lock);
> +	devfreq = find_device_devfreq(dev);
> +	mutex_unlock(&devfreq_list_lock);
> +
> +	return devfreq_resume_device(devfreq);
> +}
>  
>  /**
>   * devfreq_add_governor() - Add devfreq governor
> diff --git a/include/linux/devfreq.h b/include/linux/devfreq.h
> index 5f1ab92..f50f46e 100644
> --- a/include/linux/devfreq.h
> +++ b/include/linux/devfreq.h
> @@ -140,6 +140,8 @@ struct devfreq_governor {
>   * @trans_table:	Statistics of devfreq transitions
>   * @time_in_state:	Statistics of devfreq states
>   * @last_stat_updated:	The last time stat updated
> + * @runtime_suspend:	func to runtime suspend devfreq from runtime core
> + * @runtime_resume:	func to runtime resume devfreq from runtime core
>   *
>   * This structure stores the devfreq information for a give device.
>   *
> @@ -173,6 +175,8 @@ struct devfreq {
>  	unsigned int *trans_table;
>  	unsigned long *time_in_state;
>  	unsigned long last_stat_updated;
> +	int (*runtime_suspend)(struct device *dev);
> +	int (*runtime_resume)(struct device *dev);
>  };
>  
>  #if defined(CONFIG_PM_DEVFREQ)
> @@ -182,10 +186,6 @@ extern struct devfreq *devfreq_add_device(struct device *dev,
>  				  void *data);
>  extern int devfreq_remove_device(struct devfreq *devfreq);
>  
> -/* Supposed to be called by PM_SLEEP/PM_RUNTIME callbacks */
> -extern int devfreq_suspend_device(struct devfreq *devfreq);
> -extern int devfreq_resume_device(struct devfreq *devfreq);

Remvoing these, and removing the EXPORT_SYMBOL()s is an API change that
is not described in the changelog.

Kevin

> -
>  /* Helper functions for devfreq user device driver with OPP. */
>  extern struct opp *devfreq_recommended_opp(struct device *dev,
>  					   unsigned long *freq, u32 flags);
> @@ -228,16 +228,6 @@ static inline int devfreq_remove_device(struct devfreq *devfreq)
>  	return 0;
>  }
>  
> -static inline int devfreq_suspend_device(struct devfreq *devfreq)
> -{
> -	return 0;
> -}
> -
> -static inline int devfreq_resume_device(struct devfreq *devfreq)
> -{
> -	return 0;
> -}
> -
>  static inline struct opp *devfreq_recommended_opp(struct device *dev,
>  					   unsigned long *freq, u32 flags)
>  {
> diff --git a/include/linux/pm.h b/include/linux/pm.h
> index 03d7bb1..6097db1 100644
> --- a/include/linux/pm.h
> +++ b/include/linux/pm.h
> @@ -40,6 +40,7 @@ extern void (*pm_power_off_prepare)(void);
>   */
>  
>  struct device;
> +struct devfreq;
>  
>  #ifdef CONFIG_PM
>  extern const char power_group_name[];		/* = "power" */
> @@ -549,6 +550,7 @@ struct dev_pm_info {
>  #endif
>  	struct pm_subsys_data	*subsys_data;  /* Owned by the subsystem. */
>  	struct dev_pm_qos	*qos;
> +	struct devfreq		*devfreq;
>  };
>  
>  extern void update_pm_runtime_accounting(struct device *dev);

  parent reply	other threads:[~2013-03-22 17:34 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-03-21 13:28 [PATCH 0/2] Add devfreq runtime pm support Rajagopal Venkat
2013-03-21 13:28 ` [PATCH 1/2] PM / devfreq: Fix compiler warnings Rajagopal Venkat
2013-03-21 13:28 ` [PATCH 2/2] PM / devfreq: tie suspend/resume to runtime-pm Rajagopal Venkat
2013-03-21 15:02   ` Alan Stern
2013-03-21 15:02     ` Alan Stern
2013-03-22  9:05     ` Rajagopal Venkat
2013-03-22 17:34   ` Kevin Hilman [this message]
2013-03-25  6:13     ` Rajagopal Venkat

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=87fvzncp91.fsf@linaro.org \
    --to=khilman@linaro.org \
    --cc=linaro-kernel@lists.linaro.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pm@vger.kernel.org \
    --cc=myungjoo.ham@samsung.com \
    --cc=rajagopal.venkat@linaro.org \
    --cc=rjw@sisk.pl \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.