Date:	Tue, 20 Jan 2015 04:34:16 +0000 (GMT)
From:	MyungJoo Ham <myungjoo.ham@...sung.com>
To:	최찬우 <cw00.choi@...sung.com>,
	"kgene@...nel.org" <kgene@...nel.org>
Cc:	박경민 <kyungmin.park@...sung.com>,
	"rafael.j.wysocki@...el.com" <rafael.j.wysocki@...el.com>,
	"mark.rutland@....com" <mark.rutland@....com>,
	ABHILASH KESAVAN <a.kesavan@...sung.com>,
	"tomasz.figa@...il.com" <tomasz.figa@...il.com>,
	Krzysztof Kozlowski <k.kozlowski@...sung.com>,
	Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>,
	"robh+dt@...nel.org" <robh+dt@...nel.org>,
	대인기 <inki.dae@...sung.com>,
	"linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"linux-arm-kernel@...ts.infradead.org" 
	<linux-arm-kernel@...ts.infradead.org>,
	"linux-samsung-soc@...r.kernel.org" 
	<linux-samsung-soc@...r.kernel.org>
Subject: Re: [PATCHv8 1/9] devfreq: event: Add new devfreq_event class to
 provide basic data for devfreq governor

>   
>  This patch add new devfreq_event class for devfreq_event device which provide
> raw data (e.g., memory bus utilization/GPU utilization). This raw data from
> devfreq_event data would be used for the governor of devfreq subsystem.
> - devfreq_event device : Provide raw data for governor of existing devfreq device
> - devfreq device       : Monitor device state and change frequency/voltage of device
>                          using the raw data from devfreq_event device
> 
> The devfreq subsystem support generic DVFS(Dynamic Voltage/Frequency Scaling)
> for Non-CPU Devices. The devfreq device would dertermine current device state
> using various governor (e.g., ondemand, performance, powersave). After completed
> determination of system state, devfreq device would change the frequency/voltage
> of devfreq device according to the result of governor.
> 
> But, devfreq governor must need basic data which indicates current device state.
> Existing devfreq subsystem only consider devfreq device which check current system
> state and determine proper system state using basic data. There is no subsystem
> for device providing basic data to devfreq device.
> 
> The devfreq subsystem must need devfreq_event device(data-provider device) for
> existing devfreq device. So, this patch add new devfreq_event class for
> devfreq_event device which read various basic data(e.g, memory bus utilization,
> GPU utilization) and provide measured data to existing devfreq device through
> standard APIs of devfreq_event class.
> 
> The following description explains the feature of two kind of devfreq class:
> - devfreq class (existing)
>  : devfreq consumer device use raw data from devfreq_event device for
>    determining proper current system state and change voltage/frequency
>    dynamically using various governors.
> 
> - devfreq_event class (new)
>  : Provide measured raw data to devfreq device for governor
> 
> Cc: MyungJoo Ham <myungjoo.ham@...sung.com>
> Cc: Kyungmin Park <kyungmin.park@...sung.com>
> Signed-off-by: Chanwoo Choi <cw00.choi@...sung.com>
> ---

[]

> +/**
> + * devfreq_event_enable_edev() - Enable the devfreq-event dev and increase
> + *				 the enable_count of devfreq-event dev.
> + * @edev	: the devfreq-event device
> + *
> + * Note that this function increase the enable_count and enable the
> + * devfreq-event device. The devfreq-event device should be enabled before
> + * using it by devfreq device.
> + */
> +int devfreq_event_enable_edev(struct devfreq_event_dev *edev)
> +{
> +	int ret = 0;
> +
> +	if (!edev || !edev->desc)
> +		return -EINVAL;
> +
> +	mutex_lock(&edev->lock);
> +	if (edev->desc->ops && edev->desc->ops->enable) {
> +		ret = edev->desc->ops->enable(edev);
> +		if (ret < 0)
> +			goto err;
> +	}

Is there any reason to call enable(edev) even when enable_count is already > 0 
while you do not call disable(edev) while enable_count > 0?

I think this may incur errors in the related device drivers.
(e.g., incorrect pairing of clk/runtime-pm/regulator enable/disable
at the device driver side)

> +	edev->enable_count++;
> +err:
> +	mutex_unlock(&edev->lock);
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(devfreq_event_enable_edev);
> +
> +/**
> + * devfreq_event_disable_edev() - Disable the devfreq-event dev and decrease
> + *				  the enable_count of the devfreq-event dev.
> + * @edev	: the devfreq-event device
> + *
> + * Note that this function decrease the enable_count and disable the
> + * devfreq-event device. After the devfreq-event device is disabled,
> + * devfreq device can't use the devfreq-event device for get/set/reset
> + * operations.
> + */
> +int devfreq_event_disable_edev(struct devfreq_event_dev *edev)
> +{
> +	int ret = 0;
> +
> +	if (!edev || !edev->desc)
> +		return -EINVAL;
> +
> +	mutex_lock(&edev->lock);
> +	if (edev->enable_count > 0) {
> +		edev->enable_count--;
> +	} else {
> +		dev_warn(&edev->dev, "unbalanced enable_count\n");
> +		ret = -EINVAL;
> +		goto err;
> +	}
> +
> +	if (edev->desc->ops && edev->desc->ops->disable) {
> +		ret = edev->desc->ops->disable(edev);
> +		if (ret < 0) {
> +			edev->enable_count++;
> +			goto err;
> +		}
> +	}

You did it correctly with disable here;
not calling it when it is not required.

> +err:
> +	mutex_unlock(&edev->lock);
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(devfreq_event_disable_edev);
> +

[]
> +EXPORT_SYMBOL_GPL(devfreq_event_is_enabled);
[]

> +EXPORT_SYMBOL_GPL(devfreq_event_set_event);
[]

> +
> +/**
> + * devfreq_event_get_event() - Get event and total_event from devfreq-event dev.
> + * @edev	: the devfreq-event device
> + * @edata	: the calculated data of devfreq-event device
> + *
> + * Note that this function get the calculated event data from devfreq-event dev
> + * after stoping the progress of whole sequence of devfreq-event dev.
> + */
> +int devfreq_event_get_event(struct devfreq_event_dev *edev,
> +			    struct devfreq_event_data *edata)
> +{
> +	int ret;
> +
> +	if (!edev || !edev->desc)
> +		return -EINVAL;
> +
> +	if (!edev->desc->ops || !edev->desc->ops->get_event)
> +		return -EINVAL;
> +
> +	if (!devfreq_event_is_enabled(edev))
> +		return -EINVAL;
> +
> +	edata->event = edata->total_event = 0;
> +
> +	mutex_lock(&edev->lock);
> +	ret = edev->desc->ops->get_event(edev, edata);
> +	mutex_unlock(&edev->lock);
> +
> +	if ((edata->total_event <= 0)
> +		|| (edata->event > edata->total_event)) {
> +		edata->event = edata->total_event = 0;
> +		ret = -EINVAL;

total_event is unsigned. (cannot be < 0)

Plus, get_event() may already have returned a negative value with
the intention of giving a error different from EINVAL.

I would just simply set total_event = event = 0 if ret < 0.


> +	}
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(devfreq_event_get_event);
> +
> +/**
> + * devfreq_event_reset_event() - Reset all opeations of devfreq-event dev.
> + * @edev	: the devfreq-event device
> + *
> + * Note that this function stop all operations of devfreq-event dev and reset
> + * the current event data to make the devfreq-event device into initial state.
> + */
> +int devfreq_event_reset_event(struct devfreq_event_dev *edev)
> +{
> +	int ret = 0;
> +
> +	if (!edev || !edev->desc)
> +		return -EINVAL;
> +
> +	if (!devfreq_event_is_enabled(edev))
> +		return -EPERM;
> +
> +	mutex_lock(&edev->lock);
> +	if (edev->desc->ops && edev->desc->ops->reset)
> +		ret = edev->desc->ops->reset(edev);
> +	mutex_unlock(&edev->lock);

In the context of the get_event() handling "load",
aren't you supposed to set total_event = event = 0; here?

> +
> +	if (ret < 0)
> +		return -EINVAL;
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(devfreq_event_reset_event);
> +
> +/**
> + * devfreq_event_get_drvdata() - Return driver-data of devfreq-event dev.
> + * @edev	: the devfreq-event device
> + *
> + * Note that this function return the driver-data of devfreq-event device.
> + */
> +void *devfreq_event_get_drvdata(struct devfreq_event_dev *edev)
> +{
> +	return edev->desc->driver_data;
> +}
> +EXPORT_SYMBOL_GPL(devfreq_event_get_drvdata);

Looks like you can simply write this in the header file.
(either #define or static inline function)

[]

> +EXPORT_SYMBOL_GPL(devfreq_event_get_edev_by_phandle);
[]
> +EXPORT_SYMBOL_GPL(devfreq_event_get_edev_count);
[]

> +/**
> + * devfreq_event_add_edev() - Add new devfreq-event device.
> + * @dev		: the device owning the devfreq-event device being created
> + * @desc	: the devfreq-event device's decriptor which include essential
> + *		  data for devfreq-event device.
> + *
> + * Note that this function add new devfreq-event device to devfreq-event class
> + * list and register the device of the devfreq-event device.
> + */
> +struct devfreq_event_dev *devfreq_event_add_edev(struct device *dev,
> +						struct devfreq_event_desc *desc)
> +{
> +	struct devfreq_event_dev *edev;
> +	static atomic_t event_no = ATOMIC_INIT(0);
> +	int ret;
> +
> +	if (!dev || !desc)
> +		return ERR_PTR(-EINVAL);
> +
> +	if (!desc->name || !desc->ops)
> +		return ERR_PTR(-EINVAL);
> +
> +	if (!desc->ops->set_event || !desc->ops->get_event)
> +		return ERR_PTR(-EINVAL);
> +
> +	edev = kzalloc(sizeof(struct devfreq_event_dev), GFP_KERNEL);

devm_* here? (seeing no free)

> +	if (!edev)
> +		return ERR_PTR(-ENOMEM);
> +
> +	mutex_init(&edev->lock);
> +	edev->desc = desc;
> +	edev->dev.parent = dev;
> +	edev->dev.class = devfreq_event_class;
> +	edev->dev.release = devfreq_event_release_edev;
> +
> +	dev_set_name(&edev->dev, "event.%d", atomic_inc_return(&event_no) - 1);
> +	ret = device_register(&edev->dev);
> +	if (ret < 0) {
> +		put_device(&edev->dev);
> +		return ERR_PTR(ret);
> +	}
> +	dev_set_drvdata(&edev->dev, edev);
> +
> +	INIT_LIST_HEAD(&edev->node);
> +
> +	mutex_lock(&devfreq_event_list_lock);
> +	list_add(&edev->node, &devfreq_event_list);
> +	mutex_unlock(&devfreq_event_list_lock);
> +
> +	return edev;
> +}
> +EXPORT_SYMBOL_GPL(devfreq_event_add_edev);
> +
> +/**
> + * devfreq_event_remove_edev() - Remove the devfreq-event device registered.
> + * @dev		: the devfreq-event device
> + *
> + * Note that this function remove the registered devfreq-event device.
> + */
> +int devfreq_event_remove_edev(struct devfreq_event_dev *edev)
> +{
> +	if (!edev)
> +		return -EINVAL;
> +
> +	mutex_lock(&devfreq_event_list_lock);
> +	WARN_ON(edev->enable_count);

Let's WARN before mutex_lock();

> +	list_del(&edev->node);
> +	device_unregister(&edev->dev);

Let's unregister after mutex_unlock();

The two lines do not need to be protected by the mutex.

> +	mutex_unlock(&devfreq_event_list_lock);
> +
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(devfreq_event_remove_edev);
> +

[]

> diff --git a/drivers/devfreq/event/Kconfig b/drivers/devfreq/event/Kconfig
> new file mode 100644
> index 0000000..1ced42c
> --- /dev/null
> +++ b/drivers/devfreq/event/Kconfig
> @@ -0,0 +1,16 @@
> +menuconfig PM_DEVFREQ_EVENT
> +	bool "DEVFREQ-Event device Support"
> +	help
> +	  The devfreq-event device provide the raw data and events which
> +	  indicate the current state of devfreq-event device. The provided
> +	  data from devfreq-event device is used to monitor the state of
> +	  device and determine the suitable size of resource to reduce the
> +	  wasted resource.
> +
> +	  The devfreq-event device can support the various type of events
> +	  (e.g., raw data, utilization, latency, bandwidth). The events
> +	  may be used by devfreq governor and other subsystem.
> +
> +if PM_DEVFREQ_EVENT
> +
> +endif # PM_DEVFREQ_EVENT
> diff --git a/drivers/devfreq/event/Makefile b/drivers/devfreq/event/Makefile
> new file mode 100644
> index 0000000..dc56005
> --- /dev/null
> +++ b/drivers/devfreq/event/Makefile
> @@ -0,0 +1 @@
> +# Exynos DEVFREQ Event Drivers
> diff --git a/include/linux/devfreq-event.h b/include/linux/devfreq-event.h
> new file mode 100644
> index 0000000..b7363f5
> --- /dev/null
> +++ b/include/linux/devfreq-event.h
> @@ -0,0 +1,170 @@
> +/*
> + * devfreq-event: a framework to provide raw data and events of devfreq devices
> + *
> + * Copyright (C) 2014 Samsung Electronics
> + * Author: Chanwoo Choi <cw00.choi@...sung.com>
> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License version 2 as
> + * published by the Free Software Foundation.
> + */
> +
> +#ifndef __LINUX_DEVFREQ_EVENT_H__
> +#define __LINUX_DEVFREQ_EVENT_H__
> +
> +#include <linux/device.h>
> +
> +/**
> + * struct devfreq_event_dev - the devfreq-event device
> + *
> + * @node	: Contain the devfreq-event device that have been registered.
> + * @dev		: the device registered by devfreq-event class. dev.parent is
> + *		  the device using devfreq-event.
> + * @lock	: a mutex to protect accessing devfreq-event.
> + * @enable_count: the number of enable function have been called.
> + * @desc	: the description for devfreq-event device.
> + *
> + * This structure contains devfreq-event device information.
> + */
> +struct devfreq_event_dev {
> +	struct list_head node;
> +
> +	struct device dev;
> +	struct mutex lock;
> +	u32 enable_count;
> +
> +	const struct devfreq_event_desc *desc;
> +};
> +
> +/**
> + * struct devfreq_event_data - the devfreq-event data
> + *
> + * @event	: the load of devfreq-event device for polling period
> + * @total_event	: the total load of devfreq-event device for polling period
> + *
> + * This structure contains the data of devfreq-event device for polling period.
> + */
> +struct devfreq_event_data {
> +	unsigned long event;
> +	unsigned long total_event;
> +};

I would like to rephrase and rename as follows:

+ * @load_count		: load count of devfreq-event device for the given period.
+ * @total_count	: total count of devfreq-event device for the given period.
+ *			each count may represent a clock cycle,
+ *			a time unit (ns/us/...), or anything the device driver wants.
+ *			Generally, utilization is load_count / total_count.

[]


You may already contact me with synchronous methods for more
prompt discussion. :) Your office is not that far from mine.
(Or during lunch, dinner, or beer sessions as well)


Cheers,
MyungJoo

Powered by blists - more mailing lists