本文部分内容参考

万字长文 | Thermal框架源码剖析，

Linux Thermal机制源码分析之框架概述_不捡风筝的玖伍贰柒的博客-CSDN博客，

“热散由心静，凉生为室空” - linux温控的那些事儿_内核工匠的博客-CSDN博客，

Linux thermal governor之IPA分析_内核工匠的博客-CSDN博客

特此致谢！

接前一篇文章Linux内核Thermal框架详解十三、Thermal Governor（3）

二、具体温控策略

上一篇文章介绍并详细分析了fair_share governor的源码。本文介绍第3种温控策略：step_wise。

3. step_wise

step_wise是CPU温控常用的governor，其在温控管理上有着良好的表现。在计算target cooling state的过程中，step_wise除了根据配置的温度触发点（TripPoint）获知是否throttle外，还添加了一个参考项：trend，即温升趋势。Linux Thermal框架定义了三种trend type：上升（RAISING）、下降（DROPPING）、稳定(STABLE)。step_wise将温升趋势（trend）分为三种：上升、下降、稳定。step_wise根据配置的温度触发点及温度趋势以步进的方式调节CPU的状态，具体来讲就是调节频率。比如：温度已经超过触发点同时温度处于上升状态，则step-wise就会将CPU的coolingstate上调一级（对应操作是频率降低），然后继续轮询CPU的温度，通过判断温度趋势再执行相应动作。

一句话概括，step_wise governor根据cur_state、温升趋势trend、是否throttle去计算cooling_device的target_state，从而达到控制cooling_device来控制温升。

step_wise governor的代码在drivers/thermal/gov_step_wise.c中。它比bang_bang和fair_share复杂了不少，代码一共有200行出头。由于代码较长，在此就不全部列出了，而是直接分段贴出并进行分析。

（1）THERMAL_GOVERNOR_DECLARE相关代码

先来看THERMAL_GOVERNOR_DECLARE。它是一个宏定义，在drivers/thermal/thermal_core.h中，代码如下：

/* Init section thermal table */
extern struct thermal_governor *__governor_thermal_table[];
extern struct thermal_governor *__governor_thermal_table_end[];#define THERMAL_TABLE_ENTRY(table, name)			\static typeof(name) *__thermal_table_entry_##name	\__used __section("__" #table "_thermal_table") = &name#define THERMAL_GOVERNOR_DECLARE(name)	THERMAL_TABLE_ENTRY(governor, name)

实际上这段代码在前文Linux内核Thermal框架详解四、Thermal Core（3）中已经进行了详细分析，这里就不再赘述了。不过为了便于理解和加深印象，将step_wise governor展开后的代码再次列出：

static struct thermal_governor thermal_gov_step_wise = {.name		= "step_wise",.throttle	= step_wise_throttle,
};static struct thermal_governor *__thermal_table_entry_thermal_gov_step_wise    \__used __section("__governor_thermal_table") = &thermal_gov_step_wise

Thermal Governor都是通过THERMAL_GOVERNOR_DECLARE定义到了__governor_thermal_table这段空间内。然后在thermal core初始化时通过调用thermal_register_governors来注册到thermal_governor_list链表中。再之后通过经由“thermal_init->thermal_register_governors-> thermal_set_governor”路径和thermal zone device关联上。

（2）handle_non_critical_trips

struct thermal_governor中有一个成员throttle，其是一个函数指针：

int (*throttle)(struct thermal_zone_device *tz, int trip);

对于对象thermal_gov_step_wise来说，指向了step_wise_throttle函数。在解析step_wise_throttle函数之前，有一个问题必须弄清楚：这个函数是何时被调用的？

是在drivers/thermal/thermal_core.c的handle_non_critical_trips函数中，代码如下：

static void handle_non_critical_trips(struct thermal_zone_device *tz, int trip)
{tz->governor ? tz->governor->throttle(tz, trip) :def_governor->throttle(tz, trip);
}

那么又是哪里调用的handle_non_critical_trips？是在drivers/thermal/thermal_core.c的handle_thermal_trip函数中，代码如下：

static void handle_thermal_trip(struct thermal_zone_device *tz, int trip)
{enum thermal_trip_type type;int trip_temp, hyst = 0;/* Ignore disabled trip points */if (test_bit(trip, &tz->trips_disabled))return;tz->ops->get_trip_temp(tz, trip, &trip_temp);tz->ops->get_trip_type(tz, trip, &type);if (tz->ops->get_trip_hyst)tz->ops->get_trip_hyst(tz, trip, &hyst);if (tz->last_temperature != THERMAL_TEMP_INVALID) {if (tz->last_temperature < trip_temp &&tz->temperature >= trip_temp)thermal_notify_tz_trip_up(tz->id, trip,tz->temperature);if (tz->last_temperature >= trip_temp &&tz->temperature < (trip_temp - hyst))thermal_notify_tz_trip_down(tz->id, trip,tz->temperature);}if (type == THERMAL_TRIP_CRITICAL || type == THERMAL_TRIP_HOT)handle_critical_trips(tz, trip, type);elsehandle_non_critical_trips(tz, trip);/** Alright, we handled this trip successfully.* So, start monitoring again.*/monitor_thermal_zone(tz);
}

对于handle_thermal_trip函数的详细分析有专门的文章章节，由于本篇文章专注于bang_bang governor，故在此不深入展开。

（3）step_wise_throttle

step_wise_throttle函数无疑是step_wise governor的核心。代码如下：

/*** step_wise_throttle - throttles devices associated with the given zone* @tz: thermal_zone_device* @trip: trip point index** Throttling Logic: This uses the trend of the thermal zone to throttle.* If the thermal zone is 'heating up' this throttles all the cooling* devices associated with the zone and its particular trip point, by one* step. If the zone is 'cooling down' it brings back the performance of* the devices by one step.*/
static int step_wise_throttle(struct thermal_zone_device *tz, int trip)
{struct thermal_instance *instance;thermal_zone_trip_update(tz, trip);mutex_lock(&tz->lock);list_for_each_entry(instance, &tz->thermal_instances, tz_node)thermal_cdev_update(instance->cdev);mutex_unlock(&tz->lock);return 0;
}

函数注释已经将函数功能说得很清楚了：对与给定thermal zone关联的设备进行节流。调节逻辑如下：

本策略使用thermal zone的trend（温升趋势）。

如果thermal zone正在“升温”，则会对与thermal zone及其特定trip point相关的所有冷却设备进行节流，一步到位；

如果thermal zone正在“降温”，则将会一步到位地恢复设备的性能。

老外写东西比较拗口，没有文字功底。用更通俗的语言来解释step_wise governor对于cooling_state的计算（选择）策略：

• 当throttle发生且温升趋势为上升，使用更高一级的cooling state；
• 当throttle发生且温升趋势为下降，不改变cooling state；
• 当throttle未发生且温升趋势为上升，不改变cooling state；
• 当throttle未发生（解除）且温升趋势为下降，使用更低一级的cooling state。

再更细化一下：

• 当温升趋势为上升，分为两种情况：
  •  若发生throttle，则使用更高一级的cooling state；
  • 若未发生throttle，则不改变cooling state；
• 当温升趋势为下降，分为两种情况：
  •  若发生throttle，则不改变cooling state；
  • 若未发生（解除）throttle，则使用更低一级的cooling state。
• 当达到最高温线且发生throttle，使用最高级的cooling state；
• 当达到最低温线且发生throttle，使用最低级的cooling state。

注意： cooling state的取值范围在[instance->lower, instance->upper]。若cur_state < instance->lower，target_state则取值为THERMAL_NO_TARGET。

由此看来，step_wise governor是每个轮询周期逐级提高冷却状态，是一种相对温和的温控策略。

step_wise governor流程图如下所示：

 （4）thermal_zone_trip_update

step_wise_throttle函数中调用了thermal_zone_trip_update函数，这个函数就在step_wise_throttle函数的上边，代码如下：

static void thermal_zone_trip_update(struct thermal_zone_device *tz, int trip)
{int trip_temp;enum thermal_trip_type trip_type;enum thermal_trend trend;struct thermal_instance *instance;bool throttle = false;int old_target;tz->ops->get_trip_temp(tz, trip, &trip_temp);tz->ops->get_trip_type(tz, trip, &trip_type);trend = get_tz_trend(tz, trip);if (tz->temperature >= trip_temp) {throttle = true;trace_thermal_zone_trip(tz, trip, trip_type);}dev_dbg(&tz->device, "Trip%d[type=%d,temp=%d]:trend=%d,throttle=%d\n",trip, trip_type, trip_temp, trend, throttle);mutex_lock(&tz->lock);list_for_each_entry(instance, &tz->thermal_instances, tz_node) {if (instance->trip != trip)continue;old_target = instance->target;instance->target = get_target_state(instance, trend, throttle);dev_dbg(&instance->cdev->device, "old_target=%d, target=%d\n",old_target, (int)instance->target);if (instance->initialized && old_target == instance->target)continue;/* Activate a passive thermal instance */if (old_target == THERMAL_NO_TARGET &&instance->target != THERMAL_NO_TARGET)update_passive_instance(tz, trip_type, 1);/* Deactivate a passive thermal instance */else if (old_target != THERMAL_NO_TARGET &&instance->target == THERMAL_NO_TARGET)update_passive_instance(tz, trip_type, -1);instance->initialized = true;mutex_lock(&instance->cdev->lock);instance->cdev->updated = false; /* cdev needs update */mutex_unlock(&instance->cdev->lock);}mutex_unlock(&tz->lock);
}

thermal_zone_trip_update函数做了以下几件事情：

1）获取触发温度

通过tz->ops->get_trip_temp(tz, trip, &trip_temp)获取所属thermal zone的trip_temp。

2）获取触发类型

通过tz->ops->get_trip_type(tz, trip, &trip_type)获取所属thermal zone的trip_type。

3）获取温升趋势

通过get_tz_trend(tz, trip)获取所属thermal zone的温升趋势（trend）。get_tz_trend函数在drivers/thermal/thermal_helpers.c中，代码如下：

int get_tz_trend(struct thermal_zone_device *tz, int trip)
{enum thermal_trend trend;if (tz->emul_temperature || !tz->ops->get_trend ||tz->ops->get_trend(tz, trip, &trend)) {if (tz->temperature > tz->last_temperature)trend = THERMAL_TREND_RAISING;else if (tz->temperature < tz->last_temperature)trend = THERMAL_TREND_DROPPING;elsetrend = THERMAL_TREND_STABLE;}return trend;
}
EXPORT_SYMBOL(get_tz_trend);

tz->ops->get_trend实际指向了drivers/acpi/thermal.c中的thermal_get_trend函数。赋值函数指针的动作也是在此文件中完成的，代码如下：

static struct thermal_zone_device_ops acpi_thermal_zone_ops = {.bind = acpi_thermal_bind_cooling_device,.unbind	= acpi_thermal_unbind_cooling_device,.get_temp = thermal_get_temp,.get_trip_type = thermal_get_trip_type,.get_trip_temp = thermal_get_trip_temp,.get_crit_temp = thermal_get_crit_temp,.get_trend = thermal_get_trend,.hot = acpi_thermal_zone_device_hot,.critical = acpi_thermal_zone_device_critical,
};

thermal_get_trend函数的代码如下：

static int thermal_get_trend(struct thermal_zone_device *thermal,int trip, enum thermal_trend *trend)
{struct acpi_thermal *tz = thermal->devdata;enum thermal_trip_type type;int i;if (thermal_get_trip_type(thermal, trip, &type))return -EINVAL;if (type == THERMAL_TRIP_ACTIVE) {int trip_temp;int temp = deci_kelvin_to_millicelsius_with_offset(tz->temperature, tz->kelvin_offset);if (thermal_get_trip_temp(thermal, trip, &trip_temp))return -EINVAL;if (temp > trip_temp) {*trend = THERMAL_TREND_RAISING;return 0;} else {/* Fall back on default trend */return -EINVAL;}}/** tz->temperature has already been updated by generic thermal layer,* before this callback being invoked*/i = (tz->trips.passive.tc1 * (tz->temperature - tz->last_temperature))+ (tz->trips.passive.tc2* (tz->temperature - tz->trips.passive.temperature));if (i > 0)*trend = THERMAL_TREND_RAISING;else if (i < 0)*trend = THERMAL_TREND_DROPPING;else*trend = THERMAL_TREND_STABLE;return 0;
}

4）获取（计算）每个instance目标状态

依次通过get_target_state(instance, trend, throttle)获取所属thermal zone的每个instance的目标状态（target state）。get_target_state函数在同文件（drivers/thermal/gov_step_wise.c）中，代码如下：

/** If the temperature is higher than a trip point,*    a. if the trend is THERMAL_TREND_RAISING, use higher cooling*       state for this trip point*    b. if the trend is THERMAL_TREND_DROPPING, do nothing*    c. if the trend is THERMAL_TREND_RAISE_FULL, use upper limit*       for this trip point*    d. if the trend is THERMAL_TREND_DROP_FULL, use lower limit*       for this trip point* If the temperature is lower than a trip point,*    a. if the trend is THERMAL_TREND_RAISING, do nothing*    b. if the trend is THERMAL_TREND_DROPPING, use lower cooling*       state for this trip point, if the cooling state already*       equals lower limit, deactivate the thermal instance*    c. if the trend is THERMAL_TREND_RAISE_FULL, do nothing*    d. if the trend is THERMAL_TREND_DROP_FULL, use lower limit,*       if the cooling state already equals lower limit,*       deactivate the thermal instance*/
static unsigned long get_target_state(struct thermal_instance *instance,enum thermal_trend trend, bool throttle)
{struct thermal_cooling_device *cdev = instance->cdev;unsigned long cur_state;unsigned long next_target;/** We keep this instance the way it is by default.* Otherwise, we use the current state of the* cdev in use to determine the next_target.*/cdev->ops->get_cur_state(cdev, &cur_state);next_target = instance->target;dev_dbg(&cdev->device, "cur_state=%ld\n", cur_state);if (!instance->initialized) {if (throttle) {next_target = (cur_state + 1) >= instance->upper ?instance->upper :((cur_state + 1) < instance->lower ?instance->lower : (cur_state + 1));} else {next_target = THERMAL_NO_TARGET;}return next_target;}switch (trend) {case THERMAL_TREND_RAISING:if (throttle) {next_target = cur_state < instance->upper ?(cur_state + 1) : instance->upper;if (next_target < instance->lower)next_target = instance->lower;}break;case THERMAL_TREND_RAISE_FULL:if (throttle)next_target = instance->upper;break;case THERMAL_TREND_DROPPING:if (cur_state <= instance->lower) {if (!throttle)next_target = THERMAL_NO_TARGET;} else {if (!throttle) {next_target = cur_state - 1;if (next_target > instance->upper)next_target = instance->upper;}}break;case THERMAL_TREND_DROP_FULL:if (cur_state == instance->lower) {if (!throttle)next_target = THERMAL_NO_TARGET;} elsenext_target = instance->lower;break;default:break;}return next_target;
}

至此，step_wise governor策略就大体上分析完了。