Linux Thermal框架

Youth cowboy

于 2024-01-22 15:39:04 发布

阅读量924

点赞数 21

分类专栏： thermal Power 文章标签： linux

本文链接：https://blog.csdn.net/youthcowboy/article/details/135749357

版权

Power 同时被 2 个专栏收录

7 篇文章 3 订阅

订阅专栏

thermal

2 篇文章 0 订阅

订阅专栏

一、功能介绍

Linux的Thermal机制是基于Zone为单位的热管理机制，核心包括三个部分：获取区域温度的设备thermal_zone_device、区域降温的设备thermal_cooling_device、温控策略thermal_governor。thermal_governor从thermal_zone_device获取区域温度，然后根据当前温度，决定调用哪个降温设备来为该区域降温。

二、linux thermal框架

Linux Thermal框架可以分为Thermal Core、Thermal Governor、Thermal Cooling、Thermal Driver以及Thermal Device Tree五大部分。

Thermal Core：用于和user space、Thermal Governor、Thermal Driver交互。

Thermal Governor：主要包括gov_bang_bang、gov_fair_share、gov_power_allocator、gov_step_wise、gov_user_space等，最常用的为gov_power_allocator.

Thermal Cooling：主要包括cpufreq_cooling、cpuidle_cooling、devfreq_cooling等。

2.1 thermal core

内核将采集区域温度的设备抽象为结构体struct thermal_zone_device，主要成员包括：char type[]设备名称；int temperature当前温度；int last_temperature上次采集温度；struct thermal_governor *governor对应governor；int polling_delay温度采集时间间隔等等。其中struct thermal_zone_device_ops *ops是采集区域温度设备的操作抽象，包括绑定降温设备、获取设备温度等。

kernel/linux/thermal.h中定义了thermal_zone_device & thermal_zone_device_ops、thermal_governor、thermal_cooling_device & thermal_cooling_device_ops 结构体。

struct thermal_zone_device {
        int id;    // 设备的唯一标识符
        char type[THERMAL_NAME_LENGTH];    // 设备名称
        struct device device;    // 设备相关联的struct device结构体
        struct thermal_attr *trip_temp_attrs;    // 温度触发器（trip）的温度属性链表
        struct thermal_attr *trip_type_attrs;    // 温度触发器的触发类型属性链表
        struct thermal_attr *trip_hyst_attrs;    // 温度触发器的滞后属性链表
        void *devdata;
        int trips;
        unsigned long trips_disabled;        /* bitmap for disabled trips */
        int passive_delay;
        int polling_delay;    // 采集温度的时间间隔
        int temperature;    // 当前采集的温度
        int last_temperature;    // 上次采集的温度
        int emul_temperature;
        int passive;
        unsigned int forced_passive;    // 强制进入被动散热模式的标志
        atomic_t need_update;
        struct thermal_zone_device_ops *ops;    // 区域温度设备的操作
        struct thermal_zone_params *tzp;    // 记录一些信息，如governor name
        struct thermal_governor *governor;    // 温控策略
        void *governor_data;    
        struct list_head thermal_instances;    // 降温设备
        struct idr idr;    // 管理热区设备实例的ID
        struct mutex lock;
        struct list_head node;    // 热区设备的链表节点
        struct delayed_work poll_queue;    // 用于轮询区域温度
};

struct thermal_zone_params {
        char governor_name[THERMAL_NAME_LENGTH];

        /*
         * a boolean to indicate if the thermal to hwmon sysfs interface
         * is required. when no_hwmon == false, a hwmon sysfs interface
         * will be created. when no_hwmon == true, nothing will be done
         */
        bool no_hwmon;

        int num_tbps;        /* Number of tbp entries */
        struct thermal_bind_params *tbp;

        /*
         * Sustainable power (heat) that this thermal zone can dissipate in
         * mW
         */
        u32 sustainable_power;

        /*
         * Proportional parameter of the PID controller when
         * overshooting (i.e., when temperature is below the target)
         */
        s32 k_po;

        /*
         * Proportional parameter of the PID controller when
         * undershooting
         */
        s32 k_pu;

        /* Integral parameter of the PID controller */
        s32 k_i;

        /* Derivative parameter of the PID controller */
        s32 k_d;

        /* threshold below which the error is no longer accumulated */
        s32 integral_cutoff;

        /*
         * @slope:        slope of a linear temperature adjustment curve.
         *                 Used by thermal zone drivers.
         */
        int slope;
        /*
         * @offset:        offset of a linear temperature adjustment curve.
         *                 Used by thermal zone drivers (default 0).
         */
        int offset;
};

struct thermal_zone_device_ops {
        // 绑定一个降温设备到该热区设备
        int (*bind) (struct thermal_zone_device *,
                     struct thermal_cooling_device *);
        // 解绑一个降温设备从该热区设备
        int (*unbind) (struct thermal_zone_device *,
                       struct thermal_cooling_device *);
        // 获取当前热区设备的温度
        int (*get_temp) (struct thermal_zone_device *, int *);
        // 获取当前热区设备的工作模式
        int (*get_mode) (struct thermal_zone_device *,
                         enum thermal_device_mode *);
        // 设置当前热区设备的工作模式              
        int (*set_mode) (struct thermal_zone_device *,
                enum thermal_device_mode);
        // 获取指定温度触发器的触发类型        
        int (*get_trip_type) (struct thermal_zone_device *, int,
                enum thermal_trip_type *);
        // 获取触发等级对应的温度     
        int (*get_trip_temp) (struct thermal_zone_device *, int, int *);
        // 设置触发等级对应的温度
        int (*set_trip_temp) (struct thermal_zone_device *, int, int);
        int (*get_trip_hyst) (struct thermal_zone_device *, int, int *);
        int (*set_trip_hyst) (struct thermal_zone_device *, int, int);
        int (*get_crit_temp) (struct thermal_zone_device *, int *);
        int (*set_emul_temp) (struct thermal_zone_device *, int);
        // 获取温度的变化趋势
        int (*get_trend) (struct thermal_zone_device *, int,
                          enum thermal_trend *);
        int (*notify) (struct thermal_zone_device *, int,
                       enum thermal_trip_type);
};

// 内核将温控策略抽象为结构体struct thermal_governor，
//主要成员包括：char name[THERMAL_NAME_LENGTH]策略名称；int (*throttle)()温控决策等等。
struct thermal_governor {
        char name[THERMAL_NAME_LENGTH];
        int (*bind_to_tz)(struct thermal_zone_device *tz);
        void (*unbind_from_tz)(struct thermal_zone_device *tz);
        int (*throttle)(struct thermal_zone_device *tz, int trip);
        struct list_head        governor_list;
};
// 执行温控策略的设备成为区域降温设备，
//内核抽象为结构体struct thermal_cooling_device，struct thermal_cooling_device_ops是区域降温设备的操作集合。
struct thermal_cooling_device {
        int id;                           //每个thermal_cooling_device有独立的id
        char type[THERMAL_NAME_LENGTH];   // 名称
        struct device device;
        struct device_node *np;
        void *devdata;
        const struct thermal_cooling_device_ops *ops;
        bool updated; /* true if the cooling device does not need update */
        struct mutex lock; /* protect thermal_instances list */
        struct list_head thermal_instances;
        struct list_head node;
};

struct thermal_cooling_device_ops {
        //获取总的状态数，相当于降温等级
        int (*get_max_state) (struct thermal_cooling_device *, unsigned long *); 
        //获取当前状态  
        int (*get_cur_state) (struct thermal_cooling_device *, unsigned long *);  
        //设置状态  
        int (*set_cur_state) (struct thermal_cooling_device *, unsigned long); 
        // 获取所请求的功率     
        int (*get_requested_power)(struct thermal_cooling_device *,
                                   struct thermal_zone_device *, u32 *);
        // 将指定状态（降温等级）转换为对应的功率                          
        int (*state2power)(struct thermal_cooling_device *,
                           struct thermal_zone_device *, unsigned long, u32 *);
        // 将指定功率转换为对应的状态（降温等级）                   
        int (*power2state)(struct thermal_cooling_device *,
                           struct thermal_zone_device *, u32, unsigned long *);
};

2.1.1 初始化

2.1.1.1 thermal_governor注册

以step_wise governor为例，

int thermal_gov_step_wise_register(void)
{
        // 调用thermal_core.c中的方法
        return thermal_register_governor(&thermal_gov_step_wise);
}

static int __init thermal_init(void)
{
        int result;
        // 注册所有的governors       
        result = thermal_register_governors();
        if (result)
                goto error;

        result = class_register(&thermal_class);
        if (result)
                goto unregister_governors;

        result = genetlink_init();
        if (result)
                goto unregister_class;

        result = of_parse_thermal_zones();
        if (result)
                goto exit_netlink;

        result = register_pm_notifier(&thermal_pm_nb);
        if (result)
                pr_warn("Thermal: Can not register suspend notifier, return %d\n",
                        result);

        return 0;

exit_netlink:
        genetlink_exit();
unregister_class:
        class_unregister(&thermal_class);
unregister_governors:
        thermal_unregister_governors();
error:
        idr_destroy(&thermal_tz_idr);
        idr_destroy(&thermal_cdev_idr);
        mutex_destroy(&thermal_idr_lock);
        mutex_destroy(&thermal_list_lock);
        mutex_destroy(&thermal_governor_lock);
        return result;
}

static int __init thermal_register_governors(void)
{
        int result;
        // 调用step_wise governor中的方法，为系统默认的gov
        result = thermal_gov_step_wise_register();
        if (result)
                return result;

        result = thermal_gov_fair_share_register();
        if (result)
                return result;

        result = thermal_gov_bang_bang_register();
        if (result)
                return result;

        result = thermal_gov_user_space_register();
        if (result)
                return result;
        // 注册IPA governor
        return thermal_gov_power_allocator_register();
}

// 将第一个注册的governor设置为系统默认governor，即step_wise governor
int thermal_register_governor(struct thermal_governor *governor)
{
        int err;
        const char *name;
        struct thermal_zone_device *pos;
 
        if (!governor)
                return -EINVAL;
 
        mutex_lock(&thermal_governor_lock);
 
        err = -EBUSY;
        if (__find_governor(governor->name) == NULL) {
                err = 0;
                //链接到thermal_governor_list
                list_add(&governor->governor_list, &thermal_governor_list);
                if (!def_governor && !strncmp(governor->name,
                        DEFAULT_THERMAL_GOVERNOR, THERMAL_NAME_LENGTH))
                        def_governor = governor; //第一个设置为def_governor
        }
    .......
}

2.1.1.2 thermal_zone_device注册

struct thermal_zone_device *thermal_zone_device_register(const char *type,
        int trips, int mask, void *devdata,
        struct thermal_zone_device_ops *ops,
        struct thermal_zone_params *tzp,
        int passive_delay, int polling_delay)
{
        struct thermal_zone_device *tz;
        enum thermal_trip_type trip_type;
        int trip_temp;
        int result;
        int count;
        int passive = 0;
        struct thermal_governor *governor;
        .........................................................................
    //分配内存
        tz = kzalloc(sizeof(struct thermal_zone_device), GFP_KERNEL);
    .........................................................................
        //初始化idr，并获取id
        idr_init(&tz->idr);
        mutex_init(&tz->lock);
        result = get_idr(&thermal_tz_idr, &thermal_idr_lock, &tz->id);
        ..........................................................................
        strlcpy(tz->type, type ? : "", sizeof(tz->type));  //设置名称
        tz->ops = ops;    //操作集合
        tz->tzp = tzp;    //参数
        tz->device.class = &thermal_class;
        tz->devdata = devdata;
        tz->trips = trips;
        tz->passive_delay = passive_delay;
        tz->polling_delay = polling_delay;   //采集时间间隔
        /* A new thermal zone needs to be updated anyway. */
        atomic_set(&tz->need_update, 1);
    ........................................................................
        //根据governor name，设置降温策略
        if (tz->tzp)
                governor = __find_governor(tz->tzp->governor_name);
        else
                governor = def_governor;
    .........
        //链接到thermal_tz_list
        mutex_lock(&thermal_list_lock);
        list_add_tail(&tz->node, &thermal_tz_list);
        mutex_unlock(&thermal_list_lock);
 
        /* 尝试绑定已注册的降温设备 */
        bind_tz(tz);
 
        thermal_zone_device_reset(tz);
        /* Update the new thermal zone and mark it as already updated. */
        if (atomic_cmpxchg(&tz->need_update, 1, 0))
                thermal_zone_device_update(tz, THERMAL_EVENT_UNSPECIFIED);
 
        return tz;
    ..........
}

2.1.1.3 thermal_cooling_device注册

struct thermal_cooling_device *
thermal_cooling_device_register(char *type, void *devdata,
                                const struct thermal_cooling_device_ops *ops)
{
        return __thermal_cooling_device_register(NULL, type, devdata, ops);
}

static struct thermal_cooling_device *
__thermal_cooling_device_register(struct device_node *np,
                                  char *type, void *devdata,
                                  const struct thermal_cooling_device_ops *ops)
{
        struct thermal_cooling_device *cdev;
        struct thermal_zone_device *pos = NULL;
        int result;

        if (type && strlen(type) >= THERMAL_NAME_LENGTH)
                return ERR_PTR(-EINVAL);

        if (!ops || !ops->get_max_state || !ops->get_cur_state ||
            !ops->set_cur_state)
                return ERR_PTR(-EINVAL);
        // 分配内存
        cdev = kzalloc(sizeof(struct thermal_cooling_device), GFP_KERNEL);
        if (!cdev)
                return ERR_PTR(-ENOMEM);

        result = get_idr(&thermal_cdev_idr, &thermal_idr_lock, &cdev->id);
        if (result) {
                kfree(cdev);
                return ERR_PTR(result);
        }

        strlcpy(cdev->type, type ? : "", sizeof(cdev->type));
        mutex_init(&cdev->lock);
        INIT_LIST_HEAD(&cdev->thermal_instances);
        // 初始化成员，将mtk的ops和devdata赋值给thermal_cooling_device 
        cdev->np = np;
        cdev->ops = ops;
        cdev->updated = false;
        cdev->device.class = &thermal_class;
        cdev->device.groups = cooling_device_attr_groups;
        cdev->devdata = devdata;
        dev_set_name(&cdev->device, "cooling_device%d", cdev->id);
        // 注册device
        result = device_register(&cdev->device);
        if (result) {
                release_idr(&thermal_cdev_idr, &thermal_idr_lock, cdev->id);
                kfree(cdev);
                return ERR_PTR(result);
        }

        /* Add 'this' new cdev to the global cdev list */
        // 新的thermal_cooling_device加入到thermal_cdev_list链表
        mutex_lock(&thermal_list_lock);
        list_add(&cdev->node, &thermal_cdev_list);
        mutex_unlock(&thermal_list_lock);

        /* Update binding information for 'this' new cdev */
        // 尝试绑定到已注册的温度采集设备thermal_zone_device 
        bind_cdev(cdev);

        mutex_lock(&thermal_list_lock);
        list_for_each_entry(pos, &thermal_tz_list, node)
                if (atomic_cmpxchg(&pos->need_update, 1, 0))
                        thermal_zone_device_update(pos);
        mutex_unlock(&thermal_list_lock);

        return cdev;
}

static void bind_cdev(struct thermal_cooling_device *cdev)
{
        int i, ret;
        const struct thermal_zone_params *tzp;
        struct thermal_zone_device *pos = NULL;

        mutex_lock(&thermal_list_lock);
        // 遍历thermal_zone_device list，逐个绑定thermal_cooling_device 
        list_for_each_entry(pos, &thermal_tz_list, node) {
                if (!pos->tzp && !pos->ops->bind)
                        continue;

                if (pos->ops->bind) {
                        // 调用 thermal_zone_device中thermal_zone_device_ops成员中的bind方法
                        ret = pos->ops->bind(pos, cdev);
                        if (ret)
                                print_bind_err_msg(pos, cdev, ret);
                        continue;
                }

                tzp = pos->tzp;
                if (!tzp || !tzp->tbp)
                        continue;

                for (i = 0; i < tzp->num_tbps; i++) {
                        if (tzp->tbp[i].cdev || !tzp->tbp[i].match)
                                continue;
                        if (tzp->tbp[i].match(pos, cdev))
                                continue;
                        tzp->tbp[i].cdev = cdev;
                        __bind(pos, tzp->tbp[i].trip_mask, cdev,
                               tzp->tbp[i].binding_limits,
                               tzp->tbp[i].weight);
                }
        }

        mutex_unlock(&thermal_list_lock);
}

2.1.2 温度采集设备与降温设备的联系

同一个温度采集设备可以对应多个降温设备，结构体strcut thermal_instance用于连接温度采集设备与降温设备，成员struct thermal_zone_device *tz是对应的温度采集设备，struct thermal_cooling_device *cdev是对应的降温设备，int trip触发等级(对应一个温度)，当温度采集设备采集的温度达到一定值时，调用对应trip等级的降温设备。

struct thermal_instance {
    .................................................................
        struct thermal_zone_device *tz;      //对应温度采集设备
        struct thermal_cooling_device *cdev; //对应降温设备
        int trip;                            //触发等级
        struct list_head tz_node;            //链接到温度采集设备
        struct list_head cdev_node;          //链接到降温设备
    .................................................................
};

以温度采集设备绑定降温设备为例，当温度采集设备注册时会尝试绑定所有已经注册的降温设备。以CPU为例，bind接口对应的是tscpu_bind()，从代码中可以看出如果降温设备的名称为g_bind0--g_bind9中的一个将会绑定CPU温度采集设备和降温设备。tscpu_bind()接口中也定义了各种名称降温设备对应的触发等级。

static void bind_tz(struct thermal_zone_device *tz)
{
        int i, ret;
        struct thermal_cooling_device *pos = NULL;
        const struct thermal_zone_params *tzp = tz->tzp;
 
        if (!tzp && !tz->ops->bind)
                return;
 
        mutex_lock(&thermal_list_lock);
        if (tz->ops->bind) {
            //尝试绑定所有的已经注册的降温设备
                list_for_each_entry(pos, &thermal_cdev_list, node) {
                        ret = tz->ops->bind(tz, pos);
                        if (ret)
                                print_bind_err_msg(tz, pos, ret);
                }
                goto exit;
        }
    ...........................................................
}

static int tscpu_bind(struct thermal_zone_device *thermal, struct thermal_cooling_device *cdev)
{
        int table_val = 0;
 
        if (!strcmp(cdev->type, g_bind0)) {
                table_val = 0;
                tscpu_config_all_tc_hw_protect(trip_temp[0], tc_mid_trip);
        } else if (!strcmp(cdev->type, g_bind1)) {
                table_val = 1;
                tc_mid_trip = trip_temp[1];
                tscpu_config_all_tc_hw_protect(trip_temp[0], tc_mid_trip);
        } else if (!strcmp(cdev->type, g_bind2)) {
                table_val = 2;
        } else if (!strcmp(cdev->type, g_bind3)) {
                table_val = 3;
        } else if (!strcmp(cdev->type, g_bind4)) {
    .....................................................
        } else {
                return 0;
        }
    //以table_val为触发等级绑定发热设备和降温设备
        if (mtk_thermal_zone_bind_cooling_device(thermal, table_val, cdev)) {
                tscpu_warn("tscpu_bind error binding cooling dev\n");
                return -EINVAL;
        }
 
        tscpu_printk("tscpu_bind binding OK, %d\n", table_val);
        return 0;
}

温度采集设备知道了触发等级和降温温度，还需要知道触发等级对应的温度。thermal_zone_device_ops的get_trip_temp()用于查询触发等级对应的温度，以mtkcpu为例，所有降温设备的触发温度保存在数据中，触发等级就是该数组的下标。

static int tscpu_get_trip_temp
(struct thermal_zone_device *thermal, int trip, int *temp)
{
        *temp = trip_temp[trip];
        return 0;
}

2.1.3 cooling device

2.1.3.1 cpu coolig

cpufreq_state2power：根据cpu cooling state换算cpu power。

static int cpufreq_state2power(struct thermal_cooling_device *cdev,
                               struct thermal_zone_device *tz,
                               unsigned long state, u32 *power)
{
        unsigned int freq, num_cpus;
        cpumask_t cpumask;
        u32 static_power, dynamic_power;
        int ret;
        struct cpufreq_cooling_device *cpufreq_device = cdev->devdata;

        cpumask_and(&cpumask, &cpufreq_device->allowed_cpus, cpu_online_mask);
        // 根据cpumask得到在线cpu核数
        num_cpus = cpumask_weight(&cpumask);

        /* None of our cpus are online, so no power */
        if (num_cpus == 0) {
                *power = 0;
                return 0;
        }
        // 根据cpu state得到当前的频率
        freq = cpufreq_device->freq_table[state];
        if (!freq)
                return -EINVAL;
        // 计算当前频率下的cpu动态功耗
        dynamic_power = cpu_freq_to_power(cpufreq_device, freq) * num_cpus;
        // 计算当前频率下的cpu静态功耗
        ret = get_static_power(cpufreq_device, tz, freq, &static_power);
        if (ret)
                return ret;
        // 计算当前频率下的cpu总的功耗
        *power = static_power + dynamic_power;
        return 0;
}

cpufreq_power2state：根据cpu power换算cpu cooling state.

static int cpufreq_power2state(struct thermal_cooling_device *cdev,
                               struct thermal_zone_device *tz, u32 power,
                               unsigned long *state)
{
        unsigned int cpu, cur_freq, target_freq;
        int ret;
        s32 dyn_power;
        u32 last_load, normalised_power, static_power;
        struct cpufreq_cooling_device *cpufreq_device = cdev->devdata;

        cpu = cpumask_any_and(&cpufreq_device->allowed_cpus, cpu_online_mask);

        /* None of our cpus are online */
        if (cpu >= nr_cpu_ids)
                return -ENODEV;
        // 计算当前cpu频率
        cur_freq = cpufreq_quick_get(cpu);
        // 计算当前频率下的静态功耗
        ret = get_static_power(cpufreq_device, tz, cur_freq, &static_power);
        if (ret)
                return ret;
        // 计算当前频率下的动态功耗
        dyn_power = power - static_power;
        dyn_power = dyn_power > 0 ? dyn_power : 0;
        last_load = cpufreq_device->last_load ?: 1;
        // 计算归一化功耗
        normalised_power = (dyn_power * 100) / last_load;
        // 根据归一化功耗计算出目标频率
        target_freq = cpu_power_to_freq(cpufreq_device, normalised_power);
        // 根据目标频率得到cpu state
        *state = cpufreq_cooling_get_level(cpu, target_freq);
        if (*state == THERMAL_CSTATE_INVALID) {
                dev_warn_ratelimited(&cdev->device,
                                     "Failed to convert %dKHz for cpu %d into a cdev state\n",
                                     target_freq, cpu);
                return -EINVAL;
        }

        trace_thermal_power_cpu_limit(&cpufreq_device->allowed_cpus,
                                      target_freq, *state, power);
        return 0;
}

static u32 cpu_freq_to_power(struct cpufreq_cooling_device *cpufreq_cdev,
                             u32 freq)
{
        int i;

        for (i = cpufreq_cdev->max_level - 1; i >= 0; i--) {
                if (freq > cpufreq_cdev->em->table[i].frequency)
                        break;
        }
        // 查表获取
        return cpufreq_cdev->em->table[i + 1].power;
}

遍历了一下cpufreq_cdev里的em->table，这个table蕴含了freq和power的对应关系，这个table是跟芯片密切相关的，往往在出厂的时候厂商就已经预制好了。

2.1.4 thermal governor 实现原理

Thermal Governor：主要包括gov_bang_bang、gov_fair_share、gov_power_allocator、gov_step_wise、gov_user_space等。目前最常用的为gov_power_allocator。

IPA核心是利用PID控制器，Thermal Zone的温度作为输入，可分配功耗值作为输出，调节Allocator的频率和电压值。PID控制器在Sustainable Power基础上，根据当前温度和Control Temp之间的差值，来调节可分配功耗值的大小，进而调节Cooling设备的状态，也即调整OPP（Voltage和Frequency组合）。PID控制器的参数P、I、D也存在一定的经验值。需要测试几组不同参数，然后看温度控制效果。

Sustainable Power：是在不同OPP情境下，某一个最大OPP的温度保持基本稳定。比其大者，温度上升明显；比其小者温度保持不变或者下降。这可以通过监测不同OPP对应的温度值，得到一个Sustainable Power。

control_temp：当温度高于control_temp温度，IPA就会启动，进行OPP调节

dynamic-power-coefficient：动态功耗系数

IPA模型作用：

1）根据当前温度和控制温度差值计算出功耗，然后基于PID控制器调节OPP

2）根据当前环境预估下一个场景功耗值，包括动态功耗和静态功耗，动态功耗=c*v*v*f，静态功耗=v*F(temp)

，需要根据实测得到的数据，进行分析得到最吻合数据的一组算式。

2.4.1.1 IPA实现流程

1）IPA工作流程

实际的控制流程其实分成几部分：

1.各个器件（thermal cooling device）根据自己的performance状态计算自己需求的状态。

2.IPA governor收集各个器件的状态信息，同时收集器件的温度Tdie、Tskin等。

3.IPA governor根据器件的状态信息和温度信息，同时根据前刻的状态信息，计算出此时能够赋给各个cooling device的状态信息

4.各个cooling device得到自己的状态信息，根据governor要求调整自己的状态。

2）实际governor的工作原理

3）ARM文档手册里的图来说明功耗是如何在请求和实际需求之间分配

当需求小于给予时，大家总是能够相安无事的。矛盾总之发生在需求和给予不能相互满足的时候，即requested_power > granted_power时，我们这里可以看到各个actor的request_power，little和big的granted_power是小于其request_power的，不够分了！

这是我们把目光投到了GPU身上，大哥这个时候不用power！有多的拿来分，那可以的，于是多余的granted_power就被按需分配到了big和Little核心上。

每个actor分多少呢？那就要看他们的需求超过了多少，需求的越多，最后拿到的也就越多。

具体代码实现可参考divvy_up_power函数。

4）IPA代码流程

代码上主要由三部分组成：

1.thermal_core中主要执行Thermal_governor的轮询逻辑，即去轮询thermal_zone的温度，看看是否需要进行温控措施；

2.gov_power_allocator中主要执行thermal_governor的总体计算过程，即统计各个部件中的状态信息，再计算这个时候可以分配的功耗，最终将实际可分配的功耗计算出来下发给设备；

3.各个cooling_device设备中主要执行的是跟自身设备有关的计算逻辑，主要是计算降温设备所需的功耗、状态到功耗（state2power）、功耗到状态（power2state）；

5）IPA代码详解

核心结构体，

/**
 * struct thermal_trip - representation of a point in temperature domain
 * @temperature: temperature value in miliCelsius
 * @hysteresis: relative hysteresis in miliCelsius
 * @type: trip point type
 * @priv: pointer to driver data associated with this trip
 */
struct thermal_trip {
        int temperature;
        int hysteresis;
        enum thermal_trip_type type;
        void *priv;
};

/**
 * struct power_allocator_params - parameters for the power allocator governor
 * @allocated_tzp:        whether we have allocated tzp for this thermal zone and
 *                        it needs to be freed on unbind
 * @err_integral:        accumulated error in the PID controller.
 * @prev_err:        error in the previous iteration of the PID controller.
 *                Used to calculate the derivative term.
 * @sustainable_power:        Sustainable power (heat) that this thermal zone can
 *                        dissipate
 * @trip_switch_on:        first passive trip point of the thermal zone.  The
 *                        governor switches on when this trip point is crossed.
 *                        If the thermal zone only has one passive trip point,
 *                        @trip_switch_on should be NULL.
 * @trip_max_desired_temperature:        last passive trip point of the thermal
 *                                        zone.  The temperature we are
 *                                        controlling for.
 */
struct power_allocator_params {
        bool allocated_tzp;
        s64 err_integral;
        s32 prev_err;
        u32 sustainable_power;    // PID计算得到的power
        const struct thermal_trip *trip_switch_on;    // 触发温度
        const struct thermal_trip *trip_max_desired_temperature;    // 控制温度
};

power_allocator_throttle：IPA启动的入口函数

static int power_allocator_throttle(struct thermal_zone_device *tz,
                                    const struct thermal_trip *trip)
{
        struct power_allocator_params *params = tz->governor_data;
        bool update;

        lockdep_assert_held(&tz->lock);

        /*
         * We get called for every trip point but we only need to do
         * our calculations once
         */
        if (trip != params->trip_max_desired_temperature)
                return 0;

        trip = params->trip_switch_on;
        // 如果设备温度小于触发温度，直接return
        if (trip && tz->temperature < trip->temperature) {
                update = tz->last_temperature >= trip->temperature;
                tz->passive = 0;
                reset_pid_controller(params);
                allow_maximum_power(tz, update);
                return 0;
        }

        tz->passive = 1;
        // 如果设备温度大于触发温度（如35℃时开启控制，控制可以缓慢介入，期望控制到温度为50℃时，
        // 保持设备的温度不要继续上升），根据控制温度分配power
        return allocate_power(tz, params->trip_max_desired_temperature->temperature);
}
// allocate_power是IPA控制的核心函数，包括进行功耗的计算、功耗的限制要求以及功耗的重新分配
static int allocate_power(struct thermal_zone_device *tz, int control_temp)
{
        struct thermal_instance *instance;
        struct power_allocator_params *params = tz->governor_data;
        const struct thermal_trip *trip_max_desired_temperature =
                                        params->trip_max_desired_temperature;
        /* req_power: 各个actor根据自己的状态计算出的实际工作状态的值 */
        /* max_power: 各个actor的最大状态能够使用的power值 */
        /* granted_power：各个actor在当前状态下能够分配到的功耗值 */
        /* extra_actor_power：各个actor如果没有用完granted_power，剩下来可以分配给其他actor的power*/
        /* weighted_req_power: 各个cooling device有分配功耗的权重，经过权重计算后的功耗*/
        /* power_range: 经过PID控制后得到的power值 */                    
        u32 *req_power, *max_power, *granted_power, *extra_actor_power;
        u32 *weighted_req_power;
        u32 total_req_power, max_allocatable_power, total_weighted_req_power;
        u32 total_granted_power, power_range;
        int i, num_actors, total_weight, ret = 0;

        num_actors = 0;
        total_weight = 0;
        // 1. 遍历所有的cooling device，判断是否具备电源控制功能，统计num_actors、total_weight
        list_for_each_entry(instance, &tz->thermal_instances, tz_node) {
                if ((instance->trip == trip_max_desired_temperature) &&
                    // 判断cooling device是否具备电源控制功能
                    cdev_is_power_actor(instance->cdev)) {
                        num_actors++;
                        total_weight += instance->weight;
                }
        }

        if (!num_actors)
                return -ENODEV;

        /*
         * We need to allocate five arrays of the same size:
         * req_power, max_power, granted_power, extra_actor_power and
         * weighted_req_power.  They are going to be needed until this
         * function returns.  Allocate them all in one go to simplify
         * the allocation and deallocation logic.
         */
        BUILD_BUG_ON(sizeof(*req_power) != sizeof(*max_power));
        BUILD_BUG_ON(sizeof(*req_power) != sizeof(*granted_power));
        BUILD_BUG_ON(sizeof(*req_power) != sizeof(*extra_actor_power));
        BUILD_BUG_ON(sizeof(*req_power) != sizeof(*weighted_req_power));
        req_power = kcalloc(num_actors * 5, sizeof(*req_power), GFP_KERNEL);
        if (!req_power)
                return -ENOMEM;

        max_power = &req_power[num_actors];
        granted_power = &req_power[2 * num_actors];
        extra_actor_power = &req_power[3 * num_actors];
        weighted_req_power = &req_power[4 * num_actors];

        i = 0;
        total_weighted_req_power = 0;
        total_req_power = 0;
        max_allocatable_power = 0;

        list_for_each_entry(instance, &tz->thermal_instances, tz_node) {
                int weight;
                struct thermal_cooling_device *cdev = instance->cdev;

                if (instance->trip != trip_max_desired_temperature)
                        continue;

                if (!cdev_is_power_actor(cdev))
                        continue;
                // 获取所需power，查表获得
                if (cdev->ops->get_requested_power(cdev, &req_power[i]))
                        continue;

                if (!total_weight)
                        weight = 1 << FRAC_BITS;
                else
                        weight = instance->weight;

                weighted_req_power[i] = frac_to_int(weight * req_power[i]);
                // cooling device的state对应的power
                if (cdev->ops->state2power(cdev, instance->lower,
                                           &max_power[i]))
                        continue;

                total_req_power += req_power[i];
                max_allocatable_power += max_power[i];
                total_weighted_req_power += weighted_req_power[i];

                i++;
        }
        // 计算PID后的power
        power_range = pid_controller(tz, control_temp, max_allocatable_power);
        // 按需分配功耗
        divvy_up_power(weighted_req_power, max_power, num_actors,
                       total_weighted_req_power, power_range, granted_power,
                       extra_actor_power);

        total_granted_power = 0;
        i = 0;
        list_for_each_entry(instance, &tz->thermal_instances, tz_node) {
                if (instance->trip != trip_max_desired_temperature)
                        continue;

                if (!cdev_is_power_actor(instance->cdev))
                        continue;
                /* 这里其实就是把已经计算好的power分配到各个actor上，由各个actor来完成功耗
                * 到状态的转换工作
                */
                power_actor_set_power(instance->cdev, instance,
                                      granted_power[i]);
                total_granted_power += granted_power[i];

                i++;
        }

        trace_thermal_power_allocator(tz, req_power, total_req_power,
                                      granted_power, total_granted_power,
                                      num_actors, power_range,
                                      max_allocatable_power, tz->temperature,
                                      control_temp - tz->temperature);

        kfree(req_power);

        return ret;
}

static struct thermal_governor thermal_gov_power_allocator = {
        .name                = "power_allocator",
        .bind_to_tz        = power_allocator_bind,
        .unbind_from_tz        = power_allocator_unbind,
        .throttle        = power_allocator_throttle,
};
THERMAL_GOVERNOR_DECLARE(thermal_gov_power_allocator);

pid_controller：计算出PID参数、sustainable_power、power_range ，返回PID后的power，即power_range

/**
 * pid_controller() - PID controller
 * @tz:        thermal zone we are operating in
 * @control_temp:        the target temperature in millicelsius
 * @max_allocatable_power:        maximum allocatable power for this thermal zone
 *
 * This PID controller increases the available power budget so that the
 * temperature of the thermal zone gets as close as possible to
 * @control_temp and limits the power if it exceeds it.  k_po is the
 * proportional term when we are overshooting, k_pu is the
 * proportional term when we are undershooting.  integral_cutoff is a
 * threshold below which we stop accumulating the error.  The
 * accumulated error is only valid if the requested power will make
 * the system warmer.  If the system is mostly idle, there's no point
 * in accumulating positive error.
 *
 * Return: The power budget for the next period.
 */
static u32 pid_controller(struct thermal_zone_device *tz,
                          int control_temp,
                          u32 max_allocatable_power)
{
        s64 p, i, d, power_range;
        s32 err, max_power_frac;
        u32 sustainable_power;
        struct power_allocator_params *params = tz->governor_data;

        max_power_frac = int_to_frac(max_allocatable_power);
        // 根据控制温度，计算sustainable_power
        sustainable_power = get_sustainable_power(tz, params, control_temp);
        // 计算控制温度和实际温度的误差
        err = control_temp - tz->temperature;
        err = int_to_frac(err);

        /* Calculate the proportional term */
        // 计算出差分项的power
        p = mul_frac(err < 0 ? tz->tzp->k_po : tz->tzp->k_pu, err);

        /*
         * Calculate the integral term
         *
         * if the error is less than cut off allow integration (but
         * the integral is limited to max power)
         */
         // 计算出积分项的power
        i = mul_frac(tz->tzp->k_i, params->err_integral);

        if (err < int_to_frac(tz->tzp->integral_cutoff)) {
                s64 i_next = i + mul_frac(tz->tzp->k_i, err);

                if (abs(i_next) < max_power_frac) {
                        i = i_next;
                        params->err_integral += err;
                }
        }

        /*
         * Calculate the derivative term
         *
         * We do err - prev_err, so with a positive k_d, a decreasing
         * error (i.e. driving closer to the line) results in less
         * power being applied, slowing down the controller)
         */
         // 计算出微分项的power
        d = mul_frac(tz->tzp->k_d, err - params->prev_err);
        d = div_frac(d, jiffies_to_msecs(tz->passive_delay_jiffies));
        params->prev_err = err;
        // 累计算出PID后的power
        power_range = p + i + d;

        /* feed-forward the known sustainable dissipatable power */
        power_range = sustainable_power + frac_to_int(power_range);
        // 如果该值小于最小值，则返回最小值，如果该值大于最大值，则返回最大值，否则返回该值本身
        power_range = clamp(power_range, (s64)0, (s64)max_allocatable_power);

        trace_thermal_power_allocator_pid(tz, frac_to_int(err),
                                          frac_to_int(params->err_integral),
                                          frac_to_int(p), frac_to_int(i),
                                          frac_to_int(d), power_range);

        return power_range;
}

divvy_up_power：按需分配power

/**
 * divvy_up_power() - divvy the allocated power between the actors
 * @req_power:        each actor's requested power
 * @max_power:        each actor's maximum available power
 * @num_actors:        size of the @req_power, @max_power and @granted_power's array
 * @total_req_power: sum of @req_power
 * @power_range:        total allocated power
 * @granted_power:        output array: each actor's granted power
 * @extra_actor_power:        an appropriately sized array to be used in the
 *                        function as temporary storage of the extra power given
 *                        to the actors
 *
 * This function divides the total allocated power (@power_range)
 * fairly between the actors.  It first tries to give each actor a
 * share of the @power_range according to how much power it requested
 * compared to the rest of the actors.  For example, if only one actor
 * requests power, then it receives all the @power_range.  If
 * three actors each requests 1mW, each receives a third of the
 * @power_range.
 *
 * If any actor received more than their maximum power, then that
 * surplus is re-divvied among the actors based on how far they are
 * from their respective maximums.
 *
 * Granted power for each actor is written to @granted_power, which
 * should've been allocated by the calling function.
 */
static void divvy_up_power(u32 *req_power, u32 *max_power, int num_actors,
                           u32 total_req_power, u32 power_range,
                           u32 *granted_power, u32 *extra_actor_power)
{
        u32 extra_power, capped_extra_power;
        int i;

        /*
         * Prevent division by 0 if none of the actors request power.
         */
        if (!total_req_power)
                total_req_power = 1;

        capped_extra_power = 0;
        extra_power = 0;
        for (i = 0; i < num_actors; i++) {
                u64 req_range = (u64)req_power[i] * power_range;

                granted_power[i] = DIV_ROUND_CLOSEST_ULL(req_range,
                                                         total_req_power);

                if (granted_power[i] > max_power[i]) {
                        extra_power += granted_power[i] - max_power[i];
                        granted_power[i] = max_power[i];
                }

                extra_actor_power[i] = max_power[i] - granted_power[i];
                capped_extra_power += extra_actor_power[i];
        }

        if (!extra_power)
                return;

        /*
         * Re-divvy the reclaimed extra among actors based on
         * how far they are from the max
         */
         // 如果任何actor收到的功率超过了他们的最大功率，
         // 那么多余的部分将根据他们距离最大值的程度重新分配
        extra_power = min(extra_power, capped_extra_power);
        if (capped_extra_power > 0)
                for (i = 0; i < num_actors; i++) {
                        u64 extra_range = (u64)extra_actor_power[i] * extra_power;
                        granted_power[i] += DIV_ROUND_CLOSEST_ULL(extra_range,
                                                         capped_extra_power);
                }
}

power_actor_set_power：把已经计算好的power分配到各个actor上，由各个actor来完成功耗到状态的转换工作

/**
 * power_actor_set_power() - limit the maximum power a cooling device consumes
 * @cdev:        pointer to &thermal_cooling_device
 * @instance:        thermal instance to update
 * @power:        the power in milliwatts
 *
 * Set the cooling device to consume at most @power milliwatts. The limit is
 * expected to be a cap at the maximum power consumption.
 *
 * Return: 0 on success, -EINVAL if the cooling device does not
 * implement the power actor API or -E* for other failures.
 */
static int power_actor_set_power(struct thermal_cooling_device *cdev,
                      struct thermal_instance *instance, u32 power)
{
        unsigned long state;
        int ret;
        // power转对应的state
        ret = cdev->ops->power2state(cdev, power, &state);
        if (ret)
                return ret;

        instance->target = clamp_val(state, instance->lower, instance->upper);
        mutex_lock(&cdev->lock);
        __thermal_cdev_update(cdev);
        mutex_unlock(&cdev->lock);

        return 0;
}

Youth cowboy

关注

21
点赞
踩
27

收藏

觉得还不错? 一键收藏
1
评论
Linux Thermal框架

Linux的Thermal机制是基于Zone为单位的热管理机制，核心包括三个部分：获取区域温度的设备thermal_zone_device、区域降温的设备thermal_cooling_device、温控策略thermal_governor。thermal_governor从thermal_zone_device获取区域温度，然后根据当前温度，决定调用哪个降温设备来为该区域降温。
复制链接

扫一扫