Linux Runtime PM介绍【转】

2022-01-11 02:46:15

转自：http://blog.csdn.net/wlwl0071986/article/details/42677403

一、Runtime PM引言

1. 背景

（1）display的需求

（2）系统整机动态功耗优化的需求

（3）upstream

2. 解决方案

（1）引入debounce

（2）使用统一的workqueue来管理任务

（3）实时地关闭不需要工作的device

（4）当device作为parent时，所有的child不工作时，关闭该device

（5）引入pm_rutime

3. 性能指标

（1）快速开关屏场景，亮屏速度提升

（2）动态功耗，更为稳定；唤醒而不亮屏的场景，功耗更低

（3）有助于降低系统整机动态功耗

二、Runtime PM框架

1. Runtime PM层次结构

2. Runtime PM状态

3. Runtime PM控制流程

每个设备或者子系统都会向Runtime PM core注册3个callback。

在struct dev_pm_ops结构体中，定义了这三个callback：
struct dev_pm_ops {
...
int (*runtime_suspend)(struct device *dev);
int (*runtime_resume)(struct device *dev);
int (*runtime_idle)(struct device *dev);
.suspend
.resume
};
注：引入runtime之后，suspend 接口需和runtime接口放在同一个数据结构内；

4. rpm_status（include\linux\pm.h）

enum rpm_status {
RPM_ACTIVE = 0, /* 表示runtime_resume()被成功执行 */
RPM_RESUMING, /* 表示runtime_resume()正在被执行 */
RPM_SUSPENDED, /* 表示runtime_suspend()被成功执行 */
RPM_SUSPENDING, /* 表示runtime_suspend()正在被执行 */
};

5. rpm_request（include\linux\pm.h）

enum rpm_request {
RPM_REQ_NONE = 0,RPM_REQ_IDLE, /* 执行runtime_idle() */

RPM_REQ_SUSPEND, /* 执行runtime_suspend () */
RPM_REQ_AUTOSUSPEND, /* 延迟autosuspend_delay后执行runtime_suspend() */
RPM_REQ_RESUME, /* 执行runtime_resume() */
};
请求的类型（设置request_pending才有效）。

6. Idle Reference API

pm_runtime_put_noidle: only put
pm_runtime_idle
pm_request_idle：async
pm_runtime_put: put + async
pm_runtime_put_sync

7. Suspend Reference API

pm_schedule_suspend
pm_runtime_suspend:
pm_runtime_put_sync_suspend: put
pm_runtime_autosuspend: auto
pm_request_autosuspend：async + auto
pm_runtime_put_autosuspend: put + async + auto
pm_runtime_put_sync_autosuspend: put + auto

8. Resume Reference API

pm_runtime_get_noresume
pm_runtime_resume
pm_request_resume: async
pm_runtime_get: get + async
pm_runtime_get_sync: get

9. Device Runtime suspend 流程

触发：
pm_schedule_suspend
pm_runtime_suspend:
pm_runtime_put_sync_suspend: put
pm_runtime_autosuspend: auto
pm_request_autosuspend：async + auto
pm_runtime_put_autosuspend: put + async + auto
pm_runtime_put_sync_autosuspend: put + auto
suspend的条件：
usage_cnt= 0;
disable_depth = 0;
runtime_status = RPM_ACTIVE

10. Device Runtime PM idle流程

idle的触发：
pm_runtime_put_noidle: only put
pm_runtime_idle
pm_request_idle：async
pm_runtime_put: put + async
pm_runtime_put_sync
idle的条件：
usage_cnt= 0;
disable_depth = 0;
runtime_status = RPM_ACTIVE

11. Device Runtime PM resume流程

触发：
pm_runtime_get_noresume
pm_runtime_resume
pm_request_resume: async
pm_runtime_get: get + async
pm_runtime_get_sync: get
条件：
disable_depth = 0
runtime_status != RPM_ACTIVE

三、Runtime PM和设备模型

1. Device Runtime PM 初始化

（1）设备模型完成pm_runtime 的初始化；

（2）对pm_runtime 状态的改变，需在device_register 之后实现；

（3）设备模型，在调用drv probe, remove, shutdown 等接口时，

可以保证pm_runtime 处于disable 状态，或其他约定状态；

void pm_runtime_init(struct device *dev)
{
dev->power.runtime_status = RPM_SUSPENDED;
dev->power.idle_notification = false;
dev->power.disable_depth = 1;
atomic_set(&dev->power.usage_count, 0);
dev->power.runtime_error = 0;
atomic_set(&dev->power.child_count, 0);
pm_suspend_ignore_children(dev, false);
dev->power.runtime_auto = true;
dev->power.request_pending = false;
dev->power.request = RPM_REQ_NONE;
dev->power.deferred_resume = false;
dev->power.accounting_timestamp = jiffies;
INIT_WORK(&dev->power.work, pm_runtime_work);
dev->power.timer_expires = 0;
setup_timer(&dev->power.suspend_timer, pm_suspend_timer_fn, (unsigned long)dev);
init_waitqueue_head(&dev->power.wait_queue);
}

platform_device_register(&platform_disp_device);
pm_runtime_set_active(&platform_disp_device.dev);
pm_runtime_get_noresume(&platform_disp_device.dev);
pm_runtime_enable(&platform_disp_device.dev);
pm_runtime_set_autosuspend_delay(&platform_disp_device.dev, 5000);
pm_runtime_use_autosuspend(&platform_disp_device.dev);
platform_driver_register(&disp_driver);

2. Device Runtime PM remove流程