SuperYang`s Blog

Coding Everything.

RunLoop深度探究(二)

RunLoop 的内部逻辑

根据苹果在文档里的说明,RunLoop 内部的逻辑大致如下: 内部逻辑

内部代码整理为:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
/// 用DefaultMode启动
void CFRunLoopRun(void) {
    CFRunLoopRunSpecific(CFRunLoopGetCurrent(), kCFRunLoopDefaultMode, 1.0e10, false);
}

/// 用指定的Mode启动,允许设置RunLoop超时时间
int CFRunLoopRunInMode(CFStringRef modeName, CFTimeInterval seconds, Boolean stopAfterHandle) {
    return CFRunLoopRunSpecific(CFRunLoopGetCurrent(), modeName, seconds, returnAfterSourceHandled);
}

/// RunLoop的实现
int CFRunLoopRunSpecific(runloop, modeName, seconds, stopAfterHandle) {

    /// 首先根据modeName找到对应mode
    CFRunLoopModeRef currentMode = __CFRunLoopFindMode(runloop, modeName, false);
    /// 如果mode里没有source/timer/observer, 直接返回。
    if (__CFRunLoopModeIsEmpty(currentMode)) return;

    /// 1. 通知 Observers: RunLoop 即将进入 loop。
    __CFRunLoopDoObservers(runloop, currentMode, kCFRunLoopEntry);

    /// 内部函数,进入loop
    __CFRunLoopRun(runloop, currentMode, seconds, returnAfterSourceHandled) {

        Boolean sourceHandledThisLoop = NO;
        int retVal = 0;
        do {

            /// 2. 通知 Observers: RunLoop 即将触发 Timer 回调。
            __CFRunLoopDoObservers(runloop, currentMode, kCFRunLoopBeforeTimers);
            /// 3. 通知 Observers: RunLoop 即将触发 Source0 (非port) 回调。
            __CFRunLoopDoObservers(runloop, currentMode, kCFRunLoopBeforeSources);
            /// 执行被加入的block
            __CFRunLoopDoBlocks(runloop, currentMode);

            /// 4. RunLoop 触发 Source0 (非port) 回调。
            sourceHandledThisLoop = __CFRunLoopDoSources0(runloop, currentMode, stopAfterHandle);
            /// 执行被加入的block
            __CFRunLoopDoBlocks(runloop, currentMode);

            /// 5. 如果有 Source1 (基于port) 处于 ready 状态,直接处理这个 Source1 然后跳转去处理消息。
            if (__Source0DidDispatchPortLastTime) {
                Boolean hasMsg = __CFRunLoopServiceMachPort(dispatchPort, &msg)
                if (hasMsg) goto handle_msg;
            }

            /// 通知 Observers: RunLoop 的线程即将进入休眠(sleep)。
            if (!sourceHandledThisLoop) {
                __CFRunLoopDoObservers(runloop, currentMode, kCFRunLoopBeforeWaiting);
            }

            /// 7. 调用 mach_msg 等待接受 mach_port 的消息。线程将进入休眠, 直到被下面某一个事件唤醒。
            /// • 一个基于 port 的Source 的事件。
            /// • 一个 Timer 到时间了
            /// • RunLoop 自身的超时时间到了
            /// • 被其他什么调用者手动唤醒
            __CFRunLoopServiceMachPort(waitSet, &msg, sizeof(msg_buffer), &livePort) {
                mach_msg(msg, MACH_RCV_MSG, port); // thread wait for receive msg
            }

            /// 8. 通知 Observers: RunLoop 的线程刚刚被唤醒了。
            __CFRunLoopDoObservers(runloop, currentMode, kCFRunLoopAfterWaiting);

            /// 收到消息,处理消息。
            handle_msg:

            /// 9.1 如果一个 Timer 到时间了,触发这个Timer的回调。
            if (msg_is_timer) {
                __CFRunLoopDoTimers(runloop, currentMode, mach_absolute_time())
            }

            /// 9.2 如果有dispatch到main_queue的block,执行block。
            else if (msg_is_dispatch) {
                __CFRUNLOOP_IS_SERVICING_THE_MAIN_DISPATCH_QUEUE__(msg);
            }

            /// 9.3 如果一个 Source1 (基于port) 发出事件了,处理这个事件
            else {
                CFRunLoopSourceRef source1 = __CFRunLoopModeFindSourceForMachPort(runloop, currentMode, livePort);
                sourceHandledThisLoop = __CFRunLoopDoSource1(runloop, currentMode, source1, msg);
                if (sourceHandledThisLoop) {
                    mach_msg(reply, MACH_SEND_MSG, reply);
                }
            }

            /// 执行加入到Loop的block
            __CFRunLoopDoBlocks(runloop, currentMode);


            if (sourceHandledThisLoop && stopAfterHandle) {
                /// 进入loop时参数说处理完事件就返回。
                retVal = kCFRunLoopRunHandledSource;
            } else if (timeout) {
                /// 超出传入参数标记的超时时间了
                retVal = kCFRunLoopRunTimedOut;
            } else if (__CFRunLoopIsStopped(runloop)) {
                /// 被外部调用者强制停止了
                retVal = kCFRunLoopRunStopped;
            } else if (__CFRunLoopModeIsEmpty(runloop, currentMode)) {
                /// source/timer/observer一个都没有了
                retVal = kCFRunLoopRunFinished;
            }

            /// 如果没超时,mode里没空,loop也没被停止,那继续loop。
        } while (retVal == 0);
    }

    /// 10. 通知 Observers: RunLoop 即将退出。
    __CFRunLoopDoObservers(rl, currentMode, kCFRunLoopExit);
}

可以看到,实际上 RunLoop 就是这样一个函数,其内部是一个 do-while 循环。当你调用 CFRunLoopRun() 时,线程就会一直停留在这个循环里;直到超时或被手动停止,该函数才会返回

RunLoop 的底层实现

从上面代码可以看到,RunLoop 的核心是基于 mach port 的,其进入休眠时调用的函数是 mach_msg()。为了解释这个逻辑,下面稍微介绍一下 OSX/iOS 的系统架构。

底层逻辑

苹果官方将整个系统大致划分为上述4个层次: 应用层包括用户能接触到的图形应用,例如 Spotlight、Aqua、SpringBoard 等。 应用框架层即开发人员接触到的 Cocoa 等框架。 核心框架层包括各种核心框架、OpenGL 等内容。 Darwin 即操作系统的核心,包括系统内核、驱动、Shell 等内容,这一层是开源的,其所有源码都可以在 opensource.apple.com 里找到。

我们在深入看一下 Darwin 这个核心的架构 Darwin 其中,在硬件层上面的三个组成部分:Mach、BSD、IOKit (还包括一些上面没标注的内容),共同组成了 XNU 内核。 XNU 内核的内环被称作 Mach,其作为一个微内核,仅提供了诸如处理器调度、IPC (进程间通信)等非常少量的基础服务。 BSD 层可以看作围绕 Mach 层的一个外环,其提供了诸如进程管理、文件系统和网络等功能。 IOKit 层是为设备驱动提供了一个面向对象(C++)的一个框架。

Mach 本身提供的 API 非常有限,而且苹果也不鼓励使用 Mach 的 API,但是这些API非常基础,如果没有这些API的话,其他任何工作都无法实施。在 Mach 中,所有的东西都是通过自己的对象实现的,进程、线程和虚拟内存都被称为"对象"。和其他架构不同, Mach 的对象间不能直接调用,只能通过消息传递的方式实现对象间的通信。"消息"是 Mach 中最基础的概念,消息在两个端口 (port) 之间传递,这就是 Mach 的 IPC (进程间通信) 的核心。

Mach 的消息定义是在 <mach/message.h> 头文件的,很简单:

1
2
3
4
5
6
7
8
9
10
11
12
13
typedef struct {
  mach_msg_header_t header;
  mach_msg_body_t body;
} mach_msg_base_t;

typedef struct {
  mach_msg_bits_t msgh_bits;
  mach_msg_size_t msgh_size;
  mach_port_t msgh_remote_port;
  mach_port_t msgh_local_port;
  mach_port_name_t msgh_voucher_port;
  mach_msg_id_t msgh_id;
} mach_msg_header_t;

一条 Mach 消息实际上就是一个二进制数据包 (BLOB),其头部定义了当前端口 local_port 和目标端口 remote_port, 发送和接受消息是通过同一个 API 进行的,其 option 标记了消息传递的方向:

1
2
3
4
5
6
7
8
mach_msg_return_t mach_msg(
          mach_msg_header_t *msg,
          mach_msg_option_t option,
          mach_msg_size_t send_size,
          mach_msg_size_t rcv_size,
          mach_port_name_t rcv_name,
          mach_msg_timeout_t timeout,
          mach_port_name_t notify);

为了实现消息的发送和接收,mach_msg() 函数实际上是调用了一个 Mach 陷阱 (trap),即函数mach_msg_trap(),陷阱这个概念在 Mach 中等同于系统调用。当你在用户态调用 mach_msg_trap() 时会触发陷阱机制,切换到内核态;内核态中内核实现的 mach_msg() 函数会完成实际的工作,如下图: Darwin 这些概念可以参考维基百科:System_callTrap_(computing))

RunLoop 的核心就是一个 mach_msg() (见上面代码的第7步),RunLoop 调用这个函数去接收消息,如果没有别人发送 port 消息过来,内核会将线程置于等待状态。例如你在模拟器里跑起一个 iOS 的 App,然后在 App 静止时点击暂停,你会看到主线程调用栈是停留在 mach_msg_trap() 这个地方。

关于具体的如何利用 mach port 发送信息,可以看看 NSHipster 这一篇文章,或者这里的中文翻译 。

关于Mach的历史可以看看这篇很有趣的文章:Mac OS X 背后的故事(三)Mach 之父 Avie Tevanian

特别致谢:

http://blog.ibireme.com/ 2015/05/18/runloop/#more-41710

参考文章:

深入理解RunLoop(ibireme):

http://blog.ibireme.com/ 2015/05/18/runloop/#more-41710

Apple Document: https://developer.apple.com/library/ios/documentation/Cocoa/Conceptual/Multithreading/RunLoopManagement/RunLoopManagement.html

Apple–RunLoopManagement文档

opensource.apple.com

维基百科:

System_callTrap_(computing))

NSHipster:

inter-process- communication inter-process-communication中文

Mac OS X 背后的故事(三)Mach 之父 Avie Tevanian

Comments