Android 触摸事件分发机制(一)从内核到应用 一切的开始

写在前面

本文基于Android 7.1.1 (API 25)的源码分析编写

安卓是基于触摸操作进行交互的系统,几乎所有的操作都由对屏幕的一次次触摸完成,如何正确处理触摸事件关乎整个应用的操作体验。因此安卓对于触摸事件的分发与处理机制也是我们学习安卓开发的重中之重。同时几乎每一个安卓技术博客中都会对触摸分发机制这一块进行详解,例如比较早期也是最为出名的郭霖(《第一行代码》的作者)。现在网络上对于这一块的分析也已经比较详尽了,基本上一篇博客中遗漏的部分都可以在其他博客中找到答案。

但是无论别人的文章讲得多好,多么详细,我们都需要自己去打开源码仔细分析好好体会,一是这样一个比较复杂的过程不经历自己的思考很难完全理解,二是随着api版本的推进这部分源码也会发生很多变化,虽然大致思路相同,但是接触到新的内容总是一件好事。

这也就是我写这篇博文的原因:记录自己思考与分析的过程。

触摸事件的来源

这部分的内容与安卓本身无关,代码大部分也都是C++实现的,中间的大部分内容来自于我对相关资料的总结,不在代码层面进行详细解释,只是说明一个流程,同时也会对代码进行大部分的删减,只关注最核心的那部分。

从硬件到内核

我们从头开始,从触摸事件最初最初的来源开始,我们知道内核是以处理中断的方式处理用户的输入的,触摸事件作为一种特殊的输入事件,自然也需要这种方式进行处理,只不过触摸事件的提供的信息要稍微复杂一些。

触摸事件来自于我们对硬件的操作,最初的来源当然是硬件引起的中断。而处理特定中断的代码则来自于对应硬件的驱动:

图片来源(以下系列图片来源相同,不作标注)

当一个输入设备的驱动模块被首次载入内核的时候,会检测它应该管理的硬件设备,如果检测成功,驱动模块会调用include/linux/input.h中的input_register_device(…)函数设置一个/dev/input/eventXX为整数)来代表这个输入设备。驱动模块同时也会通过include/linux/interrupt.hrequest_irq(…)函数注册一个函数去处理这个硬件引发的中断,注册成功以后,当设备因用户交互而产生中断的时候就会交给对应的驱动模块进行处理。

驱动模块处理的细节各不相同,但最终都会将数据处理后存放进对应的/dev/input/eventX文件中。

系统对触摸事件的处理

现在驱动程序为我们收集好了原始的输入信息并存放在了eventX文件中,下一步就是系统对于这个文件的处理并发送到应用层面。

可以看到系统服务充当了从内核到应用的桥梁。

系统服务由三个组件构成:EventHubInputReaderInputDispatcher,关于它们的作用的详细分析在:

http://gityuan.com/2016/12/31/input-ipc/

下面对这个过程作简单介绍。

EventHub

文件在frameworks/native/services/inputflinger/EventHub.cpp

它的作用是监听、读取/dev/input目录下产生的新事件,并封装成RawEvent结构体供InputReader使用。

InputReader

文件在frameworks/native/services/inputflinger/InputReader.cpp

InputReader运行在一个单独的进程中,这个进程由InputManagerService的初始化而新建,具体内容请见:

http://gityuan.com/2016/12/10/input-manager/

它会在内部不断地循环调用loopOnce()方法来不断读取事件:

1
2
3
4
5
6
7
8
9
void InputReader::loopOnce() {
...
size_t count = mEventHub->getEvents(timeoutMillis, mEventBuffer, EVENT_BUFFER_SIZE);
...
if (count) {
processEventsLocked(mEventBuffer, count);
}
...
}

第3行调用了mEventHubgetEvent()方法以获取事件。

第6行调用processEventLocked()方法来处理事件,经过一系列判断之后,会执行这行代码:

1
device->process(rawEvents, count);

process函数会执行如下代码:

1
2
InputMapper* mapper = mMappers[i];
mapper->process(rawEvent);

使用mapper去处理rawEvent,不同的输入事件类型会由不同的mapper去处理,以处理触摸事件的TouchInputMapper为例:

只看核心调用的话,会依次调用如下函数:

1
2
3
4
5
6
7
8
9
10
11
void TouchInputMapper::process(const RawEvent* rawEvent)
void TouchInputMapper::sync(nsecs_t when)
void TouchInputMapper::processRawTouches(bool timeout)
void TouchInputMapper::cookAndDispatch(nsecs_t when)
void TouchInputMapper::dispatchTouches(nsecs_t when, uint32_t policyFlags)
void TouchInputMapper::dispatchMotion(nsecs_t when, uint32_t policyFlags, uint32_t source,
int32_t action, int32_t actionButton, int32_t flags,
int32_t metaState, int32_t buttonState, int32_t edgeFlags,
const PointerProperties* properties, const PointerCoords* coords,
const uint32_t* idToIndex, BitSet32 idBits, int32_t changedId,
float xPrecision, float yPrecision, nsecs_t downTime)

在最终的dispatchMotion()函数中执行以下代码:

1
2
3
4
5
NotifyMotionArgs args(when, getDeviceId(), source, policyFlags,
action, actionButton, flags, metaState, buttonState, edgeFlags,
mViewport.displayId, pointerCount, pointerProperties, pointerCoords,
xPrecision, yPrecision, downTime);
getListener()->notifyMotion(&args);

可以看到事件已经被处理成了一个args,然后调用getListener()

1
2
3
InputListenerInterface* InputReader::ContextImpl::getListener() {
return mReader->mQueuedListener.get();
}

获取的是mQueuedListener,查看notifyMotion()函数:

1
2
3
void QueuedInputListener::notifyMotion(const NotifyMotionArgs* args) {
mArgsQueue.push(new NotifyMotionArgs(*args));
}

这里的NotifyMotionArgs()只是对事件进行了一次再封装。可以看到这个args最终进入了QueuedInputListenermArgsQueue中。

我们再回到InputReaderloopOnce()函数中,函数在执行完上述调用到达最后一行时:

1
mQueuedListener->flush();

调用flush()函数:

1
2
3
4
5
6
7
8
9
void QueuedInputListener::flush() {
size_t count = mArgsQueue.size();
for (size_t i = 0; i < count; i++) {
NotifyArgs* args = mArgsQueue[i];
args->notify(mInnerListener);
delete args;
}
mArgsQueue.clear();
}

调用了各argsnotify()函数:

1
2
3
void NotifyMotionArgs::notify(const sp<InputListenerInterface>& listener) const {
listener->notifyMotion(this);
}

注意这里的listener传入的是mInnerListener,它是什么呢?

1
2
3
QueuedInputListener::QueuedInputListener(const sp<InputListenerInterface>& innerListener) :
mInnerListener(innerListener) {
}

在构造函数中初始化。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
InputReader::InputReader(const sp<EventHubInterface>& eventHub,
const sp<InputReaderPolicyInterface>& policy,
const sp<InputListenerInterface>& listener) :
mContext(this), mEventHub(eventHub), mPolicy(policy),
mGlobalMetaState(0), mGeneration(1),
mDisableVirtualKeysTimeout(LLONG_MIN), mNextTimeout(LLONG_MAX),
mConfigurationChangesToRefresh(0) {
mQueuedListener = new QueuedInputListener(listener);

{ // acquire lock
AutoMutex _l(mLock);

refreshConfigurationLocked(0);
updateGlobalMetaStateLocked();
} // release lock
}

InputReader构造函数中构造QueuedInputListener

InputReader是由InputManager类进行初始化的(线程的新建也在这个类中):

1
2
3
4
5
6
7
8
InputManager::InputManager(
const sp<EventHubInterface>& eventHub,
const sp<InputReaderPolicyInterface>& readerPolicy,
const sp<InputDispatcherPolicyInterface>& dispatcherPolicy) {
mDispatcher = new InputDispatcher(dispatcherPolicy);
mReader = new InputReader(eventHub, readerPolicy, mDispatcher);
initialize();
}

注意到第6行中,传入的listener正是mDispatcher也就是InputDispatcher对象。

所以说,listener->notifyMotion(this);调用的是InputDispatchernotifyMotion()函数,至此,InputReader的工作已经完成,它从EventHub中循环读取地rawEvent事件,并处理成args再通知InputDispatcher对事件进行进一步的分发处理。

InputDispatcher

我们直接来到InputDispatcher的源码,路径:frameworks/native/services/inputflinger/InputDispatcher.cpp

上面说到最终调用了InputDispatchernotifyMotion方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
MotionEvent event;
event.initialize(args->deviceId, args->source, args->action, args->actionButton,args->flags, args->edgeFlags, args->metaState, args->buttonState, 0, 0, args->xPrecision, args->yPrecision, args->downTime, args->eventTime, args->pointerCount, args->pointerProperties, args->pointerCoords);
...
// Just enqueue a new motion event.
MotionEntry* newEntry = new MotionEntry(args->eventTime,
args->deviceId, args->source, policyFlags,
args->action, args->actionButton, args->flags,
args->metaState, args->buttonState,
args->edgeFlags, args->xPrecision, args->yPrecision, args->downTime,
args->displayId,
args->pointerCount, args->pointerProperties, args->pointerCoords, 0, 0);

needWake = enqueueInboundEventLocked(newEntry);

里面新建并初始化了一个MotionEvent,然后包装成一个Entry,然后调用enqueueInboundEventLocked()函数:

1
mInboundQueue.enqueueAtTail(entry);

enqueueInboundEventLocked()函数中将这个entry插入到了mInboundQueue这个InputDispatcher维护的成员变量中。

到这里我们可以看到事件经过一系列的处理和传递以后最终作为一个entry插入到了InputDispatcher的队列中等待被进一步分发。

这个分发过程是在哪里进行的呢?

InputDispatcher线程的threadLoop()函数会被不断调用:

1
2
3
4
bool InputDispatcherThread::threadLoop() {
mDispatcher->dispatchOnce();
return true;
}

dispatcherOnce()中:

1
2
3
if (!haveCommandsLocked()) {
dispatchOnceInnerLocked(&nextWakeupTime);
}

在没有待执行的指令时执行dispatchOnceInnerLocked()函数:

1
mPendingEvent = mInboundQueue.dequeueAtHead();

这个函数中还包含了ANR的判断信息,关于ANR的部分之后再另开博文讲。

mInboundQueue不为空,则从中取出头部的pendingEvent

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
switch (mPendingEvent->type) {
case EventEntry::TYPE_CONFIGURATION_CHANGED: {
ConfigurationChangedEntry *typedEntry = static_cast<ConfigurationChangedEntry *>(mPendingEvent);
done = dispatchConfigurationChangedLocked(currentTime, typedEntry);
dropReason = DROP_REASON_NOT_DROPPED; // configuration changes are never dropped
break;
}

case EventEntry::TYPE_DEVICE_RESET: {
DeviceResetEntry *typedEntry = static_cast<DeviceResetEntry *>(mPendingEvent);
done = dispatchDeviceResetLocked(currentTime, typedEntry);
dropReason = DROP_REASON_NOT_DROPPED; // device resets are never dropped
break;
}

case EventEntry::TYPE_KEY: {
KeyEntry *typedEntry = static_cast<KeyEntry *>(mPendingEvent);
if (isAppSwitchDue) {
if (isAppSwitchKeyEventLocked(typedEntry)) {
resetPendingAppSwitchLocked(true);
isAppSwitchDue = false;
} else if (dropReason == DROP_REASON_NOT_DROPPED) {
dropReason = DROP_REASON_APP_SWITCH;
}
}
if (dropReason == DROP_REASON_NOT_DROPPED && isStaleEventLocked(currentTime, typedEntry)) {
dropReason = DROP_REASON_STALE;
}
if (dropReason == DROP_REASON_NOT_DROPPED && mNextUnblockedEvent) {
dropReason = DROP_REASON_BLOCKED;
}
done = dispatchKeyLocked(currentTime, typedEntry, &dropReason, nextWakeupTime);
break;
}

case EventEntry::TYPE_MOTION: {
MotionEntry *typedEntry = static_cast<MotionEntry *>(mPendingEvent);
if (dropReason == DROP_REASON_NOT_DROPPED && isAppSwitchDue) {
dropReason = DROP_REASON_APP_SWITCH;
}
if (dropReason == DROP_REASON_NOT_DROPPED && isStaleEventLocked(currentTime, typedEntry)) {
dropReason = DROP_REASON_STALE;
}
if (dropReason == DROP_REASON_NOT_DROPPED && mNextUnblockedEvent) {
dropReason = DROP_REASON_BLOCKED;
}
done = dispatchMotionLocked(currentTime, typedEntry, &dropReason, nextWakeupTime);
break;
}

default:
ALOG_ASSERT(false);
break;
}

下面对取出的mPendingEvent的类型进行判断,根据不同的类型信息把它转换回原来的Entry信息,然后调用相应的分发方法,我们还是顺着触摸事件分发这条路继续向下走,调用了bool InputDispatcher::dispatchMotionLocked()函数:

1
2
3
4
5
6
7
8
9
10
int32_t injectionResult;
if (isPointerEvent) {
// Pointer event. (eg. touchscreen)
injectionResult = findTouchedWindowTargetsLocked(currentTime,
entry, inputTargets, nextWakeupTime, &conflictingPointerActions);
} else {
// Non touch event. (eg. trackball)
injectionResult = findFocusedWindowTargetsLocked(currentTime,
entry, inputTargets, nextWakeupTime);
}

对触摸或轨迹球事件做一个判断,再调用findTouchedWindowTargesLocked()函数:

1
2
3
4
5
int32_t pointerIndex = getMotionEventActionPointerIndex(action);
int32_t x = int32_t(entry->pointerCoords[pointerIndex].
getAxisValue(AMOTION_EVENT_AXIS_X));
int32_t y = int32_t(entry->pointerCoords[pointerIndex].
getAxisValue(AMOTION_EVENT_AXIS_Y));

在这里取出了entry里面的pointerIndex与触摸点坐标的x y值。

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
// 从前向后遍历所有的window以找出触摸的window
size_t numWindows = mWindowHandles.size();
for (size_t i = 0; i < numWindows; i++) {
sp<InputWindowHandle> windowHandle = mWindowHandles.itemAt(i);
const InputWindowInfo *windowInfo = windowHandle->getInfo();
if (windowInfo->displayId != displayId) {
continue; // 错误的window(displayId不匹配)
}

int32_t flags = windowInfo->layoutParamsFlags;
if (windowInfo->visible) { // 如果window可见
if (!(flags & InputWindowInfo::FLAG_NOT_TOUCHABLE)) {
isTouchModal = (flags & (InputWindowInfo::FLAG_NOT_FOCUSABLE |
InputWindowInfo::FLAG_NOT_TOUCH_MODAL)) == 0;
// window可以被触摸
if (isTouchModal || windowInfo->touchableRegionContainsPoint(x, y)) { // (x,y)在window内可触摸区域内
newTouchedWindowHandle = windowHandle;
break; // 找到触摸的window,保存在newTouchedWindowHandle中
}
}
// 判断是否触摸window之外的区域
if (maskedAction == AMOTION_EVENT_ACTION_DOWN &&
(flags & InputWindowInfo::FLAG_WATCH_OUTSIDE_TOUCH)) {
int32_t outsideTargetFlags = InputTarget::FLAG_DISPATCH_AS_OUTSIDE;
if (isWindowObscuredAtPointLocked(windowHandle, x, y)) {
outsideTargetFlags |= InputTarget::FLAG_WINDOW_IS_OBSCURED;
} else if (isWindowObscuredLocked(windowHandle)) {
outsideTargetFlags |= InputTarget::FLAG_WINDOW_IS_PARTIALLY_OBSCURED;
}

mTempTouchState.addOrUpdateWindow(windowHandle, outsideTargetFlags, BitSet32(0));
}
}
}

这段代码的目的是为了遍历所有的window找到触摸对应的那个window

1
2
3
4
5
6
7
8
9
10
// Handle the case where we did not find a window.
if (newTouchedWindowHandle == NULL) {
// Try to assign the pointer to the first foreground window we find, if there is one.
newTouchedWindowHandle = mTempTouchState.getFirstForegroundWindowHandle();
if (newTouchedWindowHandle == NULL) {
ALOGI("Dropping event because there is no touchable window at (%d, %d).", x, y);
injectionResult = INPUT_EVENT_INJECTION_FAILED;
goto Failed;
}
}

如果遍历后没有找到合适的window,那就取第一个前台的window

然后通过addWindowTargetLocked()方法把缓存下来的结果存放入inputTargets中。

1
2
3
4
5
for (size_t i = 0; i < mTempTouchState.windows.size(); i++) {
const TouchedWindow& touchedWindow = mTempTouchState.windows.itemAt(i);
addWindowTargetLocked(touchedWindow.windowHandle, touchedWindow.targetFlags,
touchedWindow.pointerIds, inputTargets);
}
1
2
3
4
5
6
7
8
9
10
11
12
13
void InputDispatcher::addWindowTargetLocked(const sp<InputWindowHandle>& windowHandle,
int32_t targetFlags, BitSet32 pointerIds, Vector<InputTarget>& inputTargets) {
inputTargets.push();

const InputWindowInfo* windowInfo = windowHandle->getInfo();
InputTarget& target = inputTargets.editTop();
target.inputChannel = windowInfo->inputChannel;
target.flags = targetFlags;
target.xOffset = - windowInfo->frameLeft;
target.yOffset = - windowInfo->frameTop;
target.scaleFactor = windowInfo->scaleFactor;
target.pointerIds = pointerIds;
}

函数将原始的window数据进行了再次封装。

找到合适的window或是没有找到(处理错误)之后,函数返回到bool InputDispatcher::dispatchMotionLocked()中:

1
dispatchEventLocked(currentTime, entry, inputTargets);

开始向inputTargets中的目标分发事件:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
void InputDispatcher::dispatchEventLocked(nsecs_t currentTime,
EventEntry* eventEntry, const Vector<InputTarget>& inputTargets) {
...
for (size_t i = 0; i < inputTargets.size(); i++) {
const InputTarget& inputTarget = inputTargets.itemAt(i); // 遍历目标

ssize_t connectionIndex = getConnectionIndexLocked(inputTarget.inputChannel); // 见下文
if (connectionIndex >= 0) {
sp<Connection> connection = mConnectionsByFd.valueAt(connectionIndex);
prepareDispatchCycleLocked(currentTime, connection, eventEntry, &inputTarget);
} else {
...
}
}
}

inputTarget中包含的inputChannel就是后面用于与window实例通信的关键:

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
/*
* An input channel consists of a local unix domain socket used to send and receive
* input messages across processes. Each channel has a descriptive name for debugging purposes.
*
* Each endpoint has its own InputChannel object that specifies its file descriptor.
*
* The input channel is closed when all references to it are released.
*/
class InputChannel : public RefBase {
protected:
virtual ~InputChannel();

public:
InputChannel(const String8& name, int fd);

/* Creates a pair of input channels.
*
* Returns OK on success.
*/
static status_t openInputChannelPair(const String8& name,
sp<InputChannel>& outServerChannel, sp<InputChannel>& outClientChannel);

inline String8 getName() const { return mName; }
inline int getFd() const { return mFd; }

/* Sends a message to the other endpoint.
*
* If the channel is full then the message is guaranteed not to have been sent at all.
* Try again after the consumer has sent a finished signal indicating that it has
* consumed some of the pending messages from the channel.
*
* Returns OK on success.
* Returns WOULD_BLOCK if the channel is full.
* Returns DEAD_OBJECT if the channel's peer has been closed.
* Other errors probably indicate that the channel is broken.
*/
status_t sendMessage(const InputMessage* msg);

/* Receives a message sent by the other endpoint.
*
* If there is no message present, try again after poll() indicates that the fd
* is readable.
*
* Returns OK on success.
* Returns WOULD_BLOCK if there is no message present.
* Returns DEAD_OBJECT if the channel's peer has been closed.
* Other errors probably indicate that the channel is broken.
*/
status_t receiveMessage(InputMessage* msg);

/* Returns a new object that has a duplicate of this channel's fd. */
sp<InputChannel> dup() const;

private:
String8 mName;
int mFd;
};

InputChannel包含了一个本地unix socket用于跨进程发送与接收输入信息。

它的接口十分简单,我们就通过sendMessage()receiveMessage()两个函数实现跨进程通信。

回到之前,我们通过这个inputChannelFd(文件描述符)来获取一个Connection的索引,然后根据这个索引从mConnectionsByFd中获取connection对象。

1
2
3
4
5
6
7
8
9
10
ssize_t InputDispatcher::getConnectionIndexLocked(const sp<InputChannel>& inputChannel) {
ssize_t connectionIndex = mConnectionsByFd.indexOfKey(inputChannel->getFd());
if (connectionIndex >= 0) {
sp<Connection> connection = mConnectionsByFd.valueAt(connectionIndex);
if (connection->inputChannel.get() == inputChannel.get()) {
return connectionIndex;
}
}
return -1;
}

这个mConnectionByFd又是怎么建立起来的呢?在InputDispatcher中包含了一个registerInputChannel函数:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
status_t InputDispatcher::registerInputChannel(const sp<InputChannel>& inputChannel,
const sp<InputWindowHandle>& inputWindowHandle, bool monitor) {
...
{ // acquire lock
AutoMutex _l(mLock);

...
sp<Connection> connection = new Connection(inputChannel, inputWindowHandle, monitor);

int fd = inputChannel->getFd();
mConnectionsByFd.add(fd, connection);

if (monitor) {
mMonitoringChannels.push(inputChannel);
}

mLooper->addFd(fd, 0, ALOOPER_EVENT_INPUT, handleReceiveCallback, this);
} // release lock

// Wake the looper because some connections have changed.
mLooper->wake();
return OK;
}

connection对象就是在这里由inputChannel构造并加入到mConnectionsByFd中的。而mConnectionsByFd本身是一个以Fd为索引的键值对:

1
KeyedVector<int, sp<Connection> > mConnectionsByFd;

取得connection对象之后,进入到了prepareDispatchCycleLocked()函数中,这个函数对连接的状态是否正常进行检测,连接正常会调用enqueueDispatchEntriesLocked()函数:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
void InputDispatcher::enqueueDispatchEntriesLocked(nsecs_t currentTime,
const sp<Connection>& connection, EventEntry* eventEntry, const InputTarget* inputTarget) {
bool wasEmpty = connection->outboundQueue.isEmpty();

// Enqueue dispatch entries for the requested modes.
enqueueDispatchEntryLocked(connection, eventEntry, inputTarget,
InputTarget::FLAG_DISPATCH_AS_HOVER_EXIT);
enqueueDispatchEntryLocked(connection, eventEntry, inputTarget,
InputTarget::FLAG_DISPATCH_AS_OUTSIDE);
enqueueDispatchEntryLocked(connection, eventEntry, inputTarget,
InputTarget::FLAG_DISPATCH_AS_HOVER_ENTER);
enqueueDispatchEntryLocked(connection, eventEntry, inputTarget,
InputTarget::FLAG_DISPATCH_AS_IS);
enqueueDispatchEntryLocked(connection, eventEntry, inputTarget,
InputTarget::FLAG_DISPATCH_AS_SLIPPERY_EXIT);
enqueueDispatchEntryLocked(connection, eventEntry, inputTarget,
InputTarget::FLAG_DISPATCH_AS_SLIPPERY_ENTER);

// If the outbound queue was previously empty, start the dispatch cycle going.
if (wasEmpty && !connection->outboundQueue.isEmpty()) {
startDispatchCycleLocked(currentTime, connection);
}
}

中间调用了一系列的enqueueDispatchEntryLocked()函数:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
void InputDispatcher::enqueueDispatchEntryLocked(
const sp<Connection>& connection, EventEntry* eventEntry, const InputTarget* inputTarget,
int32_t dispatchMode) {
int32_t inputTargetFlags = inputTarget->flags;
if (!(inputTargetFlags & dispatchMode)) {
return;
}
inputTargetFlags = (inputTargetFlags & ~InputTarget::FLAG_DISPATCH_MASK) | dispatchMode;

// This is a new event.
// Enqueue a new dispatch entry onto the outbound queue for this connection.
DispatchEntry* dispatchEntry = new DispatchEntry(eventEntry, // increments ref
inputTargetFlags, inputTarget->xOffset, inputTarget->yOffset,
inputTarget->scaleFactor);
...
// Enqueue the dispatch entry.
connection->outboundQueue.enqueueAtTail(dispatchEntry);
traceOutboundQueueLengthLocked(connection);
...
}

省略的代码对entry进行了进一步的包装,然后在最后加入到了connection维护的outboundQueue中。

回到上面,之后调用startDispatchCycleLocked()正式开始分发事件:

1
2
3
4
5
6
7
8
9
while (connection->status == Connection::STATUS_NORMAL
&& !connection->outboundQueue.isEmpty()) {
DispatchEntry* dispatchEntry = connection->outboundQueue.head;
dispatchEntry->deliveryTime = currentTime;

// Publish the event.
status_t status;
EventEntry* eventEntry = dispatchEntry->eventEntry;
switch (eventEntry->type) {

connectionoutboundQueue取出entry之后,根据事件类型的不同对事件进一步处理:

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
case EventEntry::TYPE_MOTION: {
MotionEntry* motionEntry = static_cast<MotionEntry*>(eventEntry);

PointerCoords scaledCoords[MAX_POINTERS];
const PointerCoords* usingCoords = motionEntry->pointerCoords;

// Set the X and Y offset depending on the input source.
float xOffset, yOffset, scaleFactor;
if ((motionEntry->source & AINPUT_SOURCE_CLASS_POINTER)
&& !(dispatchEntry->targetFlags & InputTarget::FLAG_ZERO_COORDS)) {
scaleFactor = dispatchEntry->scaleFactor;
xOffset = dispatchEntry->xOffset * scaleFactor;
yOffset = dispatchEntry->yOffset * scaleFactor;
if (scaleFactor != 1.0f) {
for (uint32_t i = 0; i < motionEntry->pointerCount; i++) {
scaledCoords[i] = motionEntry->pointerCoords[i];
scaledCoords[i].scale(scaleFactor);
}
usingCoords = scaledCoords;
}
} else {
xOffset = 0.0f;
yOffset = 0.0f;
scaleFactor = 1.0f;

// We don't want the dispatch target to know.
if (dispatchEntry->targetFlags & InputTarget::FLAG_ZERO_COORDS) {
for (uint32_t i = 0; i < motionEntry->pointerCount; i++) {
scaledCoords[i].clear();
}
usingCoords = scaledCoords;
}
}

在对事件的坐标进行解析(缩放)之后,进入下面的发布过程:

1
2
3
4
5
6
7
8
9
10
11
12
    // Publish the motion event.
status = connection->inputPublisher.publishMotionEvent(dispatchEntry->seq,
motionEntry->deviceId, motionEntry->source,
dispatchEntry->resolvedAction, motionEntry->actionButton,
dispatchEntry->resolvedFlags, motionEntry->edgeFlags,
motionEntry->metaState, motionEntry->buttonState,
xOffset, yOffset, motionEntry->xPrecision, motionEntry->yPrecision,
motionEntry->downTime, motionEntry->eventTime,
motionEntry->pointerCount, motionEntry->pointerProperties,
usingCoords);
break;
}

实际上调用了InputPublisherpublishMotionEvent()函数:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
InputMessage msg;
msg.header.type = InputMessage::TYPE_MOTION;
msg.body.motion.seq = seq;
msg.body.motion.deviceId = deviceId;
msg.body.motion.source = source;
msg.body.motion.action = action;
msg.body.motion.actionButton = actionButton;
msg.body.motion.flags = flags;
msg.body.motion.edgeFlags = edgeFlags;
msg.body.motion.metaState = metaState;
msg.body.motion.buttonState = buttonState;
msg.body.motion.xOffset = xOffset;
msg.body.motion.yOffset = yOffset;
msg.body.motion.xPrecision = xPrecision;
msg.body.motion.yPrecision = yPrecision;
msg.body.motion.downTime = downTime;
msg.body.motion.eventTime = eventTime;
msg.body.motion.pointerCount = pointerCount;
for (uint32_t i = 0; i < pointerCount; i++) {
msg.body.motion.pointers[i].properties.copyFrom(pointerProperties[i]);
msg.body.motion.pointers[i].coords.copyFrom(pointerCoords[i]);
}
return mChannel->sendMessage(&msg);

函数里封装了一个msg,然后最终调用了mChannelsendMessage()方法进行跨进程通信。


现在我们到了图中的这一步:

我们的点击事件来到了建立的socket中,准备与交付给对应的app,我们知道每个app运行在自己的进程中,所以就需要使用socket来进行跨进程通信。


InputChannel连接建立过程

本段内容主要参考了

Gityuan的博客

详细分析及代码请移步上面链接。

连接的建立是在一个Activity启动时进行的。

Activity的启动是一个比较复杂的过程,会经过ActivityManagerServiceWindowManagerService的层层调用,最终到达WindowManagerGlobaladdView()方法。

1
2
3
4
5
6
7
8
9
10
11
12
13
public void addView(View view, ViewGroup.LayoutParams params,
Display display, Window parentWindow) {
...
ViewRootImpl root;
View panelParentView = null;
...

root = new ViewRootImpl(view.getContext(), display);
...

root.setView(view, wparams, panelParentView);
...
}

addView中,创建并初始化了一个ViewRootImpl对象,并调用了它的setView()方法。

ViewRootImpl的初始化过程:

1
2
3
4
5
public ViewRootImpl(Context context, Display display) {
...
mWindowSession = WindowManagerGlobal.getWindowSession();
...
}

这里我们关注的是这个mWindowSession对象的初始化。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
public static IWindowSession getWindowSession() {
synchronized (WindowManagerGlobal.class) {
if (sWindowSession == null) {
try {
InputMethodManager imm = InputMethodManager.getInstance();
IWindowManager windowManager = getWindowManagerService();
// 获取Session对象,利用Binder机制调用系统线程的方法
sWindowSession = windowManager.openSession(
new IWindowSessionCallback.Stub() {
@Override
public void onAnimatorScaleChanged(float scale) {
ValueAnimator.setDurationScale(scale);
}
},
imm.getClient(), imm.getInputContext());
} catch (RemoteException e) {
throw e.rethrowFromSystemServer();
}
}
return sWindowSession;
}
}

下面是setView()

1
2
3
4
5
6
7
8
9
10
11
12
13
14
public void setView(View view, WindowManager.LayoutParams attrs, View panelParentView) {
...
// 服务端过程
res = mWindowSession.addToDisplay(mWindow, mSeq, mWindowAttributes,
getHostVisibility(), mDisplay.getDisplayId(),
mAttachInfo.mContentInsets, mAttachInfo.mStableInsets,
mAttachInfo.mOutsets, mInputChannel);
...
// 客户端过程
if (mInputChannel != null) {
mInputEventReceiver = new WindowInputEventReceiver(mInputChannel, Looper.myLooper());
}
...
}

上面的两行语句分别的对应两个注册过程的开始,先执行服务端的注册与监听,再执行客户端的注册与监听。

下面对这两个过程分别进行追踪。

服务端过程

通过刚刚获取的mWindowSession去调用系统线程中的addToDisplay()方法:

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
    @Override
public int addToDisplay(IWindow window, int seq, WindowManager.LayoutParams attrs,
int viewVisibility, int displayId, Rect outContentInsets, Rect outStableInsets,
Rect outOutsets, InputChannel outInputChannel) {
return mService.addWindow(this, window, seq, attrs, viewVisibility, displayId,
outContentInsets, outStableInsets, outOutsets, outInputChannel);
}
...

这个`mService`自然就是之前获取它使用的`WindowManagerService`,调用它的`addWindow()`方法:

​```java
public int addWindow(Session session, IWindow client, int seq,
WindowManager.LayoutParams attrs, int viewVisibility, int displayId,
Rect outContentInsets, Rect outStableInsets, Rect outOutsets,
InputChannel outInputChannel) {
...
WindowState win = new WindowState(this, session, client, token,
attachedWindow, appOp[0], seq, attrs, viewVisibility, displayContent);
...
final boolean openInputChannels = (outInputChannel != null
&& (attrs.inputFeatures & INPUT_FEATURE_NO_INPUT_CHANNEL) == 0);
if (openInputChannels) {
win.openInputChannel(outInputChannel);
}
...
}

我们关注的是它创建并初始化了WindowState对象,然后调用了它的openInputChannel()方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
void openInputChannel(InputChannel outInputChannel) {
if (mInputChannel != null) {
throw new IllegalStateException("Window already has an input channel.");
}
String name = makeInputChannelName();
InputChannel[] inputChannels = InputChannel.openInputChannelPair(name);
mInputChannel = inputChannels[0]; // 这里将服务端的inputChannel保存在了WindowState中
mClientChannel = inputChannels[1];
mInputWindowHandle.inputChannel = inputChannels[0];
if (outInputChannel != null) {
mClientChannel.transferTo(outInputChannel);
mClientChannel.dispose();
mClientChannel = null;
} else {
// If the window died visible, we setup a dummy input channel, so that taps
// can still detected by input monitor channel, and we can relaunch the app.
// Create dummy event receiver that simply reports all events as handled.
mDeadWindowEventReceiver = new DeadWindowEventReceiver(mClientChannel);
}
mService.mInputManager.registerInputChannel(mInputChannel, mInputWindowHandle);
}

在这里创建了两个InputChannel对象,其中作为服务端存放在系统进程中的是inputChannels[0],作为客户端的存放在appui主线程中的是inputChannels[1]。它们的传递过程之后再看,我们先看InputChannel建立时调用的openInputChannelPair()方法:

1
2
3
4
5
6
7
8
9
10
public static InputChannel[] openInputChannelPair(String name) {
if (name == null) {
throw new IllegalArgumentException("name must not be null");
}

if (DEBUG) {
Slog.d(TAG, "Opening input channel pair '" + name + "'");
}
return nativeOpenInputChannelPair(name);
}

调用了native方法:

1
2
3
4
5
6
7
8
9
static jobjectArray android_view_InputChannel_nativeOpenInputChannelPair(JNIEnv* env,
jclass clazz, jstring nameObj) {
...
sp<InputChannel> serverChannel;
sp<InputChannel> clientChannel;
status_t result = InputChannel::openInputChannelPair(name, serverChannel, clientChannel);
...
return channelPair;
}

在这里就分为了serverChannelclientChannel,作为openInputChannelPair调用的两参数来成对地创建:

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
status_t InputChannel::openInputChannelPair(const String8& name,
sp<InputChannel>& outServerChannel, sp<InputChannel>& outClientChannel) {
int sockets[2];
if (socketpair(AF_UNIX, SOCK_SEQPACKET, 0, sockets)) {
status_t result = -errno;
ALOGE("channel '%s' ~ Could not create socket pair. errno=%d",
name.string(), errno);
outServerChannel.clear();
outClientChannel.clear();
return result;
}

int bufferSize = SOCKET_BUFFER_SIZE;
setsockopt(sockets[0], SOL_SOCKET, SO_SNDBUF, &bufferSize, sizeof(bufferSize));
setsockopt(sockets[0], SOL_SOCKET, SO_RCVBUF, &bufferSize, sizeof(bufferSize));
setsockopt(sockets[1], SOL_SOCKET, SO_SNDBUF, &bufferSize, sizeof(bufferSize));
setsockopt(sockets[1], SOL_SOCKET, SO_RCVBUF, &bufferSize, sizeof(bufferSize));

String8 serverChannelName = name;
serverChannelName.append(" (server)");
outServerChannel = new InputChannel(serverChannelName, sockets[0]);

String8 clientChannelName = name;
clientChannelName.append(" (client)");
outClientChannel = new InputChannel(clientChannelName, sockets[1]);
return OK;
}

我们可以看到第4行使用了系统调用socketpair()来创建一对本地socket对象,使用int数组sockets来接收创建好的fd

下面使用系统调用setsockopt()来为新建的socket进行配置,设置了socket实体的输入输出buff大小。

第21、25行调用了Native InputChannel的构造方法以name+(server)name+(client)作为名字参数与创建好的socket作为fd参数来构造InputChannel对象,它的构造方法:

1
2
3
4
5
6
7
8
9
10
11
InputChannel::InputChannel(const String8& name, int fd) :
mName(name), mFd(fd) {
#if DEBUG_CHANNEL_LIFECYCLE
ALOGD("Input channel constructed: name='%s', fd=%d",
mName.string(), fd);
#endif

int result = fcntl(mFd, F_SETFL, O_NONBLOCK);
LOG_ALWAYS_FATAL_IF(result != 0, "channel '%s' ~ Could not make socket "
"non-blocking. errno=%d", mName.string(), errno);
}

name fd保存到域中,并配置了fd

现在我们的两个InputChannel(实际上是一对本地socket的封装)就创建好了。

现在我们回到WindowStateopenInputChannel()方法中,在成功创建两个InputChannel后,调用了mService.mInputManager.registerInputChannel(mInputChannel, mInputWindowHandle);

1
2
3
4
5
6
7
8
public void registerInputChannel(InputChannel inputChannel,
InputWindowHandle inputWindowHandle) {
if (inputChannel == null) {
throw new IllegalArgumentException("inputChannel must not be null.");
}

nativeRegisterInputChannel(mPtr, inputChannel, inputWindowHandle, false);
}
1
2
3
4
5
6
7
8
9
10
11
12
13
14
static void nativeRegisterInputChannel(JNIEnv* env, jclass /* clazz */,
jlong ptr, jobject inputChannelObj, jobject inputWindowHandleObj, jboolean monitor) {
NativeInputManager* im = reinterpret_cast<NativeInputManager*>(ptr);

sp<InputChannel> inputChannel = android_view_InputChannel_getInputChannel(env,
inputChannelObj);
...
sp<InputWindowHandle> inputWindowHandle =
android_server_InputWindowHandle_getHandle(env, inputWindowHandleObj);

status_t status = im->registerInputChannel(
env, inputChannel, inputWindowHandle, monitor);
...
}
1
2
3
4
5
6
status_t NativeInputManager::registerInputChannel(JNIEnv* /* env */,
const sp<InputChannel>& inputChannel,
const sp<InputWindowHandle>& inputWindowHandle, bool monitor) {
return mInputManager->getDispatcher()->registerInputChannel(
inputChannel, inputWindowHandle, monitor);
}

getDispatcher()返回了InputDispatcher对象,这个方法就调用了我们之前提到过的InputDispatcher::registerInputchannel()方法,向InputDispatcher注册了我们创建好的InputChannel服务端。

我们回顾一下这个方法:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
status_t InputDispatcher::registerInputChannel(const sp<InputChannel>& inputChannel,
const sp<InputWindowHandle>& inputWindowHandle, bool monitor) {
...
{ // acquire lock
AutoMutex _l(mLock);

...
sp<Connection> connection = new Connection(inputChannel, inputWindowHandle, monitor);

int fd = inputChannel->getFd();
mConnectionsByFd.add(fd, connection);

if (monitor) {
mMonitoringChannels.push(inputChannel);
}

mLooper->addFd(fd, 0, ALOOPER_EVENT_INPUT, handleReceiveCallback, this);
} // release lock

// Wake the looper because some connections have changed.
mLooper->wake();
return OK;
}

InputChannelFd作为索引保存到InputDispatcher中。

然后调用mLooperaddFdNative Looper中设置自定义fd进行监听,传入了handleReceiveCallback()函数作为参数,这样一来,在服务端收到消息时就会进行回调。关于Native Looper处理事件的分析可见另一系列的博文

客户端过程

1
2
3
4
// 客户端过程
if (mInputChannel != null) {
mInputEventReceiver = new WindowInputEventReceiver(mInputChannel, Looper.myLooper());
}

这里以客户端InputChannel与当前应用的Looper作为参数,初始化了WindowInputEventReceiver对象:

1
2
3
4
5
6
7
8
9
public InputEventReceiver(InputChannel inputChannel, Looper looper) {
...
mInputChannel = inputChannel;
mMessageQueue = looper.getQueue();
mReceiverPtr = nativeInit(new WeakReference<InputEventReceiver>(this),
inputChannel, mMessageQueue);

mCloseGuard.open("dispose");
}

获取了app进程的消息队列,并调用native方法对mReceiverPtr进行初始化:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
static jlong nativeInit(JNIEnv* env, jclass clazz, jobject receiverWeak,
jobject inputChannelObj, jobject messageQueueObj) {
...
sp<InputChannel> inputChannel = android_view_InputChannel_getInputChannel(env,
inputChannelObj);

sp<MessageQueue> messageQueue = android_os_MessageQueue_getMessageQueue(env, messageQueueObj);
...

sp<NativeInputEventReceiver> receiver = new NativeInputEventReceiver(env,
receiverWeak, inputChannel, messageQueue);
status_t status = receiver->initialize();
...
receiver->incStrong(gInputEventReceiverClassInfo.clazz); // retain a reference for the object
return reinterpret_cast<jlong>(receiver.get());
}

创建了NativeInputEventReceiver对象,并调用initialize()方法进行初始化:

1
2
3
4
status_t NativeInputEventReceiver::initialize() {
setFdEvents(ALOOPER_EVENT_INPUT);
return OK;
}
1
2
3
4
5
6
7
8
9
10
11
void NativeInputEventReceiver::setFdEvents(int events) {
if (mFdEvents != events) {
mFdEvents = events;
int fd = mInputConsumer.getChannel()->getFd();
if (events) {
mMessageQueue->getLooper()->addFd(fd, 0, events, this, NULL);
} else {
mMessageQueue->getLooper()->removeFd(fd);
}
}
}

这里将客户端的InputChannel保存的Fd加入到了Native Looper中进行监听,对返回的消息进行处理:

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
int Looper::addFd(int fd, int ident, int events, const sp<LooperCallback>& callback, void* data) {
{ // acquire lock
AutoMutex _l(mLock);

Request request;
request.fd = fd;
request.ident = ident;
request.events = events;
request.seq = mNextRequestSeq++;
request.callback = callback;
request.data = data;
if (mNextRequestSeq == -1) mNextRequestSeq = 0; // reserve sequence number -1

struct epoll_event eventItem;
request.initEventItem(&eventItem);

ssize_t requestIndex = mRequests.indexOfKey(fd);
if (requestIndex < 0) {
int epollResult = epoll_ctl(mEpollFd, EPOLL_CTL_ADD, fd, & eventItem);
if (epollResult < 0) {
ALOGE("Error adding epoll events for fd %d: %s", fd, strerror(errno));
return -1;
}
mRequests.add(fd, request);
} else {
int epollResult = epoll_ctl(mEpollFd, EPOLL_CTL_MOD, fd, & eventItem);
if (epollResult < 0) {
if (errno == ENOENT) {
epollResult = epoll_ctl(mEpollFd, EPOLL_CTL_ADD, fd, & eventItem);
...
...
}
...
}

}
}

这里将NativeInputEventReceiver传入的this作为callback存入到request中, 再以fd为索引向mRequests映射表中加入request,然后以fd作为信号调用epoll_ctl系统调用,利用它对通信过程进行监听,在收到消息之后最终会根据fd找到mRequests中的request保存的callback,即NativeInputEventReceiver对象。

关于addFd方法的详细分析,可以查看Native消息机制的博文

小结

现在我们了解了从内核到应用整个触摸事件的传输过程,并且知道了InputChannel在两端的监听建立与触发的函数,至此,触摸事件已经从系统底层来到了我们的应用进程,下一篇博客将从触发函数开始讲解事件从native层真正传入java层的过程。