Android 12(S) 图形显示系统 – 示例应用(二)

1 前言

为了更深刻的理解Android图形系统抽象的概念和BufferQueue的工作机制,这篇文章我们将从Native Level入手,基于Android图形系统API写作一个简单的图形处理小程序。透过这个小程序我们将学习如何使用Native API创建Surface,如何请求图形缓冲区,如何向图形缓冲区中写入数据等知识。Talk is cheap, show me the code。让我们马上开始吧!

注:本系列文章的分析及代码均基于Android 12(S) Source Code,可参考://aospxref.com/  或 //aosp.opersys.com/

2 程序源码简介

  • 源码下载

地址://github.com/yrzroger/NativeSFDemo

注:可以git命令下载(比如git clone [email protected]:yrzroger/NativeSFDemo.git)或直接Download ZIP解压后使用

  • 源码编译

本demo程序是基于Android源码编译环境开发的,所以需要放到Android源码目录下编译。

将上一步中下载的的源码放到Android源码的合适目录下,然后执行mm进行编译,得到可执行档 NativeSFDemo

  • 源码运行

将可执行档NativeSFDemo放到目标测试平台/system/bin下(比如:adb push NativeSFDemo /system/bin/)

然后执行 adb shell NativeSFDemo

  • 效果展示

程序中去绘制单色背景: 红色->绿色->蓝色背景交替展示,如下图所示:

至此你已经收获一个可以供后续学习研究的demo小程序了 !!!


Tips:

Android源码是一个宝藏,即提供了丰富多彩的APIs供开发者使用,又可以在其中搜索到很多有价值的APIs使用实例。本文中提供的演示Demo亦是基于源码中的参考来完成的。

我把参考位置列于此:

参考1:/frameworks/av/media/libstagefright/SurfaceUtils.cpp 

参考2:/frameworks/native/libs/gui/tests/CpuConsumer_test.cpp


 

3 程序源码分析

在显示子系统中,Surface 是一个接口,供生产者与消费者交换缓冲区。通过Surface我们就能向BufferQueue请求Buffer,并和Android Native窗口系统建立连接。本文展示的demo就是基于Surface建立起来的。
 
  • 封装类NativeSurfaceWrapper

NativeSurfaceWrapper是对Surface的一层封装,用于获取屏幕参数并创建与配置Surface属性。

首先看到头文件中该类的定义:

class NativeSurfaceWrapper : public RefBase {
public:
    NativeSurfaceWrapper(const String8& name);
    virtual ~NativeSurfaceWrapper() {}

    virtual void onFirstRef();

    // Retrieves a handle to the window.
    sp<ANativeWindow>  getSurface() const;

    int width() { return mWidth; }
    int height() { return mHeight; }

private:
    DISALLOW_COPY_AND_ASSIGN(NativeSurfaceWrapper);
    ui::Size limitSurfaceSize(int width, int height) const;

    sp<SurfaceControl> mSurfaceControl;
    int mWidth;
    int mHeight;
    String8 mName;
};
 
NativeSurfaceWrapper继承自Refbase,这样就可以使用智能指针sp,wp来管理其对象,避免内存泄漏。

同时可以重写onFirstRef方法,在创建NativeSurfaceWrapper对象第一次被引用时调用onFirstRef做一些初始化操作。

下面是onFirstRef的定义:

void NativeSurfaceWrapper::onFirstRef() {
    sp<SurfaceComposerClient> surfaceComposerClient = new SurfaceComposerClient;
    status_t err = surfaceComposerClient->initCheck();
    if (err != NO_ERROR) {
        ALOGD("SurfaceComposerClient::initCheck error: %#x\n", err);
        return;
    }

    // Get main display parameters.
    sp<IBinder> displayToken = SurfaceComposerClient::getInternalDisplayToken();
    if (displayToken == nullptr)
        return;

    ui::DisplayMode displayMode;
    const status_t error =
            SurfaceComposerClient::getActiveDisplayMode(displayToken, &displayMode);
    if (error != NO_ERROR)
        return;

    ui::Size resolution = displayMode.resolution;
    resolution = limitSurfaceSize(resolution.width, resolution.height);
    // create the native surface
    sp<SurfaceControl> surfaceControl = surfaceComposerClient->createSurface(mName, resolution.getWidth(), 
                                                                             resolution.getHeight(), PIXEL_FORMAT_RGBA_8888,
                                                                             ISurfaceComposerClient::eFXSurfaceBufferState,
                                                                             /*parent*/ nullptr);

    SurfaceComposerClient::Transaction{}
            .setLayer(surfaceControl, std::numeric_limits<int32_t>::max())
            .show(surfaceControl)
            .apply();

    mSurfaceControl = surfaceControl;
    mWidth = resolution.getWidth();
    mHeight = resolution.getHeight();
}
 

onFirstRef中完成主要工作:

1. 创建一个SurfaceComposerClient对象,这是SurfaceFlinger的Client端,它将建立和SurfaceFlinger服务的通信;

2. 获取屏幕参数,SurfaceComposerClient::getActiveDisplayMode获取当前的DisplayMode,其中可以得到resolution信息;

3. 创建Surface & SurfaceControl,createSurface方法会通过Binder通信机制一直呼叫到SurfaceFlinger,SurfaceFlinger会进行创建Layer等操作;

4. createSurface时会设置width,height,format等信息;

5. setLayer,设置窗口的z-order,SurfaceFlinger根据z-Oder决定窗口的可见性及可见大小;

6. show,让当前窗口可见;

7. apply,使透过Transaction进行的设置生效,属性信息传给SurfaceFlinger;
 

Tips:

创建Surface的过程会涉及到与SurfaceFlinger的互动,SurfaceFlinger是一个系统级的服务,负责创建/管理/合成Surface对应的Layer,这部分我们本文暂不展开,之后文章中会陆续讲解。


 
limitSurfaceSize方法
该方法的作用是将width和height限制在设备GPU支持的范围内。
 
ui::Size NativeSurfaceWrapper::limitSurfaceSize(int width, int height) const {
    ui::Size limited(width, height);
    bool wasLimited = false;
    const float aspectRatio = float(width) / float(height);

    int maxWidth = android::base::GetIntProperty("ro.surface_flinger.max_graphics_width", 0);
    int maxHeight = android::base::GetIntProperty("ro.surface_flinger.max_graphics_height", 0);

    if (maxWidth != 0 && width > maxWidth) {
        limited.height = maxWidth / aspectRatio;
        limited.width = maxWidth;
        wasLimited = true;
    }
    if (maxHeight != 0 && limited.height > maxHeight) {
        limited.height = maxHeight;
        limited.width = maxHeight * aspectRatio;
        wasLimited = true;
    }
    SLOGV_IF(wasLimited, "Surface size has been limited to [%dx%d] from [%dx%d]",
             limited.width, limited.height, width, height);
    return limited;
}

该方法会将屏幕的width/height和max_graphics_width/max_graphics_height进行比较,取较小者作为创建Surface的参数。

这一点也是Android 12引入的一个新特性。getActiveDisplayMode获取到的是屏幕的真实分辨率(real display resolution),但GPU可能不支持高分辨率的UI合成,所以必须对framebuffer size做出限制。

比如设备可以4K分辨率进行video的解码和渲染,但由于硬件限制application UI只能以1080P进行合成。

 

  • NativeSFDemo的main方法

main方法比较简单

1. signal函数注册监听SIGINT信号的handler,也就是保证Ctrl+C退出程序的完整性;

2. 创建NativeSurfaceWrapper对象,并调用drawNativeSurface进行图片的绘制;

int main() {
    signal(SIGINT, sighandler);

    sp<NativeSurfaceWrapper> nativeSurface(new NativeSurfaceWrapper(String8("NativeSFDemo")));
    drawNativeSurface(nativeSurface);
    return 0;
}

按下Ctrl+C退出程序时,呼叫到sighandler将mQuit这个标志设为true,这样会使图片的while循环就可以正常流程退出了

void sighandler(int num) {
    if(num == SIGINT) {
        printf("\nSIGINT: Force to stop !\n");
        mQuit = true;
    }
}

 

  • drawNativeSurface方法

绘制图片的核心逻辑都在这个方法中,我们先看一下代码:

int drawNativeSurface(sp<NativeSurfaceWrapper> nativeSurface) {
    status_t err = NO_ERROR;
    int countFrame = 0;
    ANativeWindowBuffer *nativeBuffer = nullptr;
    ANativeWindow* nativeWindow = nativeSurface->getSurface().get();

    // 1. connect the ANativeWindow as a CPU client. Buffers will be queued after being filled using the CPU
    err = native_window_api_connect(nativeWindow, NATIVE_WINDOW_API_CPU);
    if (err != NO_ERROR) {
        ALOGE("ERROR: unable to native_window_api_connect\n");
        return EXIT_FAILURE;
    }

    // 2. set the ANativeWindow dimensions
    err = native_window_set_buffers_user_dimensions(nativeWindow, nativeSurface->width(), nativeSurface->height());
    if (err != NO_ERROR) {
        ALOGE("ERROR: unable to native_window_set_buffers_user_dimensions\n");
        return EXIT_FAILURE;
    }

    // 3. set the ANativeWindow format
    err = native_window_set_buffers_format(nativeWindow, PIXEL_FORMAT_RGBX_8888);
    if (err != NO_ERROR) {
        ALOGE("ERROR: unable to native_window_set_buffers_format\n");
        return EXIT_FAILURE;
    }

    // 4. set the ANativeWindow usage
    err = native_window_set_usage(nativeWindow, GRALLOC_USAGE_SW_WRITE_OFTEN);
    if (err != NO_ERROR) {
        ALOGE("native_window_set_usage failed: %s (%d)", strerror(-err), -err);
        return err;
    }

    // 5. set the ANativeWindow scale mode
    err = native_window_set_scaling_mode(nativeWindow, NATIVE_WINDOW_SCALING_MODE_SCALE_TO_WINDOW);
    if (err != NO_ERROR) {
        ALOGE("native_window_set_scaling_mode failed: %s (%d)", strerror(-err), -err);
        return err;
    }

    // 6. set the ANativeWindow permission to allocte new buffer, default is true
    static_cast<Surface*>(nativeWindow)->getIGraphicBufferProducer()->allowAllocation(true);

    // 7. set the ANativeWindow buffer count
    int numBufs = 0;
    int minUndequeuedBufs = 0;

    err = nativeWindow->query(nativeWindow,
            NATIVE_WINDOW_MIN_UNDEQUEUED_BUFFERS, &minUndequeuedBufs);
    if (err != NO_ERROR) {
        ALOGE("error: MIN_UNDEQUEUED_BUFFERS query "
                "failed: %s (%d)", strerror(-err), -err);
        goto handle_error;
    }

    numBufs = minUndequeuedBufs + 1;
    err = native_window_set_buffer_count(nativeWindow, numBufs);
    if (err != NO_ERROR) {
        ALOGE("error: set_buffer_count failed: %s (%d)", strerror(-err), -err);
        goto handle_error;
    }

    // 8. draw the ANativeWindow
    while(!mQuit) {
        // 9. dequeue a buffer
        int hwcFD = -1;
        err = nativeWindow->dequeueBuffer(nativeWindow, &nativeBuffer, &hwcFD);
        if (err != NO_ERROR) {
            ALOGE("error: dequeueBuffer failed: %s (%d)",
                    strerror(-err), -err);
            break;
        }

        // 10. make sure really control the dequeued buffer
        sp<Fence> hwcFence(new Fence(hwcFD));
        int waitResult = hwcFence->waitForever("dequeueBuffer_EmptyNative");
        if (waitResult != OK) {
            ALOGE("dequeueBuffer_EmptyNative: Fence::wait returned an error: %d", waitResult);
            break;
        }

        sp<GraphicBuffer> buf(GraphicBuffer::from(nativeBuffer));

        // 11. Fill the buffer with black
        uint8_t* img = nullptr;
        err = buf->lock(GRALLOC_USAGE_SW_WRITE_OFTEN, (void**)(&img));
        if (err != NO_ERROR) {
            ALOGE("error: lock failed: %s (%d)", strerror(-err), -err);
            break;
        }

        //12. Draw the window
        countFrame = (countFrame+1)%3;
        fillRGBA8Buffer(img, nativeSurface->width(), nativeSurface->height(), buf->getStride(),
                        countFrame == 0 ? 255 : 0,
                        countFrame == 1 ? 255 : 0,
                        countFrame == 2 ? 255 : 0);

        err = buf->unlock();
        if (err != NO_ERROR) {
            ALOGE("error: unlock failed: %s (%d)", strerror(-err), -err);
            break;
        }

        // 13. queue the buffer to display
        int gpuFD = -1;
        err = nativeWindow->queueBuffer(nativeWindow, buf->getNativeBuffer(), gpuFD);
        if (err != NO_ERROR) {
            ALOGE("error: queueBuffer failed: %s (%d)", strerror(-err), -err);
            break;
        }

        nativeBuffer = nullptr;
        sleep(1);
    }

handle_error:
    // 14. cancel buffer
    if (nativeBuffer != nullptr) {
        nativeWindow->cancelBuffer(nativeWindow, nativeBuffer, -1);
        nativeBuffer = nullptr;
    }

    // 15. Clean up after success or error.
    err = native_window_api_disconnect(nativeWindow, NATIVE_WINDOW_API_CPU);
    if (err != NO_ERROR) {
        ALOGE("error: api_disconnect failed: %s (%d)", strerror(-err), -err);
    }

    return err;
}

处理的大概过程

1. 获取我们已经创建Surface的窗口ANativeWindow,作为CPU客户端来连接ANativeWindow,CPU填充buffer数据后入队列进行后续处理;

2. 设置Buffer的大小尺寸native_window_set_buffers_user_dimensions;

3. 设置Buffer格式,可选,之前创建Layer的时候已经设置了;

4. 设置Buffer的usage,可能涉及protected的内容,这里我们简单设为GRALLOC_USAGE_SW_WRITE_OFTEN;

5. 设置scale模式,如果上层给的数据,比如Video,超出Buffer的大小后,怎么处理,是截取一部分还是,缩小;

6. 设置permission允许分配新buffer,默认true;

7. 设置Buffer数量,即BufferQueue中有多少个buffer可以用;

8. 下面的流程就是请求buffer并进行绘制图像的过程

9. dequeueBuffer先请求一块可用的Buffer,也就是FREE的Buffer;

10. Buffer虽然是Free的,但是在异步模式下,Buffer可能还在使用中,需要等到Fence才能确保buffer没有在被使用;

11. lock方法可以获取这块GraphicBuffer的数据地址;

12. 绘制图像,即把图像颜色数据写入Buffer里面,我们这里使用fillRGBA8Buffer来填充纯色图片;

13. 将绘制好的Buffer,queue到Buffer队列中,入队列后的buffer就可以被消费者处理或显示了;

14. 错误处理,取消掉Buffer,cancelBuffer;

15. 断开BufferQueue和窗口的连接,native_window_api_disconnect。

  • fillRGBA8Buffer
void fillRGBA8Buffer(uint8_t* img, int width, int height, int stride, int r, int g, int b) {
    for (int y = 0; y < height; y++) {
        for (int x = 0; x < width; x++) {
            uint8_t* pixel = img + (4 * (y*stride + x));
            pixel[0] = r;
            pixel[1] = g;
            pixel[2] = b;
            pixel[3] = 0;
        }
    }
}

fillRGBA8Buffer用指定的RGBA填充buffer数据,我们设置的颜色格式为PIXEL_FORMAT_RGBX_8888,所以每个像素点均由4个字节组成,前3个字节分别为R/G/B颜色分量。

 


 
我们可以通过执行 dumpsys SurfaceFlinger 来查看图层的信息
Display 4629995328241972480 HWC layers:
---------------------------------------------------------------------------------------------------------------------------------------------------------------
 Layer name
           Z |  Window Type |  Comp Type |  Transform |   Disp Frame (LTRB) |          Source Crop (LTRB) |     Frame Rate (Explicit) (Seamlessness) [Focused]
---------------------------------------------------------------------------------------------------------------------------------------------------------------
 bbq-wrapper#0
  rel      0 |            0 |     CLIENT |          0 |    0    0 1920 1080 |    0.0    0.0 1920.0 1080.0 |                                              [ ]
---------------------------------------------------------------------------------------------------------------------------------------------------------------

 
看到了没,一个名字为“bbq-wrapper#0”的Layer显示在最上层,也就是我们应用显示的图层,看到这里你一定有个疑问,我们设置的Surface Name不是“NativeSFDemo”吗 ?
 
dumpsys SurfaceFlinger信息中,我们还可以看到如下内容:
+ BufferStateLayer (NativeSFDemo#0) uid=0
  Region TransparentRegion (this=0 count=0)
  Region VisibleRegion (this=0 count=0)
  Region SurfaceDamageRegion (this=0 count=0)
      layerStack=   0, z=2147483647, pos=(0,0), size=(  -1,  -1), crop=[  0,   0,  -1,  -1], cornerRadius=0.000000, isProtected=0, isTrustedOverlay=0, isOpaque=0, invalidate=0, dataspace=Default, defaultPixelFormat=Unknown/None, backgroundBlurRadius=0, color=(0.000,0.000,0.000,1.000), flags=0x00000000, tr=[0.00, 0.00][0.00, 0.00]
      parent=none
      zOrderRelativeOf=none
      activeBuffer=[   0x   0:   0,Unknown/None], tr=[0.00, 0.00][0.00, 0.00] queued-frames=0, mRefreshPending=0, metadata={}, cornerRadiusCrop=[0.00, 0.00, 0.00, 0.00],  shadowRadius=0.000, 
+ BufferStateLayer (bbq-wrapper#0) uid=0
  Region TransparentRegion (this=0 count=0)
  Region VisibleRegion (this=0 count=1)
    [  0,   0, 1920, 1080]
  Region SurfaceDamageRegion (this=0 count=0)
      layerStack=   0, z=        0, pos=(0,0), size=(1920,1080), crop=[  0,   0,  -1,  -1], cornerRadius=0.000000, isProtected=0, isTrustedOverlay=0, isOpaque=1, invalidate=0, dataspace=Default, defaultPixelFormat=RGBx_8888, backgroundBlurRadius=0, color=(0.000,0.000,0.000,1.000), flags=0x00000100, tr=[0.00, 0.00][0.00, 0.00]
      parent=NativeSFDemo#0
      zOrderRelativeOf=none
      activeBuffer=[1920x1080:1920,RGBx_8888], tr=[0.00, 0.00][0.00, 0.00] queued-frames=0, mRefreshPending=0, metadata={dequeueTime:700243748286}, cornerRadiusCrop=[0.00, 0.00, 0.00, 0.00],  shadowRadius=0.000, 
 
两个BufferStateLayer:BufferStateLayer (NativeSFDemo#0)  和  BufferStateLayer (bbq-wrapper#0),其中bbq-wrapper#0的parent就是NativeSFDemo#0,这其中的关系我们之后的文章中会陆续分析。

 

4 小结

至此,我们已经建立起来了一个简单的图形图像处理的简单Demo,当让我们目前还是只从应用的较多介绍了基本图形APIs的使用逻辑,接下来的我们就基于此demo,深入底层逻辑探究其中的奥秘。

 

 



必读:

Android 12(S) 图形显示系统 – 开篇