当前位置：首页 > news >正文

生产者Producer往BufferQueue中写数据的过程

news 2025/7/23 5:16:35

In normal operation, the producer calls

dequeueBuffer() to get an empty buffer, fills it with data, then
calls
queueBuffer() to make it available to the consumer

代码如下：

// XXX: Tests that fork a process to hold the BufferQueue must run before tests
// that use a local BufferQueue, or else Binder will get unhappy
//
// In one instance this was a crash in the createBufferQueue where the
// binder call to create a buffer allocator apparently got garbage back.
// See b/36592665.
TEST_F(BufferQueueTest, DISABLED_BufferQueueInAnotherProcess) {const String16 PRODUCER_NAME = String16("BQTestProducer");const String16 CONSUMER_NAME = String16("BQTestConsumer");pid_t forkPid = fork();ASSERT_NE(forkPid, -1);if (forkPid == 0) {// Child processsp<IGraphicBufferProducer> producer;sp<IGraphicBufferConsumer> consumer;BufferQueue::createBufferQueue(&producer, &consumer);sp<IServiceManager> serviceManager = defaultServiceManager();serviceManager->addService(PRODUCER_NAME, IInterface::asBinder(producer));serviceManager->addService(CONSUMER_NAME, IInterface::asBinder(consumer));ProcessState::self()->startThreadPool();IPCThreadState::self()->joinThreadPool();LOG_ALWAYS_FATAL("Shouldn't be here");}sp<IServiceManager> serviceManager = defaultServiceManager();sp<IBinder> binderProducer =serviceManager->getService(PRODUCER_NAME);mProducer = interface_cast<IGraphicBufferProducer>(binderProducer);EXPECT_TRUE(mProducer != nullptr);sp<IBinder> binderConsumer =serviceManager->getService(CONSUMER_NAME);mConsumer = interface_cast<IGraphicBufferConsumer>(binderConsumer);EXPECT_TRUE(mConsumer != nullptr);sp<MockConsumer> mc(new MockConsumer);ASSERT_EQ(OK, mConsumer->consumerConnect(mc, false));IGraphicBufferProducer::QueueBufferOutput output;ASSERT_EQ(OK,mProducer->connect(nullptr, NATIVE_WINDOW_API_CPU, false, &output));int slot;sp<Fence> fence;sp<GraphicBuffer> buffer;ASSERT_EQ(IGraphicBufferProducer::BUFFER_NEEDS_REALLOCATION,mProducer->dequeueBuffer(&slot, &fence, 0, 0, 0, GRALLOC_USAGE_SW_WRITE_OFTEN,nullptr, nullptr));ASSERT_EQ(OK, mProducer->requestBuffer(slot, &buffer));uint32_t* dataIn;ASSERT_EQ(OK, buffer->lock(GraphicBuffer::USAGE_SW_WRITE_OFTEN,reinterpret_cast<void**>(&dataIn)));*dataIn = TEST_DATA;ASSERT_EQ(OK, buffer->unlock());IGraphicBufferProducer::QueueBufferInput input(0, false,HAL_DATASPACE_UNKNOWN, Rect(0, 0, 1, 1),NATIVE_WINDOW_SCALING_MODE_FREEZE, 0, Fence::NO_FENCE);ASSERT_EQ(OK, mProducer->queueBuffer(slot, input, &output));BufferItem item;ASSERT_EQ(OK, mConsumer->acquireBuffer(&item, 0));uint32_t* dataOut;ASSERT_EQ(OK, item.mGraphicBuffer->lock(GraphicBuffer::USAGE_SW_READ_OFTEN,reinterpret_cast<void**>(&dataOut)));ASSERT_EQ(*dataOut, TEST_DATA);ASSERT_EQ(OK, item.mGraphicBuffer->unlock());
}

在Android图形系统中，生产者往BufferQueue中写数据的过程通常涉及以下步骤：

创建BufferQueue：
- 生产者首先需要与BufferQueue建立联系，通常通过BufferQueue::createBufferQueue等方法创建BufferQueue的生产者和消费者接口（IGraphicBufferProducer 和 IGraphicBufferConsumer）。
请求缓冲区：
- 生产者调用IGraphicBufferProducer::dequeueBuffer方法请求一个新的缓冲区。此方法会返回一个缓冲区槽（slot）编号以及一个Fence对象，表明缓冲区何时准备就绪。
锁定缓冲区：
- 获得缓冲区后，生产者调用GraphicBuffer::lock方法锁定缓冲区，这样就可以获得指向缓冲区内存的指针。生产者会指定所需的访问模式，如GRALLOC_USAGE_SW_WRITE_OFTEN（常用于软件写入）。
填充数据：
- 生产者使用获得的内存指针来填充图像数据，例如像素数据。
解锁缓冲区：
- 数据填充完成后，生产者必须调用GraphicBuffer::unlock来解锁缓冲区，使其可以被消费者读取或被操作系统调度。
设置缓冲区属性并入队：
- 生产者创建一个IGraphicBufferProducer::QueueBufferInput结构体，设置缓冲区的属性，如时间戳、裁剪区域、缩放模式、变换等。
- 然后，调用IGraphicBufferProducer::queueBuffer方法，将填充好的缓冲区和相应的属性入队到BufferQueue中。
通知消费者：
- 当缓冲区入队后，如果设置了正确的同步机制，BufferQueue会自动或通过回调通知消费者有新的缓冲区可用。