Android 9.0 multimedia框架解析(五)ACodec配置解码器过程
概述
先来回顾下上篇文章所讲的MediaPlayerService的start过程。MediaPlayerService的start通过层层调用会调用到NuPlayer::onStart，onStart会先启动GenericSource，然后通过异步的方式调用到NuPlayer::instantiateDecoder。在instantiateDecoder中，会创建NuPlayer::Decoder（这个Decoder跟我标题里面的decoder是不一样的，标题里面的decoder指的是实际的解码器，而这里的NuPlayer::Decoder经过层层调用会调用到实际的解码器），然后会进行初始化和配置。从时序图中可以看出Decoder的配置过程会先去查找对应的解码器，然后为为解码器创建ACodec。上一节我们分析到了创建ACodec这里，没有分析ACodec的配置过程。这里我以自己集成的ffmpeg播放视频为场景，分析ACodec配置解码器的过程。
时序图
configureCodec
从时序图可以看出配置过程会调用到ACodec::configureCodec，这个configureCodec是配置过程的核心函数。
来看看配置的第一步SetComponentRole。
status_t ACodec::configureCodec(
const char *mime, const sp<AMessage> &msg) {
if (!msg->findInt32("encoder", &encoder)) { //从AMessage中取出encoder对应的值，如果不存在就认为是要初始化解码器。
encoder = false;
}
......
status_t err = setComponentRole(encoder /* isEncoder */, mime);
}
status_t ACodec::setComponentRole(
bool isEncoder, const char *mime) {
const char *role = GetComponentRole(isEncoder, mime);//根据mime找到组件的ComponentRole
......
status_t err = SetComponentRole(mOMXNode, role);//把ComponentRole传递给解码器
......
}
status_t SetComponentRole(const sp<IOMXNode> &omxNode, const char *role) {
OMX_PARAM_COMPONENTROLETYPE roleParams;
InitOMXParams(&roleParams);
strncpy((char *)roleParams.cRole,role, OMX_MAX_STRINGNAME_SIZE - 1);
roleParams.cRole[OMX_MAX_STRINGNAME_SIZE - 1] = '\0';
return omxNode->setParameter(OMX_IndexParamStandardComponentRole,&roleParams, sizeof(roleParams));//传给解码器
}
omxNode->setParameter就是调用到OMXNodeInstance::setParameter，然后OMXNodeInstance::setParameter会去调用到真正的解码器。来看看我解码器中对于OMX_IndexParamStandardComponentRole会做什么。
OMX_ERRORTYPE SoftFFmpegVideoDec::internalSetParameter(OMX_INDEXTYPE index, const OMX_PTR params)
{
switch (index){
case OMX_IndexParamStandardComponentRole:{
const OMX_PARAM_COMPONENTROLETYPE *roleParams = (const OMX_PARAM_COMPONENTROLETYPE *)params;
if (strncmp((const char *)roleParams->cRole,"video_decoder.ffmpeg",OMX_MAX_STRINGNAME_SIZE - 1)){
ALOGE("OMX_ErrorUndefined");
return OMX_ErrorUndefined;
}
return OMX_ErrorNone;
}
}
可以看出，我会识别roleParams->cRole是否是等于video_decoder.ffmpeg，如果是则返回OMX_ErrorNone，让配置过程继续。
来看看第二步setPortMode。一般来说一个解码器有两个port，一个是input，一个是output。input用于输入数据，output用于存放解码完的数据。我一般只分析output。
status_t ACodec::configureCodec(
const char *mime, const sp<AMessage> &msg) {
if (haveNativeWindow) { //如果使用nativewindow
......
err = setPortMode(kPortIndexOutput, IOMX::kPortModeDynamicANWBuffer);//设置output的端口模式为DynamicANWBuffer
}
}
status_t ACodec::setPortMode(int32_t portIndex, IOMX::PortMode mode) {
status_t err = mOMXNode->setPortMode(portIndex, mode); //调用OMXNodeInstance::setPortMode
......
mPortMode[portIndex] = mode;
return OK;
}
status_t OMXNodeInstance::setPortMode(OMX_U32 portIndex, IOMX::PortMode mode) {
switch (mode) {
case IOMX::kPortModeDynamicANWBuffer:
{
if (portIndex == kPortIndexOutput) { //由于是配置output，所以会先走这里
.....
status_t err = enableNativeBuffers_l(
portIndex, OMX_TRUE /*graphic*/, OMX_TRUE);
}
(void)enableNativeBuffers_l(portIndex, OMX_FALSE /*graphic*/, OMX_FALSE);
return storeMetaDataInBuffers_l(portIndex, OMX_TRUE, NULL);
}
}
status_t OMXNodeInstance::enableNativeBuffers_l(....){
OMX_STRING name = const_cast<OMX_STRING>(
graphic ? "OMX.google.android.index.enableAndroidNativeBuffers" //graphic为true
: "OMX.google.android.index.allocateNativeHandle");
OMX_INDEXTYPE index;
OMX_ERRORTYPE err = OMX_GetExtensionIndex(mHandle, name, &index); //获取自定义的的index（或许是为了兼容把）
if (err == OMX_ErrorNone) {
err = OMX_SetParameter(mHandle, index, ¶ms); //设置参数
}
}
来看一下我的解码器是怎么实现OMX_GetExtensionIndex对应的函数接口的
OMX_ERRORTYPE SoftFFmpegVideoDec::getExtensionIndex( const char *name, OMX_INDEXTYPE *index) {
if(strcmp(name, SPRD_INDEX_PARAM_ENABLE_ANB) == 0) {
ALOGI("getExtensionIndex:%s",SPRD_INDEX_PARAM_ENABLE_ANB);
*index = (OMX_INDEXTYPE) OMX_IndexParamEnableAndroidBuffers;
return OMX_ErrorNone;
} else......
return OMX_ErrorNotImplemented;
}
可以看出我调用OMX_GetExtensionIndex将获取到index为OMX_IndexParamEnableAndroidBuffers的值。再来看看解码器是怎么实现OMX_SetParameter对应的函数接口的。
OMX_ERRORTYPE SoftFFmpegVideoDec::internalSetParameter( OMX_INDEXTYPE index, const OMX_PTR params){
case OMX_IndexParamEnableAndroidBuffers:
{
EnableAndroidNativeBuffersParams *peanbp = (EnableAndroidNativeBuffersParams *)params;
PortInfo *pOutPort = editPortInfo(kOutputPortIndex);
if (peanbp->enable == OMX_FALSE) {
pOutPort->mDef.format.video.eColorFormat = OMX_COLOR_FormatYUV420Planar;
} else {
ALOGI("internalSetParameter, enable AndroidNativeBuffer but i not suppurt");
return OMX_ErrorUndefined;
}
return OMX_ErrorNone;
}
}
可以看出我对于想要使能AndroidNativeBuffer的操作是拒绝的。我这里直接返回OMX_ErrorUndefined。所以我实现的解码器并不支持把端口模式PortMode设置成kPortModePresetANWBuffer。不过没事下面的步骤会去尝试把PortMode设置成kPortModePresetByteBuffer，这个kPortModePresetByteBuffer我的解码器是支持的。下面的分析会分析到这里的。
来看看第三个配置步骤，这个配置步骤是我自定义的，因为需要从解析器传递一些参数给解码器。
status_t ACodec::configureCodec(const char *mime, const sp<AMessage> &msg) {
if (video) {
if (!strcasecmp(mime, MEDIA_MIMETYPE_CONTAINER_FFMPEG_VIDEO)) {
OMX_AUDIO_PARAM_FFMPEGMODETYPE FfmpegParams;
FfmpegParams.nPortIndex = kPortIndexOutput;
FfmpegParams.nDisplayWidth = displayWidth;
FfmpegParams.nDisplayHeight = displayHeight;
strcpy((char *)FfmpegParams.nName,fileName.c_str());
err = mOMXNode->setParameter(OMX_IndexParamVideoFfmpeg,
&FfmpegParams, sizeof(OMX_AUDIO_PARAM_FFMPEGMODETYPE));
}
}
来看看第四个配置步骤，这里会去设置kPortModePresetByteBuffer，前面设置ANW buffer的场景由于我解码器没有去实现相关代码所以没成功的。
diff --git a/vendor/sprd/modules/media/omx_core/sprd_omx_core.cpp b/vendor/sprd/modules/media/omx_core/sprd_omx_core.cpp
index 6d36606..7b0ab08 100644
--- a/vendor/sprd/modules/media/omx_core/sprd_omx_core.cpp
+++ b/vendor/sprd/modules/media/omx_core/sprd_omx_core.cpp
@@ -58,6 +58,7 @@ static const struct {
const char *mRole;
} kComponents[] = {
+ { "OMX.google.ffmpegVideo.decoder","soft_ffmpegVideodec","video_decoder.ffmpeg"},
{ "OMX.sprd.h263.decoder", "sprd_mpeg4dec", "video_decoder.h263" },
由于我的解码器的名字是OMX.google开头的,所以会走下面这个分支。
status_t ACodec::configureCodec(const char *mime, const sp<AMessage> &msg) {
......
if (video) {
if (haveNativeWindow && mComponentName.startsWith("OMX.google.")) {
usingSwRenderer = true;
haveNativeWindow = false;
(void)setPortMode(kPortIndexOutput, IOMX::kPortModePresetByteBuffer);
}
}
}
status_t ACodec::setPortMode(int32_t portIndex, IOMX::PortMode mode) {
status_t err = mOMXNode->setPortMode(portIndex, mode);
......
}
status_t OMXNodeInstance::setPortMode(OMX_U32 portIndex, IOMX::PortMode mode) {
......
case IOMX::kPortModePresetByteBuffer:
{
// Disable secure buffer, native buffer and metadata.
(void)enableNativeBuffers_l(portIndex, OMX_TRUE /*graphic*/, OMX_FALSE); //实现了
(void)enableNativeBuffers_l(portIndex, OMX_FALSE /*graphic*/, OMX_FALSE); //没实现
(void)storeMetaDataInBuffers_l(portIndex, OMX_FALSE, NULL); //没实现
return OK;
}
}
status_t OMXNodeInstance::enableNativeBuffers_l(OMX_U32 portIndex, OMX_BOOL graphic, OMX_BOOL enable){
OMX_STRING name = const_cast<OMX_STRING>(
graphic ? "OMX.google.android.index.enableAndroidNativeBuffers"
: "OMX.google.android.index.allocateNativeHandle");
OMX_ERRORTYPE err = OMX_GetExtensionIndex(mHandle, name, &index);
if (err == OMX_ErrorNone) {//如果解码器的getExtensionIndex有显示这个index，则对解码器进行这个index的参数设置
EnableAndroidNativeBuffersParams params;
InitOMXParams(¶ms);
params.nPortIndex = portIndex;
params.enable = enable;
err = OMX_SetParameter(mHandle, index, ¶ms);
if (!graphic) {
if (err == OMX_ErrorNone) {
mSecureBufferType[portIndex] =
enable ? kSecureBufferTypeNativeHandle : kSecureBufferTypeOpaque;
} else if (mSecureBufferType[portIndex] == kSecureBufferTypeUnknown) {
mSecureBufferType[portIndex] = kSecureBufferTypeOpaque;
}
} else {
if (err == OMX_ErrorNone) {
mGraphicBufferEnabled[portIndex] = enable; //这里会被设置成false
} else if (enable) {
mGraphicBufferEnabled[portIndex] = false;
}
}
} else {
}
这里有两个enableNativeBuffers_l，第一个对应的OMX.google.android.index.enableAndroidNativeBuffers我解码器有实现，所以会成功；第二个OMX.google.android.index.allocateNativeHandle解码器没实现，所以会失败；第三个storeMetaDataInBuffers_l对应的OMX.google.android.index.storeANWBufferInMetadata我也没有实现，所以会失败。但是这两个失败并不会直接影响解码器的配置，因为他们的返回值都被忽略了。然后mGraphicBufferEnabled[portIndex]会被设置成false，这个变量在配置数据缓冲区的时候会用来作判断。
来看看第五个配置步骤，这个步骤会通过index = OMX_IndexParamVideoPortFormat获取解码器的参数，主要是设置input和ouput的数据格式和颜色空间。对于OMX_IndexParamVideoPortFormat的设置参数的接口，可以只做状态返回。
status_t ACodec::configureCodec(const char *mime, const sp<AMessage> &msg) {
......
if (video) {
if (encoder) {
err = setupVideoEncoder(mime, msg, outputFormat, inputFormat);
} else {//解码器走这里
err = setupVideoDecoder(mime, msg, haveNativeWindow, usingSwRenderer, outputFormat);
}
}
}
status_t ACodec::setupVideoDecoder(
const char *mime, const sp<AMessage> &msg, bool haveNativeWindow,
bool usingSwRenderer, sp<AMessage> &outputFormat) {
OMX_VIDEO_CODINGTYPE compressionFormat;
//从kVideoCodingMapEntry中获取这个mime对应的OMX_VIDEO_CODINGTYPE mVideoCodingType，保存到compressionFormat中
//然后compressionFormat会拿来跟后面从解码器获取出来的eCompressionFormat作对比，如果一样则继续走。
status_t err = GetVideoCodingTypeFromMime(mime, &compressionFormat);
err = setVideoPortFormatType(
kPortIndexInput, compressionFormat, OMX_COLOR_FormatUnused);
if (msg->findInt32("color-format", &tmp)) { //我没有定义这个，所以不会走这里
......
}else{
err = setSupportedOutputFormat(!haveNativeWindow /* getLegacyFlexibleFormat */);
}
......
}
status_t ACodec::setVideoPortFormatType(
OMX_U32 portIndex,
OMX_VIDEO_CODINGTYPE compressionFormat,
OMX_COLOR_FORMATTYPE colorFormat,
bool usingNativeBuffers) {//这个函数在这里被拿来设置kPortIndexInput
for (OMX_U32 index = 0; index <= kMaxIndicesToCheck; ++index) {//我解码器对formatParams->nIndex做了限制，so只会走一次
status_t err = mOMXNode->getParameter(
OMX_IndexParamVideoPortFormat,&format, sizeof(format));
}
status_t err = mOMXNode->setParameter(
OMX_IndexParamVideoPortFormat, &format, sizeof(format));
}
status_t ACodec::setSupportedOutputFormat(bool getLegacyFlexibleFormat) {
OMX_VIDEO_PARAM_PORTFORMATTYPE format, legacyFormat;
InitOMXParams(&format);
format.nPortIndex = kPortIndexOutput; //这个函数是用来设置output的，所以这里强制赋值了
for (OMX_U32 index = 0; ; ++index) {
format.nIndex = index; //解码器只处理nIndex = 0的情况
status_t err = mOMXNode->getParameter(
OMX_IndexParamVideoPortFormat, &format, sizeof(format));
}
return mOMXNode->setParameter(
OMX_IndexParamVideoPortFormat, &format, sizeof(format));
}
OMX_ERRORTYPE SoftFFmpegVideoDec::internalGetParameter(....){
switch (index)
{
case OMX_IndexParamVideoPortFormat:
{
OMX_VIDEO_PARAM_PORTFORMATTYPE *formatParams =
(OMX_VIDEO_PARAM_PORTFORMATTYPE *)params;
if (formatParams->nPortIndex > 1) {
return OMX_ErrorUndefined;
}
if (formatParams->nIndex != 0) {
return OMX_ErrorNoMore;
}
if (formatParams->nPortIndex == 0) {
formatParams->eCompressionFormat = OMX_VIDEO_CodinFFMPEG ;//指定数据的压缩格式
formatParams->eColorFormat = OMX_COLOR_FormatUnused;//对于input不需要设置颜色类型，毕竟是需要编码的数据
formatParams->xFramerate = 0;
} else {
CHECK_EQ(formatParams->nPortIndex, 1u);
PortInfo *pOutPort = editPortInfo(kOutputPortIndex);
formatParams->eCompressionFormat = OMX_VIDEO_CodingUnused;
formatParams->eColorFormat = pOutPort->mDef.format.video.eColorFormat;
formatParams->xFramerate = 0;
}
return OMX_ErrorNone;
}
......
}
}
OMX_ERRORTYPE SoftFFmpegVideoDec::internalSetParameter(...){
case OMX_IndexParamVideoPortFormat:{//对于SetParameter，我没有作什么操作
if (formatParams->nPortIndex > 1){
return OMX_ErrorUndefined;
}
if (formatParams->nIndex != 0){
return OMX_ErrorNoMore;
}
return OMX_ErrorNone;
}
}
来看看第六个步骤，这个步骤是通过index=OMX_IndexParamPortDefinition来配置解码器的
status_t ACodec::setupVideoDecoder(
const char *mime, const sp<AMessage> &msg, bool haveNativeWindow,
bool usingSwRenderer, sp<AMessage> &outputFormat) {
OMX_VIDEO_CODINGTYPE compressionFormat;
//从kVideoCodingMapEntry中获取这个mime对应的OMX_VIDEO_CODINGTYPE mVideoCodingType，保存到compressionFormat中
//然后compressionFormat会拿来跟后面从解码器获取出来的eCompressionFormat作对比，如果一样则继续走。
status_t err = GetVideoCodingTypeFromMime(mime, &compressionFormat);
err = setVideoPortFormatType(
kPortIndexInput, compressionFormat, OMX_COLOR_FormatUnused);
if (msg->findInt32("color-format", &tmp)) { //我没有定义这个，所以不会走这里
......
}else{
err = setSupportedOutputFormat(!haveNativeWindow /* getLegacyFlexibleFormat */);
}
......
err = setVideoFormatOnPort(
kPortIndexInput, width, height, compressionFormat, frameRateFloat);
err = setVideoFormatOnPort(
kPortIndexOutput, width, height, OMX_VIDEO_CodingUnused);
if (err != OK) {
return err;
}
//下面这两个选项是可选的，也就是说
//getExtensionIndex, name: OMX.google.android.index.describeColorAspects
//getExtensionIndex, name: OMX.google.android.index.describeHDRStaticInfo
//解析器可以不实现这两个extension index
err = setColorAspectsForVideoDecoder( width, height, haveNativeWindow | usingSwRenderer, msg, outputFormat);
if (err == ERROR_UNSUPPORTED) { // support is optional
err = OK;
}
err = setHDRStaticInfoForVideoCodec(kPortIndexOutput, msg, outputFormat);
if (err == ERROR_UNSUPPORTED) { // support is optional
err = OK;
}
}
status_t ACodec::setVideoFormatOnPort(
OMX_U32 portIndex,
int32_t width, int32_t height, OMX_VIDEO_CODINGTYPE compressionFormat,
float frameRate) {
status_t err = mOMXNode->getParameter(
OMX_IndexParamPortDefinition, &def, sizeof(def));
video_def->nFrameWidth = width; //把从解析器获取的视频宽度传递给解码器
video_def->nFrameHeight = height; //把从解析器获取的视频高度传递给解码器
err = mOMXNode->setParameter(
OMX_IndexParamPortDefinition, &def, sizeof(def));
}
//由于我的解码器没有实现获取参数的OMX_IndexParamPortDefinition，所以会调用父类的internalGetParameter，
OMX_ERRORTYPE SprdSimpleOMXComponent::internalGetParameter(...){
case OMX_IndexParamPortDefinition:{
OMX_PARAM_PORTDEFINITIONTYPE *defParams = (OMX_PARAM_PORTDEFINITIONTYPE *)params;
const PortInfo *port =&mPorts.itemAt(defParams->nPortIndex);
memcpy(defParams, &port->mDef, sizeof(port->mDef));//把对应port的信息拷贝给defParams，传递给调用者
return OMX_ErrorNone;
}
}
OMX_ERRORTYPE SoftFFmpegVideoDec::internalSetParameter(...){
case OMX_IndexParamPortDefinition:
{
PortInfo *port = editPortInfo(defParams->nPortIndex);
if(defParams->nPortIndex == 1){//仅仅设置output
//对于 setVideoFormatOnPort的调用，我可以不设置mWidth和mHeight，直接使用从解析器传递过来的宽度和高度
//不过这样写也没关系，只不过这样写让setVideoFormatOnPort的OMX_IndexParamPortDefinition操作显得多余而已。嘿嘿嘿
port->mDef.format.video.nFrameWidth = mWidth;
port->mDef.format.video.nFrameHeight = mHeight;
port->mDef.format.video.nStride = mWidth;
port->mDef.format.video.nSliceHeight = mHeight;
port->mDef.nBufferSize =( mWidth*mHeight*3)/2;
}
}
}
然后回到configureCodec函数它会调用getPortFormat获取input和ouput的格式
status_t ACodec::configureCodec(
const char *mime, const sp<AMessage> &msg){
......
err = getPortFormat(kPortIndexInput, inputFormat);
if (err == OK) {
err = getPortFormat(kPortIndexOutput, outputFormat);
if (err == OK) {
mInputFormat = inputFormat; //保存格式信息
mOutputFormat = outputFormat;//保存格式信息
}
}
}
status_t ACodec::getPortFormat(OMX_U32 portIndex, sp<AMessage> ¬ify) {
status_t err = mOMXNode->getParameter(OMX_IndexParamPortDefinition, &def, sizeof(def));
switch (def.eDomain) {
case OMX_PortDomainVideo:{
OMX_VIDEO_PORTDEFINITIONTYPE *videoDef = &def.format.video;
switch ((int)videoDef->eCompressionFormat) {
case OMX_VIDEO_CodingUnused:{
notify->setInt32("stride", videoDef->nStride);
notify->setInt32("color-format", videoDef->eColorFormat);
notify->setInt32("slice-height", videoDef->nSliceHeight);
notify->setInt32("width", videoDef->nFrameWidth);
notify->setInt32("height", videoDef->nFrameHeight);
}
}
}
}
getPortFormat得到的output的格式是这样的，这些格式信息会保存在ACodec的mInputFormat和mOutputFormat中。
27531 27744 V ACodec : [OMX.google.ffmpegVideo.decoder] output format is AMessage(what = 0x00000000) = {
27531 27744 V ACodec : int32_t color-range = 2
27531 27744 V ACodec : int32_t color-standard = 4
27531 27744 V ACodec : int32_t color-transfer = 3
27531 27744 V ACodec : int32_t using-sw-renderer = 1
27531 27744 V ACodec : string mime = "video/raw"
27531 27744 V ACodec : int32_t stride = 624
27531 27744 V ACodec : int32_t slice-height = 352
27531 27744 V ACodec : int32_t color-format = 19
27531 27744 V ACodec : Buffer *image-data = 0xe32950e0
27531 27744 V ACodec : Rect crop(0, 0, 623, 351)
27531 27744 V ACodec : int32_t android._dataspace = 259
27531 27744 V ACodec : int32_t width = 624
27531 27744 V ACodec : int32_t height = 352
27531 27744 V ACodec : }
至此，ACodec::configureCodec已经分析完毕。到了这里你可以会发现没有对input和output的缓冲区进行初始化配置，别急下面就来分析。
既然configureCodec已经执行完毕，现在回到Nuplayer::Decoder 消息处理函数的kWhatConfigure，它的下一步工作就是执行mCodec->start()，从时序图可以看出它会执行到ACodec的LoadedState::onMessageReceived的ACodec::kWhatStart分支。这个分支会通过状态机执行到ACodec::LoadedToIdleState::stateEntered函数
void ACodec::LoadedToIdleState::stateEntered（）{
if ((err = allocateBuffers()) != OK) {
.......
mCodec->changeState(mCodec->mLoadedState);
}
}
status_t ACodec::LoadedToIdleState::allocateBuffers() {
status_t err = mCodec->allocateBuffersOnPort(kPortIndexInput);
err = mCodec->allocateBuffersOnPort(kPortIndexOutput);
}
status_t ACodec::allocateBuffersOnPort(OMX_U32 portIndex) {
......
OMX_PARAM_PORTDEFINITIONTYPE def;
InitOMXParams(&def);
def.nPortIndex = portIndex;
err = mOMXNode->getParameter(
OMX_IndexParamPortDefinition, &def, sizeof(def));
if (err == OK) {
size_t bufSize = def.nBufferSize;
/*这里有段log，通过这个log可以直观地知道现在正在为这4个buffer分配内存
[log] ACodec : [OMX.google.ffmpegVideo.decoder] Allocating 4 buffers of size 329472
(from 329472 using PresetByteBuffer) on output port
*/
for (OMX_U32 i = 0; i < def.nBufferCountActual && err == OK; ++i) {//执行四次,解码器定义的nBufferCountActual
......
auto transStatus = mAllocator[portIndex]->allocate(
bufSize,
[&success, &hidlMemToken](
bool s,
hidl_memory const& m) {
success = s;
hidlMemToken = m;
});//分配内存
hidlMem = mapMemory(hidlMemToken);
err = mOMXNode->useBuffer(
portIndex, hidlMemToken, &info.mBufferID);
}
}
}
status_t OMXNodeInstance::useBuffer(
OMX_U32 portIndex, const OMXBuffer &omxBuffer, IOMX::buffer_id *buffer) {
......
case OMXBuffer::kBufferTypeHidlMemory: {
sp<IHidlMemory> hidlMemory = mapMemory(omxBuffer.mHidlMemory);
return useBuffer_l(portIndex, NULL, hidlMemory, buffer);
}
}
status_t OMXNodeInstance::useBuffer_l(
OMX_U32 portIndex, const sp<IMemory> ¶ms,
const sp<IHidlMemory> &hParams, IOMX::buffer_id *buffer) {
......
if (hParams != NULL) {
paramsPointer = hParams->getPointer();
paramsSize = hParams->getSize();
}
data = static_cast<OMX_U8 *>(paramsPointer);
//这个没有用到，因为我没有实现OMX.google.android.index.storeANWBufferInMetadata
buffer_meta = new BufferMeta(params, hParams, portIndex, false /* copy */, NULL);
err = OMX_UseBuffer( //SprdSimpleOMXComponent::internalUseBuffer
mHandle, &header, portIndex, buffer_meta,
allottedSize, data); //所以解码器的pBuffer是从hidl中拿来的buffer地址 nAllocLen是从hidl中拿来的buffer大小
}
//来看看原厂是怎么实现OMX_USeBuffer的
OMX_ERRORTYPE SprdSimpleOMXComponent::internalUseBuffer(
OMX_BUFFERHEADERTYPE **header,
OMX_U32 portIndex,
OMX_PTR appPrivate,
OMX_U32 size,
OMX_U8 *ptr,
BufferPrivateStruct* bufferPrivate) {
*header = new OMX_BUFFERHEADERTYPE;
(*header)->nSize = sizeof(OMX_BUFFERHEADERTYPE);
(*header)->pBuffer = ptr; //保存从hidl拿来的buffer地址
(*header)->nAllocLen = size; //保存从hidl拿来的buffer大小
(*header)->nFilledLen = 0;
......
(*header)->nOutputPortIndex = portIndex;
(*header)->nInputPortIndex = portIndex;
PortInfo *port = &mPorts.editItemAt(portIndex);
port->mBuffers.push();
BufferInfo *buffer =
&port->mBuffers.editItemAt(port->mBuffers.size() - 1);
ALOGI("internalUseBuffer, header=0x%p, pBuffer=0x%p, size=%d",*header, ptr, size);
buffer->mHeader = *header;
buffer->mOwnedByUs = false;
if (port->mBuffers.size() == port->mDef.nBufferCountActual) {
port->mDef.bPopulated = OMX_TRUE;
checkTransitions();
}
return OMX_ErrorNone;
}
//由于我的解码器没有实现获取参数的OMX_IndexParamPortDefinition，所以会调用父类的internalGetParameter，
OMX_ERRORTYPE SprdSimpleOMXComponent::internalGetParameter(...){
case OMX_IndexParamPortDefinition:{
OMX_PARAM_PORTDEFINITIONTYPE *defParams = (OMX_PARAM_PORTDEFINITIONTYPE *)params;
const PortInfo *port =&mPorts.itemAt(defParams->nPortIndex);
memcpy(defParams, &port->mDef, sizeof(port->mDef));//把对应port的信息拷贝给defParams，传递给调用者
return OMX_ErrorNone;
}
}
可以看出allocateBuffers用户或OMX_IndexParamPortDefinition获取input和output的信息。通过SprdSimpleOMXComponent::internalGetParameter可以知道获取的是对应端口的所有信息，比如buffer大小nBufferSize、buffer个数nBufferCountActual、视频分辨率format.video.nFrameWidth和format.video.nFrameHeight。对于input，获取视频编码格式（压缩格式）eCompressionFormat、对于output获取颜色格式eColorFormat（一般是YUV420p）。然后初始化每一个端口的缓冲区，例如ouput，会分配4（nBufferCountActual）个buffer，把每一个buffer的hidlMemToken传递给omx，进而传递相关地址给解码器，这样解码器就可以通过这个地址获取到解析器解析完之后的数据了。