mozilla
/
gecko-dev
зеркало из https://github.com/mozilla/gecko-dev.git
1
0
Форкнуть 0
Код Задачи Пакеты Проекты Релизы Вики Активность
gecko-dev/dom/media/ogg/OggReader.cpp

2043 строки
70 KiB
C++
Исходник Ответственный История

/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim:set ts=2 sw=2 sts=2 et cindent: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "mozilla/DebugOnly.h"
#include "nsError.h"
#include "MediaDecoderStateMachine.h"
#include "MediaDecoder.h"
#include "OggReader.h"
#include "VideoUtils.h"
#include "theora/theoradec.h"
#include <algorithm>
#include "opus/opus.h"
extern "C" {
#include "opus/opus_multistream.h"
}
#include "mozilla/dom/TimeRanges.h"
#include "mozilla/TimeStamp.h"
#include "VorbisUtils.h"
#include "MediaMetadataManager.h"
#include "nsISeekableStream.h"
#include "gfx2DGlue.h"
using namespace mozilla::gfx;
namespace mozilla {
// On B2G estimate the buffered ranges rather than calculating them explicitly.
// This prevents us doing I/O on the main thread, which is prohibited in B2G.
#ifdef MOZ_WIDGET_GONK
#define OGG_ESTIMATE_BUFFERED 1
#endif
// Un-comment to enable logging of seek bisections.
//#define SEEK_LOGGING
#ifdef PR_LOGGING
extern PRLogModuleInfo* gMediaDecoderLog;
#define LOG(type, msg) PR_LOG(gMediaDecoderLog, type, msg)
#ifdef SEEK_LOGGING
#define SEEK_LOG(type, msg) PR_LOG(gMediaDecoderLog, type, msg)
#else
#define SEEK_LOG(type, msg)
#endif
#else
#define LOG(type, msg)
#define SEEK_LOG(type, msg)
#endif
// The number of microseconds of "fuzz" we use in a bisection search over
// HTTP. When we're seeking with fuzz, we'll stop the search if a bisection
// lands between the seek target and SEEK_FUZZ_USECS microseconds before the
// seek target. This is becaue it's usually quicker to just keep downloading
// from an exisiting connection than to do another bisection inside that
// small range, which would open a new HTTP connetion.
static const uint32_t SEEK_FUZZ_USECS = 500000;
// The number of microseconds of "pre-roll" we use for Opus streams.
// The specification recommends 80 ms.
static const int64_t SEEK_OPUS_PREROLL = 80 * USECS_PER_MS;
enum PageSyncResult {
PAGE_SYNC_ERROR = 1,
PAGE_SYNC_END_OF_RANGE= 2,
PAGE_SYNC_OK = 3
};
// Reads a page from the media resource.
static PageSyncResult
PageSync(MediaResource* aResource,
ogg_sync_state* aState,
bool aCachedDataOnly,
int64_t aOffset,
int64_t aEndOffset,
ogg_page* aPage,
int& aSkippedBytes);
// Chunk size to read when reading Ogg files. Average Ogg page length
// is about 4300 bytes, so we read the file in chunks larger than that.
static const int PAGE_STEP = 8192;
// Return the corresponding category in aKind based on the following specs.
// (https://www.whatwg.org/specs/web-apps/current-
// work/multipage/embedded-content.html#dom-audiotrack-kind) &
// (http://wiki.xiph.org/SkeletonHeaders)
static const nsString GetKind(const nsCString& aRole)
{
if (aRole.Find("audio/main") != -1 || aRole.Find("video/main") != -1) {
return NS_LITERAL_STRING("main");
} else if (aRole.Find("audio/alternate") != -1 ||
aRole.Find("video/alternate") != -1) {
return NS_LITERAL_STRING("alternative");
} else if (aRole.Find("audio/audiodesc") != -1) {
return NS_LITERAL_STRING("descriptions");
} else if (aRole.Find("audio/described") != -1) {
return NS_LITERAL_STRING("main-desc");
} else if (aRole.Find("audio/dub") != -1) {
return NS_LITERAL_STRING("translation");
} else if (aRole.Find("audio/commentary") != -1) {
return NS_LITERAL_STRING("commentary");
} else if (aRole.Find("video/sign") != -1) {
return NS_LITERAL_STRING("sign");
} else if (aRole.Find("video/captioned") != -1) {
return NS_LITERAL_STRING("captions");
} else if (aRole.Find("video/subtitled") != -1) {
return NS_LITERAL_STRING("subtitles");
}
return EmptyString();
}
static void InitTrack(MessageField* aMsgInfo, TrackInfo* aInfo, bool aEnable)
{
MOZ_ASSERT(aMsgInfo);
MOZ_ASSERT(aInfo);
nsCString* sName = aMsgInfo->mValuesStore.Get(eName);
nsCString* sRole = aMsgInfo->mValuesStore.Get(eRole);
nsCString* sTitle = aMsgInfo->mValuesStore.Get(eTitle);
nsCString* sLanguage = aMsgInfo->mValuesStore.Get(eLanguage);
aInfo->Init(sName? NS_ConvertUTF8toUTF16(*sName):EmptyString(),
sRole? GetKind(*sRole):EmptyString(),
sTitle? NS_ConvertUTF8toUTF16(*sTitle):EmptyString(),
sLanguage? NS_ConvertUTF8toUTF16(*sLanguage):EmptyString(),
aEnable);
}
OggReader::OggReader(AbstractMediaDecoder* aDecoder)
: MediaDecoderReader(aDecoder),
mMonitor("OggReader"),
mTheoraState(nullptr),
mVorbisState(nullptr),
mOpusState(nullptr),
mOpusEnabled(MediaDecoder::IsOpusEnabled()),
mSkeletonState(nullptr),
mVorbisSerial(0),
mOpusSerial(0),
mTheoraSerial(0),
mOpusPreSkip(0),
mIsChained(false),
mDecodedAudioFrames(0)
{
MOZ_COUNT_CTOR(OggReader);
memset(&mTheoraInfo, 0, sizeof(mTheoraInfo));
}
OggReader::~OggReader()
{
ogg_sync_clear(&mOggState);
MOZ_COUNT_DTOR(OggReader);
}
nsresult OggReader::Init(MediaDecoderReader* aCloneDonor) {
int ret = ogg_sync_init(&mOggState);
NS_ENSURE_TRUE(ret == 0, NS_ERROR_FAILURE);
return NS_OK;
}
nsresult OggReader::ResetDecode()
{
return ResetDecode(false);
}
nsresult OggReader::ResetDecode(bool start)
{
MOZ_ASSERT(OnTaskQueue());
nsresult res = NS_OK;
if (NS_FAILED(MediaDecoderReader::ResetDecode())) {
res = NS_ERROR_FAILURE;
}
// Discard any previously buffered packets/pages.
ogg_sync_reset(&mOggState);
if (mVorbisState && NS_FAILED(mVorbisState->Reset())) {
res = NS_ERROR_FAILURE;
}
if (mOpusState && NS_FAILED(mOpusState->Reset(start))) {
res = NS_ERROR_FAILURE;
}
if (mTheoraState && NS_FAILED(mTheoraState->Reset())) {
res = NS_ERROR_FAILURE;
}
return res;
}
bool OggReader::ReadHeaders(OggCodecState* aState)
{
while (!aState->DoneReadingHeaders()) {
ogg_packet* packet = NextOggPacket(aState);
// DecodeHeader is responsible for releasing packet.
if (!packet || !aState->DecodeHeader(packet)) {
aState->Deactivate();
return false;
}
}
return aState->Init();
}
void OggReader::BuildSerialList(nsTArray<uint32_t>& aTracks)
{
// Obtaining seek index information for currently active bitstreams.
if (HasVideo()) {
aTracks.AppendElement(mTheoraState->mSerial);
}
if (HasAudio()) {
if (mVorbisState) {
aTracks.AppendElement(mVorbisState->mSerial);
} else if (mOpusState) {
aTracks.AppendElement(mOpusState->mSerial);
}
}
}
void OggReader::SetupTargetTheora(TheoraState* aTheoraState)
{
if (mTheoraState) {
mTheoraState->Reset();
}
nsIntRect picture = nsIntRect(aTheoraState->mInfo.pic_x,
aTheoraState->mInfo.pic_y,
aTheoraState->mInfo.pic_width,
aTheoraState->mInfo.pic_height);
nsIntSize displaySize = nsIntSize(aTheoraState->mInfo.pic_width,
aTheoraState->mInfo.pic_height);
// Apply the aspect ratio to produce the intrinsic display size we report
// to the element.
ScaleDisplayByAspectRatio(displaySize, aTheoraState->mPixelAspectRatio);
nsIntSize frameSize(aTheoraState->mInfo.frame_width,
aTheoraState->mInfo.frame_height);
if (IsValidVideoRegion(frameSize, picture, displaySize)) {
// Video track's frame sizes will not overflow. Activate the video track.
mPicture = picture;
VideoFrameContainer* container = mDecoder->GetVideoFrameContainer();
if (container) {
container->SetCurrentFrame(gfxIntSize(displaySize.width, displaySize.height),
nullptr,
TimeStamp::Now());
}
// Copy Theora info data for time computations on other threads.
memcpy(&mTheoraInfo, &aTheoraState->mInfo, sizeof(mTheoraInfo));
mTheoraState = aTheoraState;
mTheoraSerial = aTheoraState->mSerial;
}
}
void OggReader::SetupTargetVorbis(VorbisState* aVorbisState)
{
if (mVorbisState) {
mVorbisState->Reset();
}
// Copy Vorbis info data for time computations on other threads.
memcpy(&mVorbisInfo, &aVorbisState->mInfo, sizeof(mVorbisInfo));
mVorbisInfo.codec_setup = nullptr;
mVorbisState = aVorbisState;
mVorbisSerial = aVorbisState->mSerial;
}
void OggReader::SetupTargetOpus(OpusState* aOpusState)
{
if (mOpusState) {
mOpusState->Reset();
}
mOpusState = aOpusState;
mOpusSerial = aOpusState->mSerial;
mOpusPreSkip = aOpusState->mPreSkip;
}
void OggReader::SetupTargetSkeleton(SkeletonState* aSkeletonState)
{
// Setup skeleton related information after mVorbisState & mTheroState
// being set (if they exist).
if (aSkeletonState) {
if (!HasAudio() && !HasVideo()) {
// We have a skeleton track, but no audio or video, may as well disable
// the skeleton, we can't do anything useful with this media.
aSkeletonState->Deactivate();
} else if (ReadHeaders(aSkeletonState) && aSkeletonState->HasIndex()) {
// Extract the duration info out of the index, so we don't need to seek to
// the end of resource to get it.
nsAutoTArray<uint32_t, 2> tracks;
BuildSerialList(tracks);
int64_t duration = 0;
if (NS_SUCCEEDED(aSkeletonState->GetDuration(tracks, duration))) {
ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor());
mDecoder->SetMediaDuration(duration);
LOG(PR_LOG_DEBUG, ("Got duration from Skeleton index %lld", duration));
}
}
}
}
void OggReader::SetupMediaTracksInfo(const nsTArray<uint32_t>& aSerials)
{
// For each serial number
// 1. Retrieve a codecState from mCodecStore by this serial number.
// 2. Retrieve a message field from mMsgFieldStore by this serial number.
// 3. For now, skip if the serial number refers to a non-primary bitstream.
// 4. Setup track and other audio/video related information per different types.
for (size_t i = 0; i < aSerials.Length(); i++) {
uint32_t serial = aSerials[i];
OggCodecState* codecState = mCodecStore.Get(serial);
MessageField* msgInfo = nullptr;
if (mSkeletonState && mSkeletonState->mMsgFieldStore.Contains(serial)) {
mSkeletonState->mMsgFieldStore.Get(serial, &msgInfo);
}
if (codecState->GetType() == OggCodecState::TYPE_THEORA) {
TheoraState* theoraState = static_cast<TheoraState*>(codecState);
if (!(mTheoraState && mTheoraState->mSerial == theoraState->mSerial)) {
continue;
}
if (msgInfo) {
InitTrack(msgInfo, &mInfo.mVideo.mTrackInfo, mTheoraState == theoraState);
}
nsIntRect picture = nsIntRect(theoraState->mInfo.pic_x,
theoraState->mInfo.pic_y,
theoraState->mInfo.pic_width,
theoraState->mInfo.pic_height);
nsIntSize displaySize = nsIntSize(theoraState->mInfo.pic_width,
theoraState->mInfo.pic_height);
nsIntSize frameSize(theoraState->mInfo.frame_width,
theoraState->mInfo.frame_height);
ScaleDisplayByAspectRatio(displaySize, theoraState->mPixelAspectRatio);
mInfo.mVideo.mDisplay = displaySize;
mInfo.mVideo.mHasVideo = IsValidVideoRegion(frameSize, picture, displaySize)? true:false;
} else if (codecState->GetType() == OggCodecState::TYPE_VORBIS) {
VorbisState* vorbisState = static_cast<VorbisState*>(codecState);
if (!(mVorbisState && mVorbisState->mSerial == vorbisState->mSerial)) {
continue;
}
if (msgInfo) {
InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, mVorbisState == vorbisState);
}
mInfo.mAudio.mHasAudio = true;
mInfo.mAudio.mRate = vorbisState->mInfo.rate;
mInfo.mAudio.mChannels = vorbisState->mInfo.channels;
} else if (codecState->GetType() == OggCodecState::TYPE_OPUS) {
OpusState* opusState = static_cast<OpusState*>(codecState);
if (!(mOpusState && mOpusState->mSerial == opusState->mSerial)) {
continue;
}
if (msgInfo) {
InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, mOpusState == opusState);
}
mInfo.mAudio.mHasAudio = true;
mInfo.mAudio.mRate = opusState->mRate;
mInfo.mAudio.mChannels = opusState->mChannels;
}
}
}
nsresult OggReader::ReadMetadata(MediaInfo* aInfo,
MetadataTags** aTags)
{
MOZ_ASSERT(OnTaskQueue());
// We read packets until all bitstreams have read all their header packets.
// We record the offset of the first non-header page so that we know
// what page to seek to when seeking to the media start.
NS_ASSERTION(aTags, "Called with null MetadataTags**.");
*aTags = nullptr;
ogg_page page;
nsAutoTArray<OggCodecState*,4> bitstreams;
nsTArray<uint32_t> serials;
bool readAllBOS = false;
while (!readAllBOS) {
if (!ReadOggPage(&page)) {
// Some kind of error...
break;
}
int serial = ogg_page_serialno(&page);
OggCodecState* codecState = 0;
if (!ogg_page_bos(&page)) {
// We've encountered a non Beginning Of Stream page. No more BOS pages
// can follow in this Ogg segment, so there will be no other bitstreams
// in the Ogg (unless it's invalid).
readAllBOS = true;
} else if (!mCodecStore.Contains(serial)) {
// We've not encountered a stream with this serial number before. Create
// an OggCodecState to demux it, and map that to the OggCodecState
// in mCodecStates.
codecState = OggCodecState::Create(&page);
mCodecStore.Add(serial, codecState);
bitstreams.AppendElement(codecState);
serials.AppendElement(serial);
}
codecState = mCodecStore.Get(serial);
NS_ENSURE_TRUE(codecState != nullptr, NS_ERROR_FAILURE);
if (NS_FAILED(codecState->PageIn(&page))) {
return NS_ERROR_FAILURE;
}
}
// We've read all BOS pages, so we know the streams contained in the media.
// 1. Process all available header packets in the Theora, Vorbis/Opus bitstreams.
// 2. Find the first encountered Theora/Vorbis/Opus bitstream, and configure
// it as the target A/V bitstream.
// 3. Deactivate the rest of bitstreams for now, until we have MediaInfo
// support multiple track infos.
for (uint32_t i = 0; i < bitstreams.Length(); ++i) {
OggCodecState* s = bitstreams[i];
if (s) {
if (s->GetType() == OggCodecState::TYPE_THEORA && ReadHeaders(s)) {
if (!mTheoraState) {
TheoraState* theoraState = static_cast<TheoraState*>(s);
SetupTargetTheora(theoraState);
} else {
s->Deactivate();
}
} else if (s->GetType() == OggCodecState::TYPE_VORBIS && ReadHeaders(s)) {
if (!mVorbisState) {
VorbisState* vorbisState = static_cast<VorbisState*>(s);
SetupTargetVorbis(vorbisState);
*aTags = vorbisState->GetTags();
} else {
s->Deactivate();
}
} else if (s->GetType() == OggCodecState::TYPE_OPUS && ReadHeaders(s)) {
if (mOpusEnabled) {
if (!mOpusState) {
OpusState* opusState = static_cast<OpusState*>(s);
SetupTargetOpus(opusState);
*aTags = opusState->GetTags();
} else {
s->Deactivate();
}
} else {
NS_WARNING("Opus decoding disabled."
" See media.opus.enabled in about:config");
}
} else if (s->GetType() == OggCodecState::TYPE_SKELETON && !mSkeletonState) {
mSkeletonState = static_cast<SkeletonState*>(s);
} else {
// Deactivate any non-primary bitstreams.
s->Deactivate();
}
}
}
SetupTargetSkeleton(mSkeletonState);
SetupMediaTracksInfo(serials);
if (HasAudio() || HasVideo()) {
ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor());
MediaResource* resource = mDecoder->GetResource();
if (mDecoder->GetMediaDuration() == -1 &&
!mDecoder->IsShutdown() &&
resource->GetLength() >= 0 &&
mDecoder->IsMediaSeekable())
{
// We didn't get a duration from the index or a Content-Duration header.
// Seek to the end of file to find the end time.
int64_t length = resource->GetLength();
NS_ASSERTION(length > 0, "Must have a content length to get end time");
int64_t endTime = 0;
{
ReentrantMonitorAutoExit exitMon(mDecoder->GetReentrantMonitor());
endTime = RangeEndTime(length);
}
if (endTime != -1) {
mDecoder->SetMediaEndTime(endTime);
LOG(PR_LOG_DEBUG, ("Got Ogg duration from seeking to end %lld", endTime));
}
}
} else {
return NS_ERROR_FAILURE;
}
*aInfo = mInfo;
return NS_OK;
}
bool
OggReader::IsMediaSeekable()
{
if (mIsChained) {
return false;
}
return true;
}
nsresult OggReader::DecodeVorbis(ogg_packet* aPacket) {
NS_ASSERTION(aPacket->granulepos != -1, "Must know vorbis granulepos!");
if (vorbis_synthesis(&mVorbisState->mBlock, aPacket) != 0) {
return NS_ERROR_FAILURE;
}
if (vorbis_synthesis_blockin(&mVorbisState->mDsp,
&mVorbisState->mBlock) != 0)
{
return NS_ERROR_FAILURE;
}
VorbisPCMValue** pcm = 0;
int32_t frames = 0;
uint32_t channels = mVorbisState->mInfo.channels;
ogg_int64_t endFrame = aPacket->granulepos;
while ((frames = vorbis_synthesis_pcmout(&mVorbisState->mDsp, &pcm)) > 0) {
mVorbisState->ValidateVorbisPacketSamples(aPacket, frames);
nsAutoArrayPtr<AudioDataValue> buffer(new AudioDataValue[frames * channels]);
for (uint32_t j = 0; j < channels; ++j) {
VorbisPCMValue* channel = pcm[j];
for (uint32_t i = 0; i < uint32_t(frames); ++i) {
buffer[i*channels + j] = MOZ_CONVERT_VORBIS_SAMPLE(channel[i]);
}
}
// No channel mapping for more than 8 channels.
if (channels > 8) {
return NS_ERROR_FAILURE;
}
int64_t duration = mVorbisState->Time((int64_t)frames);
int64_t startTime = mVorbisState->Time(endFrame - frames);
mAudioQueue.Push(new AudioData(mDecoder->GetResource()->Tell(),
startTime,
duration,
frames,
buffer.forget(),
channels,
mVorbisState->mInfo.rate));
mDecodedAudioFrames += frames;
endFrame -= frames;
if (vorbis_synthesis_read(&mVorbisState->mDsp, frames) != 0) {
return NS_ERROR_FAILURE;
}
}
return NS_OK;
}
nsresult OggReader::DecodeOpus(ogg_packet* aPacket) {
NS_ASSERTION(aPacket->granulepos != -1, "Must know opus granulepos!");
// Maximum value is 63*2880, so there's no chance of overflow.
int32_t frames_number = opus_packet_get_nb_frames(aPacket->packet,
aPacket->bytes);
if (frames_number <= 0)
return NS_ERROR_FAILURE; // Invalid packet header.
int32_t samples = opus_packet_get_samples_per_frame(aPacket->packet,
(opus_int32) mOpusState->mRate);
int32_t frames = frames_number*samples;
// A valid Opus packet must be between 2.5 and 120 ms long.
if (frames < 120 || frames > 5760)
return NS_ERROR_FAILURE;
uint32_t channels = mOpusState->mChannels;
nsAutoArrayPtr<AudioDataValue> buffer(new AudioDataValue[frames * channels]);
// Decode to the appropriate sample type.
#ifdef MOZ_SAMPLE_TYPE_FLOAT32
int ret = opus_multistream_decode_float(mOpusState->mDecoder,
aPacket->packet, aPacket->bytes,
buffer, frames, false);
#else
int ret = opus_multistream_decode(mOpusState->mDecoder,
aPacket->packet, aPacket->bytes,
buffer, frames, false);
#endif
if (ret < 0)
return NS_ERROR_FAILURE;
NS_ASSERTION(ret == frames, "Opus decoded too few audio samples");
int64_t endFrame = aPacket->granulepos;
int64_t startFrame;
// If this is the last packet, perform end trimming.
if (aPacket->e_o_s && mOpusState->mPrevPacketGranulepos != -1) {
startFrame = mOpusState->mPrevPacketGranulepos;
frames = static_cast<int32_t>(std::max(static_cast<int64_t>(0),
std::min(endFrame - startFrame,
static_cast<int64_t>(frames))));
} else {
startFrame = endFrame - frames;
}
// Trim the initial frames while the decoder is settling.
if (mOpusState->mSkip > 0) {
int32_t skipFrames = std::min(mOpusState->mSkip, frames);
if (skipFrames == frames) {
// discard the whole packet
mOpusState->mSkip -= frames;
LOG(PR_LOG_DEBUG, ("Opus decoder skipping %d frames"
" (whole packet)", frames));
return NS_OK;
}
int32_t keepFrames = frames - skipFrames;
int samples = keepFrames * channels;
nsAutoArrayPtr<AudioDataValue> trimBuffer(new AudioDataValue[samples]);
for (int i = 0; i < samples; i++)
trimBuffer[i] = buffer[skipFrames*channels + i];
startFrame = endFrame - keepFrames;
frames = keepFrames;
buffer = trimBuffer;
mOpusState->mSkip -= skipFrames;
LOG(PR_LOG_DEBUG, ("Opus decoder skipping %d frames", skipFrames));
}
// Save this packet's granule position in case we need to perform end
// trimming on the next packet.
mOpusState->mPrevPacketGranulepos = endFrame;
// Apply the header gain if one was specified.
#ifdef MOZ_SAMPLE_TYPE_FLOAT32
if (mOpusState->mGain != 1.0f) {
float gain = mOpusState->mGain;
int samples = frames * channels;
for (int i = 0; i < samples; i++) {
buffer[i] *= gain;
}
}
#else
if (mOpusState->mGain_Q16 != 65536) {
int64_t gain_Q16 = mOpusState->mGain_Q16;
int samples = frames * channels;
for (int i = 0; i < samples; i++) {
int32_t val = static_cast<int32_t>((gain_Q16*buffer[i] + 32768)>>16);
buffer[i] = static_cast<AudioDataValue>(MOZ_CLIP_TO_15(val));
}
}
#endif
// No channel mapping for more than 8 channels.
if (channels > 8) {
return NS_ERROR_FAILURE;
}
LOG(PR_LOG_DEBUG, ("Opus decoder pushing %d frames", frames));
int64_t startTime = mOpusState->Time(startFrame);
int64_t endTime = mOpusState->Time(endFrame);
mAudioQueue.Push(new AudioData(mDecoder->GetResource()->Tell(),
startTime,
endTime - startTime,
frames,
buffer.forget(),
channels,
mOpusState->mRate));
mDecodedAudioFrames += frames;
return NS_OK;
}
bool OggReader::DecodeAudioData()
{
MOZ_ASSERT(OnTaskQueue());
DebugOnly<bool> haveCodecState = mVorbisState != nullptr ||
mOpusState != nullptr;
NS_ASSERTION(haveCodecState, "Need audio codec state to decode audio");
// Read the next data packet. Skip any non-data packets we encounter.
ogg_packet* packet = 0;
OggCodecState* codecState;
if (mVorbisState)
codecState = static_cast<OggCodecState*>(mVorbisState);
else
codecState = static_cast<OggCodecState*>(mOpusState);
do {
if (packet) {
OggCodecState::ReleasePacket(packet);
}
packet = NextOggPacket(codecState);
} while (packet && codecState->IsHeader(packet));
if (!packet) {
return false;
}
NS_ASSERTION(packet && packet->granulepos != -1,
"Must have packet with known granulepos");
nsAutoRef<ogg_packet> autoRelease(packet);
if (mVorbisState) {
DecodeVorbis(packet);
} else if (mOpusState) {
DecodeOpus(packet);
}
if ((packet->e_o_s) && (!ReadOggChain())) {
// We've encountered an end of bitstream packet, or we've hit the end of
// file while trying to decode, so inform the audio queue that there'll
// be no more samples.
return false;
}
return true;
}
void OggReader::SetChained(bool aIsChained) {
{
ReentrantMonitorAutoEnter mon(mMonitor);
mIsChained = aIsChained;
}
{
ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor());
mDecoder->SetMediaSeekable(false);
}
}
bool OggReader::ReadOggChain()
{
bool chained = false;
OpusState* newOpusState = nullptr;
VorbisState* newVorbisState = nullptr;
nsAutoPtr<MetadataTags> tags;
if (HasVideo() || HasSkeleton() || !HasAudio()) {
return false;
}
ogg_page page;
if (!ReadOggPage(&page) || !ogg_page_bos(&page)) {
return false;
}
int serial = ogg_page_serialno(&page);
if (mCodecStore.Contains(serial)) {
return false;
}
nsAutoPtr<OggCodecState> codecState;
codecState = OggCodecState::Create(&page);
if (!codecState) {
return false;
}
if (mVorbisState && (codecState->GetType() == OggCodecState::TYPE_VORBIS)) {
newVorbisState = static_cast<VorbisState*>(codecState.get());
}
else if (mOpusState && (codecState->GetType() == OggCodecState::TYPE_OPUS)) {
newOpusState = static_cast<OpusState*>(codecState.get());
}
else {
return false;
}
OggCodecState* state;
mCodecStore.Add(serial, codecState.forget());
state = mCodecStore.Get(serial);
NS_ENSURE_TRUE(state != nullptr, false);
if (NS_FAILED(state->PageIn(&page))) {
return false;
}
MessageField* msgInfo = nullptr;
if (mSkeletonState && mSkeletonState->mMsgFieldStore.Contains(serial)) {
mSkeletonState->mMsgFieldStore.Get(serial, &msgInfo);
}
if ((newVorbisState && ReadHeaders(newVorbisState)) &&
(mVorbisState->mInfo.rate == newVorbisState->mInfo.rate) &&
(mVorbisState->mInfo.channels == newVorbisState->mInfo.channels)) {
SetupTargetVorbis(newVorbisState);
LOG(PR_LOG_DEBUG, ("New vorbis ogg link, serial=%d\n", mVorbisSerial));
if (msgInfo) {
InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, true);
}
mInfo.mAudio.mRate = newVorbisState->mInfo.rate;
mInfo.mAudio.mChannels = newVorbisState->mInfo.channels;
chained = true;
tags = newVorbisState->GetTags();
}
if ((newOpusState && ReadHeaders(newOpusState)) &&
(mOpusState->mRate == newOpusState->mRate) &&
(mOpusState->mChannels == newOpusState->mChannels)) {
SetupTargetOpus(newOpusState);
if (msgInfo) {
InitTrack(msgInfo, &mInfo.mAudio.mTrackInfo, true);
}
mInfo.mAudio.mRate = newOpusState->mRate;
mInfo.mAudio.mChannels = newOpusState->mChannels;
chained = true;
tags = newOpusState->GetTags();
}
if (chained) {
SetChained(true);
{
mInfo.mAudio.mHasAudio = HasAudio();
mInfo.mVideo.mHasVideo = HasVideo();
nsAutoPtr<MediaInfo> info(new MediaInfo());
*info = mInfo;
ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor());
mDecoder->QueueMetadata((mDecodedAudioFrames * USECS_PER_S) / mInfo.mAudio.mRate,
info, tags);
}
return true;
}
return false;
}
nsresult OggReader::DecodeTheora(ogg_packet* aPacket, int64_t aTimeThreshold)
{
NS_ASSERTION(aPacket->granulepos >= TheoraVersion(&mTheoraState->mInfo,3,2,1),
"Packets must have valid granulepos and packetno");
int ret = th_decode_packetin(mTheoraState->mCtx, aPacket, 0);
if (ret != 0 && ret != TH_DUPFRAME) {
return NS_ERROR_FAILURE;
}
int64_t time = mTheoraState->StartTime(aPacket->granulepos);
// Don't use the frame if it's outside the bounds of the presentation
// start time in the skeleton track. Note we still must submit the frame
// to the decoder (via th_decode_packetin), as the frames which are
// presentable may depend on this frame's data.
if (mSkeletonState && !mSkeletonState->IsPresentable(time)) {
return NS_OK;
}
int64_t endTime = mTheoraState->Time(aPacket->granulepos);
if (endTime < aTimeThreshold) {
// The end time of this frame is already before the current playback
// position. It will never be displayed, don't bother enqueing it.
return NS_OK;
}
if (ret == TH_DUPFRAME) {
nsRefPtr<VideoData> v = VideoData::CreateDuplicate(mDecoder->GetResource()->Tell(),
time,
endTime - time,
aPacket->granulepos);
mVideoQueue.Push(v);
} else if (ret == 0) {
th_ycbcr_buffer buffer;
ret = th_decode_ycbcr_out(mTheoraState->mCtx, buffer);
NS_ASSERTION(ret == 0, "th_decode_ycbcr_out failed");
bool isKeyframe = th_packet_iskeyframe(aPacket) == 1;
VideoData::YCbCrBuffer b;
for (uint32_t i=0; i < 3; ++i) {
b.mPlanes[i].mData = buffer[i].data;
b.mPlanes[i].mHeight = buffer[i].height;
b.mPlanes[i].mWidth = buffer[i].width;
b.mPlanes[i].mStride = buffer[i].stride;
b.mPlanes[i].mOffset = b.mPlanes[i].mSkip = 0;
}
nsRefPtr<VideoData> v = VideoData::Create(mInfo.mVideo,
mDecoder->GetImageContainer(),
mDecoder->GetResource()->Tell(),
time,
endTime - time,
b,
isKeyframe,
aPacket->granulepos,
ToIntRect(mPicture));
if (!v) {
// There may be other reasons for this error, but for
// simplicity just assume the worst case: out of memory.
NS_WARNING("Failed to allocate memory for video frame");
return NS_ERROR_OUT_OF_MEMORY;
}
mVideoQueue.Push(v);
}
return NS_OK;
}
bool OggReader::DecodeVideoFrame(bool &aKeyframeSkip,
int64_t aTimeThreshold)
{
MOZ_ASSERT(OnTaskQueue());
// Record number of frames decoded and parsed. Automatically update the
// stats counters using the AutoNotifyDecoded stack-based class.
AbstractMediaDecoder::AutoNotifyDecoded a(mDecoder);
// Read the next data packet. Skip any non-data packets we encounter.
ogg_packet* packet = 0;
do {
if (packet) {
OggCodecState::ReleasePacket(packet);
}
packet = NextOggPacket(mTheoraState);
} while (packet && mTheoraState->IsHeader(packet));
if (!packet) {
return false;
}
nsAutoRef<ogg_packet> autoRelease(packet);
a.mParsed++;
NS_ASSERTION(packet && packet->granulepos != -1,
"Must know first packet's granulepos");
bool eos = packet->e_o_s;
int64_t frameEndTime = mTheoraState->Time(packet->granulepos);
if (!aKeyframeSkip ||
(th_packet_iskeyframe(packet) && frameEndTime >= aTimeThreshold))
{
aKeyframeSkip = false;
nsresult res = DecodeTheora(packet, aTimeThreshold);
a.mDecoded++;
if (NS_FAILED(res)) {
return false;
}
}
if (eos) {
// We've encountered an end of bitstream packet. Inform the queue that
// there will be no more frames.
return false;
}
return true;
}
bool OggReader::ReadOggPage(ogg_page* aPage)
{
MOZ_ASSERT(OnTaskQueue());
int ret = 0;
while((ret = ogg_sync_pageseek(&mOggState, aPage)) <= 0) {
if (ret < 0) {
// Lost page sync, have to skip up to next page.
continue;
}
// Returns a buffer that can be written too
// with the given size. This buffer is stored
// in the ogg synchronisation structure.
char* buffer = ogg_sync_buffer(&mOggState, 4096);
NS_ASSERTION(buffer, "ogg_sync_buffer failed");
// Read from the resource into the buffer
uint32_t bytesRead = 0;
nsresult rv = mDecoder->GetResource()->Read(buffer, 4096, &bytesRead);
if (NS_FAILED(rv) || (bytesRead == 0 && ret == 0)) {
// End of file.
return false;
}
// Update the synchronisation layer with the number
// of bytes written to the buffer
ret = ogg_sync_wrote(&mOggState, bytesRead);
NS_ENSURE_TRUE(ret == 0, false);
}
return true;
}
ogg_packet* OggReader::NextOggPacket(OggCodecState* aCodecState)
{
MOZ_ASSERT(OnTaskQueue());
if (!aCodecState || !aCodecState->mActive) {
return nullptr;
}
ogg_packet* packet;
while ((packet = aCodecState->PacketOut()) == nullptr) {
// The codec state does not have any buffered pages, so try to read another
// page from the channel.
ogg_page page;
if (!ReadOggPage(&page)) {
return nullptr;
}
uint32_t serial = ogg_page_serialno(&page);
OggCodecState* codecState = nullptr;
codecState = mCodecStore.Get(serial);
if (codecState && NS_FAILED(codecState->PageIn(&page))) {
return nullptr;
}
}
return packet;
}
// Returns an ogg page's checksum.
static ogg_uint32_t
GetChecksum(ogg_page* page)
{
if (page == 0 || page->header == 0 || page->header_len < 25) {
return 0;
}
const unsigned char* p = page->header + 22;
uint32_t c = p[0] +
(p[1] << 8) +
(p[2] << 16) +
(p[3] << 24);
return c;
}
int64_t OggReader::RangeStartTime(int64_t aOffset)
{
MOZ_ASSERT(OnTaskQueue());
MediaResource* resource = mDecoder->GetResource();
NS_ENSURE_TRUE(resource != nullptr, 0);
nsresult res = resource->Seek(nsISeekableStream::NS_SEEK_SET, aOffset);
NS_ENSURE_SUCCESS(res, 0);
int64_t startTime = 0;
FindStartTime(startTime);
return startTime;
}
struct nsAutoOggSyncState {
nsAutoOggSyncState() {
ogg_sync_init(&mState);
}
~nsAutoOggSyncState() {
ogg_sync_clear(&mState);
}
ogg_sync_state mState;
};
int64_t OggReader::RangeEndTime(int64_t aEndOffset)
{
MOZ_ASSERT(OnTaskQueue() || mDecoder->OnStateMachineTaskQueue());
MediaResource* resource = mDecoder->GetResource();
NS_ENSURE_TRUE(resource != nullptr, -1);
int64_t position = resource->Tell();
int64_t endTime = RangeEndTime(0, aEndOffset, false);
nsresult res = resource->Seek(nsISeekableStream::NS_SEEK_SET, position);
NS_ENSURE_SUCCESS(res, -1);
return endTime;
}
int64_t OggReader::RangeEndTime(int64_t aStartOffset,
int64_t aEndOffset,
bool aCachedDataOnly)
{
MediaResource* resource = mDecoder->GetResource();
nsAutoOggSyncState sync;
// We need to find the last page which ends before aEndOffset that
// has a granulepos that we can convert to a timestamp. We do this by
// backing off from aEndOffset until we encounter a page on which we can
// interpret the granulepos. If while backing off we encounter a page which
// we've previously encountered before, we'll either backoff again if we
// haven't found an end time yet, or return the last end time found.
const int step = 5000;
const int maxOggPageSize = 65306;
int64_t readStartOffset = aEndOffset;
int64_t readLimitOffset = aEndOffset;
int64_t readHead = aEndOffset;
int64_t endTime = -1;
uint32_t checksumAfterSeek = 0;
uint32_t prevChecksumAfterSeek = 0;
bool mustBackOff = false;
while (true) {
ogg_page page;
int ret = ogg_sync_pageseek(&sync.mState, &page);
if (ret == 0) {
// We need more data if we've not encountered a page we've seen before,
// or we've read to the end of file.
if (mustBackOff || readHead == aEndOffset || readHead == aStartOffset) {
if (endTime != -1 || readStartOffset == 0) {
// We have encountered a page before, or we're at the end of file.
break;
}
mustBackOff = false;
prevChecksumAfterSeek = checksumAfterSeek;
checksumAfterSeek = 0;
ogg_sync_reset(&sync.mState);
readStartOffset = std::max(static_cast<int64_t>(0), readStartOffset - step);
// There's no point reading more than the maximum size of
// an Ogg page into data we've previously scanned. Any data
// between readLimitOffset and aEndOffset must be garbage
// and we can ignore it thereafter.
readLimitOffset = std::min(readLimitOffset,
readStartOffset + maxOggPageSize);
readHead = std::max(aStartOffset, readStartOffset);
}
int64_t limit = std::min(static_cast<int64_t>(UINT32_MAX),
aEndOffset - readHead);
limit = std::max(static_cast<int64_t>(0), limit);
limit = std::min(limit, static_cast<int64_t>(step));
uint32_t bytesToRead = static_cast<uint32_t>(limit);
uint32_t bytesRead = 0;
char* buffer = ogg_sync_buffer(&sync.mState, bytesToRead);
NS_ASSERTION(buffer, "Must have buffer");
nsresult res;
if (aCachedDataOnly) {
res = resource->ReadFromCache(buffer, readHead, bytesToRead);
NS_ENSURE_SUCCESS(res, -1);
bytesRead = bytesToRead;
} else {
NS_ASSERTION(readHead < aEndOffset,
"resource pos must be before range end");
res = resource->Seek(nsISeekableStream::NS_SEEK_SET, readHead);
NS_ENSURE_SUCCESS(res, -1);
res = resource->Read(buffer, bytesToRead, &bytesRead);
NS_ENSURE_SUCCESS(res, -1);
}
readHead += bytesRead;
if (readHead > readLimitOffset) {
mustBackOff = true;
}
// Update the synchronisation layer with the number
// of bytes written to the buffer
ret = ogg_sync_wrote(&sync.mState, bytesRead);
if (ret != 0) {
endTime = -1;
break;
}
continue;
}
if (ret < 0 || ogg_page_granulepos(&page) < 0) {
continue;
}
uint32_t checksum = GetChecksum(&page);
if (checksumAfterSeek == 0) {
// This is the first page we've decoded after a backoff/seek. Remember
// the page checksum. If we backoff further and encounter this page
// again, we'll know that we won't find a page with an end time after
// this one, so we'll know to back off again.
checksumAfterSeek = checksum;
}
if (checksum == prevChecksumAfterSeek) {
// This page has the same checksum as the first page we encountered
// after the last backoff/seek. Since we've already scanned after this
// page and failed to find an end time, we may as well backoff again and
// try to find an end time from an earlier page.
mustBackOff = true;
continue;
}
int64_t granulepos = ogg_page_granulepos(&page);
int serial = ogg_page_serialno(&page);
OggCodecState* codecState = nullptr;
codecState = mCodecStore.Get(serial);
if (!codecState) {
// This page is from a bitstream which we haven't encountered yet.
// It's probably from a new "link" in a "chained" ogg. Don't
// bother even trying to find a duration...
SetChained(true);
endTime = -1;
break;
}
int64_t t = codecState->Time(granulepos);
if (t != -1) {
endTime = t;
}
}
return endTime;
}
nsresult OggReader::GetSeekRanges(nsTArray<SeekRange>& aRanges)
{
MOZ_ASSERT(OnTaskQueue());
AutoPinned<MediaResource> resource(mDecoder->GetResource());
nsTArray<MediaByteRange> cached;
nsresult res = resource->GetCachedRanges(cached);
NS_ENSURE_SUCCESS(res, res);
for (uint32_t index = 0; index < cached.Length(); index++) {
MediaByteRange& range = cached[index];
int64_t startTime = -1;
int64_t endTime = -1;
if (NS_FAILED(ResetDecode())) {
return NS_ERROR_FAILURE;
}
int64_t startOffset = range.mStart;
int64_t endOffset = range.mEnd;
startTime = RangeStartTime(startOffset);
if (startTime != -1 &&
((endTime = RangeEndTime(endOffset)) != -1))
{
NS_WARN_IF_FALSE(startTime < endTime,
"Start time must be before end time");
aRanges.AppendElement(SeekRange(startOffset,
endOffset,
startTime,
endTime));
}
}
if (NS_FAILED(ResetDecode())) {
return NS_ERROR_FAILURE;
}
return NS_OK;
}
OggReader::SeekRange
OggReader::SelectSeekRange(const nsTArray<SeekRange>& ranges,
int64_t aTarget,
int64_t aStartTime,
int64_t aEndTime,
bool aExact)
{
MOZ_ASSERT(OnTaskQueue());
int64_t so = 0;
int64_t eo = mDecoder->GetResource()->GetLength();
int64_t st = aStartTime;
int64_t et = aEndTime;
for (uint32_t i = 0; i < ranges.Length(); i++) {
const SeekRange &r = ranges[i];
if (r.mTimeStart < aTarget) {
so = r.mOffsetStart;
st = r.mTimeStart;
}
if (r.mTimeEnd >= aTarget && r.mTimeEnd < et) {
eo = r.mOffsetEnd;
et = r.mTimeEnd;
}
if (r.mTimeStart < aTarget && aTarget <= r.mTimeEnd) {
// Target lies exactly in this range.
return ranges[i];
}
}
if (aExact || eo == -1) {
return SeekRange();
}
return SeekRange(so, eo, st, et);
}
OggReader::IndexedSeekResult OggReader::RollbackIndexedSeek(int64_t aOffset)
{
if (mSkeletonState) {
mSkeletonState->Deactivate();
}
MediaResource* resource = mDecoder->GetResource();
NS_ENSURE_TRUE(resource != nullptr, SEEK_FATAL_ERROR);
nsresult res = resource->Seek(nsISeekableStream::NS_SEEK_SET, aOffset);
NS_ENSURE_SUCCESS(res, SEEK_FATAL_ERROR);
return SEEK_INDEX_FAIL;
}
OggReader::IndexedSeekResult OggReader::SeekToKeyframeUsingIndex(int64_t aTarget)
{
MediaResource* resource = mDecoder->GetResource();
NS_ENSURE_TRUE(resource != nullptr, SEEK_FATAL_ERROR);
if (!HasSkeleton() || !mSkeletonState->HasIndex()) {
return SEEK_INDEX_FAIL;
}
// We have an index from the Skeleton track, try to use it to seek.
nsAutoTArray<uint32_t, 2> tracks;
BuildSerialList(tracks);
SkeletonState::nsSeekTarget keyframe;
if (NS_FAILED(mSkeletonState->IndexedSeekTarget(aTarget,
tracks,
keyframe)))
{
// Could not locate a keypoint for the target in the index.
return SEEK_INDEX_FAIL;
}
// Remember original resource read cursor position so we can rollback on failure.
int64_t tell = resource->Tell();
// Seek to the keypoint returned by the index.
if (keyframe.mKeyPoint.mOffset > resource->GetLength() ||
keyframe.mKeyPoint.mOffset < 0)
{
// Index must be invalid.
return RollbackIndexedSeek(tell);
}
LOG(PR_LOG_DEBUG, ("Seeking using index to keyframe at offset %lld\n",
keyframe.mKeyPoint.mOffset));
nsresult res = resource->Seek(nsISeekableStream::NS_SEEK_SET,
keyframe.mKeyPoint.mOffset);
NS_ENSURE_SUCCESS(res, SEEK_FATAL_ERROR);
// We've moved the read set, so reset decode.
res = ResetDecode();
NS_ENSURE_SUCCESS(res, SEEK_FATAL_ERROR);
// Check that the page the index thinks is exactly here is actually exactly
// here. If not, the index is invalid.
ogg_page page;
int skippedBytes = 0;
PageSyncResult syncres = PageSync(resource,
&mOggState,
false,
keyframe.mKeyPoint.mOffset,
resource->GetLength(),
&page,
skippedBytes);
NS_ENSURE_TRUE(syncres != PAGE_SYNC_ERROR, SEEK_FATAL_ERROR);
if (syncres != PAGE_SYNC_OK || skippedBytes != 0) {
LOG(PR_LOG_DEBUG, ("Indexed-seek failure: Ogg Skeleton Index is invalid "
"or sync error after seek"));
return RollbackIndexedSeek(tell);
}
uint32_t serial = ogg_page_serialno(&page);
if (serial != keyframe.mSerial) {
// Serialno of page at offset isn't what the index told us to expect.
// Assume the index is invalid.
return RollbackIndexedSeek(tell);
}
OggCodecState* codecState = mCodecStore.Get(serial);
if (codecState &&
codecState->mActive &&
ogg_stream_pagein(&codecState->mState, &page) != 0)
{
// Couldn't insert page into the ogg resource, or somehow the resource
// is no longer active.
return RollbackIndexedSeek(tell);
}
return SEEK_OK;
}
nsresult OggReader::SeekInBufferedRange(int64_t aTarget,
int64_t aAdjustedTarget,
int64_t aStartTime,
int64_t aEndTime,
const nsTArray<SeekRange>& aRanges,
const SeekRange& aRange)
{
LOG(PR_LOG_DEBUG, ("%p Seeking in buffered data to %lld using bisection search", mDecoder, aTarget));
nsresult res = NS_OK;
if (HasVideo() || aAdjustedTarget >= aTarget) {
// We know the exact byte range in which the target must lie. It must
// be buffered in the media cache. Seek there.
nsresult res = SeekBisection(aTarget, aRange, 0);
if (NS_FAILED(res) || !HasVideo()) {
return res;
}
// We have an active Theora bitstream. Decode the next Theora frame, and
// extract its keyframe's time.
bool eof;
do {
bool skip = false;
eof = !DecodeVideoFrame(skip, 0);
{
ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor());
if (mDecoder->IsShutdown()) {
return NS_ERROR_FAILURE;
}
}
} while (!eof &&
mVideoQueue.GetSize() == 0);
VideoData* video = mVideoQueue.PeekFront();
if (video && !video->mKeyframe) {
// First decoded frame isn't a keyframe, seek back to previous keyframe,
// otherwise we'll get visual artifacts.
NS_ASSERTION(video->mTimecode != -1, "Must have a granulepos");
int shift = mTheoraState->mInfo.keyframe_granule_shift;
int64_t keyframeGranulepos = (video->mTimecode >> shift) << shift;
int64_t keyframeTime = mTheoraState->StartTime(keyframeGranulepos);
SEEK_LOG(PR_LOG_DEBUG, ("Keyframe for %lld is at %lld, seeking back to it",
video->mTime, keyframeTime));
aAdjustedTarget = std::min(aAdjustedTarget, keyframeTime);
}
}
if (aAdjustedTarget < aTarget) {
SeekRange k = SelectSeekRange(aRanges,
aAdjustedTarget,
aStartTime,
aEndTime,
false);
res = SeekBisection(aAdjustedTarget, k, SEEK_FUZZ_USECS);
}
return res;
}
nsresult OggReader::SeekInUnbuffered(int64_t aTarget,
int64_t aStartTime,
int64_t aEndTime,
const nsTArray<SeekRange>& aRanges)
{
LOG(PR_LOG_DEBUG, ("%p Seeking in unbuffered data to %lld using bisection search", mDecoder, aTarget));
// If we've got an active Theora bitstream, determine the maximum possible
// time in usecs which a keyframe could be before a given interframe. We
// subtract this from our seek target, seek to the new target, and then
// will decode forward to the original seek target. We should encounter a
// keyframe in that interval. This prevents us from needing to run two
// bisections; one for the seek target frame, and another to find its
// keyframe. It's usually faster to just download this extra data, rather
// tham perform two bisections to find the seek target's keyframe. We
// don't do this offsetting when seeking in a buffered range,
// as the extra decoding causes a noticeable speed hit when all the data
// is buffered (compared to just doing a bisection to exactly find the
// keyframe).
int64_t keyframeOffsetMs = 0;
if (HasVideo() && mTheoraState) {
keyframeOffsetMs = mTheoraState->MaxKeyframeOffset();
}
// Add in the Opus pre-roll if necessary, as well.
if (HasAudio() && mOpusState) {
keyframeOffsetMs = std::max(keyframeOffsetMs, SEEK_OPUS_PREROLL);
}
int64_t seekTarget = std::max(aStartTime, aTarget - keyframeOffsetMs);
// Minimize the bisection search space using the known timestamps from the
// buffered ranges.
SeekRange k = SelectSeekRange(aRanges, seekTarget, aStartTime, aEndTime, false);
return SeekBisection(seekTarget, k, SEEK_FUZZ_USECS);
}
nsRefPtr<MediaDecoderReader::SeekPromise>
OggReader::Seek(int64_t aTarget, int64_t aEndTime)
{
nsresult res = SeekInternal(aTarget, aEndTime);
if (NS_FAILED(res)) {
return SeekPromise::CreateAndReject(res, __func__);
} else {
return SeekPromise::CreateAndResolve(aTarget, __func__);
}
}
nsresult OggReader::SeekInternal(int64_t aTarget, int64_t aEndTime)
{
MOZ_ASSERT(OnTaskQueue());
if (mIsChained)
return NS_ERROR_FAILURE;
LOG(PR_LOG_DEBUG, ("%p About to seek to %lld", mDecoder, aTarget));
nsresult res;
MediaResource* resource = mDecoder->GetResource();
NS_ENSURE_TRUE(resource != nullptr, NS_ERROR_FAILURE);
int64_t adjustedTarget = aTarget;
if (HasAudio() && mOpusState){
adjustedTarget = std::max(mStartTime, aTarget - SEEK_OPUS_PREROLL);
}
if (adjustedTarget == mStartTime) {
// We've seeked to the media start. Just seek to the offset of the first
// content page.
res = resource->Seek(nsISeekableStream::NS_SEEK_SET, 0);
NS_ENSURE_SUCCESS(res,res);
res = ResetDecode(true);
NS_ENSURE_SUCCESS(res,res);
} else {
// TODO: This may seek back unnecessarily far in the video, but we don't
// have a way of asking Skeleton to seek to a different target for each
// stream yet. Using adjustedTarget here is at least correct, if slow.
IndexedSeekResult sres = SeekToKeyframeUsingIndex(adjustedTarget);
NS_ENSURE_TRUE(sres != SEEK_FATAL_ERROR, NS_ERROR_FAILURE);
if (sres == SEEK_INDEX_FAIL) {
// No index or other non-fatal index-related failure. Try to seek
// using a bisection search. Determine the already downloaded data
// in the media cache, so we can try to seek in the cached data first.
nsAutoTArray<SeekRange, 16> ranges;
res = GetSeekRanges(ranges);
NS_ENSURE_SUCCESS(res,res);
// Figure out if the seek target lies in a buffered range.
SeekRange r = SelectSeekRange(ranges, aTarget, mStartTime, aEndTime, true);
if (!r.IsNull()) {
// We know the buffered range in which the seek target lies, do a
// bisection search in that buffered range.
res = SeekInBufferedRange(aTarget, adjustedTarget, mStartTime, aEndTime, ranges, r);
NS_ENSURE_SUCCESS(res,res);
} else {
// The target doesn't lie in a buffered range. Perform a bisection
// search over the whole media, using the known buffered ranges to
// reduce the search space.
res = SeekInUnbuffered(aTarget, mStartTime, aEndTime, ranges);
NS_ENSURE_SUCCESS(res,res);
}
}
}
if (HasVideo()) {
// Decode forwards until we find the next keyframe. This is required,
// as although the seek should finish on a page containing a keyframe,
// there may be non-keyframes in the page before the keyframe.
// When doing fastSeek we display the first frame after the seek, so
// we need to advance the decode to the keyframe otherwise we'll get
// visual artifacts in the first frame output after the seek.
// First, we must check to see if there's already a keyframe in the frames
// that we may have already decoded, and discard frames up to the
// keyframe.
VideoData* v;
while ((v = mVideoQueue.PeekFront()) && !v->mKeyframe) {
nsRefPtr<VideoData> releaseMe = mVideoQueue.PopFront();
}
if (mVideoQueue.GetSize() == 0) {
// We didn't find a keyframe in the frames already here, so decode
// forwards until we find a keyframe.
bool skip = true;
while (DecodeVideoFrame(skip, 0) && skip) {
ReentrantMonitorAutoEnter mon(mDecoder->GetReentrantMonitor());
if (mDecoder->IsShutdown()) {
return NS_ERROR_FAILURE;
}
}
}
#ifdef DEBUG
v = mVideoQueue.PeekFront();
if (!v || !v->mKeyframe) {
NS_WARNING("Ogg seek didn't end up before a key frame!");
}
#endif
}
return NS_OK;
}
// Reads a page from the media resource.
static PageSyncResult
PageSync(MediaResource* aResource,
ogg_sync_state* aState,
bool aCachedDataOnly,
int64_t aOffset,
int64_t aEndOffset,
ogg_page* aPage,
int& aSkippedBytes)
{
aSkippedBytes = 0;
// Sync to the next page.
int ret = 0;
uint32_t bytesRead = 0;
int64_t readHead = aOffset;
while (ret <= 0) {
ret = ogg_sync_pageseek(aState, aPage);
if (ret == 0) {
char* buffer = ogg_sync_buffer(aState, PAGE_STEP);
NS_ASSERTION(buffer, "Must have a buffer");
// Read from the file into the buffer
int64_t bytesToRead = std::min(static_cast<int64_t>(PAGE_STEP),
aEndOffset - readHead);
NS_ASSERTION(bytesToRead <= UINT32_MAX, "bytesToRead range check");
if (bytesToRead <= 0) {
return PAGE_SYNC_END_OF_RANGE;
}
nsresult rv = NS_OK;
if (aCachedDataOnly) {
rv = aResource->ReadFromCache(buffer, readHead,
static_cast<uint32_t>(bytesToRead));
NS_ENSURE_SUCCESS(rv,PAGE_SYNC_ERROR);
bytesRead = static_cast<uint32_t>(bytesToRead);
} else {
rv = aResource->Seek(nsISeekableStream::NS_SEEK_SET, readHead);
NS_ENSURE_SUCCESS(rv,PAGE_SYNC_ERROR);
rv = aResource->Read(buffer,
static_cast<uint32_t>(bytesToRead),
&bytesRead);
NS_ENSURE_SUCCESS(rv,PAGE_SYNC_ERROR);
}
if (bytesRead == 0 && NS_SUCCEEDED(rv)) {
// End of file.
return PAGE_SYNC_END_OF_RANGE;
}
readHead += bytesRead;
// Update the synchronisation layer with the number
// of bytes written to the buffer
ret = ogg_sync_wrote(aState, bytesRead);
NS_ENSURE_TRUE(ret == 0, PAGE_SYNC_ERROR);
continue;
}
if (ret < 0) {
NS_ASSERTION(aSkippedBytes >= 0, "Offset >= 0");
aSkippedBytes += -ret;
NS_ASSERTION(aSkippedBytes >= 0, "Offset >= 0");
continue;
}
}
return PAGE_SYNC_OK;
}
nsresult OggReader::SeekBisection(int64_t aTarget,
const SeekRange& aRange,
uint32_t aFuzz)
{
MOZ_ASSERT(OnTaskQueue());
nsresult res;
MediaResource* resource = mDecoder->GetResource();
if (aTarget == aRange.mTimeStart) {
if (NS_FAILED(ResetDecode())) {
return NS_ERROR_FAILURE;
}
res = resource->Seek(nsISeekableStream::NS_SEEK_SET, 0);
NS_ENSURE_SUCCESS(res,res);
return NS_OK;
}
// Bisection search, find start offset of last page with end time less than
// the seek target.
ogg_int64_t startOffset = aRange.mOffsetStart;
ogg_int64_t startTime = aRange.mTimeStart;
ogg_int64_t startLength = 0; // Length of the page at startOffset.
ogg_int64_t endOffset = aRange.mOffsetEnd;
ogg_int64_t endTime = aRange.mTimeEnd;
ogg_int64_t seekTarget = aTarget;
int64_t seekLowerBound = std::max(static_cast<int64_t>(0), aTarget - aFuzz);
int hops = 0;
DebugOnly<ogg_int64_t> previousGuess = -1;
int backsteps = 0;
const int maxBackStep = 10;
NS_ASSERTION(static_cast<uint64_t>(PAGE_STEP) * pow(2.0, maxBackStep) < INT32_MAX,
"Backstep calculation must not overflow");
// Seek via bisection search. Loop until we find the offset where the page
// before the offset is before the seek target, and the page after the offset
// is after the seek target.
while (true) {
ogg_int64_t duration = 0;
double target = 0;
ogg_int64_t interval = 0;
ogg_int64_t guess = 0;
ogg_page page;
int skippedBytes = 0;
ogg_int64_t pageOffset = 0;
ogg_int64_t pageLength = 0;
ogg_int64_t granuleTime = -1;
bool mustBackoff = false;
// Guess where we should bisect to, based on the bit rate and the time
// remaining in the interval. Loop until we can determine the time at
// the guess offset.
while (true) {
// Discard any previously buffered packets/pages.
if (NS_FAILED(ResetDecode())) {
return NS_ERROR_FAILURE;
}
interval = endOffset - startOffset - startLength;
if (interval == 0) {
// Our interval is empty, we've found the optimal seek point, as the
// page at the start offset is before the seek target, and the page
// at the end offset is after the seek target.
SEEK_LOG(PR_LOG_DEBUG, ("Interval narrowed, terminating bisection."));
break;
}
// Guess bisection point.
duration = endTime - startTime;
target = (double)(seekTarget - startTime) / (double)duration;
guess = startOffset + startLength +
static_cast<ogg_int64_t>((double)interval * target);
guess = std::min(guess, endOffset - PAGE_STEP);
if (mustBackoff) {
// We previously failed to determine the time at the guess offset,
// probably because we ran out of data to decode. This usually happens
// when we guess very close to the end offset. So reduce the guess
// offset using an exponential backoff until we determine the time.
SEEK_LOG(PR_LOG_DEBUG, ("Backing off %d bytes, backsteps=%d",
static_cast<int32_t>(PAGE_STEP * pow(2.0, backsteps)), backsteps));
guess -= PAGE_STEP * static_cast<ogg_int64_t>(pow(2.0, backsteps));
if (guess <= startOffset) {
// We've tried to backoff to before the start offset of our seek
// range. This means we couldn't find a seek termination position
// near the end of the seek range, so just set the seek termination
// condition, and break out of the bisection loop. We'll begin
// decoding from the start of the seek range.
interval = 0;
break;
}
backsteps = std::min(backsteps + 1, maxBackStep);
// We reset mustBackoff. If we still need to backoff further, it will
// be set to true again.
mustBackoff = false;
} else {
backsteps = 0;
}
guess = std::max(guess, startOffset + startLength);
SEEK_LOG(PR_LOG_DEBUG, ("Seek loop start[o=%lld..%lld t=%lld] "
"end[o=%lld t=%lld] "
"interval=%lld target=%lf guess=%lld",
startOffset, (startOffset+startLength), startTime,
endOffset, endTime, interval, target, guess));
NS_ASSERTION(guess >= startOffset + startLength, "Guess must be after range start");
NS_ASSERTION(guess < endOffset, "Guess must be before range end");
NS_ASSERTION(guess != previousGuess, "Guess should be different to previous");
previousGuess = guess;
hops++;
// Locate the next page after our seek guess, and then figure out the
// granule time of the audio and video bitstreams there. We can then
// make a bisection decision based on our location in the media.
PageSyncResult res = PageSync(resource,
&mOggState,
false,
guess,
endOffset,
&page,
skippedBytes);
NS_ENSURE_TRUE(res != PAGE_SYNC_ERROR, NS_ERROR_FAILURE);
if (res == PAGE_SYNC_END_OF_RANGE) {
// Our guess was too close to the end, we've ended up reading the end
// page. Backoff exponentially from the end point, in case the last
// page/frame/sample is huge.
mustBackoff = true;
SEEK_LOG(PR_LOG_DEBUG, ("Hit the end of range, backing off"));
continue;
}
// We've located a page of length |ret| at |guess + skippedBytes|.
// Remember where the page is located.
pageOffset = guess + skippedBytes;
pageLength = page.header_len + page.body_len;
// Read pages until we can determine the granule time of the audio and
// video bitstream.
ogg_int64_t audioTime = -1;
ogg_int64_t videoTime = -1;
do {
// Add the page to its codec state, determine its granule time.
uint32_t serial = ogg_page_serialno(&page);
OggCodecState* codecState = mCodecStore.Get(serial);
if (codecState && codecState->mActive) {
int ret = ogg_stream_pagein(&codecState->mState, &page);
NS_ENSURE_TRUE(ret == 0, NS_ERROR_FAILURE);
}
ogg_int64_t granulepos = ogg_page_granulepos(&page);
if (HasAudio() && granulepos > 0 && audioTime == -1) {
if (mVorbisState && serial == mVorbisState->mSerial) {
audioTime = mVorbisState->Time(granulepos);
} else if (mOpusState && serial == mOpusState->mSerial) {
audioTime = mOpusState->Time(granulepos);
}
}
if (HasVideo() &&
granulepos > 0 &&
serial == mTheoraState->mSerial &&
videoTime == -1) {
videoTime = mTheoraState->Time(granulepos);
}
if (pageOffset + pageLength >= endOffset) {
// Hit end of readable data.
break;
}
if (!ReadOggPage(&page)) {
break;
}
} while ((HasAudio() && audioTime == -1) ||
(HasVideo() && videoTime == -1));
if ((HasAudio() && audioTime == -1) ||
(HasVideo() && videoTime == -1))
{
// We don't have timestamps for all active tracks...
if (pageOffset == startOffset + startLength &&
pageOffset + pageLength >= endOffset) {
// We read the entire interval without finding timestamps for all
// active tracks. We know the interval start offset is before the seek
// target, and the interval end is after the seek target, and we can't
// terminate inside the interval, so we terminate the seek at the
// start of the interval.
interval = 0;
break;
}
// We should backoff; cause the guess to back off from the end, so
// that we've got more room to capture.
mustBackoff = true;
continue;
}
// We've found appropriate time stamps here. Proceed to bisect
// the search space.
granuleTime = std::max(audioTime, videoTime);
NS_ASSERTION(granuleTime > 0, "Must get a granuletime");
break;
} // End of "until we determine time at guess offset" loop.
if (interval == 0) {
// Seek termination condition; we've found the page boundary of the
// last page before the target, and the first page after the target.
SEEK_LOG(PR_LOG_DEBUG, ("Terminating seek at offset=%lld", startOffset));
NS_ASSERTION(startTime < aTarget, "Start time must always be less than target");
res = resource->Seek(nsISeekableStream::NS_SEEK_SET, startOffset);
NS_ENSURE_SUCCESS(res,res);
if (NS_FAILED(ResetDecode())) {
return NS_ERROR_FAILURE;
}
break;
}
SEEK_LOG(PR_LOG_DEBUG, ("Time at offset %lld is %lld", guess, granuleTime));
if (granuleTime < seekTarget && granuleTime > seekLowerBound) {
// We're within the fuzzy region in which we want to terminate the search.
res = resource->Seek(nsISeekableStream::NS_SEEK_SET, pageOffset);
NS_ENSURE_SUCCESS(res,res);
if (NS_FAILED(ResetDecode())) {
return NS_ERROR_FAILURE;
}
SEEK_LOG(PR_LOG_DEBUG, ("Terminating seek at offset=%lld", pageOffset));
break;
}
if (granuleTime >= seekTarget) {
// We've landed after the seek target.
NS_ASSERTION(pageOffset < endOffset, "offset_end must decrease");
endOffset = pageOffset;
endTime = granuleTime;
} else if (granuleTime < seekTarget) {
// Landed before seek target.
NS_ASSERTION(pageOffset >= startOffset + startLength,
"Bisection point should be at or after end of first page in interval");
startOffset = pageOffset;
startLength = pageLength;
startTime = granuleTime;
}
NS_ASSERTION(startTime < seekTarget, "Must be before seek target");
NS_ASSERTION(endTime >= seekTarget, "End must be after seek target");
}
SEEK_LOG(PR_LOG_DEBUG, ("Seek complete in %d bisections.", hops));
return NS_OK;
}
nsresult OggReader::GetBuffered(dom::TimeRanges* aBuffered)
{
MOZ_ASSERT(mStartTime != -1, "Need to finish metadata decode first");
{
mozilla::ReentrantMonitorAutoEnter mon(mMonitor);
if (mIsChained)
return NS_ERROR_FAILURE;
}
#ifdef OGG_ESTIMATE_BUFFERED
return MediaDecoderReader::GetBuffered(aBuffered);
#else
// HasAudio and HasVideo are not used here as they take a lock and cause
// a deadlock. Accessing mInfo doesn't require a lock - it doesn't change
// after metadata is read.
if (!mInfo.HasValidMedia()) {
// No need to search through the file if there are no audio or video tracks
return NS_OK;
}
AutoPinned<MediaResource> resource(mDecoder->GetResource());
nsTArray<MediaByteRange> ranges;
nsresult res = resource->GetCachedRanges(ranges);
NS_ENSURE_SUCCESS(res, res);
// Traverse across the buffered byte ranges, determining the time ranges
// they contain. MediaResource::GetNextCachedData(offset) returns -1 when
// offset is after the end of the media resource, or there's no more cached
// data after the offset. This loop will run until we've checked every
// buffered range in the media, in increasing order of offset.
nsAutoOggSyncState sync;
for (uint32_t index = 0; index < ranges.Length(); index++) {
// Ensure the offsets are after the header pages.
int64_t startOffset = ranges[index].mStart;
int64_t endOffset = ranges[index].mEnd;
// Because the granulepos time is actually the end time of the page,
// we special-case (startOffset == 0) so that the first
// buffered range always appears to be buffered from the media start
// time, rather than from the end-time of the first page.
int64_t startTime = (startOffset == 0) ? mStartTime : -1;
// Find the start time of the range. Read pages until we find one with a
// granulepos which we can convert into a timestamp to use as the time of
// the start of the buffered range.
ogg_sync_reset(&sync.mState);
while (startTime == -1) {
ogg_page page;
int32_t discard;
PageSyncResult res = PageSync(resource,
&sync.mState,
true,
startOffset,
endOffset,
&page,
discard);
if (res == PAGE_SYNC_ERROR) {
return NS_ERROR_FAILURE;
} else if (res == PAGE_SYNC_END_OF_RANGE) {
// Hit the end of range without reading a page, give up trying to
// find a start time for this buffered range, skip onto the next one.
break;
}
int64_t granulepos = ogg_page_granulepos(&page);
if (granulepos == -1) {
// Page doesn't have an end time, advance to the next page
// until we find one.
startOffset += page.header_len + page.body_len;
continue;
}
uint32_t serial = ogg_page_serialno(&page);
if (mVorbisState && serial == mVorbisSerial) {
startTime = VorbisState::Time(&mVorbisInfo, granulepos);
NS_ASSERTION(startTime > 0, "Must have positive start time");
}
else if (mOpusState && serial == mOpusSerial) {
startTime = OpusState::Time(mOpusPreSkip, granulepos);
NS_ASSERTION(startTime > 0, "Must have positive start time");
}
else if (mTheoraState && serial == mTheoraSerial) {
startTime = TheoraState::Time(&mTheoraInfo, granulepos);
NS_ASSERTION(startTime > 0, "Must have positive start time");
}
else if (mCodecStore.Contains(serial)) {
// Stream is not the theora or vorbis stream we're playing,
// but is one that we have header data for.
startOffset += page.header_len + page.body_len;
continue;
}
else {
// Page is for a stream we don't know about (possibly a chained
// ogg), return OK to abort the finding any further ranges. This
// prevents us searching through the rest of the media when we
// may not be able to extract timestamps from it.
SetChained(true);
return NS_OK;
}
}
if (startTime != -1) {
// We were able to find a start time for that range, see if we can
// find an end time.
int64_t endTime = RangeEndTime(startOffset, endOffset, true);
if (endTime != -1) {
aBuffered->Add((startTime - mStartTime) / static_cast<double>(USECS_PER_S),
(endTime - mStartTime) / static_cast<double>(USECS_PER_S));
}
}
}
return NS_OK;
#endif
}
VideoData* OggReader::FindStartTime(int64_t& aOutStartTime)
{
MOZ_ASSERT(OnTaskQueue() || mDecoder->OnStateMachineTaskQueue());
// Extract the start times of the bitstreams in order to calculate
// the duration.
int64_t videoStartTime = INT64_MAX;
int64_t audioStartTime = INT64_MAX;
VideoData* videoData = nullptr;
if (HasVideo()) {
videoData = DecodeToFirstVideoData();
if (videoData) {
videoStartTime = videoData->mTime;
LOG(PR_LOG_DEBUG, ("OggReader::FindStartTime() video=%lld", videoStartTime));
}
}
if (HasAudio()) {
AudioData* audioData = DecodeToFirstAudioData();
if (audioData) {
audioStartTime = audioData->mTime;
LOG(PR_LOG_DEBUG, ("OggReader::FindStartTime() audio=%lld", audioStartTime));
}
}
int64_t startTime = std::min(videoStartTime, audioStartTime);
if (startTime != INT64_MAX) {
aOutStartTime = startTime;
}
return videoData;
}
AudioData* OggReader::DecodeToFirstAudioData()
{
bool eof = false;
while (!eof && AudioQueue().GetSize() == 0) {
{
ReentrantMonitorAutoEnter decoderMon(mDecoder->GetReentrantMonitor());
if (mDecoder->IsShutdown()) {
return nullptr;
}
}
eof = !DecodeAudioData();
}
if (eof) {
AudioQueue().Finish();
}
AudioData* d = nullptr;
return (d = AudioQueue().PeekFront()) ? d : nullptr;
}
OggCodecStore::OggCodecStore()
: mMonitor("CodecStore")
{
}
void OggCodecStore::Add(uint32_t serial, OggCodecState* codecState)
{
MonitorAutoLock mon(mMonitor);
mCodecStates.Put(serial, codecState);
}
bool OggCodecStore::Contains(uint32_t serial)
{
MonitorAutoLock mon(mMonitor);
return mCodecStates.Get(serial, nullptr);
}
OggCodecState* OggCodecStore::Get(uint32_t serial)
{
MonitorAutoLock mon(mMonitor);
return mCodecStates.Get(serial);
}
} // namespace mozilla
Ссылка в новой задаче Показать git blame Копировать постоянную ссылку