dom/media/webm/WebMBufferedParser.cpp
author Bryce Van Dyk <bvandyk@mozilla.com>
Mon, 01 Feb 2016 13:46:02 +1300
changeset 288504 2e990a8d4553a9c2ce896681163fd33a74b78235
parent 280565 985507e906169769bf491a37e0e0a5ca865e0cdc
child 307428 39bc955ad2fa5b1a230270f8cbfdfb018fe43b59
permissions -rw-r--r--
Bug 657791 - Update WebM demuxer to clamp cueless seeks instead of failing. r=kinetik Previously if a seek time is specified outside of the buffered range for certain WebMs (particularly those without cues) the WebMDemuxer would fail out of SeekInternal() with an error code. However, this would lead to issues due to inconsistent state (recovery was not made from a failed seek). This change attemps to address this by instead seeking to the final available cluster. MozReview-Commit-ID: GZLPZDWLcT1

/* -*- 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 "nsAlgorithm.h"
#include "WebMBufferedParser.h"
#include "nsThreadUtils.h"
#include <algorithm>

#define WEBM_DEBUG(arg, ...) MOZ_LOG(gWebMDemuxerLog, mozilla::LogLevel::Debug, ("WebMBufferedParser(%p)::%s: " arg, this, __func__, ##__VA_ARGS__))

namespace mozilla {

extern LazyLogModule gWebMDemuxerLog;

static uint32_t
VIntLength(unsigned char aFirstByte, uint32_t* aMask)
{
  uint32_t count = 1;
  uint32_t mask = 1 << 7;
  while (count < 8) {
    if ((aFirstByte & mask) != 0) {
      break;
    }
    mask >>= 1;
    count += 1;
  }
  if (aMask) {
    *aMask = mask;
  }
  NS_ASSERTION(count >= 1 && count <= 8, "Insane VInt length.");
  return count;
}

void WebMBufferedParser::Append(const unsigned char* aBuffer, uint32_t aLength,
                                nsTArray<WebMTimeDataOffset>& aMapping,
                                ReentrantMonitor& aReentrantMonitor)
{
  static const uint32_t EBML_ID = 0x1a45dfa3;
  static const uint32_t SEGMENT_ID = 0x18538067;
  static const uint32_t SEGINFO_ID = 0x1549a966;
  static const uint32_t TRACKS_ID = 0x1654AE6B;
  static const uint32_t CLUSTER_ID = 0x1f43b675;
  static const uint32_t TIMECODESCALE_ID = 0x2ad7b1;
  static const unsigned char TIMECODE_ID = 0xe7;
  static const unsigned char BLOCKGROUP_ID = 0xa0;
  static const unsigned char BLOCK_ID = 0xa1;
  static const unsigned char SIMPLEBLOCK_ID = 0xa3;
  static const uint32_t BLOCK_TIMECODE_LENGTH = 2;

  static const unsigned char CLUSTER_SYNC_ID[] = { 0x1f, 0x43, 0xb6, 0x75 };

  const unsigned char* p = aBuffer;

  // Parse each byte in aBuffer one-by-one, producing timecodes and updating
  // aMapping as we go.  Parser pauses at end of stream (which may be at any
  // point within the parse) and resumes parsing the next time Append is
  // called with new data.
  while (p < aBuffer + aLength) {
    switch (mState) {
    case READ_ELEMENT_ID:
      mVIntRaw = true;
      mState = READ_VINT;
      mNextState = READ_ELEMENT_SIZE;
      break;
    case READ_ELEMENT_SIZE:
      mVIntRaw = false;
      mElement.mID = mVInt;
      mState = READ_VINT;
      mNextState = PARSE_ELEMENT;
      break;
    case FIND_CLUSTER_SYNC:
      if (*p++ == CLUSTER_SYNC_ID[mClusterSyncPos]) {
        mClusterSyncPos += 1;
      } else {
        mClusterSyncPos = 0;
      }
      if (mClusterSyncPos == sizeof(CLUSTER_SYNC_ID)) {
        mVInt.mValue = CLUSTER_ID;
        mVInt.mLength = sizeof(CLUSTER_SYNC_ID);
        mState = READ_ELEMENT_SIZE;
      }
      break;
    case PARSE_ELEMENT:
      mElement.mSize = mVInt;
      switch (mElement.mID.mValue) {
      case SEGMENT_ID:
        mState = READ_ELEMENT_ID;
        break;
      case SEGINFO_ID:
        mGotTimecodeScale = true;
        mState = READ_ELEMENT_ID;
        break;
      case TIMECODE_ID:
        mVInt = VInt();
        mVIntLeft = mElement.mSize.mValue;
        mState = READ_VINT_REST;
        mNextState = READ_CLUSTER_TIMECODE;
        break;
      case TIMECODESCALE_ID:
        mVInt = VInt();
        mVIntLeft = mElement.mSize.mValue;
        mState = READ_VINT_REST;
        mNextState = READ_TIMECODESCALE;
        break;
      case CLUSTER_ID:
        mClusterOffset = mCurrentOffset + (p - aBuffer) -
                        (mElement.mID.mLength + mElement.mSize.mLength);
        // Handle "unknown" length;
        if (mElement.mSize.mValue + 1 != uint64_t(1) << (mElement.mSize.mLength * 7)) {
          mClusterEndOffset = mClusterOffset + mElement.mID.mLength + mElement.mSize.mLength + mElement.mSize.mValue;
        } else {
          mClusterEndOffset = -1;
        }
        mState = READ_ELEMENT_ID;
        break;
      case BLOCKGROUP_ID:
        mState = READ_ELEMENT_ID;
        break;
      case SIMPLEBLOCK_ID:
        /* FALLTHROUGH */
      case BLOCK_ID:
        mBlockSize = mElement.mSize.mValue;
        mBlockTimecode = 0;
        mBlockTimecodeLength = BLOCK_TIMECODE_LENGTH;
        mBlockOffset = mCurrentOffset + (p - aBuffer) -
                       (mElement.mID.mLength + mElement.mSize.mLength);
        mState = READ_VINT;
        mNextState = READ_BLOCK_TIMECODE;
        break;
      case TRACKS_ID:
        mSkipBytes = mElement.mSize.mValue;
        mState = CHECK_INIT_FOUND;
        break;
      case EBML_ID:
        mLastInitStartOffset = mCurrentOffset + (p - aBuffer) -
                            (mElement.mID.mLength + mElement.mSize.mLength);
        MOZ_FALLTHROUGH;
      default:
        mSkipBytes = mElement.mSize.mValue;
        mState = SKIP_DATA;
        mNextState = READ_ELEMENT_ID;
        break;
      }
      break;
    case READ_VINT: {
      unsigned char c = *p++;
      uint32_t mask;
      mVInt.mLength = VIntLength(c, &mask);
      mVIntLeft = mVInt.mLength - 1;
      mVInt.mValue = mVIntRaw ? c : c & ~mask;
      mState = READ_VINT_REST;
      break;
    }
    case READ_VINT_REST:
      if (mVIntLeft) {
        mVInt.mValue <<= 8;
        mVInt.mValue |= *p++;
        mVIntLeft -= 1;
      } else {
        mState = mNextState;
      }
      break;
    case READ_TIMECODESCALE:
      MOZ_ASSERT(mGotTimecodeScale);
      mTimecodeScale = mVInt.mValue;
      mState = READ_ELEMENT_ID;
      break;
    case READ_CLUSTER_TIMECODE:
      mClusterTimecode = mVInt.mValue;
      mState = READ_ELEMENT_ID;
      break;
    case READ_BLOCK_TIMECODE:
      if (mBlockTimecodeLength) {
        mBlockTimecode <<= 8;
        mBlockTimecode |= *p++;
        mBlockTimecodeLength -= 1;
      } else {
        // It's possible we've parsed this data before, so avoid inserting
        // duplicate WebMTimeDataOffset entries.
        {
          ReentrantMonitorAutoEnter mon(aReentrantMonitor);
          int64_t endOffset = mBlockOffset + mBlockSize +
                              mElement.mID.mLength + mElement.mSize.mLength;
          uint32_t idx = aMapping.IndexOfFirstElementGt(endOffset);
          if (idx == 0 || aMapping[idx - 1] != endOffset) {
            // Don't insert invalid negative timecodes.
            if (mBlockTimecode >= 0 || mClusterTimecode >= uint16_t(abs(mBlockTimecode))) {
              MOZ_ASSERT(mGotTimecodeScale);
              uint64_t absTimecode = mClusterTimecode + mBlockTimecode;
              absTimecode *= mTimecodeScale;
              // Avoid creating an entry if the timecode is out of order
              // (invalid according to the WebM specification) so that
              // ordering invariants of aMapping are not violated.
              if (idx == 0 ||
                  aMapping[idx - 1].mTimecode <= absTimecode ||
                  (idx + 1 < aMapping.Length() &&
                   aMapping[idx + 1].mTimecode >= absTimecode)) {
                WebMTimeDataOffset entry(endOffset, absTimecode, mLastInitStartOffset,
                                         mClusterOffset, mClusterEndOffset);
                aMapping.InsertElementAt(idx, entry);
              } else {
                WEBM_DEBUG("Out of order timecode %llu in Cluster at %lld ignored",
                           absTimecode, mClusterOffset);
              }
            }
          }
        }

        // Skip rest of block header and the block's payload.
        mBlockSize -= mVInt.mLength;
        mBlockSize -= BLOCK_TIMECODE_LENGTH;
        mSkipBytes = uint32_t(mBlockSize);
        mState = SKIP_DATA;
        mNextState = READ_ELEMENT_ID;
      }
      break;
    case SKIP_DATA:
      if (mSkipBytes) {
        uint32_t left = aLength - (p - aBuffer);
        left = std::min(left, mSkipBytes);
        p += left;
        mSkipBytes -= left;
      }
      if (!mSkipBytes) {
        mBlockEndOffset = mCurrentOffset + (p - aBuffer);
        mState = mNextState;
      }
      break;
    case CHECK_INIT_FOUND:
      if (mSkipBytes) {
        uint32_t left = aLength - (p - aBuffer);
        left = std::min(left, mSkipBytes);
        p += left;
        mSkipBytes -= left;
      }
      if (!mSkipBytes) {
        if (mInitEndOffset < 0) {
          mInitEndOffset = mCurrentOffset + (p - aBuffer);
          mBlockEndOffset = mCurrentOffset + (p - aBuffer);
        }
        mState = READ_ELEMENT_ID;
      }
      break;
    }
  }

  NS_ASSERTION(p == aBuffer + aLength, "Must have parsed to end of data.");
  mCurrentOffset += aLength;
}

int64_t
WebMBufferedParser::EndSegmentOffset(int64_t aOffset)
{
  if (mLastInitStartOffset > aOffset || mClusterOffset > aOffset) {
    return std::min(mLastInitStartOffset >= 0 ? mLastInitStartOffset : INT64_MAX,
                    mClusterOffset >= 0 ? mClusterOffset : INT64_MAX);
  }
  return mBlockEndOffset;
}

// SyncOffsetComparator and TimeComparator are slightly confusing, in that
// the nsTArray they're used with (mTimeMapping) is sorted by mEndOffset and
// these comparators are used on the other fields of WebMTimeDataOffset.
// This is only valid because timecodes are required to be monotonically
// increasing within a file (thus establishing an ordering relationship with
// mTimecode), and mEndOffset is derived from mSyncOffset.
struct SyncOffsetComparator {
  bool Equals(const WebMTimeDataOffset& a, const int64_t& b) const {
    return a.mSyncOffset == b;
  }

  bool LessThan(const WebMTimeDataOffset& a, const int64_t& b) const {
    return a.mSyncOffset < b;
  }
};

struct TimeComparator {
  bool Equals(const WebMTimeDataOffset& a, const uint64_t& b) const {
    return a.mTimecode == b;
  }

  bool LessThan(const WebMTimeDataOffset& a, const uint64_t& b) const {
    return a.mTimecode < b;
  }
};

bool WebMBufferedState::CalculateBufferedForRange(int64_t aStartOffset, int64_t aEndOffset,
                                                  uint64_t* aStartTime, uint64_t* aEndTime)
{
  ReentrantMonitorAutoEnter mon(mReentrantMonitor);

  // Find the first WebMTimeDataOffset at or after aStartOffset.
  uint32_t start = mTimeMapping.IndexOfFirstElementGt(aStartOffset - 1, SyncOffsetComparator());
  if (start == mTimeMapping.Length()) {
    return false;
  }

  // Find the first WebMTimeDataOffset at or before aEndOffset.
  uint32_t end = mTimeMapping.IndexOfFirstElementGt(aEndOffset);
  if (end > 0) {
    end -= 1;
  }

  // Range is empty.
  if (end <= start) {
    return false;
  }

  NS_ASSERTION(mTimeMapping[start].mSyncOffset >= aStartOffset &&
               mTimeMapping[end].mEndOffset <= aEndOffset,
               "Computed time range must lie within data range.");
  if (start > 0) {
    NS_ASSERTION(mTimeMapping[start - 1].mSyncOffset < aStartOffset,
                 "Must have found least WebMTimeDataOffset for start");
  }
  if (end < mTimeMapping.Length() - 1) {
    NS_ASSERTION(mTimeMapping[end + 1].mEndOffset > aEndOffset,
                 "Must have found greatest WebMTimeDataOffset for end");
  }

  MOZ_ASSERT(mTimeMapping[end].mTimecode >= mTimeMapping[end - 1].mTimecode);
  uint64_t frameDuration = mTimeMapping[end].mTimecode - mTimeMapping[end - 1].mTimecode;
  *aStartTime = mTimeMapping[start].mTimecode;
  *aEndTime = mTimeMapping[end].mTimecode + frameDuration;
  return true;
}

bool WebMBufferedState::GetOffsetForTime(uint64_t aTime, int64_t* aOffset)
{
  ReentrantMonitorAutoEnter mon(mReentrantMonitor);

  if(mTimeMapping.IsEmpty()) {
    return false;
  }

  uint64_t time = aTime;
  if (time > 0) {
    time = time - 1;
  }
  uint32_t idx = mTimeMapping.IndexOfFirstElementGt(time, TimeComparator());
  if (idx == mTimeMapping.Length()) {
    // Clamp to end
    *aOffset = mTimeMapping[mTimeMapping.Length() - 1].mSyncOffset;
  } else {
    // Idx is within array or has been clamped to start
    *aOffset = mTimeMapping[idx].mSyncOffset;
  }
  return true;
}

void WebMBufferedState::NotifyDataArrived(const unsigned char* aBuffer, uint32_t aLength, int64_t aOffset)
{
  uint32_t idx = mRangeParsers.IndexOfFirstElementGt(aOffset - 1);
  if (idx == 0 || !(mRangeParsers[idx-1] == aOffset)) {
    // If the incoming data overlaps an already parsed range, adjust the
    // buffer so that we only reparse the new data.  It's also possible to
    // have an overlap where the end of the incoming data is within an
    // already parsed range, but we don't bother handling that other than by
    // avoiding storing duplicate timecodes when the parser runs.
    if (idx != mRangeParsers.Length() && mRangeParsers[idx].mStartOffset <= aOffset) {
      // Complete overlap, skip parsing.
      if (aOffset + aLength <= mRangeParsers[idx].mCurrentOffset) {
        return;
      }

      // Partial overlap, adjust the buffer to parse only the new data.
      int64_t adjust = mRangeParsers[idx].mCurrentOffset - aOffset;
      NS_ASSERTION(adjust >= 0, "Overlap detection bug.");
      aBuffer += adjust;
      aLength -= uint32_t(adjust);
    } else {
      mRangeParsers.InsertElementAt(idx, WebMBufferedParser(aOffset));
      if (idx != 0) {
        mRangeParsers[idx].SetTimecodeScale(mRangeParsers[0].GetTimecodeScale());
      }
    }
  }

  mRangeParsers[idx].Append(aBuffer,
                            aLength,
                            mTimeMapping,
                            mReentrantMonitor);

  // Merge parsers with overlapping regions and clean up the remnants.
  uint32_t i = 0;
  while (i + 1 < mRangeParsers.Length()) {
    if (mRangeParsers[i].mCurrentOffset >= mRangeParsers[i + 1].mStartOffset) {
      mRangeParsers[i + 1].mStartOffset = mRangeParsers[i].mStartOffset;
      mRangeParsers[i + 1].mInitEndOffset = mRangeParsers[i].mInitEndOffset;
      mRangeParsers.RemoveElementAt(i);
    } else {
      i += 1;
    }
  }

  if (mRangeParsers.IsEmpty()) {
    return;
  }

  ReentrantMonitorAutoEnter mon(mReentrantMonitor);
  mLastBlockOffset = mRangeParsers.LastElement().mBlockEndOffset;
}

void WebMBufferedState::Reset() {
  mRangeParsers.Clear();
  mTimeMapping.Clear();
}

void WebMBufferedState::UpdateIndex(const MediaByteRangeSet& aRanges, MediaResource* aResource)
{
  for (uint32_t index = 0; index < aRanges.Length(); index++) {
    const MediaByteRange& range = aRanges[index];
    int64_t offset = range.mStart;
    uint32_t length = range.mEnd - range.mStart;

    uint32_t idx = mRangeParsers.IndexOfFirstElementGt(offset - 1);
    if (!idx || !(mRangeParsers[idx-1] == offset)) {
      // If the incoming data overlaps an already parsed range, adjust the
      // buffer so that we only reparse the new data.  It's also possible to
      // have an overlap where the end of the incoming data is within an
      // already parsed range, but we don't bother handling that other than by
      // avoiding storing duplicate timecodes when the parser runs.
      if (idx != mRangeParsers.Length() && mRangeParsers[idx].mStartOffset <= offset) {
        // Complete overlap, skip parsing.
        if (offset + length <= mRangeParsers[idx].mCurrentOffset) {
          continue;
        }

        // Partial overlap, adjust the buffer to parse only the new data.
        int64_t adjust = mRangeParsers[idx].mCurrentOffset - offset;
        NS_ASSERTION(adjust >= 0, "Overlap detection bug.");
        offset += adjust;
        length -= uint32_t(adjust);
      } else {
        mRangeParsers.InsertElementAt(idx, WebMBufferedParser(offset));
        if (idx) {
          mRangeParsers[idx].SetTimecodeScale(mRangeParsers[0].GetTimecodeScale());
        }
      }
    }
    while (length > 0) {
      static const uint32_t BLOCK_SIZE = 1048576;
      uint32_t block = std::min(length, BLOCK_SIZE);
      RefPtr<MediaByteBuffer> bytes = aResource->MediaReadAt(offset, block);
      if (!bytes) {
        break;
      }
      NotifyDataArrived(bytes->Elements(), bytes->Length(), offset);
      length -= bytes->Length();
      offset += bytes->Length();
    }
  }
}

int64_t WebMBufferedState::GetInitEndOffset()
{
  if (mRangeParsers.IsEmpty()) {
    return -1;
  }
  return mRangeParsers[0].mInitEndOffset;
}

int64_t WebMBufferedState::GetLastBlockOffset()
{
  ReentrantMonitorAutoEnter mon(mReentrantMonitor);

  return mLastBlockOffset;
}

bool WebMBufferedState::GetStartTime(uint64_t *aTime)
{
  ReentrantMonitorAutoEnter mon(mReentrantMonitor);

  if (mTimeMapping.IsEmpty()) {
    return false;
  }

  uint32_t idx = mTimeMapping.IndexOfFirstElementGt(0, SyncOffsetComparator());
  if (idx == mTimeMapping.Length()) {
    return false;
  }

  *aTime = mTimeMapping[idx].mTimecode;
  return true;
}

bool
WebMBufferedState::GetNextKeyframeTime(uint64_t aTime, uint64_t* aKeyframeTime)
{
  ReentrantMonitorAutoEnter mon(mReentrantMonitor);
  int64_t offset = 0;
  bool rv = GetOffsetForTime(aTime, &offset);
  if (!rv) {
    return false;
  }
  uint32_t idx = mTimeMapping.IndexOfFirstElementGt(offset, SyncOffsetComparator());
  if (idx == mTimeMapping.Length()) {
    return false;
  }
  *aKeyframeTime = mTimeMapping[idx].mTimecode;
  return true;
}
} // namespace mozilla

#undef WEBM_DEBUG