Fix VP9 decoding on iOS

The current implementation of the VP9 payloader produces payloads that
are not compatible with iOS. This is because the payloader provides
only the muxing strategy called "flexible mode".

According to the VP9 RFC draft, there are two ways to wrap VP9 frames
into RTP packets: the "flexible mode" and the "non-flexible mode", with
the latter being the preferred one for live-streaming applications. In
particular, all browsers encodes VP9 RTP packets in the "non-flexible
mode", while iOS supports decoding RTP packets in this mode only, and
this is probably a problem shared by other implementations.

This patch improves the VP9 payloader by adding support for the
"non-flexible mode". The "flexible mode" is retained and a flag is
provided to perform the switch between the two modes.

codecs/vp9/bits.go

@@ -0,0 +1,62 @@
+package vp9
+
+import "fmt"
+
+func hasSpace(buf []byte, pos int, n int) error {
+	if n > ((len(buf) * 8) - pos) {
+		return fmt.Errorf("not enough bits")
+	}
+	return nil
+}
+
+func readFlag(buf []byte, pos *int) (bool, error) {
+	err := hasSpace(buf, *pos, 1)
+	if err != nil {
+		return false, err
+	}
+
+	return readFlagUnsafe(buf, pos), nil
+}
+
+func readFlagUnsafe(buf []byte, pos *int) bool {
+	b := (buf[*pos>>0x03] >> (7 - (*pos & 0x07))) & 0x01
+	*pos++
+	return b == 1
+}
+
+func readBits(buf []byte, pos *int, n int) (uint64, error) {
+	err := hasSpace(buf, *pos, n)
+	if err != nil {
+		return 0, err
+	}
+
+	return readBitsUnsafe(buf, pos, n), nil
+}
+
+func readBitsUnsafe(buf []byte, pos *int, n int) uint64 {
+	v := uint64(0)
+
+	res := 8 - (*pos & 0x07)
+	if n < res {
+		v := uint64((buf[*pos>>0x03] >> (res - n)) & (1<<n - 1))
+		*pos += n
+		return v
+	}
+
+	v = (v << res) | uint64(buf[*pos>>0x03]&(1<<res-1))
+	*pos += res
+	n -= res
+
+	for n >= 8 {
+		v = (v << 8) | uint64(buf[*pos>>0x03])
+		*pos += 8
+		n -= 8
+	}
+
+	if n > 0 {
+		v = (v << n) | uint64(buf[*pos>>0x03]>>(8-n))
+		*pos += n
+	}
+
+	return v
+}

codecs/vp9/header.go

@@ -0,0 +1,216 @@
+package vp9
+
+import (
+	"fmt"
+)
+
+// Header_ColorConfig is the color_config member of an header.
+type Header_ColorConfig struct { //nolint:revive
+	TenOrTwelveBit bool
+	BitDepth       uint8
+	ColorSpace     uint8
+	ColorRange     bool
+	SubsamplingX   bool
+	SubsamplingY   bool
+}
+
+func (c *Header_ColorConfig) unmarshal(profile uint8, buf []byte, pos *int) error {
+	if profile >= 2 {
+		var err error
+		c.TenOrTwelveBit, err = readFlag(buf, pos)
+		if err != nil {
+			return err
+		}
+
+		if c.TenOrTwelveBit {
+			c.BitDepth = 12
+		} else {
+			c.BitDepth = 10
+		}
+	} else {
+		c.BitDepth = 8
+	}
+
+	tmp, err := readBits(buf, pos, 3)
+	if err != nil {
+		return err
+	}
+	c.ColorSpace = uint8(tmp)
+
+	if c.ColorSpace != 7 {
+		var err error
+		c.ColorRange, err = readFlag(buf, pos)
+		if err != nil {
+			return err
+		}
+
+		if profile == 1 || profile == 3 {
+			err := hasSpace(buf, *pos, 3)
+			if err != nil {
+				return err
+			}
+
+			c.SubsamplingX = readFlagUnsafe(buf, pos)
+			c.SubsamplingY = readFlagUnsafe(buf, pos)
+			*pos++
+		} else {
+			c.SubsamplingX = true
+			c.SubsamplingY = true
+		}
+	} else {
+		c.ColorRange = true
+
+		if profile == 1 || profile == 3 {
+			c.SubsamplingX = false
+			c.SubsamplingY = false
+
+			err := hasSpace(buf, *pos, 1)
+			if err != nil {
+				return err
+			}
+			*pos++
+		}
+	}
+
+	return nil
+}
+
+// Header_FrameSize is the frame_size member of an header.
+type Header_FrameSize struct { //nolint:revive
+	FrameWidthMinus1  uint16
+	FrameHeightMinus1 uint16
+}
+
+func (s *Header_FrameSize) unmarshal(buf []byte, pos *int) error {
+	err := hasSpace(buf, *pos, 32)
+	if err != nil {
+		return err
+	}
+
+	s.FrameWidthMinus1 = uint16(readBitsUnsafe(buf, pos, 16))
+	s.FrameHeightMinus1 = uint16(readBitsUnsafe(buf, pos, 16))
+	return nil
+}
+
+// Header is a VP9 Frame header.
+// Specification:
+// https://storage.googleapis.com/downloads.webmproject.org/docs/vp9/vp9-bitstream-specification-v0.6-20160331-draft.pdf
+type Header struct {
+	Profile            uint8
+	ShowExistingFrame  bool
+	FrameToShowMapIdx  uint8
+	IsKeyFrame         bool
+	ShowFrame          bool
+	ErrorResilientMode bool
+	ColorConfig        *Header_ColorConfig
+	FrameSize          *Header_FrameSize
+}
+
+// Unmarshal decodes a Header.
+func (h *Header) Unmarshal(buf []byte) error {
+	pos := 0
+
+	err := hasSpace(buf, pos, 4)
+	if err != nil {
+		return err
+	}
+
+	frameMarker := readBitsUnsafe(buf, &pos, 2)
+	if frameMarker != 2 {
+		return fmt.Errorf("invalid frame marker")
+	}
+
+	profileLowBit := uint8(readBitsUnsafe(buf, &pos, 1))
+	profileHighBit := uint8(readBitsUnsafe(buf, &pos, 1))
+	h.Profile = profileHighBit<<1 + profileLowBit
+
+	if h.Profile == 3 {
+		err := hasSpace(buf, pos, 1)
+		if err != nil {
+			return err
+		}
+		pos++
+	}
+
+	h.ShowExistingFrame, err = readFlag(buf, &pos)
+	if err != nil {
+		return err
+	}
+
+	if h.ShowExistingFrame {
+		tmp, err := readBits(buf, &pos, 3)
+		if err != nil {
+			return err
+		}
+		h.FrameToShowMapIdx = uint8(tmp)
+
+		return nil
+	}
+
+	err = hasSpace(buf, pos, 3)
+	if err != nil {
+		return err
+	}
+
+	h.IsKeyFrame = !readFlagUnsafe(buf, &pos)
+	h.ShowFrame = readFlagUnsafe(buf, &pos)
+	h.ErrorResilientMode = readFlagUnsafe(buf, &pos)
+
+	if h.IsKeyFrame {
+		err := hasSpace(buf, pos, 24)
+		if err != nil {
+			return err
+		}
+
+		frameSyncByte0 := uint8(readBitsUnsafe(buf, &pos, 8))
+		if frameSyncByte0 != 0x49 {
+			return fmt.Errorf("wrong frame_sync_byte_0")
+		}
+
+		frameSyncByte1 := uint8(readBitsUnsafe(buf, &pos, 8))
+		if frameSyncByte1 != 0x83 {
+			return fmt.Errorf("wrong frame_sync_byte_1")
+		}
+
+		frameSyncByte2 := uint8(readBitsUnsafe(buf, &pos, 8))
+		if frameSyncByte2 != 0x42 {
+			return fmt.Errorf("wrong frame_sync_byte_2")
+		}
+
+		h.ColorConfig = &Header_ColorConfig{}
+		err = h.ColorConfig.unmarshal(h.Profile, buf, &pos)
+		if err != nil {
+			return err
+		}
+
+		h.FrameSize = &Header_FrameSize{}
+		err = h.FrameSize.unmarshal(buf, &pos)
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+// Width returns the video width.
+func (h Header) Width() int {
+	return int(h.FrameSize.FrameWidthMinus1) + 1
+}
+
+// Height returns the video height.
+func (h Header) Height() int {
+	return int(h.FrameSize.FrameHeightMinus1) + 1
+}
+
+// ChromaSubsampling returns the chroma subsampling format, in ISO-BMFF/vpcC format.
+func (h Header) ChromaSubsampling() uint8 {
+	switch {
+	case !h.ColorConfig.SubsamplingX && !h.ColorConfig.SubsamplingY:
+		return 3 // 4:4:4
+	case h.ColorConfig.SubsamplingX && !h.ColorConfig.SubsamplingY:
+		return 2 // 4:2:2
+	default:
+		return 1 // 4:2:0 colocated with luma
+	}
+}