This commit is contained in:
seydx
2026-01-12 03:15:48 +01:00
parent 659a042c42
commit 406159cce5
9 changed files with 1521 additions and 2309 deletions
+139 -245
View File
@@ -14,6 +14,47 @@ import (
"github.com/AlexxIT/go2rtc/pkg/wyze/tutk"
)
const (
FrameSize1080P = 0
FrameSize360P = 1
FrameSize720P = 2
FrameSize2K = 3
)
const (
BitrateMax uint16 = 0xF0
BitrateSD uint16 = 0x3C
)
const (
QualityUnknown = 0
QualityMax = 1
QualityHigh = 2
QualityMiddle = 3
QualityLow = 4
QualityMin = 5
)
const (
MediaTypeVideo = 1
MediaTypeAudio = 2
MediaTypeReturnAudio = 3
MediaTypeRDT = 4
)
const (
KCmdAuth = 10000
KCmdChallenge = 10001
KCmdChallengeResp = 10002
KCmdAuthResult = 10003
KCmdAuthWithPayload = 10008
KCmdAuthSuccess = 10009
KCmdControlChannel = 10010
KCmdControlChannelResp = 10011
KCmdSetResolution = 10056
KCmdSetResolutionResp = 10057
)
type Client struct {
conn *tutk.Conn
@@ -36,6 +77,11 @@ type Client struct {
audioChannels uint8
}
type AuthResponse struct {
ConnectionRes string `json:"connectionRes"`
CameraInfo map[string]any `json:"cameraInfo"`
}
func Dial(rawURL string) (*Client, error) {
u, err := url.Parse(rawURL)
if err != nil {
@@ -107,11 +153,11 @@ func (c *Client) SetResolution(sd bool) error {
var bitrate uint16
if sd {
frameSize = tutk.FrameSize360P
bitrate = tutk.BitrateSD
frameSize = FrameSize360P
bitrate = BitrateSD
} else {
frameSize = tutk.FrameSize2K
bitrate = tutk.BitrateMax
frameSize = FrameSize2K
bitrate = BitrateMax
}
if c.verbose {
@@ -119,120 +165,33 @@ func (c *Client) SetResolution(sd bool) error {
}
k10056 := c.buildK10056(frameSize, bitrate)
if err := c.conn.SendIOCtrl(tutk.KCmdSetResolution, k10056); err != nil {
return fmt.Errorf("wyze: K10056 send failed: %w", err)
}
// Wait for K10057 response
cmdID, data, err := c.conn.RecvIOCtrl(1 * time.Second)
if err != nil {
return err
}
if c.verbose {
fmt.Printf("[Wyze] SetResolution response: K%d (%d bytes)\n", cmdID, len(data))
}
if cmdID == tutk.KCmdSetResolutionResp && len(data) >= 17 {
result := data[16]
if c.verbose {
fmt.Printf("[Wyze] K10057 result: %d\n", result)
}
}
return nil
_, err := c.conn.WriteAndWaitIOCtrl(KCmdSetResolution, k10056, KCmdSetResolutionResp, 5*time.Second)
return err
}
func (c *Client) StartVideo() error {
k10010 := c.buildK10010(tutk.MediaTypeVideo, true)
if c.verbose {
fmt.Printf("[Wyze] TX K10010 video (%d bytes): % x\n", len(k10010), k10010)
}
if err := c.conn.SendIOCtrl(tutk.KCmdControlChannel, k10010); err != nil {
return fmt.Errorf("K10010 video send failed: %w", err)
}
// Wait for K10011 response
cmdID, data, err := c.conn.RecvIOCtrl(5 * time.Second)
if err != nil {
return fmt.Errorf("K10011 video recv failed: %w", err)
}
if c.verbose {
fmt.Printf("[Wyze] K10011 video response: cmdID=%d, len=%d\n", cmdID, len(data))
if len(data) >= 18 {
fmt.Printf("[Wyze] K10011 video: media=%d status=%d\n", data[16], data[17])
}
}
return nil
k10010 := c.buildK10010(MediaTypeVideo, true)
_, err := c.conn.WriteAndWaitIOCtrl(KCmdControlChannel, k10010, KCmdControlChannelResp, 5*time.Second)
return err
}
func (c *Client) StartAudio() error {
k10010 := c.buildK10010(tutk.MediaTypeAudio, true)
if c.verbose {
fmt.Printf("[Wyze] TX K10010 audio (%d bytes): % x\n", len(k10010), k10010)
}
if err := c.conn.SendIOCtrl(tutk.KCmdControlChannel, k10010); err != nil {
return fmt.Errorf("K10010 audio send failed: %w", err)
}
// Wait for K10011 response
cmdID, data, err := c.conn.RecvIOCtrl(5 * time.Second)
if err != nil {
return fmt.Errorf("K10011 audio recv failed: %w", err)
}
if c.verbose {
fmt.Printf("[Wyze] K10011 audio response: cmdID=%d, len=%d\n", cmdID, len(data))
if len(data) >= 18 {
fmt.Printf("[Wyze] K10011 audio: media=%d status=%d\n", data[16], data[17])
}
}
return nil
k10010 := c.buildK10010(MediaTypeAudio, true)
_, err := c.conn.WriteAndWaitIOCtrl(KCmdControlChannel, k10010, KCmdControlChannelResp, 5*time.Second)
return err
}
func (c *Client) StartIntercom() error {
if c.conn.IsBackchannelReady() {
return nil // Already enabled
return nil
}
if c.verbose {
fmt.Printf("[Wyze] Sending K10010 (enable return audio)\n")
k10010 := c.buildK10010(MediaTypeReturnAudio, true)
if _, err := c.conn.WriteAndWaitIOCtrl(KCmdControlChannel, k10010, KCmdControlChannelResp, 5*time.Second); err != nil {
return err
}
k10010 := c.buildK10010(tutk.MediaTypeReturnAudio, true)
if err := c.conn.SendIOCtrl(tutk.KCmdControlChannel, k10010); err != nil {
return fmt.Errorf("K10010 send failed: %w", err)
}
// Wait for K10011 response
cmdID, data, err := c.conn.RecvIOCtrl(5 * time.Second)
if err != nil {
return fmt.Errorf("K10011 recv failed: %w", err)
}
if c.verbose {
fmt.Printf("[Wyze] K10011 response: cmdID=%d, len=%d\n", cmdID, len(data))
}
// Perform DTLS server handshake on backchannel (camera connects to us)
if c.verbose {
fmt.Printf("[Wyze] Starting speaker channel DTLS handshake\n")
}
if err := c.conn.AVServStart(); err != nil {
return fmt.Errorf("speaker channel handshake failed: %w", err)
}
if c.verbose {
fmt.Printf("[Wyze] Backchannel ready\n")
}
return nil
return c.conn.AVServStart()
}
func (c *Client) ReadPacket() (*tutk.Packet, error) {
@@ -324,23 +283,10 @@ func (c *Client) doAVLogin() error {
}
func (c *Client) doKAuth() error {
if c.verbose {
fmt.Printf("[Wyze] Starting K-command authentication\n")
}
// Step 1: Send K10000
k10000 := c.buildK10000()
if err := c.conn.SendIOCtrl(tutk.KCmdAuth, k10000); err != nil {
return fmt.Errorf("wyze: K10000 send failed: %w", err)
}
// Step 2: Wait for K10001
cmdID, data, err := c.conn.RecvIOCtrl(10 * time.Second)
// Step 1: K10000 -> K10001
data, err := c.conn.WriteAndWaitIOCtrl(KCmdAuth, c.buildK10000(), KCmdChallenge, 10*time.Second)
if err != nil {
return fmt.Errorf("wyze: K10001 recv failed: %w", err)
}
if cmdID != tutk.KCmdChallenge {
return fmt.Errorf("wyze: expected K10001, got K%d", cmdID)
return fmt.Errorf("wyze: K10001 failed: %w", err)
}
challenge, status, err := c.parseK10001(data)
@@ -348,45 +294,18 @@ func (c *Client) doKAuth() error {
return fmt.Errorf("wyze: K10001 parse failed: %w", err)
}
if c.verbose {
fmt.Printf("[Wyze] K10001 received, status=%d\n", status)
}
// Step 3: Send K10008
k10008 := c.buildK10008(challenge, status)
if err := c.conn.SendIOCtrl(tutk.KCmdChallengeResp, k10008); err != nil {
return fmt.Errorf("wyze: K10008 send failed: %w", err)
}
// Step 4: Wait for K10009
cmdID, data, err = c.conn.RecvIOCtrl(10 * time.Second)
// Step 2: K10002 -> K10009
data, err = c.conn.WriteAndWaitIOCtrl(KCmdChallengeResp, c.buildK10002(challenge, status), KCmdAuthResult, 10*time.Second)
if err != nil {
return fmt.Errorf("wyze: K10009 recv failed: %w", err)
return fmt.Errorf("wyze: K10009 failed: %w", err)
}
if cmdID != tutk.KCmdAuthSuccess {
return fmt.Errorf("wyze: expected K10009, got K%d", cmdID)
}
authResp, err := c.parseK10009(data)
if err != nil {
return fmt.Errorf("wyze: K10009 parse failed: %w", err)
}
// Parse capabilities
authResp, _ := c.parseK10003(data)
if authResp != nil && authResp.CameraInfo != nil {
if c.verbose {
fmt.Printf("[Wyze] CameraInfo authResp: ")
b, _ := json.Marshal(authResp)
fmt.Printf("%s\n", b)
}
// Audio receiving support
if audio, ok := authResp.CameraInfo["audio"].(bool); ok {
c.hasAudio = audio
} else {
c.hasAudio = true // Default to true
c.hasAudio = true
}
} else {
c.hasAudio = true
@@ -394,9 +313,6 @@ func (c *Client) doKAuth() error {
if avResp := c.conn.GetAVLoginResponse(); avResp != nil {
c.hasIntercom = avResp.TwoWayStreaming == 1
if c.verbose {
fmt.Printf("[Wyze] two_way_streaming=%d (from AV Login Response)\n", avResp.TwoWayStreaming)
}
}
if c.verbose {
@@ -407,94 +323,72 @@ func (c *Client) doKAuth() error {
}
func (c *Client) buildK10000() []byte {
// 137 = G.711 μ-law (PCMU)
// 138 = G.711 A-law (PCMA)
// 140 = PCM 16-bit
jsonPayload := []byte(`{"cameraInfo":{"audioEncoderList":[137,138,140]}}`)
buf := make([]byte, 16+len(jsonPayload))
buf[0] = 'H'
buf[1] = 'L'
buf[2] = 5
binary.LittleEndian.PutUint16(buf[4:], tutk.KCmdAuth)
binary.LittleEndian.PutUint16(buf[6:], uint16(len(jsonPayload)))
copy(buf[16:], jsonPayload)
return buf
json := []byte(`{"cameraInfo":{"audioEncoderList":[137,138,140]}}`) // 137=PCMU, 138=PCMA, 140=PCM
b := make([]byte, 16+len(json))
copy(b, "HL") // magic
b[2] = 5 // version
binary.LittleEndian.PutUint16(b[4:], KCmdAuth) // 10000
binary.LittleEndian.PutUint16(b[6:], uint16(len(json))) // payload len
copy(b[16:], json)
return b
}
func (c *Client) buildK10002(challenge []byte, status byte) []byte {
response := crypto.GenerateChallengeResponse(challenge, c.enr, status)
buf := make([]byte, 38)
buf[0] = 'H'
buf[1] = 'L'
buf[2] = 5
binary.LittleEndian.PutUint16(buf[4:], tutk.KCmdChallengeResp)
buf[6] = 22 // Payload length
if len(response) >= 16 {
copy(buf[16:], response[:16])
}
if len(c.uid) >= 4 {
copy(buf[32:], c.uid[:4])
}
buf[36] = 1 // Video flag (0 = disabled, 1 = enabled > will start video stream immediately)
buf[37] = 1 // Audio flag (0 = disabled, 1 = enabled > will start audio stream immediately)
return buf
resp := crypto.GenerateChallengeResponse(challenge, c.enr, status)
b := make([]byte, 38)
copy(b, "HL") // magic
b[2] = 5 // version
binary.LittleEndian.PutUint16(b[4:], KCmdChallengeResp) // 10002
b[6] = 22 // payload len
copy(b[16:], resp[:16]) // challenge response
copy(b[32:], c.uid[:4]) // UID prefix
b[36] = 1 // video enabled
b[37] = 1 // audio enabled
return b
}
func (c *Client) buildK10008(challenge []byte, status byte) []byte {
response := crypto.GenerateChallengeResponse(challenge, c.enr, status)
openUserID := []byte(c.enr)
payloadLen := 16 + 4 + 1 + 1 + 1 + len(openUserID)
buf := make([]byte, 16+payloadLen)
buf[0] = 'H'
buf[1] = 'L'
buf[2] = 5 // Protocol version
binary.LittleEndian.PutUint16(buf[4:], tutk.KCmdAuthWithPayload) // 10008
binary.LittleEndian.PutUint16(buf[6:], uint16(payloadLen))
copy(buf[16:], response[:16]) // Challenge response
copy(buf[32:], c.uid[:4]) // UID prefix
buf[36] = 1 // Video enabled
buf[37] = 1 // Audio enabled
buf[38] = byte(len(openUserID))
copy(buf[39:], openUserID)
return buf
resp := crypto.GenerateChallengeResponse(challenge, c.enr, status)
userID := []byte(c.enr)
payloadLen := 16 + 4 + 1 + 1 + 1 + len(userID)
b := make([]byte, 16+payloadLen)
copy(b, "HL") // magic
b[2] = 5 // version
binary.LittleEndian.PutUint16(b[4:], KCmdAuthWithPayload) // 10008
binary.LittleEndian.PutUint16(b[6:], uint16(payloadLen)) // payload len
copy(b[16:], resp[:16]) // challenge response
copy(b[32:], c.uid[:4]) // UID prefix
b[36] = 1 // video enabled
b[37] = 1 // audio enabled
b[38] = byte(len(userID)) // userID len
copy(b[39:], userID) // userID
return b
}
func (c *Client) buildK10010(mediaType byte, enabled bool) []byte {
buf := make([]byte, 18)
buf[0] = 'H'
buf[1] = 'L'
buf[2] = 5 // Version
binary.LittleEndian.PutUint16(buf[4:], tutk.KCmdControlChannel) // 0x271a = 10010
binary.LittleEndian.PutUint16(buf[6:], 2) // Payload length = 2
buf[16] = mediaType // 1=Video, 2=Audio, 3=ReturnAudio
if enabled {
buf[17] = 1
} else {
buf[17] = 2
b := make([]byte, 18)
copy(b, "HL") // magic
b[2] = 5 // version
binary.LittleEndian.PutUint16(b[4:], KCmdControlChannel) // 10010
binary.LittleEndian.PutUint16(b[6:], 2) // payload len
b[16] = mediaType // 1=video, 2=audio, 3=return audio
b[17] = 1 // 1=enable, 2=disable
if !enabled {
b[17] = 2
}
return buf
return b
}
func (c *Client) buildK10056(frameSize uint8, bitrate uint16) []byte {
buf := make([]byte, 21)
buf[0] = 'H'
buf[1] = 'L'
buf[2] = 5 // Version
binary.LittleEndian.PutUint16(buf[4:], tutk.KCmdSetResolution) // 0x2748 = 10056
binary.LittleEndian.PutUint16(buf[6:], 5) // Payload length = 5
buf[16] = frameSize + 1 // 4 = HD
binary.LittleEndian.PutUint16(buf[17:], bitrate) // 0x00f0 = 240
// buf[19], buf[20] = FPS (0 = auto)
return buf
b := make([]byte, 21)
copy(b, "HL") // magic
b[2] = 5 // version
binary.LittleEndian.PutUint16(b[4:], KCmdSetResolution) // 10056
binary.LittleEndian.PutUint16(b[6:], 5) // payload len
b[16] = frameSize + 1 // frame size
binary.LittleEndian.PutUint16(b[17:], bitrate) // bitrate
// b[19:21] = FPS (0 = auto)
return b
}
func (c *Client) parseK10001(data []byte) (challenge []byte, status byte, err error) {
@@ -511,7 +405,7 @@ func (c *Client) parseK10001(data []byte) (challenge []byte, status byte, err er
}
cmdID := binary.LittleEndian.Uint16(data[4:])
if cmdID != tutk.KCmdChallenge {
if cmdID != KCmdChallenge {
return nil, 0, fmt.Errorf("expected cmdID 10001, got %d", cmdID)
}
@@ -522,31 +416,31 @@ func (c *Client) parseK10001(data []byte) (challenge []byte, status byte, err er
return challenge, status, nil
}
func (c *Client) parseK10003(data []byte) (*tutk.AuthResponse, error) {
func (c *Client) parseK10003(data []byte) (*AuthResponse, error) {
if c.verbose {
fmt.Printf("[Wyze] parseK10003: received %d bytes\n", len(data))
}
if len(data) < 16 {
return &tutk.AuthResponse{}, nil
return &AuthResponse{}, nil
}
if data[0] != 'H' || data[1] != 'L' {
return &tutk.AuthResponse{}, nil
return &AuthResponse{}, nil
}
cmdID := binary.LittleEndian.Uint16(data[4:])
textLen := binary.LittleEndian.Uint16(data[6:])
if cmdID != tutk.KCmdAuthResult {
return &tutk.AuthResponse{}, nil
if cmdID != KCmdAuthResult {
return &AuthResponse{}, nil
}
if len(data) > 16 && textLen > 0 {
jsonData := data[16:]
for i := range jsonData {
if jsonData[i] == '{' {
var resp tutk.AuthResponse
var resp AuthResponse
if err := json.Unmarshal(jsonData[i:], &resp); err == nil {
if c.verbose {
fmt.Printf("[Wyze] parseK10003: parsed JSON\n")
@@ -558,34 +452,34 @@ func (c *Client) parseK10003(data []byte) (*tutk.AuthResponse, error) {
}
}
return &tutk.AuthResponse{}, nil
return &AuthResponse{}, nil
}
func (c *Client) parseK10009(data []byte) (*tutk.AuthResponse, error) {
func (c *Client) parseK10009(data []byte) (*AuthResponse, error) {
if c.verbose {
fmt.Printf("[Wyze] parseK10009: received %d bytes\n", len(data))
}
if len(data) < 16 {
return &tutk.AuthResponse{}, nil
return &AuthResponse{}, nil
}
if data[0] != 'H' || data[1] != 'L' {
return &tutk.AuthResponse{}, nil
return &AuthResponse{}, nil
}
cmdID := binary.LittleEndian.Uint16(data[4:])
textLen := binary.LittleEndian.Uint16(data[6:])
if cmdID != tutk.KCmdAuthSuccess {
return &tutk.AuthResponse{}, nil
if cmdID != KCmdAuthSuccess {
return &AuthResponse{}, nil
}
if len(data) > 16 && textLen > 0 {
jsonData := data[16:]
for i := range jsonData {
if jsonData[i] == '{' {
var resp tutk.AuthResponse
var resp AuthResponse
if err := json.Unmarshal(jsonData[i:], &resp); err == nil {
if c.verbose {
fmt.Printf("[Wyze] parseK10009: parsed JSON\n")
@@ -597,5 +491,5 @@ func (c *Client) parseK10009(data []byte) (*tutk.AuthResponse, error) {
}
}
return &tutk.AuthResponse{}, nil
return &AuthResponse{}, nil
}
-126
View File
@@ -1,126 +0,0 @@
package tutk
import (
"encoding/binary"
"github.com/AlexxIT/go2rtc/pkg/aac"
)
const FrameInfoSize = 40
// Wire format (little-endian) - Wyze extended FRAMEINFO:
//
// [0-1] codec_id uint16 (0x004e=H.264, 0x0050=H.265, 0x0088=AAC)
// [2] flags uint8 (Video: 0=P/1=I, Audio: sr_idx<<2|bits16<<1|ch)
// [3] cam_index uint8
// [4] online_num uint8
// [5] framerate uint8 (FPS, e.g. 20)
// [6] frame_size uint8 (Resolution: 1=1080P, 2=360P, 4=2K)
// [7] bitrate uint8 (e.g. 0xF0=240)
// [8-11] timestamp_us uint32 (microseconds component)
// [12-15] timestamp uint32 (Unix timestamp in seconds)
// [16-19] payload_sz uint32 (frame payload size)
// [20-23] frame_no uint32 (frame number)
// [24-39] device_id 16 bytes (MAC address + padding)
type FrameInfo struct {
CodecID uint16
Flags uint8
CamIndex uint8
OnlineNum uint8
Framerate uint8 // FPS (e.g. 20)
FrameSize uint8 // Resolution: 1=1080P, 2=360P, 4=2K
Bitrate uint8 // Bitrate value (e.g. 240)
TimestampUS uint32
Timestamp uint32
PayloadSize uint32
FrameNo uint32
}
// Resolution constants (as received in FrameSize field)
// Note: Some cameras only support 2K + 360P, others support 1080P + 360P
// The actual resolution depends on camera model!
const (
ResolutionUnknown = 0
ResolutionSD = 1 // 360P (640x360) on 2K cameras, or 1080P on older cams
Resolution360P = 2 // 360P (640x360)
Resolution2K = 4 // 2K (2560x1440)
)
func (fi *FrameInfo) IsKeyframe() bool {
return fi.Flags == 0x01
}
// Resolution returns a human-readable resolution string
func (fi *FrameInfo) Resolution() string {
switch fi.FrameSize {
case ResolutionSD:
return "SD" // Could be 360P or 1080P depending on camera
case Resolution360P:
return "360P"
case Resolution2K:
return "2K"
default:
return "unknown"
}
}
func (fi *FrameInfo) SampleRate() uint32 {
srIdx := (fi.Flags >> 2) & 0x0F
return uint32(SampleRateValue(srIdx))
}
func (fi *FrameInfo) Channels() uint8 {
if fi.Flags&0x01 == 1 {
return 2
}
return 1
}
func (fi *FrameInfo) IsVideo() bool {
return IsVideoCodec(fi.CodecID)
}
func (fi *FrameInfo) IsAudio() bool {
return IsAudioCodec(fi.CodecID)
}
func ParseFrameInfo(data []byte) *FrameInfo {
if len(data) < FrameInfoSize {
return nil
}
offset := len(data) - FrameInfoSize
fi := data[offset:]
return &FrameInfo{
CodecID: binary.LittleEndian.Uint16(fi),
Flags: fi[2],
CamIndex: fi[3],
OnlineNum: fi[4],
Framerate: fi[5],
FrameSize: fi[6],
Bitrate: fi[7],
TimestampUS: binary.LittleEndian.Uint32(fi[8:]),
Timestamp: binary.LittleEndian.Uint32(fi[12:]),
PayloadSize: binary.LittleEndian.Uint32(fi[16:]),
FrameNo: binary.LittleEndian.Uint32(fi[20:]),
}
}
func ParseAudioParams(payload []byte, fi *FrameInfo) (sampleRate uint32, channels uint8) {
// Try ADTS header first (more reliable than FRAMEINFO flags)
if aac.IsADTS(payload) {
codec := aac.ADTSToCodec(payload)
if codec != nil {
return codec.ClockRate, codec.Channels
}
}
// Fallback to FRAMEINFO flags
if fi != nil {
return fi.SampleRate(), fi.Channels()
}
// Default values
return 16000, 1
}
-64
View File
@@ -1,64 +0,0 @@
package tutk
import (
"fmt"
"net"
"time"
)
type ChannelAdapter struct {
conn *Conn
channel uint8
}
func (a *ChannelAdapter) ReadFrom(p []byte) (n int, addr net.Addr, err error) {
var buf chan []byte
if a.channel == IOTCChannelMain {
buf = a.conn.mainBuf
} else {
buf = a.conn.speakerBuf
}
select {
case data := <-buf:
n = copy(p, data)
if a.conn.verbose && len(data) >= 1 {
fmt.Printf("[ChannelAdapter] ch=%d ReadFrom: len=%d contentType=%d\n",
a.channel, len(data), data[0])
}
return n, a.conn.addr, nil
case <-a.conn.done:
return 0, nil, net.ErrClosed
}
}
func (a *ChannelAdapter) WriteTo(p []byte, addr net.Addr) (n int, err error) {
if a.conn.verbose {
fmt.Printf("[IOTC TX] channel=%d size=%d\n", a.channel, len(p))
}
_, err = a.conn.sendIOTC(p, a.channel)
if err != nil {
return 0, err
}
return len(p), nil
}
func (a *ChannelAdapter) Close() error {
return nil
}
func (a *ChannelAdapter) LocalAddr() net.Addr {
return &net.UDPAddr{IP: net.IPv4(0, 0, 0, 0), Port: 0}
}
func (a *ChannelAdapter) SetDeadline(time.Time) error {
return nil
}
func (a *ChannelAdapter) SetReadDeadline(time.Time) error {
return nil
}
func (a *ChannelAdapter) SetWriteDeadline(time.Time) error {
return nil
}
+525 -1411
View File
File diff suppressed because it is too large Load Diff
-306
View File
@@ -1,306 +0,0 @@
package tutk
const (
CodecUnknown uint16 = 0x00 // Unknown codec
CodecMPEG4 uint16 = 0x4C // 76 - MPEG4
CodecH263 uint16 = 0x4D // 77 - H.263
CodecH264 uint16 = 0x4E // 78 - H.264/AVC
CodecMJPEG uint16 = 0x4F // 79 - MJPEG
CodecH265 uint16 = 0x50 // 80 - H.265/HEVC
)
const (
AudioCodecAACRaw uint16 = 0x86 // 134 - AAC raw format
AudioCodecAACADTS uint16 = 0x87 // 135 - AAC with ADTS header
AudioCodecAACLATM uint16 = 0x88 // 136 - AAC with LATM format
AudioCodecG711U uint16 = 0x89 // 137 - G.711 μ-law (PCMU)
AudioCodecG711A uint16 = 0x8A // 138 - G.711 A-law (PCMA)
AudioCodecADPCM uint16 = 0x8B // 139 - ADPCM
AudioCodecPCM uint16 = 0x8C // 140 - PCM 16-bit signed LE
AudioCodecSPEEX uint16 = 0x8D // 141 - Speex
AudioCodecMP3 uint16 = 0x8E // 142 - MP3
AudioCodecG726 uint16 = 0x8F // 143 - G.726
AudioCodecAACWyze uint16 = 0x90 // 144 - Wyze AAC
AudioCodecOpus uint16 = 0x92 // 146 - Opus codec
)
const (
SampleRate8K uint8 = 0x00 // 8000 Hz
SampleRate11K uint8 = 0x01 // 11025 Hz
SampleRate12K uint8 = 0x02 // 12000 Hz
SampleRate16K uint8 = 0x03 // 16000 Hz
SampleRate22K uint8 = 0x04 // 22050 Hz
SampleRate24K uint8 = 0x05 // 24000 Hz
SampleRate32K uint8 = 0x06 // 32000 Hz
SampleRate44K uint8 = 0x07 // 44100 Hz
SampleRate48K uint8 = 0x08 // 48000 Hz
)
var SampleRates = map[uint8]int{
SampleRate8K: 8000,
SampleRate11K: 11025,
SampleRate12K: 12000,
SampleRate16K: 16000,
SampleRate22K: 22050,
SampleRate24K: 24000,
SampleRate32K: 32000,
SampleRate44K: 44100,
SampleRate48K: 48000,
}
var SamplesPerFrame = map[uint16]uint32{
AudioCodecAACRaw: 1024, // AAC frame = 1024 samples
AudioCodecAACADTS: 1024,
AudioCodecAACLATM: 1024,
AudioCodecAACWyze: 1024,
AudioCodecG711U: 160, // G.711 typically 20ms = 160 samples at 8kHz
AudioCodecG711A: 160,
AudioCodecPCM: 160,
AudioCodecADPCM: 160,
AudioCodecSPEEX: 160,
AudioCodecMP3: 1152, // MP3 frame = 1152 samples
AudioCodecG726: 160,
AudioCodecOpus: 960, // Opus typically 20ms = 960 samples at 48kHz
}
const (
IOTypeVideoStart = 0x01FF
IOTypeVideoStop = 0x02FF
IOTypeAudioStart = 0x0300
IOTypeAudioStop = 0x0301
IOTypeSpeakerStart = 0x0350
IOTypeSpeakerStop = 0x0351
IOTypeGetAudioOutFormatReq = 0x032A
IOTypeGetAudioOutFormatRes = 0x032B
IOTypeSetStreamCtrlReq = 0x0320
IOTypeSetStreamCtrlRes = 0x0321
IOTypeGetStreamCtrlReq = 0x0322
IOTypeGetStreamCtrlRes = 0x0323
IOTypeDevInfoReq = 0x0340
IOTypeDevInfoRes = 0x0341
IOTypeGetSupportStreamReq = 0x0344
IOTypeGetSupportStreamRes = 0x0345
IOTypeSetRecordReq = 0x0310
IOTypeSetRecordRes = 0x0311
IOTypeGetRecordReq = 0x0312
IOTypeGetRecordRes = 0x0313
IOTypePTZCommand = 0x1001
IOTypeReceiveFirstFrame = 0x1002
IOTypeGetEnvironmentReq = 0x030A
IOTypeGetEnvironmentRes = 0x030B
IOTypeSetVideoModeReq = 0x030C
IOTypeSetVideoModeRes = 0x030D
IOTypeGetVideoModeReq = 0x030E
IOTypeGetVideoModeRes = 0x030F
IOTypeSetTimeReq = 0x0316
IOTypeSetTimeRes = 0x0317
IOTypeGetTimeReq = 0x0318
IOTypeGetTimeRes = 0x0319
IOTypeSetWifiReq = 0x0102
IOTypeSetWifiRes = 0x0103
IOTypeGetWifiReq = 0x0104
IOTypeGetWifiRes = 0x0105
IOTypeListWifiAPReq = 0x0106
IOTypeListWifiAPRes = 0x0107
IOTypeSetMotionDetectReq = 0x0306
IOTypeSetMotionDetectRes = 0x0307
IOTypeGetMotionDetectReq = 0x0308
IOTypeGetMotionDetectRes = 0x0309
)
// OLD DTLS Protocol (IOTC/TransCode) commands and sizes
const (
CmdDiscoReq uint16 = 0x0601
CmdDiscoRes uint16 = 0x0602
CmdSessionReq uint16 = 0x0402
CmdSessionRes uint16 = 0x0404
CmdDataTX uint16 = 0x0407
CmdDataRX uint16 = 0x0408
CmdKeepaliveReq uint16 = 0x0427
CmdKeepaliveRes uint16 = 0x0428
OldProtoHeaderSize = 16
OldProtoMinPacketSize = 16
OldProtoDiscoBodySize = 72
OldProtoDiscoPacketSize = OldProtoHeaderSize + OldProtoDiscoBodySize
OldProtoSessionBodySize = 36
OldProtoSessionPacketSize = OldProtoHeaderSize + OldProtoSessionBodySize
)
// NEW DTLS Protocol (0xCC51) commands and sizes
const (
MagicNewProto uint16 = 0xCC51
CmdNewProtoDiscovery uint16 = 0x1002
CmdNewProtoDTLS uint16 = 0x1502
NewProtoPayloadSize uint16 = 0x0028
NewProtoPacketSize = 52
NewProtoHeaderSize = 28
NewProtoAuthSize = 20
)
const (
UIDSize = 20
RandomIDSize = 8
)
const (
MagicAVLoginResp uint16 = 0x2100
MagicIOCtrl uint16 = 0x7000
MagicChannelMsg uint16 = 0x1000
MagicACK uint16 = 0x0009
MagicAVLogin1 uint16 = 0x0000
MagicAVLogin2 uint16 = 0x2000
)
const (
ProtocolVersion uint16 = 0x000c // Version 12
)
const (
DefaultCapabilities uint32 = 0x001f07fb
)
const (
KCmdAuth = 10000
KCmdChallenge = 10001
KCmdChallengeResp = 10002
KCmdAuthResult = 10003
KCmdAuthWithPayload = 10008
KCmdAuthSuccess = 10009
KCmdControlChannel = 10010
KCmdControlChannelResp = 10011
KCmdSetResolution = 10056
KCmdSetResolutionResp = 10057
)
const (
MediaTypeVideo = 1
MediaTypeAudio = 2
MediaTypeReturnAudio = 3
MediaTypeRDT = 4
)
const (
IOTCChannelMain = 0 // Main AV channel (we = DTLS Client, camera = Server)
IOTCChannelBack = 1 // Backchannel for Return Audio (we = DTLS Server, camera = Client)
)
const (
BitrateMax uint16 = 0xF0 // 240 KB/s
BitrateSD uint16 = 0x3C // 60 KB/s
)
const (
FrameSize1080P = 0
FrameSize360P = 1
FrameSize720P = 2
FrameSize2K = 3
)
const (
QualityUnknown = 0
QualityMax = 1
QualityHigh = 2
QualityMiddle = 3
QualityLow = 4
QualityMin = 5
)
func CodecName(id uint16) string {
switch id {
case CodecH264:
return "H264"
case CodecH265:
return "H265"
case CodecMPEG4:
return "MPEG4"
case CodecH263:
return "H263"
case CodecMJPEG:
return "MJPEG"
default:
return "Unknown"
}
}
func AudioCodecName(id uint16) string {
switch id {
case AudioCodecG711U:
return "PCMU"
case AudioCodecG711A:
return "PCMA"
case AudioCodecPCM:
return "PCM"
case AudioCodecAACLATM, AudioCodecAACRaw, AudioCodecAACADTS, AudioCodecAACWyze:
return "AAC"
case AudioCodecOpus:
return "Opus"
case AudioCodecSPEEX:
return "Speex"
case AudioCodecMP3:
return "MP3"
case AudioCodecG726:
return "G726"
case AudioCodecADPCM:
return "ADPCM"
default:
return "Unknown"
}
}
func SampleRateValue(enum uint8) int {
if rate, ok := SampleRates[enum]; ok {
return rate
}
return 16000 // Default
}
func SampleRateIndex(hz uint32) uint8 {
switch hz {
case 8000:
return SampleRate8K
case 11025:
return SampleRate11K
case 12000:
return SampleRate12K
case 16000:
return SampleRate16K
case 22050:
return SampleRate22K
case 24000:
return SampleRate24K
case 32000:
return SampleRate32K
case 44100:
return SampleRate44K
case 48000:
return SampleRate48K
default:
return SampleRate16K // Default
}
}
func BuildAudioFlags(sampleRate uint32, bits16 bool, stereo bool) uint8 {
flags := SampleRateIndex(sampleRate) << 2
if bits16 {
flags |= 0x02
}
if stereo {
flags |= 0x01
}
return flags
}
func IsVideoCodec(id uint16) bool {
return id >= CodecMPEG4 && id <= CodecH265
}
func IsAudioCodec(id uint16) bool {
return id >= AudioCodecAACRaw && id <= AudioCodecOpus
}
func GetSamplesPerFrame(codecID uint16) uint32 {
if samples, ok := SamplesPerFrame[codecID]; ok {
return samples
}
return 1024 // Default to AAC
}
+74
View File
@@ -0,0 +1,74 @@
package tutk
import (
"net"
"time"
"github.com/pion/dtls/v3"
)
func NewDtlsClient(c *Conn, channel uint8, psk []byte) (*dtls.Conn, error) {
adapter := &ChannelAdapter{conn: c, channel: channel}
return dtls.Client(adapter, c.addr, buildDtlsConfig(psk, false))
}
func NewDtlsServer(c *Conn, channel uint8, psk []byte) (*dtls.Conn, error) {
adapter := &ChannelAdapter{conn: c, channel: channel}
return dtls.Server(adapter, c.addr, buildDtlsConfig(psk, true))
}
func buildDtlsConfig(psk []byte, isServer bool) *dtls.Config {
config := &dtls.Config{
PSK: func(hint []byte) ([]byte, error) {
return psk, nil
},
PSKIdentityHint: []byte(PSKIdentity),
InsecureSkipVerify: true,
InsecureSkipVerifyHello: true,
MTU: 1200,
FlightInterval: 300 * time.Millisecond,
ExtendedMasterSecret: dtls.DisableExtendedMasterSecret,
}
if isServer {
config.CipherSuites = []dtls.CipherSuiteID{dtls.TLS_PSK_WITH_AES_128_CBC_SHA256}
} else {
config.CustomCipherSuites = CustomCipherSuites
}
return config
}
type ChannelAdapter struct {
conn *Conn
channel uint8
}
func (a *ChannelAdapter) ReadFrom(p []byte) (n int, addr net.Addr, err error) {
var buf chan []byte
if a.channel == IOTCChannelMain {
buf = a.conn.mainBuf
} else {
buf = a.conn.speakBuf
}
select {
case data := <-buf:
return copy(p, data), a.conn.addr, nil
case <-a.conn.ctx.Done():
return 0, nil, net.ErrClosed
}
}
func (a *ChannelAdapter) WriteTo(p []byte, _ net.Addr) (int, error) {
if err := a.conn.WriteDTLS(p, a.channel); err != nil {
return 0, err
}
return len(p), nil
}
func (a *ChannelAdapter) Close() error { return nil }
func (a *ChannelAdapter) LocalAddr() net.Addr { return &net.UDPAddr{} }
func (a *ChannelAdapter) SetDeadline(time.Time) error { return nil }
func (a *ChannelAdapter) SetReadDeadline(time.Time) error { return nil }
func (a *ChannelAdapter) SetWriteDeadline(time.Time) error { return nil }
+505
View File
@@ -0,0 +1,505 @@
package tutk
import (
"encoding/binary"
"fmt"
"github.com/AlexxIT/go2rtc/pkg/aac"
)
const (
FrameTypeStart uint8 = 0x08 // Extended start (36-byte header)
FrameTypeStartAlt uint8 = 0x09 // StartAlt (36-byte header)
FrameTypeCont uint8 = 0x00 // Continuation (28-byte header)
FrameTypeContAlt uint8 = 0x04 // Continuation alt
FrameTypeEndSingle uint8 = 0x01 // Single-packet frame (28-byte)
FrameTypeEndMulti uint8 = 0x05 // Multi-packet end (28-byte)
FrameTypeEndExt uint8 = 0x0d // Extended end (36-byte)
)
const (
ChannelIVideo uint8 = 0x05
ChannelAudio uint8 = 0x03
ChannelPVideo uint8 = 0x07
)
// Resolution constants
const (
ResolutionUnknown = 0
ResolutionSD = 1
Resolution360P = 2
Resolution2K = 4
)
const FrameInfoSize = 40
// FrameInfo - Wyze extended FRAMEINFO (40 bytes at end of packet)
type FrameInfo struct {
CodecID uint16
Flags uint8
CamIndex uint8
OnlineNum uint8
Framerate uint8
FrameSize uint8
Bitrate uint8
TimestampUS uint32
Timestamp uint32
PayloadSize uint32
FrameNo uint32
}
func (fi *FrameInfo) IsKeyframe() bool {
return fi.Flags == 0x01
}
func (fi *FrameInfo) Resolution() string {
switch fi.FrameSize {
case ResolutionSD:
return "SD"
case Resolution360P:
return "360P"
case Resolution2K:
return "2K"
default:
return "unknown"
}
}
func (fi *FrameInfo) SampleRate() uint32 {
idx := (fi.Flags >> 2) & 0x0F
return uint32(SampleRateValue(idx))
}
func (fi *FrameInfo) Channels() uint8 {
if fi.Flags&0x01 == 1 {
return 2
}
return 1
}
func (fi *FrameInfo) IsVideo() bool {
return IsVideoCodec(fi.CodecID)
}
func (fi *FrameInfo) IsAudio() bool {
return IsAudioCodec(fi.CodecID)
}
func ParseFrameInfo(data []byte) *FrameInfo {
if len(data) < FrameInfoSize {
return nil
}
offset := len(data) - FrameInfoSize
fi := data[offset:]
return &FrameInfo{
CodecID: binary.LittleEndian.Uint16(fi),
Flags: fi[2],
CamIndex: fi[3],
OnlineNum: fi[4],
Framerate: fi[5],
FrameSize: fi[6],
Bitrate: fi[7],
TimestampUS: binary.LittleEndian.Uint32(fi[8:]),
Timestamp: binary.LittleEndian.Uint32(fi[12:]),
PayloadSize: binary.LittleEndian.Uint32(fi[16:]),
FrameNo: binary.LittleEndian.Uint32(fi[20:]),
}
}
type Packet struct {
Channel uint8
Codec uint16
Timestamp uint32
Payload []byte
IsKeyframe bool
FrameNo uint32
SampleRate uint32
Channels uint8
}
func (p *Packet) IsVideo() bool {
return p.Channel == ChannelIVideo || p.Channel == ChannelPVideo
}
func (p *Packet) IsAudio() bool {
return p.Channel == ChannelAudio
}
type PacketHeader struct {
Channel byte
FrameType byte
HeaderSize int
FrameNo uint32
PktIdx uint16
PktTotal uint16
PayloadSize uint16
HasFrameInfo bool
}
func ParsePacketHeader(data []byte) *PacketHeader {
if len(data) < 28 {
return nil
}
frameType := data[1]
hdr := &PacketHeader{
Channel: data[0],
FrameType: frameType,
}
switch frameType {
case FrameTypeStart, FrameTypeStartAlt, FrameTypeEndExt:
hdr.HeaderSize = 36
default:
hdr.HeaderSize = 28
}
if len(data) < hdr.HeaderSize {
return nil
}
if hdr.HeaderSize == 28 {
hdr.PktTotal = binary.LittleEndian.Uint16(data[12:])
pktIdxOrMarker := binary.LittleEndian.Uint16(data[14:])
hdr.PayloadSize = binary.LittleEndian.Uint16(data[16:])
hdr.FrameNo = binary.LittleEndian.Uint32(data[24:])
if IsEndFrame(frameType) && pktIdxOrMarker == 0x0028 {
hdr.HasFrameInfo = true
if hdr.PktTotal > 0 {
hdr.PktIdx = hdr.PktTotal - 1
}
} else {
hdr.PktIdx = pktIdxOrMarker
}
} else {
hdr.PktTotal = binary.LittleEndian.Uint16(data[20:])
pktIdxOrMarker := binary.LittleEndian.Uint16(data[22:])
hdr.PayloadSize = binary.LittleEndian.Uint16(data[24:])
hdr.FrameNo = binary.LittleEndian.Uint32(data[32:])
if IsEndFrame(frameType) && pktIdxOrMarker == 0x0028 {
hdr.HasFrameInfo = true
if hdr.PktTotal > 0 {
hdr.PktIdx = hdr.PktTotal - 1
}
} else {
hdr.PktIdx = pktIdxOrMarker
}
}
return hdr
}
func IsStartFrame(frameType uint8) bool {
return frameType == FrameTypeStart || frameType == FrameTypeStartAlt
}
func IsEndFrame(frameType uint8) bool {
return frameType == FrameTypeEndSingle ||
frameType == FrameTypeEndMulti ||
frameType == FrameTypeEndExt
}
func IsContinuationFrame(frameType uint8) bool {
return frameType == FrameTypeCont || frameType == FrameTypeContAlt
}
type FrameAssembler struct {
FrameNo uint32
PktTotal uint16
Packets map[uint16][]byte
FrameInfo *FrameInfo
}
func ParseAudioParams(payload []byte, fi *FrameInfo) (sampleRate uint32, channels uint8) {
if aac.IsADTS(payload) {
codec := aac.ADTSToCodec(payload)
if codec != nil {
return codec.ClockRate, codec.Channels
}
}
if fi != nil {
return fi.SampleRate(), fi.Channels()
}
return 16000, 1
}
type FrameHandler struct {
assemblers map[byte]*FrameAssembler
baseTS uint64
output chan *Packet
verbose bool
}
func NewFrameHandler(verbose bool) *FrameHandler {
return &FrameHandler{
assemblers: make(map[byte]*FrameAssembler),
output: make(chan *Packet, 128),
verbose: verbose,
}
}
func (h *FrameHandler) Recv() <-chan *Packet {
return h.output
}
func (h *FrameHandler) Close() {
close(h.output)
}
func (h *FrameHandler) Handle(data []byte) {
hdr := ParsePacketHeader(data)
if hdr == nil {
return
}
if h.verbose {
h.logWireHeader(data, hdr)
}
payload, fi := h.extractPayload(data, hdr.Channel)
if payload == nil {
return
}
if h.verbose {
h.logAVPacket(hdr.Channel, hdr.FrameType, payload, fi)
}
switch hdr.Channel {
case ChannelAudio:
h.handleAudio(payload, fi)
case ChannelIVideo, ChannelPVideo:
h.handleVideo(hdr.Channel, hdr, payload, fi)
}
}
func (h *FrameHandler) extractPayload(data []byte, channel byte) ([]byte, *FrameInfo) {
if len(data) < 2 {
return nil, nil
}
frameType := data[1]
headerSize := 28
frameInfoSize := 0
switch frameType {
case FrameTypeStart:
headerSize = 36
case FrameTypeStartAlt:
headerSize = 36
if len(data) >= 22 {
pktTotal := binary.LittleEndian.Uint16(data[20:])
if pktTotal == 1 {
frameInfoSize = FrameInfoSize
}
}
case FrameTypeCont, FrameTypeContAlt:
headerSize = 28
case FrameTypeEndSingle, FrameTypeEndMulti:
headerSize = 28
frameInfoSize = FrameInfoSize
case FrameTypeEndExt:
headerSize = 36
frameInfoSize = FrameInfoSize
default:
headerSize = 28
}
if len(data) < headerSize {
return nil, nil
}
if frameInfoSize == 0 {
return data[headerSize:], nil
}
if len(data) < headerSize+frameInfoSize {
return data[headerSize:], nil
}
fi := ParseFrameInfo(data)
validCodec := false
switch channel {
case ChannelIVideo, ChannelPVideo:
validCodec = IsVideoCodec(fi.CodecID)
case ChannelAudio:
validCodec = IsAudioCodec(fi.CodecID)
}
if validCodec {
payload := data[headerSize : len(data)-frameInfoSize]
return payload, fi
}
return data[headerSize:], nil
}
func (h *FrameHandler) handleVideo(channel byte, hdr *PacketHeader, payload []byte, fi *FrameInfo) {
asm := h.assemblers[channel]
// Frame transition: new frame number = previous frame complete
if asm != nil && hdr.FrameNo != asm.FrameNo {
gotAll := uint16(len(asm.Packets)) == asm.PktTotal
if gotAll && asm.FrameInfo != nil {
h.assembleAndQueue(channel, asm)
}
asm = nil
}
// Create new assembler if needed
if asm == nil {
asm = &FrameAssembler{
FrameNo: hdr.FrameNo,
PktTotal: hdr.PktTotal,
Packets: make(map[uint16][]byte, hdr.PktTotal),
}
h.assemblers[channel] = asm
}
// Store packet (copy payload - buffer is reused by worker)
payloadCopy := make([]byte, len(payload))
copy(payloadCopy, payload)
asm.Packets[hdr.PktIdx] = payloadCopy
if fi != nil {
asm.FrameInfo = fi
}
// Check if frame is complete
if uint16(len(asm.Packets)) == asm.PktTotal && asm.FrameInfo != nil {
h.assembleAndQueue(channel, asm)
delete(h.assemblers, channel)
}
}
func (h *FrameHandler) assembleAndQueue(channel byte, asm *FrameAssembler) {
fi := asm.FrameInfo
// Assemble packets in correct order
var payload []byte
for i := uint16(0); i < asm.PktTotal; i++ {
if pkt, ok := asm.Packets[i]; ok {
payload = append(payload, pkt...)
}
}
// Size validation
if fi.PayloadSize > 0 && len(payload) != int(fi.PayloadSize) {
return
}
if len(payload) == 0 {
return
}
// Calculate RTP timestamp (90kHz for video) using relative timestamps
absoluteTS := uint64(fi.Timestamp)*1000000 + uint64(fi.TimestampUS)
if h.baseTS == 0 {
h.baseTS = absoluteTS
}
relativeUS := absoluteTS - h.baseTS
const clockRate uint64 = 90000
rtpTS := uint32(relativeUS * clockRate / 1000000)
pkt := &Packet{
Channel: channel,
Payload: payload,
Codec: fi.CodecID,
Timestamp: rtpTS,
IsKeyframe: fi.IsKeyframe(),
FrameNo: fi.FrameNo,
}
if h.verbose {
frameType := "P"
if fi.IsKeyframe() {
frameType = "I"
}
fmt.Printf("[VIDEO] #%d %s %s size=%d rtp=%d\n",
fi.FrameNo, CodecName(fi.CodecID), frameType, len(payload), rtpTS)
}
h.queue(pkt)
}
func (h *FrameHandler) handleAudio(payload []byte, fi *FrameInfo) {
if len(payload) == 0 || fi == nil {
return
}
var sampleRate uint32
var channels uint8
switch fi.CodecID {
case AudioCodecAACRaw, AudioCodecAACADTS, AudioCodecAACLATM, AudioCodecAACWyze:
sampleRate, channels = ParseAudioParams(payload, fi)
default:
sampleRate = fi.SampleRate()
channels = fi.Channels()
}
// Calculate RTP timestamp using relative timestamps (shared baseTS for A/V sync)
absoluteTS := uint64(fi.Timestamp)*1000000 + uint64(fi.TimestampUS)
if h.baseTS == 0 {
h.baseTS = absoluteTS
}
relativeUS := absoluteTS - h.baseTS
clockRate := uint64(sampleRate)
rtpTS := uint32(relativeUS * clockRate / 1000000)
pkt := &Packet{
Channel: ChannelAudio,
Payload: payload,
Codec: fi.CodecID,
Timestamp: rtpTS,
SampleRate: sampleRate,
Channels: channels,
FrameNo: fi.FrameNo,
}
if h.verbose {
fmt.Printf("[AUDIO] #%d %s size=%d rate=%d ch=%d rtp=%d\n",
fi.FrameNo, AudioCodecName(fi.CodecID), len(payload), sampleRate, channels, rtpTS)
}
h.queue(pkt)
}
func (h *FrameHandler) queue(pkt *Packet) {
select {
case h.output <- pkt:
default:
// Queue full - drop oldest
select {
case <-h.output:
default:
}
h.output <- pkt
}
}
func (h *FrameHandler) logWireHeader(data []byte, hdr *PacketHeader) {
fmt.Printf("[WIRE] ch=0x%02x type=0x%02x len=%d pkt=%d/%d frame=%d\n",
hdr.Channel, hdr.FrameType, len(data), hdr.PktIdx, hdr.PktTotal, hdr.FrameNo)
fmt.Printf(" RAW[0..35]: ")
for i := 0; i < 36 && i < len(data); i++ {
fmt.Printf("%02x ", data[i])
}
fmt.Printf("\n")
}
func (h *FrameHandler) logAVPacket(channel, frameType byte, payload []byte, fi *FrameInfo) {
fmt.Printf("[AV] ch=0x%02x type=0x%02x len=%d", channel, frameType, len(payload))
if fi != nil {
fmt.Printf(" fi={codec=0x%04x flags=0x%02x ts=%d}", fi.CodecID, fi.Flags, fi.Timestamp)
}
fmt.Printf("\n")
}
+278
View File
@@ -0,0 +1,278 @@
package tutk
type AVLoginResponse struct {
ServerType uint32
Resend int32
TwoWayStreaming int32
SyncRecvData int32
SecurityMode uint32
VideoOnConnect int32
AudioOnConnect int32
}
const (
CodecUnknown uint16 = 0x00
CodecMPEG4 uint16 = 0x4C // 76
CodecH263 uint16 = 0x4D // 77
CodecH264 uint16 = 0x4E // 78
CodecMJPEG uint16 = 0x4F // 79
CodecH265 uint16 = 0x50 // 80
)
const (
AudioCodecAACRaw uint16 = 0x86 // 134
AudioCodecAACADTS uint16 = 0x87 // 135
AudioCodecAACLATM uint16 = 0x88 // 136
AudioCodecG711U uint16 = 0x89 // 137
AudioCodecG711A uint16 = 0x8A // 138
AudioCodecADPCM uint16 = 0x8B // 139
AudioCodecPCM uint16 = 0x8C // 140
AudioCodecSPEEX uint16 = 0x8D // 141
AudioCodecMP3 uint16 = 0x8E // 142
AudioCodecG726 uint16 = 0x8F // 143
AudioCodecAACWyze uint16 = 0x90 // 144
AudioCodecOpus uint16 = 0x92 // 146
)
const (
SampleRate8K uint8 = 0x00
SampleRate11K uint8 = 0x01
SampleRate12K uint8 = 0x02
SampleRate16K uint8 = 0x03
SampleRate22K uint8 = 0x04
SampleRate24K uint8 = 0x05
SampleRate32K uint8 = 0x06
SampleRate44K uint8 = 0x07
SampleRate48K uint8 = 0x08
)
var sampleRates = map[uint8]int{
SampleRate8K: 8000,
SampleRate11K: 11025,
SampleRate12K: 12000,
SampleRate16K: 16000,
SampleRate22K: 22050,
SampleRate24K: 24000,
SampleRate32K: 32000,
SampleRate44K: 44100,
SampleRate48K: 48000,
}
var samplesPerFrame = map[uint16]uint32{
AudioCodecAACRaw: 1024,
AudioCodecAACADTS: 1024,
AudioCodecAACLATM: 1024,
AudioCodecAACWyze: 1024,
AudioCodecG711U: 160,
AudioCodecG711A: 160,
AudioCodecPCM: 160,
AudioCodecADPCM: 160,
AudioCodecSPEEX: 160,
AudioCodecMP3: 1152,
AudioCodecG726: 160,
AudioCodecOpus: 960,
}
const (
IOTypeVideoStart = 0x01FF
IOTypeVideoStop = 0x02FF
IOTypeAudioStart = 0x0300
IOTypeAudioStop = 0x0301
IOTypeSpeakerStart = 0x0350
IOTypeSpeakerStop = 0x0351
IOTypeGetAudioOutFormatReq = 0x032A
IOTypeGetAudioOutFormatRes = 0x032B
IOTypeSetStreamCtrlReq = 0x0320
IOTypeSetStreamCtrlRes = 0x0321
IOTypeGetStreamCtrlReq = 0x0322
IOTypeGetStreamCtrlRes = 0x0323
IOTypeDevInfoReq = 0x0340
IOTypeDevInfoRes = 0x0341
IOTypeGetSupportStreamReq = 0x0344
IOTypeGetSupportStreamRes = 0x0345
IOTypeSetRecordReq = 0x0310
IOTypeSetRecordRes = 0x0311
IOTypeGetRecordReq = 0x0312
IOTypeGetRecordRes = 0x0313
IOTypePTZCommand = 0x1001
IOTypeReceiveFirstFrame = 0x1002
IOTypeGetEnvironmentReq = 0x030A
IOTypeGetEnvironmentRes = 0x030B
IOTypeSetVideoModeReq = 0x030C
IOTypeSetVideoModeRes = 0x030D
IOTypeGetVideoModeReq = 0x030E
IOTypeGetVideoModeRes = 0x030F
IOTypeSetTimeReq = 0x0316
IOTypeSetTimeRes = 0x0317
IOTypeGetTimeReq = 0x0318
IOTypeGetTimeRes = 0x0319
IOTypeSetWifiReq = 0x0102
IOTypeSetWifiRes = 0x0103
IOTypeGetWifiReq = 0x0104
IOTypeGetWifiRes = 0x0105
IOTypeListWifiAPReq = 0x0106
IOTypeListWifiAPRes = 0x0107
IOTypeSetMotionDetectReq = 0x0306
IOTypeSetMotionDetectRes = 0x0307
IOTypeGetMotionDetectReq = 0x0308
IOTypeGetMotionDetectRes = 0x0309
)
// OLD Protocol (IOTC/TransCode)
const (
CmdDiscoReq uint16 = 0x0601
CmdDiscoRes uint16 = 0x0602
CmdSessionReq uint16 = 0x0402
CmdSessionRes uint16 = 0x0404
CmdDataTX uint16 = 0x0407
CmdDataRX uint16 = 0x0408
CmdKeepaliveReq uint16 = 0x0427
CmdKeepaliveRes uint16 = 0x0428
OldHeaderSize = 16
OldDiscoBodySize = 72
OldDiscoSize = OldHeaderSize + OldDiscoBodySize
OldSessionBody = 36
OldSessionSize = OldHeaderSize + OldSessionBody
)
// NEW Protocol (0xCC51)
const (
MagicNewProto uint16 = 0xCC51
CmdNewDisco uint16 = 0x1002
CmdNewDTLS uint16 = 0x1502
NewPayloadSize uint16 = 0x0028
NewPacketSize = 52
NewHeaderSize = 28
NewAuthSize = 20
)
const (
UIDSize = 20
RandIDSize = 8
)
const (
MagicAVLoginResp uint16 = 0x2100
MagicIOCtrl uint16 = 0x7000
MagicChannelMsg uint16 = 0x1000
MagicACK uint16 = 0x0009
MagicAVLogin1 uint16 = 0x0000
MagicAVLogin2 uint16 = 0x2000
)
const (
ProtoVersion uint16 = 0x000c
DefaultCaps uint32 = 0x001f07fb
)
const (
IOTCChannelMain = 0 // Main AV (we = DTLS Client)
IOTCChannelBack = 1 // Backchannel (we = DTLS Server)
)
const (
PSKIdentity = "AUTHPWD_admin"
DefaultUser = "admin"
DefaultPort = 32761
)
func CodecName(id uint16) string {
switch id {
case CodecH264:
return "H264"
case CodecH265:
return "H265"
case CodecMPEG4:
return "MPEG4"
case CodecH263:
return "H263"
case CodecMJPEG:
return "MJPEG"
default:
return "Unknown"
}
}
func AudioCodecName(id uint16) string {
switch id {
case AudioCodecG711U:
return "PCMU"
case AudioCodecG711A:
return "PCMA"
case AudioCodecPCM:
return "PCM"
case AudioCodecAACLATM, AudioCodecAACRaw, AudioCodecAACADTS, AudioCodecAACWyze:
return "AAC"
case AudioCodecOpus:
return "Opus"
case AudioCodecSPEEX:
return "Speex"
case AudioCodecMP3:
return "MP3"
case AudioCodecG726:
return "G726"
case AudioCodecADPCM:
return "ADPCM"
default:
return "Unknown"
}
}
func SampleRateValue(idx uint8) int {
if rate, ok := sampleRates[idx]; ok {
return rate
}
return 16000
}
func SampleRateIndex(hz uint32) uint8 {
switch hz {
case 8000:
return SampleRate8K
case 11025:
return SampleRate11K
case 12000:
return SampleRate12K
case 16000:
return SampleRate16K
case 22050:
return SampleRate22K
case 24000:
return SampleRate24K
case 32000:
return SampleRate32K
case 44100:
return SampleRate44K
case 48000:
return SampleRate48K
default:
return SampleRate16K
}
}
func BuildAudioFlags(sampleRate uint32, bits16, stereo bool) uint8 {
flags := SampleRateIndex(sampleRate) << 2
if bits16 {
flags |= 0x02
}
if stereo {
flags |= 0x01
}
return flags
}
func IsVideoCodec(id uint16) bool {
return id >= CodecMPEG4 && id <= CodecH265
}
func IsAudioCodec(id uint16) bool {
return id >= AudioCodecAACRaw && id <= AudioCodecOpus
}
func GetSamplesPerFrame(codecID uint16) uint32 {
if samples, ok := samplesPerFrame[codecID]; ok {
return samples
}
return 1024
}
-157
View File
@@ -1,157 +0,0 @@
package tutk
import "encoding/binary"
const (
// Start packets - first fragment of a frame
// 0x08: Extended start (36-byte header, no FrameInfo)
// 0x09: StartAlt (36-byte header, FrameInfo only if pkt_total==1)
FrameTypeStart uint8 = 0x08
FrameTypeStartAlt uint8 = 0x09
// Continuation packets - middle fragment (28-byte header, no FrameInfo)
FrameTypeCont uint8 = 0x00
FrameTypeContAlt uint8 = 0x04
// End packets - last fragment (with 40-byte FrameInfo)
// 0x01: Single-packet frame (28-byte header)
// 0x05: Multi-packet end (28-byte header)
// 0x0d: Extended end (36-byte header)
FrameTypeEndSingle uint8 = 0x01
FrameTypeEndMulti uint8 = 0x05
FrameTypeEndExt uint8 = 0x0d
)
const (
ChannelIVideo uint8 = 0x05
ChannelAudio uint8 = 0x03
ChannelPVideo uint8 = 0x07
)
type Packet struct {
Channel uint8
Codec uint16
Timestamp uint32
Payload []byte
IsKeyframe bool
FrameNo uint32
SampleRate uint32
Channels uint8
}
func (p *Packet) IsVideo() bool {
return p.Channel == ChannelIVideo || p.Channel == ChannelPVideo
}
func (p *Packet) IsAudio() bool {
return p.Channel == ChannelAudio
}
type AuthResponse struct {
ConnectionRes string `json:"connectionRes"`
CameraInfo map[string]any `json:"cameraInfo"`
}
type AVLoginResponse struct {
ServerType uint32
Resend int32
TwoWayStreaming int32
SyncRecvData int32
SecurityMode uint32
VideoOnConnect int32
AudioOnConnect int32
}
func IsStartFrame(frameType uint8) bool {
return frameType == FrameTypeStart || frameType == FrameTypeStartAlt
}
func IsEndFrame(frameType uint8) bool {
return frameType == FrameTypeEndSingle ||
frameType == FrameTypeEndMulti ||
frameType == FrameTypeEndExt
}
func IsContinuationFrame(frameType uint8) bool {
return frameType == FrameTypeCont || frameType == FrameTypeContAlt
}
type PacketHeader struct {
Channel byte
FrameType byte
HeaderSize int // 28 or 36
FrameNo uint32 // Frame number (from [24-27] for 28-byte, [32-35] for 36-byte)
PktIdx uint16 // Packet index within frame (0-based)
PktTotal uint16 // Total packets in this frame
PayloadSize uint16
HasFrameInfo bool // true if [14-15] or [22-23] == 0x0028
}
func ParsePacketHeader(data []byte) *PacketHeader {
if len(data) < 28 {
return nil
}
frameType := data[1]
hdr := &PacketHeader{
Channel: data[0],
FrameType: frameType,
}
// Header size based on FrameType (NOT magic bytes!)
switch frameType {
case FrameTypeStart, FrameTypeStartAlt, FrameTypeEndExt: // 0x08, 0x09, 0x0d
hdr.HeaderSize = 36
default: // 0x00, 0x01, 0x04, 0x05
hdr.HeaderSize = 28
}
if len(data) < hdr.HeaderSize {
return nil
}
if hdr.HeaderSize == 28 {
// 28-Byte Header Layout:
// [12-13] pkt_total
// [14-15] pkt_idx OR 0x0028 (FrameInfo marker) - ONLY 0x0028 in End packets!
// [16-17] payload_size
// [24-27] frame_no (uint32)
hdr.PktTotal = binary.LittleEndian.Uint16(data[12:])
pktIdxOrMarker := binary.LittleEndian.Uint16(data[14:])
hdr.PayloadSize = binary.LittleEndian.Uint16(data[16:])
hdr.FrameNo = binary.LittleEndian.Uint32(data[24:])
// 0x0028 is FrameInfo marker ONLY in End packets, otherwise it's pkt_idx=40
if IsEndFrame(hdr.FrameType) && pktIdxOrMarker == 0x0028 {
hdr.HasFrameInfo = true
if hdr.PktTotal > 0 {
hdr.PktIdx = hdr.PktTotal - 1 // Last packet
}
} else {
hdr.PktIdx = pktIdxOrMarker
}
} else {
// 36-Byte Header Layout:
// [20-21] pkt_total
// [22-23] pkt_idx OR 0x0028 (FrameInfo marker) - ONLY 0x0028 in End packets!
// [24-25] payload_size
// [32-35] frame_no (uint32) - GLOBAL frame counter, matches 28-byte [24-27]
// NOTE: [18-19] is channel-specific frame index, NOT used for reassembly!
hdr.PktTotal = binary.LittleEndian.Uint16(data[20:])
pktIdxOrMarker := binary.LittleEndian.Uint16(data[22:])
hdr.PayloadSize = binary.LittleEndian.Uint16(data[24:])
hdr.FrameNo = binary.LittleEndian.Uint32(data[32:])
// 0x0028 is FrameInfo marker ONLY in End packets, otherwise it's pkt_idx=40
if IsEndFrame(hdr.FrameType) && pktIdxOrMarker == 0x0028 {
hdr.HasFrameInfo = true
if hdr.PktTotal > 0 {
hdr.PktIdx = hdr.PktTotal - 1
}
} else {
hdr.PktIdx = pktIdxOrMarker
}
}
return hdr
}