Merge branch 'AlexxIT:master' into check-h265

internal/alsa/alsa.go

@@ -0,0 +1,7 @@
+//go:build !(linux && (386 || amd64 || arm || arm64 || mipsle))
+
+package alsa
+
+func Init() {
+	// not supported
+}

internal/alsa/alsa_linux.go

@@ -0,0 +1,83 @@
+//go:build linux && (386 || amd64 || arm || arm64 || mipsle)
+
+package alsa
+
+import (
+	"fmt"
+	"net/http"
+	"os"
+	"strconv"
+	"strings"
+
+	"github.com/AlexxIT/go2rtc/internal/api"
+	"github.com/AlexxIT/go2rtc/internal/streams"
+	"github.com/AlexxIT/go2rtc/pkg/alsa"
+	"github.com/AlexxIT/go2rtc/pkg/alsa/device"
+)
+
+func Init() {
+	streams.HandleFunc("alsa", alsa.Open)
+
+	api.HandleFunc("api/alsa", apiAlsa)
+}
+
+func apiAlsa(w http.ResponseWriter, r *http.Request) {
+	files, err := os.ReadDir("/dev/snd/")
+	if err != nil {
+		return
+	}
+
+	var sources []*api.Source
+
+	for _, file := range files {
+		if !strings.HasPrefix(file.Name(), "pcm") {
+			continue
+		}
+
+		path := "/dev/snd/" + file.Name()
+
+		dev, err := device.Open(path)
+		if err != nil {
+			continue
+		}
+
+		info, err := dev.Info()
+		if err == nil {
+			formats := formatsToString(dev.ListFormats())
+			r1, r2 := dev.RangeRates()
+			c1, c2 := dev.RangeChannels()
+			source := &api.Source{
+				Name: info.ID,
+				Info: fmt.Sprintf("Formats: %s, Rates: %d-%d, Channels: %d-%d", formats, r1, r2, c1, c2),
+				URL:  "alsa:device?audio=" + path,
+			}
+			if !strings.Contains(source.Name, info.Name) {
+				source.Name += ", " + info.Name
+			}
+			sources = append(sources, source)
+		}
+
+		_ = dev.Close()
+	}
+
+	api.ResponseSources(w, sources)
+}
+
+func formatsToString(formats []byte) string {
+	var s string
+	for i, format := range formats {
+		if i > 0 {
+			s += " "
+		}
+		switch format {
+		case 2:
+			s += "s16le"
+		case 10:
+			s += "s32le"
+		default:
+			s += strconv.Itoa(int(format))
+		}
+
+	}
+	return s
+}

internal/ffmpeg/ffmpeg.go

@@ -113,6 +113,7 @@ var defaults = map[string]string{
 	"pcm/48000":  "-c:a pcm_s16be -ar:a 48000 -ac:a 1",
 	"pcml":       "-c:a pcm_s16le -ar:a 8000 -ac:a 1",
 	"pcml/8000":  "-c:a pcm_s16le -ar:a 8000 -ac:a 1",
+	"pcml/16000": "-c:a pcm_s16le -ar:a 16000 -ac:a 1",
 	"pcml/44100": "-c:a pcm_s16le -ar:a 44100 -ac:a 1",
 
 	// hardware Intel and AMD on Linux

internal/ffmpeg/producer.go

@@ -42,6 +42,7 @@ func NewProducer(url string) (core.Producer, error) {
 			Codecs: []*core.Codec{
 				// OPUS will always marked as OPUS/48000/2
 				{Name: core.CodecOpus, ClockRate: 48000, Channels: 2},
+				{Name: core.CodecPCML, ClockRate: 16000},
 				{Name: core.CodecPCM, ClockRate: 16000},
 				{Name: core.CodecPCMA, ClockRate: 16000},
 				{Name: core.CodecPCMU, ClockRate: 16000},
@@ -97,6 +98,8 @@ func (p *Producer) newURL() string {
 			s += "#audio=opus"
 		case core.CodecAAC:
 			s += "#audio=aac/16000"
+		case core.CodecPCML:
+			s += "#audio=pcml/16000"
 		case core.CodecPCM:
 			s += "#audio=pcm/" + strconv.Itoa(int(codec.ClockRate))
 		case core.CodecPCMA:

internal/streams/play.go

@@ -140,10 +140,12 @@ func matchMedia(prod core.Producer, cons core.Consumer) bool {
 
 			track, err := prod.GetTrack(prodMedia, prodCodec)
 			if err != nil {
+				log.Warn().Err(err).Msg("[streams] can't get track")
 				continue
 			}
 
 			if err = cons.AddTrack(consMedia, consCodec, track); err != nil {
+				log.Warn().Err(err).Msg("[streams] can't add track")
 				continue
 			}
 

main.go

@@ -1,6 +1,7 @@
 package main
 
 import (
+	"github.com/AlexxIT/go2rtc/internal/alsa"
 	"github.com/AlexxIT/go2rtc/internal/api"
 	"github.com/AlexxIT/go2rtc/internal/api/ws"
 	"github.com/AlexxIT/go2rtc/internal/app"
@@ -90,6 +91,7 @@ func main() {
 	gopro.Init()    // gopro source
 	doorbird.Init() // doorbird source
 	v4l2.Init()     // v4l2 source
+	alsa.Init()     // alsa source
 	flussonic.Init()
 	eseecloud.Init()
 

pkg/README.md

@@ -13,6 +13,7 @@ Some formats and protocols go2rtc supports exclusively. They have no equivalent
 | Format       | Source protocols | Ingress protocols | Recevers codecs              | Senders codecs     | Example       |
 |--------------|------------------|-------------------|------------------------------|--------------------|---------------|
 | adts         | http,tcp,pipe    | http              | aac                          |                    | `http:`       |
+| alsa         | pipe             |                   |                              | pcm                | `alsa:`       |
 | bubble       | http             |                   | h264,hevc,pcm_alaw           |                    | `bubble:`     |
 | dvrip        | tcp              |                   | h264,hevc,pcm_alaw,pcm_mulaw | pcm_alaw           | `dvrip:`      |
 | flv          | http,tcp,pipe    | http              | h264,aac                     |                    | `http:`       |

pkg/aac/aac.go

@@ -53,7 +53,7 @@ func ConfigToCodec(conf []byte) *core.Codec {
 		codec.ClockRate = rd.ReadBits(24)
 	}
 
-	codec.Channels = rd.ReadBits16(4)
+	codec.Channels = rd.ReadBits8(4)
 
 	return codec
 }

pkg/aac/adts.go

@@ -28,7 +28,7 @@ func ADTSToCodec(b []byte) *core.Codec {
 	objType := rd.ReadBits8(2) + 1   // Profile, the MPEG-4 Audio Object Type minus 1
 	sampleRateIdx := rd.ReadBits8(4) // MPEG-4 Sampling Frequency Index
 	_ = rd.ReadBit()                 // Private bit, guaranteed never to be used by MPEG, set to 0 when encoding, ignore when decoding
-	channels := rd.ReadBits16(3)     // MPEG-4 Channel Configuration
+	channels := rd.ReadBits8(3)      // MPEG-4 Channel Configuration
 
 	//_ = rd.ReadBit()    // Originality, set to 1 to signal originality of the audio and 0 otherwise
 	//_ = rd.ReadBit()    // Home, set to 1 to signal home usage of the audio and 0 otherwise
@@ -43,7 +43,7 @@ func ADTSToCodec(b []byte) *core.Codec {
 	wr := bits.NewWriter(nil)
 	wr.WriteBits8(objType, 5)
 	wr.WriteBits8(sampleRateIdx, 4)
-	wr.WriteBits16(channels, 4)
+	wr.WriteBits8(channels, 4)
 	conf := wr.Bytes()
 
 	codec := &core.Codec{

pkg/alsa/README.md

@@ -0,0 +1,23 @@
+## Build
+
+```shell
+x86_64-linux-gnu-gcc -w -static asound_arch.c -o asound_amd64
+i686-linux-gnu-gcc -w -static asound_arch.c -o asound_i386
+aarch64-linux-gnu-gcc -w -static asound_arch.c -o asound_arm64
+arm-linux-gnueabihf-gcc -w -static asound_arch.c -o asound_arm
+mipsel-linux-gnu-gcc -w -static asound_arch.c -o asound_mipsle -D_TIME_BITS=32
+```
+
+## Useful links
+
+- https://github.com/torvalds/linux/blob/master/include/uapi/sound/asound.h
+- https://github.com/yobert/alsa
+- https://github.com/Narsil/alsa-go
+- https://github.com/alsa-project/alsa-lib
+- https://github.com/anisse/alsa
+- https://github.com/tinyalsa/tinyalsa
+
+**Broken pipe**
+
+- https://stackoverflow.com/questions/26545139/alsa-cannot-recovery-from-underrun-prepare-failed-broken-pipe
+- https://klipspringer.avadeaux.net/alsa-broken-pipe-errors/

pkg/alsa/capture_linux.go

@@ -0,0 +1,90 @@
+package alsa
+
+import (
+	"github.com/AlexxIT/go2rtc/pkg/alsa/device"
+	"github.com/AlexxIT/go2rtc/pkg/core"
+	"github.com/AlexxIT/go2rtc/pkg/pcm"
+	"github.com/pion/rtp"
+)
+
+type Capture struct {
+	core.Connection
+	dev    *device.Device
+	closed core.Waiter
+}
+
+func newCapture(dev *device.Device) (*Capture, error) {
+	medias := []*core.Media{
+		{
+			Kind:      core.KindAudio,
+			Direction: core.DirectionRecvonly,
+			Codecs: []*core.Codec{
+				{Name: core.CodecPCML, ClockRate: 16000},
+			},
+		},
+	}
+	return &Capture{
+		Connection: core.Connection{
+			ID:         core.NewID(),
+			FormatName: "alsa",
+			Medias:     medias,
+			Transport:  dev,
+		},
+		dev: dev,
+	}, nil
+}
+
+func (c *Capture) Start() error {
+	dst := c.Medias[0].Codecs[0]
+	src := &core.Codec{
+		Name:      dst.Name,
+		ClockRate: c.dev.GetRateNear(dst.ClockRate),
+		Channels:  c.dev.GetChannelsNear(dst.Channels),
+	}
+
+	if err := c.dev.SetHWParams(device.SNDRV_PCM_FORMAT_S16_LE, src.ClockRate, src.Channels); err != nil {
+		return err
+	}
+
+	transcode := transcodeFunc(dst, src)
+	frameBytes := int(pcm.BytesPerFrame(src))
+
+	var ts uint32
+
+	// readBufferSize for 20ms interval
+	readBufferSize := 20 * frameBytes * int(src.ClockRate) / 1000
+	b := make([]byte, readBufferSize)
+	for {
+		n, err := c.dev.Read(b)
+		if err != nil {
+			return err
+		}
+
+		c.Recv += n
+
+		if len(c.Receivers) == 0 {
+			continue
+		}
+
+		pkt := &rtp.Packet{
+			Header: rtp.Header{
+				Version:   2,
+				Marker:    true,
+				Timestamp: ts,
+			},
+			Payload: transcode(b[:n]),
+		}
+		c.Receivers[0].WriteRTP(pkt)
+
+		ts += uint32(n / frameBytes)
+	}
+}
+
+func transcodeFunc(dst, src *core.Codec) func([]byte) []byte {
+	if dst.ClockRate == src.ClockRate && dst.Channels == src.Channels {
+		return func(b []byte) []byte {
+			return b
+		}
+	}
+	return pcm.Transcode(dst, src)
+}

pkg/alsa/device/asound_32bit.go

@@ -0,0 +1,148 @@
+//go:build 386 || arm
+
+package device
+
+type unsigned_char = byte
+type signed_int = int32
+type unsigned_int = uint32
+type signed_long = int64
+type unsigned_long = uint64
+type __u32 = uint32
+type void__user = uintptr
+
+const (
+	SNDRV_PCM_STREAM_PLAYBACK = 0
+	SNDRV_PCM_STREAM_CAPTURE  = 1
+
+	SNDRV_PCM_ACCESS_MMAP_INTERLEAVED    = 0
+	SNDRV_PCM_ACCESS_MMAP_NONINTERLEAVED = 1
+	SNDRV_PCM_ACCESS_MMAP_COMPLEX        = 2
+	SNDRV_PCM_ACCESS_RW_INTERLEAVED      = 3
+	SNDRV_PCM_ACCESS_RW_NONINTERLEAVED   = 4
+
+	SNDRV_PCM_FORMAT_S8         = 0
+	SNDRV_PCM_FORMAT_U8         = 1
+	SNDRV_PCM_FORMAT_S16_LE     = 2
+	SNDRV_PCM_FORMAT_S16_BE     = 3
+	SNDRV_PCM_FORMAT_U16_LE     = 4
+	SNDRV_PCM_FORMAT_U16_BE     = 5
+	SNDRV_PCM_FORMAT_S24_LE     = 6
+	SNDRV_PCM_FORMAT_S24_BE     = 7
+	SNDRV_PCM_FORMAT_U24_LE     = 8
+	SNDRV_PCM_FORMAT_U24_BE     = 9
+	SNDRV_PCM_FORMAT_S32_LE     = 10
+	SNDRV_PCM_FORMAT_S32_BE     = 11
+	SNDRV_PCM_FORMAT_U32_LE     = 12
+	SNDRV_PCM_FORMAT_U32_BE     = 13
+	SNDRV_PCM_FORMAT_FLOAT_LE   = 14
+	SNDRV_PCM_FORMAT_FLOAT_BE   = 15
+	SNDRV_PCM_FORMAT_FLOAT64_LE = 16
+	SNDRV_PCM_FORMAT_FLOAT64_BE = 17
+	SNDRV_PCM_FORMAT_MU_LAW     = 20
+	SNDRV_PCM_FORMAT_A_LAW      = 21
+	SNDRV_PCM_FORMAT_MPEG       = 23
+
+	SNDRV_PCM_IOCTL_PVERSION      = 0x80044100
+	SNDRV_PCM_IOCTL_INFO          = 0x81204101
+	SNDRV_PCM_IOCTL_HW_REFINE     = 0xc25c4110
+	SNDRV_PCM_IOCTL_HW_PARAMS     = 0xc25c4111
+	SNDRV_PCM_IOCTL_SW_PARAMS     = 0xc0684113
+	SNDRV_PCM_IOCTL_PREPARE       = 0x00004140
+	SNDRV_PCM_IOCTL_WRITEI_FRAMES = 0x400c4150
+	SNDRV_PCM_IOCTL_READI_FRAMES  = 0x800c4151
+)
+
+type snd_pcm_info struct { // size 288
+	device           unsigned_int      // offset 0, size 4
+	subdevice        unsigned_int      // offset 4, size 4
+	stream           signed_int        // offset 8, size 4
+	card             signed_int        // offset 12, size 4
+	id               [64]unsigned_char // offset 16, size 64
+	name             [80]unsigned_char // offset 80, size 80
+	subname          [32]unsigned_char // offset 160, size 32
+	dev_class        signed_int        // offset 192, size 4
+	dev_subclass     signed_int        // offset 196, size 4
+	subdevices_count unsigned_int      // offset 200, size 4
+	subdevices_avail unsigned_int      // offset 204, size 4
+	pad1             [16]unsigned_char
+	reserved         [64]unsigned_char // offset 224, size 64
+}
+
+type snd_pcm_uframes_t = unsigned_long
+type snd_pcm_sframes_t = signed_long
+
+type snd_xferi struct { // size 12
+	result snd_pcm_sframes_t // offset 0, size 4
+	buf    void__user        // offset 4, size 4
+	frames snd_pcm_uframes_t // offset 8, size 4
+}
+
+const (
+	SNDRV_PCM_HW_PARAM_ACCESS     = 0
+	SNDRV_PCM_HW_PARAM_FORMAT     = 1
+	SNDRV_PCM_HW_PARAM_SUBFORMAT  = 2
+	SNDRV_PCM_HW_PARAM_FIRST_MASK = 0
+	SNDRV_PCM_HW_PARAM_LAST_MASK  = 2
+
+	SNDRV_PCM_HW_PARAM_SAMPLE_BITS    = 8
+	SNDRV_PCM_HW_PARAM_FRAME_BITS     = 9
+	SNDRV_PCM_HW_PARAM_CHANNELS       = 10
+	SNDRV_PCM_HW_PARAM_RATE           = 11
+	SNDRV_PCM_HW_PARAM_PERIOD_TIME    = 12
+	SNDRV_PCM_HW_PARAM_PERIOD_SIZE    = 13
+	SNDRV_PCM_HW_PARAM_PERIOD_BYTES   = 14
+	SNDRV_PCM_HW_PARAM_PERIODS        = 15
+	SNDRV_PCM_HW_PARAM_BUFFER_TIME    = 16
+	SNDRV_PCM_HW_PARAM_BUFFER_SIZE    = 17
+	SNDRV_PCM_HW_PARAM_BUFFER_BYTES   = 18
+	SNDRV_PCM_HW_PARAM_TICK_TIME      = 19
+	SNDRV_PCM_HW_PARAM_FIRST_INTERVAL = 8
+	SNDRV_PCM_HW_PARAM_LAST_INTERVAL  = 19
+
+	SNDRV_MASK_MAX = 256
+
+	SNDRV_PCM_TSTAMP_NONE   = 0
+	SNDRV_PCM_TSTAMP_ENABLE = 1
+)
+
+type snd_mask struct { // size 32
+	bits [(SNDRV_MASK_MAX + 31) / 32]__u32 // offset 0, size 32
+}
+
+type snd_interval struct { // size 12
+	min unsigned_int // offset 0, size 4
+	max unsigned_int // offset 4, size 4
+	bit unsigned_int
+}
+
+type snd_pcm_hw_params struct { // size 604
+	flags     unsigned_int                                                                           // offset 0, size 4
+	masks     [SNDRV_PCM_HW_PARAM_LAST_MASK - SNDRV_PCM_HW_PARAM_FIRST_MASK + 1]snd_mask             // offset 4, size 96
+	mres      [5]snd_mask                                                                            // offset 100, size 160
+	intervals [SNDRV_PCM_HW_PARAM_LAST_INTERVAL - SNDRV_PCM_HW_PARAM_FIRST_INTERVAL + 1]snd_interval // offset 260, size 144
+	ires      [9]snd_interval                                                                        // offset 404, size 108
+	rmask     unsigned_int                                                                           // offset 512, size 4
+	cmask     unsigned_int                                                                           // offset 516, size 4
+	info      unsigned_int                                                                           // offset 520, size 4
+	msbits    unsigned_int                                                                           // offset 524, size 4
+	rate_num  unsigned_int                                                                           // offset 528, size 4
+	rate_den  unsigned_int                                                                           // offset 532, size 4
+	fifo_size snd_pcm_uframes_t                                                                      // offset 536, size 4
+	reserved  [64]unsigned_char                                                                      // offset 540, size 64
+}
+
+type snd_pcm_sw_params struct { // size 104
+	tstamp_mode       signed_int        // offset 0, size 4
+	period_step       unsigned_int      // offset 4, size 4
+	sleep_min         unsigned_int      // offset 8, size 4
+	avail_min         snd_pcm_uframes_t // offset 12, size 4
+	xfer_align        snd_pcm_uframes_t // offset 16, size 4
+	start_threshold   snd_pcm_uframes_t // offset 20, size 4
+	stop_threshold    snd_pcm_uframes_t // offset 24, size 4
+	silence_threshold snd_pcm_uframes_t // offset 28, size 4
+	silence_size      snd_pcm_uframes_t // offset 32, size 4
+	boundary          snd_pcm_uframes_t // offset 36, size 4
+	proto             unsigned_int      // offset 40, size 4
+	tstamp_type       unsigned_int      // offset 44, size 4
+	reserved          [56]unsigned_char // offset 48, size 56
+}

pkg/alsa/device/asound_64bit.go

@@ -0,0 +1,148 @@
+//go:build amd64 || arm64
+
+package device
+
+type unsigned_char = byte
+type signed_int = int32
+type unsigned_int = uint32
+type signed_long = int64
+type unsigned_long = uint64
+type __u32 = uint32
+type void__user = uintptr
+
+const (
+	SNDRV_PCM_STREAM_PLAYBACK = 0
+	SNDRV_PCM_STREAM_CAPTURE  = 1
+
+	SNDRV_PCM_ACCESS_MMAP_INTERLEAVED    = 0
+	SNDRV_PCM_ACCESS_MMAP_NONINTERLEAVED = 1
+	SNDRV_PCM_ACCESS_MMAP_COMPLEX        = 2
+	SNDRV_PCM_ACCESS_RW_INTERLEAVED      = 3
+	SNDRV_PCM_ACCESS_RW_NONINTERLEAVED   = 4
+
+	SNDRV_PCM_FORMAT_S8         = 0
+	SNDRV_PCM_FORMAT_U8         = 1
+	SNDRV_PCM_FORMAT_S16_LE     = 2
+	SNDRV_PCM_FORMAT_S16_BE     = 3
+	SNDRV_PCM_FORMAT_U16_LE     = 4
+	SNDRV_PCM_FORMAT_U16_BE     = 5
+	SNDRV_PCM_FORMAT_S24_LE     = 6
+	SNDRV_PCM_FORMAT_S24_BE     = 7
+	SNDRV_PCM_FORMAT_U24_LE     = 8
+	SNDRV_PCM_FORMAT_U24_BE     = 9
+	SNDRV_PCM_FORMAT_S32_LE     = 10
+	SNDRV_PCM_FORMAT_S32_BE     = 11
+	SNDRV_PCM_FORMAT_U32_LE     = 12
+	SNDRV_PCM_FORMAT_U32_BE     = 13
+	SNDRV_PCM_FORMAT_FLOAT_LE   = 14
+	SNDRV_PCM_FORMAT_FLOAT_BE   = 15
+	SNDRV_PCM_FORMAT_FLOAT64_LE = 16
+	SNDRV_PCM_FORMAT_FLOAT64_BE = 17
+	SNDRV_PCM_FORMAT_MU_LAW     = 20
+	SNDRV_PCM_FORMAT_A_LAW      = 21
+	SNDRV_PCM_FORMAT_MPEG       = 23
+
+	SNDRV_PCM_IOCTL_PVERSION      = 0x80044100
+	SNDRV_PCM_IOCTL_INFO          = 0x81204101
+	SNDRV_PCM_IOCTL_HW_REFINE     = 0xc2604110
+	SNDRV_PCM_IOCTL_HW_PARAMS     = 0xc2604111
+	SNDRV_PCM_IOCTL_SW_PARAMS     = 0xc0884113
+	SNDRV_PCM_IOCTL_PREPARE       = 0x00004140
+	SNDRV_PCM_IOCTL_WRITEI_FRAMES = 0x40184150
+	SNDRV_PCM_IOCTL_READI_FRAMES  = 0x80184151
+)
+
+type snd_pcm_info struct { // size 288
+	device           unsigned_int      // offset 0, size 4
+	subdevice        unsigned_int      // offset 4, size 4
+	stream           signed_int        // offset 8, size 4
+	card             signed_int        // offset 12, size 4
+	id               [64]unsigned_char // offset 16, size 64
+	name             [80]unsigned_char // offset 80, size 80
+	subname          [32]unsigned_char // offset 160, size 32
+	dev_class        signed_int        // offset 192, size 4
+	dev_subclass     signed_int        // offset 196, size 4
+	subdevices_count unsigned_int      // offset 200, size 4
+	subdevices_avail unsigned_int      // offset 204, size 4
+	pad1             [16]unsigned_char
+	reserved         [64]unsigned_char // offset 224, size 64
+}
+
+type snd_pcm_uframes_t = unsigned_long
+type snd_pcm_sframes_t = signed_long
+
+type snd_xferi struct { // size 24
+	result snd_pcm_sframes_t // offset 0, size 8
+	buf    void__user        // offset 8, size 8
+	frames snd_pcm_uframes_t // offset 16, size 8
+}
+
+const (
+	SNDRV_PCM_HW_PARAM_ACCESS     = 0
+	SNDRV_PCM_HW_PARAM_FORMAT     = 1
+	SNDRV_PCM_HW_PARAM_SUBFORMAT  = 2
+	SNDRV_PCM_HW_PARAM_FIRST_MASK = 0
+	SNDRV_PCM_HW_PARAM_LAST_MASK  = 2
+
+	SNDRV_PCM_HW_PARAM_SAMPLE_BITS    = 8
+	SNDRV_PCM_HW_PARAM_FRAME_BITS     = 9
+	SNDRV_PCM_HW_PARAM_CHANNELS       = 10
+	SNDRV_PCM_HW_PARAM_RATE           = 11
+	SNDRV_PCM_HW_PARAM_PERIOD_TIME    = 12
+	SNDRV_PCM_HW_PARAM_PERIOD_SIZE    = 13
+	SNDRV_PCM_HW_PARAM_PERIOD_BYTES   = 14
+	SNDRV_PCM_HW_PARAM_PERIODS        = 15
+	SNDRV_PCM_HW_PARAM_BUFFER_TIME    = 16
+	SNDRV_PCM_HW_PARAM_BUFFER_SIZE    = 17
+	SNDRV_PCM_HW_PARAM_BUFFER_BYTES   = 18
+	SNDRV_PCM_HW_PARAM_TICK_TIME      = 19
+	SNDRV_PCM_HW_PARAM_FIRST_INTERVAL = 8
+	SNDRV_PCM_HW_PARAM_LAST_INTERVAL  = 19
+
+	SNDRV_MASK_MAX = 256
+
+	SNDRV_PCM_TSTAMP_NONE   = 0
+	SNDRV_PCM_TSTAMP_ENABLE = 1
+)
+
+type snd_mask struct { // size 32
+	bits [(SNDRV_MASK_MAX + 31) / 32]__u32 // offset 0, size 32
+}
+
+type snd_interval struct { // size 12
+	min unsigned_int // offset 0, size 4
+	max unsigned_int // offset 4, size 4
+	bit unsigned_int
+}
+
+type snd_pcm_hw_params struct { // size 608
+	flags     unsigned_int                                                                           // offset 0, size 4
+	masks     [SNDRV_PCM_HW_PARAM_LAST_MASK - SNDRV_PCM_HW_PARAM_FIRST_MASK + 1]snd_mask             // offset 4, size 96
+	mres      [5]snd_mask                                                                            // offset 100, size 160
+	intervals [SNDRV_PCM_HW_PARAM_LAST_INTERVAL - SNDRV_PCM_HW_PARAM_FIRST_INTERVAL + 1]snd_interval // offset 260, size 144
+	ires      [9]snd_interval                                                                        // offset 404, size 108
+	rmask     unsigned_int                                                                           // offset 512, size 4
+	cmask     unsigned_int                                                                           // offset 516, size 4
+	info      unsigned_int                                                                           // offset 520, size 4
+	msbits    unsigned_int                                                                           // offset 524, size 4
+	rate_num  unsigned_int                                                                           // offset 528, size 4
+	rate_den  unsigned_int                                                                           // offset 532, size 4
+	fifo_size snd_pcm_uframes_t                                                                      // offset 536, size 8
+	reserved  [64]unsigned_char                                                                      // offset 544, size 64
+}
+
+type snd_pcm_sw_params struct { // size 136
+	tstamp_mode       signed_int        // offset 0, size 4
+	period_step       unsigned_int      // offset 4, size 4
+	sleep_min         unsigned_int      // offset 8, size 4
+	avail_min         snd_pcm_uframes_t // offset 16, size 8
+	xfer_align        snd_pcm_uframes_t // offset 24, size 8
+	start_threshold   snd_pcm_uframes_t // offset 32, size 8
+	stop_threshold    snd_pcm_uframes_t // offset 40, size 8
+	silence_threshold snd_pcm_uframes_t // offset 48, size 8
+	silence_size      snd_pcm_uframes_t // offset 56, size 8
+	boundary          snd_pcm_uframes_t // offset 64, size 8
+	proto             unsigned_int      // offset 72, size 4
+	tstamp_type       unsigned_int      // offset 76, size 4
+	reserved          [56]unsigned_char // offset 80, size 56
+}

pkg/alsa/device/asound_arch.c

@@ -0,0 +1,163 @@
+#include <stdio.h>
+#include <stddef.h>
+#include <sys/ioctl.h>
+#include <sound/asound.h>
+
+#define print_line(text) printf("%s\n", text)
+#define print_hex_const(name) printf("\t%s = 0x%08lx\n", #name, name)
+#define print_int_const(con) printf("\t%s = %d\n", #con, con)
+
+#define print_struct_header(str) printf("type %s struct { // size %lu\n", #str, sizeof(struct str))
+#define print_struct_member(str, mem, typ) printf("\t%s %s // offset %lu, size %lu\n", #mem == "type" ? "typ" : #mem, typ, offsetof(struct str, mem), sizeof((struct str){0}.mem))
+
+// https://github.com/torvalds/linux/blob/master/include/uapi/sound/asound.h
+int main() {
+    print_line("package device\n");
+
+    print_line("type unsigned_char = byte");
+    print_line("type signed_int = int32");
+    print_line("type unsigned_int = uint32");
+    print_line("type signed_long = int64");
+    print_line("type unsigned_long = uint64");
+    print_line("type __u32 = uint32");
+    print_line("type void__user = uintptr\n");
+
+    print_line("const (");
+    print_int_const(SNDRV_PCM_STREAM_PLAYBACK);
+    print_int_const(SNDRV_PCM_STREAM_CAPTURE);
+	print_line("");
+	print_int_const(SNDRV_PCM_ACCESS_MMAP_INTERLEAVED);
+	print_int_const(SNDRV_PCM_ACCESS_MMAP_NONINTERLEAVED);
+	print_int_const(SNDRV_PCM_ACCESS_MMAP_COMPLEX);
+	print_int_const(SNDRV_PCM_ACCESS_RW_INTERLEAVED);
+	print_int_const(SNDRV_PCM_ACCESS_RW_NONINTERLEAVED);
+	print_line("");
+	print_int_const(SNDRV_PCM_FORMAT_S8);
+	print_int_const(SNDRV_PCM_FORMAT_U8);
+	print_int_const(SNDRV_PCM_FORMAT_S16_LE);
+	print_int_const(SNDRV_PCM_FORMAT_S16_BE);
+	print_int_const(SNDRV_PCM_FORMAT_U16_LE);
+	print_int_const(SNDRV_PCM_FORMAT_U16_BE);
+	print_int_const(SNDRV_PCM_FORMAT_S24_LE);
+	print_int_const(SNDRV_PCM_FORMAT_S24_BE);
+	print_int_const(SNDRV_PCM_FORMAT_U24_LE);
+	print_int_const(SNDRV_PCM_FORMAT_U24_BE);
+	print_int_const(SNDRV_PCM_FORMAT_S32_LE);
+	print_int_const(SNDRV_PCM_FORMAT_S32_BE);
+	print_int_const(SNDRV_PCM_FORMAT_U32_LE);
+	print_int_const(SNDRV_PCM_FORMAT_U32_BE);
+	print_int_const(SNDRV_PCM_FORMAT_FLOAT_LE);
+	print_int_const(SNDRV_PCM_FORMAT_FLOAT_BE);
+	print_int_const(SNDRV_PCM_FORMAT_FLOAT64_LE);
+	print_int_const(SNDRV_PCM_FORMAT_FLOAT64_BE);
+	print_int_const(SNDRV_PCM_FORMAT_MU_LAW);
+	print_int_const(SNDRV_PCM_FORMAT_A_LAW);
+	print_int_const(SNDRV_PCM_FORMAT_MPEG);
+	print_line("");
+    print_hex_const(SNDRV_PCM_IOCTL_PVERSION);        // A 0x00
+    print_hex_const(SNDRV_PCM_IOCTL_INFO);            // A 0x01
+    print_hex_const(SNDRV_PCM_IOCTL_HW_REFINE);       // A 0x10
+    print_hex_const(SNDRV_PCM_IOCTL_HW_PARAMS);       // A 0x11
+    print_hex_const(SNDRV_PCM_IOCTL_SW_PARAMS);       // A 0x13
+    print_hex_const(SNDRV_PCM_IOCTL_PREPARE);         // A 0x40
+    print_hex_const(SNDRV_PCM_IOCTL_WRITEI_FRAMES);   // A 0x50
+    print_hex_const(SNDRV_PCM_IOCTL_READI_FRAMES);    // A 0x51
+    print_line(")\n");
+
+	print_struct_header(snd_pcm_info);
+	print_struct_member(snd_pcm_info, device, "unsigned_int");
+	print_struct_member(snd_pcm_info, subdevice, "unsigned_int");
+	print_struct_member(snd_pcm_info, stream, "signed_int");
+	print_struct_member(snd_pcm_info, card, "signed_int");
+	print_struct_member(snd_pcm_info, id, "[64]unsigned_char");
+	print_struct_member(snd_pcm_info, name, "[80]unsigned_char");
+	print_struct_member(snd_pcm_info, subname, "[32]unsigned_char");
+	print_struct_member(snd_pcm_info, dev_class, "signed_int");
+	print_struct_member(snd_pcm_info, dev_subclass, "signed_int");
+	print_struct_member(snd_pcm_info, subdevices_count, "unsigned_int");
+	print_struct_member(snd_pcm_info, subdevices_avail, "unsigned_int");
+	print_line("\tpad1 [16]unsigned_char");
+	print_struct_member(snd_pcm_info, reserved, "[64]unsigned_char");
+	print_line("}\n");
+
+	print_line("type snd_pcm_uframes_t = unsigned_long");
+	print_line("type snd_pcm_sframes_t = signed_long\n");
+
+	print_struct_header(snd_xferi);
+	print_struct_member(snd_xferi, result, "snd_pcm_sframes_t");
+	print_struct_member(snd_xferi, buf, "void__user");
+	print_struct_member(snd_xferi, frames, "snd_pcm_uframes_t");
+	print_line("}\n");
+
+	print_line("const (");
+	print_int_const(SNDRV_PCM_HW_PARAM_ACCESS);
+	print_int_const(SNDRV_PCM_HW_PARAM_FORMAT);
+	print_int_const(SNDRV_PCM_HW_PARAM_SUBFORMAT);
+	print_int_const(SNDRV_PCM_HW_PARAM_FIRST_MASK);
+	print_int_const(SNDRV_PCM_HW_PARAM_LAST_MASK);
+	print_line("");
+	print_int_const(SNDRV_PCM_HW_PARAM_SAMPLE_BITS);
+	print_int_const(SNDRV_PCM_HW_PARAM_FRAME_BITS);
+	print_int_const(SNDRV_PCM_HW_PARAM_CHANNELS);
+	print_int_const(SNDRV_PCM_HW_PARAM_RATE);
+	print_int_const(SNDRV_PCM_HW_PARAM_PERIOD_TIME);
+	print_int_const(SNDRV_PCM_HW_PARAM_PERIOD_SIZE);
+	print_int_const(SNDRV_PCM_HW_PARAM_PERIOD_BYTES);
+	print_int_const(SNDRV_PCM_HW_PARAM_PERIODS);
+	print_int_const(SNDRV_PCM_HW_PARAM_BUFFER_TIME);
+	print_int_const(SNDRV_PCM_HW_PARAM_BUFFER_SIZE);
+	print_int_const(SNDRV_PCM_HW_PARAM_BUFFER_BYTES);
+	print_int_const(SNDRV_PCM_HW_PARAM_TICK_TIME);
+	print_int_const(SNDRV_PCM_HW_PARAM_FIRST_INTERVAL);
+	print_int_const(SNDRV_PCM_HW_PARAM_LAST_INTERVAL);
+	print_line("");
+	print_int_const(SNDRV_MASK_MAX);
+	print_line("");
+	print_int_const(SNDRV_PCM_TSTAMP_NONE);
+	print_int_const(SNDRV_PCM_TSTAMP_ENABLE);
+	print_line(")\n");
+
+	print_struct_header(snd_mask);
+	print_struct_member(snd_mask, bits, "[(SNDRV_MASK_MAX+31)/32]__u32");
+	print_line("}\n");
+
+	print_struct_header(snd_interval);
+	print_struct_member(snd_interval, min, "unsigned_int");
+	print_struct_member(snd_interval, max, "unsigned_int");
+	print_line("\tbit unsigned_int");
+	print_line("}\n");
+
+	print_struct_header(snd_pcm_hw_params);
+	print_struct_member(snd_pcm_hw_params, flags, "unsigned_int");
+	print_struct_member(snd_pcm_hw_params, masks, "[SNDRV_PCM_HW_PARAM_LAST_MASK-SNDRV_PCM_HW_PARAM_FIRST_MASK+1]snd_mask");
+	print_struct_member(snd_pcm_hw_params, mres, "[5]snd_mask");
+	print_struct_member(snd_pcm_hw_params, intervals, "[SNDRV_PCM_HW_PARAM_LAST_INTERVAL-SNDRV_PCM_HW_PARAM_FIRST_INTERVAL+1]snd_interval");
+	print_struct_member(snd_pcm_hw_params, ires, "[9]snd_interval");
+	print_struct_member(snd_pcm_hw_params, rmask, "unsigned_int");
+	print_struct_member(snd_pcm_hw_params, cmask, "unsigned_int");
+	print_struct_member(snd_pcm_hw_params, info, "unsigned_int");
+	print_struct_member(snd_pcm_hw_params, msbits, "unsigned_int");
+	print_struct_member(snd_pcm_hw_params, rate_num, "unsigned_int");
+	print_struct_member(snd_pcm_hw_params, rate_den, "unsigned_int");
+	print_struct_member(snd_pcm_hw_params, fifo_size, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_hw_params, reserved, "[64]unsigned_char");
+	print_line("}\n");
+
+	print_struct_header(snd_pcm_sw_params);
+	print_struct_member(snd_pcm_sw_params, tstamp_mode, "signed_int");
+	print_struct_member(snd_pcm_sw_params, period_step, "unsigned_int");
+	print_struct_member(snd_pcm_sw_params, sleep_min, "unsigned_int");
+	print_struct_member(snd_pcm_sw_params, avail_min, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_sw_params, xfer_align, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_sw_params, start_threshold, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_sw_params, stop_threshold, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_sw_params, silence_threshold, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_sw_params, silence_size, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_sw_params, boundary, "snd_pcm_uframes_t");
+	print_struct_member(snd_pcm_sw_params, proto, "unsigned_int");
+	print_struct_member(snd_pcm_sw_params, tstamp_type, "unsigned_int");
+	print_struct_member(snd_pcm_sw_params, reserved, "[56]unsigned_char");
+	print_line("}\n");
+
+	return 0;
+}

pkg/alsa/device/asound_mipsle.go

@@ -0,0 +1,146 @@
+package device
+
+type unsigned_char = byte
+type signed_int = int32
+type unsigned_int = uint32
+type signed_long = int64
+type unsigned_long = uint64
+type __u32 = uint32
+type void__user = uintptr
+
+const (
+	SNDRV_PCM_STREAM_PLAYBACK = 0
+	SNDRV_PCM_STREAM_CAPTURE  = 1
+
+	SNDRV_PCM_ACCESS_MMAP_INTERLEAVED    = 0
+	SNDRV_PCM_ACCESS_MMAP_NONINTERLEAVED = 1
+	SNDRV_PCM_ACCESS_MMAP_COMPLEX        = 2
+	SNDRV_PCM_ACCESS_RW_INTERLEAVED      = 3
+	SNDRV_PCM_ACCESS_RW_NONINTERLEAVED   = 4
+
+	SNDRV_PCM_FORMAT_S8         = 0
+	SNDRV_PCM_FORMAT_U8         = 1
+	SNDRV_PCM_FORMAT_S16_LE     = 2
+	SNDRV_PCM_FORMAT_S16_BE     = 3
+	SNDRV_PCM_FORMAT_U16_LE     = 4
+	SNDRV_PCM_FORMAT_U16_BE     = 5
+	SNDRV_PCM_FORMAT_S24_LE     = 6
+	SNDRV_PCM_FORMAT_S24_BE     = 7
+	SNDRV_PCM_FORMAT_U24_LE     = 8
+	SNDRV_PCM_FORMAT_U24_BE     = 9
+	SNDRV_PCM_FORMAT_S32_LE     = 10
+	SNDRV_PCM_FORMAT_S32_BE     = 11
+	SNDRV_PCM_FORMAT_U32_LE     = 12
+	SNDRV_PCM_FORMAT_U32_BE     = 13
+	SNDRV_PCM_FORMAT_FLOAT_LE   = 14
+	SNDRV_PCM_FORMAT_FLOAT_BE   = 15
+	SNDRV_PCM_FORMAT_FLOAT64_LE = 16
+	SNDRV_PCM_FORMAT_FLOAT64_BE = 17
+	SNDRV_PCM_FORMAT_MU_LAW     = 20
+	SNDRV_PCM_FORMAT_A_LAW      = 21
+	SNDRV_PCM_FORMAT_MPEG       = 23
+
+	SNDRV_PCM_IOCTL_PVERSION      = 0x40044100
+	SNDRV_PCM_IOCTL_INFO          = 0x41204101
+	SNDRV_PCM_IOCTL_HW_REFINE     = 0xc25c4110
+	SNDRV_PCM_IOCTL_HW_PARAMS     = 0xc25c4111
+	SNDRV_PCM_IOCTL_SW_PARAMS     = 0xc0684113
+	SNDRV_PCM_IOCTL_PREPARE       = 0x20004140
+	SNDRV_PCM_IOCTL_WRITEI_FRAMES = 0x800c4150
+	SNDRV_PCM_IOCTL_READI_FRAMES  = 0x400c4151
+)
+
+type snd_pcm_info struct { // size 288
+	device           unsigned_int      // offset 0, size 4
+	subdevice        unsigned_int      // offset 4, size 4
+	stream           signed_int        // offset 8, size 4
+	card             signed_int        // offset 12, size 4
+	id               [64]unsigned_char // offset 16, size 64
+	name             [80]unsigned_char // offset 80, size 80
+	subname          [32]unsigned_char // offset 160, size 32
+	dev_class        signed_int        // offset 192, size 4
+	dev_subclass     signed_int        // offset 196, size 4
+	subdevices_count unsigned_int      // offset 200, size 4
+	subdevices_avail unsigned_int      // offset 204, size 4
+	pad1             [16]unsigned_char
+	reserved         [64]unsigned_char // offset 224, size 64
+}
+
+type snd_pcm_uframes_t = unsigned_long
+type snd_pcm_sframes_t = signed_long
+
+type snd_xferi struct { // size 12
+	result snd_pcm_sframes_t // offset 0, size 4
+	buf    void__user        // offset 4, size 4
+	frames snd_pcm_uframes_t // offset 8, size 4
+}
+
+const (
+	SNDRV_PCM_HW_PARAM_ACCESS     = 0
+	SNDRV_PCM_HW_PARAM_FORMAT     = 1
+	SNDRV_PCM_HW_PARAM_SUBFORMAT  = 2
+	SNDRV_PCM_HW_PARAM_FIRST_MASK = 0
+	SNDRV_PCM_HW_PARAM_LAST_MASK  = 2
+
+	SNDRV_PCM_HW_PARAM_SAMPLE_BITS    = 8
+	SNDRV_PCM_HW_PARAM_FRAME_BITS     = 9
+	SNDRV_PCM_HW_PARAM_CHANNELS       = 10
+	SNDRV_PCM_HW_PARAM_RATE           = 11
+	SNDRV_PCM_HW_PARAM_PERIOD_TIME    = 12
+	SNDRV_PCM_HW_PARAM_PERIOD_SIZE    = 13
+	SNDRV_PCM_HW_PARAM_PERIOD_BYTES   = 14
+	SNDRV_PCM_HW_PARAM_PERIODS        = 15
+	SNDRV_PCM_HW_PARAM_BUFFER_TIME    = 16
+	SNDRV_PCM_HW_PARAM_BUFFER_SIZE    = 17
+	SNDRV_PCM_HW_PARAM_BUFFER_BYTES   = 18
+	SNDRV_PCM_HW_PARAM_TICK_TIME      = 19
+	SNDRV_PCM_HW_PARAM_FIRST_INTERVAL = 8
+	SNDRV_PCM_HW_PARAM_LAST_INTERVAL  = 19
+
+	SNDRV_MASK_MAX = 256
+
+	SNDRV_PCM_TSTAMP_NONE   = 0
+	SNDRV_PCM_TSTAMP_ENABLE = 1
+)
+
+type snd_mask struct { // size 32
+	bits [(SNDRV_MASK_MAX + 31) / 32]__u32 // offset 0, size 32
+}
+
+type snd_interval struct { // size 12
+	min unsigned_int // offset 0, size 4
+	max unsigned_int // offset 4, size 4
+	bit unsigned_int
+}
+
+type snd_pcm_hw_params struct { // size 604
+	flags     unsigned_int                                                                           // offset 0, size 4
+	masks     [SNDRV_PCM_HW_PARAM_LAST_MASK - SNDRV_PCM_HW_PARAM_FIRST_MASK + 1]snd_mask             // offset 4, size 96
+	mres      [5]snd_mask                                                                            // offset 100, size 160
+	intervals [SNDRV_PCM_HW_PARAM_LAST_INTERVAL - SNDRV_PCM_HW_PARAM_FIRST_INTERVAL + 1]snd_interval // offset 260, size 144
+	ires      [9]snd_interval                                                                        // offset 404, size 108
+	rmask     unsigned_int                                                                           // offset 512, size 4
+	cmask     unsigned_int                                                                           // offset 516, size 4
+	info      unsigned_int                                                                           // offset 520, size 4
+	msbits    unsigned_int                                                                           // offset 524, size 4
+	rate_num  unsigned_int                                                                           // offset 528, size 4
+	rate_den  unsigned_int                                                                           // offset 532, size 4
+	fifo_size snd_pcm_uframes_t                                                                      // offset 536, size 4
+	reserved  [64]unsigned_char                                                                      // offset 540, size 64
+}
+
+type snd_pcm_sw_params struct { // size 104
+	tstamp_mode       signed_int        // offset 0, size 4
+	period_step       unsigned_int      // offset 4, size 4
+	sleep_min         unsigned_int      // offset 8, size 4
+	avail_min         snd_pcm_uframes_t // offset 12, size 4
+	xfer_align        snd_pcm_uframes_t // offset 16, size 4
+	start_threshold   snd_pcm_uframes_t // offset 20, size 4
+	stop_threshold    snd_pcm_uframes_t // offset 24, size 4
+	silence_threshold snd_pcm_uframes_t // offset 28, size 4
+	silence_size      snd_pcm_uframes_t // offset 32, size 4
+	boundary          snd_pcm_uframes_t // offset 36, size 4
+	proto             unsigned_int      // offset 40, size 4
+	tstamp_type       unsigned_int      // offset 44, size 4
+	reserved          [56]unsigned_char // offset 48, size 56
+}

pkg/alsa/device/device_linux.go

@@ -0,0 +1,231 @@
+package device
+
+import (
+	"fmt"
+	"syscall"
+	"unsafe"
+)
+
+type Device struct {
+	fd   uintptr
+	path string
+
+	hwparams   snd_pcm_hw_params
+	frameBytes int // sample size * channels
+}
+
+func Open(path string) (*Device, error) {
+	// important to use nonblock because can get lock
+	fd, err := syscall.Open(path, syscall.O_RDWR|syscall.O_NONBLOCK, 0)
+	if err != nil {
+		return nil, err
+	}
+
+	// important to remove nonblock because better to handle reads and writes
+	if err = syscall.SetNonblock(fd, false); err != nil {
+		return nil, err
+	}
+
+	d := &Device{fd: uintptr(fd), path: path}
+	d.init()
+
+	// load all supported formats, channels, rates, etc.
+	if err = ioctl(d.fd, SNDRV_PCM_IOCTL_HW_REFINE, &d.hwparams); err != nil {
+		_ = d.Close()
+		return nil, err
+	}
+
+	d.setMask(SNDRV_PCM_HW_PARAM_ACCESS, SNDRV_PCM_ACCESS_RW_INTERLEAVED)
+
+	return d, nil
+}
+
+func (d *Device) Close() error {
+	return syscall.Close(int(d.fd))
+}
+
+func (d *Device) IsCapture() bool {
+	// path: /dev/snd/pcmC0D0c, where p - playback, c - capture
+	return d.path[len(d.path)-1] == 'c'
+}
+
+type Info struct {
+	Card      int
+	Device    int
+	SubDevice int
+	Stream    int
+	ID        string
+	Name      string
+	SubName   string
+}
+
+func (d *Device) Info() (*Info, error) {
+	var info snd_pcm_info
+	if err := ioctl(d.fd, SNDRV_PCM_IOCTL_INFO, &info); err != nil {
+		return nil, err
+	}
+	return &Info{
+		Card:      int(info.card),
+		Device:    int(info.device),
+		SubDevice: int(info.subdevice),
+		Stream:    int(info.stream),
+		ID:        str(info.id[:]),
+		Name:      str(info.name[:]),
+		SubName:   str(info.subname[:]),
+	}, nil
+}
+
+func (d *Device) CheckFormat(format byte) bool {
+	return d.checkMask(SNDRV_PCM_HW_PARAM_FORMAT, uint32(format))
+}
+
+func (d *Device) ListFormats() (formats []byte) {
+	for i := byte(0); i <= 28; i++ {
+		if d.CheckFormat(i) {
+			formats = append(formats, i)
+		}
+	}
+	return
+}
+
+func (d *Device) RangeRates() (uint32, uint32) {
+	return d.getInterval(SNDRV_PCM_HW_PARAM_RATE)
+}
+
+func (d *Device) RangeChannels() (byte, byte) {
+	minCh, maxCh := d.getInterval(SNDRV_PCM_HW_PARAM_CHANNELS)
+	return byte(minCh), byte(maxCh)
+}
+
+func (d *Device) GetRateNear(rate uint32) uint32 {
+	r1, r2 := d.RangeRates()
+	if rate < r1 {
+		return r1
+	}
+	if rate > r2 {
+		return r2
+	}
+	return rate
+}
+
+func (d *Device) GetChannelsNear(channels byte) byte {
+	c1, c2 := d.RangeChannels()
+	if channels < c1 {
+		return c1
+	}
+	if channels > c2 {
+		return c2
+	}
+	return channels
+}
+
+const bufferSize = 4096
+
+func (d *Device) SetHWParams(format byte, rate uint32, channels byte) error {
+	d.setInterval(SNDRV_PCM_HW_PARAM_CHANNELS, uint32(channels))
+	d.setInterval(SNDRV_PCM_HW_PARAM_RATE, rate)
+	d.setMask(SNDRV_PCM_HW_PARAM_FORMAT, uint32(format))
+	//d.setMask(SNDRV_PCM_HW_PARAM_SUBFORMAT, 0)
+
+	// important for smooth playback
+	d.setInterval(SNDRV_PCM_HW_PARAM_BUFFER_SIZE, bufferSize)
+	//d.setInterval(SNDRV_PCM_HW_PARAM_PERIOD_SIZE, 2000)
+
+	if err := ioctl(d.fd, SNDRV_PCM_IOCTL_HW_PARAMS, &d.hwparams); err != nil {
+		return fmt.Errorf("[alsa] set hw_params: %w", err)
+	}
+
+	_, i := d.getInterval(SNDRV_PCM_HW_PARAM_FRAME_BITS)
+	d.frameBytes = int(i / 8)
+
+	_, periods := d.getInterval(SNDRV_PCM_HW_PARAM_PERIODS)
+	_, periodSize := d.getInterval(SNDRV_PCM_HW_PARAM_PERIOD_SIZE)
+	threshold := snd_pcm_uframes_t(periods * periodSize) // same as bufferSize
+
+	swparams := snd_pcm_sw_params{
+		//tstamp_mode: SNDRV_PCM_TSTAMP_ENABLE,
+		period_step:    1,
+		avail_min:      1, // start as soon as possible
+		stop_threshold: threshold,
+	}
+
+	if d.IsCapture() {
+		swparams.start_threshold = 1
+	} else {
+		swparams.start_threshold = threshold
+	}
+
+	if err := ioctl(d.fd, SNDRV_PCM_IOCTL_SW_PARAMS, &swparams); err != nil {
+		return fmt.Errorf("[alsa] set sw_params: %w", err)
+	}
+
+	if err := ioctl(d.fd, SNDRV_PCM_IOCTL_PREPARE, nil); err != nil {
+		return fmt.Errorf("[alsa] prepare: %w", err)
+	}
+
+	return nil
+}
+
+func (d *Device) Write(b []byte) (n int, err error) {
+	xfer := &snd_xferi{
+		buf:    uintptr(unsafe.Pointer(&b[0])),
+		frames: snd_pcm_uframes_t(len(b) / d.frameBytes),
+	}
+	err = ioctl(d.fd, SNDRV_PCM_IOCTL_WRITEI_FRAMES, xfer)
+	if err == syscall.EPIPE {
+		// auto handle underrun state
+		// https://stackoverflow.com/questions/59396728/how-to-properly-handle-xrun-in-alsa-programming-when-playing-audio-with-snd-pcm
+		err = ioctl(d.fd, SNDRV_PCM_IOCTL_PREPARE, nil)
+	}
+	n = int(xfer.result) * d.frameBytes
+	return
+}
+
+func (d *Device) Read(b []byte) (n int, err error) {
+	xfer := &snd_xferi{
+		buf:    uintptr(unsafe.Pointer(&b[0])),
+		frames: snd_pcm_uframes_t(len(b) / d.frameBytes),
+	}
+	err = ioctl(d.fd, SNDRV_PCM_IOCTL_READI_FRAMES, xfer)
+	n = int(xfer.result) * d.frameBytes
+	return
+}
+
+func (d *Device) init() {
+	for i := range d.hwparams.masks {
+		d.hwparams.masks[i].bits[0] = 0xFFFFFFFF
+		d.hwparams.masks[i].bits[1] = 0xFFFFFFFF
+	}
+	for i := range d.hwparams.intervals {
+		d.hwparams.intervals[i].max = 0xFFFFFFFF
+	}
+
+	d.hwparams.rmask = 0xFFFFFFFF
+	d.hwparams.cmask = 0
+	d.hwparams.info = 0xFFFFFFFF
+}
+
+func (d *Device) setInterval(param, val uint32) {
+	d.hwparams.intervals[param-SNDRV_PCM_HW_PARAM_FIRST_INTERVAL].min = val
+	d.hwparams.intervals[param-SNDRV_PCM_HW_PARAM_FIRST_INTERVAL].max = val
+	d.hwparams.intervals[param-SNDRV_PCM_HW_PARAM_FIRST_INTERVAL].bit = 0b0100 // integer
+}
+
+func (d *Device) setIntervalMin(param, val uint32) {
+	d.hwparams.intervals[param-SNDRV_PCM_HW_PARAM_FIRST_INTERVAL].min = val
+}
+
+func (d *Device) getInterval(param uint32) (uint32, uint32) {
+	return d.hwparams.intervals[param-SNDRV_PCM_HW_PARAM_FIRST_INTERVAL].min,
+		d.hwparams.intervals[param-SNDRV_PCM_HW_PARAM_FIRST_INTERVAL].max
+}
+
+func (d *Device) setMask(mask, val uint32) {
+	d.hwparams.masks[mask].bits[0] = 0
+	d.hwparams.masks[mask].bits[1] = 0
+	d.hwparams.masks[mask].bits[val>>5] = 1 << (val & 0x1F)
+}
+
+func (d *Device) checkMask(mask, val uint32) bool {
+	return d.hwparams.masks[mask].bits[val>>5]&(1<<(val&0x1F)) > 0
+}

pkg/alsa/device/ioctl_linux.go

@@ -0,0 +1,26 @@
+package device
+
+import (
+	"bytes"
+	"reflect"
+	"syscall"
+)
+
+func ioctl(fd, req uintptr, arg any) error {
+	var ptr uintptr
+	if arg != nil {
+		ptr = reflect.ValueOf(arg).Pointer()
+	}
+	_, _, err := syscall.Syscall(syscall.SYS_IOCTL, fd, req, ptr)
+	if err != 0 {
+		return err
+	}
+	return nil
+}
+
+func str(b []byte) string {
+	if i := bytes.IndexByte(b, 0); i >= 0 {
+		return string(b[:i])
+	}
+	return string(b)
+}

pkg/alsa/open_linux.go

@@ -0,0 +1,44 @@
+package alsa
+
+import (
+	"errors"
+	"fmt"
+	"net/url"
+
+	"github.com/AlexxIT/go2rtc/pkg/alsa/device"
+	"github.com/AlexxIT/go2rtc/pkg/core"
+)
+
+func Open(rawURL string) (core.Producer, error) {
+	// Example (ffmpeg source compatible):
+	// alsa:device?audio=/dev/snd/pcmC0D0p
+	// TODO: ?audio=default
+	// TODO: ?audio=hw:0,0
+	// TODO: &sample_rate=48000&channels=2
+	// TODO: &backchannel=1
+	u, err := url.Parse(rawURL)
+	if err != nil {
+		return nil, err
+	}
+
+	path := u.Query().Get("audio")
+	dev, err := device.Open(path)
+	if err != nil {
+		return nil, err
+	}
+
+	if !dev.CheckFormat(device.SNDRV_PCM_FORMAT_S16_LE) {
+		_ = dev.Close()
+		return nil, errors.New("alsa: format S16LE not supported")
+	}
+
+	switch path[len(path)-1] {
+	case 'p': // playback
+		return newPlayback(dev)
+	case 'c': // capture
+		return newCapture(dev)
+	}
+
+	_ = dev.Close()
+	return nil, fmt.Errorf("alsa: unknown path: %s", path)
+}

pkg/alsa/playback_linux.go

@@ -0,0 +1,84 @@
+package alsa
+
+import (
+	"fmt"
+
+	"github.com/AlexxIT/go2rtc/pkg/alsa/device"
+	"github.com/AlexxIT/go2rtc/pkg/core"
+	"github.com/AlexxIT/go2rtc/pkg/pcm"
+	"github.com/pion/rtp"
+)
+
+type Playback struct {
+	core.Connection
+	dev    *device.Device
+	closed core.Waiter
+}
+
+func newPlayback(dev *device.Device) (*Playback, error) {
+	medias := []*core.Media{
+		{
+			Kind:      core.KindAudio,
+			Direction: core.DirectionSendonly,
+			Codecs: []*core.Codec{
+				{Name: core.CodecPCML},                  // support ffmpeg producer (auto transcode)
+				{Name: core.CodecPCMA, ClockRate: 8000}, // support webrtc producer
+			},
+		},
+	}
+	return &Playback{
+		Connection: core.Connection{
+			ID:         core.NewID(),
+			FormatName: "alsa",
+			Medias:     medias,
+			Transport:  dev,
+		},
+		dev: dev,
+	}, nil
+}
+
+func (p *Playback) GetTrack(media *core.Media, codec *core.Codec) (*core.Receiver, error) {
+	return nil, core.ErrCantGetTrack
+}
+
+func (p *Playback) AddTrack(media *core.Media, codec *core.Codec, track *core.Receiver) error {
+	src := track.Codec
+	dst := &core.Codec{
+		Name:      core.CodecPCML,
+		ClockRate: p.dev.GetRateNear(src.ClockRate),
+		Channels:  p.dev.GetChannelsNear(src.Channels),
+	}
+	sender := core.NewSender(media, dst)
+
+	sender.Handler = func(pkt *rtp.Packet) {
+		if n, err := p.dev.Write(pkt.Payload); err == nil {
+			p.Send += n
+		}
+	}
+
+	if sender.Handler = pcm.TranscodeHandler(dst, src, sender.Handler); sender.Handler == nil {
+		return fmt.Errorf("alsa: can't convert %s to %s", src, dst)
+	}
+
+	// typical card support:
+	// - Formats: S16_LE, S32_LE
+	// - ClockRates: 8000 - 192000
+	// - Channels: 2 - 10
+	err := p.dev.SetHWParams(device.SNDRV_PCM_FORMAT_S16_LE, dst.ClockRate, byte(dst.Channels))
+	if err != nil {
+		return err
+	}
+
+	sender.HandleRTP(track)
+	p.Senders = append(p.Senders, sender)
+	return nil
+}
+
+func (p *Playback) Start() (err error) {
+	return p.closed.Wait()
+}
+
+func (p *Playback) Stop() error {
+	p.closed.Done(nil)
+	return p.Connection.Stop()
+}

pkg/core/codec.go

@@ -13,7 +13,7 @@ import (
 type Codec struct {
 	Name        string // H264, PCMU, PCMA, opus...
 	ClockRate   uint32 // 90000, 8000, 16000...
-	Channels    uint16 // 0, 1, 2
+	Channels    uint8  // 0, 1, 2
 	FmtpLine    string
 	PayloadType uint8
 }

pkg/core/media.go

@@ -139,7 +139,7 @@ func MarshalSDP(name string, medias []*Media) ([]byte, error) {
 				Protos: []string{"RTP", "AVP"},
 			},
 		}
-		md.WithCodec(codec.PayloadType, name, codec.ClockRate, codec.Channels, codec.FmtpLine)
+		md.WithCodec(codec.PayloadType, name, codec.ClockRate, uint16(codec.Channels), codec.FmtpLine)
 
 		if media.Direction != "" {
 			md.WithPropertyAttribute(media.Direction)

pkg/homekit/helpers.go

@@ -50,7 +50,7 @@ func audioToMedia(codecs []camera.AudioCodec) *core.Media {
 				mediaCodec := &core.Codec{
 					Name:      audioCodecs[codec.CodecType],
 					ClockRate: audioSampleRates[sampleRate],
-					Channels:  uint16(param.Channels),
+					Channels:  param.Channels,
 				}
 
 				if mediaCodec.Name == core.CodecELD {

pkg/ioctl/README.md

@@ -0,0 +1,3 @@
+# IOCTL
+
+This is just an example how Linux IOCTL constants works.

pkg/ioctl/ioctl.go

@@ -0,0 +1,28 @@
+package ioctl
+
+import (
+	"bytes"
+)
+
+func Str(b []byte) string {
+	if i := bytes.IndexByte(b, 0); i >= 0 {
+		return string(b[:i])
+	}
+	return string(b)
+}
+
+func io(mode byte, type_ byte, number byte, size uint16) uintptr {
+	return uintptr(mode)<<30 | uintptr(size)<<16 | uintptr(type_)<<8 | uintptr(number)
+}
+
+func IOR(type_ byte, number byte, size uint16) uintptr {
+	return io(read, type_, number, size)
+}
+
+func IOW(type_ byte, number byte, size uint16) uintptr {
+	return io(write, type_, number, size)
+}
+
+func IORW(type_ byte, number byte, size uint16) uintptr {
+	return io(read|write, type_, number, size)
+}

pkg/ioctl/ioctl_be.go

@@ -0,0 +1,8 @@
+//go:build arm || arm64 || 386 || amd64
+
+package ioctl
+
+const (
+	write = 1
+	read  = 2
+)

pkg/ioctl/ioctl_le.go

@@ -0,0 +1,8 @@
+//go:build mipsle
+
+package ioctl
+
+const (
+	read  = 1
+	write = 2
+)

pkg/ioctl/ioctl_linux.go

@@ -0,0 +1,14 @@
+package ioctl
+
+import (
+	"syscall"
+	"unsafe"
+)
+
+func Ioctl(fd int, req uint, arg unsafe.Pointer) error {
+	_, _, err := syscall.Syscall(syscall.SYS_IOCTL, uintptr(fd), uintptr(req), uintptr(arg))
+	if err != 0 {
+		return err
+	}
+	return nil
+}

pkg/ioctl/ioctl_test.go

@@ -0,0 +1,16 @@
+package ioctl
+
+import (
+	"runtime"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestIOR(t *testing.T) {
+	// #define SNDRV_PCM_IOCTL_INFO		_IOR('A', 0x01, struct snd_pcm_info)
+	if runtime.GOARCH == "arm64" {
+		c := IOR('A', 0x01, 288)
+		require.Equal(t, uintptr(0x81204101), c)
+	}
+}

pkg/mp4/muxer.go

@@ -89,12 +89,12 @@ func (m *Muxer) GetInit() ([]byte, error) {
 			}
 
 			mv.WriteAudioTrack(
-				uint32(i+1), codec.Name, codec.ClockRate, codec.Channels, b,
+				uint32(i+1), codec.Name, codec.ClockRate, uint16(codec.Channels), b,
 			)
 
 		case core.CodecOpus, core.CodecMP3, core.CodecPCMA, core.CodecPCMU, core.CodecFLAC:
 			mv.WriteAudioTrack(
-				uint32(i+1), codec.Name, codec.ClockRate, codec.Channels, nil,
+				uint32(i+1), codec.Name, codec.ClockRate, uint16(codec.Channels), nil,
 			)
 		}
 	}

pkg/pcm/handlers.go

@@ -0,0 +1,99 @@
+package pcm
+
+import (
+	"sync"
+
+	"github.com/AlexxIT/go2rtc/pkg/core"
+	"github.com/pion/rtp"
+)
+
+// RepackG711 - Repack G.711 PCMA/PCMU into frames of size 1024
+//  1. Fixes WebRTC audio quality issue (monotonic timestamp)
+//  2. Fixes Reolink Doorbell backchannel issue (zero timestamp)
+//     https://github.com/AlexxIT/go2rtc/issues/331
+func RepackG711(zeroTS bool, handler core.HandlerFunc) core.HandlerFunc {
+	const PacketSize = 1024
+
+	var buf []byte
+	var seq uint16
+	var ts uint32
+
+	// fix https://github.com/AlexxIT/go2rtc/issues/432
+	var mu sync.Mutex
+
+	return func(packet *rtp.Packet) {
+		mu.Lock()
+
+		buf = append(buf, packet.Payload...)
+		if len(buf) < PacketSize {
+			mu.Unlock()
+			return
+		}
+
+		pkt := &rtp.Packet{
+			Header: rtp.Header{
+				Version:        2,
+				Marker:         true,               // should be true
+				PayloadType:    packet.PayloadType, // will be owerwriten
+				SequenceNumber: seq,
+				SSRC:           packet.SSRC,
+			},
+			Payload: buf[:PacketSize],
+		}
+
+		seq++
+
+		// don't know if zero TS important for Reolink Doorbell
+		// don't have this strange devices for tests
+		if !zeroTS {
+			pkt.Timestamp = ts
+			ts += PacketSize
+		}
+
+		buf = buf[PacketSize:]
+
+		mu.Unlock()
+
+		handler(pkt)
+	}
+}
+
+// LittleToBig - convert PCM little endian to PCM big endian
+func LittleToBig(handler core.HandlerFunc) core.HandlerFunc {
+	return func(packet *rtp.Packet) {
+		clone := *packet
+		clone.Payload = FlipEndian(packet.Payload)
+		handler(&clone)
+	}
+}
+
+func TranscodeHandler(dst, src *core.Codec, handler core.HandlerFunc) core.HandlerFunc {
+	var ts uint32
+	k := float32(BytesPerFrame(dst)) / float32(BytesPerFrame(src))
+	f := Transcode(dst, src)
+
+	return func(packet *rtp.Packet) {
+		ts += uint32(k * float32(len(packet.Payload)))
+
+		clone := *packet
+		clone.Payload = f(packet.Payload)
+		clone.Timestamp = ts
+		handler(&clone)
+	}
+}
+
+func BytesPerFrame(codec *core.Codec) byte {
+	channels := byte(codec.Channels)
+	if channels == 0 {
+		channels = 1
+	}
+
+	switch codec.Name {
+	case core.CodecPCML, core.CodecPCM:
+		return 2 * channels
+	case core.CodecPCMU, core.CodecPCMA:
+		return channels
+	}
+
+	return 0
+}

pkg/pcm/pcm.go

@@ -1,200 +1,187 @@
 package pcm
 
-import (
+import "github.com/AlexxIT/go2rtc/pkg/core"
-	"sync"
 
-	"github.com/AlexxIT/go2rtc/pkg/core"
+func Downsample(k float32) func([]int16) []int16 {
-	"github.com/pion/rtp"
-)
-
-// ResampleToG711 - convert PCMA/PCM/PCML to PCMA and PCMU to PCMU with decreasing sample rate
-func ResampleToG711(codec *core.Codec, sampleRate uint32, handler core.HandlerFunc) core.HandlerFunc {
-	n := float32(codec.ClockRate) / float32(sampleRate)
-
-	if codec.Channels == 2 {
-		n *= 2 // hacky way for support two channels audio
-	}
-
-	switch codec.Name {
-	case core.CodecPCMA:
-		return DownsampleByte(PCMAtoPCM, PCMtoPCMA, n, handler)
-	case core.CodecPCMU:
-		return DownsampleByte(PCMUtoPCM, PCMtoPCMU, n, handler)
-	case core.CodecPCM, core.CodecPCML:
-		if n == 1 {
-			handler = ResamplePCM(PCMtoPCMA, handler)
-		} else {
-			handler = DownsamplePCM(PCMtoPCMA, n, handler)
-		}
-
-		if codec.Name == core.CodecPCML {
-			return LittleToBig(handler)
-		}
-
-		return handler
-	}
-
-	panic(core.Caller())
-}
-
-// DownsampleByte - convert PCMA/PCMU to PCMA/PCMU with decreasing sample rate (N times)
-func DownsampleByte(
-	toPCM func(byte) int16, fromPCM func(int16) byte, n float32, handler core.HandlerFunc,
-) core.HandlerFunc {
 	var sampleN, sampleSum float32
-	var ts uint32
-
-	return func(packet *rtp.Packet) {
-		samples := len(packet.Payload)
-		newLen := uint32((float32(samples) + sampleN) / n)
-
-		oldSamples := packet.Payload
-		newSamples := make([]byte, newLen)
 
+	return func(src []int16) (dst []int16) {
 		var i int
-		for _, sample := range oldSamples {
+		dst = make([]int16, int((float32(len(src))+sampleN)/k))
-			sampleSum += float32(toPCM(sample))
+		for _, sample := range src {
-			if sampleN++; sampleN >= n {
+			sampleSum += float32(sample)
-				newSamples[i] = fromPCM(int16(sampleSum / n))
+			sampleN++
+			if sampleN >= k {
+				dst[i] = int16(sampleSum / k)
 				i++
 
 				sampleSum = 0
-				sampleN -= n
+				sampleN -= k
 			}
 		}
-
+		return
-		ts += newLen
-
-		clone := *packet
-		clone.Payload = newSamples
-		clone.Timestamp = ts
-		handler(&clone)
 	}
 }
 
-// LittleToBig - conver PCM little endian to PCM big endian
+func Upsample(k float32) func([]int16) []int16 {
-func LittleToBig(handler core.HandlerFunc) core.HandlerFunc {
+	var sampleN float32
-	return func(packet *rtp.Packet) {
-		size := len(packet.Payload)
-		b := make([]byte, size)
-		for i := 0; i < size; i += 2 {
-			b[i] = packet.Payload[i+1]
-			b[i+1] = packet.Payload[i]
-		}
 
-		clone := *packet
+	return func(src []int16) (dst []int16) {
-		clone.Payload = b
+		var i int
-		handler(&clone)
+		dst = make([]int16, int(k*float32(len(src))))
-	}
+		for _, sample := range src {
-}
+			sampleN += k
+			for sampleN > 0 {
+				dst[i] = sample
+				i++
 
-// ResamplePCM - convert PCM to PCMA/PCMU with same sample rate
+				sampleN -= 1
-func ResamplePCM(fromPCM func(int16) byte, handler core.HandlerFunc) core.HandlerFunc {
-	var ts uint32
-
-	return func(packet *rtp.Packet) {
-		len1 := len(packet.Payload)
-		len2 := len1 / 2
-
-		oldSamples := packet.Payload
-		newSamples := make([]byte, len2)
-
-		var i2 int
-		for i1 := 0; i1 < len1; i1 += 2 {
-			sample := int16(uint16(oldSamples[i1])<<8 | uint16(oldSamples[i1+1]))
-			newSamples[i2] = fromPCM(sample)
-			i2++
-		}
-
-		ts += uint32(len2)
-
-		clone := *packet
-		clone.Payload = newSamples
-		clone.Timestamp = ts
-		handler(&clone)
-	}
-}
-
-// DownsamplePCM - convert PCM to PCMA/PCMU with decreasing sample rate (N times)
-func DownsamplePCM(fromPCM func(int16) byte, n float32, handler core.HandlerFunc) core.HandlerFunc {
-	var sampleN, sampleSum float32
-	var ts uint32
-
-	return func(packet *rtp.Packet) {
-		samples := len(packet.Payload) / 2
-		newLen := uint32((float32(samples) + sampleN) / n)
-
-		oldSamples := packet.Payload
-		newSamples := make([]byte, newLen)
-
-		var i2 int
-		for i1 := 0; i1 < len(packet.Payload); i1 += 2 {
-			sampleSum += float32(int16(uint16(oldSamples[i1])<<8 | uint16(oldSamples[i1+1])))
-			if sampleN++; sampleN >= n {
-				newSamples[i2] = fromPCM(int16(sampleSum / n))
-				i2++
-
-				sampleSum = 0
-				sampleN -= n
 			}
 		}
-
+		return
-		ts += newLen
-
-		clone := *packet
-		clone.Payload = newSamples
-		clone.Timestamp = ts
-		handler(&clone)
 	}
 }
 
-// RepackG711 - Repack G.711 PCMA/PCMU into frames of size 1024
+func FlipEndian(src []byte) (dst []byte) {
-//  1. Fixes WebRTC audio quality issue (monotonic timestamp)
+	var i, j int
-//  2. Fixes Reolink Doorbell backchannel issue (zero timestamp)
+	n := len(src)
-//     https://github.com/AlexxIT/go2rtc/issues/331
+	dst = make([]byte, n)
-func RepackG711(zeroTS bool, handler core.HandlerFunc) core.HandlerFunc {
+	for i < n {
-	const PacketSize = 1024
+		x := src[i]
+		i++
+		dst[j] = src[i]
+		j++
+		i++
+		dst[j] = x
+		j++
+	}
+	return
+}
 
-	var buf []byte
+func Transcode(dst, src *core.Codec) func([]byte) []byte {
-	var seq uint16
+	var reader func([]byte) []int16
-	var ts uint32
+	var writer func([]int16) []byte
+	var filters []func([]int16) []int16
 
-	// fix https://github.com/AlexxIT/go2rtc/issues/432
+	switch src.Name {
-	var mu sync.Mutex
+	case core.CodecPCML:
-
+		reader = func(src []byte) (dst []int16) {
-	return func(packet *rtp.Packet) {
+			var i, j int
-		mu.Lock()
+			n := len(src)
-
+			dst = make([]int16, n/2)
-		buf = append(buf, packet.Payload...)
+			for i < n {
-		if len(buf) < PacketSize {
+				lo := src[i]
-			mu.Unlock()
+				i++
+				hi := src[i]
+				i++
+				dst[j] = int16(hi)<<8 | int16(lo)
+				j++
+			}
 			return
 		}
-
+	case core.CodecPCM:
-		pkt := &rtp.Packet{
+		reader = func(src []byte) (dst []int16) {
-			Header: rtp.Header{
+			var i, j int
-				Version:        2,
+			n := len(src)
-				Marker:         true,               // should be true
+			dst = make([]int16, n/2)
-				PayloadType:    packet.PayloadType, // will be owerwriten
+			for i < n {
-				SequenceNumber: seq,
+				hi := src[i]
-				SSRC:           packet.SSRC,
+				i++
-			},
+				lo := src[i]
-			Payload: buf[:PacketSize],
+				i++
+				dst[j] = int16(hi)<<8 | int16(lo)
+				j++
+			}
+			return
 		}
-
+	case core.CodecPCMU:
-		seq++
+		reader = func(src []byte) (dst []int16) {
-
+			var i int
-		// don't know if zero TS important for Reolink Doorbell
+			dst = make([]int16, len(src))
-		// don't have this strange devices for tests
+			for _, sample := range src {
-		if !zeroTS {
+				dst[i] = PCMUtoPCM(sample)
-			pkt.Timestamp = ts
+				i++
-			ts += PacketSize
+			}
+			return
 		}
+	case core.CodecPCMA:
+		reader = func(src []byte) (dst []int16) {
+			var i int
+			dst = make([]int16, len(src))
+			for _, sample := range src {
+				dst[i] = PCMAtoPCM(sample)
+				i++
+			}
+			return
+		}
+	}
 
-		buf = buf[PacketSize:]
+	if src.Channels > 1 {
+		filters = append(filters, Downsample(float32(src.Channels)))
+	}
 
-		mu.Unlock()
+	if src.ClockRate > dst.ClockRate {
+		filters = append(filters, Downsample(float32(src.ClockRate)/float32(dst.ClockRate)))
+	} else if src.ClockRate < dst.ClockRate {
+		filters = append(filters, Upsample(float32(dst.ClockRate)/float32(src.ClockRate)))
+	}
 
-		handler(pkt)
+	if dst.Channels > 1 {
+		filters = append(filters, Upsample(float32(dst.Channels)))
+	}
+
+	switch dst.Name {
+	case core.CodecPCML:
+		writer = func(src []int16) (dst []byte) {
+			var i int
+			dst = make([]byte, len(src)*2)
+			for _, sample := range src {
+				dst[i] = byte(sample)
+				i++
+				dst[i] = byte(sample >> 8)
+				i++
+			}
+			return
+		}
+	case core.CodecPCM:
+		writer = func(src []int16) (dst []byte) {
+			var i int
+			dst = make([]byte, len(src)*2)
+			for _, sample := range src {
+				dst[i] = byte(sample >> 8)
+				i++
+				dst[i] = byte(sample)
+				i++
+			}
+			return
+		}
+	case core.CodecPCMU:
+		writer = func(src []int16) (dst []byte) {
+			var i int
+			dst = make([]byte, len(src))
+			for _, sample := range src {
+				dst[i] = PCMtoPCMU(sample)
+				i++
+			}
+			return
+		}
+	case core.CodecPCMA:
+		writer = func(src []int16) (dst []byte) {
+			var i int
+			dst = make([]byte, len(src))
+			for _, sample := range src {
+				dst[i] = PCMtoPCMA(sample)
+				i++
+			}
+			return
+		}
+	}
+
+	return func(b []byte) []byte {
+		samples := reader(b)
+		for _, filter := range filters {
+			samples = filter(samples)
+		}
+		return writer(samples)
 	}
 }

pkg/pcm/pcm_test.go

@@ -0,0 +1,79 @@
+package pcm
+
+import (
+	"encoding/hex"
+	"fmt"
+	"testing"
+
+	"github.com/AlexxIT/go2rtc/pkg/core"
+	"github.com/stretchr/testify/require"
+)
+
+func TestTranscode(t *testing.T) {
+	tests := []struct {
+		name   string
+		src    core.Codec
+		dst    core.Codec
+		source string
+		expect string
+	}{
+		{
+			name:   "s16be->s16be",
+			src:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			dst:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			source: "FCCA00130343062808130B510D9E0F7610DA111113EA15BD16F2168215D41561",
+			expect: "FCCA00130343062808130B510D9E0F7610DA111113EA15BD16F2168215D41561",
+		},
+		{
+			name:   "s16be->s16le",
+			src:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			dst:    core.Codec{Name: core.CodecPCML, ClockRate: 8000, Channels: 1},
+			source: "FCCA00130343062808130B510D9E0F7610DA111113EA15BD16F2168215D41561",
+			expect: "CAFC1300430328061308510B9E0D760FDA101111EA13BD15F2168216D4156115",
+		},
+		{
+			name:   "s16be->mulaw",
+			src:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			dst:    core.Codec{Name: core.CodecPCMU, ClockRate: 8000, Channels: 1},
+			source: "FCCA00130343062808130B510D9E0F7610DA111113EA15BD16F2168215D41561",
+			expect: "52FDD1C5BEB8B3B0AEAEABA9A8A8A9AA",
+		},
+		{
+			name:   "s16be->alaw",
+			src:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			dst:    core.Codec{Name: core.CodecPCMA, ClockRate: 8000, Channels: 1},
+			source: "FCCA00130343062808130B510D9E0F7610DA111113EA15BD16F2168215D41561",
+			expect: "7CD4FFED95939E9B8584868083838080",
+		},
+		{
+			name:   "2ch->1ch",
+			src:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 2},
+			dst:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			source: "FCCAFCCA001300130343034306280628081308130B510B510D9E0D9E0F760F76",
+			expect: "FCCA00130343062808130B510D9E0F76",
+		},
+		{
+			name:   "1ch->2ch",
+			src:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			dst:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 2},
+			source: "FCCA00130343062808130B510D9E0F76",
+			expect: "FCCAFCCA001300130343034306280628081308130B510B510D9E0D9E0F760F76",
+		},
+		{
+			name:   "16khz->8khz",
+			src:    core.Codec{Name: core.CodecPCM, ClockRate: 16000, Channels: 1},
+			dst:    core.Codec{Name: core.CodecPCM, ClockRate: 8000, Channels: 1},
+			source: "FCCAFCCA001300130343034306280628081308130B510B510D9E0D9E0F760F76",
+			expect: "FCCA00130343062808130B510D9E0F76",
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			f := Transcode(&test.dst, &test.src)
+			b, _ := hex.DecodeString(test.source)
+			b = f(b)
+			s := fmt.Sprintf("%X", b)
+			require.Equal(t, test.expect, s)
+		})
+	}
+}

pkg/wav/producer.go

@@ -45,7 +45,7 @@ func Open(r io.Reader) (*Producer, error) {
 				codec.Name = core.CodecPCMU
 			}
 
-			codec.Channels = uint16(data[2])
+			codec.Channels = data[2]
 			codec.ClockRate = binary.LittleEndian.Uint32(data[4:])
 		}
 	}

pkg/webrtc/consumer.go

@@ -73,7 +73,7 @@ func (c *Conn) AddTrack(media *core.Media, codec *core.Codec, track *core.Receiv
 				codec.Name = core.CodecPCMA
 			}
 			codec.ClockRate = 8000
-			sender.Handler = pcm.ResampleToG711(track.Codec, 8000, sender.Handler)
+			sender.Handler = pcm.TranscodeHandler(codec, track.Codec, sender.Handler)
 		}
 	}
 

pkg/webrtc/producer.go

@@ -21,7 +21,7 @@ func (c *Conn) GetTrack(media *core.Media, codec *core.Codec) (*core.Receiver, e
 			RTPCodecCapability: webrtc.RTPCodecCapability{
 				MimeType:  MimeType(codec),
 				ClockRate: codec.ClockRate,
-				Channels:  codec.Channels,
+				Channels:  uint16(codec.Channels),
 			},
 			PayloadType: 0, // don't know if this necessary
 		}

www/add.html

@@ -84,6 +84,18 @@
 </script>
 
 
+<button id="alsa">ALSA (Linux audio)</button>
+<div class="module">
+    <table id="alsa-table"></table>
+</div>
+<script>
+    document.getElementById('alsa').addEventListener('click', async ev => {
+        ev.target.nextElementSibling.style.display = 'block';
+        await getSources('alsa-table', 'api/alsa');
+    });
+</script>
+
+
 <button id="homekit">Apple HomeKit</button>
 <div class="module">
     <form id="homekit-pair" style="margin-bottom: 10px">
@@ -341,7 +353,7 @@
 </script>
 
 
-<button id="v4l2">V4L2 (USB)</button>
+<button id="v4l2">V4L2 (Linux video)</button>
 <div class="module">
     <table id="v4l2-table"></table>
 </div>