/*! \internal \brief Create a test frame
* \param timestamp the time in ms of the frame
* \param seqno the frame's sequence number
* \returns a malloc'd frame
*/
static struct ast_frame *create_test_frame(long timestamp,
int seqno)
{
struct ast_frame f = {0};
f.subclass.format.id = AST_FORMAT_SLINEAR;
f.frametype = AST_FRAME_VOICE;
f.src = "TEST";
f.ts = timestamp;
f.len = DEFAULT_FRAME_MS;
f.seqno = seqno;
return ast_frisolate(&f);
}
static struct ast_frame *jpeg_read_image(int fd, int len)
{
struct ast_frame fr;
int res;
char buf[65536];
if (len > sizeof(buf) || len < 0) {
ast_log(LOG_WARNING, "JPEG image too large to read\n");
return NULL;
}
res = read(fd, buf, len);
if (res < len) {
ast_log(LOG_WARNING, "Only read %d of %d bytes: %s\n", res, len, strerror(errno));
}
memset(&fr, 0, sizeof(fr));
fr.frametype = AST_FRAME_IMAGE;
fr.subclass.format = ast_format_jpeg;
fr.data.ptr = buf;
fr.src = "JPEG Read";
fr.datalen = len;
return ast_frisolate(&fr);
}
/* And decode everything we can in the buffer */
static struct ast_frame *lintosilk_frameout(struct ast_trans_pvt *pvt)
{
struct silk_coder_pvt *coder = pvt->pvt;
SKP_int ret = 0;
SKP_int16 nBytesOut = 0;
int datalen = 0;
int samples = 0;
int numPackets = 0;
struct ast_frame *f;
/* we can only work on multiples of a 10 ms sample
* and no more than encControl->packetSize.
* So we shove in packetSize samples repeatedly until
* we are out */
/* We only do stuff if we have more than packetSize */
if (pvt->samples < coder->encControl.packetSize) {
return NULL;
}
while (pvt->samples >= coder->encControl.packetSize) {
nBytesOut = SILK_BUFFER_SIZE_BYTES - datalen;
ret = SKP_Silk_SDK_Encode(coder->psEnc,
&coder->encControl,
(SKP_int16*)(coder->buf + samples),
coder->encControl.packetSize,
(SKP_uint8*)(pvt->outbuf.ui8 + datalen),
&nBytesOut);
if (ret) {
ast_log(LOG_WARNING, "Silk_Encode returned %d\n", ret);
}
/* nBytesOut now holds the number of bytes encoded */
datalen += nBytesOut;
samples += coder->encControl.packetSize;
pvt->samples -= coder->encControl.packetSize;
if(nBytesOut > 0){
/* if stuff came out, we have encoded a packet */
numPackets++;
}
}
/* Move the remaining buffer stuff down */
if (pvt->samples) {
memmove(coder->buf, coder->buf + samples, pvt->samples *2);
}
if(datalen == 0){
/* we shoved a bunch of samples in, but got no packets
* out. We return NULL to the caller, like
* if we could not encode anything */
return NULL;
}
/* we build the frame ourselves because we have an explicit dest */
f = &pvt->f;
f->samples = coder->encControl.packetSize * numPackets;
f->datalen = datalen;
f->frametype = AST_FRAME_VOICE;
ast_format_copy(&f->subclass.format, &pvt->explicit_dst);
f->mallocd = 0;
f->offset = AST_FRIENDLY_OFFSET;
f->src = pvt->t->name;
f->data.ptr = pvt->outbuf.c;
return ast_frisolate(f);
}
/*! \brief convert work buffer and produce output frame */
static struct ast_frame *lintospeex_frameout(struct ast_trans_pvt *pvt)
{
struct speex_coder_pvt *tmp = pvt->pvt;
int is_speech=1;
int datalen = 0; /* output bytes */
int samples = 0; /* output samples */
/* We can't work on anything less than a frame in size */
if (pvt->samples < tmp->framesize)
return NULL;
speex_bits_reset(&tmp->bits);
while (pvt->samples >= tmp->framesize) {
#ifdef _SPEEX_TYPES_H
/* Preprocess audio */
if (preproc)
is_speech = speex_preprocess(tmp->pp, tmp->buf + samples, NULL);
/* Encode a frame of data */
if (is_speech) {
/* If DTX enabled speex_encode returns 0 during silence */
is_speech = speex_encode_int(tmp->speex, tmp->buf + samples, &tmp->bits) || !dtx;
} else {
/* 5 zeros interpreted by Speex as silence (submode 0) */
speex_bits_pack(&tmp->bits, 0, 5);
}
#else
{
float fbuf[1024];
int x;
/* Convert to floating point */
for (x = 0; x < tmp->framesize; x++)
fbuf[x] = tmp->buf[samples + x];
/* Encode a frame of data */
is_speech = speex_encode(tmp->speex, fbuf, &tmp->bits) || !dtx;
}
#endif
samples += tmp->framesize;
pvt->samples -= tmp->framesize;
}
/* Move the data at the end of the buffer to the front */
if (pvt->samples)
memmove(tmp->buf, tmp->buf + samples, pvt->samples * 2);
/* Use AST_FRAME_CNG to signify the start of any silence period */
if (is_speech) {
tmp->silent_state = 0;
} else {
if (tmp->silent_state) {
return NULL;
} else {
struct ast_frame frm = {
.frametype = AST_FRAME_CNG,
.src = pvt->t->name,
};
/*
* XXX I don't think the AST_FRAME_CNG code has ever
* really worked for speex. There doesn't seem to be
* any consumers of the frame type. Everyone that
* references the type seems to pass the frame on.
*/
tmp->silent_state = 1;
/* XXX what now ? format etc... */
return ast_frisolate(&frm);
}
}
/* Terminate bit stream */
speex_bits_pack(&tmp->bits, 15, 5);
datalen = speex_bits_write(&tmp->bits, pvt->outbuf.c, pvt->t->buf_size);
return ast_trans_frameout(pvt, datalen, samples);
}
static void speextolin_destroy(struct ast_trans_pvt *arg)
{
struct speex_coder_pvt *pvt = arg->pvt;
speex_decoder_destroy(pvt->speex);
speex_bits_destroy(&pvt->bits);
}
static void lintospeex_destroy(struct ast_trans_pvt *arg)
{
struct speex_coder_pvt *pvt = arg->pvt;
#ifdef _SPEEX_TYPES_H
if (preproc)
speex_preprocess_state_destroy(pvt->pp);
#endif
speex_encoder_destroy(pvt->speex);
speex_bits_destroy(&pvt->bits);
}
static struct ast_translator speextolin = {
.name = "speextolin",
.src_codec = {
.name = "speex",
.type = AST_MEDIA_TYPE_AUDIO,
.sample_rate = 8000,
},
.dst_codec = {
.name = "slin",
.type = AST_MEDIA_TYPE_AUDIO,
.sample_rate = 8000,
},
.format = "slin",
static struct ast_frame *hook_event_cb(struct ast_channel *chan, struct ast_frame *frame, enum ast_framehook_event event, void *data)
{
struct jb_framedata *framedata = data;
struct timeval now_tv;
unsigned long now;
int putframe = 0; /* signifies if audio frame was placed into the buffer or not */
switch (event) {
case AST_FRAMEHOOK_EVENT_READ:
break;
case AST_FRAMEHOOK_EVENT_ATTACHED:
case AST_FRAMEHOOK_EVENT_DETACHED:
case AST_FRAMEHOOK_EVENT_WRITE:
return frame;
}
if (ast_channel_fdno(chan) == AST_JITTERBUFFER_FD && framedata->timer) {
if (ast_timer_ack(framedata->timer, 1) < 0) {
ast_log(LOG_ERROR, "Failed to acknowledge timer in jitter buffer\n");
return frame;
}
}
if (!frame) {
return frame;
}
now_tv = ast_tvnow();
now = ast_tvdiff_ms(now_tv, framedata->start_tv);
if (frame->frametype == AST_FRAME_VOICE) {
int res;
struct ast_frame *jbframe;
if (!ast_test_flag(frame, AST_FRFLAG_HAS_TIMING_INFO) || frame->len < 2 || frame->ts < 0) {
/* only frames with timing info can enter the jitterbuffer */
return frame;
}
jbframe = ast_frisolate(frame);
ao2_replace(framedata->last_format, frame->subclass.format);
if (frame->len && (frame->len != framedata->timer_interval)) {
framedata->timer_interval = frame->len;
ast_timer_set_rate(framedata->timer, 1000 / framedata->timer_interval);
}
if (!framedata->first) {
framedata->first = 1;
res = framedata->jb_impl->put_first(framedata->jb_obj, jbframe, now);
} else {
res = framedata->jb_impl->put(framedata->jb_obj, jbframe, now);
}
if (res == AST_JB_IMPL_OK) {
if (jbframe != frame) {
ast_frfree(frame);
}
frame = &ast_null_frame;
} else if (jbframe != frame) {
ast_frfree(jbframe);
}
putframe = 1;
}
if (frame->frametype == AST_FRAME_NULL) {
int res;
long next = framedata->jb_impl->next(framedata->jb_obj);
/* If now is earlier than the next expected output frame
* from the jitterbuffer we may choose to pass on retrieving
* a frame during this read iteration. The only exception
* to this rule is when an audio frame is placed into the buffer
* and the time for the next frame to come out of the buffer is
* at least within the timer_interval of the next output frame. By
* doing this we are able to feed off the timing of the input frames
* and only rely on our jitterbuffer timer when frames are dropped.
* During testing, this hybrid form of timing gave more reliable results. */
if (now < next) {
long int diff = next - now;
if (!putframe) {
return frame;
} else if (diff >= framedata->timer_interval) {
return frame;
}
}
ast_frfree(frame);
frame = &ast_null_frame;
res = framedata->jb_impl->get(framedata->jb_obj, &frame, now, framedata->timer_interval);
switch (res) {
case AST_JB_IMPL_OK:
/* got it, and pass it through */
break;
case AST_JB_IMPL_DROP:
ast_frfree(frame);
frame = &ast_null_frame;
break;
case AST_JB_IMPL_INTERP:
if (framedata->last_format) {
struct ast_frame tmp = { 0, };
tmp.frametype = AST_FRAME_VOICE;
tmp.subclass.format = framedata->last_format;
/* example: 8000hz / (1000 / 20ms) = 160 samples */
tmp.samples = ast_format_get_sample_rate(framedata->last_format) / (1000 / framedata->timer_interval);
tmp.delivery = ast_tvadd(framedata->start_tv, ast_samp2tv(next, 1000));
tmp.offset = AST_FRIENDLY_OFFSET;
tmp.src = "func_jitterbuffer interpolation";
ast_frfree(frame);
frame = ast_frdup(&tmp);
break;
}
/* else fall through */
case AST_JB_IMPL_NOFRAME:
ast_frfree(frame);
frame = &ast_null_frame;
break;
}
}
if (frame->frametype == AST_FRAME_CONTROL) {
switch(frame->subclass.integer) {
case AST_CONTROL_HOLD:
case AST_CONTROL_UNHOLD:
case AST_CONTROL_T38_PARAMETERS:
case AST_CONTROL_SRCUPDATE:
case AST_CONTROL_SRCCHANGE:
framedata->jb_impl->force_resync(framedata->jb_obj);
break;
default:
break;
}
}
return frame;
}
