Hi guys,
Here is a first attempt at a RTP packetizer for the Xiph codecs.
This is still a work in progress; Vorbis likes to crash when rescaling
the timebase, and the Theora output only vaguely resembles what it
should. (There are some unrelated changes in there, particularly to
the depacketizer. I will separate those later.)
The only major architectural issue here is that the configuration
identifier (here hard-coded to 0xfedcba) needs to be carried across
from the SDP to the packetizer proper. Right now I don't believe there
is a clean way to do that without adding another field to
RTPMuxContext, a generic priv_data or something. The "best" way, IMO,
would be to structure RTP packetization in a way similar to how
depacketization is done -- through dynamic handlers.
We might be able to get away with a hard-coded configuration
identifier, but I don't know yet if my samples will periodically emit
in-band extradata. (The RFC provisions for this.) In that case, the
config ID needs to be refreshed, and kept across subsequent packets.
Josh
diff --git a/libavformat/Makefile b/libavformat/Makefile
index b140fe8..e150814 100644
--- a/libavformat/Makefile
+++ b/libavformat/Makefile
@@ -217,6 +217,7 @@ OBJS-$(CONFIG_RTP_MUXER) += rtp.o \
rtpenc_mpv.o \
rtpenc.o \
rtpenc_h264.o \
+ rtpenc_xiph.o \
avc.o
OBJS-$(CONFIG_RTSP_DEMUXER) += rtsp.o httpauth.o
OBJS-$(CONFIG_RTSP_MUXER) += rtsp.o rtspenc.o httpauth.o
diff --git a/libavformat/rtpdec_xiph.c b/libavformat/rtpdec_xiph.c
index 4c9cad2..3ecbb26 100644
--- a/libavformat/rtpdec_xiph.c
+++ b/libavformat/rtpdec_xiph.c
@@ -288,7 +288,17 @@ static int xiph_parse_fmtp_pair(AVStream* stream,
int result = 0;
if (!strcmp(attr, "sampling")) {
- return AVERROR_PATCHWELCOME;
+ if (!strcmp(value, "YCbCr-4:2:0")) {
+ codec->pix_fmt = PIX_FMT_YUV420P;
+ } else if (!strcmp(value, "YCbCr-4:4:2")) {
+ codec->pix_fmt = PIX_FMT_YUV422P;
+ } else if (!strcmp(value, "YCbCr-4:4:4")) {
+ codec->pix_fmt = PIX_FMT_YUV444P;
+ } else {
+ av_log(codec, AV_LOG_ERROR,
+ "Unsupported pixel format %s\n", attr);
+ return AVERROR_INVALIDDATA;
+ }
} else if (!strcmp(attr, "width")) {
/* This is an integer between 1 and 1048561
* and MUST be in multiples of 16. */
diff --git a/libavformat/rtpenc.c b/libavformat/rtpenc.c
index 4453f65..28be0bf 100644
--- a/libavformat/rtpenc.c
+++ b/libavformat/rtpenc.c
@@ -53,6 +53,8 @@ static int is_supported(enum CodecID id)
case CODEC_ID_MPEG2TS:
case CODEC_ID_AMR_NB:
case CODEC_ID_AMR_WB:
+ case CODEC_ID_VORBIS:
+ case CODEC_ID_THEORA:
return 1;
default:
return 0;
@@ -154,6 +156,11 @@ static int rtp_write_header(AVFormatContext *s1)
}
case CODEC_ID_AAC:
s->num_frames = 0;
+ case CODEC_ID_VORBIS:
+ case CODEC_ID_THEORA:
+ if (!s->max_frames_per_packet || s->max_frames_per_packet > 15)
+ s->max_frames_per_packet = 15;
+ s->max_payload_size -= 6; // ident+frag+tdt/vdt+pkt_num+pkt_length
default:
if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
av_set_pts_info(st, 32, 1, st->codec->sample_rate);
@@ -393,6 +400,10 @@ static int rtp_write_packet(AVFormatContext *s1, AVPacket *pkt)
case CODEC_ID_H263P:
ff_rtp_send_h263(s1, pkt->data, size);
break;
+ case CODEC_ID_VORBIS:
+ case CODEC_ID_THEORA:
+ ff_rtp_send_xiph(s1, pkt->data, size);
+ break;
default:
/* better than nothing : send the codec raw data */
rtp_send_raw(s1, pkt->data, size);
diff --git a/libavformat/rtpenc.h b/libavformat/rtpenc.h
index 95e70c1..d5d8b99 100644
--- a/libavformat/rtpenc.h
+++ b/libavformat/rtpenc.h
@@ -67,5 +67,6 @@ void ff_rtp_send_h263(AVFormatContext *s1, const uint8_t *buf1, int size);
void ff_rtp_send_aac(AVFormatContext *s1, const uint8_t *buff, int size);
void ff_rtp_send_amr(AVFormatContext *s1, const uint8_t *buff, int size);
void ff_rtp_send_mpegvideo(AVFormatContext *s1, const uint8_t *buf1, int size);
+void ff_rtp_send_xiph(AVFormatContext *s1, const uint8_t *buff, int size);
#endif /* AVFORMAT_RTPENC_H */
diff --git a/libavformat/rtsp.c b/libavformat/rtsp.c
index 9316f3c..6dc9c0d 100644
--- a/libavformat/rtsp.c
+++ b/libavformat/rtsp.c
@@ -1295,7 +1295,7 @@ static int rtsp_setup_output_streams(AVFormatContext *s, const char *addr)
rt->start_time = av_gettime();
/* Announce the stream */
- sdp = av_mallocz(8192);
+ sdp = av_mallocz(16384); // massive SDP buffer due to Xiph extradata
if (sdp == NULL)
return AVERROR(ENOMEM);
/* We create the SDP based on the RTSP AVFormatContext where we
@@ -1314,7 +1314,7 @@ static int rtsp_setup_output_streams(AVFormatContext *s, const char *addr)
ff_url_join(sdp_ctx.filename, sizeof(sdp_ctx.filename),
"rtsp", NULL, addr, -1, NULL);
ctx_array[0] = &sdp_ctx;
- if (avf_sdp_create(ctx_array, 1, sdp, 8192)) {
+ if (avf_sdp_create(ctx_array, 1, sdp, 16384)) {
av_free(sdp);
return AVERROR_INVALIDDATA;
}
diff --git a/libavformat/sdp.c b/libavformat/sdp.c
index b34b944..402eec1 100644
--- a/libavformat/sdp.c
+++ b/libavformat/sdp.c
@@ -21,6 +21,7 @@
#include <string.h>
#include "libavutil/avstring.h"
#include "libavutil/base64.h"
+#include "libavcodec/xiph.h"
#include "avformat.h"
#include "internal.h"
#include "avc.h"
@@ -220,6 +221,74 @@ static char *extradata2config(AVCodecContext *c)
return config;
}
+static char *xiph_extradata2config(AVCodecContext *c)
+{
+ char *config, *encoded_config;
+ uint8_t *header_start[3];
+ int headers_len, header_len[3], config_len;
+ int first_header_size;
+ uint8_t comment[] = {
+ ' ','h', 'i','r','o',' ','n','a','k','a','m','u','r','a'
+ };
+
+ switch (c->codec_id) {
+ case CODEC_ID_THEORA:
+ first_header_size = 42;
+ comment[0] = 0x81;
+ break;
+ case CODEC_ID_VORBIS:
+ first_header_size = 30;
+ comment[0] = 0x05;
+ break;
+ default:
+ av_log(c, AV_LOG_ERROR, "Unsupported Xiph codec ID\n");
+ return NULL;
+ }
+
+ if (ff_split_xiph_headers(c->extradata, c->extradata_size,
+ first_header_size, header_start,
+ header_len) < 0) {
+ av_log(c, AV_LOG_ERROR, "Extradata corrupt.");
+ return NULL;
+ }
+
+ headers_len = header_len[0]+sizeof(comment)+header_len[2];
+ config_len = 4 + // count
+ 3 + // ident
+ 2 + // packet size
+ 1 + // header count
+ 2 + // header size
+ headers_len; // and the rest
+ config = av_malloc(config_len);
+ encoded_config = av_malloc(AV_BASE64_SIZE(config_len));
+
+ if (!config || !encoded_config) {
+ av_log(c, AV_LOG_ERROR,
+ "Not enough memory for configuration string\n");
+ return NULL;
+ }
+
+ config[0] = config[1] = config[2] = 0;
+ config[3] = 1;
+ config[4] = 0xfe;
+ config[5] = 0xcd;
+ config[6] = 0xba;
+ config[7] = (headers_len >> 8) & 0xff;
+ config[8] = headers_len & 0xff;
+ config[9] = 2;
+ config[10] = header_len[0];
+ config[11] = sizeof(comment);
+ memcpy(config + 12, header_start[0], header_len[0]);
+ memcpy(config + 12 + header_len[0], comment, sizeof(comment));
+ memcpy(config + 12 + header_len[0] + sizeof(comment),
+ header_start[2], header_len[2]);
+
+ av_base64_encode(encoded_config, AV_BASE64_SIZE(config_len),
+ config, config_len);
+
+ return encoded_config;
+}
+
static char *sdp_write_media_attributes(char *buff, int size, AVCodecContext *c, int payload_type)
{
char *config = NULL;
@@ -297,6 +366,38 @@ static char *sdp_write_media_attributes(char *buff, int size, AVCodecContext *c,
payload_type, c->sample_rate, c->channels,
payload_type);
break;
+ case CODEC_ID_THEORA: {
+ const char *pix_fmt;
+ if (c->extradata_size)
+ config = xiph_extradata2config(c);
+ else
+ av_log(c, AV_LOG_ERROR, "Theora configuation info missing\n");
+ if (!config)
+ return NULL;
+
+ switch (c->pix_fmt) {
+ case PIX_FMT_YUV420P:
+ pix_fmt = "YCbCr-4:2:0";
+ break;
+ case PIX_FMT_YUV422P:
+ pix_fmt = "YCbCr-4:2:2";
+ break;
+ case PIX_FMT_YUV444P:
+ pix_fmt = "YCbCr-4:4:4";
+ break;
+ default:
+ av_log(c, AV_LOG_ERROR, "Unsupported pixel format.\n");
+ return NULL;
+ }
+
+ av_strlcatf(buff, size, "a=rtpmap:%d theora/9000\r\n"
+ "a=fmtp:%d delivery-method=inline;"
+ "width=%d; height=%d; sampling=%s;"
+ "configuration=%s;\r\n",
+ payload_type, payload_type,
+ c->width, c->height, pix_fmt, config);
+ break;
+ }
default:
/* Nothing special to do here... */
break;
_______________________________________________
FFmpeg-soc mailing list
[email protected]
https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-soc
