Hi,
On 27 July 2010 03:03, Martin Storsjö <[email protected]> wrote:
> On Mon, 26 Jul 2010, Josh Allmann wrote:
>
>> This version works better, but is not yet complete.
>>
>> -The scaffolding for Vorbis is mostly up, but I have to set its SDP properly.
Done.
>> -Theora video has problems with I-frames (or whatever the Theora
>> equivalent is), and exhibits serious blocking in areas of motion.
Fixed, although on occassion it is a bit wonky. The depacketizer will
complain about a missing start fragment at a particular position in
big buck bunny, disrupting the video briefly, but otherwise it is
fine.
>> I hope this is due to some invalid reads that Valgrind complains about.
Valgrind fixed.
>> -Packing multiple frames in a single packet is another TODO.
>
Not yet complete, but is not critical for proper operation, either.
> A few comments:
>
> + /* set xiph data type */
> + switch (*buff) {
> + case 0x01: // vorbis id
> + case 0x05: // vorbis setup
> + case 0x80: // theora header
> + case 0x82: // theora tables
> + xdt = 1; // packed config payload
> + case 0x03: // vorbis comments
> + case 0x81: // theora comments
> + xdt = 2; // comment payload
> + default:
> + xdt = 0; // raw data payload
> + }
>
> I guess you want break statements in the switch, too...
>
Fixed.
> + /* set ident
> + * Probably need a non-fixed way of generating
> + * this, but it has to be done in SDP and passed in from there. */
> + q = s->buf;
> + *q++ = 0xfe;
> + *q++ = 0xcd;
> + *q++ = 0xba;
>
> I haven't read the specs, but what's the role of this ident code? Is there
> any harm in having it hardcoded to a specific value? Is it set in the
> original stream data somewhere, so that you'd have to parse out the
> correct value from there? Or is it only used to distinguish streams if you
> have more than one vorbis/theora stream in the same presentation? In that
> case, you could use e.g. one hardcoded value for vorbis and another for
> theora - that would probably be enough for some time at lesat.
As Luca said, it is only used to make sure the extradata doesn't
change mid-stream. Different streams can have the same ident, eg a
Vorbis and a Theora can share 0xfecdba. Our depacketizer doesn't
handle changing the ident anyway.
Revised patch attached. I also had to enlarge the outgoing RTSP buffer
to handle the SDP extradata.
Josh
diff --git a/libavformat/Makefile b/libavformat/Makefile
index c32a254..dbb326c 100644
--- a/libavformat/Makefile
+++ b/libavformat/Makefile
@@ -219,6 +219,7 @@ 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/rtpenc.c b/libavformat/rtpenc.c
index 4453f65..c940643 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 = av_clip(s->max_frames_per_packet, 1, 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/rtpenc_xiph.c b/libavformat/rtpenc_xiph.c
new file mode 100644
index 0000000..d01ecdd
--- /dev/null
+++ b/libavformat/rtpenc_xiph.c
@@ -0,0 +1,91 @@
+/*
+ * RTP packetization for Xiph audio and video
+ * Copyright (c) 2010 Josh Allmann
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "avformat.h"
+#include "rtpenc.h"
+
+/**
+ * Packetize Xiph frames into RTP according to
+ * RFC 5215 (Vorbis) and the Theora RFC draft.
+ * (http://svn.xiph.org/trunk/theora/doc/draft-ietf-avt-rtp-theora-00.txt)
+ */
+void ff_rtp_send_xiph(AVFormatContext *s1, const uint8_t *buff, int size)
+{
+ RTPMuxContext *s = s1->priv_data;
+ int max_pkt_size, xdt, frag;
+ uint8_t *q;
+
+ max_pkt_size = s->max_payload_size;
+
+ /* set xiph data type */
+ switch (*buff) {
+ case 0x01: // vorbis id
+ case 0x05: // vorbis setup
+ case 0x80: // theora header
+ case 0x82: // theora tables
+ xdt = 1; // packed config payload
+ break;
+ case 0x03: // vorbis comments
+ case 0x81: // theora comments
+ xdt = 2; // comment payload
+ break;
+ default:
+ xdt = 0; // raw data payload
+ }
+
+ /* set ident
+ * Probably need a non-fixed way of generating
+ * this, but it has to be done in SDP and passed in from there. */
+ q = s->buf;
+ *q++ = 0xfe;
+ *q++ = 0xcd;
+ *q++ = 0xba;
+ s->buf_ptr = q;
+
+ /* set fragment
+ * 0 - whole frame (possibly multiple frames)
+ * 1 - first fragment
+ * 2 - fragment continuation
+ * 3 - last fragmement */
+ frag = size <= max_pkt_size ? 0 : 1;
+
+ /* TODO use s->buf_ptr, mark position for later in order to
+ * transmit multiple frames in one RTP packet. To do this,
+ * need to avoid adding in ident, frag, xdt twice */
+ while (size > 0) {
+ int len = (!frag || frag == 3) ? size : max_pkt_size;
+ int num_pkts = frag ? 0 : 1; // XXX set properly for >1 frame/pkt
+ q = s->buf_ptr;
+
+ /* set packet headers */
+ *q++ = (frag << 6) | (xdt << 4) | num_pkts;
+ *q++ = (len >> 8) & 0xff;
+ *q++ = len & 0xff;
+ /* set packet body */
+ memmove(q, buff, len);
+ q += len;
+ buff += len;
+
+ ff_rtp_send_data(s1, s->buf, q - s->buf, 0); // marker bit unused
+ size -= len;
+ frag = size <= max_pkt_size ? 3 : 2;
+ }
+}
diff --git a/libavformat/rtsp.c b/libavformat/rtsp.c
index 7dede4c..5596ae5 100644
--- a/libavformat/rtsp.c
+++ b/libavformat/rtsp.c
@@ -872,7 +872,7 @@ int ff_rtsp_send_cmd_with_content_async(AVFormatContext *s,
int send_content_length)
{
RTSPState *rt = s->priv_data;
- char buf[4096], *out_buf;
+ char buf[16384], *out_buf; // large buffer to accommodate xiph sdp
char base64buf[AV_BASE64_SIZE(sizeof(buf))];
/* Add in RTSP headers */
@@ -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..870cd90 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 +221,68 @@ 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;
+
+ switch (c->codec_id) {
+ case CODEC_ID_THEORA:
+ first_header_size = 42;
+ break;
+ case CODEC_ID_VORBIS:
+ first_header_size = 30;
+ 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]+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] = 0; // size of comment header; nonexistent
+ memcpy(config + 12, header_start[0], header_len[0]);
+ memcpy(config + 12 + header_len[0], header_start[2], header_len[2]);
+
+ av_base64_encode(encoded_config, AV_BASE64_SIZE(config_len),
+ config, config_len);
+ av_free(config);
+
+ return encoded_config;
+}
+
static char *sdp_write_media_attributes(char *buff, int size, AVCodecContext *c, int payload_type)
{
char *config = NULL;
@@ -297,6 +360,51 @@ 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_VORBIS:
+ if (c->extradata_size)
+ config = xiph_extradata2config(c);
+ else
+ av_log(c, AV_LOG_ERROR, "Vorbis configuration info missing\n");
+ if (!config)
+ return NULL;
+
+ av_strlcatf(buff, size, "a=rtpmap:%d vorbis/%d/%d\r\n"
+ "a=fmtp:%d configuration=%s\r\n",
+ payload_type, c->sample_rate, c->channels,
+ payload_type, config);
+ 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
