音视频入门基础：MPEG2-PS专题（5）——FFmpeg源码中，解析PS流中的PES流的实现

一、引言

从《音视频入门基础：MPEG2-PS专题（3）——MPEG2-PS格式简介》中可以知道，PS流由一个个pack（包装）组成。一个pack = 一个pack_header + 一个或多个PES_packet。pack_header中还可能存在system header。

但是pack_header和system header中并没有什么重要信息，所以FFmpeg源码在解析PS流时会跳过pack_header和system header，直接解析PES packet。

FFmpeg源码中通过mpegps_read_pes_header函数解析PS流中的PES packet。

二、mpegps_read_pes_header函数的定义

mpegps_read_pes_header函数定义在FFmpeg源码（本文演示用的FFmpeg源码版本为7.0.1）的源文件libavformat/mpeg.c中：

/* read the next PES header. Return its position in ppos* (if not NULL), and its start code, pts and dts.*/
static int mpegps_read_pes_header(AVFormatContext *s,int64_t *ppos, int *pstart_code,int64_t *ppts, int64_t *pdts)
{MpegDemuxContext *m = s->priv_data;int len, size, startcode, c, flags, header_len;int pes_ext, ext2_len, id_ext, skip;int64_t pts, dts;int64_t last_sync = avio_tell(s->pb);error_redo:avio_seek(s->pb, last_sync, SEEK_SET);
redo:/* next start code (should be immediately after) */m->header_state = 0xff;size      = MAX_SYNC_SIZE;startcode = find_next_start_code(s->pb, &size, &m->header_state);last_sync = avio_tell(s->pb);if (startcode < 0) {if (avio_feof(s->pb))return AVERROR_EOF;// FIXME we should remember header_statereturn FFERROR_REDO;}if (startcode == PACK_START_CODE)goto redo;if (startcode == SYSTEM_HEADER_START_CODE)goto redo;if (startcode == PADDING_STREAM) {avio_skip(s->pb, avio_rb16(s->pb));goto redo;}if (startcode == PRIVATE_STREAM_2) {if (!m->sofdec) {/* Need to detect whether this from a DVD or a 'Sofdec' stream */int len = avio_rb16(s->pb);int bytesread = 0;uint8_t *ps2buf = av_malloc(len);if (ps2buf) {bytesread = avio_read(s->pb, ps2buf, len);if (bytesread != len) {avio_skip(s->pb, len - bytesread);} else {uint8_t *p = 0;if (len >= 6)p = memchr(ps2buf, 'S', len - 5);if (p)m->sofdec = !memcmp(p+1, "ofdec", 5);m->sofdec -= !m->sofdec;if (m->sofdec < 0) {if (len == 980  && ps2buf[0] == 0) {/* PCI structure? */uint32_t startpts = AV_RB32(ps2buf + 0x0d);uint32_t endpts = AV_RB32(ps2buf + 0x11);uint8_t hours = ((ps2buf[0x19] >> 4) * 10) + (ps2buf[0x19] & 0x0f);uint8_t mins  = ((ps2buf[0x1a] >> 4) * 10) + (ps2buf[0x1a] & 0x0f);uint8_t secs  = ((ps2buf[0x1b] >> 4) * 10) + (ps2buf[0x1b] & 0x0f);m->dvd = (hours <= 23 &&mins  <= 59 &&secs  <= 59 &&(ps2buf[0x19] & 0x0f) < 10 &&(ps2buf[0x1a] & 0x0f) < 10 &&(ps2buf[0x1b] & 0x0f) < 10 &&endpts >= startpts);} else if (len == 1018 && ps2buf[0] == 1) {/* DSI structure? */uint8_t hours = ((ps2buf[0x1d] >> 4) * 10) + (ps2buf[0x1d] & 0x0f);uint8_t mins  = ((ps2buf[0x1e] >> 4) * 10) + (ps2buf[0x1e] & 0x0f);uint8_t secs  = ((ps2buf[0x1f] >> 4) * 10) + (ps2buf[0x1f] & 0x0f);m->dvd = (hours <= 23 &&mins  <= 59 &&secs  <= 59 &&(ps2buf[0x1d] & 0x0f) < 10 &&(ps2buf[0x1e] & 0x0f) < 10 &&(ps2buf[0x1f] & 0x0f) < 10);}}}av_free(ps2buf);/* If this isn't a DVD packet or no memory* could be allocated, just ignore it.* If we did, move back to the start of the* packet (plus 'length' field) */if (!m->dvd || avio_skip(s->pb, -(len + 2)) < 0) {/* Skip back failed.* This packet will be lost but that can't be helped* if we can't skip back*/goto redo;}} else {/* No memory */avio_skip(s->pb, len);goto redo;}} else if (!m->dvd) {int len = avio_rb16(s->pb);avio_skip(s->pb, len);goto redo;}}if (startcode == PROGRAM_STREAM_MAP) {mpegps_psm_parse(m, s->pb);goto redo;}/* find matching stream */if (!((startcode >= 0x1c0 && startcode <= 0x1df) ||(startcode >= 0x1e0 && startcode <= 0x1ef) ||(startcode == 0x1bd) ||(startcode == PRIVATE_STREAM_2) ||(startcode == 0x1fd)))goto redo;if (ppos) {*ppos = avio_tell(s->pb) - 4;}len = avio_rb16(s->pb);pts =dts = AV_NOPTS_VALUE;if (startcode != PRIVATE_STREAM_2){/* stuffing */for (;;) {if (len < 1)goto error_redo;c = avio_r8(s->pb);len--;/* XXX: for MPEG-1, should test only bit 7 */if (c != 0xff)break;}if ((c & 0xc0) == 0x40) {/* buffer scale & size */avio_r8(s->pb);c    = avio_r8(s->pb);len -= 2;}if ((c & 0xe0) == 0x20) {dts  =pts  = get_pts(s->pb, c);len -= 4;if (c & 0x10) {dts  = get_pts(s->pb, -1);len -= 5;}} else if ((c & 0xc0) == 0x80) {/* mpeg 2 PES */flags      = avio_r8(s->pb);header_len = avio_r8(s->pb);len       -= 2;if (header_len > len)goto error_redo;len -= header_len;if (flags & 0x80) {dts         = pts = get_pts(s->pb, -1);header_len -= 5;if (flags & 0x40) {dts         = get_pts(s->pb, -1);header_len -= 5;}}if (flags & 0x3f && header_len == 0) {flags &= 0xC0;av_log(s, AV_LOG_WARNING, "Further flags set but no bytes left\n");}if (flags & 0x01) { /* PES extension */pes_ext = avio_r8(s->pb);header_len--;/* Skip PES private data, program packet sequence counter* and P-STD buffer */skip  = (pes_ext >> 4) & 0xb;skip += skip & 0x9;if (pes_ext & 0x40 || skip > header_len) {av_log(s, AV_LOG_WARNING, "pes_ext %X is invalid\n", pes_ext);pes_ext = skip = 0;}avio_skip(s->pb, skip);header_len -= skip;if (pes_ext & 0x01) { /* PES extension 2 */ext2_len = avio_r8(s->pb);header_len--;if ((ext2_len & 0x7f) > 0) {id_ext = avio_r8(s->pb);if ((id_ext & 0x80) == 0)startcode = ((startcode & 0xff) << 8) | id_ext;header_len--;}}}if (header_len < 0)goto error_redo;avio_skip(s->pb, header_len);} else if (c != 0xf)goto redo;}if (startcode == PRIVATE_STREAM_1) {int ret = ffio_ensure_seekback(s->pb, 2);if (ret < 0)return ret;startcode = avio_r8(s->pb);m->raw_ac3 = 0;if (startcode == 0x0b) {if (avio_r8(s->pb) == 0x77) {startcode = 0x80;m->raw_ac3 = 1;avio_skip(s->pb, -2);} else {avio_skip(s->pb, -1);}} else {len--;}}if (len < 0)goto error_redo;if (dts != AV_NOPTS_VALUE && ppos) {int i;for (i = 0; i < s->nb_streams; i++) {if (startcode == s->streams[i]->id &&(s->pb->seekable & AVIO_SEEKABLE_NORMAL) /* index useless on streams anyway */) {ff_reduce_index(s, i);av_add_index_entry(s->streams[i], *ppos, dts, 0, 0,AVINDEX_KEYFRAME /* FIXME keyframe? */);}}}*pstart_code = startcode;*ppts        = pts;*pdts        = dts;return len;
}

该函数的作用就是：解析PS流中的一个PES packet，将其PES packet header里面的信息解析出来。

形参s：既是输入型参数也是输出型参数，指向一个AVFormatContext类型的变量。s->pb包含需要被解析的PS流的二进制数据。mpegps_read_pes_header函数解析的是s->pb->buf_ptr指向的PS流数据中的下一个PES packet。

形参ppos：输出型参数，*ppos为读取到的PES packet相对于文件首的偏移字节数。

形参pstart_code：输出型参数。如果读取到了PES packet，且该PES packet里面包含的不是的private_stream_1，*pstart_code为其PES packet header中的stream_id属性的值加0x100。

形参ppts：输出型参数。*ppts为从该PES packet的PES packet header中读取到的pts。

形参pdts：输出型参数。*pdts为从该PES packet的PES packet header中读取到的dts。

返回值：解析成功，返回该PES packet去掉PES packet header后的大小（即基本码流ES数据的大小）。解析失败，返回一个负数。

三、mpegps_read_pes_header函数的内部实现分析

mpegps_read_pes_header函数中，首先通过find_next_start_code函数找到s->pb->buf_ptr指向的PS流数据中的下一个pack header的起始码 或 下一个system header的起始码 或 下一个PES packet header的起始码：

redo:/* next start code (should be immediately after) */m->header_state = 0xff;size      = MAX_SYNC_SIZE;startcode = find_next_start_code(s->pb, &size, &m->header_state);last_sync = avio_tell(s->pb);if (startcode < 0) {if (avio_feof(s->pb))return AVERROR_EOF;// FIXME we should remember header_statereturn FFERROR_REDO;}

如果找到的是pack header的起始码 或 system header的起始码 或 找到的PES packet中包含padding_stream数据，通过goto语句跳转，然后重新通过find_next_start_code函数查找下一个起始码：

    if (startcode == PACK_START_CODE)goto redo;if (startcode == SYSTEM_HEADER_START_CODE)goto redo;if (startcode == PADDING_STREAM) {avio_skip(s->pb, avio_rb16(s->pb));goto redo;}

如果找到了符合要求的PES packet header的起始码，通过avio_tell函数读取该PES packet相对于文件首的偏移字节数，赋值给*ppos。关于avio_tell函数的用法可以参考：《FFmpeg源码：avio_tell函数分析》：

    /* find matching stream */if (!((startcode >= 0x1c0 && startcode <= 0x1df) ||(startcode >= 0x1e0 && startcode <= 0x1ef) ||(startcode == 0x1bd) ||(startcode == PRIVATE_STREAM_2) ||(startcode == 0x1fd)))goto redo;if (ppos) {*ppos = avio_tell(s->pb) - 4;}

读取PES packet header中的PES_packet_length属性，赋值给变量len。关于avio_rb16函数的用法可以参考：《FFmpeg源码：avio_r8、avio_rl16、avio_rl24、avio_rl32、avio_rl64函数分析》：

    len = avio_rb16(s->pb);

读取PES packet header中的PES_scrambling_control、PES_priority、data_alignment_indicator、copyright、original_or_copy属性，赋值给变量c：

        c = avio_r8(s->pb);

如果上述读取到的PES_packet_length属性后面的值为"10（(c & 0xc0) == 0x80为真），表示读取到的PES packet header的格式正确：

 执行大括号内的内容：

else if ((c & 0xc0) == 0x80) {
//...
}

读取PTS_DTS_flags、ESCR_flag、ES_rate_flag、DSM_trick_mode_flag、additional_copy_info_flag、PES_CRC_flag、PES_extension_flag这7个属性，赋值给变量flags：

        flags      = avio_r8(s->pb);

读取PES_header_data_length属性，赋值给变量header_len：

        header_len = avio_r8(s->pb);

如果PTS_DTS_flags属性的值为'10'，表示PES packet header中会存在PTS，读取PTS；值为'11'时，表示PES packet header中会同时存在PTS和DTS，读取PTS和DTS，分别赋值给变量dts和pts：

        if (flags & 0x80) {dts         = pts = get_pts(s->pb, -1);header_len -= 5;if (flags & 0x40) {dts         = get_pts(s->pb, -1);header_len -= 5;}}

如果PES_extension_flag属性的值为1，表示PES packet header有PES_extension域，读取PES_extension域：

        if (flags & 0x01) { /* PES extension */pes_ext = avio_r8(s->pb);header_len--;/* Skip PES private data, program packet sequence counter* and P-STD buffer */skip  = (pes_ext >> 4) & 0xb;skip += skip & 0x9;if (pes_ext & 0x40 || skip > header_len) {av_log(s, AV_LOG_WARNING, "pes_ext %X is invalid\n", pes_ext);pes_ext = skip = 0;}avio_skip(s->pb, skip);header_len -= skip;if (pes_ext & 0x01) { /* PES extension 2 */ext2_len = avio_r8(s->pb);header_len--;if ((ext2_len & 0x7f) > 0) {id_ext = avio_r8(s->pb);if ((id_ext & 0x80) == 0)startcode = ((startcode & 0xff) << 8) | id_ext;header_len--;}}}

最后返回从该PES packet的PES packet header中读取到的pts、dts、该PES packet去掉PES packet header后的大小等信息：

    *pstart_code = startcode;*ppts        = pts;*pdts        = dts;return len;

四、FFmpeg源码中，解析TS流中的PES流的实现

FFmpeg源码中，解析TS流中的PES流所使用的函数不一样，是通过mpegts_push_data函数进行解析的，具体可以参考：《音视频入门基础：MPEG2-TS专题（19）——FFmpeg源码中，解析TS流中的PES流的实现》。