从头用脚分析FFmpeg源码 - av_read_frame

av_read_frame作用

/**
 * Return the next frame of a stream.
 * This function returns what is stored in the file, and does not validate
 * that what is there are valid frames for the decoder. It will split what is
 * stored in the file into frames and return one for each call. It will not
 * omit invalid data between valid frames so as to give the decoder the maximum
 * information possible for decoding.
 *
 * On success, the returned packet is reference-counted (pkt->buf is set) and
 * valid indefinitely. The packet must be freed with av_packet_unref() when
 * it is no longer needed. For video, the packet contains exactly one frame.
 * For audio, it contains an integer number of frames if each frame has
 * a known fixed size (e.g. PCM or ADPCM data). If the audio frames have
 * a variable size (e.g. MPEG audio), then it contains one frame.
 *
 * pkt->pts, pkt->dts and pkt->duration are always set to correct
 * values in AVStream.time_base units (and guessed if the format cannot
 * provide them). pkt->pts can be AV_NOPTS_VALUE if the video format
 * has B-frames, so it is better to rely on pkt->dts if you do not
 * decompress the payload.
 *
 * @return 0 if OK, < 0 on error or end of file. On error, pkt will be blank
 *         (as if it came from av_packet_alloc()).
 *
 * @note pkt will be initialized, so it may be uninitialized, but it must not
 *       contain data that needs to be freed.
 */
int av_read_frame(AVFormatContext *s, AVPacket *pkt);

上面说的很复杂，简单的来讲，就是从AVFormatContext读取一个AVPacket。

再看看雷神的补充:

通过av_read_packet()，读取一个包，需要说明的是此函数必须是包含整数帧的，不存在半帧的情况，以ts流为例，是读取一个完整的PES包（一个完整pes包包含若干视频或音频es包），读取完毕后，通过av_parser_parse2()分析出视频一帧（或音频若干帧），返回，下次进入循环的时候，如果上次的数据没有完全取完，则st = s->cur_st;不会是NULL，即再此进入av_parser_parse2()流程，而不是下面的av_read_packet（）流程，这样就保证了，如果读取一次包含了N帧视频数据（以视频为例），则调用av_read_frame（**）N次都不会去读数据，而是返回第一次读取的数据，直到全部解析完毕。

av_read_frame源码

读取一个AVPacket，如果buffer里面存在，就从buffer中读，没有就调用read_frame_internal函数，从流中读。

int av_read_frame(AVFormatContext *s, AVPacket *pkt)
{
	// AVFMT_FLAG_GENPTS是指在解封装阶段就要对packet设置pts。FFmpeg默认是不会设置的，
	// FFplay在使用的时候，可以通过genpts参数设置这个flag。一般的容器都会在解封装阶段自动设置
	// pts，这个标识其实是针对在容器中没有时间戳概念的结构。（比如h264裸流？）
    const int genpts = s->flags & AVFMT_FLAG_GENPTS;
    int eof = 0;
    int ret;
    AVStream *st;

	// 一般来说，都会走这里，因为没设置AVFMT_FLAG_GENPTS标志
    if (!genpts) {
    	// 有buffer读buffer，没buffer就重新读。
    	// avformat_find_stream_info中读取的AVPacket就会存放在buffer中。
    	// read_frame_internal 才是真正从流中读取AVPacket的方法。
        ret = s->internal->packet_buffer
              ? avpriv_packet_list_get(&s->internal->packet_buffer,
                                        &s->internal->packet_buffer_end, pkt)
              : read_frame_internal(s, pkt);
        if (ret < 0)
            return ret;
        goto return_packet;
    }

	// 这里重新设置pts
    for (;;) {
        PacketList *pktl = s->internal->packet_buffer;

        if (pktl) {
            AVPacket *next_pkt = &pktl->pkt;

			// dts不能为null，pts需要从dts中推算出来
            if (next_pkt->dts != AV_NOPTS_VALUE) {
                int wrap_bits = s->streams[next_pkt->stream_index]->pts_wrap_bits;
                // last dts seen for this stream. if any of packets following
                // current one had no dts, we will set this to AV_NOPTS_VALUE.
                int64_t last_dts = next_pkt->dts;
                av_assert2(wrap_bits <= 64);
                while (pktl && next_pkt->pts == AV_NOPTS_VALUE) {
                	// 根据连续两个AVPacket之间的dts，设置pts
                    if (pktl->pkt.stream_index == next_pkt->stream_index &&
                        av_compare_mod(next_pkt->dts, pktl->pkt.dts, 2ULL << (wrap_bits - 1)) < 0) {
                        if (av_compare_mod(pktl->pkt.pts, pktl->pkt.dts, 2ULL << (wrap_bits - 1))) {
                            // not B-frame
                            next_pkt->pts = pktl->pkt.dts;
                        }
                        if (last_dts != AV_NOPTS_VALUE) {
                            // Once last dts was set to AV_NOPTS_VALUE, we don't change it.
                            last_dts = pktl->pkt.dts;
                        }
                    }
                    pktl = pktl->next;
                }
                // 尾部加上duration
                if (eof && next_pkt->pts == AV_NOPTS_VALUE && last_dts != AV_NOPTS_VALUE) {
                    // Fixing the last reference frame had none pts issue (For MXF etc).
                    // We only do this when
                    // 1. eof.
                    // 2. we are not able to resolve a pts value for current packet.
                    // 3. the packets for this stream at the end of the files had valid dts.
                    next_pkt->pts = last_dts + next_pkt->duration;
                }
                pktl = s->internal->packet_buffer;
            }

            /* read packet from packet buffer, if there is data */
            st = s->streams[next_pkt->stream_index];
            // pts设置成功，并非是AV_NOPTS_VALUE，这个时候就应用这个AVPacket
            if (!(next_pkt->pts == AV_NOPTS_VALUE && st->discard < AVDISCARD_ALL &&
                  next_pkt->dts != AV_NOPTS_VALUE && !eof)) {
                ret = avpriv_packet_list_get(&s->internal->packet_buffer,
                                               &s->internal->packet_buffer_end, pkt);
                goto return_packet;
            }
        }

		// buffer中没有，就继续读
        ret = read_frame_internal(s, pkt);
        if (ret < 0) {
            if (pktl && ret != AVERROR(EAGAIN)) {
                eof = 1;
                continue;
            } else
                return ret;
        }

		//放到buffer中，待循环的时候使用
        ret = avpriv_packet_list_put(&s->internal->packet_buffer,
                                 &s->internal->packet_buffer_end,
                                 pkt, NULL, 0);
        if (ret < 0) {
            av_packet_unref(pkt);
            return ret;
        }
    }

return_packet:

    st = s->streams[pkt->stream_index];
    if ((s->iformat->flags & AVFMT_GENERIC_INDEX) && pkt->flags & AV_PKT_FLAG_KEY) {
        ff_reduce_index(s, st->index);
        av_add_index_entry(st, pkt->pos, pkt->dts, 0, 0, AVINDEX_KEYFRAME);
    }

    if (is_relative(pkt->dts))
        pkt->dts -= RELATIVE_TS_BASE;
    if (is_relative(pkt->pts))
        pkt->pts -= RELATIVE_TS_BASE;

    return ret;
}

read_frame_internal源码

读取一个AVPacket，读取完成之后，看是否需要parse，如果需要就parse一下，然后把parse的结果重新写入AVPacket中。

static int read_frame_internal(AVFormatContext *s, AVPacket *pkt)
{
    int ret, i, got_packet = 0;
    AVDictionary *metadata = NULL;

	// packet也有可能会放到s->internal->parse_queue中
    while (!got_packet && !s->internal->parse_queue) {
        AVStream *st;

        /* read next packet */
        // 读取一个packet，这是最主要的函数
        ret = ff_read_packet(s, pkt);
        if (ret < 0) {
            if (ret == AVERROR(EAGAIN))
                return ret;
            /* flush the parsers */
            // 如果需要parser，就parser，parser后，AVPacket会放入s->internal->parse_queue中
            for (i = 0; i < s->nb_streams; i++) {
                st = s->streams[i];
                if (st->parser && st->need_parsing)
                    parse_packet(s, pkt, st->index, 1);
            }
            /* all remaining packets are now in parse_queue =>
             * really terminate parsing */
            break;
        }
        ret = 0;
        st  = s->streams[pkt->stream_index];

        st->event_flags |= AVSTREAM_EVENT_FLAG_NEW_PACKETS;

        /* update context if required */
        //更新stream中的AVCodecContext信息，在hls，webm等格式中可能会用到
        if (st->internal->need_context_update) {
            if (avcodec_is_open(st->internal->avctx)) {
                av_log(s, AV_LOG_DEBUG, "Demuxer context update while decoder is open, closing and trying to re-open\n");
                avcodec_close(st->internal->avctx);
                st->internal->info->found_decoder = 0;
            }

            /* close parser, because it depends on the codec */
            if (st->parser && st->internal->avctx->codec_id != st->codecpar->codec_id) {
                av_parser_close(st->parser);
                st->parser = NULL;
            }

            ret = avcodec_parameters_to_context(st->internal->avctx, st->codecpar);
            if (ret < 0) {
                av_packet_unref(pkt);
                return ret;
            }

#if FF_API_LAVF_AVCTX
FF_DISABLE_DEPRECATION_WARNINGS
            /* update deprecated public codec context */
            ret = avcodec_parameters_to_context(st->codec, st->codecpar);
            if (ret < 0) {
                av_packet_unref(pkt);
                return ret;
            }
FF_ENABLE_DEPRECATION_WARNINGS
#endif

            st->internal->need_context_update = 0;
        }

		// 判断pts与输出
        if (pkt->pts != AV_NOPTS_VALUE &&
            pkt->dts != AV_NOPTS_VALUE &&
            pkt->pts < pkt->dts) {
            av_log(s, AV_LOG_WARNING,
                   "Invalid timestamps stream=%d, pts=%s, dts=%s, size=%d\n",
                   pkt->stream_index,
                   av_ts2str(pkt->pts),
                   av_ts2str(pkt->dts),
                   pkt->size);
        }
        if (s->debug & FF_FDEBUG_TS)
            av_log(s, AV_LOG_DEBUG,
                   "ff_read_packet stream=%d, pts=%s, dts=%s, size=%d, duration=%"PRId64", flags=%d\n",
                   pkt->stream_index,
                   av_ts2str(pkt->pts),
                   av_ts2str(pkt->dts),
                   pkt->size, pkt->duration, pkt->flags);

        if (st->need_parsing && !st->parser && !(s->flags & AVFMT_FLAG_NOPARSE)) {
            st->parser = av_parser_init(st->codecpar->codec_id);
            if (!st->parser) {
                av_log(s, AV_LOG_VERBOSE, "parser not found for codec "
                       "%s, packets or times may be invalid.\n",
                       avcodec_get_name(st->codecpar->codec_id));
                /* no parser available: just output the raw packets */
                st->need_parsing = AVSTREAM_PARSE_NONE;
            } else if (st->need_parsing == AVSTREAM_PARSE_HEADERS)
                st->parser->flags |= PARSER_FLAG_COMPLETE_FRAMES;
            else if (st->need_parsing == AVSTREAM_PARSE_FULL_ONCE)
                st->parser->flags |= PARSER_FLAG_ONCE;
            else if (st->need_parsing == AVSTREAM_PARSE_FULL_RAW)
                st->parser->flags |= PARSER_FLAG_USE_CODEC_TS;
        }

		// 不需要parse的情况下，packet还存放在pkt变量中，没有放在parse_queue，所以got_packet = 1
        if (!st->need_parsing || !st->parser) {
            /* no parsing needed: we just output the packet as is */
            compute_pkt_fields(s, st, NULL, pkt, AV_NOPTS_VALUE, AV_NOPTS_VALUE);
            if ((s->iformat->flags & AVFMT_GENERIC_INDEX) &&
                (pkt->flags & AV_PKT_FLAG_KEY) && pkt->dts != AV_NOPTS_VALUE) {
                ff_reduce_index(s, st->index);
                av_add_index_entry(st, pkt->pos, pkt->dts,
                                   0, 0, AVINDEX_KEYFRAME);
            }
            got_packet = 1;
        } else if (st->discard < AVDISCARD_ALL) {
            if ((ret = parse_packet(s, pkt, pkt->stream_index, 0)) < 0)
                return ret;
            st->codecpar->sample_rate = st->internal->avctx->sample_rate;
            st->codecpar->bit_rate = st->internal->avctx->bit_rate;
            st->codecpar->channels = st->internal->avctx->channels;
            st->codecpar->channel_layout = st->internal->avctx->channel_layout;
            st->codecpar->codec_id = st->internal->avctx->codec_id;
        } else {
            /* free packet */
            av_packet_unref(pkt);
        }
        if (pkt->flags & AV_PKT_FLAG_KEY)
            st->internal->skip_to_keyframe = 0;
        if (st->internal->skip_to_keyframe) {
            av_packet_unref(pkt);
            got_packet = 0;
        }
    }

	//未找到，然后parse_queue中又存在的话，那就说明pkt被转到parse_queue中，所以从parse_queue中获取
    if (!got_packet && s->internal->parse_queue)
        ret = avpriv_packet_list_get(&s->internal->parse_queue, &s->internal->parse_queue_end, pkt);

    if (ret >= 0) {
        AVStream *st = s->streams[pkt->stream_index];
        int discard_padding = 0;
        if (st->internal->first_discard_sample && pkt->pts != AV_NOPTS_VALUE) {
            int64_t pts = pkt->pts - (is_relative(pkt->pts) ? RELATIVE_TS_BASE : 0);
            int64_t sample = ts_to_samples(st, pts);
            int duration = ts_to_samples(st, pkt->duration);
            int64_t end_sample = sample + duration;
            if (duration > 0 && end_sample >= st->internal->first_discard_sample &&
                sample < st->internal->last_discard_sample)
                discard_padding = FFMIN(end_sample - st->internal->first_discard_sample, duration);
        }
        if (st->internal->start_skip_samples && (pkt->pts == 0 || pkt->pts == RELATIVE_TS_BASE))
            st->internal->skip_samples = st->internal->start_skip_samples;
        if (st->internal->skip_samples || discard_padding) {
            uint8_t *p = av_packet_new_side_data(pkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
            if (p) {
                AV_WL32(p, st->internal->skip_samples);
                AV_WL32(p + 4, discard_padding);
                av_log(s, AV_LOG_DEBUG, "demuxer injecting skip %d / discard %d\n", st->internal->skip_samples, discard_padding);
            }
            st->internal->skip_samples = 0;
        }

        if (st->internal->inject_global_side_data) {
            for (i = 0; i < st->nb_side_data; i++) {
                AVPacketSideData *src_sd = &st->side_data[i];
                uint8_t *dst_data;

                if (av_packet_get_side_data(pkt, src_sd->type, NULL))
                    continue;

                dst_data = av_packet_new_side_data(pkt, src_sd->type, src_sd->size);
                if (!dst_data) {
                    av_log(s, AV_LOG_WARNING, "Could not inject global side data\n");
                    continue;
                }

                memcpy(dst_data, src_sd->data, src_sd->size);
            }
            st->internal->inject_global_side_data = 0;
        }
    }

    av_opt_get_dict_val(s, "metadata", AV_OPT_SEARCH_CHILDREN, &metadata);
    if (metadata) {
        s->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
        av_dict_copy(&s->metadata, metadata, 0);
        av_dict_free(&metadata);
        av_opt_set_dict_val(s, "metadata", NULL, AV_OPT_SEARCH_CHILDREN);
    }

#if FF_API_LAVF_AVCTX
    update_stream_avctx(s);
#endif

    if (s->debug & FF_FDEBUG_TS)
        av_log(s, AV_LOG_DEBUG,
               "read_frame_internal stream=%d, pts=%s, dts=%s, "
               "size=%d, duration=%"PRId64", flags=%d\n",
               pkt->stream_index,
               av_ts2str(pkt->pts),
               av_ts2str(pkt->dts),
               pkt->size, pkt->duration, pkt->flags);

    /* A demuxer might have returned EOF because of an IO error, let's
     * propagate this back to the user. */
    if (ret == AVERROR_EOF && s->pb && s->pb->error < 0 && s->pb->error != AVERROR(EAGAIN))
        ret = s->pb->error;

    return ret;
}

ff_read_packet源码

从AVInputFormat或者raw_packet_buffer中读取一个AVPacket，如果raw_packet_buffer中有，就从raw_packet_buffer中读取，没有的话，只能从对应的AVInputFormat中读取。

int ff_read_packet(AVFormatContext *s, AVPacket *pkt)
{
    int ret, i, err;
    AVStream *st;

#if FF_API_INIT_PACKET
FF_DISABLE_DEPRECATION_WARNINGS
    pkt->data = NULL;
    pkt->size = 0;
    av_init_packet(pkt);
FF_ENABLE_DEPRECATION_WARNINGS
#else
    av_packet_unref(pkt);
#endif

    for (;;) {
        PacketList *pktl = s->internal->raw_packet_buffer;
        const AVPacket *pkt1;

		//  probe的时候是不是有缓存在内存中的packet
        if (pktl) {
            st = s->streams[pktl->pkt.stream_index];
            if (s->internal->raw_packet_buffer_remaining_size <= 0)
                if ((err = probe_codec(s, st, NULL)) < 0)
                    return err;
            if (st->internal->request_probe <= 0) {
                avpriv_packet_list_get(&s->internal->raw_packet_buffer,
                                   &s->internal->raw_packet_buffer_end, pkt);
                s->internal->raw_packet_buffer_remaining_size += pkt->size;
                return 0;
            }
        }

		// 从对应的AVInputFormat格式中读取AVPacket， 在Mp4格式中就是读取Sample。
        ret = s->iformat->read_packet(s, pkt);
        if (ret < 0) {
            av_packet_unref(pkt);

            /* Some demuxers return FFERROR_REDO when they consume
               data and discard it (ignored streams, junk, extradata).
               We must re-call the demuxer to get the real packet. */
            if (ret == FFERROR_REDO)
                continue;
            if (!pktl || ret == AVERROR(EAGAIN))
                return ret;
            for (i = 0; i < s->nb_streams; i++) {
                st = s->streams[i];
                if (st->probe_packets || st->internal->request_probe > 0)
                    if ((err = probe_codec(s, st, NULL)) < 0)
                        return err;
                av_assert0(st->internal->request_probe <= 0);
            }
            continue;
        }

		// 把数据从 pkt->buffer->data 转移到 pkt->data
        err = av_packet_make_refcounted(pkt);
        if (err < 0) {
            av_packet_unref(pkt);
            return err;
        }

        if (pkt->flags & AV_PKT_FLAG_CORRUPT) {
            av_log(s, AV_LOG_WARNING,
                   "Packet corrupt (stream = %d, dts = %s)",
                   pkt->stream_index, av_ts2str(pkt->dts));
            if (s->flags & AVFMT_FLAG_DISCARD_CORRUPT) {
                av_log(s, AV_LOG_WARNING, ", dropping it.\n");
                av_packet_unref(pkt);
                continue;
            }
            av_log(s, AV_LOG_WARNING, ".\n");
        }

        av_assert0(pkt->stream_index < (unsigned)s->nb_streams &&
                   "Invalid stream index.\n");

        st = s->streams[pkt->stream_index];

		// 更新AVProgram中的pts_wrap_reference和pts_wrap_behavior
		// 不过Mp4格式并不用这个AVProgram
        if (update_wrap_reference(s, st, pkt->stream_index, pkt) && st->internal->pts_wrap_behavior == AV_PTS_WRAP_SUB_OFFSET) {
            // correct first time stamps to negative values
            if (!is_relative(st->first_dts))
                st->first_dts = wrap_timestamp(st, st->first_dts);
            if (!is_relative(st->start_time))
                st->start_time = wrap_timestamp(st, st->start_time);
            if (!is_relative(st->cur_dts))
                st->cur_dts = wrap_timestamp(st, st->cur_dts);
        }

		// 如果有wrap的行为，就根据pts_wrap_reference和pts_wrap_behavior来重新定义pts和dts
        pkt->dts = wrap_timestamp(st, pkt->dts);
        pkt->pts = wrap_timestamp(st, pkt->pts);

        force_codec_ids(s, st);

        /* TODO: audio: time filter; video: frame reordering (pts != dts) */
        if (s->use_wallclock_as_timestamps)
            pkt->dts = pkt->pts = av_rescale_q(av_gettime(), AV_TIME_BASE_Q, st->time_base);

		// 用的不是raw_packet_buffer中的AVPacket就在这里return
        if (!pktl && st->internal->request_probe <= 0)
            return ret;

        err = avpriv_packet_list_put(&s->internal->raw_packet_buffer,
                                 &s->internal->raw_packet_buffer_end,
                                 pkt, NULL, 0);
        if (err < 0) {
            av_packet_unref(pkt);
            return err;
        }
        pkt1 = &s->internal->raw_packet_buffer_end->pkt;
        s->internal->raw_packet_buffer_remaining_size -= pkt1->size;

        if ((err = probe_codec(s, st, pkt1)) < 0)
            return err;
    }
}