[FFmpeg-devel] [PATCH 5/6] avcodec: add h266 parser

[James Almer]

On 12/21/2020 3:07 AM, Nuo Mi wrote:
> ---
>   configure                |   3 +
>   libavcodec/Makefile      |   1 +
>   libavcodec/h2645_parse.c |  73 +++++++++-
>   libavcodec/h266_parser.c | 284 +++++++++++++++++++++++++++++++++++++++
>   libavcodec/parsers.c     |   1 +
>   5 files changed, 360 insertions(+), 2 deletions(-)
>   create mode 100644 libavcodec/h266_parser.c

[...]

> +/**
> + * Parse NAL units of found picture and decode some basic information.
> + *
> + * @param s parser context.
> + * @param avctx codec context.
> + * @param buf buffer with field/frame data.
> + * @param buf_size size of the buffer.
> + */
> +static int parse_nal_units(AVCodecParserContext *ctx, const uint8_t *buf,
> +                           int buf_size, AVCodecContext *avctx)
> +{
> +    H266ParserContext *s = ctx->priv_data;
> +    CodedBitstreamFragment *pu = &s->picture_unit;
> +    CodedBitstreamH266Context *h266 = s->cbc->priv_data;
> +    const H266RawSPS *sps;
> +    const H266RawPPS *pps;
> +    const H266RawPH *ph;
> +    int ret, num = 0, den = 0;
> +
> +    /* set some sane default values */
> +    ctx->pict_type         = AV_PICTURE_TYPE_I;
> +    ctx->key_frame         = 0;
> +    ctx->picture_structure = AV_PICTURE_STRUCTURE_FRAME;
> +
> +    s->cbc->log_ctx = avctx;
> +
> +    if (avctx->extradata_size && !s->parsed_extradata) {
> +        s->parsed_extradata = 1;
> +
> +        if ((ret = ff_cbs_read(s->cbc, pu, avctx->extradata, avctx->extradata_size)) < 0)
> +            av_log(avctx, AV_LOG_WARNING, "Failed to parse extradata.\n");
> +
> +        ff_cbs_fragment_reset(pu);
> +    }
> +
> +    if ((ret = ff_cbs_read(s->cbc, pu, buf, buf_size))< 0) {
> +        av_log(avctx, AV_LOG_ERROR, "Failed to parse picture unit.\n");
> +        goto end;
> +    }
> +
> +    sps = h266->active_sps;
> +    pps = h266->active_pps;
> +    ph = h266->active_ph;
> +    if (!h266->active_pps || !h266->active_ph) {
> +        av_log(avctx, AV_LOG_ERROR, "No pss or ph header available\n");
> +        goto end;
> +    }
> +
> +    ctx->key_frame = ph->ph_gdr_or_irap_pic_flag;
> +
> +    ctx->coded_width  = pps->pps_pic_width_in_luma_samples;
> +    ctx->coded_height = pps->pps_pic_height_in_luma_samples;
> +    ctx->width        = pps->pps_pic_width_in_luma_samples  - pps->pps_conf_win_left_offset - pps->pps_conf_win_right_offset;
> +    ctx->height       = pps->pps_pic_height_in_luma_samples - pps->pps_conf_win_top_offset  - pps->pps_conf_win_bottom_offset;
> +    ctx->pict_type    = get_pict_type(pu);
> +    ctx->format       = get_format(sps);
> +    avctx->profile  = sps->profile_tier_level.general_profile_idc;
> +    avctx->level    = sps->profile_tier_level.general_level_idc;
> +
> +
> +
> +    if(sps->sps_ptl_dpb_hrd_params_present_flag && sps->sps_timing_hrd_params_present_flag) {
> +        num = sps->sps_general_timing_hrd_parameters.num_units_in_tick;
> +        den = sps->sps_general_timing_hrd_parameters.time_scale;
> +    } else {
> +        av_log(avctx, AV_LOG_INFO, "No sps_timing_hrd_params_present_flag in sps, the fps may not right.\n");
> +        goto end;
> +    }
> +    if (num != 0 && den != 0)
> +        av_reduce(&avctx->framerate.den, &avctx->framerate.num,
> +                  num, den, 1 << 30);
> +end:
> +
> +    ff_cbs_fragment_reset(pu);
> +    s->cbc->log_ctx = NULL;
> +    return 0;
> +}
> +
> +static int h266_parser_parse(AVCodecParserContext *s, AVCodecContext *avctx,
> +                      const uint8_t **poutbuf, int *poutbuf_size,
> +                      const uint8_t *buf, int buf_size)
> +{
> +    int next;
> +    H266ParserContext *ctx = s->priv_data;
> +    ParseContext *pc = &ctx->pc;
> +    int is_dummy_buf = !buf_size;
> +    const uint8_t *dummy_buf = buf;
> +
> +    if (avctx->extradata && !ctx->parsed_extradata) {
> +        av_log(avctx, AV_LOG_INFO, "extra data is not supported yet.\n");
> +        return AVERROR_PATCHWELCOME;

You're aborting here if extradata is present, but then check again in 
parse_nal_units() and actually try to parse it? Either parse it here or 
in parse_nal_units().

Judging by this check here, i assume in your tests you didn't handle 
streams with extradata, which is expected from raw bitstreams parsed by 
the demuxer you added in patch 3/6.
To workaround this, you need to add VVC support to the extract_extradata 
bitstream filter. It should be trivial to do.

> +    }
> +
> +    if (s->flags & PARSER_FLAG_COMPLETE_FRAMES) {
> +        next = buf_size;
> +    } else {
> +        next = find_frame_end(s, buf, buf_size);
> +        if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) {
> +            *poutbuf      = NULL;
> +            *poutbuf_size = 0;
> +            return buf_size;
> +        }
> +    }
> +
> +    is_dummy_buf &= (dummy_buf == buf);
> +
> +    if (!is_dummy_buf)
> +        parse_nal_units(s, buf, buf_size, avctx);
> +
> +    *poutbuf      = buf;
> +    *poutbuf_size = buf_size;
> +    return next;
> +
> +}
> +
> +static const CodedBitstreamUnitType decompose_unit_types[] = {
> +    H266_NAL_TRAIL,
> +    H266_NAL_STSA,
> +    H266_NAL_RADL,
> +    H266_NAL_RASL,
> +    H266_NAL_IDR_W_RADL,
> +    H266_NAL_IDR_N_LP,
> +    H266_NAL_CRA_NUT,
> +    H266_NAL_GDR_NUT,
> +    H266_NAL_VPS,
> +    H266_NAL_SPS,
> +    H266_NAL_PPS,
> +    H266_NAL_PH,
> +    H266_NAL_AUD,
> +};
> +
> +static av_cold int h266_parser_init(AVCodecParserContext *ctx)
> +{
> +    H266ParserContext *s = ctx->priv_data;
> +    int ret;
> +
> +    ret = ff_cbs_init(&s->cbc, AV_CODEC_ID_VVC, NULL);
> +    if (ret < 0)
> +        return ret;
> +
> +    s->cbc->decompose_unit_types    = (CodedBitstreamUnitType *)decompose_unit_types;
> +    s->cbc->nb_decompose_unit_types = FF_ARRAY_ELEMS(decompose_unit_types);
> +
> +    return 0;
> +}
> +
> +static void h266_parser_close(AVCodecParserContext *ctx)
> +{
> +    H266ParserContext *s = ctx->priv_data;
> +
> +    ff_cbs_fragment_free(&s->picture_unit);
> +    ff_cbs_close(&s->cbc);
> +    av_freep(&s->pc.buffer);
> +}
> +
> +AVCodecParser ff_h266_parser = {
> +    .codec_ids      = { AV_CODEC_ID_VVC },
> +    .priv_data_size = sizeof(H266ParserContext),
> +    .parser_init    = h266_parser_init,
> +    .parser_close   = h266_parser_close,
> +    .parser_parse   = h266_parser_parse,
> +};
> diff --git a/libavcodec/parsers.c b/libavcodec/parsers.c
> index 83271d95a3..155aec71aa 100644
> --- a/libavcodec/parsers.c
> +++ b/libavcodec/parsers.c
> @@ -49,6 +49,7 @@ extern AVCodecParser ff_gsm_parser;
>   extern AVCodecParser ff_h261_parser;
>   extern AVCodecParser ff_h263_parser;
>   extern AVCodecParser ff_h264_parser;
> +extern AVCodecParser ff_h266_parser;
>   extern AVCodecParser ff_hevc_parser;
>   extern AVCodecParser ff_ipu_parser;
>   extern AVCodecParser ff_jpeg2000_parser;
>