Re: [FFmpeg-devel] [PATCH] avcodec/binkaudio: add support for >2 channels dct codec

[Paul B Mahol <onemda@gmail.com>]

On 3/18/22, Andreas Rheinhardt <andreas.rheinhardt@outlook.com> wrote:
> Paul B Mahol:
>> As presented in .binka files.
>>
>> Signed-off-by: Paul B Mahol <onemda@gmail.com>
>> ---
>>  libavcodec/binkaudio.c | 50 +++++++++++++++++++++++++++---------------
>>  1 file changed, 32 insertions(+), 18 deletions(-)
>>
>> diff --git a/libavcodec/binkaudio.c b/libavcodec/binkaudio.c
>> index b4ff15beeb..54b7e22854 100644
>> --- a/libavcodec/binkaudio.c
>> +++ b/libavcodec/binkaudio.c
>> @@ -51,13 +51,14 @@ typedef struct BinkAudioContext {
>>      int version_b;          ///< Bink version 'b'
>>      int first;
>>      int channels;
>> +    int ch_offset;
>>      int frame_len;          ///< transform size (samples)
>>      int overlap_len;        ///< overlap size (samples)
>>      int block_size;
>>      int num_bands;
>>      float root;
>>      unsigned int bands[26];
>> -    float previous[MAX_CHANNELS][BINK_BLOCK_MAX_SIZE / 16];  ///< coeffs
>> from previous audio block
>> +    float previous[6][BINK_BLOCK_MAX_SIZE / 16];  ///< coeffs from
>> previous audio block
>>      float quant_table[96];
>>      AVPacket *pkt;
>>      union {
>> @@ -74,6 +75,7 @@ static av_cold int decode_init(AVCodecContext *avctx)
>>      int sample_rate_half;
>>      int i, ret;
>>      int frame_len_bits;
>> +    int max_channels = avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT ?
>> MAX_CHANNELS : 6;
>
> If you allow up to six channels, then MAX_CHANNELS (i.e. two) needs to
> be renamed.
>
>>      int channels = avctx->ch_layout.nb_channels;
>>
>>      /* determine frame length */
>> @@ -85,7 +87,7 @@ static av_cold int decode_init(AVCodecContext *avctx)
>>          frame_len_bits = 11;
>>      }
>>
>> -    if (channels < 1 || channels > MAX_CHANNELS) {
>> +    if (channels < 1 || channels > max_channels) {
>>          av_log(avctx, AV_LOG_ERROR, "invalid number of channels: %d\n",
>> channels);
>>          return AVERROR_INVALIDDATA;
>>      }
>> @@ -110,7 +112,7 @@ static av_cold int decode_init(AVCodecContext *avctx)
>>
>>      s->frame_len     = 1 << frame_len_bits;
>>      s->overlap_len   = s->frame_len / 16;
>> -    s->block_size    = (s->frame_len - s->overlap_len) * s->channels;
>> +    s->block_size    = (s->frame_len - s->overlap_len) *
>> FFMIN(MAX_CHANNELS, s->channels);
>>      sample_rate_half = (sample_rate + 1LL) / 2;
>>      if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
>>          s->root = 2.0 / (sqrt(s->frame_len) * 32768.0);
>> @@ -166,7 +168,8 @@ static const uint8_t rle_length_tab[16] = {
>>   * @param[out] out Output buffer (must contain s->block_size elements)
>>   * @return 0 on success, negative error code on failure
>>   */
>> -static int decode_block(BinkAudioContext *s, float **out, int use_dct)
>> +static int decode_block(BinkAudioContext *s, float **out, int use_dct,
>> +                        int channels, int ch_offset)
>>  {
>>      int ch, i, j, k;
>>      float q, quant[25];
>> @@ -176,8 +179,8 @@ static int decode_block(BinkAudioContext *s, float
>> **out, int use_dct)
>>      if (use_dct)
>>          skip_bits(gb, 2);
>>
>> -    for (ch = 0; ch < s->channels; ch++) {
>> -        FFTSample *coeffs = out[ch];
>> +    for (ch = 0; ch < channels; ch++) {
>> +        FFTSample *coeffs = out[ch + ch_offset];
>>
>>          if (s->version_b) {
>>              if (get_bits_left(gb) < 64)
>> @@ -252,17 +255,17 @@ static int decode_block(BinkAudioContext *s, float
>> **out, int use_dct)
>>              s->trans.rdft.rdft_calc(&s->trans.rdft, coeffs);
>>      }
>>
>> -    for (ch = 0; ch < s->channels; ch++) {
>> +    for (ch = 0; ch < channels; ch++) {
>>          int j;
>> -        int count = s->overlap_len * s->channels;
>> +        int count = s->overlap_len * channels;
>>          if (!s->first) {
>>              j = ch;
>> -            for (i = 0; i < s->overlap_len; i++, j += s->channels)
>> -                out[ch][i] = (s->previous[ch][i] * (count - j) +
>> -                                      out[ch][i] *          j) / count;
>> +            for (i = 0; i < s->overlap_len; i++, j += channels)
>> +                out[ch + ch_offset][i] = (s->previous[ch + ch_offset][i]
>> * (count - j) +
>> +                                      out[ch + ch_offset][i] *
>> j) / count;
>>          }
>> -        memcpy(s->previous[ch], &out[ch][s->frame_len - s->overlap_len],
>> -               s->overlap_len * sizeof(*s->previous[ch]));
>> +        memcpy(s->previous[ch + ch_offset], &out[ch +
>> ch_offset][s->frame_len - s->overlap_len],
>> +               s->overlap_len * sizeof(*s->previous[ch + ch_offset]));
>>      }
>>
>>      s->first = 0;
>> @@ -293,6 +296,7 @@ static int binkaudio_receive_frame(AVCodecContext
>> *avctx, AVFrame *frame)
>>      GetBitContext *gb = &s->gb;
>>      int ret;
>>
>> +again:
>>      if (!s->pkt->data) {
>>          ret = ff_decode_get_packet(avctx, s->pkt);
>>          if (ret < 0)
>> @@ -313,22 +317,31 @@ static int binkaudio_receive_frame(AVCodecContext
>> *avctx, AVFrame *frame)
>>      }
>>
>>      /* get output buffer */
>> -    frame->nb_samples = s->frame_len;
>> -    if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
>> -        return ret;
>> +    if (s->ch_offset == 0) {
>> +        frame->nb_samples = s->frame_len;
>> +        if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
>> +            return ret;
>> +    }
>>
>>      if (decode_block(s, (float **)frame->extended_data,
>> -                     avctx->codec->id == AV_CODEC_ID_BINKAUDIO_DCT)) {
>> +                     avctx->codec->id == AV_CODEC_ID_BINKAUDIO_DCT,
>> +                     FFMIN(MAX_CHANNELS, s->channels), s->ch_offset)) {
>>          av_log(avctx, AV_LOG_ERROR, "Incomplete packet\n");
>>          return AVERROR_INVALIDDATA;
>>      }
>> +    s->ch_offset += MAX_CHANNELS;
>>      get_bits_align32(gb);
>>      if (!get_bits_left(gb)) {
>>          memset(gb, 0, sizeof(*gb));
>>          av_packet_unref(s->pkt);
>>      }
>> +    if (s->ch_offset >= s->channels) {
>> +        s->ch_offset = 0;
>> +    } else {
>> +        goto again;
>> +    }
>
> Is it really intended that the data for one multi-channel frame is
> divided into several input packets?

You are missing big picture here, >2 files have channels in different
packets interleaved.
Something like in XMA. (And nothing signals how are they interleaved.
so its worse than in XMA) So it is working fine. I just need another
look for possible regressions and security implications. Renaming
MAX_CHANNELS is not useful as that is not property of both codecs.

>
>>
>> -    frame->nb_samples = s->block_size / avctx->ch_layout.nb_channels;
>> +    frame->nb_samples = s->block_size /
>> FFMIN(avctx->ch_layout.nb_channels, MAX_CHANNELS);
>>
>>      return 0;
>>  fail:
>> @@ -343,6 +356,7 @@ static void decode_flush(AVCodecContext *avctx)
>>      /* s->pkt coincides with avctx->internal->in_pkt
>>       * and is unreferenced generically when flushing. */
>>      s->first = 1;
>> +    s->ch_offset = 0;
>>  }
>>
>>  const AVCodec ff_binkaudio_rdft_decoder = {
>
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel@ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".