From 4568c2bf975e51d843bf1ff6ac06060a8a6291b3 Mon Sep 17 00:00:00 2001 From: Justin Ruggles Date: Sun, 11 Sep 2011 20:17:54 -0400 Subject: [PATCH] vmdaudio: fix decoding of 16-bit audio format. The initial sample of each block is raw 16-bit PCM, not DPCM. Fixes decoding of all samples in: http://streams.videolan.org/samples/game-formats/sierra-vmd/Lighthouse/ --- libavcodec/vmdav.c | 120 +++++++++++++++++++++++++++++++++++------------------ 1 file changed, 80 insertions(+), 40 deletions(-) diff --git a/libavcodec/vmdav.c b/libavcodec/vmdav.c index 110d19c..4528996 100644 --- a/libavcodec/vmdav.c +++ b/libavcodec/vmdav.c @@ -417,9 +417,8 @@ static av_cold int vmdvideo_decode_end(AVCodecContext *avctx) #define BLOCK_TYPE_SILENCE 3 typedef struct VmdAudioContext { - AVCodecContext *avctx; int out_bps; - int predictors[2]; + int chunk_size; } VmdAudioContext; static const uint16_t vmdaudio_table[128] = { @@ -442,13 +441,23 @@ static av_cold int vmdaudio_decode_init(AVCodecContext *avctx) { VmdAudioContext *s = avctx->priv_data; - s->avctx = avctx; + if (avctx->channels < 1 || avctx->channels > 2) { + av_log(avctx, AV_LOG_ERROR, "invalid number of channels\n"); + return AVERROR(EINVAL); + } + if (avctx->block_align < 1) { + av_log(avctx, AV_LOG_ERROR, "invalid block align\n"); + return AVERROR(EINVAL); + } + if (avctx->bits_per_coded_sample == 16) avctx->sample_fmt = AV_SAMPLE_FMT_S16; else avctx->sample_fmt = AV_SAMPLE_FMT_U8; s->out_bps = av_get_bytes_per_sample(avctx->sample_fmt); + s->chunk_size = avctx->block_align + avctx->channels * (s->out_bps == 2); + av_log(avctx, AV_LOG_DEBUG, "%d channels, %d bits/sample, " "block align = %d, sample rate = %d\n", avctx->channels, avctx->bits_per_coded_sample, avctx->block_align, @@ -457,41 +466,33 @@ static av_cold int vmdaudio_decode_init(AVCodecContext *avctx) return 0; } -static void vmdaudio_decode_audio(VmdAudioContext *s, unsigned char *data, - const uint8_t *buf, int buf_size, int stereo) +static void decode_audio_s16(int16_t *out, const uint8_t *buf, int buf_size, + int channels) { - int i; - int chan = 0; - int16_t *out = (int16_t*)data; - - for(i = 0; i < buf_size; i++) { - if(buf[i] & 0x80) - s->predictors[chan] -= vmdaudio_table[buf[i] & 0x7F]; - else - s->predictors[chan] += vmdaudio_table[buf[i]]; - s->predictors[chan] = av_clip_int16(s->predictors[chan]); - out[i] = s->predictors[chan]; - chan ^= stereo; + int ch; + const uint8_t *buf_end = buf + buf_size; + int predictor[2]; + int st = channels - 1; + + /* decode initial raw sample */ + for (ch = 0; ch < channels; ch++) { + predictor[ch] = (int16_t)AV_RL16(buf); + buf += 2; + *out++ = predictor[ch]; } -} -static int vmdaudio_loadsound(VmdAudioContext *s, unsigned char *data, - const uint8_t *buf, int silent_chunks, int data_size) -{ - int silent_size = s->avctx->block_align * silent_chunks * s->out_bps; - - if (silent_chunks) { - memset(data, s->out_bps == 2 ? 0x00 : 0x80, silent_size); - data += silent_size; - } - if (s->avctx->bits_per_coded_sample == 16) - vmdaudio_decode_audio(s, data, buf, data_size, s->avctx->channels == 2); - else { - /* just copy the data */ - memcpy(data, buf, data_size); + /* decode DPCM samples */ + ch = 0; + while (buf < buf_end) { + uint8_t b = *buf++; + if (b & 0x80) + predictor[ch] -= vmdaudio_table[b & 0x7F]; + else + predictor[ch] += vmdaudio_table[b]; + predictor[ch] = av_clip_int16(predictor[ch]); + *out++ = predictor[ch]; + ch ^= st; } - - return silent_size + data_size * s->out_bps; } static int vmdaudio_decode_frame(AVCodecContext *avctx, @@ -499,10 +500,13 @@ static int vmdaudio_decode_frame(AVCodecContext *avctx, AVPacket *avpkt) { const uint8_t *buf = avpkt->data; + const uint8_t *buf_end; int buf_size = avpkt->size; VmdAudioContext *s = avctx->priv_data; - int block_type, silent_chunks; - unsigned char *output_samples = (unsigned char *)data; + int block_type, silent_chunks, audio_chunks; + int nb_samples, out_size; + uint8_t *output_samples_u8 = data; + int16_t *output_samples_s16 = data; if (buf_size < 16) { av_log(avctx, AV_LOG_WARNING, "skipping small junk packet\n"); @@ -518,10 +522,16 @@ static int vmdaudio_decode_frame(AVCodecContext *avctx, buf += 16; buf_size -= 16; + /* get number of silent chunks */ silent_chunks = 0; if (block_type == BLOCK_TYPE_INITIAL) { - uint32_t flags = AV_RB32(buf); - silent_chunks = av_popcount(flags); + uint32_t flags; + if (buf_size < 4) { + av_log(avctx, AV_LOG_ERROR, "packet is too small\n"); + return AVERROR(EINVAL); + } + flags = AV_RB32(buf); + silent_chunks = av_popcount(flags); buf += 4; buf_size -= 4; } else if (block_type == BLOCK_TYPE_SILENCE) { @@ -530,11 +540,41 @@ static int vmdaudio_decode_frame(AVCodecContext *avctx, } /* ensure output buffer is large enough */ - if (*data_size < (avctx->block_align*silent_chunks + buf_size) * s->out_bps) + audio_chunks = buf_size / s->chunk_size; + nb_samples = ((silent_chunks + audio_chunks) * avctx->block_align) / avctx->channels; + out_size = nb_samples * avctx->channels * s->out_bps; + if (*data_size < out_size) return -1; - *data_size = vmdaudio_loadsound(s, output_samples, buf, silent_chunks, buf_size); + /* decode silent chunks */ + if (silent_chunks > 0) { + int silent_size = avctx->block_align * silent_chunks; + if (s->out_bps == 2) { + memset(output_samples_s16, 0x00, silent_size * 2); + output_samples_s16 += silent_size; + } else { + memset(output_samples_u8, 0x80, silent_size); + output_samples_u8 += silent_size; + } + } + + /* decode audio chunks */ + if (audio_chunks > 0) { + buf_end = buf + buf_size; + while (buf < buf_end) { + if (s->out_bps == 2) { + decode_audio_s16(output_samples_s16, buf, s->chunk_size, + avctx->channels); + output_samples_s16 += avctx->block_align; + } else { + memcpy(output_samples_u8, buf, s->chunk_size); + output_samples_u8 += avctx->block_align; + } + buf += s->chunk_size; + } + } + *data_size = out_size; return avpkt->size; } -- 2.7.4