error_resilience: guess_dc: check malloc failure
[ffmpeg.git] / libavcodec / vorbisdec.c
index 024c8fd..8ce0fc7 100644 (file)
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+/**
+ * @file
+ * Vorbis I decoder
+ * @author Denes Balatoni  ( dbalatoni programozo hu )
+ */
+
 #include <inttypes.h>
 #include <math.h>
 
-#define ALT_BITSTREAM_READER_LE
+#define BITSTREAM_READER_LE
+#include "libavutil/float_dsp.h"
+#include "libavutil/avassert.h"
 #include "avcodec.h"
 #include "get_bits.h"
 #include "dsputil.h"
@@ -38,9 +46,6 @@
 #define V_MAX_VLCS (1 << 16)
 #define V_MAX_PARTITIONS (1 << 20)
 
-#undef NDEBUG
-#include <assert.h>
-
 typedef struct {
     uint8_t      dimensions;
     uint8_t      lookup_type;
@@ -119,8 +124,10 @@ typedef struct {
 
 typedef struct vorbis_context_s {
     AVCodecContext *avccontext;
+    AVFrame frame;
     GetBitContext gb;
     DSPContext dsp;
+    AVFloatDSPContext fdsp;
     FmtConvertContext fmt_conv;
 
     FFTContext mdct[2];
@@ -162,7 +169,7 @@ static const char idx_err_str[] = "Index value %d out of range (0 - %d) for %s a
         av_log(vc->avccontext, AV_LOG_ERROR,\
                idx_err_str,\
                (int)(idx), (int)(limit - 1), #idx, __FILE__, __LINE__);\
-        return -1;\
+        return AVERROR_INVALIDDATA;\
     }
 #define GET_VALIDATED_INDEX(idx, bits, limit) \
     {\
@@ -200,7 +207,7 @@ static void vorbis_free(vorbis_context *vc)
 
     for (i = 0; i < vc->codebook_count; ++i) {
         av_free(vc->codebooks[i].codevectors);
-        free_vlc(&vc->codebooks[i].vlc);
+        ff_free_vlc(&vc->codebooks[i].vlc);
     }
     av_freep(&vc->codebooks);
 
@@ -235,6 +242,7 @@ static int vorbis_parse_setup_hdr_codebooks(vorbis_context *vc)
     uint32_t *tmp_vlc_codes;
     GetBitContext *gb = &vc->gb;
     uint16_t *codebook_multiplicands;
+    int ret = 0;
 
     vc->codebook_count = get_bits(gb, 8) + 1;
 
@@ -254,6 +262,7 @@ static int vorbis_parse_setup_hdr_codebooks(vorbis_context *vc)
         if (get_bits(gb, 24) != 0x564342) {
             av_log(vc->avccontext, AV_LOG_ERROR,
                    " %u. Codebook setup data corrupt.\n", cb);
+            ret = AVERROR_INVALIDDATA;
             goto error;
         }
 
@@ -262,6 +271,7 @@ static int vorbis_parse_setup_hdr_codebooks(vorbis_context *vc)
             av_log(vc->avccontext, AV_LOG_ERROR,
                    " %u. Codebook's dimension is invalid (%d).\n",
                    cb, codebook_setup->dimensions);
+            ret = AVERROR_INVALIDDATA;
             goto error;
         }
         entries = get_bits(gb, 24);
@@ -269,6 +279,7 @@ static int vorbis_parse_setup_hdr_codebooks(vorbis_context *vc)
             av_log(vc->avccontext, AV_LOG_ERROR,
                    " %u. Codebook has too many entries (%u).\n",
                    cb, entries);
+            ret = AVERROR_INVALIDDATA;
             goto error;
         }
 
@@ -326,6 +337,7 @@ static int vorbis_parse_setup_hdr_codebooks(vorbis_context *vc)
             }
             if (current_entry>used_entries) {
                 av_log(vc->avccontext, AV_LOG_ERROR, " More codelengths than codes in codebook. \n");
+                ret = AVERROR_INVALIDDATA;
                 goto error;
             }
         }
@@ -393,17 +405,20 @@ static int vorbis_parse_setup_hdr_codebooks(vorbis_context *vc)
             }
             if (j != used_entries) {
                 av_log(vc->avccontext, AV_LOG_ERROR, "Bug in codevector vector building code. \n");
+                ret = AVERROR_INVALIDDATA;
                 goto error;
             }
             entries = used_entries;
         } else if (codebook_setup->lookup_type >= 2) {
             av_log(vc->avccontext, AV_LOG_ERROR, "Codebook lookup type not supported. \n");
+            ret = AVERROR_INVALIDDATA;
             goto error;
         }
 
 // Initialize VLC table
         if (ff_vorbis_len2vlc(tmp_vlc_bits, tmp_vlc_codes, entries)) {
             av_log(vc->avccontext, AV_LOG_ERROR, " Invalid code lengths while generating vlcs. \n");
+            ret = AVERROR_INVALIDDATA;
             goto error;
         }
         codebook_setup->maxdepth = 0;
@@ -418,7 +433,11 @@ static int vorbis_parse_setup_hdr_codebooks(vorbis_context *vc)
 
         codebook_setup->maxdepth = (codebook_setup->maxdepth+codebook_setup->nb_bits - 1) / codebook_setup->nb_bits;
 
-        if (init_vlc(&codebook_setup->vlc, codebook_setup->nb_bits, entries, tmp_vlc_bits, sizeof(*tmp_vlc_bits), sizeof(*tmp_vlc_bits), tmp_vlc_codes, sizeof(*tmp_vlc_codes), sizeof(*tmp_vlc_codes), INIT_VLC_LE)) {
+        if ((ret = init_vlc(&codebook_setup->vlc, codebook_setup->nb_bits,
+                            entries, tmp_vlc_bits, sizeof(*tmp_vlc_bits),
+                            sizeof(*tmp_vlc_bits), tmp_vlc_codes,
+                            sizeof(*tmp_vlc_codes), sizeof(*tmp_vlc_codes),
+                            INIT_VLC_LE))) {
             av_log(vc->avccontext, AV_LOG_ERROR, " Error generating vlc tables. \n");
             goto error;
         }
@@ -434,7 +453,7 @@ error:
     av_free(tmp_vlc_bits);
     av_free(tmp_vlc_codes);
     av_free(codebook_multiplicands);
-    return -1;
+    return ret;
 }
 
 // Process time domain transforms part (unused in Vorbis I)
@@ -452,7 +471,7 @@ static int vorbis_parse_setup_hdr_tdtransforms(vorbis_context *vc)
 
         if (vorbis_tdtransform) {
             av_log(vc->avccontext, AV_LOG_ERROR, "Vorbis time domain transform data nonzero. \n");
-            return -1;
+            return AVERROR_INVALIDDATA;
         }
     }
     return 0;
@@ -544,7 +563,7 @@ static int vorbis_parse_setup_hdr_floors(vorbis_context *vc)
                 av_log(vc->avccontext, AV_LOG_ERROR,
                        "Floor value is too large for blocksize: %u (%"PRIu32")\n",
                        rangemax, vc->blocksize[1] / 2);
-                return -1;
+                return AVERROR_INVALIDDATA;
             }
             floor_setup->data.t1.list[0].x = 0;
             floor_setup->data.t1.list[1].x = rangemax;
@@ -559,7 +578,11 @@ static int vorbis_parse_setup_hdr_floors(vorbis_context *vc)
             }
 
 // Precalculate order of x coordinates - needed for decode
-            ff_vorbis_ready_floor1_list(floor_setup->data.t1.list, floor_setup->data.t1.x_list_dim);
+            if (ff_vorbis_ready_floor1_list(vc->avccontext,
+                                            floor_setup->data.t1.list,
+                                            floor_setup->data.t1.x_list_dim)) {
+                return AVERROR_INVALIDDATA;
+            }
         } else if (floor_setup->floor_type == 0) {
             unsigned max_codebook_dim = 0;
 
@@ -574,7 +597,7 @@ static int vorbis_parse_setup_hdr_floors(vorbis_context *vc)
             if (floor_setup->data.t0.amplitude_bits == 0) {
                 av_log(vc->avccontext, AV_LOG_ERROR,
                        "Floor 0 amplitude bits is 0.\n");
-                return -1;
+                return AVERROR_INVALIDDATA;
             }
             floor_setup->data.t0.amplitude_offset = get_bits(gb, 8);
             floor_setup->data.t0.num_books        = get_bits(gb, 4) + 1;
@@ -583,7 +606,7 @@ static int vorbis_parse_setup_hdr_floors(vorbis_context *vc)
             floor_setup->data.t0.book_list =
                 av_malloc(floor_setup->data.t0.num_books);
             if (!floor_setup->data.t0.book_list)
-                return -1;
+                return AVERROR(ENOMEM);
             /* read book indexes */
             {
                 int idx;
@@ -604,7 +627,7 @@ static int vorbis_parse_setup_hdr_floors(vorbis_context *vc)
                 av_malloc((floor_setup->data.t0.order + 1 + max_codebook_dim)
                           * sizeof(*floor_setup->data.t0.lsp));
             if (!floor_setup->data.t0.lsp)
-                return -1;
+                return AVERROR(ENOMEM);
 
             /* debug output parsed headers */
             av_dlog(NULL, "floor0 order: %u\n", floor_setup->data.t0.order);
@@ -628,7 +651,7 @@ static int vorbis_parse_setup_hdr_floors(vorbis_context *vc)
             }
         } else {
             av_log(vc->avccontext, AV_LOG_ERROR, "Invalid floor type!\n");
-            return -1;
+            return AVERROR_INVALIDDATA;
         }
     }
     return 0;
@@ -660,13 +683,13 @@ static int vorbis_parse_setup_hdr_residues(vorbis_context *vc)
         res_setup->partition_size = get_bits(gb, 24) + 1;
         /* Validations to prevent a buffer overflow later. */
         if (res_setup->begin>res_setup->end ||
-            res_setup->end > vc->avccontext->channels * vc->blocksize[1] / 2 ||
+            res_setup->end > (res_setup->type == 2 ? vc->audio_channels : 1) * vc->blocksize[1] / 2 ||
             (res_setup->end-res_setup->begin) / res_setup->partition_size > V_MAX_PARTITIONS) {
             av_log(vc->avccontext, AV_LOG_ERROR,
                    "partition out of bounds: type, begin, end, size, blocksize: %"PRIu16", %"PRIu32", %"PRIu32", %u, %"PRIu32"\n",
                    res_setup->type, res_setup->begin, res_setup->end,
                    res_setup->partition_size, vc->blocksize[1] / 2);
-            return -1;
+            return AVERROR_INVALIDDATA;
         }
 
         res_setup->classifications = get_bits(gb, 6) + 1;
@@ -731,7 +754,7 @@ static int vorbis_parse_setup_hdr_mappings(vorbis_context *vc)
 
         if (get_bits(gb, 16)) {
             av_log(vc->avccontext, AV_LOG_ERROR, "Other mappings than type 0 are not compliant with the Vorbis I specification. \n");
-            return -1;
+            return AVERROR_INVALIDDATA;
         }
         if (get_bits1(gb)) {
             mapping_setup->submaps = get_bits(gb, 4) + 1;
@@ -758,7 +781,7 @@ static int vorbis_parse_setup_hdr_mappings(vorbis_context *vc)
 
         if (get_bits(gb, 2)) {
             av_log(vc->avccontext, AV_LOG_ERROR, "%u. mapping setup data invalid.\n", i);
-            return -1; // following spec.
+            return AVERROR_INVALIDDATA; // following spec.
         }
 
         if (mapping_setup->submaps>1) {
@@ -801,8 +824,7 @@ static void create_map(vorbis_context *vc, unsigned floor_number)
 
         for (idx = 0; idx < n; ++idx) {
             map[idx] = floor(BARK((vf->rate * idx) / (2.0f * n)) *
-                             ((vf->bark_map_size) /
-                              BARK(vf->rate / 2.0f)));
+                             (vf->bark_map_size / BARK(vf->rate / 2.0f)));
             if (vf->bark_map_size-1 < map[idx])
                 map[idx] = vf->bark_map_size - 1;
         }
@@ -845,41 +867,42 @@ static int vorbis_parse_setup_hdr_modes(vorbis_context *vc)
 static int vorbis_parse_setup_hdr(vorbis_context *vc)
 {
     GetBitContext *gb = &vc->gb;
+    int ret;
 
     if ((get_bits(gb, 8) != 'v') || (get_bits(gb, 8) != 'o') ||
         (get_bits(gb, 8) != 'r') || (get_bits(gb, 8) != 'b') ||
         (get_bits(gb, 8) != 'i') || (get_bits(gb, 8) != 's')) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (no vorbis signature). \n");
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
 
-    if (vorbis_parse_setup_hdr_codebooks(vc)) {
+    if ((ret = vorbis_parse_setup_hdr_codebooks(vc))) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (codebooks). \n");
-        return -2;
+        return ret;
     }
-    if (vorbis_parse_setup_hdr_tdtransforms(vc)) {
+    if ((ret = vorbis_parse_setup_hdr_tdtransforms(vc))) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (time domain transforms). \n");
-        return -3;
+        return ret;
     }
-    if (vorbis_parse_setup_hdr_floors(vc)) {
+    if ((ret = vorbis_parse_setup_hdr_floors(vc))) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (floors). \n");
-        return -4;
+        return ret;
     }
-    if (vorbis_parse_setup_hdr_residues(vc)) {
+    if ((ret = vorbis_parse_setup_hdr_residues(vc))) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (residues). \n");
-        return -5;
+        return ret;
     }
-    if (vorbis_parse_setup_hdr_mappings(vc)) {
+    if ((ret = vorbis_parse_setup_hdr_mappings(vc))) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (mappings). \n");
-        return -6;
+        return ret;
     }
-    if (vorbis_parse_setup_hdr_modes(vc)) {
+    if ((ret = vorbis_parse_setup_hdr_modes(vc))) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (modes). \n");
-        return -7;
+        return ret;
     }
     if (!get_bits1(gb)) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis setup header packet corrupt (framing flag). \n");
-        return -8; // framing flag bit unset error
+        return AVERROR_INVALIDDATA; // framing flag bit unset error
     }
 
     return 0;
@@ -896,43 +919,37 @@ static int vorbis_parse_id_hdr(vorbis_context *vc)
         (get_bits(gb, 8) != 'r') || (get_bits(gb, 8) != 'b') ||
         (get_bits(gb, 8) != 'i') || (get_bits(gb, 8) != 's')) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis id header packet corrupt (no vorbis signature). \n");
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
 
     vc->version        = get_bits_long(gb, 32);    //FIXME check 0
     vc->audio_channels = get_bits(gb, 8);
     if (vc->audio_channels <= 0) {
         av_log(vc->avccontext, AV_LOG_ERROR, "Invalid number of channels\n");
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
     vc->audio_samplerate = get_bits_long(gb, 32);
     if (vc->audio_samplerate <= 0) {
         av_log(vc->avccontext, AV_LOG_ERROR, "Invalid samplerate\n");
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
     vc->bitrate_maximum = get_bits_long(gb, 32);
     vc->bitrate_nominal = get_bits_long(gb, 32);
     vc->bitrate_minimum = get_bits_long(gb, 32);
     bl0 = get_bits(gb, 4);
     bl1 = get_bits(gb, 4);
-    vc->blocksize[0] = (1 << bl0);
-    vc->blocksize[1] = (1 << bl1);
     if (bl0 > 13 || bl0 < 6 || bl1 > 13 || bl1 < 6 || bl1 < bl0) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis id header packet corrupt (illegal blocksize). \n");
-        return -3;
-    }
-    // output format int16
-    if (vc->blocksize[1] / 2 * vc->audio_channels * 2 > AVCODEC_MAX_AUDIO_FRAME_SIZE) {
-        av_log(vc->avccontext, AV_LOG_ERROR, "Vorbis channel count makes "
-               "output packets too large.\n");
-        return -4;
+        return AVERROR_INVALIDDATA;
     }
+    vc->blocksize[0] = (1 << bl0);
+    vc->blocksize[1] = (1 << bl1);
     vc->win[0] = ff_vorbis_vwin[bl0 - 6];
     vc->win[1] = ff_vorbis_vwin[bl1 - 6];
 
     if ((get_bits1(gb)) == 0) {
         av_log(vc->avccontext, AV_LOG_ERROR, " Vorbis id header packet corrupt (framing flag not set). \n");
-        return -2;
+        return AVERROR_INVALIDDATA;
     }
 
     vc->channel_residues =  av_malloc((vc->blocksize[1]  / 2) * vc->audio_channels * sizeof(*vc->channel_residues));
@@ -960,16 +977,17 @@ static int vorbis_parse_id_hdr(vorbis_context *vc)
 
 static av_cold int vorbis_decode_init(AVCodecContext *avccontext)
 {
-    vorbis_context *vc = avccontext->priv_data ;
+    vorbis_context *vc = avccontext->priv_data;
     uint8_t *headers   = avccontext->extradata;
     int headers_len    = avccontext->extradata_size;
     uint8_t *header_start[3];
     int header_len[3];
-    GetBitContext *gb = &(vc->gb);
-    int hdr_type;
+    GetBitContext *gb = &vc->gb;
+    int hdr_type, ret;
 
     vc->avccontext = avccontext;
-    dsputil_init(&vc->dsp, avccontext);
+    ff_dsputil_init(&vc->dsp, avccontext);
+    avpriv_float_dsp_init(&vc->fdsp, avccontext->flags & CODEC_FLAG_BITEXACT);
     ff_fmt_convert_init(&vc->fmt_conv, avccontext);
 
     if (avccontext->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
@@ -982,24 +1000,24 @@ static av_cold int vorbis_decode_init(AVCodecContext *avccontext)
 
     if (!headers_len) {
         av_log(avccontext, AV_LOG_ERROR, "Extradata missing.\n");
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
 
-    if (ff_split_xiph_headers(headers, headers_len, 30, header_start, header_len) < 0) {
+    if ((ret = avpriv_split_xiph_headers(headers, headers_len, 30, header_start, header_len)) < 0) {
         av_log(avccontext, AV_LOG_ERROR, "Extradata corrupt.\n");
-        return -1;
+        return ret;
     }
 
     init_get_bits(gb, header_start[0], header_len[0]*8);
     hdr_type = get_bits(gb, 8);
     if (hdr_type != 1) {
         av_log(avccontext, AV_LOG_ERROR, "First header is not the id header.\n");
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
-    if (vorbis_parse_id_hdr(vc)) {
+    if ((ret = vorbis_parse_id_hdr(vc))) {
         av_log(avccontext, AV_LOG_ERROR, "Id header corrupt.\n");
         vorbis_free(vc);
-        return -1;
+        return ret;
     }
 
     init_get_bits(gb, header_start[2], header_len[2]*8);
@@ -1007,12 +1025,12 @@ static av_cold int vorbis_decode_init(AVCodecContext *avccontext)
     if (hdr_type != 5) {
         av_log(avccontext, AV_LOG_ERROR, "Third header is not the setup header.\n");
         vorbis_free(vc);
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
-    if (vorbis_parse_setup_hdr(vc)) {
+    if ((ret = vorbis_parse_setup_hdr(vc))) {
         av_log(avccontext, AV_LOG_ERROR, "Setup header corrupt.\n");
         vorbis_free(vc);
-        return -1;
+        return ret;
     }
 
     if (vc->audio_channels > 8)
@@ -1022,9 +1040,11 @@ static av_cold int vorbis_decode_init(AVCodecContext *avccontext)
 
     avccontext->channels    = vc->audio_channels;
     avccontext->sample_rate = vc->audio_samplerate;
-    avccontext->frame_size  = FFMIN(vc->blocksize[0], vc->blocksize[1]) >> 2;
 
-    return 0 ;
+    avcodec_get_frame_defaults(&vc->frame);
+    avccontext->coded_frame = &vc->frame;
+
+    return 0;
 }
 
 // Decode audiopackets -------------------------------------------------
@@ -1055,7 +1075,7 @@ static int vorbis_floor0_decode(vorbis_context *vc,
         codebook = vc->codebooks[vf->book_list[book_idx]];
         /* Invalid codebook! */
         if (!codebook.codevectors)
-            return -1;
+            return AVERROR_INVALIDDATA;
 
         while (lsp_len<vf->order) {
             int vec_off;
@@ -1232,20 +1252,20 @@ static int vorbis_floor1_decode(vorbis_context *vc,
             floor1_flag[i]               = 1;
             if (val >= room) {
                 if (highroom > lowroom) {
-                    floor1_Y_final[i] = val - lowroom + predicted;
+                    floor1_Y_final[i] = av_clip_uint16(val - lowroom + predicted);
                 } else {
-                    floor1_Y_final[i] = predicted - val + highroom - 1;
+                    floor1_Y_final[i] = av_clip_uint16(predicted - val + highroom - 1);
                 }
             } else {
                 if (val & 1) {
-                    floor1_Y_final[i] = predicted - (val + 1) / 2;
+                    floor1_Y_final[i] = av_clip_uint16(predicted - (val + 1) / 2);
                 } else {
-                    floor1_Y_final[i] = predicted + val / 2;
+                    floor1_Y_final[i] = av_clip_uint16(predicted + val / 2);
                 }
             }
         } else {
             floor1_flag[i]    = 0;
-            floor1_Y_final[i] = predicted;
+            floor1_Y_final[i] = av_clip_uint16(predicted);
         }
 
         av_dlog(NULL, " Decoded floor(%d) = %u / val %u\n",
@@ -1269,6 +1289,7 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
                                                            uint8_t *do_not_decode,
                                                            float *vec,
                                                            unsigned vlen,
+                                                           unsigned ch_left,
                                                            int vr_type)
 {
     GetBitContext *gb = &vc->gb;
@@ -1276,6 +1297,7 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
     unsigned ptns_to_read = vr->ptns_to_read;
     uint8_t *classifs = vr->classifs;
     unsigned pass, ch_used, i, j, k, l;
+    unsigned max_output = (ch - 1) * vlen;
 
     if (vr_type == 2) {
         for (j = 1; j < ch; ++j)
@@ -1283,8 +1305,15 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
         if (do_not_decode[0])
             return 0;
         ch_used = 1;
+        max_output += vr->end / ch;
     } else {
         ch_used = ch;
+        max_output += vr->end;
+    }
+
+    if (max_output > ch_left * vlen) {
+        av_log(vc->avccontext, AV_LOG_ERROR, "Insufficient output buffer\n");
+        return -1;
     }
 
     av_dlog(NULL, " residue type 0/1/2 decode begin, ch: %d  cpc %d  \n", ch, c_p_c);
@@ -1303,7 +1332,7 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
 
                         av_dlog(NULL, "Classword: %u\n", temp);
 
-                        assert(vr->classifications > 1 && temp <= 65536); //needed for inverse[]
+                        av_assert0(vr->classifications > 1 && temp <= 65536); //needed for inverse[]
                         for (i = 0; i < c_p_c; ++i) {
                             unsigned temp2;
 
@@ -1327,8 +1356,7 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
                         if (vqbook >= 0 && vc->codebooks[vqbook].codevectors) {
                             unsigned coffs;
                             unsigned dim  = vc->codebooks[vqbook].dimensions;
-                            unsigned step = dim == 1 ? vr->partition_size
-                                                     : FASTDIV(vr->partition_size, dim);
+                            unsigned step = FASTDIV(vr->partition_size << 1, dim << 1);
                             vorbis_codebook codebook = vc->codebooks[vqbook];
 
                             if (vr_type == 0) {
@@ -1337,14 +1365,14 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
                                 for (k = 0; k < step; ++k) {
                                     coffs = get_vlc2(gb, codebook.vlc.table, codebook.nb_bits, 3) * dim;
                                     for (l = 0; l < dim; ++l)
-                                        vec[voffs + k + l * step] += codebook.codevectors[coffs + l];  // FPMATH
+                                        vec[voffs + k + l * step] += codebook.codevectors[coffs + l];
                                 }
                             } else if (vr_type == 1) {
                                 voffs = voffset + j * vlen;
                                 for (k = 0; k < step; ++k) {
                                     coffs = get_vlc2(gb, codebook.vlc.table, codebook.nb_bits, 3) * dim;
                                     for (l = 0; l < dim; ++l, ++voffs) {
-                                        vec[voffs]+=codebook.codevectors[coffs+l];  // FPMATH
+                                        vec[voffs]+=codebook.codevectors[coffs+l];
 
                                         av_dlog(NULL, " pass %d offs: %d curr: %f change: %f cv offs.: %d  \n",
                                                 pass, voffs, vec[voffs], codebook.codevectors[coffs+l], coffs);
@@ -1356,23 +1384,23 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
                                 if (dim == 2) {
                                     for (k = 0; k < step; ++k) {
                                         coffs = get_vlc2(gb, codebook.vlc.table, codebook.nb_bits, 3) * 2;
-                                        vec[voffs + k       ] += codebook.codevectors[coffs    ];  // FPMATH
-                                        vec[voffs + k + vlen] += codebook.codevectors[coffs + 1];  // FPMATH
+                                        vec[voffs + k       ] += codebook.codevectors[coffs    ];
+                                        vec[voffs + k + vlen] += codebook.codevectors[coffs + 1];
                                     }
                                 } else if (dim == 4) {
                                     for (k = 0; k < step; ++k, voffs += 2) {
                                         coffs = get_vlc2(gb, codebook.vlc.table, codebook.nb_bits, 3) * 4;
-                                        vec[voffs           ] += codebook.codevectors[coffs    ];  // FPMATH
-                                        vec[voffs + 1       ] += codebook.codevectors[coffs + 2];  // FPMATH
-                                        vec[voffs + vlen    ] += codebook.codevectors[coffs + 1];  // FPMATH
-                                        vec[voffs + vlen + 1] += codebook.codevectors[coffs + 3];  // FPMATH
+                                        vec[voffs           ] += codebook.codevectors[coffs    ];
+                                        vec[voffs + 1       ] += codebook.codevectors[coffs + 2];
+                                        vec[voffs + vlen    ] += codebook.codevectors[coffs + 1];
+                                        vec[voffs + vlen + 1] += codebook.codevectors[coffs + 3];
                                     }
                                 } else
                                 for (k = 0; k < step; ++k) {
                                     coffs = get_vlc2(gb, codebook.vlc.table, codebook.nb_bits, 3) * dim;
                                     for (l = 0; l < dim; l += 2, voffs++) {
-                                        vec[voffs       ] += codebook.codevectors[coffs + l    ];  // FPMATH
-                                        vec[voffs + vlen] += codebook.codevectors[coffs + l + 1];  // FPMATH
+                                        vec[voffs       ] += codebook.codevectors[coffs + l    ];
+                                        vec[voffs + vlen] += codebook.codevectors[coffs + l + 1];
 
                                         av_dlog(NULL, " pass %d offs: %d curr: %f change: %f cv offs.: %d+%d  \n",
                                                 pass, voffset / ch + (voffs % ch) * vlen,
@@ -1382,17 +1410,24 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
                                 }
 
                             } else if (vr_type == 2) {
-                                voffs = voffset;
+                                unsigned voffs_div = FASTDIV(voffset << 1, ch <<1);
+                                unsigned voffs_mod = voffset - voffs_div * ch;
 
                                 for (k = 0; k < step; ++k) {
                                     coffs = get_vlc2(gb, codebook.vlc.table, codebook.nb_bits, 3) * dim;
-                                    for (l = 0; l < dim; ++l, ++voffs) {
-                                        vec[voffs / ch + (voffs % ch) * vlen] += codebook.codevectors[coffs + l];  // FPMATH FIXME use if and counter instead of / and %
+                                    for (l = 0; l < dim; ++l) {
+                                        vec[voffs_div + voffs_mod * vlen] +=
+                                            codebook.codevectors[coffs + l];
 
                                         av_dlog(NULL, " pass %d offs: %d curr: %f change: %f cv offs.: %d+%d  \n",
-                                                pass, voffset / ch + (voffs % ch) * vlen,
-                                                vec[voffset / ch + (voffs % ch) * vlen],
+                                                pass, voffs_div + voffs_mod * vlen,
+                                                vec[voffs_div + voffs_mod * vlen],
                                                 codebook.codevectors[coffs + l], coffs, l);
+
+                                        if (++voffs_mod == ch) {
+                                            voffs_div++;
+                                            voffs_mod = 0;
+                                        }
                                     }
                                 }
                             }
@@ -1411,21 +1446,22 @@ static av_always_inline int vorbis_residue_decode_internal(vorbis_context *vc,
 static inline int vorbis_residue_decode(vorbis_context *vc, vorbis_residue *vr,
                                         unsigned ch,
                                         uint8_t *do_not_decode,
-                                        float *vec, unsigned vlen)
+                                        float *vec, unsigned vlen,
+                                        unsigned ch_left)
 {
     if (vr->type == 2)
-        return vorbis_residue_decode_internal(vc, vr, ch, do_not_decode, vec, vlen, 2);
+        return vorbis_residue_decode_internal(vc, vr, ch, do_not_decode, vec, vlen, ch_left, 2);
     else if (vr->type == 1)
-        return vorbis_residue_decode_internal(vc, vr, ch, do_not_decode, vec, vlen, 1);
+        return vorbis_residue_decode_internal(vc, vr, ch, do_not_decode, vec, vlen, ch_left, 1);
     else if (vr->type == 0)
-        return vorbis_residue_decode_internal(vc, vr, ch, do_not_decode, vec, vlen, 0);
+        return vorbis_residue_decode_internal(vc, vr, ch, do_not_decode, vec, vlen, ch_left, 0);
     else {
         av_log(vc->avccontext, AV_LOG_ERROR, " Invalid residue type while residue decode?! \n");
-        return -1;
+        return AVERROR_INVALIDDATA;
     }
 }
 
-void vorbis_inverse_coupling(float *mag, float *ang, int blocksize)
+void ff_vorbis_inverse_coupling(float *mag, float *ang, int blocksize)
 {
     int i;
     for (i = 0;  i < blocksize;  i++) {
@@ -1466,10 +1502,12 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
     uint8_t res_chan[255];
     unsigned res_num = 0;
     int retlen  = 0;
+    unsigned ch_left = vc->audio_channels;
+    unsigned vlen;
 
     if (get_bits1(gb)) {
         av_log(vc->avccontext, AV_LOG_ERROR, "Not a Vorbis I audio packet.\n");
-        return -1; // packet type not audio
+        return AVERROR_INVALIDDATA; // packet type not audio
     }
 
     if (vc->mode_count == 1) {
@@ -1485,11 +1523,14 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
 
     blockflag = vc->modes[mode_number].blockflag;
     blocksize = vc->blocksize[blockflag];
-    if (blockflag)
-        skip_bits(gb, 2); // previous_window, next_window
+    vlen = blocksize / 2;
+    if (blockflag) {
+        previous_window = get_bits(gb, 1);
+        skip_bits1(gb); // next_window
+    }
 
-    memset(ch_res_ptr,   0, sizeof(float) * vc->audio_channels * blocksize / 2); //FIXME can this be removed ?
-    memset(ch_floor_ptr, 0, sizeof(float) * vc->audio_channels * blocksize / 2); //FIXME can this be removed ?
+    memset(ch_res_ptr,   0, sizeof(float) * vc->audio_channels * vlen); //FIXME can this be removed ?
+    memset(ch_floor_ptr, 0, sizeof(float) * vc->audio_channels * vlen); //FIXME can this be removed ?
 
 // Decode floor
 
@@ -1506,10 +1547,10 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
 
         if (ret < 0) {
             av_log(vc->avccontext, AV_LOG_ERROR, "Invalid codebook in vorbis_floor_decode.\n");
-            return -1;
+            return AVERROR_INVALIDDATA;
         }
         no_residue[i] = ret;
-        ch_floor_ptr += blocksize / 2;
+        ch_floor_ptr += vlen;
     }
 
 // Nonzero vector propagate
@@ -1526,6 +1567,7 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
     for (i = 0; i < mapping->submaps; ++i) {
         vorbis_residue *residue;
         unsigned ch = 0;
+        int ret;
 
         for (j = 0; j < vc->audio_channels; ++j) {
             if ((mapping->submaps == 1) || (i == mapping->mux[j])) {
@@ -1540,11 +1582,23 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
             }
         }
         residue = &vc->residues[mapping->submap_residue[i]];
-        vorbis_residue_decode(vc, residue, ch, do_not_decode, ch_res_ptr, blocksize/2);
+        if (ch_left < ch) {
+            av_log(vc->avccontext, AV_LOG_ERROR, "Too many channels in vorbis_floor_decode.\n");
+            return -1;
+        }
+        if (ch) {
+            ret = vorbis_residue_decode(vc, residue, ch, do_not_decode, ch_res_ptr, vlen, ch_left);
+            if (ret < 0)
+                return ret;
+        }
 
-        ch_res_ptr += ch * blocksize / 2;
+        ch_res_ptr += ch * vlen;
+        ch_left -= ch;
     }
 
+    if (ch_left > 0)
+        return AVERROR_INVALIDDATA;
+
 // Inverse coupling
 
     for (i = mapping->coupling_steps - 1; i >= 0; --i) { //warning: i has to be signed
@@ -1562,11 +1616,11 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
     for (j = vc->audio_channels-1;j >= 0; j--) {
         ch_floor_ptr = vc->channel_floors   + j           * blocksize / 2;
         ch_res_ptr   = vc->channel_residues + res_chan[j] * blocksize / 2;
-        vc->dsp.vector_fmul(ch_floor_ptr, ch_floor_ptr, ch_res_ptr, blocksize / 2);
+        vc->fdsp.vector_fmul(ch_floor_ptr, ch_floor_ptr, ch_res_ptr, blocksize / 2);
         mdct->imdct_half(mdct, ch_res_ptr, ch_floor_ptr);
     }
 
-// Overlap/add, save data for next overlapping  FPMATH
+// Overlap/add, save data for next overlapping
 
     retlen = (blocksize + vc->blocksize[previous_window]) / 4;
     for (j = 0; j < vc->audio_channels; j++) {
@@ -1596,40 +1650,39 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
 
 // Return the decoded audio packet through the standard api
 
-static int vorbis_decode_frame(AVCodecContext *avccontext,
-                               void *data, int *data_size,
-                               AVPacket *avpkt)
+static int vorbis_decode_frame(AVCodecContext *avccontext, void *data,
+                               int *got_frame_ptr, AVPacket *avpkt)
 {
     const uint8_t *buf = avpkt->data;
     int buf_size       = avpkt->size;
-    vorbis_context *vc = avccontext->priv_data ;
-    GetBitContext *gb = &(vc->gb);
+    vorbis_context *vc = avccontext->priv_data;
+    GetBitContext *gb = &vc->gb;
     const float *channel_ptrs[255];
-    int i, len;
-
-    if (!buf_size)
-        return 0;
+    int i, len, ret;
 
     av_dlog(NULL, "packet length %d \n", buf_size);
 
     init_get_bits(gb, buf, buf_size*8);
 
-    len = vorbis_parse_audio_packet(vc);
-
-    if (len <= 0) {
-        *data_size = 0;
-        return buf_size;
-    }
+    if ((len = vorbis_parse_audio_packet(vc)) <= 0)
+        return len;
 
     if (!vc->first_frame) {
         vc->first_frame = 1;
-        *data_size = 0;
-        return buf_size ;
+        *got_frame_ptr = 0;
+        return buf_size;
     }
 
     av_dlog(NULL, "parsed %d bytes %d bits, returned %d samples (*ch*bits) \n",
             get_bits_count(gb) / 8, get_bits_count(gb) % 8, len);
 
+    /* get output buffer */
+    vc->frame.nb_samples = len;
+    if ((ret = avccontext->get_buffer(avccontext, &vc->frame)) < 0) {
+        av_log(avccontext, AV_LOG_ERROR, "get_buffer() failed\n");
+        return ret;
+    }
+
     if (vc->audio_channels > 8) {
         for (i = 0; i < vc->audio_channels; i++)
             channel_ptrs[i] = vc->channel_floors + i * len;
@@ -1640,15 +1693,17 @@ static int vorbis_decode_frame(AVCodecContext *avccontext,
     }
 
     if (avccontext->sample_fmt == AV_SAMPLE_FMT_FLT)
-        vc->fmt_conv.float_interleave(data, channel_ptrs, len, vc->audio_channels);
+        vc->fmt_conv.float_interleave((float *)vc->frame.data[0], channel_ptrs,
+                                      len, vc->audio_channels);
     else
-        vc->fmt_conv.float_to_int16_interleave(data, channel_ptrs, len,
+        vc->fmt_conv.float_to_int16_interleave((int16_t *)vc->frame.data[0],
+                                               channel_ptrs, len,
                                                vc->audio_channels);
 
-    *data_size = len * vc->audio_channels *
-                 av_get_bytes_per_sample(avccontext->sample_fmt);
+    *got_frame_ptr   = 1;
+    *(AVFrame *)data = vc->frame;
 
-    return buf_size ;
+    return buf_size;
 }
 
 // Close decoder
@@ -1659,22 +1714,33 @@ static av_cold int vorbis_decode_close(AVCodecContext *avccontext)
 
     vorbis_free(vc);
 
-    return 0 ;
+    return 0;
+}
+
+static av_cold void vorbis_decode_flush(AVCodecContext *avccontext)
+{
+    vorbis_context *vc = avccontext->priv_data;
+
+    if (vc->saved) {
+        memset(vc->saved, 0, (vc->blocksize[1] / 4) * vc->audio_channels *
+                             sizeof(*vc->saved));
+    }
+    vc->previous_window = 0;
 }
 
 AVCodec ff_vorbis_decoder = {
-    "vorbis",
-    AVMEDIA_TYPE_AUDIO,
-    CODEC_ID_VORBIS,
-    sizeof(vorbis_context),
-    vorbis_decode_init,
-    NULL,
-    vorbis_decode_close,
-    vorbis_decode_frame,
-    .long_name = NULL_IF_CONFIG_SMALL("Vorbis"),
+    .name            = "vorbis",
+    .type            = AVMEDIA_TYPE_AUDIO,
+    .id              = AV_CODEC_ID_VORBIS,
+    .priv_data_size  = sizeof(vorbis_context),
+    .init            = vorbis_decode_init,
+    .close           = vorbis_decode_close,
+    .decode          = vorbis_decode_frame,
+    .flush           = vorbis_decode_flush,
+    .capabilities    = CODEC_CAP_DR1,
+    .long_name       = NULL_IF_CONFIG_SMALL("Vorbis"),
     .channel_layouts = ff_vorbis_channel_layouts,
-    .sample_fmts = (const enum AVSampleFormat[]) {
+    .sample_fmts     = (const enum AVSampleFormat[]) {
         AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
     },
 };
-