]> git.sesse.net Git - ffmpeg/blobdiff - libavformat/mxfenc.c
Support decoding Avid 1:1x codec, (partially) fixes issue1684.
[ffmpeg] / libavformat / mxfenc.c
index d9d8d22afa639903b6fd1ad47a67b0dd64a19b87..12dedccbbde400adf1bc3061ba44b18bc6f93f53 100644 (file)
@@ -35,7 +35,7 @@
 #include <math.h>
 #include <time.h>
 
-#include "libavutil/fifo.h"
+#include "libavutil/random_seed.h"
 #include "libavcodec/bytestream.h"
 #include "audiointerleave.h"
 #include "avformat.h"
@@ -63,12 +63,13 @@ typedef struct {
 typedef struct {
     AudioInterleaveContext aic;
     UID track_essence_element_key;
-    int index;            ///< index in mxf_essence_container_uls table
+    int index;               ///< index in mxf_essence_container_uls table
     const UID *codec_ul;
-    int order;            ///< interleaving order if dts are equal
-    int interlaced;       ///< wether picture is interlaced
+    int order;               ///< interleaving order if dts are equal
+    int interlaced;          ///< wether picture is interlaced
     int temporal_reordering;
     AVRational aspect_ratio; ///< display aspect ratio
+    int closed_gop;          ///< gop is closed, used in mpeg-2 frame parsing
 } MXFStreamContext;
 
 typedef struct {
@@ -85,7 +86,7 @@ static const struct {
     { CODEC_ID_MPEG2VIDEO, 0 },
     { CODEC_ID_PCM_S24LE,  1 },
     { CODEC_ID_PCM_S16LE,  1 },
-    { 0 }
+    { CODEC_ID_NONE }
 };
 
 static void mxf_write_wav_desc(AVFormatContext *s, AVStream *st);
@@ -187,15 +188,17 @@ typedef struct MXFContext {
     int timecode_drop_frame; ///< time code use drop frame method frop mpeg-2 essence gop header
     int edit_unit_byte_count; ///< fixed edit unit byte count
     uint64_t body_offset;
+    uint32_t instance_number;
+    uint8_t umid[16];        ///< unique material identifier
 } MXFContext;
 
 static const uint8_t uuid_base[]            = { 0xAD,0xAB,0x44,0x24,0x2f,0x25,0x4d,0xc7,0x92,0xff,0x29,0xbd };
-static const uint8_t umid_base[]            = { 0x06,0x0A,0x2B,0x34,0x01,0x01,0x01,0x05,0x01,0x01,0x0D,0x00,0x13,0x00,0x00,0x00 };
+static const uint8_t umid_ul[]              = { 0x06,0x0A,0x2B,0x34,0x01,0x01,0x01,0x05,0x01,0x01,0x0D,0x00,0x13 };
 
 /**
  * complete key for operation pattern, partitions, and primer pack
  */
-static const uint8_t op1a_ul[]                     = { 0x06,0x0E,0x2B,0x34,0x04,0x01,0x01,0x01,0x0D,0x01,0x02,0x01,0x01,0x01,0x01,0x00 };
+static const uint8_t op1a_ul[]                     = { 0x06,0x0E,0x2B,0x34,0x04,0x01,0x01,0x01,0x0D,0x01,0x02,0x01,0x01,0x01,0x09,0x00 };
 static const uint8_t footer_partition_key[]        = { 0x06,0x0E,0x2B,0x34,0x02,0x05,0x01,0x01,0x0D,0x01,0x02,0x01,0x01,0x04,0x04,0x00 }; // ClosedComplete
 static const uint8_t primer_pack_key[]             = { 0x06,0x0E,0x2B,0x34,0x02,0x05,0x01,0x01,0x0D,0x01,0x02,0x01,0x01,0x05,0x01,0x00 };
 static const uint8_t index_table_segment_key[]     = { 0x06,0x0E,0x2B,0x34,0x02,0x53,0x01,0x01,0x0d,0x01,0x02,0x01,0x01,0x10,0x01,0x00 };
@@ -295,6 +298,7 @@ static const MXFLocalTagPair mxf_local_tag_batch[] = {
     { 0x3F0A, {0x06,0x0E,0x2B,0x34,0x01,0x01,0x01,0x05,0x04,0x04,0x04,0x02,0x05,0x00,0x00,0x00}}, /* Index Entry Array */
     // MPEG video Descriptor
     { 0x8000, {0x06,0x0E,0x2B,0x34,0x01,0x01,0x01,0x05,0x04,0x01,0x06,0x02,0x01,0x0B,0x00,0x00}}, /* BitRate */
+    { 0x8007, {0x06,0x0E,0x2B,0x34,0x01,0x01,0x01,0x05,0x04,0x01,0x06,0x02,0x01,0x0A,0x00,0x00}}, /* ProfileAndLevel */
     // Wave Audio Essence Descriptor
     { 0x3D09, {0x06,0x0E,0x2B,0x34,0x01,0x01,0x01,0x05,0x04,0x02,0x03,0x03,0x05,0x00,0x00,0x00}}, /* Average Bytes Per Second */
     { 0x3D0A, {0x06,0x0E,0x2B,0x34,0x01,0x01,0x01,0x05,0x04,0x02,0x03,0x02,0x01,0x00,0x00,0x00}}, /* Block Align */
@@ -307,10 +311,13 @@ static void mxf_write_uuid(ByteIOContext *pb, enum MXFMetadataSetType type, int
     put_be16(pb, value);
 }
 
-static void mxf_write_umid(ByteIOContext *pb, enum MXFMetadataSetType type, int value)
+static void mxf_write_umid(AVFormatContext *s, int type)
 {
-    put_buffer(pb, umid_base, 16);
-    mxf_write_uuid(pb, type, value);
+    MXFContext *mxf = s->priv_data;
+    put_buffer(s->pb, umid_ul, 13);
+    put_be24(s->pb, mxf->instance_number);
+    put_buffer(s->pb, mxf->umid, 15);
+    put_byte(s->pb, type);
 }
 
 static void mxf_write_refs_count(ByteIOContext *pb, int ref_count)
@@ -452,7 +459,7 @@ static void mxf_write_preface(AVFormatContext *s)
 
     // write version
     mxf_write_local_tag(pb, 2, 0x3B05);
-    put_be16(pb, 1);
+    put_be16(pb, 258); // v1.2
 
     // write identification_refs
     mxf_write_local_tag(pb, 16 + 8, 0x3B06);
@@ -465,12 +472,7 @@ static void mxf_write_preface(AVFormatContext *s)
 
     // operational pattern
     mxf_write_local_tag(pb, 16, 0x3B09);
-    if (s->nb_streams > 1) {
-        put_buffer(pb, op1a_ul, 14);
-        put_be16(pb, 0x0900); // multi track
-    } else {
-        put_buffer(pb, op1a_ul, 16);
-    }
+    put_buffer(pb, op1a_ul, 16);
 
     // write essence_container_refs
     mxf_write_local_tag(pb, 8 + 16 * mxf->essence_container_count, 0x3B0A);
@@ -573,7 +575,7 @@ static void mxf_write_track(AVFormatContext *s, AVStream *st, enum MXFMetadataSe
 
     // write track id
     mxf_write_local_tag(pb, 4, 0x4801);
-    put_be32(pb, st->index+1);
+    put_be32(pb, st->index+2);
 
     // write track number
     mxf_write_local_tag(pb, 4, 0x4804);
@@ -698,14 +700,14 @@ static void mxf_write_structural_component(AVFormatContext *s, AVStream *st, enu
         for (i = 0; i < 4; i++)
             put_be64(pb, 0);
     } else
-        mxf_write_umid(pb, SourcePackage, 0);
+        mxf_write_umid(s, 1);
 
     // write source track id
     mxf_write_local_tag(pb, 4, 0x1102);
     if (type == SourcePackage)
         put_be32(pb, 0);
     else
-        put_be32(pb, st->index+1);
+        put_be32(pb, st->index+2);
 }
 
 static void mxf_write_multi_descriptor(AVFormatContext *s)
@@ -752,13 +754,13 @@ static void mxf_write_generic_desc(AVFormatContext *s, AVStream *st, const UID k
     ByteIOContext *pb = s->pb;
 
     put_buffer(pb, key, 16);
-    klv_encode_ber_length(pb, size+20+8+12+20);
+    klv_encode_ber4_length(pb, size+20+8+12+20);
 
     mxf_write_local_tag(pb, 16, 0x3C0A);
     mxf_write_uuid(pb, SubDescriptor, st->index);
 
     mxf_write_local_tag(pb, 4, 0x3006);
-    put_be32(pb, st->index+1);
+    put_be32(pb, st->index+2);
 
     mxf_write_local_tag(pb, 8, 0x3001);
     put_be32(pb, mxf->time_base.den);
@@ -854,12 +856,19 @@ static void mxf_write_cdci_desc(AVFormatContext *s, AVStream *st)
 static void mxf_write_mpegvideo_desc(AVFormatContext *s, AVStream *st)
 {
     ByteIOContext *pb = s->pb;
+    int profile_and_level = (st->codec->profile<<4) | st->codec->level;
 
-    mxf_write_cdci_common(s, st, mxf_mpegvideo_descriptor_key, 8);
+    mxf_write_cdci_common(s, st, mxf_mpegvideo_descriptor_key, 8+5);
 
     // bit rate
     mxf_write_local_tag(pb, 4, 0x8000);
     put_be32(pb, st->codec->bit_rate);
+
+    // profile and level
+    mxf_write_local_tag(pb, 1, 0x8007);
+    if (!st->codec->profile)
+        profile_and_level |= 0x80; // escape bit
+    put_byte(pb, profile_and_level);
 }
 
 static void mxf_write_generic_sound_common(AVFormatContext *s, AVStream *st, const UID key, unsigned size)
@@ -937,7 +946,7 @@ static void mxf_write_package(AVFormatContext *s, enum MXFMetadataSetType type)
 
     // write package umid
     mxf_write_local_tag(pb, 32, 0x4401);
-    mxf_write_umid(pb, type, 0);
+    mxf_write_umid(s, type == SourcePackage);
     PRINT_KEY(s, "package umid second part", pb->buf_ptr - 16);
 
     // package creation date
@@ -995,7 +1004,7 @@ static int mxf_write_essence_container_data(AVFormatContext *s)
     mxf_write_uuid(pb, EssenceContainerData, 0);
 
     mxf_write_local_tag(pb, 32, 0x2701); // Linked Package UID
-    mxf_write_umid(pb, SourcePackage, 0);
+    mxf_write_umid(s, 1);
 
     mxf_write_local_tag(pb, 4, 0x3F07); // BodySID
     put_be32(pb, 1);
@@ -1054,8 +1063,8 @@ static void mxf_write_index_table_segment(AVFormatContext *s)
 
     // index edit rate
     mxf_write_local_tag(pb, 8, 0x3F0B);
-    put_be32(pb, mxf->time_base.num);
     put_be32(pb, mxf->time_base.den);
+    put_be32(pb, mxf->time_base.num);
 
     // index start position
     mxf_write_local_tag(pb, 8, 0x3F0C);
@@ -1063,7 +1072,10 @@ static void mxf_write_index_table_segment(AVFormatContext *s)
 
     // index duration
     mxf_write_local_tag(pb, 8, 0x3F0D);
-    put_be64(pb, mxf->edit_units_count);
+    if (mxf->edit_unit_byte_count)
+        put_be64(pb, 0); // index table covers whole container
+    else
+        put_be64(pb, mxf->edit_units_count);
 
     // edit unit byte count
     mxf_write_local_tag(pb, 4, 0x3F05);
@@ -1111,14 +1123,14 @@ static void mxf_write_index_table_segment(AVFormatContext *s)
         put_be32(pb, mxf->edit_units_count);  // num of entries
         put_be32(pb, 11+mxf->slice_count*4);  // size of one entry
         for (i = 0; i < mxf->edit_units_count; i++) {
+            int temporal_offset = 0;
             if (temporal_reordering) {
-                int temporal_offset = 0;
                 for (j = i+1; j < mxf->edit_units_count; j++) {
                     temporal_offset++;
                     if (mxf->index_entries[j].flags & 0x10) { // backward prediction
                         // next is not b, so is reordered
                         if (!(mxf->index_entries[i+1].flags & 0x10)) {
-                            if ((mxf->index_entries[i].flags & 0x11) == 0) // i frame
+                            if ((mxf->index_entries[i].flags & 0x11) == 0) // I frame
                                 temporal_offset = 0;
                             else
                                 temporal_offset = -temporal_offset;
@@ -1126,15 +1138,17 @@ static void mxf_write_index_table_segment(AVFormatContext *s)
                         break;
                     }
                 }
-                put_byte(pb, temporal_offset);
-            } else
-                put_byte(pb, 0);
+            }
+            put_byte(pb, temporal_offset);
+
             if (!(mxf->index_entries[i].flags & 0x33)) { // I frame
+                if (mxf->index_entries[i].flags & 0x40 && // seq header
+                    (!temporal_reordering || !temporal_offset))
+                    mxf->index_entries[i].flags |= 0x80; // random access
                 mxf->last_key_index = key_index;
                 key_index = i;
             }
-            if (mxf->index_entries[i].flags & 0x10 && // backward prediction
-                !(mxf->index_entries[key_index].flags & 0x80)) { // open gop
+            if ((mxf->index_entries[i].flags & 0x30) == 0x30) { // back and forward prediction
                 put_byte(pb, mxf->last_key_index - i);
             } else {
                 put_byte(pb, key_index - i); // key frame offset
@@ -1181,7 +1195,7 @@ static void mxf_write_partition(AVFormatContext *s, int bodysid,
         index_byte_count = 85 + 12+(s->nb_streams+1)*6 +
             12+mxf->edit_units_count*(11+mxf->slice_count*4);
     else if (mxf->edit_unit_byte_count && indexsid)
-        index_byte_count = 85;
+        index_byte_count = 80;
 
     if (index_byte_count) {
         // add encoded ber length
@@ -1234,12 +1248,7 @@ static void mxf_write_partition(AVFormatContext *s, int bodysid,
     put_be32(pb, bodysid); // bodySID
 
     // operational pattern
-    if (s->nb_streams > 1) {
-        put_buffer(pb, op1a_ul, 14);
-        put_be16(pb, 0x0900); // multi track
-    } else {
-        put_buffer(pb, op1a_ul, 16);
-    }
+    put_buffer(pb, op1a_ul, 16);
 
     // essence container
     mxf_write_essence_container_refs(s);
@@ -1278,24 +1287,18 @@ static const UID mxf_mpeg2_codec_uls[] = {
 
 static const UID *mxf_get_mpeg2_codec_ul(AVCodecContext *avctx)
 {
+    int long_gop = avctx->gop_size > 1 || avctx->has_b_frames;
+
     if (avctx->profile == 4) { // Main
         if (avctx->level == 8) // Main
-            return avctx->gop_size ?
-                &mxf_mpeg2_codec_uls[1] :
-                &mxf_mpeg2_codec_uls[0];
+            return &mxf_mpeg2_codec_uls[0+long_gop];
         else if (avctx->level == 4) // High
-            return avctx->gop_size ?
-                &mxf_mpeg2_codec_uls[5] :
-                &mxf_mpeg2_codec_uls[4];
+            return &mxf_mpeg2_codec_uls[4+long_gop];
     } else if (avctx->profile == 0) { // 422
         if (avctx->level == 5) // Main
-            return avctx->gop_size ?
-                &mxf_mpeg2_codec_uls[3] :
-                &mxf_mpeg2_codec_uls[2];
+            return &mxf_mpeg2_codec_uls[2+long_gop];
         else if (avctx->level == 2) // High
-            return avctx->gop_size ?
-                &mxf_mpeg2_codec_uls[7] :
-                &mxf_mpeg2_codec_uls[6];
+            return &mxf_mpeg2_codec_uls[6+long_gop];
     }
     return NULL;
 }
@@ -1311,8 +1314,8 @@ static int mxf_parse_mpeg2_frame(AVFormatContext *s, AVStream *st, AVPacket *pkt
 
     for(i = 0; i < pkt->size - 4; i++) {
         c = (c<<8) + pkt->data[i];
-        if (c == 0x1B5) {
-            if (i + 2 < pkt->size && (pkt->data[i+1] & 0xf0) == 0x10) { // seq ext
+        if (c == 0x1b5) {
+            if ((pkt->data[i+1] & 0xf0) == 0x10) { // seq ext
                 st->codec->profile = pkt->data[i+1] & 0x07;
                 st->codec->level   = pkt->data[i+2] >> 4;
             } else if (i + 5 < pkt->size && (pkt->data[i+1] & 0xf0) == 0x80) { // pict coding ext
@@ -1320,44 +1323,46 @@ static int mxf_parse_mpeg2_frame(AVFormatContext *s, AVStream *st, AVPacket *pkt
                 break;
             }
         } else if (c == 0x1b8) { // gop
-            if (i + 4 < pkt->size) {
-                if (pkt->data[i+4]>>6 & 0x01) // closed
+            if (pkt->data[i+4]>>6 & 0x01) { // closed
+                sc->closed_gop = 1;
+                if (*flags & 0x40) // sequence header present
                     *flags |= 0x80; // random access
-                if (!mxf->header_written) {
-                    unsigned hours   =  (pkt->data[i+1]>>2) & 0x1f;
-                    unsigned minutes = ((pkt->data[i+1] & 0x03) << 4) | (pkt->data[i+2]>>4);
-                    unsigned seconds = ((pkt->data[i+2] & 0x07) << 3) | (pkt->data[i+3]>>5);
-                    unsigned frames  = ((pkt->data[i+3] & 0x1f) << 1) | (pkt->data[i+4]>>7);
-                    mxf->timecode_drop_frame = !!(pkt->data[i+1] & 0x80);
-                    mxf->timecode_start = (hours*3600 + minutes*60 + seconds) *
-                        mxf->timecode_base + frames;
-                    if (mxf->timecode_drop_frame) {
-                        unsigned tminutes = 60 * hours + minutes;
-                        mxf->timecode_start -= 2 * (tminutes - tminutes / 10);
-                    }
-                    av_log(s, AV_LOG_DEBUG, "frame %d %d:%d:%d%c%d\n", mxf->timecode_start,
-                           hours, minutes, seconds, mxf->timecode_drop_frame ? ';':':', frames);
+            }
+            if (!mxf->header_written) {
+                unsigned hours   =  (pkt->data[i+1]>>2) & 0x1f;
+                unsigned minutes = ((pkt->data[i+1] & 0x03) << 4) | (pkt->data[i+2]>>4);
+                unsigned seconds = ((pkt->data[i+2] & 0x07) << 3) | (pkt->data[i+3]>>5);
+                unsigned frames  = ((pkt->data[i+3] & 0x1f) << 1) | (pkt->data[i+4]>>7);
+                mxf->timecode_drop_frame = !!(pkt->data[i+1] & 0x80);
+                mxf->timecode_start = (hours*3600 + minutes*60 + seconds) *
+                    mxf->timecode_base + frames;
+                if (mxf->timecode_drop_frame) {
+                    unsigned tminutes = 60 * hours + minutes;
+                    mxf->timecode_start -= 2 * (tminutes - tminutes / 10);
                 }
+                av_log(s, AV_LOG_DEBUG, "frame %d %d:%d:%d%c%d\n", mxf->timecode_start,
+                       hours, minutes, seconds, mxf->timecode_drop_frame ? ';':':', frames);
             }
         } else if (c == 0x1b3) { // seq
             *flags |= 0x40;
-            if (i + 4 < pkt->size) {
-                switch ((pkt->data[i+4]>>4) & 0xf) {
-                case 2:  sc->aspect_ratio = (AVRational){  4,  3}; break;
-                case 3:  sc->aspect_ratio = (AVRational){ 16,  9}; break;
-                case 4:  sc->aspect_ratio = (AVRational){221,100}; break;
-                default:
-                    av_reduce(&sc->aspect_ratio.num, &sc->aspect_ratio.den,
-                              st->codec->width, st->codec->height, 1024*1024);
-                }
+            switch ((pkt->data[i+4]>>4) & 0xf) {
+            case 2:  sc->aspect_ratio = (AVRational){  4,  3}; break;
+            case 3:  sc->aspect_ratio = (AVRational){ 16,  9}; break;
+            case 4:  sc->aspect_ratio = (AVRational){221,100}; break;
+            default:
+                av_reduce(&sc->aspect_ratio.num, &sc->aspect_ratio.den,
+                          st->codec->width, st->codec->height, 1024*1024);
             }
         } else if (c == 0x100) { // pic
             int pict_type = (pkt->data[i+2]>>3) & 0x07;
             if (pict_type == 2) { // P frame
                 *flags |= 0x22;
-                st->codec->gop_size = 1;
+                sc->closed_gop = 0; // reset closed gop, don't matter anymore
             } else if (pict_type == 3) { // B frame
-                *flags |= 0x33;
+                if (sc->closed_gop)
+                    *flags |= 0x13; // only backward prediction
+                else
+                    *flags |= 0x33;
                 sc->temporal_reordering = -1;
             } else if (!pict_type) {
                 av_log(s, AV_LOG_ERROR, "error parsing mpeg2 frame\n");
@@ -1372,7 +1377,7 @@ static int mxf_parse_mpeg2_frame(AVFormatContext *s, AVStream *st, AVPacket *pkt
 
 static uint64_t mxf_parse_timestamp(time_t timestamp)
 {
-    struct tm *time = localtime(&timestamp);
+    struct tm *time = gmtime(&timestamp);
     return (uint64_t)(time->tm_year+1900) << 48 |
            (uint64_t)(time->tm_mon+1)     << 40 |
            (uint64_t) time->tm_mday       << 32 |
@@ -1381,6 +1386,18 @@ static uint64_t mxf_parse_timestamp(time_t timestamp)
                       time->tm_sec        << 8;
 }
 
+static void mxf_gen_umid(AVFormatContext *s)
+{
+    MXFContext *mxf = s->priv_data;
+    uint32_t seed = ff_random_get_seed();
+    uint64_t umid = seed + 0x5294713400000000LL;
+
+    AV_WB64(mxf->umid  , umid);
+    AV_WB64(mxf->umid+8, umid>>8);
+
+    mxf->instance_number = seed;
+}
+
 static int mxf_write_header(AVFormatContext *s)
 {
     MXFContext *mxf = s->priv_data;
@@ -1388,6 +1405,9 @@ static int mxf_write_header(AVFormatContext *s)
     uint8_t present[FF_ARRAY_ELEMS(mxf_essence_container_uls)] = {0};
     const int *samples_per_frame = NULL;
 
+    if (!s->nb_streams)
+        return -1;
+
     for (i = 0; i < s->nb_streams; i++) {
         AVStream *st = s->streams[i];
         MXFStreamContext *sc = av_mallocz(sizeof(*sc));
@@ -1466,21 +1486,22 @@ static int mxf_write_header(AVFormatContext *s)
 
         sc->codec_ul = &mxf_essence_container_uls[sc->index].codec_ul;
 
-        if (!present[sc->index]) {
-            present[sc->index] = 1;
-            mxf->essence_container_count++;
-        } else
-            present[sc->index]++;
-
         memcpy(sc->track_essence_element_key, mxf_essence_container_uls[sc->index].element_ul, 15);
         sc->track_essence_element_key[15] = present[sc->index];
         PRINT_KEY(s, "track essence element key", sc->track_essence_element_key);
+
+        if (!present[sc->index])
+            mxf->essence_container_count++;
+        present[sc->index]++;
     }
 
     if (s->oformat == &mxf_d10_muxer) {
         mxf->essence_container_count = 1;
     }
 
+    if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT))
+        mxf_gen_umid(s);
+
     for (i = 0; i < s->nb_streams; i++) {
         MXFStreamContext *sc = s->streams[i]->priv_data;
         // update element count
@@ -1488,7 +1509,8 @@ static int mxf_write_header(AVFormatContext *s)
         sc->order = AV_RB32(sc->track_essence_element_key+12);
     }
 
-    mxf->timestamp = mxf_parse_timestamp(s->timestamp);
+    if (s->timestamp)
+        mxf->timestamp = mxf_parse_timestamp(s->timestamp);
     mxf->duration = -1;
 
     mxf->timecode_track = av_mallocz(sizeof(*mxf->timecode_track));
@@ -1567,7 +1589,7 @@ static void mxf_write_system_item(AVFormatContext *s)
     klv_encode_ber4_length(pb, 35);
     put_byte(pb, 0x83); // UMID
     put_be16(pb, 0x20);
-    mxf_write_umid(pb, SourcePackage, 0);
+    mxf_write_umid(s, 1);
 }
 
 static void mxf_write_d10_video_packet(AVFormatContext *s, AVStream *st, AVPacket *pkt)
@@ -1604,8 +1626,7 @@ static void mxf_write_d10_audio_packet(AVFormatContext *s, AVStream *st, AVPacke
 {
     MXFContext *mxf = s->priv_data;
     ByteIOContext *pb = s->pb;
-    int frame_size = (pkt->size<<3) /
-        (st->codec->channels*av_get_bits_per_sample(st->codec->codec_id));
+    int frame_size = pkt->size / st->codec->block_align;
     uint8_t *samples = pkt->data;
     uint8_t *end = pkt->data + pkt->size;
     int i;
@@ -1620,16 +1641,16 @@ static void mxf_write_d10_audio_packet(AVFormatContext *s, AVStream *st, AVPacke
         for (i = 0; i < st->codec->channels; i++) {
             uint32_t sample;
             if (st->codec->codec_id == CODEC_ID_PCM_S24LE) {
-                sample = (AV_RL24(samples)<< 4)|((samples==pkt->data)<<3) | i;
+                sample = AV_RL24(samples)<< 4;
                 samples += 3;
             } else {
-                sample = (AV_RL16(samples)<<12)|((samples==pkt->data)<<3) | i;
+                sample = AV_RL16(samples)<<12;
                 samples += 2;
             }
-            put_le32(pb, sample);
+            put_le32(pb, sample | i);
         }
         for (; i < 8; i++)
-            put_le32(pb, 0);
+            put_le32(pb, i);
     }
 }
 
@@ -1815,6 +1836,9 @@ static int mxf_interleave_get_packet(AVFormatContext *s, AVPacket *out, AVPacket
             // purge packet queue
             while (pktl) {
                 AVPacketList *next = pktl->next;
+
+                if(s->streams[pktl->pkt.stream_index]->last_in_packet_buffer == pktl)
+                    s->streams[pktl->pkt.stream_index]->last_in_packet_buffer= NULL;
                 av_free_packet(&pktl->pkt);
                 av_freep(&pktl);
                 pktl = next;
@@ -1823,6 +1847,7 @@ static int mxf_interleave_get_packet(AVFormatContext *s, AVPacket *out, AVPacket
                 last->next = NULL;
             else {
                 s->packet_buffer = NULL;
+                s->packet_buffer_end= NULL;
                 goto out;
             }
             pktl = s->packet_buffer;
@@ -1831,6 +1856,10 @@ static int mxf_interleave_get_packet(AVFormatContext *s, AVPacket *out, AVPacket
         *out = pktl->pkt;
         //av_log(s, AV_LOG_DEBUG, "out st:%d dts:%lld\n", (*out).stream_index, (*out).dts);
         s->packet_buffer = pktl->next;
+        if(s->streams[pktl->pkt.stream_index]->last_in_packet_buffer == pktl)
+            s->streams[pktl->pkt.stream_index]->last_in_packet_buffer= NULL;
+        if(!s->packet_buffer)
+            s->packet_buffer_end= NULL;
         av_freep(&pktl);
         return 1;
     } else {