1 files changed, 252 insertions, 81 deletions

diff --git a/libavcodec/hapdec.c b/libavcodec/hapdec.c
index 0398efe6ca..f4f6ef6f2b 100644
--- a/libavcodec/hapdec.c
+++ b/libavcodec/hapdec.c
@@ -1,21 +1,22 @@
 /*
  * Vidvox Hap decoder
  * Copyright (C) 2015 Vittorio Giovara <vittorio.giovara@gmail.com>
+ * Copyright (C) 2015 Tom Butterworth <bangnoise@gmail.com>
  *
- * This file is part of Libav.
+ * This file is part of FFmpeg.
  *
- * Libav is free software; you can redistribute it and/or
+ * FFmpeg is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
  * License as published by the Free Software Foundation; either
  * version 2.1 of the License, or (at your option) any later version.
  *
- * Libav is distributed in the hope that it will be useful,
+ * FFmpeg is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  * Lesser General Public License for more details.
  *
  * You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
+ * License along with FFmpeg; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
@@ -39,105 +40,227 @@
 #include "snappy.h"
 #include "texturedsp.h"
 #include "thread.h"
+#include "memory.h"
 
 /* The first three bytes are the size of the section past the header, or zero
  * if the length is stored in the next long word. The fourth byte in the first
  * long word indicates the type of the current section. */
-static int parse_section_header(AVCodecContext *avctx)
+static int parse_section_header(GetByteContext *gbc, int *section_size,
+                                enum HapSectionType *section_type)
 {
-    HapContext *ctx = avctx->priv_data;
-    GetByteContext *gbc = &ctx->gbc;
-    int length;
-
     if (bytestream2_get_bytes_left(gbc) < 4)
         return AVERROR_INVALIDDATA;
 
-    length = bytestream2_get_le24(gbc);
+    *section_size = bytestream2_get_le24(gbc);
+    *section_type = bytestream2_get_byte(gbc);
 
-    ctx->section_type = bytestream2_get_byte(gbc);
-
-    if (length == 0) {
+    if (*section_size == 0) {
         if (bytestream2_get_bytes_left(gbc) < 4)
             return AVERROR_INVALIDDATA;
-        length = bytestream2_get_le32(gbc);
+
+        *section_size = bytestream2_get_le32(gbc);
+    }
+
+    if (*section_size > bytestream2_get_bytes_left(gbc))
+        return AVERROR_INVALIDDATA;
+    else
+        return 0;
+}
+
+static int hap_parse_decode_instructions(HapContext *ctx, int size)
+{
+    GetByteContext *gbc = &ctx->gbc;
+    int section_size;
+    enum HapSectionType section_type;
+    int is_first_table = 1, had_offsets = 0, had_compressors = 0, had_sizes = 0;
+    int i, ret;
+
+    while (size > 0) {
+        int stream_remaining = bytestream2_get_bytes_left(gbc);
+        ret = parse_section_header(gbc, &section_size, &section_type);
+        if (ret != 0)
+            return ret;
+
+        size -= stream_remaining - bytestream2_get_bytes_left(gbc);
+
+        switch (section_type) {
+            case HAP_ST_COMPRESSOR_TABLE:
+                ret = ff_hap_set_chunk_count(ctx, section_size, is_first_table);
+                if (ret != 0)
+                    return ret;
+                for (i = 0; i < section_size; i++) {
+                    ctx->chunks[i].compressor = bytestream2_get_byte(gbc) << 4;
+                }
+                had_compressors = 1;
+                is_first_table = 0;
+                break;
+            case HAP_ST_SIZE_TABLE:
+                ret = ff_hap_set_chunk_count(ctx, section_size / 4, is_first_table);
+                if (ret != 0)
+                    return ret;
+                for (i = 0; i < section_size / 4; i++) {
+                    ctx->chunks[i].compressed_size = bytestream2_get_le32(gbc);
+                }
+                had_sizes = 1;
+                is_first_table = 0;
+                break;
+            case HAP_ST_OFFSET_TABLE:
+                ret = ff_hap_set_chunk_count(ctx, section_size / 4, is_first_table);
+                if (ret != 0)
+                    return ret;
+                for (i = 0; i < section_size / 4; i++) {
+                    ctx->chunks[i].compressed_offset = bytestream2_get_le32(gbc);
+                }
+                had_offsets = 1;
+                is_first_table = 0;
+                break;
+            default:
+                break;
+        }
+        size -= section_size;
     }
 
-    if (length > bytestream2_get_bytes_left(gbc) || length == 0)
+    if (!had_sizes || !had_compressors)
         return AVERROR_INVALIDDATA;
 
-    return length;
+    /* The offsets table is optional. If not present than calculate offsets by
+     * summing the sizes of preceding chunks. */
+    if (!had_offsets) {
+        size_t running_size = 0;
+        for (i = 0; i < ctx->chunk_count; i++) {
+            ctx->chunks[i].compressed_offset = running_size;
+            running_size += ctx->chunks[i].compressed_size;
+        }
+    }
+
+    return 0;
+}
+
+static int hap_can_use_tex_in_place(HapContext *ctx)
+{
+    int i;
+    size_t running_offset = 0;
+    for (i = 0; i < ctx->chunk_count; i++) {
+        if (ctx->chunks[i].compressed_offset != running_offset
+            || ctx->chunks[i].compressor != HAP_COMP_NONE)
+            return 0;
+        running_offset += ctx->chunks[i].compressed_size;
+    }
+    return 1;
 }
 
-/* Prepare the texture to be decompressed */
-static int setup_texture(AVCodecContext *avctx, size_t length)
+static int hap_parse_frame_header(AVCodecContext *avctx)
 {
     HapContext *ctx = avctx->priv_data;
     GetByteContext *gbc = &ctx->gbc;
-    int64_t snappy_size;
-    const char *texture_name;
+    int section_size;
+    enum HapSectionType section_type;
     const char *compressorstr;
-    int ret;
+    int i, ret;
 
-    switch (ctx->section_type & 0x0F) {
-    case HAP_FMT_RGBDXT1:
-        ctx->tex_rat = 8;
-        ctx->tex_fun = ctx->dxtc.dxt1_block;
-        texture_name = "DXT1";
-        break;
-    case HAP_FMT_RGBADXT5:
-        ctx->tex_rat = 16;
-        ctx->tex_fun = ctx->dxtc.dxt5_block;
-        texture_name = "DXT5";
-        break;
-    case HAP_FMT_YCOCGDXT5:
-        ctx->tex_rat = 16;
-        ctx->tex_fun = ctx->dxtc.dxt5ys_block;
-        texture_name = "DXT5-YCoCg-scaled";
-        break;
-    default:
-        av_log(avctx, AV_LOG_ERROR,
-               "Invalid format mode %02X.\n", ctx->section_type);
+    ret = parse_section_header(gbc, &section_size, &section_type);
+    if (ret != 0)
+        return ret;
+
+    if ((avctx->codec_tag == MKTAG('H','a','p','1') && (section_type & 0x0F) != HAP_FMT_RGBDXT1)
+        || (avctx->codec_tag == MKTAG('H','a','p','5') && (section_type & 0x0F) != HAP_FMT_RGBADXT5)
+        || (avctx->codec_tag == MKTAG('H','a','p','Y') && (section_type & 0x0F) != HAP_FMT_YCOCGDXT5)) {
+        av_log(avctx, AV_LOG_ERROR, "Invalid texture format %#04x.\n", section_type & 0x0F);
         return AVERROR_INVALIDDATA;
     }
 
-    switch (ctx->section_type & 0xF0) {
-    case HAP_COMP_NONE:
-        /* Only DXTC texture compression */
-        ctx->tex_data = gbc->buffer;
-        ctx->tex_size = length;
-        compressorstr = "none";
-        break;
-    case HAP_COMP_SNAPPY:
-        snappy_size = ff_snappy_peek_uncompressed_length(gbc);
-        ret = av_reallocp(&ctx->snappied, snappy_size);
-        if (ret < 0) {
-            return ret;
+    switch (section_type & 0xF0) {
+        case HAP_COMP_NONE:
+        case HAP_COMP_SNAPPY:
+            ret = ff_hap_set_chunk_count(ctx, 1, 1);
+            if (ret == 0) {
+                ctx->chunks[0].compressor = section_type & 0xF0;
+                ctx->chunks[0].compressed_offset = 0;
+                ctx->chunks[0].compressed_size = section_size;
+            }
+            if (ctx->chunks[0].compressor == HAP_COMP_NONE) {
+                compressorstr = "none";
+            } else {
+                compressorstr = "snappy";
+            }
+            break;
+        case HAP_COMP_COMPLEX:
+            ret = parse_section_header(gbc, &section_size, &section_type);
+            if (ret == 0 && section_type != HAP_ST_DECODE_INSTRUCTIONS)
+                ret = AVERROR_INVALIDDATA;
+            if (ret == 0)
+                ret = hap_parse_decode_instructions(ctx, section_size);
+            compressorstr = "complex";
+            break;
+        default:
+            ret = AVERROR_INVALIDDATA;
+            break;
+    }
+
+    if (ret != 0)
+        return ret;
+
+    /* Check the frame is valid and read the uncompressed chunk sizes */
+    ctx->tex_size = 0;
+    for (i = 0; i < ctx->chunk_count; i++) {
+        HapChunk *chunk = &ctx->chunks[i];
+
+        /* Check the compressed buffer is valid */
+        if (chunk->compressed_offset + chunk->compressed_size > bytestream2_get_bytes_left(gbc))
+            return AVERROR_INVALIDDATA;
+
+        /* Chunks are unpacked sequentially, ctx->tex_size is the uncompressed
+         * size thus far */
+        chunk->uncompressed_offset = ctx->tex_size;
+
+        /* Fill out uncompressed size */
+        if (chunk->compressor == HAP_COMP_SNAPPY) {
+            GetByteContext gbc_tmp;
+            int64_t uncompressed_size;
+            bytestream2_init(&gbc_tmp, gbc->buffer + chunk->compressed_offset,
+                             chunk->compressed_size);
+            uncompressed_size = ff_snappy_peek_uncompressed_length(&gbc_tmp);
+            if (uncompressed_size < 0) {
+                return uncompressed_size;
+            }
+            chunk->uncompressed_size = uncompressed_size;
+        } else if (chunk->compressor == HAP_COMP_NONE) {
+            chunk->uncompressed_size = chunk->compressed_size;
+        } else {
+            return AVERROR_INVALIDDATA;
         }
+        ctx->tex_size += chunk->uncompressed_size;
+    }
+
+    av_log(avctx, AV_LOG_DEBUG, "%s compressor\n", compressorstr);
+
+    return ret;
+}
+
+static int decompress_chunks_thread(AVCodecContext *avctx, void *arg,
+                                    int chunk_nb, int thread_nb)
+{
+    HapContext *ctx = avctx->priv_data;
+
+    HapChunk *chunk = &ctx->chunks[chunk_nb];
+    GetByteContext gbc;
+    uint8_t *dst = ctx->tex_buf + chunk->uncompressed_offset;
+
+    bytestream2_init(&gbc, ctx->gbc.buffer + chunk->compressed_offset, chunk->compressed_size);
+
+    if (chunk->compressor == HAP_COMP_SNAPPY) {
+        int ret;
+        int64_t uncompressed_size = ctx->tex_size;
         /* Uncompress the frame */
-        ret = ff_snappy_uncompress(gbc, ctx->snappied, &snappy_size);
+        ret = ff_snappy_uncompress(&gbc, dst, &uncompressed_size);
         if (ret < 0) {
              av_log(avctx, AV_LOG_ERROR, "Snappy uncompress error\n");
              return ret;
         }
-
-        ctx->tex_data = ctx->snappied;
-        ctx->tex_size = snappy_size;
-        compressorstr = "snappy";
-        break;
-    case HAP_COMP_COMPLEX:
-        compressorstr = "complex";
-        avpriv_request_sample(avctx, "Complex Hap compressor");
-        return AVERROR_PATCHWELCOME;
-        break;
-    default:
-        av_log(avctx, AV_LOG_ERROR,
-               "Invalid compressor mode %02X.\n", ctx->section_type);
-        return AVERROR_INVALIDDATA;
+    } else if (chunk->compressor == HAP_COMP_NONE) {
+        bytestream2_get_buffer(&gbc, dst, chunk->compressed_size);
     }
 
-    av_log(avctx, AV_LOG_DEBUG, "%s texture with %s compressor\n",
-           texture_name, compressorstr);
-
     return 0;
 }
 
@@ -183,19 +306,13 @@ static int hap_decode(AVCodecContext *avctx, void *data,
 {
     HapContext *ctx = avctx->priv_data;
     ThreadFrame tframe;
-    int ret, length;
+    int ret, i;
+    int tex_size;
 
     bytestream2_init(&ctx->gbc, avpkt->data, avpkt->size);
 
     /* Check for section header */
-    length = parse_section_header(avctx);
-    if (length < 0) {
-        av_log(avctx, AV_LOG_ERROR, "Frame is too small.\n");
-        return length;
-    }
-
-    /* Prepare the texture buffer and decompress function */
-    ret = setup_texture(avctx, length);
+    ret = hap_parse_frame_header(avctx);
     if (ret < 0)
         return ret;
 
@@ -204,7 +321,38 @@ static int hap_decode(AVCodecContext *avctx, void *data,
     ret = ff_thread_get_buffer(avctx, &tframe, 0);
     if (ret < 0)
         return ret;
-    ff_thread_finish_setup(avctx);
+    if (avctx->codec->update_thread_context)
+        ff_thread_finish_setup(avctx);
+
+    /* Unpack the DXT texture */
+    if (hap_can_use_tex_in_place(ctx)) {
+        /* Only DXTC texture compression in a contiguous block */
+        ctx->tex_data = ctx->gbc.buffer;
+        tex_size = bytestream2_get_bytes_left(&ctx->gbc);
+    } else {
+        /* Perform the second-stage decompression */
+        ret = av_reallocp(&ctx->tex_buf, ctx->tex_size);
+        if (ret < 0)
+            return ret;
+
+        avctx->execute2(avctx, decompress_chunks_thread, NULL,
+                        ctx->chunk_results, ctx->chunk_count);
+
+        for (i = 0; i < ctx->chunk_count; i++) {
+            if (ctx->chunk_results[i] < 0)
+                return ctx->chunk_results[i];
+        }
+
+        ctx->tex_data = ctx->tex_buf;
+        tex_size = ctx->tex_size;
+    }
+
+    if (tex_size < (avctx->coded_width  / TEXTURE_BLOCK_W)
+                  *(avctx->coded_height / TEXTURE_BLOCK_H)
+                  *ctx->tex_rat) {
+        av_log(avctx, AV_LOG_ERROR, "Insufficient data\n");
+        return AVERROR_INVALIDDATA;
+    }
 
     /* Use the decompress function on the texture, one block per thread */
     avctx->execute2(avctx, decompress_texture_thread, tframe.f, NULL, ctx->slice_count);
@@ -220,6 +368,7 @@ static int hap_decode(AVCodecContext *avctx, void *data,
 static av_cold int hap_init(AVCodecContext *avctx)
 {
     HapContext *ctx = avctx->priv_data;
+    const char *texture_name;
     int ret = av_image_check_size(avctx->width, avctx->height, 0, avctx);
 
     if (ret < 0) {
@@ -237,6 +386,28 @@ static av_cold int hap_init(AVCodecContext *avctx)
 
     ff_texturedsp_init(&ctx->dxtc);
 
+    switch (avctx->codec_tag) {
+    case MKTAG('H','a','p','1'):
+        texture_name = "DXT1";
+        ctx->tex_rat = 8;
+        ctx->tex_fun = ctx->dxtc.dxt1_block;
+        break;
+    case MKTAG('H','a','p','5'):
+        texture_name = "DXT5";
+        ctx->tex_rat = 16;
+        ctx->tex_fun = ctx->dxtc.dxt5_block;
+        break;
+    case MKTAG('H','a','p','Y'):
+        texture_name = "DXT5-YCoCg-scaled";
+        ctx->tex_rat = 16;
+        ctx->tex_fun = ctx->dxtc.dxt5ys_block;
+        break;
+    default:
+        return AVERROR_DECODER_NOT_FOUND;
+    }
+
+    av_log(avctx, AV_LOG_DEBUG, "%s texture\n", texture_name);
+
     ctx->slice_count = av_clip(avctx->thread_count, 1,
                                avctx->coded_height / TEXTURE_BLOCK_H);
 
@@ -247,7 +418,7 @@ static av_cold int hap_close(AVCodecContext *avctx)
 {
     HapContext *ctx = avctx->priv_data;
 
-    av_freep(&ctx->snappied);
+    ff_hap_free_context(ctx);
 
     return 0;
 }