#ifdef WIN32 #include #include #else #include #endif #include "stdafx.h" #include "../../../../Externals/zlib/zlib.h" #include "Blob.h" #include "MappedFile.h" namespace DiscIO { const u32 kBlobCookie = 0xB10BC001; // A blob file structure: // BlobHeader // u64 offsetsToBlocks[n], top bit specifies whether the block is compressed, or not. // compressed data // Blocks that won't compress to less than 97% of the original size are stored as-is. struct BlobHeader // 32 bytes { u32 magic_cookie; //0xB10BB10B u32 sub_type; // gc image, whatever u64 compressed_data_size; u64 data_size; u32 block_size; u32 num_blocks; }; #ifdef _WIN32 class PlainFileReader : public IBlobReader { HANDLE hFile; s64 size; public: PlainFileReader(const char* filename) { hFile = CreateFile(filename, GENERIC_READ, FILE_SHARE_READ, 0, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL | FILE_FLAG_RANDOM_ACCESS, NULL); DWORD size_low, size_high; size_low = GetFileSize(hFile, &size_high); size = ((u64)size_low) | ((u64)size_high << 32); } ~PlainFileReader() { if (hFile) { CloseHandle(hFile); hFile = INVALID_HANDLE_VALUE; } } bool IsValid() const {return(hFile != INVALID_HANDLE_VALUE);} u64 GetDataSize() const {return(size);} u64 GetRawSize() const {return(size);} bool Read(u64 offset, u64 size, u8* out_ptr) { LONG offset_high = (LONG)(offset >> 32); SetFilePointer(hFile, (DWORD)(offset & 0xFFFFFFFF), &offset_high, FILE_BEGIN); if (size >= 0x100000000ULL) { return(false); // WTF, does windows really have this limitation? } DWORD unused; if (!ReadFile(hFile, out_ptr, DWORD(size & 0xFFFFFFFF), &unused, NULL)) { return(false); } else { return(true); } } }; #else // linux, 64-bit. We do not yet care about linux32 // Not optimal - will keep recreating mappings. class PlainFileReader : public IBlobReader { FILE* file_; s64 size; public: PlainFileReader(const char* filename) { file_ = fopen(filename, "rb"); #if 0 fseek64(file_, 0, SEEK_END); #else fseek(file_, 0, SEEK_END); // I don't have fseek64 with gcc 4.3 #endif size = ftell(file_); fseek(file_, 0, SEEK_SET); } ~PlainFileReader() { if (file_) { fclose(file_); } } bool IsValid() const { return(file_ != 0); } u64 GetDataSize() const { return(size); } u64 GetRawSize() const { return(size); } bool Read(u64 offset, u64 size, u8* out_ptr) { fseek(file_, offset, SEEK_SET); fread(out_ptr, size, 1, file_); return(true); } }; #endif class CompressedBlobReader : public IBlobReader { enum { CACHE_SIZE = 32 }; BlobHeader header; u64* block_pointers; int data_offset; u8* cache[CACHE_SIZE]; u64 cache_tags[CACHE_SIZE]; int cache_age[CACHE_SIZE]; u64 counter; Common::IMappedFile* mapped_file; public: CompressedBlobReader(const char* filename) { counter = 0; mapped_file = Common::IMappedFile::CreateMappedFile(); if (mapped_file) { mapped_file->Open(filename); u8* start = mapped_file->Lock(0, sizeof(BlobHeader)); memcpy(&header, start, sizeof(BlobHeader)); mapped_file->Unlock(start); block_pointers = (u64*)mapped_file->Lock(sizeof(BlobHeader), sizeof(u64) * header.num_blocks); data_offset = sizeof(BlobHeader) + sizeof(u64) * header.num_blocks; for (int i = 0; i < CACHE_SIZE; i++) { cache[i] = new u8[header.block_size]; cache_tags[i] = (u64)(s64) - 1; } } } const BlobHeader& GetHeader() const { return(header); } bool IsValid() const { return(mapped_file != 0); } ~CompressedBlobReader() { for (int i = 0; i < CACHE_SIZE; i++) { delete[] cache[i]; } mapped_file->Unlock((u8*)block_pointers); if (mapped_file) { mapped_file->Close(); } mapped_file = 0; } u64 GetDataSize() const { return(header.data_size); } u64 GetRawSize() const { return(mapped_file->GetSize()); } u64 GetBlockCompressedSize(u64 block_num) const { u64 start = block_pointers[block_num]; if (block_num != header.num_blocks - 1) { return(block_pointers[block_num + 1] - start); } else { return(header.compressed_data_size - start); } } // IMPORTANT: Calling this function invalidates all earlier pointers gotten from this function. u8* GetBlock(u64 block_num) { if (cache_tags[0] != block_num) { cache_tags[0] = block_num; //PanicAlert("here2"); // let's begin with a super naive implementation. // let's just use the start of the cache for now. bool uncompressed = false; u32 comp_block_size = (u32)GetBlockCompressedSize(block_num); u64 offset = block_pointers[block_num] + data_offset; if (offset & (1ULL << 63)) { if (comp_block_size != header.block_size) { PanicAlert("Uncompressed block with wrong size"); } uncompressed = true; offset &= ~(1ULL << 63); } u8* source = mapped_file->Lock(offset, comp_block_size); u8* dest = cache[0]; if (uncompressed) { memcpy(dest, source, comp_block_size); } else { z_stream z; memset(&z, 0, sizeof(z)); z.next_in = source; z.avail_in = comp_block_size; if (z.avail_in > header.block_size) { PanicAlert("We have a problem"); } z.next_out = dest; z.avail_out = header.block_size; inflateInit(&z); int status = inflate(&z, Z_FULL_FLUSH); int uncomp_size = header.block_size - z.avail_out; if (status != Z_STREAM_END) { // this seem to fire wrongly from time to time // to be sure, don't use compressed isos :P //PanicAlert("Failure reading block %i", block_num); } if (uncomp_size != header.block_size) { PanicAlert("Wrong block size"); } inflateEnd(&z); } mapped_file->Unlock(source); } //PanicAlert("here3"); return(cache[0]); } bool Read(u64 offset, u64 size, u8* out_ptr) { u64 startingBlock = offset / header.block_size; u64 remain = size; int positionInBlock = (int)(offset % header.block_size); u64 block = startingBlock; while (remain > 0) { u8* data = GetBlock(block); if (!data) { return(false); } int toCopy = header.block_size - positionInBlock; if (toCopy >= remain) { // yay, we are done! memcpy(out_ptr, data + positionInBlock, (size_t)remain); return(true); } else { memcpy(out_ptr, data + positionInBlock, toCopy); out_ptr += toCopy; remain -= toCopy; positionInBlock = 0; block++; } } PanicAlert("here4"); return(true); } }; bool CompressFileToBlob(const char* infile, const char* outfile, u32 sub_type, int block_size, CompressCB callback, void* arg) { //Common::IMappedFile *in = Common::IMappedFile::CreateMappedFile(); if (IsCompressedBlob(infile)) { PanicAlert("%s is already compressed!", infile); return(false); } FILE* inf = fopen(infile, "rb"); if (!inf) { return(false); } FILE* f = fopen(outfile, "wb"); if (!f) { return(false); } callback("Files opened, ready to compress.", 0, arg); fseek(inf, 0, SEEK_END); int insize = ftell(inf); fseek(inf, 0, SEEK_SET); BlobHeader header; header.magic_cookie = kBlobCookie; header.sub_type = sub_type; header.block_size = block_size; header.data_size = insize; // round upwards! header.num_blocks = (u32)((header.data_size + (block_size - 1)) / block_size); u64* offsets = new u64[header.num_blocks]; u8* out_buf = new u8[block_size]; u8* in_buf = new u8[block_size]; // seek past the header (we will write it at the end) fseek(f, sizeof(BlobHeader), SEEK_CUR); // seek past the offset table (we will write it at the end) fseek(f, sizeof(u64) * header.num_blocks, SEEK_CUR); // Now we are ready to write compressed data! u64 position = 0; int num_compressed = 0; int num_stored = 0; for (u32 i = 0; i < header.num_blocks; i++) { if (i % (header.num_blocks / 100) == 0) { u64 inpos = ftell(inf); int ratio = (int)(100 * position / inpos); char temp[512]; sprintf(temp, "%i of %i blocks. compression ratio %i%%", i, header.num_blocks, ratio); callback(temp, (float)i / (float)header.num_blocks, arg); } offsets[i] = position; // u64 start = i * header.block_size; // u64 size = header.block_size; memset(in_buf, 0, header.block_size); fread(in_buf, header.block_size, 1, inf); z_stream z; memset(&z, 0, sizeof(z)); z.zalloc = Z_NULL; z.zfree = Z_NULL; z.opaque = Z_NULL; z.next_in = in_buf; z.avail_in = header.block_size; z.next_out = out_buf; z.avail_out = block_size; int retval = deflateInit(&z, 9); if (retval != Z_OK) { PanicAlert("Deflate failed"); goto cleanup; } int status = deflate(&z, Z_FINISH); int comp_size = block_size - z.avail_out; if ((status != Z_STREAM_END) || (z.avail_out < 10)) { //PanicAlert("%i %i Store %i", i*block_size, position, comp_size); // let's store uncompressed offsets[i] |= 0x8000000000000000ULL; fwrite(in_buf, block_size, 1, f); position += block_size; num_stored++; } else { // let's store compressed //PanicAlert("Comp %i to %i", block_size, comp_size); fwrite(out_buf, comp_size, 1, f); position += comp_size; num_compressed++; } deflateEnd(&z); } header.compressed_data_size = position; // Okay, go back and fill in headers fseek(f, 0, SEEK_SET); fwrite(&header, sizeof(header), 1, f); fwrite(offsets, sizeof(u64), header.num_blocks, f); cleanup: // Cleanup delete[] in_buf; delete[] out_buf; delete[] offsets; fclose(f); fclose(inf); callback("Done.", 1.0f, arg); return(true); } bool DecompressBlobToFile(const char* infile, const char* outfile, CompressCB callback, void* arg) { if (!IsCompressedBlob(infile)) { PanicAlert("File not compressed"); return(false); } CompressedBlobReader* reader = new CompressedBlobReader(infile); FILE* f = fopen(outfile, "wb"); const BlobHeader& header = reader->GetHeader(); u8* buffer = new u8[header.block_size]; for (u64 i = 0; i < header.num_blocks; i++) { if (i % (header.num_blocks / 100) == 0) { callback("Unpacking", (float)i / (float)header.num_blocks, arg); } reader->Read(i * header.block_size, header.block_size, buffer); fwrite(buffer, header.block_size, 1, f); } delete[] buffer; #ifdef _WIN32 // TODO(ector): _chsize sucks, not 64-bit safe // F|RES: changed to _chsize_s. i think it is 64-bit safe _chsize_s(_fileno(f), (long)header.data_size); #else ftruncate(fileno(f), header.data_size); #endif fclose(f); return(true); } bool IsCompressedBlob(const char* filename) { FILE* f = fopen(filename, "rb"); if (!f) { return(0); } BlobHeader header; fread(&header, sizeof(header), 1, f); fclose(f); return(header.magic_cookie == kBlobCookie); } IBlobReader* CreateBlobReader(const char* filename) { IBlobReader* reader; if (IsCompressedBlob(filename)) { reader = new CompressedBlobReader(filename); } else { reader = new PlainFileReader(filename); } if (reader->IsValid()) { return(reader); } else { delete reader; return(0); } } } // namespace