From 57dd96199570beb29ea8b0f3934c594cd42e3043 Mon Sep 17 00:00:00 2001 From: Tianjie Xu Date: Thu, 17 Aug 2017 17:50:56 -0700 Subject: Move Image/ImageChunk/PatchChunk declaration into header files 1. Move the declaration of the Image classes to the header file to make testing easier. 2. Also move rangeset.h to bootable/recovery to allow access in imgdiff. Test: recovery component test Change-Id: I68a863e60a3f2e7ae46ee48f48eb15391f5f4330 --- applypatch/imgdiff.cpp | 335 ++++++++++--------------------------------------- 1 file changed, 65 insertions(+), 270 deletions(-) (limited to 'applypatch/imgdiff.cpp') diff --git a/applypatch/imgdiff.cpp b/applypatch/imgdiff.cpp index a81e385a3..59b600713 100644 --- a/applypatch/imgdiff.cpp +++ b/applypatch/imgdiff.cpp @@ -140,11 +140,12 @@ #include #include #include -#include - #include +#include #include +#include "applypatch/imgdiff_image.h" + using android::base::get_unaligned; static constexpr auto BUFFER_SIZE = 0x8000; @@ -161,99 +162,16 @@ static inline bool Write4(int fd, int32_t value) { return android::base::WriteFully(fd, &value, sizeof(int32_t)); } -class ImageChunk { - public: - static constexpr auto WINDOWBITS = -15; // 32kb window; negative to indicate a raw stream. - static constexpr auto MEMLEVEL = 8; // the default value. - static constexpr auto METHOD = Z_DEFLATED; - static constexpr auto STRATEGY = Z_DEFAULT_STRATEGY; - - ImageChunk(int type, size_t start, const std::vector* file_content, size_t raw_data_len, - std::string entry_name = {}) - : type_(type), - start_(start), - input_file_ptr_(file_content), - raw_data_len_(raw_data_len), - compress_level_(6), - entry_name_(std::move(entry_name)) { - CHECK(file_content != nullptr) << "input file container can't be nullptr"; - } - - int GetType() const { - return type_; - } - size_t GetRawDataLength() const { - return raw_data_len_; - } - const std::string& GetEntryName() const { - return entry_name_; - } - size_t GetStartOffset() const { - return start_; - } - int GetCompressLevel() const { - return compress_level_; - } - - // CHUNK_DEFLATE will return the uncompressed data for diff, while other types will simply return - // the raw data. - const uint8_t * DataForPatch() const; - size_t DataLengthForPatch() const; - - void Dump() const { - printf("type: %d, start: %zu, len: %zu, name: %s\n", type_, start_, DataLengthForPatch(), - entry_name_.c_str()); - } - - void SetUncompressedData(std::vector data); - bool SetBonusData(const std::vector& bonus_data); - - bool operator==(const ImageChunk& other) const; - bool operator!=(const ImageChunk& other) const { - return !(*this == other); - } - - /* - * Cause a gzip chunk to be treated as a normal chunk (ie, as a blob of uninterpreted data). - * The resulting patch will likely be about as big as the target file, but it lets us handle - * the case of images where some gzip chunks are reconstructible but others aren't (by treating - * the ones that aren't as normal chunks). - */ - void ChangeDeflateChunkToNormal(); - - /* - * Verify that we can reproduce exactly the same compressed data that we started with. Sets the - * level, method, windowBits, memLevel, and strategy fields in the chunk to the encoding - * parameters needed to produce the right output. - */ - bool ReconstructDeflateChunk(); - bool IsAdjacentNormal(const ImageChunk& other) const; - void MergeAdjacentNormal(const ImageChunk& other); - - /* - * Compute a bsdiff patch between |src| and |tgt|; Store the result in the patch_data. - * |bsdiff_cache| can be used to cache the suffix array if the same |src| chunk is used - * repeatedly, pass nullptr if not needed. - */ - static bool MakePatch(const ImageChunk& tgt, const ImageChunk& src, - std::vector* patch_data, saidx_t** bsdiff_cache); - - private: - const uint8_t* GetRawData() const; - bool TryReconstruction(int level); - - int type_; // CHUNK_NORMAL, CHUNK_DEFLATE, CHUNK_RAW - size_t start_; // offset of chunk in the original input file - const std::vector* input_file_ptr_; // ptr to the full content of original input file - size_t raw_data_len_; - - // deflate encoder parameters - int compress_level_; - - // --- for CHUNK_DEFLATE chunks only: --- - std::vector uncompressed_data_; - std::string entry_name_; // used for zip entries -}; +ImageChunk::ImageChunk(int type, size_t start, const std::vector* file_content, + size_t raw_data_len, std::string entry_name) + : type_(type), + start_(start), + input_file_ptr_(file_content), + raw_data_len_(raw_data_len), + compress_level_(6), + entry_name_(std::move(entry_name)) { + CHECK(file_content != nullptr) << "input file container can't be nullptr"; +} const uint8_t* ImageChunk::GetRawData() const { CHECK_LE(start_ + raw_data_len_, input_file_ptr_->size()); @@ -424,57 +342,28 @@ bool ImageChunk::TryReconstruction(int level) { return true; } -// PatchChunk stores the patch data between a source chunk and a target chunk. It also keeps track -// of the metadata of src&tgt chunks (e.g. offset, raw data length, uncompressed data length). -class PatchChunk { - public: - PatchChunk(const ImageChunk& tgt, const ImageChunk& src, std::vector data) - : type_(tgt.GetType()), - source_start_(src.GetStartOffset()), - source_len_(src.GetRawDataLength()), - source_uncompressed_len_(src.DataLengthForPatch()), - target_start_(tgt.GetStartOffset()), - target_len_(tgt.GetRawDataLength()), - target_uncompressed_len_(tgt.DataLengthForPatch()), - target_compress_level_(tgt.GetCompressLevel()), - data_(std::move(data)) {} - - // Construct a CHUNK_RAW patch from the target data directly. - explicit PatchChunk(const ImageChunk& tgt) - : type_(CHUNK_RAW), - source_start_(0), - source_len_(0), - source_uncompressed_len_(0), - target_start_(tgt.GetStartOffset()), - target_len_(tgt.GetRawDataLength()), - target_uncompressed_len_(tgt.DataLengthForPatch()), - target_compress_level_(tgt.GetCompressLevel()), - data_(tgt.DataForPatch(), tgt.DataForPatch() + tgt.DataLengthForPatch()) {} - - // Return true if raw data size is smaller than the patch size. - static bool RawDataIsSmaller(const ImageChunk& tgt, size_t patch_size); - - static bool WritePatchDataToFd(const std::vector& patch_chunks, int patch_fd); - - private: - size_t GetHeaderSize() const; - size_t WriteHeaderToFd(int fd, size_t offset) const; - - // The patch chunk type is the same as the target chunk type. The only exception is we change - // the |type_| to CHUNK_RAW if target length is smaller than the patch size. - int type_; - - size_t source_start_; - size_t source_len_; - size_t source_uncompressed_len_; - - size_t target_start_; // offset of the target chunk within the target file - size_t target_len_; - size_t target_uncompressed_len_; - size_t target_compress_level_; // the deflate compression level of the target chunk. - - std::vector data_; // storage for the patch data -}; +PatchChunk::PatchChunk(const ImageChunk& tgt, const ImageChunk& src, std::vector data) + : type_(tgt.GetType()), + source_start_(src.GetStartOffset()), + source_len_(src.GetRawDataLength()), + source_uncompressed_len_(src.DataLengthForPatch()), + target_start_(tgt.GetStartOffset()), + target_len_(tgt.GetRawDataLength()), + target_uncompressed_len_(tgt.DataLengthForPatch()), + target_compress_level_(tgt.GetCompressLevel()), + data_(std::move(data)) {} + +// Construct a CHUNK_RAW patch from the target data directly. +PatchChunk::PatchChunk(const ImageChunk& tgt) + : type_(CHUNK_RAW), + source_start_(0), + source_len_(0), + source_uncompressed_len_(0), + target_start_(tgt.GetStartOffset()), + target_len_(tgt.GetRawDataLength()), + target_uncompressed_len_(tgt.DataLengthForPatch()), + target_compress_level_(tgt.GetCompressLevel()), + data_(tgt.DataForPatch(), tgt.DataForPatch() + tgt.DataLengthForPatch()) {} // Return true if raw data is smaller than the patch size. bool PatchChunk::RawDataIsSmaller(const ImageChunk& tgt, size_t patch_size) { @@ -574,59 +463,15 @@ bool PatchChunk::WritePatchDataToFd(const std::vector& patch_chunks, return true; } -// Interface for zip_mode and image_mode images. We initialize the image from an input file and -// split the file content into a list of image chunks. -class Image { - public: - explicit Image(bool is_source) : is_source_(is_source) {} - - virtual ~Image() {} - - // Create a list of image chunks from input file. - virtual bool Initialize(const std::string& filename) = 0; - - // Look for runs of adjacent normal chunks and compress them down into a single chunk. (Such - // runs can be produced when deflate chunks are changed to normal chunks.) - void MergeAdjacentNormalChunks(); - - // In zip mode, find the matching deflate source chunk by entry name. Search for normal chunks - // also if |find_normal| is true. - ImageChunk* FindChunkByName(const std::string& name, bool find_normal = false); - - const ImageChunk* FindChunkByName(const std::string& name, bool find_normal = false) const; - - void DumpChunks() const; - - // Non const iterators to access the stored ImageChunks. - std::vector::iterator begin() { - return chunks_.begin(); - } - - std::vector::iterator end() { - return chunks_.end(); - } - - ImageChunk& operator[](size_t i) { - CHECK_LT(i, chunks_.size()); - return chunks_[i]; - } - - const ImageChunk& operator[](size_t i) const { - CHECK_LT(i, chunks_.size()); - return chunks_[i]; - } - - size_t NumOfChunks() const { - return chunks_.size(); - } - - protected: - bool ReadFile(const std::string& filename, std::vector* file_content); +ImageChunk& Image::operator[](size_t i) { + CHECK_LT(i, chunks_.size()); + return chunks_[i]; +} - bool is_source_; // True if it's for source chunks. - std::vector chunks_; // Internal storage of ImageChunk. - std::vector file_content_; // Store the whole input file in memory. -}; +const ImageChunk& Image::operator[](size_t i) const { + CHECK_LT(i, chunks_.size()); + return chunks_[i]; +} void Image::MergeAdjacentNormalChunks() { size_t merged_last = 0, cur = 0; @@ -650,23 +495,6 @@ void Image::MergeAdjacentNormalChunks() { } } -const ImageChunk* Image::FindChunkByName(const std::string& name, bool find_normal) const { - if (name.empty()) { - return nullptr; - } - for (auto& chunk : chunks_) { - if ((chunk.GetType() == CHUNK_DEFLATE || find_normal) && chunk.GetEntryName() == name) { - return &chunk; - } - } - return nullptr; -} - -ImageChunk* Image::FindChunkByName(const std::string& name, bool find_normal) { - return const_cast( - static_cast(this)->FindChunkByName(name, find_normal)); -} - void Image::DumpChunks() const { std::string type = is_source_ ? "source" : "target"; printf("Dumping chunks for %s\n", type.c_str()); @@ -701,39 +529,6 @@ bool Image::ReadFile(const std::string& filename, std::vector* file_con return true; } -class ZipModeImage : public Image { - public: - explicit ZipModeImage(bool is_source) : Image(is_source) {} - - bool Initialize(const std::string& filename) override; - - const ImageChunk& PseudoSource() const { - CHECK(is_source_); - CHECK(pseudo_source_ != nullptr); - return *pseudo_source_; - } - - // Verify that we can reconstruct the deflate chunks; also change the type to CHUNK_NORMAL if - // src and tgt are identical. - static bool CheckAndProcessChunks(ZipModeImage* tgt_image, ZipModeImage* src_image); - - // Compute the patch between tgt & src images, and write the data into |patch_name|. - static bool GeneratePatches(const ZipModeImage& tgt_image, const ZipModeImage& src_image, - const std::string& patch_name); - - private: - // Initialize image chunks based on the zip entries. - bool InitializeChunks(const std::string& filename, ZipArchiveHandle handle); - // Add the a zip entry to the list. - bool AddZipEntryToChunks(ZipArchiveHandle handle, const std::string& entry_name, ZipEntry* entry); - // Return the real size of the zip file. (omit the trailing zeros that used for alignment) - bool GetZipFileSize(size_t* input_file_size); - - // The pesudo source chunk for bsdiff if there's no match for the given target chunk. It's in - // fact the whole source file. - std::unique_ptr pseudo_source_; -}; - bool ZipModeImage::Initialize(const std::string& filename) { if (!ReadFile(filename, &file_content_)) { return false; @@ -754,9 +549,6 @@ bool ZipModeImage::Initialize(const std::string& filename) { return false; } - if (is_source_) { - pseudo_source_ = std::make_unique(CHUNK_NORMAL, 0, &file_content_, zipfile_size); - } if (!InitializeChunks(filename, handle)) { CloseArchive(handle); return false; @@ -895,6 +687,28 @@ bool ZipModeImage::GetZipFileSize(size_t* input_file_size) { return false; } +ImageChunk ZipModeImage::PseudoSource() const { + CHECK(is_source_); + return ImageChunk(CHUNK_NORMAL, 0, &file_content_, file_content_.size()); +} + +const ImageChunk* ZipModeImage::FindChunkByName(const std::string& name, bool find_normal) const { + if (name.empty()) { + return nullptr; + } + for (auto& chunk : chunks_) { + if ((chunk.GetType() == CHUNK_DEFLATE || find_normal) && chunk.GetEntryName() == name) { + return &chunk; + } + } + return nullptr; +} + +ImageChunk* ZipModeImage::FindChunkByName(const std::string& name, bool find_normal) { + return const_cast( + static_cast(this)->FindChunkByName(name, find_normal)); +} + bool ZipModeImage::CheckAndProcessChunks(ZipModeImage* tgt_image, ZipModeImage* src_image) { for (auto& tgt_chunk : *tgt_image) { if (tgt_chunk.GetType() != CHUNK_DEFLATE) { @@ -981,25 +795,6 @@ bool ZipModeImage::GeneratePatches(const ZipModeImage& tgt_image, const ZipModeI return PatchChunk::WritePatchDataToFd(patch_chunks, patch_fd); } -class ImageModeImage : public Image { - public: - explicit ImageModeImage(bool is_source) : Image(is_source) {} - - // Initialize the image chunks list by searching the magic numbers in an image file. - bool Initialize(const std::string& filename) override; - - bool SetBonusData(const std::vector& bonus_data); - - // In Image Mode, verify that the source and target images have the same chunk structure (ie, the - // same sequence of deflate and normal chunks). - static bool CheckAndProcessChunks(ImageModeImage* tgt_image, ImageModeImage* src_image); - - // In image mode, generate patches against the given source chunks and bonus_data; write the - // result to |patch_name|. - static bool GeneratePatches(const ImageModeImage& tgt_image, const ImageModeImage& src_image, - const std::string& patch_name); -}; - bool ImageModeImage::Initialize(const std::string& filename) { if (!ReadFile(filename, &file_content_)) { return false; -- cgit v1.2.3