| /* |
| * Copyright (C) 2014 The Android Open Source Project |
| * |
| * Licensed under the Apache License, Version 2.0 (the "License"); |
| * you may not use this file except in compliance with the License. |
| * You may obtain a copy of the License at |
| * |
| * http://www.apache.org/licenses/LICENSE-2.0 |
| * |
| * Unless required by applicable law or agreed to in writing, software |
| * distributed under the License is distributed on an "AS IS" BASIS, |
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| * See the License for the specific language governing permissions and |
| * limitations under the License. |
| */ |
| |
| #include <ctype.h> |
| #include <dirent.h> |
| #include <errno.h> |
| #include <fcntl.h> |
| #include <inttypes.h> |
| #include <linux/fs.h> |
| #include <pthread.h> |
| #include <stdarg.h> |
| #include <stdio.h> |
| #include <stdlib.h> |
| #include <string.h> |
| #include <sys/ioctl.h> |
| #include <sys/stat.h> |
| #include <sys/types.h> |
| #include <sys/wait.h> |
| #include <time.h> |
| #include <unistd.h> |
| |
| #include <functional> |
| #include <limits> |
| #include <memory> |
| #include <string> |
| #include <unordered_map> |
| #include <vector> |
| |
| #include <android-base/file.h> |
| #include <android-base/logging.h> |
| #include <android-base/parseint.h> |
| #include <android-base/strings.h> |
| #include <android-base/unique_fd.h> |
| #include <applypatch/applypatch.h> |
| #include <brotli/decode.h> |
| #include <fec/io.h> |
| #include <openssl/sha.h> |
| #include <private/android_filesystem_config.h> |
| #include <verity/hash_tree_builder.h> |
| #include <ziparchive/zip_archive.h> |
| |
| #include "edify/expr.h" |
| #include "otautil/dirutil.h" |
| #include "otautil/error_code.h" |
| #include "otautil/paths.h" |
| #include "otautil/print_sha1.h" |
| #include "otautil/rangeset.h" |
| #include "private/commands.h" |
| #include "updater/install.h" |
| #include "updater/updater.h" |
| |
| // Set this to 0 to interpret 'erase' transfers to mean do a |
| // BLKDISCARD ioctl (the normal behavior). Set to 1 to interpret |
| // erase to mean fill the region with zeroes. |
| #define DEBUG_ERASE 0 |
| |
| static constexpr size_t BLOCKSIZE = 4096; |
| static constexpr mode_t STASH_DIRECTORY_MODE = 0700; |
| static constexpr mode_t STASH_FILE_MODE = 0600; |
| static constexpr mode_t MARKER_DIRECTORY_MODE = 0700; |
| |
| static CauseCode failure_type = kNoCause; |
| static bool is_retry = false; |
| static std::unordered_map<std::string, RangeSet> stash_map; |
| |
| static void DeleteLastCommandFile() { |
| const std::string& last_command_file = Paths::Get().last_command_file(); |
| if (unlink(last_command_file.c_str()) == -1 && errno != ENOENT) { |
| PLOG(ERROR) << "Failed to unlink: " << last_command_file; |
| } |
| } |
| |
| // Parse the last command index of the last update and save the result to |last_command_index|. |
| // Return true if we successfully read the index. |
| static bool ParseLastCommandFile(size_t* last_command_index) { |
| const std::string& last_command_file = Paths::Get().last_command_file(); |
| android::base::unique_fd fd(TEMP_FAILURE_RETRY(open(last_command_file.c_str(), O_RDONLY))); |
| if (fd == -1) { |
| if (errno != ENOENT) { |
| PLOG(ERROR) << "Failed to open " << last_command_file; |
| return false; |
| } |
| |
| LOG(INFO) << last_command_file << " doesn't exist."; |
| return false; |
| } |
| |
| // Now that the last_command file exists, parse the last command index of previous update. |
| std::string content; |
| if (!android::base::ReadFdToString(fd.get(), &content)) { |
| LOG(ERROR) << "Failed to read: " << last_command_file; |
| return false; |
| } |
| |
| std::vector<std::string> lines = android::base::Split(android::base::Trim(content), "\n"); |
| if (lines.size() != 2) { |
| LOG(ERROR) << "Unexpected line counts in last command file: " << content; |
| return false; |
| } |
| |
| if (!android::base::ParseUint(lines[0], last_command_index)) { |
| LOG(ERROR) << "Failed to parse integer in: " << lines[0]; |
| return false; |
| } |
| |
| return true; |
| } |
| |
| static bool FsyncDir(const std::string& dirname) { |
| android::base::unique_fd dfd(TEMP_FAILURE_RETRY(open(dirname.c_str(), O_RDONLY | O_DIRECTORY))); |
| if (dfd == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFileOpenFailure; |
| PLOG(ERROR) << "Failed to open " << dirname; |
| return false; |
| } |
| if (fsync(dfd) == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFsyncFailure; |
| PLOG(ERROR) << "Failed to fsync " << dirname; |
| return false; |
| } |
| return true; |
| } |
| |
| // Update the last executed command index in the last_command_file. |
| static bool UpdateLastCommandIndex(size_t command_index, const std::string& command_string) { |
| const std::string& last_command_file = Paths::Get().last_command_file(); |
| std::string last_command_tmp = last_command_file + ".tmp"; |
| std::string content = std::to_string(command_index) + "\n" + command_string; |
| android::base::unique_fd wfd( |
| TEMP_FAILURE_RETRY(open(last_command_tmp.c_str(), O_WRONLY | O_CREAT | O_TRUNC, 0660))); |
| if (wfd == -1 || !android::base::WriteStringToFd(content, wfd)) { |
| PLOG(ERROR) << "Failed to update last command"; |
| return false; |
| } |
| |
| if (fsync(wfd) == -1) { |
| PLOG(ERROR) << "Failed to fsync " << last_command_tmp; |
| return false; |
| } |
| |
| if (chown(last_command_tmp.c_str(), AID_SYSTEM, AID_SYSTEM) == -1) { |
| PLOG(ERROR) << "Failed to change owner for " << last_command_tmp; |
| return false; |
| } |
| |
| if (rename(last_command_tmp.c_str(), last_command_file.c_str()) == -1) { |
| PLOG(ERROR) << "Failed to rename" << last_command_tmp; |
| return false; |
| } |
| |
| if (!FsyncDir(android::base::Dirname(last_command_file))) { |
| return false; |
| } |
| |
| return true; |
| } |
| |
| bool SetUpdatedMarker(const std::string& marker) { |
| auto dirname = android::base::Dirname(marker); |
| auto res = mkdir(dirname.c_str(), MARKER_DIRECTORY_MODE); |
| if (res == -1 && errno != EEXIST) { |
| PLOG(ERROR) << "Failed to create directory for marker: " << dirname; |
| return false; |
| } |
| |
| if (!android::base::WriteStringToFile("", marker)) { |
| PLOG(ERROR) << "Failed to write to marker file " << marker; |
| return false; |
| } |
| if (!FsyncDir(dirname)) { |
| return false; |
| } |
| LOG(INFO) << "Wrote updated marker to " << marker; |
| return true; |
| } |
| |
| static bool discard_blocks(int fd, off64_t offset, uint64_t size, bool force = false) { |
| // Don't discard blocks unless the update is a retry run or force == true |
| if (!is_retry && !force) { |
| return true; |
| } |
| |
| uint64_t args[2] = { static_cast<uint64_t>(offset), size }; |
| if (ioctl(fd, BLKDISCARD, &args) == -1) { |
| // On devices that does not support BLKDISCARD, ignore the error. |
| if (errno == EOPNOTSUPP) { |
| return true; |
| } |
| PLOG(ERROR) << "BLKDISCARD ioctl failed"; |
| return false; |
| } |
| return true; |
| } |
| |
| static bool check_lseek(int fd, off64_t offset, int whence) { |
| off64_t rc = TEMP_FAILURE_RETRY(lseek64(fd, offset, whence)); |
| if (rc == -1) { |
| failure_type = kLseekFailure; |
| PLOG(ERROR) << "lseek64 failed"; |
| return false; |
| } |
| return true; |
| } |
| |
| static void allocate(size_t size, std::vector<uint8_t>* buffer) { |
| // If the buffer's big enough, reuse it. |
| if (size <= buffer->size()) return; |
| buffer->resize(size); |
| } |
| |
| /** |
| * RangeSinkWriter reads data from the given FD, and writes them to the destination specified by the |
| * given RangeSet. |
| */ |
| class RangeSinkWriter { |
| public: |
| RangeSinkWriter(int fd, const RangeSet& tgt) |
| : fd_(fd), |
| tgt_(tgt), |
| next_range_(0), |
| current_range_left_(0), |
| bytes_written_(0) { |
| CHECK_NE(tgt.size(), static_cast<size_t>(0)); |
| }; |
| |
| bool Finished() const { |
| return next_range_ == tgt_.size() && current_range_left_ == 0; |
| } |
| |
| size_t AvailableSpace() const { |
| return tgt_.blocks() * BLOCKSIZE - bytes_written_; |
| } |
| |
| // Return number of bytes written; and 0 indicates a writing failure. |
| size_t Write(const uint8_t* data, size_t size) { |
| if (Finished()) { |
| LOG(ERROR) << "range sink write overrun; can't write " << size << " bytes"; |
| return 0; |
| } |
| |
| size_t written = 0; |
| while (size > 0) { |
| // Move to the next range as needed. |
| if (!SeekToOutputRange()) { |
| break; |
| } |
| |
| size_t write_now = size; |
| if (current_range_left_ < write_now) { |
| write_now = current_range_left_; |
| } |
| |
| if (!android::base::WriteFully(fd_, data, write_now)) { |
| failure_type = errno == EIO ? kEioFailure : kFwriteFailure; |
| PLOG(ERROR) << "Failed to write " << write_now << " bytes of data"; |
| break; |
| } |
| |
| data += write_now; |
| size -= write_now; |
| |
| current_range_left_ -= write_now; |
| written += write_now; |
| } |
| |
| bytes_written_ += written; |
| return written; |
| } |
| |
| size_t BytesWritten() const { |
| return bytes_written_; |
| } |
| |
| private: |
| // Set up the output cursor, move to next range if needed. |
| bool SeekToOutputRange() { |
| // We haven't finished the current range yet. |
| if (current_range_left_ != 0) { |
| return true; |
| } |
| // We can't write any more; let the write function return how many bytes have been written |
| // so far. |
| if (next_range_ >= tgt_.size()) { |
| return false; |
| } |
| |
| const Range& range = tgt_[next_range_]; |
| off64_t offset = static_cast<off64_t>(range.first) * BLOCKSIZE; |
| current_range_left_ = (range.second - range.first) * BLOCKSIZE; |
| next_range_++; |
| |
| if (!discard_blocks(fd_, offset, current_range_left_)) { |
| return false; |
| } |
| if (!check_lseek(fd_, offset, SEEK_SET)) { |
| return false; |
| } |
| return true; |
| } |
| |
| // The output file descriptor. |
| int fd_; |
| // The destination ranges for the data. |
| const RangeSet& tgt_; |
| // The next range that we should write to. |
| size_t next_range_; |
| // The number of bytes to write before moving to the next range. |
| size_t current_range_left_; |
| // Total bytes written by the writer. |
| size_t bytes_written_; |
| }; |
| |
| /** |
| * All of the data for all the 'new' transfers is contained in one file in the update package, |
| * concatenated together in the order in which transfers.list will need it. We want to stream it out |
| * of the archive (it's compressed) without writing it to a temp file, but we can't write each |
| * section until it's that transfer's turn to go. |
| * |
| * To achieve this, we expand the new data from the archive in a background thread, and block that |
| * threads 'receive uncompressed data' function until the main thread has reached a point where we |
| * want some new data to be written. We signal the background thread with the destination for the |
| * data and block the main thread, waiting for the background thread to complete writing that |
| * section. Then it signals the main thread to wake up and goes back to blocking waiting for a |
| * transfer. |
| * |
| * NewThreadInfo is the struct used to pass information back and forth between the two threads. When |
| * the main thread wants some data written, it sets writer to the destination location and signals |
| * the condition. When the background thread is done writing, it clears writer and signals the |
| * condition again. |
| */ |
| struct NewThreadInfo { |
| ZipArchiveHandle za; |
| ZipEntry entry; |
| bool brotli_compressed; |
| |
| std::unique_ptr<RangeSinkWriter> writer; |
| BrotliDecoderState* brotli_decoder_state; |
| bool receiver_available; |
| |
| pthread_mutex_t mu; |
| pthread_cond_t cv; |
| }; |
| |
| static bool receive_new_data(const uint8_t* data, size_t size, void* cookie) { |
| NewThreadInfo* nti = static_cast<NewThreadInfo*>(cookie); |
| |
| while (size > 0) { |
| // Wait for nti->writer to be non-null, indicating some of this data is wanted. |
| pthread_mutex_lock(&nti->mu); |
| while (nti->writer == nullptr) { |
| // End the new data receiver if we encounter an error when performing block image update. |
| if (!nti->receiver_available) { |
| pthread_mutex_unlock(&nti->mu); |
| return false; |
| } |
| pthread_cond_wait(&nti->cv, &nti->mu); |
| } |
| pthread_mutex_unlock(&nti->mu); |
| |
| // At this point nti->writer is set, and we own it. The main thread is waiting for it to |
| // disappear from nti. |
| size_t write_now = std::min(size, nti->writer->AvailableSpace()); |
| if (nti->writer->Write(data, write_now) != write_now) { |
| LOG(ERROR) << "Failed to write " << write_now << " bytes."; |
| return false; |
| } |
| |
| data += write_now; |
| size -= write_now; |
| |
| if (nti->writer->Finished()) { |
| // We have written all the bytes desired by this writer. |
| |
| pthread_mutex_lock(&nti->mu); |
| nti->writer = nullptr; |
| pthread_cond_broadcast(&nti->cv); |
| pthread_mutex_unlock(&nti->mu); |
| } |
| } |
| |
| return true; |
| } |
| |
| static bool receive_brotli_new_data(const uint8_t* data, size_t size, void* cookie) { |
| NewThreadInfo* nti = static_cast<NewThreadInfo*>(cookie); |
| |
| while (size > 0 || BrotliDecoderHasMoreOutput(nti->brotli_decoder_state)) { |
| // Wait for nti->writer to be non-null, indicating some of this data is wanted. |
| pthread_mutex_lock(&nti->mu); |
| while (nti->writer == nullptr) { |
| // End the receiver if we encounter an error when performing block image update. |
| if (!nti->receiver_available) { |
| pthread_mutex_unlock(&nti->mu); |
| return false; |
| } |
| pthread_cond_wait(&nti->cv, &nti->mu); |
| } |
| pthread_mutex_unlock(&nti->mu); |
| |
| // At this point nti->writer is set, and we own it. The main thread is waiting for it to |
| // disappear from nti. |
| |
| size_t buffer_size = std::min<size_t>(32768, nti->writer->AvailableSpace()); |
| if (buffer_size == 0) { |
| LOG(ERROR) << "No space left in output range"; |
| return false; |
| } |
| uint8_t buffer[buffer_size]; |
| size_t available_in = size; |
| size_t available_out = buffer_size; |
| uint8_t* next_out = buffer; |
| |
| // The brotli decoder will update |data|, |available_in|, |next_out| and |available_out|. |
| BrotliDecoderResult result = BrotliDecoderDecompressStream( |
| nti->brotli_decoder_state, &available_in, &data, &available_out, &next_out, nullptr); |
| |
| if (result == BROTLI_DECODER_RESULT_ERROR) { |
| LOG(ERROR) << "Decompression failed with " |
| << BrotliDecoderErrorString(BrotliDecoderGetErrorCode(nti->brotli_decoder_state)); |
| return false; |
| } |
| |
| LOG(DEBUG) << "bytes to write: " << buffer_size - available_out << ", bytes consumed " |
| << size - available_in << ", decoder status " << result; |
| |
| size_t write_now = buffer_size - available_out; |
| if (nti->writer->Write(buffer, write_now) != write_now) { |
| LOG(ERROR) << "Failed to write " << write_now << " bytes."; |
| return false; |
| } |
| |
| // Update the remaining size. The input data ptr is already updated by brotli decoder function. |
| size = available_in; |
| |
| if (nti->writer->Finished()) { |
| // We have written all the bytes desired by this writer. |
| |
| pthread_mutex_lock(&nti->mu); |
| nti->writer = nullptr; |
| pthread_cond_broadcast(&nti->cv); |
| pthread_mutex_unlock(&nti->mu); |
| } |
| } |
| |
| return true; |
| } |
| |
| static void* unzip_new_data(void* cookie) { |
| NewThreadInfo* nti = static_cast<NewThreadInfo*>(cookie); |
| if (nti->brotli_compressed) { |
| ProcessZipEntryContents(nti->za, &nti->entry, receive_brotli_new_data, nti); |
| } else { |
| ProcessZipEntryContents(nti->za, &nti->entry, receive_new_data, nti); |
| } |
| pthread_mutex_lock(&nti->mu); |
| nti->receiver_available = false; |
| if (nti->writer != nullptr) { |
| pthread_cond_broadcast(&nti->cv); |
| } |
| pthread_mutex_unlock(&nti->mu); |
| return nullptr; |
| } |
| |
| static int ReadBlocks(const RangeSet& src, std::vector<uint8_t>* buffer, int fd) { |
| size_t p = 0; |
| for (const auto& [begin, end] : src) { |
| if (!check_lseek(fd, static_cast<off64_t>(begin) * BLOCKSIZE, SEEK_SET)) { |
| return -1; |
| } |
| |
| size_t size = (end - begin) * BLOCKSIZE; |
| if (!android::base::ReadFully(fd, buffer->data() + p, size)) { |
| failure_type = errno == EIO ? kEioFailure : kFreadFailure; |
| PLOG(ERROR) << "Failed to read " << size << " bytes of data"; |
| return -1; |
| } |
| |
| p += size; |
| } |
| |
| return 0; |
| } |
| |
| static int WriteBlocks(const RangeSet& tgt, const std::vector<uint8_t>& buffer, int fd) { |
| size_t written = 0; |
| for (const auto& [begin, end] : tgt) { |
| off64_t offset = static_cast<off64_t>(begin) * BLOCKSIZE; |
| size_t size = (end - begin) * BLOCKSIZE; |
| if (!discard_blocks(fd, offset, size)) { |
| return -1; |
| } |
| |
| if (!check_lseek(fd, offset, SEEK_SET)) { |
| return -1; |
| } |
| |
| if (!android::base::WriteFully(fd, buffer.data() + written, size)) { |
| failure_type = errno == EIO ? kEioFailure : kFwriteFailure; |
| PLOG(ERROR) << "Failed to write " << size << " bytes of data"; |
| return -1; |
| } |
| |
| written += size; |
| } |
| |
| return 0; |
| } |
| |
| // Parameters for transfer list command functions |
| struct CommandParameters { |
| std::vector<std::string> tokens; |
| size_t cpos; |
| std::string cmdname; |
| std::string cmdline; |
| std::string freestash; |
| std::string stashbase; |
| bool canwrite; |
| int createdstash; |
| android::base::unique_fd fd; |
| bool foundwrites; |
| bool isunresumable; |
| int version; |
| size_t written; |
| size_t stashed; |
| NewThreadInfo nti; |
| pthread_t thread; |
| std::vector<uint8_t> buffer; |
| uint8_t* patch_start; |
| bool target_verified; // The target blocks have expected contents already. |
| }; |
| |
| // Print the hash in hex for corrupted source blocks (excluding the stashed blocks which is |
| // handled separately). |
| static void PrintHashForCorruptedSourceBlocks(const CommandParameters& params, |
| const std::vector<uint8_t>& buffer) { |
| LOG(INFO) << "unexpected contents of source blocks in cmd:\n" << params.cmdline; |
| CHECK(params.tokens[0] == "move" || params.tokens[0] == "bsdiff" || |
| params.tokens[0] == "imgdiff"); |
| |
| size_t pos = 0; |
| // Command example: |
| // move <onehash> <tgt_range> <src_blk_count> <src_range> [<loc_range> <stashed_blocks>] |
| // bsdiff <offset> <len> <src_hash> <tgt_hash> <tgt_range> <src_blk_count> <src_range> |
| // [<loc_range> <stashed_blocks>] |
| if (params.tokens[0] == "move") { |
| // src_range for move starts at the 4th position. |
| if (params.tokens.size() < 5) { |
| LOG(ERROR) << "failed to parse source range in cmd:\n" << params.cmdline; |
| return; |
| } |
| pos = 4; |
| } else { |
| // src_range for diff starts at the 7th position. |
| if (params.tokens.size() < 8) { |
| LOG(ERROR) << "failed to parse source range in cmd:\n" << params.cmdline; |
| return; |
| } |
| pos = 7; |
| } |
| |
| // Source blocks in stash only, no work to do. |
| if (params.tokens[pos] == "-") { |
| return; |
| } |
| |
| RangeSet src = RangeSet::Parse(params.tokens[pos++]); |
| if (!src) { |
| LOG(ERROR) << "Failed to parse range in " << params.cmdline; |
| return; |
| } |
| |
| RangeSet locs; |
| // If there's no stashed blocks, content in the buffer is consecutive and has the same |
| // order as the source blocks. |
| if (pos == params.tokens.size()) { |
| locs = RangeSet(std::vector<Range>{ Range{ 0, src.blocks() } }); |
| } else { |
| // Otherwise, the next token is the offset of the source blocks in the target range. |
| // Example: for the tokens <4,63946,63947,63948,63979> <4,6,7,8,39> <stashed_blocks>; |
| // We want to print SHA-1 for the data in buffer[6], buffer[8], buffer[9] ... buffer[38]; |
| // this corresponds to the 32 src blocks #63946, #63948, #63949 ... #63978. |
| locs = RangeSet::Parse(params.tokens[pos++]); |
| CHECK_EQ(src.blocks(), locs.blocks()); |
| } |
| |
| LOG(INFO) << "printing hash in hex for " << src.blocks() << " source blocks"; |
| for (size_t i = 0; i < src.blocks(); i++) { |
| size_t block_num = src.GetBlockNumber(i); |
| size_t buffer_index = locs.GetBlockNumber(i); |
| CHECK_LE((buffer_index + 1) * BLOCKSIZE, buffer.size()); |
| |
| uint8_t digest[SHA_DIGEST_LENGTH]; |
| SHA1(buffer.data() + buffer_index * BLOCKSIZE, BLOCKSIZE, digest); |
| std::string hexdigest = print_sha1(digest); |
| LOG(INFO) << " block number: " << block_num << ", SHA-1: " << hexdigest; |
| } |
| } |
| |
| // If the calculated hash for the whole stash doesn't match the stash id, print the SHA-1 |
| // in hex for each block. |
| static void PrintHashForCorruptedStashedBlocks(const std::string& id, |
| const std::vector<uint8_t>& buffer, |
| const RangeSet& src) { |
| LOG(INFO) << "printing hash in hex for stash_id: " << id; |
| CHECK_EQ(src.blocks() * BLOCKSIZE, buffer.size()); |
| |
| for (size_t i = 0; i < src.blocks(); i++) { |
| size_t block_num = src.GetBlockNumber(i); |
| |
| uint8_t digest[SHA_DIGEST_LENGTH]; |
| SHA1(buffer.data() + i * BLOCKSIZE, BLOCKSIZE, digest); |
| std::string hexdigest = print_sha1(digest); |
| LOG(INFO) << " block number: " << block_num << ", SHA-1: " << hexdigest; |
| } |
| } |
| |
| // If the stash file doesn't exist, read the source blocks this stash contains and print the |
| // SHA-1 for these blocks. |
| static void PrintHashForMissingStashedBlocks(const std::string& id, int fd) { |
| if (stash_map.find(id) == stash_map.end()) { |
| LOG(ERROR) << "No stash saved for id: " << id; |
| return; |
| } |
| |
| LOG(INFO) << "print hash in hex for source blocks in missing stash: " << id; |
| const RangeSet& src = stash_map[id]; |
| std::vector<uint8_t> buffer(src.blocks() * BLOCKSIZE); |
| if (ReadBlocks(src, &buffer, fd) == -1) { |
| LOG(ERROR) << "failed to read source blocks for stash: " << id; |
| return; |
| } |
| PrintHashForCorruptedStashedBlocks(id, buffer, src); |
| } |
| |
| static int VerifyBlocks(const std::string& expected, const std::vector<uint8_t>& buffer, |
| const size_t blocks, bool printerror) { |
| uint8_t digest[SHA_DIGEST_LENGTH]; |
| const uint8_t* data = buffer.data(); |
| |
| SHA1(data, blocks * BLOCKSIZE, digest); |
| |
| std::string hexdigest = print_sha1(digest); |
| |
| if (hexdigest != expected) { |
| if (printerror) { |
| LOG(ERROR) << "failed to verify blocks (expected " << expected << ", read " << hexdigest |
| << ")"; |
| } |
| return -1; |
| } |
| |
| return 0; |
| } |
| |
| static std::string GetStashFileName(const std::string& base, const std::string& id, |
| const std::string& postfix) { |
| if (base.empty()) { |
| return ""; |
| } |
| std::string filename = Paths::Get().stash_directory_base() + "/" + base; |
| if (id.empty() && postfix.empty()) { |
| return filename; |
| } |
| return filename + "/" + id + postfix; |
| } |
| |
| // Does a best effort enumeration of stash files. Ignores possible non-file items in the stash |
| // directory and continues despite of errors. Calls the 'callback' function for each file. |
| static void EnumerateStash(const std::string& dirname, |
| const std::function<void(const std::string&)>& callback) { |
| if (dirname.empty()) return; |
| |
| std::unique_ptr<DIR, decltype(&closedir)> directory(opendir(dirname.c_str()), closedir); |
| |
| if (directory == nullptr) { |
| if (errno != ENOENT) { |
| PLOG(ERROR) << "opendir \"" << dirname << "\" failed"; |
| } |
| return; |
| } |
| |
| dirent* item; |
| while ((item = readdir(directory.get())) != nullptr) { |
| if (item->d_type != DT_REG) continue; |
| callback(dirname + "/" + item->d_name); |
| } |
| } |
| |
| // Deletes the stash directory and all files in it. Assumes that it only |
| // contains files. There is nothing we can do about unlikely, but possible |
| // errors, so they are merely logged. |
| static void DeleteFile(const std::string& fn) { |
| if (fn.empty()) return; |
| |
| LOG(INFO) << "deleting " << fn; |
| |
| if (unlink(fn.c_str()) == -1 && errno != ENOENT) { |
| PLOG(ERROR) << "unlink \"" << fn << "\" failed"; |
| } |
| } |
| |
| static void DeleteStash(const std::string& base) { |
| if (base.empty()) return; |
| |
| LOG(INFO) << "deleting stash " << base; |
| |
| std::string dirname = GetStashFileName(base, "", ""); |
| EnumerateStash(dirname, DeleteFile); |
| |
| if (rmdir(dirname.c_str()) == -1) { |
| if (errno != ENOENT && errno != ENOTDIR) { |
| PLOG(ERROR) << "rmdir \"" << dirname << "\" failed"; |
| } |
| } |
| } |
| |
| static int LoadStash(const CommandParameters& params, const std::string& id, bool verify, |
| std::vector<uint8_t>* buffer, bool printnoent) { |
| // In verify mode, if source range_set was saved for the given hash, check contents in the source |
| // blocks first. If the check fails, search for the stashed files on /cache as usual. |
| if (!params.canwrite) { |
| if (stash_map.find(id) != stash_map.end()) { |
| const RangeSet& src = stash_map[id]; |
| allocate(src.blocks() * BLOCKSIZE, buffer); |
| |
| if (ReadBlocks(src, buffer, params.fd) == -1) { |
| LOG(ERROR) << "failed to read source blocks in stash map."; |
| return -1; |
| } |
| if (VerifyBlocks(id, *buffer, src.blocks(), true) != 0) { |
| LOG(ERROR) << "failed to verify loaded source blocks in stash map."; |
| if (!is_retry) { |
| PrintHashForCorruptedStashedBlocks(id, *buffer, src); |
| } |
| return -1; |
| } |
| return 0; |
| } |
| } |
| |
| std::string fn = GetStashFileName(params.stashbase, id, ""); |
| |
| struct stat sb; |
| if (stat(fn.c_str(), &sb) == -1) { |
| if (errno != ENOENT || printnoent) { |
| PLOG(ERROR) << "stat \"" << fn << "\" failed"; |
| PrintHashForMissingStashedBlocks(id, params.fd); |
| } |
| return -1; |
| } |
| |
| LOG(INFO) << " loading " << fn; |
| |
| if ((sb.st_size % BLOCKSIZE) != 0) { |
| LOG(ERROR) << fn << " size " << sb.st_size << " not multiple of block size " << BLOCKSIZE; |
| return -1; |
| } |
| |
| android::base::unique_fd fd(TEMP_FAILURE_RETRY(open(fn.c_str(), O_RDONLY))); |
| if (fd == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFileOpenFailure; |
| PLOG(ERROR) << "open \"" << fn << "\" failed"; |
| return -1; |
| } |
| |
| allocate(sb.st_size, buffer); |
| |
| if (!android::base::ReadFully(fd, buffer->data(), sb.st_size)) { |
| failure_type = errno == EIO ? kEioFailure : kFreadFailure; |
| PLOG(ERROR) << "Failed to read " << sb.st_size << " bytes of data"; |
| return -1; |
| } |
| |
| size_t blocks = sb.st_size / BLOCKSIZE; |
| if (verify && VerifyBlocks(id, *buffer, blocks, true) != 0) { |
| LOG(ERROR) << "unexpected contents in " << fn; |
| if (stash_map.find(id) == stash_map.end()) { |
| LOG(ERROR) << "failed to find source blocks number for stash " << id |
| << " when executing command: " << params.cmdname; |
| } else { |
| const RangeSet& src = stash_map[id]; |
| PrintHashForCorruptedStashedBlocks(id, *buffer, src); |
| } |
| DeleteFile(fn); |
| return -1; |
| } |
| |
| return 0; |
| } |
| |
| static int WriteStash(const std::string& base, const std::string& id, int blocks, |
| const std::vector<uint8_t>& buffer, bool checkspace, bool* exists) { |
| if (base.empty()) { |
| return -1; |
| } |
| |
| if (checkspace && !CheckAndFreeSpaceOnCache(blocks * BLOCKSIZE)) { |
| LOG(ERROR) << "not enough space to write stash"; |
| return -1; |
| } |
| |
| std::string fn = GetStashFileName(base, id, ".partial"); |
| std::string cn = GetStashFileName(base, id, ""); |
| |
| if (exists) { |
| struct stat sb; |
| int res = stat(cn.c_str(), &sb); |
| |
| if (res == 0) { |
| // The file already exists and since the name is the hash of the contents, |
| // it's safe to assume the contents are identical (accidental hash collisions |
| // are unlikely) |
| LOG(INFO) << " skipping " << blocks << " existing blocks in " << cn; |
| *exists = true; |
| return 0; |
| } |
| |
| *exists = false; |
| } |
| |
| LOG(INFO) << " writing " << blocks << " blocks to " << cn; |
| |
| android::base::unique_fd fd( |
| TEMP_FAILURE_RETRY(open(fn.c_str(), O_WRONLY | O_CREAT | O_TRUNC, STASH_FILE_MODE))); |
| if (fd == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFileOpenFailure; |
| PLOG(ERROR) << "failed to create \"" << fn << "\""; |
| return -1; |
| } |
| |
| if (fchown(fd, AID_SYSTEM, AID_SYSTEM) != 0) { // system user |
| PLOG(ERROR) << "failed to chown \"" << fn << "\""; |
| return -1; |
| } |
| |
| if (!android::base::WriteFully(fd, buffer.data(), blocks * BLOCKSIZE)) { |
| failure_type = errno == EIO ? kEioFailure : kFwriteFailure; |
| PLOG(ERROR) << "Failed to write " << blocks * BLOCKSIZE << " bytes of data"; |
| return -1; |
| } |
| |
| if (fsync(fd) == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFsyncFailure; |
| PLOG(ERROR) << "fsync \"" << fn << "\" failed"; |
| return -1; |
| } |
| |
| if (rename(fn.c_str(), cn.c_str()) == -1) { |
| PLOG(ERROR) << "rename(\"" << fn << "\", \"" << cn << "\") failed"; |
| return -1; |
| } |
| |
| std::string dname = GetStashFileName(base, "", ""); |
| if (!FsyncDir(dname)) { |
| return -1; |
| } |
| |
| return 0; |
| } |
| |
| // Creates a directory for storing stash files and checks if the /cache partition |
| // hash enough space for the expected amount of blocks we need to store. Returns |
| // >0 if we created the directory, zero if it existed already, and <0 of failure. |
| static int CreateStash(State* state, size_t maxblocks, const std::string& base) { |
| std::string dirname = GetStashFileName(base, "", ""); |
| struct stat sb; |
| int res = stat(dirname.c_str(), &sb); |
| if (res == -1 && errno != ENOENT) { |
| ErrorAbort(state, kStashCreationFailure, "stat \"%s\" failed: %s", dirname.c_str(), |
| strerror(errno)); |
| return -1; |
| } |
| |
| size_t max_stash_size = maxblocks * BLOCKSIZE; |
| if (res == -1) { |
| LOG(INFO) << "creating stash " << dirname; |
| res = mkdir_recursively(dirname, STASH_DIRECTORY_MODE, false, nullptr); |
| |
| if (res != 0) { |
| ErrorAbort(state, kStashCreationFailure, "mkdir \"%s\" failed: %s", dirname.c_str(), |
| strerror(errno)); |
| return -1; |
| } |
| |
| if (chown(dirname.c_str(), AID_SYSTEM, AID_SYSTEM) != 0) { // system user |
| ErrorAbort(state, kStashCreationFailure, "chown \"%s\" failed: %s", dirname.c_str(), |
| strerror(errno)); |
| return -1; |
| } |
| |
| if (!CheckAndFreeSpaceOnCache(max_stash_size)) { |
| ErrorAbort(state, kStashCreationFailure, "not enough space for stash (%zu needed)", |
| max_stash_size); |
| return -1; |
| } |
| |
| return 1; // Created directory |
| } |
| |
| LOG(INFO) << "using existing stash " << dirname; |
| |
| // If the directory already exists, calculate the space already allocated to stash files and check |
| // if there's enough for all required blocks. Delete any partially completed stash files first. |
| EnumerateStash(dirname, [](const std::string& fn) { |
| if (android::base::EndsWith(fn, ".partial")) { |
| DeleteFile(fn); |
| } |
| }); |
| |
| size_t existing = 0; |
| EnumerateStash(dirname, [&existing](const std::string& fn) { |
| if (fn.empty()) return; |
| struct stat sb; |
| if (stat(fn.c_str(), &sb) == -1) { |
| PLOG(ERROR) << "stat \"" << fn << "\" failed"; |
| return; |
| } |
| existing += static_cast<size_t>(sb.st_size); |
| }); |
| |
| if (max_stash_size > existing) { |
| size_t needed = max_stash_size - existing; |
| if (!CheckAndFreeSpaceOnCache(needed)) { |
| ErrorAbort(state, kStashCreationFailure, "not enough space for stash (%zu more needed)", |
| needed); |
| return -1; |
| } |
| } |
| |
| return 0; // Using existing directory |
| } |
| |
| static int FreeStash(const std::string& base, const std::string& id) { |
| if (base.empty() || id.empty()) { |
| return -1; |
| } |
| |
| DeleteFile(GetStashFileName(base, id, "")); |
| |
| return 0; |
| } |
| |
| // Source contains packed data, which we want to move to the locations given in locs in the dest |
| // buffer. source and dest may be the same buffer. |
| static void MoveRange(std::vector<uint8_t>& dest, const RangeSet& locs, |
| const std::vector<uint8_t>& source) { |
| const uint8_t* from = source.data(); |
| uint8_t* to = dest.data(); |
| size_t start = locs.blocks(); |
| // Must do the movement backward. |
| for (auto it = locs.crbegin(); it != locs.crend(); it++) { |
| size_t blocks = it->second - it->first; |
| start -= blocks; |
| memmove(to + (it->first * BLOCKSIZE), from + (start * BLOCKSIZE), blocks * BLOCKSIZE); |
| } |
| } |
| |
| /** |
| * We expect to parse the remainder of the parameter tokens as one of: |
| * |
| * <src_block_count> <src_range> |
| * (loads data from source image only) |
| * |
| * <src_block_count> - <[stash_id:stash_range] ...> |
| * (loads data from stashes only) |
| * |
| * <src_block_count> <src_range> <src_loc> <[stash_id:stash_range] ...> |
| * (loads data from both source image and stashes) |
| * |
| * On return, params.buffer is filled with the loaded source data (rearranged and combined with |
| * stashed data as necessary). buffer may be reallocated if needed to accommodate the source data. |
| * tgt is the target RangeSet for detecting overlaps. Any stashes required are loaded using |
| * LoadStash. |
| */ |
| static int LoadSourceBlocks(CommandParameters& params, const RangeSet& tgt, size_t* src_blocks, |
| bool* overlap) { |
| CHECK(src_blocks != nullptr); |
| CHECK(overlap != nullptr); |
| |
| // <src_block_count> |
| const std::string& token = params.tokens[params.cpos++]; |
| if (!android::base::ParseUint(token, src_blocks)) { |
| LOG(ERROR) << "invalid src_block_count \"" << token << "\""; |
| return -1; |
| } |
| |
| allocate(*src_blocks * BLOCKSIZE, ¶ms.buffer); |
| |
| // "-" or <src_range> [<src_loc>] |
| if (params.tokens[params.cpos] == "-") { |
| // no source ranges, only stashes |
| params.cpos++; |
| } else { |
| RangeSet src = RangeSet::Parse(params.tokens[params.cpos++]); |
| CHECK(static_cast<bool>(src)); |
| *overlap = src.Overlaps(tgt); |
| |
| if (ReadBlocks(src, ¶ms.buffer, params.fd) == -1) { |
| return -1; |
| } |
| |
| if (params.cpos >= params.tokens.size()) { |
| // no stashes, only source range |
| return 0; |
| } |
| |
| RangeSet locs = RangeSet::Parse(params.tokens[params.cpos++]); |
| CHECK(static_cast<bool>(locs)); |
| MoveRange(params.buffer, locs, params.buffer); |
| } |
| |
| // <[stash_id:stash_range]> |
| while (params.cpos < params.tokens.size()) { |
| // Each word is a an index into the stash table, a colon, and then a RangeSet describing where |
| // in the source block that stashed data should go. |
| std::vector<std::string> tokens = android::base::Split(params.tokens[params.cpos++], ":"); |
| if (tokens.size() != 2) { |
| LOG(ERROR) << "invalid parameter"; |
| return -1; |
| } |
| |
| std::vector<uint8_t> stash; |
| if (LoadStash(params, tokens[0], false, &stash, true) == -1) { |
| // These source blocks will fail verification if used later, but we |
| // will let the caller decide if this is a fatal failure |
| LOG(ERROR) << "failed to load stash " << tokens[0]; |
| continue; |
| } |
| |
| RangeSet locs = RangeSet::Parse(tokens[1]); |
| CHECK(static_cast<bool>(locs)); |
| MoveRange(params.buffer, locs, stash); |
| } |
| |
| return 0; |
| } |
| |
| /** |
| * Do a source/target load for move/bsdiff/imgdiff in version 3. |
| * |
| * We expect to parse the remainder of the parameter tokens as one of: |
| * |
| * <tgt_range> <src_block_count> <src_range> |
| * (loads data from source image only) |
| * |
| * <tgt_range> <src_block_count> - <[stash_id:stash_range] ...> |
| * (loads data from stashes only) |
| * |
| * <tgt_range> <src_block_count> <src_range> <src_loc> <[stash_id:stash_range] ...> |
| * (loads data from both source image and stashes) |
| * |
| * 'onehash' tells whether to expect separate source and targe block hashes, or if they are both the |
| * same and only one hash should be expected. params.isunresumable will be set to true if block |
| * verification fails in a way that the update cannot be resumed anymore. |
| * |
| * If the function is unable to load the necessary blocks or their contents don't match the hashes, |
| * the return value is -1 and the command should be aborted. |
| * |
| * If the return value is 1, the command has already been completed according to the contents of the |
| * target blocks, and should not be performed again. |
| * |
| * If the return value is 0, source blocks have expected content and the command can be performed. |
| */ |
| static int LoadSrcTgtVersion3(CommandParameters& params, RangeSet* tgt, size_t* src_blocks, |
| bool onehash) { |
| CHECK(src_blocks != nullptr); |
| |
| if (params.cpos >= params.tokens.size()) { |
| LOG(ERROR) << "missing source hash"; |
| return -1; |
| } |
| |
| std::string srchash = params.tokens[params.cpos++]; |
| std::string tgthash; |
| |
| if (onehash) { |
| tgthash = srchash; |
| } else { |
| if (params.cpos >= params.tokens.size()) { |
| LOG(ERROR) << "missing target hash"; |
| return -1; |
| } |
| tgthash = params.tokens[params.cpos++]; |
| } |
| |
| // At least it needs to provide three parameters: <tgt_range>, <src_block_count> and |
| // "-"/<src_range>. |
| if (params.cpos + 2 >= params.tokens.size()) { |
| LOG(ERROR) << "invalid parameters"; |
| return -1; |
| } |
| |
| // <tgt_range> |
| *tgt = RangeSet::Parse(params.tokens[params.cpos++]); |
| CHECK(static_cast<bool>(*tgt)); |
| |
| std::vector<uint8_t> tgtbuffer(tgt->blocks() * BLOCKSIZE); |
| if (ReadBlocks(*tgt, &tgtbuffer, params.fd) == -1) { |
| return -1; |
| } |
| |
| // Return now if target blocks already have expected content. |
| if (VerifyBlocks(tgthash, tgtbuffer, tgt->blocks(), false) == 0) { |
| return 1; |
| } |
| |
| // Load source blocks. |
| bool overlap = false; |
| if (LoadSourceBlocks(params, *tgt, src_blocks, &overlap) == -1) { |
| return -1; |
| } |
| |
| if (VerifyBlocks(srchash, params.buffer, *src_blocks, true) == 0) { |
| // If source and target blocks overlap, stash the source blocks so we can resume from possible |
| // write errors. In verify mode, we can skip stashing because the source blocks won't be |
| // overwritten. |
| if (overlap && params.canwrite) { |
| LOG(INFO) << "stashing " << *src_blocks << " overlapping blocks to " << srchash; |
| |
| bool stash_exists = false; |
| if (WriteStash(params.stashbase, srchash, *src_blocks, params.buffer, true, |
| &stash_exists) != 0) { |
| LOG(ERROR) << "failed to stash overlapping source blocks"; |
| return -1; |
| } |
| |
| params.stashed += *src_blocks; |
| // Can be deleted when the write has completed. |
| if (!stash_exists) { |
| params.freestash = srchash; |
| } |
| } |
| |
| // Source blocks have expected content, command can proceed. |
| return 0; |
| } |
| |
| if (overlap && LoadStash(params, srchash, true, ¶ms.buffer, true) == 0) { |
| // Overlapping source blocks were previously stashed, command can proceed. We are recovering |
| // from an interrupted command, so we don't know if the stash can safely be deleted after this |
| // command. |
| return 0; |
| } |
| |
| // Valid source data not available, update cannot be resumed. |
| LOG(ERROR) << "partition has unexpected contents"; |
| PrintHashForCorruptedSourceBlocks(params, params.buffer); |
| |
| params.isunresumable = true; |
| |
| return -1; |
| } |
| |
| static int PerformCommandMove(CommandParameters& params) { |
| size_t blocks = 0; |
| RangeSet tgt; |
| int status = LoadSrcTgtVersion3(params, &tgt, &blocks, true); |
| |
| if (status == -1) { |
| LOG(ERROR) << "failed to read blocks for move"; |
| return -1; |
| } |
| |
| if (status == 0) { |
| params.foundwrites = true; |
| } else { |
| params.target_verified = true; |
| if (params.foundwrites) { |
| LOG(WARNING) << "warning: commands executed out of order [" << params.cmdname << "]"; |
| } |
| } |
| |
| if (params.canwrite) { |
| if (status == 0) { |
| LOG(INFO) << " moving " << blocks << " blocks"; |
| |
| if (WriteBlocks(tgt, params.buffer, params.fd) == -1) { |
| return -1; |
| } |
| } else { |
| LOG(INFO) << "skipping " << blocks << " already moved blocks"; |
| } |
| } |
| |
| if (!params.freestash.empty()) { |
| FreeStash(params.stashbase, params.freestash); |
| params.freestash.clear(); |
| } |
| |
| params.written += tgt.blocks(); |
| |
| return 0; |
| } |
| |
| static int PerformCommandStash(CommandParameters& params) { |
| // <stash_id> <src_range> |
| if (params.cpos + 1 >= params.tokens.size()) { |
| LOG(ERROR) << "missing id and/or src range fields in stash command"; |
| return -1; |
| } |
| |
| const std::string& id = params.tokens[params.cpos++]; |
| if (LoadStash(params, id, true, ¶ms.buffer, false) == 0) { |
| // Stash file already exists and has expected contents. Do not read from source again, as the |
| // source may have been already overwritten during a previous attempt. |
| return 0; |
| } |
| |
| RangeSet src = RangeSet::Parse(params.tokens[params.cpos++]); |
| CHECK(static_cast<bool>(src)); |
| |
| size_t blocks = src.blocks(); |
| allocate(blocks * BLOCKSIZE, ¶ms.buffer); |
| if (ReadBlocks(src, ¶ms.buffer, params.fd) == -1) { |
| return -1; |
| } |
| stash_map[id] = src; |
| |
| if (VerifyBlocks(id, params.buffer, blocks, true) != 0) { |
| // Source blocks have unexpected contents. If we actually need this data later, this is an |
| // unrecoverable error. However, the command that uses the data may have already completed |
| // previously, so the possible failure will occur during source block verification. |
| LOG(ERROR) << "failed to load source blocks for stash " << id; |
| return 0; |
| } |
| |
| // In verify mode, we don't need to stash any blocks. |
| if (!params.canwrite) { |
| return 0; |
| } |
| |
| LOG(INFO) << "stashing " << blocks << " blocks to " << id; |
| int result = WriteStash(params.stashbase, id, blocks, params.buffer, false, nullptr); |
| if (result == 0) { |
| params.stashed += blocks; |
| } |
| return result; |
| } |
| |
| static int PerformCommandFree(CommandParameters& params) { |
| // <stash_id> |
| if (params.cpos >= params.tokens.size()) { |
| LOG(ERROR) << "missing stash id in free command"; |
| return -1; |
| } |
| |
| const std::string& id = params.tokens[params.cpos++]; |
| stash_map.erase(id); |
| |
| if (params.createdstash || params.canwrite) { |
| return FreeStash(params.stashbase, id); |
| } |
| |
| return 0; |
| } |
| |
| static int PerformCommandZero(CommandParameters& params) { |
| if (params.cpos >= params.tokens.size()) { |
| LOG(ERROR) << "missing target blocks for zero"; |
| return -1; |
| } |
| |
| RangeSet tgt = RangeSet::Parse(params.tokens[params.cpos++]); |
| CHECK(static_cast<bool>(tgt)); |
| |
| LOG(INFO) << " zeroing " << tgt.blocks() << " blocks"; |
| |
| allocate(BLOCKSIZE, ¶ms.buffer); |
| memset(params.buffer.data(), 0, BLOCKSIZE); |
| |
| if (params.canwrite) { |
| for (const auto& [begin, end] : tgt) { |
| off64_t offset = static_cast<off64_t>(begin) * BLOCKSIZE; |
| size_t size = (end - begin) * BLOCKSIZE; |
| if (!discard_blocks(params.fd, offset, size)) { |
| return -1; |
| } |
| |
| if (!check_lseek(params.fd, offset, SEEK_SET)) { |
| return -1; |
| } |
| |
| for (size_t j = begin; j < end; ++j) { |
| if (!android::base::WriteFully(params.fd, params.buffer.data(), BLOCKSIZE)) { |
| failure_type = errno == EIO ? kEioFailure : kFwriteFailure; |
| PLOG(ERROR) << "Failed to write " << BLOCKSIZE << " bytes of data"; |
| return -1; |
| } |
| } |
| } |
| } |
| |
| if (params.cmdname[0] == 'z') { |
| // Update only for the zero command, as the erase command will call |
| // this if DEBUG_ERASE is defined. |
| params.written += tgt.blocks(); |
| } |
| |
| return 0; |
| } |
| |
| static int PerformCommandNew(CommandParameters& params) { |
| if (params.cpos >= params.tokens.size()) { |
| LOG(ERROR) << "missing target blocks for new"; |
| return -1; |
| } |
| |
| RangeSet tgt = RangeSet::Parse(params.tokens[params.cpos++]); |
| CHECK(static_cast<bool>(tgt)); |
| |
| if (params.canwrite) { |
| LOG(INFO) << " writing " << tgt.blocks() << " blocks of new data"; |
| |
| pthread_mutex_lock(¶ms.nti.mu); |
| params.nti.writer = std::make_unique<RangeSinkWriter>(params.fd, tgt); |
| pthread_cond_broadcast(¶ms.nti.cv); |
| |
| while (params.nti.writer != nullptr) { |
| if (!params.nti.receiver_available) { |
| LOG(ERROR) << "missing " << (tgt.blocks() * BLOCKSIZE - params.nti.writer->BytesWritten()) |
| << " bytes of new data"; |
| pthread_mutex_unlock(¶ms.nti.mu); |
| return -1; |
| } |
| pthread_cond_wait(¶ms.nti.cv, ¶ms.nti.mu); |
| } |
| |
| pthread_mutex_unlock(¶ms.nti.mu); |
| } |
| |
| params.written += tgt.blocks(); |
| |
| return 0; |
| } |
| |
| static int PerformCommandDiff(CommandParameters& params) { |
| // <offset> <length> |
| if (params.cpos + 1 >= params.tokens.size()) { |
| LOG(ERROR) << "missing patch offset or length for " << params.cmdname; |
| return -1; |
| } |
| |
| size_t offset; |
| if (!android::base::ParseUint(params.tokens[params.cpos++], &offset)) { |
| LOG(ERROR) << "invalid patch offset"; |
| return -1; |
| } |
| |
| size_t len; |
| if (!android::base::ParseUint(params.tokens[params.cpos++], &len)) { |
| LOG(ERROR) << "invalid patch len"; |
| return -1; |
| } |
| |
| RangeSet tgt; |
| size_t blocks = 0; |
| int status = LoadSrcTgtVersion3(params, &tgt, &blocks, false); |
| |
| if (status == -1) { |
| LOG(ERROR) << "failed to read blocks for diff"; |
| return -1; |
| } |
| |
| if (status == 0) { |
| params.foundwrites = true; |
| } else { |
| params.target_verified = true; |
| if (params.foundwrites) { |
| LOG(WARNING) << "warning: commands executed out of order [" << params.cmdname << "]"; |
| } |
| } |
| |
| if (params.canwrite) { |
| if (status == 0) { |
| LOG(INFO) << "patching " << blocks << " blocks to " << tgt.blocks(); |
| Value patch_value( |
| Value::Type::BLOB, |
| std::string(reinterpret_cast<const char*>(params.patch_start + offset), len)); |
| |
| RangeSinkWriter writer(params.fd, tgt); |
| if (params.cmdname[0] == 'i') { // imgdiff |
| if (ApplyImagePatch(params.buffer.data(), blocks * BLOCKSIZE, patch_value, |
| std::bind(&RangeSinkWriter::Write, &writer, std::placeholders::_1, |
| std::placeholders::_2), |
| nullptr) != 0) { |
| LOG(ERROR) << "Failed to apply image patch."; |
| failure_type = kPatchApplicationFailure; |
| return -1; |
| } |
| } else { |
| if (ApplyBSDiffPatch(params.buffer.data(), blocks * BLOCKSIZE, patch_value, 0, |
| std::bind(&RangeSinkWriter::Write, &writer, std::placeholders::_1, |
| std::placeholders::_2)) != 0) { |
| LOG(ERROR) << "Failed to apply bsdiff patch."; |
| failure_type = kPatchApplicationFailure; |
| return -1; |
| } |
| } |
| |
| // We expect the output of the patcher to fill the tgt ranges exactly. |
| if (!writer.Finished()) { |
| LOG(ERROR) << "Failed to fully write target blocks (range sink underrun): Missing " |
| << writer.AvailableSpace() << " bytes"; |
| failure_type = kPatchApplicationFailure; |
| return -1; |
| } |
| } else { |
| LOG(INFO) << "skipping " << blocks << " blocks already patched to " << tgt.blocks() << " [" |
| << params.cmdline << "]"; |
| } |
| } |
| |
| if (!params.freestash.empty()) { |
| FreeStash(params.stashbase, params.freestash); |
| params.freestash.clear(); |
| } |
| |
| params.written += tgt.blocks(); |
| |
| return 0; |
| } |
| |
| static int PerformCommandErase(CommandParameters& params) { |
| if (DEBUG_ERASE) { |
| return PerformCommandZero(params); |
| } |
| |
| struct stat sb; |
| if (fstat(params.fd, &sb) == -1) { |
| PLOG(ERROR) << "failed to fstat device to erase"; |
| return -1; |
| } |
| |
| if (!S_ISBLK(sb.st_mode)) { |
| LOG(ERROR) << "not a block device; skipping erase"; |
| return -1; |
| } |
| |
| if (params.cpos >= params.tokens.size()) { |
| LOG(ERROR) << "missing target blocks for erase"; |
| return -1; |
| } |
| |
| RangeSet tgt = RangeSet::Parse(params.tokens[params.cpos++]); |
| CHECK(static_cast<bool>(tgt)); |
| |
| if (params.canwrite) { |
| LOG(INFO) << " erasing " << tgt.blocks() << " blocks"; |
| |
| for (const auto& [begin, end] : tgt) { |
| off64_t offset = static_cast<off64_t>(begin) * BLOCKSIZE; |
| size_t size = (end - begin) * BLOCKSIZE; |
| if (!discard_blocks(params.fd, offset, size, true /* force */)) { |
| return -1; |
| } |
| } |
| } |
| |
| return 0; |
| } |
| |
| static int PerformCommandAbort(CommandParameters&) { |
| LOG(INFO) << "Aborting as instructed"; |
| return -1; |
| } |
| |
| // Computes the hash_tree bytes based on the parameters, checks if the root hash of the tree |
| // matches the expected hash and writes the result to the specified range on the block_device. |
| // Hash_tree computation arguments: |
| // hash_tree_ranges |
| // source_ranges |
| // hash_algorithm |
| // salt_hex |
| // root_hash |
| static int PerformCommandComputeHashTree(CommandParameters& params) { |
| if (params.cpos + 5 != params.tokens.size()) { |
| LOG(ERROR) << "Invaild arguments count in hash computation " << params.cmdline; |
| return -1; |
| } |
| |
| // Expects the hash_tree data to be contiguous. |
| RangeSet hash_tree_ranges = RangeSet::Parse(params.tokens[params.cpos++]); |
| if (!hash_tree_ranges || hash_tree_ranges.size() != 1) { |
| LOG(ERROR) << "Invalid hash tree ranges in " << params.cmdline; |
| return -1; |
| } |
| |
| RangeSet source_ranges = RangeSet::Parse(params.tokens[params.cpos++]); |
| if (!source_ranges) { |
| LOG(ERROR) << "Invalid source ranges in " << params.cmdline; |
| return -1; |
| } |
| |
| auto hash_function = HashTreeBuilder::HashFunction(params.tokens[params.cpos++]); |
| if (hash_function == nullptr) { |
| LOG(ERROR) << "Invalid hash algorithm in " << params.cmdline; |
| return -1; |
| } |
| |
| std::vector<unsigned char> salt; |
| std::string salt_hex = params.tokens[params.cpos++]; |
| if (salt_hex.empty() || !HashTreeBuilder::ParseBytesArrayFromString(salt_hex, &salt)) { |
| LOG(ERROR) << "Failed to parse salt in " << params.cmdline; |
| return -1; |
| } |
| |
| std::string expected_root_hash = params.tokens[params.cpos++]; |
| if (expected_root_hash.empty()) { |
| LOG(ERROR) << "Invalid root hash in " << params.cmdline; |
| return -1; |
| } |
| |
| // Starts the hash_tree computation. |
| HashTreeBuilder builder(BLOCKSIZE, hash_function); |
| if (!builder.Initialize(static_cast<int64_t>(source_ranges.blocks()) * BLOCKSIZE, salt)) { |
| LOG(ERROR) << "Failed to initialize hash tree computation, source " << source_ranges.ToString() |
| << ", salt " << salt_hex; |
| return -1; |
| } |
| |
| // Iterates through every block in the source_ranges and updates the hash tree structure |
| // accordingly. |
| for (const auto& [begin, end] : source_ranges) { |
| uint8_t buffer[BLOCKSIZE]; |
| if (!check_lseek(params.fd, static_cast<off64_t>(begin) * BLOCKSIZE, SEEK_SET)) { |
| PLOG(ERROR) << "Failed to seek to block: " << begin; |
| return -1; |
| } |
| |
| <<<<<<< HEAD |
| #ifndef SUPPRESS_EMMC_WIPE |
| if (ioctl(params.fd, BLKDISCARD, &blocks) == -1) { |
| PLOG(ERROR) << "BLKDISCARD ioctl failed"; |
| ======= |
| for (size_t i = begin; i < end; i++) { |
| if (!android::base::ReadFully(params.fd, buffer, BLOCKSIZE)) { |
| failure_type = errno == EIO ? kEioFailure : kFreadFailure; |
| LOG(ERROR) << "Failed to read data in " << begin << ":" << end; |
| return -1; |
| } |
| |
| if (!builder.Update(reinterpret_cast<unsigned char*>(buffer), BLOCKSIZE)) { |
| LOG(ERROR) << "Failed to update hash tree builder"; |
| >>>>>>> android-10.0.0_r25 |
| return -1; |
| } |
| #endif |
| } |
| } |
| |
| if (!builder.BuildHashTree()) { |
| LOG(ERROR) << "Failed to build hash tree"; |
| return -1; |
| } |
| |
| std::string root_hash_hex = HashTreeBuilder::BytesArrayToString(builder.root_hash()); |
| if (root_hash_hex != expected_root_hash) { |
| LOG(ERROR) << "Root hash of the verity hash tree doesn't match the expected value. Expected: " |
| << expected_root_hash << ", actual: " << root_hash_hex; |
| return -1; |
| } |
| |
| uint64_t write_offset = static_cast<uint64_t>(hash_tree_ranges.GetBlockNumber(0)) * BLOCKSIZE; |
| if (params.canwrite && !builder.WriteHashTreeToFd(params.fd, write_offset)) { |
| LOG(ERROR) << "Failed to write hash tree to output"; |
| return -1; |
| } |
| |
| // TODO(xunchang) validates the written bytes |
| |
| return 0; |
| } |
| |
| using CommandFunction = std::function<int(CommandParameters&)>; |
| |
| using CommandMap = std::unordered_map<Command::Type, CommandFunction>; |
| |
| static bool Sha1DevicePath(const std::string& path, uint8_t digest[SHA_DIGEST_LENGTH]) { |
| auto device_name = android::base::Basename(path); |
| auto dm_target_name_path = "/sys/block/" + device_name + "/dm/name"; |
| |
| struct stat sb; |
| if (stat(dm_target_name_path.c_str(), &sb) == 0) { |
| // This is a device mapper target. Use partition name as part of the hash instead. Do not |
| // include extents as part of the hash, because the size of a partition may be shrunk after |
| // the patches are applied. |
| std::string dm_target_name; |
| if (!android::base::ReadFileToString(dm_target_name_path, &dm_target_name)) { |
| PLOG(ERROR) << "Cannot read " << dm_target_name_path; |
| return false; |
| } |
| SHA1(reinterpret_cast<const uint8_t*>(dm_target_name.data()), dm_target_name.size(), digest); |
| return true; |
| } |
| |
| if (errno != ENOENT) { |
| // This is a device mapper target, but its name cannot be retrieved. |
| PLOG(ERROR) << "Cannot get dm target name for " << path; |
| return false; |
| } |
| |
| // This doesn't appear to be a device mapper target, but if its name starts with dm-, something |
| // else might have gone wrong. |
| if (android::base::StartsWith(device_name, "dm-")) { |
| LOG(WARNING) << "Device " << path << " starts with dm- but is not mapped by device-mapper."; |
| } |
| |
| // Stash directory should be different for each partition to avoid conflicts when updating |
| // multiple partitions at the same time, so we use the hash of the block device name as the base |
| // directory. |
| SHA1(reinterpret_cast<const uint8_t*>(path.data()), path.size(), digest); |
| return true; |
| } |
| |
| static Value* PerformBlockImageUpdate(const char* name, State* state, |
| const std::vector<std::unique_ptr<Expr>>& argv, |
| const CommandMap& command_map, bool dryrun) { |
| CommandParameters params = {}; |
| stash_map.clear(); |
| params.canwrite = !dryrun; |
| |
| LOG(INFO) << "performing " << (dryrun ? "verification" : "update"); |
| if (state->is_retry) { |
| is_retry = true; |
| LOG(INFO) << "This update is a retry."; |
| } |
| if (argv.size() != 4) { |
| ErrorAbort(state, kArgsParsingFailure, "block_image_update expects 4 arguments, got %zu", |
| argv.size()); |
| return StringValue(""); |
| } |
| |
| std::vector<std::unique_ptr<Value>> args; |
| if (!ReadValueArgs(state, argv, &args)) { |
| return nullptr; |
| } |
| |
| // args: |
| // - block device (or file) to modify in-place |
| // - transfer list (blob) |
| // - new data stream (filename within package.zip) |
| // - patch stream (filename within package.zip, must be uncompressed) |
| const std::unique_ptr<Value>& blockdev_filename = args[0]; |
| const std::unique_ptr<Value>& transfer_list_value = args[1]; |
| const std::unique_ptr<Value>& new_data_fn = args[2]; |
| const std::unique_ptr<Value>& patch_data_fn = args[3]; |
| |
| if (blockdev_filename->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "blockdev_filename argument to %s must be string", name); |
| return StringValue(""); |
| } |
| if (transfer_list_value->type != Value::Type::BLOB) { |
| ErrorAbort(state, kArgsParsingFailure, "transfer_list argument to %s must be blob", name); |
| return StringValue(""); |
| } |
| if (new_data_fn->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "new_data_fn argument to %s must be string", name); |
| return StringValue(""); |
| } |
| if (patch_data_fn->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "patch_data_fn argument to %s must be string", name); |
| return StringValue(""); |
| } |
| |
| UpdaterInfo* ui = static_cast<UpdaterInfo*>(state->cookie); |
| if (ui == nullptr) { |
| return StringValue(""); |
| } |
| |
| FILE* cmd_pipe = ui->cmd_pipe; |
| ZipArchiveHandle za = ui->package_zip; |
| |
| if (cmd_pipe == nullptr || za == nullptr) { |
| return StringValue(""); |
| } |
| |
| ZipString path_data(patch_data_fn->data.c_str()); |
| ZipEntry patch_entry; |
| if (FindEntry(za, path_data, &patch_entry) != 0) { |
| LOG(ERROR) << name << "(): no file \"" << patch_data_fn->data << "\" in package"; |
| return StringValue(""); |
| } |
| |
| params.patch_start = ui->package_zip_addr + patch_entry.offset; |
| ZipString new_data(new_data_fn->data.c_str()); |
| ZipEntry new_entry; |
| if (FindEntry(za, new_data, &new_entry) != 0) { |
| LOG(ERROR) << name << "(): no file \"" << new_data_fn->data << "\" in package"; |
| return StringValue(""); |
| } |
| |
| params.fd.reset(TEMP_FAILURE_RETRY(open(blockdev_filename->data.c_str(), O_RDWR))); |
| if (params.fd == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFileOpenFailure; |
| PLOG(ERROR) << "open \"" << blockdev_filename->data << "\" failed"; |
| return StringValue(""); |
| } |
| |
| uint8_t digest[SHA_DIGEST_LENGTH]; |
| if (!Sha1DevicePath(blockdev_filename->data, digest)) { |
| return StringValue(""); |
| } |
| params.stashbase = print_sha1(digest); |
| |
| // Possibly do return early on retry, by checking the marker. If the update on this partition has |
| // been finished (but interrupted at a later point), there could be leftover on /cache that would |
| // fail the no-op retry. |
| std::string updated_marker = GetStashFileName(params.stashbase + ".UPDATED", "", ""); |
| if (is_retry) { |
| struct stat sb; |
| int result = stat(updated_marker.c_str(), &sb); |
| if (result == 0) { |
| LOG(INFO) << "Skipping already updated partition " << blockdev_filename->data |
| << " based on marker"; |
| return StringValue("t"); |
| } |
| } else { |
| // Delete the obsolete marker if any. |
| std::string err; |
| if (!android::base::RemoveFileIfExists(updated_marker, &err)) { |
| LOG(ERROR) << "Failed to remove partition updated marker " << updated_marker << ": " << err; |
| return StringValue(""); |
| } |
| } |
| |
| static constexpr size_t kTransferListHeaderLines = 4; |
| std::vector<std::string> lines = android::base::Split(transfer_list_value->data, "\n"); |
| if (lines.size() < kTransferListHeaderLines) { |
| ErrorAbort(state, kArgsParsingFailure, "too few lines in the transfer list [%zu]", |
| lines.size()); |
| return StringValue(""); |
| } |
| |
| // First line in transfer list is the version number. |
| if (!android::base::ParseInt(lines[0], ¶ms.version, 3, 4)) { |
| LOG(ERROR) << "unexpected transfer list version [" << lines[0] << "]"; |
| return StringValue(""); |
| } |
| |
| LOG(INFO) << "blockimg version is " << params.version; |
| |
| // Second line in transfer list is the total number of blocks we expect to write. |
| size_t total_blocks; |
| if (!android::base::ParseUint(lines[1], &total_blocks)) { |
| ErrorAbort(state, kArgsParsingFailure, "unexpected block count [%s]", lines[1].c_str()); |
| return StringValue(""); |
| } |
| |
| if (total_blocks == 0) { |
| return StringValue("t"); |
| } |
| |
| // Third line is how many stash entries are needed simultaneously. |
| LOG(INFO) << "maximum stash entries " << lines[2]; |
| |
| // Fourth line is the maximum number of blocks that will be stashed simultaneously |
| size_t stash_max_blocks; |
| if (!android::base::ParseUint(lines[3], &stash_max_blocks)) { |
| ErrorAbort(state, kArgsParsingFailure, "unexpected maximum stash blocks [%s]", |
| lines[3].c_str()); |
| return StringValue(""); |
| } |
| |
| int res = CreateStash(state, stash_max_blocks, params.stashbase); |
| if (res == -1) { |
| return StringValue(""); |
| } |
| params.createdstash = res; |
| |
| // Set up the new data writer. |
| if (params.canwrite) { |
| params.nti.za = za; |
| params.nti.entry = new_entry; |
| params.nti.brotli_compressed = android::base::EndsWith(new_data_fn->data, ".br"); |
| if (params.nti.brotli_compressed) { |
| // Initialize brotli decoder state. |
| params.nti.brotli_decoder_state = BrotliDecoderCreateInstance(nullptr, nullptr, nullptr); |
| } |
| params.nti.receiver_available = true; |
| |
| pthread_mutex_init(¶ms.nti.mu, nullptr); |
| pthread_cond_init(¶ms.nti.cv, nullptr); |
| pthread_attr_t attr; |
| pthread_attr_init(&attr); |
| pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE); |
| |
| int error = pthread_create(¶ms.thread, &attr, unzip_new_data, ¶ms.nti); |
| if (error != 0) { |
| LOG(ERROR) << "pthread_create failed: " << strerror(error); |
| return StringValue(""); |
| } |
| } |
| |
| // When performing an update, save the index and cmdline of the current command into the |
| // last_command_file. |
| // Upon resuming an update, read the saved index first; then |
| // 1. In verification mode, check if the 'move' or 'diff' commands before the saved index has |
| // the expected target blocks already. If not, these commands cannot be skipped and we need |
| // to attempt to execute them again. Therefore, we will delete the last_command_file so that |
| // the update will resume from the start of the transfer list. |
| // 2. In update mode, skip all commands before the saved index. Therefore, we can avoid deleting |
| // stashes with duplicate id unintentionally (b/69858743); and also speed up the update. |
| // If an update succeeds or is unresumable, delete the last_command_file. |
| bool skip_executed_command = true; |
| size_t saved_last_command_index; |
| if (!ParseLastCommandFile(&saved_last_command_index)) { |
| DeleteLastCommandFile(); |
| // We failed to parse the last command. Disallow skipping executed commands. |
| skip_executed_command = false; |
| } |
| |
| int rc = -1; |
| |
| // Subsequent lines are all individual transfer commands |
| for (size_t i = kTransferListHeaderLines; i < lines.size(); i++) { |
| const std::string& line = lines[i]; |
| if (line.empty()) continue; |
| |
| size_t cmdindex = i - kTransferListHeaderLines; |
| params.tokens = android::base::Split(line, " "); |
| params.cpos = 0; |
| params.cmdname = params.tokens[params.cpos++]; |
| params.cmdline = line; |
| params.target_verified = false; |
| |
| Command::Type cmd_type = Command::ParseType(params.cmdname); |
| if (cmd_type == Command::Type::LAST) { |
| LOG(ERROR) << "unexpected command [" << params.cmdname << "]"; |
| goto pbiudone; |
| } |
| |
| const CommandFunction& performer = command_map.at(cmd_type); |
| |
| // Skip the command if we explicitly set the corresponding function pointer to nullptr, e.g. |
| // "erase" during block_image_verify. |
| if (performer == nullptr) { |
| LOG(DEBUG) << "skip executing command [" << line << "]"; |
| continue; |
| } |
| |
| // Skip all commands before the saved last command index when resuming an update, except for |
| // "new" command. Because new commands read in the data sequentially. |
| if (params.canwrite && skip_executed_command && cmdindex <= saved_last_command_index && |
| cmd_type != Command::Type::NEW) { |
| LOG(INFO) << "Skipping already executed command: " << cmdindex |
| << ", last executed command for previous update: " << saved_last_command_index; |
| continue; |
| } |
| |
| if (performer(params) == -1) { |
| LOG(ERROR) << "failed to execute command [" << line << "]"; |
| if (cmd_type == Command::Type::COMPUTE_HASH_TREE && failure_type == kNoCause) { |
| failure_type = kHashTreeComputationFailure; |
| } |
| goto pbiudone; |
| } |
| |
| // In verify mode, check if the commands before the saved last_command_index have been executed |
| // correctly. If some target blocks have unexpected contents, delete the last command file so |
| // that we will resume the update from the first command in the transfer list. |
| if (!params.canwrite && skip_executed_command && cmdindex <= saved_last_command_index) { |
| // TODO(xunchang) check that the cmdline of the saved index is correct. |
| if ((cmd_type == Command::Type::MOVE || cmd_type == Command::Type::BSDIFF || |
| cmd_type == Command::Type::IMGDIFF) && |
| !params.target_verified) { |
| LOG(WARNING) << "Previously executed command " << saved_last_command_index << ": " |
| << params.cmdline << " doesn't produce expected target blocks."; |
| skip_executed_command = false; |
| DeleteLastCommandFile(); |
| } |
| } |
| |
| if (params.canwrite) { |
| if (fsync(params.fd) == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFsyncFailure; |
| PLOG(ERROR) << "fsync failed"; |
| goto pbiudone; |
| } |
| |
| if (!UpdateLastCommandIndex(cmdindex, params.cmdline)) { |
| LOG(WARNING) << "Failed to update the last command file."; |
| } |
| |
| fprintf(cmd_pipe, "set_progress %.4f\n", static_cast<double>(params.written) / total_blocks); |
| fflush(cmd_pipe); |
| } |
| } |
| |
| rc = 0; |
| |
| pbiudone: |
| if (params.canwrite) { |
| pthread_mutex_lock(¶ms.nti.mu); |
| if (params.nti.receiver_available) { |
| LOG(WARNING) << "new data receiver is still available after executing all commands."; |
| } |
| params.nti.receiver_available = false; |
| pthread_cond_broadcast(¶ms.nti.cv); |
| pthread_mutex_unlock(¶ms.nti.mu); |
| int ret = pthread_join(params.thread, nullptr); |
| if (ret != 0) { |
| LOG(WARNING) << "pthread join returned with " << strerror(ret); |
| } |
| |
| if (rc == 0) { |
| LOG(INFO) << "wrote " << params.written << " blocks; expected " << total_blocks; |
| LOG(INFO) << "stashed " << params.stashed << " blocks"; |
| LOG(INFO) << "max alloc needed was " << params.buffer.size(); |
| |
| const char* partition = strrchr(blockdev_filename->data.c_str(), '/'); |
| if (partition != nullptr && *(partition + 1) != 0) { |
| fprintf(cmd_pipe, "log bytes_written_%s: %" PRIu64 "\n", partition + 1, |
| static_cast<uint64_t>(params.written) * BLOCKSIZE); |
| fprintf(cmd_pipe, "log bytes_stashed_%s: %" PRIu64 "\n", partition + 1, |
| static_cast<uint64_t>(params.stashed) * BLOCKSIZE); |
| fflush(cmd_pipe); |
| } |
| // Delete stash only after successfully completing the update, as it may contain blocks needed |
| // to complete the update later. |
| DeleteStash(params.stashbase); |
| DeleteLastCommandFile(); |
| |
| // Create a marker on /cache partition, which allows skipping the update on this partition on |
| // retry. The marker will be removed once booting into normal boot, or before starting next |
| // fresh install. |
| if (!SetUpdatedMarker(updated_marker)) { |
| LOG(WARNING) << "Failed to set updated marker; continuing"; |
| } |
| } |
| |
| pthread_mutex_destroy(¶ms.nti.mu); |
| pthread_cond_destroy(¶ms.nti.cv); |
| } else if (rc == 0) { |
| LOG(INFO) << "verified partition contents; update may be resumed"; |
| } |
| |
| if (fsync(params.fd) == -1) { |
| failure_type = errno == EIO ? kEioFailure : kFsyncFailure; |
| PLOG(ERROR) << "fsync failed"; |
| } |
| // params.fd will be automatically closed because it's a unique_fd. |
| |
| if (params.nti.brotli_decoder_state != nullptr) { |
| BrotliDecoderDestroyInstance(params.nti.brotli_decoder_state); |
| } |
| |
| // Delete the last command file if the update cannot be resumed. |
| if (params.isunresumable) { |
| DeleteLastCommandFile(); |
| } |
| |
| // Only delete the stash if the update cannot be resumed, or it's a verification run and we |
| // created the stash. |
| if (params.isunresumable || (!params.canwrite && params.createdstash)) { |
| DeleteStash(params.stashbase); |
| } |
| |
| if (failure_type != kNoCause && state->cause_code == kNoCause) { |
| state->cause_code = failure_type; |
| } |
| |
| return StringValue(rc == 0 ? "t" : ""); |
| } |
| |
| /** |
| * The transfer list is a text file containing commands to transfer data from one place to another |
| * on the target partition. We parse it and execute the commands in order: |
| * |
| * zero [rangeset] |
| * - Fill the indicated blocks with zeros. |
| * |
| * new [rangeset] |
| * - Fill the blocks with data read from the new_data file. |
| * |
| * erase [rangeset] |
| * - Mark the given blocks as empty. |
| * |
| * move <...> |
| * bsdiff <patchstart> <patchlen> <...> |
| * imgdiff <patchstart> <patchlen> <...> |
| * - Read the source blocks, apply a patch (or not in the case of move), write result to target |
| * blocks. bsdiff or imgdiff specifies the type of patch; move means no patch at all. |
| * |
| * See the comments in LoadSrcTgtVersion3() for a description of the <...> format. |
| * |
| * stash <stash_id> <src_range> |
| * - Load the given source range and stash the data in the given slot of the stash table. |
| * |
| * free <stash_id> |
| * - Free the given stash data. |
| * |
| * The creator of the transfer list will guarantee that no block is read (ie, used as the source for |
| * a patch or move) after it has been written. |
| * |
| * The creator will guarantee that a given stash is loaded (with a stash command) before it's used |
| * in a move/bsdiff/imgdiff command. |
| * |
| * Within one command the source and target ranges may overlap so in general we need to read the |
| * entire source into memory before writing anything to the target blocks. |
| * |
| * All the patch data is concatenated into one patch_data file in the update package. It must be |
| * stored uncompressed because we memory-map it in directly from the archive. (Since patches are |
| * already compressed, we lose very little by not compressing their concatenation.) |
| * |
| * Commands that read data from the partition (i.e. move/bsdiff/imgdiff/stash) have one or more |
| * additional hashes before the range parameters, which are used to check if the command has already |
| * been completed and verify the integrity of the source data. |
| */ |
| Value* BlockImageVerifyFn(const char* name, State* state, |
| const std::vector<std::unique_ptr<Expr>>& argv) { |
| // Commands which are not allowed are set to nullptr to skip them completely. |
| const CommandMap command_map{ |
| // clang-format off |
| { Command::Type::ABORT, PerformCommandAbort }, |
| { Command::Type::BSDIFF, PerformCommandDiff }, |
| { Command::Type::COMPUTE_HASH_TREE, PerformCommandComputeHashTree }, |
| { Command::Type::ERASE, nullptr }, |
| { Command::Type::FREE, PerformCommandFree }, |
| { Command::Type::IMGDIFF, PerformCommandDiff }, |
| { Command::Type::MOVE, PerformCommandMove }, |
| { Command::Type::NEW, nullptr }, |
| { Command::Type::STASH, PerformCommandStash }, |
| { Command::Type::ZERO, nullptr }, |
| // clang-format on |
| }; |
| CHECK_EQ(static_cast<size_t>(Command::Type::LAST), command_map.size()); |
| |
| // Perform a dry run without writing to test if an update can proceed. |
| return PerformBlockImageUpdate(name, state, argv, command_map, true); |
| } |
| |
| Value* BlockImageUpdateFn(const char* name, State* state, |
| const std::vector<std::unique_ptr<Expr>>& argv) { |
| const CommandMap command_map{ |
| // clang-format off |
| { Command::Type::ABORT, PerformCommandAbort }, |
| { Command::Type::BSDIFF, PerformCommandDiff }, |
| { Command::Type::COMPUTE_HASH_TREE, PerformCommandComputeHashTree }, |
| { Command::Type::ERASE, PerformCommandErase }, |
| { Command::Type::FREE, PerformCommandFree }, |
| { Command::Type::IMGDIFF, PerformCommandDiff }, |
| { Command::Type::MOVE, PerformCommandMove }, |
| { Command::Type::NEW, PerformCommandNew }, |
| { Command::Type::STASH, PerformCommandStash }, |
| { Command::Type::ZERO, PerformCommandZero }, |
| // clang-format on |
| }; |
| CHECK_EQ(static_cast<size_t>(Command::Type::LAST), command_map.size()); |
| |
| return PerformBlockImageUpdate(name, state, argv, command_map, false); |
| } |
| |
| Value* RangeSha1Fn(const char* name, State* state, const std::vector<std::unique_ptr<Expr>>& argv) { |
| if (argv.size() != 2) { |
| ErrorAbort(state, kArgsParsingFailure, "range_sha1 expects 2 arguments, got %zu", argv.size()); |
| return StringValue(""); |
| } |
| |
| std::vector<std::unique_ptr<Value>> args; |
| if (!ReadValueArgs(state, argv, &args)) { |
| return nullptr; |
| } |
| |
| const std::unique_ptr<Value>& blockdev_filename = args[0]; |
| const std::unique_ptr<Value>& ranges = args[1]; |
| |
| if (blockdev_filename->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "blockdev_filename argument to %s must be string", name); |
| return StringValue(""); |
| } |
| if (ranges->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "ranges argument to %s must be string", name); |
| return StringValue(""); |
| } |
| |
| android::base::unique_fd fd(open(blockdev_filename->data.c_str(), O_RDWR)); |
| if (fd == -1) { |
| CauseCode cause_code = errno == EIO ? kEioFailure : kFileOpenFailure; |
| ErrorAbort(state, cause_code, "open \"%s\" failed: %s", blockdev_filename->data.c_str(), |
| strerror(errno)); |
| return StringValue(""); |
| } |
| |
| RangeSet rs = RangeSet::Parse(ranges->data); |
| CHECK(static_cast<bool>(rs)); |
| |
| SHA_CTX ctx; |
| SHA1_Init(&ctx); |
| |
| std::vector<uint8_t> buffer(BLOCKSIZE); |
| for (const auto& [begin, end] : rs) { |
| if (!check_lseek(fd, static_cast<off64_t>(begin) * BLOCKSIZE, SEEK_SET)) { |
| ErrorAbort(state, kLseekFailure, "failed to seek %s: %s", blockdev_filename->data.c_str(), |
| strerror(errno)); |
| return StringValue(""); |
| } |
| |
| for (size_t j = begin; j < end; ++j) { |
| if (!android::base::ReadFully(fd, buffer.data(), BLOCKSIZE)) { |
| CauseCode cause_code = errno == EIO ? kEioFailure : kFreadFailure; |
| ErrorAbort(state, cause_code, "failed to read %s: %s", blockdev_filename->data.c_str(), |
| strerror(errno)); |
| return StringValue(""); |
| } |
| |
| SHA1_Update(&ctx, buffer.data(), BLOCKSIZE); |
| } |
| } |
| uint8_t digest[SHA_DIGEST_LENGTH]; |
| SHA1_Final(digest, &ctx); |
| |
| return StringValue(print_sha1(digest)); |
| } |
| |
| // This function checks if a device has been remounted R/W prior to an incremental |
| // OTA update. This is an common cause of update abortion. The function reads the |
| // 1st block of each partition and check for mounting time/count. It return string "t" |
| // if executes successfully and an empty string otherwise. |
| |
| Value* CheckFirstBlockFn(const char* name, State* state, |
| const std::vector<std::unique_ptr<Expr>>& argv) { |
| if (argv.size() != 1) { |
| ErrorAbort(state, kArgsParsingFailure, "check_first_block expects 1 argument, got %zu", |
| argv.size()); |
| return StringValue(""); |
| } |
| |
| std::vector<std::unique_ptr<Value>> args; |
| if (!ReadValueArgs(state, argv, &args)) { |
| return nullptr; |
| } |
| |
| const std::unique_ptr<Value>& arg_filename = args[0]; |
| |
| if (arg_filename->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "filename argument to %s must be string", name); |
| return StringValue(""); |
| } |
| |
| android::base::unique_fd fd(open(arg_filename->data.c_str(), O_RDONLY)); |
| if (fd == -1) { |
| CauseCode cause_code = errno == EIO ? kEioFailure : kFileOpenFailure; |
| ErrorAbort(state, cause_code, "open \"%s\" failed: %s", arg_filename->data.c_str(), |
| strerror(errno)); |
| return StringValue(""); |
| } |
| |
| RangeSet blk0(std::vector<Range>{ Range{ 0, 1 } }); |
| std::vector<uint8_t> block0_buffer(BLOCKSIZE); |
| |
| if (ReadBlocks(blk0, &block0_buffer, fd) == -1) { |
| CauseCode cause_code = errno == EIO ? kEioFailure : kFreadFailure; |
| ErrorAbort(state, cause_code, "failed to read %s: %s", arg_filename->data.c_str(), |
| strerror(errno)); |
| return StringValue(""); |
| } |
| |
| // https://ext4.wiki.kernel.org/index.php/Ext4_Disk_Layout |
| // Super block starts from block 0, offset 0x400 |
| // 0x2C: len32 Mount time |
| // 0x30: len32 Write time |
| // 0x34: len16 Number of mounts since the last fsck |
| // 0x38: len16 Magic signature 0xEF53 |
| |
| time_t mount_time = *reinterpret_cast<uint32_t*>(&block0_buffer[0x400 + 0x2C]); |
| uint16_t mount_count = *reinterpret_cast<uint16_t*>(&block0_buffer[0x400 + 0x34]); |
| |
| if (mount_count > 0) { |
| uiPrintf(state, "Device was remounted R/W %" PRIu16 " times", mount_count); |
| uiPrintf(state, "Last remount happened on %s", ctime(&mount_time)); |
| } |
| |
| return StringValue("t"); |
| } |
| |
| Value* BlockImageRecoverFn(const char* name, State* state, |
| const std::vector<std::unique_ptr<Expr>>& argv) { |
| if (argv.size() != 2) { |
| ErrorAbort(state, kArgsParsingFailure, "block_image_recover expects 2 arguments, got %zu", |
| argv.size()); |
| return StringValue(""); |
| } |
| |
| std::vector<std::unique_ptr<Value>> args; |
| if (!ReadValueArgs(state, argv, &args)) { |
| return nullptr; |
| } |
| |
| const std::unique_ptr<Value>& filename = args[0]; |
| const std::unique_ptr<Value>& ranges = args[1]; |
| |
| if (filename->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "filename argument to %s must be string", name); |
| return StringValue(""); |
| } |
| if (ranges->type != Value::Type::STRING) { |
| ErrorAbort(state, kArgsParsingFailure, "ranges argument to %s must be string", name); |
| return StringValue(""); |
| } |
| RangeSet rs = RangeSet::Parse(ranges->data); |
| if (!rs) { |
| ErrorAbort(state, kArgsParsingFailure, "failed to parse ranges: %s", ranges->data.c_str()); |
| return StringValue(""); |
| } |
| |
| // Output notice to log when recover is attempted |
| LOG(INFO) << filename->data << " image corrupted, attempting to recover..."; |
| |
| // When opened with O_RDWR, libfec rewrites corrupted blocks when they are read |
| fec::io fh(filename->data, O_RDWR); |
| |
| if (!fh) { |
| ErrorAbort(state, kLibfecFailure, "fec_open \"%s\" failed: %s", filename->data.c_str(), |
| strerror(errno)); |
| return StringValue(""); |
| } |
| |
| if (!fh.has_ecc() || !fh.has_verity()) { |
| ErrorAbort(state, kLibfecFailure, "unable to use metadata to correct errors"); |
| return StringValue(""); |
| } |
| |
| fec_status status; |
| if (!fh.get_status(status)) { |
| ErrorAbort(state, kLibfecFailure, "failed to read FEC status"); |
| return StringValue(""); |
| } |
| |
| uint8_t buffer[BLOCKSIZE]; |
| for (const auto& [begin, end] : rs) { |
| for (size_t j = begin; j < end; ++j) { |
| // Stay within the data area, libfec validates and corrects metadata |
| if (status.data_size <= static_cast<uint64_t>(j) * BLOCKSIZE) { |
| continue; |
| } |
| |
| if (fh.pread(buffer, BLOCKSIZE, static_cast<off64_t>(j) * BLOCKSIZE) != BLOCKSIZE) { |
| ErrorAbort(state, kLibfecFailure, "failed to recover %s (block %zu): %s", |
| filename->data.c_str(), j, strerror(errno)); |
| return StringValue(""); |
| } |
| |
| // If we want to be able to recover from a situation where rewriting a corrected |
| // block doesn't guarantee the same data will be returned when re-read later, we |
| // can save a copy of corrected blocks to /cache. Note: |
| // |
| // 1. Maximum space required from /cache is the same as the maximum number of |
| // corrupted blocks we can correct. For RS(255, 253) and a 2 GiB partition, |
| // this would be ~16 MiB, for example. |
| // |
| // 2. To find out if this block was corrupted, call fec_get_status after each |
| // read and check if the errors field value has increased. |
| } |
| } |
| LOG(INFO) << "..." << filename->data << " image recovered successfully."; |
| return StringValue("t"); |
| } |
| |
| void RegisterBlockImageFunctions() { |
| RegisterFunction("block_image_verify", BlockImageVerifyFn); |
| RegisterFunction("block_image_update", BlockImageUpdateFn); |
| RegisterFunction("block_image_recover", BlockImageRecoverFn); |
| RegisterFunction("check_first_block", CheckFirstBlockFn); |
| RegisterFunction("range_sha1", RangeSha1Fn); |
| } |