| /* |
| * Copyright (C) 2010 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 "ext4_utils.h" |
| #include "output_file.h" |
| #include "sparse_format.h" |
| #include "sparse_crc32.h" |
| #include "wipe.h" |
| |
| #include <fcntl.h> |
| #include <stdbool.h> |
| #include <sys/types.h> |
| #include <sys/stat.h> |
| #include <sys/types.h> |
| #include <unistd.h> |
| #include <zlib.h> |
| |
| #ifndef USE_MINGW |
| #include <sys/mman.h> |
| #define O_BINARY 0 |
| #endif |
| |
| |
| #if defined(__APPLE__) && defined(__MACH__) |
| #define lseek64 lseek |
| #define off64_t off_t |
| #endif |
| |
| #define SPARSE_HEADER_MAJOR_VER 1 |
| #define SPARSE_HEADER_MINOR_VER 0 |
| #define SPARSE_HEADER_LEN (sizeof(sparse_header_t)) |
| #define CHUNK_HEADER_LEN (sizeof(chunk_header_t)) |
| |
| struct output_file_ops { |
| int (*seek)(struct output_file *, off64_t); |
| int (*write)(struct output_file *, u8 *, int); |
| void (*close)(struct output_file *); |
| }; |
| |
| struct output_file { |
| int fd; |
| gzFile gz_fd; |
| bool close_fd; |
| int sparse; |
| u64 cur_out_ptr; |
| u32 chunk_cnt; |
| u32 crc32; |
| struct output_file_ops *ops; |
| int use_crc; |
| }; |
| |
| static int file_seek(struct output_file *out, off64_t off) |
| { |
| off64_t ret; |
| |
| ret = lseek64(out->fd, off, SEEK_SET); |
| if (ret < 0) { |
| error_errno("lseek64"); |
| return -1; |
| } |
| return 0; |
| } |
| |
| static int file_write(struct output_file *out, u8 *data, int len) |
| { |
| int ret; |
| ret = write(out->fd, data, len); |
| if (ret < 0) { |
| error_errno("write"); |
| return -1; |
| } else if (ret < len) { |
| error("incomplete write"); |
| return -1; |
| } |
| |
| return 0; |
| } |
| |
| static void file_close(struct output_file *out) |
| { |
| if (out->close_fd) { |
| close(out->fd); |
| } |
| } |
| |
| |
| static struct output_file_ops file_ops = { |
| .seek = file_seek, |
| .write = file_write, |
| .close = file_close, |
| }; |
| |
| static int gz_file_seek(struct output_file *out, off64_t off) |
| { |
| off64_t ret; |
| |
| ret = gzseek(out->gz_fd, off, SEEK_SET); |
| if (ret < 0) { |
| error_errno("gzseek"); |
| return -1; |
| } |
| return 0; |
| } |
| |
| static int gz_file_write(struct output_file *out, u8 *data, int len) |
| { |
| int ret; |
| ret = gzwrite(out->gz_fd, data, len); |
| if (ret < 0) { |
| error_errno("gzwrite"); |
| return -1; |
| } else if (ret < len) { |
| error("incomplete gzwrite"); |
| return -1; |
| } |
| |
| return 0; |
| } |
| |
| static void gz_file_close(struct output_file *out) |
| { |
| gzclose(out->gz_fd); |
| } |
| |
| static struct output_file_ops gz_file_ops = { |
| .seek = gz_file_seek, |
| .write = gz_file_write, |
| .close = gz_file_close, |
| }; |
| |
| static sparse_header_t sparse_header = { |
| .magic = SPARSE_HEADER_MAGIC, |
| .major_version = SPARSE_HEADER_MAJOR_VER, |
| .minor_version = SPARSE_HEADER_MINOR_VER, |
| .file_hdr_sz = SPARSE_HEADER_LEN, |
| .chunk_hdr_sz = CHUNK_HEADER_LEN, |
| .blk_sz = 0, |
| .total_blks = 0, |
| .total_chunks = 0, |
| .image_checksum = 0 |
| }; |
| |
| static u8 *zero_buf; |
| |
| static int emit_skip_chunk(struct output_file *out, u64 skip_len) |
| { |
| chunk_header_t chunk_header; |
| int ret, chunk; |
| |
| //DBG printf("skip chunk: 0x%llx bytes\n", skip_len); |
| |
| if (skip_len % info.block_size) { |
| error("don't care size %llu is not a multiple of the block size %u", |
| skip_len, info.block_size); |
| return -1; |
| } |
| |
| /* We are skipping data, so emit a don't care chunk. */ |
| chunk_header.chunk_type = CHUNK_TYPE_DONT_CARE; |
| chunk_header.reserved1 = 0; |
| chunk_header.chunk_sz = skip_len / info.block_size; |
| chunk_header.total_sz = CHUNK_HEADER_LEN; |
| ret = out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header)); |
| if (ret < 0) |
| return -1; |
| |
| out->cur_out_ptr += skip_len; |
| out->chunk_cnt++; |
| |
| return 0; |
| } |
| |
| static int write_chunk_fill(struct output_file *out, u64 off, u32 fill_val, int len) |
| { |
| chunk_header_t chunk_header; |
| int rnd_up_len, zero_len, count; |
| int ret; |
| unsigned int i; |
| u32 fill_buf[4096/sizeof(u32)]; /* Maximum size of a block */ |
| |
| /* We can assume that all the chunks to be written are in |
| * ascending order, block-size aligned, and non-overlapping. |
| * So, if the offset is less than the current output pointer, |
| * throw an error, and if there is a gap, emit a "don't care" |
| * chunk. The first write (of the super block) may not be |
| * blocksize aligned, so we need to deal with that too. |
| */ |
| //DBG printf("write chunk: offset 0x%llx, length 0x%x bytes\n", off, len); |
| |
| if (off < out->cur_out_ptr) { |
| error("offset %llu is less than the current output offset %llu", |
| off, out->cur_out_ptr); |
| return -1; |
| } |
| |
| if (off > out->cur_out_ptr) { |
| emit_skip_chunk(out, off - out->cur_out_ptr); |
| } |
| |
| if (off % info.block_size) { |
| error("write chunk offset %llu is not a multiple of the block size %u", |
| off, info.block_size); |
| return -1; |
| } |
| |
| if (off != out->cur_out_ptr) { |
| error("internal error, offset accounting screwy in write_chunk_raw()"); |
| return -1; |
| } |
| |
| /* Round up the file length to a multiple of the block size */ |
| rnd_up_len = (len + (info.block_size - 1)) & (~(info.block_size -1)); |
| |
| /* Finally we can safely emit a chunk of data */ |
| chunk_header.chunk_type = CHUNK_TYPE_FILL; |
| chunk_header.reserved1 = 0; |
| chunk_header.chunk_sz = rnd_up_len / info.block_size; |
| chunk_header.total_sz = CHUNK_HEADER_LEN + sizeof(fill_val); |
| ret = out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header)); |
| |
| if (ret < 0) |
| return -1; |
| ret = out->ops->write(out, (u8 *)&fill_val, sizeof(fill_val)); |
| if (ret < 0) |
| return -1; |
| |
| if (out->use_crc) { |
| /* Initialize fill_buf with the fill_val */ |
| for (i = 0; i < (info.block_size / sizeof(u32)); i++) { |
| fill_buf[i] = fill_val; |
| } |
| |
| count = chunk_header.chunk_sz; |
| while (count) { |
| out->crc32 = sparse_crc32(out->crc32, fill_buf, info.block_size); |
| count--; |
| } |
| } |
| |
| out->cur_out_ptr += rnd_up_len; |
| out->chunk_cnt++; |
| |
| return 0; |
| } |
| |
| static int write_chunk_raw(struct output_file *out, u64 off, u8 *data, int len) |
| { |
| chunk_header_t chunk_header; |
| int rnd_up_len, zero_len; |
| int ret; |
| |
| /* We can assume that all the chunks to be written are in |
| * ascending order, block-size aligned, and non-overlapping. |
| * So, if the offset is less than the current output pointer, |
| * throw an error, and if there is a gap, emit a "don't care" |
| * chunk. The first write (of the super block) may not be |
| * blocksize aligned, so we need to deal with that too. |
| */ |
| //DBG printf("write chunk: offset 0x%llx, length 0x%x bytes\n", off, len); |
| |
| if (off < out->cur_out_ptr) { |
| error("offset %llu is less than the current output offset %llu", |
| off, out->cur_out_ptr); |
| return -1; |
| } |
| |
| if (off > out->cur_out_ptr) { |
| emit_skip_chunk(out, off - out->cur_out_ptr); |
| } |
| |
| if (off % info.block_size) { |
| error("write chunk offset %llu is not a multiple of the block size %u", |
| off, info.block_size); |
| return -1; |
| } |
| |
| if (off != out->cur_out_ptr) { |
| error("internal error, offset accounting screwy in write_chunk_raw()"); |
| return -1; |
| } |
| |
| /* Round up the file length to a multiple of the block size */ |
| rnd_up_len = (len + (info.block_size - 1)) & (~(info.block_size -1)); |
| zero_len = rnd_up_len - len; |
| |
| /* Finally we can safely emit a chunk of data */ |
| chunk_header.chunk_type = CHUNK_TYPE_RAW; |
| chunk_header.reserved1 = 0; |
| chunk_header.chunk_sz = rnd_up_len / info.block_size; |
| chunk_header.total_sz = CHUNK_HEADER_LEN + rnd_up_len; |
| ret = out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header)); |
| |
| if (ret < 0) |
| return -1; |
| ret = out->ops->write(out, data, len); |
| if (ret < 0) |
| return -1; |
| if (zero_len) { |
| ret = out->ops->write(out, zero_buf, zero_len); |
| if (ret < 0) |
| return -1; |
| } |
| |
| if (out->use_crc) { |
| out->crc32 = sparse_crc32(out->crc32, data, len); |
| if (zero_len) |
| out->crc32 = sparse_crc32(out->crc32, zero_buf, zero_len); |
| } |
| |
| out->cur_out_ptr += rnd_up_len; |
| out->chunk_cnt++; |
| |
| return 0; |
| } |
| |
| void close_output_file(struct output_file *out) |
| { |
| int ret; |
| chunk_header_t chunk_header; |
| |
| if (out->sparse) { |
| if (out->use_crc) { |
| chunk_header.chunk_type = CHUNK_TYPE_CRC32; |
| chunk_header.reserved1 = 0; |
| chunk_header.chunk_sz = 0; |
| chunk_header.total_sz = CHUNK_HEADER_LEN + 4; |
| |
| out->ops->write(out, (u8 *)&chunk_header, sizeof(chunk_header)); |
| out->ops->write(out, (u8 *)&out->crc32, 4); |
| |
| out->chunk_cnt++; |
| } |
| |
| if (out->chunk_cnt != sparse_header.total_chunks) |
| error("sparse chunk count did not match: %d %d", out->chunk_cnt, |
| sparse_header.total_chunks); |
| } |
| out->ops->close(out); |
| } |
| |
| struct output_file *open_output_fd(int fd, int gz, int sparse, |
| int chunks, int crc, int wipe) |
| { |
| int ret; |
| struct output_file *out = malloc(sizeof(struct output_file)); |
| if (!out) { |
| error_errno("malloc struct out"); |
| return NULL; |
| } |
| zero_buf = malloc(info.block_size); |
| if (!zero_buf) { |
| error_errno("malloc zero_buf"); |
| free(out); |
| return NULL; |
| } |
| memset(zero_buf, '\0', info.block_size); |
| |
| if (gz) { |
| out->ops = &gz_file_ops; |
| out->gz_fd = gzdopen(fd, "wb9"); |
| if (!out->gz_fd) { |
| error_errno("gzopen"); |
| free(out); |
| return NULL; |
| } |
| } else { |
| out->fd = fd; |
| out->ops = &file_ops; |
| } |
| out->close_fd = false; |
| out->sparse = sparse; |
| out->cur_out_ptr = 0ll; |
| out->chunk_cnt = 0; |
| |
| /* Initialize the crc32 value */ |
| out->crc32 = 0; |
| out->use_crc = crc; |
| |
| if (wipe) |
| wipe_block_device(out->fd, info.len); |
| |
| if (out->sparse) { |
| sparse_header.blk_sz = info.block_size, |
| sparse_header.total_blks = info.len / info.block_size, |
| sparse_header.total_chunks = chunks; |
| if (out->use_crc) |
| sparse_header.total_chunks++; |
| |
| ret = out->ops->write(out, (u8 *)&sparse_header, sizeof(sparse_header)); |
| if (ret < 0) |
| return NULL; |
| } |
| |
| return out; |
| } |
| |
| struct output_file *open_output_file(const char *filename, int gz, int sparse, |
| int chunks, int crc, int wipe) { |
| |
| int fd; |
| struct output_file *file; |
| |
| if (strcmp(filename, "-")) { |
| fd = open(filename, O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0644); |
| if (fd < 0) { |
| error_errno("open"); |
| return NULL; |
| } |
| } else { |
| fd = STDOUT_FILENO; |
| } |
| |
| file = open_output_fd(fd, gz, sparse, chunks, crc, wipe); |
| if (!file) { |
| close(fd); |
| return NULL; |
| } |
| |
| file->close_fd = true; // we opened descriptor thus we responsible for closing it |
| |
| return file; |
| } |
| |
| void pad_output_file(struct output_file *out, u64 len) |
| { |
| int ret; |
| |
| if (len > (u64) info.len) { |
| error("attempted to pad file %llu bytes past end of filesystem", |
| len - info.len); |
| return; |
| } |
| if (out->sparse) { |
| /* We need to emit a DONT_CARE chunk to pad out the file if the |
| * cur_out_ptr is not already at the end of the filesystem. |
| */ |
| if (len < out->cur_out_ptr) { |
| error("attempted to pad file %llu bytes less than the current output pointer", |
| out->cur_out_ptr - len); |
| return; |
| } |
| if (len > out->cur_out_ptr) { |
| emit_skip_chunk(out, len - out->cur_out_ptr); |
| } |
| } else { |
| //KEN TODO: Fixme. If the filesystem image needs no padding, |
| // this will overwrite the last byte in the file with 0 |
| // The answer is to do accounting like the sparse image |
| // code does and know if there is already data there. |
| ret = out->ops->seek(out, len - 1); |
| if (ret < 0) |
| return; |
| |
| ret = out->ops->write(out, (u8*)"", 1); |
| if (ret < 0) |
| return; |
| } |
| } |
| |
| /* Write a contiguous region of data blocks from a memory buffer */ |
| void write_data_block(struct output_file *out, u64 off, u8 *data, int len) |
| { |
| int ret; |
| |
| if (off + len > (u64) info.len) { |
| error("attempted to write block %llu past end of filesystem", |
| off + len - info.len); |
| return; |
| } |
| |
| if (out->sparse) { |
| write_chunk_raw(out, off, data, len); |
| } else { |
| ret = out->ops->seek(out, off); |
| if (ret < 0) |
| return; |
| |
| ret = out->ops->write(out, data, len); |
| if (ret < 0) |
| return; |
| } |
| } |
| |
| /* Write a contiguous region of data blocks with a fill value */ |
| void write_fill_block(struct output_file *out, u64 off, u32 fill_val, int len) |
| { |
| int ret; |
| unsigned int i; |
| int write_len; |
| u32 fill_buf[4096/sizeof(u32)]; /* Maximum size of a block */ |
| |
| if (off + len > (u64) info.len) { |
| error("attempted to write block %llu past end of filesystem", |
| off + len - info.len); |
| return; |
| } |
| |
| if (out->sparse) { |
| write_chunk_fill(out, off, fill_val, len); |
| } else { |
| /* Initialize fill_buf with the fill_val */ |
| for (i = 0; i < sizeof(fill_buf)/sizeof(u32); i++) { |
| fill_buf[i] = fill_val; |
| } |
| |
| ret = out->ops->seek(out, off); |
| if (ret < 0) |
| return; |
| |
| while (len) { |
| write_len = (len > (int)sizeof(fill_buf) ? (int)sizeof(fill_buf) : len); |
| ret = out->ops->write(out, (u8 *)fill_buf, write_len); |
| if (ret < 0) { |
| return; |
| } else { |
| len -= write_len; |
| } |
| } |
| } |
| } |
| |
| /* Write a contiguous region of data blocks from a file */ |
| void write_data_file(struct output_file *out, u64 off, const char *file, |
| off64_t offset, int len) |
| { |
| int ret; |
| off64_t aligned_offset; |
| int aligned_diff; |
| int buffer_size; |
| |
| if (off + len >= (u64) info.len) { |
| error("attempted to write block %llu past end of filesystem", |
| off + len - info.len); |
| return; |
| } |
| |
| int file_fd = open(file, O_RDONLY | O_BINARY); |
| if (file_fd < 0) { |
| error_errno("open"); |
| return; |
| } |
| |
| aligned_offset = offset & ~(4096 - 1); |
| aligned_diff = offset - aligned_offset; |
| buffer_size = len + aligned_diff; |
| |
| #ifndef USE_MINGW |
| u8 *data = mmap64(NULL, buffer_size, PROT_READ, MAP_SHARED, file_fd, |
| aligned_offset); |
| if (data == MAP_FAILED) { |
| error_errno("mmap64"); |
| close(file_fd); |
| return; |
| } |
| #else |
| u8 *data = malloc(buffer_size); |
| if (!data) { |
| error_errno("malloc"); |
| close(file_fd); |
| return; |
| } |
| memset(data, 0, buffer_size); |
| #endif |
| |
| if (out->sparse) { |
| write_chunk_raw(out, off, data + aligned_diff, len); |
| } else { |
| ret = out->ops->seek(out, off); |
| if (ret < 0) |
| goto err; |
| |
| ret = out->ops->write(out, data + aligned_diff, len); |
| if (ret < 0) |
| goto err; |
| } |
| |
| err: |
| #ifndef USE_MINGW |
| munmap(data, buffer_size); |
| #else |
| write(file_fd, data, buffer_size); |
| free(data); |
| #endif |
| close(file_fd); |
| } |