X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fresource.c;h=2dde8d55578435c39b64151746a6c1e1dbe88abd;hp=2f984d13be8ca2487f9e5c5fd031bac88ba88faa;hb=cfc2cfc859a047e24d002aa149f73d45d4979d24;hpb=6f77434ea6ff1407603410e28d1edb966c40e568 diff --git a/src/resource.c b/src/resource.c index 2f984d13..2dde8d55 100644 --- a/src/resource.c +++ b/src/resource.c @@ -31,8 +31,10 @@ #include "xpress.h" #include "sha1.h" #include "dentry.h" +#include "config.h" #include #include +#include /* @@ -59,15 +61,13 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, u64 len, u64 offset, u8 contents_ret[]) { - DEBUG2("comp size = %"PRIu64", " - "uncomp size = %"PRIu64", " - "res offset = %"PRIu64"\n", - resource_compressed_size, - resource_uncompressed_size, - resource_offset); - DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"\n", - wimlib_get_compression_type_string(resource_ctype), - len, offset); + DEBUG2("comp size = %"PRIu64", uncomp size = %"PRIu64", " + "res offset = %"PRIu64"", + resource_compressed_size, + resource_uncompressed_size, + resource_offset); + DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"", + wimlib_get_compression_type_string(resource_ctype), len, offset); /* Trivial case */ if (len == 0) return 0; @@ -157,10 +157,9 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, u64 file_offset_of_needed_chunk_entries = resource_offset + start_table_idx * chunk_entry_size; if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) { - ERROR("Failed to seek to byte %"PRIu64" " - "to read chunk table of compressed " - "resource: %m\n", - file_offset_of_needed_chunk_entries); + ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read " + "chunk table of compressed resource", + file_offset_of_needed_chunk_entries); return WIMLIB_ERR_READ; } @@ -195,10 +194,9 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, u64 file_offset_of_first_needed_chunk = resource_offset + chunk_table_size + chunk_offsets[0]; if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) { - ERROR("Failed to seek to byte %"PRIu64" " - "to read first chunk of compressed " - "resource: %m\n", - file_offset_of_first_needed_chunk); + ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read " + "first chunk of compressed resource", + file_offset_of_first_needed_chunk); return WIMLIB_ERR_READ; } @@ -218,8 +216,8 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, /* Decompress all the chunks. */ for (u64 i = start_chunk; i <= end_chunk; i++) { - DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64")\n", - i, start_chunk, end_chunk); + DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64").", + i, start_chunk, end_chunk); /* Calculate the sizes of the compressed chunk and of the * uncompressed chunk. */ @@ -254,8 +252,9 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, uncompressed_chunk_size = WIM_CHUNK_SIZE; } - DEBUG2("compressed_chunk_size = %u, uncompressed_chunk_size = %u\n", - compressed_chunk_size, uncompressed_chunk_size); + DEBUG2("compressed_chunk_size = %u, " + "uncompressed_chunk_size = %u", + compressed_chunk_size, uncompressed_chunk_size); /* Figure out how much of this chunk we actually need to read */ @@ -274,9 +273,9 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, bool is_partial_chunk = (partial_chunk_size != uncompressed_chunk_size); - DEBUG2("start_offset = %u, end_offset = %u\n", start_offset, + DEBUG2("start_offset = %u, end_offset = %u", start_offset, end_offset); - DEBUG2("partial_chunk_size = %u\n", partial_chunk_size); + DEBUG2("partial_chunk_size = %u", partial_chunk_size); /* This is undocumented, but chunks can be uncompressed. This * appears to always be the case when the compressed chunk size @@ -286,8 +285,8 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, if (start_offset != 0) { if (fseeko(fp, start_offset, SEEK_CUR) != 0) { - ERROR("Uncompressed partial chunk " - "fseek() error: %m\n"); + ERROR_WITH_ERRNO("Uncompressed partial " + "chunk fseek() error"); return WIMLIB_ERR_READ; } } @@ -317,7 +316,6 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, memcpy(out_p, uncompressed_buf + start_offset, partial_chunk_size); } else { - DEBUG2("out_p = %p\n"); ret = decompress(compressed_buf, compressed_chunk_size, out_p, @@ -336,31 +334,31 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, err: if (feof(fp)) - ERROR("Unexpected EOF in compressed file resource\n"); + ERROR("Unexpected EOF in compressed file resource"); else - ERROR("Error reading compressed file resource: %m\n"); + ERROR_WITH_ERRNO("Error reading compressed file resource"); return WIMLIB_ERR_READ; } /* * Reads uncompressed data from an open file stream. */ -int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, - u8 contents_ret[]) +int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, + u8 contents_ret[]) { if (fseeko(fp, offset, SEEK_SET) != 0) { ERROR("Failed to seek to byte %"PRIu64" of input file " - "to read uncompressed resource " - "(len = %"PRIu64")!\n", offset, len); + "to read uncompressed resource (len = %"PRIu64")", + offset, len); return WIMLIB_ERR_READ; } if (fread(contents_ret, 1, len, fp) != len) { if (feof(fp)) { - ERROR("Unexpected EOF in uncompressed file resource!\n"); + ERROR("Unexpected EOF in uncompressed file resource"); } else { ERROR("Failed to read %"PRIu64" bytes from " - "uncompressed resource at offset " - "%"PRIu64"\n", len, offset); + "uncompressed resource at offset %"PRIu64, + len, offset); } return WIMLIB_ERR_READ; } @@ -368,146 +366,7 @@ int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, } -/* - * Reads a WIM resource. - * - * @fp: The FILE* for the WIM file. - * @resource_size: The compressed size of the resource. - * @resource_original_size: The uncompressed size of the resource. - * @resource_offset: The offset of the resource in the stream @fp. - * @resource_ctype: The compression type of the resource. - * (WIM_COMPRESSION_TYPE_*) - * @len: How many bytes of the resource should be read. - * @offset: The offset within the resource at which the read - * will occur. - * - * To read the whole file resource, specify offset = - * 0 and len = resource_original_size, or call - * read_full_resource(). - * - * @contents_ret: An array, that must have length at least @len, - * into which the uncompressed contents of - * the file resource starting at @offset and - * continuing for @len bytes will be written. - * - * @return: Zero on success, nonzero on failure. Failure may be due to - * being unable to read the data from the WIM file at the - * specified length and offset, or it may be due to the - * compressed data (if the data is compressed) being - * invalid. - */ -int read_resource(FILE *fp, u64 resource_size, u64 resource_original_size, - u64 resource_offset, int resource_ctype, u64 len, - u64 offset, void *contents_ret) -{ - if (resource_ctype == WIM_COMPRESSION_TYPE_NONE) { - if (resource_size != resource_original_size) { - ERROR("Resource with original size %"PRIu64" " - "bytes is marked as uncompressed, \n", - resource_original_size); - ERROR(" but its actual size is %"PRIu64" " - "bytes!\n", - resource_size); - return WIMLIB_ERR_INVALID_RESOURCE_SIZE; - } - return read_uncompressed_resource(fp, - resource_offset + offset, - len, contents_ret); - } else { - return read_compressed_resource(fp, resource_size, - resource_original_size, resource_offset, - resource_ctype, len, offset, contents_ret); - } -} - - -/* - * Extracts the first @size bytes file resource specified by @entry to the open - * file @fd. Returns nonzero on error. - * - * XXX - * This function is somewhat redundant with uncompress_resource(). The - * main difference is that this function writes to a file descriptor using - * low-level calls to write() rather than to a FILE* with fwrite(); also this - * function allows only up to @size bytes to be extracted. - */ -int extract_resource_to_fd(WIMStruct *w, const struct resource_entry *entry, - int fd, u64 size) -{ - u64 num_chunks; - u64 n; - u8 buf[min(size, WIM_CHUNK_SIZE)]; - int res_ctype; - u64 offset; - u64 i; - int ret; - - errno = 0; - - num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; - n = WIM_CHUNK_SIZE; - res_ctype = wim_resource_compression_type(w, entry); - offset = 0; - for (i = 0; i < num_chunks; i++) { - if (i == num_chunks - 1) { - n = size % WIM_CHUNK_SIZE; - if (n == 0) { - n = WIM_CHUNK_SIZE; - } - } - - ret = read_resource(w->fp, entry->size, entry->original_size, - entry->offset, res_ctype, n, offset, buf); - if (ret != 0) - return ret; - - if (full_write(fd, buf, n) != n) - return WIMLIB_ERR_WRITE; - offset += n; - } - return ret; -} - -/* - * Copies the file resource specified by the lookup table entry @lte from the - * input WIM, pointed to by the fp field of the WIMStruct, to the output WIM, - * pointed to by the out_fp field of the WIMStruct. - * - * The output_resource_entry, out_refcnt, and part_number fields of @lte are - * updated. - * - * Metadata resources are not copied (they are handled elsewhere for joining and - * splitting). - */ -int copy_resource(struct lookup_table_entry *lte, void *w) -{ - if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) && - !((WIMStruct*)w)->write_metadata) { - return 0; - } - - FILE *in_fp = ((WIMStruct*)w)->fp; - FILE *out_fp = ((WIMStruct*)w)->out_fp; - int ret; - u64 size = lte->resource_entry.size; - u64 offset = lte->resource_entry.offset; - off_t new_offset = ftello(out_fp); - if (new_offset == -1) - return WIMLIB_ERR_WRITE; - - ret = copy_between_files(in_fp, offset, out_fp, size); - if (ret != 0) - return ret; - - memcpy(<e->output_resource_entry, <e->resource_entry, - sizeof(struct resource_entry)); - - lte->output_resource_entry.offset = new_offset; - lte->out_refcnt = lte->refcnt; - lte->part_number = ((WIMStruct*)w)->hdr.part_number; - return 0; -} /* Reads the contents of a struct resource_entry, as represented in the on-disk * format, from the memory pointed to by @p, and fills in the fields of @entry. @@ -538,363 +397,634 @@ u8 *put_resource_entry(u8 *p, const struct resource_entry *entry) return p; } -/* Given the compression type for the WIM file as a whole as the flags field of - * a resource entry, returns the compression type for that resource entry. */ -int resource_compression_type(int wim_ctype, int reshdr_flags) -{ - if (wim_ctype == WIM_COMPRESSION_TYPE_NONE) { - return WIM_COMPRESSION_TYPE_NONE; - } else { - if (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED) - return wim_ctype; - else - return WIM_COMPRESSION_TYPE_NONE; - } -} - - - /* - * Copies bytes between two file streams. + * Reads some data from the resource corresponding to a WIM lookup table entry. + * + * @lte: The WIM lookup table entry for the resource. + * @buf: Buffer into which to write the data. + * @size: Number of bytes to read. + * @offset: Offset at which to start reading the resource. + * @raw: If %true, compressed data is read literally rather than being + * decompressed first. * - * Copies @len bytes from @in to @out, at the current position in @out, and at - * an offset of @in_offset in @in. + * Returns zero on success, nonzero on failure. */ -int copy_between_files(FILE *in, off_t in_offset, FILE *out, size_t len) +int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], + size_t size, u64 offset, bool raw) { - u8 buf[BUFFER_SIZE]; - size_t n; + /* We shouldn't be allowing read over-runs in any part of the library. + * */ + if (raw) + wimlib_assert(offset + size <= lte->resource_entry.size); + else + wimlib_assert(offset + size <= lte->resource_entry.original_size); - if (fseeko(in, in_offset, SEEK_SET) != 0) { - ERROR("Failed to seek to byte %"PRIu64" of input file: %m\n", - in_offset); - return WIMLIB_ERR_READ; - } - /* To reduce memory usage and improve speed, read and write BUFFER_SIZE - * bytes at a time. */ - while (len != 0) { - n = min(len, BUFFER_SIZE); - if (fread(buf, 1, n, in) != n) { - if (feof(in)) { - ERROR("Unexpected EOF when copying data " - "between files\n"); - } else { - ERROR("Error copying data between files: %m\n"); + int ctype; + int ret; + FILE *fp; + switch (lte->resource_location) { + case RESOURCE_IN_WIM: + /* The resource is in a WIM file, and its WIMStruct is given by + * the lte->wim member. The resource may be either compressed + * or uncompressed. */ + wimlib_assert(lte->wim); + wimlib_assert(lte->wim->fp); + ctype = wim_resource_compression_type(lte); + + wimlib_assert(ctype != WIM_COMPRESSION_TYPE_NONE || + (lte->resource_entry.original_size == + lte->resource_entry.size)); + + if (raw || ctype == WIM_COMPRESSION_TYPE_NONE) + return read_uncompressed_resource(lte->wim->fp, + lte->resource_entry.offset + offset, + size, buf); + else + return read_compressed_resource(lte->wim->fp, + lte->resource_entry.size, + lte->resource_entry.original_size, + lte->resource_entry.offset, + ctype, size, offset, buf); + break; + case RESOURCE_IN_STAGING_FILE: + /* The WIM FUSE implementation needs to handle multiple open + * file descriptors per lookup table entry so it does not + * currently work with this function. */ + wimlib_assert(lte->staging_file_name); + wimlib_assert(0); + break; + case RESOURCE_IN_FILE_ON_DISK: + /* The resource is in some file on the external filesystem and + * needs to be read uncompressed */ + wimlib_assert(lte->file_on_disk); + /* Use existing file pointer if available; otherwise open one + * temporarily */ + if (lte->file_on_disk_fp) { + fp = lte->file_on_disk_fp; + } else { + fp = fopen(lte->file_on_disk, "rb"); + if (!fp) { + ERROR_WITH_ERRNO("Failed to open the file " + "`%s'", lte->file_on_disk); } - return WIMLIB_ERR_READ; } - - if (fwrite(buf, 1, n, out) != n) { - ERROR("Error copying data between files: %m\n"); - return WIMLIB_ERR_WRITE; - } - len -= n; + ret = read_uncompressed_resource(fp, offset, size, buf); + if (fp != lte->file_on_disk_fp) + fclose(fp); + return ret; + break; + case RESOURCE_IN_ATTACHED_BUFFER: + /* The resource is directly attached uncompressed in an + * in-memory buffer. */ + wimlib_assert(lte->attached_buffer); + memcpy(buf, lte->attached_buffer + offset, size); + return 0; + break; + default: + assert(0); } - return 0; } - /* - * Uncompresses a WIM file resource and writes it uncompressed to a file stream. + * Reads all the data from the resource corresponding to a WIM lookup table + * entry. * - * @in: The file stream that contains the file resource. - * @size: The size of the resource in the input file. - * @original_size: The original (uncompressed) size of the resource. - * @offset: The offset of the start of the resource in @in. - * @input_ctype: The compression type of the resource in @in. - * @out: The file stream to write the file resource to. + * @lte: The WIM lookup table entry for the resource. + * @buf: Buffer into which to write the data. It must be at least + * wim_resource_size(lte) bytes long. + * + * Returns 0 on success; nonzero on failure. */ -static int uncompress_resource(FILE *in, u64 size, u64 original_size, - off_t offset, int input_ctype, FILE *out) +int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[]) { - int ret; - u8 buf[WIM_CHUNK_SIZE]; - /* Determine how many compressed chunks the file is divided into. */ - u64 num_chunks; - u64 i; - u64 uncompressed_offset; - u64 uncompressed_chunk_size; - - num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; + return read_wim_resource(lte, buf, wim_resource_size(lte), 0, false); +} - for (i = 0; i < num_chunks; i++) { +/* Chunk table that's located at the beginning of each compressed resource in + * the WIM. (This is not the on-disk format; the on-disk format just has an + * array of offsets.) */ +struct chunk_table { + off_t file_offset; + u64 num_chunks; + u64 original_resource_size; + u64 bytes_per_chunk_entry; + u64 table_disk_size; + u64 cur_offset; + u64 *cur_offset_p; + u64 offsets[0]; +}; - uncompressed_offset = i * WIM_CHUNK_SIZE; - uncompressed_chunk_size = min(WIM_CHUNK_SIZE, - original_size - uncompressed_offset); +/* + * Allocates and initializes a chunk table, and reserves space for it in the + * output file. + */ +static int +begin_wim_resource_chunk_tab(const struct lookup_table_entry *lte, + FILE *out_fp, + off_t file_offset, + struct chunk_table **chunk_tab_ret) +{ + u64 size = wim_resource_size(lte); + u64 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; + struct chunk_table *chunk_tab = MALLOC(sizeof(struct chunk_table) + + num_chunks * sizeof(u64)); + int ret = 0; - ret = read_resource(in, size, original_size, offset, input_ctype, - uncompressed_chunk_size, - uncompressed_offset, buf); - if (ret != 0) - return ret; + wimlib_assert(size != 0); - if (fwrite(buf, 1, uncompressed_chunk_size, out) != - uncompressed_chunk_size) { - ERROR("Failed to write file resource: %m\n"); - return WIMLIB_ERR_WRITE; - } + if (!chunk_tab) { + ERROR("Failed to allocate chunk table for %"PRIu64" byte " + "resource", size); + ret = WIMLIB_ERR_NOMEM; + goto out; } - return 0; + chunk_tab->file_offset = file_offset; + chunk_tab->num_chunks = num_chunks; + chunk_tab->original_resource_size = size; + chunk_tab->bytes_per_chunk_entry = (size >= (1ULL << 32)) ? 8 : 4; + chunk_tab->table_disk_size = chunk_tab->bytes_per_chunk_entry * + (num_chunks - 1); + chunk_tab->cur_offset = 0; + chunk_tab->cur_offset_p = chunk_tab->offsets; + + if (fwrite(chunk_tab, 1, chunk_tab->table_disk_size, out_fp) != + chunk_tab->table_disk_size) { + ERROR_WITH_ERRNO("Failed to write chunk table in compressed " + "file resource"); + ret = WIMLIB_ERR_WRITE; + goto out; + } + + *chunk_tab_ret = chunk_tab; +out: + return ret; } /* - * Transfers a file resource between two files, writing it compressed. The file - * resource in the input file may be either compressed or uncompressed. - * Alternatively, the input resource may be in-memory, but it must be - * uncompressed. + * Compresses a chunk of a WIM resource. + * + * @chunk: Uncompressed data of the chunk. + * @chunk_size: Size of the uncompressed chunk in bytes. + * @compressed_chunk: Pointer to output buffer of size at least + * (@chunk_size - 1) bytes. + * @compressed_chunk_len_ret: Pointer to an unsigned int into which the size + * of the compressed chunk will be + * returned. + * @ctype: Type of compression to use. Must be WIM_COMPRESSION_TYPE_LZX + * or WIM_COMPRESSION_TYPE_XPRESS. * - * @in: The file stream that contains the file resource. Ignored - * if uncompressed_resource != NULL. - * @uncompressed_resource: If this pointer is not NULL, it points to an - * array of @original_size bytes that are - * the uncompressed input resource. - * @size: The size of the resource in the input file. - * @original_size: The original (uncompressed) size of the resource. - * @offset: The offset of the start of the resource in @in. Ignored - * if uncompressed_resource != NULL. - * @input_ctype: The compression type of the resource in @in. Ignored if - * uncompressed_resource != NULL. - * @out: The file stream to write the file resource to. - * @output_type: The compression type to use when writing the resource to - * @out. - * @new_size_ret: A location into which the new compressed size of the file - * resource in returned. + * Returns zero if compressed succeeded, and nonzero if the chunk could not be + * compressed to any smaller than @chunk_size. This function cannot fail for + * any other reasons. */ -static int recompress_resource(FILE *in, const u8 *uncompressed_resource, - u64 size, u64 original_size, - off_t offset, int input_ctype, FILE *out, - int output_ctype, u64 *new_size_ret) +static int compress_chunk(const u8 chunk[], unsigned chunk_size, + u8 compressed_chunk[], + unsigned *compressed_chunk_len_ret, + int ctype) { - int ret; - int (*compress)(const void *, uint, void *, uint *); - if (output_ctype == WIM_COMPRESSION_TYPE_LZX) + unsigned compressed_chunk_sz; + int (*compress)(const void *, unsigned, void *, unsigned *); + switch (ctype) { + case WIM_COMPRESSION_TYPE_LZX: compress = lzx_compress; - else + break; + case WIM_COMPRESSION_TYPE_XPRESS: compress = xpress_compress; + break; + default: + wimlib_assert(0); + break; + } + return (*compress)(chunk, chunk_size, compressed_chunk, + compressed_chunk_len_ret); +} - u8 uncompressed_buf[WIM_CHUNK_SIZE]; - u8 compressed_buf[WIM_CHUNK_SIZE - 1]; - - /* Determine how many compressed chunks the file needs to be divided - * into. */ - u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; - - u64 num_chunk_entries = num_chunks - 1; - - /* Size of the chunk entries--- 8 bytes for files over 4GB, otherwise 4 - * bytes */ - uint chunk_entry_size = (original_size >= (u64)1 << 32) ? 8 : 4; - - /* Array in which to construct the chunk offset table. */ - u64 chunk_offsets[num_chunk_entries]; +/* + * Writes a chunk of a WIM resource to an output file. + * + * @chunk: Uncompressed data of the chunk. + * @chunk_size: Size of the chunk (<= WIM_CHUNK_SIZE) + * @out_fp: FILE * to write tho chunk to. + * @out_ctype: Compression type to use when writing the chunk (ignored if no + * chunk table provided) + * @chunk_tab: Pointer to chunk table being created. It is updated with the + * offset of the chunk we write. + * + * Returns 0 on success; nonzero on failure. + */ +static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size, + FILE *out_fp, int out_ctype, + struct chunk_table *chunk_tab) +{ + const u8 *out_chunk; + unsigned out_chunk_size; - /* Offset of the start of the chunk table in the output file. */ - off_t chunk_tab_offset = ftello(out); + wimlib_assert(chunk_size <= WIM_CHUNK_SIZE); - /* Total size of the chunk table (as written to the file) */ - u64 chunk_tab_size = chunk_entry_size * num_chunk_entries; + if (!chunk_tab) { + out_chunk = chunk; + out_chunk_size = chunk_size; + } else { + u8 *compressed_chunk = alloca(chunk_size); + int ret; + unsigned compressed_chunk_len; - /* Reserve space for the chunk table. */ - if (fwrite(chunk_offsets, 1, chunk_tab_size, out) != chunk_tab_size) { - ERROR("Failed to write chunk offset table: %m\n"); + ret = compress_chunk(chunk, chunk_size, compressed_chunk, + &out_chunk_size, out_ctype); + if (ret == 0) { + out_chunk = compressed_chunk; + } else { + out_chunk = chunk; + out_chunk_size = chunk_size; + } + *chunk_tab->cur_offset_p++ = chunk_tab->cur_offset; + chunk_tab->cur_offset += out_chunk_size; + } + + if (fwrite(out_chunk, 1, out_chunk_size, out_fp) != out_chunk_size) { + ERROR_WITH_ERRNO("Failed to write WIM resource chunk"); return WIMLIB_ERR_WRITE; } + return 0; +} - /* Read each chunk of the file, compress it, write it to the output - * file, and update th chunk offset table. */ - u64 cur_chunk_offset = 0; - for (u64 i = 0; i < num_chunks; i++) { - - u64 uncompressed_offset = i * WIM_CHUNK_SIZE; - u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE, - original_size - uncompressed_offset); - - const u8 *uncompressed_p; - if (uncompressed_resource != NULL) { - uncompressed_p = uncompressed_resource + - uncompressed_offset; +/* + * Finishes a WIM chunk tale and writes it to the output file at the correct + * offset. + * + * The final size of the full compressed resource is returned in the + * @compressed_size_p. + */ +static int +finish_wim_resource_chunk_tab(struct chunk_table *chunk_tab, + FILE *out_fp, u64 *compressed_size_p) +{ + size_t bytes_written; + if (fseeko(out_fp, chunk_tab->file_offset, SEEK_SET) != 0) { + ERROR_WITH_ERRNO("Failed to seek to byte "PRIu64" of output " + "WIM file", chunk_tab->file_offset); + return WIMLIB_ERR_WRITE; + } - } else { - /* Read chunk i of the file into uncompressed_buf. */ - ret = read_resource(in, size, original_size, offset, input_ctype, - uncompressed_chunk_size, - uncompressed_offset, - uncompressed_buf); - if (ret != 0) - return ret; - uncompressed_p = uncompressed_buf; - } + if (chunk_tab->bytes_per_chunk_entry == 8) { + array_to_le64(chunk_tab->offsets, chunk_tab->num_chunks); + } else { + for (u64 i = 0; i < chunk_tab->num_chunks; i++) + ((u32*)chunk_tab->offsets)[i] = + to_le32(chunk_tab->offsets[i]); + } + bytes_written = fwrite((u8*)chunk_tab->offsets + + chunk_tab->bytes_per_chunk_entry, + 1, chunk_tab->table_disk_size, out_fp); + if (bytes_written != chunk_tab->table_disk_size) { + ERROR_WITH_ERRNO("Failed to write chunk table in compressed " + "file resource"); + return WIMLIB_ERR_WRITE; + } + if (fseeko(out_fp, 0, SEEK_END) != 0) { + ERROR_WITH_ERRNO("Failed to seek to end of output WIM file"); + return WIMLIB_ERR_WRITE; + } + *compressed_size_p = chunk_tab->cur_offset + chunk_tab->table_disk_size; + return 0; +} - if (i != 0) - chunk_offsets[i - 1] = cur_chunk_offset; +/* + * Writes a WIM resource to a FILE * opened for writing. The resource may be + * written uncompressed or compressed depending on the @out_ctype parameter. + * + * If by chance the resource compresses to more than the original size (this may + * happen with random data or files than are pre-compressed), the resource is + * instead written uncompressed (and this is reflected in the @out_res_entry by + * removing the WIM_RESHDR_FLAG_COMPRESSED flag). + * + * @lte: The lookup table entry for the WIM resource. + * @out_fp: The FILE * to write the resource to. + * @out_ctype: The compression type of the resource to write. Note: if this is + * the same as the compression type of the WIM resource we + * need to read, we simply copy the data (i.e. we do not + * uncompress it, then compress it again). + * @out_res_entry: If non-NULL, a resource entry that is filled in with the + * offset, original size, compressed size, and compression flag + * of the output resource. + * + * Returns 0 on success; nonzero on failure. + */ +static int write_wim_resource(struct lookup_table_entry *lte, + FILE *out_fp, int out_ctype, + struct resource_entry *out_res_entry) +{ + u64 bytes_remaining; + u64 original_size; + u64 old_compressed_size; + u64 new_compressed_size; + u64 offset = 0; + int ret = 0; + struct chunk_table *chunk_tab = NULL; + bool raw; + off_t file_offset; + + /* Original size of the resource */ + original_size = wim_resource_size(lte); + + /* Compressed size of the resource (as it exists now) */ + old_compressed_size = wim_resource_compressed_size(lte); + + /* Current offset in output file */ + file_offset = ftello(out_fp); + if (file_offset == -1) { + ERROR_WITH_ERRNO("Failed to get offset in output " + "stream"); + return WIMLIB_ERR_WRITE; + } + + /* Are the compression types the same? If so, do a raw copy (copy + * without decompressing and recompressing the data). */ + raw = (wim_resource_compression_type(lte) == out_ctype + && out_ctype != WIM_COMPRESSION_TYPE_NONE); + if (raw) + bytes_remaining = old_compressed_size; + else + bytes_remaining = original_size; - uint compressed_len; + /* Empty resource; nothing needs to be done, so just return success. */ + if (bytes_remaining == 0) + return 0; - ret = compress(uncompressed_p, uncompressed_chunk_size, - compressed_buf, &compressed_len); + /* Buffer for reading chunks for the resource */ + char buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; - /* if compress() returned nonzero, the compressed chunk would - * have been at least as large as the uncompressed chunk. In - * this situation, the WIM format requires that the uncompressed - * chunk be written instead. */ - const u8 *buf_to_write; - uint len_to_write; - if (ret == 0) { - buf_to_write = compressed_buf; - len_to_write = compressed_len; - } else { - buf_to_write = uncompressed_p; - len_to_write = uncompressed_chunk_size; - } + /* If we are writing a compressed resource and not doing a raw copy, we + * need to initialize the chunk table */ + if (out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) { + ret = begin_wim_resource_chunk_tab(lte, out_fp, file_offset, + &chunk_tab); + if (ret != 0) + goto out; + } - if (fwrite(buf_to_write, 1, len_to_write, out) != len_to_write) { - ERROR("Failed to write compressed file resource: %m\n"); - return WIMLIB_ERR_WRITE; + /* If the WIM resource is in an external file, open a FILE * to it so we + * don't have to open a temporary one in read_wim_resource() for each + * chunk. */ + if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK + && !lte->file_on_disk_fp) + { + wimlib_assert(lte->file_on_disk); + lte->file_on_disk_fp = fopen(lte->file_on_disk, "rb"); + if (!lte->file_on_disk_fp) { + ERROR_WITH_ERRNO("Failed to open the file `%s' for " + "reading", lte->file_on_disk); + ret = WIMLIB_ERR_OPEN; + goto out; } - cur_chunk_offset += len_to_write; } - /* The chunk offset after the last chunk, plus the size of the chunk - * table, gives the total compressed size of the resource. */ - *new_size_ret = cur_chunk_offset + chunk_tab_size; - - /* Now that all entries of the chunk table are determined, rewind the - * stream to where the chunk table was, and write it back out. */ - - if (fseeko(out, chunk_tab_offset, SEEK_SET) != 0) { - ERROR("Failed to seek to beginning of chunk table: %m\n"); - return WIMLIB_ERR_READ; + /* If we aren't doing a raw copy, we will compute the SHA1 message + * digest of the resource as we read it, and verify it's the same as the + * hash given in the lookup table entry once we've finished reading the + * resource. */ + SHA_CTX ctx; + if (!raw) + sha1_init(&ctx); + + /* While there are still bytes remaining in the WIM resource, read a + * chunk of the resource, update SHA1, then write that chunk using the + * desired compression type. */ + do { + u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE); + ret = read_wim_resource(lte, buf, to_read, offset, raw); + if (ret != 0) + goto out_fclose; + if (!raw) + sha1_update(&ctx, buf, to_read); + ret = write_wim_resource_chunk(buf, to_read, out_fp, + out_ctype, chunk_tab); + if (ret != 0) + goto out_fclose; + bytes_remaining -= to_read; + offset += to_read; + } while (bytes_remaining); + + /* If writing a compressed resource and not doing a raw copy, write the + * chunk table, and finish_wim_resource_chunk_tab() will provide the + * compressed size of the resource we wrote. Otherwise, the compressed + * size of the written resource is the same as the compressed size of + * the existing resource. */ + if (out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) { + ret = finish_wim_resource_chunk_tab(chunk_tab, out_fp, + &new_compressed_size); + if (ret != 0) + goto out_fclose; + } else { + new_compressed_size = old_compressed_size; } - if (chunk_entry_size == 8) { - array_to_le64(chunk_offsets, num_chunk_entries); + /* Verify SHA1 message digest of the resource, unless we are doing a raw + * write (in which case we never even saw the uncompressed data). Or, + * if the hash we had before is all 0's, just re-set it to be the new + * hash. */ + if (!raw) { + u8 md[SHA1_HASH_SIZE]; + sha1_final(md, &ctx); + if (is_zero_hash(lte->hash)) { + copy_hash(lte->hash, md); + } else if (!hashes_equal(md, lte->hash)) { + ERROR("WIM resource has incorrect hash!"); + if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK) { + ERROR("We were reading it from `%s'; maybe it changed " + "while we were reading it.", + lte->file_on_disk); + } + ret = WIMLIB_ERR_INVALID_RESOURCE_HASH; + goto out_fclose; + } + } - if (fwrite(chunk_offsets, 1, chunk_tab_size, out) != - chunk_tab_size) { - ERROR("Failed to write chunk table: %m\n"); - return WIMLIB_ERR_WRITE; + if (new_compressed_size > original_size) { + /* Oops! We compressed the resource to larger than the original + * size. Write the resource uncompressed instead. */ + if (fseeko(out_fp, file_offset, SEEK_SET) != 0) { + ERROR_WITH_ERRNO("Failed to seek to byte "PRIu64" " + "of output WIM file", file_offset); + ret = WIMLIB_ERR_WRITE; + goto out_fclose; } - } else { - u32 chunk_entries_small[num_chunk_entries]; - for (u64 i = 0; i < num_chunk_entries; i++) - chunk_entries_small[i] = to_le32(chunk_offsets[i]); - if (fwrite(chunk_entries_small, 1, chunk_tab_size, out) != - chunk_tab_size) { - ERROR("Failed to write chunk table: %m\n"); - return WIMLIB_ERR_WRITE; + ret = write_wim_resource(lte, out_fp, WIM_COMPRESSION_TYPE_NONE, + out_res_entry); + if (ret != 0) + goto out_fclose; + if (fflush(out_fp) != 0) { + ERROR_WITH_ERRNO("Failed to flush output WIM file"); + ret = WIMLIB_ERR_WRITE; + goto out_fclose; + } + if (ftruncate(fileno(out_fp), file_offset + out_res_entry->size) != 0) { + ERROR_WITH_ERRNO("Failed to truncate output WIM file"); + ret = WIMLIB_ERR_WRITE; } + goto out_fclose; } - - if (fseeko(out, 0, SEEK_END) != 0) { - ERROR("Failed to seek to end of output file: %m\n"); - return WIMLIB_ERR_WRITE; + wimlib_assert(new_compressed_size <= original_size); + if (out_res_entry) { + out_res_entry->size = new_compressed_size; + out_res_entry->original_size = original_size; + out_res_entry->offset = file_offset; + out_res_entry->flags = lte->resource_entry.flags + & ~WIM_RESHDR_FLAG_COMPRESSED; + if (out_ctype != WIM_COMPRESSION_TYPE_NONE) + out_res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED; + } +out_fclose: + if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK + && lte->file_on_disk_fp) { + fclose(lte->file_on_disk_fp); + lte->file_on_disk_fp = NULL; } +out: + FREE(chunk_tab); + return ret; +} +/* Like write_wim_resource(), but the resource is specified by a buffer of + * uncompressed data rather a lookup table entry; also writes the SHA1 hash of + * the buffer to @hash. */ +static int write_wim_resource_from_buffer(const u8 *buf, u64 buf_size, + FILE *out_fp, int out_ctype, + struct resource_entry *out_res_entry, + u8 hash[SHA1_HASH_SIZE]) +{ + /* Set up a temporary lookup table entry that we provide to + * write_wim_resource(). */ + struct lookup_table_entry lte; + int ret; + lte.resource_entry.flags = 0; + lte.resource_entry.original_size = buf_size; + lte.resource_entry.size = buf_size; + lte.resource_entry.offset = 0; + lte.resource_location = RESOURCE_IN_ATTACHED_BUFFER; + lte.attached_buffer = (u8*)buf; + + zero_hash(lte.hash); + ret = write_wim_resource(<e, out_fp, out_ctype, out_res_entry); + if (ret != 0) + return ret; + copy_hash(hash, lte.hash); return 0; } -int write_resource_from_memory(const u8 resource[], int out_ctype, - u64 resource_original_size, FILE *out, - u64 *resource_size_ret) +/* + * Extracts the first @size bytes of the WIM resource specified by @lte to the + * open file descriptor @fd. + * + * Returns 0 on success; nonzero on failure. + */ +int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd, + u64 size) { - if (out_ctype == WIM_COMPRESSION_TYPE_NONE) { - if (fwrite(resource, 1, resource_original_size, out) != - resource_original_size) { - ERROR("Failed to write resource of length " - "%"PRIu64": %m\n", - resource_original_size); + u64 bytes_remaining = size; + char buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; + u64 offset = 0; + int ret = 0; + u8 hash[SHA1_HASH_SIZE]; + + SHA_CTX ctx; + sha1_init(&ctx); + + while (bytes_remaining) { + u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE); + ret = read_wim_resource(lte, buf, to_read, offset, false); + if (ret != 0) + break; + sha1_update(&ctx, buf, to_read); + if (full_write(fd, buf, to_read) < 0) { + ERROR_WITH_ERRNO("Error extracting WIM resource"); return WIMLIB_ERR_WRITE; } - *resource_size_ret = resource_original_size; - return 0; - } else { - return recompress_resource(NULL, resource, resource_original_size, - resource_original_size, 0, 0, out, out_ctype, - resource_size_ret); + bytes_remaining -= to_read; + offset += to_read; } + sha1_final(hash, &ctx); + if (!hashes_equal(hash, lte->hash)) { + ERROR("Invalid checksum on a WIM resource " + "(detected when extracting to external file)"); + ERROR("The following WIM resource is invalid:"); + print_lookup_table_entry(lte); + return WIMLIB_ERR_INVALID_RESOURCE_HASH; + } + return 0; } +/* + * Extracts the WIM resource specified by @lte to the open file descriptor @fd. + * + * Returns 0 on success; nonzero on failure. + */ +int extract_full_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd) +{ + return extract_wim_resource_to_fd(lte, fd, wim_resource_size(lte)); +} /* - * Transfers a file resource from a FILE* opened for reading to a FILE* opened - * for writing, possibly changing the compression type. + * Copies the file resource specified by the lookup table entry @lte from the + * input WIM to the output WIM that has its FILE * given by + * ((WIMStruct*)wim)->out_fp. * - * @in: The FILE* that contains the file resource. - * @size: The (compressed) size of the file resource. - * @original_size: The uncompressed size of the file resource. - * @offset: The offset of the file resource in the input file. - * @input_ctype: The compression type of the file resource in the input - * file. - * @out: The FILE* for the output file. The file resource is - * written at the current position of @out. - * @output_ctype: The compression type to which the file resource will be - * converted. - * @output_res_entry: A pointer to a resource entry that, upon successful - * return of this function, will have the size, - * original size, offset, and flags fields filled - * in for the file resource written to the output - * file. + * The output_resource_entry, out_refcnt, and part_number fields of @lte are + * updated. + * + * Metadata resources are not copied (they are handled elsewhere for joining and + * splitting). */ -static int transfer_file_resource(FILE *in, u64 size, u64 original_size, - off_t offset, int input_ctype, FILE *out, - int output_ctype, - struct resource_entry *output_res_entry) +int copy_resource(struct lookup_table_entry *lte, void *wim) { + WIMStruct *w = wim; int ret; - /* Handle zero-length files */ - if (original_size == 0) { - memset(output_res_entry, 0, sizeof(*output_res_entry)); + if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) && + !w->write_metadata) return 0; - } - /* Get current offset in the output file. */ - output_res_entry->offset = ftello(out); - if (output_res_entry->offset == -1) { - ERROR("Failed to get output position: %m\n"); - return WIMLIB_ERR_WRITE; - } + ret = write_wim_resource(lte, w->out_fp, + wim_resource_compression_type(lte), + <e->output_resource_entry); + if (ret != 0) + return ret; + lte->out_refcnt = lte->refcnt; + lte->part_number = w->hdr.part_number; + return 0; +} - if (output_ctype == input_ctype) { - /* The same compression types; simply copy the resource. */ +/* + * Writes a dentry's resources, including the main file resource as well as all + * alternate data streams, to the output file. + * + * @dentry: The dentry for the file. + * @wim_p: A pointer to the WIMStruct containing @dentry. + * + * @return zero on success, nonzero on failure. + */ +int write_dentry_resources(struct dentry *dentry, void *wim_p) +{ + WIMStruct *w = wim_p; + int ret = 0; + struct lookup_table_entry *lte; + int ctype = wimlib_get_compression_type(w); - ret = copy_between_files(in, offset, out, size); - if (ret != 0) - return ret; - output_res_entry->size = size; - } else { - /* Different compression types. */ + if (w->write_flags & WIMLIB_WRITE_FLAG_VERBOSE) { + wimlib_assert(dentry->full_path_utf8); + printf("Writing streams for `%s'\n", dentry->full_path_utf8); + } - if (output_ctype == WIM_COMPRESSION_TYPE_NONE) { - /* Uncompress a compressed file resource */ - ret = uncompress_resource(in, size, - original_size, offset, - input_ctype, out); + for (unsigned i = 0; i <= dentry->num_ads; i++) { + lte = dentry_stream_lte(dentry, i, w->lookup_table); + if (lte && ++lte->out_refcnt == 1) { + ret = write_wim_resource(lte, w->out_fp, ctype, + <e->output_resource_entry); if (ret != 0) - return ret; - output_res_entry->size = original_size; - } else { - u64 new_size; - /* Compress an uncompressed file resource, or compress a - * compressed file resource using a different - * compression type */ - ret = recompress_resource(in, NULL, size, original_size, - offset, input_ctype, out, - output_ctype, &new_size); - if (ret != 0) - return ret; - output_res_entry->size = new_size; + break; } - } - - output_res_entry->original_size = original_size; - if (output_ctype == WIM_COMPRESSION_TYPE_NONE) - output_res_entry->flags = 0; - else - output_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED; - return 0; + return ret; } /* @@ -907,29 +1037,35 @@ static int transfer_file_resource(FILE *in, u64 size, u64 original_size, * length 8, because that's how long the 'length' field is. * * @fp: The FILE* for the input WIM file. - * @res_entry: The resource entry for the metadata resource (a.k.a the metadata - * for the metadata) * @wim_ctype: The compression type of the WIM file. - * @root_dentry_p: A pointer to a pointer to a struct dentry structure into which the - * root dentry is allocated and returned. + * @imd: Pointer to the image metadata structure. Its `metadata_lte' + * member specifies the lookup table entry for the metadata + * resource. The rest of the image metadata entry will be filled + * in by this function. * - * @return: True on success, false on failure. + * @return: Zero on success, nonzero on failure. */ -int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry, - int wim_ctype, struct image_metadata *image_metadata) +int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) { u8 *buf; int ctype; u32 dentry_offset; int ret; - struct dentry *dentry = NULL; + struct dentry *dentry; + struct wim_security_data *sd; + struct link_group_table *lgt; + const struct lookup_table_entry *metadata_lte; + const struct resource_entry *res_entry; - DEBUG("Reading metadata resource: length = %lu, offset = %lu\n", - res_entry->original_size, res_entry->offset); + metadata_lte = imd->metadata_lte; + res_entry = &metadata_lte->resource_entry; + + DEBUG("Reading metadata resource: length = %"PRIu64", " + "offset = %"PRIu64"", + res_entry->original_size, res_entry->offset); if (res_entry->original_size < 8) { - ERROR("Expected at least 8 bytes for the metadata " - "resource!\n"); + ERROR("Expected at least 8 bytes for the metadata resource"); return WIMLIB_ERR_INVALID_RESOURCE_SIZE; } @@ -938,257 +1074,151 @@ int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry, if (!buf) { ERROR("Failed to allocate %"PRIu64" bytes for uncompressed " - "metadata resource!\n", - res_entry->original_size); + "metadata resource", res_entry->original_size); return WIMLIB_ERR_NOMEM; } /* Determine the compression type of the metadata resource. */ - ctype = resource_compression_type(wim_ctype, res_entry->flags); /* Read the metadata resource into memory. (It may be compressed.) */ - ret = read_full_resource(fp, res_entry->size, - res_entry->original_size, res_entry->offset, - ctype, buf); + ret = read_full_wim_resource(metadata_lte, buf); if (ret != 0) - goto err1; + goto out_free_buf; + + DEBUG("Finished reading metadata resource into memory."); - DEBUG("Finished reading metadata resource into memory.\n"); + /* The root directory entry starts after security data, on an 8-byte + * aligned address. + * + * The security data starts with a 4-byte integer giving its total + * length. */ + /* Read the security data into a wim_security_data structure. */ + ret = read_security_data(buf, res_entry->original_size, &sd); + if (ret != 0) + goto out_free_buf; dentry = MALLOC(sizeof(struct dentry)); if (!dentry) { - ERROR("Failed to allocate %zu bytes for root dentry!\n", - sizeof(struct dentry)); + ERROR("Failed to allocate %zu bytes for root dentry", + sizeof(struct dentry)); ret = WIMLIB_ERR_NOMEM; - goto err1; + goto out_free_security_data; } - /* Read the root directory entry starts after security data, on an - * 8-byte aligned address. - * - * The security data starts with a 4-byte integer giving its total - * length. */ - - /* Read the security data into a WIMSecurityData structure. */ -#ifdef ENABLE_SECURITY_DATA - ret = read_security_data(buf, res_entry->original_size, - &image_metadata->security_data); - if (ret != 0) - goto err1; -#endif get_u32(buf, &dentry_offset); if (dentry_offset == 0) dentry_offset = 8; - dentry_offset += (8 - dentry_offset % 8) % 8; + dentry_offset = (dentry_offset + 7) & ~7; ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry); - if (ret != 0) - goto err1; - /* This is the root dentry, so set its pointers correctly. */ dentry->parent = dentry; dentry->next = dentry; dentry->prev = dentry; + if (ret != 0) + goto out_free_dentry_tree; + DEBUG("Reading dentry tree"); /* Now read the entire directory entry tree. */ ret = read_dentry_tree(buf, res_entry->original_size, dentry); if (ret != 0) - goto err2; + goto out_free_dentry_tree; + DEBUG("Calculating dentry full paths"); /* Calculate the full paths in the dentry tree. */ ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL); if (ret != 0) - goto err2; + goto out_free_dentry_tree; + + DEBUG("Building link group table"); + /* Build hash table that maps hard link group IDs to dentry sets */ + lgt = new_link_group_table(9001); + if (!lgt) + goto out_free_dentry_tree; + ret = for_dentry_in_tree(dentry, link_group_table_insert, lgt); + if (ret != 0) + goto out_free_lgt; - image_metadata->root_dentry = dentry; - FREE(buf); - return ret; -err2: - free_dentry_tree(dentry, NULL, false); -err1: + DEBUG("Freeing duplicate ADS entries in link group table"); + ret = link_groups_free_duplicate_data(lgt); + if (ret != 0) + goto out_free_lgt; + DEBUG("Done reading image metadata"); + + imd->lgt = lgt; + imd->security_data = sd; + imd->root_dentry = dentry; + goto out_free_buf; +out_free_lgt: + free_link_group_table(lgt); +out_free_dentry_tree: + free_dentry_tree(dentry, NULL); +out_free_security_data: + free_security_data(sd); +out_free_buf: FREE(buf); return ret; } -/* Write the metadata resource for the current image. */ +/* Write the metadata resource for the current WIM image. */ int write_metadata_resource(WIMStruct *w) { - FILE *out; u8 *buf; u8 *p; int ret; - off_t subdir_offset; + u64 subdir_offset; struct dentry *root; - struct lookup_table_entry *lte; - struct resource_entry *res_entry; - off_t metadata_offset; + struct lookup_table_entry *lte, *duplicate_lte; u64 metadata_original_size; - u64 metadata_compressed_size; - int metadata_ctype; - u8 hash[WIM_HASH_SIZE]; - DEBUG("Writing metadata resource for image %u\n", w->current_image); + /* + * We append 20 random bytes to the metadata resource so that we don't + * have identical metadata resources if we happen to append exactly the + * same image twice without any changes in timestamps. If this were to + * happen, it would cause confusion about the number and order of images + * in the WIM. + */ + const unsigned random_tail_len = 20; + + DEBUG("Writing metadata resource for image %d", w->current_image); - out = w->out_fp; root = wim_root_dentry(w); - metadata_ctype = wimlib_get_compression_type(w); - metadata_offset = ftello(out); - if (metadata_offset == -1) - return WIMLIB_ERR_WRITE; - #ifdef ENABLE_SECURITY_DATA struct wim_security_data *sd = wim_security_data(w); if (sd) subdir_offset = sd->total_length + root->length + 8; else - #endif subdir_offset = 8 + root->length + 8; calculate_subdir_offsets(root, &subdir_offset); - metadata_original_size = subdir_offset; + metadata_original_size = subdir_offset + random_tail_len; buf = MALLOC(metadata_original_size); if (!buf) { ERROR("Failed to allocate %"PRIu64" bytes for " - "metadata resource\n", metadata_original_size); + "metadata resource", metadata_original_size); return WIMLIB_ERR_NOMEM; } - #ifdef ENABLE_SECURITY_DATA - /* Write the security data. */ + p = write_security_data(sd, buf); - #else - p = put_u32(buf, 8); /* Total length of security data. */ - p = put_u32(p, 0); /* Number of security data entries. */ - #endif - DEBUG("Writing dentry tree.\n"); + DEBUG("Writing dentry tree."); p = write_dentry_tree(root, p); + randomize_byte_array(p, random_tail_len); + wimlib_assert(p - buf + random_tail_len == metadata_original_size); - /* Like file resources, the lookup table entry for a metadata resource - * uses for the hash code a SHA1 message digest of its uncompressed - * contents. */ - sha1_buffer(buf, metadata_original_size, hash); - - ret = write_resource_from_memory(buf, - metadata_ctype, - metadata_original_size, - out, - &metadata_compressed_size); - FREE(buf); - if (ret != 0) - return ret; - - /* Update the lookup table entry, including the hash and output resource - * entry fields, for this image's metadata resource. */ lte = wim_metadata_lookup_table_entry(w); - res_entry = <e->output_resource_entry; - lte->out_refcnt++; - if (memcmp(hash, lte->hash, WIM_HASH_SIZE) != 0) { - lookup_table_unlink(w->lookup_table, lte); - memcpy(lte->hash, hash, WIM_HASH_SIZE); - lookup_table_insert(w->lookup_table, lte); - } - res_entry->original_size = metadata_original_size; - res_entry->offset = metadata_offset; - res_entry->size = metadata_compressed_size; - res_entry->flags = WIM_RESHDR_FLAG_METADATA; - if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE) - res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED; - return 0; -} - -/* - * Writes a file resource to the output file. - * - * @dentry: The dentry for the file resource. - * @wim_p: A pointer to the WIMStruct. The fields of interest to this - * function are the input and output file streams and the lookup table. - * @return zero on success, nonzero on failure. - */ -int write_file_resource(struct dentry *dentry, void *wim_p) -{ - WIMStruct *w; - FILE *out; - FILE *in; - struct lookup_table_entry *lte; - int in_wim_ctype; - int out_wim_ctype; - struct resource_entry *output_res_entry; - u64 len; - int ret; - - w = wim_p; - out = w->out_fp; - - /* Directories don't need file resources. */ - if (dentry_is_directory(dentry)) - return 0; - - /* Get the lookup entry for the file resource. */ - lte = wim_lookup_resource(w, dentry); - if (!lte) - return 0; - - /* No need to write file resources twice. (This indicates file - * resources that are part of a hard link set.) */ - if (++lte->out_refcnt != 1) - return 0; - out_wim_ctype = wimlib_get_compression_type(w); - output_res_entry = <e->output_resource_entry; - - /* do not write empty resources */ - if (lte->resource_entry.original_size == 0) - return 0; - - /* Figure out if we can read the resource from the WIM file, or - * if we have to read it from the filesystem outside. */ - if (lte->file_on_disk) { - - /* Read from disk (uncompressed) */ - - len = lte->resource_entry.original_size; - - in = fopen(lte->file_on_disk, "rb"); - if (!in) { - ERROR("Failed to open the file `%s': %m\n", - lte->file_on_disk); - return WIMLIB_ERR_OPEN; - } - - if (w->verbose) - puts(lte->file_on_disk); - - ret = transfer_file_resource(in, len, len, 0, - WIM_COMPRESSION_TYPE_NONE, out, - out_wim_ctype, output_res_entry); - fclose(in); - } else { - - /* Read from input WIM (possibly compressed) */ - - /* It may be a different WIM file, in the case of - * exporting images from one WIM file to another */ - if (lte->other_wim_fp) { - /* Different WIM file. */ - in = lte->other_wim_fp; - in_wim_ctype = lte->other_wim_ctype; - } else { - /* Same WIM file. */ - in = w->fp; - in_wim_ctype = out_wim_ctype; - } - int input_res_ctype = resource_compression_type( - in_wim_ctype, - lte->resource_entry.flags); - - ret = transfer_file_resource(in, - lte->resource_entry.size, - lte->resource_entry.original_size, - lte->resource_entry.offset, - input_res_ctype, - out, - out_wim_ctype, - output_res_entry); - } + ret = write_wim_resource_from_buffer(buf, metadata_original_size, + w->out_fp, + wimlib_get_compression_type(w), + <e->output_resource_entry, + lte->hash); + lookup_table_unlink(w->lookup_table, lte); + lookup_table_insert(w->lookup_table, lte); + wimlib_assert(lte->out_refcnt == 0); + lte->out_refcnt++; + lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA; +out: + FREE(buf); return ret; }