X-Git-Url: https://wimlib.net/git/?a=blobdiff_plain;f=src%2Fresource.c;h=3c573f1c9131d4cfb0af8f1ff8e68b87964366b0;hb=c155e206bf806dd639b57922c3eef0b428c985df;hp=f75ce7ab01a3f0a2d96b61a863706bff14caddb6;hpb=3efd3c749f6cef737951769ace2fc802d8d8e61c;p=wimlib diff --git a/src/resource.c b/src/resource.c index f75ce7ab..3c573f1c 100644 --- a/src/resource.c +++ b/src/resource.c @@ -36,6 +36,11 @@ #include #include +#ifdef WITH_NTFS_3G +#include +#include +#include +#endif /* * Reads all or part of a compressed resource into an in-memory buffer. @@ -398,15 +403,19 @@ u8 *put_resource_entry(u8 *p, const struct resource_entry *entry) } /* - * Reads some data from a WIM resource. + * Reads some data from the resource corresponding to a WIM lookup table entry. * - * If %raw is true, compressed data is read literally rather than being - * decompressed first. + * @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. * * Returns zero on success, nonzero on failure. */ -static int __read_wim_resource(const struct lookup_table_entry *lte, - u8 buf[], size_t size, u64 offset, bool raw) +int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], + size_t size, u64 offset, bool raw) { /* We shouldn't be allowing read over-runs in any part of the library. * */ @@ -427,18 +436,9 @@ static int __read_wim_resource(const struct lookup_table_entry *lte, wimlib_assert(lte->wim->fp); ctype = wim_resource_compression_type(lte); - /* XXX This check should be moved elsewhere */ - if (ctype == WIM_COMPRESSION_TYPE_NONE && - lte->resource_entry.original_size != - lte->resource_entry.size) { - ERROR("WIM resource at offset %"PRIu64", size %"PRIu64 - "has an original size of %"PRIu64", but is " - "uncompressed", - lte->resource_entry.offset, - lte->resource_entry.size, - lte->resource_entry.original_size); - return WIMLIB_ERR_INVALID_RESOURCE_SIZE; - } + 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, @@ -452,16 +452,11 @@ static int __read_wim_resource(const struct lookup_table_entry *lte, 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); + wimlib_assert(<e->file_on_disk == <e->staging_file_name); /* Use existing file pointer if available; otherwise open one * temporarily */ if (lte->file_on_disk_fp) { @@ -471,6 +466,7 @@ static int __read_wim_resource(const struct lookup_table_entry *lte, if (!fp) { ERROR_WITH_ERRNO("Failed to open the file " "`%s'", lte->file_on_disk); + return WIMLIB_ERR_OPEN; } } ret = read_uncompressed_resource(fp, offset, size, buf); @@ -485,27 +481,27 @@ static int __read_wim_resource(const struct lookup_table_entry *lte, memcpy(buf, lte->attached_buffer + offset, size); return 0; break; +#ifdef WITH_NTFS_3G + case RESOURCE_IN_NTFS_VOLUME: + if (lte->attr) { + if (ntfs_attr_pread(lte->attr, offset, size, buf) == size) { + return 0; + } else { + ERROR_WITH_ERRNO("Error reading NTFS attribute " + "at `%s'", + lte->ntfs_loc->path_utf8); + return WIMLIB_ERR_NTFS_3G; + } + } else { + wimlib_assert(0); + } + break; +#endif default: assert(0); } } -/* - * 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. - * - * Returns 0 on success; nonzero on failure. - */ -int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], - size_t size, u64 offset) -{ - return __read_wim_resource(lte, buf, size, offset, false); -} - /* * Reads all the data from the resource corresponding to a WIM lookup table * entry. @@ -518,9 +514,12 @@ int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], */ int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[]) { - return __read_wim_resource(lte, buf, wim_resource_size(lte), 0, false); + return read_wim_resource(lte, buf, wim_resource_size(lte), 0, false); } +/* 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; @@ -532,6 +531,10 @@ struct chunk_table { u64 offsets[0]; }; +/* + * 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, @@ -574,6 +577,23 @@ out: return ret; } +/* + * 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. + * + * 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 compress_chunk(const u8 chunk[], unsigned chunk_size, u8 compressed_chunk[], unsigned *compressed_chunk_len_ret, @@ -596,6 +616,19 @@ static int compress_chunk(const u8 chunk[], unsigned chunk_size, compressed_chunk_len_ret); } +/* + * 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) @@ -603,8 +636,9 @@ static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size, const u8 *out_chunk; unsigned out_chunk_size; - if (out_ctype == WIM_COMPRESSION_TYPE_NONE) { - wimlib_assert(chunk_tab == NULL); + wimlib_assert(chunk_size <= WIM_CHUNK_SIZE); + + if (!chunk_tab) { out_chunk = chunk; out_chunk_size = chunk_size; } else { @@ -612,8 +646,6 @@ static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size, int ret; unsigned compressed_chunk_len; - wimlib_assert(chunk_tab != NULL); - ret = compress_chunk(chunk, chunk_size, compressed_chunk, &out_chunk_size, out_ctype); if (ret == 0) { @@ -633,6 +665,13 @@ static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size, return 0; } +/* + * 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) @@ -671,6 +710,11 @@ finish_wim_resource_chunk_tab(struct chunk_table *chunk_tab, * 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 @@ -696,10 +740,17 @@ static int write_wim_resource(struct lookup_table_entry *lte, struct chunk_table *chunk_tab = NULL; bool raw; off_t file_offset; +#ifdef WITH_NTFS_3G + ntfs_inode *ni; +#endif + /* 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 " @@ -707,23 +758,34 @@ static int write_wim_resource(struct lookup_table_entry *lte, return WIMLIB_ERR_WRITE; } - raw = (wim_resource_compression_type(lte) == out_ctype); + /* 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; + /* Empty resource; nothing needs to be done, so just return success. */ if (bytes_remaining == 0) return 0; + /* Buffer for reading chunks for the resource */ char buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; + /* 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 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) { @@ -736,12 +798,47 @@ static int write_wim_resource(struct lookup_table_entry *lte, goto out; } } - +#ifdef WITH_NTFS_3G + else if (lte->resource_location == RESOURCE_IN_NTFS_VOLUME + && !lte->attr) + { + struct ntfs_location *loc = lte->ntfs_loc; + wimlib_assert(loc); + ni = ntfs_pathname_to_inode(*loc->ntfs_vol_p, NULL, loc->path_utf8); + if (!ni) { + ERROR_WITH_ERRNO("Failed to open inode `%s' in NTFS " + "volume", loc->path_utf8); + } + lte->attr = ntfs_attr_open(ni, + loc->is_reparse_point ? AT_REPARSE_POINT : AT_DATA, + (ntfschar*)loc->stream_name_utf16, + loc->stream_name_utf16_num_chars); + if (!lte->attr) { + ntfs_inode_close(ni); + ERROR_WITH_ERRNO("Failed to open attribute of `%s' in " + "NTFS volume", loc->path_utf8); + } + } +#endif + + /* 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); + 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) @@ -749,6 +846,12 @@ static int write_wim_resource(struct lookup_table_entry *lte, 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); @@ -758,6 +861,27 @@ static int write_wim_resource(struct lookup_table_entry *lte, new_compressed_size = old_compressed_size; } + /* 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 (new_compressed_size > original_size) { /* Oops! We compressed the resource to larger than the original * size. Write the resource uncompressed instead. */ @@ -786,11 +910,11 @@ static int write_wim_resource(struct lookup_table_entry *lte, if (out_res_entry) { out_res_entry->size = new_compressed_size; out_res_entry->original_size = original_size; - out_res_entry->offset = file_offset; - if (out_ctype == WIM_COMPRESSION_TYPE_NONE) - out_res_entry->flags = 0; - else - out_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED; + 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 @@ -798,28 +922,50 @@ out_fclose: fclose(lte->file_on_disk_fp); lte->file_on_disk_fp = NULL; } +#ifdef WITH_NTFS_3G + else if (lte->resource_location == RESOURCE_IN_NTFS_VOLUME + && lte->attr) { + ntfs_attr_close(lte->attr); + ntfs_inode_close(ni); + } +#endif 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) + 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; - lte.resource_entry.flags = 0; + 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; - return write_wim_resource(<e, out_fp, out_ctype, out_res_entry); + 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; } /* - * Extracts the first @size bytes of the resource specified by @lte to the open - * file @fd. Returns nonzero on error. + * 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) @@ -828,12 +974,17 @@ int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd, 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); + 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; @@ -841,9 +992,22 @@ int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd, 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)); @@ -851,8 +1015,8 @@ int extract_full_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd /* * 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. + * input WIM to the output WIM that has its FILE * given by + * ((WIMStruct*)wim)->out_fp. * * The output_resource_entry, out_refcnt, and part_number fields of @lte are * updated. @@ -869,7 +1033,8 @@ int copy_resource(struct lookup_table_entry *lte, void *wim) !w->write_metadata) return 0; - ret = write_wim_resource(lte, w->out_fp, wimlib_get_compression_type(w), + ret = write_wim_resource(lte, w->out_fp, + wim_resource_compression_type(lte), <e->output_resource_entry); if (ret != 0) return ret; @@ -883,9 +1048,7 @@ int copy_resource(struct lookup_table_entry *lte, void *wim) * alternate data streams, to the output file. * * @dentry: The dentry for the file. - * @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. + * @wim_p: A pointer to the WIMStruct containing @dentry. * * @return zero on success, nonzero on failure. */ @@ -924,48 +1087,51 @@ int write_dentry_resources(struct dentry *dentry, void *wim_p) * * @fp: The FILE* for the input WIM file. * @wim_ctype: The compression type of the WIM file. - * @imd: Pointer to the image metadata structure. Its - * `lookup_table_entry' member specifies the lookup table entry for - * the metadata resource. The rest of the image metadata entry - * will be filled in by this function. + * @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: Zero on success, nonzero on failure. */ -int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) +int read_metadata_resource(WIMStruct *w, struct image_metadata *imd) { u8 *buf; int ctype; u32 dentry_offset; int ret; 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; + u64 metadata_len; + u64 metadata_offset; metadata_lte = imd->metadata_lte; - res_entry = &metadata_lte->resource_entry; + metadata_len = wim_resource_size(metadata_lte); + metadata_offset = metadata_lte->resource_entry.offset; 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"); + "offset = %"PRIu64"", metadata_len, metadata_offset); + + /* There is no way the metadata resource could possibly be less than (8 + * + WIM_DENTRY_DISK_SIZE) bytes, where the 8 is for security data (with + * no security descriptors) and WIM_DENTRY_DISK_SIZE is for the root + * dentry. */ + if (metadata_len < 8 + WIM_DENTRY_DISK_SIZE) { + ERROR("Expected at least %zu bytes for the metadata resource", + 8 + WIM_DENTRY_DISK_SIZE); return WIMLIB_ERR_INVALID_RESOURCE_SIZE; } /* Allocate memory for the uncompressed metadata resource. */ - buf = MALLOC(res_entry->original_size); + buf = MALLOC(metadata_len); if (!buf) { ERROR("Failed to allocate %"PRIu64" bytes for uncompressed " - "metadata resource", res_entry->original_size); + "metadata resource", metadata_len); return WIMLIB_ERR_NOMEM; } - /* Determine the compression type of the metadata resource. */ - /* Read the metadata resource into memory. (It may be compressed.) */ ret = read_full_wim_resource(metadata_lte, buf); if (ret != 0) @@ -973,17 +1139,27 @@ int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) DEBUG("Finished reading metadata resource into memory."); - /* The root directory entry starts after security data, on an 8-byte - * aligned address. + /* The root directory entry starts after security data, aligned on an + * 8-byte boundary within the metadata resource. * * The security data starts with a 4-byte integer giving its total - * length. */ + * length, so if we round that up to an 8-byte boundary that gives us + * the offset of the root dentry. + * + * Here we read the security data into a wim_security_data structure, + * and if successful, go ahead and calculate the offset in the metadata + * resource of the root dentry. */ - /* Read the security data into a wim_security_data structure. */ - ret = read_security_data(buf, res_entry->original_size, &sd); + ret = read_security_data(buf, metadata_len, &imd->security_data); if (ret != 0) goto out_free_buf; + get_u32(buf, &dentry_offset); + if (dentry_offset == 0) + dentry_offset = 8; + dentry_offset = (dentry_offset + 7) & ~7; + + /* Allocate memory for the root dentry and read it into memory */ dentry = MALLOC(sizeof(struct dentry)); if (!dentry) { ERROR("Failed to allocate %zu bytes for root dentry", @@ -991,13 +1167,9 @@ int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) ret = WIMLIB_ERR_NOMEM; goto out_free_security_data; } - - get_u32(buf, &dentry_offset); - if (dentry_offset == 0) - dentry_offset = 8; - dentry_offset = (dentry_offset + 7) & ~7; - ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry); + ret = read_dentry(buf, metadata_len, dentry_offset, dentry); + /* This is the root dentry, so set its pointers correctly. */ dentry->parent = dentry; dentry->next = dentry; @@ -1005,20 +1177,20 @@ int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) if (ret != 0) goto out_free_dentry_tree; + /* Now read the entire directory entry tree into memory. */ DEBUG("Reading dentry tree"); - /* Now read the entire directory entry tree. */ - ret = read_dentry_tree(buf, res_entry->original_size, dentry); + ret = read_dentry_tree(buf, metadata_len, dentry); if (ret != 0) goto out_free_dentry_tree; - DEBUG("Calculating dentry full paths"); /* Calculate the full paths in the dentry tree. */ + DEBUG("Calculating dentry full paths"); ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL); if (ret != 0) goto out_free_dentry_tree; - DEBUG("Building link group table"); /* Build hash table that maps hard link group IDs to dentry sets */ + DEBUG("Building link group table"); lgt = new_link_group_table(9001); if (!lgt) goto out_free_dentry_tree; @@ -1030,10 +1202,15 @@ int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) ret = link_groups_free_duplicate_data(lgt); if (ret != 0) goto out_free_lgt; + + DEBUG("Running miscellaneous verifications on the dentry tree"); + ret = for_dentry_in_tree(dentry, verify_dentry, w); + 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: @@ -1041,44 +1218,51 @@ out_free_lgt: out_free_dentry_tree: free_dentry_tree(dentry, NULL); out_free_security_data: - free_security_data(sd); + free_security_data(imd->security_data); + imd->security_data = NULL; 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; u64 subdir_offset; struct dentry *root; - struct lookup_table_entry *lte; - off_t metadata_offset; + struct lookup_table_entry *lte, *duplicate_lte; u64 metadata_original_size; - u64 metadata_compressed_size; - int metadata_ctype; - u8 hash[SHA1_HASH_SIZE]; + const struct wim_security_data *sd; + 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; + sd = wim_security_data(w); + + /* We do not allow the security data pointer to be NULL, although it may + * point to an empty security data with no entries. */ + wimlib_assert(sd); + + /* Offset of first child of the root dentry. It's equal to: + * - The total length of the security data, rounded to the next 8-byte + * boundary, + * - plus the total length of the root dentry, + * - plus 8 bytes for an end-of-directory entry following the root + * dentry (shouldn't really be needed, but just in case...) + */ + subdir_offset = ((sd->total_length + 7) & ~7) + dentry_total_length(root) + 8; - struct wim_security_data *sd = wim_security_data(w); - if (sd) - subdir_offset = sd->total_length + root->length + 8; - else - subdir_offset = 8 + root->length + 8; + /* Calculate the subdirectory offsets for the entire dentry tree. */ calculate_subdir_offsets(root, &subdir_offset); - metadata_original_size = subdir_offset; + + /* Total length of the metadata resource (uncompressed) */ + metadata_original_size = subdir_offset + random_tail_len; + + /* Allocate a buffer to contain the uncompressed metadata resource */ buf = MALLOC(metadata_original_size); if (!buf) { ERROR("Failed to allocate %"PRIu64" bytes for " @@ -1086,30 +1270,59 @@ int write_metadata_resource(WIMStruct *w) return WIMLIB_ERR_NOMEM; } + /* Write the security data into the resource buffer */ p = write_security_data(sd, buf); + /* Write the dentry tree into the resource buffer */ DEBUG("Writing dentry tree."); p = write_dentry_tree(root, p); - /* 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); + /* + * 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. + */ + randomize_byte_array(p, random_tail_len); + /* We MUST have exactly filled the buffer; otherwise we calculated its + * size incorrectly or wrote the data incorrectly. */ + wimlib_assert(p - buf + random_tail_len == metadata_original_size); + /* Get the lookup table entry for the metadata resource so we can update + * it. */ lte = wim_metadata_lookup_table_entry(w); + /* Write the metadata resource to the output WIM using the proper + * compression type. The lookup table entry for the metadata resource + * is updated. */ ret = write_wim_resource_from_buffer(buf, metadata_original_size, - out, metadata_ctype, - <e->output_resource_entry); - FREE(buf); + w->out_fp, + wimlib_get_compression_type(w), + <e->output_resource_entry, + lte->hash); if (ret != 0) - return ret; + goto out; - lte->out_refcnt++; - lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA; + /* It's very likely the SHA1 message digest of the metadata resource, so + * re-insert the lookup table entry into the lookup table. */ lookup_table_unlink(w->lookup_table, lte); - copy_hash(lte->hash, hash); lookup_table_insert(w->lookup_table, lte); - return 0; + + /* We do not allow a metadata resource to be referenced multiple times, + * and the 20 random bytes appended to it should make it extremely + * likely for each metadata resource to be unique, even if the exact + * same image is captured. */ + wimlib_assert(lte->out_refcnt == 0); + lte->out_refcnt = 1; + + /* Make sure that the resource entry is written marked with the metadata + * flag. */ + lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA; +out: + /* All the data has been written to the new WIM; no need for the buffer + * anymore */ + FREE(buf); + return ret; }