X-Git-Url: https://wimlib.net/git/?a=blobdiff_plain;f=src%2Fdentry.c;h=af27c68776459c2d1272656b42a2b4b9e84039df;hb=d4f8cc3998aebb08c6cd12d248173323b43c07b8;hp=45093069001ea6a94c7df3614690675e6e914cb0;hpb=01c6f902b2a48a33af3f94e29791f85c68df91d0;p=wimlib diff --git a/src/dentry.c b/src/dentry.c index 45093069..af27c687 100644 --- a/src/dentry.c +++ b/src/dentry.c @@ -38,6 +38,7 @@ #include "wimlib/lookup_table.h" #include "wimlib/metadata.h" #include "wimlib/resource.h" +#include "wimlib/security.h" #include "wimlib/sha1.h" #include "wimlib/timestamp.h" @@ -72,18 +73,107 @@ struct wim_ads_entry_on_disk { #define WIM_ADS_ENTRY_DISK_SIZE 38 -/* WIM directory entry (on-disk format) */ +/* On-disk format of a WIM dentry (directory entry), located in the metadata + * resource for a WIM image. */ struct wim_dentry_on_disk { + + /* Length of this directory entry in bytes, not including any alternate + * data stream entries. Should be a multiple of 8 so that the following + * dentry or alternate data stream entry is aligned on an 8-byte + * boundary. (If not, wimlib will round it up.) It must be at least as + * long as the fixed-length fields of the dentry (WIM_DENTRY_DISK_SIZE), + * plus the lengths of the file name and/or short name if present. + * + * It is also possible for this field to be 0. This situation, which is + * undocumented, indicates the end of a list of sibling nodes in a + * directory. It also means the real length is 8, because the dentry + * included only the length field, but that takes up 8 bytes. */ le64 length; + + /* Attributes of the file or directory. This is a bitwise OR of the + * FILE_ATTRIBUTE_* constants and should correspond to the value + * retrieved by GetFileAttributes() on Windows. */ le32 attributes; + + /* A value that specifies the security descriptor for this file or + * directory. If -1, the file or directory has no security descriptor. + * Otherwise, it is a 0-based index into the WIM image's table of + * security descriptors (see: `struct wim_security_data') */ sle32 security_id; + + /* Offset, in bytes, from the start of the uncompressed metadata + * resource of this directory's child directory entries, or 0 if this + * directory entry does not correspond to a directory or otherwise does + * not have any children. */ le64 subdir_offset; + + /* Reserved fields */ le64 unused_1; le64 unused_2; + + + /* Creation time, last access time, and last write time, in + * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601. They + * should correspond to the times gotten by calling GetFileTime() on + * Windows. */ le64 creation_time; le64 last_access_time; le64 last_write_time; + + /* Vaguely, the SHA-1 message digest ("hash") of the file's contents. + * More specifically, this is for the "unnamed data stream" rather than + * any "alternate data streams". This hash value is used to look up the + * corresponding entry in the WIM's stream lookup table to actually find + * the file contents within the WIM. + * + * If the file has no unnamed data stream (e.g. is a directory), then + * this field will be all zeroes. If the unnamed data stream is empty + * (i.e. an "empty file"), then this field is also expected to be all + * zeroes. (It will be if wimlib created the WIM image, at least; + * otherwise it can't be ruled out that the SHA-1 message digest of 0 + * bytes of data is given explicitly.) + * + * If the file has reparse data, then this field will instead specify + * the SHA-1 message digest of the reparse data. If it is somehow + * possible for a file to have both an unnamed data stream and reparse + * data, then this is not handled by wimlib. + * + * As a further special case, if this field is all zeroes but there is + * an alternate data stream entry with no name and a nonzero SHA-1 + * message digest field, then that hash must be used instead of this + * one. (wimlib does not use this quirk on WIM images it creates.) + */ u8 unnamed_stream_hash[SHA1_HASH_SIZE]; + + /* The format of the following data is not yet completely known and they + * do not correspond to Microsoft's documentation. + * + * If this directory entry is for a reparse point (has + * FILE_ATTRIBUTE_REPARSE_POINT set in the attributes field), then the + * version of the following fields containing the reparse tag is valid. + * Furthermore, the field notated as not_rpfixed, as far as I can tell, + * is supposed to be set to 1 if reparse point fixups (a.k.a. fixing the + * targets of absolute symbolic links) were *not* done, and otherwise 0. + * + * If this directory entry is not for a reparse point, then the version + * of the following fields containing the hard_link_group_id is valid. + * All MS says about this field is that "If this file is part of a hard + * link set, all the directory entries in the set will share the same + * value in this field.". However, more specifically I have observed + * the following: + * - If the file is part of a hard link set of size 1, then the + * hard_link_group_id should be set to either 0, which is treated + * specially as indicating "not hardlinked", or any unique value. + * - The specific nonzero values used to identity hard link sets do + * not matter, as long as they are unique. + * - However, due to bugs in Microsoft's software, it is actually NOT + * guaranteed that directory entries that share the same hard link + * group ID are actually hard linked to each either. We have to + * handle this by using special code to use distinguishing features + * (which is possible because some information about the underlying + * inode is repeated in each dentry) to split up these fake hard link + * groups into what they actually are supposed to be. + */ union { struct { le32 rp_unknown_1; @@ -96,14 +186,28 @@ struct wim_dentry_on_disk { le64 hard_link_group_id; } _packed_attribute nonreparse; }; + + /* Number of alternate data stream entries that directly follow this + * dentry on-disk. */ le16 num_alternate_data_streams; + + /* Length of this file's UTF-16LE encoded short name (8.3 DOS-compatible + * name), if present, in bytes, excluding the null terminator. If this + * file has no short name, then this field should be 0. */ le16 short_name_nbytes; + + /* Length of this file's UTF-16LE encoded "long" name, excluding the + * null terminator. If this file has no short name, then this field + * should be 0. It's expected that only the root dentry has this field + * set to 0. */ le16 file_name_nbytes; - /* Follewed by variable length file name, if file_name_nbytes != 0 */ - utf16lechar file_name[]; + /* Followed by variable length file name, in UTF16-LE, if + * file_name_nbytes != 0. Includes null terminator. */ + /*utf16lechar file_name[];*/ - /* Followed by variable length short name, if short_name_nbytes != 0 */ + /* Followed by variable length short name, in UTF16-LE, if + * short_name_nbytes != 0. Includes null terminator. */ /*utf16lechar short_name[];*/ } _packed_attribute; @@ -687,8 +791,12 @@ get_dentry_utf16le(WIMStruct *wim, const utf16lechar *path) return cur_dentry; } -/* Returns the dentry corresponding to the @path, or NULL if there is no such - * dentry. */ +/* + * Returns the dentry in the currently selected WIM image named by @path + * starting from the root of the WIM image, or NULL if there is no such dentry. + * + * On Windows, the search is done case-insensitively. + */ struct wim_dentry * get_dentry(WIMStruct *wim, const tchar *path) { @@ -880,13 +988,6 @@ new_timeless_inode(void) inode->i_next_stream_id = 1; inode->i_not_rpfixed = 1; INIT_LIST_HEAD(&inode->i_list); - #ifdef WITH_FUSE - if (pthread_mutex_init(&inode->i_mutex, NULL) != 0) { - ERROR_WITH_ERRNO("Error initializing mutex"); - FREE(inode); - return NULL; - } - #endif INIT_LIST_HEAD(&inode->i_dentry); } return inode; @@ -985,6 +1086,19 @@ new_filler_directory(const tchar *name, struct wim_dentry **dentry_ret) return 0; } +static int +dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore) +{ + dentry->d_inode->i_visited = 0; + return 0; +} + +void +dentry_tree_clear_inode_visited(struct wim_dentry *root) +{ + for_dentry_in_tree(root, dentry_clear_inode_visited, NULL); +} + static int init_ads_entry(struct wim_ads_entry *ads_entry, const void *name, size_t name_nbytes, bool is_utf16le) @@ -1025,15 +1139,10 @@ free_inode(struct wim_inode *inode) destroy_ads_entry(&inode->i_ads_entries[i]); FREE(inode->i_ads_entries); } - #ifdef WITH_FUSE - wimlib_assert(inode->i_num_opened_fds == 0); - FREE(inode->i_fds); - pthread_mutex_destroy(&inode->i_mutex); - #endif /* HACK: This may instead delete the inode from i_list, but the * hlist_del() behaves the same as list_del(). */ - hlist_del(&inode->i_hlist); - FREE(inode->i_extracted_file); + if (!hlist_unhashed(&inode->i_hlist)) + hlist_del(&inode->i_hlist); FREE(inode); } } @@ -1096,11 +1205,13 @@ do_free_dentry(struct wim_dentry *dentry, void *_lookup_table) /* * Unlinks and frees a dentry tree. * - * @root: The root of the tree. - * @lookup_table: The lookup table for dentries. If non-NULL, the - * reference counts in the lookup table for the lookup - * table entries corresponding to the dentries will be - * decremented. + * @root: + * The root of the tree. + * + * @lookup_table: + * The lookup table for dentries. If non-NULL, the reference counts in the + * lookup table for the lookup table entries corresponding to the dentries + * will be decremented. */ void free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table) @@ -1348,7 +1459,7 @@ add_stream_from_data_buffer(const void *buffer, size_t size, struct wim_lookup_table_entry *lte, *existing_lte; sha1_buffer(buffer, size, hash); - existing_lte = __lookup_resource(lookup_table, hash); + existing_lte = lookup_resource(lookup_table, hash); if (existing_lte) { wimlib_assert(wim_resource_size(existing_lte) == size); lte = existing_lte; @@ -1395,6 +1506,15 @@ inode_add_ads_with_data(struct wim_inode *inode, const tchar *name, return 0; } +bool +inode_has_named_stream(const struct wim_inode *inode) +{ + for (u16 i = 0; i < inode->i_num_ads; i++) + if (ads_entry_is_named_stream(&inode->i_ads_entries[i])) + return true; + return false; +} + /* Set the unnamed stream of a WIM inode, given a data buffer containing the * stream contents. */ int @@ -1435,6 +1555,15 @@ inode_remove_ads(struct wim_inode *inode, u16 idx, inode->i_num_ads--; } +bool +inode_has_unix_data(const struct wim_inode *inode) +{ + for (u16 i = 0; i < inode->i_num_ads; i++) + if (ads_entry_is_unix_data(&inode->i_ads_entries[i])) + return true; + return false; +} + #ifndef __WIN32__ int inode_get_unix_data(const struct wim_inode *inode, @@ -1510,19 +1639,24 @@ inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode, /* * Reads the alternate data stream entries of a WIM dentry. * - * @p: Pointer to buffer that starts with the first alternate stream entry. + * @p: + * Pointer to buffer that starts with the first alternate stream entry. * - * @inode: Inode to load the alternate data streams into. - * @inode->i_num_ads must have been set to the number of - * alternate data streams that are expected. + * @inode: + * Inode to load the alternate data streams into. @inode->i_num_ads must + * have been set to the number of alternate data streams that are expected. * - * @remaining_size: Number of bytes of data remaining in the buffer pointed - * to by @p. + * @remaining_size: + * Number of bytes of data remaining in the buffer pointed to by @p. * + * On success, inode->i_ads_entries is set to an array of `struct + * wim_ads_entry's of length inode->i_num_ads. On failure, @inode is not + * modified. * - * Return 0 on success or nonzero on failure. On success, inode->i_ads_entries - * is set to an array of `struct wim_ads_entry's of length inode->i_num_ads. On - * failure, @inode is not modified. + * Return values: + * WIMLIB_ERR_SUCCESS (0) + * WIMLIB_ERR_INVALID_METADATA_RESOURCE + * WIMLIB_ERR_NOMEM */ static int read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode, @@ -1626,7 +1760,7 @@ out_of_memory: goto out_free_ads_entries; out_invalid: ERROR("An alternate data stream entry is invalid"); - ret = WIMLIB_ERR_INVALID_DENTRY; + ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE; out_free_ads_entries: if (ads_entries) { for (u16 i = 0; i < num_ads; i++) @@ -1647,7 +1781,7 @@ out: * @metadata_resource_len: * Length of the metadata resource buffer, in bytes. * - * @offset: Offset of the dentry within the metadata resource. + * @offset: Offset of the dentry within the metadata resource. * * @dentry: A `struct wim_dentry' that will be filled in by this function. * @@ -1657,9 +1791,10 @@ out: * this was a special "end of directory" dentry and not a real dentry. If * nonzero, this was a real dentry. * - * Possible errors include: - * WIMLIB_ERR_NOMEM - * WIMLIB_ERR_INVALID_DENTRY + * Return values: + * WIMLIB_ERR_SUCCESS (0) + * WIMLIB_ERR_INVALID_METADATA_RESOURCE + * WIMLIB_ERR_NOMEM */ int read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, @@ -1693,7 +1828,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, ERROR("Directory entry starting at %"PRIu64" ends past the " "end of the metadata resource (size %"PRIu64")", offset, metadata_resource_len); - return WIMLIB_ERR_INVALID_DENTRY; + return WIMLIB_ERR_INVALID_METADATA_RESOURCE; } dentry->length = le64_to_cpu(disk_dentry->length); @@ -1715,7 +1850,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, "%"PRIu64" ends past the end of the metadata resource " "(size %"PRIu64")", offset, dentry->length, metadata_resource_len); - return WIMLIB_ERR_INVALID_DENTRY; + return WIMLIB_ERR_INVALID_METADATA_RESOURCE; } /* Make sure the dentry length is at least as large as the number of @@ -1723,7 +1858,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, if (dentry->length < sizeof(struct wim_dentry_on_disk)) { ERROR("Directory entry has invalid length of %"PRIu64" bytes", dentry->length); - return WIMLIB_ERR_INVALID_DENTRY; + return WIMLIB_ERR_INVALID_METADATA_RESOURCE; } /* Allocate a `struct wim_inode' for this `struct wim_dentry'. */ @@ -1769,7 +1904,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, if ((short_name_nbytes & 1) | (file_name_nbytes & 1)) { ERROR("Dentry name is not valid UTF-16LE (odd number of bytes)!"); - ret = WIMLIB_ERR_INVALID_DENTRY; + ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE; goto out_free_inode; } @@ -1786,7 +1921,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, ERROR("Unexpected end of directory entry! (Expected " "at least %"PRIu64" bytes, got %"PRIu64" bytes.)", calculated_size, dentry->length); - ret = WIMLIB_ERR_INVALID_DENTRY; + ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE; goto out_free_inode; } @@ -1840,7 +1975,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, * be included in the dentry->length field for some reason. */ if (inode->i_num_ads != 0) { - ret = WIMLIB_ERR_INVALID_DENTRY; + ret = WIMLIB_ERR_INVALID_METADATA_RESOURCE; if (offset + dentry->length > metadata_resource_len || (ret = read_ads_entries(&metadata_resource[offset + dentry->length], inode, @@ -1885,26 +2020,32 @@ dentry_get_file_type_string(const struct wim_dentry *dentry) /* Reads the children of a dentry, and all their children, ..., etc. from the * metadata resource and into the dentry tree. * - * @metadata_resource: An array that contains the uncompressed metadata - * resource for the WIM file. + * @metadata_resource: + * An array that contains the uncompressed metadata resource for the WIM + * file. * - * @metadata_resource_len: The length of the uncompressed metadata resource, in - * bytes. + * @metadata_resource_len: + * The length of the uncompressed metadata resource, in bytes. * - * @dentry: A pointer to a `struct wim_dentry' that is the root of the directory - * tree and has already been read from the metadata resource. It - * does not need to be the real root because this procedure is - * called recursively. + * @dentry: + * A pointer to a `struct wim_dentry' that is the root of the directory + * tree and has already been read from the metadata resource. It does not + * need to be the real root because this procedure is called recursively. * - * Returns zero on success; nonzero on failure. + * Return values: + * WIMLIB_ERR_SUCCESS (0) + * WIMLIB_ERR_INVALID_METADATA_RESOURCE + * WIMLIB_ERR_NOMEM */ int -read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, - struct wim_dentry *dentry) +read_dentry_tree(const u8 * restrict metadata_resource, + u64 metadata_resource_len, + struct wim_dentry * restrict dentry) { u64 cur_offset = dentry->subdir_offset; struct wim_dentry *child; struct wim_dentry *duplicate; + struct wim_dentry *parent; struct wim_dentry cur_child; int ret; @@ -1917,6 +2058,18 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, if (cur_offset == 0) return 0; + /* Check for cyclic directory structure */ + for (parent = dentry->parent; !dentry_is_root(parent); parent = parent->parent) + { + if (unlikely(parent->subdir_offset == cur_offset)) { + ERROR("Cyclic directory structure directed: children " + "of \"%"TS"\" coincide with children of \"%"TS"\"", + dentry_full_path(dentry), + dentry_full_path(parent)); + return WIMLIB_ERR_INVALID_METADATA_RESOURCE; + } + } + /* Find and read all the children of @dentry. */ for (;;) { @@ -1946,8 +2099,16 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, * entries. */ cur_offset += dentry_total_length(child); + if (unlikely(!dentry_has_long_name(child))) { + WARNING("Ignoring unnamed dentry in " + "directory \"%"TS"\"", + dentry_full_path(dentry)); + free_dentry(child); + continue; + } + duplicate = dentry_add_child(dentry, child); - if (duplicate) { + if (unlikely(duplicate)) { const tchar *child_type, *duplicate_type; child_type = dentry_get_file_type_string(child); duplicate_type = dentry_get_file_type_string(duplicate); @@ -1956,23 +2117,24 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, "at that path with the exact same name)", child_type, dentry_full_path(duplicate), duplicate_type); - } else { - inode_add_dentry(child, child->d_inode); - /* If there are children of this child, call this - * procedure recursively. */ - if (child->subdir_offset != 0) { - if (dentry_is_directory(child)) { - ret = read_dentry_tree(metadata_resource, - metadata_resource_len, - child); - if (ret) - break; - } else { - WARNING("Ignoring children of non-directory \"%"TS"\"", - dentry_full_path(child)); - } - } + free_dentry(child); + continue; + } + inode_add_dentry(child, child->d_inode); + /* If there are children of this child, call this + * procedure recursively. */ + if (child->subdir_offset != 0) { + if (likely(dentry_is_directory(child))) { + ret = read_dentry_tree(metadata_resource, + metadata_resource_len, + child); + if (ret) + break; + } else { + WARNING("Ignoring children of non-directory \"%"TS"\"", + dentry_full_path(child)); + } } } return ret; @@ -1998,7 +2160,7 @@ write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p) wimlib_assert(((uintptr_t)p & 7) == 0); /* 8 byte aligned */ orig_p = p; - inode = dentry->d_inode; + inode = dentry->d_inode; disk_dentry = (struct wim_dentry_on_disk*)p; disk_dentry->attributes = cpu_to_le32(inode->i_attributes); @@ -2026,6 +2188,8 @@ write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p) disk_dentry->file_name_nbytes = cpu_to_le16(dentry->file_name_nbytes); p += sizeof(struct wim_dentry_on_disk); + wimlib_assert(dentry_is_root(dentry) != dentry_has_long_name(dentry)); + if (dentry_has_long_name(dentry)) p = mempcpy(p, dentry->file_name, dentry->file_name_nbytes + 2); @@ -2122,7 +2286,7 @@ write_dentry_tree_recursive(const struct wim_dentry *parent, u8 *p) * Returns pointer to the byte after the last byte we wrote. */ u8 * -write_dentry_tree(const struct wim_dentry *root, u8 *p) +write_dentry_tree(const struct wim_dentry * restrict root, u8 * restrict p) { DEBUG("Writing dentry tree."); wimlib_assert(dentry_is_root(root)); @@ -2139,3 +2303,422 @@ write_dentry_tree(const struct wim_dentry *root, u8 *p) /* Recursively write the rest of the dentry tree. */ return write_dentry_tree_recursive(root, p); } + + +static int +init_wimlib_dentry(struct wimlib_dir_entry *wdentry, + struct wim_dentry *dentry, + const WIMStruct *wim, + int flags) +{ + int ret; + size_t dummy; + const struct wim_inode *inode = dentry->d_inode; + struct wim_lookup_table_entry *lte; + const u8 *hash; + +#if TCHAR_IS_UTF16LE + wdentry->filename = dentry->file_name; + wdentry->dos_name = dentry->short_name; +#else + if (dentry_has_long_name(dentry)) { + ret = utf16le_to_tstr(dentry->file_name, + dentry->file_name_nbytes, + (tchar**)&wdentry->filename, + &dummy); + if (ret) + return ret; + } + if (dentry_has_short_name(dentry)) { + ret = utf16le_to_tstr(dentry->short_name, + dentry->short_name_nbytes, + (tchar**)&wdentry->dos_name, + &dummy); + if (ret) + return ret; + } +#endif + ret = calculate_dentry_full_path(dentry); + if (ret) + return ret; + wdentry->full_path = dentry->_full_path; + + for (struct wim_dentry *d = dentry; !dentry_is_root(d); d = d->parent) + wdentry->depth++; + + if (inode->i_security_id >= 0) { + const struct wim_security_data *sd = wim_const_security_data(wim); + wdentry->security_descriptor = sd->descriptors[inode->i_security_id]; + wdentry->security_descriptor_size = sd->sizes[inode->i_security_id]; + } + wdentry->reparse_tag = inode->i_reparse_tag; + wdentry->num_links = inode->i_nlink; + wdentry->attributes = inode->i_attributes; + wdentry->hard_link_group_id = inode->i_ino; + wdentry->creation_time = wim_timestamp_to_timespec(inode->i_creation_time); + wdentry->last_write_time = wim_timestamp_to_timespec(inode->i_last_write_time); + wdentry->last_access_time = wim_timestamp_to_timespec(inode->i_last_access_time); + + lte = inode_unnamed_lte(inode, wim->lookup_table); + if (lte) { + lte_to_wimlib_resource_entry(lte, &wdentry->streams[0].resource); + } else if (!is_zero_hash(hash = inode_unnamed_stream_hash(inode))) { + if (flags & WIMLIB_ITERATE_DIR_TREE_FLAG_RESOURCES_NEEDED) + return resource_not_found_error(inode, hash); + copy_hash(wdentry->streams[0].resource.sha1_hash, hash); + wdentry->streams[0].resource.is_missing = 1; + } + + for (unsigned i = 0; i < inode->i_num_ads; i++) { + if (!ads_entry_is_named_stream(&inode->i_ads_entries[i])) + continue; + lte = inode_stream_lte(inode, i + 1, wim->lookup_table); + wdentry->num_named_streams++; + if (lte) { + lte_to_wimlib_resource_entry(lte, &wdentry->streams[ + wdentry->num_named_streams].resource); + } else if (!is_zero_hash(hash = inode_stream_hash(inode, i + 1))) { + if (flags & WIMLIB_ITERATE_DIR_TREE_FLAG_RESOURCES_NEEDED) + return resource_not_found_error(inode, hash); + copy_hash(wdentry->streams[ + wdentry->num_named_streams].resource.sha1_hash, hash); + wdentry->streams[ + wdentry->num_named_streams].resource.is_missing = 1; + } + #if TCHAR_IS_UTF16LE + wdentry->streams[wdentry->num_named_streams].stream_name = + inode->i_ads_entries[i].stream_name; + #else + size_t dummy; + + ret = utf16le_to_tstr(inode->i_ads_entries[i].stream_name, + inode->i_ads_entries[i].stream_name_nbytes, + (tchar**)&wdentry->streams[ + wdentry->num_named_streams].stream_name, + &dummy); + if (ret) + return ret; + #endif + } + return 0; +} + +static void +free_wimlib_dentry(struct wimlib_dir_entry *wdentry) +{ +#if !TCHAR_IS_UTF16LE + FREE((tchar*)wdentry->filename); + FREE((tchar*)wdentry->dos_name); + for (unsigned i = 1; i <= wdentry->num_named_streams; i++) + FREE((tchar*)wdentry->streams[i].stream_name); +#endif + FREE(wdentry); +} + +struct iterate_dir_tree_ctx { + WIMStruct *wim; + int flags; + wimlib_iterate_dir_tree_callback_t cb; + void *user_ctx; +}; + +static int +do_iterate_dir_tree(WIMStruct *wim, + struct wim_dentry *dentry, int flags, + wimlib_iterate_dir_tree_callback_t cb, + void *user_ctx); + +static int +call_do_iterate_dir_tree(struct wim_dentry *dentry, void *_ctx) +{ + struct iterate_dir_tree_ctx *ctx = _ctx; + return do_iterate_dir_tree(ctx->wim, dentry, ctx->flags, + ctx->cb, ctx->user_ctx); +} + +static int +do_iterate_dir_tree(WIMStruct *wim, + struct wim_dentry *dentry, int flags, + wimlib_iterate_dir_tree_callback_t cb, + void *user_ctx) +{ + struct wimlib_dir_entry *wdentry; + int ret = WIMLIB_ERR_NOMEM; + + + wdentry = CALLOC(1, sizeof(struct wimlib_dir_entry) + + (1 + dentry->d_inode->i_num_ads) * + sizeof(struct wimlib_stream_entry)); + if (!wdentry) + goto out; + + ret = init_wimlib_dentry(wdentry, dentry, wim, flags); + if (ret) + goto out_free_wimlib_dentry; + + if (!(flags & WIMLIB_ITERATE_DIR_TREE_FLAG_CHILDREN)) { + ret = (*cb)(wdentry, user_ctx); + if (ret) + goto out_free_wimlib_dentry; + } + + if (flags & (WIMLIB_ITERATE_DIR_TREE_FLAG_RECURSIVE | + WIMLIB_ITERATE_DIR_TREE_FLAG_CHILDREN)) + { + struct iterate_dir_tree_ctx ctx = { + .wim = wim, + .flags = flags &= ~WIMLIB_ITERATE_DIR_TREE_FLAG_CHILDREN, + .cb = cb, + .user_ctx = user_ctx, + }; + ret = for_dentry_child(dentry, call_do_iterate_dir_tree, &ctx); + } +out_free_wimlib_dentry: + free_wimlib_dentry(wdentry); +out: + return ret; +} + +struct image_iterate_dir_tree_ctx { + const tchar *path; + int flags; + wimlib_iterate_dir_tree_callback_t cb; + void *user_ctx; +}; + + +static int +image_do_iterate_dir_tree(WIMStruct *wim) +{ + struct image_iterate_dir_tree_ctx *ctx = wim->private; + struct wim_dentry *dentry; + + dentry = get_dentry(wim, ctx->path); + if (!dentry) + return WIMLIB_ERR_PATH_DOES_NOT_EXIST; + return do_iterate_dir_tree(wim, dentry, ctx->flags, ctx->cb, ctx->user_ctx); +} + +/* API function documented in wimlib.h */ +WIMLIBAPI int +wimlib_iterate_dir_tree(WIMStruct *wim, int image, const tchar *path, + int flags, + wimlib_iterate_dir_tree_callback_t cb, void *user_ctx) +{ + struct image_iterate_dir_tree_ctx ctx = { + .path = path, + .flags = flags, + .cb = cb, + .user_ctx = user_ctx, + }; + wim->private = &ctx; + return for_image(wim, image, image_do_iterate_dir_tree); +} + +/* Returns %true iff the metadata of @inode and @template_inode are reasonably + * consistent with them being the same, unmodified file. */ +static bool +inode_metadata_consistent(const struct wim_inode *inode, + const struct wim_inode *template_inode, + const struct wim_lookup_table *template_lookup_table) +{ + /* Must have exact same creation time and last write time. */ + if (inode->i_creation_time != template_inode->i_creation_time || + inode->i_last_write_time != template_inode->i_last_write_time) + return false; + + /* Last access time may have stayed the same or increased, but certainly + * shouldn't have decreased. */ + if (inode->i_last_access_time < template_inode->i_last_access_time) + return false; + + /* Must have same number of alternate data stream entries. */ + if (inode->i_num_ads != template_inode->i_num_ads) + return false; + + /* If the stream entries for the inode are for some reason not resolved, + * then the hashes are already available and the point of this function + * is defeated. */ + if (!inode->i_resolved) + return false; + + /* Iterate through each stream and do some more checks. */ + for (unsigned i = 0; i <= inode->i_num_ads; i++) { + const struct wim_lookup_table_entry *lte, *template_lte; + + lte = inode_stream_lte_resolved(inode, i); + template_lte = inode_stream_lte(template_inode, i, + template_lookup_table); + + /* Compare stream sizes. */ + if (lte && template_lte) { + if (wim_resource_size(lte) != wim_resource_size(template_lte)) + return false; + + /* If hash happens to be available, compare with template. */ + if (!lte->unhashed && !template_lte->unhashed && + !hashes_equal(lte->hash, template_lte->hash)) + return false; + + } else if (lte && wim_resource_size(lte)) { + return false; + } else if (template_lte && wim_resource_size(template_lte)) { + return false; + } + } + + /* All right, barring a full checksum and given that the inodes share a + * path and the user isn't trying to trick us, these inodes most likely + * refer to the same file. */ + return true; +} + +/** + * Given an inode @inode that has been determined to be "the same" as another + * inode @template_inode in either the same WIM or another WIM, retrieve some + * useful stream information (e.g. checksums) from @template_inode. + * + * This assumes that the streams for @inode have been resolved (to point + * directly to the appropriate `struct wim_lookup_table_entry's) but do not + * necessarily have checksum information filled in. + */ +static int +inode_copy_checksums(struct wim_inode *inode, + struct wim_inode *template_inode, + WIMStruct *wim, + WIMStruct *template_wim) +{ + for (unsigned i = 0; i <= inode->i_num_ads; i++) { + struct wim_lookup_table_entry *lte, *template_lte; + struct wim_lookup_table_entry *replace_lte; + + lte = inode_stream_lte_resolved(inode, i); + template_lte = inode_stream_lte(template_inode, i, + template_wim->lookup_table); + + /* Only take action if both entries exist, the entry for @inode + * has no checksum calculated, but the entry for @template_inode + * does. */ + if (!lte || !template_lte || + !lte->unhashed || template_lte->unhashed) + continue; + + wimlib_assert(lte->refcnt == inode->i_nlink); + + /* If the WIM of the template image is the same as the WIM of + * the new image, then @template_lte can be used directly. + * + * Otherwise, look for a stream with the same hash in the WIM of + * the new image. If found, use it; otherwise re-use the entry + * being discarded, filling in the hash. */ + + if (wim == template_wim) + replace_lte = template_lte; + else + replace_lte = lookup_resource(wim->lookup_table, + template_lte->hash); + + list_del(<e->unhashed_list); + if (replace_lte) { + free_lookup_table_entry(lte); + } else { + copy_hash(lte->hash, template_lte->hash); + lte->unhashed = 0; + lookup_table_insert(wim->lookup_table, lte); + lte->refcnt = 0; + replace_lte = lte; + } + + if (i == 0) + inode->i_lte = replace_lte; + else + inode->i_ads_entries[i - 1].lte = replace_lte; + + replace_lte->refcnt += inode->i_nlink; + } + return 0; +} + +struct reference_template_args { + WIMStruct *wim; + WIMStruct *template_wim; +}; + +static int +dentry_reference_template(struct wim_dentry *dentry, void *_args) +{ + int ret; + struct wim_dentry *template_dentry; + struct wim_inode *inode, *template_inode; + struct reference_template_args *args = _args; + WIMStruct *wim = args->wim; + WIMStruct *template_wim = args->template_wim; + + if (dentry->d_inode->i_visited) + return 0; + + ret = calculate_dentry_full_path(dentry); + if (ret) + return ret; + + template_dentry = get_dentry(template_wim, dentry->_full_path); + if (!template_dentry) { + DEBUG("\"%"TS"\": newly added file", dentry->_full_path); + return 0; + } + + inode = dentry->d_inode; + template_inode = template_dentry->d_inode; + + if (inode_metadata_consistent(inode, template_inode, + template_wim->lookup_table)) { + /*DEBUG("\"%"TS"\": No change detected", dentry->_full_path);*/ + ret = inode_copy_checksums(inode, template_inode, + wim, template_wim); + inode->i_visited = 1; + } else { + DEBUG("\"%"TS"\": change detected!", dentry->_full_path); + ret = 0; + } + return ret; +} + +/* API function documented in wimlib.h */ +WIMLIBAPI int +wimlib_reference_template_image(WIMStruct *wim, int new_image, + WIMStruct *template_wim, int template_image, + int flags, wimlib_progress_func_t progress_func) +{ + int ret; + struct wim_image_metadata *new_imd; + + if (wim == NULL || template_wim == NULL) + return WIMLIB_ERR_INVALID_PARAM; + + if (wim == template_wim && new_image == template_image) + return WIMLIB_ERR_INVALID_PARAM; + + if (new_image < 1 || new_image > wim->hdr.image_count) + return WIMLIB_ERR_INVALID_IMAGE; + + if (!wim_has_metadata(wim)) + return WIMLIB_ERR_METADATA_NOT_FOUND; + + new_imd = wim->image_metadata[new_image - 1]; + if (!new_imd->modified) + return WIMLIB_ERR_INVALID_PARAM; + + ret = select_wim_image(template_wim, template_image); + if (ret) + return ret; + + struct reference_template_args args = { + .wim = wim, + .template_wim = template_wim, + }; + + ret = for_dentry_in_tree(new_imd->root_dentry, + dentry_reference_template, &args); + dentry_tree_clear_inode_visited(new_imd->root_dentry); + return ret; +}