X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fdentry.c;h=e87bfefc0b46046ecdef6622e24e9f7f5933e191;hp=04ed1092900889fc2917d0ffea28c226c55221c2;hb=d7aa64b64fc9836418293781b2b093f090414b6e;hpb=74faf510afbc0c3130035a7002cf5bc0d6d8fdcc diff --git a/src/dentry.c b/src/dentry.c index 04ed1092..e87bfefc 100644 --- a/src/dentry.c +++ b/src/dentry.c @@ -37,7 +37,9 @@ #include "wimlib/error.h" #include "wimlib/lookup_table.h" #include "wimlib/metadata.h" +#include "wimlib/paths.h" #include "wimlib/resource.h" +#include "wimlib/security.h" #include "wimlib/sha1.h" #include "wimlib/timestamp.h" @@ -72,18 +74,109 @@ 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. In fact, when named data streams are present, some versions of + * Windows PE contain a bug where they only look in the alternate data + * stream entries for the unnamed data stream, not here. + */ 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 +189,28 @@ 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; @@ -113,7 +220,7 @@ struct wim_dentry_on_disk { * a file name and short name that take the specified numbers of bytes. This * excludes any alternate data stream entries that may follow the dentry. */ static u64 -_dentry_correct_length_unaligned(u16 file_name_nbytes, u16 short_name_nbytes) +dentry_correct_length_unaligned(u16 file_name_nbytes, u16 short_name_nbytes) { u64 length = sizeof(struct wim_dentry_on_disk); if (file_name_nbytes) @@ -128,10 +235,13 @@ _dentry_correct_length_unaligned(u16 file_name_nbytes, u16 short_name_nbytes) * ignored; also, this excludes any alternate data stream entries that may * follow the dentry. */ static u64 -dentry_correct_length_unaligned(const struct wim_dentry *dentry) +dentry_correct_length_aligned(const struct wim_dentry *dentry) { - return _dentry_correct_length_unaligned(dentry->file_name_nbytes, - dentry->short_name_nbytes); + u64 len; + + len = dentry_correct_length_unaligned(dentry->file_name_nbytes, + dentry->short_name_nbytes); + return (len + 7) & ~7; } /* Duplicates a string of system-dependent encoding into a UTF-16LE string and @@ -147,7 +257,7 @@ get_utf16le_name(const tchar *name, utf16lechar **name_utf16le_ret, #if TCHAR_IS_UTF16LE name_utf16le_nbytes = tstrlen(name) * sizeof(utf16lechar); name_utf16le = MALLOC(name_utf16le_nbytes + sizeof(utf16lechar)); - if (!name_utf16le) + if (name_utf16le == NULL) return WIMLIB_ERR_NOMEM; memcpy(name_utf16le, name, name_utf16le_nbytes + sizeof(utf16lechar)); ret = 0; @@ -201,31 +311,58 @@ ads_entry_total_length(const struct wim_ads_entry *entry) return (len + 7) & ~7; } +/* + * Determine whether to include a "dummy" stream when writing a WIM dentry: + * + * Some versions of Microsoft's WIM software (the boot driver(s) in WinPE 3.0, + * for example) contain a bug where they assume the first alternate data stream + * (ADS) entry of a dentry with a nonzero ADS count specifies the unnamed + * stream, even if it has a name and the unnamed stream is already specified in + * the hash field of the dentry itself. + * + * wimlib has to work around this behavior by carefully emulating the behavior + * of (most versions of) ImageX/WIMGAPI, which move the unnamed stream reference + * into the alternate stream entries whenever there are named data streams, even + * though there is already a field in the dentry itself for the unnamed stream + * reference, which then goes to waste. + */ +static inline bool inode_needs_dummy_stream(const struct wim_inode *inode) +{ + return (inode->i_num_ads > 0 && + inode->i_num_ads < 0xffff && /* overflow check */ + inode->i_canonical_streams); /* assume the dentry is okay if it + already had an unnamed ADS entry + when it was read in */ +} -static u64 -_dentry_total_length(const struct wim_dentry *dentry, u64 length) +/* Calculate the total number of bytes that will be consumed when a WIM dentry + * is written. This includes base dentry and name fields as well as all + * alternate data stream entries and alignment bytes. */ +u64 +dentry_out_total_length(const struct wim_dentry *dentry) { + u64 length = dentry_correct_length_aligned(dentry); const struct wim_inode *inode = dentry->d_inode; + + if (inode_needs_dummy_stream(inode)) + length += ads_entry_total_length(&(struct wim_ads_entry){}); + for (u16 i = 0; i < inode->i_num_ads; i++) length += ads_entry_total_length(&inode->i_ads_entries[i]); - return (length + 7) & ~7; -} -/* Calculate the aligned *total* length of an on-disk WIM dentry. This includes - * all alternate data streams. */ -u64 -dentry_correct_total_length(const struct wim_dentry *dentry) -{ - return _dentry_total_length(dentry, - dentry_correct_length_unaligned(dentry)); + return length; } -/* Like dentry_correct_total_length(), but use the existing dentry->length field - * instead of calculating its "correct" value. */ +/* Calculate the aligned, total length of a dentry, including all alternate data + * stream entries. Uses dentry->length. */ static u64 -dentry_total_length(const struct wim_dentry *dentry) +dentry_in_total_length(const struct wim_dentry *dentry) { - return _dentry_total_length(dentry, dentry->length); + u64 length = dentry->length; + const struct wim_inode *inode = dentry->d_inode; + for (u16 i = 0; i < inode->i_num_ads; i++) + length += ads_entry_total_length(&inode->i_ads_entries[i]); + return (length + 7) & ~7; } int @@ -299,6 +436,24 @@ for_dentry_tree_in_rbtree(struct rb_node *node, return 0; } +/* + * Iterate over all children of @dentry, calling the function @visitor, passing + * it a child dentry and the extra argument @arg. + * + * Note: this function iterates over ALL child dentries, even those with the + * same case-insensitive name. + * + * Note: this function clobbers the tmp_list field of the child dentries. */ +int +for_dentry_child(const struct wim_dentry *dentry, + int (*visitor)(struct wim_dentry *, void *), + void *arg) +{ + return for_dentry_in_rbtree(dentry->d_inode->i_children.rb_node, + visitor, + arg); +} + /* Calls a function on all directory entries in a WIM dentry tree. Logically, * this is a pre-order traversal (the function is called on a parent dentry * before its children), but sibling dentries will be visited in order as well. @@ -309,7 +464,7 @@ for_dentry_in_tree(struct wim_dentry *root, { int ret; - if (!root) + if (root == NULL) return 0; ret = (*visitor)(root, arg); if (ret) @@ -327,7 +482,7 @@ for_dentry_in_tree_depth(struct wim_dentry *root, { int ret; - if (!root) + if (root == NULL) return 0; ret = for_dentry_tree_in_rbtree_depth(root->d_inode->i_children.rb_node, visitor, arg); @@ -338,7 +493,7 @@ for_dentry_in_tree_depth(struct wim_dentry *root, /* Calculate the full path of @dentry. The full path of its parent must have * already been calculated, or it must be the root dentry. */ -static int +int calculate_dentry_full_path(struct wim_dentry *dentry) { tchar *full_path; @@ -349,8 +504,9 @@ calculate_dentry_full_path(struct wim_dentry *dentry) return 0; if (dentry_is_root(dentry)) { - full_path = TSTRDUP(T("/")); - if (!full_path) + static const tchar _root_path[] = {WIM_PATH_SEPARATOR, T('\0')}; + full_path = TSTRDUP(_root_path); + if (full_path == NULL) return WIMLIB_ERR_NOMEM; full_path_nbytes = 1 * sizeof(tchar); } else { @@ -364,7 +520,7 @@ calculate_dentry_full_path(struct wim_dentry *dentry) parent_full_path = T(""); parent_full_path_nbytes = 0; } else { - if (!parent->_full_path) { + if (parent->_full_path == NULL) { ret = calculate_dentry_full_path(parent); if (ret) return ret; @@ -390,10 +546,10 @@ calculate_dentry_full_path(struct wim_dentry *dentry) full_path_nbytes = parent_full_path_nbytes + sizeof(tchar) + filename_nbytes; full_path = MALLOC(full_path_nbytes + sizeof(tchar)); - if (!full_path) + if (full_path == NULL) return WIMLIB_ERR_NOMEM; memcpy(full_path, parent_full_path, parent_full_path_nbytes); - full_path[parent_full_path_nbytes / sizeof(tchar)] = T('/'); + full_path[parent_full_path_nbytes / sizeof(tchar)] = WIM_PATH_SEPARATOR; #if TCHAR_IS_UTF16LE memcpy(&full_path[parent_full_path_nbytes / sizeof(tchar) + 1], dentry->file_name, @@ -432,7 +588,7 @@ dentry_full_path(struct wim_dentry *dentry) static int increment_subdir_offset(struct wim_dentry *dentry, void *subdir_offset_p) { - *(u64*)subdir_offset_p += dentry_correct_total_length(dentry); + *(u64*)subdir_offset_p += dentry_out_total_length(dentry); return 0; } @@ -479,114 +635,134 @@ calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p) } } -/* UNIX: Case-sensitive UTF-16LE dentry or stream name comparison. We call this - * on Windows as well to distinguish true duplicates from names differing by - * case only. */ -static int -compare_utf16le_names_case_sensitive(const utf16lechar *name1, size_t nbytes1, - const utf16lechar *name2, size_t nbytes2) -{ - /* Return the result if the strings differ up to their minimum length. - * Note that we cannot strcmp() or strncmp() here, as the strings are in - * UTF-16LE format. */ - int result = memcmp(name1, name2, min(nbytes1, nbytes2)); - if (result) - return result; - /* The strings are the same up to their minimum length, so return a - * result based on their lengths. */ - if (nbytes1 < nbytes2) - return -1; - else if (nbytes1 > nbytes2) - return 1; - else - return 0; -} - -#ifdef __WIN32__ -/* Windoze: Case-insensitive UTF-16LE dentry or stream name comparison */ -static int -compare_utf16le_names_case_insensitive(const utf16lechar *name1, size_t nbytes1, - const utf16lechar *name2, size_t nbytes2) -{ - /* Only call _wcsicmp() if both strings are of nonzero length; otherwise - * one could be NULL. */ - if (nbytes1 && nbytes2) - return _wcsicmp((const wchar_t*)name1, (const wchar_t*)name2); - /* The strings are the same up to their minimum length, so return a - * result based on their lengths. */ - if (nbytes1 < nbytes2) - return -1; - else if (nbytes1 > nbytes2) - return 1; - else - return 0; -} -#endif /* __WIN32__ */ - -#ifdef __WIN32__ -# define compare_utf16le_names compare_utf16le_names_case_insensitive -#else -# define compare_utf16le_names compare_utf16le_names_case_sensitive -#endif - - -#ifdef __WIN32__ static int dentry_compare_names_case_insensitive(const struct wim_dentry *d1, const struct wim_dentry *d2) { - return compare_utf16le_names_case_insensitive(d1->file_name, - d1->file_name_nbytes, - d2->file_name, - d2->file_name_nbytes); + return cmp_utf16le_strings(d1->file_name, + d1->file_name_nbytes / 2, + d2->file_name, + d2->file_name_nbytes / 2, + true); } -#endif /* __WIN32__ */ static int dentry_compare_names_case_sensitive(const struct wim_dentry *d1, const struct wim_dentry *d2) { - return compare_utf16le_names_case_sensitive(d1->file_name, - d1->file_name_nbytes, - d2->file_name, - d2->file_name_nbytes); + return cmp_utf16le_strings(d1->file_name, + d1->file_name_nbytes / 2, + d2->file_name, + d2->file_name_nbytes / 2, + false); } -#ifdef __WIN32__ -# define dentry_compare_names dentry_compare_names_case_insensitive -#else -# define dentry_compare_names dentry_compare_names_case_sensitive -#endif - /* Return %true iff the alternate data stream entry @entry has the UTF-16LE * stream name @name that has length @name_nbytes bytes. */ static inline bool ads_entry_has_name(const struct wim_ads_entry *entry, - const utf16lechar *name, size_t name_nbytes) + const utf16lechar *name, size_t name_nbytes, + bool ignore_case) { - return !compare_utf16le_names(name, name_nbytes, - entry->stream_name, - entry->stream_name_nbytes); + return 0 == cmp_utf16le_strings(name, + name_nbytes / 2, + entry->stream_name, + entry->stream_name_nbytes / 2, + ignore_case); } +/* Default case sensitivity behavior for searches with + * WIMLIB_CASE_PLATFORM_DEFAULT specified. This can be modified by + * wimlib_global_init(). */ +bool default_ignore_case = +#ifdef __WIN32__ + true +#else + false +#endif +; + +static bool +will_ignore_case(CASE_SENSITIVITY_TYPE case_type) +{ + if (case_type == WIMLIB_CASE_SENSITIVE) + return false; + if (case_type == WIMLIB_CASE_INSENSITIVE) + return true; + + return default_ignore_case; +} + + +/* Given a UTF-16LE filename and a directory, look up the dentry for the file. + * Return it if found, otherwise NULL. This is case-sensitive on UNIX and + * case-insensitive on Windows. */ struct wim_dentry * get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry, const utf16lechar *name, - size_t name_nbytes) + size_t name_nbytes, + CASE_SENSITIVITY_TYPE case_ctype) { - struct rb_node *node = dentry->d_inode->i_children.rb_node; + struct rb_node *node; + + bool ignore_case = will_ignore_case(case_ctype); + + if (ignore_case) + node = dentry->d_inode->i_children_case_insensitive.rb_node; + else + node = dentry->d_inode->i_children.rb_node; + struct wim_dentry *child; while (node) { - child = rbnode_dentry(node); - int result = compare_utf16le_names(name, name_nbytes, - child->file_name, - child->file_name_nbytes); - if (result < 0) + if (ignore_case) + child = rb_entry(node, struct wim_dentry, rb_node_case_insensitive); + else + child = rb_entry(node, struct wim_dentry, rb_node); + + int result = cmp_utf16le_strings(name, + name_nbytes / 2, + child->file_name, + child->file_name_nbytes / 2, + ignore_case); + if (result < 0) { node = node->rb_left; - else if (result > 0) + } else if (result > 0) { node = node->rb_right; - else + } else if (!ignore_case || + list_empty(&child->case_insensitive_conflict_list)) { return child; + } else { + /* Multiple dentries have the same case-insensitive + * name, and a case-insensitive lookup is being + * performed. Choose the dentry with the same + * case-sensitive name, if one exists; otherwise print a + * warning and choose one arbitrarily. */ + struct wim_dentry *alt = child; + size_t num_alts = 0; + + do { + num_alts++; + if (0 == cmp_utf16le_strings(name, + name_nbytes / 2, + alt->file_name, + alt->file_name_nbytes / 2, + false)) + return alt; + alt = list_entry(alt->case_insensitive_conflict_list.next, + struct wim_dentry, + case_insensitive_conflict_list); + } while (alt != child); + + WARNING("Result of case-insensitive lookup is ambiguous\n" + " (returning \"%"TS"\" of %zu " + "possible files, including \"%"TS"\")", + dentry_full_path(child), + num_alts, + dentry_full_path(list_entry(child->case_insensitive_conflict_list.next, + struct wim_dentry, + case_insensitive_conflict_list))); + return child; + } } return NULL; } @@ -594,11 +770,13 @@ get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry, /* Returns the child of @dentry that has the file name @name. Returns NULL if * no child has the name. */ struct wim_dentry * -get_dentry_child_with_name(const struct wim_dentry *dentry, const tchar *name) +get_dentry_child_with_name(const struct wim_dentry *dentry, const tchar *name, + CASE_SENSITIVITY_TYPE case_type) { #if TCHAR_IS_UTF16LE return get_dentry_child_with_utf16le_name(dentry, name, - tstrlen(name) * sizeof(tchar)); + tstrlen(name) * sizeof(tchar), + case_type); #else utf16lechar *utf16le_name; size_t utf16le_name_nbytes; @@ -612,7 +790,8 @@ get_dentry_child_with_name(const struct wim_dentry *dentry, const tchar *name) } else { child = get_dentry_child_with_utf16le_name(dentry, utf16le_name, - utf16le_name_nbytes); + utf16le_name_nbytes, + case_type); FREE(utf16le_name); } return child; @@ -620,49 +799,111 @@ get_dentry_child_with_name(const struct wim_dentry *dentry, const tchar *name) } static struct wim_dentry * -get_dentry_utf16le(WIMStruct *wim, const utf16lechar *path) +get_dentry_utf16le(WIMStruct *wim, const utf16lechar *path, + CASE_SENSITIVITY_TYPE case_type) { - struct wim_dentry *cur_dentry, *parent_dentry; - const utf16lechar *p, *pp; + struct wim_dentry *cur_dentry; + const utf16lechar *name_start, *name_end; - cur_dentry = parent_dentry = wim_root_dentry(wim); - if (!cur_dentry) { - errno = ENOENT; - return NULL; - } - p = path; - while (1) { - while (*p == cpu_to_le16('/')) - p++; - if (*p == cpu_to_le16('\0')) - break; - pp = p; - while (*pp != cpu_to_le16('/') && *pp != cpu_to_le16('\0')) - pp++; + /* Start with the root directory of the image. Note: this will be NULL + * if an image has been added directly with wimlib_add_empty_image() but + * no files have been added yet; in that case we fail with ENOENT. */ + cur_dentry = wim_root_dentry(wim); - cur_dentry = get_dentry_child_with_utf16le_name(parent_dentry, p, - (void*)pp - (void*)p); - if (cur_dentry == NULL) - break; - p = pp; - parent_dentry = cur_dentry; - } - if (cur_dentry == NULL) { - if (dentry_is_directory(parent_dentry)) + name_start = path; + for (;;) { + if (cur_dentry == NULL) { errno = ENOENT; - else + return NULL; + } + + if (*name_start && !dentry_is_directory(cur_dentry)) { errno = ENOTDIR; + return NULL; + } + + while (*name_start == cpu_to_le16(WIM_PATH_SEPARATOR)) + name_start++; + + if (!*name_start) + return cur_dentry; + + name_end = name_start; + do { + ++name_end; + } while (*name_end != cpu_to_le16(WIM_PATH_SEPARATOR) && *name_end); + + cur_dentry = get_dentry_child_with_utf16le_name(cur_dentry, + name_start, + (u8*)name_end - (u8*)name_start, + case_type); + name_start = name_end; } - return cur_dentry; } -/* Returns the dentry corresponding to the @path, or NULL if there is no such - * dentry. */ +/* + * WIM path lookup: translate a path in the currently selected WIM image to the + * corresponding dentry, if it exists. + * + * @wim + * The WIMStruct for the WIM. The search takes place in the currently + * selected image. + * + * @path + * The path to look up, given relative to the root of the WIM image. + * Characters with value WIM_PATH_SEPARATOR are taken to be path + * separators. Leading path separators are ignored, whereas one or more + * trailing path separators cause the path to only match a directory. + * + * @case_type + * The case-sensitivity behavior of this function, as one of the following + * constants: + * + * - WIMLIB_CASE_SENSITIVE: Perform the search case sensitively. This means + * that names must match exactly. + * + * - WIMLIB_CASE_INSENSITIVE: Perform the search case insensitively. This + * means that names are considered to match if they are equal when + * transformed to upper case. If a path component matches multiple names + * case-insensitively, the name that matches the path component + * case-sensitively is chosen, if existent; otherwise one + * case-insensitively matching name is chosen arbitrarily. + * + * - WIMLIB_CASE_PLATFORM_DEFAULT: Perform either case-sensitive or + * case-insensitive search, depending on the value of the global variable + * default_ignore_case. + * + * In any case, no Unicode normalization is done before comparing strings. + * + * Returns a pointer to the dentry that is the result of the lookup, or NULL if + * no such dentry exists. If NULL is returned, errno is set to one of the + * following values: + * + * ENOTDIR if one of the path components used as a directory existed but + * was not, in fact, a directory. + * + * ENOENT otherwise. + * + * Additional notes: + * + * - This function does not consider a reparse point to be a directory, even + * if it has FILE_ATTRIBUTE_DIRECTORY set. + * + * - This function does not dereference symbolic links or junction points + * when performing the search. + * + * - Since this function ignores leading slashes, the empty path is valid and + * names the root directory of the WIM image. + * + * - An image added with wimlib_add_empty_image() does not have a root + * directory yet, and this function will fail with ENOENT for any path on + * such an image. + */ struct wim_dentry * -get_dentry(WIMStruct *wim, const tchar *path) +get_dentry(WIMStruct *wim, const tchar *path, CASE_SENSITIVITY_TYPE case_type) { #if TCHAR_IS_UTF16LE - return get_dentry_utf16le(wim, path); + return get_dentry_utf16le(wim, path, case_type); #else utf16lechar *path_utf16le; size_t path_utf16le_nbytes; @@ -673,49 +914,41 @@ get_dentry(WIMStruct *wim, const tchar *path) &path_utf16le, &path_utf16le_nbytes); if (ret) return NULL; - dentry = get_dentry_utf16le(wim, path_utf16le); + dentry = get_dentry_utf16le(wim, path_utf16le, case_type); FREE(path_utf16le); return dentry; #endif } -struct wim_inode * -wim_pathname_to_inode(WIMStruct *wim, const tchar *path) -{ - struct wim_dentry *dentry; - dentry = get_dentry(wim, path); - if (dentry) - return dentry->d_inode; - else - return NULL; -} - /* Takes in a path of length @len in @buf, and transforms it into a string for * the path of its parent directory. */ static void to_parent_name(tchar *buf, size_t len) { ssize_t i = (ssize_t)len - 1; - while (i >= 0 && buf[i] == T('/')) + while (i >= 0 && buf[i] == WIM_PATH_SEPARATOR) i--; - while (i >= 0 && buf[i] != T('/')) + while (i >= 0 && buf[i] != WIM_PATH_SEPARATOR) i--; - while (i >= 0 && buf[i] == T('/')) + while (i >= 0 && buf[i] == WIM_PATH_SEPARATOR) i--; buf[i + 1] = T('\0'); } -/* Returns the dentry that corresponds to the parent directory of @path, or NULL - * if the dentry is not found. */ +/* Similar to get_dentry(), but returns the dentry named by @path with the last + * component stripped off. + * + * Note: The returned dentry is NOT guaranteed to be a directory. */ struct wim_dentry * -get_parent_dentry(WIMStruct *wim, const tchar *path) +get_parent_dentry(WIMStruct *wim, const tchar *path, + CASE_SENSITIVITY_TYPE case_type) { size_t path_len = tstrlen(path); tchar buf[path_len + 1]; tmemcpy(buf, path, path_len + 1); to_parent_name(buf, path_len); - return get_dentry(wim, buf); + return get_dentry(wim, buf, case_type); } /* Prints the full path of a dentry. */ @@ -848,14 +1081,8 @@ new_timeless_inode(void) inode->i_nlink = 1; inode->i_next_stream_id = 1; inode->i_not_rpfixed = 1; + inode->i_canonical_streams = 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; @@ -882,7 +1109,7 @@ new_dentry(const tchar *name, struct wim_dentry **dentry_ret) int ret; dentry = MALLOC(sizeof(struct wim_dentry)); - if (!dentry) + if (dentry == NULL) return WIMLIB_ERR_NOMEM; dentry_common_init(dentry); @@ -914,7 +1141,7 @@ _new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret, dentry->d_inode = new_timeless_inode(); else dentry->d_inode = new_inode(); - if (!dentry->d_inode) { + if (dentry->d_inode == NULL) { free_dentry(dentry); return WIMLIB_ERR_NOMEM; } @@ -954,6 +1181,19 @@ 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) @@ -963,7 +1203,7 @@ init_ads_entry(struct wim_ads_entry *ads_entry, const void *name, if (is_utf16le) { utf16lechar *p = MALLOC(name_nbytes + sizeof(utf16lechar)); - if (!p) + if (p == NULL) return WIMLIB_ERR_NOMEM; memcpy(p, name, name_nbytes); p[name_nbytes / 2] = cpu_to_le16(0); @@ -994,15 +1234,10 @@ 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); } } @@ -1065,11 +1300,13 @@ 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) @@ -1077,6 +1314,42 @@ free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table) for_dentry_in_tree_depth(root, do_free_dentry, lookup_table); } +/* Insert a dentry into the case insensitive index for a directory. + * + * This is a red-black tree, but when multiple dentries share the same + * case-insensitive name, only one is inserted into the tree itself; the rest + * are connected in a list. + */ +static struct wim_dentry * +dentry_add_child_case_insensitive(struct wim_dentry *parent, + struct wim_dentry *child) +{ + struct rb_root *root; + struct rb_node **new; + struct rb_node *rb_parent; + + root = &parent->d_inode->i_children_case_insensitive; + new = &root->rb_node; + rb_parent = NULL; + while (*new) { + struct wim_dentry *this = container_of(*new, struct wim_dentry, + rb_node_case_insensitive); + int result = dentry_compare_names_case_insensitive(child, this); + + rb_parent = *new; + + if (result < 0) + new = &((*new)->rb_left); + else if (result > 0) + new = &((*new)->rb_right); + else + return this; + } + rb_link_node(&child->rb_node_case_insensitive, rb_parent, new); + rb_insert_color(&child->rb_node_case_insensitive, root); + return NULL; +} + /* * Links a dentry into the directory tree. * @@ -1084,23 +1357,27 @@ free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table) * @child: The dentry to link. * * Returns NULL if successful. If @parent already contains a dentry with the - * same name as @child (see compare_utf16le_names() for what names are - * considered the "same"), the pointer to this duplicate dentry is returned. + * same case-sensitive name as @child, the pointer to this duplicate dentry is + * returned. */ struct wim_dentry * dentry_add_child(struct wim_dentry * restrict parent, struct wim_dentry * restrict child) { + struct rb_root *root; + struct rb_node **new; + struct rb_node *rb_parent; + wimlib_assert(dentry_is_directory(parent)); wimlib_assert(parent != child); - struct rb_root *root = &parent->d_inode->i_children; - struct rb_node **new = &(root->rb_node); - struct rb_node *rb_parent = NULL; - + /* Case sensitive child dentry index */ + root = &parent->d_inode->i_children; + new = &root->rb_node; + rb_parent = NULL; while (*new) { struct wim_dentry *this = rbnode_dentry(*new); - int result = dentry_compare_names(child, this); + int result = dentry_compare_names_case_sensitive(child, this); rb_parent = *new; @@ -1114,6 +1391,19 @@ dentry_add_child(struct wim_dentry * restrict parent, child->parent = parent; rb_link_node(&child->rb_node, rb_parent, new); rb_insert_color(&child->rb_node, root); + + /* Case insensitive child dentry index */ + { + struct wim_dentry *existing; + existing = dentry_add_child_case_insensitive(parent, child); + if (existing) { + list_add(&child->case_insensitive_conflict_list, + &existing->case_insensitive_conflict_list); + child->rb_node_case_insensitive.__rb_parent_color = 0; + } else { + INIT_LIST_HEAD(&child->case_insensitive_conflict_list); + } + } return NULL; } @@ -1121,13 +1411,109 @@ dentry_add_child(struct wim_dentry * restrict parent, void unlink_dentry(struct wim_dentry *dentry) { - if (!dentry_is_root(dentry)) - rb_erase(&dentry->rb_node, &dentry->parent->d_inode->i_children); + struct wim_dentry *parent = dentry->parent; + + if (parent == dentry) + return; + rb_erase(&dentry->rb_node, &parent->d_inode->i_children); + + if (dentry->rb_node_case_insensitive.__rb_parent_color) { + /* This dentry was in the case-insensitive red-black tree. */ + rb_erase(&dentry->rb_node_case_insensitive, + &parent->d_inode->i_children_case_insensitive); + if (!list_empty(&dentry->case_insensitive_conflict_list)) { + /* Make a different case-insensitively-the-same dentry + * be the "representative" in the red-black tree. */ + struct list_head *next; + struct wim_dentry *other; + struct wim_dentry *existing; + + next = dentry->case_insensitive_conflict_list.next; + other = list_entry(next, struct wim_dentry, case_insensitive_conflict_list); + existing = dentry_add_child_case_insensitive(parent, other); + wimlib_assert(existing == NULL); + } + } + list_del(&dentry->case_insensitive_conflict_list); +} + +static int +free_dentry_full_path(struct wim_dentry *dentry, void *_ignore) +{ + FREE(dentry->_full_path); + dentry->_full_path = NULL; + return 0; +} + +/* Rename a file or directory in the WIM. */ +int +rename_wim_path(WIMStruct *wim, const tchar *from, const tchar *to, + CASE_SENSITIVITY_TYPE case_type) +{ + struct wim_dentry *src; + struct wim_dentry *dst; + struct wim_dentry *parent_of_dst; + int ret; + + /* This rename() implementation currently only supports actual files + * (not alternate data streams) */ + + src = get_dentry(wim, from, case_type); + if (!src) + return -errno; + + dst = get_dentry(wim, to, case_type); + + if (dst) { + /* Destination file exists */ + + if (src == dst) /* Same file */ + return 0; + + if (!dentry_is_directory(src)) { + /* Cannot rename non-directory to directory. */ + if (dentry_is_directory(dst)) + return -EISDIR; + } else { + /* Cannot rename directory to a non-directory or a non-empty + * directory */ + if (!dentry_is_directory(dst)) + return -ENOTDIR; + if (dentry_has_children(dst)) + return -ENOTEMPTY; + } + parent_of_dst = dst->parent; + } else { + /* Destination does not exist */ + parent_of_dst = get_parent_dentry(wim, to, case_type); + if (!parent_of_dst) + return -errno; + + if (!dentry_is_directory(parent_of_dst)) + return -ENOTDIR; + } + + ret = set_dentry_name(src, path_basename(to)); + if (ret) + return -ENOMEM; + if (dst) { + unlink_dentry(dst); + free_dentry_tree(dst, wim->lookup_table); + } + unlink_dentry(src); + dentry_add_child(parent_of_dst, src); + if (src->_full_path) + for_dentry_in_tree(src, free_dentry_full_path, NULL); + return 0; } /* * Returns the alternate data stream entry belonging to @inode that has the - * stream name @stream_name. + * stream name @stream_name, or NULL if the inode has no alternate data stream + * with that name. + * + * If @p stream_name is the empty string, NULL is returned --- that is, this + * function will not return "unnamed" alternate data stream entries. */ struct wim_ads_entry * inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name, @@ -1140,6 +1526,9 @@ inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name, u16 i; struct wim_ads_entry *result; + if (stream_name[0] == T('\0')) + return NULL; + #if TCHAR_IS_UTF16LE const utf16lechar *stream_name_utf16le; @@ -1163,7 +1552,8 @@ inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name, do { if (ads_entry_has_name(&inode->i_ads_entries[i], stream_name_utf16le, - stream_name_utf16le_nbytes)) + stream_name_utf16le_nbytes, + default_ignore_case)) { if (idx_ret) *idx_ret = i; @@ -1186,6 +1576,8 @@ do_inode_add_ads(struct wim_inode *inode, const void *stream_name, struct wim_ads_entry *ads_entries; struct wim_ads_entry *new_entry; + wimlib_assert(stream_name_nbytes != 0); + if (inode->i_num_ads >= 0xfffe) { ERROR("Too many alternate data streams in one inode!"); return NULL; @@ -1193,7 +1585,7 @@ do_inode_add_ads(struct wim_inode *inode, const void *stream_name, num_ads = inode->i_num_ads + 1; ads_entries = REALLOC(inode->i_ads_entries, num_ads * sizeof(inode->i_ads_entries[0])); - if (!ads_entries) { + if (ads_entries == NULL) { ERROR("Failed to allocate memory for new alternate data stream"); return NULL; } @@ -1217,8 +1609,10 @@ inode_add_ads_utf16le(struct wim_inode *inode, } /* - * Add an alternate stream entry to a WIM inode and return a pointer to it, or - * NULL if memory could not be allocated. + * Add an alternate stream entry to a WIM inode. On success, returns a pointer + * to the new entry; on failure, returns NULL. + * + * @stream_name must be a nonempty string. */ struct wim_ads_entry * inode_add_ads(struct wim_inode *inode, const tchar *stream_name) @@ -1237,24 +1631,24 @@ 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); + wimlib_assert(existing_lte->size == size); lte = existing_lte; lte->refcnt++; } else { void *buffer_copy; lte = new_lookup_table_entry(); - if (!lte) + if (lte == NULL) return NULL; buffer_copy = memdup(buffer, size); - if (!buffer_copy) { + if (buffer_copy == NULL) { free_lookup_table_entry(lte); return NULL; } - lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER; - lte->attached_buffer = buffer_copy; - lte->resource_entry.original_size = size; + lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER; + lte->attached_buffer = buffer_copy; + lte->size = size; copy_hash(lte->hash, hash); lookup_table_insert(lookup_table, lte); } @@ -1271,12 +1665,12 @@ inode_add_ads_with_data(struct wim_inode *inode, const tchar *name, wimlib_assert(inode->i_resolved); new_ads_entry = inode_add_ads(inode, name); - if (!new_ads_entry) + if (new_ads_entry == NULL) return WIMLIB_ERR_NOMEM; new_ads_entry->lte = add_stream_from_data_buffer(value, size, lookup_table); - if (!new_ads_entry->lte) { + if (new_ads_entry->lte == NULL) { inode_remove_ads(inode, new_ads_entry - inode->i_ads_entries, lookup_table); return WIMLIB_ERR_NOMEM; @@ -1284,6 +1678,15 @@ 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 @@ -1291,7 +1694,7 @@ inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len, struct wim_lookup_table *lookup_table) { inode->i_lte = add_stream_from_data_buffer(data, len, lookup_table); - if (!inode->i_lte) + if (inode->i_lte == NULL) return WIMLIB_ERR_NOMEM; inode->i_resolved = 1; return 0; @@ -1324,6 +1727,15 @@ 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, @@ -1339,21 +1751,21 @@ inode_get_unix_data(const struct wim_inode *inode, ads_entry = inode_get_ads_entry((struct wim_inode*)inode, WIMLIB_UNIX_DATA_TAG, NULL); - if (!ads_entry) + if (ads_entry == NULL) return NO_UNIX_DATA; if (stream_idx_ret) *stream_idx_ret = ads_entry - inode->i_ads_entries; lte = ads_entry->lte; - if (!lte) + if (lte == NULL) return NO_UNIX_DATA; - size = wim_resource_size(lte); + size = lte->size; if (size != sizeof(struct wimlib_unix_data)) return BAD_UNIX_DATA; - ret = read_full_resource_into_buf(lte, unix_data); + ret = read_full_stream_into_buf(lte, unix_data); if (ret) return ret; @@ -1396,55 +1808,27 @@ inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode, } #endif /* !__WIN32__ */ -/* Replace weird characters in filenames and alternate data stream names. - * - * In particular we do not want the path separator to appear in any names, as - * that would make it possible for a "malicious" WIM to extract itself to any - * location it wanted to. */ -static void -replace_forbidden_characters(utf16lechar *name) -{ - utf16lechar *p; - - for (p = name; *p; p++) { - #ifdef __WIN32__ - if (wcschr(L"<>:\"/\\|?*", (wchar_t)*p)) - #else - if (*p == cpu_to_le16('/')) - #endif - { - if (name) { - WARNING("File, directory, or stream name \"%"WS"\"\n" - " contains forbidden characters; " - "substituting replacement characters.", - name); - name = NULL; - } - #ifdef __WIN32__ - *p = cpu_to_le16(0xfffd); - #else - *p = cpu_to_le16('?'); - #endif - } - } -} - /* * 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, @@ -1460,7 +1844,7 @@ read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode, * data stream entries. */ num_ads = inode->i_num_ads; ads_entries = CALLOC(num_ads, sizeof(inode->i_ads_entries[0])); - if (!ads_entries) + if (ads_entries == NULL) goto out_of_memory; /* Read the entries into our newly allocated buffer. */ @@ -1515,14 +1899,16 @@ read_ads_entries(const u8 * restrict p, struct wim_inode * restrict inode, goto out_invalid; cur_entry->stream_name = MALLOC(cur_entry->stream_name_nbytes + 2); - if (!cur_entry->stream_name) + if (cur_entry->stream_name == NULL) goto out_of_memory; memcpy(cur_entry->stream_name, disk_entry->stream_name, cur_entry->stream_name_nbytes); cur_entry->stream_name[cur_entry->stream_name_nbytes / 2] = cpu_to_le16(0); - replace_forbidden_characters(cur_entry->stream_name); + } else { + /* Mark inode as having weird stream entries. */ + inode->i_canonical_streams = 0; } /* It's expected that the size of every ADS entry is a multiple @@ -1549,7 +1935,7 @@ 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++) @@ -1570,7 +1956,7 @@ 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. * @@ -1580,9 +1966,10 @@ 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, @@ -1616,7 +2003,7 @@ 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); @@ -1638,7 +2025,7 @@ 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 @@ -1646,12 +2033,12 @@ 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'. */ inode = new_timeless_inode(); - if (!inode) + if (inode == NULL) return WIMLIB_ERR_NOMEM; /* Read more fields; some into the dentry, and some into the inode. */ @@ -1692,7 +2079,7 @@ 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; } @@ -1702,14 +2089,14 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, * The calculated length here is unaligned to allow for the possibility * that the dentry->length names an unaligned length, although this * would be unexpected. */ - calculated_size = _dentry_correct_length_unaligned(file_name_nbytes, - short_name_nbytes); + calculated_size = dentry_correct_length_unaligned(file_name_nbytes, + short_name_nbytes); if (dentry->length < calculated_size) { 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; } @@ -1719,7 +2106,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, * is no null terminator following it. */ if (file_name_nbytes) { file_name = MALLOC(file_name_nbytes + 2); - if (!file_name) { + if (file_name == NULL) { ERROR("Failed to allocate %d bytes for dentry file name", file_name_nbytes + 2); ret = WIMLIB_ERR_NOMEM; @@ -1728,7 +2115,6 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, memcpy(file_name, p, file_name_nbytes); p += file_name_nbytes + 2; file_name[file_name_nbytes / 2] = cpu_to_le16(0); - replace_forbidden_characters(file_name); } else { file_name = NULL; } @@ -1738,7 +2124,7 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, * filename, there is no null terminator following it. */ if (short_name_nbytes) { short_name = MALLOC(short_name_nbytes + 2); - if (!short_name) { + if (short_name == NULL) { ERROR("Failed to allocate %d bytes for dentry short name", short_name_nbytes + 2); ret = WIMLIB_ERR_NOMEM; @@ -1747,7 +2133,6 @@ read_dentry(const u8 * restrict metadata_resource, u64 metadata_resource_len, memcpy(short_name, p, short_name_nbytes); p += short_name_nbytes + 2; short_name[short_name_nbytes / 2] = cpu_to_le16(0); - replace_forbidden_characters(short_name); } else { short_name = NULL; } @@ -1765,7 +2150,7 @@ 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, @@ -1810,26 +2195,32 @@ 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; @@ -1842,6 +2233,18 @@ 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 (;;) { @@ -1858,7 +2261,7 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, /* Not end of directory. Allocate this child permanently and * link it to the parent and previous child. */ child = memdup(&cur_child, sizeof(struct wim_dentry)); - if (!child) { + if (child == NULL) { ERROR("Failed to allocate new dentry!"); ret = WIMLIB_ERR_NOMEM; break; @@ -1869,59 +2272,81 @@ read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len, * cur_child.length, which although it does take into account * the padding, it DOES NOT take into account alternate stream * entries. */ - cur_offset += dentry_total_length(child); + cur_offset += dentry_in_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); - /* On UNIX, duplicates are exact. On Windows, - * duplicates may differ by case and we wish to provide - * a different warning message in this case. */ - #ifdef __WIN32__ - if (dentry_compare_names_case_sensitive(child, duplicate)) - { - child->parent = dentry; - WARNING("Ignoring %ls \"%ls\", which differs " - "only in case from %ls \"%ls\"", - child_type, - dentry_full_path(child), - duplicate_type, - dentry_full_path(duplicate)); - } - else - #endif - { - WARNING("Ignoring duplicate %"TS" \"%"TS"\" " - "(the WIM image already contains a %"TS" " - "at that path with the exact same name)", - child_type, dentry_full_path(duplicate), - duplicate_type); - } + WARNING("Ignoring duplicate %"TS" \"%"TS"\" " + "(the WIM image already contains a %"TS" " + "at that path with the exact same name)", + child_type, dentry_full_path(duplicate), + duplicate_type); free_dentry(child); - } 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)); - } - } + 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; } +/* + * Writes a WIM alternate data stream (ADS) entry to an output buffer. + * + * @ads_entry: The ADS entry structure. + * @hash: The hash field to use (instead of the one in the ADS entry). + * @p: The memory location to write the data to. + * + * Returns a pointer to the byte after the last byte written. + */ +static u8 * +write_ads_entry(const struct wim_ads_entry *ads_entry, + const u8 *hash, u8 * restrict p) +{ + struct wim_ads_entry_on_disk *disk_ads_entry = + (struct wim_ads_entry_on_disk*)p; + u8 *orig_p = p; + + disk_ads_entry->reserved = cpu_to_le64(ads_entry->reserved); + copy_hash(disk_ads_entry->hash, hash); + disk_ads_entry->stream_name_nbytes = cpu_to_le16(ads_entry->stream_name_nbytes); + p += sizeof(struct wim_ads_entry_on_disk); + if (ads_entry->stream_name_nbytes) { + p = mempcpy(p, ads_entry->stream_name, + ads_entry->stream_name_nbytes + 2); + } + /* Align to 8-byte boundary */ + while ((uintptr_t)p & 7) + *p++ = 0; + disk_ads_entry->length = cpu_to_le64(p - orig_p); + return p; +} + /* * Writes a WIM dentry to an output buffer. * @@ -1938,11 +2363,14 @@ write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p) struct wim_dentry_on_disk *disk_dentry; const u8 *orig_p; const u8 *hash; + bool use_dummy_stream; + u16 num_ads; wimlib_assert(((uintptr_t)p & 7) == 0); /* 8 byte aligned */ orig_p = p; - inode = dentry->d_inode; + inode = dentry->d_inode; + use_dummy_stream = inode_needs_dummy_stream(inode); disk_dentry = (struct wim_dentry_on_disk*)p; disk_dentry->attributes = cpu_to_le32(inode->i_attributes); @@ -1953,7 +2381,10 @@ write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p) disk_dentry->creation_time = cpu_to_le64(inode->i_creation_time); disk_dentry->last_access_time = cpu_to_le64(inode->i_last_access_time); disk_dentry->last_write_time = cpu_to_le64(inode->i_last_write_time); - hash = inode_stream_hash(inode, 0); + if (use_dummy_stream) + hash = zero_hash; + else + hash = inode_stream_hash(inode, 0); copy_hash(disk_dentry->unnamed_stream_hash, hash); if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { disk_dentry->reparse.rp_unknown_1 = cpu_to_le32(inode->i_rp_unknown_1); @@ -1965,11 +2396,16 @@ write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p) disk_dentry->nonreparse.hard_link_group_id = cpu_to_le64((inode->i_nlink == 1) ? 0 : inode->i_ino); } - disk_dentry->num_alternate_data_streams = cpu_to_le16(inode->i_num_ads); + num_ads = inode->i_num_ads; + if (use_dummy_stream) + num_ads++; + disk_dentry->num_alternate_data_streams = cpu_to_le16(num_ads); disk_dentry->short_name_nbytes = cpu_to_le16(dentry->short_name_nbytes); 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); @@ -1987,29 +2423,17 @@ write_dentry(const struct wim_dentry * restrict dentry, u8 * restrict p) * have been renamed, thus changing its needed length. */ disk_dentry->length = cpu_to_le64(p - orig_p); + if (use_dummy_stream) { + hash = inode_unnamed_stream_hash(inode); + p = write_ads_entry(&(struct wim_ads_entry){}, hash, p); + } + /* Write the alternate data streams entries, if any. */ for (u16 i = 0; i < inode->i_num_ads; i++) { - const struct wim_ads_entry *ads_entry = - &inode->i_ads_entries[i]; - struct wim_ads_entry_on_disk *disk_ads_entry = - (struct wim_ads_entry_on_disk*)p; - orig_p = p; - - disk_ads_entry->reserved = cpu_to_le64(ads_entry->reserved); - hash = inode_stream_hash(inode, i + 1); - copy_hash(disk_ads_entry->hash, hash); - disk_ads_entry->stream_name_nbytes = cpu_to_le16(ads_entry->stream_name_nbytes); - p += sizeof(struct wim_ads_entry_on_disk); - if (ads_entry->stream_name_nbytes) { - p = mempcpy(p, ads_entry->stream_name, - ads_entry->stream_name_nbytes + 2); - } - /* Align to 8-byte boundary */ - while ((uintptr_t)p & 7) - *p++ = 0; - disk_ads_entry->length = cpu_to_le64(p - orig_p); + p = write_ads_entry(&inode->i_ads_entries[i], hash, p); } + return p; } @@ -2066,7 +2490,7 @@ 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)); @@ -2083,3 +2507,423 @@ 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 == NULL) + 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, WIMLIB_CASE_PLATFORM_DEFAULT); + if (dentry == NULL) + 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 (lte->size != template_lte->size) + 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 && lte->size) { + return false; + } else if (template_lte && template_lte->size) { + 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 == NULL || template_lte == NULL || + !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, + WIMLIB_CASE_SENSITIVE); + if (template_dentry == NULL) { + 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; +}