4 * In the WIM file format, the dentries are stored in the "metadata resource"
5 * section right after the security data. Each image in the WIM file has its
6 * own metadata resource with its own security data and dentry tree. Dentries
7 * in different images may share file resources by referring to the same lookup
12 * Copyright (C) 2012, 2013 Eric Biggers
14 * This file is part of wimlib, a library for working with WIM files.
16 * wimlib is free software; you can redistribute it and/or modify it under the
17 * terms of the GNU General Public License as published by the Free Software
18 * Foundation; either version 3 of the License, or (at your option) any later
21 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
22 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
23 * A PARTICULAR PURPOSE. See the GNU General Public License for more details.
25 * You should have received a copy of the GNU General Public License along with
26 * wimlib; if not, see http://www.gnu.org/licenses/.
29 #include "buffer_io.h"
31 #include "lookup_table.h"
32 #include "timestamp.h"
33 #include "wimlib_internal.h"
36 /* Calculates the unaligned length, in bytes, of an on-disk WIM dentry that has
37 * a file name and short name that take the specified numbers of bytes. This
38 * excludes any alternate data stream entries that may follow the dentry. */
40 __dentry_correct_length_unaligned(u16 file_name_nbytes, u16 short_name_nbytes)
42 u64 length = WIM_DENTRY_DISK_SIZE;
44 length += file_name_nbytes + 2;
45 if (short_name_nbytes)
46 length += short_name_nbytes + 2;
50 /* Calculates the unaligned length, in bytes, of an on-disk WIM dentry, based on
51 * the file name length and short name length. Note that dentry->length is
52 * ignored; also, this excludes any alternate data stream entries that may
53 * follow the dentry. */
55 dentry_correct_length_unaligned(const struct wim_dentry *dentry)
57 return __dentry_correct_length_unaligned(dentry->file_name_nbytes,
58 dentry->short_name_nbytes);
61 /* Return the "correct" value to write in the length field of a WIM dentry,
62 * based on the file name length and short name length. */
64 dentry_correct_length(const struct wim_dentry *dentry)
66 return (dentry_correct_length_unaligned(dentry) + 7) & ~7;
69 /* Return %true iff the alternate data stream entry @entry has the UTF-16LE
70 * stream name @name that has length @name_nbytes bytes. */
72 ads_entry_has_name(const struct wim_ads_entry *entry,
73 const utf16lechar *name, size_t name_nbytes)
75 return entry->stream_name_nbytes == name_nbytes &&
76 memcmp(entry->stream_name, name, name_nbytes) == 0;
79 /* Duplicates a string of system-dependent encoding into a UTF-16LE string and
80 * returns the string and its length, in bytes, in the pointer arguments. Frees
81 * any existing string at the return location before overwriting it. */
83 get_utf16le_name(const tchar *name, utf16lechar **name_utf16le_ret,
84 u16 *name_utf16le_nbytes_ret)
86 utf16lechar *name_utf16le;
87 size_t name_utf16le_nbytes;
90 name_utf16le_nbytes = tstrlen(name) * sizeof(utf16lechar);
91 name_utf16le = MALLOC(name_utf16le_nbytes + sizeof(utf16lechar));
93 return WIMLIB_ERR_NOMEM;
94 memcpy(name_utf16le, name, name_utf16le_nbytes + sizeof(utf16lechar));
98 ret = tstr_to_utf16le(name, tstrlen(name), &name_utf16le,
99 &name_utf16le_nbytes);
101 if (name_utf16le_nbytes > 0xffff) {
103 ERROR("Multibyte string \"%"TS"\" is too long!", name);
104 ret = WIMLIB_ERR_INVALID_UTF8_STRING;
109 FREE(*name_utf16le_ret);
110 *name_utf16le_ret = name_utf16le;
111 *name_utf16le_nbytes_ret = name_utf16le_nbytes;
116 /* Sets the name of a WIM dentry from a multibyte string. */
118 set_dentry_name(struct wim_dentry *dentry, const tchar *new_name)
121 ret = get_utf16le_name(new_name, &dentry->file_name,
122 &dentry->file_name_nbytes);
124 /* Clear the short name and recalculate the dentry length */
125 if (dentry_has_short_name(dentry)) {
126 FREE(dentry->short_name);
127 dentry->short_name = NULL;
128 dentry->short_name_nbytes = 0;
130 dentry->length = dentry_correct_length(dentry);
135 /* Returns the total length of a WIM alternate data stream entry on-disk,
136 * including the stream name, the null terminator, AND the padding after the
137 * entry to align the next ADS entry or dentry on an 8-byte boundary. */
139 ads_entry_total_length(const struct wim_ads_entry *entry)
141 u64 len = WIM_ADS_ENTRY_DISK_SIZE;
142 if (entry->stream_name_nbytes)
143 len += entry->stream_name_nbytes + 2;
144 return (len + 7) & ~7;
149 __dentry_total_length(const struct wim_dentry *dentry, u64 length)
151 const struct wim_inode *inode = dentry->d_inode;
152 for (u16 i = 0; i < inode->i_num_ads; i++)
153 length += ads_entry_total_length(&inode->i_ads_entries[i]);
154 return (length + 7) & ~7;
157 /* Calculate the aligned *total* length of an on-disk WIM dentry. This includes
158 * all alternate data streams. */
160 dentry_correct_total_length(const struct wim_dentry *dentry)
162 return __dentry_total_length(dentry,
163 dentry_correct_length_unaligned(dentry));
166 /* Like dentry_correct_total_length(), but use the existing dentry->length field
167 * instead of calculating its "correct" value. */
169 dentry_total_length(const struct wim_dentry *dentry)
171 return __dentry_total_length(dentry, dentry->length);
175 for_dentry_in_rbtree(struct rb_node *root,
176 int (*visitor)(struct wim_dentry *, void *),
180 struct rb_node *node = root;
184 list_add(&rbnode_dentry(node)->tmp_list, &stack);
185 node = node->rb_left;
187 struct list_head *next;
188 struct wim_dentry *dentry;
193 dentry = container_of(next, struct wim_dentry, tmp_list);
195 ret = visitor(dentry, arg);
198 node = dentry->rb_node.rb_right;
204 for_dentry_tree_in_rbtree_depth(struct rb_node *node,
205 int (*visitor)(struct wim_dentry*, void*),
210 ret = for_dentry_tree_in_rbtree_depth(node->rb_left,
214 ret = for_dentry_tree_in_rbtree_depth(node->rb_right,
218 ret = for_dentry_in_tree_depth(rbnode_dentry(node), visitor, arg);
226 for_dentry_tree_in_rbtree(struct rb_node *node,
227 int (*visitor)(struct wim_dentry*, void*),
232 ret = for_dentry_tree_in_rbtree(node->rb_left, visitor, arg);
235 ret = for_dentry_in_tree(rbnode_dentry(node), visitor, arg);
238 ret = for_dentry_tree_in_rbtree(node->rb_right, visitor, arg);
245 /* Calls a function on all directory entries in a WIM dentry tree. Logically,
246 * this is a pre-order traversal (the function is called on a parent dentry
247 * before its children), but sibling dentries will be visited in order as well.
250 for_dentry_in_tree(struct wim_dentry *root,
251 int (*visitor)(struct wim_dentry*, void*), void *arg)
253 int ret = visitor(root, arg);
255 ret = for_dentry_tree_in_rbtree(root->d_inode->i_children.rb_node,
262 /* Like for_dentry_in_tree(), but the visitor function is always called on a
263 * dentry's children before on itself. */
265 for_dentry_in_tree_depth(struct wim_dentry *root,
266 int (*visitor)(struct wim_dentry*, void*), void *arg)
269 ret = for_dentry_tree_in_rbtree_depth(root->d_inode->i_children.rb_node,
272 ret = visitor(root, arg);
276 /* Calculate the full path of @dentry. The full path of its parent must have
277 * already been calculated, or it must be the root dentry. */
279 calculate_dentry_full_path(struct wim_dentry *dentry)
282 u32 full_path_nbytes;
285 if (dentry->_full_path)
288 if (dentry_is_root(dentry)) {
289 full_path = TSTRDUP(T("/"));
291 return WIMLIB_ERR_NOMEM;
292 full_path_nbytes = 1 * sizeof(tchar);
294 struct wim_dentry *parent;
295 tchar *parent_full_path;
296 u32 parent_full_path_nbytes;
297 size_t filename_nbytes;
299 parent = dentry->parent;
300 if (dentry_is_root(parent)) {
301 parent_full_path = T("");
302 parent_full_path_nbytes = 0;
304 if (!parent->_full_path) {
305 ret = calculate_dentry_full_path(parent);
309 parent_full_path = parent->_full_path;
310 parent_full_path_nbytes = parent->full_path_nbytes;
313 /* Append this dentry's name as a tchar string to the full path
314 * of the parent followed by the path separator */
316 filename_nbytes = dentry->file_name_nbytes;
319 int ret = utf16le_to_tstr_nbytes(dentry->file_name,
320 dentry->file_name_nbytes,
327 full_path_nbytes = parent_full_path_nbytes + sizeof(tchar) +
329 full_path = MALLOC(full_path_nbytes + sizeof(tchar));
331 return WIMLIB_ERR_NOMEM;
332 memcpy(full_path, parent_full_path, parent_full_path_nbytes);
333 full_path[parent_full_path_nbytes / sizeof(tchar)] = T('/');
335 memcpy(&full_path[parent_full_path_nbytes / sizeof(tchar) + 1],
337 filename_nbytes + sizeof(tchar));
339 utf16le_to_tstr_buf(dentry->file_name,
340 dentry->file_name_nbytes,
341 &full_path[parent_full_path_nbytes /
345 dentry->_full_path = full_path;
346 dentry->full_path_nbytes= full_path_nbytes;
351 do_calculate_dentry_full_path(struct wim_dentry *dentry, void *_ignore)
353 return calculate_dentry_full_path(dentry);
357 calculate_dentry_tree_full_paths(struct wim_dentry *root)
359 return for_dentry_in_tree(root, do_calculate_dentry_full_path, NULL);
363 dentry_full_path(struct wim_dentry *dentry)
365 calculate_dentry_full_path(dentry);
366 return dentry->_full_path;
370 increment_subdir_offset(struct wim_dentry *dentry, void *subdir_offset_p)
372 *(u64*)subdir_offset_p += dentry_correct_total_length(dentry);
377 call_calculate_subdir_offsets(struct wim_dentry *dentry, void *subdir_offset_p)
379 calculate_subdir_offsets(dentry, subdir_offset_p);
384 * Recursively calculates the subdir offsets for a directory tree.
386 * @dentry: The root of the directory tree.
387 * @subdir_offset_p: The current subdirectory offset; i.e., the subdirectory
388 * offset for @dentry.
391 calculate_subdir_offsets(struct wim_dentry *dentry, u64 *subdir_offset_p)
393 struct rb_node *node;
395 dentry->subdir_offset = *subdir_offset_p;
396 node = dentry->d_inode->i_children.rb_node;
398 /* Advance the subdir offset by the amount of space the children
399 * of this dentry take up. */
400 for_dentry_in_rbtree(node, increment_subdir_offset, subdir_offset_p);
402 /* End-of-directory dentry on disk. */
403 *subdir_offset_p += 8;
405 /* Recursively call calculate_subdir_offsets() on all the
407 for_dentry_in_rbtree(node, call_calculate_subdir_offsets, subdir_offset_p);
409 /* On disk, childless directories have a valid subdir_offset
410 * that points to an 8-byte end-of-directory dentry. Regular
411 * files or reparse points have a subdir_offset of 0. */
412 if (dentry_is_directory(dentry))
413 *subdir_offset_p += 8;
415 dentry->subdir_offset = 0;
420 compare_utf16le_names(const utf16lechar *name1, size_t nbytes1,
421 const utf16lechar *name2, size_t nbytes2)
423 int result = memcmp(name1, name2, min(nbytes1, nbytes2));
427 return (int)nbytes1 - (int)nbytes2;
431 dentry_compare_names(const struct wim_dentry *d1, const struct wim_dentry *d2)
433 return compare_utf16le_names(d1->file_name, d1->file_name_nbytes,
434 d2->file_name, d2->file_name_nbytes);
439 get_dentry_child_with_utf16le_name(const struct wim_dentry *dentry,
440 const utf16lechar *name,
443 struct rb_node *node = dentry->d_inode->i_children.rb_node;
444 struct wim_dentry *child;
446 child = rbnode_dentry(node);
447 int result = compare_utf16le_names(name, name_nbytes,
449 child->file_name_nbytes);
451 node = node->rb_left;
453 node = node->rb_right;
460 /* Returns the child of @dentry that has the file name @name. Returns NULL if
461 * no child has the name. */
463 get_dentry_child_with_name(const struct wim_dentry *dentry, const tchar *name)
466 return get_dentry_child_with_utf16le_name(dentry, name,
467 tstrlen(name) * sizeof(tchar));
469 utf16lechar *utf16le_name;
470 size_t utf16le_name_nbytes;
472 struct wim_dentry *child;
474 ret = tstr_to_utf16le(name, tstrlen(name) * sizeof(tchar),
475 &utf16le_name, &utf16le_name_nbytes);
479 child = get_dentry_child_with_utf16le_name(dentry,
481 utf16le_name_nbytes);
488 static struct wim_dentry *
489 get_dentry_utf16le(WIMStruct *w, const utf16lechar *path,
492 struct wim_dentry *cur_dentry, *parent_dentry;
493 const utf16lechar *p, *pp;
495 cur_dentry = parent_dentry = wim_root_dentry(w);
498 while (*p == cpu_to_le16('/'))
503 while (*pp != cpu_to_le16('/') && *pp != cpu_to_le16('\0'))
506 cur_dentry = get_dentry_child_with_utf16le_name(parent_dentry, p,
507 (void*)pp - (void*)p);
508 if (cur_dentry == NULL)
511 parent_dentry = cur_dentry;
513 if (cur_dentry == NULL) {
514 if (dentry_is_directory(parent_dentry))
522 /* Returns the dentry corresponding to the @path, or NULL if there is no such
525 get_dentry(WIMStruct *w, const tchar *path)
528 return get_dentry_utf16le(w, path, tstrlen(path) * sizeof(tchar));
530 utf16lechar *path_utf16le;
531 size_t path_utf16le_nbytes;
533 struct wim_dentry *dentry;
535 ret = tstr_to_utf16le(path, tstrlen(path) * sizeof(tchar),
536 &path_utf16le, &path_utf16le_nbytes);
539 dentry = get_dentry_utf16le(w, path_utf16le, path_utf16le_nbytes);
546 wim_pathname_to_inode(WIMStruct *w, const tchar *path)
548 struct wim_dentry *dentry;
549 dentry = get_dentry(w, path);
551 return dentry->d_inode;
556 /* Takes in a path of length @len in @buf, and transforms it into a string for
557 * the path of its parent directory. */
559 to_parent_name(tchar *buf, size_t len)
561 ssize_t i = (ssize_t)len - 1;
562 while (i >= 0 && buf[i] == T('/'))
564 while (i >= 0 && buf[i] != T('/'))
566 while (i >= 0 && buf[i] == T('/'))
568 buf[i + 1] = T('\0');
571 /* Returns the dentry that corresponds to the parent directory of @path, or NULL
572 * if the dentry is not found. */
574 get_parent_dentry(WIMStruct *w, const tchar *path)
576 size_t path_len = tstrlen(path);
577 tchar buf[path_len + 1];
579 tmemcpy(buf, path, path_len + 1);
580 to_parent_name(buf, path_len);
581 return get_dentry(w, buf);
584 /* Prints the full path of a dentry. */
586 print_dentry_full_path(struct wim_dentry *dentry, void *_ignore)
588 int ret = calculate_dentry_full_path(dentry);
591 tprintf(T("%"TS"\n"), dentry->_full_path);
595 /* We want to be able to show the names of the file attribute flags that are
597 struct file_attr_flag {
601 struct file_attr_flag file_attr_flags[] = {
602 {FILE_ATTRIBUTE_READONLY, T("READONLY")},
603 {FILE_ATTRIBUTE_HIDDEN, T("HIDDEN")},
604 {FILE_ATTRIBUTE_SYSTEM, T("SYSTEM")},
605 {FILE_ATTRIBUTE_DIRECTORY, T("DIRECTORY")},
606 {FILE_ATTRIBUTE_ARCHIVE, T("ARCHIVE")},
607 {FILE_ATTRIBUTE_DEVICE, T("DEVICE")},
608 {FILE_ATTRIBUTE_NORMAL, T("NORMAL")},
609 {FILE_ATTRIBUTE_TEMPORARY, T("TEMPORARY")},
610 {FILE_ATTRIBUTE_SPARSE_FILE, T("SPARSE_FILE")},
611 {FILE_ATTRIBUTE_REPARSE_POINT, T("REPARSE_POINT")},
612 {FILE_ATTRIBUTE_COMPRESSED, T("COMPRESSED")},
613 {FILE_ATTRIBUTE_OFFLINE, T("OFFLINE")},
614 {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,T("NOT_CONTENT_INDEXED")},
615 {FILE_ATTRIBUTE_ENCRYPTED, T("ENCRYPTED")},
616 {FILE_ATTRIBUTE_VIRTUAL, T("VIRTUAL")},
619 /* Prints a directory entry. @lookup_table is a pointer to the lookup table, if
620 * available. If the dentry is unresolved and the lookup table is NULL, the
621 * lookup table entries will not be printed. Otherwise, they will be. */
623 print_dentry(struct wim_dentry *dentry, void *lookup_table)
626 struct wim_lookup_table_entry *lte;
627 const struct wim_inode *inode = dentry->d_inode;
630 tprintf(T("[DENTRY]\n"));
631 tprintf(T("Length = %"PRIu64"\n"), dentry->length);
632 tprintf(T("Attributes = 0x%x\n"), inode->i_attributes);
633 for (size_t i = 0; i < ARRAY_LEN(file_attr_flags); i++)
634 if (file_attr_flags[i].flag & inode->i_attributes)
635 tprintf(T(" FILE_ATTRIBUTE_%"TS" is set\n"),
636 file_attr_flags[i].name);
637 tprintf(T("Security ID = %d\n"), inode->i_security_id);
638 tprintf(T("Subdir offset = %"PRIu64"\n"), dentry->subdir_offset);
640 wim_timestamp_to_str(inode->i_creation_time, buf, sizeof(buf));
641 tprintf(T("Creation Time = %"TS"\n"), buf);
643 wim_timestamp_to_str(inode->i_last_access_time, buf, sizeof(buf));
644 tprintf(T("Last Access Time = %"TS"\n"), buf);
646 wim_timestamp_to_str(inode->i_last_write_time, buf, sizeof(buf));
647 tprintf(T("Last Write Time = %"TS"\n"), buf);
649 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
650 tprintf(T("Reparse Tag = 0x%"PRIx32"\n"), inode->i_reparse_tag);
651 tprintf(T("Reparse Point Flags = 0x%"PRIx16"\n"),
652 inode->i_not_rpfixed);
653 tprintf(T("Reparse Point Unknown 2 = 0x%"PRIx32"\n"),
654 inode->i_rp_unknown_2);
656 tprintf(T("Reparse Point Unknown 1 = 0x%"PRIx32"\n"),
657 inode->i_rp_unknown_1);
658 tprintf(T("Hard Link Group = 0x%"PRIx64"\n"), inode->i_ino);
659 tprintf(T("Hard Link Group Size = %"PRIu32"\n"), inode->i_nlink);
660 tprintf(T("Number of Alternate Data Streams = %hu\n"), inode->i_num_ads);
661 if (dentry_has_long_name(dentry))
662 wimlib_printf(T("Filename = \"%"WS"\"\n"), dentry->file_name);
663 if (dentry_has_short_name(dentry))
664 wimlib_printf(T("Short Name \"%"WS"\"\n"), dentry->short_name);
665 if (dentry->_full_path)
666 tprintf(T("Full Path = \"%"TS"\"\n"), dentry->_full_path);
668 lte = inode_stream_lte(dentry->d_inode, 0, lookup_table);
670 print_lookup_table_entry(lte, stdout);
672 hash = inode_stream_hash(inode, 0);
674 tprintf(T("Hash = 0x"));
675 print_hash(hash, stdout);
680 for (u16 i = 0; i < inode->i_num_ads; i++) {
681 tprintf(T("[Alternate Stream Entry %u]\n"), i);
682 wimlib_printf(T("Name = \"%"WS"\"\n"),
683 inode->i_ads_entries[i].stream_name);
684 tprintf(T("Name Length (UTF16 bytes) = %hu\n"),
685 inode->i_ads_entries[i].stream_name_nbytes);
686 hash = inode_stream_hash(inode, i + 1);
688 tprintf(T("Hash = 0x"));
689 print_hash(hash, stdout);
692 print_lookup_table_entry(inode_stream_lte(inode, i + 1, lookup_table),
698 /* Initializations done on every `struct wim_dentry'. */
700 dentry_common_init(struct wim_dentry *dentry)
702 memset(dentry, 0, sizeof(struct wim_dentry));
708 struct wim_inode *inode = CALLOC(1, sizeof(struct wim_inode));
710 inode->i_security_id = -1;
712 inode->i_next_stream_id = 1;
713 inode->i_not_rpfixed = 1;
715 if (pthread_mutex_init(&inode->i_mutex, NULL) != 0) {
716 ERROR_WITH_ERRNO("Error initializing mutex");
721 INIT_LIST_HEAD(&inode->i_dentry);
726 static struct wim_inode *
729 struct wim_inode *inode = new_timeless_inode();
731 u64 now = get_wim_timestamp();
732 inode->i_creation_time = now;
733 inode->i_last_access_time = now;
734 inode->i_last_write_time = now;
739 /* Creates an unlinked directory entry. */
741 new_dentry(const tchar *name, struct wim_dentry **dentry_ret)
743 struct wim_dentry *dentry;
746 dentry = MALLOC(sizeof(struct wim_dentry));
748 return WIMLIB_ERR_NOMEM;
750 dentry_common_init(dentry);
751 ret = set_dentry_name(dentry, name);
753 dentry->parent = dentry;
754 *dentry_ret = dentry;
757 ERROR("Failed to set name on new dentry with name \"%"TS"\"",
765 __new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret,
768 struct wim_dentry *dentry;
771 ret = new_dentry(name, &dentry);
776 dentry->d_inode = new_timeless_inode();
778 dentry->d_inode = new_inode();
779 if (!dentry->d_inode) {
781 return WIMLIB_ERR_NOMEM;
784 inode_add_dentry(dentry, dentry->d_inode);
785 *dentry_ret = dentry;
790 new_dentry_with_timeless_inode(const tchar *name, struct wim_dentry **dentry_ret)
792 return __new_dentry_with_inode(name, dentry_ret, true);
796 new_dentry_with_inode(const tchar *name, struct wim_dentry **dentry_ret)
798 return __new_dentry_with_inode(name, dentry_ret, false);
803 init_ads_entry(struct wim_ads_entry *ads_entry, const void *name,
804 size_t name_nbytes, bool is_utf16le)
807 memset(ads_entry, 0, sizeof(*ads_entry));
810 utf16lechar *p = MALLOC(name_nbytes + sizeof(utf16lechar));
812 return WIMLIB_ERR_NOMEM;
813 memcpy(p, name, name_nbytes);
814 p[name_nbytes / 2] = 0;
815 ads_entry->stream_name = p;
816 ads_entry->stream_name_nbytes = name_nbytes;
818 if (name && *(const tchar*)name != T('\0')) {
819 ret = get_utf16le_name(name, &ads_entry->stream_name,
820 &ads_entry->stream_name_nbytes);
827 destroy_ads_entry(struct wim_ads_entry *ads_entry)
829 FREE(ads_entry->stream_name);
832 /* Frees an inode. */
834 free_inode(struct wim_inode *inode)
837 if (inode->i_ads_entries) {
838 for (u16 i = 0; i < inode->i_num_ads; i++)
839 destroy_ads_entry(&inode->i_ads_entries[i]);
840 FREE(inode->i_ads_entries);
843 wimlib_assert(inode->i_num_opened_fds == 0);
845 pthread_mutex_destroy(&inode->i_mutex);
846 if (inode->i_hlist.pprev)
847 hlist_del(&inode->i_hlist);
849 FREE(inode->i_extracted_file);
854 /* Decrements link count on an inode and frees it if the link count reaches 0.
857 put_inode(struct wim_inode *inode)
859 wimlib_assert(inode->i_nlink != 0);
860 if (--inode->i_nlink == 0) {
862 if (inode->i_num_opened_fds == 0)
870 /* Frees a WIM dentry.
872 * The corresponding inode (if any) is freed only if its link count is
876 free_dentry(struct wim_dentry *dentry)
878 FREE(dentry->file_name);
879 FREE(dentry->short_name);
880 FREE(dentry->_full_path);
882 put_inode(dentry->d_inode);
886 /* This function is passed as an argument to for_dentry_in_tree_depth() in order
887 * to free a directory tree. */
889 do_free_dentry(struct wim_dentry *dentry, void *__lookup_table)
891 struct wim_lookup_table *lookup_table = __lookup_table;
895 struct wim_lookup_table_entry *lte;
896 struct wim_inode *inode = dentry->d_inode;
897 wimlib_assert(inode->i_nlink != 0);
898 for (i = 0; i <= inode->i_num_ads; i++) {
899 lte = inode_stream_lte(inode, i, lookup_table);
901 lte_decrement_refcnt(lte, lookup_table);
909 * Unlinks and frees a dentry tree.
911 * @root: The root of the tree.
912 * @lookup_table: The lookup table for dentries. If non-NULL, the
913 * reference counts in the lookup table for the lookup
914 * table entries corresponding to the dentries will be
918 free_dentry_tree(struct wim_dentry *root, struct wim_lookup_table *lookup_table)
921 for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
925 * Links a dentry into the directory tree.
927 * @parent: The dentry that will be the parent of @dentry.
928 * @dentry: The dentry to link.
931 dentry_add_child(struct wim_dentry * restrict parent,
932 struct wim_dentry * restrict child)
934 wimlib_assert(dentry_is_directory(parent));
936 struct rb_root *root = &parent->d_inode->i_children;
937 struct rb_node **new = &(root->rb_node);
938 struct rb_node *rb_parent = NULL;
941 struct wim_dentry *this = rbnode_dentry(*new);
942 int result = dentry_compare_names(child, this);
947 new = &((*new)->rb_left);
949 new = &((*new)->rb_right);
953 child->parent = parent;
954 rb_link_node(&child->rb_node, rb_parent, new);
955 rb_insert_color(&child->rb_node, root);
959 /* Unlink a WIM dentry from the directory entry tree. */
961 unlink_dentry(struct wim_dentry *dentry)
963 struct wim_dentry *parent = dentry->parent;
964 if (parent == dentry)
966 rb_erase(&dentry->rb_node, &parent->d_inode->i_children);
970 * Returns the alternate data stream entry belonging to @inode that has the
971 * stream name @stream_name.
973 struct wim_ads_entry *
974 inode_get_ads_entry(struct wim_inode *inode, const tchar *stream_name,
977 if (inode->i_num_ads == 0) {
980 size_t stream_name_utf16le_nbytes;
982 struct wim_ads_entry *result;
985 const utf16lechar *stream_name_utf16le;
987 stream_name_utf16le = stream_name;
988 stream_name_utf16le_nbytes = tstrlen(stream_name) * sizeof(tchar);
990 utf16lechar *stream_name_utf16le;
993 int ret = tstr_to_utf16le(stream_name,
994 tstrlen(stream_name) *
996 &stream_name_utf16le,
997 &stream_name_utf16le_nbytes);
1005 if (ads_entry_has_name(&inode->i_ads_entries[i],
1006 stream_name_utf16le,
1007 stream_name_utf16le_nbytes))
1011 result = &inode->i_ads_entries[i];
1014 } while (++i != inode->i_num_ads);
1015 #if !TCHAR_IS_UTF16LE
1016 FREE(stream_name_utf16le);
1022 static struct wim_ads_entry *
1023 do_inode_add_ads(struct wim_inode *inode, const void *stream_name,
1024 size_t stream_name_nbytes, bool is_utf16le)
1027 struct wim_ads_entry *ads_entries;
1028 struct wim_ads_entry *new_entry;
1030 if (inode->i_num_ads >= 0xfffe) {
1031 ERROR("Too many alternate data streams in one inode!");
1034 num_ads = inode->i_num_ads + 1;
1035 ads_entries = REALLOC(inode->i_ads_entries,
1036 num_ads * sizeof(inode->i_ads_entries[0]));
1038 ERROR("Failed to allocate memory for new alternate data stream");
1041 inode->i_ads_entries = ads_entries;
1043 new_entry = &inode->i_ads_entries[num_ads - 1];
1044 if (init_ads_entry(new_entry, stream_name, stream_name_nbytes, is_utf16le))
1046 new_entry->stream_id = inode->i_next_stream_id++;
1047 inode->i_num_ads = num_ads;
1051 struct wim_ads_entry *
1052 inode_add_ads_utf16le(struct wim_inode *inode,
1053 const utf16lechar *stream_name,
1054 size_t stream_name_nbytes)
1056 DEBUG("Add alternate data stream \"%"WS"\"", stream_name);
1057 return do_inode_add_ads(inode, stream_name, stream_name_nbytes, true);
1061 * Add an alternate stream entry to a WIM inode and return a pointer to it, or
1062 * NULL if memory could not be allocated.
1064 struct wim_ads_entry *
1065 inode_add_ads(struct wim_inode *inode, const tchar *stream_name)
1067 DEBUG("Add alternate data stream \"%"TS"\"", stream_name);
1068 return do_inode_add_ads(inode, stream_name,
1069 tstrlen(stream_name) * sizeof(tchar),
1074 inode_add_ads_with_data(struct wim_inode *inode, const tchar *name,
1075 const void *value, size_t size,
1076 struct wim_lookup_table *lookup_table)
1078 int ret = WIMLIB_ERR_NOMEM;
1079 struct wim_ads_entry *new_ads_entry;
1080 struct wim_lookup_table_entry *existing_lte;
1081 struct wim_lookup_table_entry *lte;
1082 u8 value_hash[SHA1_HASH_SIZE];
1084 wimlib_assert(inode->i_resolved);
1085 new_ads_entry = inode_add_ads(inode, name);
1088 sha1_buffer((const u8*)value, size, value_hash);
1089 existing_lte = __lookup_resource(lookup_table, value_hash);
1095 lte = new_lookup_table_entry();
1097 goto out_remove_ads_entry;
1098 value_copy = MALLOC(size);
1101 goto out_remove_ads_entry;
1103 memcpy(value_copy, value, size);
1104 lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER;
1105 lte->attached_buffer = value_copy;
1106 lte->resource_entry.original_size = size;
1107 lte->resource_entry.size = size;
1108 copy_hash(lte->hash, value_hash);
1109 lookup_table_insert(lookup_table, lte);
1111 new_ads_entry->lte = lte;
1114 out_remove_ads_entry:
1115 inode_remove_ads(inode, new_ads_entry - inode->i_ads_entries,
1121 /* Set the unnamed stream of a WIM inode, given a data buffer containing the
1122 * stream contents. */
1124 inode_set_unnamed_stream(struct wim_inode *inode, const void *data, size_t len,
1125 struct wim_lookup_table *lookup_table)
1127 struct wim_lookup_table_entry *lte, *existing_lte;
1128 u8 hash[SHA1_HASH_SIZE];
1131 sha1_buffer(data, len, hash);
1133 existing_lte = __lookup_resource(lookup_table, hash);
1136 wimlib_assert(wim_resource_size(existing_lte) == len);
1142 lte = new_lookup_table_entry();
1144 return WIMLIB_ERR_NOMEM;
1147 free_lookup_table_entry(lte);
1148 return WIMLIB_ERR_NOMEM;
1150 memcpy(buf, data, len);
1151 lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER;
1152 lte->attached_buffer = buf;
1153 lte->resource_entry.original_size = len;
1154 copy_hash(lte->hash, hash);
1155 lookup_table_insert(lookup_table, lte);
1158 inode->i_resolved = 1;
1162 /* Remove an alternate data stream from a WIM inode */
1164 inode_remove_ads(struct wim_inode *inode, u16 idx,
1165 struct wim_lookup_table *lookup_table)
1167 struct wim_ads_entry *ads_entry;
1168 struct wim_lookup_table_entry *lte;
1170 wimlib_assert(idx < inode->i_num_ads);
1171 wimlib_assert(inode->i_resolved);
1173 ads_entry = &inode->i_ads_entries[idx];
1175 DEBUG("Remove alternate data stream \"%"WS"\"", ads_entry->stream_name);
1177 lte = ads_entry->lte;
1179 lte_decrement_refcnt(lte, lookup_table);
1181 destroy_ads_entry(ads_entry);
1183 memmove(&inode->i_ads_entries[idx],
1184 &inode->i_ads_entries[idx + 1],
1185 (inode->i_num_ads - idx - 1) * sizeof(inode->i_ads_entries[0]));
1191 inode_get_unix_data(const struct wim_inode *inode,
1192 struct wimlib_unix_data *unix_data,
1193 u16 *stream_idx_ret)
1195 const struct wim_ads_entry *ads_entry;
1196 const struct wim_lookup_table_entry *lte;
1200 wimlib_assert(inode->i_resolved);
1202 ads_entry = inode_get_ads_entry((struct wim_inode*)inode,
1203 WIMLIB_UNIX_DATA_TAG, NULL);
1205 return NO_UNIX_DATA;
1208 *stream_idx_ret = ads_entry - inode->i_ads_entries;
1210 lte = ads_entry->lte;
1212 return NO_UNIX_DATA;
1214 size = wim_resource_size(lte);
1215 if (size != sizeof(struct wimlib_unix_data))
1216 return BAD_UNIX_DATA;
1218 ret = read_full_resource_into_buf(lte, unix_data, true);
1222 if (unix_data->version != 0)
1223 return BAD_UNIX_DATA;
1228 inode_set_unix_data(struct wim_inode *inode, uid_t uid, gid_t gid, mode_t mode,
1229 struct wim_lookup_table *lookup_table, int which)
1231 struct wimlib_unix_data unix_data;
1233 bool have_good_unix_data = false;
1234 bool have_unix_data = false;
1237 if (!(which & UNIX_DATA_CREATE)) {
1238 ret = inode_get_unix_data(inode, &unix_data, &stream_idx);
1239 if (ret == 0 || ret == BAD_UNIX_DATA || ret > 0)
1240 have_unix_data = true;
1242 have_good_unix_data = true;
1244 unix_data.version = 0;
1245 if (which & UNIX_DATA_UID || !have_good_unix_data)
1246 unix_data.uid = uid;
1247 if (which & UNIX_DATA_GID || !have_good_unix_data)
1248 unix_data.gid = gid;
1249 if (which & UNIX_DATA_MODE || !have_good_unix_data)
1250 unix_data.mode = mode;
1251 ret = inode_add_ads_with_data(inode, WIMLIB_UNIX_DATA_TAG,
1253 sizeof(struct wimlib_unix_data),
1255 if (ret == 0 && have_unix_data)
1256 inode_remove_ads(inode, stream_idx, lookup_table);
1259 #endif /* !__WIN32__ */
1261 /* Replace weird characters in filenames and alternate data stream names.
1263 * In particular we do not want the path separator to appear in any names, as
1264 * that would make it possible for a "malicious" WIM to extract itself to any
1265 * location it wanted to. */
1267 replace_forbidden_characters(utf16lechar *name)
1271 for (p = name; *p; p++) {
1273 if (wcschr(L"<>:\"/\\|?*", (wchar_t)*p))
1279 *p = cpu_to_le16(0xfffd);
1284 WARNING("File, directory, or stream name \"%"WS"\"\n"
1285 " contains forbidden characters; "
1286 "substituting replacement characters.",
1295 * Reads the alternate data stream entries of a WIM dentry.
1297 * @p: Pointer to buffer that starts with the first alternate stream entry.
1299 * @inode: Inode to load the alternate data streams into.
1300 * @inode->i_num_ads must have been set to the number of
1301 * alternate data streams that are expected.
1303 * @remaining_size: Number of bytes of data remaining in the buffer pointed
1306 * The format of the on-disk alternate stream entries is as follows:
1308 * struct wim_ads_entry_on_disk {
1309 * u64 length; // Length of the entry, in bytes. This includes
1310 * all fields (including the stream name and
1311 * null terminator if present, AND the padding!).
1312 * u64 reserved; // Seems to be unused
1313 * u8 hash[20]; // SHA1 message digest of the uncompressed stream
1314 * u16 stream_name_len; // Length of the stream name, in bytes
1315 * char stream_name[]; // Stream name in UTF-16LE, @stream_name_len bytes long,
1316 * not including null terminator
1317 * u16 zero; // UTF-16 null terminator for the stream name, NOT
1318 * included in @stream_name_len. Based on what
1319 * I've observed from filenames in dentries,
1320 * this field should not exist when
1321 * (@stream_name_len == 0), but you can't
1322 * actually tell because of the padding anyway
1323 * (provided that the padding is zeroed, which
1324 * it always seems to be).
1325 * char padding[]; // Padding to make the size a multiple of 8 bytes.
1328 * In addition, the entries are 8-byte aligned.
1330 * Return 0 on success or nonzero on failure. On success, inode->i_ads_entries
1331 * is set to an array of `struct wim_ads_entry's of length inode->i_num_ads. On
1332 * failure, @inode is not modified.
1335 read_ads_entries(const u8 *p, struct wim_inode *inode, u64 remaining_size)
1338 struct wim_ads_entry *ads_entries;
1341 num_ads = inode->i_num_ads;
1342 ads_entries = CALLOC(num_ads, sizeof(inode->i_ads_entries[0]));
1344 ERROR("Could not allocate memory for %"PRIu16" "
1345 "alternate data stream entries", num_ads);
1346 return WIMLIB_ERR_NOMEM;
1349 for (u16 i = 0; i < num_ads; i++) {
1350 struct wim_ads_entry *cur_entry;
1352 u64 length_no_padding;
1354 const u8 *p_save = p;
1356 cur_entry = &ads_entries[i];
1359 ads_entries[i].stream_id = i + 1;
1362 /* Read the base stream entry, excluding the stream name. */
1363 if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
1364 ERROR("Stream entries go past end of metadata resource");
1365 ERROR("(remaining_size = %"PRIu64")", remaining_size);
1366 ret = WIMLIB_ERR_INVALID_DENTRY;
1367 goto out_free_ads_entries;
1370 p = get_u64(p, &length);
1371 p = get_u64(p, &cur_entry->unused);
1372 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
1373 p = get_u16(p, &cur_entry->stream_name_nbytes);
1375 cur_entry->stream_name = NULL;
1377 /* Length including neither the null terminator nor the padding
1379 length_no_padding = WIM_ADS_ENTRY_DISK_SIZE +
1380 cur_entry->stream_name_nbytes;
1382 /* Length including the null terminator and the padding */
1383 total_length = ((length_no_padding + 2) + 7) & ~7;
1385 wimlib_assert(total_length == ads_entry_total_length(cur_entry));
1387 if (remaining_size < length_no_padding) {
1388 ERROR("Stream entries go past end of metadata resource");
1389 ERROR("(remaining_size = %"PRIu64" bytes, "
1390 "length_no_padding = %"PRIu64" bytes)",
1391 remaining_size, length_no_padding);
1392 ret = WIMLIB_ERR_INVALID_DENTRY;
1393 goto out_free_ads_entries;
1396 /* The @length field in the on-disk ADS entry is expected to be
1397 * equal to @total_length, which includes all of the entry and
1398 * the padding that follows it to align the next ADS entry to an
1399 * 8-byte boundary. However, to be safe, we'll accept the
1400 * length field as long as it's not less than the un-padded
1401 * total length and not more than the padded total length. */
1402 if (length < length_no_padding || length > total_length) {
1403 ERROR("Stream entry has unexpected length "
1404 "field (length field = %"PRIu64", "
1405 "unpadded total length = %"PRIu64", "
1406 "padded total length = %"PRIu64")",
1407 length, length_no_padding, total_length);
1408 ret = WIMLIB_ERR_INVALID_DENTRY;
1409 goto out_free_ads_entries;
1412 if (cur_entry->stream_name_nbytes) {
1413 cur_entry->stream_name = MALLOC(cur_entry->stream_name_nbytes + 2);
1414 if (!cur_entry->stream_name) {
1415 ret = WIMLIB_ERR_NOMEM;
1416 goto out_free_ads_entries;
1418 get_bytes(p, cur_entry->stream_name_nbytes,
1419 cur_entry->stream_name);
1420 cur_entry->stream_name[cur_entry->stream_name_nbytes / 2] = 0;
1421 replace_forbidden_characters(cur_entry->stream_name);
1423 /* It's expected that the size of every ADS entry is a multiple
1424 * of 8. However, to be safe, I'm allowing the possibility of
1425 * an ADS entry at the very end of the metadata resource ending
1426 * un-aligned. So although we still need to increment the input
1427 * pointer by @total_length to reach the next ADS entry, it's
1428 * possible that less than @total_length is actually remaining
1429 * in the metadata resource. We should set the remaining size to
1430 * 0 bytes if this happens. */
1431 p = p_save + total_length;
1432 if (remaining_size < total_length)
1435 remaining_size -= total_length;
1437 inode->i_ads_entries = ads_entries;
1439 inode->i_next_stream_id = inode->i_num_ads + 1;
1442 out_free_ads_entries:
1443 for (u16 i = 0; i < num_ads; i++)
1444 destroy_ads_entry(&ads_entries[i]);
1450 * Reads a WIM directory entry, including all alternate data stream entries that
1451 * follow it, from the WIM image's metadata resource.
1453 * @metadata_resource: Buffer containing the uncompressed metadata resource.
1454 * @metadata_resource_len: Length of the metadata resource.
1455 * @offset: Offset of this directory entry in the metadata resource.
1456 * @dentry: A `struct wim_dentry' that will be filled in by this function.
1458 * Return 0 on success or nonzero on failure. On failure, @dentry will have
1459 * been modified, but it will not be left with pointers to any allocated
1460 * buffers. On success, the dentry->length field must be examined. If zero,
1461 * this was a special "end of directory" dentry and not a real dentry. If
1462 * nonzero, this was a real dentry.
1465 read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
1466 u64 offset, struct wim_dentry *dentry)
1469 u64 calculated_size;
1470 utf16lechar *file_name = NULL;
1471 utf16lechar *short_name = NULL;
1472 u16 short_name_nbytes;
1473 u16 file_name_nbytes;
1475 struct wim_inode *inode = NULL;
1477 dentry_common_init(dentry);
1479 /*Make sure the dentry really fits into the metadata resource.*/
1480 if (offset + 8 > metadata_resource_len || offset + 8 < offset) {
1481 ERROR("Directory entry starting at %"PRIu64" ends past the "
1482 "end of the metadata resource (size %"PRIu64")",
1483 offset, metadata_resource_len);
1484 return WIMLIB_ERR_INVALID_DENTRY;
1487 /* Before reading the whole dentry, we need to read just the length.
1488 * This is because a dentry of length 8 (that is, just the length field)
1489 * terminates the list of sibling directory entries. */
1491 p = get_u64(&metadata_resource[offset], &dentry->length);
1493 /* A zero length field (really a length of 8, since that's how big the
1494 * directory entry is...) indicates that this is the end of directory
1495 * dentry. We do not read it into memory as an actual dentry, so just
1496 * return successfully in that case. */
1497 if (dentry->length == 0)
1500 /* If the dentry does not overflow the metadata resource buffer and is
1501 * not too short, read the rest of it (excluding the alternate data
1502 * streams, but including the file name and short name variable-length
1503 * fields) into memory. */
1504 if (offset + dentry->length >= metadata_resource_len
1505 || offset + dentry->length < offset)
1507 ERROR("Directory entry at offset %"PRIu64" and with size "
1508 "%"PRIu64" ends past the end of the metadata resource "
1510 offset, dentry->length, metadata_resource_len);
1511 return WIMLIB_ERR_INVALID_DENTRY;
1514 if (dentry->length < WIM_DENTRY_DISK_SIZE) {
1515 ERROR("Directory entry has invalid length of %"PRIu64" bytes",
1517 return WIMLIB_ERR_INVALID_DENTRY;
1520 inode = new_timeless_inode();
1522 return WIMLIB_ERR_NOMEM;
1524 p = get_u32(p, &inode->i_attributes);
1525 p = get_u32(p, (u32*)&inode->i_security_id);
1526 p = get_u64(p, &dentry->subdir_offset);
1528 p = get_u64(p, &inode->i_unused_1);
1529 p = get_u64(p, &inode->i_unused_2);
1531 p = get_u64(p, &inode->i_creation_time);
1532 p = get_u64(p, &inode->i_last_access_time);
1533 p = get_u64(p, &inode->i_last_write_time);
1535 p = get_bytes(p, SHA1_HASH_SIZE, inode->i_hash);
1537 /* I don't know what's going on here. It seems like M$ screwed up the
1538 * reparse points, then put the fields in the same place and didn't
1540 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1541 p = get_u32(p, &inode->i_rp_unknown_1);
1542 p = get_u32(p, &inode->i_reparse_tag);
1543 p = get_u16(p, &inode->i_rp_unknown_2);
1544 p = get_u16(p, &inode->i_not_rpfixed);
1546 p = get_u32(p, &inode->i_rp_unknown_1);
1547 p = get_u64(p, &inode->i_ino);
1550 /* By the way, the reparse_reserved field does not actually exist (at
1551 * least when the file is not a reparse point) */
1553 p = get_u16(p, &inode->i_num_ads);
1555 p = get_u16(p, &short_name_nbytes);
1556 p = get_u16(p, &file_name_nbytes);
1558 /* We now know the length of the file name and short name. Make sure
1559 * the length of the dentry is large enough to actually hold them.
1561 * The calculated length here is unaligned to allow for the possibility
1562 * that the dentry->length names an unaligned length, although this
1563 * would be unexpected. */
1564 calculated_size = __dentry_correct_length_unaligned(file_name_nbytes,
1567 if (dentry->length < calculated_size) {
1568 ERROR("Unexpected end of directory entry! (Expected "
1569 "at least %"PRIu64" bytes, got %"PRIu64" bytes. "
1570 "short_name_nbytes = %hu, file_name_nbytes = %hu)",
1571 calculated_size, dentry->length,
1572 short_name_nbytes, file_name_nbytes);
1573 ret = WIMLIB_ERR_INVALID_DENTRY;
1574 goto out_free_inode;
1577 /* Read the filename if present. Note: if the filename is empty, there
1578 * is no null terminator following it. */
1579 if (file_name_nbytes) {
1580 file_name = MALLOC(file_name_nbytes + 2);
1582 ERROR("Failed to allocate %d bytes for dentry file name",
1583 file_name_nbytes + 2);
1584 ret = WIMLIB_ERR_NOMEM;
1585 goto out_free_inode;
1587 p = get_bytes(p, file_name_nbytes + 2, file_name);
1588 if (file_name[file_name_nbytes / 2] != 0) {
1589 file_name[file_name_nbytes / 2] = 0;
1590 WARNING("File name in WIM dentry \"%"WS"\" is not "
1591 "null-terminated!", file_name);
1593 replace_forbidden_characters(file_name);
1596 /* Align the calculated size */
1597 calculated_size = (calculated_size + 7) & ~7;
1599 if (dentry->length > calculated_size) {
1600 /* Weird; the dentry says it's longer than it should be. Note
1601 * that the length field does NOT include the size of the
1602 * alternate stream entries. */
1604 /* Strangely, some directory entries inexplicably have a little
1605 * over 70 bytes of extra data. The exact amount of data seems
1606 * to be 72 bytes, but it is aligned on the next 8-byte
1607 * boundary. It does NOT seem to be alternate data stream
1608 * entries. Here's an example of the aligned data:
1610 * 01000000 40000000 6c786bba c58ede11 b0bb0026 1870892a b6adb76f
1611 * e63a3e46 8fca8653 0d2effa1 6c786bba c58ede11 b0bb0026 1870892a
1612 * 00000000 00000000 00000000 00000000
1614 * Here's one interpretation of how the data is laid out.
1617 * u32 field1; (always 0x00000001)
1618 * u32 field2; (always 0x40000000)
1619 * u8 data[48]; (???)
1620 * u64 reserved1; (always 0)
1621 * u64 reserved2; (always 0)
1623 /*DEBUG("Dentry for file or directory `%"WS"' has %"PRIu64" "*/
1624 /*"extra bytes of data", file_name,*/
1625 /*dentry->length - calculated_size);*/
1628 /* Read the short filename if present. Note: if there is no short
1629 * filename, there is no null terminator following it. */
1630 if (short_name_nbytes) {
1631 short_name = MALLOC(short_name_nbytes + 2);
1633 ERROR("Failed to allocate %d bytes for dentry short name",
1634 short_name_nbytes + 2);
1635 ret = WIMLIB_ERR_NOMEM;
1636 goto out_free_file_name;
1638 p = get_bytes(p, short_name_nbytes + 2, short_name);
1639 if (short_name[short_name_nbytes / 2] != 0) {
1640 short_name[short_name_nbytes / 2] = 0;
1641 WARNING("Short name in WIM dentry \"%"WS"\" is not "
1642 "null-terminated!", file_name);
1644 replace_forbidden_characters(short_name);
1648 * Read the alternate data streams, if present. dentry->num_ads tells
1649 * us how many they are, and they will directly follow the dentry
1652 * Note that each alternate data stream entry begins on an 8-byte
1653 * aligned boundary, and the alternate data stream entries are NOT
1654 * included in the dentry->length field for some reason.
1656 if (inode->i_num_ads != 0) {
1658 /* Trying different lengths is just a hack to make sure we have
1659 * a chance of reading the ADS entries correctly despite the
1660 * poor documentation. */
1662 if (calculated_size != dentry->length) {
1663 WARNING("Trying calculated dentry length (%"PRIu64") "
1664 "instead of dentry->length field (%"PRIu64") "
1665 "to read ADS entries",
1666 calculated_size, dentry->length);
1668 u64 lengths_to_try[3] = {calculated_size,
1669 (dentry->length + 7) & ~7,
1671 ret = WIMLIB_ERR_INVALID_DENTRY;
1672 for (size_t i = 0; i < ARRAY_LEN(lengths_to_try); i++) {
1673 if (lengths_to_try[i] > metadata_resource_len - offset)
1675 ret = read_ads_entries(&metadata_resource[offset + lengths_to_try[i]],
1677 metadata_resource_len - offset - lengths_to_try[i]);
1681 ERROR("Failed to read alternate data stream "
1682 "entries of WIM dentry \"%"WS"\"", file_name);
1683 goto out_free_short_name;
1686 /* We've read all the data for this dentry. Set the names and their
1687 * lengths, and we've done. */
1688 dentry->d_inode = inode;
1689 dentry->file_name = file_name;
1690 dentry->short_name = short_name;
1691 dentry->file_name_nbytes = file_name_nbytes;
1692 dentry->short_name_nbytes = short_name_nbytes;
1694 out_free_short_name:
1703 /* Reads the children of a dentry, and all their children, ..., etc. from the
1704 * metadata resource and into the dentry tree.
1706 * @metadata_resource: An array that contains the uncompressed metadata
1707 * resource for the WIM file.
1709 * @metadata_resource_len: The length of the uncompressed metadata resource, in
1712 * @dentry: A pointer to a `struct wim_dentry' that is the root of the directory
1713 * tree and has already been read from the metadata resource. It
1714 * does not need to be the real root because this procedure is
1715 * called recursively.
1717 * Returns zero on success; nonzero on failure.
1720 read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
1721 struct wim_dentry *dentry)
1723 u64 cur_offset = dentry->subdir_offset;
1724 struct wim_dentry *child;
1725 struct wim_dentry cur_child;
1729 * If @dentry has no child dentries, nothing more needs to be done for
1730 * this branch. This is the case for regular files, symbolic links, and
1731 * *possibly* empty directories (although an empty directory may also
1732 * have one child dentry that is the special end-of-directory dentry)
1734 if (cur_offset == 0)
1737 /* Find and read all the children of @dentry. */
1740 /* Read next child of @dentry into @cur_child. */
1741 ret = read_dentry(metadata_resource, metadata_resource_len,
1742 cur_offset, &cur_child);
1746 /* Check for end of directory. */
1747 if (cur_child.length == 0)
1750 /* Not end of directory. Allocate this child permanently and
1751 * link it to the parent and previous child. */
1752 child = MALLOC(sizeof(struct wim_dentry));
1754 ERROR("Failed to allocate %zu bytes for new dentry",
1755 sizeof(struct wim_dentry));
1756 ret = WIMLIB_ERR_NOMEM;
1759 memcpy(child, &cur_child, sizeof(struct wim_dentry));
1760 dentry_add_child(dentry, child);
1761 inode_add_dentry(child, child->d_inode);
1763 /* If there are children of this child, call this procedure
1765 if (child->subdir_offset != 0) {
1766 ret = read_dentry_tree(metadata_resource,
1767 metadata_resource_len, child);
1772 /* Advance to the offset of the next child. Note: We need to
1773 * advance by the TOTAL length of the dentry, not by the length
1774 * child->length, which although it does take into account the
1775 * padding, it DOES NOT take into account alternate stream
1777 cur_offset += dentry_total_length(child);
1783 * Writes a WIM dentry to an output buffer.
1785 * @dentry: The dentry structure.
1786 * @p: The memory location to write the data to.
1787 * @return: Pointer to the byte after the last byte we wrote as part of the
1791 write_dentry(const struct wim_dentry *dentry, u8 *p)
1795 const struct wim_inode *inode = dentry->d_inode;
1797 /* We calculate the correct length of the dentry ourselves because the
1798 * dentry->length field may been set to an unexpected value from when we
1799 * read the dentry in (for example, there may have been unknown data
1800 * appended to the end of the dentry...) */
1801 u64 length = dentry_correct_length(dentry);
1803 p = put_u64(p, length);
1804 p = put_u32(p, inode->i_attributes);
1805 p = put_u32(p, inode->i_security_id);
1806 p = put_u64(p, dentry->subdir_offset);
1807 p = put_u64(p, inode->i_unused_1);
1808 p = put_u64(p, inode->i_unused_2);
1809 p = put_u64(p, inode->i_creation_time);
1810 p = put_u64(p, inode->i_last_access_time);
1811 p = put_u64(p, inode->i_last_write_time);
1812 hash = inode_stream_hash(inode, 0);
1813 p = put_bytes(p, SHA1_HASH_SIZE, hash);
1814 if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1815 p = put_u32(p, inode->i_rp_unknown_1);
1816 p = put_u32(p, inode->i_reparse_tag);
1817 p = put_u16(p, inode->i_rp_unknown_2);
1818 p = put_u16(p, inode->i_not_rpfixed);
1821 p = put_u32(p, inode->i_rp_unknown_1);
1822 if (inode->i_nlink == 1)
1825 link_group_id = inode->i_ino;
1826 p = put_u64(p, link_group_id);
1828 p = put_u16(p, inode->i_num_ads);
1829 p = put_u16(p, dentry->short_name_nbytes);
1830 p = put_u16(p, dentry->file_name_nbytes);
1831 if (dentry_has_long_name(dentry)) {
1832 p = put_bytes(p, dentry->file_name_nbytes + 2,
1835 if (dentry_has_short_name(dentry)) {
1836 p = put_bytes(p, dentry->short_name_nbytes + 2,
1837 dentry->short_name);
1840 /* Align to 8-byte boundary */
1841 wimlib_assert(length >= (p - orig_p) && length - (p - orig_p) <= 7);
1842 p = put_zeroes(p, length - (p - orig_p));
1844 /* Write the alternate data streams, if there are any. Please see
1845 * read_ads_entries() for comments about the format of the on-disk
1846 * alternate data stream entries. */
1847 for (u16 i = 0; i < inode->i_num_ads; i++) {
1848 p = put_u64(p, ads_entry_total_length(&inode->i_ads_entries[i]));
1849 p = put_u64(p, inode->i_ads_entries[i].unused);
1850 hash = inode_stream_hash(inode, i + 1);
1851 p = put_bytes(p, SHA1_HASH_SIZE, hash);
1852 p = put_u16(p, inode->i_ads_entries[i].stream_name_nbytes);
1853 if (inode->i_ads_entries[i].stream_name_nbytes) {
1855 inode->i_ads_entries[i].stream_name_nbytes + 2,
1856 inode->i_ads_entries[i].stream_name);
1858 p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1860 wimlib_assert(p - orig_p == __dentry_total_length(dentry, length));
1865 write_dentry_cb(struct wim_dentry *dentry, void *_p)
1868 *p = write_dentry(dentry, *p);
1873 write_dentry_tree_recursive(const struct wim_dentry *parent, u8 *p);
1876 write_dentry_tree_recursive_cb(struct wim_dentry *dentry, void *_p)
1879 *p = write_dentry_tree_recursive(dentry, *p);
1883 /* Recursive function that writes a dentry tree rooted at @parent, not including
1884 * @parent itself, which has already been written. */
1886 write_dentry_tree_recursive(const struct wim_dentry *parent, u8 *p)
1888 /* Nothing to do if this dentry has no children. */
1889 if (parent->subdir_offset == 0)
1892 /* Write child dentries and end-of-directory entry.
1894 * Note: we need to write all of this dentry's children before
1895 * recursively writing the directory trees rooted at each of the child
1896 * dentries, since the on-disk dentries for a dentry's children are
1897 * always located at consecutive positions in the metadata resource! */
1898 for_dentry_child(parent, write_dentry_cb, &p);
1900 /* write end of directory entry */
1903 /* Recurse on children. */
1904 for_dentry_child(parent, write_dentry_tree_recursive_cb, &p);
1908 /* Writes a directory tree to the metadata resource.
1910 * @root: Root of the dentry tree.
1911 * @p: Pointer to a buffer with enough space for the dentry tree.
1913 * Returns pointer to the byte after the last byte we wrote.
1916 write_dentry_tree(const struct wim_dentry *root, u8 *p)
1918 DEBUG("Writing dentry tree.");
1919 wimlib_assert(dentry_is_root(root));
1921 /* If we're the root dentry, we have no parent that already
1922 * wrote us, so we need to write ourselves. */
1923 p = write_dentry(root, p);
1925 /* Write end of directory entry after the root dentry just to be safe;
1926 * however the root dentry obviously cannot have any siblings. */
1929 /* Recursively write the rest of the dentry tree. */
1930 return write_dentry_tree_recursive(root, p);