4 * A dentry (directory entry) contains the metadata for a file. In the WIM file
5 * format, the dentries are stored in the "metadata resource" section right
6 * after the security data. Each image in the WIM file has its own metadata
7 * resource with its own security data and dentry tree. Dentries in different
8 * images may share file resources by referring to the same lookup table
14 * Copyright (C) 2010 Carl Thijssen
15 * Copyright (C) 2012 Eric Biggers
17 * This file is part of wimlib, a library for working with WIM files.
19 * wimlib is free software; you can redistribute it and/or modify it under the
20 * terms of the GNU Lesser General Public License as published by the Free
21 * Software Foundation; either version 2.1 of the License, or (at your option)
24 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
25 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
26 * A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
29 * You should have received a copy of the GNU Lesser General Public License
30 * along with wimlib; if not, see http://www.gnu.org/licenses/.
38 #include "wimlib_internal.h"
41 #include "timestamp.h"
42 #include "lookup_table.h"
46 * Returns true if @dentry has the UTF-8 file name @name that has length
49 static bool dentry_has_name(const struct dentry *dentry, const char *name,
52 if (dentry->file_name_utf8_len != name_len)
54 return memcmp(dentry->file_name_utf8, name, name_len) == 0;
57 /* Real length of a dentry, including the alternate data stream entries, which
58 * are not included in the dentry->length field... */
59 u64 dentry_total_length(const struct dentry *dentry)
61 u64 length = (dentry->length + 7) & ~7;
62 for (u16 i = 0; i < dentry->num_ads; i++)
63 length += ads_entry_length(&dentry->ads_entries[i]);
67 /* Transfers file attributes from a `stat' buffer to a struct dentry. */
68 void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry)
70 if (S_ISLNK(stbuf->st_mode)) {
71 dentry->attributes = FILE_ATTRIBUTE_REPARSE_POINT;
72 dentry->reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
73 } else if (S_ISDIR(stbuf->st_mode)) {
74 dentry->attributes = FILE_ATTRIBUTE_DIRECTORY;
76 dentry->attributes = FILE_ATTRIBUTE_NORMAL;
78 if (sizeof(ino_t) >= 8)
79 dentry->hard_link = (u64)stbuf->st_ino;
81 dentry->hard_link = (u64)stbuf->st_ino |
82 ((u64)stbuf->st_dev << (sizeof(ino_t) * 8));
84 dentry->creation_time = timespec_to_wim_timestamp(&stbuf->st_mtim);
85 dentry->last_write_time = timespec_to_wim_timestamp(&stbuf->st_mtim);
86 dentry->last_access_time = timespec_to_wim_timestamp(&stbuf->st_atim);
90 /* Makes all timestamp fields for the dentry be the current time. */
91 void dentry_update_all_timestamps(struct dentry *dentry)
93 u64 now = get_wim_timestamp();
94 dentry->creation_time = now;
95 dentry->last_access_time = now;
96 dentry->last_write_time = now;
99 struct ads_entry *dentry_get_ads_entry(struct dentry *dentry,
100 const char *stream_name)
102 size_t stream_name_len;
105 if (dentry->num_ads) {
107 stream_name_len = strlen(stream_name);
109 if (ads_entry_has_name(&dentry->ads_entries[i],
110 stream_name, stream_name_len))
111 return &dentry->ads_entries[i];
112 } while (++i != dentry->num_ads);
117 static void ads_entry_init(struct ads_entry *ads_entry)
119 memset(ads_entry, 0, sizeof(struct ads_entry));
120 INIT_LIST_HEAD(&ads_entry->lte_group_list.list);
121 ads_entry->lte_group_list.type = STREAM_TYPE_ADS;
124 /* Add an alternate stream entry to a dentry and return a pointer to it, or NULL
126 struct ads_entry *dentry_add_ads(struct dentry *dentry, const char *stream_name)
129 struct ads_entry *ads_entries;
130 struct ads_entry *new_entry;
132 DEBUG("Add alternate data stream %s:%s",
133 dentry->file_name_utf8, stream_name);
135 if (dentry->num_ads == 0xffff)
137 num_ads = dentry->num_ads + 1;
138 ads_entries = REALLOC(dentry->ads_entries,
139 num_ads * sizeof(struct ads_entry));
142 if (ads_entries != dentry->ads_entries) {
143 /* We moved the ADS entries. Adjust the stream lists. */
144 for (u16 i = 0; i < dentry->num_ads; i++) {
145 struct list_head *cur = &ads_entries[i].lte_group_list.list;
146 cur->prev->next = cur;
147 cur->next->prev = cur;
151 new_entry = &ads_entries[num_ads - 1];
152 ads_entry_init(new_entry);
153 if (change_ads_name(new_entry, stream_name) != 0)
155 dentry->ads_entries = ads_entries;
156 dentry->num_ads = num_ads;
160 void dentry_remove_ads(struct dentry *dentry, struct ads_entry *ads_entry)
165 wimlib_assert(dentry->num_ads);
166 idx = ads_entry - dentry->ads_entries;
167 wimlib_assert(idx < dentry->num_ads);
168 following = dentry->num_ads - idx - 1;
170 destroy_ads_entry(ads_entry);
171 memcpy(ads_entry, ads_entry + 1, following * sizeof(struct ads_entry));
173 /* We moved the ADS entries. Adjust the stream lists. */
174 for (u16 i = 0; i < following; i++) {
175 struct list_head *cur = &ads_entry[i].lte_group_list.list;
176 cur->prev->next = cur;
177 cur->next->prev = cur;
184 * Calls a function on all directory entries in a directory tree. It is called
185 * on a parent before its children.
187 int for_dentry_in_tree(struct dentry *root,
188 int (*visitor)(struct dentry*, void*), void *arg)
191 struct dentry *child;
193 ret = visitor(root, arg);
198 child = root->children;
204 ret = for_dentry_in_tree(child, visitor, arg);
208 } while (child != root->children);
213 * Like for_dentry_in_tree(), but the visitor function is always called on a
214 * dentry's children before on itself.
216 int for_dentry_in_tree_depth(struct dentry *root,
217 int (*visitor)(struct dentry*, void*), void *arg)
220 struct dentry *child;
223 child = root->children;
227 ret = for_dentry_in_tree_depth(child, visitor, arg);
231 } while (child != root->children);
233 return visitor(root, arg);
237 * Calculate the full path of @dentry, based on its parent's full path and on
238 * its UTF-8 file name.
240 int calculate_dentry_full_path(struct dentry *dentry, void *ignore)
244 if (dentry_is_root(dentry)) {
245 full_path = MALLOC(2);
252 char *parent_full_path;
253 u32 parent_full_path_len;
254 const struct dentry *parent = dentry->parent;
256 if (dentry_is_root(parent)) {
257 parent_full_path = "";
258 parent_full_path_len = 0;
260 parent_full_path = parent->full_path_utf8;
261 parent_full_path_len = parent->full_path_utf8_len;
264 full_path_len = parent_full_path_len + 1 +
265 dentry->file_name_utf8_len;
266 full_path = MALLOC(full_path_len + 1);
270 memcpy(full_path, parent_full_path, parent_full_path_len);
271 full_path[parent_full_path_len] = '/';
272 memcpy(full_path + parent_full_path_len + 1,
273 dentry->file_name_utf8,
274 dentry->file_name_utf8_len);
275 full_path[full_path_len] = '\0';
277 FREE(dentry->full_path_utf8);
278 dentry->full_path_utf8 = full_path;
279 dentry->full_path_utf8_len = full_path_len;
282 ERROR("Out of memory while calculating dentry full path");
283 return WIMLIB_ERR_NOMEM;
287 * Recursively calculates the subdir offsets for a directory tree.
289 * @dentry: The root of the directory tree.
290 * @subdir_offset_p: The current subdirectory offset; i.e., the subdirectory
291 * offset for @dentry.
293 void calculate_subdir_offsets(struct dentry *dentry, u64 *subdir_offset_p)
295 struct dentry *child;
297 child = dentry->children;
298 dentry->subdir_offset = *subdir_offset_p;
301 /* Advance the subdir offset by the amount of space the children
302 * of this dentry take up. */
304 *subdir_offset_p += dentry_total_length(child);
306 } while (child != dentry->children);
308 /* End-of-directory dentry on disk. */
309 *subdir_offset_p += 8;
311 /* Recursively call calculate_subdir_offsets() on all the
314 calculate_subdir_offsets(child, subdir_offset_p);
316 } while (child != dentry->children);
318 /* On disk, childless directories have a valid subdir_offset
319 * that points to an 8-byte end-of-directory dentry. Regular
320 * files have a subdir_offset of 0. */
321 if (dentry_is_directory(dentry))
322 *subdir_offset_p += 8;
324 dentry->subdir_offset = 0;
329 /* Returns the child of @dentry that has the file name @name.
330 * Returns NULL if no child has the name. */
331 struct dentry *get_dentry_child_with_name(const struct dentry *dentry,
334 struct dentry *child;
337 child = dentry->children;
339 name_len = strlen(name);
341 if (dentry_has_name(child, name, name_len))
344 } while (child != dentry->children);
349 /* Retrieves the dentry that has the UTF-8 @path relative to the dentry
350 * @cur_dir. Returns NULL if no dentry having the path is found. */
351 static struct dentry *get_dentry_relative_path(struct dentry *cur_dir, const char *path)
353 struct dentry *child;
355 const char *new_path;
360 child = cur_dir->children;
362 new_path = path_next_part(path, &base_len);
364 if (dentry_has_name(child, path, base_len))
365 return get_dentry_relative_path(child, new_path);
367 } while (child != cur_dir->children);
372 /* Returns the dentry corresponding to the UTF-8 @path, or NULL if there is no
374 struct dentry *get_dentry(WIMStruct *w, const char *path)
376 struct dentry *root = wim_root_dentry(w);
379 return get_dentry_relative_path(root, path);
382 /* Returns the parent directory for the @path. */
383 struct dentry *get_parent_dentry(WIMStruct *w, const char *path)
385 size_t path_len = strlen(path);
386 char buf[path_len + 1];
388 memcpy(buf, path, path_len + 1);
390 to_parent_name(buf, path_len);
392 return get_dentry(w, buf);
395 /* Prints the full path of a dentry. */
396 int print_dentry_full_path(struct dentry *dentry, void *ignore)
398 if (dentry->full_path_utf8)
399 puts(dentry->full_path_utf8);
403 struct file_attr_flag {
407 struct file_attr_flag file_attr_flags[] = {
408 {FILE_ATTRIBUTE_READONLY, "READONLY"},
409 {FILE_ATTRIBUTE_HIDDEN, "HIDDEN"},
410 {FILE_ATTRIBUTE_SYSTEM, "SYSTEM"},
411 {FILE_ATTRIBUTE_DIRECTORY, "DIRECTORY"},
412 {FILE_ATTRIBUTE_ARCHIVE, "ARCHIVE"},
413 {FILE_ATTRIBUTE_DEVICE, "DEVICE"},
414 {FILE_ATTRIBUTE_NORMAL, "NORMAL"},
415 {FILE_ATTRIBUTE_TEMPORARY, "TEMPORARY"},
416 {FILE_ATTRIBUTE_SPARSE_FILE, "SPARSE_FILE"},
417 {FILE_ATTRIBUTE_REPARSE_POINT, "REPARSE_POINT"},
418 {FILE_ATTRIBUTE_COMPRESSED, "COMPRESSED"},
419 {FILE_ATTRIBUTE_OFFLINE, "OFFLINE"},
420 {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,"NOT_CONTENT_INDEXED"},
421 {FILE_ATTRIBUTE_ENCRYPTED, "ENCRYPTED"},
422 {FILE_ATTRIBUTE_VIRTUAL, "VIRTUAL"},
425 /* Prints a directory entry. @lookup_table is a pointer to the lookup table, or
426 * NULL if the resource entry for the dentry is not to be printed. */
427 int print_dentry(struct dentry *dentry, void *lookup_table)
430 struct lookup_table_entry *lte;
432 printf("[DENTRY]\n");
433 printf("Length = %"PRIu64"\n", dentry->length);
434 printf("Attributes = 0x%x\n", dentry->attributes);
435 for (unsigned i = 0; i < ARRAY_LEN(file_attr_flags); i++)
436 if (file_attr_flags[i].flag & dentry->attributes)
437 printf(" FILE_ATTRIBUTE_%s is set\n",
438 file_attr_flags[i].name);
439 printf("Security ID = %d\n", dentry->security_id);
440 printf("Subdir offset = %"PRIu64"\n", dentry->subdir_offset);
442 printf("Unused1 = 0x%"PRIu64"\n", dentry->unused1);
443 printf("Unused2 = %"PRIu64"\n", dentry->unused2);
446 printf("Creation Time = 0x%"PRIx64"\n");
447 printf("Last Access Time = 0x%"PRIx64"\n");
448 printf("Last Write Time = 0x%"PRIx64"\n");
451 time_t creat_time = wim_timestamp_to_unix(dentry->creation_time);
452 time_t access_time = wim_timestamp_to_unix(dentry->last_access_time);
453 time_t mod_time = wim_timestamp_to_unix(dentry->last_write_time);
454 printf("Creation Time = %s", asctime(localtime(&creat_time)));
455 printf("Last Access Time = %s", asctime(localtime(&access_time)));
456 printf("Last Write Time = %s", asctime(localtime(&mod_time)));
458 printf("Reparse Tag = 0x%"PRIx32"\n", dentry->reparse_tag);
459 printf("Hard Link Group = 0x%"PRIx64"\n", dentry->hard_link);
460 printf("Number of Alternate Data Streams = %hu\n", dentry->num_ads);
461 printf("Filename = \"");
462 print_string(dentry->file_name, dentry->file_name_len);
464 printf("Filename Length = %hu\n", dentry->file_name_len);
465 printf("Filename (UTF-8) = \"%s\"\n", dentry->file_name_utf8);
466 printf("Filename (UTF-8) Length = %hu\n", dentry->file_name_utf8_len);
467 printf("Short Name = \"");
468 print_string(dentry->short_name, dentry->short_name_len);
470 printf("Short Name Length = %hu\n", dentry->short_name_len);
471 printf("Full Path (UTF-8) = \"%s\"\n", dentry->full_path_utf8);
472 lte = dentry_stream_lte(dentry, 0, lookup_table);
474 print_lookup_table_entry(lte);
476 hash = dentry_stream_hash(dentry, 0);
483 for (u16 i = 0; i < dentry->num_ads; i++) {
484 printf("[Alternate Stream Entry %u]\n", i);
485 printf("Name = \"%s\"\n", dentry->ads_entries[i].stream_name_utf8);
486 printf("Name Length (UTF-16) = %u\n",
487 dentry->ads_entries[i].stream_name_len);
488 hash = dentry_stream_hash(dentry, i + 1);
494 print_lookup_table_entry(dentry_stream_lte(dentry, i + 1,
500 static inline void dentry_common_init(struct dentry *dentry)
502 memset(dentry, 0, sizeof(struct dentry));
504 dentry->security_id = -1;
505 dentry->ads_entries_status = ADS_ENTRIES_DEFAULT;
506 dentry->lte_group_list.type = STREAM_TYPE_NORMAL;
510 * Creates an unlinked directory entry.
512 * @name: The base name of the new dentry.
513 * @return: A pointer to the new dentry, or NULL if out of memory.
515 struct dentry *new_dentry(const char *name)
517 struct dentry *dentry;
519 dentry = MALLOC(sizeof(struct dentry));
523 dentry_common_init(dentry);
524 if (change_dentry_name(dentry, name) != 0)
527 dentry_update_all_timestamps(dentry);
528 dentry->next = dentry;
529 dentry->prev = dentry;
530 dentry->parent = dentry;
531 INIT_LIST_HEAD(&dentry->link_group_list);
535 ERROR("Failed to allocate new dentry");
539 void dentry_free_ads_entries(struct dentry *dentry)
541 for (u16 i = 0; i < dentry->num_ads; i++)
542 destroy_ads_entry(&dentry->ads_entries[i]);
543 FREE(dentry->ads_entries);
544 dentry->ads_entries = NULL;
548 static void __destroy_dentry(struct dentry *dentry)
550 FREE(dentry->file_name);
551 FREE(dentry->file_name_utf8);
552 FREE(dentry->short_name);
553 FREE(dentry->full_path_utf8);
554 FREE(dentry->extracted_file);
557 void free_dentry(struct dentry *dentry)
559 wimlib_assert(dentry);
560 __destroy_dentry(dentry);
561 if (dentry->ads_entries_status != ADS_ENTRIES_USER)
562 dentry_free_ads_entries(dentry);
566 /* Like free_dentry(), but assigns a new ADS entries owner if this dentry was
567 * the previous owner, and also deletes the dentry from its link_group_list */
568 void put_dentry(struct dentry *dentry)
570 if (dentry->ads_entries_status == ADS_ENTRIES_OWNER) {
571 struct dentry *new_owner;
572 list_for_each_entry(new_owner, &dentry->link_group_list,
575 if (new_owner->ads_entries_status == ADS_ENTRIES_USER) {
576 new_owner->ads_entries_status = ADS_ENTRIES_OWNER;
580 dentry->ads_entries_status = ADS_ENTRIES_USER;
582 struct list_head *next;
583 list_del(&dentry->link_group_list);
591 * - memory for file names is not cloned
592 * - next, prev, and children pointers and not touched
593 * - stream entries are not cloned.
595 struct dentry *clone_dentry(struct dentry *old)
597 struct dentry *new = MALLOC(sizeof(struct dentry));
600 memcpy(new, old, sizeof(struct dentry));
601 new->file_name = NULL;
602 new->file_name_len = 0;
603 new->file_name_utf8 = NULL;
604 new->file_name_utf8_len = 0;
605 new->short_name = NULL;
606 new->short_name_len = 0;
611 * This function is passed as an argument to for_dentry_in_tree_depth() in order
612 * to free a directory tree. __args is a pointer to a `struct free_dentry_args'.
614 static int do_free_dentry(struct dentry *dentry, void *__lookup_table)
616 struct lookup_table *lookup_table = __lookup_table;
618 struct lookup_table_entry *lte;
619 if (dentry->resolved)
622 lte = __lookup_resource(lookup_table, dentry->hash);
623 lte_decrement_refcnt(lte, lookup_table);
626 wimlib_assert(dentry->refcnt != 0);
627 if (--dentry->refcnt == 0)
633 * Unlinks and frees a dentry tree.
635 * @root: The root of the tree.
636 * @lookup_table: The lookup table for dentries. If non-NULL, the
637 * reference counts in the lookup table for the lookup
638 * table entries corresponding to the dentries will be
641 void free_dentry_tree(struct dentry *root, struct lookup_table *lookup_table)
643 if (!root || !root->parent)
645 for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
648 int increment_dentry_refcnt(struct dentry *dentry, void *ignore)
655 * Links a dentry into the directory tree.
657 * @dentry: The dentry to link.
658 * @parent: The dentry that will be the parent of @dentry.
660 void link_dentry(struct dentry *dentry, struct dentry *parent)
662 wimlib_assert(dentry_is_directory(parent));
663 dentry->parent = parent;
664 if (parent->children) {
665 /* Not an only child; link to siblings. */
666 dentry->next = parent->children;
667 dentry->prev = parent->children->prev;
668 dentry->next->prev = dentry;
669 dentry->prev->next = dentry;
671 /* Only child; link to parent. */
672 parent->children = dentry;
673 dentry->next = dentry;
674 dentry->prev = dentry;
679 /* Unlink a dentry from the directory tree.
681 * Note: This merely removes it from the in-memory tree structure. See
682 * remove_dentry() in mount.c for a function implemented on top of this one that
683 * frees the dentry and implements reference counting for the lookup table
685 void unlink_dentry(struct dentry *dentry)
687 if (dentry_is_root(dentry))
689 if (dentry_is_only_child(dentry)) {
690 dentry->parent->children = NULL;
692 if (dentry_is_first_sibling(dentry))
693 dentry->parent->children = dentry->next;
694 dentry->next->prev = dentry->prev;
695 dentry->prev->next = dentry->next;
700 /* Recalculates the length of @dentry based on its file name length and short
702 static inline void recalculate_dentry_size(struct dentry *dentry)
704 dentry->length = WIM_DENTRY_DISK_SIZE + dentry->file_name_len +
705 2 + dentry->short_name_len;
706 /* Must be multiple of 8. */
707 dentry->length = (dentry->length + 7) & ~7;
710 /* Duplicates a UTF-8 name into UTF-8 and UTF-16 strings and returns the strings
711 * and their lengths in the pointer arguments */
712 int get_names(char **name_utf16_ret, char **name_utf8_ret,
713 u16 *name_utf16_len_ret, u16 *name_utf8_len_ret,
718 char *name_utf16, *name_utf8;
720 utf8_len = strlen(name);
722 name_utf16 = utf8_to_utf16(name, utf8_len, &utf16_len);
725 return WIMLIB_ERR_NOMEM;
727 name_utf8 = MALLOC(utf8_len + 1);
730 return WIMLIB_ERR_NOMEM;
732 memcpy(name_utf8, name, utf8_len + 1);
733 FREE(*name_utf8_ret);
734 FREE(*name_utf16_ret);
735 *name_utf8_ret = name_utf8;
736 *name_utf16_ret = name_utf16;
737 *name_utf8_len_ret = utf8_len;
738 *name_utf16_len_ret = utf16_len;
742 /* Changes the name of a dentry to @new_name. Only changes the file_name and
743 * file_name_utf8 fields; does not change the short_name, short_name_utf8, or
744 * full_path_utf8 fields. Also recalculates its length. */
745 int change_dentry_name(struct dentry *dentry, const char *new_name)
749 ret = get_names(&dentry->file_name, &dentry->file_name_utf8,
750 &dentry->file_name_len, &dentry->file_name_utf8_len,
753 recalculate_dentry_size(dentry);
757 int change_ads_name(struct ads_entry *entry, const char *new_name)
759 return get_names(&entry->stream_name, &entry->stream_name_utf8,
760 &entry->stream_name_len,
761 &entry->stream_name_utf8_len,
765 /* Parameters for calculate_dentry_statistics(). */
766 struct image_statistics {
767 struct lookup_table *lookup_table;
771 u64 *hard_link_bytes;
774 static int calculate_dentry_statistics(struct dentry *dentry, void *arg)
776 struct image_statistics *stats;
777 struct lookup_table_entry *lte;
782 if (dentry_is_directory(dentry) && !dentry_is_root(dentry))
785 ++*stats->file_count;
787 if (dentry->resolved)
790 lte = __lookup_resource(stats->lookup_table, dentry->hash);
794 u64 size = lte->resource_entry.original_size;
795 *stats->total_bytes += size;
796 if (++lte->out_refcnt == 1)
797 *stats->hard_link_bytes += size;
799 if (i == dentry->num_ads)
801 lte = __lookup_resource(stats->lookup_table,
802 dentry->ads_entries[i].hash);
809 void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table,
810 u64 *dir_count_ret, u64 *file_count_ret,
811 u64 *total_bytes_ret,
812 u64 *hard_link_bytes_ret)
814 struct image_statistics stats;
817 *total_bytes_ret = 0;
818 *hard_link_bytes_ret = 0;
819 stats.lookup_table = table;
820 stats.dir_count = dir_count_ret;
821 stats.file_count = file_count_ret;
822 stats.total_bytes = total_bytes_ret;
823 stats.hard_link_bytes = hard_link_bytes_ret;
824 for_lookup_table_entry(table, zero_out_refcnts, NULL);
825 for_dentry_in_tree(root, calculate_dentry_statistics, &stats);
828 static int read_ads_entries(const u8 *p, struct dentry *dentry,
831 u16 num_ads = dentry->num_ads;
832 struct ads_entry *ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
835 ERROR("Could not allocate memory for %"PRIu16" "
836 "alternate data stream entries", num_ads);
837 return WIMLIB_ERR_NOMEM;
839 DEBUG2("Reading %"PRIu16" alternate data streams "
840 "(remaining size = %"PRIu64")", num_ads, remaining_size);
842 for (u16 i = 0; i < num_ads; i++) {
843 struct ads_entry *cur_entry = &ads_entries[i];
846 const char *p_save = p;
847 /* Read the base stream entry, excluding the stream name. */
848 if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
849 ERROR("Stream entries go past end of metadata resource");
850 ERROR("(remaining_size = %"PRIu64")", remaining_size);
851 ret = WIMLIB_ERR_INVALID_DENTRY;
852 goto out_free_ads_entries;
854 remaining_size -= WIM_ADS_ENTRY_DISK_SIZE;
856 p = get_u64(p, &length); /* ADS entry length */
858 DEBUG2("ADS length = %"PRIu64, length);
861 p = get_bytes(p, SHA1_HASH_SIZE, (u8*)cur_entry->hash);
862 p = get_u16(p, &cur_entry->stream_name_len);
864 DEBUG2("Stream name length = %u", cur_entry->stream_name_len);
866 cur_entry->stream_name = NULL;
867 cur_entry->stream_name_utf8 = NULL;
869 if (remaining_size < cur_entry->stream_name_len + 2) {
870 ERROR("Stream entries go past end of metadata resource");
871 ERROR("(remaining_size = %"PRIu64" bytes, stream_name_len "
872 "= %"PRIu16" bytes", remaining_size,
873 cur_entry->stream_name_len);
874 ret = WIMLIB_ERR_INVALID_DENTRY;
875 goto out_free_ads_entries;
877 remaining_size -= cur_entry->stream_name_len + 2;
879 cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
880 if (!cur_entry->stream_name) {
881 ret = WIMLIB_ERR_NOMEM;
882 goto out_free_ads_entries;
884 get_bytes(p, cur_entry->stream_name_len,
885 (u8*)cur_entry->stream_name);
886 cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
887 cur_entry->stream_name_len,
889 cur_entry->stream_name_utf8_len = utf8_len;
891 if (!cur_entry->stream_name_utf8) {
892 ret = WIMLIB_ERR_NOMEM;
893 goto out_free_ads_entries;
895 p = p_save + ads_entry_length(cur_entry);
897 dentry->ads_entries = ads_entries;
899 out_free_ads_entries:
900 for (u16 i = 0; i < num_ads; i++) {
901 FREE(ads_entries[i].stream_name);
902 FREE(ads_entries[i].stream_name_utf8);
909 * Reads a directory entry from the metadata resource.
911 int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
912 u64 offset, struct dentry *dentry)
917 char *file_name_utf8;
921 size_t file_name_utf8_len;
924 dentry_common_init(dentry);
926 /*Make sure the dentry really fits into the metadata resource.*/
927 if (offset + 8 > metadata_resource_len) {
928 ERROR("Directory entry starting at %"PRIu64" ends past the "
929 "end of the metadata resource (size %"PRIu64")",
930 offset, metadata_resource_len);
931 return WIMLIB_ERR_INVALID_DENTRY;
934 /* Before reading the whole entry, we need to read just the length.
935 * This is because an entry of length 8 (that is, just the length field)
936 * terminates the list of sibling directory entries. */
938 p = get_u64(&metadata_resource[offset], &dentry->length);
940 /* A zero length field (really a length of 8, since that's how big the
941 * directory entry is...) indicates that this is the end of directory
942 * dentry. We do not read it into memory as an actual dentry, so just
943 * return true in that case. */
944 if (dentry->length == 0)
947 if (offset + dentry->length >= metadata_resource_len) {
948 ERROR("Directory entry at offset %"PRIu64" and with size "
949 "%"PRIu64" ends past the end of the metadata resource "
951 offset, dentry->length, metadata_resource_len);
952 return WIMLIB_ERR_INVALID_DENTRY;
955 /* If it is a recognized length, read the rest of the directory entry.
956 * Note: The root directory entry has no name, and its length does not
957 * include the short name length field. */
958 if (dentry->length < WIM_DENTRY_DISK_SIZE) {
959 ERROR("Directory entry has invalid length of %"PRIu64" bytes",
961 return WIMLIB_ERR_INVALID_DENTRY;
964 p = get_u32(p, &dentry->attributes);
965 p = get_u32(p, (u32*)&dentry->security_id);
966 p = get_u64(p, &dentry->subdir_offset);
968 /* 2 unused fields */
969 p += 2 * sizeof(u64);
970 /*p = get_u64(p, &dentry->unused1);*/
971 /*p = get_u64(p, &dentry->unused2);*/
973 p = get_u64(p, &dentry->creation_time);
974 p = get_u64(p, &dentry->last_access_time);
975 p = get_u64(p, &dentry->last_write_time);
977 p = get_bytes(p, SHA1_HASH_SIZE, dentry->hash);
980 * I don't know what's going on here. It seems like M$ screwed up the
981 * reparse points, then put the fields in the same place and didn't
982 * document it. The WIM_HDR_FLAG_RP_FIX flag in the WIM header might
983 * have something to do with this, but it's not documented.
985 if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
988 p = get_u32(p, &dentry->reparse_tag);
991 p = get_u32(p, &dentry->reparse_tag);
992 p = get_u64(p, &dentry->hard_link);
995 /* By the way, the reparse_reserved field does not actually exist (at
996 * least when the file is not a reparse point) */
999 p = get_u16(p, &dentry->num_ads);
1001 p = get_u16(p, &short_name_len);
1002 p = get_u16(p, &file_name_len);
1004 calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
1007 if (dentry->length < calculated_size) {
1008 ERROR("Unexpected end of directory entry! (Expected "
1009 "%"PRIu64" bytes, got %"PRIu64" bytes. "
1010 "short_name_len = %hu, file_name_len = %hu)",
1011 calculated_size, dentry->length,
1012 short_name_len, file_name_len);
1013 return WIMLIB_ERR_INVALID_DENTRY;
1016 /* Read the filename. */
1017 file_name = MALLOC(file_name_len);
1019 ERROR("Failed to allocate %hu bytes for dentry file name",
1021 return WIMLIB_ERR_NOMEM;
1023 p = get_bytes(p, file_name_len, file_name);
1025 /* Convert filename to UTF-8. */
1026 file_name_utf8 = utf16_to_utf8(file_name, file_name_len,
1027 &file_name_utf8_len);
1029 if (!file_name_utf8) {
1030 ERROR("Failed to allocate memory to convert UTF-16 "
1031 "filename (%hu bytes) to UTF-8", file_name_len);
1032 ret = WIMLIB_ERR_NOMEM;
1033 goto out_free_file_name;
1036 /* Undocumented padding between file name and short name. This probably
1037 * is supposed to be a terminating null character. */
1040 /* Read the short filename. */
1041 short_name = MALLOC(short_name_len);
1043 ERROR("Failed to allocate %hu bytes for short filename",
1045 ret = WIMLIB_ERR_NOMEM;
1046 goto out_free_file_name_utf8;
1049 p = get_bytes(p, short_name_len, short_name);
1051 /* Some directory entries inexplicibly have a little over 70 bytes of
1052 * extra data. The exact amount of data seems to be 72 bytes, but it is
1053 * aligned on the next 8-byte boundary. Here's an example of the
1056 * 01000000 40000000 6c786bba c58ede11 b0bb0026 1870892a b6adb76f
1057 * e63a3e46 8fca8653 0d2effa1 6c786bba c58ede11 b0bb0026 1870892a
1058 * 00000000 00000000 00000000 00000000
1060 * Here's one interpretation of how the data is laid out.
1063 * u32 field1; (always 0x00000001)
1064 * u32 field2; (always 0x40000000)
1065 * u8 data[48]; (???)
1066 * u64 reserved1; (always 0)
1067 * u64 reserved2; (always 0)
1070 if (dentry->length - calculated_size >= WIM_ADS_ENTRY_DISK_SIZE) {
1071 printf("%s: %lu / %lu (", file_name_utf8,
1072 calculated_size, dentry->length);
1073 print_string(p + WIM_ADS_ENTRY_DISK_SIZE, dentry->length - calculated_size - WIM_ADS_ENTRY_DISK_SIZE);
1075 print_byte_field(p, dentry->length - calculated_size);
1080 if (dentry->num_ads != 0) {
1081 calculated_size = (calculated_size + 7) & ~7;
1082 if (calculated_size > metadata_resource_len - offset) {
1083 ERROR("Not enough space in metadata resource for "
1084 "alternate stream entries");
1085 ret = WIMLIB_ERR_INVALID_DENTRY;
1086 goto out_free_short_name;
1088 ret = read_ads_entries(&metadata_resource[offset + calculated_size],
1090 metadata_resource_len - offset - calculated_size);
1092 goto out_free_short_name;
1095 dentry->file_name = file_name;
1096 dentry->file_name_utf8 = file_name_utf8;
1097 dentry->short_name = short_name;
1098 dentry->file_name_len = file_name_len;
1099 dentry->file_name_utf8_len = file_name_utf8_len;
1100 dentry->short_name_len = short_name_len;
1102 out_free_short_name:
1104 out_free_file_name_utf8:
1105 FREE(file_name_utf8);
1112 * Writes a dentry to an output buffer.
1114 * @dentry: The dentry structure.
1115 * @p: The memory location to write the data to.
1116 * @return: Pointer to the byte after the last byte we wrote as part of the
1119 static u8 *write_dentry(const struct dentry *dentry, u8 *p)
1125 p = put_u64(p, dentry->length);
1126 p = put_u32(p, dentry->attributes);
1127 p = put_u32(p, dentry->security_id);
1128 p = put_u64(p, dentry->subdir_offset);
1129 p = put_u64(p, 0); /* unused1 */
1130 p = put_u64(p, 0); /* unused2 */
1131 p = put_u64(p, dentry->creation_time);
1132 p = put_u64(p, dentry->last_access_time);
1133 p = put_u64(p, dentry->last_write_time);
1134 if (dentry->resolved && dentry->lte)
1135 hash = dentry->lte->hash;
1137 hash = dentry->hash;
1138 p = put_bytes(p, SHA1_HASH_SIZE, hash);
1139 if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1140 p = put_zeroes(p, 4);
1141 p = put_u32(p, dentry->reparse_tag);
1142 p = put_zeroes(p, 4);
1145 p = put_u32(p, dentry->reparse_tag);
1146 if (dentry->link_group_list.next == &dentry->link_group_list)
1149 hard_link = dentry->hard_link;
1150 p = put_u64(p, hard_link);
1152 p = put_u16(p, dentry->num_ads);
1153 p = put_u16(p, dentry->short_name_len);
1154 p = put_u16(p, dentry->file_name_len);
1155 p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
1156 p = put_u16(p, 0); /* filename padding, 2 bytes. */
1157 p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
1159 wimlib_assert(p - orig_p <= dentry->length);
1160 if (p - orig_p < dentry->length)
1161 p = put_zeroes(p, dentry->length - (p - orig_p));
1163 p = put_zeroes(p, (8 - dentry->length % 8) % 8);
1165 for (u16 i = 0; i < dentry->num_ads; i++) {
1166 p = put_u64(p, ads_entry_length(&dentry->ads_entries[i]));
1167 p = put_u64(p, 0); /* Unused */
1168 if (dentry->resolved && dentry->ads_entries[i].lte)
1169 hash = dentry->ads_entries[i].lte->hash;
1171 hash = dentry->ads_entries[i].hash;
1172 p = put_bytes(p, SHA1_HASH_SIZE, hash);
1173 p = put_u16(p, dentry->ads_entries[i].stream_name_len);
1174 p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
1175 (u8*)dentry->ads_entries[i].stream_name);
1177 p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1182 /* Recursive function that writes a dentry tree rooted at @tree, not including
1183 * @tree itself, which has already been written, except in the case of the root
1184 * dentry, which is written right away, along with an end-of-directory entry. */
1185 u8 *write_dentry_tree(const struct dentry *tree, u8 *p)
1187 const struct dentry *child;
1189 if (dentry_is_root(tree)) {
1190 p = write_dentry(tree, p);
1192 /* write end of directory entry */
1195 /* Nothing to do for non-directories */
1196 if (!dentry_is_directory(tree))
1200 /* Write child dentries and end-of-directory entry. */
1201 child = tree->children;
1204 p = write_dentry(child, p);
1205 child = child->next;
1206 } while (child != tree->children);
1209 /* write end of directory entry */
1212 /* Recurse on children. */
1215 p = write_dentry_tree(child, p);
1216 child = child->next;
1217 } while (child != tree->children);
1222 /* Reads the children of a dentry, and all their children, ..., etc. from the
1223 * metadata resource and into the dentry tree.
1225 * @metadata_resource: An array that contains the uncompressed metadata
1226 * resource for the WIM file.
1227 * @metadata_resource_len: The length of @metadata_resource.
1228 * @dentry: A pointer to a struct dentry that is the root of the directory
1229 * tree and has already been read from the metadata resource. It
1230 * does not need to be the real root because this procedure is
1231 * called recursively.
1233 * @return: Zero on success, nonzero on failure.
1235 int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
1236 struct dentry *dentry)
1238 u64 cur_offset = dentry->subdir_offset;
1239 struct dentry *prev_child = NULL;
1240 struct dentry *first_child = NULL;
1241 struct dentry *child;
1242 struct dentry cur_child;
1245 /* If @dentry is a regular file, nothing more needs to be done for this
1247 if (cur_offset == 0)
1250 /* Find and read all the children of @dentry. */
1253 /* Read next child of @dentry into @cur_child. */
1254 ret = read_dentry(metadata_resource, metadata_resource_len,
1255 cur_offset, &cur_child);
1259 /* Check for end of directory. */
1260 if (cur_child.length == 0) {
1265 /* Not end of directory. Allocate this child permanently and
1266 * link it to the parent and previous child. */
1267 child = MALLOC(sizeof(struct dentry));
1269 ERROR("Failed to allocate %zu bytes for new dentry",
1270 sizeof(struct dentry));
1271 ret = WIMLIB_ERR_NOMEM;
1274 memcpy(child, &cur_child, sizeof(struct dentry));
1277 prev_child->next = child;
1278 child->prev = prev_child;
1280 first_child = child;
1283 child->parent = dentry;
1286 /* If there are children of this child, call this procedure
1288 if (child->subdir_offset != 0) {
1289 ret = read_dentry_tree(metadata_resource,
1290 metadata_resource_len, child);
1295 /* Advance to the offset of the next child. */
1296 cur_offset += dentry_total_length(child);
1299 /* Link last child to first one, and set parent's
1300 * children pointer to the first child. */
1302 prev_child->next = first_child;
1303 first_child->prev = prev_child;
1305 dentry->children = first_child;