2f0cabd3d577f4eef9581c15fb58e370337e2cb0
[wimlib] / src / dentry.c
1 /*
2  * dentry.c
3  *
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
9  * entries.
10  */
11
12 /*
13  *
14  * Copyright (C) 2010 Carl Thijssen
15  * Copyright (C) 2012 Eric Biggers
16  *
17  * This file is part of wimlib, a library for working with WIM files.
18  *
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)
22  * any later version.
23  *
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
27  * details.
28  *
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/.
31  */
32
33 #include "wimlib_internal.h"
34 #include "dentry.h"
35 #include "io.h"
36 #include "timestamp.h"
37 #include "lookup_table.h"
38 #include "sha1.h"
39 #include <errno.h>
40 #include <unistd.h>
41 #include <sys/stat.h>
42
43 /*
44  * Returns true if @dentry has the UTF-8 file name @name that has length
45  * @name_len.
46  */
47 static bool dentry_has_name(const struct dentry *dentry, const char *name, 
48                             size_t name_len)
49 {
50         if (dentry->file_name_utf8_len != name_len)
51                 return false;
52         return memcmp(dentry->file_name_utf8, name, name_len) == 0;
53 }
54
55 /* Real length of a dentry, including the alternate data stream entries, which
56  * are not included in the dentry->length field... */
57 u64 dentry_total_length(const struct dentry *dentry)
58 {
59         u64 length = (dentry->length + 7) & ~7;
60         for (u16 i = 0 ; i < dentry->num_ads; i++)
61                 length += ads_entry_length(&dentry->ads_entries[i]);
62         return length;
63 }
64
65 /* Transfers file attributes from a `stat' buffer to a struct dentry. */
66 void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry)
67 {
68         if (S_ISLNK(stbuf->st_mode)) {
69                 dentry->attributes = FILE_ATTRIBUTE_REPARSE_POINT;
70                 dentry->reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
71         } else if (S_ISDIR(stbuf->st_mode)) {
72                 dentry->attributes = FILE_ATTRIBUTE_DIRECTORY;
73         } else {
74                 dentry->attributes = FILE_ATTRIBUTE_NORMAL;
75         }
76         if (sizeof(ino_t) >= 8)
77                 dentry->hard_link = (u64)stbuf->st_ino;
78         else
79                 dentry->hard_link = (u64)stbuf->st_ino |
80                                    ((u64)stbuf->st_dev << (sizeof(ino_t) * 8));
81 }
82
83
84 /* Makes all timestamp fields for the dentry be the current time. */
85 void dentry_update_all_timestamps(struct dentry *dentry)
86 {
87         u64 now = get_timestamp();
88         dentry->creation_time    = now;
89         dentry->last_access_time = now;
90         dentry->last_write_time  = now;
91 }
92
93 struct ads_entry *dentry_get_ads_entry(struct dentry *dentry,
94                                        const char *stream_name)
95 {
96         size_t stream_name_len = strlen(stream_name);
97         if (!stream_name)
98                 return NULL;
99         for (u16 i = 0; i < dentry->num_ads; i++)
100                 if (ads_entry_has_name(&dentry->ads_entries[i],
101                                        stream_name, stream_name_len))
102                         return &dentry->ads_entries[i];
103         return NULL;
104 }
105
106 static void ads_entry_init(struct ads_entry *ads_entry)
107 {
108         memset(ads_entry, 0, sizeof(struct ads_entry));
109         INIT_LIST_HEAD(&ads_entry->lte_group_list.list);
110         ads_entry->lte_group_list.type = STREAM_TYPE_ADS;
111 }
112
113 /* Add an alternate stream entry to a dentry and return a pointer to it, or NULL
114  * on failure. */
115 struct ads_entry *dentry_add_ads(struct dentry *dentry, const char *stream_name)
116 {
117         u16 num_ads;
118         struct ads_entry *ads_entries;
119         struct ads_entry *new_entry;
120
121         if (dentry->num_ads == 0xffff)
122                 return NULL;
123         num_ads = dentry->num_ads + 1;
124         ads_entries = REALLOC(dentry->ads_entries,
125                               num_ads * sizeof(struct ads_entry));
126         if (!ads_entries)
127                 return NULL;
128         if (ads_entries != dentry->ads_entries) {
129                 /* We moved the ADS entries.  Adjust the stream lists. */
130                 for (u16 i = 0; i < dentry->num_ads; i++) {
131                         struct list_head *cur = &ads_entries[i].lte_group_list.list;
132                         cur->prev->next = cur;
133                         cur->next->prev = cur;
134                 }
135         }
136         dentry->ads_entries = ads_entries;
137
138         new_entry = &ads_entries[num_ads - 1];
139         if (change_ads_name(new_entry, stream_name) != 0)
140                 return NULL;
141         dentry->num_ads = num_ads;
142         ads_entry_init(new_entry);
143         return new_entry;
144 }
145
146 void dentry_remove_ads(struct dentry *dentry, struct ads_entry *ads_entry)
147 {
148         u16 idx;
149         u16 following;
150
151         wimlib_assert(dentry->num_ads);
152         idx = ads_entry - dentry->ads_entries;
153         wimlib_assert(idx < dentry->num_ads);
154         following = dentry->num_ads - idx - 1;
155
156         destroy_ads_entry(ads_entry);
157         memcpy(ads_entry, ads_entry + 1, following * sizeof(struct ads_entry));
158
159         /* We moved the ADS entries.  Adjust the stream lists. */
160         for (u16 i = 0; i < following; i++) {
161                 struct list_head *cur = &ads_entry[i].lte_group_list.list;
162                 cur->prev->next = cur;
163                 cur->next->prev = cur;
164         }
165
166         dentry->num_ads--;
167 }
168
169 /* 
170  * Calls a function on all directory entries in a directory tree.  It is called
171  * on a parent before its children.
172  */
173 int for_dentry_in_tree(struct dentry *root, 
174                        int (*visitor)(struct dentry*, void*), void *arg)
175 {
176         int ret;
177         struct dentry *child;
178
179         ret = visitor(root, arg);
180
181         if (ret != 0)
182                 return ret;
183
184         child = root->children;
185
186         if (!child)
187                 return 0;
188
189         do {
190                 ret = for_dentry_in_tree(child, visitor, arg);
191                 if (ret != 0)
192                         return ret;
193                 child = child->next;
194         } while (child != root->children);
195         return 0;
196 }
197
198 /* 
199  * Like for_dentry_in_tree(), but the visitor function is always called on a
200  * dentry's children before on itself.
201  */
202 int for_dentry_in_tree_depth(struct dentry *root, 
203                              int (*visitor)(struct dentry*, void*), void *arg)
204 {
205         int ret;
206         struct dentry *child;
207         struct dentry *next;
208
209         child = root->children;
210         if (child) {
211                 do {
212                         next = child->next;
213                         ret = for_dentry_in_tree_depth(child, visitor, arg);
214                         if (ret != 0)
215                                 return ret;
216                         child = next;
217                 } while (child != root->children);
218         }
219         return visitor(root, arg);
220 }
221
222 /* 
223  * Calculate the full path of @dentry, based on its parent's full path and on
224  * its UTF-8 file name. 
225  */
226 int calculate_dentry_full_path(struct dentry *dentry, void *ignore)
227 {
228         char *full_path;
229         u32 full_path_len;
230         if (dentry_is_root(dentry)) {
231                 full_path = MALLOC(2);
232                 if (!full_path)
233                         goto oom;
234                 full_path[0] = '/';
235                 full_path[1] = '\0';
236                 full_path_len = 1;
237         } else {
238                 char *parent_full_path;
239                 u32 parent_full_path_len;
240                 const struct dentry *parent = dentry->parent;
241
242                 if (dentry_is_root(parent)) {
243                         parent_full_path = "";
244                         parent_full_path_len = 0;
245                 } else {
246                         parent_full_path = parent->full_path_utf8;
247                         parent_full_path_len = parent->full_path_utf8_len;
248                 }
249
250                 full_path_len = parent_full_path_len + 1 +
251                                 dentry->file_name_utf8_len;
252                 full_path = MALLOC(full_path_len + 1);
253                 if (!full_path)
254                         goto oom;
255
256                 memcpy(full_path, parent_full_path, parent_full_path_len);
257                 full_path[parent_full_path_len] = '/';
258                 memcpy(full_path + parent_full_path_len + 1,
259                        dentry->file_name_utf8,
260                        dentry->file_name_utf8_len);
261                 full_path[full_path_len] = '\0';
262         }
263         FREE(dentry->full_path_utf8);
264         dentry->full_path_utf8 = full_path;
265         dentry->full_path_utf8_len = full_path_len;
266         return 0;
267 oom:
268         ERROR("Out of memory while calculating dentry full path");
269         return WIMLIB_ERR_NOMEM;
270 }
271
272 /* 
273  * Recursively calculates the subdir offsets for a directory tree. 
274  *
275  * @dentry:  The root of the directory tree.
276  * @subdir_offset_p:  The current subdirectory offset; i.e., the subdirectory
277  *      offset for @dentry. 
278  */
279 void calculate_subdir_offsets(struct dentry *dentry, u64 *subdir_offset_p)
280 {
281         struct dentry *child;
282
283         child = dentry->children;
284         dentry->subdir_offset = *subdir_offset_p;
285         if (child) {
286
287                 /* Advance the subdir offset by the amount of space the children
288                  * of this dentry take up. */
289                 do {
290                         *subdir_offset_p += dentry_total_length(child);
291                         child = child->next;
292                 } while (child != dentry->children);
293
294                 /* End-of-directory dentry on disk. */
295                 *subdir_offset_p += 8;
296
297                 /* Recursively call calculate_subdir_offsets() on all the
298                  * children. */
299                 do {
300                         calculate_subdir_offsets(child, subdir_offset_p);
301                         child = child->next;
302                 } while (child != dentry->children);
303         } else {
304                 /* On disk, childless directories have a valid subdir_offset
305                  * that points to an 8-byte end-of-directory dentry.  Regular
306                  * files have a subdir_offset of 0. */
307                 if (dentry_is_directory(dentry))
308                         *subdir_offset_p += 8;
309                 else
310                         dentry->subdir_offset = 0;
311         }
312 }
313
314
315 /* Returns the child of @dentry that has the file name @name.  
316  * Returns NULL if no child has the name. */
317 struct dentry *get_dentry_child_with_name(const struct dentry *dentry, 
318                                                         const char *name)
319 {
320         struct dentry *child;
321         size_t name_len;
322         
323         child = dentry->children;
324         if (child) {
325                 name_len = strlen(name);
326                 do {
327                         if (dentry_has_name(child, name, name_len))
328                                 return child;
329                         child = child->next;
330                 } while (child != dentry->children);
331         }
332         return NULL;
333 }
334
335 /* Retrieves the dentry that has the UTF-8 @path relative to the dentry
336  * @cur_dir.  Returns NULL if no dentry having the path is found. */
337 static struct dentry *get_dentry_relative_path(struct dentry *cur_dir, const char *path)
338 {
339         struct dentry *child;
340         size_t base_len;
341         const char *new_path;
342
343         if (*path == '\0')
344                 return cur_dir;
345
346         child = cur_dir->children;
347         if (child) {
348                 new_path = path_next_part(path, &base_len);
349                 do {
350                         if (dentry_has_name(child, path, base_len))
351                                 return get_dentry_relative_path(child, new_path);
352                         child = child->next;
353                 } while (child != cur_dir->children);
354         }
355         return NULL;
356 }
357
358 /* Returns the dentry corresponding to the UTF-8 @path, or NULL if there is no
359  * such dentry. */
360 struct dentry *get_dentry(WIMStruct *w, const char *path)
361 {
362         struct dentry *root = wim_root_dentry(w);
363         while (*path == '/')
364                 path++;
365         return get_dentry_relative_path(root, path);
366 }
367
368 /* Returns the parent directory for the @path. */
369 struct dentry *get_parent_dentry(WIMStruct *w, const char *path)
370 {
371         size_t path_len = strlen(path);
372         char buf[path_len + 1];
373
374         memcpy(buf, path, path_len + 1);
375
376         to_parent_name(buf, path_len);
377
378         return get_dentry(w, buf);
379 }
380
381 /* Prints the full path of a dentry. */
382 int print_dentry_full_path(struct dentry *dentry, void *ignore)
383 {
384         if (dentry->full_path_utf8)
385                 puts(dentry->full_path_utf8);
386         return 0;
387 }
388
389 struct file_attr_flag {
390         u32 flag;
391         const char *name;
392 };
393 struct file_attr_flag file_attr_flags[] = {
394         {FILE_ATTRIBUTE_READONLY,               "READONLY"},
395         {FILE_ATTRIBUTE_HIDDEN,         "HIDDEN"},
396         {FILE_ATTRIBUTE_SYSTEM,         "SYSTEM"},
397         {FILE_ATTRIBUTE_DIRECTORY,              "DIRECTORY"},
398         {FILE_ATTRIBUTE_ARCHIVE,                "ARCHIVE"},
399         {FILE_ATTRIBUTE_DEVICE,         "DEVICE"},
400         {FILE_ATTRIBUTE_NORMAL,         "NORMAL"},
401         {FILE_ATTRIBUTE_TEMPORARY,              "TEMPORARY"},
402         {FILE_ATTRIBUTE_SPARSE_FILE,    "SPARSE_FILE"},
403         {FILE_ATTRIBUTE_REPARSE_POINT,  "REPARSE_POINT"},
404         {FILE_ATTRIBUTE_COMPRESSED,             "COMPRESSED"},
405         {FILE_ATTRIBUTE_OFFLINE,                "OFFLINE"},
406         {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,"NOT_CONTENT_INDEXED"},
407         {FILE_ATTRIBUTE_ENCRYPTED,              "ENCRYPTED"},
408         {FILE_ATTRIBUTE_VIRTUAL,                "VIRTUAL"},
409 };
410
411 /* Prints a directory entry.  @lookup_table is a pointer to the lookup table, or
412  * NULL if the resource entry for the dentry is not to be printed. */
413 int print_dentry(struct dentry *dentry, void *lookup_table)
414 {
415         const u8 *hash;
416
417         printf("[DENTRY]\n");
418         printf("Length            = %"PRIu64"\n", dentry->length);
419         printf("Attributes        = 0x%x\n", dentry->attributes);
420         for (unsigned i = 0; i < ARRAY_LEN(file_attr_flags); i++)
421                 if (file_attr_flags[i].flag & dentry->attributes)
422                         printf("    FILE_ATTRIBUTE_%s is set\n",
423                                 file_attr_flags[i].name);
424         printf("Security ID       = %d\n", dentry->security_id);
425         printf("Subdir offset     = %"PRIu64"\n", dentry->subdir_offset);
426         /*printf("Unused1           = 0x%"PRIu64"\n", dentry->unused1);*/
427         /*printf("Unused2           = %"PRIu64"\n", dentry->unused2);*/
428         printf("Creation Time     = 0x%"PRIx64"\n", dentry->creation_time);
429         printf("Last Access Time  = 0x%"PRIx64"\n", dentry->last_access_time);
430         printf("Last Write Time   = 0x%"PRIx64"\n", dentry->last_write_time);
431         hash = dentry_stream_hash(dentry, 0);
432         if (hash) {
433                 printf("Hash              = 0x"); 
434                 print_hash(hash);
435                 putchar('\n');
436         }
437         printf("Reparse Tag       = 0x%"PRIx32"\n", dentry->reparse_tag);
438         printf("Hard Link Group   = 0x%"PRIx64"\n", dentry->hard_link);
439         printf("Number of Alternate Data Streams = %hu\n", dentry->num_ads);
440         printf("Filename          = \"");
441         print_string(dentry->file_name, dentry->file_name_len);
442         puts("\"");
443         printf("Filename Length   = %hu\n", dentry->file_name_len);
444         printf("Filename (UTF-8)  = \"%s\"\n", dentry->file_name_utf8);
445         printf("Filename (UTF-8) Length = %hu\n", dentry->file_name_utf8_len);
446         printf("Short Name        = \"");
447         print_string(dentry->short_name, dentry->short_name_len);
448         puts("\"");
449         printf("Short Name Length = %hu\n", dentry->short_name_len);
450         printf("Full Path (UTF-8) = \"%s\"\n", dentry->full_path_utf8);
451         print_lookup_table_entry(dentry_stream_lte(dentry, 0, lookup_table));
452         for (u16 i = 0; i < dentry->num_ads; i++) {
453                 printf("[Alternate Stream Entry %u]\n", i);
454                 printf("Name = \"%s\"\n", dentry->ads_entries[i].stream_name_utf8);
455                 printf("Name Length (UTF-16) = %u\n",
456                                 dentry->ads_entries[i].stream_name_len);
457                 hash = dentry_stream_hash(dentry, i + 1);
458                 if (hash) {
459                         printf("Hash              = 0x"); 
460                         print_hash(hash);
461                         putchar('\n');
462                 }
463                 print_lookup_table_entry(dentry_stream_lte(dentry, i + 1,
464                                                            lookup_table));
465         }
466         return 0;
467 }
468
469 static inline void dentry_common_init(struct dentry *dentry)
470 {
471         memset(dentry, 0, sizeof(struct dentry));
472         dentry->refcnt = 1;
473         dentry->security_id = -1;
474         dentry->ads_entries_status = ADS_ENTRIES_DEFAULT;
475         dentry->lte_group_list.type = STREAM_TYPE_NORMAL;
476 }
477
478 /* 
479  * Creates an unlinked directory entry.
480  *
481  * @name:    The base name of the new dentry.
482  * @return:  A pointer to the new dentry, or NULL if out of memory.
483  */
484 struct dentry *new_dentry(const char *name)
485 {
486         struct dentry *dentry;
487         
488         dentry = MALLOC(sizeof(struct dentry));
489         if (!dentry)
490                 goto err;
491
492         dentry_common_init(dentry);
493         if (change_dentry_name(dentry, name) != 0)
494                 goto err;
495
496         dentry_update_all_timestamps(dentry);
497         dentry->next   = dentry;
498         dentry->prev   = dentry;
499         dentry->parent = dentry;
500         INIT_LIST_HEAD(&dentry->link_group_list);
501         return dentry;
502 err:
503         FREE(dentry);
504         ERROR("Failed to allocate new dentry");
505         return NULL;
506 }
507
508 void dentry_free_ads_entries(struct dentry *dentry)
509 {
510         for (u16 i = 0; i < dentry->num_ads; i++)
511                 destroy_ads_entry(&dentry->ads_entries[i]);
512         FREE(dentry->ads_entries);
513         dentry->ads_entries = NULL;
514         dentry->num_ads = 0;
515 }
516
517 static void __destroy_dentry(struct dentry *dentry)
518 {
519         FREE(dentry->file_name);
520         FREE(dentry->file_name_utf8);
521         FREE(dentry->short_name);
522         FREE(dentry->full_path_utf8);
523         FREE(dentry->extracted_file);
524 }
525
526 void free_dentry(struct dentry *dentry)
527 {
528         wimlib_assert(dentry);
529         __destroy_dentry(dentry);
530         if (dentry->ads_entries_status != ADS_ENTRIES_USER)
531                 dentry_free_ads_entries(dentry);
532         FREE(dentry);
533 }
534
535 /* Like free_dentry(), but assigns a new ADS entries owner if this dentry was
536  * the previous owner, and also deletes the dentry from its link_group_list */
537 void put_dentry(struct dentry *dentry)
538 {
539         if (dentry->ads_entries_status == ADS_ENTRIES_OWNER) {
540                 struct dentry *new_owner;
541                 list_for_each_entry(new_owner, &dentry->link_group_list,
542                                     link_group_list)
543                 {
544                         if (new_owner->ads_entries_status == ADS_ENTRIES_USER) {
545                                 new_owner->ads_entries_status = ADS_ENTRIES_OWNER;
546                                 break;
547                         }
548                 }
549                 dentry->ads_entries_status = ADS_ENTRIES_USER;
550         }
551         struct list_head *next;
552         list_del(&dentry->link_group_list);
553         free_dentry(dentry);
554 }
555
556
557 /* clones a dentry.
558  *
559  * Beware:
560  *      - memory for file names is not cloned
561  *      - next, prev, and children pointers and not touched
562  *      - stream entries are not cloned.
563  */
564 struct dentry *clone_dentry(struct dentry *old)
565 {
566         struct dentry *new = MALLOC(sizeof(struct dentry));
567         if (!new)
568                 return NULL;
569         memcpy(new, old, sizeof(struct dentry));
570         new->file_name          = NULL;
571         new->file_name_len      = 0;
572         new->file_name_utf8     = NULL;
573         new->file_name_utf8_len = 0;
574         new->short_name         = NULL;
575         new->short_name_len     = 0;
576         return new;
577 }
578
579 /* 
580  * This function is passed as an argument to for_dentry_in_tree_depth() in order
581  * to free a directory tree.  __args is a pointer to a `struct free_dentry_args'.
582  */
583 static int do_free_dentry(struct dentry *dentry, void *__lookup_table)
584 {
585         struct lookup_table *lookup_table = __lookup_table;
586         if (lookup_table) {
587                 struct lookup_table_entry *lte;
588                 if (dentry->resolved)
589                         lte = dentry->lte;
590                 else
591                         lte = __lookup_resource(lookup_table, dentry->hash);
592                 lte_decrement_refcnt(lte, lookup_table);
593         }
594
595         wimlib_assert(dentry->refcnt != 0);
596         if (--dentry->refcnt == 0)
597                 free_dentry(dentry);
598         return 0;
599 }
600
601 /* 
602  * Unlinks and frees a dentry tree.
603  *
604  * @root:               The root of the tree.
605  * @lookup_table:       The lookup table for dentries.  If non-NULL, the
606  *                      reference counts in the lookup table for the lookup
607  *                      table entries corresponding to the dentries will be
608  *                      decremented.
609  */
610 void free_dentry_tree(struct dentry *root, struct lookup_table *lookup_table)
611 {
612         if (!root || !root->parent)
613                 return;
614         for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
615 }
616
617 int increment_dentry_refcnt(struct dentry *dentry, void *ignore)
618 {
619         dentry->refcnt++;
620         return 0;
621 }
622
623 /* 
624  * Links a dentry into the directory tree.
625  *
626  * @dentry: The dentry to link.
627  * @parent: The dentry that will be the parent of @dentry.
628  */
629 void link_dentry(struct dentry *dentry, struct dentry *parent)
630 {
631         wimlib_assert(dentry_is_directory(parent));
632         dentry->parent = parent;
633         if (parent->children) {
634                 /* Not an only child; link to siblings. */
635                 dentry->next = parent->children;
636                 dentry->prev = parent->children->prev;
637                 dentry->next->prev = dentry;
638                 dentry->prev->next = dentry;
639         } else {
640                 /* Only child; link to parent. */
641                 parent->children = dentry;
642                 dentry->next = dentry;
643                 dentry->prev = dentry;
644         }
645 }
646
647
648 /* Unlink a dentry from the directory tree. 
649  *
650  * Note: This merely removes it from the in-memory tree structure.  See
651  * remove_dentry() in mount.c for a function implemented on top of this one that
652  * frees the dentry and implements reference counting for the lookup table
653  * entries. */
654 void unlink_dentry(struct dentry *dentry)
655 {
656         if (dentry_is_root(dentry))
657                 return;
658         if (dentry_is_only_child(dentry)) {
659                 dentry->parent->children = NULL;
660         } else {
661                 if (dentry_is_first_sibling(dentry))
662                         dentry->parent->children = dentry->next;
663                 dentry->next->prev = dentry->prev;
664                 dentry->prev->next = dentry->next;
665         }
666 }
667
668
669 /* Recalculates the length of @dentry based on its file name length and short
670  * name length.  */
671 static inline void recalculate_dentry_size(struct dentry *dentry)
672 {
673         dentry->length = WIM_DENTRY_DISK_SIZE + dentry->file_name_len + 
674                          2 + dentry->short_name_len;
675         /* Must be multiple of 8. */
676         dentry->length = (dentry->length + 7) & ~7;
677 }
678
679 /* Duplicates a UTF-8 name into UTF-8 and UTF-16 strings and returns the strings
680  * and their lengths in the pointer arguments */
681 int get_names(char **name_utf16_ret, char **name_utf8_ret,
682               u16 *name_utf16_len_ret, u16 *name_utf8_len_ret,
683               const char *name)
684 {
685         size_t utf8_len;
686         size_t utf16_len;
687         char *name_utf16, *name_utf8;
688
689         utf8_len = strlen(name);
690
691         name_utf16 = utf8_to_utf16(name, utf8_len, &utf16_len);
692
693         if (!name_utf16)
694                 return WIMLIB_ERR_NOMEM;
695
696         name_utf8 = MALLOC(utf8_len + 1);
697         if (!name_utf8) {
698                 FREE(name_utf8);
699                 return WIMLIB_ERR_NOMEM;
700         }
701         memcpy(name_utf8, name, utf8_len + 1);
702         FREE(*name_utf8_ret);
703         FREE(*name_utf16_ret);
704         *name_utf8_ret      = name_utf8;
705         *name_utf16_ret     = name_utf16;
706         *name_utf8_len_ret  = utf8_len;
707         *name_utf16_len_ret = utf16_len;
708         return 0;
709 }
710
711 /* Changes the name of a dentry to @new_name.  Only changes the file_name and
712  * file_name_utf8 fields; does not change the short_name, short_name_utf8, or
713  * full_path_utf8 fields.  Also recalculates its length. */
714 int change_dentry_name(struct dentry *dentry, const char *new_name)
715 {
716         int ret;
717
718         ret = get_names(&dentry->file_name, &dentry->file_name_utf8,
719                         &dentry->file_name_len, &dentry->file_name_utf8_len,
720                          new_name);
721         if (ret == 0)
722                 recalculate_dentry_size(dentry);
723         return ret;
724 }
725
726 int change_ads_name(struct ads_entry *entry, const char *new_name)
727 {
728         return get_names(&entry->stream_name, &entry->stream_name_utf8,
729                          &entry->stream_name_len,
730                          &entry->stream_name_utf8_len,
731                           new_name);
732 }
733
734 /* Parameters for calculate_dentry_statistics(). */
735 struct image_statistics {
736         struct lookup_table *lookup_table;
737         u64 *dir_count;
738         u64 *file_count;
739         u64 *total_bytes;
740         u64 *hard_link_bytes;
741 };
742
743 static int calculate_dentry_statistics(struct dentry *dentry, void *arg)
744 {
745         struct image_statistics *stats;
746         struct lookup_table_entry *lte; 
747         u16 i;
748         
749         stats = arg;
750
751         if (dentry_is_directory(dentry) && !dentry_is_root(dentry))
752                 ++*stats->dir_count;
753         else
754                 ++*stats->file_count;
755
756         if (dentry->resolved)
757                 lte = dentry->lte;
758         else
759                 lte = __lookup_resource(stats->lookup_table, dentry->hash);
760         i = 0;
761         while (1) {
762                 if (lte) {
763                         u64 size = lte->resource_entry.original_size;
764                         *stats->total_bytes += size;
765                         if (++lte->out_refcnt == 1)
766                                 *stats->hard_link_bytes += size;
767                 }
768                 if (i == dentry->num_ads)
769                         break;
770                 lte = __lookup_resource(stats->lookup_table,
771                                         dentry->ads_entries[i].hash);
772                 i++;
773         }
774
775         return 0;
776 }
777
778 void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table, 
779                                    u64 *dir_count_ret, u64 *file_count_ret, 
780                                    u64 *total_bytes_ret, 
781                                    u64 *hard_link_bytes_ret)
782 {
783         struct image_statistics stats;
784         *dir_count_ret         = 0;
785         *file_count_ret        = 0;
786         *total_bytes_ret       = 0;
787         *hard_link_bytes_ret   = 0;
788         stats.lookup_table     = table;
789         stats.dir_count       = dir_count_ret;
790         stats.file_count      = file_count_ret;
791         stats.total_bytes     = total_bytes_ret;
792         stats.hard_link_bytes = hard_link_bytes_ret;
793         for_lookup_table_entry(table, zero_out_refcnts, NULL);
794         for_dentry_in_tree(root, calculate_dentry_statistics, &stats);
795 }
796
797 static int read_ads_entries(const u8 *p, struct dentry *dentry,
798                             u64 remaining_size)
799 {
800         u16 num_ads = dentry->num_ads;
801         struct ads_entry *ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
802         int ret;
803         if (!ads_entries) {
804                 ERROR("Could not allocate memory for %"PRIu16" "
805                       "alternate data stream entries", num_ads);
806                 return WIMLIB_ERR_NOMEM;
807         }
808         DEBUG2("Reading %"PRIu16" alternate data streams "
809                "(remaining size = %"PRIu64")", num_ads, remaining_size);
810
811         for (u16 i = 0; i < num_ads; i++) {
812                 struct ads_entry *cur_entry = &ads_entries[i];
813                 u64 length;
814                 size_t utf8_len;
815                 const char *p_save = p;
816                 /* Read the base stream entry, excluding the stream name. */
817                 if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
818                         ERROR("Stream entries go past end of metadata resource");
819                         ERROR("(remaining_size = %"PRIu64")", remaining_size);
820                         ret = WIMLIB_ERR_INVALID_DENTRY;
821                         goto out_free_ads_entries;
822                 }
823                 remaining_size -= WIM_ADS_ENTRY_DISK_SIZE;
824
825                 p = get_u64(p, &length); /* ADS entry length */
826
827                 DEBUG2("ADS length = %"PRIu64, length);
828
829                 p += 8; /* Unused */
830                 p = get_bytes(p, SHA1_HASH_SIZE, (u8*)cur_entry->hash);
831                 p = get_u16(p, &cur_entry->stream_name_len);
832
833                 DEBUG2("Stream name length = %u", cur_entry->stream_name_len);
834
835                 cur_entry->stream_name = NULL;
836                 cur_entry->stream_name_utf8 = NULL;
837
838                 if (remaining_size < cur_entry->stream_name_len + 2) {
839                         ERROR("Stream entries go past end of metadata resource");
840                         ERROR("(remaining_size = %"PRIu64" bytes, stream_name_len "
841                               "= %"PRIu16" bytes", remaining_size,
842                               cur_entry->stream_name_len);
843                         ret = WIMLIB_ERR_INVALID_DENTRY;
844                         goto out_free_ads_entries;
845                 }
846                 remaining_size -= cur_entry->stream_name_len + 2;
847
848                 cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
849                 if (!cur_entry->stream_name) {
850                         ret = WIMLIB_ERR_NOMEM;
851                         goto out_free_ads_entries;
852                 }
853                 get_bytes(p, cur_entry->stream_name_len,
854                           (u8*)cur_entry->stream_name);
855                 cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
856                                                             cur_entry->stream_name_len,
857                                                             &utf8_len);
858                 cur_entry->stream_name_utf8_len = utf8_len;
859
860                 if (!cur_entry->stream_name_utf8) {
861                         ret = WIMLIB_ERR_NOMEM;
862                         goto out_free_ads_entries;
863                 }
864                 p = p_save + ads_entry_length(cur_entry);
865         }
866         dentry->ads_entries = ads_entries;
867         return 0;
868 out_free_ads_entries:
869         for (u16 i = 0; i < num_ads; i++) {
870                 FREE(ads_entries[i].stream_name);
871                 FREE(ads_entries[i].stream_name_utf8);
872         }
873         FREE(ads_entries);
874         return ret;
875 }
876
877 /* 
878  * Reads a directory entry from the metadata resource.
879  */
880 int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, 
881                 u64 offset, struct dentry *dentry)
882 {
883         const u8 *p;
884         u64 calculated_size;
885         char *file_name;
886         char *file_name_utf8;
887         char *short_name;
888         u16 short_name_len;
889         u16 file_name_len;
890         size_t file_name_utf8_len;
891         int ret;
892
893         dentry_common_init(dentry);
894
895         /*Make sure the dentry really fits into the metadata resource.*/
896         if (offset + 8 > metadata_resource_len) {
897                 ERROR("Directory entry starting at %"PRIu64" ends past the "
898                       "end of the metadata resource (size %"PRIu64")",
899                       offset, metadata_resource_len);
900                 return WIMLIB_ERR_INVALID_DENTRY;
901         }
902
903         /* Before reading the whole entry, we need to read just the length.
904          * This is because an entry of length 8 (that is, just the length field)
905          * terminates the list of sibling directory entries. */
906
907         p = get_u64(&metadata_resource[offset], &dentry->length);
908
909         /* A zero length field (really a length of 8, since that's how big the
910          * directory entry is...) indicates that this is the end of directory
911          * dentry.  We do not read it into memory as an actual dentry, so just
912          * return true in that case. */
913         if (dentry->length == 0)
914                 return 0;
915
916         if (offset + dentry->length >= metadata_resource_len) {
917                 ERROR("Directory entry at offset %"PRIu64" and with size "
918                       "%"PRIu64" ends past the end of the metadata resource "
919                       "(size %"PRIu64")",
920                       offset, dentry->length, metadata_resource_len);
921                 return WIMLIB_ERR_INVALID_DENTRY;
922         }
923
924         /* If it is a recognized length, read the rest of the directory entry.
925          * Note: The root directory entry has no name, and its length does not
926          * include the short name length field.  */
927         if (dentry->length < WIM_DENTRY_DISK_SIZE) {
928                 ERROR("Directory entry has invalid length of %"PRIu64" bytes",
929                       dentry->length);
930                 return WIMLIB_ERR_INVALID_DENTRY;
931         }
932
933         p = get_u32(p, &dentry->attributes);
934         p = get_u32(p, (u32*)&dentry->security_id);
935         p = get_u64(p, &dentry->subdir_offset);
936
937         /* 2 unused fields */
938         p += 2 * sizeof(u64);
939         /*p = get_u64(p, &dentry->unused1);*/
940         /*p = get_u64(p, &dentry->unused2);*/
941
942         p = get_u64(p, &dentry->creation_time);
943         p = get_u64(p, &dentry->last_access_time);
944         p = get_u64(p, &dentry->last_write_time);
945
946         p = get_bytes(p, SHA1_HASH_SIZE, dentry->hash);
947         
948         /*
949          * I don't know what's going on here.  It seems like M$ screwed up the
950          * reparse points, then put the fields in the same place and didn't
951          * document it.  The WIM_HDR_FLAG_RP_FIX flag in the WIM header might
952          * have something to do with this, but it's not documented.
953          */
954         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
955                 /* ??? */
956                 p += 4;
957                 p = get_u32(p, &dentry->reparse_tag);
958                 p += 4;
959         } else {
960                 p = get_u32(p, &dentry->reparse_tag);
961                 p = get_u64(p, &dentry->hard_link);
962         }
963
964         /* By the way, the reparse_reserved field does not actually exist (at
965          * least when the file is not a reparse point) */
966
967         
968         p = get_u16(p, &dentry->num_ads);
969
970         p = get_u16(p, &short_name_len);
971         p = get_u16(p, &file_name_len);
972
973         calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
974                           short_name_len;
975
976         if (dentry->length < calculated_size) {
977                 ERROR("Unexpected end of directory entry! (Expected "
978                       "%"PRIu64" bytes, got %"PRIu64" bytes. "
979                       "short_name_len = %hu, file_name_len = %hu)", 
980                       calculated_size, dentry->length,
981                       short_name_len, file_name_len);
982                 return WIMLIB_ERR_INVALID_DENTRY;
983         }
984
985         /* Read the filename. */
986         file_name = MALLOC(file_name_len);
987         if (!file_name) {
988                 ERROR("Failed to allocate %hu bytes for dentry file name",
989                       file_name_len);
990                 return WIMLIB_ERR_NOMEM;
991         }
992         p = get_bytes(p, file_name_len, file_name);
993
994         /* Convert filename to UTF-8. */
995         file_name_utf8 = utf16_to_utf8(file_name, file_name_len, 
996                                        &file_name_utf8_len);
997
998         if (!file_name_utf8) {
999                 ERROR("Failed to allocate memory to convert UTF-16 "
1000                       "filename (%hu bytes) to UTF-8", file_name_len);
1001                 ret = WIMLIB_ERR_NOMEM;
1002                 goto out_free_file_name;
1003         }
1004
1005         /* Undocumented padding between file name and short name.  This probably
1006          * is supposed to be a terminating null character. */
1007         p += 2;
1008
1009         /* Read the short filename. */
1010         short_name = MALLOC(short_name_len);
1011         if (!short_name) {
1012                 ERROR("Failed to allocate %hu bytes for short filename",
1013                       short_name_len);
1014                 ret = WIMLIB_ERR_NOMEM;
1015                 goto out_free_file_name_utf8;
1016         }
1017
1018         p = get_bytes(p, short_name_len, short_name);
1019
1020         /* Some directory entries inexplicibly have a little over 70 bytes of
1021          * extra data.  The exact amount of data seems to be 72 bytes, but it is
1022          * aligned on the next 8-byte boundary.  Here's an example of the
1023          * aligned data:
1024          *
1025          * 01000000400000006c786bbac58ede11b0bb00261870892ab6adb76fe63a3
1026          * e468fca86530d2effa16c786bbac58ede11b0bb00261870892a0000000000
1027          * 0000000000000000000000
1028          *
1029          * Here's one interpretation of how the data is laid out.
1030          *
1031          * struct unknown {
1032          *      u32 field1; (always 0x00000001)
1033          *      u32 field2; (always 0x40000000)
1034          *      u16 field3;
1035          *      u32 field4;
1036          *      u32 field5;
1037          *      u32 field6;
1038          *      u8  data[48]; (???)
1039          *      u64 reserved1; (always 0)
1040          *      u64 reserved2; (always 0)
1041          * };*/
1042 #if 0
1043         if (dentry->length - calculated_size >= WIM_ADS_ENTRY_DISK_SIZE) {
1044                 printf("%s: %lu / %lu (", file_name_utf8, 
1045                                 calculated_size, dentry->length);
1046                 print_string(p + WIM_ADS_ENTRY_DISK_SIZE, dentry->length - calculated_size - WIM_ADS_ENTRY_DISK_SIZE);
1047                 puts(")");
1048                 print_byte_field(p, dentry->length - calculated_size);
1049                 putchar('\n');
1050         }
1051 #endif
1052
1053         if (dentry->num_ads != 0) {
1054                 calculated_size = (calculated_size + 7) & ~7;
1055                 if (calculated_size > metadata_resource_len - offset) {
1056                         ERROR("Not enough space in metadata resource for "
1057                               "alternate stream entries");
1058                         ret = WIMLIB_ERR_INVALID_DENTRY;
1059                         goto out_free_short_name;
1060                 }
1061                 ret = read_ads_entries(&metadata_resource[offset + calculated_size],
1062                                        dentry,
1063                                        metadata_resource_len - offset - calculated_size);
1064                 if (ret != 0)
1065                         goto out_free_short_name;
1066         }
1067
1068         dentry->file_name          = file_name;
1069         dentry->file_name_utf8     = file_name_utf8;
1070         dentry->short_name         = short_name;
1071         dentry->file_name_len      = file_name_len;
1072         dentry->file_name_utf8_len = file_name_utf8_len;
1073         dentry->short_name_len     = short_name_len;
1074         return 0;
1075 out_free_short_name:
1076         FREE(short_name);
1077 out_free_file_name_utf8:
1078         FREE(file_name_utf8);
1079 out_free_file_name:
1080         FREE(file_name);
1081         return ret;
1082 }
1083
1084 /* 
1085  * Writes a dentry to an output buffer.
1086  *
1087  * @dentry:  The dentry structure.
1088  * @p:       The memory location to write the data to.
1089  * @return:  Pointer to the byte after the last byte we wrote as part of the
1090  *              dentry.
1091  */
1092 static u8 *write_dentry(const struct dentry *dentry, u8 *p)
1093 {
1094         u8 *orig_p = p;
1095         unsigned padding;
1096         const u8 *hash;
1097
1098         p = put_u64(p, dentry->length);
1099         p = put_u32(p, dentry->attributes);
1100         p = put_u32(p, dentry->security_id);
1101         p = put_u64(p, dentry->subdir_offset);
1102         p = put_u64(p, 0); /* unused1 */
1103         p = put_u64(p, 0); /* unused2 */
1104         p = put_u64(p, dentry->creation_time);
1105         p = put_u64(p, dentry->last_access_time);
1106         p = put_u64(p, dentry->last_write_time);
1107         if (dentry->resolved && dentry->lte)
1108                 hash = dentry->lte->hash;
1109         else
1110                 hash = dentry->hash;
1111         p = put_bytes(p, SHA1_HASH_SIZE, hash);
1112         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1113                 p = put_zeroes(p, 4);
1114                 p = put_u32(p, dentry->reparse_tag);
1115                 p = put_zeroes(p, 4);
1116         } else {
1117                 u64 hard_link;
1118                 p = put_u32(p, dentry->reparse_tag);
1119                 if (dentry->link_group_list.next == &dentry->link_group_list)
1120                         hard_link = 0;
1121                 else
1122                         hard_link = dentry->hard_link;
1123                 p = put_u64(p, hard_link);
1124         }
1125         p = put_u16(p, dentry->num_ads);
1126         p = put_u16(p, dentry->short_name_len);
1127         p = put_u16(p, dentry->file_name_len);
1128         p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
1129         p = put_u16(p, 0); /* filename padding, 2 bytes. */
1130         p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
1131
1132         wimlib_assert(p - orig_p <= dentry->length);
1133         if (p - orig_p < dentry->length)
1134                 p = put_zeroes(p, dentry->length - (p - orig_p));
1135
1136         p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1137
1138         for (u16 i = 0; i < dentry->num_ads; i++) {
1139                 p = put_u64(p, ads_entry_length(&dentry->ads_entries[i]));
1140                 p = put_u64(p, 0); /* Unused */
1141                 if (dentry->resolved && dentry->ads_entries[i].lte)
1142                         hash = dentry->ads_entries[i].lte->hash;
1143                 else
1144                         hash = dentry->ads_entries[i].hash;
1145                 p = put_bytes(p, SHA1_HASH_SIZE, hash);
1146                 p = put_u16(p, dentry->ads_entries[i].stream_name_len);
1147                 p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
1148                                  (u8*)dentry->ads_entries[i].stream_name);
1149                 p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1150         }
1151         return p;
1152 }
1153
1154 /* Recursive function that writes a dentry tree rooted at @tree, not including
1155  * @tree itself, which has already been written, except in the case of the root
1156  * dentry, which is written right away, along with an end-of-directory entry. */
1157 u8 *write_dentry_tree(const struct dentry *tree, u8 *p)
1158 {
1159         const struct dentry *child;
1160
1161         if (dentry_is_root(tree)) {
1162                 p = write_dentry(tree, p);
1163
1164                 /* write end of directory entry */
1165                 p = put_u64(p, 0);
1166         } else {
1167                 /* Nothing to do for non-directories */
1168                 if (!dentry_is_directory(tree))
1169                         return p;
1170         }
1171
1172         /* Write child dentries and end-of-directory entry. */
1173         child = tree->children;
1174         if (child) {
1175                 do {
1176                         p = write_dentry(child, p);
1177                         child = child->next;
1178                 } while (child != tree->children);
1179         }
1180
1181         /* write end of directory entry */
1182         p = put_u64(p, 0);
1183
1184         /* Recurse on children. */
1185         if (child) {
1186                 do {
1187                         p = write_dentry_tree(child, p);
1188                         child = child->next;
1189                 } while (child != tree->children);
1190         }
1191         return p;
1192 }
1193
1194 /* Reads the children of a dentry, and all their children, ..., etc. from the
1195  * metadata resource and into the dentry tree.
1196  *
1197  * @metadata_resource:  An array that contains the uncompressed metadata
1198  *                      resource for the WIM file.
1199  * @metadata_resource_len:      The length of @metadata_resource.
1200  * @dentry:     A pointer to a struct dentry that is the root of the directory
1201  *              tree and has already been read from the metadata resource.  It
1202  *              does not need to be the real root because this procedure is
1203  *              called recursively.
1204  *
1205  * @return:     Zero on success, nonzero on failure.
1206  */
1207 int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
1208                      struct dentry *dentry)
1209 {
1210         u64 cur_offset = dentry->subdir_offset;
1211         struct dentry *prev_child = NULL;
1212         struct dentry *first_child = NULL;
1213         struct dentry *child;
1214         struct dentry cur_child;
1215         int ret;
1216
1217         /* If @dentry is a regular file, nothing more needs to be done for this
1218          * branch. */
1219         if (cur_offset == 0)
1220                 return 0;
1221
1222         /* Find and read all the children of @dentry. */
1223         while (1) {
1224
1225                 /* Read next child of @dentry into @cur_child. */
1226                 ret = read_dentry(metadata_resource, metadata_resource_len, 
1227                                   cur_offset, &cur_child);
1228                 if (ret != 0)
1229                         break;
1230
1231                 /* Check for end of directory. */
1232                 if (cur_child.length == 0) {
1233                         ret = 0;
1234                         break;
1235                 }
1236
1237                 /* Not end of directory.  Allocate this child permanently and
1238                  * link it to the parent and previous child. */
1239                 child = MALLOC(sizeof(struct dentry));
1240                 if (!child) {
1241                         ERROR("Failed to allocate %zu bytes for new dentry",
1242                               sizeof(struct dentry));
1243                         ret = WIMLIB_ERR_NOMEM;
1244                         break;
1245                 }
1246                 memcpy(child, &cur_child, sizeof(struct dentry));
1247
1248                 if (prev_child) {
1249                         prev_child->next = child;
1250                         child->prev = prev_child;
1251                 } else {
1252                         first_child = child;
1253                 }
1254
1255                 child->parent = dentry;
1256                 prev_child = child;
1257
1258                 /* If there are children of this child, call this procedure
1259                  * recursively. */
1260                 if (child->subdir_offset != 0) {
1261                         ret = read_dentry_tree(metadata_resource, 
1262                                                metadata_resource_len, child);
1263                         if (ret != 0)
1264                                 break;
1265                 }
1266
1267                 /* Advance to the offset of the next child. */
1268                 cur_offset += dentry_total_length(child);
1269         }
1270
1271         /* Link last child to first one, and set parent's
1272          * children pointer to the first child.  */
1273         if (prev_child) {
1274                 prev_child->next = first_child;
1275                 first_child->prev = prev_child;
1276         }
1277         dentry->children = first_child;
1278         return ret;
1279 }