]> wimlib.net Git - wimlib/blob - src/dentry.c
Make lookup table use hlist
[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         struct lookup_table_entry *lte;
416         unsigned i;
417
418         printf("[DENTRY]\n");
419         printf("Length            = %"PRIu64"\n", dentry->length);
420         printf("Attributes        = 0x%x\n", dentry->attributes);
421         for (i = 0; i < ARRAY_LEN(file_attr_flags); i++)
422                 if (file_attr_flags[i].flag & dentry->attributes)
423                         printf("    FILE_ATTRIBUTE_%s is set\n",
424                                 file_attr_flags[i].name);
425         printf("Security ID       = %d\n", dentry->security_id);
426         printf("Subdir offset     = %"PRIu64"\n", dentry->subdir_offset);
427         /*printf("Unused1           = 0x%"PRIu64"\n", dentry->unused1);*/
428         /*printf("Unused2           = %"PRIu64"\n", dentry->unused2);*/
429         printf("Creation Time     = 0x%"PRIx64"\n", dentry->creation_time);
430         printf("Last Access Time  = 0x%"PRIx64"\n", dentry->last_access_time);
431         printf("Last Write Time   = 0x%"PRIx64"\n", dentry->last_write_time);
432         printf("Hash              = 0x"); 
433         print_hash(dentry->hash); 
434         putchar('\n');
435         printf("Reparse Tag       = 0x%"PRIx32"\n", dentry->reparse_tag);
436         printf("Hard Link Group   = 0x%"PRIx64"\n", dentry->hard_link);
437         printf("Number of Alternate Data Streams = %hu\n", dentry->num_ads);
438         printf("Filename          = \"");
439         print_string(dentry->file_name, dentry->file_name_len);
440         puts("\"");
441         printf("Filename Length   = %hu\n", dentry->file_name_len);
442         printf("Filename (UTF-8)  = \"%s\"\n", dentry->file_name_utf8);
443         printf("Filename (UTF-8) Length = %hu\n", dentry->file_name_utf8_len);
444         printf("Short Name        = \"");
445         print_string(dentry->short_name, dentry->short_name_len);
446         puts("\"");
447         printf("Short Name Length = %hu\n", dentry->short_name_len);
448         printf("Full Path (UTF-8) = \"%s\"\n", dentry->full_path_utf8);
449         if (lookup_table && (lte = __lookup_resource(lookup_table, dentry->hash)))
450                 print_lookup_table_entry(lte, NULL);
451         else
452                 putchar('\n');
453         for (u16 i = 0; i < dentry->num_ads; i++) {
454                 printf("[Alternate Stream Entry %u]\n", i);
455                 printf("Name = \"%s\"\n", dentry->ads_entries[i].stream_name_utf8);
456                 printf("Name Length (UTF-16) = %u\n",
457                                 dentry->ads_entries[i].stream_name_len);
458                 printf("Hash              = 0x"); 
459                 print_hash(dentry->ads_entries[i].hash); 
460                 if (lookup_table &&
461                      (lte = __lookup_resource(lookup_table,
462                                               dentry->ads_entries[i].hash)))
463                 {
464                         print_lookup_table_entry(lte, NULL);
465                 } else {
466                         putchar('\n');
467                 }
468         }
469         return 0;
470 }
471
472 static inline void dentry_common_init(struct dentry *dentry)
473 {
474         memset(dentry, 0, sizeof(struct dentry));
475         dentry->refcnt = 1;
476         dentry->security_id = -1;
477         dentry->ads_entries_status = ADS_ENTRIES_DEFAULT;
478         dentry->lte_group_list.type = STREAM_TYPE_NORMAL;
479 }
480
481 /* 
482  * Creates an unlinked directory entry.
483  *
484  * @name:    The base name of the new dentry.
485  * @return:  A pointer to the new dentry, or NULL if out of memory.
486  */
487 struct dentry *new_dentry(const char *name)
488 {
489         struct dentry *dentry;
490         
491         dentry = MALLOC(sizeof(struct dentry));
492         if (!dentry)
493                 goto err;
494
495         dentry_common_init(dentry);
496         if (change_dentry_name(dentry, name) != 0)
497                 goto err;
498
499         dentry_update_all_timestamps(dentry);
500         dentry->next   = dentry;
501         dentry->prev   = dentry;
502         dentry->parent = dentry;
503         INIT_LIST_HEAD(&dentry->link_group_list);
504         return dentry;
505 err:
506         FREE(dentry);
507         ERROR("Failed to allocate new dentry");
508         return NULL;
509 }
510
511 void dentry_free_ads_entries(struct dentry *dentry)
512 {
513         for (u16 i = 0; i < dentry->num_ads; i++)
514                 destroy_ads_entry(&dentry->ads_entries[i]);
515         FREE(dentry->ads_entries);
516         dentry->ads_entries = NULL;
517         dentry->num_ads = 0;
518 }
519
520 static void __destroy_dentry(struct dentry *dentry)
521 {
522         FREE(dentry->file_name);
523         FREE(dentry->file_name_utf8);
524         FREE(dentry->short_name);
525         FREE(dentry->full_path_utf8);
526         FREE(dentry->extracted_file);
527 }
528
529 void free_dentry(struct dentry *dentry)
530 {
531         wimlib_assert(dentry);
532         __destroy_dentry(dentry);
533         if (dentry->ads_entries_status != ADS_ENTRIES_USER)
534                 dentry_free_ads_entries(dentry);
535         FREE(dentry);
536 }
537
538 /* Like free_dentry(), but assigns a new ADS entries owner if this dentry was
539  * the previous owner, and also deletes the dentry from its link_group_list */
540 void put_dentry(struct dentry *dentry)
541 {
542         if (dentry->ads_entries_status == ADS_ENTRIES_OWNER) {
543                 struct dentry *new_owner;
544                 list_for_each_entry(new_owner, &dentry->link_group_list,
545                                     link_group_list)
546                 {
547                         if (new_owner->ads_entries_status == ADS_ENTRIES_USER) {
548                                 new_owner->ads_entries_status = ADS_ENTRIES_OWNER;
549                                 break;
550                         }
551                 }
552                 dentry->ads_entries_status = ADS_ENTRIES_USER;
553         }
554         struct list_head *next;
555         list_del(&dentry->link_group_list);
556         free_dentry(dentry);
557 }
558
559
560 /* clones a dentry.
561  *
562  * Beware:
563  *      - memory for file names is not cloned
564  *      - next, prev, and children pointers and not touched
565  *      - stream entries are not cloned.
566  */
567 struct dentry *clone_dentry(struct dentry *old)
568 {
569         struct dentry *new = MALLOC(sizeof(struct dentry));
570         if (!new)
571                 return NULL;
572         memcpy(new, old, sizeof(struct dentry));
573         new->file_name          = NULL;
574         new->file_name_len      = 0;
575         new->file_name_utf8     = NULL;
576         new->file_name_utf8_len = 0;
577         new->short_name         = NULL;
578         new->short_name_len     = 0;
579         return new;
580 }
581
582 /* 
583  * This function is passed as an argument to for_dentry_in_tree_depth() in order
584  * to free a directory tree.  __args is a pointer to a `struct free_dentry_args'.
585  */
586 static int do_free_dentry(struct dentry *dentry, void *__lookup_table)
587 {
588         struct lookup_table *lookup_table = __lookup_table;
589         if (lookup_table) {
590                 struct lookup_table_entry *lte;
591                 if (dentry->resolved)
592                         lte = dentry->lte;
593                 else
594                         lte = __lookup_resource(lookup_table, dentry->hash);
595                 lte_decrement_refcnt(lte, lookup_table);
596         }
597
598         wimlib_assert(dentry->refcnt != 0);
599         if (--dentry->refcnt == 0)
600                 free_dentry(dentry);
601         return 0;
602 }
603
604 /* 
605  * Unlinks and frees a dentry tree.
606  *
607  * @root:               The root of the tree.
608  * @lookup_table:       The lookup table for dentries.  If non-NULL, the
609  *                      reference counts in the lookup table for the lookup
610  *                      table entries corresponding to the dentries will be
611  *                      decremented.
612  */
613 void free_dentry_tree(struct dentry *root, struct lookup_table *lookup_table)
614 {
615         if (!root || !root->parent)
616                 return;
617         for_dentry_in_tree_depth(root, do_free_dentry, lookup_table);
618 }
619
620 int increment_dentry_refcnt(struct dentry *dentry, void *ignore)
621 {
622         dentry->refcnt++;
623         return 0;
624 }
625
626 /* 
627  * Links a dentry into the directory tree.
628  *
629  * @dentry: The dentry to link.
630  * @parent: The dentry that will be the parent of @dentry.
631  */
632 void link_dentry(struct dentry *dentry, struct dentry *parent)
633 {
634         wimlib_assert(dentry_is_directory(parent));
635         dentry->parent = parent;
636         if (parent->children) {
637                 /* Not an only child; link to siblings. */
638                 dentry->next = parent->children;
639                 dentry->prev = parent->children->prev;
640                 dentry->next->prev = dentry;
641                 dentry->prev->next = dentry;
642         } else {
643                 /* Only child; link to parent. */
644                 parent->children = dentry;
645                 dentry->next = dentry;
646                 dentry->prev = dentry;
647         }
648 }
649
650
651 /* Unlink a dentry from the directory tree. 
652  *
653  * Note: This merely removes it from the in-memory tree structure.  See
654  * remove_dentry() in mount.c for a function implemented on top of this one that
655  * frees the dentry and implements reference counting for the lookup table
656  * entries. */
657 void unlink_dentry(struct dentry *dentry)
658 {
659         if (dentry_is_root(dentry))
660                 return;
661         if (dentry_is_only_child(dentry)) {
662                 dentry->parent->children = NULL;
663         } else {
664                 if (dentry_is_first_sibling(dentry))
665                         dentry->parent->children = dentry->next;
666                 dentry->next->prev = dentry->prev;
667                 dentry->prev->next = dentry->next;
668         }
669 }
670
671
672 /* Recalculates the length of @dentry based on its file name length and short
673  * name length.  */
674 static inline void recalculate_dentry_size(struct dentry *dentry)
675 {
676         dentry->length = WIM_DENTRY_DISK_SIZE + dentry->file_name_len + 
677                          2 + dentry->short_name_len;
678         /* Must be multiple of 8. */
679         dentry->length = (dentry->length + 7) & ~7;
680 }
681
682 /* Duplicates a UTF-8 name into UTF-8 and UTF-16 strings and returns the strings
683  * and their lengths in the pointer arguments */
684 int get_names(char **name_utf16_ret, char **name_utf8_ret,
685               u16 *name_utf16_len_ret, u16 *name_utf8_len_ret,
686               const char *name)
687 {
688         size_t utf8_len;
689         size_t utf16_len;
690         char *name_utf16, *name_utf8;
691
692         utf8_len = strlen(name);
693
694         name_utf16 = utf8_to_utf16(name, utf8_len, &utf16_len);
695
696         if (!name_utf16)
697                 return WIMLIB_ERR_NOMEM;
698
699         name_utf8 = MALLOC(utf8_len + 1);
700         if (!name_utf8) {
701                 FREE(name_utf8);
702                 return WIMLIB_ERR_NOMEM;
703         }
704         memcpy(name_utf8, name, utf8_len + 1);
705         FREE(*name_utf8_ret);
706         FREE(*name_utf16_ret);
707         *name_utf8_ret      = name_utf8;
708         *name_utf16_ret     = name_utf16;
709         *name_utf8_len_ret  = utf8_len;
710         *name_utf16_len_ret = utf16_len;
711         return 0;
712 }
713
714 /* Changes the name of a dentry to @new_name.  Only changes the file_name and
715  * file_name_utf8 fields; does not change the short_name, short_name_utf8, or
716  * full_path_utf8 fields.  Also recalculates its length. */
717 int change_dentry_name(struct dentry *dentry, const char *new_name)
718 {
719         int ret;
720
721         ret = get_names(&dentry->file_name, &dentry->file_name_utf8,
722                         &dentry->file_name_len, &dentry->file_name_utf8_len,
723                          new_name);
724         if (ret == 0)
725                 recalculate_dentry_size(dentry);
726         return ret;
727 }
728
729 int change_ads_name(struct ads_entry *entry, const char *new_name)
730 {
731         return get_names(&entry->stream_name, &entry->stream_name_utf8,
732                          &entry->stream_name_len,
733                          &entry->stream_name_utf8_len,
734                           new_name);
735 }
736
737 /* Parameters for calculate_dentry_statistics(). */
738 struct image_statistics {
739         struct lookup_table *lookup_table;
740         u64 *dir_count;
741         u64 *file_count;
742         u64 *total_bytes;
743         u64 *hard_link_bytes;
744 };
745
746 static int calculate_dentry_statistics(struct dentry *dentry, void *arg)
747 {
748         struct image_statistics *stats;
749         struct lookup_table_entry *lte; 
750         u16 i;
751         
752         stats = arg;
753
754         if (dentry_is_directory(dentry) && !dentry_is_root(dentry))
755                 ++*stats->dir_count;
756         else
757                 ++*stats->file_count;
758
759         if (dentry->resolved)
760                 lte = dentry->lte;
761         else
762                 lte = __lookup_resource(stats->lookup_table, dentry->hash);
763         i = 0;
764         while (1) {
765                 if (lte) {
766                         u64 size = lte->resource_entry.original_size;
767                         *stats->total_bytes += size;
768                         if (++lte->out_refcnt == 1)
769                                 *stats->hard_link_bytes += size;
770                 }
771                 if (i == dentry->num_ads)
772                         break;
773                 lte = __lookup_resource(stats->lookup_table,
774                                         dentry->ads_entries[i].hash);
775                 i++;
776         }
777
778         return 0;
779 }
780
781 void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table, 
782                                    u64 *dir_count_ret, u64 *file_count_ret, 
783                                    u64 *total_bytes_ret, 
784                                    u64 *hard_link_bytes_ret)
785 {
786         struct image_statistics stats;
787         *dir_count_ret         = 0;
788         *file_count_ret        = 0;
789         *total_bytes_ret       = 0;
790         *hard_link_bytes_ret   = 0;
791         stats.lookup_table     = table;
792         stats.dir_count       = dir_count_ret;
793         stats.file_count      = file_count_ret;
794         stats.total_bytes     = total_bytes_ret;
795         stats.hard_link_bytes = hard_link_bytes_ret;
796         for_lookup_table_entry(table, zero_out_refcnts, NULL);
797         for_dentry_in_tree(root, calculate_dentry_statistics, &stats);
798 }
799
800 static int read_ads_entries(const u8 *p, struct dentry *dentry,
801                             u64 remaining_size)
802 {
803         u16 num_ads = dentry->num_ads;
804         struct ads_entry *ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
805         int ret;
806         if (!ads_entries) {
807                 ERROR("Could not allocate memory for %"PRIu16" "
808                       "alternate data stream entries", num_ads);
809                 return WIMLIB_ERR_NOMEM;
810         }
811         DEBUG2("Reading %"PRIu16" alternate data streams "
812                "(remaining size = %"PRIu64")", num_ads, remaining_size);
813
814         for (u16 i = 0; i < num_ads; i++) {
815                 struct ads_entry *cur_entry = &ads_entries[i];
816                 u64 length;
817                 size_t utf8_len;
818                 const char *p_save = p;
819                 /* Read the base stream entry, excluding the stream name. */
820                 if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
821                         ERROR("Stream entries go past end of metadata resource");
822                         ERROR("(remaining_size = %"PRIu64")", remaining_size);
823                         ret = WIMLIB_ERR_INVALID_DENTRY;
824                         goto out_free_ads_entries;
825                 }
826                 remaining_size -= WIM_ADS_ENTRY_DISK_SIZE;
827
828                 p = get_u64(p, &length); /* ADS entry length */
829
830                 DEBUG2("ADS length = %"PRIu64, length);
831
832                 p += 8; /* Unused */
833                 p = get_bytes(p, WIM_HASH_SIZE, (u8*)cur_entry->hash);
834                 p = get_u16(p, &cur_entry->stream_name_len);
835
836                 DEBUG2("Stream name length = %u", cur_entry->stream_name_len);
837
838                 cur_entry->stream_name = NULL;
839                 cur_entry->stream_name_utf8 = NULL;
840
841                 if (remaining_size < cur_entry->stream_name_len + 2) {
842                         ERROR("Stream entries go past end of metadata resource");
843                         ERROR("(remaining_size = %"PRIu64" bytes, stream_name_len "
844                               "= %"PRIu16" bytes", remaining_size,
845                               cur_entry->stream_name_len);
846                         ret = WIMLIB_ERR_INVALID_DENTRY;
847                         goto out_free_ads_entries;
848                 }
849                 remaining_size -= cur_entry->stream_name_len + 2;
850
851                 cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
852                 if (!cur_entry->stream_name) {
853                         ret = WIMLIB_ERR_NOMEM;
854                         goto out_free_ads_entries;
855                 }
856                 get_bytes(p, cur_entry->stream_name_len,
857                           (u8*)cur_entry->stream_name);
858                 cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
859                                                             cur_entry->stream_name_len,
860                                                             &utf8_len);
861                 cur_entry->stream_name_utf8_len = utf8_len;
862
863                 if (!cur_entry->stream_name_utf8) {
864                         ret = WIMLIB_ERR_NOMEM;
865                         goto out_free_ads_entries;
866                 }
867                 p = p_save + ads_entry_length(cur_entry);
868         }
869         dentry->ads_entries = ads_entries;
870         return 0;
871 out_free_ads_entries:
872         for (u16 i = 0; i < num_ads; i++) {
873                 FREE(ads_entries[i].stream_name);
874                 FREE(ads_entries[i].stream_name_utf8);
875         }
876         FREE(ads_entries);
877         return ret;
878 }
879
880 /* 
881  * Reads a directory entry from the metadata resource.
882  */
883 int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, 
884                 u64 offset, struct dentry *dentry)
885 {
886         const u8 *p;
887         u64 calculated_size;
888         char *file_name;
889         char *file_name_utf8;
890         char *short_name;
891         u16 short_name_len;
892         u16 file_name_len;
893         size_t file_name_utf8_len;
894         int ret;
895
896         dentry_common_init(dentry);
897
898         /*Make sure the dentry really fits into the metadata resource.*/
899         if (offset + 8 > metadata_resource_len) {
900                 ERROR("Directory entry starting at %"PRIu64" ends past the "
901                       "end of the metadata resource (size %"PRIu64")",
902                       offset, metadata_resource_len);
903                 return WIMLIB_ERR_INVALID_DENTRY;
904         }
905
906         /* Before reading the whole entry, we need to read just the length.
907          * This is because an entry of length 8 (that is, just the length field)
908          * terminates the list of sibling directory entries. */
909
910         p = get_u64(&metadata_resource[offset], &dentry->length);
911
912         /* A zero length field (really a length of 8, since that's how big the
913          * directory entry is...) indicates that this is the end of directory
914          * dentry.  We do not read it into memory as an actual dentry, so just
915          * return true in that case. */
916         if (dentry->length == 0)
917                 return 0;
918
919         if (offset + dentry->length >= metadata_resource_len) {
920                 ERROR("Directory entry at offset %"PRIu64" and with size "
921                       "%"PRIu64" ends past the end of the metadata resource "
922                       "(size %"PRIu64")",
923                       offset, dentry->length, metadata_resource_len);
924                 return WIMLIB_ERR_INVALID_DENTRY;
925         }
926
927         /* If it is a recognized length, read the rest of the directory entry.
928          * Note: The root directory entry has no name, and its length does not
929          * include the short name length field.  */
930         if (dentry->length < WIM_DENTRY_DISK_SIZE) {
931                 ERROR("Directory entry has invalid length of %"PRIu64" bytes",
932                       dentry->length);
933                 return WIMLIB_ERR_INVALID_DENTRY;
934         }
935
936         p = get_u32(p, &dentry->attributes);
937         p = get_u32(p, (u32*)&dentry->security_id);
938         p = get_u64(p, &dentry->subdir_offset);
939
940         /* 2 unused fields */
941         p += 2 * sizeof(u64);
942         /*p = get_u64(p, &dentry->unused1);*/
943         /*p = get_u64(p, &dentry->unused2);*/
944
945         p = get_u64(p, &dentry->creation_time);
946         p = get_u64(p, &dentry->last_access_time);
947         p = get_u64(p, &dentry->last_write_time);
948
949         p = get_bytes(p, WIM_HASH_SIZE, dentry->hash);
950         
951         /*
952          * I don't know what's going on here.  It seems like M$ screwed up the
953          * reparse points, then put the fields in the same place and didn't
954          * document it.  The WIM_HDR_FLAG_RP_FIX flag in the WIM header might
955          * have something to do with this, but it's not documented.
956          */
957         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
958                 /* ??? */
959                 p += 4;
960                 p = get_u32(p, &dentry->reparse_tag);
961                 p += 4;
962         } else {
963                 p = get_u32(p, &dentry->reparse_tag);
964                 p = get_u64(p, &dentry->hard_link);
965         }
966
967         /* By the way, the reparse_reserved field does not actually exist (at
968          * least when the file is not a reparse point) */
969
970         
971         p = get_u16(p, &dentry->num_ads);
972
973         p = get_u16(p, &short_name_len);
974         p = get_u16(p, &file_name_len);
975
976         calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
977                           short_name_len;
978
979         if (dentry->length < calculated_size) {
980                 ERROR("Unexpected end of directory entry! (Expected "
981                       "%"PRIu64" bytes, got %"PRIu64" bytes. "
982                       "short_name_len = %hu, file_name_len = %hu)", 
983                       calculated_size, dentry->length,
984                       short_name_len, file_name_len);
985                 return WIMLIB_ERR_INVALID_DENTRY;
986         }
987
988         /* Read the filename. */
989         file_name = MALLOC(file_name_len);
990         if (!file_name) {
991                 ERROR("Failed to allocate %hu bytes for dentry file name",
992                       file_name_len);
993                 return WIMLIB_ERR_NOMEM;
994         }
995         p = get_bytes(p, file_name_len, file_name);
996
997         /* Convert filename to UTF-8. */
998         file_name_utf8 = utf16_to_utf8(file_name, file_name_len, 
999                                        &file_name_utf8_len);
1000
1001         if (!file_name_utf8) {
1002                 ERROR("Failed to allocate memory to convert UTF-16 "
1003                       "filename (%hu bytes) to UTF-8", file_name_len);
1004                 ret = WIMLIB_ERR_NOMEM;
1005                 goto out_free_file_name;
1006         }
1007
1008         /* Undocumented padding between file name and short name.  This probably
1009          * is supposed to be a terminating null character. */
1010         p += 2;
1011
1012         /* Read the short filename. */
1013         short_name = MALLOC(short_name_len);
1014         if (!short_name) {
1015                 ERROR("Failed to allocate %hu bytes for short filename",
1016                       short_name_len);
1017                 ret = WIMLIB_ERR_NOMEM;
1018                 goto out_free_file_name_utf8;
1019         }
1020
1021         p = get_bytes(p, short_name_len, short_name);
1022
1023         /* Some directory entries inexplicibly have a little over 70 bytes of
1024          * extra data.  The exact amount of data seems to be 72 bytes, but it is
1025          * aligned on the next 8-byte boundary.  Here's an example of the
1026          * aligned data:
1027          *
1028          * 01000000400000006c786bbac58ede11b0bb00261870892ab6adb76fe63a3
1029          * e468fca86530d2effa16c786bbac58ede11b0bb00261870892a0000000000
1030          * 0000000000000000000000
1031          *
1032          * Here's one interpretation of how the data is laid out.
1033          *
1034          * struct unknown {
1035          *      u32 field1; (always 0x00000001)
1036          *      u32 field2; (always 0x40000000)
1037          *      u16 field3;
1038          *      u32 field4;
1039          *      u32 field5;
1040          *      u32 field6;
1041          *      u8  data[48]; (???)
1042          *      u64 reserved1; (always 0)
1043          *      u64 reserved2; (always 0)
1044          * };*/
1045 #if 0
1046         if (dentry->length - calculated_size >= WIM_ADS_ENTRY_DISK_SIZE) {
1047                 printf("%s: %lu / %lu (", file_name_utf8, 
1048                                 calculated_size, dentry->length);
1049                 print_string(p + WIM_ADS_ENTRY_DISK_SIZE, dentry->length - calculated_size - WIM_ADS_ENTRY_DISK_SIZE);
1050                 puts(")");
1051                 print_byte_field(p, dentry->length - calculated_size);
1052                 putchar('\n');
1053         }
1054 #endif
1055
1056         if (dentry->num_ads != 0) {
1057                 calculated_size = (calculated_size + 7) & ~7;
1058                 if (calculated_size > metadata_resource_len - offset) {
1059                         ERROR("Not enough space in metadata resource for "
1060                               "alternate stream entries");
1061                         ret = WIMLIB_ERR_INVALID_DENTRY;
1062                         goto out_free_short_name;
1063                 }
1064                 ret = read_ads_entries(&metadata_resource[offset + calculated_size],
1065                                        dentry,
1066                                        metadata_resource_len - offset - calculated_size);
1067                 if (ret != 0)
1068                         goto out_free_short_name;
1069         }
1070
1071         dentry->file_name          = file_name;
1072         dentry->file_name_utf8     = file_name_utf8;
1073         dentry->short_name         = short_name;
1074         dentry->file_name_len      = file_name_len;
1075         dentry->file_name_utf8_len = file_name_utf8_len;
1076         dentry->short_name_len     = short_name_len;
1077         return 0;
1078 out_free_short_name:
1079         FREE(short_name);
1080 out_free_file_name_utf8:
1081         FREE(file_name_utf8);
1082 out_free_file_name:
1083         FREE(file_name);
1084         return ret;
1085 }
1086
1087 /* 
1088  * Writes a dentry to an output buffer.
1089  *
1090  * @dentry:  The dentry structure.
1091  * @p:       The memory location to write the data to.
1092  * @return:  Pointer to the byte after the last byte we wrote as part of the
1093  *              dentry.
1094  */
1095 static u8 *write_dentry(const struct dentry *dentry, u8 *p)
1096 {
1097         u8 *orig_p = p;
1098         unsigned padding;
1099         const u8 *hash;
1100
1101         p = put_u64(p, dentry->length);
1102         p = put_u32(p, dentry->attributes);
1103         p = put_u32(p, dentry->security_id);
1104         p = put_u64(p, dentry->subdir_offset);
1105         p = put_u64(p, 0); /* unused1 */
1106         p = put_u64(p, 0); /* unused2 */
1107         p = put_u64(p, dentry->creation_time);
1108         p = put_u64(p, dentry->last_access_time);
1109         p = put_u64(p, dentry->last_write_time);
1110         if (dentry->resolved && dentry->lte)
1111                 hash = dentry->lte->hash;
1112         else
1113                 hash = dentry->hash;
1114         p = put_bytes(p, WIM_HASH_SIZE, hash);
1115         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1116                 p = put_zeroes(p, 4);
1117                 p = put_u32(p, dentry->reparse_tag);
1118                 p = put_zeroes(p, 4);
1119         } else {
1120                 u64 hard_link;
1121                 p = put_u32(p, dentry->reparse_tag);
1122                 if (dentry->link_group_list.next == &dentry->link_group_list)
1123                         hard_link = 0;
1124                 else
1125                         hard_link = dentry->hard_link;
1126                 p = put_u64(p, hard_link);
1127         }
1128         p = put_u16(p, dentry->num_ads);
1129         p = put_u16(p, dentry->short_name_len);
1130         p = put_u16(p, dentry->file_name_len);
1131         p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
1132         p = put_u16(p, 0); /* filename padding, 2 bytes. */
1133         p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
1134
1135         wimlib_assert(p - orig_p <= dentry->length);
1136         if (p - orig_p < dentry->length)
1137                 p = put_zeroes(p, dentry->length - (p - orig_p));
1138
1139         p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1140
1141         for (u16 i = 0; i < dentry->num_ads; i++) {
1142                 p = put_u64(p, ads_entry_length(&dentry->ads_entries[i]));
1143                 p = put_u64(p, 0); /* Unused */
1144                 p = put_bytes(p, WIM_HASH_SIZE, dentry->ads_entries[i].hash);
1145                 p = put_u16(p, dentry->ads_entries[i].stream_name_len);
1146                 p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
1147                                  (u8*)dentry->ads_entries[i].stream_name);
1148                 p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1149         }
1150         return p;
1151 }
1152
1153 /* Recursive function that writes a dentry tree rooted at @tree, not including
1154  * @tree itself, which has already been written, except in the case of the root
1155  * dentry, which is written right away, along with an end-of-directory entry. */
1156 u8 *write_dentry_tree(const struct dentry *tree, u8 *p)
1157 {
1158         const struct dentry *child;
1159
1160         if (dentry_is_root(tree)) {
1161                 p = write_dentry(tree, p);
1162
1163                 /* write end of directory entry */
1164                 p = put_u64(p, 0);
1165         } else {
1166                 /* Nothing to do for non-directories */
1167                 if (!dentry_is_directory(tree))
1168                         return p;
1169         }
1170
1171         /* Write child dentries and end-of-directory entry. */
1172         child = tree->children;
1173         if (child) {
1174                 do {
1175                         p = write_dentry(child, p);
1176                         child = child->next;
1177                 } while (child != tree->children);
1178         }
1179
1180         /* write end of directory entry */
1181         p = put_u64(p, 0);
1182
1183         /* Recurse on children. */
1184         if (child) {
1185                 do {
1186                         p = write_dentry_tree(child, p);
1187                         child = child->next;
1188                 } while (child != tree->children);
1189         }
1190         return p;
1191 }
1192
1193 /* Reads the children of a dentry, and all their children, ..., etc. from the
1194  * metadata resource and into the dentry tree.
1195  *
1196  * @metadata_resource:  An array that contains the uncompressed metadata
1197  *                      resource for the WIM file.
1198  * @metadata_resource_len:      The length of @metadata_resource.
1199  * @dentry:     A pointer to a struct dentry that is the root of the directory
1200  *              tree and has already been read from the metadata resource.  It
1201  *              does not need to be the real root because this procedure is
1202  *              called recursively.
1203  *
1204  * @return:     Zero on success, nonzero on failure.
1205  */
1206 int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
1207                      struct dentry *dentry)
1208 {
1209         u64 cur_offset = dentry->subdir_offset;
1210         struct dentry *prev_child = NULL;
1211         struct dentry *first_child = NULL;
1212         struct dentry *child;
1213         struct dentry cur_child;
1214         int ret;
1215
1216         /* If @dentry is a regular file, nothing more needs to be done for this
1217          * branch. */
1218         if (cur_offset == 0)
1219                 return 0;
1220
1221         /* Find and read all the children of @dentry. */
1222         while (1) {
1223
1224                 /* Read next child of @dentry into @cur_child. */
1225                 ret = read_dentry(metadata_resource, metadata_resource_len, 
1226                                   cur_offset, &cur_child);
1227                 if (ret != 0)
1228                         break;
1229
1230                 /* Check for end of directory. */
1231                 if (cur_child.length == 0) {
1232                         ret = 0;
1233                         break;
1234                 }
1235
1236                 /* Not end of directory.  Allocate this child permanently and
1237                  * link it to the parent and previous child. */
1238                 child = MALLOC(sizeof(struct dentry));
1239                 if (!child) {
1240                         ERROR("Failed to allocate %zu bytes for new dentry",
1241                               sizeof(struct dentry));
1242                         ret = WIMLIB_ERR_NOMEM;
1243                         break;
1244                 }
1245                 memcpy(child, &cur_child, sizeof(struct dentry));
1246
1247                 if (prev_child) {
1248                         prev_child->next = child;
1249                         child->prev = prev_child;
1250                 } else {
1251                         first_child = child;
1252                 }
1253
1254                 child->parent = dentry;
1255                 prev_child = child;
1256
1257                 /* If there are children of this child, call this procedure
1258                  * recursively. */
1259                 if (child->subdir_offset != 0) {
1260                         ret = read_dentry_tree(metadata_resource, 
1261                                                metadata_resource_len, child);
1262                         if (ret != 0)
1263                                 break;
1264                 }
1265
1266                 /* Advance to the offset of the next child. */
1267                 cur_offset += dentry_total_length(child);
1268         }
1269
1270         /* Link last child to first one, and set parent's
1271          * children pointer to the first child.  */
1272         if (prev_child) {
1273                 prev_child->next = first_child;
1274                 first_child->prev = prev_child;
1275         }
1276         dentry->children = first_child;
1277         return ret;
1278 }