cf77adc765c6180100c43cbea5726a844ca86881
[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 <unistd.h>
40 #include <sys/stat.h>
41
42 /*
43  * Returns true if @dentry has the UTF-8 file name @name that has length
44  * @name_len.
45  */
46 static bool dentry_has_name(const struct dentry *dentry, const char *name, 
47                             size_t name_len)
48 {
49         if (dentry->file_name_utf8_len != name_len)
50                 return false;
51         return memcmp(dentry->file_name_utf8, name, name_len) == 0;
52 }
53
54 static bool ads_entry_has_name(const struct ads_entry *entry,
55                                const char *name, size_t name_len)
56 {
57         if (entry->stream_name_utf8_len != name_len)
58                 return false;
59         return memcmp(entry->stream_name_utf8, name, name_len) == 0;
60 }
61
62 /* Real length of a dentry, including the alternate data stream entries, which
63  * are not included in the dentry->length field... */
64 u64 dentry_total_length(const struct dentry *dentry)
65 {
66         u64 length = (dentry->length + 7) & ~7;
67         for (u16 i = 0 ; i < dentry->num_ads; i++)
68                 length += ads_entry_length(&dentry->ads_entries[i]);
69         return length;
70 }
71
72 /* Transfers file attributes from a `stat' buffer to a struct dentry. */
73 void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry)
74 {
75         if (S_ISLNK(stbuf->st_mode)) {
76                 dentry->attributes = FILE_ATTRIBUTE_REPARSE_POINT;
77                 dentry->reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
78         } else if (S_ISDIR(stbuf->st_mode)) {
79                 dentry->attributes = FILE_ATTRIBUTE_DIRECTORY;
80         } else {
81                 dentry->attributes = FILE_ATTRIBUTE_NORMAL;
82         }
83         if (sizeof(ino_t) >= 8)
84                 dentry->hard_link = (u64)stbuf->st_ino;
85         else
86                 dentry->hard_link = (u64)stbuf->st_ino |
87                                    ((u64)stbuf->st_dev << (sizeof(ino_t) * 8));
88 }
89
90 /* Transfers file attributes from a struct dentry to a `stat' buffer. */
91 void dentry_to_stbuf(const struct dentry *dentry, struct stat *stbuf, 
92                      const struct lookup_table *table)
93 {
94         struct lookup_table_entry *lte;
95
96         if (dentry_is_symlink(dentry))
97                 stbuf->st_mode = S_IFLNK | 0777;
98         else if (dentry_is_directory(dentry))
99                 stbuf->st_mode = S_IFDIR | 0755;
100         else
101                 stbuf->st_mode = S_IFREG | 0644;
102
103         if (table)
104                 lte = __lookup_resource(table, dentry_hash(dentry));
105         else
106                 lte = NULL;
107
108         if (lte) {
109                 stbuf->st_nlink = lte->refcnt;
110                 stbuf->st_size = lte->resource_entry.original_size;
111         } else {
112                 stbuf->st_nlink = 1;
113                 stbuf->st_size = 0;
114         }
115         stbuf->st_uid     = getuid();
116         stbuf->st_gid     = getgid();
117         stbuf->st_atime   = ms_timestamp_to_unix(dentry->last_access_time);
118         stbuf->st_mtime   = ms_timestamp_to_unix(dentry->last_write_time);
119         stbuf->st_ctime   = ms_timestamp_to_unix(dentry->creation_time);
120         stbuf->st_blocks  = (stbuf->st_size + 511) / 512;
121 }
122
123 /* Makes all timestamp fields for the dentry be the current time. */
124 void dentry_update_all_timestamps(struct dentry *dentry)
125 {
126         u64 now = get_timestamp();
127         dentry->creation_time    = now;
128         dentry->last_access_time = now;
129         dentry->last_write_time  = now;
130 }
131
132 struct ads_entry *dentry_get_ads_entry(struct dentry *dentry,
133                                        const char *stream_name)
134 {
135         size_t stream_name_len = strlen(stream_name);
136         if (!stream_name)
137                 return NULL;
138         for (u16 i = 0; i < dentry->num_ads; i++)
139                 if (ads_entry_has_name(&dentry->ads_entries[i],
140                                        stream_name, stream_name_len))
141                         return &dentry->ads_entries[i];
142         return NULL;
143 }
144
145 /* Add an alternate stream entry to a dentry and return a pointer to it, or NULL
146  * on failure. */
147 struct ads_entry *dentry_add_ads(struct dentry *dentry, const char *stream_name)
148 {
149         u16 num_ads = dentry->num_ads + 1;
150         struct ads_entry *ads_entries;
151         struct ads_entry *new_entry;
152         if (num_ads == 0xffff)
153                 return NULL;
154         ads_entries = MALLOC(num_ads * sizeof(struct ads_entry));
155         if (!ads_entries)
156                 return NULL;
157
158         new_entry = &ads_entries[num_ads - 1];
159         if (change_ads_name(new_entry, stream_name) != 0) {
160                 FREE(ads_entries);
161                 return NULL;
162         }
163
164         memcpy(ads_entries, dentry->ads_entries,
165                (num_ads - 1) * sizeof(struct ads_entry));
166         FREE(dentry->ads_entries);
167         dentry->ads_entries = ads_entries;
168         dentry->num_ads = num_ads;
169         return memset(new_entry, 0, sizeof(struct ads_entry));
170 }
171
172 void dentry_remove_ads(struct dentry *dentry, struct ads_entry *sentry)
173 {
174         destroy_ads_entry(sentry);
175         memcpy(sentry, sentry + 1,
176                (dentry->num_ads - (sentry - dentry->ads_entries))
177                  * sizeof(struct ads_entry));
178         dentry->num_ads--;
179 }
180
181 /* 
182  * Calls a function on all directory entries in a directory tree.  It is called
183  * on a parent before its children.
184  */
185 int for_dentry_in_tree(struct dentry *root, 
186                        int (*visitor)(struct dentry*, void*), void *arg)
187 {
188         int ret;
189         struct dentry *child;
190
191         ret = visitor(root, arg);
192
193         if (ret != 0)
194                 return ret;
195
196         child = root->children;
197
198         if (!child)
199                 return 0;
200
201         do {
202                 ret = for_dentry_in_tree(child, visitor, arg);
203                 if (ret != 0)
204                         return ret;
205                 child = child->next;
206         } while (child != root->children);
207         return 0;
208 }
209
210 /* 
211  * Like for_dentry_in_tree(), but the visitor function is always called on a
212  * dentry's children before on itself.
213  */
214 int for_dentry_in_tree_depth(struct dentry *root, 
215                              int (*visitor)(struct dentry*, void*), void *arg)
216 {
217         int ret;
218         struct dentry *child;
219         struct dentry *next;
220
221         child = root->children;
222         if (child) {
223                 do {
224                         next = child->next;
225                         ret = for_dentry_in_tree_depth(child, visitor, arg);
226                         if (ret != 0)
227                                 return ret;
228                         child = next;
229                 } while (child != root->children);
230         }
231         return visitor(root, arg);
232 }
233
234 /* 
235  * Calculate the full path of @dentry, based on its parent's full path and on
236  * its UTF-8 file name. 
237  */
238 int calculate_dentry_full_path(struct dentry *dentry, void *ignore)
239 {
240         char *full_path;
241         u32 full_path_len;
242         if (dentry_is_root(dentry)) {
243                 full_path = MALLOC(2);
244                 if (!full_path)
245                         goto oom;
246                 full_path[0] = '/';
247                 full_path[1] = '\0';
248                 full_path_len = 1;
249         } else {
250                 char *parent_full_path;
251                 u32 parent_full_path_len;
252                 const struct dentry *parent = dentry->parent;
253
254                 if (dentry_is_root(parent)) {
255                         parent_full_path = "";
256                         parent_full_path_len = 0;
257                 } else {
258                         parent_full_path = parent->full_path_utf8;
259                         parent_full_path_len = parent->full_path_utf8_len;
260                 }
261
262                 full_path_len = parent_full_path_len + 1 +
263                                 dentry->file_name_utf8_len;
264                 full_path = MALLOC(full_path_len + 1);
265                 if (!full_path)
266                         goto oom;
267
268                 memcpy(full_path, parent_full_path, parent_full_path_len);
269                 full_path[parent_full_path_len] = '/';
270                 memcpy(full_path + parent_full_path_len + 1,
271                        dentry->file_name_utf8,
272                        dentry->file_name_utf8_len);
273                 full_path[full_path_len] = '\0';
274         }
275         FREE(dentry->full_path_utf8);
276         dentry->full_path_utf8 = full_path;
277         dentry->full_path_utf8_len = full_path_len;
278         return 0;
279 oom:
280         ERROR("Out of memory while calculating dentry full path");
281         return WIMLIB_ERR_NOMEM;
282 }
283
284 /* 
285  * Recursively calculates the subdir offsets for a directory tree. 
286  *
287  * @dentry:  The root of the directory tree.
288  * @subdir_offset_p:  The current subdirectory offset; i.e., the subdirectory
289  *      offset for @dentry. 
290  */
291 void calculate_subdir_offsets(struct dentry *dentry, u64 *subdir_offset_p)
292 {
293         struct dentry *child;
294
295         child = dentry->children;
296         dentry->subdir_offset = *subdir_offset_p;
297         if (child) {
298
299                 /* Advance the subdir offset by the amount of space the children
300                  * of this dentry take up. */
301                 do {
302                         *subdir_offset_p += dentry_total_length(child);
303                         child = child->next;
304                 } while (child != dentry->children);
305
306                 /* End-of-directory dentry on disk. */
307                 *subdir_offset_p += 8;
308
309                 /* Recursively call calculate_subdir_offsets() on all the
310                  * children. */
311                 do {
312                         calculate_subdir_offsets(child, subdir_offset_p);
313                         child = child->next;
314                 } while (child != dentry->children);
315         } else {
316                 /* On disk, childless directories have a valid subdir_offset
317                  * that points to an 8-byte end-of-directory dentry.  Regular
318                  * files have a subdir_offset of 0. */
319                 if (dentry_is_directory(dentry))
320                         *subdir_offset_p += 8;
321                 else
322                         dentry->subdir_offset = 0;
323         }
324 }
325
326
327 /* Returns the child of @dentry that has the file name @name.  
328  * Returns NULL if no child has the name. */
329 struct dentry *get_dentry_child_with_name(const struct dentry *dentry, 
330                                                         const char *name)
331 {
332         struct dentry *child;
333         size_t name_len;
334         
335         child = dentry->children;
336         if (child) {
337                 name_len = strlen(name);
338                 do {
339                         if (dentry_has_name(child, name, name_len))
340                                 return child;
341                         child = child->next;
342                 } while (child != dentry->children);
343         }
344         return NULL;
345 }
346
347 /* Retrieves the dentry that has the UTF-8 @path relative to the dentry
348  * @cur_dir.  Returns NULL if no dentry having the path is found. */
349 static struct dentry *get_dentry_relative_path(struct dentry *cur_dir, const char *path)
350 {
351         struct dentry *child;
352         size_t base_len;
353         const char *new_path;
354
355         if (*path == '\0')
356                 return cur_dir;
357
358         child = cur_dir->children;
359         if (child) {
360                 new_path = path_next_part(path, &base_len);
361                 do {
362                         if (dentry_has_name(child, path, base_len))
363                                 return get_dentry_relative_path(child, new_path);
364                         child = child->next;
365                 } while (child != cur_dir->children);
366         }
367         return NULL;
368 }
369
370 /* Returns the dentry corresponding to the UTF-8 @path, or NULL if there is no
371  * such dentry. */
372 struct dentry *get_dentry(WIMStruct *w, const char *path)
373 {
374         struct dentry *root = wim_root_dentry(w);
375         while (*path == '/')
376                 path++;
377         return get_dentry_relative_path(root, path);
378 }
379
380 /* Returns the parent directory for the @path. */
381 struct dentry *get_parent_dentry(WIMStruct *w, const char *path)
382 {
383         size_t path_len = strlen(path);
384         char buf[path_len + 1];
385
386         memcpy(buf, path, path_len + 1);
387
388         to_parent_name(buf, path_len);
389
390         return get_dentry(w, buf);
391 }
392
393 /* Prints the full path of a dentry. */
394 int print_dentry_full_path(struct dentry *dentry, void *ignore)
395 {
396         if (dentry->full_path_utf8)
397                 puts(dentry->full_path_utf8);
398         return 0;
399 }
400
401 struct file_attr_flag {
402         u32 flag;
403         const char *name;
404 };
405 struct file_attr_flag file_attr_flags[] = {
406         {FILE_ATTRIBUTE_READONLY,               "READONLY"},
407         {FILE_ATTRIBUTE_HIDDEN,         "HIDDEN"},
408         {FILE_ATTRIBUTE_SYSTEM,         "SYSTEM"},
409         {FILE_ATTRIBUTE_DIRECTORY,              "DIRECTORY"},
410         {FILE_ATTRIBUTE_ARCHIVE,                "ARCHIVE"},
411         {FILE_ATTRIBUTE_DEVICE,         "DEVICE"},
412         {FILE_ATTRIBUTE_NORMAL,         "NORMAL"},
413         {FILE_ATTRIBUTE_TEMPORARY,              "TEMPORARY"},
414         {FILE_ATTRIBUTE_SPARSE_FILE,    "SPARSE_FILE"},
415         {FILE_ATTRIBUTE_REPARSE_POINT,  "REPARSE_POINT"},
416         {FILE_ATTRIBUTE_COMPRESSED,             "COMPRESSED"},
417         {FILE_ATTRIBUTE_OFFLINE,                "OFFLINE"},
418         {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,"NOT_CONTENT_INDEXED"},
419         {FILE_ATTRIBUTE_ENCRYPTED,              "ENCRYPTED"},
420         {FILE_ATTRIBUTE_VIRTUAL,                "VIRTUAL"},
421 };
422
423 /* Prints a directory entry.  @lookup_table is a pointer to the lookup table, or
424  * NULL if the resource entry for the dentry is not to be printed. */
425 int print_dentry(struct dentry *dentry, void *lookup_table)
426 {
427         struct lookup_table_entry *lte;
428         unsigned i;
429
430         printf("[DENTRY]\n");
431         printf("Length            = %"PRIu64"\n", dentry->length);
432         printf("Attributes        = 0x%x\n", dentry->attributes);
433         for (i = 0; i < ARRAY_LEN(file_attr_flags); i++)
434                 if (file_attr_flags[i].flag & dentry->attributes)
435                         printf("    FILE_ATTRIBUTE_%s is set\n",
436                                 file_attr_flags[i].name);
437         printf("Security ID       = %d\n", dentry->security_id);
438         printf("Subdir offset     = %"PRIu64"\n", dentry->subdir_offset);
439         /*printf("Unused1           = 0x%"PRIu64"\n", dentry->unused1);*/
440         /*printf("Unused2           = %"PRIu64"\n", dentry->unused2);*/
441         printf("Creation Time     = 0x%"PRIx64"\n", dentry->creation_time);
442         printf("Last Access Time  = 0x%"PRIx64"\n", dentry->last_access_time);
443         printf("Last Write Time   = 0x%"PRIx64"\n", dentry->last_write_time);
444         printf("Hash              = 0x"); 
445         print_hash(dentry->hash); 
446         putchar('\n');
447         printf("Reparse Tag       = 0x%"PRIx32"\n", dentry->reparse_tag);
448         printf("Hard Link Group   = 0x%"PRIx64"\n", dentry->hard_link);
449         printf("Number of Alternate Data Streams = %hu\n", dentry->num_ads);
450         printf("Filename          = \"");
451         print_string(dentry->file_name, dentry->file_name_len);
452         puts("\"");
453         printf("Filename Length   = %hu\n", dentry->file_name_len);
454         printf("Filename (UTF-8)  = \"%s\"\n", dentry->file_name_utf8);
455         printf("Filename (UTF-8) Length = %hu\n", dentry->file_name_utf8_len);
456         printf("Short Name        = \"");
457         print_string(dentry->short_name, dentry->short_name_len);
458         puts("\"");
459         printf("Short Name Length = %hu\n", dentry->short_name_len);
460         printf("Full Path (UTF-8) = \"%s\"\n", dentry->full_path_utf8);
461         if (lookup_table && (lte = __lookup_resource(lookup_table, dentry->hash)))
462                 print_lookup_table_entry(lte, NULL);
463         else
464                 putchar('\n');
465         for (u16 i = 0; i < dentry->num_ads; i++) {
466                 printf("[Alternate Stream Entry %u]\n", i);
467                 printf("Name = \"%s\"\n", dentry->ads_entries[i].stream_name_utf8);
468                 printf("Name Length (UTF-16) = %u\n",
469                                 dentry->ads_entries[i].stream_name_len);
470                 printf("Hash              = 0x"); 
471                 print_hash(dentry->ads_entries[i].hash); 
472                 if (lookup_table &&
473                      (lte = __lookup_resource(lookup_table,
474                                               dentry->ads_entries[i].hash)))
475                 {
476                         print_lookup_table_entry(lte, NULL);
477                 } else {
478                         putchar('\n');
479                 }
480         }
481         return 0;
482 }
483
484 static inline void dentry_common_init(struct dentry *dentry)
485 {
486         memset(dentry, 0, sizeof(struct dentry));
487         dentry->refcnt = 1;
488         dentry->security_id = -1;
489 }
490
491 /* 
492  * Creates an unlinked directory entry.
493  *
494  * @name:    The base name of the new dentry.
495  * @return:  A pointer to the new dentry, or NULL if out of memory.
496  */
497 struct dentry *new_dentry(const char *name)
498 {
499         struct dentry *dentry;
500         
501         dentry = MALLOC(sizeof(struct dentry));
502         if (!dentry)
503                 goto err;
504
505         dentry_common_init(dentry);
506         if (change_dentry_name(dentry, name) != 0)
507                 goto err;
508
509         dentry_update_all_timestamps(dentry);
510         dentry->next   = dentry;
511         dentry->prev   = dentry;
512         dentry->parent = dentry;
513         return dentry;
514 err:
515         FREE(dentry);
516         ERROR("Failed to allocate new dentry");
517         return NULL;
518 }
519
520 static void dentry_free_ads_entries(struct dentry *dentry)
521 {
522         for (u16 i = 0; i < dentry->num_ads; i++)
523                 destroy_ads_entry(&dentry->ads_entries[i]);
524         FREE(dentry->ads_entries);
525         dentry->ads_entries = NULL;
526         dentry->num_ads = 0;
527 }
528
529
530 void free_dentry(struct dentry *dentry)
531 {
532         FREE(dentry->file_name);
533         FREE(dentry->file_name_utf8);
534         FREE(dentry->short_name);
535         FREE(dentry->full_path_utf8);
536         dentry_free_ads_entries(dentry);
537         FREE(dentry);
538 }
539
540 /* Arguments for do_free_dentry(). */
541 struct free_dentry_args {
542         struct lookup_table *lookup_table;
543         bool lt_decrement_refcnt;
544 };
545
546 /* 
547  * This function is passed as an argument to for_dentry_in_tree_depth() in order
548  * to free a directory tree.  __args is a pointer to a `struct free_dentry_args'.
549  */
550 static int do_free_dentry(struct dentry *dentry, void *__args)
551 {
552         struct free_dentry_args *args = (struct free_dentry_args*)__args;
553
554         if (args->lt_decrement_refcnt && !dentry_is_directory(dentry)) {
555                 lookup_table_decrement_refcnt(args->lookup_table, 
556                                               dentry->hash);
557         }
558
559         wimlib_assert(dentry->refcnt >= 1);
560         if (--dentry->refcnt == 0)
561                 free_dentry(dentry);
562         return 0;
563 }
564
565 /* 
566  * Unlinks and frees a dentry tree.
567  *
568  * @root:               The root of the tree.
569  * @lookup_table:       The lookup table for dentries.
570  * @decrement_refcnt:   True if the dentries in the tree are to have their 
571  *                      reference counts in the lookup table decremented.
572  */
573 void free_dentry_tree(struct dentry *root, struct lookup_table *lookup_table, 
574                       bool lt_decrement_refcnt)
575 {
576         if (!root || !root->parent)
577                 return;
578
579         struct free_dentry_args args;
580         args.lookup_table        = lookup_table;
581         args.lt_decrement_refcnt = lt_decrement_refcnt;
582         for_dentry_in_tree_depth(root, do_free_dentry, &args);
583 }
584
585 int increment_dentry_refcnt(struct dentry *dentry, void *ignore)
586 {
587         dentry->refcnt++;
588         return 0;
589 }
590
591 /* 
592  * Links a dentry into the directory tree.
593  *
594  * @dentry: The dentry to link.
595  * @parent: The dentry that will be the parent of @dentry.
596  */
597 void link_dentry(struct dentry *dentry, struct dentry *parent)
598 {
599         dentry->parent = parent;
600         if (parent->children) {
601                 /* Not an only child; link to siblings. */
602                 dentry->next = parent->children;
603                 dentry->prev = parent->children->prev;
604                 dentry->next->prev = dentry;
605                 dentry->prev->next = dentry;
606         } else {
607                 /* Only child; link to parent. */
608                 parent->children = dentry;
609                 dentry->next = dentry;
610                 dentry->prev = dentry;
611         }
612 }
613
614 /* Unlink a dentry from the directory tree. */
615 void unlink_dentry(struct dentry *dentry)
616 {
617         if (dentry_is_root(dentry))
618                 return;
619         if (dentry_is_only_child(dentry)) {
620                 dentry->parent->children = NULL;
621         } else {
622                 if (dentry_is_first_sibling(dentry))
623                         dentry->parent->children = dentry->next;
624                 dentry->next->prev = dentry->prev;
625                 dentry->prev->next = dentry->next;
626         }
627 }
628
629
630 /* Recalculates the length of @dentry based on its file name length and short
631  * name length.  */
632 static inline void recalculate_dentry_size(struct dentry *dentry)
633 {
634         dentry->length = WIM_DENTRY_DISK_SIZE + dentry->file_name_len + 
635                          2 + dentry->short_name_len;
636         /* Must be multiple of 8. */
637         dentry->length = (dentry->length + 7) & ~7;
638 }
639
640 static int do_name_change(char **file_name_ret,
641                           char **file_name_utf8_ret,
642                           u16 *file_name_len_ret,
643                           u16 *file_name_utf8_len_ret,
644                           const char *new_name)
645 {
646         size_t utf8_len;
647         size_t utf16_len;
648         char *file_name, *file_name_utf8;
649
650         utf8_len = strlen(new_name);
651
652         file_name = utf8_to_utf16(new_name, utf8_len, &utf16_len);
653
654         if (!file_name)
655                 return WIMLIB_ERR_NOMEM;
656
657         file_name_utf8 = MALLOC(utf8_len + 1);
658         if (!file_name_utf8) {
659                 FREE(file_name);
660                 return WIMLIB_ERR_NOMEM;
661         }
662         memcpy(file_name_utf8, new_name, utf8_len + 1);
663
664         FREE(*file_name_ret);
665         FREE(*file_name_utf8_ret);
666         *file_name_ret          = file_name;
667         *file_name_utf8_ret     = file_name_utf8;
668         *file_name_len_ret      = utf16_len;
669         *file_name_utf8_len_ret = utf8_len;
670         return 0;
671 }
672
673 /* Changes the name of a dentry to @new_name.  Only changes the file_name and
674  * file_name_utf8 fields; does not change the short_name, short_name_utf8, or
675  * full_path_utf8 fields.  Also recalculates its length. */
676 int change_dentry_name(struct dentry *dentry, const char *new_name)
677 {
678         int ret;
679
680         ret = do_name_change(&dentry->file_name, &dentry->file_name_utf8,
681                              &dentry->file_name_len, &dentry->file_name_utf8_len,
682                              new_name);
683         if (ret == 0)
684                 recalculate_dentry_size(dentry);
685         return ret;
686 }
687
688 int change_ads_name(struct ads_entry *entry, const char *new_name)
689 {
690         return do_name_change(&entry->stream_name, &entry->stream_name_utf8,
691                               &entry->stream_name_len,
692                               &entry->stream_name_utf8_len,
693                               new_name);
694 }
695
696 /* Parameters for calculate_dentry_statistics(). */
697 struct image_statistics {
698         struct lookup_table *lookup_table;
699         u64 *dir_count;
700         u64 *file_count;
701         u64 *total_bytes;
702         u64 *hard_link_bytes;
703 };
704
705 static int calculate_dentry_statistics(struct dentry *dentry, void *arg)
706 {
707         struct image_statistics *stats;
708         struct lookup_table_entry *lte; 
709         u16 i;
710         
711         stats = arg;
712
713         if (dentry_is_directory(dentry) && !dentry_is_root(dentry))
714                 ++*stats->dir_count;
715         else
716                 ++*stats->file_count;
717
718         lte = __lookup_resource(stats->lookup_table, dentry->hash);
719         i = 0;
720         while (1) {
721                 if (lte) {
722                         u64 size = lte->resource_entry.original_size;
723                         *stats->total_bytes += size;
724                         if (++lte->out_refcnt == 1)
725                                 *stats->hard_link_bytes += size;
726                 }
727                 if (i == dentry->num_ads)
728                         break;
729                 lte = __lookup_resource(stats->lookup_table,
730                                         dentry->ads_entries[i].hash);
731                 i++;
732         }
733
734         return 0;
735 }
736
737 void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table, 
738                                    u64 *dir_count_ret, u64 *file_count_ret, 
739                                    u64 *total_bytes_ret, 
740                                    u64 *hard_link_bytes_ret)
741 {
742         struct image_statistics stats;
743         *dir_count_ret         = 0;
744         *file_count_ret        = 0;
745         *total_bytes_ret       = 0;
746         *hard_link_bytes_ret   = 0;
747         stats.lookup_table     = table;
748         stats.dir_count       = dir_count_ret;
749         stats.file_count      = file_count_ret;
750         stats.total_bytes     = total_bytes_ret;
751         stats.hard_link_bytes = hard_link_bytes_ret;
752         for_lookup_table_entry(table, zero_out_refcnts, NULL);
753         for_dentry_in_tree(root, calculate_dentry_statistics, &stats);
754 }
755
756 static int read_ads_entries(const u8 *p, struct dentry *dentry,
757                             u64 remaining_size)
758 {
759         u16 num_ads = dentry->num_ads;
760         struct ads_entry *ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
761         int ret;
762         if (!ads_entries) {
763                 ERROR("Could not allocate memory for %"PRIu16" "
764                       "alternate data stream entries", num_ads);
765                 return WIMLIB_ERR_NOMEM;
766         }
767         DEBUG2("Reading %"PRIu16" alternate data streams "
768                "(remaining size = %"PRIu64")", num_ads, remaining_size);
769
770         for (u16 i = 0; i < num_ads; i++) {
771                 struct ads_entry *cur_entry = &ads_entries[i];
772                 u64 length;
773                 size_t utf8_len;
774                 const char *p_save = p;
775                 /* Read the base stream entry, excluding the stream name. */
776                 if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
777                         ERROR("Stream entries go past end of metadata resource");
778                         ERROR("(remaining_size = %"PRIu64")", remaining_size);
779                         ret = WIMLIB_ERR_INVALID_DENTRY;
780                         goto out_free_ads_entries;
781                 }
782                 remaining_size -= WIM_ADS_ENTRY_DISK_SIZE;
783
784                 p = get_u64(p, &length); /* ADS entry length */
785
786                 DEBUG2("ADS length = %"PRIu64, length);
787
788                 p += 8; /* Unused */
789                 p = get_bytes(p, WIM_HASH_SIZE, (u8*)cur_entry->hash);
790                 p = get_u16(p, &cur_entry->stream_name_len);
791
792                 DEBUG2("Stream name length = %u", cur_entry->stream_name_len);
793
794                 cur_entry->stream_name = NULL;
795                 cur_entry->stream_name_utf8 = NULL;
796
797                 if (remaining_size < cur_entry->stream_name_len + 2) {
798                         ERROR("Stream entries go past end of metadata resource");
799                         ERROR("(remaining_size = %"PRIu64" bytes, stream_name_len "
800                               "= %"PRIu16" bytes", remaining_size,
801                               cur_entry->stream_name_len);
802                         ret = WIMLIB_ERR_INVALID_DENTRY;
803                         goto out_free_ads_entries;
804                 }
805                 remaining_size -= cur_entry->stream_name_len + 2;
806
807                 cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
808                 if (!cur_entry->stream_name) {
809                         ret = WIMLIB_ERR_NOMEM;
810                         goto out_free_ads_entries;
811                 }
812                 get_bytes(p, cur_entry->stream_name_len,
813                           (u8*)cur_entry->stream_name);
814                 cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
815                                                             cur_entry->stream_name_len,
816                                                             &utf8_len);
817                 cur_entry->stream_name_utf8_len = utf8_len;
818
819                 if (!cur_entry->stream_name_utf8) {
820                         ret = WIMLIB_ERR_NOMEM;
821                         goto out_free_ads_entries;
822                 }
823                 p = p_save + ads_entry_length(cur_entry);
824         }
825         dentry->ads_entries = ads_entries;
826         return 0;
827 out_free_ads_entries:
828         for (u16 i = 0; i < num_ads; i++) {
829                 FREE(ads_entries[i].stream_name);
830                 FREE(ads_entries[i].stream_name_utf8);
831         }
832         FREE(ads_entries);
833         return ret;
834 }
835
836 /* 
837  * Reads a directory entry from the metadata resource.
838  */
839 int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, 
840                 u64 offset, struct dentry *dentry)
841 {
842         const u8 *p;
843         u64 calculated_size;
844         char *file_name;
845         char *file_name_utf8;
846         char *short_name;
847         u16 short_name_len;
848         u16 file_name_len;
849         size_t file_name_utf8_len;
850         int ret;
851
852         dentry_common_init(dentry);
853
854         /*Make sure the dentry really fits into the metadata resource.*/
855         if (offset + 8 > metadata_resource_len) {
856                 ERROR("Directory entry starting at %"PRIu64" ends past the "
857                       "end of the metadata resource (size %"PRIu64")",
858                       offset, metadata_resource_len);
859                 return WIMLIB_ERR_INVALID_DENTRY;
860         }
861
862         /* Before reading the whole entry, we need to read just the length.
863          * This is because an entry of length 8 (that is, just the length field)
864          * terminates the list of sibling directory entries. */
865
866         p = get_u64(&metadata_resource[offset], &dentry->length);
867
868         /* A zero length field (really a length of 8, since that's how big the
869          * directory entry is...) indicates that this is the end of directory
870          * dentry.  We do not read it into memory as an actual dentry, so just
871          * return true in that case. */
872         if (dentry->length == 0)
873                 return 0;
874
875         if (offset + dentry->length >= metadata_resource_len) {
876                 ERROR("Directory entry at offset %"PRIu64" and with size "
877                       "%"PRIu64" ends past the end of the metadata resource "
878                       "(size %"PRIu64")",
879                       offset, dentry->length, metadata_resource_len);
880                 return WIMLIB_ERR_INVALID_DENTRY;
881         }
882
883         /* If it is a recognized length, read the rest of the directory entry.
884          * Note: The root directory entry has no name, and its length does not
885          * include the short name length field.  */
886         if (dentry->length < WIM_DENTRY_DISK_SIZE) {
887                 ERROR("Directory entry has invalid length of %"PRIu64" bytes",
888                       dentry->length);
889                 return WIMLIB_ERR_INVALID_DENTRY;
890         }
891
892         p = get_u32(p, &dentry->attributes);
893         p = get_u32(p, (u32*)&dentry->security_id);
894         p = get_u64(p, &dentry->subdir_offset);
895
896         /* 2 unused fields */
897         p += 2 * sizeof(u64);
898         /*p = get_u64(p, &dentry->unused1);*/
899         /*p = get_u64(p, &dentry->unused2);*/
900
901         p = get_u64(p, &dentry->creation_time);
902         p = get_u64(p, &dentry->last_access_time);
903         p = get_u64(p, &dentry->last_write_time);
904
905         p = get_bytes(p, WIM_HASH_SIZE, dentry->hash);
906         
907         /*
908          * I don't know what's going on here.  It seems like M$ screwed up the
909          * reparse points, then put the fields in the same place and didn't
910          * document it.  The WIM_HDR_FLAG_RP_FIX flag in the WIM header might
911          * have something to do with this, but it's not documented.
912          */
913         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
914                 /* ??? */
915                 u32 u1, u2;
916                 p = get_u32(p, &u1);
917                 /*p += 4;*/
918                 p = get_u32(p, &dentry->reparse_tag);
919                 p = get_u32(p, &u2);
920                 /*p += 4;*/
921                 dentry->hard_link = (u64)(u1) | ((u64)(u2) << 32);
922         } else {
923                 p = get_u32(p, &dentry->reparse_tag);
924                 p = get_u64(p, &dentry->hard_link);
925         }
926
927         /* By the way, the reparse_reserved field does not actually exist (at
928          * least when the file is not a reparse point) */
929
930         
931         p = get_u16(p, &dentry->num_ads);
932
933         p = get_u16(p, &short_name_len);
934         p = get_u16(p, &file_name_len);
935
936         calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
937                           short_name_len;
938
939         if (dentry->length < calculated_size) {
940                 ERROR("Unexpected end of directory entry! (Expected "
941                       "%"PRIu64" bytes, got %"PRIu64" bytes. "
942                       "short_name_len = %hu, file_name_len = %hu)", 
943                       calculated_size, dentry->length,
944                       short_name_len, file_name_len);
945                 return WIMLIB_ERR_INVALID_DENTRY;
946         }
947
948         /* Read the filename. */
949         file_name = MALLOC(file_name_len);
950         if (!file_name) {
951                 ERROR("Failed to allocate %hu bytes for dentry file name",
952                       file_name_len);
953                 return WIMLIB_ERR_NOMEM;
954         }
955         p = get_bytes(p, file_name_len, file_name);
956
957         /* Convert filename to UTF-8. */
958         file_name_utf8 = utf16_to_utf8(file_name, file_name_len, 
959                                        &file_name_utf8_len);
960
961         if (!file_name_utf8) {
962                 ERROR("Failed to allocate memory to convert UTF-16 "
963                       "filename (%hu bytes) to UTF-8", file_name_len);
964                 ret = WIMLIB_ERR_NOMEM;
965                 goto out_free_file_name;
966         }
967
968         /* Undocumented padding between file name and short name.  This probably
969          * is supposed to be a terminating null character. */
970         p += 2;
971
972         /* Read the short filename. */
973         short_name = MALLOC(short_name_len);
974         if (!short_name) {
975                 ERROR("Failed to allocate %hu bytes for short filename",
976                       short_name_len);
977                 ret = WIMLIB_ERR_NOMEM;
978                 goto out_free_file_name_utf8;
979         }
980
981         p = get_bytes(p, short_name_len, short_name);
982
983         /* Some directory entries inexplicibly have a little over 70 bytes of
984          * extra data.  The exact amount of data seems to be 72 bytes, but it is
985          * aligned on the next 8-byte boundary.  Here's an example of the
986          * aligned data:
987          *
988          * 01000000400000006c786bbac58ede11b0bb00261870892ab6adb76fe63a3
989          * e468fca86530d2effa16c786bbac58ede11b0bb00261870892a0000000000
990          * 0000000000000000000000
991          *
992          * Here's one interpretation of how the data is laid out.
993          *
994          * struct unknown {
995          *      u32 field1; (always 0x00000001)
996          *      u32 field2; (always 0x40000000)
997          *      u16 field3;
998          *      u32 field4;
999          *      u32 field5;
1000          *      u32 field6;
1001          *      u8  data[48]; (???)
1002          *      u64 reserved1; (always 0)
1003          *      u64 reserved2; (always 0)
1004          * };*/
1005 #if 0
1006         if (dentry->length - calculated_size >= WIM_ADS_ENTRY_DISK_SIZE) {
1007                 printf("%s: %lu / %lu (", file_name_utf8, 
1008                                 calculated_size, dentry->length);
1009                 print_string(p + WIM_ADS_ENTRY_DISK_SIZE, dentry->length - calculated_size - WIM_ADS_ENTRY_DISK_SIZE);
1010                 puts(")");
1011                 print_byte_field(p, dentry->length - calculated_size);
1012                 putchar('\n');
1013         }
1014 #endif
1015
1016         if (dentry->num_ads != 0) {
1017                 calculated_size = (calculated_size + 7) & ~7;
1018                 if (calculated_size > metadata_resource_len - offset) {
1019                         ERROR("Not enough space in metadata resource for "
1020                               "alternate stream entries");
1021                         ret = WIMLIB_ERR_INVALID_DENTRY;
1022                         goto out_free_short_name;
1023                 }
1024                 ret = read_ads_entries(&metadata_resource[offset + calculated_size],
1025                                        dentry,
1026                                        metadata_resource_len - offset - calculated_size);
1027                 if (ret != 0)
1028                         goto out_free_short_name;
1029         }
1030
1031         dentry->file_name          = file_name;
1032         dentry->file_name_utf8     = file_name_utf8;
1033         dentry->short_name         = short_name;
1034         dentry->file_name_len      = file_name_len;
1035         dentry->file_name_utf8_len = file_name_utf8_len;
1036         dentry->short_name_len     = short_name_len;
1037         return 0;
1038 out_free_short_name:
1039         FREE(short_name);
1040 out_free_file_name_utf8:
1041         FREE(file_name_utf8);
1042 out_free_file_name:
1043         FREE(file_name);
1044         return ret;
1045 }
1046
1047 /* 
1048  * Writes a dentry to an output buffer.
1049  *
1050  * @dentry:  The dentry structure.
1051  * @p:       The memory location to write the data to.
1052  * @return:  Pointer to the byte after the last byte we wrote as part of the
1053  *              dentry.
1054  */
1055 static u8 *write_dentry(const struct dentry *dentry, u8 *p)
1056 {
1057         u8 *orig_p = p;
1058         unsigned padding;
1059         memset(p, 0, dentry->length);
1060         p = put_u64(p, dentry->length);
1061         p = put_u32(p, dentry->attributes);
1062         p = put_u32(p, dentry->security_id);
1063         p = put_u64(p, dentry->subdir_offset);
1064         p = put_u64(p, 0); /* unused1 */
1065         p = put_u64(p, 0); /* unused2 */
1066         p = put_u64(p, dentry->creation_time);
1067         p = put_u64(p, dentry->last_access_time);
1068         p = put_u64(p, dentry->last_write_time);
1069         memcpy(p, dentry->hash, WIM_HASH_SIZE);
1070         p += WIM_HASH_SIZE;
1071         p = put_u32(p, dentry->reparse_tag);
1072         p = put_u64(p, dentry->hard_link);
1073         p = put_u16(p, dentry->num_ads); /*streams */
1074         p = put_u16(p, dentry->short_name_len);
1075         p = put_u16(p, dentry->file_name_len);
1076         p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
1077         p = put_u16(p, 0); /* filename padding, 2 bytes. */
1078         p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
1079
1080         wimlib_assert(p - orig_p <= dentry->length);
1081         if (p - orig_p < dentry->length)
1082                 p = put_zeroes(p, dentry->length - (p - orig_p));
1083
1084         p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1085
1086         for (u16 i = 0; i < dentry->num_ads; i++) {
1087                 p = put_u64(p, ads_entry_length(&dentry->ads_entries[i]));
1088                 p = put_u64(p, 0); /* Unused */
1089                 p = put_bytes(p, WIM_HASH_SIZE, dentry->ads_entries[i].hash);
1090                 p = put_u16(p, dentry->ads_entries[i].stream_name_len);
1091                 p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
1092                                  (u8*)dentry->ads_entries[i].stream_name);
1093                 p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1094         }
1095         return p;
1096 }
1097
1098 /* Recursive function that writes a dentry tree rooted at @tree, not including
1099  * @tree itself, which has already been written, except in the case of the root
1100  * dentry, which is written right away, along with an end-of-directory entry. */
1101 u8 *write_dentry_tree(const struct dentry *tree, u8 *p)
1102 {
1103         const struct dentry *child;
1104
1105         if (dentry_is_root(tree)) {
1106                 p = write_dentry(tree, p);
1107
1108                 /* write end of directory entry */
1109                 p = put_u64(p, 0);
1110         } else {
1111                 /* Nothing to do for a regular file. */
1112                 if (dentry_is_regular_file(tree))
1113                         return p;
1114         }
1115
1116         /* Write child dentries and end-of-directory entry. */
1117         child = tree->children;
1118         if (child) {
1119                 do {
1120                         p = write_dentry(child, p);
1121                         child = child->next;
1122                 } while (child != tree->children);
1123         }
1124
1125         /* write end of directory entry */
1126         p = put_u64(p, 0);
1127
1128         /* Recurse on children. */
1129         if (child) {
1130                 do {
1131                         p = write_dentry_tree(child, p);
1132                         child = child->next;
1133                 } while (child != tree->children);
1134         }
1135         return p;
1136 }
1137
1138 /* Reads the children of a dentry, and all their children, ..., etc. from the
1139  * metadata resource and into the dentry tree.
1140  *
1141  * @metadata_resource:  An array that contains the uncompressed metadata
1142  *                      resource for the WIM file.
1143  * @metadata_resource_len:      The length of @metadata_resource.
1144  * @dentry:     A pointer to a struct dentry that is the root of the directory
1145  *              tree and has already been read from the metadata resource.  It
1146  *              does not need to be the real root because this procedure is
1147  *              called recursively.
1148  *
1149  * @return:     Zero on success, nonzero on failure.
1150  */
1151 int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
1152                      struct dentry *dentry)
1153 {
1154         u64 cur_offset = dentry->subdir_offset;
1155         struct dentry *prev_child = NULL;
1156         struct dentry *first_child = NULL;
1157         struct dentry *child;
1158         struct dentry cur_child;
1159         int ret;
1160
1161         /* If @dentry is a regular file, nothing more needs to be done for this
1162          * branch. */
1163         if (cur_offset == 0)
1164                 return 0;
1165
1166         /* Find and read all the children of @dentry. */
1167         while (1) {
1168
1169                 /* Read next child of @dentry into @cur_child. */
1170                 ret = read_dentry(metadata_resource, metadata_resource_len, 
1171                                   cur_offset, &cur_child);
1172                 if (ret != 0)
1173                         break;
1174
1175                 /* Check for end of directory. */
1176                 if (cur_child.length == 0) {
1177                         ret = 0;
1178                         break;
1179                 }
1180
1181                 /* Not end of directory.  Allocate this child permanently and
1182                  * link it to the parent and previous child. */
1183                 child = MALLOC(sizeof(struct dentry));
1184                 if (!child) {
1185                         ERROR("Failed to allocate %zu bytes for new dentry",
1186                               sizeof(struct dentry));
1187                         ret = WIMLIB_ERR_NOMEM;
1188                         break;
1189                 }
1190                 memcpy(child, &cur_child, sizeof(struct dentry));
1191
1192                 if (prev_child) {
1193                         prev_child->next = child;
1194                         child->prev = prev_child;
1195                 } else {
1196                         first_child = child;
1197                 }
1198
1199                 child->parent = dentry;
1200                 prev_child = child;
1201
1202                 /* If there are children of this child, call this procedure
1203                  * recursively. */
1204                 if (child->subdir_offset != 0) {
1205                         ret = read_dentry_tree(metadata_resource, 
1206                                                metadata_resource_len, child);
1207                         if (ret != 0)
1208                                 break;
1209                 }
1210
1211                 /* Advance to the offset of the next child. */
1212                 cur_offset += dentry_total_length(child);
1213         }
1214
1215         /* Link last child to first one, and set parent's
1216          * children pointer to the first child.  */
1217         if (prev_child) {
1218                 prev_child->next = first_child;
1219                 first_child->prev = prev_child;
1220         }
1221         dentry->children = first_child;
1222         return ret;
1223 }
1224
1225 int dentry_set_symlink_buf(struct dentry *dentry, const u8 symlink_buf_hash[])
1226 {
1227         struct ads_entry *ads_entries;
1228
1229         ads_entries = CALLOC(2, sizeof(struct ads_entry));
1230         if (!ads_entries)
1231                 return WIMLIB_ERR_NOMEM;
1232         memcpy(ads_entries[1].hash, symlink_buf_hash, WIM_HASH_SIZE);
1233         dentry_free_ads_entries(dentry);
1234         dentry->num_ads = 2;
1235         dentry->ads_entries = ads_entries;
1236         return 0;
1237 }