clone_dentry(): Set name fields to NULL
[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 /* Real length of a dentry, including the alternate data stream entries, which
55  * are not included in the dentry->length field... */
56 u64 dentry_total_length(const struct dentry *dentry)
57 {
58         u64 length = (dentry->length + 7) & ~7;
59         for (u16 i = 0 ; i < dentry->num_ads; i++)
60                 length += ads_entry_length(&dentry->ads_entries[i]);
61         return length;
62 }
63
64 /* Transfers file attributes from a `stat' buffer to a struct dentry. */
65 void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry)
66 {
67         if (S_ISLNK(stbuf->st_mode)) {
68                 dentry->attributes = FILE_ATTRIBUTE_REPARSE_POINT;
69                 dentry->reparse_tag = WIM_IO_REPARSE_TAG_SYMLINK;
70         } else if (S_ISDIR(stbuf->st_mode)) {
71                 dentry->attributes = FILE_ATTRIBUTE_DIRECTORY;
72         } else {
73                 dentry->attributes = FILE_ATTRIBUTE_NORMAL;
74         }
75         if (sizeof(ino_t) >= 8)
76                 dentry->hard_link = (u64)stbuf->st_ino;
77         else
78                 dentry->hard_link = (u64)stbuf->st_ino |
79                                    ((u64)stbuf->st_dev << (sizeof(ino_t) * 8));
80 }
81
82 /* Transfers file attributes from a struct dentry to a `stat' buffer. */
83 void dentry_to_stbuf(const struct dentry *dentry, struct stat *stbuf, 
84                      const struct lookup_table *table)
85 {
86         struct lookup_table_entry *lte;
87
88         if (dentry_is_symlink(dentry))
89                 stbuf->st_mode = S_IFLNK | 0777;
90         else if (dentry_is_directory(dentry))
91                 stbuf->st_mode = S_IFDIR | 0755;
92         else
93                 stbuf->st_mode = S_IFREG | 0644;
94
95         /* Use the size of the unnamed (default) file stream. */
96         if (table && (lte = __lookup_resource(table, dentry_hash(dentry))))
97                 stbuf->st_size = lte->resource_entry.original_size;
98         else
99                 stbuf->st_size = 0;
100
101         stbuf->st_nlink   = dentry_link_group_size(dentry);
102         stbuf->st_ino     = dentry->hard_link;
103         stbuf->st_uid     = getuid();
104         stbuf->st_gid     = getgid();
105         stbuf->st_atime   = ms_timestamp_to_unix(dentry->last_access_time);
106         stbuf->st_mtime   = ms_timestamp_to_unix(dentry->last_write_time);
107         stbuf->st_ctime   = ms_timestamp_to_unix(dentry->creation_time);
108         stbuf->st_blocks  = (stbuf->st_size + 511) / 512;
109 }
110
111 /* Makes all timestamp fields for the dentry be the current time. */
112 void dentry_update_all_timestamps(struct dentry *dentry)
113 {
114         u64 now = get_timestamp();
115         dentry->creation_time    = now;
116         dentry->last_access_time = now;
117         dentry->last_write_time  = now;
118 }
119
120 struct ads_entry *dentry_get_ads_entry(struct dentry *dentry,
121                                        const char *stream_name)
122 {
123         size_t stream_name_len = strlen(stream_name);
124         if (!stream_name)
125                 return NULL;
126         for (u16 i = 0; i < dentry->num_ads; i++)
127                 if (ads_entry_has_name(&dentry->ads_entries[i],
128                                        stream_name, stream_name_len))
129                         return &dentry->ads_entries[i];
130         return NULL;
131 }
132
133 /* Add an alternate stream entry to a dentry and return a pointer to it, or NULL
134  * on failure. */
135 struct ads_entry *dentry_add_ads(struct dentry *dentry, const char *stream_name)
136 {
137         u16 num_ads = dentry->num_ads + 1;
138         struct ads_entry *ads_entries;
139         struct ads_entry *new_entry;
140         if (num_ads == 0xffff)
141                 return NULL;
142         ads_entries = MALLOC(num_ads * sizeof(struct ads_entry));
143         if (!ads_entries)
144                 return NULL;
145
146         new_entry = &ads_entries[num_ads - 1];
147         if (change_ads_name(new_entry, stream_name) != 0) {
148                 FREE(ads_entries);
149                 return NULL;
150         }
151
152         memcpy(ads_entries, dentry->ads_entries,
153                (num_ads - 1) * sizeof(struct ads_entry));
154         FREE(dentry->ads_entries);
155         dentry->ads_entries = ads_entries;
156         dentry->num_ads = num_ads;
157         return memset(new_entry, 0, sizeof(struct ads_entry));
158 }
159
160 void dentry_remove_ads(struct dentry *dentry, struct ads_entry *sentry)
161 {
162         destroy_ads_entry(sentry);
163         memcpy(sentry, sentry + 1,
164                (dentry->num_ads - (sentry - dentry->ads_entries))
165                  * sizeof(struct ads_entry));
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 }
478
479 /* 
480  * Creates an unlinked directory entry.
481  *
482  * @name:    The base name of the new dentry.
483  * @return:  A pointer to the new dentry, or NULL if out of memory.
484  */
485 struct dentry *new_dentry(const char *name)
486 {
487         struct dentry *dentry;
488         
489         dentry = MALLOC(sizeof(struct dentry));
490         if (!dentry)
491                 goto err;
492
493         dentry_common_init(dentry);
494         if (change_dentry_name(dentry, name) != 0)
495                 goto err;
496
497         dentry_update_all_timestamps(dentry);
498         dentry->next   = dentry;
499         dentry->prev   = dentry;
500         dentry->parent = dentry;
501         return dentry;
502 err:
503         FREE(dentry);
504         ERROR("Failed to allocate new dentry");
505         return NULL;
506 }
507
508 void dentry_free_ads_entries(struct dentry *dentry)
509 {
510         for (u16 i = 0; i < dentry->num_ads; i++)
511                 destroy_ads_entry(&dentry->ads_entries[i]);
512         FREE(dentry->ads_entries);
513         dentry->ads_entries = NULL;
514         dentry->num_ads = 0;
515 }
516
517
518 void free_dentry(struct dentry *dentry)
519 {
520         FREE(dentry->file_name);
521         FREE(dentry->file_name_utf8);
522         FREE(dentry->short_name);
523         FREE(dentry->full_path_utf8);
524         dentry_free_ads_entries(dentry);
525         FREE(dentry);
526 }
527
528 /* clones a dentry.
529  *
530  * Beware:
531  *      - memory for file names is not cloned
532  *      - next, prev, and children pointers and not touched
533  *      - stream entries are not cloned.
534  */
535 struct dentry *clone_dentry(struct dentry *old)
536 {
537         struct dentry *new = MALLOC(sizeof(struct dentry));
538         if (!new)
539                 return NULL;
540         memcpy(new, old, sizeof(struct dentry));
541         new->file_name          = NULL;
542         new->file_name_len      = 0;
543         new->file_name_utf8     = NULL;
544         new->file_name_utf8_len = 0;
545         new->short_name         = NULL;
546         new->short_name_len     = 0;
547         return new;
548 }
549
550 /* Arguments for do_free_dentry(). */
551 struct free_dentry_args {
552         struct lookup_table *lookup_table;
553         bool lt_decrement_refcnt;
554 };
555
556 /* 
557  * This function is passed as an argument to for_dentry_in_tree_depth() in order
558  * to free a directory tree.  __args is a pointer to a `struct free_dentry_args'.
559  */
560 static int do_free_dentry(struct dentry *dentry, void *__args)
561 {
562         struct free_dentry_args *args = (struct free_dentry_args*)__args;
563
564         if (args->lt_decrement_refcnt && !dentry_is_directory(dentry)) {
565                 lookup_table_decrement_refcnt(args->lookup_table, 
566                                               dentry->hash);
567         }
568
569         wimlib_assert(dentry->refcnt >= 1);
570         if (--dentry->refcnt == 0)
571                 free_dentry(dentry);
572         return 0;
573 }
574
575 /* 
576  * Unlinks and frees a dentry tree.
577  *
578  * @root:               The root of the tree.
579  * @lookup_table:       The lookup table for dentries.
580  * @decrement_refcnt:   True if the dentries in the tree are to have their 
581  *                      reference counts in the lookup table decremented.
582  */
583 void free_dentry_tree(struct dentry *root, struct lookup_table *lookup_table, 
584                       bool lt_decrement_refcnt)
585 {
586         if (!root || !root->parent)
587                 return;
588
589         struct free_dentry_args args;
590         args.lookup_table        = lookup_table;
591         args.lt_decrement_refcnt = lt_decrement_refcnt;
592         for_dentry_in_tree_depth(root, do_free_dentry, &args);
593 }
594
595 int increment_dentry_refcnt(struct dentry *dentry, void *ignore)
596 {
597         dentry->refcnt++;
598         return 0;
599 }
600
601 /* 
602  * Links a dentry into the directory tree.
603  *
604  * @dentry: The dentry to link.
605  * @parent: The dentry that will be the parent of @dentry.
606  */
607 void link_dentry(struct dentry *dentry, struct dentry *parent)
608 {
609         dentry->parent = parent;
610         if (parent->children) {
611                 /* Not an only child; link to siblings. */
612                 dentry->next = parent->children;
613                 dentry->prev = parent->children->prev;
614                 dentry->next->prev = dentry;
615                 dentry->prev->next = dentry;
616         } else {
617                 /* Only child; link to parent. */
618                 parent->children = dentry;
619                 dentry->next = dentry;
620                 dentry->prev = dentry;
621         }
622 }
623
624 /* Unlink a dentry from the directory tree. */
625 void unlink_dentry(struct dentry *dentry)
626 {
627         if (dentry_is_root(dentry))
628                 return;
629         if (dentry_is_only_child(dentry)) {
630                 dentry->parent->children = NULL;
631         } else {
632                 if (dentry_is_first_sibling(dentry))
633                         dentry->parent->children = dentry->next;
634                 dentry->next->prev = dentry->prev;
635                 dentry->prev->next = dentry->next;
636         }
637 }
638
639
640 /* Recalculates the length of @dentry based on its file name length and short
641  * name length.  */
642 static inline void recalculate_dentry_size(struct dentry *dentry)
643 {
644         dentry->length = WIM_DENTRY_DISK_SIZE + dentry->file_name_len + 
645                          2 + dentry->short_name_len;
646         /* Must be multiple of 8. */
647         dentry->length = (dentry->length + 7) & ~7;
648 }
649
650 static int do_name_change(char **file_name_ret,
651                           char **file_name_utf8_ret,
652                           u16 *file_name_len_ret,
653                           u16 *file_name_utf8_len_ret,
654                           const char *new_name)
655 {
656         size_t utf8_len;
657         size_t utf16_len;
658         char *file_name, *file_name_utf8;
659
660         utf8_len = strlen(new_name);
661
662         file_name = utf8_to_utf16(new_name, utf8_len, &utf16_len);
663
664         if (!file_name)
665                 return WIMLIB_ERR_NOMEM;
666
667         file_name_utf8 = MALLOC(utf8_len + 1);
668         if (!file_name_utf8) {
669                 FREE(file_name);
670                 return WIMLIB_ERR_NOMEM;
671         }
672         memcpy(file_name_utf8, new_name, utf8_len + 1);
673
674         FREE(*file_name_ret);
675         FREE(*file_name_utf8_ret);
676         *file_name_ret          = file_name;
677         *file_name_utf8_ret     = file_name_utf8;
678         *file_name_len_ret      = utf16_len;
679         *file_name_utf8_len_ret = utf8_len;
680         return 0;
681 }
682
683 /* Changes the name of a dentry to @new_name.  Only changes the file_name and
684  * file_name_utf8 fields; does not change the short_name, short_name_utf8, or
685  * full_path_utf8 fields.  Also recalculates its length. */
686 int change_dentry_name(struct dentry *dentry, const char *new_name)
687 {
688         int ret;
689
690         ret = do_name_change(&dentry->file_name, &dentry->file_name_utf8,
691                              &dentry->file_name_len, &dentry->file_name_utf8_len,
692                              new_name);
693         if (ret == 0)
694                 recalculate_dentry_size(dentry);
695         return ret;
696 }
697
698 int change_ads_name(struct ads_entry *entry, const char *new_name)
699 {
700         return do_name_change(&entry->stream_name, &entry->stream_name_utf8,
701                               &entry->stream_name_len,
702                               &entry->stream_name_utf8_len,
703                               new_name);
704 }
705
706 /* Parameters for calculate_dentry_statistics(). */
707 struct image_statistics {
708         struct lookup_table *lookup_table;
709         u64 *dir_count;
710         u64 *file_count;
711         u64 *total_bytes;
712         u64 *hard_link_bytes;
713 };
714
715 static int calculate_dentry_statistics(struct dentry *dentry, void *arg)
716 {
717         struct image_statistics *stats;
718         struct lookup_table_entry *lte; 
719         u16 i;
720         
721         stats = arg;
722
723         if (dentry_is_directory(dentry) && !dentry_is_root(dentry))
724                 ++*stats->dir_count;
725         else
726                 ++*stats->file_count;
727
728         lte = __lookup_resource(stats->lookup_table, dentry->hash);
729         i = 0;
730         while (1) {
731                 if (lte) {
732                         u64 size = lte->resource_entry.original_size;
733                         *stats->total_bytes += size;
734                         if (++lte->out_refcnt == 1)
735                                 *stats->hard_link_bytes += size;
736                 }
737                 if (i == dentry->num_ads)
738                         break;
739                 lte = __lookup_resource(stats->lookup_table,
740                                         dentry->ads_entries[i].hash);
741                 i++;
742         }
743
744         return 0;
745 }
746
747 void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table, 
748                                    u64 *dir_count_ret, u64 *file_count_ret, 
749                                    u64 *total_bytes_ret, 
750                                    u64 *hard_link_bytes_ret)
751 {
752         struct image_statistics stats;
753         *dir_count_ret         = 0;
754         *file_count_ret        = 0;
755         *total_bytes_ret       = 0;
756         *hard_link_bytes_ret   = 0;
757         stats.lookup_table     = table;
758         stats.dir_count       = dir_count_ret;
759         stats.file_count      = file_count_ret;
760         stats.total_bytes     = total_bytes_ret;
761         stats.hard_link_bytes = hard_link_bytes_ret;
762         for_lookup_table_entry(table, zero_out_refcnts, NULL);
763         for_dentry_in_tree(root, calculate_dentry_statistics, &stats);
764 }
765
766 static int read_ads_entries(const u8 *p, struct dentry *dentry,
767                             u64 remaining_size)
768 {
769         u16 num_ads = dentry->num_ads;
770         struct ads_entry *ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
771         int ret;
772         if (!ads_entries) {
773                 ERROR("Could not allocate memory for %"PRIu16" "
774                       "alternate data stream entries", num_ads);
775                 return WIMLIB_ERR_NOMEM;
776         }
777         DEBUG2("Reading %"PRIu16" alternate data streams "
778                "(remaining size = %"PRIu64")", num_ads, remaining_size);
779
780         for (u16 i = 0; i < num_ads; i++) {
781                 struct ads_entry *cur_entry = &ads_entries[i];
782                 u64 length;
783                 size_t utf8_len;
784                 const char *p_save = p;
785                 /* Read the base stream entry, excluding the stream name. */
786                 if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
787                         ERROR("Stream entries go past end of metadata resource");
788                         ERROR("(remaining_size = %"PRIu64")", remaining_size);
789                         ret = WIMLIB_ERR_INVALID_DENTRY;
790                         goto out_free_ads_entries;
791                 }
792                 remaining_size -= WIM_ADS_ENTRY_DISK_SIZE;
793
794                 p = get_u64(p, &length); /* ADS entry length */
795
796                 DEBUG2("ADS length = %"PRIu64, length);
797
798                 p += 8; /* Unused */
799                 p = get_bytes(p, WIM_HASH_SIZE, (u8*)cur_entry->hash);
800                 p = get_u16(p, &cur_entry->stream_name_len);
801
802                 DEBUG2("Stream name length = %u", cur_entry->stream_name_len);
803
804                 cur_entry->stream_name = NULL;
805                 cur_entry->stream_name_utf8 = NULL;
806
807                 if (remaining_size < cur_entry->stream_name_len + 2) {
808                         ERROR("Stream entries go past end of metadata resource");
809                         ERROR("(remaining_size = %"PRIu64" bytes, stream_name_len "
810                               "= %"PRIu16" bytes", remaining_size,
811                               cur_entry->stream_name_len);
812                         ret = WIMLIB_ERR_INVALID_DENTRY;
813                         goto out_free_ads_entries;
814                 }
815                 remaining_size -= cur_entry->stream_name_len + 2;
816
817                 cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
818                 if (!cur_entry->stream_name) {
819                         ret = WIMLIB_ERR_NOMEM;
820                         goto out_free_ads_entries;
821                 }
822                 get_bytes(p, cur_entry->stream_name_len,
823                           (u8*)cur_entry->stream_name);
824                 cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
825                                                             cur_entry->stream_name_len,
826                                                             &utf8_len);
827                 cur_entry->stream_name_utf8_len = utf8_len;
828
829                 if (!cur_entry->stream_name_utf8) {
830                         ret = WIMLIB_ERR_NOMEM;
831                         goto out_free_ads_entries;
832                 }
833                 p = p_save + ads_entry_length(cur_entry);
834         }
835         dentry->ads_entries = ads_entries;
836         return 0;
837 out_free_ads_entries:
838         for (u16 i = 0; i < num_ads; i++) {
839                 FREE(ads_entries[i].stream_name);
840                 FREE(ads_entries[i].stream_name_utf8);
841         }
842         FREE(ads_entries);
843         return ret;
844 }
845
846 /* 
847  * Reads a directory entry from the metadata resource.
848  */
849 int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, 
850                 u64 offset, struct dentry *dentry)
851 {
852         const u8 *p;
853         u64 calculated_size;
854         char *file_name;
855         char *file_name_utf8;
856         char *short_name;
857         u16 short_name_len;
858         u16 file_name_len;
859         size_t file_name_utf8_len;
860         int ret;
861
862         dentry_common_init(dentry);
863
864         /*Make sure the dentry really fits into the metadata resource.*/
865         if (offset + 8 > metadata_resource_len) {
866                 ERROR("Directory entry starting at %"PRIu64" ends past the "
867                       "end of the metadata resource (size %"PRIu64")",
868                       offset, metadata_resource_len);
869                 return WIMLIB_ERR_INVALID_DENTRY;
870         }
871
872         /* Before reading the whole entry, we need to read just the length.
873          * This is because an entry of length 8 (that is, just the length field)
874          * terminates the list of sibling directory entries. */
875
876         p = get_u64(&metadata_resource[offset], &dentry->length);
877
878         /* A zero length field (really a length of 8, since that's how big the
879          * directory entry is...) indicates that this is the end of directory
880          * dentry.  We do not read it into memory as an actual dentry, so just
881          * return true in that case. */
882         if (dentry->length == 0)
883                 return 0;
884
885         if (offset + dentry->length >= metadata_resource_len) {
886                 ERROR("Directory entry at offset %"PRIu64" and with size "
887                       "%"PRIu64" ends past the end of the metadata resource "
888                       "(size %"PRIu64")",
889                       offset, dentry->length, metadata_resource_len);
890                 return WIMLIB_ERR_INVALID_DENTRY;
891         }
892
893         /* If it is a recognized length, read the rest of the directory entry.
894          * Note: The root directory entry has no name, and its length does not
895          * include the short name length field.  */
896         if (dentry->length < WIM_DENTRY_DISK_SIZE) {
897                 ERROR("Directory entry has invalid length of %"PRIu64" bytes",
898                       dentry->length);
899                 return WIMLIB_ERR_INVALID_DENTRY;
900         }
901
902         p = get_u32(p, &dentry->attributes);
903         p = get_u32(p, (u32*)&dentry->security_id);
904         p = get_u64(p, &dentry->subdir_offset);
905
906         /* 2 unused fields */
907         p += 2 * sizeof(u64);
908         /*p = get_u64(p, &dentry->unused1);*/
909         /*p = get_u64(p, &dentry->unused2);*/
910
911         p = get_u64(p, &dentry->creation_time);
912         p = get_u64(p, &dentry->last_access_time);
913         p = get_u64(p, &dentry->last_write_time);
914
915         p = get_bytes(p, WIM_HASH_SIZE, dentry->hash);
916         
917         /*
918          * I don't know what's going on here.  It seems like M$ screwed up the
919          * reparse points, then put the fields in the same place and didn't
920          * document it.  The WIM_HDR_FLAG_RP_FIX flag in the WIM header might
921          * have something to do with this, but it's not documented.
922          */
923         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
924                 /* ??? */
925                 p += 4;
926                 p = get_u32(p, &dentry->reparse_tag);
927                 p += 4;
928         } else {
929                 p = get_u32(p, &dentry->reparse_tag);
930                 p = get_u64(p, &dentry->hard_link);
931         }
932
933         /* By the way, the reparse_reserved field does not actually exist (at
934          * least when the file is not a reparse point) */
935
936         
937         p = get_u16(p, &dentry->num_ads);
938
939         p = get_u16(p, &short_name_len);
940         p = get_u16(p, &file_name_len);
941
942         calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
943                           short_name_len;
944
945         if (dentry->length < calculated_size) {
946                 ERROR("Unexpected end of directory entry! (Expected "
947                       "%"PRIu64" bytes, got %"PRIu64" bytes. "
948                       "short_name_len = %hu, file_name_len = %hu)", 
949                       calculated_size, dentry->length,
950                       short_name_len, file_name_len);
951                 return WIMLIB_ERR_INVALID_DENTRY;
952         }
953
954         /* Read the filename. */
955         file_name = MALLOC(file_name_len);
956         if (!file_name) {
957                 ERROR("Failed to allocate %hu bytes for dentry file name",
958                       file_name_len);
959                 return WIMLIB_ERR_NOMEM;
960         }
961         p = get_bytes(p, file_name_len, file_name);
962
963         /* Convert filename to UTF-8. */
964         file_name_utf8 = utf16_to_utf8(file_name, file_name_len, 
965                                        &file_name_utf8_len);
966
967         if (!file_name_utf8) {
968                 ERROR("Failed to allocate memory to convert UTF-16 "
969                       "filename (%hu bytes) to UTF-8", file_name_len);
970                 ret = WIMLIB_ERR_NOMEM;
971                 goto out_free_file_name;
972         }
973
974         /* Undocumented padding between file name and short name.  This probably
975          * is supposed to be a terminating null character. */
976         p += 2;
977
978         /* Read the short filename. */
979         short_name = MALLOC(short_name_len);
980         if (!short_name) {
981                 ERROR("Failed to allocate %hu bytes for short filename",
982                       short_name_len);
983                 ret = WIMLIB_ERR_NOMEM;
984                 goto out_free_file_name_utf8;
985         }
986
987         p = get_bytes(p, short_name_len, short_name);
988
989         /* Some directory entries inexplicibly have a little over 70 bytes of
990          * extra data.  The exact amount of data seems to be 72 bytes, but it is
991          * aligned on the next 8-byte boundary.  Here's an example of the
992          * aligned data:
993          *
994          * 01000000400000006c786bbac58ede11b0bb00261870892ab6adb76fe63a3
995          * e468fca86530d2effa16c786bbac58ede11b0bb00261870892a0000000000
996          * 0000000000000000000000
997          *
998          * Here's one interpretation of how the data is laid out.
999          *
1000          * struct unknown {
1001          *      u32 field1; (always 0x00000001)
1002          *      u32 field2; (always 0x40000000)
1003          *      u16 field3;
1004          *      u32 field4;
1005          *      u32 field5;
1006          *      u32 field6;
1007          *      u8  data[48]; (???)
1008          *      u64 reserved1; (always 0)
1009          *      u64 reserved2; (always 0)
1010          * };*/
1011 #if 0
1012         if (dentry->length - calculated_size >= WIM_ADS_ENTRY_DISK_SIZE) {
1013                 printf("%s: %lu / %lu (", file_name_utf8, 
1014                                 calculated_size, dentry->length);
1015                 print_string(p + WIM_ADS_ENTRY_DISK_SIZE, dentry->length - calculated_size - WIM_ADS_ENTRY_DISK_SIZE);
1016                 puts(")");
1017                 print_byte_field(p, dentry->length - calculated_size);
1018                 putchar('\n');
1019         }
1020 #endif
1021
1022         if (dentry->num_ads != 0) {
1023                 calculated_size = (calculated_size + 7) & ~7;
1024                 if (calculated_size > metadata_resource_len - offset) {
1025                         ERROR("Not enough space in metadata resource for "
1026                               "alternate stream entries");
1027                         ret = WIMLIB_ERR_INVALID_DENTRY;
1028                         goto out_free_short_name;
1029                 }
1030                 ret = read_ads_entries(&metadata_resource[offset + calculated_size],
1031                                        dentry,
1032                                        metadata_resource_len - offset - calculated_size);
1033                 if (ret != 0)
1034                         goto out_free_short_name;
1035         }
1036
1037         dentry->file_name          = file_name;
1038         dentry->file_name_utf8     = file_name_utf8;
1039         dentry->short_name         = short_name;
1040         dentry->file_name_len      = file_name_len;
1041         dentry->file_name_utf8_len = file_name_utf8_len;
1042         dentry->short_name_len     = short_name_len;
1043         return 0;
1044 out_free_short_name:
1045         FREE(short_name);
1046 out_free_file_name_utf8:
1047         FREE(file_name_utf8);
1048 out_free_file_name:
1049         FREE(file_name);
1050         return ret;
1051 }
1052
1053 /* 
1054  * Writes a dentry to an output buffer.
1055  *
1056  * @dentry:  The dentry structure.
1057  * @p:       The memory location to write the data to.
1058  * @return:  Pointer to the byte after the last byte we wrote as part of the
1059  *              dentry.
1060  */
1061 static u8 *write_dentry(const struct dentry *dentry, u8 *p)
1062 {
1063         u8 *orig_p = p;
1064         unsigned padding;
1065
1066         p = put_u64(p, dentry->length);
1067         p = put_u32(p, dentry->attributes);
1068         p = put_u32(p, dentry->security_id);
1069         p = put_u64(p, dentry->subdir_offset);
1070         p = put_u64(p, 0); /* unused1 */
1071         p = put_u64(p, 0); /* unused2 */
1072         p = put_u64(p, dentry->creation_time);
1073         p = put_u64(p, dentry->last_access_time);
1074         p = put_u64(p, dentry->last_write_time);
1075         p = put_bytes(p, WIM_HASH_SIZE, dentry->hash);
1076         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
1077                 p = put_zeroes(p, 4);
1078                 p = put_u32(p, dentry->reparse_tag);
1079                 p = put_zeroes(p, 4);
1080         } else {
1081                 p = put_u32(p, dentry->reparse_tag);
1082                 p = put_u64(p, dentry->hard_link);
1083         }
1084         p = put_u16(p, dentry->num_ads);
1085         p = put_u16(p, dentry->short_name_len);
1086         p = put_u16(p, dentry->file_name_len);
1087         p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
1088         p = put_u16(p, 0); /* filename padding, 2 bytes. */
1089         p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
1090
1091         wimlib_assert(p - orig_p <= dentry->length);
1092         if (p - orig_p < dentry->length)
1093                 p = put_zeroes(p, dentry->length - (p - orig_p));
1094
1095         p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1096
1097         for (u16 i = 0; i < dentry->num_ads; i++) {
1098                 p = put_u64(p, ads_entry_length(&dentry->ads_entries[i]));
1099                 p = put_u64(p, 0); /* Unused */
1100                 p = put_bytes(p, WIM_HASH_SIZE, dentry->ads_entries[i].hash);
1101                 p = put_u16(p, dentry->ads_entries[i].stream_name_len);
1102                 p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
1103                                  (u8*)dentry->ads_entries[i].stream_name);
1104                 p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
1105         }
1106         return p;
1107 }
1108
1109 /* Recursive function that writes a dentry tree rooted at @tree, not including
1110  * @tree itself, which has already been written, except in the case of the root
1111  * dentry, which is written right away, along with an end-of-directory entry. */
1112 u8 *write_dentry_tree(const struct dentry *tree, u8 *p)
1113 {
1114         const struct dentry *child;
1115
1116         if (dentry_is_root(tree)) {
1117                 p = write_dentry(tree, p);
1118
1119                 /* write end of directory entry */
1120                 p = put_u64(p, 0);
1121         } else {
1122                 /* Nothing to do for non-directories */
1123                 if (!dentry_is_directory(tree))
1124                         return p;
1125         }
1126
1127         /* Write child dentries and end-of-directory entry. */
1128         child = tree->children;
1129         if (child) {
1130                 do {
1131                         p = write_dentry(child, p);
1132                         child = child->next;
1133                 } while (child != tree->children);
1134         }
1135
1136         /* write end of directory entry */
1137         p = put_u64(p, 0);
1138
1139         /* Recurse on children. */
1140         if (child) {
1141                 do {
1142                         p = write_dentry_tree(child, p);
1143                         child = child->next;
1144                 } while (child != tree->children);
1145         }
1146         return p;
1147 }
1148
1149 /* Reads the children of a dentry, and all their children, ..., etc. from the
1150  * metadata resource and into the dentry tree.
1151  *
1152  * @metadata_resource:  An array that contains the uncompressed metadata
1153  *                      resource for the WIM file.
1154  * @metadata_resource_len:      The length of @metadata_resource.
1155  * @dentry:     A pointer to a struct dentry that is the root of the directory
1156  *              tree and has already been read from the metadata resource.  It
1157  *              does not need to be the real root because this procedure is
1158  *              called recursively.
1159  *
1160  * @return:     Zero on success, nonzero on failure.
1161  */
1162 int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
1163                      struct dentry *dentry)
1164 {
1165         u64 cur_offset = dentry->subdir_offset;
1166         struct dentry *prev_child = NULL;
1167         struct dentry *first_child = NULL;
1168         struct dentry *child;
1169         struct dentry cur_child;
1170         int ret;
1171
1172         /* If @dentry is a regular file, nothing more needs to be done for this
1173          * branch. */
1174         if (cur_offset == 0)
1175                 return 0;
1176
1177         /* Find and read all the children of @dentry. */
1178         while (1) {
1179
1180                 /* Read next child of @dentry into @cur_child. */
1181                 ret = read_dentry(metadata_resource, metadata_resource_len, 
1182                                   cur_offset, &cur_child);
1183                 if (ret != 0)
1184                         break;
1185
1186                 /* Check for end of directory. */
1187                 if (cur_child.length == 0) {
1188                         ret = 0;
1189                         break;
1190                 }
1191
1192                 /* Not end of directory.  Allocate this child permanently and
1193                  * link it to the parent and previous child. */
1194                 child = MALLOC(sizeof(struct dentry));
1195                 if (!child) {
1196                         ERROR("Failed to allocate %zu bytes for new dentry",
1197                               sizeof(struct dentry));
1198                         ret = WIMLIB_ERR_NOMEM;
1199                         break;
1200                 }
1201                 memcpy(child, &cur_child, sizeof(struct dentry));
1202
1203                 if (prev_child) {
1204                         prev_child->next = child;
1205                         child->prev = prev_child;
1206                 } else {
1207                         first_child = child;
1208                 }
1209
1210                 child->parent = dentry;
1211                 prev_child = child;
1212
1213                 /* If there are children of this child, call this procedure
1214                  * recursively. */
1215                 if (child->subdir_offset != 0) {
1216                         ret = read_dentry_tree(metadata_resource, 
1217                                                metadata_resource_len, child);
1218                         if (ret != 0)
1219                                 break;
1220                 }
1221
1222                 /* Advance to the offset of the next child. */
1223                 cur_offset += dentry_total_length(child);
1224         }
1225
1226         /* Link last child to first one, and set parent's
1227          * children pointer to the first child.  */
1228         if (prev_child) {
1229                 prev_child->next = first_child;
1230                 first_child->prev = prev_child;
1231         }
1232         dentry->children = first_child;
1233         return ret;
1234 }