Alignment of ADS entries
[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 /* Real length of a dentry, including the alternate data stream entries, which
43  * are not included in the dentry->length field... */
44 u64 dentry_total_length(const struct dentry *dentry)
45 {
46         u64 length = dentry->length;
47         for (u16 i = 0 ; i < dentry->num_ads; i++)
48                 length += ads_entry_length(&dentry->ads_entries[i]);
49
50         /* Round to 8 byte boundary. */
51         return (length + 7) & ~7;
52 }
53
54 /* Transfers file attributes from a `stat' buffer to a struct dentry. */
55 void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry)
56 {
57         if (S_ISDIR(stbuf->st_mode))
58                 dentry->attributes = FILE_ATTRIBUTE_DIRECTORY;
59         else
60                 dentry->attributes = FILE_ATTRIBUTE_NORMAL;
61 }
62
63 /* Transfers file attributes from a struct dentry to a `stat' buffer. */
64 void dentry_to_stbuf(const struct dentry *dentry, struct stat *stbuf, 
65                      const struct lookup_table *table)
66 {
67         struct lookup_table_entry *lte;
68
69         if (dentry_is_directory(dentry))
70                 stbuf->st_mode = S_IFDIR | 0755;
71         else
72                 stbuf->st_mode = S_IFREG | 0644;
73
74         if (table)
75                 lte = lookup_resource(table, dentry->hash);
76         else
77                 lte = NULL;
78
79         if (lte) {
80                 stbuf->st_nlink = lte->refcnt;
81                 stbuf->st_size = lte->resource_entry.original_size;
82         } else {
83                 stbuf->st_nlink = 1;
84                 stbuf->st_size = 0;
85         }
86         stbuf->st_uid     = getuid();
87         stbuf->st_gid     = getgid();
88         stbuf->st_atime   = ms_timestamp_to_unix(dentry->last_access_time);
89         stbuf->st_mtime   = ms_timestamp_to_unix(dentry->last_write_time);
90         stbuf->st_ctime   = ms_timestamp_to_unix(dentry->creation_time);
91         stbuf->st_blocks  = (stbuf->st_size + 511) / 512;
92 }
93
94 /* Makes all timestamp fields for the dentry be the current time. */
95 void dentry_update_all_timestamps(struct dentry *dentry)
96 {
97         u64 now = get_timestamp();
98         dentry->creation_time    = now;
99         dentry->last_access_time = now;
100         dentry->last_write_time  = now;
101 }
102
103 /* 
104  * Calls a function on all directory entries in a directory tree.  It is called
105  * on a parent before its children.
106  */
107 int for_dentry_in_tree(struct dentry *root, 
108                        int (*visitor)(struct dentry*, void*), void *arg)
109 {
110         int ret;
111         struct dentry *child;
112
113         ret = visitor(root, arg);
114
115         if (ret != 0)
116                 return ret;
117
118         child = root->children;
119
120         if (!child)
121                 return 0;
122
123         do {
124                 ret = for_dentry_in_tree(child, visitor, arg);
125                 if (ret != 0)
126                         return ret;
127                 child = child->next;
128         } while (child != root->children);
129         return 0;
130 }
131
132 /* 
133  * Like for_dentry_in_tree(), but the visitor function is always called on a
134  * dentry's children before on itself.
135  */
136 int for_dentry_in_tree_depth(struct dentry *root, 
137                              int (*visitor)(struct dentry*, void*), void *arg)
138 {
139         int ret;
140         struct dentry *child;
141         struct dentry *next;
142
143         child = root->children;
144         if (child) {
145                 do {
146                         next = child->next;
147                         ret = for_dentry_in_tree_depth(child, visitor, arg);
148                         if (ret != 0)
149                                 return ret;
150                         child = next;
151                 } while (child != root->children);
152         }
153         return visitor(root, arg);
154 }
155
156 /* 
157  * Calculate the full path of @dentry, based on its parent's full path and on
158  * its UTF-8 file name. 
159  */
160 int calculate_dentry_full_path(struct dentry *dentry, void *ignore)
161 {
162         char *full_path;
163         u32 full_path_len;
164         if (dentry_is_root(dentry)) {
165                 full_path = MALLOC(2);
166                 if (!full_path)
167                         goto oom;
168                 full_path[0] = '/';
169                 full_path[1] = '\0';
170                 full_path_len = 1;
171         } else {
172                 char *parent_full_path;
173                 u32 parent_full_path_len;
174                 const struct dentry *parent = dentry->parent;
175
176                 if (dentry_is_root(parent)) {
177                         parent_full_path = "";
178                         parent_full_path_len = 0;
179                 } else {
180                         parent_full_path = parent->full_path_utf8;
181                         parent_full_path_len = parent->full_path_utf8_len;
182                 }
183
184                 full_path_len = parent_full_path_len + 1 +
185                                 dentry->file_name_utf8_len;
186                 full_path = MALLOC(full_path_len + 1);
187                 if (!full_path)
188                         goto oom;
189
190                 memcpy(full_path, parent_full_path, parent_full_path_len);
191                 full_path[parent_full_path_len] = '/';
192                 memcpy(full_path + parent_full_path_len + 1,
193                        dentry->file_name_utf8,
194                        dentry->file_name_utf8_len);
195                 full_path[full_path_len] = '\0';
196         }
197         FREE(dentry->full_path_utf8);
198         dentry->full_path_utf8 = full_path;
199         dentry->full_path_utf8_len = full_path_len;
200         return 0;
201 oom:
202         ERROR("Out of memory while calculating dentry full path");
203         return WIMLIB_ERR_NOMEM;
204 }
205
206 /* 
207  * Recursively calculates the subdir offsets for a directory tree. 
208  *
209  * @dentry:  The root of the directory tree.
210  * @subdir_offset_p:  The current subdirectory offset; i.e., the subdirectory
211  *      offset for @dentry. 
212  */
213 void calculate_subdir_offsets(struct dentry *dentry, u64 *subdir_offset_p)
214 {
215         struct dentry *child;
216
217         child = dentry->children;
218         dentry->subdir_offset = *subdir_offset_p;
219         if (child) {
220
221                 /* Advance the subdir offset by the amount of space the children
222                  * of this dentry take up. */
223                 do {
224                         *subdir_offset_p += dentry_total_length(child);
225                         child = child->next;
226                 } while (child != dentry->children);
227
228                 /* End-of-directory dentry on disk. */
229                 *subdir_offset_p += 8;
230
231                 /* Recursively call calculate_subdir_offsets() on all the
232                  * children. */
233                 do {
234                         calculate_subdir_offsets(child, subdir_offset_p);
235                         child = child->next;
236                 } while (child != dentry->children);
237         } else {
238                 /* On disk, childless directories have a valid subdir_offset
239                  * that points to an 8-byte end-of-directory dentry.  Regular
240                  * files have a subdir_offset of 0. */
241                 if (dentry_is_directory(dentry))
242                         *subdir_offset_p += 8;
243                 else
244                         dentry->subdir_offset = 0;
245         }
246 }
247
248
249 /* Returns the child of @dentry that has the file name @name.  
250  * Returns NULL if no child has the name. */
251 struct dentry *get_dentry_child_with_name(const struct dentry *dentry, 
252                                                         const char *name)
253 {
254         struct dentry *child;
255         size_t name_len;
256         
257         child = dentry->children;
258         if (child) {
259                 name_len = strlen(name);
260                 do {
261                         if (dentry_has_name(child, name, name_len))
262                                 return child;
263                         child = child->next;
264                 } while (child != dentry->children);
265         }
266         return NULL;
267 }
268
269 /* Retrieves the dentry that has the UTF-8 @path relative to the dentry
270  * @cur_dir.  Returns NULL if no dentry having the path is found. */
271 static struct dentry *get_dentry_relative_path(struct dentry *cur_dir, const char *path)
272 {
273         struct dentry *child;
274         size_t base_len;
275         const char *new_path;
276
277         if (*path == '\0')
278                 return cur_dir;
279
280         child = cur_dir->children;
281         if (child) {
282                 new_path = path_next_part(path, &base_len);
283                 do {
284                         if (dentry_has_name(child, path, base_len))
285                                 return get_dentry_relative_path(child, new_path);
286                         child = child->next;
287                 } while (child != cur_dir->children);
288         }
289         return NULL;
290 }
291
292 /* Returns the dentry corresponding to the UTF-8 @path, or NULL if there is no
293  * such dentry. */
294 struct dentry *get_dentry(WIMStruct *w, const char *path)
295 {
296         struct dentry *root = wim_root_dentry(w);
297         while (*path == '/')
298                 path++;
299         return get_dentry_relative_path(root, path);
300 }
301
302 /* Returns the parent directory for the @path. */
303 struct dentry *get_parent_dentry(WIMStruct *w, const char *path)
304 {
305         size_t path_len = strlen(path);
306         char buf[path_len + 1];
307
308         memcpy(buf, path, path_len + 1);
309
310         to_parent_name(buf, path_len);
311
312         return get_dentry(w, buf);
313 }
314
315 /* Prints the full path of a dentry. */
316 int print_dentry_full_path(struct dentry *dentry, void *ignore)
317 {
318         if (dentry->full_path_utf8)
319                 puts(dentry->full_path_utf8);
320         return 0;
321 }
322
323 struct file_attr_flag {
324         u32 flag;
325         const char *name;
326 };
327 struct file_attr_flag file_attr_flags[] = {
328         {FILE_ATTRIBUTE_READONLY,               "READONLY"},
329         {FILE_ATTRIBUTE_HIDDEN,         "HIDDEN"},
330         {FILE_ATTRIBUTE_SYSTEM,         "SYSTEM"},
331         {FILE_ATTRIBUTE_DIRECTORY,              "DIRECTORY"},
332         {FILE_ATTRIBUTE_ARCHIVE,                "ARCHIVE"},
333         {FILE_ATTRIBUTE_DEVICE,         "DEVICE"},
334         {FILE_ATTRIBUTE_NORMAL,         "NORMAL"},
335         {FILE_ATTRIBUTE_TEMPORARY,              "TEMPORARY"},
336         {FILE_ATTRIBUTE_SPARSE_FILE,    "SPARSE_FILE"},
337         {FILE_ATTRIBUTE_REPARSE_POINT,  "REPARSE_POINT"},
338         {FILE_ATTRIBUTE_COMPRESSED,             "COMPRESSED"},
339         {FILE_ATTRIBUTE_OFFLINE,                "OFFLINE"},
340         {FILE_ATTRIBUTE_NOT_CONTENT_INDEXED,"NOT_CONTENT_INDEXED"},
341         {FILE_ATTRIBUTE_ENCRYPTED,              "ENCRYPTED"},
342         {FILE_ATTRIBUTE_VIRTUAL,                "VIRTUAL"},
343 };
344
345 /* Prints a directory entry.  @lookup_table is a pointer to the lookup table, or
346  * NULL if the resource entry for the dentry is not to be printed. */
347 int print_dentry(struct dentry *dentry, void *lookup_table)
348 {
349         struct lookup_table_entry *lte;
350         unsigned i;
351
352         printf("[DENTRY]\n");
353         printf("Length            = %"PRIu64"\n", dentry->length);
354         printf("Attributes        = 0x%x\n", dentry->attributes);
355         for (i = 0; i < ARRAY_LEN(file_attr_flags); i++)
356                 if (file_attr_flags[i].flag & dentry->attributes)
357                         printf("    FILE_ATTRIBUTE_%s is set\n",
358                                 file_attr_flags[i].name);
359         printf("Security ID       = %d\n", dentry->security_id);
360         printf("Subdir offset     = %"PRIu64"\n", dentry->subdir_offset);
361         /*printf("Unused1           = 0x%"PRIu64"\n", dentry->unused1);*/
362         /*printf("Unused2           = %"PRIu64"\n", dentry->unused2);*/
363         printf("Creation Time     = %"PRIu64"\n", dentry->creation_time);
364         printf("Last Access Time  = %"PRIu64"\n", dentry->last_access_time);
365         printf("Last Write Time   = %"PRIu64"\n", dentry->last_write_time);
366         printf("Creation Time     = 0x%"PRIx64"\n", dentry->creation_time);
367         printf("Hash              = "); 
368         print_hash(dentry->hash); 
369         putchar('\n');
370         printf("Reparse Tag       = 0x%"PRIx32"\n", dentry->reparse_tag);
371         printf("Hard Link Group   = 0x%"PRIx64"\n", dentry->hard_link);
372         printf("Number of Alternate Data Streams = %hu\n", dentry->num_ads);
373         printf("Filename          = \"");
374         print_string(dentry->file_name, dentry->file_name_len);
375         puts("\"");
376         printf("Filename Length   = %hu\n", dentry->file_name_len);
377         printf("Filename (UTF-8)  = \"%s\"\n", dentry->file_name_utf8);
378         printf("Filename (UTF-8) Length = %hu\n", dentry->file_name_utf8_len);
379         printf("Short Name        = \"");
380         print_string(dentry->short_name, dentry->short_name_len);
381         puts("\"");
382         printf("Short Name Length = %hu\n", dentry->short_name_len);
383         printf("Full Path (UTF-8) = \"%s\"\n", dentry->full_path_utf8);
384         if (lookup_table) {
385                 lte = lookup_resource(lookup_table, dentry->hash);
386                 if (lte)
387                         print_lookup_table_entry(lte, NULL);
388                 else
389                         putchar('\n');
390         } else {
391                 putchar('\n');
392         }
393         for (u16 i = 0; i < dentry->num_ads; i++) {
394                 printf("[Alternate Stream Entry %u]\n", i);
395                 printf("Name = \"%s\"\n", dentry->ads_entries[i].stream_name_utf8);
396                 lte = lookup_resource(lookup_table, dentry->ads_entries[i].hash);
397                 if (lte)
398                         print_lookup_table_entry(lte, NULL);
399                 else
400                         putchar('\n');
401         }
402         return 0;
403 }
404
405 static inline void dentry_common_init(struct dentry *dentry)
406 {
407         memset(dentry, 0, sizeof(struct dentry));
408         dentry->refcnt = 1;
409         dentry->security_id = -1;
410 }
411
412 /* 
413  * Creates an unlinked directory entry.
414  *
415  * @name:    The base name of the new dentry.
416  * @return:  A pointer to the new dentry, or NULL if out of memory.
417  */
418 struct dentry *new_dentry(const char *name)
419 {
420         struct dentry *dentry;
421         
422         dentry = MALLOC(sizeof(struct dentry));
423         if (!dentry)
424                 return NULL;
425
426         dentry_common_init(dentry);
427         if (change_dentry_name(dentry, name) != 0) {
428                 FREE(dentry);
429                 return NULL;
430         }
431
432         dentry_update_all_timestamps(dentry);
433         dentry->next   = dentry;
434         dentry->prev   = dentry;
435         dentry->parent = dentry;
436         return dentry;
437 }
438
439
440 void free_dentry(struct dentry *dentry)
441 {
442         FREE(dentry->file_name);
443         FREE(dentry->file_name_utf8);
444         FREE(dentry->short_name);
445         FREE(dentry->full_path_utf8);
446         FREE(dentry);
447 }
448
449 /* Arguments for do_free_dentry(). */
450 struct free_dentry_args {
451         struct lookup_table *lookup_table;
452         bool lt_decrement_refcnt;
453 };
454
455 /* 
456  * This function is passed as an argument to for_dentry_in_tree_depth() in order
457  * to free a directory tree.  __args is a pointer to a `struct free_dentry_args'.
458  */
459 static int do_free_dentry(struct dentry *dentry, void *__args)
460 {
461         struct free_dentry_args *args = (struct free_dentry_args*)__args;
462
463         if (args->lt_decrement_refcnt && !dentry_is_directory(dentry)) {
464                 lookup_table_decrement_refcnt(args->lookup_table, 
465                                               dentry->hash);
466         }
467
468         wimlib_assert(dentry->refcnt >= 1);
469         if (--dentry->refcnt == 0)
470                 free_dentry(dentry);
471         return 0;
472 }
473
474 /* 
475  * Unlinks and frees a dentry tree.
476  *
477  * @root:               The root of the tree.
478  * @lookup_table:       The lookup table for dentries.
479  * @decrement_refcnt:   True if the dentries in the tree are to have their 
480  *                      reference counts in the lookup table decremented.
481  */
482 void free_dentry_tree(struct dentry *root, struct lookup_table *lookup_table, 
483                       bool lt_decrement_refcnt)
484 {
485         if (!root || !root->parent)
486                 return;
487
488         struct free_dentry_args args;
489         args.lookup_table        = lookup_table;
490         args.lt_decrement_refcnt = lt_decrement_refcnt;
491         for_dentry_in_tree_depth(root, do_free_dentry, &args);
492 }
493
494 int increment_dentry_refcnt(struct dentry *dentry, void *ignore)
495 {
496         dentry->refcnt++;
497         return 0;
498 }
499
500 /* 
501  * Links a dentry into the directory tree.
502  *
503  * @dentry: The dentry to link.
504  * @parent: The dentry that will be the parent of @dentry.
505  */
506 void link_dentry(struct dentry *dentry, struct dentry *parent)
507 {
508         dentry->parent = parent;
509         if (parent->children) {
510                 /* Not an only child; link to siblings. */
511                 dentry->next = parent->children;
512                 dentry->prev = parent->children->prev;
513                 dentry->next->prev = dentry;
514                 dentry->prev->next = dentry;
515         } else {
516                 /* Only child; link to parent. */
517                 parent->children = dentry;
518                 dentry->next = dentry;
519                 dentry->prev = dentry;
520         }
521 }
522
523 /* Unlink a dentry from the directory tree. */
524 void unlink_dentry(struct dentry *dentry)
525 {
526         if (dentry_is_root(dentry))
527                 return;
528         if (dentry_is_only_child(dentry)) {
529                 dentry->parent->children = NULL;
530         } else {
531                 if (dentry_is_first_sibling(dentry))
532                         dentry->parent->children = dentry->next;
533                 dentry->next->prev = dentry->prev;
534                 dentry->prev->next = dentry->next;
535         }
536 }
537
538
539 /* Recalculates the length of @dentry based on its file name length and short
540  * name length.  */
541 static inline void recalculate_dentry_size(struct dentry *dentry)
542 {
543         dentry->length = WIM_DENTRY_DISK_SIZE + dentry->file_name_len + 
544                          2 + dentry->short_name_len;
545         for (u16 i = 0; i < dentry->num_ads; i++)
546                 dentry->length += ads_entry_length(&dentry->ads_entries[i]);
547         /* Must be multiple of 8. */
548         dentry->length = (dentry->length + 7) & ~7;
549 }
550
551 /* Changes the name of a dentry to @new_name.  Only changes the file_name and
552  * file_name_utf8 fields; does not change the short_name, short_name_utf8, or
553  * full_path_utf8 fields.  Also recalculates its length. */
554 int change_dentry_name(struct dentry *dentry, const char *new_name)
555 {
556         size_t utf8_len;
557         size_t utf16_len;
558
559         FREE(dentry->file_name);
560
561         utf8_len = strlen(new_name);
562
563         dentry->file_name = utf8_to_utf16(new_name, utf8_len, &utf16_len);
564
565         if (!dentry->file_name)
566                 return WIMLIB_ERR_NOMEM;
567
568         FREE(dentry->file_name_utf8);
569         dentry->file_name_utf8 = MALLOC(utf8_len + 1);
570         if (!dentry->file_name_utf8) {
571                 FREE(dentry->file_name);
572                 dentry->file_name = NULL;
573                 return WIMLIB_ERR_NOMEM;
574         }
575
576         dentry->file_name_len = utf16_len;
577         dentry->file_name_utf8_len = utf8_len;
578         memcpy(dentry->file_name_utf8, new_name, utf8_len + 1);
579         recalculate_dentry_size(dentry);
580         return 0;
581 }
582
583 /* Parameters for calculate_dentry_statistics(). */
584 struct image_statistics {
585         struct lookup_table *lookup_table;
586         u64 *dir_count;
587         u64 *file_count;
588         u64 *total_bytes;
589         u64 *hard_link_bytes;
590 };
591
592 static int calculate_dentry_statistics(struct dentry *dentry, void *arg)
593 {
594         struct image_statistics *stats;
595         struct lookup_table_entry *lte; 
596         
597         stats = arg;
598         lte = lookup_resource(stats->lookup_table, dentry->hash);
599
600         if (dentry_is_directory(dentry) && !dentry_is_root(dentry))
601                 ++*stats->dir_count;
602         else
603                 ++*stats->file_count;
604
605         if (lte) {
606                 u64 size = lte->resource_entry.original_size;
607                 *stats->total_bytes += size;
608                 if (++lte->out_refcnt == 1)
609                         *stats->hard_link_bytes += size;
610         }
611         return 0;
612 }
613
614 void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *table, 
615                                    u64 *dir_count_ret, u64 *file_count_ret, 
616                                    u64 *total_bytes_ret, 
617                                    u64 *hard_link_bytes_ret)
618 {
619         struct image_statistics stats;
620         *dir_count_ret         = 0;
621         *file_count_ret        = 0;
622         *total_bytes_ret       = 0;
623         *hard_link_bytes_ret   = 0;
624         stats.lookup_table     = table;
625         stats.dir_count       = dir_count_ret;
626         stats.file_count      = file_count_ret;
627         stats.total_bytes     = total_bytes_ret;
628         stats.hard_link_bytes = hard_link_bytes_ret;
629         for_lookup_table_entry(table, zero_out_refcnts, NULL);
630         for_dentry_in_tree(root, calculate_dentry_statistics, &stats);
631 }
632
633 static int read_ads_entries(const u8 *p, struct dentry *dentry,
634                             u64 remaining_size)
635 {
636         u16 num_ads = dentry->num_ads;
637         struct ads_entry *ads_entries = CALLOC(num_ads, sizeof(struct ads_entry));
638         int ret;
639         if (!ads_entries) {
640                 ERROR("Could not allocate memory for %"PRIu16" "
641                       "alternate data stream entries", num_ads);
642                 return WIMLIB_ERR_NOMEM;
643         }
644         DEBUG2("Reading %"PRIu16" alternate data streams "
645                "(remaining size = %"PRIu64")", num_ads, remaining_size);
646
647         for (u16 i = 0; i < num_ads; i++) {
648                 struct ads_entry *cur_entry = &ads_entries[i];
649                 u64 length;
650                 size_t utf8_len;
651                 const char *p_save = p;
652                 /* Read the base stream entry, excluding the stream name. */
653                 if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
654                         ERROR("Stream entries go past end of metadata resource");
655                         ERROR("(remaining_size = %"PRIu64")", remaining_size);
656                         ret = WIMLIB_ERR_INVALID_DENTRY;
657                         goto out_free_ads_entries;
658                 }
659                 remaining_size -= WIM_ADS_ENTRY_DISK_SIZE;
660                 /*print_string(p + 40, 10);*/
661                 /*print_byte_field(p, 50);*/
662
663                 p = get_u64(p, &length); /* ADS entry length */
664
665                 DEBUG2("ADS length = %"PRIu64, length);
666
667                 p += 8; /* Unused */
668                 p = get_bytes(p, WIM_HASH_SIZE, (u8*)cur_entry->hash);
669                 p = get_u16(p, &cur_entry->stream_name_len);
670
671                 DEBUG2("Stream name length = %u", cur_entry->stream_name_len);
672
673                 cur_entry->stream_name = NULL;
674                 cur_entry->stream_name_utf8 = NULL;
675
676                 if (remaining_size < cur_entry->stream_name_len + 2) {
677                         ERROR("Stream entries go past end of metadata resource");
678                         ERROR("(remaining_size = %"PRIu64" bytes, stream_name_len "
679                               "= %"PRIu16" bytes", remaining_size,
680                               cur_entry->stream_name_len);
681                         ret = WIMLIB_ERR_INVALID_DENTRY;
682                         goto out_free_ads_entries;
683                 }
684                 remaining_size -= cur_entry->stream_name_len + 2;
685
686                 cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
687                 if (!cur_entry->stream_name) {
688                         ret = WIMLIB_ERR_NOMEM;
689                         goto out_free_ads_entries;
690                 }
691                 get_bytes(p, cur_entry->stream_name_len,
692                           (u8*)cur_entry->stream_name);
693                 cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
694                                                             cur_entry->stream_name_len,
695                                                             &utf8_len);
696                 cur_entry->stream_name_len_utf8 = utf8_len;
697
698                 if (!cur_entry->stream_name_utf8) {
699                         ret = WIMLIB_ERR_NOMEM;
700                         goto out_free_ads_entries;
701                 }
702                 p = p_save + ads_entry_length(cur_entry);
703         }
704         dentry->ads_entries = ads_entries;
705         return 0;
706 out_free_ads_entries:
707         for (u16 i = 0; i < num_ads; i++) {
708                 FREE(ads_entries[i].stream_name);
709                 FREE(ads_entries[i].stream_name_utf8);
710         }
711         FREE(ads_entries);
712         return ret;
713 }
714
715 /* 
716  * Reads a directory entry from the metadata resource.
717  */
718 int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len, 
719                 u64 offset, struct dentry *dentry)
720 {
721         const u8 *p;
722         u64 calculated_size;
723         char *file_name;
724         char *file_name_utf8;
725         char *short_name;
726         u16 short_name_len;
727         u16 file_name_len;
728         size_t file_name_utf8_len;
729         int ret;
730
731         dentry_common_init(dentry);
732
733         /*Make sure the dentry really fits into the metadata resource.*/
734         if (offset + 8 > metadata_resource_len) {
735                 ERROR("Directory entry starting at %"PRIu64" ends past the "
736                       "end of the metadata resource (size %"PRIu64")",
737                       offset, metadata_resource_len);
738                 return WIMLIB_ERR_INVALID_DENTRY;
739         }
740
741         /* Before reading the whole entry, we need to read just the length.
742          * This is because an entry of length 8 (that is, just the length field)
743          * terminates the list of sibling directory entries. */
744
745         p = get_u64(&metadata_resource[offset], &dentry->length);
746
747         /* A zero length field (really a length of 8, since that's how big the
748          * directory entry is...) indicates that this is the end of directory
749          * dentry.  We do not read it into memory as an actual dentry, so just
750          * return true in that case. */
751         if (dentry->length == 0)
752                 return 0;
753
754         if (offset + dentry->length >= metadata_resource_len) {
755                 ERROR("Directory entry at offset %"PRIu64" and with size "
756                       "%"PRIu64" ends past the end of the metadata resource "
757                       "(size %"PRIu64")",
758                       offset, dentry->length, metadata_resource_len);
759                 return WIMLIB_ERR_INVALID_DENTRY;
760         }
761
762         /* If it is a recognized length, read the rest of the directory entry.
763          * Note: The root directory entry has no name, and its length does not
764          * include the short name length field.  */
765         if (dentry->length < WIM_DENTRY_DISK_SIZE) {
766                 ERROR("Directory entry has invalid length of %"PRIu64" bytes",
767                       dentry->length);
768                 return WIMLIB_ERR_INVALID_DENTRY;
769         }
770
771         p = get_u32(p, &dentry->attributes);
772         p = get_u32(p, (u32*)&dentry->security_id);
773         p = get_u64(p, &dentry->subdir_offset);
774
775         /* 2 unused fields */
776         p += 2 * sizeof(u64);
777         /*p = get_u64(p, &dentry->unused1);*/
778         /*p = get_u64(p, &dentry->unused2);*/
779
780         p = get_u64(p, &dentry->creation_time);
781         p = get_u64(p, &dentry->last_access_time);
782         p = get_u64(p, &dentry->last_write_time);
783
784         p = get_bytes(p, WIM_HASH_SIZE, dentry->hash);
785         
786         /*
787          * I don't know what's going on here.  It seems like M$ screwed up the
788          * reparse points, then put the fields in the same place and didn't
789          * document it.  The WIM_HDR_FLAG_RP_FIX flag in the WIM header might
790          * have something to do with this, but it's not documented.
791          */
792         if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
793                 /* ??? */
794                 u32 u1, u2;
795                 p = get_u32(p, &u1);
796                 /*p += 4;*/
797                 p = get_u32(p, &dentry->reparse_tag);
798                 p = get_u32(p, &u2);
799                 /*p += 4;*/
800                 dentry->hard_link = (u64)(u1) | ((u64)(u2) << 32);
801         } else {
802                 p = get_u32(p, &dentry->reparse_tag);
803                 p = get_u64(p, &dentry->hard_link);
804         }
805
806         /* By the way, the reparse_reserved field does not actually exist (at
807          * least when the file is not a reparse point) */
808
809         
810         p = get_u16(p, &dentry->num_ads);
811
812         p = get_u16(p, &short_name_len);
813         p = get_u16(p, &file_name_len);
814
815         calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
816                           short_name_len;
817
818         if (dentry->length < calculated_size) {
819                 ERROR("Unexpected end of directory entry! (Expected "
820                       "%"PRIu64" bytes, got %"PRIu64" bytes. "
821                       "short_name_len = %hu, file_name_len = %hu)", 
822                       calculated_size, dentry->length,
823                       short_name_len, file_name_len);
824                 return WIMLIB_ERR_INVALID_DENTRY;
825         }
826
827         /* Read the filename. */
828         file_name = MALLOC(file_name_len);
829         if (!file_name) {
830                 ERROR("Failed to allocate %hu bytes for dentry file name",
831                       file_name_len);
832                 return WIMLIB_ERR_NOMEM;
833         }
834         p = get_bytes(p, file_name_len, file_name);
835
836         /* Convert filename to UTF-8. */
837         file_name_utf8 = utf16_to_utf8(file_name, file_name_len, 
838                                        &file_name_utf8_len);
839
840         if (!file_name_utf8) {
841                 ERROR("Failed to allocate memory to convert UTF-16 "
842                       "filename (%hu bytes) to UTF-8", file_name_len);
843                 ret = WIMLIB_ERR_NOMEM;
844                 goto out_free_file_name;
845         }
846
847         /* Undocumented padding between file name and short name.  This probably
848          * is supposed to be a terminating null character. */
849         p += 2;
850
851         /* Read the short filename. */
852         short_name = MALLOC(short_name_len);
853         if (!short_name) {
854                 ERROR("Failed to allocate %hu bytes for short filename",
855                       short_name_len);
856                 ret = WIMLIB_ERR_NOMEM;
857                 goto out_free_file_name_utf8;
858         }
859
860         p = get_bytes(p, short_name_len, short_name);
861
862         /* Some directory entries inexplicibly have a little over 70 bytes of
863          * extra data.  The exact amount of data seems to be 72 bytes, but it is
864          * aligned on the next 8-byte boundary.  Here's an example of the
865          * aligned data:
866          *
867          * 01000000400000006c786bbac58ede11b0bb00261870892ab6adb76fe63a3
868          * e468fca86530d2effa16c786bbac58ede11b0bb00261870892a0000000000
869          * 0000000000000000000000
870          *
871          * Here's one interpretation of how the data is laid out.
872          *
873          * struct unknown {
874          *      u32 field1; (always 0x00000001)
875          *      u32 field2; (always 0x40000000)
876          *      u16 field3;
877          *      u32 field4;
878          *      u32 field5;
879          *      u32 field6;
880          *      u8  data[48]; (???)
881          *      u64 reserved1; (always 0)
882          *      u64 reserved2; (always 0)
883          * };*/
884 #if 0
885         if (dentry->length - calculated_size >= WIM_ADS_ENTRY_DISK_SIZE) {
886                 printf("%s: %lu / %lu (", file_name_utf8, 
887                                 calculated_size, dentry->length);
888                 print_string(p + WIM_ADS_ENTRY_DISK_SIZE, dentry->length - calculated_size - WIM_ADS_ENTRY_DISK_SIZE);
889                 puts(")");
890                 print_byte_field(p, dentry->length - calculated_size);
891                 putchar('\n');
892         }
893 #endif
894
895         if (dentry->num_ads != 0) {
896                 ret = read_ads_entries(p, dentry,
897                                        metadata_resource_len - offset - calculated_size);
898                 if (ret != 0)
899                         goto out_free_short_name;
900         }
901
902         dentry->file_name          = file_name;
903         dentry->file_name_utf8     = file_name_utf8;
904         dentry->short_name         = short_name;
905         dentry->file_name_len      = file_name_len;
906         dentry->file_name_utf8_len = file_name_utf8_len;
907         dentry->short_name_len     = short_name_len;
908         return 0;
909 out_free_short_name:
910         FREE(short_name);
911 out_free_file_name_utf8:
912         FREE(file_name_utf8);
913 out_free_file_name:
914         FREE(file_name);
915         return ret;
916 }
917
918 /* 
919  * Writes a dentry to an output buffer.
920  *
921  * @dentry:  The dentry structure.
922  * @p:       The memory location to write the data to.
923  * @return:  Pointer to the byte after the last byte we wrote as part of the
924  *              dentry.
925  */
926 static u8 *write_dentry(const struct dentry *dentry, u8 *p)
927 {
928         u8 *orig_p = p;
929         memset(p, 0, dentry->length);
930         p = put_u64(p, dentry->length);
931         p = put_u32(p, dentry->attributes);
932         p = put_u32(p, dentry->security_id);
933         p = put_u64(p, dentry->subdir_offset);
934         p = put_u64(p, 0); /* unused1 */
935         p = put_u64(p, 0); /* unused2 */
936         p = put_u64(p, dentry->creation_time);
937         p = put_u64(p, dentry->last_access_time);
938         p = put_u64(p, dentry->last_write_time);
939         memcpy(p, dentry->hash, WIM_HASH_SIZE);
940         p += WIM_HASH_SIZE;
941         p = put_u32(p, dentry->reparse_tag);
942         p = put_u64(p, dentry->hard_link);
943         p = put_u16(p, dentry->num_ads); /*streams */
944         p = put_u16(p, dentry->short_name_len);
945         p = put_u16(p, dentry->file_name_len);
946         p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
947         p = put_u16(p, 0); /* filename padding, 2 bytes. */
948         p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
949         for (u16 i = 0; i < dentry->num_ads; i++) {
950                 p = put_u64(p, ads_entry_length(&dentry->ads_entries[i]));
951                 p = put_u64(p, 0); /* Unused */
952                 p = put_bytes(p, WIM_HASH_SIZE, dentry->ads_entries[i].hash);
953                 p = put_u16(p, dentry->ads_entries[i].stream_name_len);
954                 p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
955                                  (u8*)dentry->ads_entries[i].stream_name);
956         }
957         return orig_p + dentry->length;
958 }
959
960 /* Recursive function that writes a dentry tree rooted at @tree, not including
961  * @tree itself, which has already been written, except in the case of the root
962  * dentry, which is written right away, along with an end-of-directory entry. */
963 u8 *write_dentry_tree(const struct dentry *tree, u8 *p)
964 {
965         const struct dentry *child;
966
967         if (dentry_is_root(tree)) {
968                 p = write_dentry(tree, p);
969
970                 /* write end of directory entry */
971                 p = put_u64(p, 0);
972         } else {
973                 /* Nothing to do for a regular file. */
974                 if (dentry_is_regular_file(tree))
975                         return p;
976         }
977
978         /* Write child dentries and end-of-directory entry. */
979         child = tree->children;
980         if (child) {
981                 do {
982                         p = write_dentry(child, p);
983                         child = child->next;
984                 } while (child != tree->children);
985         }
986
987         /* write end of directory entry */
988         p = put_u64(p, 0);
989
990         /* Recurse on children. */
991         if (child) {
992                 do {
993                         p = write_dentry_tree(child, p);
994                         child = child->next;
995                 } while (child != tree->children);
996         }
997         return p;
998 }
999
1000 /* Reads the children of a dentry, and all their children, ..., etc. from the
1001  * metadata resource and into the dentry tree.
1002  *
1003  * @metadata_resource:  An array that contains the uncompressed metadata
1004  *                      resource for the WIM file.
1005  * @metadata_resource_len:      The length of @metadata_resource.
1006  * @dentry:     A pointer to a struct dentry that is the root of the directory
1007  *              tree and has already been read from the metadata resource.  It
1008  *              does not need to be the real root because this procedure is
1009  *              called recursively.
1010  *
1011  * @return:     Zero on success, nonzero on failure.
1012  */
1013 int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
1014                      struct dentry *dentry)
1015 {
1016         u64 cur_offset = dentry->subdir_offset;
1017         struct dentry *prev_child = NULL;
1018         struct dentry *first_child = NULL;
1019         struct dentry *child;
1020         struct dentry cur_child;
1021         int ret;
1022
1023         /* If @dentry is a regular file, nothing more needs to be done for this
1024          * branch. */
1025         if (cur_offset == 0)
1026                 return 0;
1027
1028         /* Find and read all the children of @dentry. */
1029         while (1) {
1030
1031                 /* Read next child of @dentry into @cur_child. */
1032                 ret = read_dentry(metadata_resource, metadata_resource_len, 
1033                                   cur_offset, &cur_child);
1034                 if (ret != 0)
1035                         break;
1036
1037                 /* Check for end of directory. */
1038                 if (cur_child.length == 0) {
1039                         ret = 0;
1040                         break;
1041                 }
1042
1043                 /* Not end of directory.  Allocate this child permanently and
1044                  * link it to the parent and previous child. */
1045                 child = MALLOC(sizeof(struct dentry));
1046                 if (!child) {
1047                         ERROR("Failed to allocate %zu bytes for new dentry",
1048                               sizeof(struct dentry));
1049                         ret = WIMLIB_ERR_NOMEM;
1050                         break;
1051                 }
1052                 memcpy(child, &cur_child, sizeof(struct dentry));
1053
1054                 if (prev_child) {
1055                         prev_child->next = child;
1056                         child->prev = prev_child;
1057                 } else {
1058                         first_child = child;
1059                 }
1060
1061                 child->parent = dentry;
1062                 prev_child = child;
1063
1064                 /* If there are children of this child, call this procedure
1065                  * recursively. */
1066                 if (child->subdir_offset != 0) {
1067                         ret = read_dentry_tree(metadata_resource, 
1068                                                metadata_resource_len, child);
1069                         if (ret != 0)
1070                                 break;
1071                 }
1072
1073                 /* Advance to the offset of the next child. */
1074                 cur_offset += dentry_total_length(child);
1075         }
1076
1077         /* Link last child to first one, and set parent's
1078          * children pointer to the first child.  */
1079         if (prev_child) {
1080                 prev_child->next = first_child;
1081                 first_child->prev = prev_child;
1082         }
1083         dentry->children = first_child;
1084         return ret;
1085 }