Consolidate `struct lookup_table_entry'
[wimlib] / src / lookup_table.c
1 /*
2  * lookup_table.c
3  *
4  * Lookup table, implemented as a hash table, that maps dentries to file
5  * resources.
6  */
7
8 /*
9  * Copyright (C) 2012 Eric Biggers
10  *
11  * This file is part of wimlib, a library for working with WIM files.
12  *
13  * wimlib is free software; you can redistribute it and/or modify it under the
14  * terms of the GNU General Public License as published by the Free
15  * Software Foundation; either version 3 of the License, or (at your option)
16  * any later version.
17  *
18  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
19  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
20  * A PARTICULAR PURPOSE. See the GNU General Public License for more
21  * details.
22  *
23  * You should have received a copy of the GNU General Public License
24  * along with wimlib; if not, see http://www.gnu.org/licenses/.
25  */
26
27 #include "wimlib_internal.h"
28 #include "lookup_table.h"
29 #include "io.h"
30 #include <errno.h>
31
32 #ifdef WITH_FUSE
33 #include <unistd.h>
34 #endif
35
36 struct lookup_table *new_lookup_table(size_t capacity)
37 {
38         struct lookup_table *table;
39         struct hlist_head *array;
40
41         table = MALLOC(sizeof(struct lookup_table));
42         if (!table)
43                 goto err;
44         array = CALLOC(capacity, sizeof(array[0]));
45         if (!array) {
46                 FREE(table);
47                 goto err;
48         }
49         table->num_entries = 0;
50         table->capacity = capacity;
51         table->array = array;
52         return table;
53 err:
54         ERROR("Failed to allocate memory for lookup table with capacity %zu",
55               capacity);
56         return NULL;
57 }
58
59 struct lookup_table_entry *new_lookup_table_entry()
60 {
61         struct lookup_table_entry *lte;
62
63         lte = CALLOC(1, sizeof(struct lookup_table_entry));
64         if (lte) {
65                 lte->part_number  = 1;
66                 lte->refcnt       = 1;
67         } else {
68                 ERROR("Out of memory (tried to allocate %zu bytes for "
69                       "lookup table entry)",
70                       sizeof(struct lookup_table_entry));
71         }
72         return lte;
73 }
74
75 struct lookup_table_entry *
76 clone_lookup_table_entry(const struct lookup_table_entry *old)
77 {
78         struct lookup_table_entry *new;
79
80         new = MALLOC(sizeof(*new));
81         if (!new)
82                 return NULL;
83
84         memcpy(new, old, sizeof(*old));
85         new->extracted_file = NULL;
86         switch (new->resource_location) {
87         case RESOURCE_IN_STAGING_FILE:
88         case RESOURCE_IN_FILE_ON_DISK:
89                 wimlib_assert((void*)&old->file_on_disk ==
90                               (void*)&old->staging_file_name);
91                 new->staging_file_name = STRDUP(old->staging_file_name);
92                 if (!new->staging_file_name)
93                         goto out_free;
94                 break;
95         case RESOURCE_IN_ATTACHED_BUFFER:
96                 new->attached_buffer = MALLOC(wim_resource_size(old));
97                 if (!new->attached_buffer)
98                         goto out_free;
99                 memcpy(new->attached_buffer, old->attached_buffer,
100                        wim_resource_size(old));
101                 break;
102 #ifdef WITH_NTFS_3G
103         case RESOURCE_IN_NTFS_VOLUME:
104                 if (old->ntfs_loc) {
105                         struct ntfs_location *loc;
106                         loc = MALLOC(sizeof(*loc));
107                         if (!loc)
108                                 goto out_free;
109                         memcpy(loc, old->ntfs_loc, sizeof(*loc));
110                         loc->path_utf8 = NULL;
111                         loc->stream_name_utf16 = NULL;
112                         new->ntfs_loc = loc;
113                         loc->path_utf8 = STRDUP(old->ntfs_loc->path_utf8);
114                         if (!loc->path_utf8)
115                                 goto out_free;
116                         loc->stream_name_utf16 = MALLOC(loc->stream_name_utf16_num_chars * 2);
117                         if (!loc->stream_name_utf16)
118                                 goto out_free;
119                         memcpy(loc->stream_name_utf16,
120                                old->ntfs_loc->stream_name_utf16,
121                                loc->stream_name_utf16_num_chars * 2);
122                 }
123                 break;
124 #endif
125         default:
126                 break;
127         }
128         return new;
129 out_free:
130         free_lookup_table_entry(new);
131         return NULL;
132 }
133
134 void free_lookup_table_entry(struct lookup_table_entry *lte)
135 {
136         if (lte) {
137                 switch (lte->resource_location) {
138                 case RESOURCE_IN_STAGING_FILE:
139                 case RESOURCE_IN_ATTACHED_BUFFER:
140                 case RESOURCE_IN_FILE_ON_DISK:
141                         wimlib_assert(((void*)&lte->file_on_disk ==
142                                       (void*)&lte->staging_file_name)
143                                       && ((void*)&lte->file_on_disk ==
144                                       (void*)&lte->attached_buffer));
145                         FREE(lte->file_on_disk);
146                         break;
147 #ifdef WITH_NTFS_3G
148                 case RESOURCE_IN_NTFS_VOLUME:
149                         if (lte->ntfs_loc) {
150                                 FREE(lte->ntfs_loc->path_utf8);
151                                 FREE(lte->ntfs_loc->stream_name_utf16);
152                                 FREE(lte->ntfs_loc);
153                         }
154                         break;
155 #endif
156                 default:
157                         break;
158                 }
159                 FREE(lte);
160         }
161 }
162
163 static int do_free_lookup_table_entry(struct lookup_table_entry *entry,
164                                       void *ignore)
165 {
166         free_lookup_table_entry(entry);
167         return 0;
168 }
169
170
171 void free_lookup_table(struct lookup_table *table)
172 {
173         DEBUG2("Freeing lookup table");
174         if (table) {
175                 if (table->array) {
176                         for_lookup_table_entry(table,
177                                                do_free_lookup_table_entry,
178                                                NULL);
179                         FREE(table->array);
180                 }
181                 FREE(table);
182         }
183 }
184
185 /*
186  * Inserts an entry into the lookup table.
187  *
188  * @table:      A pointer to the lookup table.
189  * @entry:      A pointer to the entry to insert.
190  */
191 void lookup_table_insert(struct lookup_table *table,
192                          struct lookup_table_entry *lte)
193 {
194         size_t i = lte->hash_short % table->capacity;
195         hlist_add_head(&lte->hash_list, &table->array[i]);
196
197         /* XXX Make the table grow when too many entries have been inserted. */
198         table->num_entries++;
199 }
200
201 static void finalize_lte(struct lookup_table_entry *lte)
202 {
203         #ifdef WITH_FUSE
204         if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
205                 unlink(lte->staging_file_name);
206                 wimlib_assert(lte->staging_list.next);
207                 wimlib_assert(lte->staging_list.prev);
208                 list_del(&lte->staging_list);
209         }
210         #endif
211         free_lookup_table_entry(lte);
212 }
213
214 /* Decrements the reference count for the lookup table entry @lte.  If its
215  * reference count reaches 0, it is unlinked from the lookup table.  If,
216  * furthermore, the entry has no opened file descriptors associated with it, the
217  * entry is freed.  */
218 void lte_decrement_refcnt(struct lookup_table_entry *lte,
219                           struct lookup_table *table)
220 {
221         wimlib_assert(lte != NULL);
222         wimlib_assert(lte->refcnt != 0);
223         if (--lte->refcnt == 0) {
224                 lookup_table_unlink(table, lte);
225         #ifdef WITH_FUSE
226                 if (lte->num_opened_fds == 0)
227         #endif
228                         finalize_lte(lte);
229         }
230 }
231
232 #ifdef WITH_FUSE
233 void lte_decrement_num_opened_fds(struct lookup_table_entry *lte)
234 {
235         wimlib_assert(lte != NULL);
236         wimlib_assert(lte->num_opened_fds != 0);
237         if (atomic_dec(&lte->num_opened_fds) == 0 && lte->refcnt == 0)
238                 finalize_lte(lte);
239 }
240 #endif
241
242 /*
243  * Calls a function on all the entries in the lookup table.  Stop early and
244  * return nonzero if any call to the function returns nonzero.
245  */
246 int for_lookup_table_entry(struct lookup_table *table,
247                            int (*visitor)(struct lookup_table_entry *, void *),
248                            void *arg)
249 {
250         struct lookup_table_entry *lte;
251         struct hlist_node *pos, *tmp;
252         int ret;
253
254         for (size_t i = 0; i < table->capacity; i++) {
255                 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
256                                           hash_list)
257                 {
258                         ret = visitor(lte, arg);
259                         if (ret != 0)
260                                 return ret;
261                 }
262         }
263         return 0;
264 }
265
266
267 /*
268  * Reads the lookup table from a WIM file.
269  */
270 int read_lookup_table(WIMStruct *w)
271 {
272         u64    num_entries;
273         u8     buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
274         int    ret;
275         struct lookup_table *table;
276         struct lookup_table_entry *cur_entry = NULL, *duplicate_entry;
277
278         DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
279               w->hdr.lookup_table_res_entry.offset,
280               w->hdr.lookup_table_res_entry.original_size);
281
282         if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0)
283         {
284                 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
285                                  "lookup table",
286                                  w->hdr.lookup_table_res_entry.offset);
287                 return WIMLIB_ERR_READ;
288         }
289
290         num_entries = w->hdr.lookup_table_res_entry.original_size /
291                       WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
292         table = new_lookup_table(num_entries * 2 + 1);
293         if (!table)
294                 return WIMLIB_ERR_NOMEM;
295
296         while (num_entries--) {
297                 const u8 *p;
298
299                 if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
300                         if (feof(w->fp)) {
301                                 ERROR("Unexpected EOF in WIM lookup table!");
302                         } else {
303                                 ERROR_WITH_ERRNO("Error reading WIM lookup "
304                                                  "table");
305                         }
306                         ret = WIMLIB_ERR_READ;
307                         goto out;
308                 }
309                 cur_entry = new_lookup_table_entry();
310                 if (!cur_entry) {
311                         ret = WIMLIB_ERR_NOMEM;
312                         goto out;
313                 }
314                 cur_entry->wim = w;
315                 cur_entry->resource_location = RESOURCE_IN_WIM;
316
317                 p = get_resource_entry(buf, &cur_entry->resource_entry);
318                 p = get_u16(p, &cur_entry->part_number);
319                 p = get_u32(p, &cur_entry->refcnt);
320                 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
321
322                 if (cur_entry->part_number != w->hdr.part_number) {
323                         ERROR("A lookup table entry in part %hu of the WIM "
324                               "points to part %hu",
325                               w->hdr.part_number, cur_entry->part_number);
326                         ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
327                         goto out_free_cur_entry;
328                 }
329
330                 if (is_zero_hash(cur_entry->hash)) {
331                         ERROR("The WIM lookup table contains an entry with a "
332                               "SHA1 message digest of all 0's");
333                         ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
334                         goto out_free_cur_entry;
335                 }
336
337                 /* Ordinarily, no two streams should share the same SHA1 message
338                  * digest.  However, this constraint can be broken for metadata
339                  * resources--- two identical images will have the same metadata
340                  * resource, but their lookup table entries are not shared. */
341                 duplicate_entry = __lookup_resource(table, cur_entry->hash);
342                 if (duplicate_entry
343                     && !((duplicate_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
344                           && cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
345                 {
346                         ERROR("The WIM lookup table contains two entries with the "
347                               "same SHA1 message digest!");
348                         ERROR("The first entry is:");
349                         print_lookup_table_entry(duplicate_entry);
350                         ERROR("The second entry is:");
351                         print_lookup_table_entry(cur_entry);
352                         ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
353                         goto out_free_cur_entry;
354                 }
355
356                 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
357                     && (cur_entry->resource_entry.size !=
358                         cur_entry->resource_entry.original_size))
359                 {
360                         ERROR("Found uncompressed resource with original size "
361                               "not the same as compressed size");
362                         ERROR("The lookup table entry for the resource is as follows:");
363                         print_lookup_table_entry(cur_entry);
364                         ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
365                         goto out_free_cur_entry;
366                 }
367                 lookup_table_insert(table, cur_entry);
368
369         }
370         DEBUG("Done reading lookup table.");
371         w->lookup_table = table;
372         return 0;
373 out_free_cur_entry:
374         FREE(cur_entry);
375 out:
376         free_lookup_table(table);
377         return ret;
378 }
379
380
381 /*
382  * Writes a lookup table entry to the output file.
383  */
384 int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out)
385 {
386         FILE *out;
387         u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
388         u8 *p;
389
390         out = __out;
391
392         /* Don't write entries that have not had file resources or metadata
393          * resources written for them. */
394         if (lte->out_refcnt == 0)
395                 return 0;
396
397         if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
398                 DEBUG("Writing metadata entry at %lu (orig size = %zu)",
399                       ftello(out), lte->output_resource_entry.original_size);
400
401         p = put_resource_entry(buf, &lte->output_resource_entry);
402         p = put_u16(p, lte->part_number);
403         p = put_u32(p, lte->out_refcnt);
404         p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
405         if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
406                 ERROR_WITH_ERRNO("Failed to write lookup table entry");
407                 return WIMLIB_ERR_WRITE;
408         }
409         return 0;
410 }
411
412
413 int lte_zero_real_refcnt(struct lookup_table_entry *lte, void *ignore)
414 {
415         lte->real_refcnt = 0;
416         return 0;
417 }
418
419 int lte_zero_out_refcnt(struct lookup_table_entry *lte, void *ignore)
420 {
421         lte->out_refcnt = 0;
422         return 0;
423 }
424
425 int lte_zero_extracted_file(struct lookup_table_entry *lte, void *ignore)
426 {
427         lte->extracted_file = NULL;
428         return 0;
429 }
430
431 int lte_free_extracted_file(struct lookup_table_entry *lte, void *ignore)
432 {
433         if (lte->extracted_file != NULL) {
434                 FREE(lte->extracted_file);
435                 lte->extracted_file = NULL;
436         }
437         return 0;
438 }
439
440 void print_lookup_table_entry(const struct lookup_table_entry *lte)
441 {
442         if (!lte) {
443                 putchar('\n');
444                 return;
445         }
446         printf("Offset            = %"PRIu64" bytes\n",
447                lte->resource_entry.offset);
448         printf("Size              = %"PRIu64" bytes\n",
449                (u64)lte->resource_entry.size);
450         printf("Original size     = %"PRIu64" bytes\n",
451                lte->resource_entry.original_size);
452         printf("Part Number       = %hu\n", lte->part_number);
453         printf("Reference Count   = %u\n", lte->refcnt);
454         printf("Hash              = 0x");
455         print_hash(lte->hash);
456         putchar('\n');
457         printf("Flags             = ");
458         u8 flags = lte->resource_entry.flags;
459         if (flags & WIM_RESHDR_FLAG_COMPRESSED)
460                 fputs("WIM_RESHDR_FLAG_COMPRESSED, ", stdout);
461         if (flags & WIM_RESHDR_FLAG_FREE)
462                 fputs("WIM_RESHDR_FLAG_FREE, ", stdout);
463         if (flags & WIM_RESHDR_FLAG_METADATA)
464                 fputs("WIM_RESHDR_FLAG_METADATA, ", stdout);
465         if (flags & WIM_RESHDR_FLAG_SPANNED)
466                 fputs("WIM_RESHDR_FLAG_SPANNED, ", stdout);
467         putchar('\n');
468         switch (lte->resource_location) {
469         case RESOURCE_IN_WIM:
470                 if (lte->wim->filename) {
471                         printf("WIM file          = `%s'\n",
472                                lte->wim->filename);
473                 }
474                 break;
475         case RESOURCE_IN_FILE_ON_DISK:
476                 printf("File on Disk      = `%s'\n", lte->file_on_disk);
477                 break;
478         case RESOURCE_IN_STAGING_FILE:
479                 printf("Staging File      = `%s'\n", lte->staging_file_name);
480                 break;
481         default:
482                 break;
483         }
484         putchar('\n');
485 }
486
487 static int do_print_lookup_table_entry(struct lookup_table_entry *lte,
488                                        void *ignore)
489 {
490         print_lookup_table_entry(lte);
491         return 0;
492 }
493
494 /*
495  * Prints the lookup table of a WIM file.
496  */
497 WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
498 {
499         for_lookup_table_entry(w->lookup_table,
500                                do_print_lookup_table_entry,
501                                NULL);
502 }
503
504 /*
505  * Looks up an entry in the lookup table.
506  */
507 struct lookup_table_entry *
508 __lookup_resource(const struct lookup_table *table, const u8 hash[])
509 {
510         size_t i;
511         struct lookup_table_entry *lte;
512         struct hlist_node *pos;
513
514         wimlib_assert(table != NULL);
515         wimlib_assert(hash != NULL);
516
517         i = *(size_t*)hash % table->capacity;
518         hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
519                 if (hashes_equal(hash, lte->hash))
520                         return lte;
521         return NULL;
522 }
523
524 #ifdef WITH_FUSE
525 /*
526  * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
527  * given a path name.
528  *
529  * This is only for pre-resolved inodes.
530  */
531 int lookup_resource(WIMStruct *w, const char *path,
532                     int lookup_flags,
533                     struct dentry **dentry_ret,
534                     struct lookup_table_entry **lte_ret,
535                     u16 *stream_idx_ret)
536 {
537         struct dentry *dentry;
538         struct lookup_table_entry *lte;
539         u16 stream_idx;
540         const char *stream_name = NULL;
541         struct inode *inode;
542         char *p = NULL;
543
544         if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
545                 stream_name = path_stream_name(path);
546                 if (stream_name) {
547                         p = (char*)stream_name - 1;
548                         *p = '\0';
549                 }
550         }
551
552         dentry = get_dentry(w, path);
553         if (p)
554                 *p = ':';
555         if (!dentry)
556                 return -ENOENT;
557
558         inode = dentry->d_inode;
559
560         wimlib_assert(inode->resolved);
561
562         if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
563               && inode_is_directory(inode))
564                 return -EISDIR;
565
566         if (stream_name) {
567                 struct ads_entry *ads_entry;
568                 u16 ads_idx;
569                 ads_entry = inode_get_ads_entry(inode, stream_name,
570                                                 &ads_idx);
571                 if (ads_entry) {
572                         stream_idx = ads_idx + 1;
573                         lte = ads_entry->lte;
574                         goto out;
575                 } else {
576                         return -ENOENT;
577                 }
578         } else {
579                 lte = inode->lte;
580                 stream_idx = 0;
581         }
582 out:
583         if (dentry_ret)
584                 *dentry_ret = dentry;
585         if (lte_ret)
586                 *lte_ret = lte;
587         if (stream_idx_ret)
588                 *stream_idx_ret = stream_idx;
589         return 0;
590 }
591 #endif
592
593 static void inode_resolve_ltes(struct inode *inode, struct lookup_table *table)
594 {
595         struct lookup_table_entry *lte;
596
597         wimlib_assert(!inode->resolved);
598
599         /* Resolve the default file stream */
600         lte = __lookup_resource(table, inode->hash);
601         inode->lte = lte;
602         inode->resolved = true;
603
604         /* Resolve the alternate data streams */
605         for (u16 i = 0; i < inode->num_ads; i++) {
606                 struct ads_entry *cur_entry = &inode->ads_entries[i];
607                 lte = __lookup_resource(table, cur_entry->hash);
608                 cur_entry->lte = lte;
609         }
610 }
611
612 static void inode_unresolve_ltes(struct inode *inode)
613 {
614         wimlib_assert(inode->resolved);
615         if (inode->lte)
616                 copy_hash(inode->hash, inode->lte->hash);
617         else
618                 zero_out_hash(inode->hash);
619
620         for (u16 i = 0; i < inode->num_ads; i++) {
621                 if (inode->ads_entries[i].lte)
622                         copy_hash(inode->ads_entries[i].hash,
623                                   inode->ads_entries[i].lte->hash);
624                 else
625                         zero_out_hash(inode->ads_entries[i].hash);
626         }
627         inode->resolved = false;
628 }
629
630 /* Resolve a dentry's lookup table entries
631  *
632  * This replaces the SHA1 hash fields (which are used to lookup an entry in the
633  * lookup table) with pointers directly to the lookup table entries.  A circular
634  * linked list of streams sharing the same lookup table entry is created.
635  *
636  * This function always succeeds; unresolved lookup table entries are given a
637  * NULL pointer.
638  */
639 int dentry_resolve_ltes(struct dentry *dentry, void *table)
640 {
641         if (!dentry->d_inode->resolved)
642                 inode_resolve_ltes(dentry->d_inode, table);
643         return 0;
644 }
645
646 int dentry_unresolve_ltes(struct dentry *dentry, void *ignore)
647 {
648         if (dentry->d_inode->resolved)
649                 inode_unresolve_ltes(dentry->d_inode);
650         return 0;
651 }
652
653 /* Return the lookup table entry for the unnamed data stream of an inode, or
654  * NULL if there is none.
655  *
656  * You'd think this would be easier than it actually is, since the unnamed data
657  * stream should be the one referenced from the inode itself.  Alas, if there
658  * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
659  * data stream in one of the alternate data streams instead of inside the WIM
660  * dentry itself.  So we need to check the alternate data streams too.
661  *
662  * Also, note that a dentry may appear to have more than one unnamed stream, but
663  * if the SHA1 message digest is all 0's then the corresponding stream does not
664  * really "count" (this is the case for the inode's own file stream when the
665  * file stream that should be there is actually in one of the alternate stream
666  * entries.).  This is despite the fact that we may need to extract such a
667  * missing entry as an empty file or empty named data stream.
668  */
669 struct lookup_table_entry *
670 inode_unnamed_lte(const struct inode *inode,
671                    const struct lookup_table *table)
672 {
673         if (inode->resolved)
674                 return inode_unnamed_lte_resolved(inode);
675         else
676                 return inode_unnamed_lte_unresolved(inode, table);
677 }
678