c1bd00d33e3e25caf0bf31dd81f3eeebe51702b2
[wimlib] / src / lookup_table.c
1 /*
2  * lookup_table.c
3  *
4  * Lookup table, implemented as a hash table, that maps SHA1 message digests to
5  * data streams.
6  */
7
8 /*
9  * Copyright (C) 2012, 2013 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 #ifdef HAVE_CONFIG_H
28 #  include "config.h"
29 #endif
30
31 #include "wimlib/endianness.h"
32 #include "wimlib/error.h"
33 #include "wimlib/file_io.h"
34 #include "wimlib/lookup_table.h"
35 #include "wimlib/metadata.h"
36 #include "wimlib/paths.h"
37 #include "wimlib/resource.h"
38 #include "wimlib/util.h"
39
40 #include <errno.h>
41 #include <stdlib.h>
42 #ifdef WITH_FUSE
43 #  include <unistd.h> /* for unlink() */
44 #endif
45
46 struct wim_lookup_table *
47 new_lookup_table(size_t capacity)
48 {
49         struct wim_lookup_table *table;
50         struct hlist_head *array;
51
52         table = CALLOC(1, sizeof(struct wim_lookup_table));
53         if (table) {
54                 array = CALLOC(capacity, sizeof(array[0]));
55                 if (array) {
56                         table->num_entries = 0;
57                         table->capacity = capacity;
58                         table->array = array;
59                 } else {
60                         FREE(table);
61                         table = NULL;
62                         ERROR("Failed to allocate memory for lookup table "
63                               "with capacity %zu", capacity);
64                 }
65         }
66         return table;
67 }
68
69 struct wim_lookup_table_entry *
70 new_lookup_table_entry(void)
71 {
72         struct wim_lookup_table_entry *lte;
73
74         lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
75         if (lte) {
76                 lte->part_number  = 1;
77                 lte->refcnt       = 1;
78         } else {
79                 ERROR("Out of memory (tried to allocate %zu bytes for "
80                       "lookup table entry)",
81                       sizeof(struct wim_lookup_table_entry));
82         }
83         return lte;
84 }
85
86 struct wim_lookup_table_entry *
87 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
88 {
89         struct wim_lookup_table_entry *new;
90
91         new = memdup(old, sizeof(struct wim_lookup_table_entry));
92         if (!new)
93                 return NULL;
94
95         new->extracted_file = NULL;
96         switch (new->resource_location) {
97 #ifdef __WIN32__
98         case RESOURCE_WIN32:
99         case RESOURCE_WIN32_ENCRYPTED:
100 #else
101         case RESOURCE_IN_FILE_ON_DISK:
102 #endif
103 #ifdef WITH_FUSE
104         case RESOURCE_IN_STAGING_FILE:
105                 BUILD_BUG_ON((void*)&old->file_on_disk !=
106                              (void*)&old->staging_file_name);
107 #endif
108                 new->file_on_disk = TSTRDUP(old->file_on_disk);
109                 if (!new->file_on_disk)
110                         goto out_free;
111                 break;
112         case RESOURCE_IN_ATTACHED_BUFFER:
113                 new->attached_buffer = memdup(old->attached_buffer,
114                                               wim_resource_size(old));
115                 if (!new->attached_buffer)
116                         goto out_free;
117                 break;
118 #ifdef WITH_NTFS_3G
119         case RESOURCE_IN_NTFS_VOLUME:
120                 if (old->ntfs_loc) {
121                         struct ntfs_location *loc;
122                         loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
123                         if (!loc)
124                                 goto out_free;
125                         loc->path = NULL;
126                         loc->stream_name = NULL;
127                         new->ntfs_loc = loc;
128                         loc->path = STRDUP(old->ntfs_loc->path);
129                         if (!loc->path)
130                                 goto out_free;
131                         if (loc->stream_name_nchars) {
132                                 loc->stream_name = memdup(old->ntfs_loc->stream_name,
133                                                           loc->stream_name_nchars * 2);
134                                 if (!loc->stream_name)
135                                         goto out_free;
136                         }
137                 }
138                 break;
139 #endif
140         default:
141                 break;
142         }
143         return new;
144 out_free:
145         free_lookup_table_entry(new);
146         return NULL;
147 }
148
149 void
150 free_lookup_table_entry(struct wim_lookup_table_entry *lte)
151 {
152         if (lte) {
153                 switch (lte->resource_location) {
154         #ifdef __WIN32__
155                 case RESOURCE_WIN32:
156                 case RESOURCE_WIN32_ENCRYPTED:
157         #else
158                 case RESOURCE_IN_FILE_ON_DISK:
159         #endif
160         #ifdef WITH_FUSE
161                 case RESOURCE_IN_STAGING_FILE:
162                         BUILD_BUG_ON((void*)&lte->file_on_disk !=
163                                      (void*)&lte->staging_file_name);
164         #endif
165                 case RESOURCE_IN_ATTACHED_BUFFER:
166                         BUILD_BUG_ON((void*)&lte->file_on_disk !=
167                                      (void*)&lte->attached_buffer);
168                         FREE(lte->file_on_disk);
169                         break;
170 #ifdef WITH_NTFS_3G
171                 case RESOURCE_IN_NTFS_VOLUME:
172                         if (lte->ntfs_loc) {
173                                 FREE(lte->ntfs_loc->path);
174                                 FREE(lte->ntfs_loc->stream_name);
175                                 FREE(lte->ntfs_loc);
176                         }
177                         break;
178 #endif
179                 default:
180                         break;
181                 }
182                 FREE(lte);
183         }
184 }
185
186 static int
187 do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
188 {
189         free_lookup_table_entry(entry);
190         return 0;
191 }
192
193
194 void
195 free_lookup_table(struct wim_lookup_table *table)
196 {
197         DEBUG2("Freeing lookup table");
198         if (table) {
199                 if (table->array) {
200                         for_lookup_table_entry(table,
201                                                do_free_lookup_table_entry,
202                                                NULL);
203                         FREE(table->array);
204                 }
205                 FREE(table);
206         }
207 }
208
209 /*
210  * Inserts an entry into the lookup table.
211  *
212  * @table:      A pointer to the lookup table.
213  * @lte:        A pointer to the entry to insert.
214  */
215 void
216 lookup_table_insert(struct wim_lookup_table *table,
217                     struct wim_lookup_table_entry *lte)
218 {
219         size_t i = lte->hash_short % table->capacity;
220         hlist_add_head(&lte->hash_list, &table->array[i]);
221
222         /* XXX Make the table grow when too many entries have been inserted. */
223         table->num_entries++;
224 }
225
226 static void
227 finalize_lte(struct wim_lookup_table_entry *lte)
228 {
229         #ifdef WITH_FUSE
230         if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
231                 unlink(lte->staging_file_name);
232                 list_del(&lte->unhashed_list);
233         }
234         #endif
235         free_lookup_table_entry(lte);
236 }
237
238 /* Decrements the reference count for the lookup table entry @lte.  If its
239  * reference count reaches 0, it is unlinked from the lookup table.  If,
240  * furthermore, the entry has no opened file descriptors associated with it, the
241  * entry is freed.  */
242 void
243 lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
244                      struct wim_lookup_table *table)
245 {
246         wimlib_assert(lte != NULL);
247         wimlib_assert(lte->refcnt != 0);
248         if (--lte->refcnt == 0) {
249                 if (lte->unhashed)
250                         list_del(&lte->unhashed_list);
251                 else
252                         lookup_table_unlink(table, lte);
253         #ifdef WITH_FUSE
254                 if (lte->num_opened_fds == 0)
255         #endif
256                         finalize_lte(lte);
257         }
258 }
259
260 #ifdef WITH_FUSE
261 void
262 lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
263 {
264         if (lte->num_opened_fds != 0)
265                 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
266                         finalize_lte(lte);
267 }
268 #endif
269
270 /* Calls a function on all the entries in the WIM lookup table.  Stop early and
271  * return nonzero if any call to the function returns nonzero. */
272 int
273 for_lookup_table_entry(struct wim_lookup_table *table,
274                        int (*visitor)(struct wim_lookup_table_entry *, void *),
275                        void *arg)
276 {
277         struct wim_lookup_table_entry *lte;
278         struct hlist_node *pos, *tmp;
279         int ret;
280
281         for (size_t i = 0; i < table->capacity; i++) {
282                 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
283                                           hash_list)
284                 {
285                         wimlib_assert2(!(lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA));
286                         ret = visitor(lte, arg);
287                         if (ret)
288                                 return ret;
289                 }
290         }
291         return 0;
292 }
293
294 int
295 cmp_streams_by_wim_position(const void *p1, const void *p2)
296 {
297         const struct wim_lookup_table_entry *lte1, *lte2;
298         lte1 = *(const struct wim_lookup_table_entry**)p1;
299         lte2 = *(const struct wim_lookup_table_entry**)p2;
300         if (lte1->resource_entry.offset < lte2->resource_entry.offset)
301                 return -1;
302         else if (lte1->resource_entry.offset > lte2->resource_entry.offset)
303                 return 1;
304         else
305                 return 0;
306 }
307
308
309 static int
310 add_lte_to_array(struct wim_lookup_table_entry *lte,
311                  void *_pp)
312 {
313         struct wim_lookup_table_entry ***pp = _pp;
314         *(*pp)++ = lte;
315         return 0;
316 }
317
318 /* Iterate through the lookup table entries, but first sort them by stream
319  * offset in the WIM.  Caution: this is intended to be used when the stream
320  * offset field has actually been set. */
321 int
322 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
323                                   int (*visitor)(struct wim_lookup_table_entry *,
324                                                  void *),
325                                   void *arg)
326 {
327         struct wim_lookup_table_entry **lte_array, **p;
328         size_t num_streams = table->num_entries;
329         int ret;
330
331         lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
332         if (!lte_array)
333                 return WIMLIB_ERR_NOMEM;
334         p = lte_array;
335         for_lookup_table_entry(table, add_lte_to_array, &p);
336
337         wimlib_assert(p == lte_array + num_streams);
338
339         qsort(lte_array, num_streams, sizeof(lte_array[0]),
340               cmp_streams_by_wim_position);
341         ret = 0;
342         for (size_t i = 0; i < num_streams; i++) {
343                 ret = visitor(lte_array[i], arg);
344                 if (ret)
345                         break;
346         }
347         FREE(lte_array);
348         return ret;
349 }
350
351 /* On-disk format of a WIM lookup table entry (stream entry). */
352 struct wim_lookup_table_entry_disk {
353         /* Location, offset, compression status, and metadata status of the
354          * stream. */
355         struct resource_entry_disk resource_entry;
356
357         /* Which part of the split WIM this stream is in; indexed from 1. */
358         le16 part_number;
359
360         /* Reference count of this stream over all WIM images. */
361         le32 refcnt;
362
363         /* SHA1 message digest of the uncompressed data of this stream, or
364          * optionally all zeroes if this stream is of zero length. */
365         u8 hash[SHA1_HASH_SIZE];
366 } _packed_attribute;
367
368 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
369
370 /*
371  * Reads the lookup table from a WIM file.
372  *
373  * Saves lookup table entries for non-metadata streams in a hash table, and
374  * saves the metadata entry for each image in a special per-image location (the
375  * image_metadata array).
376  */
377 int
378 read_lookup_table(WIMStruct *w)
379 {
380         int ret;
381         size_t num_entries;
382         struct wim_lookup_table *table;
383         struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
384         struct wim_lookup_table_entry_disk
385                         table_buf[BUFFER_SIZE / sizeof(struct wim_lookup_table_entry_disk)]
386                                 _aligned_attribute(8);
387
388         BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
389                      WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
390
391         off_t offset;
392         size_t buf_entries_remaining;
393         const struct wim_lookup_table_entry_disk *disk_entry;
394
395         DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
396               w->hdr.lookup_table_res_entry.offset,
397               w->hdr.lookup_table_res_entry.original_size);
398
399         if (resource_is_compressed(&w->hdr.lookup_table_res_entry)) {
400                 ERROR("Didn't expect a compressed lookup table!");
401                 ERROR("Ask the author to implement support for this.");
402                 return WIMLIB_ERR_COMPRESSED_LOOKUP_TABLE;
403         }
404
405         num_entries = w->hdr.lookup_table_res_entry.size /
406                       sizeof(struct wim_lookup_table_entry_disk);
407         table = new_lookup_table(num_entries * 2 + 1);
408         if (!table) {
409                 ERROR("Failed to allocate stream hash table of size %zu",
410                       num_entries * 2 + 1);
411                 return WIMLIB_ERR_NOMEM;
412         }
413
414         w->current_image = 0;
415         offset = w->hdr.lookup_table_res_entry.offset;
416         buf_entries_remaining = 0;
417         for (; num_entries != 0;
418              num_entries--, buf_entries_remaining--, disk_entry++)
419         {
420                 if (buf_entries_remaining == 0) {
421                         size_t entries_to_read, bytes_to_read;
422
423                         entries_to_read = min(ARRAY_LEN(table_buf), num_entries);
424                         bytes_to_read = entries_to_read * sizeof(struct wim_lookup_table_entry_disk);
425                         if (full_pread(w->in_fd, table_buf,
426                                        bytes_to_read, offset) != bytes_to_read)
427                         {
428                                 ERROR_WITH_ERRNO("Error reading lookup table "
429                                                  "(offset=%"PRIu64")", offset);
430                                 ret = WIMLIB_ERR_READ;
431                                 goto out_free_lookup_table;
432                         }
433                         offset += bytes_to_read;
434                         disk_entry = table_buf;
435                         buf_entries_remaining = entries_to_read;
436                 }
437                 cur_entry = new_lookup_table_entry();
438                 if (!cur_entry) {
439                         ret = WIMLIB_ERR_NOMEM;
440                         goto out_free_lookup_table;
441                 }
442
443                 cur_entry->wim = w;
444                 cur_entry->resource_location = RESOURCE_IN_WIM;
445                 get_resource_entry(&disk_entry->resource_entry, &cur_entry->resource_entry);
446                 cur_entry->part_number = le16_to_cpu(disk_entry->part_number);
447                 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
448                 copy_hash(cur_entry->hash, disk_entry->hash);
449
450                 if (cur_entry->part_number != w->hdr.part_number) {
451                         WARNING("A lookup table entry in part %hu of the WIM "
452                                 "points to part %hu (ignoring it)",
453                                 w->hdr.part_number, cur_entry->part_number);
454                         free_lookup_table_entry(cur_entry);
455                         continue;
456                 }
457
458                 if (is_zero_hash(cur_entry->hash)) {
459                         WARNING("The WIM lookup table contains an entry with a "
460                                 "SHA1 message digest of all 0's (ignoring it)");
461                         free_lookup_table_entry(cur_entry);
462                         continue;
463                 }
464
465                 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
466                     && (cur_entry->resource_entry.size !=
467                         cur_entry->resource_entry.original_size))
468                 {
469                         if (wimlib_print_errors) {
470                                 WARNING("Found uncompressed resource with "
471                                         "original size (%"PRIu64") not the same "
472                                         "as compressed size (%"PRIu64")",
473                                         cur_entry->resource_entry.original_size,
474                                         cur_entry->resource_entry.size);
475                                 if (cur_entry->resource_entry.original_size) {
476                                         WARNING("Overriding compressed size with original size.");
477                                         cur_entry->resource_entry.size =
478                                                 cur_entry->resource_entry.original_size;
479                                 } else {
480                                         WARNING("Overriding original size with compressed size");
481                                         cur_entry->resource_entry.original_size =
482                                                 cur_entry->resource_entry.size;
483                                 }
484                         }
485                 }
486
487                 if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
488                         /* Lookup table entry for a metadata resource */
489                         if (cur_entry->refcnt != 1) {
490                                 if (wimlib_print_errors) {
491                                         ERROR("Found metadata resource with refcnt != 1:");
492                                         print_lookup_table_entry(cur_entry, stderr);
493                                 }
494                                 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
495                                 goto out_free_cur_entry;
496                         }
497
498                         if (w->hdr.part_number != 1) {
499                                 WARNING("Ignoring metadata resource found in a "
500                                         "non-first part of the split WIM");
501                                 free_lookup_table_entry(cur_entry);
502                                 continue;
503                         }
504                         if (w->current_image == w->hdr.image_count) {
505                                 WARNING("The WIM header says there are %u images "
506                                         "in the WIM, but we found more metadata "
507                                         "resources than this (ignoring the extra)",
508                                         w->hdr.image_count);
509                                 free_lookup_table_entry(cur_entry);
510                                 continue;
511                         }
512
513                         /* Notice very carefully:  We are assigning the metadata
514                          * resources in the exact order mirrored by their lookup
515                          * table entries on disk, which is the behavior of
516                          * Microsoft's software.  In particular, this overrides
517                          * the actual locations of the metadata resources
518                          * themselves in the WIM file as well as any information
519                          * written in the XML data. */
520                         DEBUG("Found metadata resource for image %u at "
521                               "offset %"PRIu64".",
522                               w->current_image + 1,
523                               cur_entry->resource_entry.offset);
524                         w->image_metadata[
525                                 w->current_image++]->metadata_lte = cur_entry;
526                 } else {
527                         /* Lookup table entry for a stream that is not a
528                          * metadata resource */
529                         duplicate_entry = __lookup_resource(table, cur_entry->hash);
530                         if (duplicate_entry) {
531                                 if (wimlib_print_errors) {
532                                         WARNING("The WIM lookup table contains two entries with the "
533                                               "same SHA1 message digest!");
534                                         WARNING("The first entry is:");
535                                         print_lookup_table_entry(duplicate_entry, stderr);
536                                         WARNING("The second entry is:");
537                                         print_lookup_table_entry(cur_entry, stderr);
538                                 }
539                                 free_lookup_table_entry(cur_entry);
540                                 continue;
541                         } else {
542                                 lookup_table_insert(table, cur_entry);
543                         }
544                 }
545         }
546
547         if (w->hdr.part_number == 1 && w->current_image != w->hdr.image_count)
548         {
549                 ERROR("The WIM header says there are %u images "
550                       "in the WIM, but we only found %d metadata "
551                       "resources!", w->hdr.image_count, w->current_image);
552                 ret = WIMLIB_ERR_IMAGE_COUNT;
553                 goto out_free_lookup_table;
554         }
555         DEBUG("Done reading lookup table.");
556         w->lookup_table = table;
557         ret = 0;
558         goto out;
559 out_free_cur_entry:
560         FREE(cur_entry);
561 out_free_lookup_table:
562         free_lookup_table(table);
563 out:
564         w->current_image = 0;
565         return ret;
566 }
567
568
569 static void
570 write_lookup_table_entry(const struct wim_lookup_table_entry *lte,
571                          struct wim_lookup_table_entry_disk *disk_entry)
572 {
573         put_resource_entry(&lte->output_resource_entry, &disk_entry->resource_entry);
574         disk_entry->part_number = cpu_to_le16(lte->part_number);
575         disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
576         copy_hash(disk_entry->hash, lte->hash);
577 }
578
579 int
580 write_lookup_table_from_stream_list(struct list_head *stream_list,
581                                     int out_fd,
582                                     struct resource_entry *out_res_entry)
583 {
584         int ret;
585         off_t start_offset;
586         struct wim_lookup_table_entry_disk
587                         table_buf[BUFFER_SIZE / sizeof(struct wim_lookup_table_entry_disk)]
588                                 _aligned_attribute(8);
589         size_t table_size;
590         size_t bytes_to_write;
591         struct wim_lookup_table_entry *lte;
592         size_t cur_idx;
593
594         start_offset = filedes_offset(out_fd);
595         if (start_offset == -1)
596                 goto write_error;
597
598         table_size = 0;
599         cur_idx = 0;
600         list_for_each_entry(lte, stream_list, lookup_table_list) {
601                 if (cur_idx == ARRAY_LEN(table_buf)) {
602                         bytes_to_write = sizeof(table_buf);
603                         if (full_write(out_fd, table_buf,
604                                        bytes_to_write) != bytes_to_write)
605                                 goto write_error;
606                         table_size += bytes_to_write;
607                         cur_idx = 0;
608                 }
609                 write_lookup_table_entry(lte, &table_buf[cur_idx]);
610                 cur_idx++;
611         }
612         if (cur_idx != 0) {
613                 bytes_to_write = cur_idx * sizeof(struct wim_lookup_table_entry_disk);
614                 if (full_write(out_fd, table_buf,
615                                bytes_to_write) != bytes_to_write)
616                         goto write_error;
617                 table_size += bytes_to_write;
618         }
619         out_res_entry->offset        = start_offset;
620         out_res_entry->size          = table_size;
621         out_res_entry->original_size = table_size;
622         out_res_entry->flags         = WIM_RESHDR_FLAG_METADATA;
623         ret = 0;
624 out:
625         return ret;
626 write_error:
627         ERROR_WITH_ERRNO("Failed to write lookup table");
628         ret = WIMLIB_ERR_WRITE;
629         goto out;
630 }
631
632 static int
633 append_lookup_table_entry(struct wim_lookup_table_entry *lte, void *_list)
634 {
635         if (lte->out_refcnt != 0)
636                 list_add_tail(&lte->lookup_table_list, (struct list_head*)_list);
637         return 0;
638 }
639
640 /* Writes the WIM lookup table to the output file. */
641 int
642 write_lookup_table(WIMStruct *w, int image, struct resource_entry *out_res_entry)
643 {
644         LIST_HEAD(stream_list);
645         int start_image;
646         int end_image;
647
648         if (image == WIMLIB_ALL_IMAGES) {
649                 start_image = 1;
650                 end_image = w->hdr.image_count;
651         } else {
652                 start_image = image;
653                 end_image = image;
654         }
655
656         for (int i = start_image; i <= end_image; i++) {
657                 struct wim_lookup_table_entry *metadata_lte;
658
659                 metadata_lte = w->image_metadata[i - 1]->metadata_lte;
660                 metadata_lte->out_refcnt = 1;
661                 metadata_lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
662                 append_lookup_table_entry(metadata_lte, &stream_list);
663         }
664         for_lookup_table_entry(w->lookup_table,
665                                append_lookup_table_entry,
666                                &stream_list);
667         return write_lookup_table_from_stream_list(&stream_list,
668                                                    w->out_fd,
669                                                    out_res_entry);
670 }
671
672 int
673 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
674 {
675         lte->real_refcnt = 0;
676         return 0;
677 }
678
679 int
680 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
681 {
682         lte->out_refcnt = 0;
683         return 0;
684 }
685
686 int
687 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
688 {
689         if (lte->extracted_file != NULL) {
690                 FREE(lte->extracted_file);
691                 lte->extracted_file = NULL;
692         }
693         return 0;
694 }
695
696 void
697 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
698 {
699         if (!lte) {
700                 tputc(T('\n'), out);
701                 return;
702         }
703         tfprintf(out, T("Offset            = %"PRIu64" bytes\n"),
704                  lte->resource_entry.offset);
705
706         tfprintf(out, T("Size              = %"PRIu64" bytes\n"),
707                  (u64)lte->resource_entry.size);
708
709         tfprintf(out, T("Original size     = %"PRIu64" bytes\n"),
710                  lte->resource_entry.original_size);
711
712         tfprintf(out, T("Part Number       = %hu\n"), lte->part_number);
713         tfprintf(out, T("Reference Count   = %u\n"), lte->refcnt);
714
715         if (lte->unhashed) {
716                 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
717                          lte->back_inode, lte->back_stream_id);
718         } else {
719                 tfprintf(out, T("Hash              = 0x"));
720                 print_hash(lte->hash, out);
721                 tputc(T('\n'), out);
722         }
723
724         tfprintf(out, T("Flags             = "));
725         u8 flags = lte->resource_entry.flags;
726         if (flags & WIM_RESHDR_FLAG_COMPRESSED)
727                 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
728         if (flags & WIM_RESHDR_FLAG_FREE)
729                 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
730         if (flags & WIM_RESHDR_FLAG_METADATA)
731                 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
732         if (flags & WIM_RESHDR_FLAG_SPANNED)
733                 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
734         tputc(T('\n'), out);
735         switch (lte->resource_location) {
736         case RESOURCE_IN_WIM:
737                 if (lte->wim->filename) {
738                         tfprintf(out, T("WIM file          = `%"TS"'\n"),
739                                  lte->wim->filename);
740                 }
741                 break;
742 #ifdef __WIN32__
743         case RESOURCE_WIN32:
744         case RESOURCE_WIN32_ENCRYPTED:
745 #else
746         case RESOURCE_IN_FILE_ON_DISK:
747 #endif
748                 tfprintf(out, T("File on Disk      = `%"TS"'\n"),
749                          lte->file_on_disk);
750                 break;
751 #ifdef WITH_FUSE
752         case RESOURCE_IN_STAGING_FILE:
753                 tfprintf(out, T("Staging File      = `%"TS"'\n"),
754                                 lte->staging_file_name);
755                 break;
756 #endif
757         default:
758                 break;
759         }
760         tputc(T('\n'), out);
761 }
762
763 static int
764 do_print_lookup_table_entry(struct wim_lookup_table_entry *lte, void *fp)
765 {
766         print_lookup_table_entry(lte, (FILE*)fp);
767         return 0;
768 }
769
770 /*
771  * Prints the lookup table of a WIM file.
772  */
773 WIMLIBAPI void
774 wimlib_print_lookup_table(WIMStruct *wim)
775 {
776         for (int i = 0; i < wim->hdr.image_count; i++)
777                 print_lookup_table_entry(wim->image_metadata[i]->metadata_lte, stdout);
778         for_lookup_table_entry(wim->lookup_table,
779                                do_print_lookup_table_entry,
780                                stdout);
781 }
782
783 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
784  * lookup table, or NULL if there is none.  */
785 struct wim_lookup_table_entry *
786 __lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
787 {
788         size_t i;
789         struct wim_lookup_table_entry *lte;
790         struct hlist_node *pos;
791
792         wimlib_assert(table != NULL);
793         wimlib_assert(hash != NULL);
794
795         i = *(size_t*)hash % table->capacity;
796         hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
797                 if (hashes_equal(hash, lte->hash))
798                         return lte;
799         return NULL;
800 }
801
802 #ifdef WITH_FUSE
803 /*
804  * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
805  * given a path name.
806  *
807  * This is only for pre-resolved inodes.
808  */
809 int
810 lookup_resource(WIMStruct *w,
811                 const tchar *path,
812                 int lookup_flags,
813                 struct wim_dentry **dentry_ret,
814                 struct wim_lookup_table_entry **lte_ret,
815                 u16 *stream_idx_ret)
816 {
817         struct wim_dentry *dentry;
818         struct wim_lookup_table_entry *lte;
819         u16 stream_idx;
820         const tchar *stream_name = NULL;
821         struct wim_inode *inode;
822         tchar *p = NULL;
823
824         if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
825                 stream_name = path_stream_name(path);
826                 if (stream_name) {
827                         p = (tchar*)stream_name - 1;
828                         *p = T('\0');
829                 }
830         }
831
832         dentry = get_dentry(w, path);
833         if (p)
834                 *p = T(':');
835         if (!dentry)
836                 return -errno;
837
838         inode = dentry->d_inode;
839
840         if (!inode->i_resolved)
841                 if (inode_resolve_ltes(inode, w->lookup_table))
842                         return -EIO;
843
844         if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
845               && inode_is_directory(inode))
846                 return -EISDIR;
847
848         if (stream_name) {
849                 struct wim_ads_entry *ads_entry;
850                 u16 ads_idx;
851                 ads_entry = inode_get_ads_entry(inode, stream_name,
852                                                 &ads_idx);
853                 if (ads_entry) {
854                         stream_idx = ads_idx + 1;
855                         lte = ads_entry->lte;
856                         goto out;
857                 } else {
858                         return -ENOENT;
859                 }
860         } else {
861                 lte = inode->i_lte;
862                 stream_idx = 0;
863         }
864 out:
865         if (dentry_ret)
866                 *dentry_ret = dentry;
867         if (lte_ret)
868                 *lte_ret = lte;
869         if (stream_idx_ret)
870                 *stream_idx_ret = stream_idx;
871         return 0;
872 }
873 #endif
874
875 /*
876  * XXX Probably should store the compression type directly in the lookup table
877  * entry
878  */
879 int
880 wim_resource_compression_type(const struct wim_lookup_table_entry *lte)
881 {
882         if (!(lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
883             || lte->resource_location != RESOURCE_IN_WIM)
884                 return WIMLIB_COMPRESSION_TYPE_NONE;
885         return wimlib_get_compression_type(lte->wim);
886 }
887
888 /* Resolve an inode's lookup table entries
889  *
890  * This replaces the SHA1 hash fields (which are used to lookup an entry in the
891  * lookup table) with pointers directly to the lookup table entries.  A circular
892  * linked list of streams sharing the same lookup table entry is created.
893  *
894  * This function always succeeds; unresolved lookup table entries are given a
895  * NULL pointer.
896  */
897 int
898 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
899 {
900         const u8 *hash;
901
902         if (!inode->i_resolved) {
903                 struct wim_lookup_table_entry *lte, *ads_lte;
904
905                 /* Resolve the default file stream */
906                 lte = NULL;
907                 hash = inode->i_hash;
908                 if (!is_zero_hash(hash)) {
909                         lte = __lookup_resource(table, hash);
910                         if (unlikely(!lte))
911                                 goto resource_not_found;
912                 }
913
914                 /* Resolve the alternate data streams */
915                 struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
916                 for (u16 i = 0; i < inode->i_num_ads; i++) {
917                         struct wim_ads_entry *cur_entry;
918
919                         ads_lte = NULL;
920                         cur_entry = &inode->i_ads_entries[i];
921                         hash = cur_entry->hash;
922                         if (!is_zero_hash(hash)) {
923                                 ads_lte = __lookup_resource(table, hash);
924                                 if (unlikely(!ads_lte))
925                                         goto resource_not_found;
926                         }
927                         ads_ltes[i] = ads_lte;
928                 }
929                 inode->i_lte = lte;
930                 for (u16 i = 0; i < inode->i_num_ads; i++)
931                         inode->i_ads_entries[i].lte = ads_ltes[i];
932                 inode->i_resolved = 1;
933         }
934         return 0;
935 resource_not_found:
936         if (wimlib_print_errors) {
937                 ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
938                 tfprintf(stderr, T("        SHA-1 message digest of missing resource:\n        "));
939                 print_hash(hash, stderr);
940                 tputc(T('\n'), stderr);
941         }
942         return WIMLIB_ERR_RESOURCE_NOT_FOUND;
943 }
944
945 void
946 inode_unresolve_ltes(struct wim_inode *inode)
947 {
948         if (inode->i_resolved) {
949                 if (inode->i_lte)
950                         copy_hash(inode->i_hash, inode->i_lte->hash);
951                 else
952                         zero_out_hash(inode->i_hash);
953
954                 for (u16 i = 0; i < inode->i_num_ads; i++) {
955                         if (inode->i_ads_entries[i].lte)
956                                 copy_hash(inode->i_ads_entries[i].hash,
957                                           inode->i_ads_entries[i].lte->hash);
958                         else
959                                 zero_out_hash(inode->i_ads_entries[i].hash);
960                 }
961                 inode->i_resolved = 0;
962         }
963 }
964
965 /*
966  * Returns the lookup table entry for stream @stream_idx of the inode, where
967  * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
968  * corresponds to an alternate data stream.
969  *
970  * This works for both resolved and un-resolved inodes.
971  */
972 struct wim_lookup_table_entry *
973 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
974                  const struct wim_lookup_table *table)
975 {
976         if (inode->i_resolved)
977                 return inode_stream_lte_resolved(inode, stream_idx);
978         else
979                 return inode_stream_lte_unresolved(inode, stream_idx, table);
980 }
981
982 struct wim_lookup_table_entry *
983 inode_unnamed_lte_resolved(const struct wim_inode *inode)
984 {
985         wimlib_assert(inode->i_resolved);
986         for (unsigned i = 0; i <= inode->i_num_ads; i++) {
987                 if (inode_stream_name_nbytes(inode, i) == 0 &&
988                     !is_zero_hash(inode_stream_hash_resolved(inode, i)))
989                 {
990                         return inode_stream_lte_resolved(inode, i);
991                 }
992         }
993         return NULL;
994 }
995
996 struct wim_lookup_table_entry *
997 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
998                              const struct wim_lookup_table *table)
999 {
1000         wimlib_assert(!inode->i_resolved);
1001         for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1002                 if (inode_stream_name_nbytes(inode, i) == 0 &&
1003                     !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
1004                 {
1005                         return inode_stream_lte_unresolved(inode, i, table);
1006                 }
1007         }
1008         return NULL;
1009 }
1010
1011 /* Return the lookup table entry for the unnamed data stream of an inode, or
1012  * NULL if there is none.
1013  *
1014  * You'd think this would be easier than it actually is, since the unnamed data
1015  * stream should be the one referenced from the inode itself.  Alas, if there
1016  * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
1017  * data stream in one of the alternate data streams instead of inside the WIM
1018  * dentry itself.  So we need to check the alternate data streams too.
1019  *
1020  * Also, note that a dentry may appear to have more than one unnamed stream, but
1021  * if the SHA1 message digest is all 0's then the corresponding stream does not
1022  * really "count" (this is the case for the inode's own file stream when the
1023  * file stream that should be there is actually in one of the alternate stream
1024  * entries.).  This is despite the fact that we may need to extract such a
1025  * missing entry as an empty file or empty named data stream.
1026  */
1027 struct wim_lookup_table_entry *
1028 inode_unnamed_lte(const struct wim_inode *inode,
1029                   const struct wim_lookup_table *table)
1030 {
1031         if (inode->i_resolved)
1032                 return inode_unnamed_lte_resolved(inode);
1033         else
1034                 return inode_unnamed_lte_unresolved(inode, table);
1035 }
1036
1037 static int
1038 lte_add_stream_size(struct wim_lookup_table_entry *lte, void *total_bytes_p)
1039 {
1040         *(u64*)total_bytes_p += lte->resource_entry.size;
1041         return 0;
1042 }
1043
1044 u64
1045 lookup_table_total_stream_size(struct wim_lookup_table *table)
1046 {
1047         u64 total_size = 0;
1048         for_lookup_table_entry(table, lte_add_stream_size, &total_size);
1049         return total_size;
1050 }
1051
1052 struct wim_lookup_table_entry **
1053 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1054 {
1055         wimlib_assert(lte->unhashed);
1056         struct wim_inode *inode = lte->back_inode;
1057         u32 stream_id = lte->back_stream_id;
1058         if (stream_id == 0)
1059                 return &inode->i_lte;
1060         else
1061                 for (u16 i = 0; i < inode->i_num_ads; i++)
1062                         if (inode->i_ads_entries[i].stream_id == stream_id)
1063                                 return &inode->i_ads_entries[i].lte;
1064         wimlib_assert(0);
1065         return NULL;
1066 }
1067
1068 /* Calculate the SHA1 message digest of a stream and move it from the list of
1069  * unhashed streams to the stream lookup table, possibly joining it with an
1070  * existing lookup table entry for an identical stream.
1071  *
1072  * @lte:  An unhashed lookup table entry.
1073  * @lookup_table:  Lookup table for the WIM.
1074  * @lte_ret:  On success, write a pointer to the resulting lookup table
1075  *            entry to this location.  This will be the same as @lte
1076  *            if it was inserted into the lookup table, or different if
1077  *            a duplicate stream was found.
1078  *
1079  * Returns 0 on success; nonzero if there is an error reading the stream.
1080  */
1081 int
1082 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1083                      struct wim_lookup_table *lookup_table,
1084                      struct wim_lookup_table_entry **lte_ret)
1085 {
1086         int ret;
1087         struct wim_lookup_table_entry *duplicate_lte;
1088         struct wim_lookup_table_entry **back_ptr;
1089
1090         wimlib_assert(lte->unhashed);
1091
1092         /* back_ptr must be saved because @back_inode and @back_stream_id are in
1093          * union with the SHA1 message digest and will no longer be valid once
1094          * the SHA1 has been calculated. */
1095         back_ptr = retrieve_lte_pointer(lte);
1096
1097         ret = sha1_resource(lte);
1098         if (ret)
1099                 return ret;
1100
1101         /* Look for a duplicate stream */
1102         duplicate_lte = __lookup_resource(lookup_table, lte->hash);
1103         list_del(&lte->unhashed_list);
1104         if (duplicate_lte) {
1105                 /* We have a duplicate stream.  Transfer the reference counts
1106                  * from this stream to the duplicate, update the reference to
1107                  * this stream (in an inode or ads_entry) to point to the
1108                  * duplicate, then free this stream. */
1109                 wimlib_assert(!(duplicate_lte->unhashed));
1110                 duplicate_lte->refcnt += lte->refcnt;
1111                 duplicate_lte->out_refcnt += lte->refcnt;
1112                 *back_ptr = duplicate_lte;
1113                 free_lookup_table_entry(lte);
1114                 lte = duplicate_lte;
1115         } else {
1116                 /* No duplicate stream, so we need to insert
1117                  * this stream into the lookup table and treat
1118                  * it as a hashed stream. */
1119                 lookup_table_insert(lookup_table, lte);
1120                 lte->unhashed = 0;
1121         }
1122         if (lte_ret)
1123                 *lte_ret = lte;
1124         return 0;
1125 }
1126