4 * Lookup table, implemented as a hash table, that maps SHA1 message digests to
5 * data streams; plus code to read and write the corresponding on-disk data.
9 * Copyright (C) 2012, 2013 Eric Biggers
11 * This file is part of wimlib, a library for working with WIM files.
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)
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
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/.
31 #include "wimlib/assert.h"
32 #include "wimlib/endianness.h"
33 #include "wimlib/error.h"
34 #include "wimlib/lookup_table.h"
35 #include "wimlib/metadata.h"
36 #include "wimlib/ntfs_3g.h"
37 #include "wimlib/resource.h"
38 #include "wimlib/util.h"
39 #include "wimlib/write.h"
43 #include <unistd.h> /* for unlink() */
47 * This is a logical mapping from SHA1 message digests to the data streams
50 * Here it is implemented as a hash table.
52 * Note: Everything will break horribly if there is a SHA1 collision.
54 struct wim_lookup_table {
55 struct hlist_head *array;
60 struct wim_lookup_table *
61 new_lookup_table(size_t capacity)
63 struct wim_lookup_table *table;
64 struct hlist_head *array;
66 table = MALLOC(sizeof(struct wim_lookup_table));
70 array = CALLOC(capacity, sizeof(array[0]));
76 table->num_entries = 0;
77 table->capacity = capacity;
82 ERROR("Failed to allocate memory for lookup table "
83 "with capacity %zu", capacity);
88 do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
90 free_lookup_table_entry(entry);
95 free_lookup_table(struct wim_lookup_table *table)
97 DEBUG("Freeing lookup table.");
102 for_lookup_table_entry(table,
103 do_free_lookup_table_entry,
110 struct wim_lookup_table_entry *
111 new_lookup_table_entry(void)
113 struct wim_lookup_table_entry *lte;
115 lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
121 /* lte->resource_location = RESOURCE_NONEXISTENT */
122 BUILD_BUG_ON(RESOURCE_NONEXISTENT != 0);
127 struct wim_lookup_table_entry *
128 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
130 struct wim_lookup_table_entry *new;
132 new = memdup(old, sizeof(struct wim_lookup_table_entry));
136 new->extracted_file = NULL;
137 switch (new->resource_location) {
138 case RESOURCE_IN_WIM:
139 list_add(&new->rspec_node, &new->rspec->stream_list);
142 case RESOURCE_IN_FILE_ON_DISK:
144 case RESOURCE_WIN32_ENCRYPTED:
147 case RESOURCE_IN_STAGING_FILE:
148 BUILD_BUG_ON((void*)&old->file_on_disk !=
149 (void*)&old->staging_file_name);
151 new->file_on_disk = TSTRDUP(old->file_on_disk);
152 if (new->file_on_disk == NULL)
155 case RESOURCE_IN_ATTACHED_BUFFER:
156 new->attached_buffer = memdup(old->attached_buffer, old->size);
157 if (new->attached_buffer == NULL)
161 case RESOURCE_IN_NTFS_VOLUME:
163 struct ntfs_location *loc;
164 loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
168 loc->stream_name = NULL;
170 loc->path = STRDUP(old->ntfs_loc->path);
171 if (loc->path == NULL)
173 if (loc->stream_name_nchars != 0) {
174 loc->stream_name = memdup(old->ntfs_loc->stream_name,
175 loc->stream_name_nchars * 2);
176 if (loc->stream_name == NULL)
188 free_lookup_table_entry(new);
193 lte_put_resource(struct wim_lookup_table_entry *lte)
195 switch (lte->resource_location) {
196 case RESOURCE_IN_WIM:
197 list_del(<e->rspec_node);
198 if (list_empty(<e->rspec->stream_list))
201 case RESOURCE_IN_FILE_ON_DISK:
203 case RESOURCE_WIN32_ENCRYPTED:
206 case RESOURCE_IN_STAGING_FILE:
207 BUILD_BUG_ON((void*)<e->file_on_disk !=
208 (void*)<e->staging_file_name);
210 case RESOURCE_IN_ATTACHED_BUFFER:
211 BUILD_BUG_ON((void*)<e->file_on_disk !=
212 (void*)<e->attached_buffer);
213 FREE(lte->file_on_disk);
216 case RESOURCE_IN_NTFS_VOLUME:
218 FREE(lte->ntfs_loc->path);
219 FREE(lte->ntfs_loc->stream_name);
230 free_lookup_table_entry(struct wim_lookup_table_entry *lte)
233 lte_put_resource(lte);
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
243 lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
244 struct wim_lookup_table *table)
246 wimlib_assert(lte->refcnt != 0);
248 if (--lte->refcnt == 0) {
250 list_del(<e->unhashed_list);
252 /* If the stream has been extracted to a staging file
253 * for a FUSE mount, unlink the staging file. (Note
254 * that there still may be open file descriptors to it.)
256 if (lte->resource_location == RESOURCE_IN_STAGING_FILE)
257 unlink(lte->staging_file_name);
260 lookup_table_unlink(table, lte);
263 /* If FUSE mounts are enabled, we don't actually free the entry
264 * until the last file descriptor has been closed by
265 * lte_decrement_num_opened_fds(). */
267 if (lte->num_opened_fds == 0)
269 free_lookup_table_entry(lte);
275 lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
277 wimlib_assert(lte->num_opened_fds != 0);
279 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
280 free_lookup_table_entry(lte);
285 lookup_table_insert_raw(struct wim_lookup_table *table,
286 struct wim_lookup_table_entry *lte)
288 size_t i = lte->hash_short % table->capacity;
290 hlist_add_head(<e->hash_list, &table->array[i]);
294 enlarge_lookup_table(struct wim_lookup_table *table)
296 size_t old_capacity, new_capacity;
297 struct hlist_head *old_array, *new_array;
298 struct wim_lookup_table_entry *lte;
299 struct hlist_node *cur, *tmp;
302 old_capacity = table->capacity;
303 new_capacity = old_capacity * 2;
304 new_array = CALLOC(new_capacity, sizeof(struct hlist_head));
305 if (new_array == NULL)
307 old_array = table->array;
308 table->array = new_array;
309 table->capacity = new_capacity;
311 for (i = 0; i < old_capacity; i++) {
312 hlist_for_each_entry_safe(lte, cur, tmp, &old_array[i], hash_list) {
313 hlist_del(<e->hash_list);
314 lookup_table_insert_raw(table, lte);
320 /* Inserts an entry into the lookup table. */
322 lookup_table_insert(struct wim_lookup_table *table,
323 struct wim_lookup_table_entry *lte)
325 lookup_table_insert_raw(table, lte);
326 if (++table->num_entries > table->capacity)
327 enlarge_lookup_table(table);
330 /* Unlinks a lookup table entry from the table; does not free it. */
332 lookup_table_unlink(struct wim_lookup_table *table,
333 struct wim_lookup_table_entry *lte)
335 wimlib_assert(!lte->unhashed);
336 wimlib_assert(table->num_entries != 0);
338 hlist_del(<e->hash_list);
339 table->num_entries--;
342 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
343 * lookup table, or NULL if there is none. */
344 struct wim_lookup_table_entry *
345 lookup_stream(const struct wim_lookup_table *table, const u8 hash[])
348 struct wim_lookup_table_entry *lte;
349 struct hlist_node *pos;
351 i = *(size_t*)hash % table->capacity;
352 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
353 if (hashes_equal(hash, lte->hash))
358 /* Calls a function on all the entries in the WIM lookup table. Stop early and
359 * return nonzero if any call to the function returns nonzero. */
361 for_lookup_table_entry(struct wim_lookup_table *table,
362 int (*visitor)(struct wim_lookup_table_entry *, void *),
365 struct wim_lookup_table_entry *lte;
366 struct hlist_node *pos, *tmp;
369 for (size_t i = 0; i < table->capacity; i++) {
370 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
373 ret = visitor(lte, arg);
381 /* qsort() callback that sorts streams (represented by `struct
382 * wim_lookup_table_entry's) into an order optimized for reading.
384 * Sorting is done primarily by resource location, then secondarily by a
385 * per-resource location order. For example, resources in WIM files are sorted
386 * primarily by part number, then secondarily by offset, as to implement optimal
387 * reading of either a standalone or split WIM. */
389 cmp_streams_by_sequential_order(const void *p1, const void *p2)
391 const struct wim_lookup_table_entry *lte1, *lte2;
393 WIMStruct *wim1, *wim2;
395 lte1 = *(const struct wim_lookup_table_entry**)p1;
396 lte2 = *(const struct wim_lookup_table_entry**)p2;
398 v = (int)lte1->resource_location - (int)lte2->resource_location;
400 /* Different resource locations? */
404 switch (lte1->resource_location) {
405 case RESOURCE_IN_WIM:
406 wim1 = lte1->rspec->wim;
407 wim2 = lte2->rspec->wim;
409 /* Different (possibly split) WIMs? */
411 v = memcmp(wim1->hdr.guid, wim2->hdr.guid, WIM_GID_LEN);
416 /* Different part numbers in the same WIM? */
417 v = (int)wim1->hdr.part_number - (int)wim2->hdr.part_number;
421 if (lte1->rspec->offset_in_wim != lte2->rspec->offset_in_wim)
422 return cmp_u64(lte1->rspec->offset_in_wim,
423 lte2->rspec->offset_in_wim);
425 return cmp_u64(lte1->offset_in_res, lte2->offset_in_res);
427 case RESOURCE_IN_FILE_ON_DISK:
429 case RESOURCE_IN_STAGING_FILE:
432 case RESOURCE_WIN32_ENCRYPTED:
434 /* Compare files by path: just a heuristic that will place files
435 * in the same directory next to each other. */
436 return tstrcmp(lte1->file_on_disk, lte2->file_on_disk);
438 case RESOURCE_IN_NTFS_VOLUME:
439 return tstrcmp(lte1->ntfs_loc->path, lte2->ntfs_loc->path);
442 /* No additional sorting order defined for this resource
443 * location (e.g. RESOURCE_IN_ATTACHED_BUFFER); simply compare
444 * everything equal to each other. */
450 sort_stream_list(struct list_head *stream_list,
451 size_t list_head_offset,
452 int (*compar)(const void *, const void*))
454 struct list_head *cur;
455 struct wim_lookup_table_entry **array;
458 size_t num_streams = 0;
460 list_for_each(cur, stream_list)
463 if (num_streams <= 1)
466 array_size = num_streams * sizeof(array[0]);
467 array = MALLOC(array_size);
469 return WIMLIB_ERR_NOMEM;
471 cur = stream_list->next;
472 for (i = 0; i < num_streams; i++) {
473 array[i] = (struct wim_lookup_table_entry*)((u8*)cur -
478 qsort(array, num_streams, sizeof(array[0]), compar);
480 INIT_LIST_HEAD(stream_list);
481 for (i = 0; i < num_streams; i++) {
482 list_add_tail((struct list_head*)
483 ((u8*)array[i] + list_head_offset),
490 /* Sort the specified list of streams in an order optimized for reading. */
492 sort_stream_list_by_sequential_order(struct list_head *stream_list,
493 size_t list_head_offset)
495 return sort_stream_list(stream_list, list_head_offset,
496 cmp_streams_by_sequential_order);
501 add_lte_to_array(struct wim_lookup_table_entry *lte,
504 struct wim_lookup_table_entry ***pp = _pp;
509 /* Iterate through the lookup table entries, but first sort them by stream
510 * offset in the WIM. Caution: this is intended to be used when the stream
511 * offset field has actually been set. */
513 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
514 int (*visitor)(struct wim_lookup_table_entry *,
518 struct wim_lookup_table_entry **lte_array, **p;
519 size_t num_streams = table->num_entries;
522 lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
524 return WIMLIB_ERR_NOMEM;
526 for_lookup_table_entry(table, add_lte_to_array, &p);
528 wimlib_assert(p == lte_array + num_streams);
530 qsort(lte_array, num_streams, sizeof(lte_array[0]),
531 cmp_streams_by_sequential_order);
533 for (size_t i = 0; i < num_streams; i++) {
534 ret = visitor(lte_array[i], arg);
542 /* On-disk format of a WIM lookup table entry (stream entry). */
543 struct wim_lookup_table_entry_disk {
544 /* Size, offset, and flags of the stream. */
545 struct wim_reshdr_disk reshdr;
547 /* Which part of the split WIM this stream is in; indexed from 1. */
550 /* Reference count of this stream over all WIM images. */
553 /* SHA1 message digest of the uncompressed data of this stream, or
554 * optionally all zeroes if this stream is of zero length. */
555 u8 hash[SHA1_HASH_SIZE];
558 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
560 /* Validate the size and location of a WIM resource. */
562 validate_resource(const struct wim_resource_spec *rspec)
564 struct wim_lookup_table_entry *lte;
567 /* Verify that calculating the offset of the end of the resource doesn't
569 if (rspec->offset_in_wim + rspec->size_in_wim < rspec->size_in_wim)
572 /* Verify that each stream in the resource has a valid offset and size,
573 * and that no streams overlap, and that the streams were added in order
574 * of increasing offset. */
576 list_for_each_entry(lte, &rspec->stream_list, rspec_node) {
577 if (lte->offset_in_res + lte->size < lte->size ||
578 lte->offset_in_res + lte->size > rspec->uncompressed_size ||
579 lte->offset_in_res < cur_offset)
582 cur_offset = lte->offset_in_res + lte->size;
588 ERROR("Invalid resource entry!");
589 return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
593 * Reads the lookup table from a WIM file. Each entry specifies a stream that
594 * the WIM file contains, along with its location and SHA1 message digest.
596 * Saves lookup table entries for non-metadata streams in a hash table, and
597 * saves the metadata entry for each image in a special per-image location (the
598 * image_metadata array).
601 * WIMLIB_ERR_SUCCESS (0)
602 * WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
603 * WIMLIB_ERR_RESOURCE_NOT_FOUND
605 * Or an error code caused by failure to read the lookup table into memory.
608 read_wim_lookup_table(WIMStruct *wim)
613 struct wim_lookup_table *table;
614 struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
615 struct wim_resource_spec *cur_rspec;
617 bool back_to_back_pack;
619 DEBUG("Reading lookup table.");
621 /* Sanity check: lookup table entries are 50 bytes each. */
622 BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
623 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
625 /* Calculate number of entries in the lookup table. */
626 num_entries = wim->hdr.lookup_table_reshdr.uncompressed_size /
627 sizeof(struct wim_lookup_table_entry_disk);
629 /* Read the lookup table into a buffer. */
630 ret = wim_reshdr_to_data(&wim->hdr.lookup_table_reshdr, wim, &buf);
634 /* Allocate a hash table to map SHA1 message digests into stream
635 * specifications. This is the in-memory "lookup table". */
636 table = new_lookup_table(num_entries * 2 + 1);
638 ERROR("Not enough memory to read lookup table.");
639 ret = WIMLIB_ERR_NOMEM;
643 /* Allocate and initalize stream entries from the raw lookup table
645 wim->current_image = 0;
647 for (i = 0; i < num_entries; i++) {
648 const struct wim_lookup_table_entry_disk *disk_entry =
649 &((const struct wim_lookup_table_entry_disk*)buf)[i];
651 struct wim_reshdr reshdr;
653 get_wim_reshdr(&disk_entry->reshdr, &reshdr);
655 DEBUG("reshdr: size_in_wim=%"PRIu64", "
656 "uncompressed_size=%"PRIu64", "
657 "offset_in_wim=%"PRIu64", "
659 reshdr.size_in_wim, reshdr.uncompressed_size,
660 reshdr.offset_in_wim, reshdr.flags);
662 if (wim->hdr.wim_version == WIM_VERSION_DEFAULT)
663 reshdr.flags &= ~WIM_RESHDR_FLAG_PACKED_STREAMS;
665 cur_entry = new_lookup_table_entry();
666 if (cur_entry == NULL) {
667 ERROR("Not enough memory to read lookup table!");
668 ret = WIMLIB_ERR_NOMEM;
672 part_number = le16_to_cpu(disk_entry->part_number);
673 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
674 copy_hash(cur_entry->hash, disk_entry->hash);
676 if (part_number != wim->hdr.part_number) {
677 WARNING("A lookup table entry in part %hu of the WIM "
678 "points to part %hu (ignoring it)",
679 wim->hdr.part_number, part_number);
680 free_lookup_table_entry(cur_entry);
684 if (!(reshdr.flags & (WIM_RESHDR_FLAG_PACKED_STREAMS |
685 WIM_RESHDR_FLAG_COMPRESSED))) {
686 if (reshdr.uncompressed_size != reshdr.size_in_wim) {
687 ERROR("Invalid resource entry!");
688 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
693 back_to_back_pack = false;
694 if (!(reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) ||
697 ((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
698 reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER &&
700 cur_rspec->size_in_wim != 0)))
702 /* Starting new run of streams that share the same WIM
704 struct wim_lookup_table_entry *prev_entry = NULL;
706 if (back_to_back_pack &&
707 !list_empty(&cur_rspec->stream_list))
709 prev_entry = list_entry(cur_rspec->stream_list.prev,
710 struct wim_lookup_table_entry,
712 lte_unbind_wim_resource_spec(prev_entry);
714 if (cur_rspec != NULL) {
715 ret = validate_resource(cur_rspec);
720 /* Allocate the resource specification and initialize it
721 * with values from the current stream entry. */
722 cur_rspec = MALLOC(sizeof(*cur_rspec));
723 if (cur_rspec == NULL) {
724 ERROR("Not enough memory to read lookup table!");
725 ret = WIMLIB_ERR_NOMEM;
728 wim_res_hdr_to_spec(&reshdr, wim, cur_rspec);
730 /* If this is a packed run, the current stream entry may
731 * specify a stream within the resource, and not the
732 * resource itself. Zero possibly irrelevant data until
733 * it is read for certain. (Note that the computation
734 * of 'back_to_back_pack' tests if 'size_in_wim' is
735 * nonzero to see if the resource info has been read;
736 * hence we need to set it to 0 here.) */
737 if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
738 cur_rspec->size_in_wim = 0;
739 cur_rspec->uncompressed_size = 0;
740 cur_rspec->offset_in_wim = 0;
744 lte_bind_wim_resource_spec(prev_entry, cur_rspec);
747 if ((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
748 reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER)
750 /* Found the specification for the packed resource.
751 * Transfer the values to the `struct
752 * wim_resource_spec', and discard the current stream
753 * since this lookup table entry did not, in fact,
754 * correspond to a "stream".
757 /* Uncompressed size of the resource pack is actually
758 * stored in the header of the resource itself. Read
759 * it, and also grab the chunk size and compression type
760 * (which are not necessarily the defaults from the WIM
762 struct alt_chunk_table_header_disk hdr;
764 ret = full_pread(&wim->in_fd, &hdr,
765 sizeof(hdr), reshdr.offset_in_wim);
769 cur_rspec->uncompressed_size = le64_to_cpu(hdr.res_usize);
770 cur_rspec->offset_in_wim = reshdr.offset_in_wim;
771 cur_rspec->size_in_wim = reshdr.size_in_wim;
772 cur_rspec->flags = reshdr.flags;
774 /* Compression format numbers must be the same as in
775 * WIMGAPI to be compatible here. */
776 BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
777 BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZX != 1);
778 BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_XPRESS != 2);
779 BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_LZMS != 3);
780 cur_rspec->compression_type = le32_to_cpu(hdr.compression_format);
782 cur_rspec->chunk_size = le32_to_cpu(hdr.chunk_size);
784 DEBUG("Full pack is %"PRIu64" compressed bytes "
785 "at file offset %"PRIu64" (flags 0x%02x)",
786 cur_rspec->size_in_wim,
787 cur_rspec->offset_in_wim,
789 free_lookup_table_entry(cur_entry);
793 if (is_zero_hash(cur_entry->hash)) {
794 free_lookup_table_entry(cur_entry);
798 if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
799 /* Continuing the pack with another stream. */
800 DEBUG("Continuing pack with stream: "
801 "%"PRIu64" uncompressed bytes @ "
802 "resource offset %"PRIu64")",
803 reshdr.size_in_wim, reshdr.offset_in_wim);
806 lte_bind_wim_resource_spec(cur_entry, cur_rspec);
807 if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
808 /* In packed runs, the offset field is used for
809 * in-resource offset, not the in-WIM offset, and the
810 * size field is used for the uncompressed size, not the
811 * compressed size. */
812 cur_entry->offset_in_res = reshdr.offset_in_wim;
813 cur_entry->size = reshdr.size_in_wim;
814 cur_entry->flags = reshdr.flags;
816 /* Normal case: The stream corresponds one-to-one with
817 * the resource entry. */
818 cur_entry->offset_in_res = 0;
819 cur_entry->size = reshdr.uncompressed_size;
820 cur_entry->flags = reshdr.flags;
824 if (cur_entry->flags & WIM_RESHDR_FLAG_METADATA) {
825 /* Lookup table entry for a metadata resource */
827 /* Metadata entries with no references must be ignored;
828 * see for example the WinPE WIMs from the WAIK v2.1.
830 if (cur_entry->refcnt == 0) {
831 free_lookup_table_entry(cur_entry);
835 if (cur_entry->refcnt != 1) {
836 ERROR("Found metadata resource with refcnt != 1");
837 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
841 if (wim->hdr.part_number != 1) {
842 WARNING("Ignoring metadata resource found in a "
843 "non-first part of the split WIM");
844 free_lookup_table_entry(cur_entry);
847 if (wim->current_image == wim->hdr.image_count) {
848 WARNING("The WIM header says there are %u images "
849 "in the WIM, but we found more metadata "
850 "resources than this (ignoring the extra)",
851 wim->hdr.image_count);
852 free_lookup_table_entry(cur_entry);
856 /* Notice very carefully: We are assigning the metadata
857 * resources in the exact order mirrored by their lookup
858 * table entries on disk, which is the behavior of
859 * Microsoft's software. In particular, this overrides
860 * the actual locations of the metadata resources
861 * themselves in the WIM file as well as any information
862 * written in the XML data. */
863 DEBUG("Found metadata resource for image %u at "
865 wim->current_image + 1,
866 cur_entry->rspec->offset_in_wim);
868 wim->current_image++]->metadata_lte = cur_entry;
872 /* Lookup table entry for a stream that is not a metadata
874 duplicate_entry = lookup_stream(table, cur_entry->hash);
875 if (duplicate_entry) {
876 WARNING("The WIM lookup table contains two entries "
877 "with the same SHA1 message digest!");
878 free_lookup_table_entry(cur_entry);
882 /* Finally, insert the stream into the lookup table, keyed by
883 * its SHA1 message digest. */
884 lookup_table_insert(table, cur_entry);
888 /* Validate the last resource. */
889 if (cur_rspec != NULL) {
890 ret = validate_resource(cur_rspec);
895 if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
896 WARNING("The header of \"%"TS"\" says there are %u images in\n"
897 " the WIM, but we only found %d metadata resources! Acting as if\n"
898 " the header specified only %d images instead.",
899 wim->filename, wim->hdr.image_count,
900 wim->current_image, wim->current_image);
901 for (int i = wim->current_image; i < wim->hdr.image_count; i++)
902 put_image_metadata(wim->image_metadata[i], NULL);
903 wim->hdr.image_count = wim->current_image;
905 DEBUG("Done reading lookup table.");
906 wim->lookup_table = table;
911 if (cur_rspec && list_empty(&cur_rspec->stream_list))
913 free_lookup_table_entry(cur_entry);
914 free_lookup_table(table);
918 wim->current_image = 0;
923 put_wim_lookup_table_entry(struct wim_lookup_table_entry_disk *disk_entry,
924 const struct wim_reshdr *out_reshdr,
925 u16 part_number, u32 refcnt, const u8 *hash)
927 put_wim_reshdr(out_reshdr, &disk_entry->reshdr);
928 disk_entry->part_number = cpu_to_le16(part_number);
929 disk_entry->refcnt = cpu_to_le32(refcnt);
930 copy_hash(disk_entry->hash, hash);
934 write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
935 struct filedes *out_fd,
937 struct wim_reshdr *out_reshdr,
938 int write_resource_flags)
941 struct wim_lookup_table_entry *lte;
942 struct wim_lookup_table_entry_disk *table_buf;
943 struct wim_lookup_table_entry_disk *table_buf_ptr;
945 u64 prev_res_offset_in_wim = ~0ULL;
948 list_for_each_entry(lte, stream_list, lookup_table_list) {
949 table_size += sizeof(struct wim_lookup_table_entry_disk);
951 if (lte->out_reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
952 lte->out_res_offset_in_wim != prev_res_offset_in_wim)
954 table_size += sizeof(struct wim_lookup_table_entry_disk);
955 prev_res_offset_in_wim = lte->out_res_offset_in_wim;
959 DEBUG("Writing WIM lookup table (size=%zu, offset=%"PRIu64")",
960 table_size, out_fd->offset);
962 table_buf = MALLOC(table_size);
963 if (table_buf == NULL) {
964 ERROR("Failed to allocate %zu bytes for temporary lookup table",
966 return WIMLIB_ERR_NOMEM;
968 table_buf_ptr = table_buf;
970 prev_res_offset_in_wim = ~0ULL;
971 list_for_each_entry(lte, stream_list, lookup_table_list) {
973 put_wim_lookup_table_entry(table_buf_ptr++,
978 if (lte->out_reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
979 lte->out_res_offset_in_wim != prev_res_offset_in_wim)
981 /* Put the main resource entry for the pack. */
983 struct wim_reshdr reshdr;
985 reshdr.offset_in_wim = lte->out_res_offset_in_wim;
986 reshdr.size_in_wim = lte->out_res_size_in_wim;
987 reshdr.uncompressed_size = WIM_PACK_MAGIC_NUMBER;
988 reshdr.flags = WIM_RESHDR_FLAG_PACKED_STREAMS;
990 DEBUG("Putting main entry for pack: "
991 "size_in_wim=%"PRIu64", "
992 "offset_in_wim=%"PRIu64", "
993 "uncompressed_size=%"PRIu64,
995 reshdr.offset_in_wim,
996 reshdr.uncompressed_size);
998 put_wim_lookup_table_entry(table_buf_ptr++,
1002 prev_res_offset_in_wim = lte->out_res_offset_in_wim;
1006 wimlib_assert((u8*)table_buf_ptr - (u8*)table_buf == table_size);
1008 /* Write the lookup table uncompressed. Although wimlib can handle a
1009 * compressed lookup table, MS software cannot. */
1010 ret = write_wim_resource_from_buffer(table_buf,
1012 WIM_RESHDR_FLAG_METADATA,
1014 WIMLIB_COMPRESSION_TYPE_NONE,
1018 write_resource_flags);
1020 DEBUG("ret=%d", ret);
1025 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
1027 lte->real_refcnt = 0;
1032 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
1034 lte->out_refcnt = 0;
1039 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
1041 if (lte->extracted_file != NULL) {
1042 FREE(lte->extracted_file);
1043 lte->extracted_file = NULL;
1048 /* Allocate a stream entry for the contents of the buffer, or re-use an existing
1049 * entry in @lookup_table for the same stream. */
1050 struct wim_lookup_table_entry *
1051 new_stream_from_data_buffer(const void *buffer, size_t size,
1052 struct wim_lookup_table *lookup_table)
1054 u8 hash[SHA1_HASH_SIZE];
1055 struct wim_lookup_table_entry *lte, *existing_lte;
1057 sha1_buffer(buffer, size, hash);
1058 existing_lte = lookup_stream(lookup_table, hash);
1060 wimlib_assert(existing_lte->size == size);
1065 lte = new_lookup_table_entry();
1068 buffer_copy = memdup(buffer, size);
1069 if (buffer_copy == NULL) {
1070 free_lookup_table_entry(lte);
1073 lte->resource_location = RESOURCE_IN_ATTACHED_BUFFER;
1074 lte->attached_buffer = buffer_copy;
1076 copy_hash(lte->hash, hash);
1077 lookup_table_insert(lookup_table, lte);
1082 /* Calculate the SHA1 message digest of a stream and move it from the list of
1083 * unhashed streams to the stream lookup table, possibly joining it with an
1084 * existing lookup table entry for an identical stream.
1086 * @lte: An unhashed lookup table entry.
1087 * @lookup_table: Lookup table for the WIM.
1088 * @lte_ret: On success, write a pointer to the resulting lookup table
1089 * entry to this location. This will be the same as @lte
1090 * if it was inserted into the lookup table, or different if
1091 * a duplicate stream was found.
1093 * Returns 0 on success; nonzero if there is an error reading the stream.
1096 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1097 struct wim_lookup_table *lookup_table,
1098 struct wim_lookup_table_entry **lte_ret)
1101 struct wim_lookup_table_entry *duplicate_lte;
1102 struct wim_lookup_table_entry **back_ptr;
1104 wimlib_assert(lte->unhashed);
1106 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1107 * union with the SHA1 message digest and will no longer be valid once
1108 * the SHA1 has been calculated. */
1109 back_ptr = retrieve_lte_pointer(lte);
1111 ret = sha1_stream(lte);
1115 /* Look for a duplicate stream */
1116 duplicate_lte = lookup_stream(lookup_table, lte->hash);
1117 list_del(<e->unhashed_list);
1118 if (duplicate_lte) {
1119 /* We have a duplicate stream. Transfer the reference counts
1120 * from this stream to the duplicate and update the reference to
1121 * this stream (in an inode or ads_entry) to point to the
1122 * duplicate. The caller is responsible for freeing @lte if
1124 wimlib_assert(!(duplicate_lte->unhashed));
1125 wimlib_assert(duplicate_lte->size == lte->size);
1126 duplicate_lte->refcnt += lte->refcnt;
1128 *back_ptr = duplicate_lte;
1129 lte = duplicate_lte;
1131 /* No duplicate stream, so we need to insert this stream into
1132 * the lookup table and treat it as a hashed stream. */
1133 lookup_table_insert(lookup_table, lte);
1141 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
1142 struct wimlib_resource_entry *wentry)
1144 memset(wentry, 0, sizeof(*wentry));
1146 wentry->uncompressed_size = lte->size;
1147 if (lte->resource_location == RESOURCE_IN_WIM) {
1148 wentry->part_number = lte->rspec->wim->hdr.part_number;
1149 if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
1150 wentry->compressed_size = 0;
1151 wentry->offset = lte->offset_in_res;
1153 wentry->compressed_size = lte->rspec->size_in_wim;
1154 wentry->offset = lte->rspec->offset_in_wim;
1156 wentry->raw_resource_offset_in_wim = lte->rspec->offset_in_wim;
1157 /*wentry->raw_resource_uncompressed_size = lte->rspec->uncompressed_size;*/
1158 wentry->raw_resource_compressed_size = lte->rspec->size_in_wim;
1160 copy_hash(wentry->sha1_hash, lte->hash);
1161 wentry->reference_count = lte->refcnt;
1162 wentry->is_compressed = (lte->flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
1163 wentry->is_metadata = (lte->flags & WIM_RESHDR_FLAG_METADATA) != 0;
1164 wentry->is_free = (lte->flags & WIM_RESHDR_FLAG_FREE) != 0;
1165 wentry->is_spanned = (lte->flags & WIM_RESHDR_FLAG_SPANNED) != 0;
1166 wentry->packed = (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) != 0;
1169 struct iterate_lte_context {
1170 wimlib_iterate_lookup_table_callback_t cb;
1175 do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
1177 struct iterate_lte_context *ctx = _ctx;
1178 struct wimlib_resource_entry entry;
1180 lte_to_wimlib_resource_entry(lte, &entry);
1181 return (*ctx->cb)(&entry, ctx->user_ctx);
1184 /* API function documented in wimlib.h */
1186 wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
1187 wimlib_iterate_lookup_table_callback_t cb,
1191 return WIMLIB_ERR_INVALID_PARAM;
1193 struct iterate_lte_context ctx = {
1195 .user_ctx = user_ctx,
1197 if (wim->hdr.part_number == 1) {
1199 for (int i = 0; i < wim->hdr.image_count; i++) {
1200 ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
1206 return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);