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/endianness.h"
32 #include "wimlib/error.h"
33 #include "wimlib/file_io.h"
34 #include "wimlib/glob.h"
35 #include "wimlib/lookup_table.h"
36 #include "wimlib/metadata.h"
37 #include "wimlib/paths.h"
38 #include "wimlib/resource.h"
39 #include "wimlib/util.h"
40 #include "wimlib/write.h"
45 # include <unistd.h> /* for unlink() */
48 struct wim_lookup_table *
49 new_lookup_table(size_t capacity)
51 struct wim_lookup_table *table;
52 struct hlist_head *array;
54 table = CALLOC(1, sizeof(struct wim_lookup_table));
56 array = CALLOC(capacity, sizeof(array[0]));
58 table->num_entries = 0;
59 table->capacity = capacity;
64 ERROR("Failed to allocate memory for lookup table "
65 "with capacity %zu", capacity);
71 struct wim_lookup_table_entry *
72 new_lookup_table_entry(void)
74 struct wim_lookup_table_entry *lte;
76 lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
78 ERROR("Out of memory (tried to allocate %zu bytes for "
79 "lookup table entry)",
80 sizeof(struct wim_lookup_table_entry));
84 BUILD_BUG_ON(RESOURCE_NONEXISTENT != 0);
88 struct wim_lookup_table_entry *
89 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
91 struct wim_lookup_table_entry *new;
93 new = memdup(old, sizeof(struct wim_lookup_table_entry));
97 new->extracted_file = NULL;
98 switch (new->resource_location) {
100 list_add(&new->rspec_node, &new->rspec->stream_list);
103 case RESOURCE_IN_FILE_ON_DISK:
105 case RESOURCE_WIN32_ENCRYPTED:
108 case RESOURCE_IN_STAGING_FILE:
109 BUILD_BUG_ON((void*)&old->file_on_disk !=
110 (void*)&old->staging_file_name);
112 new->file_on_disk = TSTRDUP(old->file_on_disk);
113 if (new->file_on_disk == NULL)
116 case RESOURCE_IN_ATTACHED_BUFFER:
117 new->attached_buffer = memdup(old->attached_buffer, old->size);
118 if (new->attached_buffer == NULL)
122 case RESOURCE_IN_NTFS_VOLUME:
124 struct ntfs_location *loc;
125 loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
129 loc->stream_name = NULL;
131 loc->path = STRDUP(old->ntfs_loc->path);
132 if (loc->path == NULL)
134 if (loc->stream_name_nchars != 0) {
135 loc->stream_name = memdup(old->ntfs_loc->stream_name,
136 loc->stream_name_nchars * 2);
137 if (loc->stream_name == NULL)
148 free_lookup_table_entry(new);
153 free_lookup_table_entry(struct wim_lookup_table_entry *lte)
156 switch (lte->resource_location) {
157 case RESOURCE_IN_WIM:
158 list_del(<e->rspec_node);
159 if (list_empty(<e->rspec->stream_list))
162 case RESOURCE_IN_FILE_ON_DISK:
164 case RESOURCE_WIN32_ENCRYPTED:
167 case RESOURCE_IN_STAGING_FILE:
168 BUILD_BUG_ON((void*)<e->file_on_disk !=
169 (void*)<e->staging_file_name);
171 case RESOURCE_IN_ATTACHED_BUFFER:
172 BUILD_BUG_ON((void*)<e->file_on_disk !=
173 (void*)<e->attached_buffer);
174 FREE(lte->file_on_disk);
177 case RESOURCE_IN_NTFS_VOLUME:
179 FREE(lte->ntfs_loc->path);
180 FREE(lte->ntfs_loc->stream_name);
193 do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
195 free_lookup_table_entry(entry);
201 free_lookup_table(struct wim_lookup_table *table)
203 DEBUG("Freeing lookup table.");
206 for_lookup_table_entry(table,
207 do_free_lookup_table_entry,
216 * Inserts an entry into the lookup table.
218 * @table: A pointer to the lookup table.
219 * @lte: A pointer to the entry to insert.
222 lookup_table_insert(struct wim_lookup_table *table,
223 struct wim_lookup_table_entry *lte)
225 size_t i = lte->hash_short % table->capacity;
226 hlist_add_head(<e->hash_list, &table->array[i]);
228 /* XXX Make the table grow when too many entries have been inserted. */
229 table->num_entries++;
233 finalize_lte(struct wim_lookup_table_entry *lte)
236 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
237 unlink(lte->staging_file_name);
238 list_del(<e->unhashed_list);
241 free_lookup_table_entry(lte);
244 /* Decrements the reference count for the lookup table entry @lte. If its
245 * reference count reaches 0, it is unlinked from the lookup table. If,
246 * furthermore, the entry has no opened file descriptors associated with it, the
249 lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
250 struct wim_lookup_table *table)
252 wimlib_assert(lte != NULL);
253 wimlib_assert(lte->refcnt != 0);
254 if (--lte->refcnt == 0) {
256 list_del(<e->unhashed_list);
258 lookup_table_unlink(table, lte);
260 if (lte->num_opened_fds == 0)
268 lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
270 if (lte->num_opened_fds != 0)
271 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
276 /* Calls a function on all the entries in the WIM lookup table. Stop early and
277 * return nonzero if any call to the function returns nonzero. */
279 for_lookup_table_entry(struct wim_lookup_table *table,
280 int (*visitor)(struct wim_lookup_table_entry *, void *),
283 struct wim_lookup_table_entry *lte;
284 struct hlist_node *pos, *tmp;
287 for (size_t i = 0; i < table->capacity; i++) {
288 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
291 wimlib_assert2(!(lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA));
292 ret = visitor(lte, arg);
300 /* qsort() callback that sorts streams (represented by `struct
301 * wim_lookup_table_entry's) into an order optimized for reading.
303 * Sorting is done primarily by resource location, then secondarily by a
304 * per-resource location order. For example, resources in WIM files are sorted
305 * primarily by part number, then secondarily by offset, as to implement optimal
306 * reading of either a standalone or split WIM. */
308 cmp_streams_by_sequential_order(const void *p1, const void *p2)
310 const struct wim_lookup_table_entry *lte1, *lte2;
312 WIMStruct *wim1, *wim2;
314 lte1 = *(const struct wim_lookup_table_entry**)p1;
315 lte2 = *(const struct wim_lookup_table_entry**)p2;
317 v = (int)lte1->resource_location - (int)lte2->resource_location;
319 /* Different resource locations? */
323 switch (lte1->resource_location) {
324 case RESOURCE_IN_WIM:
325 wim1 = lte1->rspec->wim;
326 wim2 = lte2->rspec->wim;
328 /* Different (possibly split) WIMs? */
330 v = memcmp(wim1->hdr.guid, wim2->hdr.guid, WIM_GID_LEN);
335 /* Different part numbers in the same WIM? */
336 v = (int)wim1->hdr.part_number - (int)wim2->hdr.part_number;
340 if (lte1->rspec->offset_in_wim != lte2->rspec->offset_in_wim)
341 return cmp_u64(lte1->rspec->offset_in_wim,
342 lte2->rspec->offset_in_wim);
344 return cmp_u64(lte1->offset_in_res, lte2->offset_in_res);
346 case RESOURCE_IN_FILE_ON_DISK:
348 case RESOURCE_IN_STAGING_FILE:
351 case RESOURCE_WIN32_ENCRYPTED:
353 /* Compare files by path: just a heuristic that will place files
354 * in the same directory next to each other. */
355 return tstrcmp(lte1->file_on_disk, lte2->file_on_disk);
357 case RESOURCE_IN_NTFS_VOLUME:
358 return tstrcmp(lte1->ntfs_loc->path, lte2->ntfs_loc->path);
361 /* No additional sorting order defined for this resource
362 * location (e.g. RESOURCE_IN_ATTACHED_BUFFER); simply compare
363 * everything equal to each other. */
369 sort_stream_list(struct list_head *stream_list,
370 size_t list_head_offset,
371 int (*compar)(const void *, const void*))
373 struct list_head *cur;
374 struct wim_lookup_table_entry **array;
377 size_t num_streams = 0;
379 list_for_each(cur, stream_list)
382 if (num_streams <= 1)
385 array_size = num_streams * sizeof(array[0]);
386 array = MALLOC(array_size);
388 return WIMLIB_ERR_NOMEM;
390 cur = stream_list->next;
391 for (i = 0; i < num_streams; i++) {
392 array[i] = (struct wim_lookup_table_entry*)((u8*)cur -
397 qsort(array, num_streams, sizeof(array[0]), compar);
399 INIT_LIST_HEAD(stream_list);
400 for (i = 0; i < num_streams; i++) {
401 list_add_tail((struct list_head*)
402 ((u8*)array[i] + list_head_offset),
409 /* Sort the specified list of streams in an order optimized for reading. */
411 sort_stream_list_by_sequential_order(struct list_head *stream_list,
412 size_t list_head_offset)
414 return sort_stream_list(stream_list, list_head_offset,
415 cmp_streams_by_sequential_order);
420 add_lte_to_array(struct wim_lookup_table_entry *lte,
423 struct wim_lookup_table_entry ***pp = _pp;
428 /* Iterate through the lookup table entries, but first sort them by stream
429 * offset in the WIM. Caution: this is intended to be used when the stream
430 * offset field has actually been set. */
432 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
433 int (*visitor)(struct wim_lookup_table_entry *,
437 struct wim_lookup_table_entry **lte_array, **p;
438 size_t num_streams = table->num_entries;
441 lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
443 return WIMLIB_ERR_NOMEM;
445 for_lookup_table_entry(table, add_lte_to_array, &p);
447 wimlib_assert(p == lte_array + num_streams);
449 qsort(lte_array, num_streams, sizeof(lte_array[0]),
450 cmp_streams_by_sequential_order);
452 for (size_t i = 0; i < num_streams; i++) {
453 ret = visitor(lte_array[i], arg);
461 /* On-disk format of a WIM lookup table entry (stream entry). */
462 struct wim_lookup_table_entry_disk {
463 /* Size, offset, and flags of the stream. */
464 struct wim_reshdr_disk reshdr;
466 /* Which part of the split WIM this stream is in; indexed from 1. */
469 /* Reference count of this stream over all WIM images. */
472 /* SHA1 message digest of the uncompressed data of this stream, or
473 * optionally all zeroes if this stream is of zero length. */
474 u8 hash[SHA1_HASH_SIZE];
477 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
479 /* Validate the size and location of a WIM resource. */
481 validate_resource(const struct wim_resource_spec *rspec)
483 struct wim_lookup_table_entry *lte;
486 /* Verify that calculating the offset of the end of the resource doesn't
488 if (rspec->offset_in_wim + rspec->size_in_wim < rspec->size_in_wim)
491 /* Verify that each stream in the resource has a valid offset and size,
492 * and that no streams overlap, and that the streams were added in order
493 * of increasing offset. */
495 list_for_each_entry(lte, &rspec->stream_list, rspec_node) {
496 if (lte->offset_in_res + lte->size < lte->size ||
497 lte->offset_in_res + lte->size > rspec->uncompressed_size ||
498 lte->offset_in_res < cur_offset)
501 cur_offset = lte->offset_in_res + lte->size;
507 ERROR("Invalid resource entry!");
508 return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
512 * Reads the lookup table from a WIM file. Each entry specifies a stream that
513 * the WIM file contains, along with its location and SHA1 message digest.
515 * Saves lookup table entries for non-metadata streams in a hash table, and
516 * saves the metadata entry for each image in a special per-image location (the
517 * image_metadata array).
520 * WIMLIB_ERR_SUCCESS (0)
521 * WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
522 * WIMLIB_ERR_RESOURCE_NOT_FOUND
524 * Or an error code caused by failure to read the lookup table into memory.
527 read_wim_lookup_table(WIMStruct *wim)
532 struct wim_lookup_table *table;
533 struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
534 struct wim_resource_spec *cur_rspec;
536 bool back_to_back_pack;
538 DEBUG("Reading lookup table.");
540 /* Sanity check: lookup table entries are 50 bytes each. */
541 BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
542 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
544 /* Calculate number of entries in the lookup table. */
545 num_entries = wim->hdr.lookup_table_reshdr.uncompressed_size /
546 sizeof(struct wim_lookup_table_entry_disk);
548 /* Read the lookup table into a buffer. */
549 ret = wim_reshdr_to_data(&wim->hdr.lookup_table_reshdr, wim, &buf);
553 /* Allocate a hash table to map SHA1 message digests into stream
554 * specifications. This is the in-memory "lookup table". */
555 table = new_lookup_table(num_entries * 2 + 1);
557 ERROR("Not enough memory to read lookup table.");
558 ret = WIMLIB_ERR_NOMEM;
562 /* Allocate and initalize stream entries from the raw lookup table
564 wim->current_image = 0;
566 for (i = 0; i < num_entries; i++) {
567 const struct wim_lookup_table_entry_disk *disk_entry =
568 &((const struct wim_lookup_table_entry_disk*)buf)[i];
570 struct wim_reshdr reshdr;
572 get_wim_reshdr(&disk_entry->reshdr, &reshdr);
574 DEBUG("reshdr: size_in_wim=%"PRIu64", "
575 "uncompressed_size=%"PRIu64", "
576 "offset_in_wim=%"PRIu64", "
578 reshdr.size_in_wim, reshdr.uncompressed_size,
579 reshdr.offset_in_wim, reshdr.flags);
581 if (wim->hdr.wim_version == WIM_VERSION_DEFAULT)
582 reshdr.flags &= ~WIM_RESHDR_FLAG_PACKED_STREAMS;
584 cur_entry = new_lookup_table_entry();
585 if (cur_entry == NULL) {
586 ERROR("Not enough memory to read lookup table!");
587 ret = WIMLIB_ERR_NOMEM;
588 goto out_free_lookup_table;
591 part_number = le16_to_cpu(disk_entry->part_number);
592 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
593 copy_hash(cur_entry->hash, disk_entry->hash);
595 if (part_number != wim->hdr.part_number) {
596 WARNING("A lookup table entry in part %hu of the WIM "
597 "points to part %hu (ignoring it)",
598 wim->hdr.part_number, part_number);
599 free_lookup_table_entry(cur_entry);
603 if (!(reshdr.flags & (WIM_RESHDR_FLAG_PACKED_STREAMS |
604 WIM_RESHDR_FLAG_COMPRESSED))) {
605 if (reshdr.uncompressed_size != reshdr.size_in_wim) {
606 ERROR("Invalid resource entry!");
607 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
608 goto out_free_cur_entry;
612 back_to_back_pack = false;
613 if (!(reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) ||
616 ((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
617 reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER &&
619 cur_rspec->size_in_wim != 0)))
621 /* Starting new run of streams that share the same WIM
623 struct wim_lookup_table_entry *prev_entry = NULL;
625 if (back_to_back_pack) {
626 prev_entry = list_entry(cur_rspec->stream_list.prev,
627 struct wim_lookup_table_entry,
629 lte_unbind_wim_resource_spec(prev_entry);
630 cur_rspec->uncompressed_size -= prev_entry->size;
632 if (cur_rspec != NULL) {
633 ret = validate_resource(cur_rspec);
635 goto out_free_cur_entry;
638 /* Allocate the resource specification and initialize it
639 * with values from the current stream entry. */
640 cur_rspec = MALLOC(sizeof(*cur_rspec));
641 if (cur_rspec == NULL) {
642 ERROR("Not enough memory to read lookup table!");
643 ret = WIMLIB_ERR_NOMEM;
644 goto out_free_cur_entry;
646 wim_res_hdr_to_spec(&reshdr, wim, cur_rspec);
648 /* If this is a packed run, the current stream entry may
649 * specify a stream within the resource, and not the
650 * resource itself. Zero possibly irrelevant data until
651 * it is read for certain. */
652 if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
653 cur_rspec->size_in_wim = 0;
654 cur_rspec->uncompressed_size = 0;
655 cur_rspec->offset_in_wim = 0;
659 lte_bind_wim_resource_spec(prev_entry, cur_rspec);
660 cur_rspec->uncompressed_size = prev_entry->size;
664 if ((reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) &&
665 reshdr.uncompressed_size == WIM_PACK_MAGIC_NUMBER)
667 /* Found the specification for the packed resource.
668 * Transfer the values to the `struct
669 * wim_resource_spec', and discard the current stream
670 * since this lookup table entry did not, in fact,
671 * correspond to a "stream". */
673 cur_rspec->offset_in_wim = reshdr.offset_in_wim;
674 cur_rspec->size_in_wim = reshdr.size_in_wim;
675 cur_rspec->flags = reshdr.flags;
676 DEBUG("Full pack is %"PRIu64" compressed bytes "
677 "at file offset %"PRIu64" (flags 0x%02x)",
678 cur_rspec->size_in_wim,
679 cur_rspec->offset_in_wim,
681 free_lookup_table_entry(cur_entry);
685 if (is_zero_hash(cur_entry->hash)) {
686 free_lookup_table_entry(cur_entry);
690 if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
691 /* Continuing the pack with another stream. */
692 DEBUG("Continuing packed run with stream: "
693 "%"PRIu64" uncompressed bytes @ resource offset %"PRIu64")",
694 reshdr.size_in_wim, reshdr.offset_in_wim);
695 cur_rspec->uncompressed_size += reshdr.size_in_wim;
698 lte_bind_wim_resource_spec(cur_entry, cur_rspec);
699 if (reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
700 /* In packed runs, the offset field is used for
701 * in-resource offset, not the in-WIM offset, and the
702 * size field is used for the uncompressed size, not the
703 * compressed size. */
704 cur_entry->offset_in_res = reshdr.offset_in_wim;
705 cur_entry->size = reshdr.size_in_wim;
706 cur_entry->flags = reshdr.flags;
708 /* Normal case: The stream corresponds one-to-one with
709 * the resource entry. */
710 cur_entry->offset_in_res = 0;
711 cur_entry->size = reshdr.uncompressed_size;
712 cur_entry->flags = reshdr.flags;
716 if (cur_entry->flags & WIM_RESHDR_FLAG_METADATA) {
717 /* Lookup table entry for a metadata resource */
719 /* Metadata entries with no references must be ignored;
720 * see for example the WinPE WIMs from the WAIK v2.1.
722 if (cur_entry->refcnt == 0) {
723 free_lookup_table_entry(cur_entry);
727 if (cur_entry->refcnt != 1) {
728 if (wimlib_print_errors) {
729 ERROR("Found metadata resource with refcnt != 1:");
730 print_lookup_table_entry(cur_entry, stderr);
732 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
733 goto out_free_cur_entry;
736 if (wim->hdr.part_number != 1) {
737 WARNING("Ignoring metadata resource found in a "
738 "non-first part of the split WIM");
739 free_lookup_table_entry(cur_entry);
742 if (wim->current_image == wim->hdr.image_count) {
743 WARNING("The WIM header says there are %u images "
744 "in the WIM, but we found more metadata "
745 "resources than this (ignoring the extra)",
746 wim->hdr.image_count);
747 free_lookup_table_entry(cur_entry);
751 /* Notice very carefully: We are assigning the metadata
752 * resources in the exact order mirrored by their lookup
753 * table entries on disk, which is the behavior of
754 * Microsoft's software. In particular, this overrides
755 * the actual locations of the metadata resources
756 * themselves in the WIM file as well as any information
757 * written in the XML data. */
758 DEBUG("Found metadata resource for image %u at "
760 wim->current_image + 1,
761 cur_entry->rspec->offset_in_wim);
763 wim->current_image++]->metadata_lte = cur_entry;
767 /* Lookup table entry for a stream that is not a metadata
769 duplicate_entry = lookup_resource(table, cur_entry->hash);
770 if (duplicate_entry) {
771 if (wimlib_print_errors) {
772 WARNING("The WIM lookup table contains two entries with the "
773 "same SHA1 message digest!");
774 WARNING("The first entry is:");
775 print_lookup_table_entry(duplicate_entry, stderr);
776 WARNING("The second entry is:");
777 print_lookup_table_entry(cur_entry, stderr);
779 free_lookup_table_entry(cur_entry);
783 /* Finally, insert the stream into the lookup table, keyed by
784 * its SHA1 message digest. */
785 lookup_table_insert(table, cur_entry);
788 /* Validate the last resource. */
789 if (cur_rspec != NULL) {
790 ret = validate_resource(cur_rspec);
792 goto out_free_lookup_table;
795 if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
796 WARNING("The header of \"%"TS"\" says there are %u images in\n"
797 " the WIM, but we only found %d metadata resources! Acting as if\n"
798 " the header specified only %d images instead.",
799 wim->filename, wim->hdr.image_count,
800 wim->current_image, wim->current_image);
801 for (int i = wim->current_image; i < wim->hdr.image_count; i++)
802 put_image_metadata(wim->image_metadata[i], NULL);
803 wim->hdr.image_count = wim->current_image;
805 DEBUG("Done reading lookup table.");
806 wim->lookup_table = table;
812 out_free_lookup_table:
813 free_lookup_table(table);
817 wim->current_image = 0;
822 put_wim_lookup_table_entry(struct wim_lookup_table_entry_disk *disk_entry,
823 const struct wim_reshdr *out_reshdr,
824 u16 part_number, u32 refcnt, const u8 *hash)
826 put_wim_reshdr(out_reshdr, &disk_entry->reshdr);
827 disk_entry->part_number = cpu_to_le16(part_number);
828 disk_entry->refcnt = cpu_to_le32(refcnt);
829 copy_hash(disk_entry->hash, hash);
833 write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
834 struct filedes *out_fd,
836 struct wim_reshdr *out_reshdr,
837 int write_resource_flags,
838 struct wimlib_lzx_context **comp_ctx)
841 struct wim_lookup_table_entry *lte;
842 struct wim_lookup_table_entry_disk *table_buf;
843 struct wim_lookup_table_entry_disk *table_buf_ptr;
845 u64 prev_res_offset_in_wim = ~0ULL;
848 list_for_each_entry(lte, stream_list, lookup_table_list) {
849 table_size += sizeof(struct wim_lookup_table_entry_disk);
851 if (lte->out_reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
852 lte->out_res_offset_in_wim != prev_res_offset_in_wim)
854 table_size += sizeof(struct wim_lookup_table_entry_disk);
855 prev_res_offset_in_wim = lte->out_res_offset_in_wim;
859 DEBUG("Writing WIM lookup table (size=%zu, offset=%"PRIu64")",
860 table_size, out_fd->offset);
862 table_buf = MALLOC(table_size);
863 if (table_buf == NULL) {
864 ERROR("Failed to allocate %zu bytes for temporary lookup table",
866 return WIMLIB_ERR_NOMEM;
868 table_buf_ptr = table_buf;
870 prev_res_offset_in_wim = ~0ULL;
871 list_for_each_entry(lte, stream_list, lookup_table_list) {
873 put_wim_lookup_table_entry(table_buf_ptr++,
878 if (lte->out_reshdr.flags & WIM_RESHDR_FLAG_PACKED_STREAMS &&
879 lte->out_res_offset_in_wim != prev_res_offset_in_wim)
881 /* Put the main resource entry for the pack. */
883 struct wim_reshdr reshdr;
885 reshdr.offset_in_wim = lte->out_res_offset_in_wim;
886 reshdr.size_in_wim = lte->out_res_size_in_wim;
887 reshdr.uncompressed_size = WIM_PACK_MAGIC_NUMBER;
888 reshdr.flags = WIM_RESHDR_FLAG_PACKED_STREAMS;
890 DEBUG("Putting main entry for pack: "
891 "size_in_wim=%"PRIu64", "
892 "offset_in_wim=%"PRIu64", "
893 "uncompressed_size=%"PRIu64,
895 reshdr.offset_in_wim,
896 reshdr.uncompressed_size);
898 put_wim_lookup_table_entry(table_buf_ptr++,
902 prev_res_offset_in_wim = lte->out_res_offset_in_wim;
906 wimlib_assert((u8*)table_buf_ptr - (u8*)table_buf == table_size);
908 /* Write the lookup table uncompressed. Although wimlib can handle a
909 * compressed lookup table, MS software cannot. */
910 ret = write_wim_resource_from_buffer(table_buf,
912 WIM_RESHDR_FLAG_METADATA,
914 WIMLIB_COMPRESSION_TYPE_NONE,
918 write_resource_flags,
921 DEBUG("ret=%d", ret);
926 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
928 lte->real_refcnt = 0;
933 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
940 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
942 if (lte->extracted_file != NULL) {
943 FREE(lte->extracted_file);
944 lte->extracted_file = NULL;
950 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
958 tprintf(T("Uncompressed size = %"PRIu64" bytes\n"),
960 if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
961 tprintf(T("Offset = %"PRIu64" bytes\n"),
964 tprintf(T("Raw uncompressed size = %"PRIu64" bytes\n"),
965 lte->rspec->uncompressed_size);
967 tprintf(T("Raw compressed size = %"PRIu64" bytes\n"),
968 lte->rspec->size_in_wim);
970 tprintf(T("Raw offset = %"PRIu64" bytes\n"),
971 lte->rspec->offset_in_wim);
972 } else if (lte->resource_location == RESOURCE_IN_WIM) {
973 tprintf(T("Compressed size = %"PRIu64" bytes\n"),
974 lte->rspec->size_in_wim);
976 tprintf(T("Offset = %"PRIu64" bytes\n"),
977 lte->rspec->offset_in_wim);
980 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
983 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
984 lte->back_inode, lte->back_stream_id);
986 tfprintf(out, T("Hash = 0x"));
987 print_hash(lte->hash, out);
991 tfprintf(out, T("Flags = "));
992 u8 flags = lte->flags;
993 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
994 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
995 if (flags & WIM_RESHDR_FLAG_FREE)
996 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
997 if (flags & WIM_RESHDR_FLAG_METADATA)
998 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
999 if (flags & WIM_RESHDR_FLAG_SPANNED)
1000 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
1001 if (flags & WIM_RESHDR_FLAG_PACKED_STREAMS)
1002 tfputs(T("WIM_RESHDR_FLAG_PACKED_STREAMS, "), out);
1003 tputc(T('\n'), out);
1004 switch (lte->resource_location) {
1005 case RESOURCE_IN_WIM:
1006 if (lte->rspec->wim->filename) {
1007 tfprintf(out, T("WIM file = `%"TS"'\n"),
1008 lte->rspec->wim->filename);
1012 case RESOURCE_WIN32_ENCRYPTED:
1014 case RESOURCE_IN_FILE_ON_DISK:
1015 tfprintf(out, T("File on Disk = `%"TS"'\n"),
1019 case RESOURCE_IN_STAGING_FILE:
1020 tfprintf(out, T("Staging File = `%"TS"'\n"),
1021 lte->staging_file_name);
1027 tputc(T('\n'), out);
1031 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
1032 struct wimlib_resource_entry *wentry)
1034 memset(wentry, 0, sizeof(*wentry));
1036 wentry->uncompressed_size = lte->size;
1037 if (lte->resource_location == RESOURCE_IN_WIM) {
1038 wentry->part_number = lte->rspec->wim->hdr.part_number;
1039 if (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) {
1040 wentry->compressed_size = 0;
1041 wentry->offset = lte->offset_in_res;
1043 wentry->compressed_size = lte->rspec->size_in_wim;
1044 wentry->offset = lte->rspec->offset_in_wim;
1046 wentry->raw_resource_offset_in_wim = lte->rspec->offset_in_wim;
1047 /*wentry->raw_resource_uncompressed_size = lte->rspec->uncompressed_size;*/
1048 wentry->raw_resource_compressed_size = lte->rspec->size_in_wim;
1050 copy_hash(wentry->sha1_hash, lte->hash);
1051 wentry->reference_count = lte->refcnt;
1052 wentry->is_compressed = (lte->flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
1053 wentry->is_metadata = (lte->flags & WIM_RESHDR_FLAG_METADATA) != 0;
1054 wentry->is_free = (lte->flags & WIM_RESHDR_FLAG_FREE) != 0;
1055 wentry->is_spanned = (lte->flags & WIM_RESHDR_FLAG_SPANNED) != 0;
1056 wentry->packed = (lte->flags & WIM_RESHDR_FLAG_PACKED_STREAMS) != 0;
1059 struct iterate_lte_context {
1060 wimlib_iterate_lookup_table_callback_t cb;
1065 do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
1067 struct iterate_lte_context *ctx = _ctx;
1068 struct wimlib_resource_entry entry;
1070 lte_to_wimlib_resource_entry(lte, &entry);
1071 return (*ctx->cb)(&entry, ctx->user_ctx);
1074 /* API function documented in wimlib.h */
1076 wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
1077 wimlib_iterate_lookup_table_callback_t cb,
1080 struct iterate_lte_context ctx = {
1082 .user_ctx = user_ctx,
1084 if (wim->hdr.part_number == 1) {
1086 for (int i = 0; i < wim->hdr.image_count; i++) {
1087 ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
1093 return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
1096 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
1097 * lookup table, or NULL if there is none. */
1098 struct wim_lookup_table_entry *
1099 lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
1102 struct wim_lookup_table_entry *lte;
1103 struct hlist_node *pos;
1105 wimlib_assert(table != NULL);
1106 wimlib_assert(hash != NULL);
1108 i = *(size_t*)hash % table->capacity;
1109 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
1110 if (hashes_equal(hash, lte->hash))
1117 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
1118 * given a path name.
1120 * This is only for pre-resolved inodes.
1123 wim_pathname_to_stream(WIMStruct *wim,
1126 struct wim_dentry **dentry_ret,
1127 struct wim_lookup_table_entry **lte_ret,
1128 u16 *stream_idx_ret)
1130 struct wim_dentry *dentry;
1131 struct wim_lookup_table_entry *lte;
1133 const tchar *stream_name = NULL;
1134 struct wim_inode *inode;
1137 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
1138 stream_name = path_stream_name(path);
1140 p = (tchar*)stream_name - 1;
1145 dentry = get_dentry(wim, path);
1151 inode = dentry->d_inode;
1153 if (!inode->i_resolved)
1154 if (inode_resolve_ltes(inode, wim->lookup_table, false))
1157 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
1158 && inode_is_directory(inode))
1162 struct wim_ads_entry *ads_entry;
1164 ads_entry = inode_get_ads_entry(inode, stream_name,
1167 stream_idx = ads_idx + 1;
1168 lte = ads_entry->lte;
1174 lte = inode_unnamed_stream_resolved(inode, &stream_idx);
1178 *dentry_ret = dentry;
1182 *stream_idx_ret = stream_idx;
1188 resource_not_found_error(const struct wim_inode *inode, const u8 *hash)
1190 if (wimlib_print_errors) {
1191 ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
1192 tfprintf(stderr, T(" SHA-1 message digest of missing resource:\n "));
1193 print_hash(hash, stderr);
1194 tputc(T('\n'), stderr);
1196 return WIMLIB_ERR_RESOURCE_NOT_FOUND;
1200 * Resolve an inode's lookup table entries.
1202 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
1203 * lookup table) with pointers directly to the lookup table entries.
1205 * If @force is %false:
1206 * If any needed SHA1 message digests are not found in the lookup table,
1207 * WIMLIB_ERR_RESOURCE_NOT_FOUND is returned and the inode is left
1209 * If @force is %true:
1210 * If any needed SHA1 message digests are not found in the lookup table,
1211 * new entries are allocated and inserted into the lookup table.
1214 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table,
1219 if (!inode->i_resolved) {
1220 struct wim_lookup_table_entry *lte, *ads_lte;
1222 /* Resolve the default file stream */
1224 hash = inode->i_hash;
1225 if (!is_zero_hash(hash)) {
1226 lte = lookup_resource(table, hash);
1229 lte = new_lookup_table_entry();
1231 return WIMLIB_ERR_NOMEM;
1232 copy_hash(lte->hash, hash);
1233 lookup_table_insert(table, lte);
1235 goto resource_not_found;
1240 /* Resolve the alternate data streams */
1241 struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
1242 for (u16 i = 0; i < inode->i_num_ads; i++) {
1243 struct wim_ads_entry *cur_entry;
1246 cur_entry = &inode->i_ads_entries[i];
1247 hash = cur_entry->hash;
1248 if (!is_zero_hash(hash)) {
1249 ads_lte = lookup_resource(table, hash);
1252 ads_lte = new_lookup_table_entry();
1254 return WIMLIB_ERR_NOMEM;
1255 copy_hash(ads_lte->hash, hash);
1256 lookup_table_insert(table, ads_lte);
1258 goto resource_not_found;
1262 ads_ltes[i] = ads_lte;
1265 for (u16 i = 0; i < inode->i_num_ads; i++)
1266 inode->i_ads_entries[i].lte = ads_ltes[i];
1267 inode->i_resolved = 1;
1272 return resource_not_found_error(inode, hash);
1276 inode_unresolve_ltes(struct wim_inode *inode)
1278 if (inode->i_resolved) {
1280 copy_hash(inode->i_hash, inode->i_lte->hash);
1282 zero_out_hash(inode->i_hash);
1284 for (u16 i = 0; i < inode->i_num_ads; i++) {
1285 if (inode->i_ads_entries[i].lte)
1286 copy_hash(inode->i_ads_entries[i].hash,
1287 inode->i_ads_entries[i].lte->hash);
1289 zero_out_hash(inode->i_ads_entries[i].hash);
1291 inode->i_resolved = 0;
1296 * Returns the lookup table entry for stream @stream_idx of the inode, where
1297 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
1298 * corresponds to an alternate data stream.
1300 * This works for both resolved and un-resolved inodes.
1302 struct wim_lookup_table_entry *
1303 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
1304 const struct wim_lookup_table *table)
1306 if (inode->i_resolved)
1307 return inode_stream_lte_resolved(inode, stream_idx);
1309 return inode_stream_lte_unresolved(inode, stream_idx, table);
1312 struct wim_lookup_table_entry *
1313 inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
1315 wimlib_assert(inode->i_resolved);
1316 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1317 if (inode_stream_name_nbytes(inode, i) == 0 &&
1318 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
1320 *stream_idx_ret = i;
1321 return inode_stream_lte_resolved(inode, i);
1324 *stream_idx_ret = 0;
1328 struct wim_lookup_table_entry *
1329 inode_unnamed_lte_resolved(const struct wim_inode *inode)
1332 return inode_unnamed_stream_resolved(inode, &stream_idx);
1335 struct wim_lookup_table_entry *
1336 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
1337 const struct wim_lookup_table *table)
1339 wimlib_assert(!inode->i_resolved);
1340 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1341 if (inode_stream_name_nbytes(inode, i) == 0 &&
1342 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
1344 return inode_stream_lte_unresolved(inode, i, table);
1350 /* Return the lookup table entry for the unnamed data stream of an inode, or
1351 * NULL if there is none.
1353 * You'd think this would be easier than it actually is, since the unnamed data
1354 * stream should be the one referenced from the inode itself. Alas, if there
1355 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
1356 * data stream in one of the alternate data streams instead of inside the WIM
1357 * dentry itself. So we need to check the alternate data streams too.
1359 * Also, note that a dentry may appear to have more than one unnamed stream, but
1360 * if the SHA1 message digest is all 0's then the corresponding stream does not
1361 * really "count" (this is the case for the inode's own file stream when the
1362 * file stream that should be there is actually in one of the alternate stream
1363 * entries.). This is despite the fact that we may need to extract such a
1364 * missing entry as an empty file or empty named data stream.
1366 struct wim_lookup_table_entry *
1367 inode_unnamed_lte(const struct wim_inode *inode,
1368 const struct wim_lookup_table *table)
1370 if (inode->i_resolved)
1371 return inode_unnamed_lte_resolved(inode);
1373 return inode_unnamed_lte_unresolved(inode, table);
1376 /* Returns the SHA1 message digest of the unnamed data stream of a WIM inode, or
1377 * 'zero_hash' if the unnamed data stream is missing has all zeroes in its SHA1
1378 * message digest field. */
1380 inode_unnamed_stream_hash(const struct wim_inode *inode)
1384 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1385 if (inode_stream_name_nbytes(inode, i) == 0) {
1386 hash = inode_stream_hash(inode, i);
1387 if (!is_zero_hash(hash))
1394 struct wim_lookup_table_entry **
1395 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1397 wimlib_assert(lte->unhashed);
1398 struct wim_inode *inode = lte->back_inode;
1399 u32 stream_id = lte->back_stream_id;
1401 return &inode->i_lte;
1403 for (u16 i = 0; i < inode->i_num_ads; i++)
1404 if (inode->i_ads_entries[i].stream_id == stream_id)
1405 return &inode->i_ads_entries[i].lte;
1410 /* Calculate the SHA1 message digest of a stream and move it from the list of
1411 * unhashed streams to the stream lookup table, possibly joining it with an
1412 * existing lookup table entry for an identical stream.
1414 * @lte: An unhashed lookup table entry.
1415 * @lookup_table: Lookup table for the WIM.
1416 * @lte_ret: On success, write a pointer to the resulting lookup table
1417 * entry to this location. This will be the same as @lte
1418 * if it was inserted into the lookup table, or different if
1419 * a duplicate stream was found.
1421 * Returns 0 on success; nonzero if there is an error reading the stream.
1424 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1425 struct wim_lookup_table *lookup_table,
1426 struct wim_lookup_table_entry **lte_ret)
1429 struct wim_lookup_table_entry *duplicate_lte;
1430 struct wim_lookup_table_entry **back_ptr;
1432 wimlib_assert(lte->unhashed);
1434 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1435 * union with the SHA1 message digest and will no longer be valid once
1436 * the SHA1 has been calculated. */
1437 back_ptr = retrieve_lte_pointer(lte);
1439 ret = sha1_stream(lte);
1443 /* Look for a duplicate stream */
1444 duplicate_lte = lookup_resource(lookup_table, lte->hash);
1445 list_del(<e->unhashed_list);
1446 if (duplicate_lte) {
1447 /* We have a duplicate stream. Transfer the reference counts
1448 * from this stream to the duplicate and update the reference to
1449 * this stream (in an inode or ads_entry) to point to the
1450 * duplicate. The caller is responsible for freeing @lte if
1452 wimlib_assert(!(duplicate_lte->unhashed));
1453 wimlib_assert(duplicate_lte->size == lte->size);
1454 duplicate_lte->refcnt += lte->refcnt;
1456 *back_ptr = duplicate_lte;
1457 lte = duplicate_lte;
1459 /* No duplicate stream, so we need to insert this stream into
1460 * the lookup table and treat it as a hashed stream. */
1461 lookup_table_insert(lookup_table, lte);
1469 lte_clone_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1471 struct wim_lookup_table *lookup_table = _lookup_table;
1473 if (lookup_resource(lookup_table, lte->hash))
1474 return 0; /* Resource already present. */
1476 lte = clone_lookup_table_entry(lte);
1478 return WIMLIB_ERR_NOMEM;
1479 lte->out_refcnt = 1;
1480 lookup_table_insert(lookup_table, lte);
1485 lte_delete_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1487 struct wim_lookup_table *lookup_table = _lookup_table;
1489 if (lte->out_refcnt) {
1490 lookup_table_unlink(lookup_table, lte);
1491 free_lookup_table_entry(lte);
1496 /* API function documented in wimlib.h */
1498 wimlib_reference_resources(WIMStruct *wim,
1499 WIMStruct **resource_wims, unsigned num_resource_wims,
1506 return WIMLIB_ERR_INVALID_PARAM;
1508 if (num_resource_wims != 0 && resource_wims == NULL)
1509 return WIMLIB_ERR_INVALID_PARAM;
1511 for (i = 0; i < num_resource_wims; i++)
1512 if (resource_wims[i] == NULL)
1513 return WIMLIB_ERR_INVALID_PARAM;
1515 for_lookup_table_entry(wim->lookup_table, lte_zero_out_refcnt, NULL);
1517 for (i = 0; i < num_resource_wims; i++) {
1518 ret = for_lookup_table_entry(resource_wims[i]->lookup_table,
1527 for_lookup_table_entry(wim->lookup_table, lte_delete_if_new,
1533 reference_resource_paths(WIMStruct *wim,
1534 const tchar * const *resource_wimfiles,
1535 unsigned num_resource_wimfiles,
1538 wimlib_progress_func_t progress_func)
1540 WIMStruct **resource_wims;
1544 resource_wims = CALLOC(num_resource_wimfiles, sizeof(resource_wims[0]));
1546 return WIMLIB_ERR_NOMEM;
1548 for (i = 0; i < num_resource_wimfiles; i++) {
1549 DEBUG("Referencing resources from path \"%"TS"\"",
1550 resource_wimfiles[i]);
1551 ret = wimlib_open_wim(resource_wimfiles[i], open_flags,
1552 &resource_wims[i], progress_func);
1554 goto out_free_resource_wims;
1557 ret = wimlib_reference_resources(wim, resource_wims,
1558 num_resource_wimfiles, ref_flags);
1560 goto out_free_resource_wims;
1562 for (i = 0; i < num_resource_wimfiles; i++)
1563 list_add_tail(&resource_wims[i]->subwim_node, &wim->subwims);
1566 goto out_free_array;
1568 out_free_resource_wims:
1569 for (i = 0; i < num_resource_wimfiles; i++)
1570 wimlib_free(resource_wims[i]);
1572 FREE(resource_wims);
1577 reference_resource_glob(WIMStruct *wim, const tchar *refglob,
1578 int ref_flags, int open_flags,
1579 wimlib_progress_func_t progress_func)
1584 /* Note: glob() is replaced in Windows native builds. */
1585 ret = tglob(refglob, GLOB_ERR | GLOB_NOSORT, NULL, &globbuf);
1587 if (ret == GLOB_NOMATCH) {
1588 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ERR_ON_NOMATCH) {
1589 ERROR("Found no files for glob \"%"TS"\"", refglob);
1590 return WIMLIB_ERR_GLOB_HAD_NO_MATCHES;
1592 return reference_resource_paths(wim,
1600 ERROR_WITH_ERRNO("Failed to process glob \"%"TS"\"", refglob);
1601 if (ret == GLOB_NOSPACE)
1602 return WIMLIB_ERR_NOMEM;
1604 return WIMLIB_ERR_READ;
1608 ret = reference_resource_paths(wim,
1609 (const tchar * const *)globbuf.gl_pathv,
1618 /* API function documented in wimlib.h */
1620 wimlib_reference_resource_files(WIMStruct *wim,
1621 const tchar * const * resource_wimfiles_or_globs,
1625 wimlib_progress_func_t progress_func)
1630 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ENABLE) {
1631 for (i = 0; i < count; i++) {
1632 ret = reference_resource_glob(wim,
1633 resource_wimfiles_or_globs[i],
1642 return reference_resource_paths(wim, resource_wimfiles_or_globs,
1644 open_flags, progress_func);