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);
85 BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
89 struct wim_lookup_table_entry *
90 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
92 struct wim_lookup_table_entry *new;
94 new = memdup(old, sizeof(struct wim_lookup_table_entry));
98 new->extracted_file = NULL;
99 switch (new->resource_location) {
100 case RESOURCE_IN_WIM:
101 list_add(&new->wim_resource_list, &new->rspec->lte_list);
104 case RESOURCE_IN_FILE_ON_DISK:
106 case RESOURCE_WIN32_ENCRYPTED:
109 case RESOURCE_IN_STAGING_FILE:
110 BUILD_BUG_ON((void*)&old->file_on_disk !=
111 (void*)&old->staging_file_name);
113 new->file_on_disk = TSTRDUP(old->file_on_disk);
114 if (!new->file_on_disk)
117 case RESOURCE_IN_ATTACHED_BUFFER:
118 new->attached_buffer = memdup(old->attached_buffer, old->size);
119 if (!new->attached_buffer)
123 case RESOURCE_IN_NTFS_VOLUME:
125 struct ntfs_location *loc;
126 loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
130 loc->stream_name = NULL;
132 loc->path = STRDUP(old->ntfs_loc->path);
135 if (loc->stream_name_nchars) {
136 loc->stream_name = memdup(old->ntfs_loc->stream_name,
137 loc->stream_name_nchars * 2);
138 if (!loc->stream_name)
149 free_lookup_table_entry(new);
154 free_lookup_table_entry(struct wim_lookup_table_entry *lte)
157 switch (lte->resource_location) {
158 case RESOURCE_IN_WIM:
159 list_del(<e->wim_resource_list);
160 if (list_empty(<e->rspec->lte_list))
163 case RESOURCE_IN_FILE_ON_DISK:
165 case RESOURCE_WIN32_ENCRYPTED:
168 case RESOURCE_IN_STAGING_FILE:
169 BUILD_BUG_ON((void*)<e->file_on_disk !=
170 (void*)<e->staging_file_name);
172 case RESOURCE_IN_ATTACHED_BUFFER:
173 BUILD_BUG_ON((void*)<e->file_on_disk !=
174 (void*)<e->attached_buffer);
175 FREE(lte->file_on_disk);
178 case RESOURCE_IN_NTFS_VOLUME:
180 FREE(lte->ntfs_loc->path);
181 FREE(lte->ntfs_loc->stream_name);
194 do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
196 free_lookup_table_entry(entry);
202 free_lookup_table(struct wim_lookup_table *table)
204 DEBUG2("Freeing lookup table");
207 for_lookup_table_entry(table,
208 do_free_lookup_table_entry,
217 * Inserts an entry into the lookup table.
219 * @table: A pointer to the lookup table.
220 * @lte: A pointer to the entry to insert.
223 lookup_table_insert(struct wim_lookup_table *table,
224 struct wim_lookup_table_entry *lte)
226 size_t i = lte->hash_short % table->capacity;
227 hlist_add_head(<e->hash_list, &table->array[i]);
229 /* XXX Make the table grow when too many entries have been inserted. */
230 table->num_entries++;
234 finalize_lte(struct wim_lookup_table_entry *lte)
237 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
238 unlink(lte->staging_file_name);
239 list_del(<e->unhashed_list);
242 free_lookup_table_entry(lte);
245 /* Decrements the reference count for the lookup table entry @lte. If its
246 * reference count reaches 0, it is unlinked from the lookup table. If,
247 * furthermore, the entry has no opened file descriptors associated with it, the
250 lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
251 struct wim_lookup_table *table)
253 wimlib_assert(lte != NULL);
254 wimlib_assert(lte->refcnt != 0);
255 if (--lte->refcnt == 0) {
257 list_del(<e->unhashed_list);
259 lookup_table_unlink(table, lte);
261 if (lte->num_opened_fds == 0)
269 lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
271 if (lte->num_opened_fds != 0)
272 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
277 /* Calls a function on all the entries in the WIM lookup table. Stop early and
278 * return nonzero if any call to the function returns nonzero. */
280 for_lookup_table_entry(struct wim_lookup_table *table,
281 int (*visitor)(struct wim_lookup_table_entry *, void *),
284 struct wim_lookup_table_entry *lte;
285 struct hlist_node *pos, *tmp;
288 for (size_t i = 0; i < table->capacity; i++) {
289 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
292 wimlib_assert2(!(lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA));
293 ret = visitor(lte, arg);
301 /* qsort() callback that sorts streams (represented by `struct
302 * wim_lookup_table_entry's) into an order optimized for reading and writing.
304 * Sorting is done primarily by resource location, then secondarily by a
305 * per-resource location order. For example, resources in WIM files are sorted
306 * primarily by part number, then secondarily by offset, as to implement optimal
307 * reading of either a standalone or split WIM. */
309 cmp_streams_by_sequential_order(const void *p1, const void *p2)
311 const struct wim_lookup_table_entry *lte1, *lte2;
313 WIMStruct *wim1, *wim2;
315 lte1 = *(const struct wim_lookup_table_entry**)p1;
316 lte2 = *(const struct wim_lookup_table_entry**)p2;
318 v = (int)lte1->resource_location - (int)lte2->resource_location;
320 /* Different resource locations? */
324 switch (lte1->resource_location) {
325 case RESOURCE_IN_WIM:
326 wim1 = lte1->rspec->wim;
327 wim2 = lte2->rspec->wim;
329 /* Different (possibly split) WIMs? */
331 v = memcmp(wim1->hdr.guid, wim2->hdr.guid, WIM_GID_LEN);
336 /* Different part numbers in the same WIM? */
337 v = (int)wim1->hdr.part_number - (int)wim2->hdr.part_number;
341 /* Compare by offset. */
342 if (lte1->rspec->offset_in_wim < lte2->rspec->offset_in_wim)
344 if (lte1->rspec->offset_in_wim > lte2->rspec->offset_in_wim)
347 case RESOURCE_IN_FILE_ON_DISK:
349 case RESOURCE_IN_STAGING_FILE:
352 case RESOURCE_WIN32_ENCRYPTED:
354 /* Compare files by path: just a heuristic that will place files
355 * in the same directory next to each other. */
356 return tstrcmp(lte1->file_on_disk, lte2->file_on_disk);
358 case RESOURCE_IN_NTFS_VOLUME:
359 return tstrcmp(lte1->ntfs_loc->path, lte2->ntfs_loc->path);
362 /* No additional sorting order defined for this resource
363 * location (e.g. RESOURCE_IN_ATTACHED_BUFFER); simply compare
364 * everything equal to each other. */
370 sort_stream_list_by_sequential_order(struct list_head *stream_list,
371 size_t list_head_offset)
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 array_size = num_streams * sizeof(array[0]);
383 array = MALLOC(array_size);
385 return WIMLIB_ERR_NOMEM;
386 cur = stream_list->next;
387 for (i = 0; i < num_streams; i++) {
388 array[i] = (struct wim_lookup_table_entry*)((u8*)cur -
393 qsort(array, num_streams, sizeof(array[0]),
394 cmp_streams_by_sequential_order);
396 INIT_LIST_HEAD(stream_list);
397 for (i = 0; i < num_streams; i++) {
398 list_add_tail((struct list_head*)
399 ((u8*)array[i] + list_head_offset),
408 add_lte_to_array(struct wim_lookup_table_entry *lte,
411 struct wim_lookup_table_entry ***pp = _pp;
416 /* Iterate through the lookup table entries, but first sort them by stream
417 * offset in the WIM. Caution: this is intended to be used when the stream
418 * offset field has actually been set. */
420 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
421 int (*visitor)(struct wim_lookup_table_entry *,
425 struct wim_lookup_table_entry **lte_array, **p;
426 size_t num_streams = table->num_entries;
429 lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
431 return WIMLIB_ERR_NOMEM;
433 for_lookup_table_entry(table, add_lte_to_array, &p);
435 wimlib_assert(p == lte_array + num_streams);
437 qsort(lte_array, num_streams, sizeof(lte_array[0]),
438 cmp_streams_by_sequential_order);
440 for (size_t i = 0; i < num_streams; i++) {
441 ret = visitor(lte_array[i], arg);
449 /* On-disk format of a WIM lookup table entry (stream entry). */
450 struct wim_lookup_table_entry_disk {
451 /* Size, offset, and flags of the stream. */
452 struct wim_reshdr_disk reshdr;
454 /* Which part of the split WIM this stream is in; indexed from 1. */
457 /* Reference count of this stream over all WIM images. */
460 /* SHA1 message digest of the uncompressed data of this stream, or
461 * optionally all zeroes if this stream is of zero length. */
462 u8 hash[SHA1_HASH_SIZE];
465 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
468 validate_resource(const struct wim_resource_spec *rspec)
470 struct wim_lookup_table_entry *lte;
471 if (!list_is_singular(&rspec->lte_list)) {
472 list_for_each_entry(lte, &rspec->lte_list, wim_resource_list) {
473 if (rspec->flags & WIM_RESHDR_FLAG_COMPRESSED)
474 lte->flags |= WIM_RESHDR_FLAG_COMPRESSED;
476 lte->flags &= ~WIM_RESHDR_FLAG_COMPRESSED;
478 if (lte->offset_in_res + lte->size < lte->size ||
479 lte->offset_in_res + lte->size > rspec->uncompressed_size)
481 return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
489 * Reads the lookup table from a WIM file.
491 * Saves lookup table entries for non-metadata streams in a hash table, and
492 * saves the metadata entry for each image in a special per-image location (the
493 * image_metadata array).
496 * WIMLIB_ERR_SUCCESS (0)
497 * WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
498 * WIMLIB_ERR_RESOURCE_NOT_FOUND
501 read_wim_lookup_table(WIMStruct *wim)
506 struct wim_lookup_table *table;
507 struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
508 struct wim_resource_spec *cur_rspec;
511 BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
512 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
514 DEBUG("Reading lookup table.");
516 /* Calculate number of entries in the lookup table. */
517 num_entries = wim->hdr.lookup_table_reshdr.uncompressed_size /
518 sizeof(struct wim_lookup_table_entry_disk);
520 /* Read the lookup table into a buffer. */
521 ret = wim_reshdr_to_data(&wim->hdr.lookup_table_reshdr, wim, &buf);
525 /* Allocate hash table. */
526 table = new_lookup_table(num_entries * 2 + 1);
528 ERROR("Not enough memory to read lookup table.");
529 ret = WIMLIB_ERR_NOMEM;
533 /* Allocate and initalize `struct wim_lookup_table_entry's from the
534 * on-disk lookup table. */
535 wim->current_image = 0;
537 for (i = 0; i < num_entries; i++) {
538 const struct wim_lookup_table_entry_disk *disk_entry =
539 &((const struct wim_lookup_table_entry_disk*)buf)[i];
541 struct wim_reshdr reshdr;
543 ret = get_wim_reshdr(&disk_entry->reshdr, &reshdr);
545 ERROR("Resource header is invalid!");
546 goto out_free_lookup_table;
549 DEBUG("reshdr: size_in_wim=%"PRIu64", "
550 "uncompressed_size=%"PRIu64", "
551 "offset_in_wim=%"PRIu64", "
553 reshdr.size_in_wim, reshdr.uncompressed_size,
554 reshdr.offset_in_wim, reshdr.flags);
556 cur_entry = new_lookup_table_entry();
557 if (cur_entry == NULL) {
558 ERROR("Not enough memory to read lookup table!");
559 ret = WIMLIB_ERR_NOMEM;
560 goto out_free_lookup_table;
563 part_number = le16_to_cpu(disk_entry->part_number);
564 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
565 copy_hash(cur_entry->hash, disk_entry->hash);
567 if (part_number != wim->hdr.part_number) {
568 WARNING("A lookup table entry in part %hu of the WIM "
569 "points to part %hu (ignoring it)",
570 wim->hdr.part_number, part_number);
571 free_lookup_table_entry(cur_entry);
575 if (cur_rspec == NULL ||
576 !(reshdr.flags & WIM_RESHDR_FLAG_CONCAT))
578 /* Starting new run of stream entries that all share the
579 * same WIM resource (streams concatenated together); or
580 * simply a single normal entry by itself. */
582 if (cur_rspec != NULL) {
583 ret = validate_resource(cur_rspec);
585 goto out_free_cur_entry;
588 cur_rspec = MALLOC(sizeof(struct wim_resource_spec));
589 if (cur_rspec == NULL) {
590 ERROR("Not enough memory to read lookup table!");
591 ret = WIMLIB_ERR_NOMEM;
592 goto out_free_cur_entry;
594 wim_res_hdr_to_spec(&reshdr, wim, cur_rspec);
595 if (reshdr.flags & WIM_RESHDR_FLAG_CONCAT) {
596 cur_rspec->size_in_wim = 0;
597 cur_rspec->uncompressed_size = 0;
599 } else if (is_zero_hash(cur_entry->hash)) {
600 /* Found the resource specification for the run. */
601 cur_rspec->offset_in_wim = reshdr.offset_in_wim;
602 cur_rspec->size_in_wim = reshdr.size_in_wim;
603 cur_rspec->flags = reshdr.flags;
604 DEBUG("Full run is %"PRIu64" compressed bytes "
605 "at file offset %"PRIu64" (flags 0x%02x)",
606 cur_rspec->size_in_wim,
607 cur_rspec->offset_in_wim,
609 free_lookup_table_entry(cur_entry);
613 if (reshdr.flags & WIM_RESHDR_FLAG_CONCAT) {
614 /* Continuing the run with another stream. */
615 DEBUG("Continuing concat run with stream: "
616 "%"PRIu64" uncompressed bytes @ resource offset %"PRIu64")",
617 reshdr.size_in_wim, reshdr.offset_in_wim);
618 cur_rspec->uncompressed_size += reshdr.size_in_wim;
621 lte_bind_wim_resource_spec(cur_entry, cur_rspec);
622 if (reshdr.flags & WIM_RESHDR_FLAG_CONCAT) {
623 /* In concatenation runs, the offset field is used for
624 * in-resource offset, not the in-WIM offset, and the
625 * size field is used for the uncompressed size, not the
626 * compressed size. */
627 cur_entry->offset_in_res = reshdr.offset_in_wim;
628 cur_entry->size = reshdr.size_in_wim;
629 cur_entry->flags = reshdr.flags;
631 cur_entry->offset_in_res = 0;
632 cur_entry->size = reshdr.uncompressed_size;
633 cur_entry->flags = reshdr.flags;
637 if (is_zero_hash(cur_entry->hash)) {
638 WARNING("The WIM lookup table contains an entry with a "
639 "SHA1 message digest of all 0's (ignoring it)");
640 free_lookup_table_entry(cur_entry);
644 if (cur_entry->flags & WIM_RESHDR_FLAG_METADATA) {
645 /* Lookup table entry for a metadata resource */
646 if (cur_entry->refcnt != 1) {
647 /* Metadata entries with no references must be
648 * ignored. See for example the WinPE WIMs from
650 if (cur_entry->refcnt == 0) {
651 free_lookup_table_entry(cur_entry);
654 if (wimlib_print_errors) {
655 ERROR("Found metadata resource with refcnt != 1:");
656 print_lookup_table_entry(cur_entry, stderr);
658 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
659 goto out_free_cur_entry;
662 if (wim->hdr.part_number != 1) {
663 WARNING("Ignoring metadata resource found in a "
664 "non-first part of the split WIM");
665 free_lookup_table_entry(cur_entry);
668 if (wim->current_image == wim->hdr.image_count) {
669 WARNING("The WIM header says there are %u images "
670 "in the WIM, but we found more metadata "
671 "resources than this (ignoring the extra)",
672 wim->hdr.image_count);
673 free_lookup_table_entry(cur_entry);
677 /* Notice very carefully: We are assigning the metadata
678 * resources in the exact order mirrored by their lookup
679 * table entries on disk, which is the behavior of
680 * Microsoft's software. In particular, this overrides
681 * the actual locations of the metadata resources
682 * themselves in the WIM file as well as any information
683 * written in the XML data. */
684 DEBUG("Found metadata resource for image %u at "
686 wim->current_image + 1,
687 cur_entry->rspec->offset_in_wim);
689 wim->current_image++]->metadata_lte = cur_entry;
691 /* Lookup table entry for a stream that is not a
692 * metadata resource */
693 duplicate_entry = lookup_resource(table, cur_entry->hash);
694 if (duplicate_entry) {
695 if (wimlib_print_errors) {
696 WARNING("The WIM lookup table contains two entries with the "
697 "same SHA1 message digest!");
698 WARNING("The first entry is:");
699 print_lookup_table_entry(duplicate_entry, stderr);
700 WARNING("The second entry is:");
701 print_lookup_table_entry(cur_entry, stderr);
703 free_lookup_table_entry(cur_entry);
706 lookup_table_insert(table, cur_entry);
711 if (cur_rspec != NULL) {
712 ret = validate_resource(cur_rspec);
714 goto out_free_cur_entry;
717 if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
718 WARNING("The header of \"%"TS"\" says there are %u images in\n"
719 " the WIM, but we only found %d metadata resources! Acting as if\n"
720 " the header specified only %d images instead.",
721 wim->filename, wim->hdr.image_count,
722 wim->current_image, wim->current_image);
723 for (int i = wim->current_image; i < wim->hdr.image_count; i++)
724 put_image_metadata(wim->image_metadata[i], NULL);
725 wim->hdr.image_count = wim->current_image;
727 DEBUG("Done reading lookup table.");
728 wim->lookup_table = table;
734 out_free_lookup_table:
735 free_lookup_table(table);
739 wim->current_image = 0;
745 write_wim_lookup_table_entry(const struct wim_lookup_table_entry *lte,
746 struct wim_lookup_table_entry_disk *disk_entry,
749 put_wim_reshdr(<e->out_reshdr, &disk_entry->reshdr);
750 disk_entry->part_number = cpu_to_le16(part_number);
751 disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
752 copy_hash(disk_entry->hash, lte->hash);
756 write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
757 struct filedes *out_fd,
759 struct wim_reshdr *out_reshdr,
760 int write_resource_flags,
761 struct wimlib_lzx_context **comp_ctx)
764 struct wim_lookup_table_entry *lte;
765 struct wim_lookup_table_entry_disk *table_buf;
766 struct wim_lookup_table_entry_disk *table_buf_ptr;
770 list_for_each_entry(lte, stream_list, lookup_table_list)
771 table_size += sizeof(struct wim_lookup_table_entry_disk);
773 DEBUG("Writing WIM lookup table (size=%zu, offset=%"PRIu64")",
774 table_size, out_fd->offset);
776 table_buf = MALLOC(table_size);
778 ERROR("Failed to allocate %zu bytes for temporary lookup table",
780 return WIMLIB_ERR_NOMEM;
782 table_buf_ptr = table_buf;
783 list_for_each_entry(lte, stream_list, lookup_table_list)
784 write_wim_lookup_table_entry(lte, table_buf_ptr++, part_number);
786 /* Write the lookup table uncompressed. Although wimlib can handle a
787 * compressed lookup table, MS software cannot. */
788 ret = write_wim_resource_from_buffer(table_buf,
790 WIM_RESHDR_FLAG_METADATA,
792 WIMLIB_COMPRESSION_TYPE_NONE,
796 write_resource_flags,
799 DEBUG("ret=%d", ret);
804 append_lookup_table_entry(struct wim_lookup_table_entry *lte, void *_list)
806 /* Lookup table entries with 'out_refcnt' == 0 correspond to streams not
807 * written and not present in the resulting WIM file, and should not be
808 * included in the lookup table.
810 * Lookup table entries marked as filtered (EXTERNAL_WIM) with
811 * 'out_refcnt != 0' were referenced as part of the logical write but
812 * correspond to streams that were not in fact written, and should not
813 * be included in the lookup table.
815 * Lookup table entries marked as filtered (SAME_WIM) with 'out_refcnt
816 * != 0' were referenced as part of the logical write but correspond to
817 * streams that were not in fact written, but nevertheless were already
818 * present in the WIM being overwritten in-place. These entries must be
819 * included in the lookup table, and the resource information to write
820 * needs to be copied from the resource information read originally.
822 if (lte->out_refcnt != 0 && !(lte->filtered & FILTERED_EXTERNAL_WIM)) {
823 if (lte->filtered & FILTERED_SAME_WIM)
824 wim_res_spec_to_hdr(lte->rspec, <e->out_reshdr);
825 list_add_tail(<e->lookup_table_list, (struct list_head*)_list);
831 write_wim_lookup_table(WIMStruct *wim, int image, int write_flags,
832 struct wim_reshdr *out_reshdr,
833 struct list_head *stream_list_override)
835 int write_resource_flags;
836 struct list_head _stream_list;
837 struct list_head *stream_list;
839 if (stream_list_override) {
840 stream_list = stream_list_override;
842 stream_list = &_stream_list;
843 INIT_LIST_HEAD(stream_list);
846 if (!(write_flags & WIMLIB_WRITE_FLAG_NO_METADATA)) {
850 if (image == WIMLIB_ALL_IMAGES) {
852 end_image = wim->hdr.image_count;
858 /* Push metadata resource lookup table entries onto the front of
859 * the list in reverse order, so that they're written in order.
861 for (int i = end_image; i >= start_image; i--) {
862 struct wim_lookup_table_entry *metadata_lte;
864 metadata_lte = wim->image_metadata[i - 1]->metadata_lte;
865 metadata_lte->out_refcnt = 1;
866 metadata_lte->out_reshdr.flags |= WIM_RESHDR_FLAG_METADATA;
867 list_add(&metadata_lte->lookup_table_list, stream_list);
871 /* Append additional lookup table entries that need to be written, with
872 * some special handling for streams that have been marked as filtered.
874 if (!stream_list_override) {
875 for_lookup_table_entry(wim->lookup_table,
876 append_lookup_table_entry, stream_list);
879 write_resource_flags = 0;
880 if (write_flags & WIMLIB_WRITE_FLAG_PIPABLE)
881 write_resource_flags |= WIMLIB_WRITE_RESOURCE_FLAG_PIPABLE;
882 return write_wim_lookup_table_from_stream_list(stream_list,
884 wim->hdr.part_number,
886 write_resource_flags,
892 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
894 lte->real_refcnt = 0;
899 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
906 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
908 if (lte->extracted_file != NULL) {
909 FREE(lte->extracted_file);
910 lte->extracted_file = NULL;
916 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
924 tprintf(T("Uncompressed size = %"PRIu64" bytes\n"),
926 if (lte_is_partial(lte)) {
927 tprintf(T("Offset = %"PRIu64" bytes\n"),
930 tprintf(T("Raw uncompressed size = %"PRIu64" bytes\n"),
931 lte->rspec->uncompressed_size);
933 tprintf(T("Raw compressed size = %"PRIu64" bytes\n"),
934 lte->rspec->size_in_wim);
936 tprintf(T("Raw offset = %"PRIu64" bytes\n"),
937 lte->rspec->offset_in_wim);
938 } else if (lte->resource_location == RESOURCE_IN_WIM) {
939 tprintf(T("Compressed size = %"PRIu64" bytes\n"),
940 lte->rspec->size_in_wim);
942 tprintf(T("Offset = %"PRIu64" bytes\n"),
943 lte->rspec->offset_in_wim);
946 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
949 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
950 lte->back_inode, lte->back_stream_id);
952 tfprintf(out, T("Hash = 0x"));
953 print_hash(lte->hash, out);
957 tfprintf(out, T("Flags = "));
958 u8 flags = lte->flags;
959 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
960 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
961 if (flags & WIM_RESHDR_FLAG_FREE)
962 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
963 if (flags & WIM_RESHDR_FLAG_METADATA)
964 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
965 if (flags & WIM_RESHDR_FLAG_SPANNED)
966 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
967 if (flags & WIM_RESHDR_FLAG_CONCAT)
968 tfputs(T("WIM_RESHDR_FLAG_CONCAT, "), out);
970 switch (lte->resource_location) {
971 case RESOURCE_IN_WIM:
972 if (lte->rspec->wim->filename) {
973 tfprintf(out, T("WIM file = `%"TS"'\n"),
974 lte->rspec->wim->filename);
978 case RESOURCE_WIN32_ENCRYPTED:
980 case RESOURCE_IN_FILE_ON_DISK:
981 tfprintf(out, T("File on Disk = `%"TS"'\n"),
985 case RESOURCE_IN_STAGING_FILE:
986 tfprintf(out, T("Staging File = `%"TS"'\n"),
987 lte->staging_file_name);
997 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
998 struct wimlib_resource_entry *wentry)
1000 memset(wentry, 0, sizeof(*wentry));
1002 wentry->uncompressed_size = lte->size;
1003 if (lte->resource_location == RESOURCE_IN_WIM) {
1004 wentry->part_number = lte->rspec->wim->hdr.part_number;
1005 if (lte_is_partial(lte)) {
1006 wentry->compressed_size = 0;
1007 wentry->offset = lte->offset_in_res;
1009 wentry->compressed_size = lte->rspec->size_in_wim;
1010 wentry->offset = lte->rspec->offset_in_wim;
1012 wentry->raw_resource_offset_in_wim = lte->rspec->offset_in_wim;
1013 wentry->raw_resource_uncompressed_size = lte->rspec->uncompressed_size;
1014 wentry->raw_resource_compressed_size = lte->rspec->size_in_wim;
1016 copy_hash(wentry->sha1_hash, lte->hash);
1017 wentry->reference_count = lte->refcnt;
1018 wentry->is_compressed = (lte->flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
1019 wentry->is_metadata = (lte->flags & WIM_RESHDR_FLAG_METADATA) != 0;
1020 wentry->is_free = (lte->flags & WIM_RESHDR_FLAG_FREE) != 0;
1021 wentry->is_spanned = (lte->flags & WIM_RESHDR_FLAG_SPANNED) != 0;
1022 wentry->is_partial = lte_is_partial(lte);
1025 struct iterate_lte_context {
1026 wimlib_iterate_lookup_table_callback_t cb;
1031 do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
1033 struct iterate_lte_context *ctx = _ctx;
1034 struct wimlib_resource_entry entry;
1036 lte_to_wimlib_resource_entry(lte, &entry);
1037 return (*ctx->cb)(&entry, ctx->user_ctx);
1040 /* API function documented in wimlib.h */
1042 wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
1043 wimlib_iterate_lookup_table_callback_t cb,
1046 struct iterate_lte_context ctx = {
1048 .user_ctx = user_ctx,
1050 if (wim->hdr.part_number == 1) {
1052 for (int i = 0; i < wim->hdr.image_count; i++) {
1053 ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
1059 return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
1062 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
1063 * lookup table, or NULL if there is none. */
1064 struct wim_lookup_table_entry *
1065 lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
1068 struct wim_lookup_table_entry *lte;
1069 struct hlist_node *pos;
1071 wimlib_assert(table != NULL);
1072 wimlib_assert(hash != NULL);
1074 i = *(size_t*)hash % table->capacity;
1075 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
1076 if (hashes_equal(hash, lte->hash))
1083 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
1084 * given a path name.
1086 * This is only for pre-resolved inodes.
1089 wim_pathname_to_stream(WIMStruct *wim,
1092 struct wim_dentry **dentry_ret,
1093 struct wim_lookup_table_entry **lte_ret,
1094 u16 *stream_idx_ret)
1096 struct wim_dentry *dentry;
1097 struct wim_lookup_table_entry *lte;
1099 const tchar *stream_name = NULL;
1100 struct wim_inode *inode;
1103 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
1104 stream_name = path_stream_name(path);
1106 p = (tchar*)stream_name - 1;
1111 dentry = get_dentry(wim, path);
1117 inode = dentry->d_inode;
1119 if (!inode->i_resolved)
1120 if (inode_resolve_ltes(inode, wim->lookup_table, false))
1123 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
1124 && inode_is_directory(inode))
1128 struct wim_ads_entry *ads_entry;
1130 ads_entry = inode_get_ads_entry(inode, stream_name,
1133 stream_idx = ads_idx + 1;
1134 lte = ads_entry->lte;
1140 lte = inode_unnamed_stream_resolved(inode, &stream_idx);
1144 *dentry_ret = dentry;
1148 *stream_idx_ret = stream_idx;
1154 resource_not_found_error(const struct wim_inode *inode, const u8 *hash)
1156 if (wimlib_print_errors) {
1157 ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
1158 tfprintf(stderr, T(" SHA-1 message digest of missing resource:\n "));
1159 print_hash(hash, stderr);
1160 tputc(T('\n'), stderr);
1162 return WIMLIB_ERR_RESOURCE_NOT_FOUND;
1166 * Resolve an inode's lookup table entries.
1168 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
1169 * lookup table) with pointers directly to the lookup table entries.
1171 * If @force is %false:
1172 * If any needed SHA1 message digests are not found in the lookup table,
1173 * WIMLIB_ERR_RESOURCE_NOT_FOUND is returned and the inode is left
1175 * If @force is %true:
1176 * If any needed SHA1 message digests are not found in the lookup table,
1177 * new entries are allocated and inserted into the lookup table.
1180 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table,
1185 if (!inode->i_resolved) {
1186 struct wim_lookup_table_entry *lte, *ads_lte;
1188 /* Resolve the default file stream */
1190 hash = inode->i_hash;
1191 if (!is_zero_hash(hash)) {
1192 lte = lookup_resource(table, hash);
1195 lte = new_lookup_table_entry();
1197 return WIMLIB_ERR_NOMEM;
1198 copy_hash(lte->hash, hash);
1199 lookup_table_insert(table, lte);
1201 goto resource_not_found;
1206 /* Resolve the alternate data streams */
1207 struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
1208 for (u16 i = 0; i < inode->i_num_ads; i++) {
1209 struct wim_ads_entry *cur_entry;
1212 cur_entry = &inode->i_ads_entries[i];
1213 hash = cur_entry->hash;
1214 if (!is_zero_hash(hash)) {
1215 ads_lte = lookup_resource(table, hash);
1218 ads_lte = new_lookup_table_entry();
1220 return WIMLIB_ERR_NOMEM;
1221 copy_hash(ads_lte->hash, hash);
1222 lookup_table_insert(table, ads_lte);
1224 goto resource_not_found;
1228 ads_ltes[i] = ads_lte;
1231 for (u16 i = 0; i < inode->i_num_ads; i++)
1232 inode->i_ads_entries[i].lte = ads_ltes[i];
1233 inode->i_resolved = 1;
1238 return resource_not_found_error(inode, hash);
1242 inode_unresolve_ltes(struct wim_inode *inode)
1244 if (inode->i_resolved) {
1246 copy_hash(inode->i_hash, inode->i_lte->hash);
1248 zero_out_hash(inode->i_hash);
1250 for (u16 i = 0; i < inode->i_num_ads; i++) {
1251 if (inode->i_ads_entries[i].lte)
1252 copy_hash(inode->i_ads_entries[i].hash,
1253 inode->i_ads_entries[i].lte->hash);
1255 zero_out_hash(inode->i_ads_entries[i].hash);
1257 inode->i_resolved = 0;
1262 * Returns the lookup table entry for stream @stream_idx of the inode, where
1263 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
1264 * corresponds to an alternate data stream.
1266 * This works for both resolved and un-resolved inodes.
1268 struct wim_lookup_table_entry *
1269 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
1270 const struct wim_lookup_table *table)
1272 if (inode->i_resolved)
1273 return inode_stream_lte_resolved(inode, stream_idx);
1275 return inode_stream_lte_unresolved(inode, stream_idx, table);
1278 struct wim_lookup_table_entry *
1279 inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
1281 wimlib_assert(inode->i_resolved);
1282 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1283 if (inode_stream_name_nbytes(inode, i) == 0 &&
1284 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
1286 *stream_idx_ret = i;
1287 return inode_stream_lte_resolved(inode, i);
1290 *stream_idx_ret = 0;
1294 struct wim_lookup_table_entry *
1295 inode_unnamed_lte_resolved(const struct wim_inode *inode)
1298 return inode_unnamed_stream_resolved(inode, &stream_idx);
1301 struct wim_lookup_table_entry *
1302 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
1303 const struct wim_lookup_table *table)
1305 wimlib_assert(!inode->i_resolved);
1306 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1307 if (inode_stream_name_nbytes(inode, i) == 0 &&
1308 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
1310 return inode_stream_lte_unresolved(inode, i, table);
1316 /* Return the lookup table entry for the unnamed data stream of an inode, or
1317 * NULL if there is none.
1319 * You'd think this would be easier than it actually is, since the unnamed data
1320 * stream should be the one referenced from the inode itself. Alas, if there
1321 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
1322 * data stream in one of the alternate data streams instead of inside the WIM
1323 * dentry itself. So we need to check the alternate data streams too.
1325 * Also, note that a dentry may appear to have more than one unnamed stream, but
1326 * if the SHA1 message digest is all 0's then the corresponding stream does not
1327 * really "count" (this is the case for the inode's own file stream when the
1328 * file stream that should be there is actually in one of the alternate stream
1329 * entries.). This is despite the fact that we may need to extract such a
1330 * missing entry as an empty file or empty named data stream.
1332 struct wim_lookup_table_entry *
1333 inode_unnamed_lte(const struct wim_inode *inode,
1334 const struct wim_lookup_table *table)
1336 if (inode->i_resolved)
1337 return inode_unnamed_lte_resolved(inode);
1339 return inode_unnamed_lte_unresolved(inode, table);
1342 /* Returns the SHA1 message digest of the unnamed data stream of a WIM inode, or
1343 * 'zero_hash' if the unnamed data stream is missing has all zeroes in its SHA1
1344 * message digest field. */
1346 inode_unnamed_stream_hash(const struct wim_inode *inode)
1350 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1351 if (inode_stream_name_nbytes(inode, i) == 0) {
1352 hash = inode_stream_hash(inode, i);
1353 if (!is_zero_hash(hash))
1360 struct wim_lookup_table_entry **
1361 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1363 wimlib_assert(lte->unhashed);
1364 struct wim_inode *inode = lte->back_inode;
1365 u32 stream_id = lte->back_stream_id;
1367 return &inode->i_lte;
1369 for (u16 i = 0; i < inode->i_num_ads; i++)
1370 if (inode->i_ads_entries[i].stream_id == stream_id)
1371 return &inode->i_ads_entries[i].lte;
1376 /* Calculate the SHA1 message digest of a stream and move it from the list of
1377 * unhashed streams to the stream lookup table, possibly joining it with an
1378 * existing lookup table entry for an identical stream.
1380 * @lte: An unhashed lookup table entry.
1381 * @lookup_table: Lookup table for the WIM.
1382 * @lte_ret: On success, write a pointer to the resulting lookup table
1383 * entry to this location. This will be the same as @lte
1384 * if it was inserted into the lookup table, or different if
1385 * a duplicate stream was found.
1387 * Returns 0 on success; nonzero if there is an error reading the stream.
1390 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1391 struct wim_lookup_table *lookup_table,
1392 struct wim_lookup_table_entry **lte_ret)
1395 struct wim_lookup_table_entry *duplicate_lte;
1396 struct wim_lookup_table_entry **back_ptr;
1398 wimlib_assert(lte->unhashed);
1400 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1401 * union with the SHA1 message digest and will no longer be valid once
1402 * the SHA1 has been calculated. */
1403 back_ptr = retrieve_lte_pointer(lte);
1405 ret = sha1_stream(lte);
1409 /* Look for a duplicate stream */
1410 duplicate_lte = lookup_resource(lookup_table, lte->hash);
1411 list_del(<e->unhashed_list);
1412 if (duplicate_lte) {
1413 /* We have a duplicate stream. Transfer the reference counts
1414 * from this stream to the duplicate, update the reference to
1415 * this stream (in an inode or ads_entry) to point to the
1416 * duplicate, then free this stream. */
1417 wimlib_assert(!(duplicate_lte->unhashed));
1418 duplicate_lte->refcnt += lte->refcnt;
1419 duplicate_lte->out_refcnt += lte->out_refcnt;
1420 *back_ptr = duplicate_lte;
1421 free_lookup_table_entry(lte);
1422 lte = duplicate_lte;
1424 /* No duplicate stream, so we need to insert
1425 * this stream into the lookup table and treat
1426 * it as a hashed stream. */
1427 lookup_table_insert(lookup_table, lte);
1436 lte_clone_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1438 struct wim_lookup_table *lookup_table = _lookup_table;
1440 if (lookup_resource(lookup_table, lte->hash))
1441 return 0; /* Resource already present. */
1443 lte = clone_lookup_table_entry(lte);
1445 return WIMLIB_ERR_NOMEM;
1446 lte->out_refcnt = 1;
1447 lookup_table_insert(lookup_table, lte);
1452 lte_delete_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1454 struct wim_lookup_table *lookup_table = _lookup_table;
1456 if (lte->out_refcnt) {
1457 lookup_table_unlink(lookup_table, lte);
1458 free_lookup_table_entry(lte);
1463 /* API function documented in wimlib.h */
1465 wimlib_reference_resources(WIMStruct *wim,
1466 WIMStruct **resource_wims, unsigned num_resource_wims,
1473 return WIMLIB_ERR_INVALID_PARAM;
1475 if (num_resource_wims != 0 && resource_wims == NULL)
1476 return WIMLIB_ERR_INVALID_PARAM;
1478 for (i = 0; i < num_resource_wims; i++)
1479 if (resource_wims[i] == NULL)
1480 return WIMLIB_ERR_INVALID_PARAM;
1482 for_lookup_table_entry(wim->lookup_table, lte_zero_out_refcnt, NULL);
1484 for (i = 0; i < num_resource_wims; i++) {
1485 ret = for_lookup_table_entry(resource_wims[i]->lookup_table,
1494 for_lookup_table_entry(wim->lookup_table, lte_delete_if_new,
1500 reference_resource_paths(WIMStruct *wim,
1501 const tchar * const *resource_wimfiles,
1502 unsigned num_resource_wimfiles,
1505 wimlib_progress_func_t progress_func)
1507 WIMStruct **resource_wims;
1511 resource_wims = CALLOC(num_resource_wimfiles, sizeof(resource_wims[0]));
1513 return WIMLIB_ERR_NOMEM;
1515 for (i = 0; i < num_resource_wimfiles; i++) {
1516 DEBUG("Referencing resources from path \"%"TS"\"",
1517 resource_wimfiles[i]);
1518 ret = wimlib_open_wim(resource_wimfiles[i], open_flags,
1519 &resource_wims[i], progress_func);
1521 goto out_free_resource_wims;
1524 ret = wimlib_reference_resources(wim, resource_wims,
1525 num_resource_wimfiles, ref_flags);
1527 goto out_free_resource_wims;
1529 for (i = 0; i < num_resource_wimfiles; i++)
1530 list_add_tail(&resource_wims[i]->subwim_node, &wim->subwims);
1533 goto out_free_array;
1535 out_free_resource_wims:
1536 for (i = 0; i < num_resource_wimfiles; i++)
1537 wimlib_free(resource_wims[i]);
1539 FREE(resource_wims);
1544 reference_resource_glob(WIMStruct *wim, const tchar *refglob,
1545 int ref_flags, int open_flags,
1546 wimlib_progress_func_t progress_func)
1551 /* Note: glob() is replaced in Windows native builds. */
1552 ret = tglob(refglob, GLOB_ERR | GLOB_NOSORT, NULL, &globbuf);
1554 if (ret == GLOB_NOMATCH) {
1555 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ERR_ON_NOMATCH) {
1556 ERROR("Found no files for glob \"%"TS"\"", refglob);
1557 return WIMLIB_ERR_GLOB_HAD_NO_MATCHES;
1559 return reference_resource_paths(wim,
1567 ERROR_WITH_ERRNO("Failed to process glob \"%"TS"\"", refglob);
1568 if (ret == GLOB_NOSPACE)
1569 return WIMLIB_ERR_NOMEM;
1571 return WIMLIB_ERR_READ;
1575 ret = reference_resource_paths(wim,
1576 (const tchar * const *)globbuf.gl_pathv,
1585 /* API function documented in wimlib.h */
1587 wimlib_reference_resource_files(WIMStruct *wim,
1588 const tchar * const * resource_wimfiles_or_globs,
1592 wimlib_progress_func_t progress_func)
1597 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ENABLE) {
1598 for (i = 0; i < count; i++) {
1599 ret = reference_resource_glob(wim,
1600 resource_wimfiles_or_globs[i],
1609 return reference_resource_paths(wim, resource_wimfiles_or_globs,
1611 open_flags, progress_func);