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,
469 u64 offset_save, u64 size_save)
471 struct wim_lookup_table_entry *lte;
472 if (!list_is_singular(&rspec->lte_list)) {
473 list_for_each_entry(lte, &rspec->lte_list, wim_resource_list) {
474 if (rspec->flags & WIM_RESHDR_FLAG_COMPRESSED)
475 lte->flags |= WIM_RESHDR_FLAG_COMPRESSED;
477 lte->flags &= ~WIM_RESHDR_FLAG_COMPRESSED;
479 if (!(lte->flags & WIM_RESHDR_FLAG_CONCAT)) {
480 lte->offset_in_res = offset_save;
481 lte->size = size_save;
484 if (lte->offset_in_res + lte->size < lte->size ||
485 lte->offset_in_res + lte->size > rspec->uncompressed_size)
487 return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
495 * Reads the lookup table from a WIM file.
497 * Saves lookup table entries for non-metadata streams in a hash table, and
498 * saves the metadata entry for each image in a special per-image location (the
499 * image_metadata array).
502 * WIMLIB_ERR_SUCCESS (0)
503 * WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
504 * WIMLIB_ERR_RESOURCE_NOT_FOUND
507 read_wim_lookup_table(WIMStruct *wim)
512 struct wim_lookup_table *table;
513 struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
514 struct wim_resource_spec *cur_rspec;
519 BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
520 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
522 DEBUG("Reading lookup table.");
524 /* Calculate number of entries in the lookup table. */
525 num_entries = wim->hdr.lookup_table_reshdr.uncompressed_size /
526 sizeof(struct wim_lookup_table_entry_disk);
528 /* Read the lookup table into a buffer. */
529 ret = wim_reshdr_to_data(&wim->hdr.lookup_table_reshdr, wim, &buf);
533 /* Allocate hash table. */
534 table = new_lookup_table(num_entries * 2 + 1);
536 ERROR("Not enough memory to read lookup table.");
537 ret = WIMLIB_ERR_NOMEM;
541 /* Allocate and initalize `struct wim_lookup_table_entry's from the
542 * on-disk lookup table. */
543 wim->current_image = 0;
545 for (i = 0; i < num_entries; i++) {
546 const struct wim_lookup_table_entry_disk *disk_entry =
547 &((const struct wim_lookup_table_entry_disk*)buf)[i];
549 struct wim_reshdr reshdr;
551 ret = get_wim_reshdr(&disk_entry->reshdr, &reshdr);
553 ERROR("Resource header is invalid!");
554 goto out_free_lookup_table;
557 DEBUG("reshdr: size_in_wim=%"PRIu64", "
558 "uncompressed_size=%"PRIu64", "
559 "offset_in_wim=%"PRIu64", "
561 reshdr.size_in_wim, reshdr.uncompressed_size,
562 reshdr.offset_in_wim, reshdr.flags);
564 cur_entry = new_lookup_table_entry();
565 if (cur_entry == NULL) {
566 ERROR("Not enough memory to read lookup table!");
567 ret = WIMLIB_ERR_NOMEM;
568 goto out_free_lookup_table;
571 part_number = le16_to_cpu(disk_entry->part_number);
572 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
573 copy_hash(cur_entry->hash, disk_entry->hash);
575 if (part_number != wim->hdr.part_number) {
576 WARNING("A lookup table entry in part %hu of the WIM "
577 "points to part %hu (ignoring it)",
578 wim->hdr.part_number, part_number);
579 free_lookup_table_entry(cur_entry);
583 if (cur_rspec == NULL ||
584 !(reshdr.flags & WIM_RESHDR_FLAG_CONCAT))
586 /* Starting new run of stream entries that all share the
587 * same WIM resource (streams concatenated together); or
588 * simply a single normal entry by itself. */
590 if (cur_rspec != NULL) {
591 ret = validate_resource(cur_rspec, offset_save,
594 goto out_free_cur_entry;
597 cur_rspec = MALLOC(sizeof(struct wim_resource_spec));
598 if (cur_rspec == NULL) {
599 ERROR("Not enough memory to read lookup table!");
600 ret = WIMLIB_ERR_NOMEM;
601 goto out_free_cur_entry;
603 offset_save = reshdr.offset_in_wim;
604 size_save = reshdr.size_in_wim;
605 wim_res_hdr_to_spec(&reshdr, wim, cur_rspec);
606 } else if (is_zero_hash(cur_entry->hash)) {
607 /* Found the resource specification for the run. */
608 cur_rspec->offset_in_wim = reshdr.offset_in_wim;
609 cur_rspec->size_in_wim = reshdr.size_in_wim;
610 cur_rspec->flags = reshdr.flags;
611 DEBUG("Full run is %"PRIu64" compressed bytes "
612 "at file offset %"PRIu64" (flags 0x%02x)",
613 cur_rspec->size_in_wim,
614 cur_rspec->offset_in_wim,
616 free_lookup_table_entry(cur_entry);
619 /* Continuing the run with another stream. */
620 DEBUG("Continuing concat run with stream: "
621 "%"PRIu64" uncompressed bytes @ resource offset %"PRIu64")",
622 reshdr.size_in_wim, reshdr.offset_in_wim);
623 cur_rspec->uncompressed_size += reshdr.size_in_wim;
626 lte_bind_wim_resource_spec(cur_entry, cur_rspec);
627 if (reshdr.flags & WIM_RESHDR_FLAG_CONCAT) {
628 /* In concatenation runs, the offset field is used for
629 * in-resource offset, not the in-WIM offset, and the
630 * size field is used for the uncompressed size, not the
631 * compressed size. */
632 cur_entry->offset_in_res = reshdr.offset_in_wim;
633 cur_entry->size = reshdr.size_in_wim;
634 cur_entry->flags = reshdr.flags;
636 /* These may be overwritten in validate_resource() if
637 * the run turns out to be a concatenation. */
638 cur_entry->offset_in_res = 0;
639 cur_entry->size = reshdr.uncompressed_size;
640 cur_entry->flags = reshdr.flags;
643 if (is_zero_hash(cur_entry->hash)) {
644 WARNING("The WIM lookup table contains an entry with a "
645 "SHA1 message digest of all 0's (ignoring it)");
646 free_lookup_table_entry(cur_entry);
650 if (cur_entry->flags & WIM_RESHDR_FLAG_METADATA) {
651 /* Lookup table entry for a metadata resource */
652 if (cur_entry->refcnt != 1) {
653 /* Metadata entries with no references must be
654 * ignored. See for example the WinPE WIMs from
656 if (cur_entry->refcnt == 0) {
657 free_lookup_table_entry(cur_entry);
660 if (wimlib_print_errors) {
661 ERROR("Found metadata resource with refcnt != 1:");
662 print_lookup_table_entry(cur_entry, stderr);
664 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
665 goto out_free_cur_entry;
668 if (wim->hdr.part_number != 1) {
669 WARNING("Ignoring metadata resource found in a "
670 "non-first part of the split WIM");
671 free_lookup_table_entry(cur_entry);
674 if (wim->current_image == wim->hdr.image_count) {
675 WARNING("The WIM header says there are %u images "
676 "in the WIM, but we found more metadata "
677 "resources than this (ignoring the extra)",
678 wim->hdr.image_count);
679 free_lookup_table_entry(cur_entry);
683 /* Notice very carefully: We are assigning the metadata
684 * resources in the exact order mirrored by their lookup
685 * table entries on disk, which is the behavior of
686 * Microsoft's software. In particular, this overrides
687 * the actual locations of the metadata resources
688 * themselves in the WIM file as well as any information
689 * written in the XML data. */
690 DEBUG("Found metadata resource for image %u at "
692 wim->current_image + 1,
693 cur_entry->rspec->offset_in_wim);
695 wim->current_image++]->metadata_lte = cur_entry;
697 /* Lookup table entry for a stream that is not a
698 * metadata resource */
699 duplicate_entry = lookup_resource(table, cur_entry->hash);
700 if (duplicate_entry) {
701 if (wimlib_print_errors) {
702 WARNING("The WIM lookup table contains two entries with the "
703 "same SHA1 message digest!");
704 WARNING("The first entry is:");
705 print_lookup_table_entry(duplicate_entry, stderr);
706 WARNING("The second entry is:");
707 print_lookup_table_entry(cur_entry, stderr);
709 free_lookup_table_entry(cur_entry);
712 lookup_table_insert(table, cur_entry);
717 if (cur_rspec != NULL) {
718 ret = validate_resource(cur_rspec, offset_save, size_save);
720 goto out_free_cur_entry;
723 if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
724 WARNING("The header of \"%"TS"\" says there are %u images in\n"
725 " the WIM, but we only found %d metadata resources! Acting as if\n"
726 " the header specified only %d images instead.",
727 wim->filename, wim->hdr.image_count,
728 wim->current_image, wim->current_image);
729 for (int i = wim->current_image; i < wim->hdr.image_count; i++)
730 put_image_metadata(wim->image_metadata[i], NULL);
731 wim->hdr.image_count = wim->current_image;
733 DEBUG("Done reading lookup table.");
734 wim->lookup_table = table;
740 out_free_lookup_table:
741 free_lookup_table(table);
745 wim->current_image = 0;
751 write_wim_lookup_table_entry(const struct wim_lookup_table_entry *lte,
752 struct wim_lookup_table_entry_disk *disk_entry,
755 put_wim_reshdr(<e->out_reshdr, &disk_entry->reshdr);
756 disk_entry->part_number = cpu_to_le16(part_number);
757 disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
758 copy_hash(disk_entry->hash, lte->hash);
762 write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
763 struct filedes *out_fd,
765 struct wim_reshdr *out_reshdr,
766 int write_resource_flags,
767 struct wimlib_lzx_context **comp_ctx)
770 struct wim_lookup_table_entry *lte;
771 struct wim_lookup_table_entry_disk *table_buf;
772 struct wim_lookup_table_entry_disk *table_buf_ptr;
776 list_for_each_entry(lte, stream_list, lookup_table_list)
777 table_size += sizeof(struct wim_lookup_table_entry_disk);
779 DEBUG("Writing WIM lookup table (size=%zu, offset=%"PRIu64")",
780 table_size, out_fd->offset);
782 table_buf = MALLOC(table_size);
784 ERROR("Failed to allocate %zu bytes for temporary lookup table",
786 return WIMLIB_ERR_NOMEM;
788 table_buf_ptr = table_buf;
789 list_for_each_entry(lte, stream_list, lookup_table_list)
790 write_wim_lookup_table_entry(lte, table_buf_ptr++, part_number);
792 /* Write the lookup table uncompressed. Although wimlib can handle a
793 * compressed lookup table, MS software cannot. */
794 ret = write_wim_resource_from_buffer(table_buf,
796 WIM_RESHDR_FLAG_METADATA,
798 WIMLIB_COMPRESSION_TYPE_NONE,
802 write_resource_flags,
805 DEBUG("ret=%d", ret);
810 append_lookup_table_entry(struct wim_lookup_table_entry *lte, void *_list)
812 /* Lookup table entries with 'out_refcnt' == 0 correspond to streams not
813 * written and not present in the resulting WIM file, and should not be
814 * included in the lookup table.
816 * Lookup table entries marked as filtered (EXTERNAL_WIM) with
817 * 'out_refcnt != 0' were referenced as part of the logical write but
818 * correspond to streams that were not in fact written, and should not
819 * be included in the lookup table.
821 * Lookup table entries marked as filtered (SAME_WIM) with 'out_refcnt
822 * != 0' were referenced as part of the logical write but correspond to
823 * streams that were not in fact written, but nevertheless were already
824 * present in the WIM being overwritten in-place. These entries must be
825 * included in the lookup table, and the resource information to write
826 * needs to be copied from the resource information read originally.
828 if (lte->out_refcnt != 0 && !(lte->filtered & FILTERED_EXTERNAL_WIM)) {
829 if (lte->filtered & FILTERED_SAME_WIM)
830 wim_res_spec_to_hdr(lte->rspec, <e->out_reshdr);
831 list_add_tail(<e->lookup_table_list, (struct list_head*)_list);
837 write_wim_lookup_table(WIMStruct *wim, int image, int write_flags,
838 struct wim_reshdr *out_reshdr,
839 struct list_head *stream_list_override)
841 int write_resource_flags;
842 struct list_head _stream_list;
843 struct list_head *stream_list;
845 if (stream_list_override) {
846 stream_list = stream_list_override;
848 stream_list = &_stream_list;
849 INIT_LIST_HEAD(stream_list);
852 if (!(write_flags & WIMLIB_WRITE_FLAG_NO_METADATA)) {
856 if (image == WIMLIB_ALL_IMAGES) {
858 end_image = wim->hdr.image_count;
864 /* Push metadata resource lookup table entries onto the front of
865 * the list in reverse order, so that they're written in order.
867 for (int i = end_image; i >= start_image; i--) {
868 struct wim_lookup_table_entry *metadata_lte;
870 metadata_lte = wim->image_metadata[i - 1]->metadata_lte;
871 metadata_lte->out_refcnt = 1;
872 metadata_lte->out_reshdr.flags |= WIM_RESHDR_FLAG_METADATA;
873 list_add(&metadata_lte->lookup_table_list, stream_list);
877 /* Append additional lookup table entries that need to be written, with
878 * some special handling for streams that have been marked as filtered.
880 if (!stream_list_override) {
881 for_lookup_table_entry(wim->lookup_table,
882 append_lookup_table_entry, stream_list);
885 write_resource_flags = 0;
886 if (write_flags & WIMLIB_WRITE_FLAG_PIPABLE)
887 write_resource_flags |= WIMLIB_WRITE_RESOURCE_FLAG_PIPABLE;
888 return write_wim_lookup_table_from_stream_list(stream_list,
890 wim->hdr.part_number,
892 write_resource_flags,
898 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
900 lte->real_refcnt = 0;
905 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
912 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
914 if (lte->extracted_file != NULL) {
915 FREE(lte->extracted_file);
916 lte->extracted_file = NULL;
922 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
930 tprintf(T("Uncompressed size = %"PRIu64" bytes\n"),
932 if (lte_is_partial(lte)) {
933 tprintf(T("Offset = %"PRIu64" bytes\n"),
936 tprintf(T("Raw uncompressed size = %"PRIu64" bytes\n"),
937 lte->rspec->uncompressed_size);
939 tprintf(T("Raw compressed size = %"PRIu64" bytes\n"),
940 lte->rspec->size_in_wim);
942 tprintf(T("Raw offset = %"PRIu64" bytes\n"),
943 lte->rspec->offset_in_wim);
944 } else if (lte->resource_location == RESOURCE_IN_WIM) {
945 tprintf(T("Compressed size = %"PRIu64" bytes\n"),
946 lte->rspec->size_in_wim);
948 tprintf(T("Offset = %"PRIu64" bytes\n"),
949 lte->rspec->offset_in_wim);
952 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
955 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
956 lte->back_inode, lte->back_stream_id);
958 tfprintf(out, T("Hash = 0x"));
959 print_hash(lte->hash, out);
963 tfprintf(out, T("Flags = "));
964 u8 flags = lte->flags;
965 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
966 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
967 if (flags & WIM_RESHDR_FLAG_FREE)
968 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
969 if (flags & WIM_RESHDR_FLAG_METADATA)
970 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
971 if (flags & WIM_RESHDR_FLAG_SPANNED)
972 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
973 if (flags & WIM_RESHDR_FLAG_CONCAT)
974 tfputs(T("WIM_RESHDR_FLAG_CONCAT, "), out);
976 switch (lte->resource_location) {
977 case RESOURCE_IN_WIM:
978 if (lte->rspec->wim->filename) {
979 tfprintf(out, T("WIM file = `%"TS"'\n"),
980 lte->rspec->wim->filename);
984 case RESOURCE_WIN32_ENCRYPTED:
986 case RESOURCE_IN_FILE_ON_DISK:
987 tfprintf(out, T("File on Disk = `%"TS"'\n"),
991 case RESOURCE_IN_STAGING_FILE:
992 tfprintf(out, T("Staging File = `%"TS"'\n"),
993 lte->staging_file_name);
1003 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
1004 struct wimlib_resource_entry *wentry)
1006 memset(wentry, 0, sizeof(*wentry));
1008 wentry->uncompressed_size = lte->size;
1009 if (lte->resource_location == RESOURCE_IN_WIM) {
1010 wentry->part_number = lte->rspec->wim->hdr.part_number;
1011 if (lte_is_partial(lte)) {
1012 wentry->compressed_size = 0;
1013 wentry->offset = lte->offset_in_res;
1015 wentry->compressed_size = lte->rspec->size_in_wim;
1016 wentry->offset = lte->rspec->offset_in_wim;
1018 wentry->raw_resource_offset_in_wim = lte->rspec->offset_in_wim;
1019 wentry->raw_resource_uncompressed_size = lte->rspec->uncompressed_size;
1020 wentry->raw_resource_compressed_size = lte->rspec->size_in_wim;
1022 copy_hash(wentry->sha1_hash, lte->hash);
1023 wentry->reference_count = lte->refcnt;
1024 wentry->is_compressed = (lte->flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
1025 wentry->is_metadata = (lte->flags & WIM_RESHDR_FLAG_METADATA) != 0;
1026 wentry->is_free = (lte->flags & WIM_RESHDR_FLAG_FREE) != 0;
1027 wentry->is_spanned = (lte->flags & WIM_RESHDR_FLAG_SPANNED) != 0;
1028 wentry->is_partial = lte_is_partial(lte);
1031 struct iterate_lte_context {
1032 wimlib_iterate_lookup_table_callback_t cb;
1037 do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
1039 struct iterate_lte_context *ctx = _ctx;
1040 struct wimlib_resource_entry entry;
1042 lte_to_wimlib_resource_entry(lte, &entry);
1043 return (*ctx->cb)(&entry, ctx->user_ctx);
1046 /* API function documented in wimlib.h */
1048 wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
1049 wimlib_iterate_lookup_table_callback_t cb,
1052 struct iterate_lte_context ctx = {
1054 .user_ctx = user_ctx,
1056 if (wim->hdr.part_number == 1) {
1058 for (int i = 0; i < wim->hdr.image_count; i++) {
1059 ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
1065 return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
1068 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
1069 * lookup table, or NULL if there is none. */
1070 struct wim_lookup_table_entry *
1071 lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
1074 struct wim_lookup_table_entry *lte;
1075 struct hlist_node *pos;
1077 wimlib_assert(table != NULL);
1078 wimlib_assert(hash != NULL);
1080 i = *(size_t*)hash % table->capacity;
1081 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
1082 if (hashes_equal(hash, lte->hash))
1089 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
1090 * given a path name.
1092 * This is only for pre-resolved inodes.
1095 wim_pathname_to_stream(WIMStruct *wim,
1098 struct wim_dentry **dentry_ret,
1099 struct wim_lookup_table_entry **lte_ret,
1100 u16 *stream_idx_ret)
1102 struct wim_dentry *dentry;
1103 struct wim_lookup_table_entry *lte;
1105 const tchar *stream_name = NULL;
1106 struct wim_inode *inode;
1109 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
1110 stream_name = path_stream_name(path);
1112 p = (tchar*)stream_name - 1;
1117 dentry = get_dentry(wim, path);
1123 inode = dentry->d_inode;
1125 if (!inode->i_resolved)
1126 if (inode_resolve_ltes(inode, wim->lookup_table, false))
1129 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
1130 && inode_is_directory(inode))
1134 struct wim_ads_entry *ads_entry;
1136 ads_entry = inode_get_ads_entry(inode, stream_name,
1139 stream_idx = ads_idx + 1;
1140 lte = ads_entry->lte;
1146 lte = inode_unnamed_stream_resolved(inode, &stream_idx);
1150 *dentry_ret = dentry;
1154 *stream_idx_ret = stream_idx;
1160 resource_not_found_error(const struct wim_inode *inode, const u8 *hash)
1162 if (wimlib_print_errors) {
1163 ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
1164 tfprintf(stderr, T(" SHA-1 message digest of missing resource:\n "));
1165 print_hash(hash, stderr);
1166 tputc(T('\n'), stderr);
1168 return WIMLIB_ERR_RESOURCE_NOT_FOUND;
1172 * Resolve an inode's lookup table entries.
1174 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
1175 * lookup table) with pointers directly to the lookup table entries.
1177 * If @force is %false:
1178 * If any needed SHA1 message digests are not found in the lookup table,
1179 * WIMLIB_ERR_RESOURCE_NOT_FOUND is returned and the inode is left
1181 * If @force is %true:
1182 * If any needed SHA1 message digests are not found in the lookup table,
1183 * new entries are allocated and inserted into the lookup table.
1186 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table,
1191 if (!inode->i_resolved) {
1192 struct wim_lookup_table_entry *lte, *ads_lte;
1194 /* Resolve the default file stream */
1196 hash = inode->i_hash;
1197 if (!is_zero_hash(hash)) {
1198 lte = lookup_resource(table, hash);
1201 lte = new_lookup_table_entry();
1203 return WIMLIB_ERR_NOMEM;
1204 copy_hash(lte->hash, hash);
1205 lookup_table_insert(table, lte);
1207 goto resource_not_found;
1212 /* Resolve the alternate data streams */
1213 struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
1214 for (u16 i = 0; i < inode->i_num_ads; i++) {
1215 struct wim_ads_entry *cur_entry;
1218 cur_entry = &inode->i_ads_entries[i];
1219 hash = cur_entry->hash;
1220 if (!is_zero_hash(hash)) {
1221 ads_lte = lookup_resource(table, hash);
1224 ads_lte = new_lookup_table_entry();
1226 return WIMLIB_ERR_NOMEM;
1227 copy_hash(ads_lte->hash, hash);
1228 lookup_table_insert(table, ads_lte);
1230 goto resource_not_found;
1234 ads_ltes[i] = ads_lte;
1237 for (u16 i = 0; i < inode->i_num_ads; i++)
1238 inode->i_ads_entries[i].lte = ads_ltes[i];
1239 inode->i_resolved = 1;
1244 return resource_not_found_error(inode, hash);
1248 inode_unresolve_ltes(struct wim_inode *inode)
1250 if (inode->i_resolved) {
1252 copy_hash(inode->i_hash, inode->i_lte->hash);
1254 zero_out_hash(inode->i_hash);
1256 for (u16 i = 0; i < inode->i_num_ads; i++) {
1257 if (inode->i_ads_entries[i].lte)
1258 copy_hash(inode->i_ads_entries[i].hash,
1259 inode->i_ads_entries[i].lte->hash);
1261 zero_out_hash(inode->i_ads_entries[i].hash);
1263 inode->i_resolved = 0;
1268 * Returns the lookup table entry for stream @stream_idx of the inode, where
1269 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
1270 * corresponds to an alternate data stream.
1272 * This works for both resolved and un-resolved inodes.
1274 struct wim_lookup_table_entry *
1275 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
1276 const struct wim_lookup_table *table)
1278 if (inode->i_resolved)
1279 return inode_stream_lte_resolved(inode, stream_idx);
1281 return inode_stream_lte_unresolved(inode, stream_idx, table);
1284 struct wim_lookup_table_entry *
1285 inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
1287 wimlib_assert(inode->i_resolved);
1288 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1289 if (inode_stream_name_nbytes(inode, i) == 0 &&
1290 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
1292 *stream_idx_ret = i;
1293 return inode_stream_lte_resolved(inode, i);
1296 *stream_idx_ret = 0;
1300 struct wim_lookup_table_entry *
1301 inode_unnamed_lte_resolved(const struct wim_inode *inode)
1304 return inode_unnamed_stream_resolved(inode, &stream_idx);
1307 struct wim_lookup_table_entry *
1308 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
1309 const struct wim_lookup_table *table)
1311 wimlib_assert(!inode->i_resolved);
1312 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1313 if (inode_stream_name_nbytes(inode, i) == 0 &&
1314 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
1316 return inode_stream_lte_unresolved(inode, i, table);
1322 /* Return the lookup table entry for the unnamed data stream of an inode, or
1323 * NULL if there is none.
1325 * You'd think this would be easier than it actually is, since the unnamed data
1326 * stream should be the one referenced from the inode itself. Alas, if there
1327 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
1328 * data stream in one of the alternate data streams instead of inside the WIM
1329 * dentry itself. So we need to check the alternate data streams too.
1331 * Also, note that a dentry may appear to have more than one unnamed stream, but
1332 * if the SHA1 message digest is all 0's then the corresponding stream does not
1333 * really "count" (this is the case for the inode's own file stream when the
1334 * file stream that should be there is actually in one of the alternate stream
1335 * entries.). This is despite the fact that we may need to extract such a
1336 * missing entry as an empty file or empty named data stream.
1338 struct wim_lookup_table_entry *
1339 inode_unnamed_lte(const struct wim_inode *inode,
1340 const struct wim_lookup_table *table)
1342 if (inode->i_resolved)
1343 return inode_unnamed_lte_resolved(inode);
1345 return inode_unnamed_lte_unresolved(inode, table);
1348 /* Returns the SHA1 message digest of the unnamed data stream of a WIM inode, or
1349 * 'zero_hash' if the unnamed data stream is missing has all zeroes in its SHA1
1350 * message digest field. */
1352 inode_unnamed_stream_hash(const struct wim_inode *inode)
1356 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1357 if (inode_stream_name_nbytes(inode, i) == 0) {
1358 hash = inode_stream_hash(inode, i);
1359 if (!is_zero_hash(hash))
1366 struct wim_lookup_table_entry **
1367 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1369 wimlib_assert(lte->unhashed);
1370 struct wim_inode *inode = lte->back_inode;
1371 u32 stream_id = lte->back_stream_id;
1373 return &inode->i_lte;
1375 for (u16 i = 0; i < inode->i_num_ads; i++)
1376 if (inode->i_ads_entries[i].stream_id == stream_id)
1377 return &inode->i_ads_entries[i].lte;
1382 /* Calculate the SHA1 message digest of a stream and move it from the list of
1383 * unhashed streams to the stream lookup table, possibly joining it with an
1384 * existing lookup table entry for an identical stream.
1386 * @lte: An unhashed lookup table entry.
1387 * @lookup_table: Lookup table for the WIM.
1388 * @lte_ret: On success, write a pointer to the resulting lookup table
1389 * entry to this location. This will be the same as @lte
1390 * if it was inserted into the lookup table, or different if
1391 * a duplicate stream was found.
1393 * Returns 0 on success; nonzero if there is an error reading the stream.
1396 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1397 struct wim_lookup_table *lookup_table,
1398 struct wim_lookup_table_entry **lte_ret)
1401 struct wim_lookup_table_entry *duplicate_lte;
1402 struct wim_lookup_table_entry **back_ptr;
1404 wimlib_assert(lte->unhashed);
1406 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1407 * union with the SHA1 message digest and will no longer be valid once
1408 * the SHA1 has been calculated. */
1409 back_ptr = retrieve_lte_pointer(lte);
1411 ret = sha1_stream(lte);
1415 /* Look for a duplicate stream */
1416 duplicate_lte = lookup_resource(lookup_table, lte->hash);
1417 list_del(<e->unhashed_list);
1418 if (duplicate_lte) {
1419 /* We have a duplicate stream. Transfer the reference counts
1420 * from this stream to the duplicate, update the reference to
1421 * this stream (in an inode or ads_entry) to point to the
1422 * duplicate, then free this stream. */
1423 wimlib_assert(!(duplicate_lte->unhashed));
1424 duplicate_lte->refcnt += lte->refcnt;
1425 duplicate_lte->out_refcnt += lte->out_refcnt;
1426 *back_ptr = duplicate_lte;
1427 free_lookup_table_entry(lte);
1428 lte = duplicate_lte;
1430 /* No duplicate stream, so we need to insert
1431 * this stream into the lookup table and treat
1432 * it as a hashed stream. */
1433 lookup_table_insert(lookup_table, lte);
1442 lte_clone_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1444 struct wim_lookup_table *lookup_table = _lookup_table;
1446 if (lookup_resource(lookup_table, lte->hash))
1447 return 0; /* Resource already present. */
1449 lte = clone_lookup_table_entry(lte);
1451 return WIMLIB_ERR_NOMEM;
1452 lte->out_refcnt = 1;
1453 lookup_table_insert(lookup_table, lte);
1458 lte_delete_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1460 struct wim_lookup_table *lookup_table = _lookup_table;
1462 if (lte->out_refcnt) {
1463 lookup_table_unlink(lookup_table, lte);
1464 free_lookup_table_entry(lte);
1469 /* API function documented in wimlib.h */
1471 wimlib_reference_resources(WIMStruct *wim,
1472 WIMStruct **resource_wims, unsigned num_resource_wims,
1479 return WIMLIB_ERR_INVALID_PARAM;
1481 if (num_resource_wims != 0 && resource_wims == NULL)
1482 return WIMLIB_ERR_INVALID_PARAM;
1484 for (i = 0; i < num_resource_wims; i++)
1485 if (resource_wims[i] == NULL)
1486 return WIMLIB_ERR_INVALID_PARAM;
1488 for_lookup_table_entry(wim->lookup_table, lte_zero_out_refcnt, NULL);
1490 for (i = 0; i < num_resource_wims; i++) {
1491 ret = for_lookup_table_entry(resource_wims[i]->lookup_table,
1500 for_lookup_table_entry(wim->lookup_table, lte_delete_if_new,
1506 reference_resource_paths(WIMStruct *wim,
1507 const tchar * const *resource_wimfiles,
1508 unsigned num_resource_wimfiles,
1511 wimlib_progress_func_t progress_func)
1513 WIMStruct **resource_wims;
1517 resource_wims = CALLOC(num_resource_wimfiles, sizeof(resource_wims[0]));
1519 return WIMLIB_ERR_NOMEM;
1521 for (i = 0; i < num_resource_wimfiles; i++) {
1522 DEBUG("Referencing resources from path \"%"TS"\"",
1523 resource_wimfiles[i]);
1524 ret = wimlib_open_wim(resource_wimfiles[i], open_flags,
1525 &resource_wims[i], progress_func);
1527 goto out_free_resource_wims;
1530 ret = wimlib_reference_resources(wim, resource_wims,
1531 num_resource_wimfiles, ref_flags);
1533 goto out_free_resource_wims;
1535 for (i = 0; i < num_resource_wimfiles; i++)
1536 list_add_tail(&resource_wims[i]->subwim_node, &wim->subwims);
1539 goto out_free_array;
1541 out_free_resource_wims:
1542 for (i = 0; i < num_resource_wimfiles; i++)
1543 wimlib_free(resource_wims[i]);
1545 FREE(resource_wims);
1550 reference_resource_glob(WIMStruct *wim, const tchar *refglob,
1551 int ref_flags, int open_flags,
1552 wimlib_progress_func_t progress_func)
1557 /* Note: glob() is replaced in Windows native builds. */
1558 ret = tglob(refglob, GLOB_ERR | GLOB_NOSORT, NULL, &globbuf);
1560 if (ret == GLOB_NOMATCH) {
1561 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ERR_ON_NOMATCH) {
1562 ERROR("Found no files for glob \"%"TS"\"", refglob);
1563 return WIMLIB_ERR_GLOB_HAD_NO_MATCHES;
1565 return reference_resource_paths(wim,
1573 ERROR_WITH_ERRNO("Failed to process glob \"%"TS"\"", refglob);
1574 if (ret == GLOB_NOSPACE)
1575 return WIMLIB_ERR_NOMEM;
1577 return WIMLIB_ERR_READ;
1581 ret = reference_resource_paths(wim,
1582 (const tchar * const *)globbuf.gl_pathv,
1591 /* API function documented in wimlib.h */
1593 wimlib_reference_resource_files(WIMStruct *wim,
1594 const tchar * const * resource_wimfiles_or_globs,
1598 wimlib_progress_func_t progress_func)
1603 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ENABLE) {
1604 for (i = 0; i < count; i++) {
1605 ret = reference_resource_glob(wim,
1606 resource_wimfiles_or_globs[i],
1615 return reference_resource_paths(wim, resource_wimfiles_or_globs,
1617 open_flags, progress_func);