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));
80 BUILD_BUG_ON(RESOURCE_NONEXISTENT != 0);
81 BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
83 ERROR("Out of memory (tried to allocate %zu bytes for "
84 "lookup table entry)",
85 sizeof(struct wim_lookup_table_entry));
90 struct wim_lookup_table_entry *
91 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
93 struct wim_lookup_table_entry *new;
95 new = memdup(old, sizeof(struct wim_lookup_table_entry));
99 new->extracted_file = NULL;
100 switch (new->resource_location) {
101 case RESOURCE_IN_FILE_ON_DISK:
103 case RESOURCE_WIN32_ENCRYPTED:
106 case RESOURCE_IN_STAGING_FILE:
107 BUILD_BUG_ON((void*)&old->file_on_disk !=
108 (void*)&old->staging_file_name);
110 new->file_on_disk = TSTRDUP(old->file_on_disk);
111 if (!new->file_on_disk)
114 case RESOURCE_IN_ATTACHED_BUFFER:
115 new->attached_buffer = memdup(old->attached_buffer,
116 wim_resource_size(old));
117 if (!new->attached_buffer)
121 case RESOURCE_IN_NTFS_VOLUME:
123 struct ntfs_location *loc;
124 loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
128 loc->stream_name = NULL;
130 loc->path = STRDUP(old->ntfs_loc->path);
133 if (loc->stream_name_nchars) {
134 loc->stream_name = memdup(old->ntfs_loc->stream_name,
135 loc->stream_name_nchars * 2);
136 if (!loc->stream_name)
147 free_lookup_table_entry(new);
152 free_lookup_table_entry(struct wim_lookup_table_entry *lte)
155 switch (lte->resource_location) {
156 case RESOURCE_IN_FILE_ON_DISK:
158 case RESOURCE_WIN32_ENCRYPTED:
161 case RESOURCE_IN_STAGING_FILE:
162 BUILD_BUG_ON((void*)<e->file_on_disk !=
163 (void*)<e->staging_file_name);
165 case RESOURCE_IN_ATTACHED_BUFFER:
166 BUILD_BUG_ON((void*)<e->file_on_disk !=
167 (void*)<e->attached_buffer);
168 FREE(lte->file_on_disk);
171 case RESOURCE_IN_NTFS_VOLUME:
173 FREE(lte->ntfs_loc->path);
174 FREE(lte->ntfs_loc->stream_name);
187 do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
189 free_lookup_table_entry(entry);
195 free_lookup_table(struct wim_lookup_table *table)
197 DEBUG2("Freeing lookup table");
200 for_lookup_table_entry(table,
201 do_free_lookup_table_entry,
210 * Inserts an entry into the lookup table.
212 * @table: A pointer to the lookup table.
213 * @lte: A pointer to the entry to insert.
216 lookup_table_insert(struct wim_lookup_table *table,
217 struct wim_lookup_table_entry *lte)
219 size_t i = lte->hash_short % table->capacity;
220 hlist_add_head(<e->hash_list, &table->array[i]);
222 /* XXX Make the table grow when too many entries have been inserted. */
223 table->num_entries++;
227 finalize_lte(struct wim_lookup_table_entry *lte)
230 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
231 unlink(lte->staging_file_name);
232 list_del(<e->unhashed_list);
235 free_lookup_table_entry(lte);
238 /* Decrements the reference count for the lookup table entry @lte. If its
239 * reference count reaches 0, it is unlinked from the lookup table. If,
240 * furthermore, the entry has no opened file descriptors associated with it, the
243 lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
244 struct wim_lookup_table *table)
246 wimlib_assert(lte != NULL);
247 wimlib_assert(lte->refcnt != 0);
248 if (--lte->refcnt == 0) {
250 list_del(<e->unhashed_list);
252 lookup_table_unlink(table, lte);
254 if (lte->num_opened_fds == 0)
262 lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
264 if (lte->num_opened_fds != 0)
265 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
270 /* Calls a function on all the entries in the WIM lookup table. Stop early and
271 * return nonzero if any call to the function returns nonzero. */
273 for_lookup_table_entry(struct wim_lookup_table *table,
274 int (*visitor)(struct wim_lookup_table_entry *, void *),
277 struct wim_lookup_table_entry *lte;
278 struct hlist_node *pos, *tmp;
281 for (size_t i = 0; i < table->capacity; i++) {
282 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
285 wimlib_assert2(!(lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA));
286 ret = visitor(lte, arg);
294 /* qsort() callback that sorts streams (represented by `struct
295 * wim_lookup_table_entry's) into an order optimized for reading and writing.
297 * Sorting is done primarily by resource location, then secondarily by a
298 * per-resource location order. For example, resources in WIM files are sorted
299 * primarily by part number, then secondarily by offset, as to implement optimal
300 * reading of either a standalone or split WIM. */
302 cmp_streams_by_sequential_order(const void *p1, const void *p2)
304 const struct wim_lookup_table_entry *lte1, *lte2;
307 lte1 = *(const struct wim_lookup_table_entry**)p1;
308 lte2 = *(const struct wim_lookup_table_entry**)p2;
310 v = (int)lte1->resource_location - (int)lte2->resource_location;
312 /* Different resource locations? */
316 switch (lte1->resource_location) {
317 case RESOURCE_IN_WIM:
319 /* Different (possibly split) WIMs? */
320 if (lte1->wim != lte2->wim) {
321 v = memcmp(lte1->wim->hdr.guid, lte2->wim->hdr.guid,
327 /* Different part numbers in the same WIM? */
328 v = (int)lte1->wim->hdr.part_number - (int)lte2->wim->hdr.part_number;
332 /* Compare by offset. */
333 if (lte1->resource_entry.offset < lte2->resource_entry.offset)
335 else if (lte1->resource_entry.offset > lte2->resource_entry.offset)
338 case RESOURCE_IN_FILE_ON_DISK:
340 case RESOURCE_IN_STAGING_FILE:
343 case RESOURCE_WIN32_ENCRYPTED:
345 /* Compare files by path: just a heuristic that will place files
346 * in the same directory next to each other. */
347 return tstrcmp(lte1->file_on_disk, lte2->file_on_disk);
349 case RESOURCE_IN_NTFS_VOLUME:
350 return tstrcmp(lte1->ntfs_loc->path, lte2->ntfs_loc->path);
353 /* No additional sorting order defined for this resource
354 * location (e.g. RESOURCE_IN_ATTACHED_BUFFER); simply compare
355 * everything equal to each other. */
361 sort_stream_list_by_sequential_order(struct list_head *stream_list,
362 size_t list_head_offset)
364 struct list_head *cur;
365 struct wim_lookup_table_entry **array;
368 size_t num_streams = 0;
370 list_for_each(cur, stream_list)
373 array_size = num_streams * sizeof(array[0]);
374 array = MALLOC(array_size);
376 return WIMLIB_ERR_NOMEM;
377 cur = stream_list->next;
378 for (i = 0; i < num_streams; i++) {
379 array[i] = (struct wim_lookup_table_entry*)((u8*)cur -
384 qsort(array, num_streams, sizeof(array[0]),
385 cmp_streams_by_sequential_order);
387 INIT_LIST_HEAD(stream_list);
388 for (i = 0; i < num_streams; i++) {
389 list_add_tail((struct list_head*)
390 ((u8*)array[i] + list_head_offset),
399 add_lte_to_array(struct wim_lookup_table_entry *lte,
402 struct wim_lookup_table_entry ***pp = _pp;
407 /* Iterate through the lookup table entries, but first sort them by stream
408 * offset in the WIM. Caution: this is intended to be used when the stream
409 * offset field has actually been set. */
411 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
412 int (*visitor)(struct wim_lookup_table_entry *,
416 struct wim_lookup_table_entry **lte_array, **p;
417 size_t num_streams = table->num_entries;
420 lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
422 return WIMLIB_ERR_NOMEM;
424 for_lookup_table_entry(table, add_lte_to_array, &p);
426 wimlib_assert(p == lte_array + num_streams);
428 qsort(lte_array, num_streams, sizeof(lte_array[0]),
429 cmp_streams_by_sequential_order);
431 for (size_t i = 0; i < num_streams; i++) {
432 ret = visitor(lte_array[i], arg);
440 /* On-disk format of a WIM lookup table entry (stream entry). */
441 struct wim_lookup_table_entry_disk {
442 /* Location, offset, compression status, and metadata status of the
444 struct resource_entry_disk resource_entry;
446 /* Which part of the split WIM this stream is in; indexed from 1. */
449 /* Reference count of this stream over all WIM images. */
452 /* SHA1 message digest of the uncompressed data of this stream, or
453 * optionally all zeroes if this stream is of zero length. */
454 u8 hash[SHA1_HASH_SIZE];
457 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
460 lte_init_wim(struct wim_lookup_table_entry *lte, WIMStruct *wim)
462 lte->resource_location = RESOURCE_IN_WIM;
464 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
465 lte->compression_type = wim->compression_type;
467 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
469 if (wim_is_pipable(wim))
474 * Reads the lookup table from a WIM file.
476 * Saves lookup table entries for non-metadata streams in a hash table, and
477 * saves the metadata entry for each image in a special per-image location (the
478 * image_metadata array).
481 * WIMLIB_ERR_SUCCESS (0)
482 * WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
483 * WIMLIB_ERR_RESOURCE_NOT_FOUND
486 read_wim_lookup_table(WIMStruct *wim)
491 struct wim_lookup_table *table;
492 struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
495 BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
496 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
498 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
499 wim->hdr.lookup_table_res_entry.offset,
500 wim->hdr.lookup_table_res_entry.size);
502 /* Calculate number of entries in the lookup table. */
503 num_entries = wim->hdr.lookup_table_res_entry.size /
504 sizeof(struct wim_lookup_table_entry_disk);
507 /* Read the lookup table into a buffer. */
508 ret = res_entry_to_data(&wim->hdr.lookup_table_res_entry, wim, &buf);
512 /* Allocate hash table. */
513 table = new_lookup_table(num_entries * 2 + 1);
515 ERROR("Not enough memory to read lookup table.");
516 ret = WIMLIB_ERR_NOMEM;
520 /* Allocate and initalize `struct wim_lookup_table_entry's from the
521 * on-disk lookup table. */
522 wim->current_image = 0;
523 for (i = 0; i < num_entries; i++) {
524 const struct wim_lookup_table_entry_disk *disk_entry =
525 &((const struct wim_lookup_table_entry_disk*)buf)[i];
527 cur_entry = new_lookup_table_entry();
529 ERROR("Not enough memory to read lookup table.");
530 ret = WIMLIB_ERR_NOMEM;
531 goto out_free_lookup_table;
534 cur_entry->wim = wim;
535 cur_entry->resource_location = RESOURCE_IN_WIM;
536 get_resource_entry(&disk_entry->resource_entry, &cur_entry->resource_entry);
537 cur_entry->part_number = le16_to_cpu(disk_entry->part_number);
538 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
539 copy_hash(cur_entry->hash, disk_entry->hash);
540 lte_init_wim(cur_entry, wim);
542 if (cur_entry->part_number != wim->hdr.part_number) {
543 WARNING("A lookup table entry in part %hu of the WIM "
544 "points to part %hu (ignoring it)",
545 wim->hdr.part_number, cur_entry->part_number);
546 free_lookup_table_entry(cur_entry);
550 if (is_zero_hash(cur_entry->hash)) {
551 WARNING("The WIM lookup table contains an entry with a "
552 "SHA1 message digest of all 0's (ignoring it)");
553 free_lookup_table_entry(cur_entry);
557 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
558 && (cur_entry->resource_entry.size !=
559 cur_entry->resource_entry.original_size))
561 if (wimlib_print_errors) {
562 WARNING("Found uncompressed resource with "
563 "original size (%"PRIu64") not the same "
564 "as compressed size (%"PRIu64")",
565 cur_entry->resource_entry.original_size,
566 cur_entry->resource_entry.size);
567 if (cur_entry->resource_entry.original_size) {
568 WARNING("Overriding compressed size with original size.");
569 cur_entry->resource_entry.size =
570 cur_entry->resource_entry.original_size;
572 WARNING("Overriding original size with compressed size");
573 cur_entry->resource_entry.original_size =
574 cur_entry->resource_entry.size;
579 if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
580 /* Lookup table entry for a metadata resource */
581 if (cur_entry->refcnt != 1) {
582 /* Metadata entries with no references must be
583 * ignored. See for example the WinPE WIMs from
585 if (cur_entry->refcnt == 0) {
586 free_lookup_table_entry(cur_entry);
589 if (wimlib_print_errors) {
590 ERROR("Found metadata resource with refcnt != 1:");
591 print_lookup_table_entry(cur_entry, stderr);
593 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
594 goto out_free_cur_entry;
597 if (wim->hdr.part_number != 1) {
598 WARNING("Ignoring metadata resource found in a "
599 "non-first part of the split WIM");
600 free_lookup_table_entry(cur_entry);
603 if (wim->current_image == wim->hdr.image_count) {
604 WARNING("The WIM header says there are %u images "
605 "in the WIM, but we found more metadata "
606 "resources than this (ignoring the extra)",
607 wim->hdr.image_count);
608 free_lookup_table_entry(cur_entry);
612 /* Notice very carefully: We are assigning the metadata
613 * resources in the exact order mirrored by their lookup
614 * table entries on disk, which is the behavior of
615 * Microsoft's software. In particular, this overrides
616 * the actual locations of the metadata resources
617 * themselves in the WIM file as well as any information
618 * written in the XML data. */
619 DEBUG("Found metadata resource for image %u at "
621 wim->current_image + 1,
622 cur_entry->resource_entry.offset);
624 wim->current_image++]->metadata_lte = cur_entry;
626 /* Lookup table entry for a stream that is not a
627 * metadata resource */
628 duplicate_entry = lookup_resource(table, cur_entry->hash);
629 if (duplicate_entry) {
630 if (wimlib_print_errors) {
631 WARNING("The WIM lookup table contains two entries with the "
632 "same SHA1 message digest!");
633 WARNING("The first entry is:");
634 print_lookup_table_entry(duplicate_entry, stderr);
635 WARNING("The second entry is:");
636 print_lookup_table_entry(cur_entry, stderr);
638 free_lookup_table_entry(cur_entry);
641 lookup_table_insert(table, cur_entry);
646 if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
647 WARNING("The header of \"%"TS"\" says there are %u images in\n"
648 " the WIM, but we only found %d metadata resources! Acting as if\n"
649 " the header specified only %d images instead.",
650 wim->filename, wim->hdr.image_count,
651 wim->current_image, wim->current_image);
652 for (int i = wim->current_image; i < wim->hdr.image_count; i++)
653 put_image_metadata(wim->image_metadata[i], NULL);
654 wim->hdr.image_count = wim->current_image;
656 DEBUG("Done reading lookup table.");
657 wim->lookup_table = table;
662 out_free_lookup_table:
663 free_lookup_table(table);
667 wim->current_image = 0;
673 write_wim_lookup_table_entry(const struct wim_lookup_table_entry *lte,
674 struct wim_lookup_table_entry_disk *disk_entry)
676 put_resource_entry(<e->output_resource_entry, &disk_entry->resource_entry);
677 disk_entry->part_number = cpu_to_le16(lte->part_number);
678 disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
679 copy_hash(disk_entry->hash, lte->hash);
683 write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
684 struct filedes *out_fd,
685 struct resource_entry *out_res_entry,
686 int write_resource_flags,
687 struct wimlib_lzx_context **comp_ctx)
690 struct wim_lookup_table_entry *lte;
691 struct wim_lookup_table_entry_disk *table_buf;
692 struct wim_lookup_table_entry_disk *table_buf_ptr;
696 list_for_each_entry(lte, stream_list, lookup_table_list)
697 table_size += sizeof(struct wim_lookup_table_entry_disk);
699 DEBUG("Writing WIM lookup table (size=%zu, offset=%"PRIu64")",
700 table_size, out_fd->offset);
702 table_buf = MALLOC(table_size);
704 ERROR("Failed to allocate %zu bytes for temporary lookup table",
706 return WIMLIB_ERR_NOMEM;
708 table_buf_ptr = table_buf;
709 list_for_each_entry(lte, stream_list, lookup_table_list)
710 write_wim_lookup_table_entry(lte, table_buf_ptr++);
712 /* Write the lookup table uncompressed. Although wimlib can handle a
713 * compressed lookup table, MS software cannot. */
714 ret = write_wim_resource_from_buffer(table_buf,
716 WIM_RESHDR_FLAG_METADATA,
718 WIMLIB_COMPRESSION_TYPE_NONE,
722 write_resource_flags,
725 DEBUG("ret=%d", ret);
730 append_lookup_table_entry(struct wim_lookup_table_entry *lte, void *_list)
732 /* Lookup table entries with 'out_refcnt' == 0 correspond to streams not
733 * written and not present in the resulting WIM file, and should not be
734 * included in the lookup table.
736 * Lookup table entries marked as filtered (EXTERNAL_WIM) with
737 * 'out_refcnt != 0' were referenced as part of the logical write but
738 * correspond to streams that were not in fact written, and should not
739 * be included in the lookup table.
741 * Lookup table entries marked as filtered (SAME_WIM) with 'out_refcnt
742 * != 0' were referenced as part of the logical write but correspond to
743 * streams that were not in fact written, but nevertheless were already
744 * present in the WIM being overwritten in-place. These entries must be
745 * included in the lookup table, and the resource information to write
746 * needs to be copied from the resource information read originally.
748 if (lte->out_refcnt != 0 && !(lte->filtered & FILTERED_EXTERNAL_WIM)) {
749 if (lte->filtered & FILTERED_SAME_WIM) {
750 copy_resource_entry(<e->output_resource_entry,
751 <e->resource_entry);
753 list_add_tail(<e->lookup_table_list, (struct list_head*)_list);
759 write_wim_lookup_table(WIMStruct *wim, int image, int write_flags,
760 struct resource_entry *out_res_entry,
761 struct list_head *stream_list_override)
763 int write_resource_flags;
764 struct list_head _stream_list;
765 struct list_head *stream_list;
767 if (stream_list_override) {
768 stream_list = stream_list_override;
770 stream_list = &_stream_list;
771 INIT_LIST_HEAD(stream_list);
774 if (!(write_flags & WIMLIB_WRITE_FLAG_NO_METADATA)) {
778 if (image == WIMLIB_ALL_IMAGES) {
780 end_image = wim->hdr.image_count;
786 /* Push metadata resource lookup table entries onto the front of
787 * the list in reverse order, so that they're written in order.
789 for (int i = end_image; i >= start_image; i--) {
790 struct wim_lookup_table_entry *metadata_lte;
792 metadata_lte = wim->image_metadata[i - 1]->metadata_lte;
793 metadata_lte->out_refcnt = 1;
794 metadata_lte->part_number = wim->hdr.part_number;
795 metadata_lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
797 list_add(&metadata_lte->lookup_table_list, stream_list);
801 /* Append additional lookup table entries that need to be written, with
802 * some special handling for streams that have been marked as filtered.
804 if (!stream_list_override) {
805 for_lookup_table_entry(wim->lookup_table,
806 append_lookup_table_entry, stream_list);
809 write_resource_flags = 0;
810 if (write_flags & WIMLIB_WRITE_FLAG_PIPABLE)
811 write_resource_flags |= WIMLIB_WRITE_RESOURCE_FLAG_PIPABLE;
812 return write_wim_lookup_table_from_stream_list(stream_list,
815 write_resource_flags,
821 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
823 lte->real_refcnt = 0;
828 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
835 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
837 if (lte->extracted_file != NULL) {
838 FREE(lte->extracted_file);
839 lte->extracted_file = NULL;
845 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
851 tfprintf(out, T("Offset = %"PRIu64" bytes\n"),
852 lte->resource_entry.offset);
854 tfprintf(out, T("Size = %"PRIu64" bytes\n"),
855 (u64)lte->resource_entry.size);
857 tfprintf(out, T("Original size = %"PRIu64" bytes\n"),
858 lte->resource_entry.original_size);
860 tfprintf(out, T("Part Number = %hu\n"), lte->part_number);
861 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
864 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
865 lte->back_inode, lte->back_stream_id);
867 tfprintf(out, T("Hash = 0x"));
868 print_hash(lte->hash, out);
872 tfprintf(out, T("Flags = "));
873 u8 flags = lte->resource_entry.flags;
874 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
875 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
876 if (flags & WIM_RESHDR_FLAG_FREE)
877 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
878 if (flags & WIM_RESHDR_FLAG_METADATA)
879 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
880 if (flags & WIM_RESHDR_FLAG_SPANNED)
881 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
883 switch (lte->resource_location) {
884 case RESOURCE_IN_WIM:
885 if (lte->wim->filename) {
886 tfprintf(out, T("WIM file = `%"TS"'\n"),
891 case RESOURCE_WIN32_ENCRYPTED:
893 case RESOURCE_IN_FILE_ON_DISK:
894 tfprintf(out, T("File on Disk = `%"TS"'\n"),
898 case RESOURCE_IN_STAGING_FILE:
899 tfprintf(out, T("Staging File = `%"TS"'\n"),
900 lte->staging_file_name);
910 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
911 struct wimlib_resource_entry *wentry)
913 wentry->uncompressed_size = lte->resource_entry.original_size;
914 wentry->compressed_size = lte->resource_entry.size;
915 wentry->offset = lte->resource_entry.offset;
916 copy_hash(wentry->sha1_hash, lte->hash);
917 wentry->part_number = lte->part_number;
918 wentry->reference_count = lte->refcnt;
919 wentry->is_compressed = (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
920 wentry->is_metadata = (lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) != 0;
921 wentry->is_free = (lte->resource_entry.flags & WIM_RESHDR_FLAG_FREE) != 0;
922 wentry->is_spanned = (lte->resource_entry.flags & WIM_RESHDR_FLAG_SPANNED) != 0;
925 struct iterate_lte_context {
926 wimlib_iterate_lookup_table_callback_t cb;
931 do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
933 struct iterate_lte_context *ctx = _ctx;
934 struct wimlib_resource_entry entry;
936 lte_to_wimlib_resource_entry(lte, &entry);
937 return (*ctx->cb)(&entry, ctx->user_ctx);
940 /* API function documented in wimlib.h */
942 wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
943 wimlib_iterate_lookup_table_callback_t cb,
946 struct iterate_lte_context ctx = {
948 .user_ctx = user_ctx,
950 if (wim->hdr.part_number == 1) {
952 for (int i = 0; i < wim->hdr.image_count; i++) {
953 ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
959 return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
962 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
963 * lookup table, or NULL if there is none. */
964 struct wim_lookup_table_entry *
965 lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
968 struct wim_lookup_table_entry *lte;
969 struct hlist_node *pos;
971 wimlib_assert(table != NULL);
972 wimlib_assert(hash != NULL);
974 i = *(size_t*)hash % table->capacity;
975 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
976 if (hashes_equal(hash, lte->hash))
983 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
986 * This is only for pre-resolved inodes.
989 wim_pathname_to_stream(WIMStruct *wim,
992 struct wim_dentry **dentry_ret,
993 struct wim_lookup_table_entry **lte_ret,
996 struct wim_dentry *dentry;
997 struct wim_lookup_table_entry *lte;
999 const tchar *stream_name = NULL;
1000 struct wim_inode *inode;
1003 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
1004 stream_name = path_stream_name(path);
1006 p = (tchar*)stream_name - 1;
1011 dentry = get_dentry(wim, path);
1017 inode = dentry->d_inode;
1019 if (!inode->i_resolved)
1020 if (inode_resolve_ltes(inode, wim->lookup_table, false))
1023 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
1024 && inode_is_directory(inode))
1028 struct wim_ads_entry *ads_entry;
1030 ads_entry = inode_get_ads_entry(inode, stream_name,
1033 stream_idx = ads_idx + 1;
1034 lte = ads_entry->lte;
1040 lte = inode_unnamed_stream_resolved(inode, &stream_idx);
1044 *dentry_ret = dentry;
1048 *stream_idx_ret = stream_idx;
1054 resource_not_found_error(const struct wim_inode *inode, const u8 *hash)
1056 if (wimlib_print_errors) {
1057 ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
1058 tfprintf(stderr, T(" SHA-1 message digest of missing resource:\n "));
1059 print_hash(hash, stderr);
1060 tputc(T('\n'), stderr);
1062 return WIMLIB_ERR_RESOURCE_NOT_FOUND;
1066 * Resolve an inode's lookup table entries.
1068 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
1069 * lookup table) with pointers directly to the lookup table entries.
1071 * If @force is %false:
1072 * If any needed SHA1 message digests are not found in the lookup table,
1073 * WIMLIB_ERR_RESOURCE_NOT_FOUND is returned and the inode is left
1075 * If @force is %true:
1076 * If any needed SHA1 message digests are not found in the lookup table,
1077 * new entries are allocated and inserted into the lookup table.
1080 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table,
1085 if (!inode->i_resolved) {
1086 struct wim_lookup_table_entry *lte, *ads_lte;
1088 /* Resolve the default file stream */
1090 hash = inode->i_hash;
1091 if (!is_zero_hash(hash)) {
1092 lte = lookup_resource(table, hash);
1095 lte = new_lookup_table_entry();
1097 return WIMLIB_ERR_NOMEM;
1098 copy_hash(lte->hash, hash);
1099 lookup_table_insert(table, lte);
1101 goto resource_not_found;
1106 /* Resolve the alternate data streams */
1107 struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
1108 for (u16 i = 0; i < inode->i_num_ads; i++) {
1109 struct wim_ads_entry *cur_entry;
1112 cur_entry = &inode->i_ads_entries[i];
1113 hash = cur_entry->hash;
1114 if (!is_zero_hash(hash)) {
1115 ads_lte = lookup_resource(table, hash);
1118 ads_lte = new_lookup_table_entry();
1120 return WIMLIB_ERR_NOMEM;
1121 copy_hash(ads_lte->hash, hash);
1122 lookup_table_insert(table, ads_lte);
1124 goto resource_not_found;
1128 ads_ltes[i] = ads_lte;
1131 for (u16 i = 0; i < inode->i_num_ads; i++)
1132 inode->i_ads_entries[i].lte = ads_ltes[i];
1133 inode->i_resolved = 1;
1138 return resource_not_found_error(inode, hash);
1142 inode_unresolve_ltes(struct wim_inode *inode)
1144 if (inode->i_resolved) {
1146 copy_hash(inode->i_hash, inode->i_lte->hash);
1148 zero_out_hash(inode->i_hash);
1150 for (u16 i = 0; i < inode->i_num_ads; i++) {
1151 if (inode->i_ads_entries[i].lte)
1152 copy_hash(inode->i_ads_entries[i].hash,
1153 inode->i_ads_entries[i].lte->hash);
1155 zero_out_hash(inode->i_ads_entries[i].hash);
1157 inode->i_resolved = 0;
1162 * Returns the lookup table entry for stream @stream_idx of the inode, where
1163 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
1164 * corresponds to an alternate data stream.
1166 * This works for both resolved and un-resolved inodes.
1168 struct wim_lookup_table_entry *
1169 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
1170 const struct wim_lookup_table *table)
1172 if (inode->i_resolved)
1173 return inode_stream_lte_resolved(inode, stream_idx);
1175 return inode_stream_lte_unresolved(inode, stream_idx, table);
1178 struct wim_lookup_table_entry *
1179 inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
1181 wimlib_assert(inode->i_resolved);
1182 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1183 if (inode_stream_name_nbytes(inode, i) == 0 &&
1184 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
1186 *stream_idx_ret = i;
1187 return inode_stream_lte_resolved(inode, i);
1190 *stream_idx_ret = 0;
1194 struct wim_lookup_table_entry *
1195 inode_unnamed_lte_resolved(const struct wim_inode *inode)
1198 return inode_unnamed_stream_resolved(inode, &stream_idx);
1201 struct wim_lookup_table_entry *
1202 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
1203 const struct wim_lookup_table *table)
1205 wimlib_assert(!inode->i_resolved);
1206 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1207 if (inode_stream_name_nbytes(inode, i) == 0 &&
1208 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
1210 return inode_stream_lte_unresolved(inode, i, table);
1216 /* Return the lookup table entry for the unnamed data stream of an inode, or
1217 * NULL if there is none.
1219 * You'd think this would be easier than it actually is, since the unnamed data
1220 * stream should be the one referenced from the inode itself. Alas, if there
1221 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
1222 * data stream in one of the alternate data streams instead of inside the WIM
1223 * dentry itself. So we need to check the alternate data streams too.
1225 * Also, note that a dentry may appear to have more than one unnamed stream, but
1226 * if the SHA1 message digest is all 0's then the corresponding stream does not
1227 * really "count" (this is the case for the inode's own file stream when the
1228 * file stream that should be there is actually in one of the alternate stream
1229 * entries.). This is despite the fact that we may need to extract such a
1230 * missing entry as an empty file or empty named data stream.
1232 struct wim_lookup_table_entry *
1233 inode_unnamed_lte(const struct wim_inode *inode,
1234 const struct wim_lookup_table *table)
1236 if (inode->i_resolved)
1237 return inode_unnamed_lte_resolved(inode);
1239 return inode_unnamed_lte_unresolved(inode, table);
1242 /* Returns the SHA1 message digest of the unnamed data stream of a WIM inode, or
1243 * 'zero_hash' if the unnamed data stream is missing has all zeroes in its SHA1
1244 * message digest field. */
1246 inode_unnamed_stream_hash(const struct wim_inode *inode)
1250 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1251 if (inode_stream_name_nbytes(inode, i) == 0) {
1252 hash = inode_stream_hash(inode, i);
1253 if (!is_zero_hash(hash))
1262 lte_add_stream_size(struct wim_lookup_table_entry *lte, void *total_bytes_p)
1264 *(u64*)total_bytes_p += lte->resource_entry.size;
1269 lookup_table_total_stream_size(struct wim_lookup_table *table)
1272 for_lookup_table_entry(table, lte_add_stream_size, &total_size);
1276 struct wim_lookup_table_entry **
1277 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1279 wimlib_assert(lte->unhashed);
1280 struct wim_inode *inode = lte->back_inode;
1281 u32 stream_id = lte->back_stream_id;
1283 return &inode->i_lte;
1285 for (u16 i = 0; i < inode->i_num_ads; i++)
1286 if (inode->i_ads_entries[i].stream_id == stream_id)
1287 return &inode->i_ads_entries[i].lte;
1292 /* Calculate the SHA1 message digest of a stream and move it from the list of
1293 * unhashed streams to the stream lookup table, possibly joining it with an
1294 * existing lookup table entry for an identical stream.
1296 * @lte: An unhashed lookup table entry.
1297 * @lookup_table: Lookup table for the WIM.
1298 * @lte_ret: On success, write a pointer to the resulting lookup table
1299 * entry to this location. This will be the same as @lte
1300 * if it was inserted into the lookup table, or different if
1301 * a duplicate stream was found.
1303 * Returns 0 on success; nonzero if there is an error reading the stream.
1306 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1307 struct wim_lookup_table *lookup_table,
1308 struct wim_lookup_table_entry **lte_ret)
1311 struct wim_lookup_table_entry *duplicate_lte;
1312 struct wim_lookup_table_entry **back_ptr;
1314 wimlib_assert(lte->unhashed);
1316 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1317 * union with the SHA1 message digest and will no longer be valid once
1318 * the SHA1 has been calculated. */
1319 back_ptr = retrieve_lte_pointer(lte);
1321 ret = sha1_resource(lte);
1325 /* Look for a duplicate stream */
1326 duplicate_lte = lookup_resource(lookup_table, lte->hash);
1327 list_del(<e->unhashed_list);
1328 if (duplicate_lte) {
1329 /* We have a duplicate stream. Transfer the reference counts
1330 * from this stream to the duplicate, update the reference to
1331 * this stream (in an inode or ads_entry) to point to the
1332 * duplicate, then free this stream. */
1333 wimlib_assert(!(duplicate_lte->unhashed));
1334 duplicate_lte->refcnt += lte->refcnt;
1335 duplicate_lte->out_refcnt += lte->out_refcnt;
1336 *back_ptr = duplicate_lte;
1337 free_lookup_table_entry(lte);
1338 lte = duplicate_lte;
1340 /* No duplicate stream, so we need to insert
1341 * this stream into the lookup table and treat
1342 * it as a hashed stream. */
1343 lookup_table_insert(lookup_table, lte);
1352 lte_clone_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1354 struct wim_lookup_table *lookup_table = _lookup_table;
1356 if (lookup_resource(lookup_table, lte->hash))
1357 return 0; /* Resource already present. */
1359 lte = clone_lookup_table_entry(lte);
1361 return WIMLIB_ERR_NOMEM;
1362 lte->out_refcnt = 1;
1363 lookup_table_insert(lookup_table, lte);
1368 lte_delete_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1370 struct wim_lookup_table *lookup_table = _lookup_table;
1372 if (lte->out_refcnt) {
1373 lookup_table_unlink(lookup_table, lte);
1374 free_lookup_table_entry(lte);
1379 /* API function documented in wimlib.h */
1381 wimlib_reference_resources(WIMStruct *wim,
1382 WIMStruct **resource_wims, unsigned num_resource_wims,
1389 return WIMLIB_ERR_INVALID_PARAM;
1391 if (num_resource_wims != 0 && resource_wims == NULL)
1392 return WIMLIB_ERR_INVALID_PARAM;
1394 for (i = 0; i < num_resource_wims; i++)
1395 if (resource_wims[i] == NULL)
1396 return WIMLIB_ERR_INVALID_PARAM;
1398 for_lookup_table_entry(wim->lookup_table, lte_zero_out_refcnt, NULL);
1400 for (i = 0; i < num_resource_wims; i++) {
1401 ret = for_lookup_table_entry(resource_wims[i]->lookup_table,
1410 for_lookup_table_entry(wim->lookup_table, lte_delete_if_new,
1416 reference_resource_paths(WIMStruct *wim,
1417 const tchar * const *resource_wimfiles,
1418 unsigned num_resource_wimfiles,
1421 wimlib_progress_func_t progress_func)
1423 WIMStruct **resource_wims;
1427 resource_wims = CALLOC(num_resource_wimfiles, sizeof(resource_wims[0]));
1429 return WIMLIB_ERR_NOMEM;
1431 for (i = 0; i < num_resource_wimfiles; i++) {
1432 DEBUG("Referencing resources from path \"%"TS"\"",
1433 resource_wimfiles[i]);
1434 ret = wimlib_open_wim(resource_wimfiles[i], open_flags,
1435 &resource_wims[i], progress_func);
1437 goto out_free_resource_wims;
1440 ret = wimlib_reference_resources(wim, resource_wims,
1441 num_resource_wimfiles, ref_flags);
1443 goto out_free_resource_wims;
1445 for (i = 0; i < num_resource_wimfiles; i++)
1446 list_add_tail(&resource_wims[i]->subwim_node, &wim->subwims);
1449 goto out_free_array;
1451 out_free_resource_wims:
1452 for (i = 0; i < num_resource_wimfiles; i++)
1453 wimlib_free(resource_wims[i]);
1455 FREE(resource_wims);
1460 reference_resource_glob(WIMStruct *wim, const tchar *refglob,
1461 int ref_flags, int open_flags,
1462 wimlib_progress_func_t progress_func)
1467 /* Note: glob() is replaced in Windows native builds. */
1468 ret = tglob(refglob, GLOB_ERR | GLOB_NOSORT, NULL, &globbuf);
1470 if (ret == GLOB_NOMATCH) {
1471 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ERR_ON_NOMATCH) {
1472 ERROR("Found no files for glob \"%"TS"\"", refglob);
1473 return WIMLIB_ERR_GLOB_HAD_NO_MATCHES;
1475 return reference_resource_paths(wim,
1483 ERROR_WITH_ERRNO("Failed to process glob \"%"TS"\"", refglob);
1484 if (ret == GLOB_NOSPACE)
1485 return WIMLIB_ERR_NOMEM;
1487 return WIMLIB_ERR_READ;
1491 ret = reference_resource_paths(wim,
1492 (const tchar * const *)globbuf.gl_pathv,
1501 /* API function documented in wimlib.h */
1503 wimlib_reference_resource_files(WIMStruct *wim,
1504 const tchar * const * resource_wimfiles_or_globs,
1508 wimlib_progress_func_t progress_func)
1513 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ENABLE) {
1514 for (i = 0; i < count; i++) {
1515 ret = reference_resource_glob(wim,
1516 resource_wimfiles_or_globs[i],
1525 return reference_resource_paths(wim, resource_wimfiles_or_globs,
1527 open_flags, progress_func);