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_WIN32_ENCRYPTED:
342 /* Compare files by path: just a heuristic that will place files
343 * in the same directory next to each other. */
344 return tstrcmp(lte1->file_on_disk, lte2->file_on_disk);
346 case RESOURCE_IN_NTFS_VOLUME:
347 return tstrcmp(lte1->ntfs_loc->path, lte2->ntfs_loc->path);
350 /* No additional sorting order defined for this resource
351 * location (e.g. RESOURCE_IN_ATTACHED_BUFFER); simply compare
352 * everything equal to each other. */
358 sort_stream_list_by_sequential_order(struct list_head *stream_list,
359 size_t list_head_offset)
361 struct list_head *cur;
362 struct wim_lookup_table_entry **array;
365 size_t num_streams = 0;
367 list_for_each(cur, stream_list)
370 array_size = num_streams * sizeof(array[0]);
371 array = MALLOC(array_size);
373 return WIMLIB_ERR_NOMEM;
374 cur = stream_list->next;
375 for (i = 0; i < num_streams; i++) {
376 array[i] = (struct wim_lookup_table_entry*)((u8*)cur -
381 qsort(array, num_streams, sizeof(array[0]),
382 cmp_streams_by_sequential_order);
384 INIT_LIST_HEAD(stream_list);
385 for (i = 0; i < num_streams; i++) {
386 list_add_tail((struct list_head*)
387 ((u8*)array[i] + list_head_offset),
396 add_lte_to_array(struct wim_lookup_table_entry *lte,
399 struct wim_lookup_table_entry ***pp = _pp;
404 /* Iterate through the lookup table entries, but first sort them by stream
405 * offset in the WIM. Caution: this is intended to be used when the stream
406 * offset field has actually been set. */
408 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
409 int (*visitor)(struct wim_lookup_table_entry *,
413 struct wim_lookup_table_entry **lte_array, **p;
414 size_t num_streams = table->num_entries;
417 lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
419 return WIMLIB_ERR_NOMEM;
421 for_lookup_table_entry(table, add_lte_to_array, &p);
423 wimlib_assert(p == lte_array + num_streams);
425 qsort(lte_array, num_streams, sizeof(lte_array[0]),
426 cmp_streams_by_sequential_order);
428 for (size_t i = 0; i < num_streams; i++) {
429 ret = visitor(lte_array[i], arg);
437 /* On-disk format of a WIM lookup table entry (stream entry). */
438 struct wim_lookup_table_entry_disk {
439 /* Location, offset, compression status, and metadata status of the
441 struct resource_entry_disk resource_entry;
443 /* Which part of the split WIM this stream is in; indexed from 1. */
446 /* Reference count of this stream over all WIM images. */
449 /* SHA1 message digest of the uncompressed data of this stream, or
450 * optionally all zeroes if this stream is of zero length. */
451 u8 hash[SHA1_HASH_SIZE];
454 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
457 lte_init_wim(struct wim_lookup_table_entry *lte, WIMStruct *wim)
459 lte->resource_location = RESOURCE_IN_WIM;
461 if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
462 lte->compression_type = wim->compression_type;
464 lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE;
466 if (wim_is_pipable(wim))
471 * Reads the lookup table from a WIM file.
473 * Saves lookup table entries for non-metadata streams in a hash table, and
474 * saves the metadata entry for each image in a special per-image location (the
475 * image_metadata array).
478 * WIMLIB_ERR_SUCCESS (0)
479 * WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY
480 * WIMLIB_ERR_RESOURCE_NOT_FOUND
483 read_wim_lookup_table(WIMStruct *wim)
488 struct wim_lookup_table *table;
489 struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
492 BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
493 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
495 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
496 wim->hdr.lookup_table_res_entry.offset,
497 wim->hdr.lookup_table_res_entry.size);
499 /* Calculate number of entries in the lookup table. */
500 num_entries = wim->hdr.lookup_table_res_entry.size /
501 sizeof(struct wim_lookup_table_entry_disk);
504 /* Read the lookup table into a buffer. */
505 ret = res_entry_to_data(&wim->hdr.lookup_table_res_entry, wim, &buf);
509 /* Allocate hash table. */
510 table = new_lookup_table(num_entries * 2 + 1);
512 ERROR("Not enough memory to read lookup table.");
513 ret = WIMLIB_ERR_NOMEM;
517 /* Allocate and initalize `struct wim_lookup_table_entry's from the
518 * on-disk lookup table. */
519 wim->current_image = 0;
520 for (i = 0; i < num_entries; i++) {
521 const struct wim_lookup_table_entry_disk *disk_entry =
522 &((const struct wim_lookup_table_entry_disk*)buf)[i];
524 cur_entry = new_lookup_table_entry();
526 ERROR("Not enough memory to read lookup table.");
527 ret = WIMLIB_ERR_NOMEM;
528 goto out_free_lookup_table;
531 cur_entry->wim = wim;
532 cur_entry->resource_location = RESOURCE_IN_WIM;
533 get_resource_entry(&disk_entry->resource_entry, &cur_entry->resource_entry);
534 cur_entry->part_number = le16_to_cpu(disk_entry->part_number);
535 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
536 copy_hash(cur_entry->hash, disk_entry->hash);
537 lte_init_wim(cur_entry, wim);
539 if (cur_entry->part_number != wim->hdr.part_number) {
540 WARNING("A lookup table entry in part %hu of the WIM "
541 "points to part %hu (ignoring it)",
542 wim->hdr.part_number, cur_entry->part_number);
543 free_lookup_table_entry(cur_entry);
547 if (is_zero_hash(cur_entry->hash)) {
548 WARNING("The WIM lookup table contains an entry with a "
549 "SHA1 message digest of all 0's (ignoring it)");
550 free_lookup_table_entry(cur_entry);
554 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
555 && (cur_entry->resource_entry.size !=
556 cur_entry->resource_entry.original_size))
558 if (wimlib_print_errors) {
559 WARNING("Found uncompressed resource with "
560 "original size (%"PRIu64") not the same "
561 "as compressed size (%"PRIu64")",
562 cur_entry->resource_entry.original_size,
563 cur_entry->resource_entry.size);
564 if (cur_entry->resource_entry.original_size) {
565 WARNING("Overriding compressed size with original size.");
566 cur_entry->resource_entry.size =
567 cur_entry->resource_entry.original_size;
569 WARNING("Overriding original size with compressed size");
570 cur_entry->resource_entry.original_size =
571 cur_entry->resource_entry.size;
576 if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
577 /* Lookup table entry for a metadata resource */
578 if (cur_entry->refcnt != 1) {
579 /* Metadata entries with no references must be
580 * ignored. See for example the WinPE WIMs from
582 if (cur_entry->refcnt == 0) {
583 free_lookup_table_entry(cur_entry);
586 if (wimlib_print_errors) {
587 ERROR("Found metadata resource with refcnt != 1:");
588 print_lookup_table_entry(cur_entry, stderr);
590 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
591 goto out_free_cur_entry;
594 if (wim->hdr.part_number != 1) {
595 WARNING("Ignoring metadata resource found in a "
596 "non-first part of the split WIM");
597 free_lookup_table_entry(cur_entry);
600 if (wim->current_image == wim->hdr.image_count) {
601 WARNING("The WIM header says there are %u images "
602 "in the WIM, but we found more metadata "
603 "resources than this (ignoring the extra)",
604 wim->hdr.image_count);
605 free_lookup_table_entry(cur_entry);
609 /* Notice very carefully: We are assigning the metadata
610 * resources in the exact order mirrored by their lookup
611 * table entries on disk, which is the behavior of
612 * Microsoft's software. In particular, this overrides
613 * the actual locations of the metadata resources
614 * themselves in the WIM file as well as any information
615 * written in the XML data. */
616 DEBUG("Found metadata resource for image %u at "
618 wim->current_image + 1,
619 cur_entry->resource_entry.offset);
621 wim->current_image++]->metadata_lte = cur_entry;
623 /* Lookup table entry for a stream that is not a
624 * metadata resource */
625 duplicate_entry = lookup_resource(table, cur_entry->hash);
626 if (duplicate_entry) {
627 if (wimlib_print_errors) {
628 WARNING("The WIM lookup table contains two entries with the "
629 "same SHA1 message digest!");
630 WARNING("The first entry is:");
631 print_lookup_table_entry(duplicate_entry, stderr);
632 WARNING("The second entry is:");
633 print_lookup_table_entry(cur_entry, stderr);
635 free_lookup_table_entry(cur_entry);
638 lookup_table_insert(table, cur_entry);
643 if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
644 WARNING("The header of \"%"TS"\" says there are %u images in\n"
645 " the WIM, but we only found %d metadata resources! Acting as if\n"
646 " the header specified only %d images instead.",
647 wim->filename, wim->hdr.image_count,
648 wim->current_image, wim->current_image);
649 for (int i = wim->current_image; i < wim->hdr.image_count; i++)
650 put_image_metadata(wim->image_metadata[i], NULL);
651 wim->hdr.image_count = wim->current_image;
653 DEBUG("Done reading lookup table.");
654 wim->lookup_table = table;
659 out_free_lookup_table:
660 free_lookup_table(table);
664 wim->current_image = 0;
670 write_wim_lookup_table_entry(const struct wim_lookup_table_entry *lte,
671 struct wim_lookup_table_entry_disk *disk_entry)
673 put_resource_entry(<e->output_resource_entry, &disk_entry->resource_entry);
674 disk_entry->part_number = cpu_to_le16(lte->part_number);
675 disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
676 copy_hash(disk_entry->hash, lte->hash);
680 write_wim_lookup_table_from_stream_list(struct list_head *stream_list,
681 struct filedes *out_fd,
682 struct resource_entry *out_res_entry,
683 int write_resource_flags,
684 struct wimlib_lzx_context **comp_ctx)
687 struct wim_lookup_table_entry *lte;
688 struct wim_lookup_table_entry_disk *table_buf;
689 struct wim_lookup_table_entry_disk *table_buf_ptr;
693 list_for_each_entry(lte, stream_list, lookup_table_list)
694 table_size += sizeof(struct wim_lookup_table_entry_disk);
696 DEBUG("Writing WIM lookup table (size=%zu, offset=%"PRIu64")",
697 table_size, out_fd->offset);
699 table_buf = MALLOC(table_size);
701 ERROR("Failed to allocate %zu bytes for temporary lookup table",
703 return WIMLIB_ERR_NOMEM;
705 table_buf_ptr = table_buf;
706 list_for_each_entry(lte, stream_list, lookup_table_list)
707 write_wim_lookup_table_entry(lte, table_buf_ptr++);
709 /* Write the lookup table uncompressed. Although wimlib can handle a
710 * compressed lookup table, MS software cannot. */
711 ret = write_wim_resource_from_buffer(table_buf,
713 WIM_RESHDR_FLAG_METADATA,
715 WIMLIB_COMPRESSION_TYPE_NONE,
719 write_resource_flags,
722 DEBUG("ret=%d", ret);
727 append_lookup_table_entry(struct wim_lookup_table_entry *lte, void *_list)
729 /* Lookup table entries with 'out_refcnt' == 0 correspond to streams not
730 * written and not present in the resulting WIM file, and should not be
731 * included in the lookup table.
733 * Lookup table entries marked as filtered (EXTERNAL_WIM) with
734 * 'out_refcnt != 0' were referenced as part of the logical write but
735 * correspond to streams that were not in fact written, and should not
736 * be included in the lookup table.
738 * Lookup table entries marked as filtered (SAME_WIM) with 'out_refcnt
739 * != 0' were referenced as part of the logical write but correspond to
740 * streams that were not in fact written, but nevertheless were already
741 * present in the WIM being overwritten in-place. These entries must be
742 * included in the lookup table, and the resource information to write
743 * needs to be copied from the resource information read originally.
745 if (lte->out_refcnt != 0 && !(lte->filtered & FILTERED_EXTERNAL_WIM)) {
746 if (lte->filtered & FILTERED_SAME_WIM) {
747 copy_resource_entry(<e->output_resource_entry,
748 <e->resource_entry);
750 list_add_tail(<e->lookup_table_list, (struct list_head*)_list);
756 write_wim_lookup_table(WIMStruct *wim, int image, int write_flags,
757 struct resource_entry *out_res_entry,
758 struct list_head *stream_list_override)
760 int write_resource_flags;
761 struct list_head _stream_list;
762 struct list_head *stream_list;
764 if (stream_list_override) {
765 stream_list = stream_list_override;
767 stream_list = &_stream_list;
768 INIT_LIST_HEAD(stream_list);
771 if (!(write_flags & WIMLIB_WRITE_FLAG_NO_METADATA)) {
775 if (image == WIMLIB_ALL_IMAGES) {
777 end_image = wim->hdr.image_count;
783 /* Push metadata resource lookup table entries onto the front of
784 * the list in reverse order, so that they're written in order.
786 for (int i = end_image; i >= start_image; i--) {
787 struct wim_lookup_table_entry *metadata_lte;
789 metadata_lte = wim->image_metadata[i - 1]->metadata_lte;
790 metadata_lte->out_refcnt = 1;
791 metadata_lte->part_number = wim->hdr.part_number;
792 metadata_lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
794 list_add(&metadata_lte->lookup_table_list, stream_list);
798 /* Append additional lookup table entries that need to be written, with
799 * some special handling for streams that have been marked as filtered.
801 if (!stream_list_override) {
802 for_lookup_table_entry(wim->lookup_table,
803 append_lookup_table_entry, stream_list);
806 write_resource_flags = 0;
807 if (write_flags & WIMLIB_WRITE_FLAG_PIPABLE)
808 write_resource_flags |= WIMLIB_WRITE_RESOURCE_FLAG_PIPABLE;
809 return write_wim_lookup_table_from_stream_list(stream_list,
812 write_resource_flags,
818 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
820 lte->real_refcnt = 0;
825 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
832 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
834 if (lte->extracted_file != NULL) {
835 FREE(lte->extracted_file);
836 lte->extracted_file = NULL;
842 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
848 tfprintf(out, T("Offset = %"PRIu64" bytes\n"),
849 lte->resource_entry.offset);
851 tfprintf(out, T("Size = %"PRIu64" bytes\n"),
852 (u64)lte->resource_entry.size);
854 tfprintf(out, T("Original size = %"PRIu64" bytes\n"),
855 lte->resource_entry.original_size);
857 tfprintf(out, T("Part Number = %hu\n"), lte->part_number);
858 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
861 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
862 lte->back_inode, lte->back_stream_id);
864 tfprintf(out, T("Hash = 0x"));
865 print_hash(lte->hash, out);
869 tfprintf(out, T("Flags = "));
870 u8 flags = lte->resource_entry.flags;
871 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
872 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
873 if (flags & WIM_RESHDR_FLAG_FREE)
874 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
875 if (flags & WIM_RESHDR_FLAG_METADATA)
876 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
877 if (flags & WIM_RESHDR_FLAG_SPANNED)
878 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
880 switch (lte->resource_location) {
881 case RESOURCE_IN_WIM:
882 if (lte->wim->filename) {
883 tfprintf(out, T("WIM file = `%"TS"'\n"),
888 case RESOURCE_WIN32_ENCRYPTED:
890 case RESOURCE_IN_FILE_ON_DISK:
891 tfprintf(out, T("File on Disk = `%"TS"'\n"),
895 case RESOURCE_IN_STAGING_FILE:
896 tfprintf(out, T("Staging File = `%"TS"'\n"),
897 lte->staging_file_name);
907 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
908 struct wimlib_resource_entry *wentry)
910 wentry->uncompressed_size = lte->resource_entry.original_size;
911 wentry->compressed_size = lte->resource_entry.size;
912 wentry->offset = lte->resource_entry.offset;
913 copy_hash(wentry->sha1_hash, lte->hash);
914 wentry->part_number = lte->part_number;
915 wentry->reference_count = lte->refcnt;
916 wentry->is_compressed = (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
917 wentry->is_metadata = (lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) != 0;
918 wentry->is_free = (lte->resource_entry.flags & WIM_RESHDR_FLAG_FREE) != 0;
919 wentry->is_spanned = (lte->resource_entry.flags & WIM_RESHDR_FLAG_SPANNED) != 0;
922 struct iterate_lte_context {
923 wimlib_iterate_lookup_table_callback_t cb;
928 do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
930 struct iterate_lte_context *ctx = _ctx;
931 struct wimlib_resource_entry entry;
933 lte_to_wimlib_resource_entry(lte, &entry);
934 return (*ctx->cb)(&entry, ctx->user_ctx);
937 /* API function documented in wimlib.h */
939 wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
940 wimlib_iterate_lookup_table_callback_t cb,
943 struct iterate_lte_context ctx = {
945 .user_ctx = user_ctx,
947 if (wim->hdr.part_number == 1) {
949 for (int i = 0; i < wim->hdr.image_count; i++) {
950 ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
956 return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
959 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
960 * lookup table, or NULL if there is none. */
961 struct wim_lookup_table_entry *
962 lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
965 struct wim_lookup_table_entry *lte;
966 struct hlist_node *pos;
968 wimlib_assert(table != NULL);
969 wimlib_assert(hash != NULL);
971 i = *(size_t*)hash % table->capacity;
972 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
973 if (hashes_equal(hash, lte->hash))
980 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
983 * This is only for pre-resolved inodes.
986 wim_pathname_to_stream(WIMStruct *wim,
989 struct wim_dentry **dentry_ret,
990 struct wim_lookup_table_entry **lte_ret,
993 struct wim_dentry *dentry;
994 struct wim_lookup_table_entry *lte;
996 const tchar *stream_name = NULL;
997 struct wim_inode *inode;
1000 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
1001 stream_name = path_stream_name(path);
1003 p = (tchar*)stream_name - 1;
1008 dentry = get_dentry(wim, path);
1014 inode = dentry->d_inode;
1016 if (!inode->i_resolved)
1017 if (inode_resolve_ltes(inode, wim->lookup_table, false))
1020 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
1021 && inode_is_directory(inode))
1025 struct wim_ads_entry *ads_entry;
1027 ads_entry = inode_get_ads_entry(inode, stream_name,
1030 stream_idx = ads_idx + 1;
1031 lte = ads_entry->lte;
1037 lte = inode_unnamed_stream_resolved(inode, &stream_idx);
1041 *dentry_ret = dentry;
1045 *stream_idx_ret = stream_idx;
1051 resource_not_found_error(const struct wim_inode *inode, const u8 *hash)
1053 if (wimlib_print_errors) {
1054 ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
1055 tfprintf(stderr, T(" SHA-1 message digest of missing resource:\n "));
1056 print_hash(hash, stderr);
1057 tputc(T('\n'), stderr);
1059 return WIMLIB_ERR_RESOURCE_NOT_FOUND;
1063 * Resolve an inode's lookup table entries.
1065 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
1066 * lookup table) with pointers directly to the lookup table entries.
1068 * If @force is %false:
1069 * If any needed SHA1 message digests are not found in the lookup table,
1070 * WIMLIB_ERR_RESOURCE_NOT_FOUND is returned and the inode is left
1072 * If @force is %true:
1073 * If any needed SHA1 message digests are not found in the lookup table,
1074 * new entries are allocated and inserted into the lookup table.
1077 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table,
1082 if (!inode->i_resolved) {
1083 struct wim_lookup_table_entry *lte, *ads_lte;
1085 /* Resolve the default file stream */
1087 hash = inode->i_hash;
1088 if (!is_zero_hash(hash)) {
1089 lte = lookup_resource(table, hash);
1092 lte = new_lookup_table_entry();
1094 return WIMLIB_ERR_NOMEM;
1095 copy_hash(lte->hash, hash);
1096 lookup_table_insert(table, lte);
1098 goto resource_not_found;
1103 /* Resolve the alternate data streams */
1104 struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
1105 for (u16 i = 0; i < inode->i_num_ads; i++) {
1106 struct wim_ads_entry *cur_entry;
1109 cur_entry = &inode->i_ads_entries[i];
1110 hash = cur_entry->hash;
1111 if (!is_zero_hash(hash)) {
1112 ads_lte = lookup_resource(table, hash);
1115 ads_lte = new_lookup_table_entry();
1117 return WIMLIB_ERR_NOMEM;
1118 copy_hash(ads_lte->hash, hash);
1119 lookup_table_insert(table, ads_lte);
1121 goto resource_not_found;
1125 ads_ltes[i] = ads_lte;
1128 for (u16 i = 0; i < inode->i_num_ads; i++)
1129 inode->i_ads_entries[i].lte = ads_ltes[i];
1130 inode->i_resolved = 1;
1135 return resource_not_found_error(inode, hash);
1139 inode_unresolve_ltes(struct wim_inode *inode)
1141 if (inode->i_resolved) {
1143 copy_hash(inode->i_hash, inode->i_lte->hash);
1145 zero_out_hash(inode->i_hash);
1147 for (u16 i = 0; i < inode->i_num_ads; i++) {
1148 if (inode->i_ads_entries[i].lte)
1149 copy_hash(inode->i_ads_entries[i].hash,
1150 inode->i_ads_entries[i].lte->hash);
1152 zero_out_hash(inode->i_ads_entries[i].hash);
1154 inode->i_resolved = 0;
1159 * Returns the lookup table entry for stream @stream_idx of the inode, where
1160 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
1161 * corresponds to an alternate data stream.
1163 * This works for both resolved and un-resolved inodes.
1165 struct wim_lookup_table_entry *
1166 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
1167 const struct wim_lookup_table *table)
1169 if (inode->i_resolved)
1170 return inode_stream_lte_resolved(inode, stream_idx);
1172 return inode_stream_lte_unresolved(inode, stream_idx, table);
1175 struct wim_lookup_table_entry *
1176 inode_unnamed_stream_resolved(const struct wim_inode *inode, u16 *stream_idx_ret)
1178 wimlib_assert(inode->i_resolved);
1179 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1180 if (inode_stream_name_nbytes(inode, i) == 0 &&
1181 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
1183 *stream_idx_ret = i;
1184 return inode_stream_lte_resolved(inode, i);
1187 *stream_idx_ret = 0;
1191 struct wim_lookup_table_entry *
1192 inode_unnamed_lte_resolved(const struct wim_inode *inode)
1195 return inode_unnamed_stream_resolved(inode, &stream_idx);
1198 struct wim_lookup_table_entry *
1199 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
1200 const struct wim_lookup_table *table)
1202 wimlib_assert(!inode->i_resolved);
1203 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1204 if (inode_stream_name_nbytes(inode, i) == 0 &&
1205 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
1207 return inode_stream_lte_unresolved(inode, i, table);
1213 /* Return the lookup table entry for the unnamed data stream of an inode, or
1214 * NULL if there is none.
1216 * You'd think this would be easier than it actually is, since the unnamed data
1217 * stream should be the one referenced from the inode itself. Alas, if there
1218 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
1219 * data stream in one of the alternate data streams instead of inside the WIM
1220 * dentry itself. So we need to check the alternate data streams too.
1222 * Also, note that a dentry may appear to have more than one unnamed stream, but
1223 * if the SHA1 message digest is all 0's then the corresponding stream does not
1224 * really "count" (this is the case for the inode's own file stream when the
1225 * file stream that should be there is actually in one of the alternate stream
1226 * entries.). This is despite the fact that we may need to extract such a
1227 * missing entry as an empty file or empty named data stream.
1229 struct wim_lookup_table_entry *
1230 inode_unnamed_lte(const struct wim_inode *inode,
1231 const struct wim_lookup_table *table)
1233 if (inode->i_resolved)
1234 return inode_unnamed_lte_resolved(inode);
1236 return inode_unnamed_lte_unresolved(inode, table);
1239 /* Returns the SHA1 message digest of the unnamed data stream of a WIM inode, or
1240 * 'zero_hash' if the unnamed data stream is missing has all zeroes in its SHA1
1241 * message digest field. */
1243 inode_unnamed_stream_hash(const struct wim_inode *inode)
1247 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1248 if (inode_stream_name_nbytes(inode, i) == 0) {
1249 hash = inode_stream_hash(inode, i);
1250 if (!is_zero_hash(hash))
1259 lte_add_stream_size(struct wim_lookup_table_entry *lte, void *total_bytes_p)
1261 *(u64*)total_bytes_p += lte->resource_entry.size;
1266 lookup_table_total_stream_size(struct wim_lookup_table *table)
1269 for_lookup_table_entry(table, lte_add_stream_size, &total_size);
1273 struct wim_lookup_table_entry **
1274 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1276 wimlib_assert(lte->unhashed);
1277 struct wim_inode *inode = lte->back_inode;
1278 u32 stream_id = lte->back_stream_id;
1280 return &inode->i_lte;
1282 for (u16 i = 0; i < inode->i_num_ads; i++)
1283 if (inode->i_ads_entries[i].stream_id == stream_id)
1284 return &inode->i_ads_entries[i].lte;
1289 /* Calculate the SHA1 message digest of a stream and move it from the list of
1290 * unhashed streams to the stream lookup table, possibly joining it with an
1291 * existing lookup table entry for an identical stream.
1293 * @lte: An unhashed lookup table entry.
1294 * @lookup_table: Lookup table for the WIM.
1295 * @lte_ret: On success, write a pointer to the resulting lookup table
1296 * entry to this location. This will be the same as @lte
1297 * if it was inserted into the lookup table, or different if
1298 * a duplicate stream was found.
1300 * Returns 0 on success; nonzero if there is an error reading the stream.
1303 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1304 struct wim_lookup_table *lookup_table,
1305 struct wim_lookup_table_entry **lte_ret)
1308 struct wim_lookup_table_entry *duplicate_lte;
1309 struct wim_lookup_table_entry **back_ptr;
1311 wimlib_assert(lte->unhashed);
1313 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1314 * union with the SHA1 message digest and will no longer be valid once
1315 * the SHA1 has been calculated. */
1316 back_ptr = retrieve_lte_pointer(lte);
1318 ret = sha1_resource(lte);
1322 /* Look for a duplicate stream */
1323 duplicate_lte = lookup_resource(lookup_table, lte->hash);
1324 list_del(<e->unhashed_list);
1325 if (duplicate_lte) {
1326 /* We have a duplicate stream. Transfer the reference counts
1327 * from this stream to the duplicate, update the reference to
1328 * this stream (in an inode or ads_entry) to point to the
1329 * duplicate, then free this stream. */
1330 wimlib_assert(!(duplicate_lte->unhashed));
1331 duplicate_lte->refcnt += lte->refcnt;
1332 duplicate_lte->out_refcnt += lte->out_refcnt;
1333 *back_ptr = duplicate_lte;
1334 free_lookup_table_entry(lte);
1335 lte = duplicate_lte;
1337 /* No duplicate stream, so we need to insert
1338 * this stream into the lookup table and treat
1339 * it as a hashed stream. */
1340 lookup_table_insert(lookup_table, lte);
1349 lte_clone_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1351 struct wim_lookup_table *lookup_table = _lookup_table;
1353 if (lookup_resource(lookup_table, lte->hash))
1354 return 0; /* Resource already present. */
1356 lte = clone_lookup_table_entry(lte);
1358 return WIMLIB_ERR_NOMEM;
1359 lte->out_refcnt = 1;
1360 lookup_table_insert(lookup_table, lte);
1365 lte_delete_if_new(struct wim_lookup_table_entry *lte, void *_lookup_table)
1367 struct wim_lookup_table *lookup_table = _lookup_table;
1369 if (lte->out_refcnt) {
1370 lookup_table_unlink(lookup_table, lte);
1371 free_lookup_table_entry(lte);
1376 /* API function documented in wimlib.h */
1378 wimlib_reference_resources(WIMStruct *wim,
1379 WIMStruct **resource_wims, unsigned num_resource_wims,
1386 return WIMLIB_ERR_INVALID_PARAM;
1388 if (num_resource_wims != 0 && resource_wims == NULL)
1389 return WIMLIB_ERR_INVALID_PARAM;
1391 for (i = 0; i < num_resource_wims; i++)
1392 if (resource_wims[i] == NULL)
1393 return WIMLIB_ERR_INVALID_PARAM;
1395 for_lookup_table_entry(wim->lookup_table, lte_zero_out_refcnt, NULL);
1397 for (i = 0; i < num_resource_wims; i++) {
1398 ret = for_lookup_table_entry(resource_wims[i]->lookup_table,
1407 for_lookup_table_entry(wim->lookup_table, lte_delete_if_new,
1413 reference_resource_paths(WIMStruct *wim,
1414 const tchar * const *resource_wimfiles,
1415 unsigned num_resource_wimfiles,
1418 wimlib_progress_func_t progress_func)
1420 WIMStruct **resource_wims;
1424 resource_wims = CALLOC(num_resource_wimfiles, sizeof(resource_wims[0]));
1426 return WIMLIB_ERR_NOMEM;
1428 for (i = 0; i < num_resource_wimfiles; i++) {
1429 DEBUG("Referencing resources from path \"%"TS"\"",
1430 resource_wimfiles[i]);
1431 ret = wimlib_open_wim(resource_wimfiles[i], open_flags,
1432 &resource_wims[i], progress_func);
1434 goto out_free_resource_wims;
1437 ret = wimlib_reference_resources(wim, resource_wims,
1438 num_resource_wimfiles, ref_flags);
1440 goto out_free_resource_wims;
1442 for (i = 0; i < num_resource_wimfiles; i++)
1443 list_add_tail(&resource_wims[i]->subwim_node, &wim->subwims);
1446 goto out_free_array;
1448 out_free_resource_wims:
1449 for (i = 0; i < num_resource_wimfiles; i++)
1450 wimlib_free(resource_wims[i]);
1452 FREE(resource_wims);
1457 reference_resource_glob(WIMStruct *wim, const tchar *refglob,
1458 int ref_flags, int open_flags,
1459 wimlib_progress_func_t progress_func)
1464 /* Note: glob() is replaced in Windows native builds. */
1465 ret = tglob(refglob, GLOB_ERR | GLOB_NOSORT, NULL, &globbuf);
1467 if (ret == GLOB_NOMATCH) {
1468 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ERR_ON_NOMATCH) {
1469 ERROR("Found no files for glob \"%"TS"\"", refglob);
1470 return WIMLIB_ERR_GLOB_HAD_NO_MATCHES;
1472 return reference_resource_paths(wim,
1480 ERROR_WITH_ERRNO("Failed to process glob \"%"TS"\"", refglob);
1481 if (ret == GLOB_NOSPACE)
1482 return WIMLIB_ERR_NOMEM;
1484 return WIMLIB_ERR_READ;
1488 ret = reference_resource_paths(wim,
1489 (const tchar * const *)globbuf.gl_pathv,
1498 /* API function documented in wimlib.h */
1500 wimlib_reference_resource_files(WIMStruct *wim,
1501 const tchar * const * resource_wimfiles_or_globs,
1505 wimlib_progress_func_t progress_func)
1510 if (ref_flags & WIMLIB_REF_FLAG_GLOB_ENABLE) {
1511 for (i = 0; i < count; i++) {
1512 ret = reference_resource_glob(wim,
1513 resource_wimfiles_or_globs[i],
1522 return reference_resource_paths(wim, resource_wimfiles_or_globs,
1524 open_flags, progress_func);