4 * Lookup table, implemented as a hash table, that maps SHA1 message digests to
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/lookup_table.h"
35 #include "wimlib/metadata.h"
36 #include "wimlib/paths.h"
37 #include "wimlib/resource.h"
38 #include "wimlib/util.h"
43 # include <unistd.h> /* for unlink() */
46 struct wim_lookup_table *
47 new_lookup_table(size_t capacity)
49 struct wim_lookup_table *table;
50 struct hlist_head *array;
52 table = CALLOC(1, sizeof(struct wim_lookup_table));
54 array = CALLOC(capacity, sizeof(array[0]));
56 table->num_entries = 0;
57 table->capacity = capacity;
62 ERROR("Failed to allocate memory for lookup table "
63 "with capacity %zu", capacity);
69 struct wim_lookup_table_entry *
70 new_lookup_table_entry(void)
72 struct wim_lookup_table_entry *lte;
74 lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
79 ERROR("Out of memory (tried to allocate %zu bytes for "
80 "lookup table entry)",
81 sizeof(struct wim_lookup_table_entry));
86 struct wim_lookup_table_entry *
87 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
89 struct wim_lookup_table_entry *new;
91 new = memdup(old, sizeof(struct wim_lookup_table_entry));
95 new->extracted_file = NULL;
96 switch (new->resource_location) {
99 case RESOURCE_WIN32_ENCRYPTED:
101 case RESOURCE_IN_FILE_ON_DISK:
104 case RESOURCE_IN_STAGING_FILE:
105 BUILD_BUG_ON((void*)&old->file_on_disk !=
106 (void*)&old->staging_file_name);
108 new->file_on_disk = TSTRDUP(old->file_on_disk);
109 if (!new->file_on_disk)
112 case RESOURCE_IN_ATTACHED_BUFFER:
113 new->attached_buffer = memdup(old->attached_buffer,
114 wim_resource_size(old));
115 if (!new->attached_buffer)
119 case RESOURCE_IN_NTFS_VOLUME:
121 struct ntfs_location *loc;
122 loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
126 loc->stream_name = NULL;
128 loc->path = STRDUP(old->ntfs_loc->path);
131 if (loc->stream_name_nchars) {
132 loc->stream_name = memdup(old->ntfs_loc->stream_name,
133 loc->stream_name_nchars * 2);
134 if (!loc->stream_name)
145 free_lookup_table_entry(new);
150 free_lookup_table_entry(struct wim_lookup_table_entry *lte)
153 switch (lte->resource_location) {
156 case RESOURCE_WIN32_ENCRYPTED:
158 case RESOURCE_IN_FILE_ON_DISK:
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);
295 cmp_streams_by_wim_position(const void *p1, const void *p2)
297 const struct wim_lookup_table_entry *lte1, *lte2;
298 lte1 = *(const struct wim_lookup_table_entry**)p1;
299 lte2 = *(const struct wim_lookup_table_entry**)p2;
300 if (lte1->resource_entry.offset < lte2->resource_entry.offset)
302 else if (lte1->resource_entry.offset > lte2->resource_entry.offset)
310 add_lte_to_array(struct wim_lookup_table_entry *lte,
313 struct wim_lookup_table_entry ***pp = _pp;
318 /* Iterate through the lookup table entries, but first sort them by stream
319 * offset in the WIM. Caution: this is intended to be used when the stream
320 * offset field has actually been set. */
322 for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
323 int (*visitor)(struct wim_lookup_table_entry *,
327 struct wim_lookup_table_entry **lte_array, **p;
328 size_t num_streams = table->num_entries;
331 lte_array = MALLOC(num_streams * sizeof(lte_array[0]));
333 return WIMLIB_ERR_NOMEM;
335 for_lookup_table_entry(table, add_lte_to_array, &p);
337 wimlib_assert(p == lte_array + num_streams);
339 qsort(lte_array, num_streams, sizeof(lte_array[0]),
340 cmp_streams_by_wim_position);
342 for (size_t i = 0; i < num_streams; i++) {
343 ret = visitor(lte_array[i], arg);
351 /* On-disk format of a WIM lookup table entry (stream entry). */
352 struct wim_lookup_table_entry_disk {
353 /* Location, offset, compression status, and metadata status of the
355 struct resource_entry_disk resource_entry;
357 /* Which part of the split WIM this stream is in; indexed from 1. */
360 /* Reference count of this stream over all WIM images. */
363 /* SHA1 message digest of the uncompressed data of this stream, or
364 * optionally all zeroes if this stream is of zero length. */
365 u8 hash[SHA1_HASH_SIZE];
368 #define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
371 * Reads the lookup table from a WIM file.
373 * Saves lookup table entries for non-metadata streams in a hash table, and
374 * saves the metadata entry for each image in a special per-image location (the
375 * image_metadata array).
378 read_lookup_table(WIMStruct *w)
382 struct wim_lookup_table *table;
383 struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
384 struct wim_lookup_table_entry_disk
385 table_buf[BUFFER_SIZE / sizeof(struct wim_lookup_table_entry_disk)]
386 _aligned_attribute(8);
388 BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
389 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
392 size_t buf_entries_remaining;
393 const struct wim_lookup_table_entry_disk *disk_entry;
395 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
396 w->hdr.lookup_table_res_entry.offset,
397 w->hdr.lookup_table_res_entry.original_size);
399 if (resource_is_compressed(&w->hdr.lookup_table_res_entry)) {
400 ERROR("Didn't expect a compressed lookup table!");
401 ERROR("Ask the author to implement support for this.");
402 return WIMLIB_ERR_COMPRESSED_LOOKUP_TABLE;
405 num_entries = w->hdr.lookup_table_res_entry.size /
406 sizeof(struct wim_lookup_table_entry_disk);
407 table = new_lookup_table(num_entries * 2 + 1);
409 return WIMLIB_ERR_NOMEM;
411 w->current_image = 0;
412 offset = w->hdr.lookup_table_res_entry.offset;
413 buf_entries_remaining = 0;
414 for (; num_entries != 0;
415 num_entries--, buf_entries_remaining--, disk_entry++)
417 if (buf_entries_remaining == 0) {
418 size_t entries_to_read, bytes_to_read;
420 entries_to_read = min(ARRAY_LEN(table_buf), num_entries);
421 bytes_to_read = entries_to_read * sizeof(struct wim_lookup_table_entry_disk);
422 if (full_pread(w->in_fd, table_buf,
423 bytes_to_read, offset) != bytes_to_read)
425 ERROR_WITH_ERRNO("Error reading lookup table "
426 "(offset=%"PRIu64")", offset);
427 ret = WIMLIB_ERR_READ;
428 goto out_free_lookup_table;
430 offset += bytes_to_read;
431 disk_entry = table_buf;
432 buf_entries_remaining = entries_to_read;
434 cur_entry = new_lookup_table_entry();
436 ret = WIMLIB_ERR_NOMEM;
437 goto out_free_lookup_table;
441 cur_entry->resource_location = RESOURCE_IN_WIM;
442 get_resource_entry(&disk_entry->resource_entry, &cur_entry->resource_entry);
443 cur_entry->part_number = le16_to_cpu(disk_entry->part_number);
444 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
445 copy_hash(cur_entry->hash, disk_entry->hash);
447 if (cur_entry->part_number != w->hdr.part_number) {
448 ERROR("A lookup table entry in part %hu of the WIM "
449 "points to part %hu",
450 w->hdr.part_number, cur_entry->part_number);
451 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
452 goto out_free_cur_entry;
455 if (is_zero_hash(cur_entry->hash)) {
456 ERROR("The WIM lookup table contains an entry with a "
457 "SHA1 message digest of all 0's");
458 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
459 goto out_free_cur_entry;
462 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
463 && (cur_entry->resource_entry.size !=
464 cur_entry->resource_entry.original_size))
466 #ifdef ENABLE_ERROR_MESSAGES
467 ERROR("Found uncompressed resource with original size "
468 "not the same as compressed size");
469 ERROR("The lookup table entry for the resource is as follows:");
470 print_lookup_table_entry(cur_entry, stderr);
472 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
473 goto out_free_cur_entry;
476 if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
477 /* Lookup table entry for a metadata resource */
478 if (cur_entry->refcnt != 1) {
479 #ifdef ENABLE_ERROR_MESSAGES
480 ERROR("Found metadata resource with refcnt != 1:");
481 print_lookup_table_entry(cur_entry, stderr);
483 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
484 goto out_free_cur_entry;
487 if (w->hdr.part_number != 1) {
488 ERROR("Found a metadata resource in a "
489 "non-first part of the split WIM!");
490 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
491 goto out_free_cur_entry;
493 if (w->current_image == w->hdr.image_count) {
494 ERROR("The WIM header says there are %u images "
495 "in the WIM, but we found more metadata "
496 "resources than this", w->hdr.image_count);
497 ret = WIMLIB_ERR_IMAGE_COUNT;
498 goto out_free_cur_entry;
501 /* Notice very carefully: We are assigning the metadata
502 * resources in the exact order mirrored by their lookup
503 * table entries on disk, which is the behavior of
504 * Microsoft's software. In particular, this overrides
505 * the actual locations of the metadata resources
506 * themselves in the WIM file as well as any information
507 * written in the XML data. */
508 DEBUG("Found metadata resource for image %u at "
510 w->current_image + 1,
511 cur_entry->resource_entry.offset);
513 w->current_image++]->metadata_lte = cur_entry;
515 /* Lookup table entry for a stream that is not a
516 * metadata resource */
517 duplicate_entry = __lookup_resource(table, cur_entry->hash);
518 if (duplicate_entry) {
519 #ifdef ENABLE_ERROR_MESSAGES
520 ERROR("The WIM lookup table contains two entries with the "
521 "same SHA1 message digest!");
522 ERROR("The first entry is:");
523 print_lookup_table_entry(duplicate_entry, stderr);
524 ERROR("The second entry is:");
525 print_lookup_table_entry(cur_entry, stderr);
527 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
528 goto out_free_cur_entry;
530 lookup_table_insert(table, cur_entry);
534 if (w->hdr.part_number == 1 && w->current_image != w->hdr.image_count)
536 ERROR("The WIM header says there are %u images "
537 "in the WIM, but we only found %d metadata "
538 "resources!", w->hdr.image_count, w->current_image);
539 ret = WIMLIB_ERR_IMAGE_COUNT;
540 goto out_free_lookup_table;
542 DEBUG("Done reading lookup table.");
543 w->lookup_table = table;
548 out_free_lookup_table:
549 free_lookup_table(table);
551 w->current_image = 0;
557 write_lookup_table_entry(const struct wim_lookup_table_entry *lte,
558 struct wim_lookup_table_entry_disk *disk_entry)
560 put_resource_entry(<e->output_resource_entry, &disk_entry->resource_entry);
561 disk_entry->part_number = cpu_to_le16(lte->part_number);
562 disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
563 copy_hash(disk_entry->hash, lte->hash);
567 write_lookup_table_from_stream_list(struct list_head *stream_list,
569 struct resource_entry *out_res_entry)
573 struct wim_lookup_table_entry_disk
574 table_buf[BUFFER_SIZE / sizeof(struct wim_lookup_table_entry_disk)]
575 _aligned_attribute(8);
577 size_t bytes_to_write;
578 struct wim_lookup_table_entry *lte;
581 start_offset = filedes_offset(out_fd);
582 if (start_offset == -1)
587 list_for_each_entry(lte, stream_list, lookup_table_list) {
588 if (cur_idx == ARRAY_LEN(table_buf)) {
589 bytes_to_write = sizeof(table_buf);
590 if (full_write(out_fd, table_buf,
591 bytes_to_write) != bytes_to_write)
593 table_size += bytes_to_write;
596 write_lookup_table_entry(lte, &table_buf[cur_idx]);
600 bytes_to_write = cur_idx * sizeof(struct wim_lookup_table_entry_disk);
601 if (full_write(out_fd, table_buf,
602 bytes_to_write) != bytes_to_write)
604 table_size += bytes_to_write;
606 out_res_entry->offset = start_offset;
607 out_res_entry->size = table_size;
608 out_res_entry->original_size = table_size;
609 out_res_entry->flags = WIM_RESHDR_FLAG_METADATA;
614 ERROR_WITH_ERRNO("Failed to write lookup table");
615 ret = WIMLIB_ERR_WRITE;
620 append_lookup_table_entry(struct wim_lookup_table_entry *lte, void *_list)
622 if (lte->out_refcnt != 0)
623 list_add_tail(<e->lookup_table_list, (struct list_head*)_list);
627 /* Writes the WIM lookup table to the output file. */
629 write_lookup_table(WIMStruct *w, int image, struct resource_entry *out_res_entry)
631 LIST_HEAD(stream_list);
635 if (image == WIMLIB_ALL_IMAGES) {
637 end_image = w->hdr.image_count;
643 for (int i = start_image; i <= end_image; i++) {
644 struct wim_lookup_table_entry *metadata_lte;
646 metadata_lte = w->image_metadata[i - 1]->metadata_lte;
647 metadata_lte->out_refcnt = 1;
648 metadata_lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
649 append_lookup_table_entry(metadata_lte, &stream_list);
651 for_lookup_table_entry(w->lookup_table,
652 append_lookup_table_entry,
654 return write_lookup_table_from_stream_list(&stream_list,
660 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
662 lte->real_refcnt = 0;
667 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
674 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
676 if (lte->extracted_file != NULL) {
677 FREE(lte->extracted_file);
678 lte->extracted_file = NULL;
684 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
690 tfprintf(out, T("Offset = %"PRIu64" bytes\n"),
691 lte->resource_entry.offset);
693 tfprintf(out, T("Size = %"PRIu64" bytes\n"),
694 (u64)lte->resource_entry.size);
696 tfprintf(out, T("Original size = %"PRIu64" bytes\n"),
697 lte->resource_entry.original_size);
699 tfprintf(out, T("Part Number = %hu\n"), lte->part_number);
700 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
703 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
704 lte->back_inode, lte->back_stream_id);
706 tfprintf(out, T("Hash = 0x"));
707 print_hash(lte->hash, out);
711 tfprintf(out, T("Flags = "));
712 u8 flags = lte->resource_entry.flags;
713 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
714 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
715 if (flags & WIM_RESHDR_FLAG_FREE)
716 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
717 if (flags & WIM_RESHDR_FLAG_METADATA)
718 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
719 if (flags & WIM_RESHDR_FLAG_SPANNED)
720 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
722 switch (lte->resource_location) {
723 case RESOURCE_IN_WIM:
724 if (lte->wim->filename) {
725 tfprintf(out, T("WIM file = `%"TS"'\n"),
731 case RESOURCE_WIN32_ENCRYPTED:
733 case RESOURCE_IN_FILE_ON_DISK:
735 tfprintf(out, T("File on Disk = `%"TS"'\n"),
739 case RESOURCE_IN_STAGING_FILE:
740 tfprintf(out, T("Staging File = `%"TS"'\n"),
741 lte->staging_file_name);
751 do_print_lookup_table_entry(struct wim_lookup_table_entry *lte, void *fp)
753 print_lookup_table_entry(lte, (FILE*)fp);
758 * Prints the lookup table of a WIM file.
761 wimlib_print_lookup_table(WIMStruct *w)
763 for_lookup_table_entry(w->lookup_table,
764 do_print_lookup_table_entry,
768 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
769 * lookup table, or NULL if there is none. */
770 struct wim_lookup_table_entry *
771 __lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
774 struct wim_lookup_table_entry *lte;
775 struct hlist_node *pos;
777 wimlib_assert(table != NULL);
778 wimlib_assert(hash != NULL);
780 i = *(size_t*)hash % table->capacity;
781 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
782 if (hashes_equal(hash, lte->hash))
789 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
792 * This is only for pre-resolved inodes.
795 lookup_resource(WIMStruct *w,
798 struct wim_dentry **dentry_ret,
799 struct wim_lookup_table_entry **lte_ret,
802 struct wim_dentry *dentry;
803 struct wim_lookup_table_entry *lte;
805 const tchar *stream_name = NULL;
806 struct wim_inode *inode;
809 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
810 stream_name = path_stream_name(path);
812 p = (tchar*)stream_name - 1;
817 dentry = get_dentry(w, path);
823 inode = dentry->d_inode;
825 wimlib_assert(inode->i_resolved);
827 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
828 && inode_is_directory(inode))
832 struct wim_ads_entry *ads_entry;
834 ads_entry = inode_get_ads_entry(inode, stream_name,
837 stream_idx = ads_idx + 1;
838 lte = ads_entry->lte;
849 *dentry_ret = dentry;
853 *stream_idx_ret = stream_idx;
859 * XXX Probably should store the compression type directly in the lookup table
863 wim_resource_compression_type(const struct wim_lookup_table_entry *lte)
865 if (!(lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
866 || lte->resource_location != RESOURCE_IN_WIM)
867 return WIMLIB_COMPRESSION_TYPE_NONE;
868 return wimlib_get_compression_type(lte->wim);
871 /* Resolve an inode's lookup table entries
873 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
874 * lookup table) with pointers directly to the lookup table entries. A circular
875 * linked list of streams sharing the same lookup table entry is created.
877 * This function always succeeds; unresolved lookup table entries are given a
881 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
884 if (!inode->i_resolved) {
885 struct wim_lookup_table_entry *lte;
886 /* Resolve the default file stream */
887 lte = __lookup_resource(table, inode->i_hash);
889 inode->i_resolved = 1;
891 /* Resolve the alternate data streams */
892 for (u16 i = 0; i < inode->i_num_ads; i++) {
893 struct wim_ads_entry *cur_entry = &inode->i_ads_entries[i];
894 lte = __lookup_resource(table, cur_entry->hash);
895 cur_entry->lte = lte;
901 inode_unresolve_ltes(struct wim_inode *inode)
903 if (inode->i_resolved) {
905 copy_hash(inode->i_hash, inode->i_lte->hash);
907 zero_out_hash(inode->i_hash);
909 for (u16 i = 0; i < inode->i_num_ads; i++) {
910 if (inode->i_ads_entries[i].lte)
911 copy_hash(inode->i_ads_entries[i].hash,
912 inode->i_ads_entries[i].lte->hash);
914 zero_out_hash(inode->i_ads_entries[i].hash);
916 inode->i_resolved = 0;
921 * Returns the lookup table entry for stream @stream_idx of the inode, where
922 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
923 * corresponds to an alternate data stream.
925 * This works for both resolved and un-resolved inodes.
927 struct wim_lookup_table_entry *
928 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
929 const struct wim_lookup_table *table)
931 if (inode->i_resolved)
932 return inode_stream_lte_resolved(inode, stream_idx);
934 return inode_stream_lte_unresolved(inode, stream_idx, table);
937 struct wim_lookup_table_entry *
938 inode_unnamed_lte_resolved(const struct wim_inode *inode)
940 wimlib_assert(inode->i_resolved);
941 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
942 if (inode_stream_name_nbytes(inode, i) == 0 &&
943 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
945 return inode_stream_lte_resolved(inode, i);
951 struct wim_lookup_table_entry *
952 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
953 const struct wim_lookup_table *table)
955 wimlib_assert(!inode->i_resolved);
956 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
957 if (inode_stream_name_nbytes(inode, i) == 0 &&
958 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
960 return inode_stream_lte_unresolved(inode, i, table);
966 /* Return the lookup table entry for the unnamed data stream of an inode, or
967 * NULL if there is none.
969 * You'd think this would be easier than it actually is, since the unnamed data
970 * stream should be the one referenced from the inode itself. Alas, if there
971 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
972 * data stream in one of the alternate data streams instead of inside the WIM
973 * dentry itself. So we need to check the alternate data streams too.
975 * Also, note that a dentry may appear to have more than one unnamed stream, but
976 * if the SHA1 message digest is all 0's then the corresponding stream does not
977 * really "count" (this is the case for the inode's own file stream when the
978 * file stream that should be there is actually in one of the alternate stream
979 * entries.). This is despite the fact that we may need to extract such a
980 * missing entry as an empty file or empty named data stream.
982 struct wim_lookup_table_entry *
983 inode_unnamed_lte(const struct wim_inode *inode,
984 const struct wim_lookup_table *table)
986 if (inode->i_resolved)
987 return inode_unnamed_lte_resolved(inode);
989 return inode_unnamed_lte_unresolved(inode, table);
993 lte_add_stream_size(struct wim_lookup_table_entry *lte, void *total_bytes_p)
995 *(u64*)total_bytes_p += lte->resource_entry.size;
1000 lookup_table_total_stream_size(struct wim_lookup_table *table)
1003 for_lookup_table_entry(table, lte_add_stream_size, &total_size);
1007 struct wim_lookup_table_entry **
1008 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1010 wimlib_assert(lte->unhashed);
1011 struct wim_inode *inode = lte->back_inode;
1012 u32 stream_id = lte->back_stream_id;
1014 return &inode->i_lte;
1016 for (u16 i = 0; i < inode->i_num_ads; i++)
1017 if (inode->i_ads_entries[i].stream_id == stream_id)
1018 return &inode->i_ads_entries[i].lte;
1023 /* Calculate the SHA1 message digest of a stream and move it from the list of
1024 * unhashed streams to the stream lookup table, possibly joining it with an
1025 * existing lookup table entry for an identical stream.
1027 * @lte: An unhashed lookup table entry.
1028 * @lookup_table: Lookup table for the WIM.
1029 * @lte_ret: On success, write a pointer to the resulting lookup table
1030 * entry to this location. This will be the same as @lte
1031 * if it was inserted into the lookup table, or different if
1032 * a duplicate stream was found.
1034 * Returns 0 on success; nonzero if there is an error reading the stream.
1037 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1038 struct wim_lookup_table *lookup_table,
1039 struct wim_lookup_table_entry **lte_ret)
1042 struct wim_lookup_table_entry *duplicate_lte;
1043 struct wim_lookup_table_entry **back_ptr;
1045 wimlib_assert(lte->unhashed);
1047 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1048 * union with the SHA1 message digest and will no longer be valid once
1049 * the SHA1 has been calculated. */
1050 back_ptr = retrieve_lte_pointer(lte);
1052 ret = sha1_resource(lte);
1056 /* Look for a duplicate stream */
1057 duplicate_lte = __lookup_resource(lookup_table, lte->hash);
1058 list_del(<e->unhashed_list);
1059 if (duplicate_lte) {
1060 /* We have a duplicate stream. Transfer the reference counts
1061 * from this stream to the duplicate, update the reference to
1062 * this stream (in an inode or ads_entry) to point to the
1063 * duplicate, then free this stream. */
1064 wimlib_assert(!(duplicate_lte->unhashed));
1065 duplicate_lte->refcnt += lte->refcnt;
1066 duplicate_lte->out_refcnt += lte->refcnt;
1067 *back_ptr = duplicate_lte;
1068 free_lookup_table_entry(lte);
1069 lte = duplicate_lte;
1071 /* No duplicate stream, so we need to insert
1072 * this stream into the lookup table and treat
1073 * it as a hashed stream. */
1074 lookup_table_insert(lookup_table, lte);