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 *wim)
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 wim->hdr.lookup_table_res_entry.offset,
397 wim->hdr.lookup_table_res_entry.original_size);
399 if (resource_is_compressed(&wim->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 = wim->hdr.lookup_table_res_entry.size /
406 sizeof(struct wim_lookup_table_entry_disk);
407 table = new_lookup_table(num_entries * 2 + 1);
409 ERROR("Failed to allocate stream hash table of size %zu",
410 num_entries * 2 + 1);
411 return WIMLIB_ERR_NOMEM;
414 wim->current_image = 0;
415 offset = wim->hdr.lookup_table_res_entry.offset;
416 buf_entries_remaining = 0;
417 for (; num_entries != 0;
418 num_entries--, buf_entries_remaining--, disk_entry++)
420 if (buf_entries_remaining == 0) {
421 size_t entries_to_read, bytes_to_read;
423 entries_to_read = min(ARRAY_LEN(table_buf), num_entries);
424 bytes_to_read = entries_to_read * sizeof(struct wim_lookup_table_entry_disk);
425 if (full_pread(wim->in_fd, table_buf,
426 bytes_to_read, offset) != bytes_to_read)
428 ERROR_WITH_ERRNO("Error reading lookup table "
429 "(offset=%"PRIu64")", offset);
430 ret = WIMLIB_ERR_READ;
431 goto out_free_lookup_table;
433 offset += bytes_to_read;
434 disk_entry = table_buf;
435 buf_entries_remaining = entries_to_read;
437 cur_entry = new_lookup_table_entry();
439 ret = WIMLIB_ERR_NOMEM;
440 goto out_free_lookup_table;
443 cur_entry->wim = wim;
444 cur_entry->resource_location = RESOURCE_IN_WIM;
445 get_resource_entry(&disk_entry->resource_entry, &cur_entry->resource_entry);
446 cur_entry->part_number = le16_to_cpu(disk_entry->part_number);
447 cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
448 copy_hash(cur_entry->hash, disk_entry->hash);
450 if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
451 cur_entry->compression_type = wim->compression_type;
453 BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
455 if (cur_entry->part_number != wim->hdr.part_number) {
456 WARNING("A lookup table entry in part %hu of the WIM "
457 "points to part %hu (ignoring it)",
458 wim->hdr.part_number, cur_entry->part_number);
459 free_lookup_table_entry(cur_entry);
463 if (is_zero_hash(cur_entry->hash)) {
464 WARNING("The WIM lookup table contains an entry with a "
465 "SHA1 message digest of all 0's (ignoring it)");
466 free_lookup_table_entry(cur_entry);
470 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
471 && (cur_entry->resource_entry.size !=
472 cur_entry->resource_entry.original_size))
474 if (wimlib_print_errors) {
475 WARNING("Found uncompressed resource with "
476 "original size (%"PRIu64") not the same "
477 "as compressed size (%"PRIu64")",
478 cur_entry->resource_entry.original_size,
479 cur_entry->resource_entry.size);
480 if (cur_entry->resource_entry.original_size) {
481 WARNING("Overriding compressed size with original size.");
482 cur_entry->resource_entry.size =
483 cur_entry->resource_entry.original_size;
485 WARNING("Overriding original size with compressed size");
486 cur_entry->resource_entry.original_size =
487 cur_entry->resource_entry.size;
492 if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
493 /* Lookup table entry for a metadata resource */
494 if (cur_entry->refcnt != 1) {
495 if (wimlib_print_errors) {
496 ERROR("Found metadata resource with refcnt != 1:");
497 print_lookup_table_entry(cur_entry, stderr);
499 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
500 goto out_free_cur_entry;
503 if (wim->hdr.part_number != 1) {
504 WARNING("Ignoring metadata resource found in a "
505 "non-first part of the split WIM");
506 free_lookup_table_entry(cur_entry);
509 if (wim->current_image == wim->hdr.image_count) {
510 WARNING("The WIM header says there are %u images "
511 "in the WIM, but we found more metadata "
512 "resources than this (ignoring the extra)",
513 wim->hdr.image_count);
514 free_lookup_table_entry(cur_entry);
518 /* Notice very carefully: We are assigning the metadata
519 * resources in the exact order mirrored by their lookup
520 * table entries on disk, which is the behavior of
521 * Microsoft's software. In particular, this overrides
522 * the actual locations of the metadata resources
523 * themselves in the WIM file as well as any information
524 * written in the XML data. */
525 DEBUG("Found metadata resource for image %u at "
527 wim->current_image + 1,
528 cur_entry->resource_entry.offset);
530 wim->current_image++]->metadata_lte = cur_entry;
532 /* Lookup table entry for a stream that is not a
533 * metadata resource */
534 duplicate_entry = __lookup_resource(table, cur_entry->hash);
535 if (duplicate_entry) {
536 if (wimlib_print_errors) {
537 WARNING("The WIM lookup table contains two entries with the "
538 "same SHA1 message digest!");
539 WARNING("The first entry is:");
540 print_lookup_table_entry(duplicate_entry, stderr);
541 WARNING("The second entry is:");
542 print_lookup_table_entry(cur_entry, stderr);
544 free_lookup_table_entry(cur_entry);
547 lookup_table_insert(table, cur_entry);
552 if (wim->hdr.part_number == 1 && wim->current_image != wim->hdr.image_count) {
553 WARNING("The header of \"%"TS"\" says there are %u images in\n"
554 " the WIM, but we only found %d metadata resources! Acting as if\n"
555 " the header specified only %d images instead.",
556 wim->filename, wim->hdr.image_count,
557 wim->current_image, wim->current_image);
558 for (int i = wim->current_image; i < wim->hdr.image_count; i++)
559 put_image_metadata(wim->image_metadata[i], NULL);
560 wim->hdr.image_count = wim->current_image;
562 DEBUG("Done reading lookup table.");
563 wim->lookup_table = table;
568 out_free_lookup_table:
569 free_lookup_table(table);
571 wim->current_image = 0;
577 write_lookup_table_entry(const struct wim_lookup_table_entry *lte,
578 struct wim_lookup_table_entry_disk *disk_entry)
580 put_resource_entry(<e->output_resource_entry, &disk_entry->resource_entry);
581 disk_entry->part_number = cpu_to_le16(lte->part_number);
582 disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
583 copy_hash(disk_entry->hash, lte->hash);
587 write_lookup_table_from_stream_list(struct list_head *stream_list,
589 struct resource_entry *out_res_entry)
593 struct wim_lookup_table_entry_disk
594 table_buf[BUFFER_SIZE / sizeof(struct wim_lookup_table_entry_disk)]
595 _aligned_attribute(8);
597 size_t bytes_to_write;
598 struct wim_lookup_table_entry *lte;
601 start_offset = filedes_offset(out_fd);
602 if (start_offset == -1)
607 list_for_each_entry(lte, stream_list, lookup_table_list) {
608 if (cur_idx == ARRAY_LEN(table_buf)) {
609 bytes_to_write = sizeof(table_buf);
610 if (full_write(out_fd, table_buf,
611 bytes_to_write) != bytes_to_write)
613 table_size += bytes_to_write;
616 write_lookup_table_entry(lte, &table_buf[cur_idx]);
620 bytes_to_write = cur_idx * sizeof(struct wim_lookup_table_entry_disk);
621 if (full_write(out_fd, table_buf,
622 bytes_to_write) != bytes_to_write)
624 table_size += bytes_to_write;
626 out_res_entry->offset = start_offset;
627 out_res_entry->size = table_size;
628 out_res_entry->original_size = table_size;
629 out_res_entry->flags = WIM_RESHDR_FLAG_METADATA;
634 ERROR_WITH_ERRNO("Failed to write lookup table");
635 ret = WIMLIB_ERR_WRITE;
640 append_lookup_table_entry(struct wim_lookup_table_entry *lte, void *_list)
642 if (lte->out_refcnt != 0)
643 list_add_tail(<e->lookup_table_list, (struct list_head*)_list);
647 /* Writes the WIM lookup table to the output file. */
649 write_lookup_table(WIMStruct *wim, int image, struct resource_entry *out_res_entry)
651 LIST_HEAD(stream_list);
655 if (image == WIMLIB_ALL_IMAGES) {
657 end_image = wim->hdr.image_count;
663 for (int i = start_image; i <= end_image; i++) {
664 struct wim_lookup_table_entry *metadata_lte;
666 metadata_lte = wim->image_metadata[i - 1]->metadata_lte;
667 metadata_lte->out_refcnt = 1;
668 metadata_lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA;
669 append_lookup_table_entry(metadata_lte, &stream_list);
671 for_lookup_table_entry(wim->lookup_table,
672 append_lookup_table_entry,
674 return write_lookup_table_from_stream_list(&stream_list,
680 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
682 lte->real_refcnt = 0;
687 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *_ignore)
694 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *_ignore)
696 if (lte->extracted_file != NULL) {
697 FREE(lte->extracted_file);
698 lte->extracted_file = NULL;
704 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
710 tfprintf(out, T("Offset = %"PRIu64" bytes\n"),
711 lte->resource_entry.offset);
713 tfprintf(out, T("Size = %"PRIu64" bytes\n"),
714 (u64)lte->resource_entry.size);
716 tfprintf(out, T("Original size = %"PRIu64" bytes\n"),
717 lte->resource_entry.original_size);
719 tfprintf(out, T("Part Number = %hu\n"), lte->part_number);
720 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
723 tfprintf(out, T("(Unhashed: inode %p, stream_id = %u)\n"),
724 lte->back_inode, lte->back_stream_id);
726 tfprintf(out, T("Hash = 0x"));
727 print_hash(lte->hash, out);
731 tfprintf(out, T("Flags = "));
732 u8 flags = lte->resource_entry.flags;
733 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
734 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
735 if (flags & WIM_RESHDR_FLAG_FREE)
736 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
737 if (flags & WIM_RESHDR_FLAG_METADATA)
738 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
739 if (flags & WIM_RESHDR_FLAG_SPANNED)
740 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
742 switch (lte->resource_location) {
743 case RESOURCE_IN_WIM:
744 if (lte->wim->filename) {
745 tfprintf(out, T("WIM file = `%"TS"'\n"),
751 case RESOURCE_WIN32_ENCRYPTED:
753 case RESOURCE_IN_FILE_ON_DISK:
755 tfprintf(out, T("File on Disk = `%"TS"'\n"),
759 case RESOURCE_IN_STAGING_FILE:
760 tfprintf(out, T("Staging File = `%"TS"'\n"),
761 lte->staging_file_name);
771 lte_to_wimlib_resource_entry(const struct wim_lookup_table_entry *lte,
772 struct wimlib_resource_entry *wentry)
774 wentry->uncompressed_size = lte->resource_entry.original_size;
775 wentry->compressed_size = lte->resource_entry.size;
776 wentry->offset = lte->resource_entry.offset;
777 copy_hash(wentry->sha1_hash, lte->hash);
778 wentry->part_number = lte->part_number;
779 wentry->reference_count = lte->refcnt;
780 wentry->is_compressed = (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) != 0;
781 wentry->is_metadata = (lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) != 0;
782 wentry->is_free = (lte->resource_entry.flags & WIM_RESHDR_FLAG_FREE) != 0;
783 wentry->is_spanned = (lte->resource_entry.flags & WIM_RESHDR_FLAG_SPANNED) != 0;
786 struct iterate_lte_context {
787 wimlib_iterate_lookup_table_callback_t cb;
792 do_iterate_lte(struct wim_lookup_table_entry *lte, void *_ctx)
794 struct iterate_lte_context *ctx = _ctx;
795 struct wimlib_resource_entry entry;
797 lte_to_wimlib_resource_entry(lte, &entry);
798 return (*ctx->cb)(&entry, ctx->user_ctx);
802 wimlib_iterate_lookup_table(WIMStruct *wim, int flags,
803 wimlib_iterate_lookup_table_callback_t cb,
806 struct iterate_lte_context ctx = {
808 .user_ctx = user_ctx,
810 if (wim->hdr.part_number == 1) {
812 for (int i = 0; i < wim->hdr.image_count; i++) {
813 ret = do_iterate_lte(wim->image_metadata[i]->metadata_lte,
819 return for_lookup_table_entry(wim->lookup_table, do_iterate_lte, &ctx);
823 do_print_lookup_table_entry(struct wim_lookup_table_entry *lte, void *fp)
825 print_lookup_table_entry(lte, (FILE*)fp);
833 wimlib_print_lookup_table(WIMStruct *wim)
835 for (int i = 0; i < wim->hdr.image_count; i++)
836 print_lookup_table_entry(wim->image_metadata[i]->metadata_lte, stdout);
837 for_lookup_table_entry(wim->lookup_table,
838 do_print_lookup_table_entry,
842 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
843 * lookup table, or NULL if there is none. */
844 struct wim_lookup_table_entry *
845 __lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
848 struct wim_lookup_table_entry *lte;
849 struct hlist_node *pos;
851 wimlib_assert(table != NULL);
852 wimlib_assert(hash != NULL);
854 i = *(size_t*)hash % table->capacity;
855 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
856 if (hashes_equal(hash, lte->hash))
863 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
866 * This is only for pre-resolved inodes.
869 lookup_resource(WIMStruct *wim,
872 struct wim_dentry **dentry_ret,
873 struct wim_lookup_table_entry **lte_ret,
876 struct wim_dentry *dentry;
877 struct wim_lookup_table_entry *lte;
879 const tchar *stream_name = NULL;
880 struct wim_inode *inode;
883 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
884 stream_name = path_stream_name(path);
886 p = (tchar*)stream_name - 1;
891 dentry = get_dentry(wim, path);
897 inode = dentry->d_inode;
899 if (!inode->i_resolved)
900 if (inode_resolve_ltes(inode, wim->lookup_table))
903 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
904 && inode_is_directory(inode))
908 struct wim_ads_entry *ads_entry;
910 ads_entry = inode_get_ads_entry(inode, stream_name,
913 stream_idx = ads_idx + 1;
914 lte = ads_entry->lte;
925 *dentry_ret = dentry;
929 *stream_idx_ret = stream_idx;
934 /* Resolve an inode's lookup table entries
936 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
937 * lookup table) with pointers directly to the lookup table entries. A circular
938 * linked list of streams sharing the same lookup table entry is created.
940 * This function always succeeds; unresolved lookup table entries are given a
944 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
948 if (!inode->i_resolved) {
949 struct wim_lookup_table_entry *lte, *ads_lte;
951 /* Resolve the default file stream */
953 hash = inode->i_hash;
954 if (!is_zero_hash(hash)) {
955 lte = __lookup_resource(table, hash);
957 goto resource_not_found;
960 /* Resolve the alternate data streams */
961 struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
962 for (u16 i = 0; i < inode->i_num_ads; i++) {
963 struct wim_ads_entry *cur_entry;
966 cur_entry = &inode->i_ads_entries[i];
967 hash = cur_entry->hash;
968 if (!is_zero_hash(hash)) {
969 ads_lte = __lookup_resource(table, hash);
970 if (unlikely(!ads_lte))
971 goto resource_not_found;
973 ads_ltes[i] = ads_lte;
976 for (u16 i = 0; i < inode->i_num_ads; i++)
977 inode->i_ads_entries[i].lte = ads_ltes[i];
978 inode->i_resolved = 1;
982 if (wimlib_print_errors) {
983 ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
984 tfprintf(stderr, T(" SHA-1 message digest of missing resource:\n "));
985 print_hash(hash, stderr);
986 tputc(T('\n'), stderr);
988 return WIMLIB_ERR_RESOURCE_NOT_FOUND;
992 inode_unresolve_ltes(struct wim_inode *inode)
994 if (inode->i_resolved) {
996 copy_hash(inode->i_hash, inode->i_lte->hash);
998 zero_out_hash(inode->i_hash);
1000 for (u16 i = 0; i < inode->i_num_ads; i++) {
1001 if (inode->i_ads_entries[i].lte)
1002 copy_hash(inode->i_ads_entries[i].hash,
1003 inode->i_ads_entries[i].lte->hash);
1005 zero_out_hash(inode->i_ads_entries[i].hash);
1007 inode->i_resolved = 0;
1012 * Returns the lookup table entry for stream @stream_idx of the inode, where
1013 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
1014 * corresponds to an alternate data stream.
1016 * This works for both resolved and un-resolved inodes.
1018 struct wim_lookup_table_entry *
1019 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
1020 const struct wim_lookup_table *table)
1022 if (inode->i_resolved)
1023 return inode_stream_lte_resolved(inode, stream_idx);
1025 return inode_stream_lte_unresolved(inode, stream_idx, table);
1028 struct wim_lookup_table_entry *
1029 inode_unnamed_lte_resolved(const struct wim_inode *inode)
1031 wimlib_assert(inode->i_resolved);
1032 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1033 if (inode_stream_name_nbytes(inode, i) == 0 &&
1034 !is_zero_hash(inode_stream_hash_resolved(inode, i)))
1036 return inode_stream_lte_resolved(inode, i);
1042 struct wim_lookup_table_entry *
1043 inode_unnamed_lte_unresolved(const struct wim_inode *inode,
1044 const struct wim_lookup_table *table)
1046 wimlib_assert(!inode->i_resolved);
1047 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
1048 if (inode_stream_name_nbytes(inode, i) == 0 &&
1049 !is_zero_hash(inode_stream_hash_unresolved(inode, i)))
1051 return inode_stream_lte_unresolved(inode, i, table);
1057 /* Return the lookup table entry for the unnamed data stream of an inode, or
1058 * NULL if there is none.
1060 * You'd think this would be easier than it actually is, since the unnamed data
1061 * stream should be the one referenced from the inode itself. Alas, if there
1062 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
1063 * data stream in one of the alternate data streams instead of inside the WIM
1064 * dentry itself. So we need to check the alternate data streams too.
1066 * Also, note that a dentry may appear to have more than one unnamed stream, but
1067 * if the SHA1 message digest is all 0's then the corresponding stream does not
1068 * really "count" (this is the case for the inode's own file stream when the
1069 * file stream that should be there is actually in one of the alternate stream
1070 * entries.). This is despite the fact that we may need to extract such a
1071 * missing entry as an empty file or empty named data stream.
1073 struct wim_lookup_table_entry *
1074 inode_unnamed_lte(const struct wim_inode *inode,
1075 const struct wim_lookup_table *table)
1077 if (inode->i_resolved)
1078 return inode_unnamed_lte_resolved(inode);
1080 return inode_unnamed_lte_unresolved(inode, table);
1084 lte_add_stream_size(struct wim_lookup_table_entry *lte, void *total_bytes_p)
1086 *(u64*)total_bytes_p += lte->resource_entry.size;
1091 lookup_table_total_stream_size(struct wim_lookup_table *table)
1094 for_lookup_table_entry(table, lte_add_stream_size, &total_size);
1098 struct wim_lookup_table_entry **
1099 retrieve_lte_pointer(struct wim_lookup_table_entry *lte)
1101 wimlib_assert(lte->unhashed);
1102 struct wim_inode *inode = lte->back_inode;
1103 u32 stream_id = lte->back_stream_id;
1105 return &inode->i_lte;
1107 for (u16 i = 0; i < inode->i_num_ads; i++)
1108 if (inode->i_ads_entries[i].stream_id == stream_id)
1109 return &inode->i_ads_entries[i].lte;
1114 /* Calculate the SHA1 message digest of a stream and move it from the list of
1115 * unhashed streams to the stream lookup table, possibly joining it with an
1116 * existing lookup table entry for an identical stream.
1118 * @lte: An unhashed lookup table entry.
1119 * @lookup_table: Lookup table for the WIM.
1120 * @lte_ret: On success, write a pointer to the resulting lookup table
1121 * entry to this location. This will be the same as @lte
1122 * if it was inserted into the lookup table, or different if
1123 * a duplicate stream was found.
1125 * Returns 0 on success; nonzero if there is an error reading the stream.
1128 hash_unhashed_stream(struct wim_lookup_table_entry *lte,
1129 struct wim_lookup_table *lookup_table,
1130 struct wim_lookup_table_entry **lte_ret)
1133 struct wim_lookup_table_entry *duplicate_lte;
1134 struct wim_lookup_table_entry **back_ptr;
1136 wimlib_assert(lte->unhashed);
1138 /* back_ptr must be saved because @back_inode and @back_stream_id are in
1139 * union with the SHA1 message digest and will no longer be valid once
1140 * the SHA1 has been calculated. */
1141 back_ptr = retrieve_lte_pointer(lte);
1143 ret = sha1_resource(lte);
1147 /* Look for a duplicate stream */
1148 duplicate_lte = __lookup_resource(lookup_table, lte->hash);
1149 list_del(<e->unhashed_list);
1150 if (duplicate_lte) {
1151 /* We have a duplicate stream. Transfer the reference counts
1152 * from this stream to the duplicate, update the reference to
1153 * this stream (in an inode or ads_entry) to point to the
1154 * duplicate, then free this stream. */
1155 wimlib_assert(!(duplicate_lte->unhashed));
1156 duplicate_lte->refcnt += lte->refcnt;
1157 duplicate_lte->out_refcnt += lte->refcnt;
1158 *back_ptr = duplicate_lte;
1159 free_lookup_table_entry(lte);
1160 lte = duplicate_lte;
1162 /* No duplicate stream, so we need to insert
1163 * this stream into the lookup table and treat
1164 * it as a hashed stream. */
1165 lookup_table_insert(lookup_table, lte);