4 * Lookup table, implemented as a hash table, that maps dentries to file
9 * Copyright (C) 2012 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/.
27 #include "wimlib_internal.h"
28 #include "lookup_table.h"
36 struct lookup_table *new_lookup_table(size_t capacity)
38 struct lookup_table *table;
39 struct hlist_head *array;
41 table = MALLOC(sizeof(struct lookup_table));
44 array = CALLOC(capacity, sizeof(array[0]));
49 table->num_entries = 0;
50 table->capacity = capacity;
54 ERROR("Failed to allocate memory for lookup table with capacity %zu",
59 struct lookup_table_entry *new_lookup_table_entry()
61 struct lookup_table_entry *lte;
63 lte = CALLOC(1, sizeof(struct lookup_table_entry));
68 ERROR("Out of memory (tried to allocate %zu bytes for "
69 "lookup table entry)",
70 sizeof(struct lookup_table_entry));
75 struct lookup_table_entry *
76 clone_lookup_table_entry(const struct lookup_table_entry *old)
78 struct lookup_table_entry *new;
80 new = MALLOC(sizeof(*new));
84 memcpy(new, old, sizeof(*old));
85 new->extracted_file = NULL;
86 switch (new->resource_location) {
87 case RESOURCE_IN_STAGING_FILE:
88 case RESOURCE_IN_FILE_ON_DISK:
89 wimlib_assert((void*)&old->file_on_disk ==
90 (void*)&old->staging_file_name);
91 new->staging_file_name = STRDUP(old->staging_file_name);
92 if (!new->staging_file_name)
95 case RESOURCE_IN_ATTACHED_BUFFER:
96 new->attached_buffer = MALLOC(wim_resource_size(old));
97 if (!new->attached_buffer)
99 memcpy(new->attached_buffer, old->attached_buffer,
100 wim_resource_size(old));
103 case RESOURCE_IN_NTFS_VOLUME:
105 struct ntfs_location *loc;
106 loc = MALLOC(sizeof(*loc));
109 memcpy(loc, old->ntfs_loc, sizeof(*loc));
110 loc->path_utf8 = NULL;
111 loc->stream_name_utf16 = NULL;
113 loc->path_utf8 = STRDUP(old->ntfs_loc->path_utf8);
116 loc->stream_name_utf16 = MALLOC(loc->stream_name_utf16_num_chars * 2);
117 if (!loc->stream_name_utf16)
119 memcpy(loc->stream_name_utf16,
120 old->ntfs_loc->stream_name_utf16,
121 loc->stream_name_utf16_num_chars * 2);
128 free_lookup_table_entry(new);
132 void free_lookup_table_entry(struct lookup_table_entry *lte)
135 switch (lte->resource_location) {
136 case RESOURCE_IN_STAGING_FILE:
137 case RESOURCE_IN_ATTACHED_BUFFER:
138 case RESOURCE_IN_FILE_ON_DISK:
139 wimlib_assert(((void*)<e->file_on_disk ==
140 (void*)<e->staging_file_name)
141 && ((void*)<e->file_on_disk ==
142 (void*)<e->attached_buffer));
143 FREE(lte->file_on_disk);
146 case RESOURCE_IN_NTFS_VOLUME:
148 FREE(lte->ntfs_loc->path_utf8);
149 FREE(lte->ntfs_loc->stream_name_utf16);
157 FREE(lte->extracted_file);
162 static int do_free_lookup_table_entry(struct lookup_table_entry *entry,
165 free_lookup_table_entry(entry);
170 void free_lookup_table(struct lookup_table *table)
172 DEBUG2("Freeing lookup table");
175 for_lookup_table_entry(table,
176 do_free_lookup_table_entry,
185 * Inserts an entry into the lookup table.
187 * @table: A pointer to the lookup table.
188 * @entry: A pointer to the entry to insert.
190 void lookup_table_insert(struct lookup_table *table,
191 struct lookup_table_entry *lte)
193 size_t i = lte->hash_short % table->capacity;
194 hlist_add_head(<e->hash_list, &table->array[i]);
196 /* XXX Make the table grow when too many entries have been inserted. */
197 table->num_entries++;
200 static void finalize_lte(struct lookup_table_entry *lte)
203 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
204 unlink(lte->staging_file_name);
205 wimlib_assert(lte->staging_list.next);
206 wimlib_assert(lte->staging_list.prev);
207 list_del(<e->staging_list);
210 free_lookup_table_entry(lte);
213 /* Decrements the reference count for the lookup table entry @lte. If its
214 * reference count reaches 0, it is unlinked from the lookup table. If,
215 * furthermore, the entry has no opened file descriptors associated with it, the
217 void lte_decrement_refcnt(struct lookup_table_entry *lte,
218 struct lookup_table *table)
221 wimlib_assert(lte->refcnt);
222 if (--lte->refcnt == 0) {
223 lookup_table_unlink(table, lte);
225 if (lte->num_opened_fds == 0)
232 void lte_decrement_num_opened_fds(struct lookup_table_entry *lte)
235 wimlib_assert(lte->num_opened_fds);
236 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
242 * Calls a function on all the entries in the lookup table. Stop early and
243 * return nonzero if any call to the function returns nonzero.
245 int for_lookup_table_entry(struct lookup_table *table,
246 int (*visitor)(struct lookup_table_entry *, void *),
249 struct lookup_table_entry *lte;
250 struct hlist_node *pos, *tmp;
253 for (size_t i = 0; i < table->capacity; i++) {
254 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
257 ret = visitor(lte, arg);
267 * Reads the lookup table from a WIM file.
269 int read_lookup_table(WIMStruct *w)
272 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
274 struct lookup_table *table;
275 struct lookup_table_entry *cur_entry = NULL, *duplicate_entry;
277 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
278 w->hdr.lookup_table_res_entry.offset,
279 w->hdr.lookup_table_res_entry.original_size);
281 if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0)
283 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
285 w->hdr.lookup_table_res_entry.offset);
286 return WIMLIB_ERR_READ;
289 num_entries = w->hdr.lookup_table_res_entry.original_size /
290 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
291 table = new_lookup_table(num_entries * 2 + 1);
293 return WIMLIB_ERR_NOMEM;
295 while (num_entries--) {
298 if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
300 ERROR("Unexpected EOF in WIM lookup table!");
302 ERROR_WITH_ERRNO("Error reading WIM lookup "
305 ret = WIMLIB_ERR_READ;
308 cur_entry = new_lookup_table_entry();
310 ret = WIMLIB_ERR_NOMEM;
314 cur_entry->resource_location = RESOURCE_IN_WIM;
316 p = get_resource_entry(buf, &cur_entry->resource_entry);
317 p = get_u16(p, &cur_entry->part_number);
318 p = get_u32(p, &cur_entry->refcnt);
319 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
321 if (cur_entry->part_number != w->hdr.part_number) {
322 ERROR("A lookup table entry in part %hu of the WIM "
323 "points to part %hu",
324 w->hdr.part_number, cur_entry->part_number);
325 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
326 goto out_free_cur_entry;
329 if (is_zero_hash(cur_entry->hash)) {
330 ERROR("The WIM lookup table contains an entry with a "
331 "SHA1 message digest of all 0's");
332 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
333 goto out_free_cur_entry;
336 /* Ordinarily, no two streams should share the same SHA1 message
337 * digest. However, this constraint can be broken for metadata
338 * resources--- two identical images will have the same metadata
339 * resource, but their lookup table entries are not shared. */
340 duplicate_entry = __lookup_resource(table, cur_entry->hash);
342 && !((duplicate_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
343 && cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
345 ERROR("The WIM lookup table contains two entries with the "
346 "same SHA1 message digest!");
347 ERROR("The first entry is:");
348 print_lookup_table_entry(duplicate_entry);
349 ERROR("The second entry is:");
350 print_lookup_table_entry(cur_entry);
351 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
352 goto out_free_cur_entry;
355 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
356 && (cur_entry->resource_entry.size !=
357 cur_entry->resource_entry.original_size))
359 ERROR("Found uncompressed resource with original size "
360 "not the same as compressed size");
361 ERROR("The lookup table entry for the resource is as follows:");
362 print_lookup_table_entry(cur_entry);
363 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
364 goto out_free_cur_entry;
366 lookup_table_insert(table, cur_entry);
369 DEBUG("Done reading lookup table.");
370 w->lookup_table = table;
375 free_lookup_table(table);
381 * Writes a lookup table entry to the output file.
383 int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out)
386 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
391 /* Don't write entries that have not had file resources or metadata
392 * resources written for them. */
393 if (lte->out_refcnt == 0)
396 if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
397 DEBUG("Writing metadata entry at %lu (orig size = %zu)",
398 ftello(out), lte->output_resource_entry.original_size);
400 p = put_resource_entry(buf, <e->output_resource_entry);
401 p = put_u16(p, lte->part_number);
402 p = put_u32(p, lte->out_refcnt);
403 p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
404 if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
405 ERROR_WITH_ERRNO("Failed to write lookup table entry");
406 return WIMLIB_ERR_WRITE;
412 int lte_zero_real_refcnt(struct lookup_table_entry *lte, void *ignore)
414 lte->real_refcnt = 0;
418 int lte_zero_out_refcnt(struct lookup_table_entry *lte, void *ignore)
424 int lte_free_extracted_file(struct lookup_table_entry *lte, void *ignone)
426 FREE(lte->extracted_file);
427 lte->extracted_file = NULL;
431 void print_lookup_table_entry(const struct lookup_table_entry *lte)
437 printf("Offset = %"PRIu64" bytes\n",
438 lte->resource_entry.offset);
439 printf("Size = %"PRIu64" bytes\n",
440 (u64)lte->resource_entry.size);
441 printf("Original size = %"PRIu64" bytes\n",
442 lte->resource_entry.original_size);
443 printf("Part Number = %hu\n", lte->part_number);
444 printf("Reference Count = %u\n", lte->refcnt);
446 print_hash(lte->hash);
449 u8 flags = lte->resource_entry.flags;
450 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
451 fputs("WIM_RESHDR_FLAG_COMPRESSED, ", stdout);
452 if (flags & WIM_RESHDR_FLAG_FREE)
453 fputs("WIM_RESHDR_FLAG_FREE, ", stdout);
454 if (flags & WIM_RESHDR_FLAG_METADATA)
455 fputs("WIM_RESHDR_FLAG_METADATA, ", stdout);
456 if (flags & WIM_RESHDR_FLAG_SPANNED)
457 fputs("WIM_RESHDR_FLAG_SPANNED, ", stdout);
459 switch (lte->resource_location) {
460 case RESOURCE_IN_WIM:
461 if (lte->wim->filename) {
462 printf("WIM file = `%s'\n",
466 case RESOURCE_IN_FILE_ON_DISK:
467 printf("File on Disk = `%s'\n", lte->file_on_disk);
469 case RESOURCE_IN_STAGING_FILE:
470 printf("Staging File = `%s'\n", lte->staging_file_name);
478 static int do_print_lookup_table_entry(struct lookup_table_entry *lte,
481 print_lookup_table_entry(lte);
486 * Prints the lookup table of a WIM file.
488 WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
490 for_lookup_table_entry(w->lookup_table,
491 do_print_lookup_table_entry,
496 * Looks up an entry in the lookup table.
498 struct lookup_table_entry *
499 __lookup_resource(const struct lookup_table *table, const u8 hash[])
502 struct lookup_table_entry *lte;
503 struct hlist_node *pos;
505 wimlib_assert(table != NULL);
506 wimlib_assert(hash != NULL);
508 i = *(size_t*)hash % table->capacity;
509 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
510 if (hashes_equal(hash, lte->hash))
517 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
520 * This is only for pre-resolved inodes.
522 int lookup_resource(WIMStruct *w, const char *path,
524 struct dentry **dentry_ret,
525 struct lookup_table_entry **lte_ret,
528 struct dentry *dentry;
529 struct lookup_table_entry *lte;
531 const char *stream_name = NULL;
535 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
536 stream_name = path_stream_name(path);
538 p = (char*)stream_name - 1;
543 dentry = get_dentry(w, path);
549 inode = dentry->d_inode;
551 wimlib_assert(inode->resolved);
553 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
554 && inode_is_directory(inode))
558 struct ads_entry *ads_entry;
560 ads_entry = inode_get_ads_entry(inode, stream_name,
563 stream_idx = ads_idx + 1;
564 lte = ads_entry->lte;
575 *dentry_ret = dentry;
579 *stream_idx_ret = stream_idx;
584 static void inode_resolve_ltes(struct inode *inode, struct lookup_table *table)
586 struct lookup_table_entry *lte;
588 wimlib_assert(!inode->resolved);
590 /* Resolve the default file stream */
591 lte = __lookup_resource(table, inode->hash);
593 inode->resolved = true;
595 /* Resolve the alternate data streams */
596 for (u16 i = 0; i < inode->num_ads; i++) {
597 struct ads_entry *cur_entry = &inode->ads_entries[i];
598 lte = __lookup_resource(table, cur_entry->hash);
599 cur_entry->lte = lte;
603 static void inode_unresolve_ltes(struct inode *inode)
605 wimlib_assert(inode->resolved);
607 copy_hash(inode->hash, inode->lte->hash);
609 zero_out_hash(inode->hash);
611 for (u16 i = 0; i < inode->num_ads; i++) {
612 if (inode->ads_entries[i].lte)
613 copy_hash(inode->ads_entries[i].hash,
614 inode->ads_entries[i].lte->hash);
616 zero_out_hash(inode->ads_entries[i].hash);
618 inode->resolved = false;
621 /* Resolve a dentry's lookup table entries
623 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
624 * lookup table) with pointers directly to the lookup table entries. A circular
625 * linked list of streams sharing the same lookup table entry is created.
627 * This function always succeeds; unresolved lookup table entries are given a
630 int dentry_resolve_ltes(struct dentry *dentry, void *table)
632 if (!dentry->d_inode->resolved)
633 inode_resolve_ltes(dentry->d_inode, table);
637 int dentry_unresolve_ltes(struct dentry *dentry, void *ignore)
639 if (dentry->d_inode->resolved)
640 inode_unresolve_ltes(dentry->d_inode);
644 /* Return the lookup table entry for the unnamed data stream of an inode, or
645 * NULL if there is none.
647 * You'd think this would be easier than it actually is, since the unnamed data
648 * stream should be the one referenced from the inode itself. Alas, if there
649 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
650 * data stream in one of the alternate data streams instead of inside the WIM
651 * dentry itself. So we need to check the alternate data streams too.
653 * Also, note that a dentry may appear to have more than one unnamed stream, but
654 * if the SHA1 message digest is all 0's then the corresponding stream does not
655 * really "count" (this is the case for the inode's own file stream when the
656 * file stream that should be there is actually in one of the alternate stream
657 * entries.). This is despite the fact that we may need to extract such a
658 * missing entry as an empty file or empty named data stream.
660 struct lookup_table_entry *
661 inode_unnamed_lte(const struct inode *inode,
662 const struct lookup_table *table)
665 return inode_unnamed_lte_resolved(inode);
667 return inode_unnamed_lte_unresolved(inode, table);