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));
65 ERROR("Out of memory (tried to allocate %zu bytes for "
66 "lookup table entry)",
67 sizeof(struct lookup_table_entry));
76 struct lookup_table_entry *
77 clone_lookup_table_entry(const struct lookup_table_entry *old)
79 struct lookup_table_entry *new;
81 new = MALLOC(sizeof(*new));
85 memcpy(new, old, sizeof(*old));
86 new->extracted_file = NULL;
87 switch (new->resource_location) {
88 case RESOURCE_IN_STAGING_FILE:
89 case RESOURCE_IN_FILE_ON_DISK:
90 wimlib_assert((void*)&old->file_on_disk ==
91 (void*)&old->staging_file_name);
92 new->staging_file_name = STRDUP(old->staging_file_name);
93 if (!new->staging_file_name)
96 case RESOURCE_IN_ATTACHED_BUFFER:
97 new->attached_buffer = MALLOC(wim_resource_size(old));
98 if (!new->attached_buffer)
100 memcpy(new->attached_buffer, old->attached_buffer,
101 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);
127 free_lookup_table_entry(new);
131 void free_lookup_table_entry(struct lookup_table_entry *lte)
134 switch (lte->resource_location) {
135 case RESOURCE_IN_STAGING_FILE:
136 case RESOURCE_IN_ATTACHED_BUFFER:
137 case RESOURCE_IN_FILE_ON_DISK:
138 wimlib_assert(((void*)<e->file_on_disk ==
139 (void*)<e->staging_file_name)
140 && ((void*)<e->file_on_disk ==
141 (void*)<e->attached_buffer));
142 FREE(lte->file_on_disk);
145 case RESOURCE_IN_NTFS_VOLUME:
147 FREE(lte->ntfs_loc->path_utf8);
148 FREE(lte->ntfs_loc->stream_name_utf16);
156 FREE(lte->extracted_file);
161 static int do_free_lookup_table_entry(struct lookup_table_entry *entry,
164 free_lookup_table_entry(entry);
169 void free_lookup_table(struct lookup_table *table)
171 DEBUG2("Freeing lookup table");
174 for_lookup_table_entry(table,
175 do_free_lookup_table_entry,
184 * Inserts an entry into the lookup table.
186 * @table: A pointer to the lookup table.
187 * @entry: A pointer to the entry to insert.
189 void lookup_table_insert(struct lookup_table *table,
190 struct lookup_table_entry *lte)
192 size_t i = lte->hash_short % table->capacity;
193 hlist_add_head(<e->hash_list, &table->array[i]);
195 /* XXX Make the table grow when too many entries have been inserted. */
196 table->num_entries++;
199 static void finalize_lte(struct lookup_table_entry *lte)
202 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
203 unlink(lte->staging_file_name);
204 wimlib_assert(lte->staging_list.next);
205 wimlib_assert(lte->staging_list.prev);
206 list_del(<e->staging_list);
209 free_lookup_table_entry(lte);
212 /* Decrements the reference count for the lookup table entry @lte. If its
213 * reference count reaches 0, it is unlinked from the lookup table. If,
214 * furthermore, the entry has no opened file descriptors associated with it, the
216 void lte_decrement_refcnt(struct lookup_table_entry *lte,
217 struct lookup_table *table)
220 wimlib_assert(lte->refcnt);
221 if (--lte->refcnt == 0) {
222 lookup_table_unlink(table, lte);
224 if (lte->num_opened_fds == 0)
231 void lte_decrement_num_opened_fds(struct lookup_table_entry *lte,
232 struct lookup_table *table)
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) {
282 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
284 w->hdr.lookup_table_res_entry.offset);
285 return WIMLIB_ERR_READ;
288 num_entries = w->hdr.lookup_table_res_entry.original_size /
289 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
290 table = new_lookup_table(num_entries * 2 + 1);
292 return WIMLIB_ERR_NOMEM;
294 while (num_entries--) {
297 if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
299 ERROR("Unexpected EOF in WIM lookup table!");
301 ERROR_WITH_ERRNO("Error reading WIM lookup "
304 ret = WIMLIB_ERR_READ;
307 cur_entry = new_lookup_table_entry();
309 ret = WIMLIB_ERR_NOMEM;
313 cur_entry->resource_location = RESOURCE_IN_WIM;
315 p = get_resource_entry(buf, &cur_entry->resource_entry);
316 p = get_u16(p, &cur_entry->part_number);
317 p = get_u32(p, &cur_entry->refcnt);
318 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
320 if (cur_entry->part_number != w->hdr.part_number) {
321 ERROR("A lookup table entry in part %hu of the WIM "
322 "points to part %hu",
323 w->hdr.part_number, cur_entry->part_number);
324 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
325 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 duplicate_entry = __lookup_resource(table, cur_entry->hash);
337 if (duplicate_entry) {
338 ERROR("The WIM lookup table contains two entries with the "
339 "same SHA1 message digest!");
340 ERROR("The first entry is:");
341 print_lookup_table_entry(duplicate_entry);
342 ERROR("The second entry is:");
343 print_lookup_table_entry(cur_entry);
344 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
345 goto out_free_cur_entry;
348 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
349 && (cur_entry->resource_entry.size !=
350 cur_entry->resource_entry.original_size))
352 ERROR("Found uncompressed resource with original size "
353 "not the same as compressed size");
354 ERROR("The lookup table entry for the resource is as follows:");
355 print_lookup_table_entry(cur_entry);
356 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
357 goto out_free_cur_entry;
359 lookup_table_insert(table, cur_entry);
362 DEBUG("Done reading lookup table.");
363 w->lookup_table = table;
368 free_lookup_table(table);
374 * Writes a lookup table entry to the output file.
376 int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out)
379 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
384 /* Don't write entries that have not had file resources or metadata
385 * resources written for them. */
386 if (lte->out_refcnt == 0)
389 if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
390 DEBUG("Writing metadata entry at %lu (orig size = %zu)",
391 ftello(out), lte->output_resource_entry.original_size);
393 p = put_resource_entry(buf, <e->output_resource_entry);
394 p = put_u16(p, lte->part_number);
395 p = put_u32(p, lte->out_refcnt);
396 p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
397 if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
398 ERROR_WITH_ERRNO("Failed to write lookup table entry");
399 return WIMLIB_ERR_WRITE;
405 int lte_zero_real_refcnt(struct lookup_table_entry *lte, void *ignore)
407 lte->real_refcnt = 0;
411 int lte_zero_out_refcnt(struct lookup_table_entry *lte, void *ignore)
417 int lte_free_extracted_file(struct lookup_table_entry *lte, void *ignone)
419 FREE(lte->extracted_file);
420 lte->extracted_file = NULL;
424 void print_lookup_table_entry(const struct lookup_table_entry *lte)
430 printf("Offset = %"PRIu64" bytes\n",
431 lte->resource_entry.offset);
432 printf("Size = %"PRIu64" bytes\n",
433 (u64)lte->resource_entry.size);
434 printf("Original size = %"PRIu64" bytes\n",
435 lte->resource_entry.original_size);
436 printf("Part Number = %hu\n", lte->part_number);
437 printf("Reference Count = %u\n", lte->refcnt);
439 print_hash(lte->hash);
442 u8 flags = lte->resource_entry.flags;
443 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
444 fputs("WIM_RESHDR_FLAG_COMPRESSED, ", stdout);
445 if (flags & WIM_RESHDR_FLAG_FREE)
446 fputs("WIM_RESHDR_FLAG_FREE, ", stdout);
447 if (flags & WIM_RESHDR_FLAG_METADATA)
448 fputs("WIM_RESHDR_FLAG_METADATA, ", stdout);
449 if (flags & WIM_RESHDR_FLAG_SPANNED)
450 fputs("WIM_RESHDR_FLAG_SPANNED, ", stdout);
452 switch (lte->resource_location) {
453 case RESOURCE_IN_WIM:
454 if (lte->wim->filename) {
455 printf("WIM file = `%s'\n",
459 case RESOURCE_IN_FILE_ON_DISK:
460 printf("File on Disk = `%s'\n", lte->file_on_disk);
462 case RESOURCE_IN_STAGING_FILE:
463 printf("Staging File = `%s'\n", lte->staging_file_name);
471 static int do_print_lookup_table_entry(struct lookup_table_entry *lte,
474 print_lookup_table_entry(lte);
479 * Prints the lookup table of a WIM file.
481 WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
483 for_lookup_table_entry(w->lookup_table,
484 do_print_lookup_table_entry,
489 * Looks up an entry in the lookup table.
491 struct lookup_table_entry *
492 __lookup_resource(const struct lookup_table *table, const u8 hash[])
495 struct lookup_table_entry *lte;
496 struct hlist_node *pos;
498 wimlib_assert(table != NULL);
499 wimlib_assert(hash != NULL);
501 i = *(size_t*)hash % table->capacity;
502 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
503 if (hashes_equal(hash, lte->hash))
510 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
513 * This is only for pre-resolved inodes.
515 int lookup_resource(WIMStruct *w, const char *path,
517 struct dentry **dentry_ret,
518 struct lookup_table_entry **lte_ret,
521 struct dentry *dentry;
522 struct lookup_table_entry *lte;
524 const char *stream_name = NULL;
528 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
529 stream_name = path_stream_name(path);
531 p = (char*)stream_name - 1;
536 dentry = get_dentry(w, path);
542 inode = dentry->d_inode;
544 wimlib_assert(inode->resolved);
546 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
547 && inode_is_directory(inode))
551 struct ads_entry *ads_entry;
553 ads_entry = inode_get_ads_entry(inode, stream_name,
556 stream_idx = ads_idx + 1;
557 lte = ads_entry->lte;
568 *dentry_ret = dentry;
572 *stream_idx_ret = stream_idx;
577 static void inode_resolve_ltes(struct inode *inode, struct lookup_table *table)
579 struct lookup_table_entry *lte;
581 wimlib_assert(!inode->resolved);
583 /* Resolve the default file stream */
584 lte = __lookup_resource(table, inode->hash);
586 inode->resolved = true;
588 /* Resolve the alternate data streams */
589 for (u16 i = 0; i < inode->num_ads; i++) {
590 struct ads_entry *cur_entry = &inode->ads_entries[i];
591 lte = __lookup_resource(table, cur_entry->hash);
592 cur_entry->lte = lte;
596 static void inode_unresolve_ltes(struct inode *inode)
598 wimlib_assert(inode->resolved);
600 copy_hash(inode->hash, inode->lte->hash);
602 zero_out_hash(inode->hash);
604 for (u16 i = 0; i < inode->num_ads; i++) {
605 if (inode->ads_entries[i].lte)
606 copy_hash(inode->ads_entries[i].hash,
607 inode->ads_entries[i].lte->hash);
609 zero_out_hash(inode->ads_entries[i].hash);
611 inode->resolved = false;
614 /* Resolve a dentry's lookup table entries
616 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
617 * lookup table) with pointers directly to the lookup table entries. A circular
618 * linked list of streams sharing the same lookup table entry is created.
620 * This function always succeeds; unresolved lookup table entries are given a
623 int dentry_resolve_ltes(struct dentry *dentry, void *table)
625 if (!dentry->d_inode->resolved)
626 inode_resolve_ltes(dentry->d_inode, table);
630 int dentry_unresolve_ltes(struct dentry *dentry, void *ignore)
632 if (dentry->d_inode->resolved)
633 inode_unresolve_ltes(dentry->d_inode);
637 /* Return the lookup table entry for the unnamed data stream of an inode, or
638 * NULL if there is none.
640 * You'd think this would be easier than it actually is, since the unnamed data
641 * stream should be the one referenced from the inode itself. Alas, if there
642 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
643 * data stream in one of the alternate data streams instead of inside the WIM
644 * dentry itself. So we need to check the alternate data streams too.
646 * Also, note that a dentry may appear to have more than one unnamed stream, but
647 * if the SHA1 message digest is all 0's then the corresponding stream does not
648 * really "count" (this is the case for the inode's own file stream when the
649 * file stream that should be there is actually in one of the alternate stream
650 * entries.). This is despite the fact that we may need to extract such a
651 * missing entry as an empty file or empty named data stream.
653 struct lookup_table_entry *
654 inode_unnamed_lte(const struct inode *inode,
655 const struct lookup_table *table)
658 return inode_unnamed_lte_resolved(inode);
660 return inode_unnamed_lte_unresolved(inode, table);