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));
43 array = CALLOC(capacity, sizeof(array[0]));
45 table->num_entries = 0;
46 table->capacity = capacity;
51 ERROR("Failed to allocate memory for lookup table with capacity %zu",
58 struct lookup_table_entry *new_lookup_table_entry()
60 struct lookup_table_entry *lte;
62 lte = CALLOC(1, sizeof(struct lookup_table_entry));
67 ERROR("Out of memory (tried to allocate %zu bytes for "
68 "lookup table entry)",
69 sizeof(struct lookup_table_entry));
74 struct lookup_table_entry *
75 clone_lookup_table_entry(const struct lookup_table_entry *old)
77 struct lookup_table_entry *new;
79 new = MALLOC(sizeof(*new));
83 memcpy(new, old, sizeof(*old));
84 new->extracted_file = NULL;
85 switch (new->resource_location) {
86 case RESOURCE_IN_STAGING_FILE:
87 case RESOURCE_IN_FILE_ON_DISK:
88 wimlib_assert((void*)&old->file_on_disk ==
89 (void*)&old->staging_file_name);
90 new->staging_file_name = STRDUP(old->staging_file_name);
91 if (!new->staging_file_name)
94 case RESOURCE_IN_ATTACHED_BUFFER:
95 new->attached_buffer = MALLOC(wim_resource_size(old));
96 if (!new->attached_buffer)
98 memcpy(new->attached_buffer, old->attached_buffer,
99 wim_resource_size(old));
102 case RESOURCE_IN_NTFS_VOLUME:
104 struct ntfs_location *loc;
105 loc = MALLOC(sizeof(*loc));
108 memcpy(loc, old->ntfs_loc, sizeof(*loc));
109 loc->path_utf8 = NULL;
110 loc->stream_name_utf16 = NULL;
112 loc->path_utf8 = STRDUP(old->ntfs_loc->path_utf8);
115 loc->stream_name_utf16 = MALLOC(loc->stream_name_utf16_num_chars * 2);
116 if (!loc->stream_name_utf16)
118 memcpy(loc->stream_name_utf16,
119 old->ntfs_loc->stream_name_utf16,
120 loc->stream_name_utf16_num_chars * 2);
129 free_lookup_table_entry(new);
133 void free_lookup_table_entry(struct lookup_table_entry *lte)
136 switch (lte->resource_location) {
137 case RESOURCE_IN_STAGING_FILE:
138 case RESOURCE_IN_ATTACHED_BUFFER:
139 case RESOURCE_IN_FILE_ON_DISK:
140 wimlib_assert(((void*)<e->file_on_disk ==
141 (void*)<e->staging_file_name)
142 && ((void*)<e->file_on_disk ==
143 (void*)<e->attached_buffer));
144 FREE(lte->file_on_disk);
147 case RESOURCE_IN_NTFS_VOLUME:
149 FREE(lte->ntfs_loc->path_utf8);
150 FREE(lte->ntfs_loc->stream_name_utf16);
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)
220 wimlib_assert(lte != NULL);
221 wimlib_assert(lte->refcnt != 0);
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)
234 wimlib_assert(lte != NULL);
235 if (lte->num_opened_fds != 0) {
236 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
243 * Calls a function on all the entries in the lookup table. Stop early and
244 * return nonzero if any call to the function returns nonzero.
246 int for_lookup_table_entry(struct lookup_table *table,
247 int (*visitor)(struct lookup_table_entry *, void *),
250 struct lookup_table_entry *lte;
251 struct hlist_node *pos, *tmp;
254 for (size_t i = 0; i < table->capacity; i++) {
255 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
258 ret = visitor(lte, arg);
268 * Reads the lookup table from a WIM file.
270 int read_lookup_table(WIMStruct *w)
273 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
275 struct lookup_table *table;
276 struct lookup_table_entry *cur_entry = NULL, *duplicate_entry;
278 if (resource_is_compressed(&w->hdr.lookup_table_res_entry)) {
279 ERROR("Didn't expect a compressed lookup table!");
280 ERROR("Ask the author to implement support for this.");
281 return WIMLIB_ERR_COMPRESSED_LOOKUP_TABLE;
284 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
285 w->hdr.lookup_table_res_entry.offset,
286 w->hdr.lookup_table_res_entry.original_size);
288 if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0)
290 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
292 w->hdr.lookup_table_res_entry.offset);
293 return WIMLIB_ERR_READ;
296 num_entries = w->hdr.lookup_table_res_entry.original_size /
297 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
298 table = new_lookup_table(num_entries * 2 + 1);
300 return WIMLIB_ERR_NOMEM;
302 while (num_entries--) {
305 if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
307 ERROR("Unexpected EOF in WIM lookup table!");
309 ERROR_WITH_ERRNO("Error reading WIM lookup "
312 ret = WIMLIB_ERR_READ;
315 cur_entry = new_lookup_table_entry();
317 ret = WIMLIB_ERR_NOMEM;
321 cur_entry->resource_location = RESOURCE_IN_WIM;
323 p = get_resource_entry(buf, &cur_entry->resource_entry);
324 p = get_u16(p, &cur_entry->part_number);
325 p = get_u32(p, &cur_entry->refcnt);
326 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
328 if (cur_entry->part_number != w->hdr.part_number) {
329 ERROR("A lookup table entry in part %hu of the WIM "
330 "points to part %hu",
331 w->hdr.part_number, cur_entry->part_number);
332 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
333 goto out_free_cur_entry;
336 if (is_zero_hash(cur_entry->hash)) {
337 ERROR("The WIM lookup table contains an entry with a "
338 "SHA1 message digest of all 0's");
339 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
340 goto out_free_cur_entry;
343 /* Ordinarily, no two streams should share the same SHA1 message
344 * digest. However, this constraint can be broken for metadata
345 * resources--- two identical images will have the same metadata
346 * resource, but their lookup table entries are not shared. */
347 duplicate_entry = __lookup_resource(table, cur_entry->hash);
349 && !((duplicate_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
350 && cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
352 ERROR("The WIM lookup table contains two entries with the "
353 "same SHA1 message digest!");
354 ERROR("The first entry is:");
355 print_lookup_table_entry(duplicate_entry);
356 ERROR("The second entry is:");
357 print_lookup_table_entry(cur_entry);
358 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
359 goto out_free_cur_entry;
362 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
363 && (cur_entry->resource_entry.size !=
364 cur_entry->resource_entry.original_size))
366 ERROR("Found uncompressed resource with original size "
367 "not the same as compressed size");
368 ERROR("The lookup table entry for the resource is as follows:");
369 print_lookup_table_entry(cur_entry);
370 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
371 goto out_free_cur_entry;
373 if ((cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
374 && cur_entry->refcnt != 1)
376 ERROR("Found metadata resource with refcnt != 1:");
377 print_lookup_table_entry(cur_entry);
378 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
379 goto out_free_cur_entry;
381 lookup_table_insert(table, cur_entry);
384 DEBUG("Done reading lookup table.");
385 w->lookup_table = table;
390 free_lookup_table(table);
396 * Writes a lookup table entry to the output file.
398 int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out)
401 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
406 /* Don't write entries that have not had file resources or metadata
407 * resources written for them. */
408 if (lte->out_refcnt == 0)
411 if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
412 DEBUG("Writing metadata entry at %lu (orig size = %zu)",
413 ftello(out), lte->output_resource_entry.original_size);
415 p = put_resource_entry(buf, <e->output_resource_entry);
416 p = put_u16(p, lte->part_number);
417 p = put_u32(p, lte->out_refcnt);
418 p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
419 if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
420 ERROR_WITH_ERRNO("Failed to write lookup table entry");
421 return WIMLIB_ERR_WRITE;
426 /* Writes the lookup table to the output file. */
427 int write_lookup_table(struct lookup_table *table, FILE *out,
428 struct resource_entry *out_res_entry)
430 off_t start_offset, end_offset;
433 start_offset = ftello(out);
434 if (start_offset == -1)
435 return WIMLIB_ERR_WRITE;
437 ret = for_lookup_table_entry(table, write_lookup_table_entry, out);
441 end_offset = ftello(out);
442 if (end_offset == -1)
443 return WIMLIB_ERR_WRITE;
445 out_res_entry->offset = start_offset;
446 out_res_entry->size = end_offset - start_offset;
447 out_res_entry->original_size = end_offset - start_offset;
448 out_res_entry->flags = WIM_RESHDR_FLAG_METADATA;
454 int lte_zero_real_refcnt(struct lookup_table_entry *lte, void *ignore)
456 lte->real_refcnt = 0;
460 int lte_zero_out_refcnt(struct lookup_table_entry *lte, void *ignore)
466 int lte_zero_extracted_file(struct lookup_table_entry *lte, void *ignore)
468 lte->extracted_file = NULL;
472 int lte_free_extracted_file(struct lookup_table_entry *lte, void *ignore)
474 if (lte->extracted_file != NULL) {
475 FREE(lte->extracted_file);
476 lte->extracted_file = NULL;
481 void print_lookup_table_entry(const struct lookup_table_entry *lte)
487 printf("Offset = %"PRIu64" bytes\n",
488 lte->resource_entry.offset);
489 printf("Size = %"PRIu64" bytes\n",
490 (u64)lte->resource_entry.size);
491 printf("Original size = %"PRIu64" bytes\n",
492 lte->resource_entry.original_size);
493 printf("Part Number = %hu\n", lte->part_number);
494 printf("Reference Count = %u\n", lte->refcnt);
496 print_hash(lte->hash);
499 u8 flags = lte->resource_entry.flags;
500 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
501 fputs("WIM_RESHDR_FLAG_COMPRESSED, ", stdout);
502 if (flags & WIM_RESHDR_FLAG_FREE)
503 fputs("WIM_RESHDR_FLAG_FREE, ", stdout);
504 if (flags & WIM_RESHDR_FLAG_METADATA)
505 fputs("WIM_RESHDR_FLAG_METADATA, ", stdout);
506 if (flags & WIM_RESHDR_FLAG_SPANNED)
507 fputs("WIM_RESHDR_FLAG_SPANNED, ", stdout);
509 switch (lte->resource_location) {
510 case RESOURCE_IN_WIM:
511 if (lte->wim->filename) {
512 printf("WIM file = `%s'\n",
516 case RESOURCE_IN_FILE_ON_DISK:
517 printf("File on Disk = `%s'\n", lte->file_on_disk);
519 case RESOURCE_IN_STAGING_FILE:
520 printf("Staging File = `%s'\n", lte->staging_file_name);
528 static int do_print_lookup_table_entry(struct lookup_table_entry *lte,
531 print_lookup_table_entry(lte);
536 * Prints the lookup table of a WIM file.
538 WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
540 for_lookup_table_entry(w->lookup_table,
541 do_print_lookup_table_entry,
546 * Looks up an entry in the lookup table.
548 struct lookup_table_entry *
549 __lookup_resource(const struct lookup_table *table, const u8 hash[])
552 struct lookup_table_entry *lte;
553 struct hlist_node *pos;
555 wimlib_assert(table != NULL);
556 wimlib_assert(hash != NULL);
558 i = *(size_t*)hash % table->capacity;
559 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
560 if (hashes_equal(hash, lte->hash))
567 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
570 * This is only for pre-resolved inodes.
572 int lookup_resource(WIMStruct *w, const char *path,
574 struct dentry **dentry_ret,
575 struct lookup_table_entry **lte_ret,
578 struct dentry *dentry;
579 struct lookup_table_entry *lte;
581 const char *stream_name = NULL;
585 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
586 stream_name = path_stream_name(path);
588 p = (char*)stream_name - 1;
593 dentry = get_dentry(w, path);
599 inode = dentry->d_inode;
601 wimlib_assert(inode->resolved);
603 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
604 && inode_is_directory(inode))
608 struct ads_entry *ads_entry;
610 ads_entry = inode_get_ads_entry(inode, stream_name,
613 stream_idx = ads_idx + 1;
614 lte = ads_entry->lte;
625 *dentry_ret = dentry;
629 *stream_idx_ret = stream_idx;
634 void inode_resolve_ltes(struct inode *inode, struct lookup_table *table)
636 struct lookup_table_entry *lte;
638 wimlib_assert(!inode->resolved);
640 /* Resolve the default file stream */
641 lte = __lookup_resource(table, inode->hash);
643 inode->resolved = true;
645 /* Resolve the alternate data streams */
646 for (u16 i = 0; i < inode->num_ads; i++) {
647 struct ads_entry *cur_entry = &inode->ads_entries[i];
648 lte = __lookup_resource(table, cur_entry->hash);
649 cur_entry->lte = lte;
653 static void inode_unresolve_ltes(struct inode *inode)
655 wimlib_assert(inode->resolved);
657 copy_hash(inode->hash, inode->lte->hash);
659 zero_out_hash(inode->hash);
661 for (u16 i = 0; i < inode->num_ads; i++) {
662 if (inode->ads_entries[i].lte)
663 copy_hash(inode->ads_entries[i].hash,
664 inode->ads_entries[i].lte->hash);
666 zero_out_hash(inode->ads_entries[i].hash);
668 inode->resolved = false;
671 /* Resolve a dentry's lookup table entries
673 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
674 * lookup table) with pointers directly to the lookup table entries. A circular
675 * linked list of streams sharing the same lookup table entry is created.
677 * This function always succeeds; unresolved lookup table entries are given a
680 int dentry_resolve_ltes(struct dentry *dentry, void *table)
682 if (!dentry->d_inode->resolved)
683 inode_resolve_ltes(dentry->d_inode, table);
687 int dentry_unresolve_ltes(struct dentry *dentry, void *ignore)
689 if (dentry->d_inode->resolved)
690 inode_unresolve_ltes(dentry->d_inode);
695 * Returns the lookup table entry for stream @stream_idx of the inode, where
696 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
697 * corresponds to an alternate data stream.
699 * This works for both resolved and un-resolved dentries.
701 struct lookup_table_entry *
702 inode_stream_lte(const struct inode *inode, unsigned stream_idx,
703 const struct lookup_table *table)
706 return inode_stream_lte_resolved(inode, stream_idx);
708 return inode_stream_lte_unresolved(inode, stream_idx, table);
712 /* Return the lookup table entry for the unnamed data stream of an inode, or
713 * NULL if there is none.
715 * You'd think this would be easier than it actually is, since the unnamed data
716 * stream should be the one referenced from the inode itself. Alas, if there
717 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
718 * data stream in one of the alternate data streams instead of inside the WIM
719 * dentry itself. So we need to check the alternate data streams too.
721 * Also, note that a dentry may appear to have more than one unnamed stream, but
722 * if the SHA1 message digest is all 0's then the corresponding stream does not
723 * really "count" (this is the case for the inode's own file stream when the
724 * file stream that should be there is actually in one of the alternate stream
725 * entries.). This is despite the fact that we may need to extract such a
726 * missing entry as an empty file or empty named data stream.
728 struct lookup_table_entry *
729 inode_unnamed_lte(const struct inode *inode,
730 const struct lookup_table *table)
733 return inode_unnamed_lte_resolved(inode);
735 return inode_unnamed_lte_unresolved(inode, table);
738 static int lte_add_stream_size(struct lookup_table_entry *lte,
741 *(u64*)total_bytes_p += lte->resource_entry.size;
745 u64 lookup_table_total_stream_size(struct lookup_table *table)
748 for_lookup_table_entry(table, lte_add_stream_size, &total_size);