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/.
27 #include "wimlib_internal.h"
28 #include "lookup_table.h"
29 #include "buffer_io.h"
36 struct wim_lookup_table *new_lookup_table(size_t capacity)
38 struct wim_lookup_table *table;
39 struct hlist_head *array;
41 table = MALLOC(sizeof(struct wim_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 wim_lookup_table_entry *
59 new_lookup_table_entry()
61 struct wim_lookup_table_entry *lte;
63 lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
68 ERROR("Out of memory (tried to allocate %zu bytes for "
69 "lookup table entry)",
70 sizeof(struct wim_lookup_table_entry));
75 struct wim_lookup_table_entry *
76 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
78 struct wim_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 BUILD_BUG_ON((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);
130 free_lookup_table_entry(new);
134 void free_lookup_table_entry(struct wim_lookup_table_entry *lte)
137 switch (lte->resource_location) {
138 case RESOURCE_IN_STAGING_FILE:
139 case RESOURCE_IN_ATTACHED_BUFFER:
140 case RESOURCE_IN_FILE_ON_DISK:
141 BUILD_BUG_ON((void*)<e->file_on_disk !=
142 (void*)<e->staging_file_name);
143 BUILD_BUG_ON((void*)<e->file_on_disk !=
144 (void*)<e->attached_buffer);
145 FREE(lte->file_on_disk);
148 case RESOURCE_IN_NTFS_VOLUME:
150 FREE(lte->ntfs_loc->path_utf8);
151 FREE(lte->ntfs_loc->stream_name_utf16);
163 static int do_free_lookup_table_entry(struct wim_lookup_table_entry *entry,
166 free_lookup_table_entry(entry);
171 void free_lookup_table(struct wim_lookup_table *table)
173 DEBUG2("Freeing lookup table");
176 for_lookup_table_entry(table,
177 do_free_lookup_table_entry,
186 * Inserts an entry into the lookup table.
188 * @table: A pointer to the lookup table.
189 * @lte: A pointer to the entry to insert.
191 void lookup_table_insert(struct wim_lookup_table *table,
192 struct wim_lookup_table_entry *lte)
194 size_t i = lte->hash_short % table->capacity;
195 hlist_add_head(<e->hash_list, &table->array[i]);
197 /* XXX Make the table grow when too many entries have been inserted. */
198 table->num_entries++;
201 static void finalize_lte(struct wim_lookup_table_entry *lte)
204 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
205 unlink(lte->staging_file_name);
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 wim_lookup_table_entry *lte,
217 struct wim_lookup_table *table)
219 wimlib_assert(lte != NULL);
220 wimlib_assert(lte->refcnt != 0);
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 wim_lookup_table_entry *lte)
233 if (lte->num_opened_fds != 0)
234 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
239 /* Calls a function on all the entries in the WIM lookup table. Stop early and
240 * return nonzero if any call to the function returns nonzero. */
241 int for_lookup_table_entry(struct wim_lookup_table *table,
242 int (*visitor)(struct wim_lookup_table_entry *, void *),
245 struct wim_lookup_table_entry *lte;
246 struct hlist_node *pos, *tmp;
249 for (size_t i = 0; i < table->capacity; i++) {
250 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
253 ret = visitor(lte, arg);
263 * Reads the lookup table from a WIM file.
265 int read_lookup_table(WIMStruct *w)
268 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
270 struct wim_lookup_table *table;
271 struct wim_lookup_table_entry *cur_entry = NULL, *duplicate_entry;
273 if (resource_is_compressed(&w->hdr.lookup_table_res_entry)) {
274 ERROR("Didn't expect a compressed lookup table!");
275 ERROR("Ask the author to implement support for this.");
276 return WIMLIB_ERR_COMPRESSED_LOOKUP_TABLE;
279 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
280 w->hdr.lookup_table_res_entry.offset,
281 w->hdr.lookup_table_res_entry.original_size);
283 if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0)
285 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
287 w->hdr.lookup_table_res_entry.offset);
288 return WIMLIB_ERR_READ;
291 num_entries = w->hdr.lookup_table_res_entry.original_size /
292 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
293 table = new_lookup_table(num_entries * 2 + 1);
295 return WIMLIB_ERR_NOMEM;
297 while (num_entries--) {
300 if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
302 ERROR("Unexpected EOF in WIM lookup table!");
304 ERROR_WITH_ERRNO("Error reading WIM lookup "
307 ret = WIMLIB_ERR_READ;
310 cur_entry = new_lookup_table_entry();
312 ret = WIMLIB_ERR_NOMEM;
316 cur_entry->resource_location = RESOURCE_IN_WIM;
318 p = get_resource_entry(buf, &cur_entry->resource_entry);
319 p = get_u16(p, &cur_entry->part_number);
320 p = get_u32(p, &cur_entry->refcnt);
321 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
323 if (cur_entry->part_number != w->hdr.part_number) {
324 ERROR("A lookup table entry in part %hu of the WIM "
325 "points to part %hu",
326 w->hdr.part_number, cur_entry->part_number);
327 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
328 goto out_free_cur_entry;
331 if (is_zero_hash(cur_entry->hash)) {
332 ERROR("The WIM lookup table contains an entry with a "
333 "SHA1 message digest of all 0's");
334 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
335 goto out_free_cur_entry;
338 /* Ordinarily, no two streams should share the same SHA1 message
339 * digest. However, this constraint can be broken for metadata
340 * resources--- two identical images will have the same metadata
341 * resource, but their lookup table entries are not shared. */
342 duplicate_entry = __lookup_resource(table, cur_entry->hash);
344 && !((duplicate_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
345 && cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
347 #ifdef ENABLE_ERROR_MESSAGES
348 ERROR("The WIM lookup table contains two entries with the "
349 "same SHA1 message digest!");
350 ERROR("The first entry is:");
351 print_lookup_table_entry(duplicate_entry, stderr);
352 ERROR("The second entry is:");
353 print_lookup_table_entry(cur_entry, stderr);
355 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
356 goto out_free_cur_entry;
359 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
360 && (cur_entry->resource_entry.size !=
361 cur_entry->resource_entry.original_size))
363 #ifdef ENABLE_ERROR_MESSAGES
364 ERROR("Found uncompressed resource with original size "
365 "not the same as compressed size");
366 ERROR("The lookup table entry for the resource is as follows:");
367 print_lookup_table_entry(cur_entry, stderr);
369 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
370 goto out_free_cur_entry;
372 if ((cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
373 && cur_entry->refcnt != 1)
375 #ifdef ENABLE_ERROR_MESSAGES
376 ERROR("Found metadata resource with refcnt != 1:");
377 print_lookup_table_entry(cur_entry, stderr);
379 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
380 goto out_free_cur_entry;
382 lookup_table_insert(table, cur_entry);
385 DEBUG("Done reading lookup table.");
386 w->lookup_table = table;
391 free_lookup_table(table);
397 * Writes a lookup table entry to the output file.
399 int write_lookup_table_entry(struct wim_lookup_table_entry *lte, void *__out)
402 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
407 /* Don't write entries that have not had file resources or metadata
408 * resources written for them. */
409 if (lte->out_refcnt == 0)
412 if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
413 DEBUG("Writing metadata entry at %"PRIu64" "
414 "(orig size = %"PRIu64")",
415 ftello(out), lte->output_resource_entry.original_size);
418 p = put_resource_entry(buf, <e->output_resource_entry);
419 p = put_u16(p, lte->part_number);
420 p = put_u32(p, lte->out_refcnt);
421 p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
422 if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
423 ERROR_WITH_ERRNO("Failed to write lookup table entry");
424 return WIMLIB_ERR_WRITE;
429 /* Writes the lookup table to the output file. */
430 int write_lookup_table(struct wim_lookup_table *table, FILE *out,
431 struct resource_entry *out_res_entry)
433 off_t start_offset, end_offset;
436 start_offset = ftello(out);
437 if (start_offset == -1)
438 return WIMLIB_ERR_WRITE;
440 ret = for_lookup_table_entry(table, write_lookup_table_entry, out);
444 end_offset = ftello(out);
445 if (end_offset == -1)
446 return WIMLIB_ERR_WRITE;
448 out_res_entry->offset = start_offset;
449 out_res_entry->size = end_offset - start_offset;
450 out_res_entry->original_size = end_offset - start_offset;
451 out_res_entry->flags = WIM_RESHDR_FLAG_METADATA;
457 int lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *ignore)
459 lte->real_refcnt = 0;
463 int lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *ignore)
469 int lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *ignore)
471 if (lte->extracted_file != NULL) {
472 FREE(lte->extracted_file);
473 lte->extracted_file = NULL;
478 void print_lookup_table_entry(const struct wim_lookup_table_entry *lte,
485 fprintf(out, "Offset = %"PRIu64" bytes\n",
486 lte->resource_entry.offset);
487 fprintf(out, "Size = %"PRIu64" bytes\n",
488 (u64)lte->resource_entry.size);
489 fprintf(out, "Original size = %"PRIu64" bytes\n",
490 lte->resource_entry.original_size);
491 fprintf(out, "Part Number = %hu\n", lte->part_number);
492 fprintf(out, "Reference Count = %u\n", lte->refcnt);
493 fprintf(out, "Hash = 0x");
494 print_hash(lte->hash);
496 fprintf(out, "Flags = ");
497 u8 flags = lte->resource_entry.flags;
498 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
499 fputs("WIM_RESHDR_FLAG_COMPRESSED, ", out);
500 if (flags & WIM_RESHDR_FLAG_FREE)
501 fputs("WIM_RESHDR_FLAG_FREE, ", out);
502 if (flags & WIM_RESHDR_FLAG_METADATA)
503 fputs("WIM_RESHDR_FLAG_METADATA, ", out);
504 if (flags & WIM_RESHDR_FLAG_SPANNED)
505 fputs("WIM_RESHDR_FLAG_SPANNED, ", out);
507 switch (lte->resource_location) {
508 case RESOURCE_IN_WIM:
509 if (lte->wim->filename) {
510 fprintf(out, "WIM file = `%s'\n",
514 case RESOURCE_IN_FILE_ON_DISK:
515 fprintf(out, "File on Disk = `%s'\n", lte->file_on_disk);
517 case RESOURCE_IN_STAGING_FILE:
518 fprintf(out, "Staging File = `%s'\n", lte->staging_file_name);
526 static int do_print_lookup_table_entry(struct wim_lookup_table_entry *lte,
529 print_lookup_table_entry(lte, (FILE*)fp);
534 * Prints the lookup table of a WIM file.
536 WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
538 for_lookup_table_entry(w->lookup_table,
539 do_print_lookup_table_entry,
543 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
544 * lookup table, or NULL if there is none. */
545 struct wim_lookup_table_entry *
546 __lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
549 struct wim_lookup_table_entry *lte;
550 struct hlist_node *pos;
552 wimlib_assert(table != NULL);
553 wimlib_assert(hash != NULL);
555 i = *(size_t*)hash % table->capacity;
556 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
557 if (hashes_equal(hash, lte->hash))
564 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
567 * This is only for pre-resolved inodes.
569 int lookup_resource(WIMStruct *w, const char *path,
571 struct wim_dentry **dentry_ret,
572 struct wim_lookup_table_entry **lte_ret,
575 struct wim_dentry *dentry;
576 struct wim_lookup_table_entry *lte;
578 const char *stream_name = NULL;
579 struct wim_inode *inode;
582 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
583 stream_name = path_stream_name(path);
585 p = (char*)stream_name - 1;
590 dentry = get_dentry(w, path);
596 inode = dentry->d_inode;
598 wimlib_assert(inode->i_resolved);
600 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
601 && inode_is_directory(inode))
605 struct wim_ads_entry *ads_entry;
607 ads_entry = inode_get_ads_entry(inode, stream_name,
610 stream_idx = ads_idx + 1;
611 lte = ads_entry->lte;
622 *dentry_ret = dentry;
626 *stream_idx_ret = stream_idx;
631 /* Resolve an inode's lookup table entries
633 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
634 * lookup table) with pointers directly to the lookup table entries. A circular
635 * linked list of streams sharing the same lookup table entry is created.
637 * This function always succeeds; unresolved lookup table entries are given a
640 void inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
643 if (!inode->i_resolved) {
644 struct wim_lookup_table_entry *lte;
645 /* Resolve the default file stream */
646 lte = __lookup_resource(table, inode->i_hash);
648 inode->i_resolved = 1;
650 /* Resolve the alternate data streams */
651 for (u16 i = 0; i < inode->i_num_ads; i++) {
652 struct wim_ads_entry *cur_entry = &inode->i_ads_entries[i];
653 lte = __lookup_resource(table, cur_entry->hash);
654 cur_entry->lte = lte;
659 void inode_unresolve_ltes(struct wim_inode *inode)
661 if (inode->i_resolved) {
663 copy_hash(inode->i_hash, inode->i_lte->hash);
665 zero_out_hash(inode->i_hash);
667 for (u16 i = 0; i < inode->i_num_ads; i++) {
668 if (inode->i_ads_entries[i].lte)
669 copy_hash(inode->i_ads_entries[i].hash,
670 inode->i_ads_entries[i].lte->hash);
672 zero_out_hash(inode->i_ads_entries[i].hash);
674 inode->i_resolved = 0;
679 * Returns the lookup table entry for stream @stream_idx of the inode, where
680 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
681 * corresponds to an alternate data stream.
683 * This works for both resolved and un-resolved dentries.
685 struct wim_lookup_table_entry *
686 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
687 const struct wim_lookup_table *table)
689 if (inode->i_resolved)
690 return inode_stream_lte_resolved(inode, stream_idx);
692 return inode_stream_lte_unresolved(inode, stream_idx, table);
696 /* Return the lookup table entry for the unnamed data stream of an inode, or
697 * NULL if there is none.
699 * You'd think this would be easier than it actually is, since the unnamed data
700 * stream should be the one referenced from the inode itself. Alas, if there
701 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
702 * data stream in one of the alternate data streams instead of inside the WIM
703 * dentry itself. So we need to check the alternate data streams too.
705 * Also, note that a dentry may appear to have more than one unnamed stream, but
706 * if the SHA1 message digest is all 0's then the corresponding stream does not
707 * really "count" (this is the case for the inode's own file stream when the
708 * file stream that should be there is actually in one of the alternate stream
709 * entries.). This is despite the fact that we may need to extract such a
710 * missing entry as an empty file or empty named data stream.
712 struct wim_lookup_table_entry *
713 inode_unnamed_lte(const struct wim_inode *inode,
714 const struct wim_lookup_table *table)
716 if (inode->i_resolved)
717 return inode_unnamed_lte_resolved(inode);
719 return inode_unnamed_lte_unresolved(inode, table);
722 static int lte_add_stream_size(struct wim_lookup_table_entry *lte,
725 *(u64*)total_bytes_p += lte->resource_entry.size;
729 u64 lookup_table_total_stream_size(struct wim_lookup_table *table)
732 for_lookup_table_entry(table, lte_add_stream_size, &total_size);