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) {
90 size_t nbytes = utf16le_strlen(old->win32_file_on_disk);
91 new->win32_file_on_disk = MALLOC(nbytes + 2);
92 if (!new->win32_file_on_disk)
94 memcpy(new->win32_file_on_disk, old->win32_file_on_disk,
99 case RESOURCE_IN_STAGING_FILE:
100 case RESOURCE_IN_FILE_ON_DISK:
101 BUILD_BUG_ON((void*)&old->file_on_disk !=
102 (void*)&old->staging_file_name);
103 new->staging_file_name = STRDUP(old->staging_file_name);
104 if (!new->staging_file_name)
107 case RESOURCE_IN_ATTACHED_BUFFER:
108 new->attached_buffer = MALLOC(wim_resource_size(old));
109 if (!new->attached_buffer)
111 memcpy(new->attached_buffer, old->attached_buffer,
112 wim_resource_size(old));
115 case RESOURCE_IN_NTFS_VOLUME:
117 struct ntfs_location *loc;
118 loc = MALLOC(sizeof(*loc));
121 memcpy(loc, old->ntfs_loc, sizeof(*loc));
123 loc->stream_name = NULL;
125 loc->path = STRDUP(old->ntfs_loc->path);
128 loc->stream_name = MALLOC((loc->stream_name_nchars + 1) * 2);
129 if (!loc->stream_name)
131 memcpy(loc->stream_name,
132 old->ntfs_loc->stream_name,
133 (loc->stream_name_nchars + 1) * 2);
142 free_lookup_table_entry(new);
146 void free_lookup_table_entry(struct wim_lookup_table_entry *lte)
149 switch (lte->resource_location) {
150 case RESOURCE_IN_STAGING_FILE:
151 case RESOURCE_IN_ATTACHED_BUFFER:
152 case RESOURCE_IN_FILE_ON_DISK:
156 BUILD_BUG_ON((void*)<e->file_on_disk !=
157 (void*)<e->staging_file_name);
158 BUILD_BUG_ON((void*)<e->file_on_disk !=
159 (void*)<e->attached_buffer);
160 FREE(lte->file_on_disk);
163 case RESOURCE_IN_NTFS_VOLUME:
165 FREE(lte->ntfs_loc->path);
166 FREE(lte->ntfs_loc->stream_name);
178 static int do_free_lookup_table_entry(struct wim_lookup_table_entry *entry,
181 free_lookup_table_entry(entry);
186 void free_lookup_table(struct wim_lookup_table *table)
188 DEBUG2("Freeing lookup table");
191 for_lookup_table_entry(table,
192 do_free_lookup_table_entry,
201 * Inserts an entry into the lookup table.
203 * @table: A pointer to the lookup table.
204 * @lte: A pointer to the entry to insert.
206 void lookup_table_insert(struct wim_lookup_table *table,
207 struct wim_lookup_table_entry *lte)
209 size_t i = lte->hash_short % table->capacity;
210 hlist_add_head(<e->hash_list, &table->array[i]);
212 /* XXX Make the table grow when too many entries have been inserted. */
213 table->num_entries++;
216 static void finalize_lte(struct wim_lookup_table_entry *lte)
219 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
220 unlink(lte->staging_file_name);
221 list_del(<e->staging_list);
224 free_lookup_table_entry(lte);
227 /* Decrements the reference count for the lookup table entry @lte. If its
228 * reference count reaches 0, it is unlinked from the lookup table. If,
229 * furthermore, the entry has no opened file descriptors associated with it, the
231 void lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
232 struct wim_lookup_table *table)
234 wimlib_assert(lte != NULL);
235 wimlib_assert(lte->refcnt != 0);
236 if (--lte->refcnt == 0) {
237 lookup_table_unlink(table, lte);
239 if (lte->num_opened_fds == 0)
246 void lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
248 if (lte->num_opened_fds != 0)
249 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
254 /* Calls a function on all the entries in the WIM lookup table. Stop early and
255 * return nonzero if any call to the function returns nonzero. */
256 int for_lookup_table_entry(struct wim_lookup_table *table,
257 int (*visitor)(struct wim_lookup_table_entry *, void *),
260 struct wim_lookup_table_entry *lte;
261 struct hlist_node *pos, *tmp;
264 for (size_t i = 0; i < table->capacity; i++) {
265 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
268 ret = visitor(lte, arg);
278 * Reads the lookup table from a WIM file.
280 int read_lookup_table(WIMStruct *w)
283 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
285 struct wim_lookup_table *table;
286 struct wim_lookup_table_entry *cur_entry = NULL, *duplicate_entry;
288 if (resource_is_compressed(&w->hdr.lookup_table_res_entry)) {
289 ERROR("Didn't expect a compressed lookup table!");
290 ERROR("Ask the author to implement support for this.");
291 return WIMLIB_ERR_COMPRESSED_LOOKUP_TABLE;
294 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
295 w->hdr.lookup_table_res_entry.offset,
296 w->hdr.lookup_table_res_entry.original_size);
298 if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0)
300 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
302 w->hdr.lookup_table_res_entry.offset);
303 return WIMLIB_ERR_READ;
306 num_entries = w->hdr.lookup_table_res_entry.original_size /
307 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
308 table = new_lookup_table(num_entries * 2 + 1);
310 return WIMLIB_ERR_NOMEM;
312 while (num_entries--) {
315 if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
317 ERROR("Unexpected EOF in WIM lookup table!");
319 ERROR_WITH_ERRNO("Error reading WIM lookup "
322 ret = WIMLIB_ERR_READ;
325 cur_entry = new_lookup_table_entry();
327 ret = WIMLIB_ERR_NOMEM;
331 cur_entry->resource_location = RESOURCE_IN_WIM;
333 p = get_resource_entry(buf, &cur_entry->resource_entry);
334 p = get_u16(p, &cur_entry->part_number);
335 p = get_u32(p, &cur_entry->refcnt);
336 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
338 if (cur_entry->part_number != w->hdr.part_number) {
339 ERROR("A lookup table entry in part %hu of the WIM "
340 "points to part %hu",
341 w->hdr.part_number, cur_entry->part_number);
342 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
343 goto out_free_cur_entry;
346 if (is_zero_hash(cur_entry->hash)) {
347 ERROR("The WIM lookup table contains an entry with a "
348 "SHA1 message digest of all 0's");
349 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
350 goto out_free_cur_entry;
353 /* Ordinarily, no two streams should share the same SHA1 message
354 * digest. However, this constraint can be broken for metadata
355 * resources--- two identical images will have the same metadata
356 * resource, but their lookup table entries are not shared. */
357 duplicate_entry = __lookup_resource(table, cur_entry->hash);
359 && !((duplicate_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
360 && cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA))
362 #ifdef ENABLE_ERROR_MESSAGES
363 ERROR("The WIM lookup table contains two entries with the "
364 "same SHA1 message digest!");
365 ERROR("The first entry is:");
366 print_lookup_table_entry(duplicate_entry, stderr);
367 ERROR("The second entry is:");
368 print_lookup_table_entry(cur_entry, stderr);
370 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
371 goto out_free_cur_entry;
374 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
375 && (cur_entry->resource_entry.size !=
376 cur_entry->resource_entry.original_size))
378 #ifdef ENABLE_ERROR_MESSAGES
379 ERROR("Found uncompressed resource with original size "
380 "not the same as compressed size");
381 ERROR("The lookup table entry for the resource is as follows:");
382 print_lookup_table_entry(cur_entry, stderr);
384 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
385 goto out_free_cur_entry;
387 if ((cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
388 && cur_entry->refcnt != 1)
390 #ifdef ENABLE_ERROR_MESSAGES
391 ERROR("Found metadata resource with refcnt != 1:");
392 print_lookup_table_entry(cur_entry, stderr);
394 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
395 goto out_free_cur_entry;
397 lookup_table_insert(table, cur_entry);
400 DEBUG("Done reading lookup table.");
401 w->lookup_table = table;
406 free_lookup_table(table);
412 * Writes a lookup table entry to the output file.
414 int write_lookup_table_entry(struct wim_lookup_table_entry *lte, void *__out)
417 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
422 /* Don't write entries that have not had file resources or metadata
423 * resources written for them. */
424 if (lte->out_refcnt == 0)
427 if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
428 DEBUG("Writing metadata entry at %"PRIu64" "
429 "(orig size = %"PRIu64")",
430 ftello(out), lte->output_resource_entry.original_size);
433 p = put_resource_entry(buf, <e->output_resource_entry);
434 p = put_u16(p, lte->part_number);
435 p = put_u32(p, lte->out_refcnt);
436 p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
437 if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
438 ERROR_WITH_ERRNO("Failed to write lookup table entry");
439 return WIMLIB_ERR_WRITE;
444 /* Writes the lookup table to the output file. */
445 int write_lookup_table(struct wim_lookup_table *table, FILE *out,
446 struct resource_entry *out_res_entry)
448 off_t start_offset, end_offset;
451 start_offset = ftello(out);
452 if (start_offset == -1)
453 return WIMLIB_ERR_WRITE;
455 ret = for_lookup_table_entry(table, write_lookup_table_entry, out);
459 end_offset = ftello(out);
460 if (end_offset == -1)
461 return WIMLIB_ERR_WRITE;
463 out_res_entry->offset = start_offset;
464 out_res_entry->size = end_offset - start_offset;
465 out_res_entry->original_size = end_offset - start_offset;
466 out_res_entry->flags = WIM_RESHDR_FLAG_METADATA;
472 int lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *ignore)
474 lte->real_refcnt = 0;
478 int lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *ignore)
484 int lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *ignore)
486 if (lte->extracted_file != NULL) {
487 FREE(lte->extracted_file);
488 lte->extracted_file = NULL;
493 void print_lookup_table_entry(const struct wim_lookup_table_entry *lte,
500 fprintf(out, "Offset = %"PRIu64" bytes\n",
501 lte->resource_entry.offset);
502 fprintf(out, "Size = %"PRIu64" bytes\n",
503 (u64)lte->resource_entry.size);
504 fprintf(out, "Original size = %"PRIu64" bytes\n",
505 lte->resource_entry.original_size);
506 fprintf(out, "Part Number = %hu\n", lte->part_number);
507 fprintf(out, "Reference Count = %u\n", lte->refcnt);
508 fprintf(out, "Hash = 0x");
509 print_hash(lte->hash);
511 fprintf(out, "Flags = ");
512 u8 flags = lte->resource_entry.flags;
513 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
514 fputs("WIM_RESHDR_FLAG_COMPRESSED, ", out);
515 if (flags & WIM_RESHDR_FLAG_FREE)
516 fputs("WIM_RESHDR_FLAG_FREE, ", out);
517 if (flags & WIM_RESHDR_FLAG_METADATA)
518 fputs("WIM_RESHDR_FLAG_METADATA, ", out);
519 if (flags & WIM_RESHDR_FLAG_SPANNED)
520 fputs("WIM_RESHDR_FLAG_SPANNED, ", out);
522 switch (lte->resource_location) {
523 case RESOURCE_IN_WIM:
524 if (lte->wim->filename) {
525 fprintf(out, "WIM file = `%s'\n",
529 case RESOURCE_IN_FILE_ON_DISK:
530 fprintf(out, "File on Disk = `%s'\n", lte->file_on_disk);
532 case RESOURCE_IN_STAGING_FILE:
533 fprintf(out, "Staging File = `%s'\n", lte->staging_file_name);
541 static int do_print_lookup_table_entry(struct wim_lookup_table_entry *lte,
544 print_lookup_table_entry(lte, (FILE*)fp);
549 * Prints the lookup table of a WIM file.
551 WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
553 for_lookup_table_entry(w->lookup_table,
554 do_print_lookup_table_entry,
558 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
559 * lookup table, or NULL if there is none. */
560 struct wim_lookup_table_entry *
561 __lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
564 struct wim_lookup_table_entry *lte;
565 struct hlist_node *pos;
567 wimlib_assert(table != NULL);
568 wimlib_assert(hash != NULL);
570 i = *(size_t*)hash % table->capacity;
571 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
572 if (hashes_equal(hash, lte->hash))
579 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
582 * This is only for pre-resolved inodes.
585 lookup_resource(WIMStruct *w,
588 struct wim_dentry **dentry_ret,
589 struct wim_lookup_table_entry **lte_ret,
592 struct wim_dentry *dentry;
593 struct wim_lookup_table_entry *lte;
595 const mbchar *stream_name = NULL;
596 struct wim_inode *inode;
599 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
600 stream_name = path_stream_name(path);
602 p = (mbchar*)stream_name - 1;
607 dentry = get_dentry(w, path);
613 inode = dentry->d_inode;
615 wimlib_assert(inode->i_resolved);
617 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
618 && inode_is_directory(inode))
622 struct wim_ads_entry *ads_entry;
624 ads_entry = inode_get_ads_entry(inode, stream_name,
627 stream_idx = ads_idx + 1;
628 lte = ads_entry->lte;
639 *dentry_ret = dentry;
643 *stream_idx_ret = stream_idx;
648 /* Resolve an inode's lookup table entries
650 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
651 * lookup table) with pointers directly to the lookup table entries. A circular
652 * linked list of streams sharing the same lookup table entry is created.
654 * This function always succeeds; unresolved lookup table entries are given a
657 void inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
660 if (!inode->i_resolved) {
661 struct wim_lookup_table_entry *lte;
662 /* Resolve the default file stream */
663 lte = __lookup_resource(table, inode->i_hash);
665 inode->i_resolved = 1;
667 /* Resolve the alternate data streams */
668 for (u16 i = 0; i < inode->i_num_ads; i++) {
669 struct wim_ads_entry *cur_entry = &inode->i_ads_entries[i];
670 lte = __lookup_resource(table, cur_entry->hash);
671 cur_entry->lte = lte;
676 void inode_unresolve_ltes(struct wim_inode *inode)
678 if (inode->i_resolved) {
680 copy_hash(inode->i_hash, inode->i_lte->hash);
682 zero_out_hash(inode->i_hash);
684 for (u16 i = 0; i < inode->i_num_ads; i++) {
685 if (inode->i_ads_entries[i].lte)
686 copy_hash(inode->i_ads_entries[i].hash,
687 inode->i_ads_entries[i].lte->hash);
689 zero_out_hash(inode->i_ads_entries[i].hash);
691 inode->i_resolved = 0;
696 * Returns the lookup table entry for stream @stream_idx of the inode, where
697 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
698 * corresponds to an alternate data stream.
700 * This works for both resolved and un-resolved dentries.
702 struct wim_lookup_table_entry *
703 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
704 const struct wim_lookup_table *table)
706 if (inode->i_resolved)
707 return inode_stream_lte_resolved(inode, stream_idx);
709 return inode_stream_lte_unresolved(inode, stream_idx, table);
713 /* Return the lookup table entry for the unnamed data stream of an inode, or
714 * NULL if there is none.
716 * You'd think this would be easier than it actually is, since the unnamed data
717 * stream should be the one referenced from the inode itself. Alas, if there
718 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
719 * data stream in one of the alternate data streams instead of inside the WIM
720 * dentry itself. So we need to check the alternate data streams too.
722 * Also, note that a dentry may appear to have more than one unnamed stream, but
723 * if the SHA1 message digest is all 0's then the corresponding stream does not
724 * really "count" (this is the case for the inode's own file stream when the
725 * file stream that should be there is actually in one of the alternate stream
726 * entries.). This is despite the fact that we may need to extract such a
727 * missing entry as an empty file or empty named data stream.
729 struct wim_lookup_table_entry *
730 inode_unnamed_lte(const struct wim_inode *inode,
731 const struct wim_lookup_table *table)
733 if (inode->i_resolved)
734 return inode_unnamed_lte_resolved(inode);
736 return inode_unnamed_lte_unresolved(inode, table);
739 static int lte_add_stream_size(struct wim_lookup_table_entry *lte,
742 *(u64*)total_bytes_p += lte->resource_entry.size;
746 u64 lookup_table_total_stream_size(struct wim_lookup_table *table)
749 for_lookup_table_entry(table, lte_add_stream_size, &total_size);