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 *
37 new_lookup_table(size_t capacity)
39 struct wim_lookup_table *table;
40 struct hlist_head *array;
42 table = MALLOC(sizeof(struct wim_lookup_table));
44 array = CALLOC(capacity, sizeof(array[0]));
46 table->num_entries = 0;
47 table->capacity = capacity;
52 ERROR("Failed to allocate memory for lookup table with capacity %zu",
59 struct wim_lookup_table_entry *
60 new_lookup_table_entry()
62 struct wim_lookup_table_entry *lte;
64 lte = CALLOC(1, sizeof(struct wim_lookup_table_entry));
69 ERROR("Out of memory (tried to allocate %zu bytes for "
70 "lookup table entry)",
71 sizeof(struct wim_lookup_table_entry));
76 struct wim_lookup_table_entry *
77 clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
79 struct wim_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) {
91 case RESOURCE_IN_STAGING_FILE:
92 case RESOURCE_IN_FILE_ON_DISK:
93 BUILD_BUG_ON((void*)&old->file_on_disk !=
94 (void*)&old->staging_file_name);
95 new->staging_file_name = TSTRDUP(old->staging_file_name);
96 if (!new->staging_file_name)
99 case RESOURCE_IN_ATTACHED_BUFFER:
100 new->attached_buffer = MALLOC(wim_resource_size(old));
101 if (!new->attached_buffer)
103 memcpy(new->attached_buffer, old->attached_buffer,
104 wim_resource_size(old));
107 case RESOURCE_IN_NTFS_VOLUME:
109 struct ntfs_location *loc;
110 loc = MALLOC(sizeof(*loc));
113 memcpy(loc, old->ntfs_loc, sizeof(*loc));
115 loc->stream_name = NULL;
117 loc->path = STRDUP(old->ntfs_loc->path);
120 loc->stream_name = MALLOC((loc->stream_name_nchars + 1) * 2);
121 if (!loc->stream_name)
123 memcpy(loc->stream_name,
124 old->ntfs_loc->stream_name,
125 (loc->stream_name_nchars + 1) * 2);
134 free_lookup_table_entry(new);
139 free_lookup_table_entry(struct wim_lookup_table_entry *lte)
142 switch (lte->resource_location) {
143 case RESOURCE_IN_STAGING_FILE:
144 case RESOURCE_IN_ATTACHED_BUFFER:
145 case RESOURCE_IN_FILE_ON_DISK:
149 BUILD_BUG_ON((void*)<e->file_on_disk !=
150 (void*)<e->staging_file_name);
151 BUILD_BUG_ON((void*)<e->file_on_disk !=
152 (void*)<e->attached_buffer);
153 FREE(lte->file_on_disk);
156 case RESOURCE_IN_NTFS_VOLUME:
158 FREE(lte->ntfs_loc->path);
159 FREE(lte->ntfs_loc->stream_name);
172 do_free_lookup_table_entry(struct wim_lookup_table_entry *entry, void *ignore)
174 free_lookup_table_entry(entry);
180 free_lookup_table(struct wim_lookup_table *table)
182 DEBUG2("Freeing lookup table");
185 for_lookup_table_entry(table,
186 do_free_lookup_table_entry,
195 * Inserts an entry into the lookup table.
197 * @table: A pointer to the lookup table.
198 * @lte: A pointer to the entry to insert.
201 lookup_table_insert(struct wim_lookup_table *table,
202 struct wim_lookup_table_entry *lte)
204 size_t i = lte->hash_short % table->capacity;
205 hlist_add_head(<e->hash_list, &table->array[i]);
207 /* XXX Make the table grow when too many entries have been inserted. */
208 table->num_entries++;
212 finalize_lte(struct wim_lookup_table_entry *lte)
215 if (lte->resource_location == RESOURCE_IN_STAGING_FILE) {
216 unlink(lte->staging_file_name);
217 list_del(<e->staging_list);
220 free_lookup_table_entry(lte);
223 /* Decrements the reference count for the lookup table entry @lte. If its
224 * reference count reaches 0, it is unlinked from the lookup table. If,
225 * furthermore, the entry has no opened file descriptors associated with it, the
228 lte_decrement_refcnt(struct wim_lookup_table_entry *lte,
229 struct wim_lookup_table *table)
231 wimlib_assert(lte != NULL);
232 wimlib_assert(lte->refcnt != 0);
233 if (--lte->refcnt == 0) {
234 lookup_table_unlink(table, lte);
236 if (lte->num_opened_fds == 0)
244 lte_decrement_num_opened_fds(struct wim_lookup_table_entry *lte)
246 if (lte->num_opened_fds != 0)
247 if (--lte->num_opened_fds == 0 && lte->refcnt == 0)
252 /* Calls a function on all the entries in the WIM lookup table. Stop early and
253 * return nonzero if any call to the function returns nonzero. */
255 for_lookup_table_entry(struct wim_lookup_table *table,
256 int (*visitor)(struct wim_lookup_table_entry *, void *),
259 struct wim_lookup_table_entry *lte;
260 struct hlist_node *pos, *tmp;
263 for (size_t i = 0; i < table->capacity; i++) {
264 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
267 ret = visitor(lte, arg);
277 * Reads the lookup table from a WIM file.
280 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.
415 write_lookup_table_entry(struct wim_lookup_table_entry *lte, void *__out)
418 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
423 /* Don't write entries that have not had file resources or metadata
424 * resources written for them. */
425 if (lte->out_refcnt == 0)
428 if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
429 DEBUG("Writing metadata entry at %"PRIu64" "
430 "(orig size = %"PRIu64")",
431 ftello(out), lte->output_resource_entry.original_size);
434 p = put_resource_entry(buf, <e->output_resource_entry);
435 p = put_u16(p, lte->part_number);
436 p = put_u32(p, lte->out_refcnt);
437 p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
438 if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
439 ERROR_WITH_ERRNO("Failed to write lookup table entry");
440 return WIMLIB_ERR_WRITE;
445 /* Writes the lookup table to the output file. */
447 write_lookup_table(struct wim_lookup_table *table, FILE *out,
448 struct resource_entry *out_res_entry)
450 off_t start_offset, end_offset;
453 start_offset = ftello(out);
454 if (start_offset == -1)
455 return WIMLIB_ERR_WRITE;
457 ret = for_lookup_table_entry(table, write_lookup_table_entry, out);
461 end_offset = ftello(out);
462 if (end_offset == -1)
463 return WIMLIB_ERR_WRITE;
465 out_res_entry->offset = start_offset;
466 out_res_entry->size = end_offset - start_offset;
467 out_res_entry->original_size = end_offset - start_offset;
468 out_res_entry->flags = WIM_RESHDR_FLAG_METADATA;
475 lte_zero_real_refcnt(struct wim_lookup_table_entry *lte, void *ignore)
477 lte->real_refcnt = 0;
482 lte_zero_out_refcnt(struct wim_lookup_table_entry *lte, void *ignore)
489 lte_free_extracted_file(struct wim_lookup_table_entry *lte, void *ignore)
491 if (lte->extracted_file != NULL) {
492 FREE(lte->extracted_file);
493 lte->extracted_file = NULL;
499 print_lookup_table_entry(const struct wim_lookup_table_entry *lte, FILE *out)
505 tfprintf(out, T("Offset = %"PRIu64" bytes\n"),
506 lte->resource_entry.offset);
508 tfprintf(out, T("Size = %"PRIu64" bytes\n"),
509 (u64)lte->resource_entry.size);
511 tfprintf(out, T("Original size = %"PRIu64" bytes\n"),
512 lte->resource_entry.original_size);
514 tfprintf(out, T("Part Number = %hu\n"), lte->part_number);
515 tfprintf(out, T("Reference Count = %u\n"), lte->refcnt);
517 tfprintf(out, T("Hash = 0x"));
518 print_hash(lte->hash);
521 tfprintf(out, T("Flags = "));
522 u8 flags = lte->resource_entry.flags;
523 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
524 tfputs(T("WIM_RESHDR_FLAG_COMPRESSED, "), out);
525 if (flags & WIM_RESHDR_FLAG_FREE)
526 tfputs(T("WIM_RESHDR_FLAG_FREE, "), out);
527 if (flags & WIM_RESHDR_FLAG_METADATA)
528 tfputs(T("WIM_RESHDR_FLAG_METADATA, "), out);
529 if (flags & WIM_RESHDR_FLAG_SPANNED)
530 tfputs(T("WIM_RESHDR_FLAG_SPANNED, "), out);
532 switch (lte->resource_location) {
533 case RESOURCE_IN_WIM:
534 if (lte->wim->filename) {
535 tfprintf(out, T("WIM file = `%"TS"'\n"),
542 case RESOURCE_IN_FILE_ON_DISK:
543 tfprintf(out, T("File on Disk = `%"TS"'\n"),
546 case RESOURCE_IN_STAGING_FILE:
547 tfprintf(out, T("Staging File = `%"TS"'\n"),
548 lte->staging_file_name);
557 do_print_lookup_table_entry(struct wim_lookup_table_entry *lte, void *fp)
559 print_lookup_table_entry(lte, (FILE*)fp);
564 * Prints the lookup table of a WIM file.
567 wimlib_print_lookup_table(WIMStruct *w)
569 for_lookup_table_entry(w->lookup_table,
570 do_print_lookup_table_entry,
574 /* Given a SHA1 message digest, return the corresponding entry in the WIM's
575 * lookup table, or NULL if there is none. */
576 struct wim_lookup_table_entry *
577 __lookup_resource(const struct wim_lookup_table *table, const u8 hash[])
580 struct wim_lookup_table_entry *lte;
581 struct hlist_node *pos;
583 wimlib_assert(table != NULL);
584 wimlib_assert(hash != NULL);
586 i = *(size_t*)hash % table->capacity;
587 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
588 if (hashes_equal(hash, lte->hash))
595 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
598 * This is only for pre-resolved inodes.
601 lookup_resource(WIMStruct *w,
604 struct wim_dentry **dentry_ret,
605 struct wim_lookup_table_entry **lte_ret,
608 struct wim_dentry *dentry;
609 struct wim_lookup_table_entry *lte;
611 const tchar *stream_name = NULL;
612 struct wim_inode *inode;
615 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
616 stream_name = path_stream_name(path);
618 p = (tchar*)stream_name - 1;
623 dentry = get_dentry(w, path);
629 inode = dentry->d_inode;
631 wimlib_assert(inode->i_resolved);
633 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
634 && inode_is_directory(inode))
638 struct wim_ads_entry *ads_entry;
640 ads_entry = inode_get_ads_entry(inode, stream_name,
643 stream_idx = ads_idx + 1;
644 lte = ads_entry->lte;
655 *dentry_ret = dentry;
659 *stream_idx_ret = stream_idx;
664 /* Resolve an inode's lookup table entries
666 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
667 * lookup table) with pointers directly to the lookup table entries. A circular
668 * linked list of streams sharing the same lookup table entry is created.
670 * This function always succeeds; unresolved lookup table entries are given a
674 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
677 if (!inode->i_resolved) {
678 struct wim_lookup_table_entry *lte;
679 /* Resolve the default file stream */
680 lte = __lookup_resource(table, inode->i_hash);
682 inode->i_resolved = 1;
684 /* Resolve the alternate data streams */
685 for (u16 i = 0; i < inode->i_num_ads; i++) {
686 struct wim_ads_entry *cur_entry = &inode->i_ads_entries[i];
687 lte = __lookup_resource(table, cur_entry->hash);
688 cur_entry->lte = lte;
694 inode_unresolve_ltes(struct wim_inode *inode)
696 if (inode->i_resolved) {
698 copy_hash(inode->i_hash, inode->i_lte->hash);
700 zero_out_hash(inode->i_hash);
702 for (u16 i = 0; i < inode->i_num_ads; i++) {
703 if (inode->i_ads_entries[i].lte)
704 copy_hash(inode->i_ads_entries[i].hash,
705 inode->i_ads_entries[i].lte->hash);
707 zero_out_hash(inode->i_ads_entries[i].hash);
709 inode->i_resolved = 0;
714 * Returns the lookup table entry for stream @stream_idx of the inode, where
715 * stream_idx = 0 means the default un-named file stream, and stream_idx >= 1
716 * corresponds to an alternate data stream.
718 * This works for both resolved and un-resolved inodes.
720 struct wim_lookup_table_entry *
721 inode_stream_lte(const struct wim_inode *inode, unsigned stream_idx,
722 const struct wim_lookup_table *table)
724 if (inode->i_resolved)
725 return inode_stream_lte_resolved(inode, stream_idx);
727 return inode_stream_lte_unresolved(inode, stream_idx, table);
731 /* Return the lookup table entry for the unnamed data stream of an inode, or
732 * NULL if there is none.
734 * You'd think this would be easier than it actually is, since the unnamed data
735 * stream should be the one referenced from the inode itself. Alas, if there
736 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
737 * data stream in one of the alternate data streams instead of inside the WIM
738 * dentry itself. So we need to check the alternate data streams too.
740 * Also, note that a dentry may appear to have more than one unnamed stream, but
741 * if the SHA1 message digest is all 0's then the corresponding stream does not
742 * really "count" (this is the case for the inode's own file stream when the
743 * file stream that should be there is actually in one of the alternate stream
744 * entries.). This is despite the fact that we may need to extract such a
745 * missing entry as an empty file or empty named data stream.
747 struct wim_lookup_table_entry *
748 inode_unnamed_lte(const struct wim_inode *inode,
749 const struct wim_lookup_table *table)
751 if (inode->i_resolved)
752 return inode_unnamed_lte_resolved(inode);
754 return inode_unnamed_lte_unresolved(inode, table);
758 lte_add_stream_size(struct wim_lookup_table_entry *lte, void *total_bytes_p)
760 *(u64*)total_bytes_p += lte->resource_entry.size;
765 lookup_table_total_stream_size(struct wim_lookup_table *table)
768 for_lookup_table_entry(table, lte_add_stream_size, &total_size);