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 Lesser General Public License as published by the Free
15 * Software Foundation; either version 2.1 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 Lesser General Public License for more
23 * You should have received a copy of the GNU Lesser 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"
32 struct lookup_table *new_lookup_table(size_t capacity)
34 struct lookup_table *table;
35 struct hlist_head *array;
37 table = MALLOC(sizeof(struct lookup_table));
40 array = CALLOC(capacity, sizeof(array[0]));
45 table->num_entries = 0;
46 table->capacity = capacity;
50 ERROR("Failed to allocate memory for lookup table with capacity %zu",
55 struct lookup_table_entry *new_lookup_table_entry()
57 struct lookup_table_entry *lte;
59 lte = CALLOC(1, sizeof(struct lookup_table_entry));
61 ERROR("Out of memory (tried to allocate %zu bytes for "
62 "lookup table entry)",
63 sizeof(struct lookup_table_entry));
69 INIT_LIST_HEAD(<e->lte_group_list);
73 void free_lookup_table_entry(struct lookup_table_entry *lte)
76 if (lte->staging_list.next)
77 list_del(<e->staging_list);
78 if (lte->resource_location != RESOURCE_IN_WIM &&
79 lte->resource_location != RESOURCE_NONEXISTENT)
80 FREE(lte->file_on_disk);
85 static int do_free_lookup_table_entry(struct lookup_table_entry *entry,
88 free_lookup_table_entry(entry);
93 void free_lookup_table(struct lookup_table *table)
95 DEBUG("Freeing lookup table");
98 for_lookup_table_entry(table,
99 do_free_lookup_table_entry,
108 * Inserts an entry into the lookup table.
110 * @table: A pointer to the lookup table.
111 * @entry: A pointer to the entry to insert.
113 void lookup_table_insert(struct lookup_table *table,
114 struct lookup_table_entry *lte)
116 size_t i = lte->hash_short % table->capacity;
117 hlist_add_head(<e->hash_list, &table->array[i]);
119 /* XXX Make the table grow when too many entries have been inserted. */
120 table->num_entries++;
125 /* Decrements the reference count for the lookup table entry @lte. If its
126 * reference count reaches 0, it is unlinked from the lookup table. If,
127 * furthermore, the entry has no opened file descriptors associated with it, the
129 struct lookup_table_entry *
130 lte_decrement_refcnt(struct lookup_table_entry *lte, struct lookup_table *table)
133 wimlib_assert(lte->refcnt);
134 if (--lte->refcnt == 0) {
135 lookup_table_unlink(table, lte);
137 if (lte->num_opened_fds == 0)
140 free_lookup_table_entry(lte);
149 * Calls a function on all the entries in the lookup table. Stop early and
150 * return nonzero if any call to the function returns nonzero.
152 int for_lookup_table_entry(struct lookup_table *table,
153 int (*visitor)(struct lookup_table_entry *, void *),
156 struct lookup_table_entry *lte;
157 struct hlist_node *pos, *tmp;
160 for (size_t i = 0; i < table->capacity; i++) {
161 hlist_for_each_entry_safe(lte, pos, tmp, &table->array[i],
164 ret = visitor(lte, arg);
174 * Reads the lookup table from a WIM file.
176 int read_lookup_table(WIMStruct *w)
179 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
181 struct lookup_table *table;
183 DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
184 w->hdr.lookup_table_res_entry.offset,
185 w->hdr.lookup_table_res_entry.original_size);
187 if (fseeko(w->fp, w->hdr.lookup_table_res_entry.offset, SEEK_SET) != 0) {
188 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read "
190 w->hdr.lookup_table_res_entry.offset);
191 return WIMLIB_ERR_READ;
194 num_entries = w->hdr.lookup_table_res_entry.original_size /
195 WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
196 table = new_lookup_table(num_entries * 2 + 1);
198 return WIMLIB_ERR_NOMEM;
200 while (num_entries--) {
202 struct lookup_table_entry *cur_entry, *duplicate_entry;
204 if (fread(buf, 1, sizeof(buf), w->fp) != sizeof(buf)) {
206 ERROR("Unexpected EOF in WIM lookup table!");
208 ERROR_WITH_ERRNO("Error reading WIM lookup "
211 ret = WIMLIB_ERR_READ;
214 cur_entry = new_lookup_table_entry();
216 ret = WIMLIB_ERR_NOMEM;
220 cur_entry->resource_location = RESOURCE_IN_WIM;
222 p = get_resource_entry(buf, &cur_entry->resource_entry);
223 p = get_u16(p, &cur_entry->part_number);
224 p = get_u32(p, &cur_entry->refcnt);
225 p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
227 if (is_zero_hash(cur_entry->hash)) {
228 ERROR("The WIM lookup table contains an entry with a "
229 "SHA1 message digest of all 0's");
230 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
235 duplicate_entry = __lookup_resource(table, cur_entry->hash);
236 if (duplicate_entry) {
237 ERROR("The WIM lookup table contains two entries with the "
238 "same SHA1 message digest!");
239 ERROR("The first entry is:");
240 print_lookup_table_entry(duplicate_entry);
241 ERROR("The second entry is:");
242 print_lookup_table_entry(cur_entry);
243 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
247 lookup_table_insert(table, cur_entry);
249 if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
250 && (cur_entry->resource_entry.size !=
251 cur_entry->resource_entry.original_size))
253 ERROR("Found uncompressed resource with original size "
254 "not the same as compressed size");
255 ERROR("The lookup table entry for the resource is as follows:");
256 print_lookup_table_entry(cur_entry);
257 ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
261 DEBUG("Done reading lookup table.");
262 w->lookup_table = table;
265 free_lookup_table(table);
271 * Writes a lookup table entry to the output file.
273 int write_lookup_table_entry(struct lookup_table_entry *lte, void *__out)
276 u8 buf[WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
281 /* Don't write entries that have not had file resources or metadata
282 * resources written for them. */
283 if (lte->out_refcnt == 0)
286 if (lte->output_resource_entry.flags & WIM_RESHDR_FLAG_METADATA)
287 DEBUG("Writing metadata entry at %lu (orig size = %zu)",
288 ftello(out), lte->output_resource_entry.original_size);
290 p = put_resource_entry(buf, <e->output_resource_entry);
291 p = put_u16(p, lte->part_number);
292 p = put_u32(p, lte->out_refcnt);
293 p = put_bytes(p, SHA1_HASH_SIZE, lte->hash);
294 if (fwrite(buf, 1, sizeof(buf), out) != sizeof(buf)) {
295 ERROR_WITH_ERRNO("Failed to write lookup table entry");
296 return WIMLIB_ERR_WRITE;
303 int zero_out_refcnts(struct lookup_table_entry *entry, void *ignore)
305 entry->out_refcnt = 0;
309 void print_lookup_table_entry(const struct lookup_table_entry *lte)
315 printf("Offset = %"PRIu64" bytes\n",
316 lte->resource_entry.offset);
317 printf("Size = %"PRIu64" bytes\n",
318 (u64)lte->resource_entry.size);
319 printf("Original size = %"PRIu64" bytes\n",
320 lte->resource_entry.original_size);
321 printf("Part Number = %hu\n", lte->part_number);
322 printf("Reference Count = %u\n", lte->refcnt);
324 print_hash(lte->hash);
327 u8 flags = lte->resource_entry.flags;
328 if (flags & WIM_RESHDR_FLAG_COMPRESSED)
329 fputs("WIM_RESHDR_FLAG_COMPRESSED, ", stdout);
330 if (flags & WIM_RESHDR_FLAG_FREE)
331 fputs("WIM_RESHDR_FLAG_FREE, ", stdout);
332 if (flags & WIM_RESHDR_FLAG_METADATA)
333 fputs("WIM_RESHDR_FLAG_METADATA, ", stdout);
334 if (flags & WIM_RESHDR_FLAG_SPANNED)
335 fputs("WIM_RESHDR_FLAG_SPANNED, ", stdout);
337 switch (lte->resource_location) {
338 case RESOURCE_IN_WIM:
339 if (lte->wim->filename) {
340 printf("WIM file = `%s'\n",
344 case RESOURCE_IN_FILE_ON_DISK:
345 printf("File on Disk = `%s'\n", lte->file_on_disk);
347 case RESOURCE_IN_STAGING_FILE:
348 printf("Staging File = `%s'\n", lte->staging_file_name);
356 static int do_print_lookup_table_entry(struct lookup_table_entry *lte,
359 print_lookup_table_entry(lte);
364 * Prints the lookup table of a WIM file.
366 WIMLIBAPI void wimlib_print_lookup_table(WIMStruct *w)
368 for_lookup_table_entry(w->lookup_table,
369 do_print_lookup_table_entry,
374 * Looks up an entry in the lookup table.
376 struct lookup_table_entry *
377 __lookup_resource(const struct lookup_table *table, const u8 hash[])
380 struct lookup_table_entry *lte;
381 struct hlist_node *pos;
383 i = *(size_t*)hash % table->capacity;
384 hlist_for_each_entry(lte, pos, &table->array[i], hash_list)
385 if (hashes_equal(hash, lte->hash))
391 * Finds the dentry, lookup table entry, and stream index for a WIM file stream,
394 * This is only for pre-resolved dentries.
396 int lookup_resource(WIMStruct *w, const char *path,
398 struct dentry **dentry_ret,
399 struct lookup_table_entry **lte_ret,
400 unsigned *stream_idx_ret)
402 struct dentry *dentry;
403 struct lookup_table_entry *lte;
405 const char *stream_name = NULL;
408 if (lookup_flags & LOOKUP_FLAG_ADS_OK) {
409 stream_name = path_stream_name(path);
411 p = (char*)stream_name - 1;
416 dentry = get_dentry(w, path);
422 wimlib_assert(dentry->resolved);
425 if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
426 && dentry_is_directory(dentry))
430 size_t stream_name_len = strlen(stream_name);
431 for (u16 i = 0; i < dentry->num_ads; i++) {
432 if (ads_entry_has_name(&dentry->ads_entries[i],
437 lte = dentry->ads_entries[i].lte;
445 *dentry_ret = dentry;
449 *stream_idx_ret = stream_idx;
453 /* Resolve a dentry's lookup table entries
455 * This replaces the SHA1 hash fields (which are used to lookup an entry in the
456 * lookup table) with pointers directly to the lookup table entries. A circular
457 * linked list of streams sharing the same lookup table entry is created.
459 * This function always succeeds; unresolved lookup table entries are given a
462 int dentry_resolve_ltes(struct dentry *dentry, void *__table)
464 struct lookup_table *table = __table;
465 struct lookup_table_entry *lte;
467 if (dentry->resolved)
470 /* Resolve the default file stream */
471 lte = __lookup_resource(table, dentry->hash);
473 list_add(&dentry->lte_group_list.list, <e->lte_group_list);
475 INIT_LIST_HEAD(&dentry->lte_group_list.list);
477 dentry->lte_group_list.type = STREAM_TYPE_NORMAL;
478 dentry->resolved = true;
480 /* Resolve the alternate data streams */
481 if (dentry->ads_entries_status != ADS_ENTRIES_USER) {
482 for (u16 i = 0; i < dentry->num_ads; i++) {
483 struct ads_entry *cur_entry = &dentry->ads_entries[i];
485 lte = __lookup_resource(table, cur_entry->hash);
487 list_add(&cur_entry->lte_group_list.list,
488 <e->lte_group_list);
490 INIT_LIST_HEAD(&cur_entry->lte_group_list.list);
491 cur_entry->lte = lte;
492 cur_entry->lte_group_list.type = STREAM_TYPE_ADS;
498 /* Return the lookup table entry for the unnamed data stream of a dentry, or
499 * NULL if there is none.
501 * You'd think this would be easier than it actually is, since the unnamed data
502 * stream should be the one referenced from the dentry itself. Alas, if there
503 * are named data streams, Microsoft's "imagex.exe" program will put the unnamed
504 * data stream in one of the alternate data streams instead of inside the
505 * dentry. So we need to check the alternate data streams too.
507 * Also, note that a dentry may appear to have than one unnamed stream, but if
508 * the SHA1 message digest is all 0's then the corresponding stream does not
509 * really "count" (this is the case for the dentry's own file stream when the
510 * file stream that should be there is actually in one of the alternate stream
511 * entries.). This is despite the fact that we may need to extract such a
512 * missing entry as an empty file or empty named data stream.
514 struct lookup_table_entry *
515 dentry_unnamed_lte(const struct dentry *dentry,
516 const struct lookup_table *table)
518 if (dentry->resolved)
519 return dentry_unnamed_lte_resolved(dentry);
521 return dentry_unnamed_lte_unresolved(dentry, table);