4 * Some functions to verify that stuff in the WIM is valid. Of course, not
5 * *all* the verifications of the input data are in this file.
9 * Copyright (C) 2012 Eric Biggers
11 * wimlib - Library for working with WIM files
13 * This file is part of wimlib, a library for working with WIM files.
15 * wimlib is free software; you can redistribute it and/or modify it under the
16 * terms of the GNU General Public License as published by the Free
17 * Software Foundation; either version 3 of the License, or (at your option)
20 * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
21 * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
22 * A PARTICULAR PURPOSE. See the GNU General Public License for more
25 * You should have received a copy of the GNU General Public License
26 * along with wimlib; if not, see http://www.gnu.org/licenses/.
29 #include "wimlib_internal.h"
31 #include "lookup_table.h"
33 static int verify_inode(struct wim_inode *inode, const WIMStruct *w)
35 const struct wim_lookup_table *table = w->lookup_table;
36 const struct wim_security_data *sd = wim_const_security_data(w);
37 const struct wim_dentry *first_dentry = inode_first_dentry(inode);
38 int ret = WIMLIB_ERR_INVALID_DENTRY;
40 /* Check the security ID */
41 if (inode->i_security_id < -1) {
42 ERROR("Dentry `%s' has an invalid security ID (%d)",
43 first_dentry->full_path_utf8, inode->i_security_id);
46 if (inode->i_security_id >= sd->num_entries) {
47 ERROR("Dentry `%s' has an invalid security ID (%d) "
48 "(there are only %u entries in the security table)",
49 first_dentry->full_path_utf8, inode->i_security_id,
54 /* Check that lookup table entries for all the resources exist, except
55 * if the SHA1 message digest is all 0's, which indicates there is
56 * intentionally no resource there. */
57 if (w->hdr.total_parts == 1) {
58 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
59 struct wim_lookup_table_entry *lte;
61 hash = inode_stream_hash_unresolved(inode, i);
62 lte = __lookup_resource(table, hash);
63 if (!lte && !is_zero_hash(hash)) {
64 ERROR("Could not find lookup table entry for stream "
65 "%u of dentry `%s'", i, first_dentry->full_path_utf8);
69 lte->real_refcnt += inode->i_nlink;
71 /* The following is now done when required by
72 * wim_run_full_verifications(). */
75 if (lte && !w->full_verification_in_progress &&
76 lte->real_refcnt > lte->refcnt)
78 #ifdef ENABLE_ERROR_MESSAGES
79 WARNING("The following lookup table entry "
80 "has a reference count of %u, but",
82 WARNING("We found %u references to it",
84 WARNING("(One dentry referencing it is at `%s')",
85 first_dentry->full_path_utf8);
87 print_lookup_table_entry(lte);
89 /* Guess what! install.wim for Windows 8
90 * contains many streams referenced by more
91 * dentries than the refcnt stated in the lookup
92 * table entry. So we will need to handle this
93 * case and not just make it be an error... I'm
94 * just setting the reference count to the
95 * number of references we found.
96 * (Unfortunately, even after doing this, the
97 * reference count could be too low if it's also
98 * referenced in other WIM images) */
101 lte->refcnt = lte->real_refcnt;
102 WARNING("Fixing reference count");
111 /* Make sure there is only one un-named stream. */
112 unsigned num_unnamed_streams = 0;
113 for (unsigned i = 0; i <= inode->i_num_ads; i++) {
115 hash = inode_stream_hash_unresolved(inode, i);
116 if (inode_stream_name_len(inode, i) == 0 && !is_zero_hash(hash))
117 num_unnamed_streams++;
119 if (num_unnamed_streams > 1) {
120 ERROR("Dentry `%s' has multiple (%u) un-named streams",
121 first_dentry->full_path_utf8, num_unnamed_streams);
124 inode->i_verified = 1;
130 /* Run some miscellaneous verifications on a WIM dentry */
131 int verify_dentry(struct wim_dentry *dentry, void *wim)
135 if (!dentry->d_inode->i_verified) {
136 ret = verify_inode(dentry->d_inode, wim);
141 /* Cannot have a short name but no long name */
142 if (dentry->short_name_len && !dentry->file_name_len) {
143 ERROR("Dentry `%s' has a short name but no long name",
144 dentry->full_path_utf8);
145 return WIMLIB_ERR_INVALID_DENTRY;
148 /* Make sure root dentry is unnamed */
149 if (dentry_is_root(dentry)) {
150 if (dentry->file_name_len) {
151 ERROR("The root dentry is named `%s', but it must "
152 "be unnamed", dentry->file_name_utf8);
153 return WIMLIB_ERR_INVALID_DENTRY;
158 /* Check timestamps */
159 if (inode->i_last_access_time < inode->i_creation_time ||
160 inode->i_last_write_time < inode->i_creation_time) {
161 WARNING("Dentry `%s' was created after it was last accessed or "
162 "written to", dentry->full_path_utf8);
169 static int image_run_full_verifications(WIMStruct *w)
171 return for_dentry_in_tree(wim_root_dentry(w), verify_dentry, w);
174 static int lte_fix_refcnt(struct wim_lookup_table_entry *lte, void *ctr)
176 if (lte->refcnt != lte->real_refcnt) {
177 WARNING("The following lookup table entry has a reference "
178 "count of %u, but", lte->refcnt);
179 WARNING("We found %u references to it",
181 print_lookup_table_entry(lte);
182 lte->refcnt = lte->real_refcnt;
183 ++*(unsigned long *)ctr;
188 /* Ideally this would be unnecessary... however, the WIMs for Windows 8 are
189 * screwed up because some lookup table entries are referenced more times than
190 * their stated reference counts. So theoretically, if we delete all the
191 * references to a stream and then remove it, it might still be referenced
192 * somewhere else, making a file be missing from the WIM... So, work around this
193 * problem by looking at ALL the images to re-calculate the reference count of
194 * EVERY lookup table entry. This only absolutely has to be done before an image
195 * is deleted or before an image is mounted read-write. */
196 int wim_run_full_verifications(WIMStruct *w)
200 for_lookup_table_entry(w->lookup_table, lte_zero_real_refcnt, NULL);
201 w->all_images_verified = 1;
202 w->full_verification_in_progress = 1;
203 ret = for_image(w, WIMLIB_ALL_IMAGES, image_run_full_verifications);
204 w->full_verification_in_progress = 0;
206 unsigned long num_ltes_with_bogus_refcnt = 0;
207 for (int i = 0; i < w->hdr.image_count; i++)
208 w->image_metadata[i].metadata_lte->real_refcnt++;
209 for_lookup_table_entry(w->lookup_table, lte_fix_refcnt,
210 &num_ltes_with_bogus_refcnt);
211 if (num_ltes_with_bogus_refcnt != 0) {
212 WARNING("A total of %lu entries in the WIM's stream "
213 "lookup table had to have\n"
214 " their reference counts fixed.",
215 num_ltes_with_bogus_refcnt);
218 w->all_images_verified = 0;
224 * verify_swm_set: - Sanity checks to make sure a set of WIMs correctly
225 * correspond to a spanned set.
231 * All parts of the set other than part 1.
233 * @num_additional_swms:
234 * Number of WIMStructs in @additional_swms. Or, the total number of parts
235 * in the set minus 1.
238 * 0 on success; WIMLIB_ERR_SPLIT_INVALID if the set is not valid.
240 int verify_swm_set(WIMStruct *w, WIMStruct **additional_swms,
241 unsigned num_additional_swms)
243 unsigned total_parts = w->hdr.total_parts;
247 if (total_parts != num_additional_swms + 1) {
248 ERROR("`%s' says there are %u parts in the spanned set, "
249 "but %s%u part%s provided",
250 w->filename, total_parts,
251 (num_additional_swms + 1 < total_parts) ? "only " : "",
252 num_additional_swms + 1,
253 (num_additional_swms) ? "s were" : " was");
254 return WIMLIB_ERR_SPLIT_INVALID;
256 if (w->hdr.part_number != 1) {
257 ERROR("WIM `%s' is not the first part of the split WIM.",
259 return WIMLIB_ERR_SPLIT_INVALID;
261 for (unsigned i = 0; i < num_additional_swms; i++) {
262 if (additional_swms[i]->hdr.total_parts != total_parts) {
263 ERROR("WIM `%s' says there are %u parts in the spanned set, "
264 "but %u parts were provided",
265 additional_swms[i]->filename,
266 additional_swms[i]->hdr.total_parts,
268 return WIMLIB_ERR_SPLIT_INVALID;
272 /* keep track of ctype and guid just to make sure they are the same for
274 ctype = wimlib_get_compression_type(w);
277 WIMStruct *parts_to_swms[num_additional_swms];
278 ZERO_ARRAY(parts_to_swms);
279 for (unsigned i = 0; i < num_additional_swms; i++) {
281 WIMStruct *swm = additional_swms[i];
283 if (wimlib_get_compression_type(swm) != ctype) {
284 ERROR("The split WIMs do not all have the same "
286 return WIMLIB_ERR_SPLIT_INVALID;
288 if (memcmp(guid, swm->hdr.guid, WIM_GID_LEN) != 0) {
289 ERROR("The split WIMs do not all have the same "
291 return WIMLIB_ERR_SPLIT_INVALID;
293 if (swm->hdr.part_number == 1) {
294 ERROR("WIMs `%s' and `%s' both are marked as the "
295 "first WIM in the spanned set",
296 w->filename, swm->filename);
297 return WIMLIB_ERR_SPLIT_INVALID;
299 if (swm->hdr.part_number == 0 ||
300 swm->hdr.part_number > total_parts)
302 ERROR("WIM `%s' says it is part %u in the spanned set, "
303 "but the part number must be in the range "
305 swm->filename, swm->hdr.part_number, total_parts);
306 return WIMLIB_ERR_SPLIT_INVALID;
308 if (parts_to_swms[swm->hdr.part_number - 2])
310 ERROR("`%s' and `%s' are both marked as part %u of %u "
311 "in the spanned set",
312 parts_to_swms[swm->hdr.part_number - 2]->filename,
314 swm->hdr.part_number,
316 return WIMLIB_ERR_SPLIT_INVALID;
318 parts_to_swms[swm->hdr.part_number - 2] = swm;