Various minor changes and fixes.
[wimlib] / src / integrity.c
1 /*
2  * integrity.c
3  *
4  * WIM files can optionally contain an array of SHA1 message digests at the end,
5  * one digest for each 1 MB of the file.  This file implements the checking of
6  * the digests, and the writing of the digests for new WIM files.
7  */
8
9 /*
10  * Copyright (C) 2012 Eric Biggers
11  *
12  * This file is part of wimlib, a library for working with WIM files.
13  *
14  * wimlib is free software; you can redistribute it and/or modify it under the
15  * terms of the GNU Lesser General Public License as published by the Free
16  * Software Foundation; either version 2.1 of the License, or (at your option)
17  * any later version.
18  *
19  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
20  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
21  * A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
22  * details.
23  *
24  * You should have received a copy of the GNU Lesser General Public License
25  * along with wimlib; if not, see http://www.gnu.org/licenses/.
26  */
27
28 #include "wimlib_internal.h"
29 #include "io.h"
30 #include "sha1.h"
31
32 /* Size, in bytes, of each SHA1-summed chunk, when wimlib writes integrity
33  * information. */
34 #define INTEGRITY_CHUNK_SIZE 10485760
35
36 /*
37  * Verifies the integrity of a WIM. 
38  *
39  * @fp:            FILE* of the WIM, currently positioned at the end of the header. 
40  * @num_bytes:     Number of bytes to verify the integrity of.
41  * @chunk_size:    Chunk size per SHA1 message digest.
42  * @sha1sums:      Array of SHA1 message digests; 20 bytes each, one per chunk.
43  * @show_progress: Nonzero if the percent complete is to be printed after every
44  *                      chunk.
45  * @status:        On success, set to WIM_INTEGRITY_OK or WIM_INTEGRITY_NOT_OK 
46  *                      based on whether the WIM is intact or not.
47  */
48 static int verify_integrity(FILE *fp, u64 num_bytes, u32 chunk_size, 
49                             const u8 *sha1sums, int show_progress,
50                             int *status)
51 {
52         char  *chunk_buf;
53         u8     resblock[WIM_HASH_SIZE];
54         u64    bytes_remaining;
55         size_t bytes_to_read;
56         uint   percent_done;
57         int    ret;
58
59         chunk_buf = MALLOC(chunk_size);
60         if (!chunk_buf) {
61                 ERROR("Failed to allocate %u byte buffer for integrity chunks",
62                       chunk_size);
63                 return WIMLIB_ERR_NOMEM;
64         }
65         bytes_remaining = num_bytes;
66         while (bytes_remaining != 0) {
67                 if (show_progress) {
68                         percent_done = (num_bytes - bytes_remaining) * 100 / 
69                                         num_bytes;
70                         printf("Verifying integrity of WIM (%"PRIu64" bytes "
71                                         "remaining, %u%% done)       \r", 
72                                         bytes_remaining, percent_done);
73                         fflush(stdout);
74                 }
75                 bytes_to_read = min(chunk_size, bytes_remaining);
76                 if (fread(chunk_buf, 1, bytes_to_read, fp) != bytes_to_read) {
77                         if (feof(fp)) {
78                                 ERROR("Unexpected EOF while verifying "
79                                       "integrity of WIM");
80                         } else {
81                                 ERROR_WITH_ERRNO("File stream error while "
82                                                  "verifying integrity of WIM");
83                         }
84                         ret = WIMLIB_ERR_READ;
85                         goto verify_integrity_error;
86                 }
87                 sha1_buffer(chunk_buf, bytes_to_read, resblock);
88                 if (memcmp(resblock, sha1sums, WIM_HASH_SIZE) != 0) {
89                         *status = WIM_INTEGRITY_NOT_OK;
90                         goto verify_integrity_done;
91                 }
92                 sha1sums += WIM_HASH_SIZE;
93                 bytes_remaining -= bytes_to_read;
94         }
95         *status = WIM_INTEGRITY_OK;
96 verify_integrity_done:
97         ret = 0;
98 verify_integrity_error:
99         FREE(chunk_buf);
100         if (show_progress)
101                 putchar('\n');
102         return ret;
103 }
104
105 /*
106  * Verifies the integrity of the WIM. 
107  *
108  * @show_progress: Nonzero if the percent complete is to be printed after every
109  *                      chunk.
110  * @status:        On success, set to WIM_INTEGRITY_OK, WIM_INTEGRITY_NOT_OK,
111  *                      or WIM_INTEGRITY_NONEXISTENT.
112  *
113  * Returns: 0, WIMLIB_ERR_INVALID_INTEGRITY_TABLE, WIMLIB_ERR_NOMEM, or
114  * WIMLIB_ERR_READ.  If nonzero, the boolean pointed to by @ok is not changed.
115  */
116 int check_wim_integrity(WIMStruct *w, int show_progress, int *status)
117 {
118
119         struct resource_entry *res_entry;
120         int ctype;
121         u8 *buf = NULL;
122         int ret;
123         u32 integrity_table_size;
124         u32 num_entries;
125         u32 chunk_size;
126         const u8 *p;
127         u64 expected_size;
128         u64 end_lookup_table_offset;
129         u64 bytes_to_check;
130         u64 expected_num_entries;
131
132         res_entry = &w->hdr.integrity;
133         if (res_entry->size == 0) {
134                 DEBUG("No integrity information.");
135                 *status = WIM_INTEGRITY_NONEXISTENT;
136                 return 0;
137         }
138         ctype = wim_resource_compression_type(w, res_entry);
139         if (res_entry->original_size < 12) {
140                 ERROR("Integrity table is too short");
141                 return WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
142         }
143
144         /* Read the integrity table into memory. */
145         buf = MALLOC(res_entry->original_size);
146         if (!buf) {
147                 ERROR("Out of memory (needed %zu bytes for integrity table)",
148                       res_entry->original_size);
149                 ret = WIMLIB_ERR_NOMEM;
150                 goto out;
151         }
152         ret = read_full_resource(w->fp, res_entry->size,
153                                  res_entry->original_size,
154                                  res_entry->offset, ctype, buf);
155         if (ret != 0) {
156                 ERROR("Failed to read integrity table (size = %"PRIu64", "
157                       "original_size = %"PRIu64", offset = "
158                       "%"PRIu64", ctype = %d",
159                       (u64)res_entry->size, res_entry->original_size,
160                       res_entry->offset, ctype);
161                 goto out;
162         }
163
164         p = get_u32(buf, &integrity_table_size);
165         p = get_u32(p, &num_entries);
166         p = get_u32(p, &chunk_size);
167
168         /* p now points to the array of SHA1 message digests for the WIM. */
169
170         /* Make sure the integrity table is the right size. */
171         if (integrity_table_size != res_entry->original_size) {
172                 ERROR("Inconsistent integrity table sizes: header says %u "
173                       "bytes but resource entry says "
174                       "%"PRIu64" bytes",
175                       integrity_table_size, res_entry->original_size);
176                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
177                 goto out;
178         }
179
180         DEBUG("integrity_table_size = %u, num_entries = %u, chunk_size = %u",
181               integrity_table_size, num_entries, chunk_size);
182
183
184         expected_size = num_entries * WIM_HASH_SIZE + 12;
185
186         if (integrity_table_size != expected_size) {
187                 ERROR("Integrity table is %u bytes, but expected %"PRIu64" "
188                       "bytes to hold %u entries", 
189                       integrity_table_size, expected_size, num_entries);
190                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
191                 goto out;
192         }
193
194         if (chunk_size == 0) {
195                 ERROR("Cannot use integrity chunk size of 0");
196                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
197                 goto out;
198         }
199
200         end_lookup_table_offset = w->hdr.lookup_table_res_entry.offset +
201                                   w->hdr.lookup_table_res_entry.size;
202
203         bytes_to_check = end_lookup_table_offset - WIM_HEADER_DISK_SIZE;
204
205         expected_num_entries = (bytes_to_check + chunk_size - 1) / chunk_size;
206
207         if (num_entries != expected_num_entries) {
208                 ERROR("%"PRIu64" entries would be required to checksum "
209                       "the %"PRIu64" bytes from the end of the header to the",
210                       expected_num_entries, bytes_to_check);
211                 ERROR("end of the lookup table with a chunk size of %u, but "
212                       "there were only %u entries", expected_num_entries,
213                       bytes_to_check, chunk_size, num_entries);
214                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
215                 goto out;
216         }
217
218         /* The integrity checking starts after the header, so seek to the offset
219          * in the WIM after the header. */
220
221         if (fseeko(w->fp, WIM_HEADER_DISK_SIZE, SEEK_SET) != 0) {
222                 ERROR_WITH_ERRNO("Failed to seek to byte %u of WIM to check "
223                                  "integrity", WIM_HEADER_DISK_SIZE);
224                 ret = WIMLIB_ERR_READ;
225                 goto out;
226         }
227         /* call verify_integrity(), which does the actual checking of the SHA1
228          * message digests. */
229         ret = verify_integrity(w->fp, bytes_to_check, chunk_size, p, 
230                                show_progress, status);
231 out:
232         FREE(buf);
233         return ret;
234 }
235
236 /* 
237  * Writes integrity information to the output stream for a WIM file being
238  * written. 
239  *
240  * @end_header_offset is the offset of the byte after the header, which is the
241  *      beginning of the region that is checksummed.
242  *
243  * @end_lookup_table_offset is the offset of the byte after the lookup table,
244  *      which is the end of the region that is checksummed. 
245  */
246 int write_integrity_table(FILE *out, u64 end_header_offset, 
247                           u64 end_lookup_table_offset, int show_progress)
248 {
249         u64   bytes_to_check;
250         u64   bytes_remaining;
251         u8   *buf;
252         u8   *p;
253         char *chunk_buf;
254         u32   num_entries;
255         u32   integrity_table_size;
256         int   ret;
257
258         DEBUG("Writing integrity table");
259         if (fseeko(out, end_header_offset, SEEK_SET) != 0) {
260                 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" of WIM to "
261                                  "calculate integrity data", end_header_offset);
262                 return WIMLIB_ERR_WRITE;
263         }
264
265         bytes_to_check = end_lookup_table_offset - end_header_offset;
266         num_entries = bytes_to_check / INTEGRITY_CHUNK_SIZE +
267                         (bytes_to_check % INTEGRITY_CHUNK_SIZE != 0);
268         integrity_table_size = num_entries * WIM_HASH_SIZE + 3 * sizeof(u32);
269
270         DEBUG("integrity table size = %u", integrity_table_size);
271
272
273         buf = MALLOC(integrity_table_size);
274         if (!buf) {
275                 ERROR("Failed to allocate %u bytes for integrity table",
276                       integrity_table_size);
277                 return WIMLIB_ERR_NOMEM;
278         }
279
280         p = put_u32(buf, integrity_table_size);
281         p = put_u32(p, num_entries);
282         p = put_u32(p, INTEGRITY_CHUNK_SIZE);
283
284         chunk_buf = MALLOC(INTEGRITY_CHUNK_SIZE);
285         if (!chunk_buf) {
286                 ERROR("Failed to allocate %u bytes for integrity chunk buffer",
287                       INTEGRITY_CHUNK_SIZE);
288                 ret = WIMLIB_ERR_NOMEM;
289                 goto err2;
290         }
291
292         bytes_remaining = bytes_to_check;
293
294         DEBUG("Bytes to check = %"PRIu64, bytes_to_check);
295
296         while (bytes_remaining != 0) {
297
298                 uint percent_done = (bytes_to_check - bytes_remaining) * 
299                                     100 / bytes_to_check;
300
301                 if (show_progress) {
302                         printf("Calculating integrity checksums for WIM "
303                                         "(%"PRIu64" bytes remaining, %u%% "
304                                         "done)      \r", 
305                                         bytes_remaining, percent_done);
306                         fflush(stdout);
307                 }
308
309
310                 size_t bytes_to_read = min(INTEGRITY_CHUNK_SIZE, bytes_remaining);
311                 size_t bytes_read = fread(chunk_buf, 1, bytes_to_read, out);
312                 if (bytes_read != bytes_to_read) {
313                         if (feof(out)) {
314                                 ERROR("Unexpected EOF while calculating "
315                                       "integrity checksums");
316                         } else {
317                                 ERROR_WITH_ERRNO("File stream error while "
318                                                  "calculating integrity "
319                                                  "checksums");
320                         }
321                         ret = WIMLIB_ERR_READ;
322                         goto err2;
323                 }
324                 sha1_buffer(chunk_buf, bytes_read, p);
325                 p += WIM_HASH_SIZE;
326                 bytes_remaining -= bytes_read;
327         }
328         if (show_progress)
329                 puts("Calculating integrity checksums for WIM "
330                                 "(0 bytes remaining, 100% done)"
331                                 "                       ");
332
333         if (fseeko(out, 0, SEEK_END) != 0) {
334                 ERROR_WITH_ERRNO("Failed to seek to end of WIM to write "
335                                  "integrity table");
336                 ret = WIMLIB_ERR_WRITE;
337                 goto err1;
338         }
339
340         if (fwrite(buf, 1, integrity_table_size, out) != integrity_table_size) {
341                 ERROR_WITH_ERRNO("Failed to write integrity table to end of "
342                                  "WIM");
343                 ret = WIMLIB_ERR_WRITE;
344                 goto err1;
345         }
346         ret = 0;
347 err1:
348         FREE(chunk_buf);
349 err2:
350         FREE(buf);
351         return ret;
352 }