12968bf91004a0b2e54d6b59c677c5e8c5f261c2
[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[SHA1_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 (!hashes_equal(resblock, sha1sums)) {
89                         *status = WIM_INTEGRITY_NOT_OK;
90                         goto verify_integrity_done;
91                 }
92                 sha1sums += SHA1_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         if (res_entry->original_size < 12) {
139                 ERROR("Integrity table is too short");
140                 return WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
141         }
142         if (res_entry->flags & WIM_RESHDR_FLAG_COMPRESSED) {
143                 ERROR("Didn't expect a compressed integrity table");
144                 return WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
145         }
146
147         /* Read the integrity table into memory. */
148         buf = MALLOC(res_entry->original_size);
149         if (!buf) {
150                 ERROR("Out of memory (needed %zu bytes for integrity table)",
151                       res_entry->original_size);
152                 ret = WIMLIB_ERR_NOMEM;
153                 goto out;
154         }
155         ret = read_uncompressed_resource(w->fp, res_entry->offset,
156                                          res_entry->original_size, buf);
157         if (ret != 0) {
158                 ERROR("Failed to read integrity table (size = %"PRIu64", "
159                       "original_size = %"PRIu64", offset = "
160                       "%"PRIu64")",
161                       (u64)res_entry->size, res_entry->original_size,
162                       res_entry->offset);
163                 goto out;
164         }
165
166         p = get_u32(buf, &integrity_table_size);
167         p = get_u32(p, &num_entries);
168         p = get_u32(p, &chunk_size);
169
170         /* p now points to the array of SHA1 message digests for the WIM. */
171
172         /* Make sure the integrity table is the right size. */
173         if (integrity_table_size != res_entry->original_size) {
174                 ERROR("Inconsistent integrity table sizes: header says %u "
175                       "bytes but resource entry says "
176                       "%"PRIu64" bytes",
177                       integrity_table_size, res_entry->original_size);
178                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
179                 goto out;
180         }
181
182         DEBUG("integrity_table_size = %u, num_entries = %u, chunk_size = %u",
183               integrity_table_size, num_entries, chunk_size);
184
185
186         expected_size = num_entries * SHA1_HASH_SIZE + 12;
187
188         if (integrity_table_size != expected_size) {
189                 ERROR("Integrity table is %u bytes, but expected %"PRIu64" "
190                       "bytes to hold %u entries", 
191                       integrity_table_size, expected_size, num_entries);
192                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
193                 goto out;
194         }
195
196         if (chunk_size == 0) {
197                 ERROR("Cannot use integrity chunk size of 0");
198                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
199                 goto out;
200         }
201
202         end_lookup_table_offset = w->hdr.lookup_table_res_entry.offset +
203                                   w->hdr.lookup_table_res_entry.size;
204
205         bytes_to_check = end_lookup_table_offset - WIM_HEADER_DISK_SIZE;
206
207         expected_num_entries = (bytes_to_check + chunk_size - 1) / chunk_size;
208
209         if (num_entries != expected_num_entries) {
210                 ERROR("%"PRIu64" entries would be required to checksum "
211                       "the %"PRIu64" bytes from the end of the header to the",
212                       expected_num_entries, bytes_to_check);
213                 ERROR("end of the lookup table with a chunk size of %u, but "
214                       "there were only %u entries", chunk_size, num_entries);
215                 ret = WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
216                 goto out;
217         }
218
219         /* The integrity checking starts after the header, so seek to the offset
220          * in the WIM after the header. */
221
222         if (fseeko(w->fp, WIM_HEADER_DISK_SIZE, SEEK_SET) != 0) {
223                 ERROR_WITH_ERRNO("Failed to seek to byte %u of WIM to check "
224                                  "integrity", WIM_HEADER_DISK_SIZE);
225                 ret = WIMLIB_ERR_READ;
226                 goto out;
227         }
228         /* call verify_integrity(), which does the actual checking of the SHA1
229          * message digests. */
230         ret = verify_integrity(w->fp, bytes_to_check, chunk_size, p, 
231                                show_progress, status);
232 out:
233         FREE(buf);
234         return ret;
235 }
236
237 /* 
238  * Writes integrity information to the output stream for a WIM file being
239  * written. 
240  *
241  * @end_header_offset is the offset of the byte after the header, which is the
242  *      beginning of the region that is checksummed.
243  *
244  * @end_lookup_table_offset is the offset of the byte after the lookup table,
245  *      which is the end of the region that is checksummed. 
246  */
247 int write_integrity_table(FILE *out, u64 end_header_offset, 
248                           u64 end_lookup_table_offset, int show_progress)
249 {
250         u64   bytes_to_check;
251         u64   bytes_remaining;
252         u8   *buf;
253         u8   *p;
254         char *chunk_buf;
255         u32   num_entries;
256         u32   integrity_table_size;
257         int   ret;
258
259         DEBUG("Writing integrity table");
260         if (fseeko(out, end_header_offset, SEEK_SET) != 0) {
261                 ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" of WIM to "
262                                  "calculate integrity data", end_header_offset);
263                 return WIMLIB_ERR_WRITE;
264         }
265
266         bytes_to_check = end_lookup_table_offset - end_header_offset;
267         num_entries = bytes_to_check / INTEGRITY_CHUNK_SIZE +
268                         (bytes_to_check % INTEGRITY_CHUNK_SIZE != 0);
269         integrity_table_size = num_entries * SHA1_HASH_SIZE + 3 * sizeof(u32);
270
271         DEBUG("integrity table size = %u", integrity_table_size);
272
273
274         buf = MALLOC(integrity_table_size);
275         if (!buf) {
276                 ERROR("Failed to allocate %u bytes for integrity table",
277                       integrity_table_size);
278                 return WIMLIB_ERR_NOMEM;
279         }
280
281         p = put_u32(buf, integrity_table_size);
282         p = put_u32(p, num_entries);
283         p = put_u32(p, INTEGRITY_CHUNK_SIZE);
284
285         chunk_buf = MALLOC(INTEGRITY_CHUNK_SIZE);
286         if (!chunk_buf) {
287                 ERROR("Failed to allocate %u bytes for integrity chunk buffer",
288                       INTEGRITY_CHUNK_SIZE);
289                 ret = WIMLIB_ERR_NOMEM;
290                 goto err2;
291         }
292
293         bytes_remaining = bytes_to_check;
294
295         DEBUG("Bytes to check = %"PRIu64, bytes_to_check);
296
297         while (bytes_remaining != 0) {
298
299                 uint percent_done = (bytes_to_check - bytes_remaining) * 
300                                     100 / bytes_to_check;
301
302                 if (show_progress) {
303                         printf("Calculating integrity checksums for WIM "
304                                         "(%"PRIu64" bytes remaining, %u%% "
305                                         "done)      \r", 
306                                         bytes_remaining, percent_done);
307                         fflush(stdout);
308                 }
309
310
311                 size_t bytes_to_read = min(INTEGRITY_CHUNK_SIZE, bytes_remaining);
312                 size_t bytes_read = fread(chunk_buf, 1, bytes_to_read, out);
313                 if (bytes_read != bytes_to_read) {
314                         if (feof(out)) {
315                                 ERROR("Unexpected EOF while calculating "
316                                       "integrity checksums");
317                         } else {
318                                 ERROR_WITH_ERRNO("File stream error while "
319                                                  "calculating integrity "
320                                                  "checksums");
321                         }
322                         ret = WIMLIB_ERR_READ;
323                         goto err2;
324                 }
325                 sha1_buffer(chunk_buf, bytes_read, p);
326                 p += SHA1_HASH_SIZE;
327                 bytes_remaining -= bytes_read;
328         }
329         if (show_progress)
330                 puts("Calculating integrity checksums for WIM "
331                                 "(0 bytes remaining, 100% done)"
332                                 "                       ");
333
334         if (fseeko(out, 0, SEEK_END) != 0) {
335                 ERROR_WITH_ERRNO("Failed to seek to end of WIM to write "
336                                  "integrity table");
337                 ret = WIMLIB_ERR_WRITE;
338                 goto err1;
339         }
340
341         if (fwrite(buf, 1, integrity_table_size, out) != integrity_table_size) {
342                 ERROR_WITH_ERRNO("Failed to write integrity table to end of "
343                                  "WIM");
344                 ret = WIMLIB_ERR_WRITE;
345                 goto err1;
346         }
347         ret = 0;
348 err1:
349         FREE(chunk_buf);
350 err2:
351         FREE(buf);
352         return ret;
353 }