Remove remaining DEBUG() messages
[wimlib] / src / integrity.c
1 /*
2  * integrity.c
3  *
4  * WIM files can optionally contain a table of SHA1 message digests at the end,
5  * one digest for each chunk of the file of some specified size (often 10 MB).
6  * This file implements the checking and writing of this table.
7  */
8
9 /*
10  * Copyright (C) 2012, 2013 Eric Biggers
11  *
12  * This file is free software; you can redistribute it and/or modify it under
13  * the terms of the GNU Lesser General Public License as published by the Free
14  * Software Foundation; either version 3 of the License, or (at your option) any
15  * later version.
16  *
17  * This file is distributed in the hope that it will be useful, but WITHOUT
18  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
19  * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
20  * details.
21  *
22  * You should have received a copy of the GNU Lesser General Public License
23  * along with this file; if not, see http://www.gnu.org/licenses/.
24  */
25
26 #ifdef HAVE_CONFIG_H
27 #  include "config.h"
28 #endif
29
30 #include "wimlib/assert.h"
31 #include "wimlib/endianness.h"
32 #include "wimlib/error.h"
33 #include "wimlib/file_io.h"
34 #include "wimlib/integrity.h"
35 #include "wimlib/progress.h"
36 #include "wimlib/resource.h"
37 #include "wimlib/sha1.h"
38 #include "wimlib/wim.h"
39 #include "wimlib/write.h"
40
41 /* Size, in bytes, of each SHA1-summed chunk, when wimlib writes integrity
42  * information. */
43 #define INTEGRITY_CHUNK_SIZE 10485760
44
45 /* Only use a different chunk size for compatiblity with an existing integrity
46  * table if the chunk size is between these two numbers. */
47 #define INTEGRITY_MIN_CHUNK_SIZE 4096
48 #define INTEGRITY_MAX_CHUNK_SIZE 134217728
49
50 struct integrity_table {
51         u32 size;
52         u32 num_entries;
53         u32 chunk_size;
54         u8  sha1sums[][20];
55 } _packed_attribute;
56
57 static int
58 calculate_chunk_sha1(struct filedes *in_fd, size_t this_chunk_size,
59                      off_t offset, u8 sha1_md[])
60 {
61         u8 buf[BUFFER_SIZE];
62         SHA_CTX ctx;
63         size_t bytes_remaining;
64         size_t bytes_to_read;
65         int ret;
66
67         bytes_remaining = this_chunk_size;
68         sha1_init(&ctx);
69         do {
70                 bytes_to_read = min(bytes_remaining, sizeof(buf));
71                 ret = full_pread(in_fd, buf, bytes_to_read, offset);
72                 if (ret) {
73                         ERROR_WITH_ERRNO("Read error while calculating "
74                                          "integrity checksums");
75                         return ret;
76                 }
77                 sha1_update(&ctx, buf, bytes_to_read);
78                 bytes_remaining -= bytes_to_read;
79                 offset += bytes_to_read;
80         } while (bytes_remaining);
81         sha1_final(sha1_md, &ctx);
82         return 0;
83 }
84
85
86 /*
87  * read_integrity_table: -  Reads the integrity table from a WIM file.
88  *
89  * @wim:
90  *      WIMStruct for the WIM file; @wim->hdr.integrity_table_reshdr specifies
91  *      the location of the integrity table.  @wim->in_fd is expected to be a
92  *      seekable file descriptor to the WIM file opened for reading.
93  *
94  * @num_checked_bytes:
95  *      Number of bytes of data that should be checked by the integrity table.
96  *
97  * @table_ret:
98  *      On success, a pointer to an in-memory structure containing the integrity
99  *      information is written to this location.
100  *
101  * Return values:
102  *      WIMLIB_ERR_SUCCESS (0)
103  *      WIMLIB_ERR_INVALID_INTEGRITY_TABLE
104  *      WIMLIB_ERR_NOMEM
105  *      WIMLIB_ERR_READ
106  *      WIMLIB_ERR_UNEXPECTED_END_OF_FILE
107  */
108 int
109 read_integrity_table(WIMStruct *wim, u64 num_checked_bytes,
110                      struct integrity_table **table_ret)
111 {
112         void *buf;
113         struct integrity_table *table;
114         int ret;
115
116         if (wim->hdr.integrity_table_reshdr.uncompressed_size < 8)
117                 goto invalid;
118
119         ret = wim_reshdr_to_data(&wim->hdr.integrity_table_reshdr, wim, &buf);
120         if (ret)
121                 return ret;
122         table = buf;
123
124         table->size        = le32_to_cpu(table->size);
125         table->num_entries = le32_to_cpu(table->num_entries);
126         table->chunk_size  = le32_to_cpu(table->chunk_size);
127
128         if (table->size != wim->hdr.integrity_table_reshdr.uncompressed_size ||
129             table->size != (u64)table->num_entries * SHA1_HASH_SIZE + 12 ||
130             table->chunk_size == 0 ||
131             table->num_entries != DIV_ROUND_UP(num_checked_bytes, table->chunk_size))
132         {
133                 FREE(table);
134                 goto invalid;
135         }
136
137         *table_ret = table;
138         return 0;
139
140 invalid:
141         return WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
142 }
143
144 /*
145  * calculate_integrity_table():
146  *
147  * Calculates an integrity table for the data in a file beginning at offset 208
148  * (WIM_HEADER_DISK_SIZE).
149  *
150  * @in_fd:
151  *      File descriptor for the file to be checked, opened for reading.  Does
152  *      not need to be at any specific location in the file.
153  *
154  * @new_check_end:
155  *      Offset of byte after the last byte to be checked.
156  *
157  * @old_table:
158  *      If non-NULL, a pointer to the table containing the previously calculated
159  *      integrity data for a prefix of this file.
160  *
161  * @old_check_end:
162  *      If @old_table is non-NULL, the byte after the last byte that was checked
163  *      in the old table.  Must be less than or equal to new_check_end.
164  *
165  * @integrity_table_ret:
166  *      On success, a pointer to the calculated integrity table is written into
167  *      this location.
168  *
169  * Return values:
170  *      WIMLIB_ERR_SUCCESS (0)
171  *      WIMLIB_ERR_NOMEM
172  *      WIMLIB_ERR_READ
173  *      WIMLIB_ERR_UNEXPECTED_END_OF_FILE
174  */
175 static int
176 calculate_integrity_table(struct filedes *in_fd,
177                           off_t new_check_end,
178                           const struct integrity_table *old_table,
179                           off_t old_check_end,
180                           struct integrity_table **integrity_table_ret,
181                           wimlib_progress_func_t progfunc,
182                           void *progctx)
183 {
184         int ret;
185         size_t chunk_size = INTEGRITY_CHUNK_SIZE;
186
187         /* If an old table is provided, set the chunk size to be compatible with
188          * the old chunk size, unless the old chunk size was weird. */
189         if (old_table != NULL) {
190                 if (old_table->num_entries == 0 ||
191                     old_table->chunk_size < INTEGRITY_MIN_CHUNK_SIZE ||
192                     old_table->chunk_size > INTEGRITY_MAX_CHUNK_SIZE)
193                         old_table = NULL;
194                 else
195                         chunk_size = old_table->chunk_size;
196         }
197
198
199         u64 old_check_bytes = old_check_end - WIM_HEADER_DISK_SIZE;
200         u64 new_check_bytes = new_check_end - WIM_HEADER_DISK_SIZE;
201
202         u32 old_num_chunks = DIV_ROUND_UP(old_check_bytes, chunk_size);
203         u32 new_num_chunks = DIV_ROUND_UP(new_check_bytes, chunk_size);
204
205         size_t old_last_chunk_size = MODULO_NONZERO(old_check_bytes, chunk_size);
206         size_t new_last_chunk_size = MODULO_NONZERO(new_check_bytes, chunk_size);
207
208         size_t new_table_size = 12 + new_num_chunks * SHA1_HASH_SIZE;
209
210         struct integrity_table *new_table = MALLOC(new_table_size);
211         if (!new_table)
212                 return WIMLIB_ERR_NOMEM;
213         new_table->num_entries = new_num_chunks;
214         new_table->size = new_table_size;
215         new_table->chunk_size = chunk_size;
216
217         u64 offset = WIM_HEADER_DISK_SIZE;
218         union wimlib_progress_info progress;
219
220         progress.integrity.total_bytes      = new_check_bytes;
221         progress.integrity.total_chunks     = new_num_chunks;
222         progress.integrity.completed_chunks = 0;
223         progress.integrity.completed_bytes  = 0;
224         progress.integrity.chunk_size       = chunk_size;
225         progress.integrity.filename         = NULL;
226
227         ret = call_progress(progfunc, WIMLIB_PROGRESS_MSG_CALC_INTEGRITY,
228                             &progress, progctx);
229         if (ret)
230                 goto out_free_new_table;
231
232         for (u32 i = 0; i < new_num_chunks; i++) {
233                 size_t this_chunk_size;
234                 if (i == new_num_chunks - 1)
235                         this_chunk_size = new_last_chunk_size;
236                 else
237                         this_chunk_size = chunk_size;
238                 if (old_table &&
239                     ((this_chunk_size == chunk_size && i < old_num_chunks - 1) ||
240                       (i == old_num_chunks - 1 && this_chunk_size == old_last_chunk_size)))
241                 {
242                         /* Can use SHA1 message digest from old integrity table
243                          * */
244                         copy_hash(new_table->sha1sums[i], old_table->sha1sums[i]);
245                 } else {
246                         /* Calculate the SHA1 message digest of this chunk */
247                         ret = calculate_chunk_sha1(in_fd, this_chunk_size,
248                                                    offset, new_table->sha1sums[i]);
249                         if (ret)
250                                 goto out_free_new_table;
251                 }
252                 offset += this_chunk_size;
253
254                 progress.integrity.completed_chunks++;
255                 progress.integrity.completed_bytes += this_chunk_size;
256                 ret = call_progress(progfunc, WIMLIB_PROGRESS_MSG_CALC_INTEGRITY,
257                                     &progress, progctx);
258                 if (ret)
259                         goto out_free_new_table;
260         }
261         *integrity_table_ret = new_table;
262         return 0;
263
264 out_free_new_table:
265         FREE(new_table);
266         return ret;
267 }
268
269 /*
270  * write_integrity_table():
271  *
272  * Writes a WIM integrity table (a list of SHA1 message digests of raw 10 MiB
273  * chunks of the file).
274  *
275  * This function can optionally re-use entries from an older integrity table.
276  * To do this, specify old_blob_table_end and old_table.
277  *
278  * On success, @wim->out_hdr.integrity_table_reshdr will be filled in with
279  * information about the integrity table that was written.
280  *
281  * @wim:
282  *      WIMStruct for the WIM file.  @wim->out_fd must be a seekable descriptor
283  *      to the new WIM file, opened read-write, positioned at the location at
284  *      which the integrity table is to be written.
285  *
286  * @new_blob_table_end:
287  *      The offset of the byte directly following the blob table in the WIM
288  *      being written.
289  *
290  * @old_blob_table_end:
291  *      If nonzero, the offset of the byte directly following the old blob table
292  *      in the WIM.
293  *
294  * @old_table
295  *      Pointer to the old integrity table read into memory, or NULL if not
296  *      specified.
297  */
298 int
299 write_integrity_table(WIMStruct *wim,
300                       off_t new_blob_table_end,
301                       off_t old_blob_table_end,
302                       struct integrity_table *old_table)
303 {
304         struct integrity_table *new_table;
305         int ret;
306         u32 new_table_size;
307
308         wimlib_assert(old_blob_table_end <= new_blob_table_end);
309
310         ret = calculate_integrity_table(&wim->out_fd, new_blob_table_end,
311                                         old_table, old_blob_table_end,
312                                         &new_table, wim->progfunc, wim->progctx);
313         if (ret)
314                 return ret;
315
316         new_table_size = new_table->size;
317
318         new_table->size        = cpu_to_le32(new_table->size);
319         new_table->num_entries = cpu_to_le32(new_table->num_entries);
320         new_table->chunk_size  = cpu_to_le32(new_table->chunk_size);
321
322         ret = write_wim_resource_from_buffer(new_table,
323                                              new_table_size,
324                                              false,
325                                              &wim->out_fd,
326                                              WIMLIB_COMPRESSION_TYPE_NONE,
327                                              0,
328                                              &wim->out_hdr.integrity_table_reshdr,
329                                              NULL,
330                                              0);
331         FREE(new_table);
332         return ret;
333 }
334
335 /*
336  * verify_integrity():
337  *
338  * Checks a WIM for consistency with the integrity table.
339  *
340  * @in_fd:
341  *      File descriptor to the WIM file, opened for reading.
342  *
343  * @table:
344  *      The integrity table for the WIM, read into memory.
345  *
346  * @bytes_to_check:
347  *      Number of bytes in the WIM that need to be checked (offset of end of the
348  *      blob table minus offset of end of the header).
349  *
350  * Returns:
351  *      > 0 (WIMLIB_ERR_READ, WIMLIB_ERR_UNEXPECTED_END_OF_FILE) on error
352  *      0 (WIM_INTEGRITY_OK) if the integrity was checked successfully and there
353  *      were no inconsistencies.
354  *      -1 (WIM_INTEGRITY_NOT_OK) if the WIM failed the integrity check.
355  */
356 static int
357 verify_integrity(struct filedes *in_fd, const tchar *filename,
358                  const struct integrity_table *table,
359                  u64 bytes_to_check,
360                  wimlib_progress_func_t progfunc, void *progctx)
361 {
362         int ret;
363         u64 offset = WIM_HEADER_DISK_SIZE;
364         u8 sha1_md[SHA1_HASH_SIZE];
365         union wimlib_progress_info progress;
366
367         progress.integrity.total_bytes      = bytes_to_check;
368         progress.integrity.total_chunks     = table->num_entries;
369         progress.integrity.completed_chunks = 0;
370         progress.integrity.completed_bytes  = 0;
371         progress.integrity.chunk_size       = table->chunk_size;
372         progress.integrity.filename         = filename;
373
374         ret = call_progress(progfunc, WIMLIB_PROGRESS_MSG_VERIFY_INTEGRITY,
375                             &progress, progctx);
376         if (ret)
377                 return ret;
378
379         for (u32 i = 0; i < table->num_entries; i++) {
380                 size_t this_chunk_size;
381                 if (i == table->num_entries - 1)
382                         this_chunk_size = MODULO_NONZERO(bytes_to_check,
383                                                          table->chunk_size);
384                 else
385                         this_chunk_size = table->chunk_size;
386
387                 ret = calculate_chunk_sha1(in_fd, this_chunk_size, offset, sha1_md);
388                 if (ret)
389                         return ret;
390
391                 if (!hashes_equal(sha1_md, table->sha1sums[i]))
392                         return WIM_INTEGRITY_NOT_OK;
393
394                 offset += this_chunk_size;
395                 progress.integrity.completed_chunks++;
396                 progress.integrity.completed_bytes += this_chunk_size;
397
398                 ret = call_progress(progfunc, WIMLIB_PROGRESS_MSG_VERIFY_INTEGRITY,
399                                     &progress, progctx);
400                 if (ret)
401                         return ret;
402         }
403         return WIM_INTEGRITY_OK;
404 }
405
406
407 /*
408  * check_wim_integrity():
409  *
410  * Verifies the integrity of the WIM by making sure the SHA1 message digests of
411  * ~10 MiB chunks of the WIM match up with the values given in the integrity
412  * table.
413  *
414  * @wim:
415  *      The WIM, opened for reading.
416  *
417  * Returns:
418  *      > 0 (WIMLIB_ERR_INVALID_INTEGRITY_TABLE, WIMLIB_ERR_READ,
419  *           WIMLIB_ERR_UNEXPECTED_END_OF_FILE) on error
420  *      0 (WIM_INTEGRITY_OK) if the integrity was checked successfully and there
421  *      were no inconsistencies.
422  *      -1 (WIM_INTEGRITY_NOT_OK) if the WIM failed the integrity check.
423  *      -2 (WIM_INTEGRITY_NONEXISTENT) if the WIM contains no integrity
424  *      information.
425  */
426 int
427 check_wim_integrity(WIMStruct *wim)
428 {
429         int ret;
430         u64 bytes_to_check;
431         struct integrity_table *table;
432         u64 end_blob_table_offset;
433
434         if (!wim_has_integrity_table(wim))
435                 return WIM_INTEGRITY_NONEXISTENT;
436
437         end_blob_table_offset = wim->hdr.blob_table_reshdr.offset_in_wim +
438                                 wim->hdr.blob_table_reshdr.size_in_wim;
439
440         if (end_blob_table_offset < WIM_HEADER_DISK_SIZE) {
441                 ERROR("WIM blob table ends before WIM header ends!");
442                 return WIMLIB_ERR_INVALID_INTEGRITY_TABLE;
443         }
444
445         bytes_to_check = end_blob_table_offset - WIM_HEADER_DISK_SIZE;
446
447         ret = read_integrity_table(wim, bytes_to_check, &table);
448         if (ret)
449                 return ret;
450         ret = verify_integrity(&wim->in_fd, wim->filename, table,
451                                bytes_to_check, wim->progfunc, wim->progctx);
452         FREE(table);
453         return ret;
454 }