]>
Commit | Line | Data |
---|---|---|
61983f67 | 1 | /* |
56281ed4 | 2 | * Copyright (c) 2007, 2011 SGI |
61983f67 BN |
3 | * All Rights Reserved. |
4 | * | |
5 | * This program is free software; you can redistribute it and/or | |
6 | * modify it under the terms of the GNU General Public License as | |
7 | * published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope that it would be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
17 | */ | |
18 | ||
19 | #include <libxfs.h> | |
20 | #include "bmap.h" | |
21 | #include "command.h" | |
22 | #include "metadump.h" | |
23 | #include "io.h" | |
24 | #include "output.h" | |
25 | #include "type.h" | |
26 | #include "init.h" | |
27 | #include "sig.h" | |
28 | #include "xfs_metadump.h" | |
a2ceac1f DC |
29 | #include "fprint.h" |
30 | #include "faddr.h" | |
31 | #include "field.h" | |
32 | #include "dir2.h" | |
61983f67 | 33 | |
7431d134 BN |
34 | #define DEFAULT_MAX_EXT_SIZE 1000 |
35 | ||
88b1fe2a AE |
36 | /* |
37 | * It's possible that multiple files in a directory (or attributes | |
38 | * in a file) produce the same obfuscated name. If that happens, we | |
39 | * try to create another one. After several rounds of this though, | |
40 | * we just give up and leave the original name as-is. | |
41 | */ | |
42 | #define DUP_MAX 5 /* Max duplicates before we give up */ | |
43 | ||
61983f67 BN |
44 | /* copy all metadata structures to/from a file */ |
45 | ||
46 | static int metadump_f(int argc, char **argv); | |
47 | static void metadump_help(void); | |
48 | ||
49 | /* | |
50 | * metadump commands issue info/wornings/errors to standard error as | |
51 | * metadump supports stdout as a destination. | |
52 | * | |
53 | * All static functions return zero on failure, while the public functions | |
54 | * return zero on success. | |
55 | */ | |
56 | ||
57 | static const cmdinfo_t metadump_cmd = | |
58 | { "metadump", NULL, metadump_f, 0, -1, 0, | |
9ee7055c AM |
59 | N_("[-e] [-g] [-m max_extent] [-w] [-o] filename"), |
60 | N_("dump metadata to a file"), metadump_help }; | |
61983f67 BN |
61 | |
62 | static FILE *outf; /* metadump file */ | |
63 | ||
64 | static xfs_metablock_t *metablock; /* header + index + buffers */ | |
65 | static __be64 *block_index; | |
66 | static char *block_buffer; | |
67 | ||
68 | static int num_indicies; | |
69 | static int cur_index; | |
70 | ||
71 | static xfs_ino_t cur_ino; | |
72 | ||
73 | static int show_progress = 0; | |
74 | static int stop_on_read_error = 0; | |
7431d134 | 75 | static int max_extent_size = DEFAULT_MAX_EXT_SIZE; |
61983f67 BN |
76 | static int dont_obfuscate = 0; |
77 | static int show_warnings = 0; | |
78 | static int progress_since_warning = 0; | |
79 | ||
80 | void | |
81 | metadump_init(void) | |
82 | { | |
83 | add_command(&metadump_cmd); | |
84 | } | |
85 | ||
86 | static void | |
87 | metadump_help(void) | |
88 | { | |
9ee7055c | 89 | dbprintf(_( |
61983f67 BN |
90 | "\n" |
91 | " The 'metadump' command dumps the known metadata to a compact file suitable\n" | |
92 | " for compressing and sending to an XFS maintainer for corruption analysis \n" | |
93 | " or xfs_repair failures.\n\n" | |
88b8e1d6 | 94 | " Options:\n" |
61983f67 BN |
95 | " -e -- Ignore read errors and keep going\n" |
96 | " -g -- Display dump progress\n" | |
7431d134 | 97 | " -m -- Specify max extent size in blocks to copy (default = %d blocks)\n" |
61983f67 BN |
98 | " -o -- Don't obfuscate names and extended attributes\n" |
99 | " -w -- Show warnings of bad metadata information\n" | |
9ee7055c | 100 | "\n"), DEFAULT_MAX_EXT_SIZE); |
61983f67 BN |
101 | } |
102 | ||
103 | static void | |
104 | print_warning(const char *fmt, ...) | |
105 | { | |
106 | char buf[200]; | |
107 | va_list ap; | |
108 | ||
109 | if (seenint()) | |
110 | return; | |
111 | ||
112 | va_start(ap, fmt); | |
113 | vsnprintf(buf, sizeof(buf), fmt, ap); | |
114 | va_end(ap); | |
115 | buf[sizeof(buf)-1] = '\0'; | |
116 | ||
117 | fprintf(stderr, "%s%s: %s\n", progress_since_warning ? "\n" : "", | |
118 | progname, buf); | |
119 | progress_since_warning = 0; | |
120 | } | |
121 | ||
122 | static void | |
123 | print_progress(const char *fmt, ...) | |
124 | { | |
125 | char buf[60]; | |
126 | va_list ap; | |
127 | FILE *f; | |
128 | ||
129 | if (seenint()) | |
130 | return; | |
131 | ||
132 | va_start(ap, fmt); | |
133 | vsnprintf(buf, sizeof(buf), fmt, ap); | |
134 | va_end(ap); | |
135 | buf[sizeof(buf)-1] = '\0'; | |
136 | ||
137 | f = (outf == stdout) ? stderr : stdout; | |
138 | fprintf(f, "\r%-59s", buf); | |
139 | fflush(f); | |
140 | progress_since_warning = 1; | |
141 | } | |
142 | ||
143 | /* | |
144 | * A complete dump file will have a "zero" entry in the last index block, | |
145 | * even if the dump is exactly aligned, the last index will be full of | |
146 | * zeros. If the last index entry is non-zero, the dump is incomplete. | |
147 | * Correspondingly, the last chunk will have a count < num_indicies. | |
148 | */ | |
149 | ||
150 | static int | |
151 | write_index(void) | |
152 | { | |
153 | /* | |
154 | * write index block and following data blocks (streaming) | |
155 | */ | |
156 | metablock->mb_count = cpu_to_be16(cur_index); | |
157 | if (fwrite(metablock, (cur_index + 1) << BBSHIFT, 1, outf) != 1) { | |
158 | print_warning("error writing to file: %s", strerror(errno)); | |
159 | return 0; | |
160 | } | |
161 | ||
162 | memset(block_index, 0, num_indicies * sizeof(__be64)); | |
163 | cur_index = 0; | |
164 | return 1; | |
165 | } | |
166 | ||
167 | static int | |
168 | write_buf( | |
169 | iocur_t *buf) | |
170 | { | |
171 | char *data; | |
172 | __int64_t off; | |
173 | int i; | |
174 | ||
175 | for (i = 0, off = buf->bb, data = buf->data; | |
176 | i < buf->blen; | |
177 | i++, off++, data += BBSIZE) { | |
178 | block_index[cur_index] = cpu_to_be64(off); | |
179 | memcpy(&block_buffer[cur_index << BBSHIFT], data, BBSIZE); | |
180 | if (++cur_index == num_indicies) { | |
181 | if (!write_index()) | |
182 | return 0; | |
183 | } | |
184 | } | |
185 | return !seenint(); | |
186 | } | |
187 | ||
188 | ||
189 | static int | |
190 | scan_btree( | |
191 | xfs_agnumber_t agno, | |
192 | xfs_agblock_t agbno, | |
193 | int level, | |
194 | typnm_t btype, | |
195 | void *arg, | |
b194c7d8 | 196 | int (*func)(struct xfs_btree_block *block, |
61983f67 BN |
197 | xfs_agnumber_t agno, |
198 | xfs_agblock_t agbno, | |
199 | int level, | |
200 | typnm_t btype, | |
201 | void *arg)) | |
202 | { | |
d24c0a90 BN |
203 | int rval = 0; |
204 | ||
61983f67 BN |
205 | push_cur(); |
206 | set_cur(&typtab[btype], XFS_AGB_TO_DADDR(mp, agno, agbno), blkbb, | |
207 | DB_RING_IGN, NULL); | |
208 | if (iocur_top->data == NULL) { | |
209 | print_warning("cannot read %s block %u/%u", typtab[btype].name, | |
210 | agno, agbno); | |
d24c0a90 BN |
211 | rval = !stop_on_read_error; |
212 | goto pop_out; | |
61983f67 BN |
213 | } |
214 | if (!write_buf(iocur_top)) | |
d24c0a90 | 215 | goto pop_out; |
61983f67 BN |
216 | |
217 | if (!(*func)(iocur_top->data, agno, agbno, level - 1, btype, arg)) | |
d24c0a90 BN |
218 | goto pop_out; |
219 | rval = 1; | |
220 | pop_out: | |
61983f67 | 221 | pop_cur(); |
d24c0a90 | 222 | return rval; |
61983f67 BN |
223 | } |
224 | ||
225 | /* free space tree copy routines */ | |
226 | ||
227 | static int | |
228 | valid_bno( | |
61983f67 | 229 | xfs_agnumber_t agno, |
88b8e1d6 | 230 | xfs_agblock_t agbno) |
61983f67 | 231 | { |
88b8e1d6 BN |
232 | if (agno < (mp->m_sb.sb_agcount - 1) && agbno > 0 && |
233 | agbno <= mp->m_sb.sb_agblocks) | |
234 | return 1; | |
235 | if (agno == (mp->m_sb.sb_agcount - 1) && agbno > 0 && | |
236 | agbno <= (mp->m_sb.sb_dblocks - | |
66be354e ES |
237 | (xfs_drfsbno_t)(mp->m_sb.sb_agcount - 1) * |
238 | mp->m_sb.sb_agblocks)) | |
61983f67 BN |
239 | return 1; |
240 | ||
61983f67 BN |
241 | return 0; |
242 | } | |
243 | ||
88b8e1d6 | 244 | |
61983f67 BN |
245 | static int |
246 | scanfunc_freesp( | |
b194c7d8 | 247 | struct xfs_btree_block *block, |
61983f67 BN |
248 | xfs_agnumber_t agno, |
249 | xfs_agblock_t agbno, | |
250 | int level, | |
251 | typnm_t btype, | |
252 | void *arg) | |
253 | { | |
254 | xfs_alloc_ptr_t *pp; | |
255 | int i; | |
88b8e1d6 | 256 | int numrecs; |
61983f67 BN |
257 | |
258 | if (level == 0) | |
259 | return 1; | |
260 | ||
b194c7d8 | 261 | numrecs = be16_to_cpu(block->bb_numrecs); |
88b8e1d6 | 262 | if (numrecs > mp->m_alloc_mxr[1]) { |
61983f67 | 263 | if (show_warnings) |
88b8e1d6 BN |
264 | print_warning("invalid numrecs (%u) in %s block %u/%u", |
265 | numrecs, typtab[btype].name, agno, agbno); | |
61983f67 BN |
266 | return 1; |
267 | } | |
268 | ||
b3563c19 | 269 | pp = XFS_ALLOC_PTR_ADDR(mp, block, 1, mp->m_alloc_mxr[1]); |
88b8e1d6 BN |
270 | for (i = 0; i < numrecs; i++) { |
271 | if (!valid_bno(agno, be32_to_cpu(pp[i]))) { | |
272 | if (show_warnings) | |
273 | print_warning("invalid block number (%u/%u) " | |
274 | "in %s block %u/%u", | |
275 | agno, be32_to_cpu(pp[i]), | |
276 | typtab[btype].name, agno, agbno); | |
61983f67 | 277 | continue; |
88b8e1d6 | 278 | } |
61983f67 BN |
279 | if (!scan_btree(agno, be32_to_cpu(pp[i]), level, btype, arg, |
280 | scanfunc_freesp)) | |
281 | return 0; | |
282 | } | |
283 | return 1; | |
284 | } | |
285 | ||
286 | static int | |
287 | copy_free_bno_btree( | |
288 | xfs_agnumber_t agno, | |
289 | xfs_agf_t *agf) | |
290 | { | |
291 | xfs_agblock_t root; | |
292 | int levels; | |
293 | ||
294 | root = be32_to_cpu(agf->agf_roots[XFS_BTNUM_BNO]); | |
295 | levels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_BNO]); | |
296 | ||
297 | /* validate root and levels before processing the tree */ | |
298 | if (root == 0 || root > mp->m_sb.sb_agblocks) { | |
299 | if (show_warnings) | |
300 | print_warning("invalid block number (%u) in bnobt " | |
301 | "root in agf %u", root, agno); | |
302 | return 1; | |
303 | } | |
304 | if (levels >= XFS_BTREE_MAXLEVELS) { | |
305 | if (show_warnings) | |
306 | print_warning("invalid level (%u) in bnobt root " | |
307 | "in agf %u", levels, agno); | |
308 | return 1; | |
309 | } | |
310 | ||
311 | return scan_btree(agno, root, levels, TYP_BNOBT, agf, scanfunc_freesp); | |
312 | } | |
313 | ||
314 | static int | |
315 | copy_free_cnt_btree( | |
316 | xfs_agnumber_t agno, | |
317 | xfs_agf_t *agf) | |
318 | { | |
319 | xfs_agblock_t root; | |
320 | int levels; | |
321 | ||
322 | root = be32_to_cpu(agf->agf_roots[XFS_BTNUM_CNT]); | |
323 | levels = be32_to_cpu(agf->agf_levels[XFS_BTNUM_CNT]); | |
324 | ||
325 | /* validate root and levels before processing the tree */ | |
326 | if (root == 0 || root > mp->m_sb.sb_agblocks) { | |
327 | if (show_warnings) | |
328 | print_warning("invalid block number (%u) in cntbt " | |
329 | "root in agf %u", root, agno); | |
330 | return 1; | |
331 | } | |
332 | if (levels >= XFS_BTREE_MAXLEVELS) { | |
333 | if (show_warnings) | |
334 | print_warning("invalid level (%u) in cntbt root " | |
335 | "in agf %u", levels, agno); | |
336 | return 1; | |
337 | } | |
338 | ||
339 | return scan_btree(agno, root, levels, TYP_CNTBT, agf, scanfunc_freesp); | |
340 | } | |
341 | ||
342 | /* filename and extended attribute obfuscation routines */ | |
343 | ||
78027d48 | 344 | struct name_ent { |
61983f67 BN |
345 | struct name_ent *next; |
346 | xfs_dahash_t hash; | |
78027d48 AE |
347 | int namelen; |
348 | uchar_t name[1]; | |
349 | }; | |
61983f67 BN |
350 | |
351 | #define NAME_TABLE_SIZE 4096 | |
352 | ||
a85f8b0a | 353 | static struct name_ent *nametable[NAME_TABLE_SIZE]; |
61983f67 BN |
354 | |
355 | static void | |
a85f8b0a | 356 | nametable_clear(void) |
61983f67 | 357 | { |
a85f8b0a | 358 | int i; |
78027d48 | 359 | struct name_ent *ent; |
61983f67 BN |
360 | |
361 | for (i = 0; i < NAME_TABLE_SIZE; i++) { | |
a85f8b0a AE |
362 | while ((ent = nametable[i])) { |
363 | nametable[i] = ent->next; | |
364 | free(ent); | |
61983f67 BN |
365 | } |
366 | } | |
367 | } | |
368 | ||
a85f8b0a AE |
369 | /* |
370 | * See if the given name is already in the name table. If so, | |
371 | * return a pointer to its entry, otherwise return a null pointer. | |
372 | */ | |
373 | static struct name_ent * | |
374 | nametable_find(xfs_dahash_t hash, int namelen, uchar_t *name) | |
375 | { | |
376 | struct name_ent *ent; | |
377 | ||
378 | for (ent = nametable[hash % NAME_TABLE_SIZE]; ent; ent = ent->next) { | |
379 | if (ent->hash == hash && ent->namelen == namelen && | |
380 | !memcmp(ent->name, name, namelen)) | |
381 | return ent; | |
382 | } | |
383 | return NULL; | |
384 | } | |
385 | ||
386 | /* | |
387 | * Add the given name to the name table. Returns a pointer to the | |
388 | * name's new entry, or a null pointer if an error occurs. | |
389 | */ | |
390 | static struct name_ent * | |
391 | nametable_add(xfs_dahash_t hash, int namelen, uchar_t *name) | |
392 | { | |
393 | struct name_ent *ent; | |
394 | ||
395 | ent = malloc(sizeof *ent + namelen); | |
396 | if (!ent) | |
397 | return NULL; | |
398 | ||
399 | ent->namelen = namelen; | |
400 | memcpy(ent->name, name, namelen); | |
401 | ent->hash = hash; | |
402 | ent->next = nametable[hash % NAME_TABLE_SIZE]; | |
403 | ||
404 | nametable[hash % NAME_TABLE_SIZE] = ent; | |
405 | ||
406 | return ent; | |
407 | } | |
61983f67 BN |
408 | |
409 | #define is_invalid_char(c) ((c) == '/' || (c) == '\0') | |
410 | #define rol32(x,y) (((x) << (y)) | ((x) >> (32 - (y)))) | |
411 | ||
412 | static inline uchar_t | |
413 | random_filename_char(void) | |
414 | { | |
7c3a9916 AE |
415 | static uchar_t filename_alphabet[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZ" |
416 | "abcdefghijklmnopqrstuvwxyz" | |
417 | "0123456789-_"; | |
61983f67 | 418 | |
7c3a9916 | 419 | return filename_alphabet[random() % (sizeof filename_alphabet - 1)]; |
61983f67 BN |
420 | } |
421 | ||
56281ed4 AE |
422 | #define ORPHANAGE "lost+found" |
423 | #define ORPHANAGE_LEN (sizeof (ORPHANAGE) - 1) | |
424 | ||
425 | static inline int | |
426 | is_orphanage_dir( | |
427 | struct xfs_mount *mp, | |
428 | xfs_ino_t dir_ino, | |
429 | size_t name_len, | |
430 | uchar_t *name) | |
431 | { | |
432 | return dir_ino == mp->m_sb.sb_rootino && | |
433 | name_len == ORPHANAGE_LEN && | |
434 | !memcmp(name, ORPHANAGE, ORPHANAGE_LEN); | |
435 | } | |
436 | ||
437 | /* | |
438 | * Determine whether a name is one we shouldn't obfuscate because | |
439 | * it's an orphan (or the "lost+found" directory itself). Note | |
440 | * "cur_ino" is the inode for the directory currently being | |
441 | * processed. | |
442 | * | |
443 | * Returns 1 if the name should NOT be obfuscated or 0 otherwise. | |
444 | */ | |
61983f67 | 445 | static int |
56281ed4 | 446 | in_lost_found( |
61983f67 BN |
447 | xfs_ino_t ino, |
448 | int namelen, | |
449 | uchar_t *name) | |
450 | { | |
451 | static xfs_ino_t orphanage_ino = 0; | |
56281ed4 | 452 | char s[24]; /* 21 is enough (64 bits in decimal) */ |
61983f67 BN |
453 | int slen; |
454 | ||
56281ed4 AE |
455 | /* Record the "lost+found" inode if we haven't done so already */ |
456 | ||
457 | ASSERT(ino != 0); | |
458 | if (!orphanage_ino && is_orphanage_dir(mp, cur_ino, namelen, name)) | |
459 | orphanage_ino = ino; | |
460 | ||
461 | /* We don't obfuscate the "lost+found" directory itself */ | |
462 | ||
463 | if (ino == orphanage_ino) | |
61983f67 BN |
464 | return 1; |
465 | ||
56281ed4 AE |
466 | /* Most files aren't in "lost+found" at all */ |
467 | ||
468 | if (cur_ino != orphanage_ino) | |
61983f67 BN |
469 | return 0; |
470 | ||
471 | /* | |
56281ed4 AE |
472 | * Within "lost+found", we don't obfuscate any file whose |
473 | * name is the same as its inode number. Any others are | |
474 | * stray files and can be obfuscated. | |
61983f67 | 475 | */ |
56281ed4 | 476 | slen = snprintf(s, sizeof (s), "%llu", (unsigned long long) ino); |
61983f67 | 477 | |
56281ed4 | 478 | return slen == namelen && !memcmp(name, s, namelen); |
61983f67 BN |
479 | } |
480 | ||
da7daaf2 AE |
481 | /* |
482 | * Given a name and its hash value, massage the name in such a way | |
483 | * that the result is another name of equal length which shares the | |
484 | * same hash value. | |
485 | */ | |
61983f67 | 486 | static void |
da7daaf2 AE |
487 | obfuscate_name( |
488 | xfs_dahash_t hash, | |
489 | size_t name_len, | |
490 | uchar_t *name) | |
61983f67 | 491 | { |
002c6e02 | 492 | uchar_t *newp = name; |
da7daaf2 AE |
493 | int i; |
494 | xfs_dahash_t new_hash = 0; | |
495 | uchar_t *first; | |
496 | uchar_t high_bit; | |
497 | int shift; | |
61983f67 | 498 | |
56281ed4 AE |
499 | /* |
500 | * Our obfuscation algorithm requires at least 5-character | |
501 | * names, so don't bother if the name is too short. We | |
502 | * work backward from a hash value to determine the last | |
503 | * five bytes in a name required to produce a new name | |
504 | * with the same hash. | |
505 | */ | |
da7daaf2 | 506 | if (name_len < 5) |
61983f67 BN |
507 | return; |
508 | ||
da7daaf2 AE |
509 | /* |
510 | * The beginning of the obfuscated name can be pretty much | |
511 | * anything, so fill it in with random characters. | |
512 | * Accumulate its new hash value as we go. | |
513 | */ | |
514 | for (i = 0; i < name_len - 5; i++) { | |
515 | *newp = random_filename_char(); | |
516 | new_hash = *newp ^ rol32(new_hash, 7); | |
517 | newp++; | |
518 | } | |
519 | ||
520 | /* | |
521 | * Compute which five bytes need to be used at the end of | |
522 | * the name so the hash of the obfuscated name is the same | |
523 | * as the hash of the original. If any result in an invalid | |
524 | * character, flip a bit and arrange for a corresponding bit | |
525 | * in a neighboring byte to be flipped as well. For the | |
526 | * last byte, the "neighbor" to change is the first byte | |
527 | * we're computing here. | |
528 | */ | |
529 | new_hash = rol32(new_hash, 3) ^ hash; | |
530 | ||
531 | first = newp; | |
532 | high_bit = 0; | |
533 | for (shift = 28; shift >= 0; shift -= 7) { | |
534 | *newp = (new_hash >> shift & 0x7f) ^ high_bit; | |
535 | if (is_invalid_char(*newp)) { | |
536 | *newp ^= 1; | |
537 | high_bit = 0x80; | |
538 | } else | |
539 | high_bit = 0; | |
540 | ASSERT(!is_invalid_char(*newp)); | |
541 | newp++; | |
542 | } | |
543 | ||
544 | /* | |
545 | * If we flipped a bit on the last byte, we need to fix up | |
546 | * the matching bit in the first byte. The result will | |
547 | * be a valid character, because we know that first byte | |
548 | * has 0's in its upper four bits (it was produced by a | |
549 | * 28-bit right-shift of a 32-bit unsigned value). | |
550 | */ | |
551 | if (high_bit) { | |
552 | *first ^= 0x10; | |
553 | ASSERT(!is_invalid_char(*first)); | |
554 | } | |
002c6e02 | 555 | ASSERT(libxfs_da_hashname(name, name_len) == hash); |
da7daaf2 AE |
556 | } |
557 | ||
1167ddc4 AE |
558 | /* |
559 | * Flip a bit in each of two bytes at the end of the given name. | |
560 | * This is used in generating a series of alternate names to be used | |
561 | * in the event a duplicate is found. | |
562 | * | |
563 | * The bits flipped are selected such that they both affect the same | |
564 | * bit in the name's computed hash value, so flipping them both will | |
565 | * preserve the hash. | |
566 | * | |
567 | * The following diagram aims to show the portion of a computed | |
568 | * hash that a given byte of a name affects. | |
569 | * | |
570 | * 31 28 24 21 14 8 7 3 0 | |
571 | * +-+-+-+-+-+-+-+-|-+-+-+-+-+-+-+-|-+-+-+-+-+-+-+-|-+-+-+-+-+-+-+-+ | |
572 | * hash: | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
573 | * +-+-+-+-+-+-+-+-|-+-+-+-+-+-+-+-|-+-+-+-+-+-+-+-|-+-+-+-+-+-+-+-+ | |
574 | * last-4 ->| |<-- last-2 --->| |<--- last ---->| | |
575 | * |<-- last-3 --->| |<-- last-1 --->| |<- last-4 | |
576 | * |<-- last-7 --->| |<-- last-5 --->| | |
577 | * |<-- last-8 --->| |<-- last-6 --->| | |
578 | * . . . and so on | |
579 | * | |
580 | * The last byte of the name directly affects the low-order byte of | |
581 | * the hash. The next-to-last affects bits 7-14, the next one back | |
582 | * affects bits 14-21, and so on. The effect wraps around when it | |
583 | * goes beyond the top of the hash (as happens for byte last-4). | |
584 | * | |
585 | * Bits that are flipped together "overlap" on the hash value. As | |
586 | * an example of overlap, the last two bytes both affect bit 7 in | |
587 | * the hash. That pair of bytes (and their overlapping bits) can be | |
588 | * used for this "flip bit" operation (it's the first pair tried, | |
589 | * actually). | |
590 | * | |
591 | * A table defines overlapping pairs--the bytes involved and bits | |
592 | * within them--that can be used this way. The byte offset is | |
593 | * relative to a starting point within the name, which will be set | |
594 | * to affect the bytes at the end of the name. The function is | |
595 | * called with a "bitseq" value which indicates which bit flip is | |
596 | * desired, and this translates directly into selecting which entry | |
597 | * in the bit_to_flip[] table to apply. | |
598 | * | |
599 | * The function returns 1 if the operation was successful. It | |
600 | * returns 0 if the result produced a character that's not valid in | |
601 | * a name (either '/' or a '\0'). Finally, it returns -1 if the bit | |
602 | * sequence number is beyond what is supported for a name of this | |
603 | * length. | |
604 | * | |
605 | * Discussion | |
606 | * ---------- | |
607 | * (Also see the discussion above find_alternate(), below.) | |
608 | * | |
609 | * In order to make this function work for any length name, the | |
610 | * table is ordered by increasing byte offset, so that the earliest | |
611 | * entries can apply to the shortest strings. This way all names | |
612 | * are done consistently. | |
613 | * | |
614 | * When bit flips occur, they can convert printable characters | |
615 | * into non-printable ones. In an effort to reduce the impact of | |
616 | * this, the first bit flips are chosen to affect bytes the end of | |
617 | * the name (and furthermore, toward the low bits of a byte). Those | |
618 | * bytes are often non-printable anyway because of the way they are | |
619 | * initially selected by obfuscate_name()). This is accomplished, | |
620 | * using later table entries first. | |
621 | * | |
622 | * Each row in the table doubles the number of alternates that | |
623 | * can be generated. A two-byte name is limited to using only | |
624 | * the first row, so it's possible to generate two alternates | |
625 | * (the original name, plus the alternate produced by flipping | |
626 | * the one pair of bits). In a 5-byte name, the effect of the | |
627 | * first byte overlaps the last by 4 its, and there are 8 bits | |
628 | * to flip, allowing for 256 possible alternates. | |
629 | * | |
630 | * Short names (less than 5 bytes) are never even obfuscated, so for | |
631 | * such names the relatively small number of alternates should never | |
632 | * really be a problem. | |
633 | * | |
634 | * Long names (more than 6 bytes, say) are not likely to exhaust | |
635 | * the number of available alternates. In fact, the table could | |
636 | * probably have stopped at 8 entries, on the assumption that 256 | |
637 | * alternates should be enough for most any situation. The entries | |
638 | * beyond those are present mostly for demonstration of how it could | |
639 | * be populated with more entries, should it ever be necessary to do | |
640 | * so. | |
641 | */ | |
642 | static int | |
643 | flip_bit( | |
644 | size_t name_len, | |
645 | uchar_t *name, | |
646 | uint32_t bitseq) | |
647 | { | |
648 | int index; | |
649 | size_t offset; | |
650 | uchar_t *p0, *p1; | |
651 | uchar_t m0, m1; | |
652 | struct { | |
653 | int byte; /* Offset from start within name */ | |
654 | uchar_t bit; /* Bit within that byte */ | |
655 | } bit_to_flip[][2] = { /* Sorted by second entry's byte */ | |
656 | { { 0, 0 }, { 1, 7 } }, /* Each row defines a pair */ | |
657 | { { 1, 0 }, { 2, 7 } }, /* of bytes and a bit within */ | |
658 | { { 2, 0 }, { 3, 7 } }, /* each byte. Each bit in */ | |
659 | { { 0, 4 }, { 4, 0 } }, /* a pair affects the same */ | |
660 | { { 0, 5 }, { 4, 1 } }, /* bit in the hash, so flipping */ | |
661 | { { 0, 6 }, { 4, 2 } }, /* both will change the name */ | |
662 | { { 0, 7 }, { 4, 3 } }, /* while preserving the hash. */ | |
663 | { { 3, 0 }, { 4, 7 } }, | |
664 | { { 0, 0 }, { 5, 3 } }, /* The first entry's byte offset */ | |
665 | { { 0, 1 }, { 5, 4 } }, /* must be less than the second. */ | |
666 | { { 0, 2 }, { 5, 5 } }, | |
667 | { { 0, 3 }, { 5, 6 } }, /* The table can be extended to */ | |
668 | { { 0, 4 }, { 5, 7 } }, /* an arbitrary number of entries */ | |
669 | { { 4, 0 }, { 5, 7 } }, /* but there's not much point. */ | |
670 | /* . . . */ | |
671 | }; | |
672 | ||
673 | /* Find the first entry *not* usable for name of this length */ | |
674 | ||
675 | for (index = 0; index < ARRAY_SIZE(bit_to_flip); index++) | |
676 | if (bit_to_flip[index][1].byte >= name_len) | |
677 | break; | |
678 | ||
679 | /* | |
680 | * Back up to the last usable entry. If that number is | |
681 | * smaller than the bit sequence number, inform the caller | |
682 | * that nothing this large (or larger) will work. | |
683 | */ | |
684 | if (bitseq > --index) | |
685 | return -1; | |
686 | ||
687 | /* | |
688 | * We will be switching bits at the end of name, with a | |
689 | * preference for affecting the last bytes first. Compute | |
690 | * where in the name we'll start applying the changes. | |
691 | */ | |
692 | offset = name_len - (bit_to_flip[index][1].byte + 1); | |
693 | index -= bitseq; /* Use later table entries first */ | |
694 | ||
695 | p0 = name + offset + bit_to_flip[index][0].byte; | |
696 | p1 = name + offset + bit_to_flip[index][1].byte; | |
697 | m0 = 1 << bit_to_flip[index][0].bit; | |
698 | m1 = 1 << bit_to_flip[index][1].bit; | |
699 | ||
700 | /* Only change the bytes if it produces valid characters */ | |
701 | ||
702 | if (is_invalid_char(*p0 ^ m0) || is_invalid_char(*p1 ^ m1)) | |
703 | return 0; | |
704 | ||
705 | *p0 ^= m0; | |
706 | *p1 ^= m1; | |
707 | ||
708 | return 1; | |
709 | } | |
710 | ||
711 | /* | |
712 | * This function generates a well-defined sequence of "alternate" | |
713 | * names for a given name. An alternate is a name having the same | |
714 | * length and same hash value as the original name. This is needed | |
715 | * because the algorithm produces only one obfuscated name to use | |
716 | * for a given original name, and it's possible that result matches | |
717 | * a name already seen. This function checks for this, and if it | |
718 | * occurs, finds another suitable obfuscated name to use. | |
719 | * | |
720 | * Each bit in the binary representation of the sequence number is | |
721 | * used to select one possible "bit flip" operation to perform on | |
722 | * the name. So for example: | |
723 | * seq = 0: selects no bits to flip | |
724 | * seq = 1: selects the 0th bit to flip | |
725 | * seq = 2: selects the 1st bit to flip | |
726 | * seq = 3: selects the 0th and 1st bit to flip | |
727 | * ... and so on. | |
728 | * | |
729 | * The flip_bit() function takes care of the details of the bit | |
730 | * flipping within the name. Note that the "1st bit" in this | |
731 | * context is a bit sequence number; i.e. it doesn't necessarily | |
732 | * mean bit 0x02 will be changed. | |
733 | * | |
734 | * If a valid name (one that contains no '/' or '\0' characters) is | |
735 | * produced by this process for the given sequence number, this | |
736 | * function returns 1. If the result is not valid, it returns 0. | |
737 | * Returns -1 if the sequence number is beyond the the maximum for | |
738 | * names of the given length. | |
739 | * | |
740 | * | |
741 | * Discussion | |
742 | * ---------- | |
743 | * The number of alternates available for a given name is dependent | |
744 | * on its length. A "bit flip" involves inverting two bits in | |
745 | * a name--the two bits being selected such that their values | |
746 | * affect the name's hash value in the same way. Alternates are | |
747 | * thus generated by inverting the value of pairs of such | |
748 | * "overlapping" bits in the original name. Each byte after the | |
749 | * first in a name adds at least one bit of overlap to work with. | |
750 | * (See comments above flip_bit() for more discussion on this.) | |
751 | * | |
752 | * So the number of alternates is dependent on the number of such | |
753 | * overlapping bits in a name. If there are N bit overlaps, there | |
754 | * 2^N alternates for that hash value. | |
755 | * | |
756 | * Here are the number of overlapping bits available for generating | |
757 | * alternates for names of specific lengths: | |
758 | * 1 0 (must have 2 bytes to have any overlap) | |
759 | * 2 1 One bit overlaps--so 2 possible alternates | |
760 | * 3 2 Two bits overlap--so 4 possible alternates | |
761 | * 4 4 Three bits overlap, so 2^3 alternates | |
762 | * 5 8 8 bits overlap (due to wrapping), 256 alternates | |
763 | * 6 18 2^18 alternates | |
764 | * 7 28 2^28 alternates | |
765 | * ... | |
766 | * It's clear that the number of alternates grows very quickly with | |
767 | * the length of the name. But note that the set of alternates | |
768 | * includes invalid names. And for certain (contrived) names, the | |
769 | * number of valid names is a fairly small fraction of the total | |
770 | * number of alternates. | |
771 | * | |
772 | * The main driver for this infrastructure for coming up with | |
773 | * alternate names is really related to names 5 (or possibly 6) | |
774 | * bytes in length. 5-byte obfuscated names contain no randomly- | |
775 | * generated bytes in them, and the chance of an obfuscated name | |
776 | * matching an already-seen name is too high to just ignore. This | |
777 | * methodical selection of alternates ensures we don't produce | |
778 | * duplicate names unless we have exhausted our options. | |
779 | */ | |
780 | static int | |
781 | find_alternate( | |
782 | size_t name_len, | |
783 | uchar_t *name, | |
784 | uint32_t seq) | |
785 | { | |
786 | uint32_t bitseq = 0; | |
787 | uint32_t bits = seq; | |
788 | ||
789 | if (!seq) | |
790 | return 1; /* alternate 0 is the original name */ | |
791 | if (name_len < 2) /* Must have 2 bytes to flip */ | |
792 | return -1; | |
793 | ||
794 | for (bitseq = 0; bits; bitseq++) { | |
795 | uint32_t mask = 1 << bitseq; | |
796 | int fb; | |
797 | ||
798 | if (!(bits & mask)) | |
799 | continue; | |
800 | ||
801 | fb = flip_bit(name_len, name, bitseq); | |
802 | if (fb < 1) | |
803 | return fb ? -1 : 0; | |
804 | bits ^= mask; | |
805 | } | |
806 | ||
807 | return 1; | |
808 | } | |
809 | ||
fcb63670 AE |
810 | /* |
811 | * Look up the given name in the name table. If it is already | |
1167ddc4 AE |
812 | * present, iterate through a well-defined sequence of alternate |
813 | * names and attempt to use an alternate name instead. | |
fcb63670 AE |
814 | * |
815 | * Returns 1 if the (possibly modified) name is not present in the | |
1167ddc4 AE |
816 | * name table. Returns 0 if the name and all possible alternates |
817 | * are already in the table. | |
fcb63670 AE |
818 | */ |
819 | static int | |
820 | handle_duplicate_name(xfs_dahash_t hash, size_t name_len, uchar_t *name) | |
821 | { | |
1167ddc4 AE |
822 | uchar_t new_name[name_len + 1]; |
823 | uint32_t seq = 1; | |
fcb63670 AE |
824 | |
825 | if (!nametable_find(hash, name_len, name)) | |
1167ddc4 | 826 | return 1; /* No duplicate */ |
fcb63670 AE |
827 | |
828 | /* Name is already in use. Need to find an alternate. */ | |
829 | ||
830 | do { | |
1167ddc4 | 831 | int found; |
fcb63670 | 832 | |
1167ddc4 AE |
833 | /* Only change incoming name if we find an alternate */ |
834 | do { | |
835 | memcpy(new_name, name, name_len); | |
836 | found = find_alternate(name_len, new_name, seq++); | |
837 | if (found < 0) | |
838 | return 0; /* No more to check */ | |
839 | } while (!found); | |
840 | } while (nametable_find(hash, name_len, new_name)); | |
fcb63670 | 841 | |
1167ddc4 AE |
842 | /* |
843 | * The alternate wasn't in the table already. Pass it back | |
844 | * to the caller. | |
845 | */ | |
846 | memcpy(name, new_name, name_len); | |
847 | ||
848 | return 1; | |
fcb63670 AE |
849 | } |
850 | ||
da7daaf2 AE |
851 | static void |
852 | generate_obfuscated_name( | |
853 | xfs_ino_t ino, | |
854 | int namelen, | |
855 | uchar_t *name) | |
856 | { | |
857 | xfs_dahash_t hash; | |
da7daaf2 | 858 | |
56281ed4 AE |
859 | /* |
860 | * We don't obfuscate "lost+found" or any orphan files | |
861 | * therein. When the name table is used for extended | |
862 | * attributes, the inode number provided is 0, in which | |
863 | * case we don't need to make this check. | |
864 | */ | |
865 | if (ino && in_lost_found(ino, namelen, name)) | |
866 | return; | |
61983f67 | 867 | |
ad6bb839 | 868 | /* |
fcb63670 AE |
869 | * If the name starts with a slash, just skip over it. It |
870 | * isn't included in the hash and we don't record it in the | |
871 | * name table. Note that the namelen value passed in does | |
872 | * not count the leading slash (if one is present). | |
ad6bb839 AE |
873 | */ |
874 | if (*name == '/') | |
875 | name++; | |
61983f67 | 876 | |
fcb63670 | 877 | /* Obfuscate the name (if possible) */ |
61983f67 | 878 | |
fcb63670 AE |
879 | hash = libxfs_da_hashname(name, namelen); |
880 | obfuscate_name(hash, namelen, name); | |
88b1fe2a AE |
881 | |
882 | /* | |
fcb63670 AE |
883 | * Make sure the name is not something already seen. If we |
884 | * fail to find a suitable alternate, we're dealing with a | |
885 | * very pathological situation, and we may end up creating | |
886 | * a duplicate name in the metadump, so issue a warning. | |
88b1fe2a | 887 | */ |
fcb63670 | 888 | if (!handle_duplicate_name(hash, namelen, name)) { |
88b1fe2a AE |
889 | print_warning("duplicate name for inode %llu " |
890 | "in dir inode %llu\n", | |
891 | (unsigned long long) ino, | |
892 | (unsigned long long) cur_ino); | |
fcb63670 AE |
893 | return; |
894 | } | |
895 | ||
896 | /* Create an entry for the new name in the name table. */ | |
61983f67 | 897 | |
a85f8b0a AE |
898 | if (!nametable_add(hash, namelen, name)) |
899 | print_warning("unable to record name for inode %llu " | |
900 | "in dir inode %llu\n", | |
901 | (unsigned long long) ino, | |
902 | (unsigned long long) cur_ino); | |
61983f67 BN |
903 | } |
904 | ||
905 | static void | |
906 | obfuscate_sf_dir( | |
907 | xfs_dinode_t *dip) | |
908 | { | |
eb0cb950 | 909 | struct xfs_dir2_sf_hdr *sfp; |
61983f67 | 910 | xfs_dir2_sf_entry_t *sfep; |
5e656dbb | 911 | __uint64_t ino_dir_size; |
61983f67 BN |
912 | int i; |
913 | ||
eb0cb950 | 914 | sfp = (struct xfs_dir2_sf_hdr *)XFS_DFORK_DPTR(dip); |
56b2de80 | 915 | ino_dir_size = be64_to_cpu(dip->di_size); |
61983f67 BN |
916 | if (ino_dir_size > XFS_DFORK_DSIZE(dip, mp)) { |
917 | ino_dir_size = XFS_DFORK_DSIZE(dip, mp); | |
918 | if (show_warnings) | |
88b8e1d6 | 919 | print_warning("invalid size in dir inode %llu", |
61983f67 BN |
920 | (long long)cur_ino); |
921 | } | |
922 | ||
eb0cb950 DC |
923 | sfep = xfs_dir2_sf_firstentry(sfp); |
924 | for (i = 0; (i < sfp->count) && | |
61983f67 BN |
925 | ((char *)sfep - (char *)sfp < ino_dir_size); i++) { |
926 | ||
927 | /* | |
928 | * first check for bad name lengths. If they are bad, we | |
929 | * have limitations to how much can be obfuscated. | |
930 | */ | |
931 | int namelen = sfep->namelen; | |
932 | ||
933 | if (namelen == 0) { | |
934 | if (show_warnings) | |
935 | print_warning("zero length entry in dir inode " | |
936 | "%llu", (long long)cur_ino); | |
eb0cb950 | 937 | if (i != sfp->count - 1) |
61983f67 BN |
938 | break; |
939 | namelen = ino_dir_size - ((char *)&sfep->name[0] - | |
940 | (char *)sfp); | |
941 | } else if ((char *)sfep - (char *)sfp + | |
494434d7 | 942 | xfs_dir3_sf_entsize(mp, sfp, sfep->namelen) > |
61983f67 BN |
943 | ino_dir_size) { |
944 | if (show_warnings) | |
945 | print_warning("entry length in dir inode %llu " | |
946 | "overflows space", (long long)cur_ino); | |
eb0cb950 | 947 | if (i != sfp->count - 1) |
61983f67 BN |
948 | break; |
949 | namelen = ino_dir_size - ((char *)&sfep->name[0] - | |
950 | (char *)sfp); | |
951 | } | |
952 | ||
494434d7 | 953 | generate_obfuscated_name(xfs_dir3_sfe_get_ino(mp, sfp, sfep), |
a2ceac1f | 954 | namelen, &sfep->name[0]); |
61983f67 BN |
955 | |
956 | sfep = (xfs_dir2_sf_entry_t *)((char *)sfep + | |
494434d7 | 957 | xfs_dir3_sf_entsize(mp, sfp, namelen)); |
61983f67 BN |
958 | } |
959 | } | |
960 | ||
b249a9f0 ES |
961 | static void |
962 | obfuscate_path_components( | |
963 | char *buf, | |
964 | __uint64_t len) | |
965 | { | |
966 | uchar_t *comp; | |
967 | xfs_dahash_t hash; | |
968 | ||
969 | comp = (uchar_t *)buf; | |
970 | while (comp < (uchar_t *)buf + len) { | |
971 | char *slash; | |
972 | int namelen; | |
973 | ||
974 | /* find slash at end of this component */ | |
975 | slash = strchr((char *)comp, '/'); | |
976 | if (!slash) { | |
977 | /* last (or single) component */ | |
978 | namelen = strlen((char *)comp); | |
979 | hash = libxfs_da_hashname(comp, namelen); | |
980 | obfuscate_name(hash, namelen, comp); | |
981 | break; | |
982 | } | |
983 | namelen = slash - (char *)comp; | |
984 | /* handle leading or consecutive slashes */ | |
985 | if (!namelen) { | |
986 | comp++; | |
987 | continue; | |
988 | } | |
989 | hash = libxfs_da_hashname(comp, namelen); | |
990 | obfuscate_name(hash, namelen, comp); | |
991 | comp += namelen + 1; | |
992 | } | |
993 | } | |
994 | ||
61983f67 BN |
995 | static void |
996 | obfuscate_sf_symlink( | |
997 | xfs_dinode_t *dip) | |
998 | { | |
5e656dbb | 999 | __uint64_t len; |
56b2de80 | 1000 | char *buf; |
88b8e1d6 | 1001 | |
56b2de80 | 1002 | len = be64_to_cpu(dip->di_size); |
88b8e1d6 BN |
1003 | if (len > XFS_DFORK_DSIZE(dip, mp)) { |
1004 | if (show_warnings) | |
1005 | print_warning("invalid size (%d) in symlink inode %llu", | |
1006 | len, (long long)cur_ino); | |
1007 | len = XFS_DFORK_DSIZE(dip, mp); | |
1008 | } | |
61983f67 | 1009 | |
56b2de80 | 1010 | buf = (char *)XFS_DFORK_DPTR(dip); |
b249a9f0 | 1011 | obfuscate_path_components(buf, len); |
61983f67 BN |
1012 | } |
1013 | ||
1014 | static void | |
1015 | obfuscate_sf_attr( | |
1016 | xfs_dinode_t *dip) | |
1017 | { | |
1018 | /* | |
1019 | * with extended attributes, obfuscate the names and zero the actual | |
1020 | * values. | |
1021 | */ | |
1022 | ||
1023 | xfs_attr_shortform_t *asfp; | |
1024 | xfs_attr_sf_entry_t *asfep; | |
1025 | int ino_attr_size; | |
1026 | int i; | |
1027 | ||
1028 | asfp = (xfs_attr_shortform_t *)XFS_DFORK_APTR(dip); | |
1029 | if (asfp->hdr.count == 0) | |
1030 | return; | |
1031 | ||
1032 | ino_attr_size = be16_to_cpu(asfp->hdr.totsize); | |
1033 | if (ino_attr_size > XFS_DFORK_ASIZE(dip, mp)) { | |
1034 | ino_attr_size = XFS_DFORK_ASIZE(dip, mp); | |
1035 | if (show_warnings) | |
1036 | print_warning("invalid attr size in inode %llu", | |
1037 | (long long)cur_ino); | |
1038 | } | |
1039 | ||
1040 | asfep = &asfp->list[0]; | |
1041 | for (i = 0; (i < asfp->hdr.count) && | |
1042 | ((char *)asfep - (char *)asfp < ino_attr_size); i++) { | |
1043 | ||
1044 | int namelen = asfep->namelen; | |
1045 | ||
1046 | if (namelen == 0) { | |
1047 | if (show_warnings) | |
1048 | print_warning("zero length attr entry in inode " | |
1049 | "%llu", (long long)cur_ino); | |
1050 | break; | |
1051 | } else if ((char *)asfep - (char *)asfp + | |
1052 | XFS_ATTR_SF_ENTSIZE(asfep) > ino_attr_size) { | |
1053 | if (show_warnings) | |
1054 | print_warning("attr entry length in inode %llu " | |
1055 | "overflows space", (long long)cur_ino); | |
1056 | break; | |
1057 | } | |
1058 | ||
1059 | generate_obfuscated_name(0, asfep->namelen, &asfep->nameval[0]); | |
1060 | memset(&asfep->nameval[asfep->namelen], 0, asfep->valuelen); | |
1061 | ||
1062 | asfep = (xfs_attr_sf_entry_t *)((char *)asfep + | |
1063 | XFS_ATTR_SF_ENTSIZE(asfep)); | |
1064 | } | |
1065 | } | |
1066 | ||
1067 | /* | |
1068 | * dir_data structure is used to track multi-fsblock dir2 blocks between extent | |
1069 | * processing calls. | |
1070 | */ | |
1071 | ||
1072 | static struct dir_data_s { | |
1073 | int end_of_data; | |
1074 | int block_index; | |
1075 | int offset_to_entry; | |
1076 | int bad_block; | |
1077 | } dir_data; | |
1078 | ||
1079 | static void | |
1080 | obfuscate_dir_data_blocks( | |
1081 | char *block, | |
1082 | xfs_dfiloff_t offset, | |
1083 | xfs_dfilblks_t count, | |
1084 | int is_block_format) | |
1085 | { | |
1086 | /* | |
1087 | * we have to rely on the fileoffset and signature of the block to | |
1088 | * handle it's contents. If it's invalid, leave it alone. | |
1089 | * for multi-fsblock dir blocks, if a name crosses an extent boundary, | |
1090 | * ignore it and continue. | |
1091 | */ | |
1092 | int c; | |
1093 | int dir_offset; | |
1094 | char *ptr; | |
1095 | char *endptr; | |
1096 | ||
1097 | if (is_block_format && count != mp->m_dirblkfsbs) | |
1098 | return; /* too complex to handle this rare case */ | |
1099 | ||
1100 | for (c = 0, endptr = block; c < count; c++) { | |
1101 | ||
1102 | if (dir_data.block_index == 0) { | |
1103 | int wantmagic; | |
eb0cb950 DC |
1104 | struct xfs_dir2_data_hdr *datahdr; |
1105 | ||
1106 | datahdr = (struct xfs_dir2_data_hdr *)block; | |
61983f67 BN |
1107 | |
1108 | if (offset % mp->m_dirblkfsbs != 0) | |
1109 | return; /* corrupted, leave it alone */ | |
1110 | ||
1111 | dir_data.bad_block = 0; | |
1112 | ||
1113 | if (is_block_format) { | |
1114 | xfs_dir2_leaf_entry_t *blp; | |
1115 | xfs_dir2_block_tail_t *btp; | |
1116 | ||
eb0cb950 | 1117 | btp = xfs_dir2_block_tail_p(mp, datahdr); |
5e656dbb | 1118 | blp = xfs_dir2_block_leaf_p(btp); |
61983f67 BN |
1119 | if ((char *)blp > (char *)btp) |
1120 | blp = (xfs_dir2_leaf_entry_t *)btp; | |
1121 | ||
1122 | dir_data.end_of_data = (char *)blp - block; | |
1123 | wantmagic = XFS_DIR2_BLOCK_MAGIC; | |
1124 | } else { /* leaf/node format */ | |
1125 | dir_data.end_of_data = mp->m_dirblkfsbs << | |
1126 | mp->m_sb.sb_blocklog; | |
1127 | wantmagic = XFS_DIR2_DATA_MAGIC; | |
1128 | } | |
eb0cb950 DC |
1129 | dir_data.offset_to_entry = |
1130 | xfs_dir3_data_entry_offset(datahdr); | |
61983f67 | 1131 | |
eb0cb950 | 1132 | if (be32_to_cpu(datahdr->magic) != wantmagic) { |
61983f67 BN |
1133 | if (show_warnings) |
1134 | print_warning("invalid magic in dir " | |
1135 | "inode %llu block %ld", | |
1136 | (long long)cur_ino, | |
1137 | (long)offset); | |
1138 | dir_data.bad_block = 1; | |
1139 | } | |
1140 | } | |
1141 | dir_data.block_index++; | |
1142 | if (dir_data.block_index == mp->m_dirblkfsbs) | |
1143 | dir_data.block_index = 0; | |
1144 | ||
1145 | if (dir_data.bad_block) | |
1146 | continue; | |
1147 | ||
1148 | dir_offset = (dir_data.block_index << mp->m_sb.sb_blocklog) + | |
1149 | dir_data.offset_to_entry; | |
1150 | ||
1151 | ptr = endptr + dir_data.offset_to_entry; | |
1152 | endptr += mp->m_sb.sb_blocksize; | |
1153 | ||
1154 | while (ptr < endptr && dir_offset < dir_data.end_of_data) { | |
1155 | xfs_dir2_data_entry_t *dep; | |
1156 | xfs_dir2_data_unused_t *dup; | |
1157 | int length; | |
1158 | ||
1159 | dup = (xfs_dir2_data_unused_t *)ptr; | |
1160 | ||
1161 | if (be16_to_cpu(dup->freetag) == XFS_DIR2_DATA_FREE_TAG) { | |
1162 | int length = be16_to_cpu(dup->length); | |
1163 | if (dir_offset + length > dir_data.end_of_data || | |
1164 | length == 0 || (length & | |
1165 | (XFS_DIR2_DATA_ALIGN - 1))) { | |
1166 | if (show_warnings) | |
1167 | print_warning("invalid length " | |
1168 | "for dir free space in " | |
1169 | "inode %llu", | |
1170 | (long long)cur_ino); | |
1171 | dir_data.bad_block = 1; | |
1172 | break; | |
1173 | } | |
5e656dbb | 1174 | if (be16_to_cpu(*xfs_dir2_data_unused_tag_p(dup)) != |
61983f67 BN |
1175 | dir_offset) { |
1176 | dir_data.bad_block = 1; | |
1177 | break; | |
1178 | } | |
1179 | dir_offset += length; | |
1180 | ptr += length; | |
1181 | if (dir_offset >= dir_data.end_of_data || | |
1182 | ptr >= endptr) | |
1183 | break; | |
1184 | } | |
1185 | ||
1186 | dep = (xfs_dir2_data_entry_t *)ptr; | |
494434d7 | 1187 | length = xfs_dir3_data_entsize(mp, dep->namelen); |
61983f67 BN |
1188 | |
1189 | if (dir_offset + length > dir_data.end_of_data || | |
1190 | ptr + length > endptr) { | |
1191 | if (show_warnings) | |
1192 | print_warning("invalid length for " | |
1193 | "dir entry name in inode %llu", | |
1194 | (long long)cur_ino); | |
1195 | break; | |
1196 | } | |
494434d7 | 1197 | if (be16_to_cpu(*xfs_dir3_data_entry_tag_p(mp, dep)) != |
61983f67 BN |
1198 | dir_offset) { |
1199 | dir_data.bad_block = 1; | |
1200 | break; | |
1201 | } | |
1202 | generate_obfuscated_name(be64_to_cpu(dep->inumber), | |
1203 | dep->namelen, &dep->name[0]); | |
1204 | dir_offset += length; | |
1205 | ptr += length; | |
1206 | } | |
1207 | dir_data.offset_to_entry = dir_offset & | |
1208 | (mp->m_sb.sb_blocksize - 1); | |
1209 | } | |
1210 | } | |
1211 | ||
1212 | static void | |
1213 | obfuscate_symlink_blocks( | |
1214 | char *block, | |
1215 | xfs_dfilblks_t count) | |
1216 | { | |
61983f67 | 1217 | count <<= mp->m_sb.sb_blocklog; |
b249a9f0 | 1218 | obfuscate_path_components(block, count); |
61983f67 BN |
1219 | } |
1220 | ||
1221 | #define MAX_REMOTE_VALS 4095 | |
1222 | ||
1223 | static struct attr_data_s { | |
1224 | int remote_val_count; | |
1225 | xfs_dablk_t remote_vals[MAX_REMOTE_VALS]; | |
1226 | } attr_data; | |
1227 | ||
1228 | static inline void | |
1229 | add_remote_vals( | |
1230 | xfs_dablk_t blockidx, | |
1231 | int length) | |
1232 | { | |
1233 | while (length > 0 && attr_data.remote_val_count < MAX_REMOTE_VALS) { | |
1234 | attr_data.remote_vals[attr_data.remote_val_count] = blockidx; | |
1235 | attr_data.remote_val_count++; | |
1236 | blockidx++; | |
1237 | length -= XFS_LBSIZE(mp); | |
1238 | } | |
1239 | } | |
1240 | ||
1241 | static void | |
1242 | obfuscate_attr_blocks( | |
1243 | char *block, | |
1244 | xfs_dfiloff_t offset, | |
1245 | xfs_dfilblks_t count) | |
1246 | { | |
1247 | xfs_attr_leafblock_t *leaf; | |
1248 | int c; | |
1249 | int i; | |
1250 | int nentries; | |
1251 | xfs_attr_leaf_entry_t *entry; | |
1252 | xfs_attr_leaf_name_local_t *local; | |
1253 | xfs_attr_leaf_name_remote_t *remote; | |
1254 | ||
1255 | for (c = 0; c < count; c++, offset++, block += XFS_LBSIZE(mp)) { | |
1256 | ||
1257 | leaf = (xfs_attr_leafblock_t *)block; | |
1258 | ||
1259 | if (be16_to_cpu(leaf->hdr.info.magic) != XFS_ATTR_LEAF_MAGIC) { | |
1260 | for (i = 0; i < attr_data.remote_val_count; i++) { | |
1261 | if (attr_data.remote_vals[i] == offset) | |
1262 | memset(block, 0, XFS_LBSIZE(mp)); | |
1263 | } | |
1264 | continue; | |
1265 | } | |
1266 | ||
1267 | nentries = be16_to_cpu(leaf->hdr.count); | |
1268 | if (nentries * sizeof(xfs_attr_leaf_entry_t) + | |
1269 | sizeof(xfs_attr_leaf_hdr_t) > XFS_LBSIZE(mp)) { | |
1270 | if (show_warnings) | |
1271 | print_warning("invalid attr count in inode %llu", | |
1272 | (long long)cur_ino); | |
1273 | continue; | |
1274 | } | |
1275 | ||
1276 | for (i = 0, entry = &leaf->entries[0]; i < nentries; | |
1277 | i++, entry++) { | |
1278 | if (be16_to_cpu(entry->nameidx) > XFS_LBSIZE(mp)) { | |
1279 | if (show_warnings) | |
1280 | print_warning("invalid attr nameidx " | |
1281 | "in inode %llu", | |
1282 | (long long)cur_ino); | |
1283 | break; | |
1284 | } | |
1285 | if (entry->flags & XFS_ATTR_LOCAL) { | |
a24374f4 | 1286 | local = xfs_attr3_leaf_name_local(leaf, i); |
61983f67 BN |
1287 | if (local->namelen == 0) { |
1288 | if (show_warnings) | |
1289 | print_warning("zero length for " | |
1290 | "attr name in inode %llu", | |
1291 | (long long)cur_ino); | |
1292 | break; | |
1293 | } | |
1294 | generate_obfuscated_name(0, local->namelen, | |
1295 | &local->nameval[0]); | |
1296 | memset(&local->nameval[local->namelen], 0, | |
1297 | be16_to_cpu(local->valuelen)); | |
1298 | } else { | |
a24374f4 | 1299 | remote = xfs_attr3_leaf_name_remote(leaf, i); |
61983f67 BN |
1300 | if (remote->namelen == 0 || |
1301 | remote->valueblk == 0) { | |
1302 | if (show_warnings) | |
1303 | print_warning("invalid attr " | |
1304 | "entry in inode %llu", | |
1305 | (long long)cur_ino); | |
1306 | break; | |
1307 | } | |
1308 | generate_obfuscated_name(0, remote->namelen, | |
1309 | &remote->name[0]); | |
1310 | add_remote_vals(be32_to_cpu(remote->valueblk), | |
1311 | be32_to_cpu(remote->valuelen)); | |
1312 | } | |
1313 | } | |
1314 | } | |
1315 | } | |
1316 | ||
1317 | /* inode copy routines */ | |
1318 | ||
1319 | static int | |
1320 | process_bmbt_reclist( | |
1321 | xfs_bmbt_rec_t *rp, | |
1322 | int numrecs, | |
1323 | typnm_t btype) | |
1324 | { | |
1325 | int i; | |
95c20099 | 1326 | xfs_dfiloff_t o, op = NULLDFILOFF; |
61983f67 | 1327 | xfs_dfsbno_t s; |
95c20099 | 1328 | xfs_dfilblks_t c, cp = NULLDFILOFF; |
61983f67 BN |
1329 | int f; |
1330 | xfs_dfiloff_t last; | |
88b8e1d6 BN |
1331 | xfs_agnumber_t agno; |
1332 | xfs_agblock_t agbno; | |
61983f67 BN |
1333 | |
1334 | if (btype == TYP_DATA) | |
1335 | return 1; | |
1336 | ||
1337 | convert_extent(&rp[numrecs - 1], &o, &s, &c, &f); | |
1338 | last = o + c; | |
1339 | ||
1340 | for (i = 0; i < numrecs; i++, rp++) { | |
1341 | convert_extent(rp, &o, &s, &c, &f); | |
1342 | ||
88b8e1d6 BN |
1343 | /* |
1344 | * ignore extents that are clearly bogus, and if a bogus | |
1345 | * one is found, stop processing remaining extents | |
1346 | */ | |
1347 | if (i > 0 && op + cp > o) { | |
1348 | if (show_warnings) | |
1349 | print_warning("bmap extent %d in %s ino %llu " | |
1350 | "starts at %llu, previous extent " | |
1351 | "ended at %llu", i, | |
1352 | typtab[btype].name, (long long)cur_ino, | |
1353 | o, op + cp - 1); | |
1354 | break; | |
1355 | } | |
1356 | ||
1357 | if (c > max_extent_size) { | |
1358 | /* | |
1359 | * since we are only processing non-data extents, | |
1360 | * large numbers of blocks in a metadata extent is | |
1361 | * extremely rare and more than likely to be corrupt. | |
1362 | */ | |
1363 | if (show_warnings) | |
1364 | print_warning("suspicious count %u in bmap " | |
1365 | "extent %d in %s ino %llu", c, i, | |
1366 | typtab[btype].name, (long long)cur_ino); | |
1367 | break; | |
1368 | } | |
1369 | ||
1370 | op = o; | |
1371 | cp = c; | |
1372 | ||
1373 | agno = XFS_FSB_TO_AGNO(mp, s); | |
1374 | agbno = XFS_FSB_TO_AGBNO(mp, s); | |
1375 | ||
1376 | if (!valid_bno(agno, agbno)) { | |
1377 | if (show_warnings) | |
1378 | print_warning("invalid block number %u/%u " | |
1379 | "(%llu) in bmap extent %d in %s ino " | |
1380 | "%llu", agno, agbno, s, i, | |
1381 | typtab[btype].name, (long long)cur_ino); | |
1382 | break; | |
1383 | } | |
1384 | ||
1385 | if (!valid_bno(agno, agbno + c - 1)) { | |
1386 | if (show_warnings) | |
1387 | print_warning("bmap extent %i in %s inode %llu " | |
1388 | "overflows AG (end is %u/%u)", i, | |
1389 | typtab[btype].name, (long long)cur_ino, | |
1390 | agno, agbno + c - 1); | |
1391 | break; | |
1392 | } | |
1393 | ||
61983f67 BN |
1394 | push_cur(); |
1395 | set_cur(&typtab[btype], XFS_FSB_TO_DADDR(mp, s), c * blkbb, | |
1396 | DB_RING_IGN, NULL); | |
1397 | if (iocur_top->data == NULL) { | |
88b8e1d6 BN |
1398 | print_warning("cannot read %s block %u/%u (%llu)", |
1399 | typtab[btype].name, agno, agbno, s); | |
d24c0a90 BN |
1400 | if (stop_on_read_error) { |
1401 | pop_cur(); | |
61983f67 | 1402 | return 0; |
d24c0a90 | 1403 | } |
61983f67 BN |
1404 | } else { |
1405 | if (!dont_obfuscate) | |
1406 | switch (btype) { | |
1407 | case TYP_DIR2: | |
1408 | if (o < mp->m_dirleafblk) | |
1409 | obfuscate_dir_data_blocks( | |
1410 | iocur_top->data, o, c, | |
1411 | last == mp->m_dirblkfsbs); | |
1412 | break; | |
1413 | ||
1414 | case TYP_SYMLINK: | |
1415 | obfuscate_symlink_blocks( | |
1416 | iocur_top->data, c); | |
1417 | break; | |
1418 | ||
1419 | case TYP_ATTR: | |
1420 | obfuscate_attr_blocks(iocur_top->data, | |
1421 | o, c); | |
1422 | break; | |
1423 | ||
1424 | default: ; | |
1425 | } | |
d24c0a90 BN |
1426 | if (!write_buf(iocur_top)) { |
1427 | pop_cur(); | |
61983f67 | 1428 | return 0; |
d24c0a90 | 1429 | } |
61983f67 BN |
1430 | } |
1431 | pop_cur(); | |
1432 | } | |
1433 | ||
1434 | return 1; | |
1435 | } | |
1436 | ||
1437 | static int | |
1438 | scanfunc_bmap( | |
b194c7d8 | 1439 | struct xfs_btree_block *block, |
61983f67 BN |
1440 | xfs_agnumber_t agno, |
1441 | xfs_agblock_t agbno, | |
1442 | int level, | |
1443 | typnm_t btype, | |
1444 | void *arg) /* ptr to itype */ | |
1445 | { | |
1446 | int i; | |
1447 | xfs_bmbt_ptr_t *pp; | |
61983f67 BN |
1448 | int nrecs; |
1449 | ||
b194c7d8 | 1450 | nrecs = be16_to_cpu(block->bb_numrecs); |
61983f67 BN |
1451 | |
1452 | if (level == 0) { | |
1453 | if (nrecs > mp->m_bmap_dmxr[0]) { | |
1454 | if (show_warnings) | |
1455 | print_warning("invalid numrecs (%u) in %s " | |
1456 | "block %u/%u", nrecs, | |
1457 | typtab[btype].name, agno, agbno); | |
1458 | return 1; | |
1459 | } | |
b3563c19 BN |
1460 | return process_bmbt_reclist(XFS_BMBT_REC_ADDR(mp, block, 1), |
1461 | nrecs, *(typnm_t*)arg); | |
61983f67 BN |
1462 | } |
1463 | ||
1464 | if (nrecs > mp->m_bmap_dmxr[1]) { | |
1465 | if (show_warnings) | |
1466 | print_warning("invalid numrecs (%u) in %s block %u/%u", | |
1467 | nrecs, typtab[btype].name, agno, agbno); | |
1468 | return 1; | |
1469 | } | |
b3563c19 | 1470 | pp = XFS_BMBT_PTR_ADDR(mp, block, 1, mp->m_bmap_dmxr[1]); |
61983f67 BN |
1471 | for (i = 0; i < nrecs; i++) { |
1472 | xfs_agnumber_t ag; | |
1473 | xfs_agblock_t bno; | |
1474 | ||
1475 | ag = XFS_FSB_TO_AGNO(mp, be64_to_cpu(pp[i])); | |
1476 | bno = XFS_FSB_TO_AGBNO(mp, be64_to_cpu(pp[i])); | |
1477 | ||
1478 | if (bno == 0 || bno > mp->m_sb.sb_agblocks || | |
1479 | ag > mp->m_sb.sb_agcount) { | |
1480 | if (show_warnings) | |
1481 | print_warning("invalid block number (%u/%u) " | |
1482 | "in %s block %u/%u", ag, bno, | |
1483 | typtab[btype].name, agno, agbno); | |
1484 | continue; | |
1485 | } | |
1486 | ||
1487 | if (!scan_btree(ag, bno, level, btype, arg, scanfunc_bmap)) | |
1488 | return 0; | |
1489 | } | |
1490 | return 1; | |
1491 | } | |
1492 | ||
1493 | static int | |
1494 | process_btinode( | |
1495 | xfs_dinode_t *dip, | |
1496 | typnm_t itype) | |
1497 | { | |
1498 | xfs_bmdr_block_t *dib; | |
1499 | int i; | |
1500 | xfs_bmbt_ptr_t *pp; | |
61983f67 BN |
1501 | int level; |
1502 | int nrecs; | |
1503 | int maxrecs; | |
1504 | int whichfork; | |
1505 | typnm_t btype; | |
1506 | ||
1507 | whichfork = (itype == TYP_ATTR) ? XFS_ATTR_FORK : XFS_DATA_FORK; | |
1508 | btype = (itype == TYP_ATTR) ? TYP_BMAPBTA : TYP_BMAPBTD; | |
1509 | ||
1510 | dib = (xfs_bmdr_block_t *)XFS_DFORK_PTR(dip, whichfork); | |
1511 | level = be16_to_cpu(dib->bb_level); | |
1512 | nrecs = be16_to_cpu(dib->bb_numrecs); | |
1513 | ||
1514 | if (level > XFS_BM_MAXLEVELS(mp, whichfork)) { | |
1515 | if (show_warnings) | |
1516 | print_warning("invalid level (%u) in inode %lld %s " | |
1517 | "root", level, (long long)cur_ino, | |
1518 | typtab[btype].name); | |
1519 | return 1; | |
1520 | } | |
1521 | ||
b3563c19 BN |
1522 | if (level == 0) { |
1523 | return process_bmbt_reclist(XFS_BMDR_REC_ADDR(dib, 1), | |
1524 | nrecs, itype); | |
1525 | } | |
61983f67 | 1526 | |
b3563c19 | 1527 | maxrecs = xfs_bmdr_maxrecs(mp, XFS_DFORK_SIZE(dip, mp, whichfork), 0); |
61983f67 BN |
1528 | if (nrecs > maxrecs) { |
1529 | if (show_warnings) | |
1530 | print_warning("invalid numrecs (%u) in inode %lld %s " | |
1531 | "root", nrecs, (long long)cur_ino, | |
1532 | typtab[btype].name); | |
1533 | return 1; | |
1534 | } | |
1535 | ||
b3563c19 | 1536 | pp = XFS_BMDR_PTR_ADDR(dib, 1, maxrecs); |
61983f67 BN |
1537 | for (i = 0; i < nrecs; i++) { |
1538 | xfs_agnumber_t ag; | |
1539 | xfs_agblock_t bno; | |
1540 | ||
1541 | ag = XFS_FSB_TO_AGNO(mp, be64_to_cpu(pp[i])); | |
1542 | bno = XFS_FSB_TO_AGBNO(mp, be64_to_cpu(pp[i])); | |
1543 | ||
1544 | if (bno == 0 || bno > mp->m_sb.sb_agblocks || | |
1545 | ag > mp->m_sb.sb_agcount) { | |
1546 | if (show_warnings) | |
1547 | print_warning("invalid block number (%u/%u) " | |
1548 | "in inode %llu %s root", ag, | |
1549 | bno, (long long)cur_ino, | |
1550 | typtab[btype].name); | |
1551 | continue; | |
1552 | } | |
1553 | ||
1554 | if (!scan_btree(ag, bno, level, btype, &itype, scanfunc_bmap)) | |
1555 | return 0; | |
1556 | } | |
1557 | return 1; | |
1558 | } | |
1559 | ||
1560 | static int | |
1561 | process_exinode( | |
1562 | xfs_dinode_t *dip, | |
1563 | typnm_t itype) | |
1564 | { | |
1565 | int whichfork; | |
88b8e1d6 | 1566 | xfs_extnum_t nex; |
61983f67 BN |
1567 | |
1568 | whichfork = (itype == TYP_ATTR) ? XFS_ATTR_FORK : XFS_DATA_FORK; | |
1569 | ||
5e656dbb BN |
1570 | nex = XFS_DFORK_NEXTENTS(dip, whichfork); |
1571 | if (nex < 0 || nex > XFS_DFORK_SIZE(dip, mp, whichfork) / | |
1572 | sizeof(xfs_bmbt_rec_t)) { | |
88b8e1d6 BN |
1573 | if (show_warnings) |
1574 | print_warning("bad number of extents %d in inode %lld", | |
1575 | nex, (long long)cur_ino); | |
1576 | return 1; | |
1577 | } | |
1578 | ||
1579 | return process_bmbt_reclist((xfs_bmbt_rec_t *)XFS_DFORK_PTR(dip, | |
1580 | whichfork), nex, itype); | |
61983f67 BN |
1581 | } |
1582 | ||
1583 | static int | |
1584 | process_inode_data( | |
1585 | xfs_dinode_t *dip, | |
1586 | typnm_t itype) | |
1587 | { | |
56b2de80 | 1588 | switch (dip->di_format) { |
61983f67 BN |
1589 | case XFS_DINODE_FMT_LOCAL: |
1590 | if (!dont_obfuscate) | |
1591 | switch (itype) { | |
1592 | case TYP_DIR2: | |
1593 | obfuscate_sf_dir(dip); | |
1594 | break; | |
1595 | ||
1596 | case TYP_SYMLINK: | |
1597 | obfuscate_sf_symlink(dip); | |
1598 | break; | |
1599 | ||
1600 | default: ; | |
1601 | } | |
1602 | break; | |
1603 | ||
1604 | case XFS_DINODE_FMT_EXTENTS: | |
1605 | return process_exinode(dip, itype); | |
1606 | ||
1607 | case XFS_DINODE_FMT_BTREE: | |
1608 | return process_btinode(dip, itype); | |
1609 | } | |
1610 | return 1; | |
1611 | } | |
1612 | ||
1613 | static int | |
1614 | process_inode( | |
1615 | xfs_agnumber_t agno, | |
1616 | xfs_agino_t agino, | |
1617 | xfs_dinode_t *dip) | |
1618 | { | |
61983f67 BN |
1619 | int success; |
1620 | ||
61983f67 BN |
1621 | success = 1; |
1622 | cur_ino = XFS_AGINO_TO_INO(mp, agno, agino); | |
1623 | ||
61983f67 | 1624 | /* copy appropriate data fork metadata */ |
56b2de80 | 1625 | switch (be16_to_cpu(dip->di_mode) & S_IFMT) { |
61983f67 BN |
1626 | case S_IFDIR: |
1627 | memset(&dir_data, 0, sizeof(dir_data)); | |
1628 | success = process_inode_data(dip, TYP_DIR2); | |
1629 | break; | |
1630 | case S_IFLNK: | |
1631 | success = process_inode_data(dip, TYP_SYMLINK); | |
1632 | break; | |
88b8e1d6 | 1633 | case S_IFREG: |
61983f67 | 1634 | success = process_inode_data(dip, TYP_DATA); |
88b8e1d6 BN |
1635 | break; |
1636 | default: ; | |
61983f67 | 1637 | } |
a85f8b0a | 1638 | nametable_clear(); |
61983f67 | 1639 | |
88b8e1d6 | 1640 | /* copy extended attributes if they exist and forkoff is valid */ |
49f693fa DC |
1641 | if (success && |
1642 | XFS_DFORK_DSIZE(dip, mp) < XFS_LITINO(mp, dip->di_version)) { | |
61983f67 | 1643 | attr_data.remote_val_count = 0; |
56b2de80 | 1644 | switch (dip->di_aformat) { |
61983f67 BN |
1645 | case XFS_DINODE_FMT_LOCAL: |
1646 | if (!dont_obfuscate) | |
1647 | obfuscate_sf_attr(dip); | |
1648 | break; | |
1649 | ||
1650 | case XFS_DINODE_FMT_EXTENTS: | |
1651 | success = process_exinode(dip, TYP_ATTR); | |
1652 | break; | |
1653 | ||
1654 | case XFS_DINODE_FMT_BTREE: | |
1655 | success = process_btinode(dip, TYP_ATTR); | |
1656 | break; | |
1657 | } | |
a85f8b0a | 1658 | nametable_clear(); |
61983f67 | 1659 | } |
61983f67 BN |
1660 | return success; |
1661 | } | |
1662 | ||
1663 | static __uint32_t inodes_copied = 0; | |
1664 | ||
1665 | static int | |
1666 | copy_inode_chunk( | |
1667 | xfs_agnumber_t agno, | |
1668 | xfs_inobt_rec_t *rp) | |
1669 | { | |
1670 | xfs_agino_t agino; | |
1671 | int off; | |
1672 | xfs_agblock_t agbno; | |
1673 | int i; | |
d24c0a90 | 1674 | int rval = 0; |
61983f67 BN |
1675 | |
1676 | agino = be32_to_cpu(rp->ir_startino); | |
1677 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1678 | off = XFS_INO_TO_OFFSET(mp, agino); | |
1679 | ||
88b8e1d6 BN |
1680 | if (agino == 0 || agino == NULLAGINO || !valid_bno(agno, agbno) || |
1681 | !valid_bno(agno, XFS_AGINO_TO_AGBNO(mp, | |
1682 | agino + XFS_INODES_PER_CHUNK - 1))) { | |
1683 | if (show_warnings) | |
1684 | print_warning("bad inode number %llu (%u/%u)", | |
1685 | XFS_AGINO_TO_INO(mp, agno, agino), agno, agino); | |
1686 | return 1; | |
1687 | } | |
1688 | ||
61983f67 BN |
1689 | push_cur(); |
1690 | set_cur(&typtab[TYP_INODE], XFS_AGB_TO_DADDR(mp, agno, agbno), | |
1691 | XFS_FSB_TO_BB(mp, XFS_IALLOC_BLOCKS(mp)), | |
1692 | DB_RING_IGN, NULL); | |
1693 | if (iocur_top->data == NULL) { | |
1694 | print_warning("cannot read inode block %u/%u", agno, agbno); | |
d24c0a90 BN |
1695 | rval = !stop_on_read_error; |
1696 | goto pop_out; | |
61983f67 BN |
1697 | } |
1698 | ||
88b8e1d6 BN |
1699 | /* |
1700 | * check for basic assumptions about inode chunks, and if any | |
1701 | * assumptions fail, don't process the inode chunk. | |
1702 | */ | |
1703 | ||
1704 | if ((mp->m_sb.sb_inopblock <= XFS_INODES_PER_CHUNK && off != 0) || | |
1705 | (mp->m_sb.sb_inopblock > XFS_INODES_PER_CHUNK && | |
1706 | off % XFS_INODES_PER_CHUNK != 0) || | |
5e656dbb | 1707 | (xfs_sb_version_hasalign(&mp->m_sb) && |
88b8e1d6 BN |
1708 | agbno % mp->m_sb.sb_inoalignmt != 0)) { |
1709 | if (show_warnings) | |
1710 | print_warning("badly aligned inode (start = %llu)", | |
1711 | XFS_AGINO_TO_INO(mp, agno, agino)); | |
1712 | goto skip_processing; | |
1713 | } | |
1714 | ||
61983f67 BN |
1715 | /* |
1716 | * scan through inodes and copy any btree extent lists, directory | |
1717 | * contents and extended attributes. | |
1718 | */ | |
61983f67 BN |
1719 | for (i = 0; i < XFS_INODES_PER_CHUNK; i++) { |
1720 | xfs_dinode_t *dip; | |
1721 | ||
1722 | if (XFS_INOBT_IS_FREE_DISK(rp, i)) | |
1723 | continue; | |
1724 | ||
1725 | dip = (xfs_dinode_t *)((char *)iocur_top->data + | |
1726 | ((off + i) << mp->m_sb.sb_inodelog)); | |
1727 | ||
1728 | if (!process_inode(agno, agino + i, dip)) | |
d24c0a90 | 1729 | goto pop_out; |
61983f67 | 1730 | } |
88b8e1d6 | 1731 | skip_processing: |
61983f67 | 1732 | if (!write_buf(iocur_top)) |
d24c0a90 | 1733 | goto pop_out; |
61983f67 BN |
1734 | |
1735 | inodes_copied += XFS_INODES_PER_CHUNK; | |
1736 | ||
1737 | if (show_progress) | |
1738 | print_progress("Copied %u of %u inodes (%u of %u AGs)", | |
1739 | inodes_copied, mp->m_sb.sb_icount, agno, | |
1740 | mp->m_sb.sb_agcount); | |
d24c0a90 BN |
1741 | rval = 1; |
1742 | pop_out: | |
61983f67 | 1743 | pop_cur(); |
d24c0a90 | 1744 | return rval; |
61983f67 BN |
1745 | } |
1746 | ||
1747 | static int | |
1748 | scanfunc_ino( | |
b194c7d8 | 1749 | struct xfs_btree_block *block, |
61983f67 BN |
1750 | xfs_agnumber_t agno, |
1751 | xfs_agblock_t agbno, | |
1752 | int level, | |
1753 | typnm_t btype, | |
1754 | void *arg) | |
1755 | { | |
1756 | xfs_inobt_rec_t *rp; | |
1757 | xfs_inobt_ptr_t *pp; | |
1758 | int i; | |
88b8e1d6 BN |
1759 | int numrecs; |
1760 | ||
b194c7d8 | 1761 | numrecs = be16_to_cpu(block->bb_numrecs); |
61983f67 BN |
1762 | |
1763 | if (level == 0) { | |
88b8e1d6 BN |
1764 | if (numrecs > mp->m_inobt_mxr[0]) { |
1765 | if (show_warnings) | |
1766 | print_warning("invalid numrecs %d in %s " | |
1767 | "block %u/%u", numrecs, | |
1768 | typtab[btype].name, agno, agbno); | |
1769 | numrecs = mp->m_inobt_mxr[0]; | |
1770 | } | |
b3563c19 | 1771 | rp = XFS_INOBT_REC_ADDR(mp, block, 1); |
88b8e1d6 | 1772 | for (i = 0; i < numrecs; i++, rp++) { |
61983f67 BN |
1773 | if (!copy_inode_chunk(agno, rp)) |
1774 | return 0; | |
1775 | } | |
88b8e1d6 BN |
1776 | return 1; |
1777 | } | |
1778 | ||
1779 | if (numrecs > mp->m_inobt_mxr[1]) { | |
1780 | if (show_warnings) | |
1781 | print_warning("invalid numrecs %d in %s block %u/%u", | |
1782 | numrecs, typtab[btype].name, agno, agbno); | |
1783 | numrecs = mp->m_inobt_mxr[1]; | |
1784 | } | |
1785 | ||
b3563c19 | 1786 | pp = XFS_INOBT_PTR_ADDR(mp, block, 1, mp->m_inobt_mxr[1]); |
88b8e1d6 BN |
1787 | for (i = 0; i < numrecs; i++) { |
1788 | if (!valid_bno(agno, be32_to_cpu(pp[i]))) { | |
1789 | if (show_warnings) | |
1790 | print_warning("invalid block number (%u/%u) " | |
1791 | "in %s block %u/%u", | |
1792 | agno, be32_to_cpu(pp[i]), | |
1793 | typtab[btype].name, agno, agbno); | |
1794 | continue; | |
61983f67 | 1795 | } |
88b8e1d6 BN |
1796 | if (!scan_btree(agno, be32_to_cpu(pp[i]), level, |
1797 | btype, arg, scanfunc_ino)) | |
1798 | return 0; | |
61983f67 BN |
1799 | } |
1800 | return 1; | |
1801 | } | |
1802 | ||
1803 | static int | |
1804 | copy_inodes( | |
1805 | xfs_agnumber_t agno, | |
1806 | xfs_agi_t *agi) | |
1807 | { | |
1808 | xfs_agblock_t root; | |
1809 | int levels; | |
1810 | ||
1811 | root = be32_to_cpu(agi->agi_root); | |
1812 | levels = be32_to_cpu(agi->agi_level); | |
1813 | ||
1814 | /* validate root and levels before processing the tree */ | |
1815 | if (root == 0 || root > mp->m_sb.sb_agblocks) { | |
1816 | if (show_warnings) | |
1817 | print_warning("invalid block number (%u) in inobt " | |
1818 | "root in agi %u", root, agno); | |
1819 | return 1; | |
1820 | } | |
1821 | if (levels >= XFS_BTREE_MAXLEVELS) { | |
1822 | if (show_warnings) | |
1823 | print_warning("invalid level (%u) in inobt root " | |
1824 | "in agi %u", levels, agno); | |
1825 | return 1; | |
1826 | } | |
1827 | ||
1828 | return scan_btree(agno, root, levels, TYP_INOBT, agi, scanfunc_ino); | |
1829 | } | |
1830 | ||
1831 | static int | |
1832 | scan_ag( | |
1833 | xfs_agnumber_t agno) | |
1834 | { | |
1835 | xfs_agf_t *agf; | |
1836 | xfs_agi_t *agi; | |
d24c0a90 BN |
1837 | int stack_count = 0; |
1838 | int rval = 0; | |
61983f67 BN |
1839 | |
1840 | /* copy the superblock of the AG */ | |
1841 | push_cur(); | |
d24c0a90 | 1842 | stack_count++; |
61983f67 BN |
1843 | set_cur(&typtab[TYP_SB], XFS_AG_DADDR(mp, agno, XFS_SB_DADDR), |
1844 | XFS_FSS_TO_BB(mp, 1), DB_RING_IGN, NULL); | |
1845 | if (!iocur_top->data) { | |
1846 | print_warning("cannot read superblock for ag %u", agno); | |
1847 | if (stop_on_read_error) | |
d24c0a90 | 1848 | goto pop_out; |
61983f67 BN |
1849 | } else { |
1850 | if (!write_buf(iocur_top)) | |
d24c0a90 | 1851 | goto pop_out; |
61983f67 BN |
1852 | } |
1853 | ||
1854 | /* copy the AG free space btree root */ | |
1855 | push_cur(); | |
d24c0a90 | 1856 | stack_count++; |
61983f67 BN |
1857 | set_cur(&typtab[TYP_AGF], XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)), |
1858 | XFS_FSS_TO_BB(mp, 1), DB_RING_IGN, NULL); | |
1859 | agf = iocur_top->data; | |
1860 | if (iocur_top->data == NULL) { | |
1861 | print_warning("cannot read agf block for ag %u", agno); | |
1862 | if (stop_on_read_error) | |
d24c0a90 | 1863 | goto pop_out; |
61983f67 BN |
1864 | } else { |
1865 | if (!write_buf(iocur_top)) | |
d24c0a90 | 1866 | goto pop_out; |
61983f67 BN |
1867 | } |
1868 | ||
1869 | /* copy the AG inode btree root */ | |
1870 | push_cur(); | |
d24c0a90 | 1871 | stack_count++; |
61983f67 BN |
1872 | set_cur(&typtab[TYP_AGI], XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)), |
1873 | XFS_FSS_TO_BB(mp, 1), DB_RING_IGN, NULL); | |
1874 | agi = iocur_top->data; | |
1875 | if (iocur_top->data == NULL) { | |
1876 | print_warning("cannot read agi block for ag %u", agno); | |
1877 | if (stop_on_read_error) | |
d24c0a90 | 1878 | goto pop_out; |
61983f67 BN |
1879 | } else { |
1880 | if (!write_buf(iocur_top)) | |
d24c0a90 | 1881 | goto pop_out; |
61983f67 BN |
1882 | } |
1883 | ||
1884 | /* copy the AG free list header */ | |
1885 | push_cur(); | |
d24c0a90 | 1886 | stack_count++; |
61983f67 BN |
1887 | set_cur(&typtab[TYP_AGFL], XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)), |
1888 | XFS_FSS_TO_BB(mp, 1), DB_RING_IGN, NULL); | |
1889 | if (iocur_top->data == NULL) { | |
1890 | print_warning("cannot read agfl block for ag %u", agno); | |
1891 | if (stop_on_read_error) | |
d24c0a90 | 1892 | goto pop_out; |
61983f67 BN |
1893 | } else { |
1894 | if (!write_buf(iocur_top)) | |
d24c0a90 | 1895 | goto pop_out; |
61983f67 | 1896 | } |
61983f67 BN |
1897 | |
1898 | /* copy AG free space btrees */ | |
1899 | if (agf) { | |
1900 | if (show_progress) | |
1901 | print_progress("Copying free space trees of AG %u", | |
1902 | agno); | |
1903 | if (!copy_free_bno_btree(agno, agf)) | |
d24c0a90 | 1904 | goto pop_out; |
61983f67 | 1905 | if (!copy_free_cnt_btree(agno, agf)) |
d24c0a90 | 1906 | goto pop_out; |
61983f67 BN |
1907 | } |
1908 | ||
1909 | /* copy inode btrees and the inodes and their associated metadata */ | |
1910 | if (agi) { | |
1911 | if (!copy_inodes(agno, agi)) | |
d24c0a90 | 1912 | goto pop_out; |
61983f67 | 1913 | } |
d24c0a90 BN |
1914 | rval = 1; |
1915 | pop_out: | |
1916 | while (stack_count--) | |
1917 | pop_cur(); | |
1918 | return rval; | |
61983f67 BN |
1919 | } |
1920 | ||
1921 | static int | |
1922 | copy_ino( | |
1923 | xfs_ino_t ino, | |
1924 | typnm_t itype) | |
1925 | { | |
1926 | xfs_agnumber_t agno; | |
1927 | xfs_agblock_t agbno; | |
1928 | xfs_agino_t agino; | |
61983f67 | 1929 | int offset; |
d24c0a90 | 1930 | int rval = 0; |
61983f67 BN |
1931 | |
1932 | if (ino == 0) | |
1933 | return 1; | |
1934 | ||
1935 | agno = XFS_INO_TO_AGNO(mp, ino); | |
1936 | agino = XFS_INO_TO_AGINO(mp, ino); | |
1937 | agbno = XFS_AGINO_TO_AGBNO(mp, agino); | |
1938 | offset = XFS_AGINO_TO_OFFSET(mp, agino); | |
1939 | ||
1940 | if (agno >= mp->m_sb.sb_agcount || agbno >= mp->m_sb.sb_agblocks || | |
1941 | offset >= mp->m_sb.sb_inopblock) { | |
1942 | if (show_warnings) | |
1943 | print_warning("invalid %s inode number (%lld)", | |
1944 | typtab[itype].name, (long long)ino); | |
1945 | return 1; | |
1946 | } | |
1947 | ||
1948 | push_cur(); | |
1949 | set_cur(&typtab[TYP_INODE], XFS_AGB_TO_DADDR(mp, agno, agbno), | |
1950 | blkbb, DB_RING_IGN, NULL); | |
1951 | if (iocur_top->data == NULL) { | |
1952 | print_warning("cannot read %s inode %lld", | |
1953 | typtab[itype].name, (long long)ino); | |
d24c0a90 BN |
1954 | rval = !stop_on_read_error; |
1955 | goto pop_out; | |
61983f67 BN |
1956 | } |
1957 | off_cur(offset << mp->m_sb.sb_inodelog, mp->m_sb.sb_inodesize); | |
1958 | ||
61983f67 | 1959 | cur_ino = ino; |
5e656dbb | 1960 | rval = process_inode_data(iocur_top->data, itype); |
d24c0a90 BN |
1961 | pop_out: |
1962 | pop_cur(); | |
1963 | return rval; | |
61983f67 BN |
1964 | } |
1965 | ||
1966 | ||
1967 | static int | |
1968 | copy_sb_inodes(void) | |
1969 | { | |
1970 | if (!copy_ino(mp->m_sb.sb_rbmino, TYP_RTBITMAP)) | |
1971 | return 0; | |
1972 | ||
1973 | if (!copy_ino(mp->m_sb.sb_rsumino, TYP_RTSUMMARY)) | |
1974 | return 0; | |
1975 | ||
1976 | if (!copy_ino(mp->m_sb.sb_uquotino, TYP_DQBLK)) | |
1977 | return 0; | |
1978 | ||
0340d706 CS |
1979 | if (!copy_ino(mp->m_sb.sb_gquotino, TYP_DQBLK)) |
1980 | return 0; | |
1981 | ||
1982 | return copy_ino(mp->m_sb.sb_pquotino, TYP_DQBLK); | |
61983f67 BN |
1983 | } |
1984 | ||
1985 | static int | |
1986 | copy_log(void) | |
1987 | { | |
1988 | if (show_progress) | |
1989 | print_progress("Copying log"); | |
1990 | ||
1991 | push_cur(); | |
1992 | set_cur(&typtab[TYP_LOG], XFS_FSB_TO_DADDR(mp, mp->m_sb.sb_logstart), | |
1993 | mp->m_sb.sb_logblocks * blkbb, DB_RING_IGN, NULL); | |
1994 | if (iocur_top->data == NULL) { | |
d24c0a90 | 1995 | pop_cur(); |
61983f67 BN |
1996 | print_warning("cannot read log"); |
1997 | return !stop_on_read_error; | |
1998 | } | |
1999 | return write_buf(iocur_top); | |
2000 | } | |
2001 | ||
2002 | static int | |
2003 | metadump_f( | |
2004 | int argc, | |
2005 | char **argv) | |
2006 | { | |
2007 | xfs_agnumber_t agno; | |
2008 | int c; | |
2009 | int start_iocur_sp; | |
88b8e1d6 | 2010 | char *p; |
61983f67 BN |
2011 | |
2012 | exitcode = 1; | |
2013 | show_progress = 0; | |
2014 | show_warnings = 0; | |
2015 | stop_on_read_error = 0; | |
2016 | ||
2017 | if (mp->m_sb.sb_magicnum != XFS_SB_MAGIC) { | |
2018 | print_warning("bad superblock magic number %x, giving up", | |
2019 | mp->m_sb.sb_magicnum); | |
2020 | return 0; | |
2021 | } | |
2022 | ||
88b8e1d6 | 2023 | while ((c = getopt(argc, argv, "egm:ow")) != EOF) { |
61983f67 BN |
2024 | switch (c) { |
2025 | case 'e': | |
2026 | stop_on_read_error = 1; | |
2027 | break; | |
2028 | case 'g': | |
2029 | show_progress = 1; | |
2030 | break; | |
88b8e1d6 BN |
2031 | case 'm': |
2032 | max_extent_size = (int)strtol(optarg, &p, 0); | |
2033 | if (*p != '\0' || max_extent_size <= 0) { | |
2034 | print_warning("bad max extent size %s", | |
2035 | optarg); | |
2036 | return 0; | |
2037 | } | |
2038 | break; | |
61983f67 BN |
2039 | case 'o': |
2040 | dont_obfuscate = 1; | |
2041 | break; | |
2042 | case 'w': | |
2043 | show_warnings = 1; | |
2044 | break; | |
2045 | default: | |
2046 | print_warning("bad option for metadump command"); | |
2047 | return 0; | |
2048 | } | |
2049 | } | |
2050 | ||
2051 | if (optind != argc - 1) { | |
2052 | print_warning("too few options for metadump (no filename given)"); | |
2053 | return 0; | |
2054 | } | |
2055 | ||
1a03f4a6 DC |
2056 | if (xfs_sb_version_hascrc(&mp->m_sb) && dont_obfuscate == 0) { |
2057 | print_warning("Can't obfuscate CRC enabled filesystems yet."); | |
2058 | return 0; | |
2059 | } | |
2060 | ||
61983f67 BN |
2061 | metablock = (xfs_metablock_t *)calloc(BBSIZE + 1, BBSIZE); |
2062 | if (metablock == NULL) { | |
2063 | print_warning("memory allocation failure"); | |
2064 | return 0; | |
2065 | } | |
2066 | metablock->mb_blocklog = BBSHIFT; | |
2067 | metablock->mb_magic = cpu_to_be32(XFS_MD_MAGIC); | |
2068 | ||
61983f67 BN |
2069 | block_index = (__be64 *)((char *)metablock + sizeof(xfs_metablock_t)); |
2070 | block_buffer = (char *)metablock + BBSIZE; | |
2071 | num_indicies = (BBSIZE - sizeof(xfs_metablock_t)) / sizeof(__be64); | |
2072 | cur_index = 0; | |
2073 | start_iocur_sp = iocur_sp; | |
2074 | ||
2075 | if (strcmp(argv[optind], "-") == 0) { | |
2076 | if (isatty(fileno(stdout))) { | |
2077 | print_warning("cannot write to a terminal"); | |
61983f67 BN |
2078 | free(metablock); |
2079 | return 0; | |
2080 | } | |
2081 | outf = stdout; | |
2082 | } else { | |
2083 | outf = fopen(argv[optind], "wb"); | |
2084 | if (outf == NULL) { | |
2085 | print_warning("cannot create dump file"); | |
61983f67 BN |
2086 | free(metablock); |
2087 | return 0; | |
2088 | } | |
2089 | } | |
2090 | ||
2091 | exitcode = 0; | |
2092 | ||
2093 | for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) { | |
2094 | if (!scan_ag(agno)) { | |
2095 | exitcode = 1; | |
2096 | break; | |
2097 | } | |
2098 | } | |
2099 | ||
2100 | /* copy realtime and quota inode contents */ | |
2101 | if (!exitcode) | |
2102 | exitcode = !copy_sb_inodes(); | |
2103 | ||
2104 | /* copy log if it's internal */ | |
2105 | if ((mp->m_sb.sb_logstart != 0) && !exitcode) | |
2106 | exitcode = !copy_log(); | |
2107 | ||
2108 | /* write the remaining index */ | |
2109 | if (!exitcode) | |
2110 | exitcode = !write_index(); | |
2111 | ||
2112 | if (progress_since_warning) | |
2113 | fputc('\n', (outf == stdout) ? stderr : stdout); | |
2114 | ||
2115 | if (outf != stdout) | |
2116 | fclose(outf); | |
2117 | ||
2118 | /* cleanup iocur stack */ | |
2119 | while (iocur_sp > start_iocur_sp) | |
2120 | pop_cur(); | |
2121 | ||
61983f67 BN |
2122 | free(metablock); |
2123 | ||
2124 | return 0; | |
2125 | } |