Remove unnecessary $O from gengetopt target.
[osl.git] / fsck.c
1 /*
2 * Copyright (C) 2007-2009 Andre Noll <maan@systemlinux.org>
3 *
4 * Licensed under the GPL v2. For licencing details see COPYING.
5 */
6
7 /** \file fsck.c The program used to check an osl table. */
8
9 #include <sys/stat.h>
10 #include <sys/types.h>
11 #include <dirent.h>
12 #include <assert.h>
13 #include <pwd.h>
14
15 #include "log.h"
16 #include "osl.h"
17 #include "util.h"
18 #include "osl_core.h"
19 #include "fsck.cmdline.h"
20
21 /** version text used by various commands if -V switch was given */
22 #define VERSION_TEXT(prefix) "osl_" prefix " " VERSION " " "\n" \
23 "Copyright (C) 2008-2009 Andre Noll\n" \
24 "This is free software with ABSOLUTELY NO WARRANTY." \
25 " See COPYING for details.\n" \
26 "Written by Andre Noll.\n" \
27 "Report bugs to <maan@systemlinux.org>.\n"
28
29 /** print out \p VERSION_TEXT and exit if version flag was given */
30 #define HANDLE_VERSION_FLAG(_prefix, _args_info_struct) \
31 if (_args_info_struct.version_given) { \
32 printf("%s", VERSION_TEXT(_prefix)); \
33 exit(EXIT_SUCCESS); \
34 }
35
36 static struct fsck_args_info conf;
37
38 #define FSCK_ERRORS \
39 FSCK_ERROR(RANGE_VIOLATION, "range violation detected, very bad"), \
40 FSCK_ERROR(NOT_A_REGULAR_FILE, "not a regular file"), \
41 FSCK_ERROR(SYNTAX, "fsck syntax error"), \
42 FSCK_ERROR(ACCESS, "permission denied"), \
43 FSCK_ERROR(CHDIR, "could not change directory"), \
44 FSCK_ERROR(OPENDIR, "could not open directory"),
45
46 #define FSCK_ERROR_BIT 29
47 #define FSCK_ERROR(num, txt) E_FSCK_ ## num
48 enum {
49 FSCK_DUMMY = (1 << FSCK_ERROR_BIT) - 1,
50 FSCK_ERRORS
51 };
52 #undef FSCK_ERROR
53 #define FSCK_ERROR(num, txt) txt
54 static const char *fsck_errors[] = {
55 FSCK_ERRORS
56 };
57
58 static const char *fsck_strerror(int num)
59 {
60 if (num & (1 << FSCK_ERROR_BIT))
61 return fsck_errors[num & ((1 << FSCK_ERROR_BIT) - 1)];
62 return osl_strerror(num);
63 }
64
65 extern int loglevel;
66
67 /* taken from git */
68 static signed char hexval_table[256] = {
69 -1, -1, -1, -1, -1, -1, -1, -1, /* 00-07 */
70 -1, -1, -1, -1, -1, -1, -1, -1, /* 08-0f */
71 -1, -1, -1, -1, -1, -1, -1, -1, /* 10-17 */
72 -1, -1, -1, -1, -1, -1, -1, -1, /* 18-1f */
73 -1, -1, -1, -1, -1, -1, -1, -1, /* 20-27 */
74 -1, -1, -1, -1, -1, -1, -1, -1, /* 28-2f */
75 0, 1, 2, 3, 4, 5, 6, 7, /* 30-37 */
76 8, 9, -1, -1, -1, -1, -1, -1, /* 38-3f */
77 -1, 10, 11, 12, 13, 14, 15, -1, /* 40-47 */
78 -1, -1, -1, -1, -1, -1, -1, -1, /* 48-4f */
79 -1, -1, -1, -1, -1, -1, -1, -1, /* 50-57 */
80 -1, -1, -1, -1, -1, -1, -1, -1, /* 58-5f */
81 -1, 10, 11, 12, 13, 14, 15, -1, /* 60-67 */
82 -1, -1, -1, -1, -1, -1, -1, -1, /* 68-67 */
83 -1, -1, -1, -1, -1, -1, -1, -1, /* 70-77 */
84 -1, -1, -1, -1, -1, -1, -1, -1, /* 78-7f */
85 -1, -1, -1, -1, -1, -1, -1, -1, /* 80-87 */
86 -1, -1, -1, -1, -1, -1, -1, -1, /* 88-8f */
87 -1, -1, -1, -1, -1, -1, -1, -1, /* 90-97 */
88 -1, -1, -1, -1, -1, -1, -1, -1, /* 98-9f */
89 -1, -1, -1, -1, -1, -1, -1, -1, /* a0-a7 */
90 -1, -1, -1, -1, -1, -1, -1, -1, /* a8-af */
91 -1, -1, -1, -1, -1, -1, -1, -1, /* b0-b7 */
92 -1, -1, -1, -1, -1, -1, -1, -1, /* b8-bf */
93 -1, -1, -1, -1, -1, -1, -1, -1, /* c0-c7 */
94 -1, -1, -1, -1, -1, -1, -1, -1, /* c8-cf */
95 -1, -1, -1, -1, -1, -1, -1, -1, /* d0-d7 */
96 -1, -1, -1, -1, -1, -1, -1, -1, /* d8-df */
97 -1, -1, -1, -1, -1, -1, -1, -1, /* e0-e7 */
98 -1, -1, -1, -1, -1, -1, -1, -1, /* e8-ef */
99 -1, -1, -1, -1, -1, -1, -1, -1, /* f0-f7 */
100 -1, -1, -1, -1, -1, -1, -1, -1, /* f8-ff */
101 };
102
103 static int asc_to_hash(const char *asc_hash, int len, HASH_TYPE *hash)
104 {
105 int i = 0;
106 const unsigned char *asc = (const unsigned char *) asc_hash;
107
108 while (*asc && i++ < len) {
109 unsigned int val = (hexval_table[asc[0]] << 4) | hexval_table[asc[1]];
110 if (val & ~0xff)
111 return -1;
112 *hash++ = val;
113 asc += 2;
114
115 }
116 return 1;
117 }
118
119 static int _write_all(int fd, const char *buf, size_t len)
120 {
121 return write_all(fd, buf, &len);
122 }
123
124 /*
125 * Wrapper for malloc().
126 *
127 * \param size The desired new size.
128 *
129 * A wrapper for malloc(3) which exits on errors.
130 *
131 * \return A pointer to the allocated memory, which is suitably aligned for any
132 * kind of variable.
133 *
134 * \sa malloc(3).
135 */
136 __must_check __malloc static void *fsck_malloc(size_t size)
137 {
138 assert(size);
139 void *p = malloc(size);
140
141 if (!p) {
142 EMERG_LOG("malloc failed (size = %zu), aborting\n",
143 size);
144 exit(EXIT_FAILURE);
145 }
146 return p;
147 }
148
149 /**
150 * Allocate memory and fill with zeros.
151 *
152 * \param size The desired new size.
153 *
154 * A wrapper for calloc(3) which exits on errors.
155 *
156 * \return A pointer to the allocated and zeroed-out memory, which is suitably
157 * aligned for any kind of variable.
158 *
159 * \sa calloc(3)
160 */
161 __must_check __malloc static void *fsck_calloc(size_t size)
162 {
163 void *ret = fsck_malloc(size);
164
165 memset(ret, 0, size);
166 return ret;
167 }
168
169 /**
170 * Save version of strdup().
171 *
172 * \param s The string to be duplicated.
173 *
174 * A wrapper for strdup(3). It calls \p exit(EXIT_FAILURE) on errors, i.e.
175 * there is no need to check the return value in the caller.
176 *
177 * \return A pointer to the duplicated string. If \p s was the NULL pointer,
178 * an pointer to an empty string is returned.
179 *
180 * \sa strdup(3)
181 */
182 __must_check __malloc static char *fsck_strdup(const char *s)
183 {
184 char *ret;
185
186 if ((ret = strdup(s? s: "")))
187 return ret;
188 EMERG_LOG("strdup failed, aborting\n");
189 exit(EXIT_FAILURE);
190 }
191
192 /**
193 * Compare two osl objects pointing to unsigned integers of 32 bit size.
194 *
195 * \param obj1 Pointer to the first integer.
196 * \param obj2 Pointer to the second integer.
197 *
198 * \return The values required for an osl compare function.
199 *
200 * \sa osl_compare_func, osl_hash_compare().
201 */
202 static int uint32_compare(const struct osl_object *obj1, const struct osl_object *obj2)
203 {
204 uint32_t d1 = read_u32((const char *)obj1->data);
205 uint32_t d2 = read_u32((const char *)obj2->data);
206
207 if (d1 < d2)
208 return 1;
209 if (d1 > d2)
210 return -1;
211 return 0;
212 }
213
214 /**
215 * A wrapper for fchdir().
216 *
217 * \param fd An open file descriptor.
218 *
219 * \return Standard.
220 */
221 static inline int __fchdir(int fd)
222 {
223 if (fchdir(fd) >= 0)
224 return 1;
225 return errno == EACCES? -E_FSCK_ACCESS : -E_FSCK_CHDIR;
226 }
227
228 /**
229 * Wrapper for chdir(2).
230 *
231 * \param path The specified directory.
232 *
233 * \return Standard.
234 */
235 _static_inline_ int __chdir(const char *path)
236 {
237 if (chdir(path) >= 0)
238 return 1;
239 return errno == EACCES? -E_FSCK_ACCESS : -E_FSCK_CHDIR;
240 }
241
242 /**
243 * Save the cwd and open a given directory.
244 *
245 * \param dirname Path to the directory to open.
246 * \param dir Result pointer.
247 * \param cwd File descriptor of the current working directory.
248 *
249 * \return Standard.
250 *
251 * Opening the current directory (".") and calling fchdir() to return is
252 * usually faster and more reliable than saving cwd in some buffer and calling
253 * chdir() afterwards.
254 *
255 * If \a cwd is not \p NULL "." is opened and the resulting file descriptor is
256 * stored in \a cwd. If the function returns success, and \a cwd is not \p
257 * NULL, the caller must close this file descriptor (probably after calling
258 * fchdir(*cwd)).
259 *
260 * On errors, the function undos everything, so the caller needs neither close
261 * any files, nor change back to the original working directory.
262 *
263 * \sa getcwd(3).
264 *
265 */
266 static int fsck_opendir(const char *dirname, DIR **dir, int *cwd)
267 {
268 int ret;
269
270 if (cwd) {
271 ret = osl_open(".", O_RDONLY, 0);
272 if (ret < 0)
273 return ret;
274 *cwd = ret;
275 }
276 ret = __chdir(dirname);
277 if (ret < 0)
278 goto close_cwd;
279 *dir = opendir(".");
280 if (*dir)
281 return 1;
282 ret = errno == EACCES? -E_FSCK_ACCESS : -E_FSCK_OPENDIR;
283 /* Ignore return value of fchdir() and close(). We're busted anyway. */
284 if (cwd)
285 fchdir(*cwd);
286 close_cwd:
287 if (cwd)
288 close(*cwd);
289 return ret;
290 }
291
292
293
294 /**
295 * Traverse the given directory recursively.
296 *
297 * \param dirname The directory to traverse.
298 * \param func The function to call for each entry.
299 * \param private_data Pointer to an arbitrary data structure.
300 *
301 * For each regular file under \a dirname, the supplied function \a func is
302 * called. The full path of the regular file and the \a private_data pointer
303 * are passed to \a func. Directories for which the calling process has no
304 * permissions to change to are silently ignored.
305 *
306 * \return Standard.
307 */
308 static int for_each_file_in_dir(const char *dirname,
309 int (*func)(const char *, void *), void *private_data)
310 {
311 DIR *dir;
312 struct dirent *entry;
313 int cwd_fd, ret2, ret = fsck_opendir(dirname, &dir, &cwd_fd);
314
315 if (ret < 0)
316 return ret == -E_FSCK_ACCESS? 1 : ret;
317 /* scan cwd recursively */
318 while ((entry = readdir(dir))) {
319 mode_t m;
320 char *tmp;
321 struct stat s;
322
323 if (!strcmp(entry->d_name, "."))
324 continue;
325 if (!strcmp(entry->d_name, ".."))
326 continue;
327 if (lstat(entry->d_name, &s) == -1)
328 continue;
329 m = s.st_mode;
330 if (!S_ISREG(m) && !S_ISDIR(m))
331 continue;
332 tmp = make_message("%s/%s", dirname, entry->d_name);
333 if (!S_ISDIR(m)) {
334 ret = func(tmp, private_data);
335 free(tmp);
336 if (ret < 0)
337 goto out;
338 continue;
339 }
340 /* directory */
341 ret = for_each_file_in_dir(tmp, func, private_data);
342 free(tmp);
343 if (ret < 0)
344 goto out;
345 }
346 ret = 1;
347 out:
348 closedir(dir);
349 ret2 = __fchdir(cwd_fd);
350 if (ret2 < 0 && ret >= 0)
351 ret = ret2;
352 close(cwd_fd);
353 return ret;
354 }
355
356 /*
357 * check for object boundary violations
358 *
359 * test whether the range pointed to by the index entry for a given cell is
360 * contained in mapped data file. This should always be the case. Otherwise
361 * we are in real trouble.
362 */
363 static int check_range(struct osl_table *t, uint32_t row_num, uint32_t col_num)
364 {
365 char *index_entry;
366 struct osl_object obj;
367 struct osl_column *col;
368 int ret;
369 char *map_start, *obj_start;
370
371 ret = get_cell_index(t, row_num, col_num, &index_entry);
372 if (ret < 0)
373 return ret;
374 ret = get_mapped_object(t, col_num, row_num, &obj);
375 if (ret < 0)
376 return ret;
377 col = t->columns + col_num;
378 obj_start = obj.data;
379 map_start = col->data_map.data;
380 // INFO_LOG("obj: %p..%p\n", obj_start, obj_start + obj.size);
381 // INFO_LOG("map: %p..%p\n", map_start, map_start + col->data_map.size);
382 if (obj_start < map_start || obj_start + obj.size > map_start + col->data_map.size) {
383 CRIT_LOG("range violation in row %u, col %u\n", row_num,
384 col_num);
385 return -E_FSCK_RANGE_VIOLATION;
386 }
387 DEBUG_LOG("col %u: ok\n", col_num);
388 return 1;
389 }
390
391 static int fsck_mark_row_invalid(struct osl_table *t, int i)
392 {
393 if (conf.dry_run_given)
394 return 0;
395 return mark_row_invalid(t, i);
396 }
397
398 /*
399 * check all cells of the given table for boundary violations
400 */
401 static int check_index_ranges(struct osl_table *t)
402 {
403 int i, j, ret;
404
405 INFO_LOG("checking for range violations in index\n");
406 //DEBUG_LOG("%d rows. %d columns\n", t->num_rows, t->desc->num_columns);
407 t->num_invalid_rows = 0;
408 for (i = 0; i < t->num_rows; i++) {
409 const struct osl_column_description *cd;
410
411 if (row_is_invalid(t, i)) {
412 t->num_invalid_rows++;
413 continue;
414 }
415 FOR_EACH_MAPPED_COLUMN(j, t, cd) {
416 ret = check_range(t, i, j);
417 if (ret < 0) {
418 if (ret != -E_FSCK_RANGE_VIOLATION)
419 goto err;
420 ret = fsck_mark_row_invalid(t, i);
421 if (ret < 0)
422 goto err;
423 t->num_invalid_rows++;
424 break;
425 }
426 }
427
428 }
429 if (t->num_invalid_rows)
430 NOTICE_LOG("ranges OK. %d invalid row(s) detected\n",
431 t->num_invalid_rows);
432 else
433 INFO_LOG("no invalid rows, no range violations, good\n");
434 return 1;
435 err:
436 return ret;
437 }
438
439 static int move_index_entry(struct osl_table *t, uint32_t dest, uint32_t src)
440 {
441 char *dest_ie, *src_ie;
442 int ret = get_row_index(t, dest, &dest_ie);
443
444 if (ret < 0)
445 return ret;
446 ret = get_row_index(t, src, &src_ie);
447 if (ret < 0)
448 return ret;
449 INFO_LOG("moving entry #%u to position %u\n", src, dest);
450 memcpy(dest_ie, src_ie, t->row_index_size);
451 return 1;
452 }
453
454 static int map_index(const struct osl_table_description *desc, struct osl_object *map)
455 {
456 char *filename = index_filename(desc);
457 int ret;
458
459 ret = mmap_full_file(filename, O_RDWR, &map->data, &map->size, NULL);
460 DEBUG_LOG("mapping index %s: ret: %d, size: %zu\n", filename, ret, map->size);
461 free(filename);
462 return ret;
463 }
464
465 static int prune_invalid_rows_from_index(struct osl_table *t)
466 {
467 uint32_t top = 0, bottom;
468 char *filename;
469 int ret;
470
471 if (!t->num_invalid_rows) {
472 INFO_LOG("all rows are valid, good\n");
473 return 1;
474 }
475 NOTICE_LOG("index contains %u invalid row(s) (%d bytes)\n",
476 t->num_invalid_rows, t->row_index_size * t->num_invalid_rows);
477 if (conf.dry_run_given)
478 return 0;
479 NOTICE_LOG("removing invalid rows from index\n");
480 bottom = t->num_rows - 1;
481 while (top < bottom) {
482 if (!row_is_invalid(t, top)) {
483 top++;
484 continue;
485 }
486 while (bottom > top) {
487 if (row_is_invalid(t, bottom)) {
488 bottom--;
489 continue;
490 }
491 /* move bottom index entry to top */
492 move_index_entry(t, top, bottom);
493 bottom--;
494 top++;
495 break;
496 }
497 }
498 DEBUG_LOG("unmapping index\n");
499 osl_munmap(t->index_map.data, t->index_map.size);
500 filename = index_filename(t->desc);
501 ret = truncate_file(filename, t->row_index_size
502 * t->num_invalid_rows);
503 free(filename);
504 if (ret < 0)
505 return ret;
506 ret = map_index(t->desc, &t->index_map);
507 if (ret < 0)
508 return ret;
509 t->num_rows = table_num_rows(t);
510 return 1;
511 }
512
513 static int check_for_invalid_objects(struct osl_table *t, uint32_t **lost_bytes)
514 {
515 int i, j, ret;
516 const struct osl_column_description *cd;
517 uint32_t *loss = fsck_malloc(sizeof(uint32_t) * t->desc->num_columns);
518
519 INFO_LOG("looking for mapped objects not contained in index\n");
520 /* first count used bytes */
521 FOR_EACH_MAPPED_COLUMN(i, t, cd) {
522 loss[i] = t->columns[i].data_map.size;
523 DEBUG_LOG("column %i data map: %zu bytes\n", i,
524 t->columns[i].data_map.size);
525 for (j = 0; j < t->num_rows; j++) {
526 struct osl_object obj;
527 ret = get_mapped_object(t, i, j, &obj);
528 if (ret < 0)
529 goto err;
530 loss[i] -= obj.size;
531 }
532 }
533 ret = 0;
534 FOR_EACH_MAPPED_COLUMN(i, t, cd) {
535 if (loss[i]) {
536 NOTICE_LOG("column %u contains %u lost bytes\n",
537 i, loss[i]);
538 ret = 1;
539 }
540 }
541 if (!ret)
542 INFO_LOG("all mapped objects are valid, good\n");
543 *lost_bytes = loss;
544 return ret;
545 err:
546 free(loss);
547 return ret;
548 }
549
550 /* prune_invalid_rows() must be run on the table before calling this */
551 static int prune_mapped_column(struct osl_table *t, uint32_t col_num, int fd)
552 {
553 int i, ret;
554 uint32_t written = 0;
555 struct osl_column *col = t->columns + col_num;
556
557 INFO_LOG("pruning col %u\n", col_num);
558 for (i = 0; i < t->num_rows; i++) {
559 struct osl_object obj;
560 char *index_entry;
561
562 DEBUG_LOG("checking row %u/%u\n", i, t->num_rows);
563 ret = get_mapped_object(t, col_num, i, &obj);
564 if (ret < 0)
565 return ret;
566 ret = _write_all(fd, (char *)(obj.data), obj.size);
567 if (ret < 0)
568 return ret;
569 written += obj.size;
570 ret = get_row_index(t, i, &index_entry);
571 if (ret < 0)
572 return ret;
573 update_cell_index(index_entry, col, written, obj.size);
574 }
575 return 1;
576 }
577
578 static int prune_objects(struct osl_table *t, uint32_t *lost_bytes)
579 {
580 int i, ret;
581 const struct osl_column_description *cd;
582 char **col_filenames = fsck_calloc(t->desc->num_columns * sizeof(char *));
583 char **new_col_filenames = fsck_calloc(t->desc->num_columns * sizeof(char *));
584 char *idx_filename = index_filename(t->desc);
585 char *old_idx_filename = make_message("%s.bak", idx_filename);
586 int fd;
587
588 NOTICE_LOG("removing unreferenced objects from data files\n");
589 /* first make a copy of the index */
590 DEBUG_LOG("opening %s\n", old_idx_filename);
591 ret = osl_open(old_idx_filename, O_WRONLY | O_CREAT | O_TRUNC, 0644);
592 if (ret < 0)
593 goto out_free;
594 fd = ret;
595 ret = _write_all(fd, t->index_map.data, t->index_map.size);
596 close(fd);
597 if (ret < 0)
598 goto out_free;
599 FOR_EACH_MAPPED_COLUMN(i, t, cd) {
600 if (!lost_bytes[i])
601 continue;
602 col_filenames[i] = column_filename(t, i);
603 new_col_filenames[i] = make_message("%s.fsck", col_filenames[i]);
604 ret = osl_open(new_col_filenames[i], O_WRONLY | O_CREAT | O_EXCL, 0644);
605 if (ret < 0)
606 goto out_unlink_data;
607 fd = ret;
608 ret = prune_mapped_column(t, i, fd);
609 close(fd);
610 if (ret < 0)
611 goto out_unlink_data;
612 }
613 ret = unmap_table(t, OSL_MARK_CLEAN);
614 if (ret < 0)
615 goto out_unlink_data;
616 FOR_EACH_MAPPED_COLUMN(i, t, cd) {
617 if (!lost_bytes[i])
618 continue;
619 ret = osl_rename(new_col_filenames[i], col_filenames[i]);
620 if (ret < 0) { /* we're kinda screwed here */
621 CRIT_LOG("rename of col %i failed: %s\n", i,
622 osl_strerror(errno));
623 goto out_free;
624 }
625 }
626 unlink(old_idx_filename);
627 ret = map_table(t, 0);
628 goto out_free;
629 out_unlink_data:
630 FOR_EACH_MAPPED_COLUMN(i, t, cd)
631 unlink(new_col_filenames[i]);
632 out_free:
633 free(old_idx_filename);
634 free(idx_filename);
635 FOR_EACH_MAPPED_COLUMN(i, t, cd) {
636 free(col_filenames[i]);
637 free(new_col_filenames[i]);
638 }
639 free(col_filenames);
640 free(new_col_filenames);
641 return ret;
642 }
643
644 static struct osl_column_description hash_tree_table_cols[] = {
645 {
646 .storage_type = OSL_NO_STORAGE,
647 .storage_flags = OSL_RBTREE | OSL_FIXED_SIZE | OSL_UNIQUE,
648 .name = "hash",
649 .compare_function = uint32_compare,
650 .data_size = HASH_SIZE
651 },
652 };
653
654 static const struct osl_table_description hash_tree_table_desc = {
655 .dir = "/", /* irrelevant */
656 .name = "hash_tree",
657 .num_columns = 1,
658 .flags = 0,
659 .column_descriptions = hash_tree_table_cols
660 };
661
662 /**
663 * The hash_tree table contains all hashes of the disk storage name column.
664 * of each row. It is used for checking if a disk storage file has a reference
665 * in the table.
666 */
667 static struct osl_table *hash_tree_table;
668 static HASH_TYPE *hashes;
669
670 static int check_disk_storage_column(struct osl_table *t, int row_num,
671 int col_num, char *ds_name, unsigned *num_missing_objects)
672 {
673 int ret;
674 struct stat statbuf;
675 char *path = disk_storage_path(t, col_num, ds_name);
676 unsigned dsnc = t->disk_storage_name_column;
677 struct osl_object obj;
678
679 DEBUG_LOG("checking if %s is a regular file\n", path);
680 ret = stat(path, &statbuf);
681 if (ret < 0 && errno == ENOENT) {
682 struct osl_row *row;
683 (*num_missing_objects)++;
684 ERROR_LOG("row %d: object %s is missing\n", row_num, path);
685 ret = 0;
686 if (conf.dry_run_given)
687 goto out;
688 NOTICE_LOG("trying to delete row %d\n", row_num);
689 ret = osl_get_row(t, dsnc, &obj, &row);
690 if (ret < 0) {
691 CRIT_LOG("unable to get row %d\n", row_num);
692 fsck_mark_row_invalid(t, row_num);
693 CRIT_LOG("Please re-run fsck\n");
694 goto out;
695 }
696 ret = osl_del_row(t, row);
697 if (ret < 0)
698 goto out;
699 }
700 out:
701 free(path);
702 if (ret < 0)
703 return ret;
704 ret = -E_FSCK_NOT_A_REGULAR_FILE;
705 if (!(S_IFREG & statbuf.st_mode))
706 return ret;
707 return 1;
708 }
709
710 static int check_disk_storage_presence(struct osl_table *t)
711 {
712 int ret, i, j;
713 struct osl_object obj, hash_obj = {.size = HASH_SIZE};
714 char *ds_name;
715 const struct osl_column_description *cd;
716 unsigned dsnc = t->disk_storage_name_column, missing_objects = 0;
717
718 if (!t->num_rows)
719 return 1;
720 hashes = fsck_malloc(t->num_rows * HASH_SIZE);
721 INFO_LOG("looking for missing disk storage objects\n");
722 for (i = 0; i < t->num_rows; i++) {
723 if (row_is_invalid(t, i))
724 continue;
725 ret = get_mapped_object(t, dsnc, i, &obj);
726 if (ret < 0)
727 return ret;
728 hash_object(&obj, hashes + i * HASH_SIZE);
729 hash_obj.data = hashes + i * HASH_SIZE;
730 osl_add_row(hash_tree_table, &hash_obj);
731 ds_name = disk_storage_name_of_hash(t, hashes + i * HASH_SIZE);
732 FOR_EACH_DISK_STORAGE_COLUMN(j, t, cd) {
733 ret = check_disk_storage_column(t, i, j, ds_name,
734 &missing_objects);
735 if (ret < 0)
736 goto err;
737 }
738 free(ds_name);
739 }
740 if (!missing_objects)
741 INFO_LOG("all referenced disk storage objects exist, good\n");
742 else
743 NOTICE_LOG("%d missing object(s)\n", missing_objects);
744 return missing_objects;
745 err:
746 free(ds_name);
747 return ret;
748 }
749
750 static int dummy_compare(const struct osl_object *obj1, const struct osl_object *obj2)
751 {
752 if (obj1 < obj2)
753 return -1;
754 if (obj1 > obj2)
755 return 1;
756 return 0;
757 }
758
759 static unsigned files_pruned;
760
761 static int prune_disk_storage_file(const char *path, void *private_data)
762 {
763 HASH_TYPE hash[HASH_SIZE];
764 uint8_t flags = *(uint8_t *)private_data;
765 struct osl_object obj = {.data = hash, .size = HASH_SIZE};
766 struct osl_row *row;
767 int ret = -1;
768 size_t len = strlen(path);
769
770
771 DEBUG_LOG("path: %s\n", path);
772 if (flags & OSL_LARGE_TABLE) {
773 if (len < HASH_SIZE * 2 + 2)
774 goto invalid;
775 // NOTICE_LOG("p: %s\n", path + len - 2 * HASH_SIZE - 1);
776 ret = asc_to_hash(path + len - 2 * HASH_SIZE - 1, 1, hash);
777 if (ret < 0)
778 goto invalid;
779 ret = asc_to_hash(path + len - 2 * HASH_SIZE + 2, HASH_SIZE - 1,
780 hash + 1);
781 if (ret < 0)
782 goto invalid;
783 // INFO_LOG("high: %x, low: %x, hash: %x\n", high, low, hash);
784 } else {
785 if (len < 2 * HASH_SIZE + 1)
786 goto invalid;
787 ret = asc_to_hash(path + len - 2 * HASH_SIZE, 2 * HASH_SIZE, hash);
788 if (ret < 0)
789 goto invalid;
790 // INFO_LOG("hash: %x\n", hash);
791 }
792 #if 0
793 {
794 char asc[2 * HASH_SIZE + 1];
795 hash_to_asc(hash, asc);
796 NOTICE_LOG("before: %s\nafter: %s\n", path, asc);
797 }
798 #endif
799 ret = osl_get_row(hash_tree_table, 0, &obj, &row);
800 if (ret >= 0)
801 return 1;
802 NOTICE_LOG("unreferenced file in hash dir: %s\n", path);
803 goto remove;
804 invalid:
805 ERROR_LOG("could not read hash value of %s\n", path);
806 remove:
807 if (conf.dry_run_given)
808 return 0;
809 NOTICE_LOG("removing %s\n", path);
810 unlink(path);
811 files_pruned++;
812 return 1;
813 }
814
815 static int prune_disk_storage_files(struct osl_table *t)
816 {
817 int i, ret = 1;
818 const struct osl_column_description *cd;
819
820 INFO_LOG("looking for unreferenced disk storage files\n");
821 FOR_EACH_DISK_STORAGE_COLUMN(i, t, cd) {
822 char *dirname = column_filename(t, i);
823 uint8_t flags = t->desc->flags;
824 ret = for_each_file_in_dir(dirname, prune_disk_storage_file,
825 &flags);
826 free(dirname);
827 }
828 if (files_pruned)
829 NOTICE_LOG("%u disk storage files deleted\n",
830 files_pruned);
831 else
832 INFO_LOG("all files are are referenced, good\n");
833 return ret;
834 }
835
836 static int check_disk_storage_columns(struct osl_table *t)
837 {
838 int ret, i;
839 const struct osl_column_description *cd;
840
841 if (!t->num_disk_storage_columns) {
842 INFO_LOG("no disk storage columns in table '%s', "
843 "skipping checks\n", t->desc->name);
844 return 1;
845 }
846 FOR_EACH_COLUMN(i, t->desc, cd)
847 t->desc->column_descriptions[i].compare_function = dummy_compare;
848 ret = init_rbtrees(t);
849 if (ret < 0)
850 return ret;
851 INFO_LOG("creating rbtree for disk storage hash values\n");
852 ret = osl_open_table(&hash_tree_table_desc, &hash_tree_table);
853 if (ret < 0)
854 goto out;
855 ret = check_disk_storage_presence(t);
856 if (ret < 0)
857 goto out_close_hash_tree;
858 ret = prune_disk_storage_files(t);
859 out_close_hash_tree:
860 osl_close_table(hash_tree_table, 0);
861 free(hashes);
862 hashes = NULL;
863 out:
864 clear_rbtrees(t); /* TODO why are we doing that here? Seems odd */
865 return ret;
866 }
867
868 static void set_dummy_contents(struct osl_table_description *desc)
869 {
870 int i;
871 struct osl_column_description *cd;
872
873 for (i = 0; i < desc->num_columns; i++) {
874 cd = get_column_description(desc, i);
875 cd->compare_function = dummy_compare;
876 }
877 }
878
879 static int fsck_init(struct osl_table_description *desc, struct osl_table **t)
880 {
881 struct osl_object map;
882 int ret = map_index(desc, &map);
883
884 if (ret < 0)
885 goto out;
886 ret = read_table_desc(&map, desc);
887 if (ret < 0) {
888 osl_munmap(map.data, map.size);
889 goto out;
890 }
891 set_dummy_contents(desc);
892 ret = init_table_structure(desc, t);
893 if (ret < 0) {
894 osl_munmap(map.data, map.size);
895 goto out;
896 }
897 DEBUG_LOG("unmapping index\n");
898 osl_munmap(map.data, map.size);
899 if (conf.force_given)
900 ret = map_table(*t, (MAP_TBL_FL_IGNORE_DIRTY));
901 else
902 ret = map_table(*t, 0);
903 if (ret >= 0) {
904 (*t)->num_rows = table_num_rows(*t);
905 DEBUG_LOG("index header size: %d\n", (*t)->index_header_size);
906 DEBUG_LOG("row index size: %d\n", (*t)->row_index_size);
907 }
908 out:
909 return ret;
910 }
911
912 static void fsck_cleanup(struct osl_table *t)
913 {
914 int i;
915
916 if (!t)
917 return;
918 if (t->desc->column_descriptions) {
919 struct osl_column_description *cd;
920 for (i = 0; i < t->desc->num_columns; i++) {
921 cd = get_column_description(t->desc, i);
922 free((char*)cd->name);
923 }
924 free(t->desc->column_descriptions);
925 }
926 free(t->columns);
927 free(t);
928
929 }
930
931 #define ST_CASE(st) case st: return #st
932
933 static const char *get_asc_storage_type(enum osl_storage_type st)
934 {
935 switch (st) {
936 ST_CASE(OSL_MAPPED_STORAGE);
937 ST_CASE(OSL_DISK_STORAGE);
938 ST_CASE(OSL_NO_STORAGE);
939 }
940 return NULL;
941 }
942
943 #define APPEND_ASC_SF(sf, flag, str) do { if (sf & flag) { \
944 if (str) str = make_message("%s%s", str, " | " # flag); \
945 else str = fsck_strdup(#flag); }} while (0)
946
947
948 static char *get_asc_storage_flags(enum osl_storage_type sf)
949 {
950 char *asc_sf = NULL;
951
952 APPEND_ASC_SF(sf, OSL_RBTREE, asc_sf);
953 APPEND_ASC_SF(sf, OSL_FIXED_SIZE, asc_sf);
954 APPEND_ASC_SF(sf, OSL_UNIQUE, asc_sf);
955 return asc_sf;
956 }
957
958 static int dump_table_desc(struct osl_table *t, int fd)
959 {
960 const struct osl_table_description *desc = t->desc;
961 int ret, i;
962 struct osl_column_description *cd;
963 char *msg = make_message("static struct osl_column_description cols[] = {\n");
964 ret = _write_all(fd, msg, strlen(msg));
965 if (ret < 0)
966 return ret;
967 free(msg);
968 FOR_EACH_COLUMN(i, desc, cd) {
969 const char *asc_st;
970 msg = make_message("\t[%d] = {\n", i);
971 ret = _write_all(fd, msg, strlen(msg));
972 if (ret < 0)
973 return ret;
974 free(msg);
975 asc_st = get_asc_storage_type(cd->storage_type);
976 msg = make_message("\t\t.storage_type = %s,\n", asc_st);
977 ret = _write_all(fd, msg, strlen(msg));
978 if (ret < 0)
979 return ret;
980 free(msg);
981 if (cd->storage_flags) {
982 char *asc_sf = get_asc_storage_flags(cd->storage_flags);
983 msg = make_message("\t\t,storage_flags = %s,\n", asc_sf);
984 free(asc_sf);
985 ret = _write_all(fd, msg, strlen(msg));
986 if (ret < 0)
987 return ret;
988 free(msg);
989 }
990 if (cd->storage_flags & OSL_FIXED_SIZE) {
991 msg = make_message("\t\t.data_size = %u,\n", cd->data_size);
992 ret = _write_all(fd, msg, strlen(msg));
993 if (ret < 0)
994 return ret;
995 free(msg);
996 }
997 msg = make_message("\t\t.name = \"%s\",\n", cd->name);
998 ret = _write_all(fd, msg, strlen(msg));
999 if (ret < 0)
1000 return ret;
1001 free(msg);
1002 if (cd->storage_flags & OSL_RBTREE) {
1003 msg = make_message("\t\t.compare_function = compare_func,\n");
1004 ret = _write_all(fd, msg, strlen(msg));
1005 if (ret < 0)
1006 return ret;
1007 free(msg);
1008 }
1009 msg = make_message("\t},\n");
1010 ret = _write_all(fd, msg, strlen(msg));
1011 if (ret < 0)
1012 return ret;
1013 free(msg);
1014 }
1015 msg = make_message("};\n");
1016 ret = _write_all(fd, msg, strlen(msg));
1017 if (ret < 0)
1018 return ret;
1019 free(msg);
1020 return 1;
1021 }
1022
1023 static int dump_row(struct osl_table *t, unsigned row_num, const char *row_dir)
1024 {
1025 int ret, i;
1026 const struct osl_column_description *cd;
1027 unsigned dsnc;
1028 struct osl_object obj;
1029 char *ds_name;
1030 HASH_TYPE hash[HASH_SIZE];
1031 char *filename;
1032
1033 FOR_EACH_MAPPED_COLUMN(i, t, cd) {
1034 ret = get_mapped_object(t, i, row_num, &obj);
1035 if (ret < 0)
1036 return ret;
1037 filename = make_message("%s/col_%03u", row_dir, i);
1038 ret = write_file(filename, obj.data, obj.size);
1039 free(filename);
1040 if (ret < 0)
1041 return ret;
1042 }
1043 if (!t->num_disk_storage_columns)
1044 return 1;
1045 dsnc = t->disk_storage_name_column;
1046 ret = get_mapped_object(t, dsnc, row_num, &obj);
1047 if (ret < 0)
1048 return ret;
1049 hash_object(&obj, hash);
1050 ds_name = disk_storage_name_of_hash(t, hash);
1051 FOR_EACH_DISK_STORAGE_COLUMN(i, t, cd) {
1052 filename = disk_storage_path(t, i, ds_name);
1053 ret = mmap_full_file(filename, O_RDONLY, &obj.data, &obj.size, NULL);
1054 free(filename);
1055 if (ret < 0)
1056 goto out;
1057 filename = make_message("%s/col_%03u", row_dir, i);
1058 ret = write_file(filename, obj.data, obj.size);
1059 free(filename);
1060 if (ret < 0)
1061 goto out;
1062 }
1063 ret = 1;
1064 out:
1065 free(ds_name);
1066 return ret;
1067 }
1068
1069 static int dump_rows(char *dump_dir, struct osl_table *t)
1070 {
1071 unsigned i;
1072 char *current_dir = NULL;
1073 int ret = 0;
1074
1075 for (i = 0; i < t->num_rows; i++) {
1076 char *row_dir;
1077 if (row_is_invalid(t, i))
1078 continue;
1079 if (!(i % 1000)) {
1080 free(current_dir);
1081 current_dir = make_message("%s/rows_%u-%u", dump_dir, i, i + 999);
1082 NOTICE_LOG("dumping rows %u - %u\n", i, i + 999);
1083 ret = osl_mkdir(current_dir, 0777);
1084 if (ret < 0 && ret != -E_OSL_DIR_EXISTS)
1085 goto out;
1086 }
1087 row_dir = make_message("%s/row_%03u", current_dir, i);
1088 ret = osl_mkdir(row_dir, 0777);
1089 if (ret < 0 && ret != -E_OSL_DIR_EXISTS) {
1090 free(row_dir);
1091 goto out;
1092 }
1093 ret = dump_row(t, i, row_dir);
1094 free(row_dir);
1095 if (ret < 0)
1096 goto out;
1097 }
1098 out:
1099 free(current_dir);
1100 return ret;
1101 }
1102
1103 static int dump_table(char *dump_dir, struct osl_table_description *desc)
1104 {
1105 struct osl_table *t = NULL;
1106 int fd, ret = fsck_init(desc, &t);
1107 char *desc_file;
1108 char *table_dump_dir = NULL;
1109
1110 if (ret < 0)
1111 goto out;
1112 ret = osl_mkdir(dump_dir, 0777);
1113 if (ret < 0 && ret != -E_OSL_DIR_EXISTS)
1114 goto out;
1115 table_dump_dir = make_message("%s/%s", dump_dir, desc->name);
1116 ret = osl_mkdir(table_dump_dir, 0777);
1117 if (ret < 0 && ret != -E_OSL_DIR_EXISTS)
1118 goto out;
1119 desc_file = make_message("%s/table_description.c", table_dump_dir);
1120 ret = osl_open(desc_file, O_WRONLY | O_CREAT | O_EXCL, 0644);
1121 free(desc_file);
1122 if (ret < 0)
1123 goto out;
1124 fd = ret;
1125 ret = dump_table_desc(t, fd);
1126 close(fd);
1127 if (ret < 0)
1128 goto out;
1129 ret = dump_rows(table_dump_dir, t);
1130 out:
1131 free(table_dump_dir);
1132 fsck_cleanup(t);
1133 return ret;
1134 }
1135
1136 static int fsck(struct osl_table_description *desc)
1137 {
1138 int ret;
1139 struct osl_table *t = NULL;
1140 uint32_t *lost_bytes = NULL;
1141
1142 ret = fsck_init(desc, &t);
1143 if (ret < 0)
1144 goto out;
1145 ret = check_index_ranges(t);
1146 if (ret < 0)
1147 goto out_unmap;
1148 ret = check_disk_storage_columns(t);
1149 if (ret < 0)
1150 goto out_unmap;
1151 ret = prune_invalid_rows_from_index(t);
1152 if (ret < 0)
1153 goto out_unmap;
1154 ret = check_for_invalid_objects(t, &lost_bytes);
1155 if (ret < 0)
1156 goto out_unmap;
1157 if (ret > 0 && !conf.dry_run_given) { /* at least one mapped data file needs pruning */
1158 ret = prune_objects(t, lost_bytes);
1159 if (ret < 0)
1160 goto out_unmap;
1161 }
1162 free(lost_bytes);
1163 out_unmap:
1164 unmap_table(t, OSL_MARK_CLEAN);
1165 out:
1166 fsck_cleanup(t);
1167 return ret;
1168 }
1169
1170 static int check_table(char *db_dir, char *table_name)
1171 {
1172 struct osl_table_description desc = {
1173 .column_descriptions = NULL,
1174 .dir = db_dir,
1175 .name = table_name
1176 };
1177 int ret;
1178
1179 INFO_LOG("checking table %s\n", table_name);
1180 if (!conf.no_fsck_given) {
1181 ret = fsck(&desc);
1182 if (ret < 0)
1183 goto out;
1184 }
1185 ret = 1;
1186 if (!conf.dump_dir_given || !*conf.dump_dir_arg)
1187 goto out;
1188 ret = dump_table(conf.dump_dir_arg, &desc);
1189 out:
1190 if (ret < 0)
1191 ERROR_LOG("failed to check table %s: %s\n", table_name,
1192 fsck_strerror(-ret));
1193 else
1194 NOTICE_LOG("successfully checked table %s\n", table_name);
1195 return ret;
1196 }
1197
1198 static int check_all_tables(char *db_dir)
1199 {
1200 DIR *dir;
1201 struct dirent *entry;
1202 int cwd_fd, ret2, ret = fsck_opendir(db_dir, &dir, &cwd_fd);
1203
1204 if (ret < 0)
1205 return ret;
1206 while ((entry = readdir(dir))) {
1207 mode_t m;
1208 struct stat s;
1209 if (!strcmp(entry->d_name, "."))
1210 continue;
1211 if (!strcmp(entry->d_name, ".."))
1212 continue;
1213 if (lstat(entry->d_name, &s) == -1)
1214 continue;
1215 m = s.st_mode;
1216 if (!S_ISDIR(m))
1217 continue;
1218 ret = check_table(db_dir, entry->d_name);
1219 if (ret < 0)
1220 break;
1221 }
1222 closedir(dir);
1223 ret2 = __fchdir(cwd_fd);
1224 if (ret2 < 0 && ret >= 0)
1225 ret = ret2;
1226 close(cwd_fd);
1227 return ret;
1228 }
1229
1230 int main(int argc, char **argv)
1231 {
1232 int i, ret;
1233 struct fsck_cmdline_parser_params params = {
1234 .override = 0,
1235 .initialize = 1,
1236 .check_required = 1,
1237 .check_ambiguity = 1,
1238 .print_errors = 1
1239 };
1240 ret = fsck_cmdline_parser_ext(argc, argv, &conf, &params);
1241 if (ret < 0) {
1242 loglevel = EMERG;
1243 ret = -E_FSCK_SYNTAX;
1244 goto out;
1245 }
1246 loglevel = conf.loglevel_arg;
1247 HANDLE_VERSION_FLAG("fsck", conf);
1248 INFO_LOG("database dir: %s\n", conf.database_dir_arg);
1249 if (!conf.inputs_num) {
1250 ret = check_all_tables(conf.database_dir_arg);
1251 goto out;
1252 }
1253 for (i = 0; i < conf.inputs_num; i++) {
1254 ret = check_table(conf.database_dir_arg, conf.inputs[i]);
1255 if (ret < 0)
1256 break;
1257 }
1258 out:
1259 if (ret < 0) {
1260 ERROR_LOG("%s\n", fsck_strerror(-ret));
1261 if (conf.loglevel_arg > 1)
1262 EMERG_LOG("re-run with \"--loglevel %d\" to increase verbosity\n",
1263 conf.loglevel_arg - 1);
1264 } else
1265 NOTICE_LOG("success\n");
1266 return ret < 0? EXIT_FAILURE : EXIT_SUCCESS;
1267 }