2 * Unix SMB/CIFS implementation.
4 * A general tdb content validation mechanism
6 * Copyright (C) Michael Adam 2007
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 3 of the License, or
11 * (at your option) any later version.
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
18 * You should have received a copy of the GNU General Public License
19 * along with this program. If not, see <http://www.gnu.org/licenses/>.
23 #include "system/filesys.h"
25 #include "tdb_validate.h"
28 * internal validation function, executed by the child.
30 static int tdb_validate_child(struct tdb_context *tdb,
31 tdb_validate_data_func validate_fn)
35 struct tdb_validation_status v_status;
37 v_status.tdb_error = False;
38 v_status.bad_freelist = False;
39 v_status.bad_entry = False;
40 v_status.unknown_key = False;
41 v_status.success = True;
44 v_status.tdb_error = True;
45 v_status.success = False;
50 * we can simplify this by passing a check function,
51 * but I don't want to change all the callers...
53 ret = tdb_check(tdb, NULL, NULL);
55 v_status.tdb_error = True;
56 v_status.success = False;
60 /* Check if the tdb's freelist is good. */
61 if (tdb_validate_freelist(tdb, &num_entries) == -1) {
62 v_status.bad_freelist = True;
63 v_status.success = False;
67 DEBUG(10,("tdb_validate_child: tdb %s freelist has %d entries\n",
68 tdb_name(tdb), num_entries));
70 /* Now traverse the tdb to validate it. */
71 num_entries = tdb_traverse(tdb, validate_fn, (void *)&v_status);
72 if (!v_status.success) {
74 } else if (num_entries < 0) {
75 v_status.tdb_error = True;
76 v_status.success = False;
80 DEBUG(10,("tdb_validate_child: tdb %s is good with %d entries\n",
81 tdb_name(tdb), num_entries));
82 ret = 0; /* Cache is good. */
85 DEBUG(10, ("tdb_validate_child: summary of validation status:\n"));
86 DEBUGADD(10,(" * tdb error: %s\n", v_status.tdb_error ? "yes" : "no"));
87 DEBUGADD(10,(" * bad freelist: %s\n",v_status.bad_freelist?"yes":"no"));
88 DEBUGADD(10,(" * bad entry: %s\n", v_status.bad_entry ? "yes" : "no"));
89 DEBUGADD(10,(" * unknown key: %s\n", v_status.unknown_key?"yes":"no"));
90 DEBUGADD(10,(" => overall success: %s\n", v_status.success?"yes":"no"));
96 * tdb validation function.
97 * returns 0 if tdb is ok, != 0 if it isn't.
98 * this function expects an opened tdb.
100 int tdb_validate(struct tdb_context *tdb, tdb_validate_data_func validate_fn)
102 pid_t child_pid = -1;
103 int child_status = 0;
108 DEBUG(1, ("Error: tdb_validate called with tdb == NULL\n"));
112 DEBUG(5, ("tdb_validate called for tdb '%s'\n", tdb_name(tdb)));
114 /* fork and let the child do the validation.
115 * benefit: no need to twist signal handlers and panic functions.
116 * just let the child panic. we catch the signal. */
118 DEBUG(10, ("tdb_validate: forking to let child do validation.\n"));
120 if (child_pid == 0) {
122 DEBUG(10, ("tdb_validate (validation child): created\n"));
123 DEBUG(10, ("tdb_validate (validation child): "
124 "calling tdb_validate_child\n"));
125 exit(tdb_validate_child(tdb, validate_fn));
127 else if (child_pid < 0) {
128 DEBUG(1, ("tdb_validate: fork for validation failed.\n"));
134 DEBUG(10, ("tdb_validate: fork succeeded, child PID = %u\n",
135 (unsigned int)child_pid));
137 DEBUG(10, ("tdb_validate: waiting for child to finish...\n"));
138 while ((wait_pid = sys_waitpid(child_pid, &child_status, 0)) < 0) {
139 if (errno == EINTR) {
140 DEBUG(10, ("tdb_validate: got signal during waitpid, "
145 DEBUG(1, ("tdb_validate: waitpid failed with error '%s'.\n",
149 if (wait_pid != child_pid) {
150 DEBUG(1, ("tdb_validate: waitpid returned pid %d, "
151 "but %u was expected\n", wait_pid, (unsigned int)child_pid));
155 DEBUG(10, ("tdb_validate: validating child returned.\n"));
156 if (WIFEXITED(child_status)) {
157 DEBUG(10, ("tdb_validate: child exited, code %d.\n",
158 WEXITSTATUS(child_status)));
159 ret = WEXITSTATUS(child_status);
161 if (WIFSIGNALED(child_status)) {
162 DEBUG(10, ("tdb_validate: child terminated by signal %d\n",
163 WTERMSIG(child_status)));
165 if (WCOREDUMP(child_status)) {
166 DEBUGADD(10, ("core dumped\n"));
169 ret = WTERMSIG(child_status);
171 if (WIFSTOPPED(child_status)) {
172 DEBUG(10, ("tdb_validate: child was stopped by signal %d\n",
173 WSTOPSIG(child_status)));
174 ret = WSTOPSIG(child_status);
178 DEBUG(5, ("tdb_validate returning code '%d' for tdb '%s'\n", ret,
185 * tdb validation function.
186 * returns 0 if tdb is ok, != 0 if it isn't.
187 * this is a wrapper around the actual validation function that opens and closes
190 int tdb_validate_open(const char *tdb_path, tdb_validate_data_func validate_fn)
192 TDB_CONTEXT *tdb = NULL;
195 DEBUG(5, ("tdb_validate_open called for tdb '%s'\n", tdb_path));
197 tdb = tdb_open_log(tdb_path, 0, TDB_DEFAULT, O_RDWR, 0);
199 DEBUG(1, ("Error opening tdb %s\n", tdb_path));
203 ret = tdb_validate(tdb, validate_fn);
209 * tdb backup function and helpers for tdb_validate wrapper with backup
213 /* this structure eliminates the need for a global overall status for
214 * the traverse-copy */
215 struct tdb_copy_data {
216 struct tdb_context *dst;
220 static int traverse_copy_fn(struct tdb_context *tdb, TDB_DATA key,
221 TDB_DATA dbuf, void *private_data)
223 struct tdb_copy_data *data = (struct tdb_copy_data *)private_data;
225 if (tdb_store(data->dst, key, dbuf, TDB_INSERT) != 0) {
226 DEBUG(4, ("Failed to insert into %s: %s\n", tdb_name(data->dst),
228 data->success = False;
234 static int tdb_copy(struct tdb_context *src, struct tdb_context *dst)
236 struct tdb_copy_data data;
242 count = tdb_traverse(src, traverse_copy_fn, (void *)(&data));
243 if ((count < 0) || (data.success == False)) {
249 static int tdb_verify_basic(struct tdb_context *tdb)
251 return tdb_traverse(tdb, NULL, NULL);
254 /* this backup function is essentially taken from lib/tdb/tools/tdbbackup.tdb
256 static int tdb_backup(TALLOC_CTX *ctx, const char *src_path,
257 const char *dst_path, int hash_size)
259 struct tdb_context *src_tdb = NULL;
260 struct tdb_context *dst_tdb = NULL;
261 char *tmp_path = NULL;
267 if (stat(src_path, &st) != 0) {
268 DEBUG(3, ("Could not stat '%s': %s\n", src_path,
273 /* open old tdb RDWR - so we can lock it */
274 src_tdb = tdb_open_log(src_path, 0, TDB_DEFAULT, O_RDWR, 0);
275 if (src_tdb == NULL) {
276 DEBUG(3, ("Failed to open tdb '%s'\n", src_path));
280 if (tdb_lockall(src_tdb) != 0) {
281 DEBUG(3, ("Failed to lock tdb '%s'\n", src_path));
285 tmp_path = talloc_asprintf(ctx, "%s%s", dst_path, ".tmp");
287 DEBUG(3, ("talloc fail\n"));
294 hash_size = tdb_hash_size(src_tdb);
297 dst_tdb = tdb_open_log(tmp_path, hash_size,
298 TDB_DEFAULT, O_RDWR | O_CREAT | O_EXCL,
300 if (dst_tdb == NULL) {
301 DEBUG(3, ("Error creating tdb '%s': %s\n", tmp_path,
308 count1 = tdb_copy(src_tdb, dst_tdb);
310 DEBUG(3, ("Failed to copy tdb '%s': %s\n", src_path,
316 /* reopen ro and do basic verification */
318 dst_tdb = tdb_open_log(tmp_path, 0, TDB_DEFAULT, O_RDONLY, 0);
320 DEBUG(3, ("Failed to reopen tdb '%s': %s\n", tmp_path,
324 count2 = tdb_verify_basic(dst_tdb);
325 if (count2 != count1) {
326 DEBUG(3, ("Failed to verify result of copying tdb '%s'.\n",
332 DEBUG(10, ("tdb_backup: successfully copied %d entries\n", count1));
334 /* make sure the new tdb has reached stable storage
335 * then rename it to its destination */
336 fsync(tdb_fd(dst_tdb));
339 if (rename(tmp_path, dst_path) != 0) {
340 DEBUG(3, ("Failed to rename '%s' to '%s': %s\n",
341 tmp_path, dst_path, strerror(errno)));
349 if (src_tdb != NULL) {
352 if (tmp_path != NULL) {
354 TALLOC_FREE(tmp_path);
356 if (saved_errno != 0) {
362 static int rename_file_with_suffix(TALLOC_CTX *ctx, const char *path,
368 dst_path = talloc_asprintf(ctx, "%s%s", path, suffix);
369 if (dst_path == NULL) {
370 DEBUG(3, ("error out of memory\n"));
374 ret = (rename(path, dst_path) != 0);
377 DEBUG(5, ("moved '%s' to '%s'\n", path, dst_path));
378 } else if (errno == ENOENT) {
379 DEBUG(3, ("file '%s' does not exist - so not moved\n", path));
382 DEBUG(3, ("error renaming %s to %s: %s\n", path, dst_path,
386 TALLOC_FREE(dst_path);
391 * do a backup of a tdb, moving the destination out of the way first
393 static int tdb_backup_with_rotate(TALLOC_CTX *ctx, const char *src_path,
394 const char *dst_path, int hash_size,
395 const char *rotate_suffix,
396 bool retry_norotate_if_nospc,
397 bool rename_as_last_resort_if_nospc)
401 rename_file_with_suffix(ctx, dst_path, rotate_suffix);
403 ret = tdb_backup(ctx, src_path, dst_path, hash_size);
406 DEBUG(10, ("backup of %s failed: %s\n", src_path, strerror(errno)));
408 if ((ret != 0) && (errno == ENOSPC) && retry_norotate_if_nospc)
410 char *rotate_path = talloc_asprintf(ctx, "%s%s", dst_path,
412 if (rotate_path == NULL) {
413 DEBUG(10, ("talloc fail\n"));
416 DEBUG(10, ("backup of %s failed due to lack of space\n",
418 DEBUGADD(10, ("trying to free some space by removing rotated "
419 "dst %s\n", rotate_path));
420 if (unlink(rotate_path) == -1) {
421 DEBUG(10, ("unlink of %s failed: %s\n", rotate_path,
424 ret = tdb_backup(ctx, src_path, dst_path, hash_size);
426 TALLOC_FREE(rotate_path);
429 if ((ret != 0) && (errno == ENOSPC) && rename_as_last_resort_if_nospc)
431 DEBUG(10, ("backup of %s failed due to lack of space\n",
433 DEBUGADD(10, ("using 'rename' as a last resort\n"));
434 ret = rename(src_path, dst_path);
441 * validation function with backup handling:
443 * - calls tdb_validate
444 * - if the tdb is ok, create a backup "name.bak", possibly moving
445 * existing backup to name.bak.old,
446 * return 0 (success) even if the backup fails
447 * - if the tdb is corrupt:
448 * - move the tdb to "name.corrupt"
449 * - check if there is valid backup.
450 * if so, restore the backup.
451 * if restore is successful, return 0 (success),
452 * - otherwise return -1 (failure)
454 int tdb_validate_and_backup(const char *tdb_path,
455 tdb_validate_data_func validate_fn)
458 const char *backup_suffix = ".bak";
459 const char *corrupt_suffix = ".corrupt";
460 const char *rotate_suffix = ".old";
461 char *tdb_path_backup;
463 TALLOC_CTX *ctx = NULL;
465 ctx = talloc_new(NULL);
467 DEBUG(0, ("tdb_validate_and_backup: out of memory\n"));
471 tdb_path_backup = talloc_asprintf(ctx, "%s%s", tdb_path, backup_suffix);
472 if (!tdb_path_backup) {
473 DEBUG(0, ("tdb_validate_and_backup: out of memory\n"));
477 ret = tdb_validate_open(tdb_path, validate_fn);
480 DEBUG(1, ("tdb '%s' is valid\n", tdb_path));
481 ret = tdb_backup_with_rotate(ctx, tdb_path, tdb_path_backup, 0,
482 rotate_suffix, True, False);
484 DEBUG(1, ("Error creating backup of tdb '%s'\n",
486 /* the actual validation was successful: */
489 DEBUG(1, ("Created backup '%s' of tdb '%s'\n",
490 tdb_path_backup, tdb_path));
493 DEBUG(1, ("tdb '%s' is invalid\n", tdb_path));
495 ret =stat(tdb_path_backup, &st);
497 DEBUG(5, ("Could not stat '%s': %s\n", tdb_path_backup,
499 DEBUG(1, ("No backup found.\n"));
501 DEBUG(1, ("backup '%s' found.\n", tdb_path_backup));
502 ret = tdb_validate_open(tdb_path_backup, validate_fn);
504 DEBUG(1, ("Backup '%s' is invalid.\n",
510 int renamed = rename_file_with_suffix(ctx, tdb_path,
513 DEBUG(1, ("Error moving tdb to '%s%s'\n",
514 tdb_path, corrupt_suffix));
516 DEBUG(1, ("Corrupt tdb stored as '%s%s'\n",
517 tdb_path, corrupt_suffix));
522 DEBUG(1, ("valid backup '%s' found\n", tdb_path_backup));
523 ret = tdb_backup_with_rotate(ctx, tdb_path_backup, tdb_path, 0,
524 corrupt_suffix, True, True);
526 DEBUG(1, ("Error restoring backup from '%s'\n",
529 DEBUG(1, ("Restored tdb backup from '%s'\n",