docs: Fix variable list in man vfs_crossrename.
[Samba.git] / lib / ntdb / ntdb.c
blob5d56b33b5a123a5da087741b5e1cc03e5de6a511
1 /*
2 Trivial Database 2: fetch, store and misc routines.
3 Copyright (C) Rusty Russell 2010
5 This library is free software; you can redistribute it and/or
6 modify it under the terms of the GNU Lesser General Public
7 License as published by the Free Software Foundation; either
8 version 3 of the License, or (at your option) any later version.
10 This library is distributed in the hope that it will be useful,
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 Lesser General Public License for more details.
15 You should have received a copy of the GNU Lesser General Public
16 License along with this library; if not, see <http://www.gnu.org/licenses/>.
18 #include "private.h"
19 #ifndef HAVE_LIBREPLACE
20 #include <ccan/asprintf/asprintf.h>
21 #include <stdarg.h>
22 #endif
24 static enum NTDB_ERROR update_rec_hdr(struct ntdb_context *ntdb,
25 ntdb_off_t off,
26 ntdb_len_t keylen,
27 ntdb_len_t datalen,
28 struct ntdb_used_record *rec)
30 uint64_t dataroom = rec_data_length(rec) + rec_extra_padding(rec);
31 enum NTDB_ERROR ecode;
33 ecode = set_header(ntdb, rec, NTDB_USED_MAGIC, keylen, datalen,
34 keylen + dataroom);
35 if (ecode == NTDB_SUCCESS) {
36 ecode = ntdb_write_convert(ntdb, off, rec, sizeof(*rec));
38 return ecode;
41 static enum NTDB_ERROR replace_data(struct ntdb_context *ntdb,
42 struct hash_info *h,
43 NTDB_DATA key, NTDB_DATA dbuf,
44 ntdb_off_t old_off, ntdb_len_t old_room,
45 bool growing)
47 ntdb_off_t new_off;
48 enum NTDB_ERROR ecode;
50 /* Allocate a new record. */
51 new_off = alloc(ntdb, key.dsize, dbuf.dsize, NTDB_USED_MAGIC, growing);
52 if (NTDB_OFF_IS_ERR(new_off)) {
53 return NTDB_OFF_TO_ERR(new_off);
56 /* We didn't like the existing one: remove it. */
57 if (old_off) {
58 ntdb->stats.frees++;
59 ecode = add_free_record(ntdb, old_off,
60 sizeof(struct ntdb_used_record)
61 + key.dsize + old_room,
62 NTDB_LOCK_WAIT, true);
63 if (ecode == NTDB_SUCCESS)
64 ecode = replace_in_hash(ntdb, h, new_off);
65 } else {
66 ecode = add_to_hash(ntdb, h, new_off);
68 if (ecode != NTDB_SUCCESS) {
69 return ecode;
72 new_off += sizeof(struct ntdb_used_record);
73 ecode = ntdb->io->twrite(ntdb, new_off, key.dptr, key.dsize);
74 if (ecode != NTDB_SUCCESS) {
75 return ecode;
78 new_off += key.dsize;
79 ecode = ntdb->io->twrite(ntdb, new_off, dbuf.dptr, dbuf.dsize);
80 if (ecode != NTDB_SUCCESS) {
81 return ecode;
84 if (ntdb->flags & NTDB_SEQNUM)
85 ntdb_inc_seqnum(ntdb);
87 return NTDB_SUCCESS;
90 static enum NTDB_ERROR update_data(struct ntdb_context *ntdb,
91 ntdb_off_t off,
92 NTDB_DATA dbuf,
93 ntdb_len_t extra)
95 enum NTDB_ERROR ecode;
97 ecode = ntdb->io->twrite(ntdb, off, dbuf.dptr, dbuf.dsize);
98 if (ecode == NTDB_SUCCESS && extra) {
99 /* Put a zero in; future versions may append other data. */
100 ecode = ntdb->io->twrite(ntdb, off + dbuf.dsize, "", 1);
102 if (ntdb->flags & NTDB_SEQNUM)
103 ntdb_inc_seqnum(ntdb);
105 return ecode;
108 _PUBLIC_ enum NTDB_ERROR ntdb_store(struct ntdb_context *ntdb,
109 NTDB_DATA key, NTDB_DATA dbuf, int flag)
111 struct hash_info h;
112 ntdb_off_t off;
113 ntdb_len_t old_room = 0;
114 struct ntdb_used_record rec;
115 enum NTDB_ERROR ecode;
117 off = find_and_lock(ntdb, key, F_WRLCK, &h, &rec, NULL);
118 if (NTDB_OFF_IS_ERR(off)) {
119 return NTDB_OFF_TO_ERR(off);
122 /* Now we have lock on this hash bucket. */
123 if (flag == NTDB_INSERT) {
124 if (off) {
125 ecode = NTDB_ERR_EXISTS;
126 goto out;
128 } else {
129 if (off) {
130 old_room = rec_data_length(&rec)
131 + rec_extra_padding(&rec);
132 if (old_room >= dbuf.dsize) {
133 /* Can modify in-place. Easy! */
134 ecode = update_rec_hdr(ntdb, off,
135 key.dsize, dbuf.dsize,
136 &rec);
137 if (ecode != NTDB_SUCCESS) {
138 goto out;
140 ecode = update_data(ntdb,
141 off + sizeof(rec)
142 + key.dsize, dbuf,
143 old_room - dbuf.dsize);
144 if (ecode != NTDB_SUCCESS) {
145 goto out;
147 ntdb_unlock_hash(ntdb, h.h, F_WRLCK);
148 return NTDB_SUCCESS;
150 } else {
151 if (flag == NTDB_MODIFY) {
152 /* if the record doesn't exist and we
153 are in NTDB_MODIFY mode then we should fail
154 the store */
155 ecode = NTDB_ERR_NOEXIST;
156 goto out;
161 /* If we didn't use the old record, this implies we're growing. */
162 ecode = replace_data(ntdb, &h, key, dbuf, off, old_room, off);
163 out:
164 ntdb_unlock_hash(ntdb, h.h, F_WRLCK);
165 return ecode;
168 _PUBLIC_ enum NTDB_ERROR ntdb_append(struct ntdb_context *ntdb,
169 NTDB_DATA key, NTDB_DATA dbuf)
171 struct hash_info h;
172 ntdb_off_t off;
173 struct ntdb_used_record rec;
174 ntdb_len_t old_room = 0, old_dlen;
175 unsigned char *newdata;
176 NTDB_DATA new_dbuf;
177 enum NTDB_ERROR ecode;
179 off = find_and_lock(ntdb, key, F_WRLCK, &h, &rec, NULL);
180 if (NTDB_OFF_IS_ERR(off)) {
181 return NTDB_OFF_TO_ERR(off);
184 if (off) {
185 old_dlen = rec_data_length(&rec);
186 old_room = old_dlen + rec_extra_padding(&rec);
188 /* Fast path: can append in place. */
189 if (rec_extra_padding(&rec) >= dbuf.dsize) {
190 ecode = update_rec_hdr(ntdb, off, key.dsize,
191 old_dlen + dbuf.dsize, &rec);
192 if (ecode != NTDB_SUCCESS) {
193 goto out;
196 off += sizeof(rec) + key.dsize + old_dlen;
197 ecode = update_data(ntdb, off, dbuf,
198 rec_extra_padding(&rec));
199 goto out;
202 /* Slow path. */
203 newdata = ntdb->alloc_fn(ntdb, key.dsize + old_dlen + dbuf.dsize,
204 ntdb->alloc_data);
205 if (!newdata) {
206 ecode = ntdb_logerr(ntdb, NTDB_ERR_OOM, NTDB_LOG_ERROR,
207 "ntdb_append:"
208 " failed to allocate %zu bytes",
209 (size_t)(key.dsize + old_dlen
210 + dbuf.dsize));
211 goto out;
213 ecode = ntdb->io->tread(ntdb, off + sizeof(rec) + key.dsize,
214 newdata, old_dlen);
215 if (ecode != NTDB_SUCCESS) {
216 goto out_free_newdata;
218 memcpy(newdata + old_dlen, dbuf.dptr, dbuf.dsize);
219 new_dbuf.dptr = newdata;
220 new_dbuf.dsize = old_dlen + dbuf.dsize;
221 } else {
222 newdata = NULL;
223 new_dbuf = dbuf;
226 /* If they're using ntdb_append(), it implies they're growing record. */
227 ecode = replace_data(ntdb, &h, key, new_dbuf, off, old_room, true);
229 out_free_newdata:
230 ntdb->free_fn(newdata, ntdb->alloc_data);
231 out:
232 ntdb_unlock_hash(ntdb, h.h, F_WRLCK);
233 return ecode;
236 _PUBLIC_ enum NTDB_ERROR ntdb_fetch(struct ntdb_context *ntdb, NTDB_DATA key,
237 NTDB_DATA *data)
239 ntdb_off_t off;
240 struct ntdb_used_record rec;
241 struct hash_info h;
242 enum NTDB_ERROR ecode;
243 const char *keyp;
245 off = find_and_lock(ntdb, key, F_RDLCK, &h, &rec, &keyp);
246 if (NTDB_OFF_IS_ERR(off)) {
247 return NTDB_OFF_TO_ERR(off);
250 if (!off) {
251 ecode = NTDB_ERR_NOEXIST;
252 } else {
253 data->dsize = rec_data_length(&rec);
254 data->dptr = ntdb->alloc_fn(ntdb, data->dsize, ntdb->alloc_data);
255 if (unlikely(!data->dptr)) {
256 ecode = NTDB_ERR_OOM;
257 } else {
258 memcpy(data->dptr, keyp + key.dsize, data->dsize);
259 ecode = NTDB_SUCCESS;
261 ntdb_access_release(ntdb, keyp);
264 ntdb_unlock_hash(ntdb, h.h, F_RDLCK);
265 return ecode;
268 _PUBLIC_ bool ntdb_exists(struct ntdb_context *ntdb, NTDB_DATA key)
270 ntdb_off_t off;
271 struct ntdb_used_record rec;
272 struct hash_info h;
274 off = find_and_lock(ntdb, key, F_RDLCK, &h, &rec, NULL);
275 if (NTDB_OFF_IS_ERR(off)) {
276 return false;
278 ntdb_unlock_hash(ntdb, h.h, F_RDLCK);
280 return off ? true : false;
283 _PUBLIC_ enum NTDB_ERROR ntdb_delete(struct ntdb_context *ntdb, NTDB_DATA key)
285 ntdb_off_t off;
286 struct ntdb_used_record rec;
287 struct hash_info h;
288 enum NTDB_ERROR ecode;
290 off = find_and_lock(ntdb, key, F_WRLCK, &h, &rec, NULL);
291 if (NTDB_OFF_IS_ERR(off)) {
292 return NTDB_OFF_TO_ERR(off);
295 if (!off) {
296 ecode = NTDB_ERR_NOEXIST;
297 goto unlock;
300 ecode = delete_from_hash(ntdb, &h);
301 if (ecode != NTDB_SUCCESS) {
302 goto unlock;
305 /* Free the deleted entry. */
306 ntdb->stats.frees++;
307 ecode = add_free_record(ntdb, off,
308 sizeof(struct ntdb_used_record)
309 + rec_key_length(&rec)
310 + rec_data_length(&rec)
311 + rec_extra_padding(&rec),
312 NTDB_LOCK_WAIT, true);
314 if (ntdb->flags & NTDB_SEQNUM)
315 ntdb_inc_seqnum(ntdb);
317 unlock:
318 ntdb_unlock_hash(ntdb, h.h, F_WRLCK);
319 return ecode;
322 _PUBLIC_ unsigned int ntdb_get_flags(struct ntdb_context *ntdb)
324 return ntdb->flags;
327 static bool inside_transaction(const struct ntdb_context *ntdb)
329 return ntdb->transaction != NULL;
332 static bool readonly_changable(struct ntdb_context *ntdb, const char *caller)
334 if (inside_transaction(ntdb)) {
335 ntdb_logerr(ntdb, NTDB_ERR_EINVAL, NTDB_LOG_USE_ERROR,
336 "%s: can't change"
337 " NTDB_RDONLY inside transaction",
338 caller);
339 return false;
341 return true;
344 _PUBLIC_ void ntdb_add_flag(struct ntdb_context *ntdb, unsigned flag)
346 if (ntdb->flags & NTDB_INTERNAL) {
347 ntdb_logerr(ntdb, NTDB_ERR_EINVAL, NTDB_LOG_USE_ERROR,
348 "ntdb_add_flag: internal db");
349 return;
351 switch (flag) {
352 case NTDB_NOLOCK:
353 ntdb->flags |= NTDB_NOLOCK;
354 break;
355 case NTDB_NOMMAP:
356 if (ntdb->file->direct_count) {
357 ntdb_logerr(ntdb, NTDB_ERR_EINVAL, NTDB_LOG_USE_ERROR,
358 "ntdb_add_flag: Can't get NTDB_NOMMAP from"
359 " ntdb_parse_record!");
360 return;
362 ntdb->flags |= NTDB_NOMMAP;
363 #ifndef HAVE_INCOHERENT_MMAP
364 ntdb_munmap(ntdb);
365 #endif
366 break;
367 case NTDB_NOSYNC:
368 ntdb->flags |= NTDB_NOSYNC;
369 break;
370 case NTDB_SEQNUM:
371 ntdb->flags |= NTDB_SEQNUM;
372 break;
373 case NTDB_ALLOW_NESTING:
374 ntdb->flags |= NTDB_ALLOW_NESTING;
375 break;
376 case NTDB_RDONLY:
377 if (readonly_changable(ntdb, "ntdb_add_flag"))
378 ntdb->flags |= NTDB_RDONLY;
379 break;
380 default:
381 ntdb_logerr(ntdb, NTDB_ERR_EINVAL, NTDB_LOG_USE_ERROR,
382 "ntdb_add_flag: Unknown flag %u", flag);
386 _PUBLIC_ void ntdb_remove_flag(struct ntdb_context *ntdb, unsigned flag)
388 if (ntdb->flags & NTDB_INTERNAL) {
389 ntdb_logerr(ntdb, NTDB_ERR_EINVAL, NTDB_LOG_USE_ERROR,
390 "ntdb_remove_flag: internal db");
391 return;
393 switch (flag) {
394 case NTDB_NOLOCK:
395 ntdb->flags &= ~NTDB_NOLOCK;
396 break;
397 case NTDB_NOMMAP:
398 ntdb->flags &= ~NTDB_NOMMAP;
399 #ifndef HAVE_INCOHERENT_MMAP
400 /* If mmap incoherent, we were mmaping anyway. */
401 ntdb_mmap(ntdb);
402 #endif
403 break;
404 case NTDB_NOSYNC:
405 ntdb->flags &= ~NTDB_NOSYNC;
406 break;
407 case NTDB_SEQNUM:
408 ntdb->flags &= ~NTDB_SEQNUM;
409 break;
410 case NTDB_ALLOW_NESTING:
411 ntdb->flags &= ~NTDB_ALLOW_NESTING;
412 break;
413 case NTDB_RDONLY:
414 if ((ntdb->open_flags & O_ACCMODE) == O_RDONLY) {
415 ntdb_logerr(ntdb, NTDB_ERR_EINVAL, NTDB_LOG_USE_ERROR,
416 "ntdb_remove_flag: can't"
417 " remove NTDB_RDONLY on ntdb"
418 " opened with O_RDONLY");
419 break;
421 if (readonly_changable(ntdb, "ntdb_remove_flag"))
422 ntdb->flags &= ~NTDB_RDONLY;
423 break;
424 default:
425 ntdb_logerr(ntdb, NTDB_ERR_EINVAL, NTDB_LOG_USE_ERROR,
426 "ntdb_remove_flag: Unknown flag %u",
427 flag);
431 _PUBLIC_ const char *ntdb_errorstr(enum NTDB_ERROR ecode)
433 /* Gcc warns if you miss a case in the switch, so use that. */
434 switch (NTDB_ERR_TO_OFF(ecode)) {
435 case NTDB_ERR_TO_OFF(NTDB_SUCCESS): return "Success";
436 case NTDB_ERR_TO_OFF(NTDB_ERR_CORRUPT): return "Corrupt database";
437 case NTDB_ERR_TO_OFF(NTDB_ERR_IO): return "IO Error";
438 case NTDB_ERR_TO_OFF(NTDB_ERR_LOCK): return "Locking error";
439 case NTDB_ERR_TO_OFF(NTDB_ERR_OOM): return "Out of memory";
440 case NTDB_ERR_TO_OFF(NTDB_ERR_EXISTS): return "Record exists";
441 case NTDB_ERR_TO_OFF(NTDB_ERR_EINVAL): return "Invalid parameter";
442 case NTDB_ERR_TO_OFF(NTDB_ERR_NOEXIST): return "Record does not exist";
443 case NTDB_ERR_TO_OFF(NTDB_ERR_RDONLY): return "write not permitted";
445 return "Invalid error code";
448 enum NTDB_ERROR COLD ntdb_logerr(struct ntdb_context *ntdb,
449 enum NTDB_ERROR ecode,
450 enum ntdb_log_level level,
451 const char *fmt, ...)
453 char *message;
454 va_list ap;
455 size_t len;
456 /* ntdb_open paths care about errno, so save it. */
457 int saved_errno = errno;
459 if (!ntdb->log_fn)
460 return ecode;
462 va_start(ap, fmt);
463 len = vsnprintf(NULL, 0, fmt, ap);
464 va_end(ap);
466 message = ntdb->alloc_fn(ntdb, len + 1, ntdb->alloc_data);
467 if (!message) {
468 ntdb->log_fn(ntdb, NTDB_LOG_ERROR, NTDB_ERR_OOM,
469 "out of memory formatting message:", ntdb->log_data);
470 ntdb->log_fn(ntdb, level, ecode, fmt, ntdb->log_data);
471 } else {
472 va_start(ap, fmt);
473 vsnprintf(message, len+1, fmt, ap);
474 va_end(ap);
475 ntdb->log_fn(ntdb, level, ecode, message, ntdb->log_data);
476 ntdb->free_fn(message, ntdb->alloc_data);
478 errno = saved_errno;
479 return ecode;
482 _PUBLIC_ enum NTDB_ERROR ntdb_parse_record_(struct ntdb_context *ntdb,
483 NTDB_DATA key,
484 enum NTDB_ERROR (*parse)(NTDB_DATA k,
485 NTDB_DATA d,
486 void *data),
487 void *data)
489 ntdb_off_t off;
490 struct ntdb_used_record rec;
491 struct hash_info h;
492 enum NTDB_ERROR ecode;
493 const char *keyp;
495 off = find_and_lock(ntdb, key, F_RDLCK, &h, &rec, &keyp);
496 if (NTDB_OFF_IS_ERR(off)) {
497 return NTDB_OFF_TO_ERR(off);
500 if (!off) {
501 ecode = NTDB_ERR_NOEXIST;
502 } else {
503 unsigned int old_flags;
504 NTDB_DATA d = ntdb_mkdata(keyp + key.dsize,
505 rec_data_length(&rec));
508 * Make sure they don't try to write db, since they
509 * have read lock! They can if they've done
510 * ntdb_lockall(): if it was ntdb_lockall_read, that'll
511 * stop them doing a write operation anyway.
513 old_flags = ntdb->flags;
514 if (!ntdb->file->allrecord_lock.count &&
515 !(ntdb->flags & NTDB_NOLOCK)) {
516 ntdb->flags |= NTDB_RDONLY;
518 ecode = parse(key, d, data);
519 ntdb->flags = old_flags;
520 ntdb_access_release(ntdb, keyp);
523 ntdb_unlock_hash(ntdb, h.h, F_RDLCK);
524 return ecode;
527 _PUBLIC_ const char *ntdb_name(const struct ntdb_context *ntdb)
529 return ntdb->name;
532 _PUBLIC_ int64_t ntdb_get_seqnum(struct ntdb_context *ntdb)
534 return ntdb_read_off(ntdb, offsetof(struct ntdb_header, seqnum));
538 _PUBLIC_ int ntdb_fd(const struct ntdb_context *ntdb)
540 return ntdb->file->fd;
543 struct traverse_state {
544 enum NTDB_ERROR error;
545 struct ntdb_context *dest_db;
549 traverse function for repacking
551 static int repack_traverse(struct ntdb_context *ntdb, NTDB_DATA key, NTDB_DATA data,
552 struct traverse_state *state)
554 state->error = ntdb_store(state->dest_db, key, data, NTDB_INSERT);
555 if (state->error != NTDB_SUCCESS) {
556 return -1;
558 return 0;
561 _PUBLIC_ enum NTDB_ERROR ntdb_repack(struct ntdb_context *ntdb)
563 struct ntdb_context *tmp_db;
564 struct traverse_state state;
566 state.error = ntdb_transaction_start(ntdb);
567 if (state.error != NTDB_SUCCESS) {
568 return state.error;
571 tmp_db = ntdb_open("tmpdb", NTDB_INTERNAL, O_RDWR|O_CREAT, 0, NULL);
572 if (tmp_db == NULL) {
573 state.error = ntdb_logerr(ntdb, NTDB_ERR_OOM, NTDB_LOG_ERROR,
574 __location__
575 " Failed to create tmp_db");
576 ntdb_transaction_cancel(ntdb);
577 return state.error;
580 state.dest_db = tmp_db;
581 if (ntdb_traverse(ntdb, repack_traverse, &state) < 0) {
582 goto fail;
585 state.error = ntdb_wipe_all(ntdb);
586 if (state.error != NTDB_SUCCESS) {
587 goto fail;
590 state.dest_db = ntdb;
591 if (ntdb_traverse(tmp_db, repack_traverse, &state) < 0) {
592 goto fail;
595 ntdb_close(tmp_db);
596 return ntdb_transaction_commit(ntdb);
598 fail:
599 ntdb_transaction_cancel(ntdb);
600 ntdb_close(tmp_db);
601 return state.error;