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/>.
20 #include <ccan/asprintf/asprintf.h>
24 static enum TDB_ERROR update_rec_hdr(struct tdb_context *tdb,
28 struct tdb_used_record *rec,
31 uint64_t dataroom = rec_data_length(rec) + rec_extra_padding(rec);
34 ecode = set_header(tdb, rec, TDB_USED_MAGIC, keylen, datalen,
35 keylen + dataroom, h);
36 if (ecode == TDB_SUCCESS) {
37 ecode = tdb_write_convert(tdb, off, rec, sizeof(*rec));
42 static enum TDB_ERROR replace_data(struct tdb_context *tdb,
44 struct tdb_data key, struct tdb_data dbuf,
45 tdb_off_t old_off, tdb_len_t old_room,
51 /* Allocate a new record. */
52 new_off = alloc(tdb, key.dsize, dbuf.dsize, h->h, TDB_USED_MAGIC,
54 if (TDB_OFF_IS_ERR(new_off)) {
58 /* We didn't like the existing one: remove it. */
61 ecode = add_free_record(tdb, old_off,
62 sizeof(struct tdb_used_record)
63 + key.dsize + old_room,
65 if (ecode == TDB_SUCCESS)
66 ecode = replace_in_hash(tdb, h, new_off);
68 ecode = add_to_hash(tdb, h, new_off);
70 if (ecode != TDB_SUCCESS) {
74 new_off += sizeof(struct tdb_used_record);
75 ecode = tdb->tdb2.io->twrite(tdb, new_off, key.dptr, key.dsize);
76 if (ecode != TDB_SUCCESS) {
81 ecode = tdb->tdb2.io->twrite(tdb, new_off, dbuf.dptr, dbuf.dsize);
82 if (ecode != TDB_SUCCESS) {
86 if (tdb->flags & TDB_SEQNUM)
92 static enum TDB_ERROR update_data(struct tdb_context *tdb,
99 ecode = tdb->tdb2.io->twrite(tdb, off, dbuf.dptr, dbuf.dsize);
100 if (ecode == TDB_SUCCESS && extra) {
101 /* Put a zero in; future versions may append other data. */
102 ecode = tdb->tdb2.io->twrite(tdb, off + dbuf.dsize, "", 1);
104 if (tdb->flags & TDB_SEQNUM)
110 enum TDB_ERROR tdb_store(struct tdb_context *tdb,
111 struct tdb_data key, struct tdb_data dbuf, int flag)
115 tdb_len_t old_room = 0;
116 struct tdb_used_record rec;
117 enum TDB_ERROR ecode;
119 if (tdb->flags & TDB_VERSION1) {
120 if (tdb1_store(tdb, key, dbuf, flag) == -1)
121 return tdb->last_error;
125 off = find_and_lock(tdb, key, F_WRLCK, &h, &rec, NULL);
126 if (TDB_OFF_IS_ERR(off)) {
127 return tdb->last_error = off;
130 /* Now we have lock on this hash bucket. */
131 if (flag == TDB_INSERT) {
133 ecode = TDB_ERR_EXISTS;
138 old_room = rec_data_length(&rec)
139 + rec_extra_padding(&rec);
140 if (old_room >= dbuf.dsize) {
141 /* Can modify in-place. Easy! */
142 ecode = update_rec_hdr(tdb, off,
143 key.dsize, dbuf.dsize,
145 if (ecode != TDB_SUCCESS) {
148 ecode = update_data(tdb,
151 old_room - dbuf.dsize);
152 if (ecode != TDB_SUCCESS) {
155 tdb_unlock_hashes(tdb, h.hlock_start,
156 h.hlock_range, F_WRLCK);
157 return tdb->last_error = TDB_SUCCESS;
160 if (flag == TDB_MODIFY) {
161 /* if the record doesn't exist and we
162 are in TDB_MODIFY mode then we should fail
164 ecode = TDB_ERR_NOEXIST;
170 /* If we didn't use the old record, this implies we're growing. */
171 ecode = replace_data(tdb, &h, key, dbuf, off, old_room, off);
173 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_WRLCK);
174 return tdb->last_error = ecode;
177 enum TDB_ERROR tdb_append(struct tdb_context *tdb,
178 struct tdb_data key, struct tdb_data dbuf)
182 struct tdb_used_record rec;
183 tdb_len_t old_room = 0, old_dlen;
184 unsigned char *newdata;
185 struct tdb_data new_dbuf;
186 enum TDB_ERROR ecode;
188 off = find_and_lock(tdb, key, F_WRLCK, &h, &rec, NULL);
189 if (TDB_OFF_IS_ERR(off)) {
190 return tdb->last_error = off;
194 old_dlen = rec_data_length(&rec);
195 old_room = old_dlen + rec_extra_padding(&rec);
197 /* Fast path: can append in place. */
198 if (rec_extra_padding(&rec) >= dbuf.dsize) {
199 ecode = update_rec_hdr(tdb, off, key.dsize,
200 old_dlen + dbuf.dsize, &rec,
202 if (ecode != TDB_SUCCESS) {
206 off += sizeof(rec) + key.dsize + old_dlen;
207 ecode = update_data(tdb, off, dbuf,
208 rec_extra_padding(&rec));
213 newdata = malloc(key.dsize + old_dlen + dbuf.dsize);
215 ecode = tdb_logerr(tdb, TDB_ERR_OOM, TDB_LOG_ERROR,
217 " failed to allocate %zu bytes",
218 (size_t)(key.dsize + old_dlen
222 ecode = tdb->tdb2.io->tread(tdb, off + sizeof(rec) + key.dsize,
224 if (ecode != TDB_SUCCESS) {
225 goto out_free_newdata;
227 memcpy(newdata + old_dlen, dbuf.dptr, dbuf.dsize);
228 new_dbuf.dptr = newdata;
229 new_dbuf.dsize = old_dlen + dbuf.dsize;
235 /* If they're using tdb_append(), it implies they're growing record. */
236 ecode = replace_data(tdb, &h, key, new_dbuf, off, old_room, true);
241 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_WRLCK);
242 return tdb->last_error = ecode;
245 enum TDB_ERROR tdb_fetch(struct tdb_context *tdb, struct tdb_data key,
246 struct tdb_data *data)
249 struct tdb_used_record rec;
251 enum TDB_ERROR ecode;
253 if (tdb->flags & TDB_VERSION1)
254 return tdb1_fetch(tdb, key, data);
256 off = find_and_lock(tdb, key, F_RDLCK, &h, &rec, NULL);
257 if (TDB_OFF_IS_ERR(off)) {
258 return tdb->last_error = off;
262 ecode = TDB_ERR_NOEXIST;
264 data->dsize = rec_data_length(&rec);
265 data->dptr = tdb_alloc_read(tdb, off + sizeof(rec) + key.dsize,
267 if (TDB_PTR_IS_ERR(data->dptr)) {
268 ecode = TDB_PTR_ERR(data->dptr);
273 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_RDLCK);
274 return tdb->last_error = ecode;
277 bool tdb_exists(struct tdb_context *tdb, TDB_DATA key)
280 struct tdb_used_record rec;
283 off = find_and_lock(tdb, key, F_RDLCK, &h, &rec, NULL);
284 if (TDB_OFF_IS_ERR(off)) {
285 tdb->last_error = off;
288 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_RDLCK);
290 tdb->last_error = TDB_SUCCESS;
291 return off ? true : false;
294 enum TDB_ERROR tdb_delete(struct tdb_context *tdb, struct tdb_data key)
297 struct tdb_used_record rec;
299 enum TDB_ERROR ecode;
301 off = find_and_lock(tdb, key, F_WRLCK, &h, &rec, NULL);
302 if (TDB_OFF_IS_ERR(off)) {
303 return tdb->last_error = off;
307 ecode = TDB_ERR_NOEXIST;
311 ecode = delete_from_hash(tdb, &h);
312 if (ecode != TDB_SUCCESS) {
316 /* Free the deleted entry. */
318 ecode = add_free_record(tdb, off,
319 sizeof(struct tdb_used_record)
320 + rec_key_length(&rec)
321 + rec_data_length(&rec)
322 + rec_extra_padding(&rec),
323 TDB_LOCK_WAIT, true);
325 if (tdb->flags & TDB_SEQNUM)
329 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_WRLCK);
330 return tdb->last_error = ecode;
333 unsigned int tdb_get_flags(struct tdb_context *tdb)
338 static bool inside_transaction(const struct tdb_context *tdb)
340 if (tdb->flags & TDB_VERSION1)
341 return tdb->tdb1.transaction != NULL;
343 return tdb->tdb2.transaction != NULL;
346 static bool readonly_changable(struct tdb_context *tdb, const char *caller)
348 if (inside_transaction(tdb)) {
349 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
352 " TDB_RDONLY inside transaction",
357 if (tdb->file->allrecord_lock.count != 0
358 || tdb->file->num_lockrecs != 0) {
359 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
362 " TDB_RDONLY holding locks",
369 void tdb_add_flag(struct tdb_context *tdb, unsigned flag)
371 if (tdb->flags & TDB_INTERNAL) {
372 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
374 "tdb_add_flag: internal db");
379 tdb->flags |= TDB_NOLOCK;
382 tdb->flags |= TDB_NOMMAP;
383 tdb_munmap(tdb->file);
386 tdb->flags |= TDB_NOSYNC;
389 tdb->flags |= TDB_SEQNUM;
391 case TDB_ALLOW_NESTING:
392 tdb->flags |= TDB_ALLOW_NESTING;
395 if (readonly_changable(tdb, "tdb_add_flag"))
396 tdb->flags |= TDB_RDONLY;
399 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
401 "tdb_add_flag: Unknown flag %u",
406 void tdb_remove_flag(struct tdb_context *tdb, unsigned flag)
408 if (tdb->flags & TDB_INTERNAL) {
409 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
411 "tdb_remove_flag: internal db");
416 tdb->flags &= ~TDB_NOLOCK;
419 tdb->flags &= ~TDB_NOMMAP;
423 tdb->flags &= ~TDB_NOSYNC;
426 tdb->flags &= ~TDB_SEQNUM;
428 case TDB_ALLOW_NESTING:
429 tdb->flags &= ~TDB_ALLOW_NESTING;
432 if ((tdb->open_flags & O_ACCMODE) == O_RDONLY) {
433 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
435 "tdb_remove_flag: can't"
436 " remove TDB_RDONLY on tdb"
437 " opened with O_RDONLY");
440 if (readonly_changable(tdb, "tdb_remove_flag"))
441 tdb->flags &= ~TDB_RDONLY;
444 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
446 "tdb_remove_flag: Unknown flag %u",
451 const char *tdb_errorstr(enum TDB_ERROR ecode)
453 /* Gcc warns if you miss a case in the switch, so use that. */
455 case TDB_SUCCESS: return "Success";
456 case TDB_ERR_CORRUPT: return "Corrupt database";
457 case TDB_ERR_IO: return "IO Error";
458 case TDB_ERR_LOCK: return "Locking error";
459 case TDB_ERR_OOM: return "Out of memory";
460 case TDB_ERR_EXISTS: return "Record exists";
461 case TDB_ERR_EINVAL: return "Invalid parameter";
462 case TDB_ERR_NOEXIST: return "Record does not exist";
463 case TDB_ERR_RDONLY: return "write not permitted";
465 return "Invalid error code";
468 enum TDB_ERROR tdb_error(struct tdb_context *tdb)
470 return tdb->last_error;
473 enum TDB_ERROR COLD tdb_logerr(struct tdb_context *tdb,
474 enum TDB_ERROR ecode,
475 enum tdb_log_level level,
476 const char *fmt, ...)
481 /* tdb_open paths care about errno, so save it. */
482 int saved_errno = errno;
488 len = vasprintf(&message, fmt, ap);
492 tdb->log_fn(tdb, TDB_LOG_ERROR, TDB_ERR_OOM,
493 "out of memory formatting message:", tdb->log_data);
494 tdb->log_fn(tdb, level, ecode, fmt, tdb->log_data);
496 tdb->log_fn(tdb, level, ecode, message, tdb->log_data);
503 enum TDB_ERROR tdb_parse_record_(struct tdb_context *tdb,
505 enum TDB_ERROR (*parse)(TDB_DATA k,
511 struct tdb_used_record rec;
513 enum TDB_ERROR ecode;
515 off = find_and_lock(tdb, key, F_RDLCK, &h, &rec, NULL);
516 if (TDB_OFF_IS_ERR(off)) {
517 return tdb->last_error = off;
521 ecode = TDB_ERR_NOEXIST;
524 dptr = tdb_access_read(tdb, off + sizeof(rec) + key.dsize,
525 rec_data_length(&rec), false);
526 if (TDB_PTR_IS_ERR(dptr)) {
527 ecode = TDB_PTR_ERR(dptr);
529 TDB_DATA d = tdb_mkdata(dptr, rec_data_length(&rec));
531 ecode = parse(key, d, data);
532 tdb_access_release(tdb, dptr);
536 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_RDLCK);
537 return tdb->last_error = ecode;
540 const char *tdb_name(const struct tdb_context *tdb)
545 int64_t tdb_get_seqnum(struct tdb_context *tdb)
547 tdb_off_t off = tdb_read_off(tdb, offsetof(struct tdb_header, seqnum));
548 if (TDB_OFF_IS_ERR(off))
549 tdb->last_error = off;
551 tdb->last_error = TDB_SUCCESS;
556 int tdb_fd(const struct tdb_context *tdb)
558 return tdb->file->fd;