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/>.
19 #include <ccan/asprintf/asprintf.h>
22 static enum TDB_ERROR update_rec_hdr(struct tdb_context *tdb,
26 struct tdb_used_record *rec,
29 uint64_t dataroom = rec_data_length(rec) + rec_extra_padding(rec);
32 ecode = set_header(tdb, rec, TDB_USED_MAGIC, keylen, datalen,
33 keylen + dataroom, h);
34 if (ecode == TDB_SUCCESS) {
35 ecode = tdb_write_convert(tdb, off, rec, sizeof(*rec));
40 static enum TDB_ERROR replace_data(struct tdb_context *tdb,
42 struct tdb_data key, struct tdb_data dbuf,
43 tdb_off_t old_off, tdb_len_t old_room,
49 /* Allocate a new record. */
50 new_off = alloc(tdb, key.dsize, dbuf.dsize, h->h, TDB_USED_MAGIC,
52 if (TDB_OFF_IS_ERR(new_off)) {
56 /* We didn't like the existing one: remove it. */
59 ecode = add_free_record(tdb, old_off,
60 sizeof(struct tdb_used_record)
61 + key.dsize + old_room,
63 if (ecode == TDB_SUCCESS)
64 ecode = replace_in_hash(tdb, h, new_off);
66 ecode = add_to_hash(tdb, h, new_off);
68 if (ecode != TDB_SUCCESS) {
72 new_off += sizeof(struct tdb_used_record);
73 ecode = tdb->tdb2.io->twrite(tdb, new_off, key.dptr, key.dsize);
74 if (ecode != TDB_SUCCESS) {
79 ecode = tdb->tdb2.io->twrite(tdb, new_off, dbuf.dptr, dbuf.dsize);
80 if (ecode != TDB_SUCCESS) {
84 if (tdb->flags & TDB_SEQNUM)
90 static enum TDB_ERROR update_data(struct tdb_context *tdb,
97 ecode = tdb->tdb2.io->twrite(tdb, off, dbuf.dptr, dbuf.dsize);
98 if (ecode == TDB_SUCCESS && extra) {
99 /* Put a zero in; future versions may append other data. */
100 ecode = tdb->tdb2.io->twrite(tdb, off + dbuf.dsize, "", 1);
102 if (tdb->flags & TDB_SEQNUM)
108 enum TDB_ERROR tdb_store(struct tdb_context *tdb,
109 struct tdb_data key, struct tdb_data dbuf, int flag)
113 tdb_len_t old_room = 0;
114 struct tdb_used_record rec;
115 enum TDB_ERROR ecode;
117 if (tdb->flags & TDB_VERSION1) {
118 if (tdb1_store(tdb, key, dbuf, flag) == -1)
119 return tdb->last_error;
123 off = find_and_lock(tdb, key, F_WRLCK, &h, &rec, NULL);
124 if (TDB_OFF_IS_ERR(off)) {
125 return tdb->last_error = off;
128 /* Now we have lock on this hash bucket. */
129 if (flag == TDB_INSERT) {
131 ecode = TDB_ERR_EXISTS;
136 old_room = rec_data_length(&rec)
137 + rec_extra_padding(&rec);
138 if (old_room >= dbuf.dsize) {
139 /* Can modify in-place. Easy! */
140 ecode = update_rec_hdr(tdb, off,
141 key.dsize, dbuf.dsize,
143 if (ecode != TDB_SUCCESS) {
146 ecode = update_data(tdb,
149 old_room - dbuf.dsize);
150 if (ecode != TDB_SUCCESS) {
153 tdb_unlock_hashes(tdb, h.hlock_start,
154 h.hlock_range, F_WRLCK);
155 return tdb->last_error = TDB_SUCCESS;
158 if (flag == TDB_MODIFY) {
159 /* if the record doesn't exist and we
160 are in TDB_MODIFY mode then we should fail
162 ecode = TDB_ERR_NOEXIST;
168 /* If we didn't use the old record, this implies we're growing. */
169 ecode = replace_data(tdb, &h, key, dbuf, off, old_room, off);
171 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_WRLCK);
172 return tdb->last_error = ecode;
175 enum TDB_ERROR tdb_append(struct tdb_context *tdb,
176 struct tdb_data key, struct tdb_data dbuf)
180 struct tdb_used_record rec;
181 tdb_len_t old_room = 0, old_dlen;
182 unsigned char *newdata;
183 struct tdb_data new_dbuf;
184 enum TDB_ERROR ecode;
186 off = find_and_lock(tdb, key, F_WRLCK, &h, &rec, NULL);
187 if (TDB_OFF_IS_ERR(off)) {
188 return tdb->last_error = off;
192 old_dlen = rec_data_length(&rec);
193 old_room = old_dlen + rec_extra_padding(&rec);
195 /* Fast path: can append in place. */
196 if (rec_extra_padding(&rec) >= dbuf.dsize) {
197 ecode = update_rec_hdr(tdb, off, key.dsize,
198 old_dlen + dbuf.dsize, &rec,
200 if (ecode != TDB_SUCCESS) {
204 off += sizeof(rec) + key.dsize + old_dlen;
205 ecode = update_data(tdb, off, dbuf,
206 rec_extra_padding(&rec));
211 newdata = malloc(key.dsize + old_dlen + dbuf.dsize);
213 ecode = tdb_logerr(tdb, TDB_ERR_OOM, TDB_LOG_ERROR,
215 " failed to allocate %zu bytes",
216 (size_t)(key.dsize + old_dlen
220 ecode = tdb->tdb2.io->tread(tdb, off + sizeof(rec) + key.dsize,
222 if (ecode != TDB_SUCCESS) {
223 goto out_free_newdata;
225 memcpy(newdata + old_dlen, dbuf.dptr, dbuf.dsize);
226 new_dbuf.dptr = newdata;
227 new_dbuf.dsize = old_dlen + dbuf.dsize;
233 /* If they're using tdb_append(), it implies they're growing record. */
234 ecode = replace_data(tdb, &h, key, new_dbuf, off, old_room, true);
239 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_WRLCK);
240 return tdb->last_error = ecode;
243 enum TDB_ERROR tdb_fetch(struct tdb_context *tdb, struct tdb_data key,
244 struct tdb_data *data)
247 struct tdb_used_record rec;
249 enum TDB_ERROR ecode;
251 if (tdb->flags & TDB_VERSION1)
252 return tdb1_fetch(tdb, key, data);
254 off = find_and_lock(tdb, key, F_RDLCK, &h, &rec, NULL);
255 if (TDB_OFF_IS_ERR(off)) {
256 return tdb->last_error = off;
260 ecode = TDB_ERR_NOEXIST;
262 data->dsize = rec_data_length(&rec);
263 data->dptr = tdb_alloc_read(tdb, off + sizeof(rec) + key.dsize,
265 if (TDB_PTR_IS_ERR(data->dptr)) {
266 ecode = TDB_PTR_ERR(data->dptr);
271 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_RDLCK);
272 return tdb->last_error = ecode;
275 bool tdb_exists(struct tdb_context *tdb, TDB_DATA key)
278 struct tdb_used_record rec;
281 off = find_and_lock(tdb, key, F_RDLCK, &h, &rec, NULL);
282 if (TDB_OFF_IS_ERR(off)) {
283 tdb->last_error = off;
286 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_RDLCK);
288 tdb->last_error = TDB_SUCCESS;
289 return off ? true : false;
292 enum TDB_ERROR tdb_delete(struct tdb_context *tdb, struct tdb_data key)
295 struct tdb_used_record rec;
297 enum TDB_ERROR ecode;
299 off = find_and_lock(tdb, key, F_WRLCK, &h, &rec, NULL);
300 if (TDB_OFF_IS_ERR(off)) {
301 return tdb->last_error = off;
305 ecode = TDB_ERR_NOEXIST;
309 ecode = delete_from_hash(tdb, &h);
310 if (ecode != TDB_SUCCESS) {
314 /* Free the deleted entry. */
316 ecode = add_free_record(tdb, off,
317 sizeof(struct tdb_used_record)
318 + rec_key_length(&rec)
319 + rec_data_length(&rec)
320 + rec_extra_padding(&rec),
321 TDB_LOCK_WAIT, true);
323 if (tdb->flags & TDB_SEQNUM)
327 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_WRLCK);
328 return tdb->last_error = ecode;
331 unsigned int tdb_get_flags(struct tdb_context *tdb)
336 static bool inside_transaction(const struct tdb_context *tdb)
338 if (tdb->flags & TDB_VERSION1)
339 return tdb->tdb1.transaction != NULL;
341 return tdb->tdb2.transaction != NULL;
344 static bool readonly_changable(struct tdb_context *tdb, const char *caller)
346 if (inside_transaction(tdb)) {
347 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
350 " TDB_RDONLY inside transaction",
355 if (tdb->file->allrecord_lock.count != 0
356 || tdb->file->num_lockrecs != 0) {
357 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
360 " TDB_RDONLY holding locks",
367 void tdb_add_flag(struct tdb_context *tdb, unsigned flag)
369 if (tdb->flags & TDB_INTERNAL) {
370 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
372 "tdb_add_flag: internal db");
377 tdb->flags |= TDB_NOLOCK;
380 tdb->flags |= TDB_NOMMAP;
381 tdb_munmap(tdb->file);
384 tdb->flags |= TDB_NOSYNC;
387 tdb->flags |= TDB_SEQNUM;
389 case TDB_ALLOW_NESTING:
390 tdb->flags |= TDB_ALLOW_NESTING;
393 if (readonly_changable(tdb, "tdb_add_flag"))
394 tdb->flags |= TDB_RDONLY;
397 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
399 "tdb_add_flag: Unknown flag %u",
404 void tdb_remove_flag(struct tdb_context *tdb, unsigned flag)
406 if (tdb->flags & TDB_INTERNAL) {
407 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
409 "tdb_remove_flag: internal db");
414 tdb->flags &= ~TDB_NOLOCK;
417 tdb->flags &= ~TDB_NOMMAP;
421 tdb->flags &= ~TDB_NOSYNC;
424 tdb->flags &= ~TDB_SEQNUM;
426 case TDB_ALLOW_NESTING:
427 tdb->flags &= ~TDB_ALLOW_NESTING;
430 if ((tdb->open_flags & O_ACCMODE) == O_RDONLY) {
431 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
433 "tdb_remove_flag: can't"
434 " remove TDB_RDONLY on tdb"
435 " opened with O_RDONLY");
438 if (readonly_changable(tdb, "tdb_remove_flag"))
439 tdb->flags &= ~TDB_RDONLY;
442 tdb->last_error = tdb_logerr(tdb, TDB_ERR_EINVAL,
444 "tdb_remove_flag: Unknown flag %u",
449 const char *tdb_errorstr(enum TDB_ERROR ecode)
451 /* Gcc warns if you miss a case in the switch, so use that. */
453 case TDB_SUCCESS: return "Success";
454 case TDB_ERR_CORRUPT: return "Corrupt database";
455 case TDB_ERR_IO: return "IO Error";
456 case TDB_ERR_LOCK: return "Locking error";
457 case TDB_ERR_OOM: return "Out of memory";
458 case TDB_ERR_EXISTS: return "Record exists";
459 case TDB_ERR_EINVAL: return "Invalid parameter";
460 case TDB_ERR_NOEXIST: return "Record does not exist";
461 case TDB_ERR_RDONLY: return "write not permitted";
463 return "Invalid error code";
466 enum TDB_ERROR tdb_error(struct tdb_context *tdb)
468 return tdb->last_error;
471 enum TDB_ERROR COLD tdb_logerr(struct tdb_context *tdb,
472 enum TDB_ERROR ecode,
473 enum tdb_log_level level,
474 const char *fmt, ...)
479 /* tdb_open paths care about errno, so save it. */
480 int saved_errno = errno;
486 len = vasprintf(&message, fmt, ap);
490 tdb->log_fn(tdb, TDB_LOG_ERROR, TDB_ERR_OOM,
491 "out of memory formatting message:", tdb->log_data);
492 tdb->log_fn(tdb, level, ecode, fmt, tdb->log_data);
494 tdb->log_fn(tdb, level, ecode, message, tdb->log_data);
501 enum TDB_ERROR tdb_parse_record_(struct tdb_context *tdb,
503 enum TDB_ERROR (*parse)(TDB_DATA k,
509 struct tdb_used_record rec;
511 enum TDB_ERROR ecode;
513 off = find_and_lock(tdb, key, F_RDLCK, &h, &rec, NULL);
514 if (TDB_OFF_IS_ERR(off)) {
515 return tdb->last_error = off;
519 ecode = TDB_ERR_NOEXIST;
522 dptr = tdb_access_read(tdb, off + sizeof(rec) + key.dsize,
523 rec_data_length(&rec), false);
524 if (TDB_PTR_IS_ERR(dptr)) {
525 ecode = TDB_PTR_ERR(dptr);
527 TDB_DATA d = tdb_mkdata(dptr, rec_data_length(&rec));
529 ecode = parse(key, d, data);
530 tdb_access_release(tdb, dptr);
534 tdb_unlock_hashes(tdb, h.hlock_start, h.hlock_range, F_RDLCK);
535 return tdb->last_error = ecode;
538 const char *tdb_name(const struct tdb_context *tdb)
543 int64_t tdb_get_seqnum(struct tdb_context *tdb)
545 tdb_off_t off = tdb_read_off(tdb, offsetof(struct tdb_header, seqnum));
546 if (TDB_OFF_IS_ERR(off))
547 tdb->last_error = off;
549 tdb->last_error = TDB_SUCCESS;
554 int tdb_fd(const struct tdb_context *tdb)
556 return tdb->file->fd;