4 Trivial Database 2: private types and prototypes
5 Copyright (C) Rusty Russell 2010
7 This library is free software; you can redistribute it and/or
8 modify it under the terms of the GNU Lesser General Public
9 License as published by the Free Software Foundation; either
10 version 3 of the License, or (at your option) any later version.
12 This library is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 Lesser General Public License for more details.
17 You should have received a copy of the GNU Lesser General Public
18 License along with this library; if not, see <http://www.gnu.org/licenses/>.
22 #if HAVE_FILE_OFFSET_BITS
23 #define _FILE_OFFSET_BITS 64
38 #include <ccan/tdb2/tdb2.h>
39 #include <ccan/likely/likely.h>
40 #include <ccan/compiler/compiler.h>
49 /* #define TDB_TRACE 1 */
52 #define __STRING(x) #x
55 #ifndef __STRINGSTRING
56 #define __STRINGSTRING(x) __STRING(x)
60 #define __location__ __FILE__ ":" __STRINGSTRING(__LINE__)
63 typedef uint64_t tdb_len_t;
64 typedef uint64_t tdb_off_t;
66 #define TDB_MAGIC_FOOD "TDB file\n"
67 #define TDB_VERSION ((uint64_t)(0x26011967 + 7))
68 #define TDB_USED_MAGIC ((uint64_t)0x1999)
69 #define TDB_HTABLE_MAGIC ((uint64_t)0x1888)
70 #define TDB_CHAIN_MAGIC ((uint64_t)0x1777)
71 #define TDB_FTABLE_MAGIC ((uint64_t)0x1666)
72 #define TDB_FREE_MAGIC ((uint64_t)0xFE)
73 #define TDB_HASH_MAGIC (0xA1ABE11A01092008ULL)
74 #define TDB_RECOVERY_MAGIC (0xf53bc0e7ad124589ULL)
75 #define TDB_RECOVERY_INVALID_MAGIC (0x0ULL)
77 #define TDB_OFF_IS_ERR(off) unlikely(off >= (tdb_off_t)TDB_ERR_LAST)
79 /* Packing errors into pointers and v.v. */
80 #define TDB_PTR_IS_ERR(ptr) \
81 unlikely((unsigned long)(ptr) >= (unsigned long)TDB_ERR_LAST)
82 #define TDB_PTR_ERR(p) ((enum TDB_ERROR)(long)(p))
83 #define TDB_ERR_PTR(err) ((void *)(long)(err))
85 /* Common case of returning true, false or -ve error. */
86 typedef int tdb_bool_err;
88 /* Prevent others from opening the file. */
89 #define TDB_OPEN_LOCK 0
90 /* Doing a transaction. */
91 #define TDB_TRANSACTION_LOCK 1
93 #define TDB_EXPANSION_LOCK 2
94 /* Hash chain locks. */
95 #define TDB_HASH_LOCK_START 64
97 /* Range for hash locks. */
98 #define TDB_HASH_LOCK_RANGE_BITS 30
99 #define TDB_HASH_LOCK_RANGE (1 << TDB_HASH_LOCK_RANGE_BITS)
101 /* We have 1024 entries in the top level. */
102 #define TDB_TOPLEVEL_HASH_BITS 10
103 /* And 64 entries in each sub-level: thus 64 bits exactly after 9 levels. */
104 #define TDB_SUBLEVEL_HASH_BITS 6
105 /* And 8 entries in each group, ie 8 groups per sublevel. */
106 #define TDB_HASH_GROUP_BITS 3
107 /* This is currently 10: beyond this we chain. */
108 #define TDB_MAX_LEVELS (1+(64-TDB_TOPLEVEL_HASH_BITS) / TDB_SUBLEVEL_HASH_BITS)
110 /* Extend file by least 100 times larger than needed. */
111 #define TDB_EXTENSION_FACTOR 100
113 /* We steal bits from the offsets to store hash info. */
114 #define TDB_OFF_HASH_GROUP_MASK ((1ULL << TDB_HASH_GROUP_BITS) - 1)
115 /* We steal this many upper bits, giving a maximum offset of 64 exabytes. */
116 #define TDB_OFF_UPPER_STEAL 8
117 #define TDB_OFF_UPPER_STEAL_EXTRA 7
118 /* The bit number where we store extra hash bits. */
119 #define TDB_OFF_HASH_EXTRA_BIT 57
120 #define TDB_OFF_UPPER_STEAL_SUBHASH_BIT 56
122 /* Additional features we understand. Currently: none. */
123 #define TDB_FEATURE_MASK ((uint64_t)0)
125 /* The bit number where we store the extra hash bits. */
126 /* Convenience mask to get actual offset. */
127 #define TDB_OFF_MASK \
128 (((1ULL << (64 - TDB_OFF_UPPER_STEAL)) - 1) - TDB_OFF_HASH_GROUP_MASK)
130 /* How many buckets in a free list: see size_to_bucket(). */
131 #define TDB_FREE_BUCKETS (64 - TDB_OFF_UPPER_STEAL)
133 /* We have to be able to fit a free record here. */
134 #define TDB_MIN_DATA_LEN \
135 (sizeof(struct tdb_free_record) - sizeof(struct tdb_used_record))
137 /* Indicates this entry is not on an flist (can happen during coalescing) */
138 #define TDB_FTABLE_NONE ((1ULL << TDB_OFF_UPPER_STEAL) - 1)
141 static inline uint64_t bswap_64(uint64_t x)
143 return (((x&0x000000FFULL)<<56)
144 | ((x&0x0000FF00ULL)<<48)
145 | ((x&0x00FF0000ULL)<<40)
146 | ((x&0xFF000000ULL)<<32)
147 | ((x>>8)&0xFF000000ULL)
148 | ((x>>16)&0x00FF0000ULL)
149 | ((x>>24)&0x0000FF00ULL)
150 | ((x>>32)&0x000000FFULL));
154 struct tdb_used_record {
155 /* For on-disk compatibility, we avoid bitfields:
161 uint64_t magic_and_meta;
162 /* The bottom key_len_bits*2 are key length, rest is data length. */
163 uint64_t key_and_data_len;
166 static inline unsigned rec_key_bits(const struct tdb_used_record *r)
168 return ((r->magic_and_meta >> 43) & ((1 << 5)-1)) * 2;
171 static inline uint64_t rec_key_length(const struct tdb_used_record *r)
173 return r->key_and_data_len & ((1ULL << rec_key_bits(r)) - 1);
176 static inline uint64_t rec_data_length(const struct tdb_used_record *r)
178 return r->key_and_data_len >> rec_key_bits(r);
181 static inline uint64_t rec_extra_padding(const struct tdb_used_record *r)
183 return (r->magic_and_meta >> 11) & 0xFFFFFFFF;
186 static inline uint32_t rec_hash(const struct tdb_used_record *r)
188 return r->magic_and_meta & ((1 << 11) - 1);
191 static inline uint16_t rec_magic(const struct tdb_used_record *r)
193 return (r->magic_and_meta >> 48);
196 struct tdb_free_record {
197 uint64_t magic_and_prev; /* TDB_OFF_UPPER_STEAL bits magic, then prev */
198 uint64_t ftable_and_len; /* Len not counting these two fields. */
199 /* This is why the minimum record size is 8 bytes. */
203 static inline uint64_t frec_prev(const struct tdb_free_record *f)
205 return f->magic_and_prev & ((1ULL << (64 - TDB_OFF_UPPER_STEAL)) - 1);
208 static inline uint64_t frec_magic(const struct tdb_free_record *f)
210 return f->magic_and_prev >> (64 - TDB_OFF_UPPER_STEAL);
213 static inline uint64_t frec_len(const struct tdb_free_record *f)
215 return f->ftable_and_len & ((1ULL << (64 - TDB_OFF_UPPER_STEAL))-1);
218 static inline unsigned frec_ftable(const struct tdb_free_record *f)
220 return f->ftable_and_len >> (64 - TDB_OFF_UPPER_STEAL);
223 struct tdb_recovery_record {
225 /* Length of record. */
229 /* Old length of file before transaction. */
233 /* If we bottom out of the subhashes, we chain. */
235 tdb_off_t rec[1 << TDB_HASH_GROUP_BITS];
239 /* this is stored at the front of every database */
241 char magic_food[64]; /* for /etc/magic */
242 /* FIXME: Make me 32 bit? */
243 uint64_t version; /* version of the code */
244 uint64_t hash_test; /* result of hashing HASH_MAGIC. */
245 uint64_t hash_seed; /* "random" seed written at creation time. */
246 tdb_off_t free_table; /* (First) free table. */
247 tdb_off_t recovery; /* Transaction recovery area. */
249 uint64_t features_used; /* Features all writers understand */
250 uint64_t features_offered; /* Features offered */
252 uint64_t seqnum; /* Sequence number for TDB_SEQNUM */
254 tdb_off_t reserved[23];
256 /* Top level hash table. */
257 tdb_off_t hashtable[1ULL << TDB_TOPLEVEL_HASH_BITS];
260 struct tdb_freetable {
261 struct tdb_used_record hdr;
263 tdb_off_t buckets[TDB_FREE_BUCKETS];
266 /* Information about a particular (locked) hash entry. */
268 /* Full hash value of entry. */
270 /* Start and length of lock acquired. */
271 tdb_off_t hlock_start;
272 tdb_len_t hlock_range;
273 /* Start of hash group. */
274 tdb_off_t group_start;
275 /* Bucket we belong in. */
276 unsigned int home_bucket;
277 /* Bucket we (or an empty space) were found in. */
278 unsigned int found_bucket;
279 /* How many bits of the hash are already used. */
280 unsigned int hash_used;
281 /* Current working group. */
282 tdb_off_t group[1 << TDB_HASH_GROUP_BITS];
285 struct traverse_info {
286 struct traverse_level {
288 /* We ignore groups here, and treat it as a big array. */
290 unsigned int total_buckets;
291 } levels[TDB_MAX_LEVELS + 1];
292 unsigned int num_levels;
293 unsigned int toplevel_group;
294 /* This makes delete-everything-inside-traverse work as expected. */
298 enum tdb_lock_flags {
299 /* WAIT == F_SETLKW, NOWAIT == F_SETLK */
302 /* If set, don't log an error on failure. */
304 /* If set, don't check for recovery (used by recovery code). */
305 TDB_LOCK_NOCHECK = 4,
309 struct tdb_context *owner;
315 /* This is only needed for tdb_access_commit, but used everywhere to
317 struct tdb_access_hdr {
318 struct tdb_access_hdr *next;
325 /* Single list of all TDBs, to detect multiple opens. */
326 struct tdb_file *next;
328 /* How many are sharing us? */
331 /* Mmap (if any), or malloc (for TDB_INTERNAL). */
334 /* How much space has been mapped (<= current file size) */
337 /* The file descriptor (-1 for TDB_INTERNAL). */
340 /* Lock information */
342 struct tdb_lock allrecord_lock;
344 struct tdb_lock *lockrecs;
346 /* Identity of this file. */
352 /* Filename of the database. */
355 /* Are we accessing directly? (debugging check). */
358 /* Operating read-only? (Opened O_RDONLY, or in traverse_read) */
361 /* mmap read only? */
364 /* the flags passed to tdb_open, for tdb_reopen. */
367 /* Logging function */
368 void (*log_fn)(struct tdb_context *tdb,
369 enum tdb_log_level level,
375 uint64_t (*hash_fn)(const void *key, size_t len, uint64_t seed, void *);
379 /* low level (fnctl) lock functions. */
380 int (*lock_fn)(int fd, int rw, off_t off, off_t len, bool w, void *);
381 int (*unlock_fn)(int fd, int rw, off_t off, off_t len, void *);
384 /* Set if we are in a transaction. */
385 struct tdb_transaction *transaction;
387 /* What free table are we using? */
388 tdb_off_t ftable_off;
391 /* IO methods: changes for transactions. */
392 const struct tdb_methods *methods;
394 /* Our statistics. */
395 struct tdb_attribute_stats stats;
397 /* Direct access information */
398 struct tdb_access_hdr *access;
400 /* Last error we returned. */
401 enum TDB_ERROR last_error;
403 /* The actual file information */
404 struct tdb_file *file;
408 enum TDB_ERROR (*tread)(struct tdb_context *, tdb_off_t, void *,
410 enum TDB_ERROR (*twrite)(struct tdb_context *, tdb_off_t, const void *,
412 enum TDB_ERROR (*oob)(struct tdb_context *, tdb_off_t, bool);
413 enum TDB_ERROR (*expand_file)(struct tdb_context *, tdb_len_t);
414 void *(*direct)(struct tdb_context *, tdb_off_t, size_t, bool);
421 tdb_bool_err first_in_hash(struct tdb_context *tdb,
422 struct traverse_info *tinfo,
423 TDB_DATA *kbuf, size_t *dlen);
425 tdb_bool_err next_in_hash(struct tdb_context *tdb,
426 struct traverse_info *tinfo,
427 TDB_DATA *kbuf, size_t *dlen);
429 /* Hash random memory. */
430 uint64_t tdb_hash(struct tdb_context *tdb, const void *ptr, size_t len);
433 uint64_t hash_record(struct tdb_context *tdb, tdb_off_t off);
435 /* Find and lock a hash entry (or where it would be). */
436 tdb_off_t find_and_lock(struct tdb_context *tdb,
440 struct tdb_used_record *rec,
441 struct traverse_info *tinfo);
443 enum TDB_ERROR replace_in_hash(struct tdb_context *tdb,
447 enum TDB_ERROR add_to_hash(struct tdb_context *tdb, struct hash_info *h,
450 enum TDB_ERROR delete_from_hash(struct tdb_context *tdb, struct hash_info *h);
453 bool is_subhash(tdb_off_t val);
456 enum TDB_ERROR tdb_ftable_init(struct tdb_context *tdb);
458 /* check.c needs these to iterate through free lists. */
459 tdb_off_t first_ftable(struct tdb_context *tdb);
460 tdb_off_t next_ftable(struct tdb_context *tdb, tdb_off_t ftable);
462 /* This returns space or -ve error number. */
463 tdb_off_t alloc(struct tdb_context *tdb, size_t keylen, size_t datalen,
464 uint64_t hash, unsigned magic, bool growing);
466 /* Put this record in a free list. */
467 enum TDB_ERROR add_free_record(struct tdb_context *tdb,
468 tdb_off_t off, tdb_len_t len_with_header,
469 enum tdb_lock_flags waitflag,
472 /* Set up header for a used/ftable/htable/chain record. */
473 enum TDB_ERROR set_header(struct tdb_context *tdb,
474 struct tdb_used_record *rec,
475 unsigned magic, uint64_t keylen, uint64_t datalen,
476 uint64_t actuallen, unsigned hashlow);
478 /* Used by tdb_check to verify. */
479 unsigned int size_to_bucket(tdb_len_t data_len);
480 tdb_off_t bucket_off(tdb_off_t ftable_off, unsigned bucket);
482 /* Used by tdb_summary */
483 tdb_off_t dead_space(struct tdb_context *tdb, tdb_off_t off);
486 /* Initialize tdb->methods. */
487 void tdb_io_init(struct tdb_context *tdb);
489 /* Convert endian of the buffer if required. */
490 void *tdb_convert(const struct tdb_context *tdb, void *buf, tdb_len_t size);
492 /* Unmap and try to map the tdb. */
493 void tdb_munmap(struct tdb_file *file);
494 void tdb_mmap(struct tdb_context *tdb);
496 /* Either alloc a copy, or give direct access. Release frees or noop. */
497 const void *tdb_access_read(struct tdb_context *tdb,
498 tdb_off_t off, tdb_len_t len, bool convert);
499 void *tdb_access_write(struct tdb_context *tdb,
500 tdb_off_t off, tdb_len_t len, bool convert);
502 /* Release result of tdb_access_read/write. */
503 void tdb_access_release(struct tdb_context *tdb, const void *p);
504 /* Commit result of tdb_acces_write. */
505 enum TDB_ERROR tdb_access_commit(struct tdb_context *tdb, void *p);
507 /* Convenience routine to get an offset. */
508 tdb_off_t tdb_read_off(struct tdb_context *tdb, tdb_off_t off);
510 /* Write an offset at an offset. */
511 enum TDB_ERROR tdb_write_off(struct tdb_context *tdb, tdb_off_t off,
514 /* Clear an ondisk area. */
515 enum TDB_ERROR zero_out(struct tdb_context *tdb, tdb_off_t off, tdb_len_t len);
517 /* Return a non-zero offset between >= start < end in this array (or end). */
518 tdb_off_t tdb_find_nonzero_off(struct tdb_context *tdb,
523 /* Return a zero offset in this array, or num. */
524 tdb_off_t tdb_find_zero_off(struct tdb_context *tdb, tdb_off_t off,
527 /* Allocate and make a copy of some offset. */
528 void *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len);
530 /* Writes a converted copy of a record. */
531 enum TDB_ERROR tdb_write_convert(struct tdb_context *tdb, tdb_off_t off,
532 const void *rec, size_t len);
534 /* Reads record and converts it */
535 enum TDB_ERROR tdb_read_convert(struct tdb_context *tdb, tdb_off_t off,
536 void *rec, size_t len);
538 /* Bump the seqnum (caller checks for tdb->flags & TDB_SEQNUM) */
539 void tdb_inc_seqnum(struct tdb_context *tdb);
542 /* Lock/unlock a range of hashes. */
543 enum TDB_ERROR tdb_lock_hashes(struct tdb_context *tdb,
544 tdb_off_t hash_lock, tdb_len_t hash_range,
545 int ltype, enum tdb_lock_flags waitflag);
546 enum TDB_ERROR tdb_unlock_hashes(struct tdb_context *tdb,
548 tdb_len_t hash_range, int ltype);
550 /* For closing the file. */
551 void tdb_lock_cleanup(struct tdb_context *tdb);
553 /* Lock/unlock a particular free bucket. */
554 enum TDB_ERROR tdb_lock_free_bucket(struct tdb_context *tdb, tdb_off_t b_off,
555 enum tdb_lock_flags waitflag);
556 void tdb_unlock_free_bucket(struct tdb_context *tdb, tdb_off_t b_off);
558 /* Serialize transaction start. */
559 enum TDB_ERROR tdb_transaction_lock(struct tdb_context *tdb, int ltype);
560 void tdb_transaction_unlock(struct tdb_context *tdb, int ltype);
562 /* Do we have any hash locks (ie. via tdb_chainlock) ? */
563 bool tdb_has_hash_locks(struct tdb_context *tdb);
565 /* Lock entire database. */
566 enum TDB_ERROR tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
567 enum tdb_lock_flags flags, bool upgradable);
568 void tdb_allrecord_unlock(struct tdb_context *tdb, int ltype);
569 enum TDB_ERROR tdb_allrecord_upgrade(struct tdb_context *tdb);
571 /* Serialize db open. */
572 enum TDB_ERROR tdb_lock_open(struct tdb_context *tdb,
573 enum tdb_lock_flags flags);
574 void tdb_unlock_open(struct tdb_context *tdb);
575 bool tdb_has_open_lock(struct tdb_context *tdb);
577 /* Serialize db expand. */
578 enum TDB_ERROR tdb_lock_expand(struct tdb_context *tdb, int ltype);
579 void tdb_unlock_expand(struct tdb_context *tdb, int ltype);
580 bool tdb_has_expansion_lock(struct tdb_context *tdb);
582 /* If it needs recovery, grab all the locks and do it. */
583 enum TDB_ERROR tdb_lock_and_recover(struct tdb_context *tdb);
585 /* Default lock and unlock functions. */
586 int tdb_fcntl_lock(int fd, int rw, off_t off, off_t len, bool waitflag, void *);
587 int tdb_fcntl_unlock(int fd, int rw, off_t off, off_t len, void *);
590 enum TDB_ERROR tdb_transaction_recover(struct tdb_context *tdb);
591 tdb_bool_err tdb_needs_recovery(struct tdb_context *tdb);
594 enum TDB_ERROR COLD tdb_logerr(struct tdb_context *tdb,
595 enum TDB_ERROR ecode,
596 enum tdb_log_level level,
597 const char *fmt, ...);
600 void tdb_trace(struct tdb_context *tdb, const char *op);
601 void tdb_trace_seqnum(struct tdb_context *tdb, uint32_t seqnum, const char *op);
602 void tdb_trace_open(struct tdb_context *tdb, const char *op,
603 unsigned hash_size, unsigned tdb_flags, unsigned open_flags);
604 void tdb_trace_ret(struct tdb_context *tdb, const char *op, int ret);
605 void tdb_trace_retrec(struct tdb_context *tdb, const char *op, TDB_DATA ret);
606 void tdb_trace_1rec(struct tdb_context *tdb, const char *op,
608 void tdb_trace_1rec_ret(struct tdb_context *tdb, const char *op,
609 TDB_DATA rec, int ret);
610 void tdb_trace_1rec_retrec(struct tdb_context *tdb, const char *op,
611 TDB_DATA rec, TDB_DATA ret);
612 void tdb_trace_2rec_flag_ret(struct tdb_context *tdb, const char *op,
613 TDB_DATA rec1, TDB_DATA rec2, unsigned flag,
615 void tdb_trace_2rec_retrec(struct tdb_context *tdb, const char *op,
616 TDB_DATA rec1, TDB_DATA rec2, TDB_DATA ret);
618 #define tdb_trace(tdb, op)
619 #define tdb_trace_seqnum(tdb, seqnum, op)
620 #define tdb_trace_open(tdb, op, hash_size, tdb_flags, open_flags)
621 #define tdb_trace_ret(tdb, op, ret)
622 #define tdb_trace_retrec(tdb, op, ret)
623 #define tdb_trace_1rec(tdb, op, rec)
624 #define tdb_trace_1rec_ret(tdb, op, rec, ret)
625 #define tdb_trace_1rec_retrec(tdb, op, rec, ret)
626 #define tdb_trace_2rec_flag_ret(tdb, op, rec1, rec2, flag, ret)
627 #define tdb_trace_2rec_retrec(tdb, op, rec1, rec2, ret)
628 #endif /* !TDB_TRACE */