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/>.
21 #define _XOPEN_SOURCE 500
22 #define _FILE_OFFSET_BITS 64
36 #include <ccan/tdb2/tdb2.h>
37 #include <ccan/likely/likely.h>
38 #ifdef HAVE_BYTESWAP_H
46 /* #define TDB_TRACE 1 */
49 #define __STRING(x) #x
52 #ifndef __STRINGSTRING
53 #define __STRINGSTRING(x) __STRING(x)
57 #define __location__ __FILE__ ":" __STRINGSTRING(__LINE__)
60 typedef uint64_t tdb_len_t;
61 typedef uint64_t tdb_off_t;
64 #define offsetof(t,f) ((unsigned int)&((t *)0)->f)
67 #define TDB_MAGIC_FOOD "TDB file\n"
68 #define TDB_VERSION ((uint64_t)(0x26011967 + 7))
69 #define TDB_MAGIC ((uint64_t)0x1999)
70 #define TDB_FREE_MAGIC ((~(uint64_t)TDB_MAGIC) << 6)
71 #define TDB_HASH_MAGIC (0xA1ABE11A01092008ULL)
72 #define TDB_RECOVERY_MAGIC (0xf53bc0e7U)
73 #define TDB_RECOVERY_INVALID_MAGIC (0x0)
75 #define TDB_OFF_ERR ((tdb_off_t)-1)
77 /* Prevent others from opening the file. */
78 #define TDB_OPEN_LOCK 0
79 /* Doing a transaction. */
80 #define TDB_TRANSACTION_LOCK 1
82 #define TDB_EXPANSION_LOCK 2
83 /* Hash chain locks. */
84 #define TDB_HASH_LOCK_START 3
86 /* We start wih 256 hash buckets, and a 64k-sized zone. */
87 #define INITIAL_HASH_BITS 8
88 #define INITIAL_ZONE_BITS 16
90 /* Try to create zones at least 32 times larger than allocations. */
91 #define TDB_COMFORT_FACTOR_BITS 5
93 /* We ensure buckets up to size 1 << (zone_bits - TDB_COMFORT_FACTOR_BITS). */
94 /* FIXME: test this matches size_to_bucket! */
95 #define BUCKETS_FOR_ZONE(zone_bits) ((zone_bits) + 2 - TDB_COMFORT_FACTOR_BITS)
98 static inline uint64_t bswap_64(uint64_t x)
100 return (((x&0x000000FFULL)<<56)
101 | ((x&0x0000FF00ULL)<<48)
102 | ((x&0x00FF0000ULL)<<40)
103 | ((x&0xFF000000ULL)<<32)
104 | ((x>>8)&0xFF000000ULL)
105 | ((x>>16)&0x00FF0000ULL)
106 | ((x>>24)&0x0000FF00ULL)
107 | ((x>>32)&0x000000FFULL));
111 struct tdb_used_record {
112 /* For on-disk compatibility, we avoid bitfields:
117 zone_bits: 6 (lowest)
119 uint64_t magic_and_meta;
120 /* The bottom key_len_bits*2 are key length, rest is data length. */
121 uint64_t key_and_data_len;
124 static inline unsigned rec_key_bits(const struct tdb_used_record *r)
126 return ((r->magic_and_meta >> 43) & ((1 << 5)-1)) * 2;
129 static inline uint64_t rec_key_length(const struct tdb_used_record *r)
131 return r->key_and_data_len & ((1ULL << rec_key_bits(r)) - 1);
134 static inline uint64_t rec_data_length(const struct tdb_used_record *r)
136 return r->key_and_data_len >> rec_key_bits(r);
139 static inline uint64_t rec_extra_padding(const struct tdb_used_record *r)
141 return (r->magic_and_meta >> 11) & 0xFFFFFFFF;
144 static inline unsigned int rec_zone_bits(const struct tdb_used_record *r)
146 return r->magic_and_meta & ((1 << 6) - 1);
149 static inline uint32_t rec_hash(const struct tdb_used_record *r)
151 return (r->magic_and_meta >> 6) & ((1 << 5) - 1);
154 static inline uint16_t rec_magic(const struct tdb_used_record *r)
156 return (r->magic_and_meta >> 48);
159 struct tdb_free_record {
160 uint64_t magic_and_meta; /* Bottom 6 bits are zone bits. */
161 uint64_t data_len; /* Not counting these two fields. */
162 /* This is why the minimum record size is 16 bytes. */
166 static inline unsigned int frec_zone_bits(const struct tdb_free_record *f)
168 return f->magic_and_meta & ((1 << 6) - 1);
171 static inline uint64_t frec_magic(const struct tdb_free_record *f)
173 return f->magic_and_meta & ~((1ULL << 6) - 1);
176 /* These parts can change while we have db open. */
177 struct tdb_header_volatile {
178 uint64_t generation; /* Makes sure it changes on every update. */
179 uint64_t hash_bits; /* Entries in hash table. */
180 uint64_t hash_off; /* Offset of hash table. */
183 /* this is stored at the front of every database */
185 char magic_food[32]; /* for /etc/magic */
186 /* FIXME: Make me 32 bit? */
187 uint64_t version; /* version of the code */
188 uint64_t hash_test; /* result of hashing HASH_MAGIC. */
189 uint64_t hash_seed; /* "random" seed written at creation time. */
191 struct tdb_header_volatile v;
193 tdb_off_t reserved[19];
196 /* Each zone has its set of free lists at the head.
198 * For each zone we have a series of per-size buckets, and a final bucket for
200 struct free_zone_header {
201 /* How much does this zone cover? */
203 /* tdb_off_t buckets[free_buckets + 1] */
206 enum tdb_lock_flags {
207 /* WAIT == F_SETLKW, NOWAIT == F_SETLK */
210 /* If set, don't log an error on failure. */
214 struct tdb_lock_type {
221 /* Filename of the database. */
224 /* Mmap (if any), or malloc (for TDB_INTERNAL). */
227 /* Open file descriptor (undefined for TDB_INTERNAL). */
230 /* How much space has been mapped (<= current file size) */
233 /* Opened read-only? */
236 /* Error code for last tdb error. */
237 enum TDB_ERROR ecode;
239 /* A cached copy of the header */
240 struct tdb_header header;
241 /* (for debugging). */
242 bool header_uptodate;
244 /* the flags passed to tdb_open, for tdb_reopen. */
247 /* Logging function */
255 /* Set if we are in a transaction. */
256 struct tdb_transaction *transaction;
258 /* What zone of the tdb to use, for spreading load. */
260 /* Cached copy of zone header. */
261 struct free_zone_header zhdr;
263 /* IO methods: changes for transactions. */
264 const struct tdb_methods *methods;
266 /* Lock information */
267 struct tdb_lock_type allrecord_lock;
268 uint64_t num_lockrecs;
269 struct tdb_lock_type *lockrecs;
271 /* Single list of all TDBs, to avoid multiple opens. */
272 struct tdb_context *next;
278 int (*read)(struct tdb_context *, tdb_off_t, void *, tdb_len_t);
279 int (*write)(struct tdb_context *, tdb_off_t, const void *, tdb_len_t);
280 int (*oob)(struct tdb_context *, tdb_off_t, bool);
281 int (*expand_file)(struct tdb_context *, tdb_len_t);
288 /* Returns true if header changed (and updates it). */
289 bool header_changed(struct tdb_context *tdb);
291 /* Commit header to disk. */
292 int write_header(struct tdb_context *tdb);
294 /* Hash random memory. */
295 uint64_t tdb_hash(struct tdb_context *tdb, const void *ptr, size_t len);
297 /* offset of hash table entry for this list/hash value */
298 tdb_off_t hash_off(struct tdb_context *tdb, uint64_t list);
302 int tdb_zone_init(struct tdb_context *tdb);
304 /* If this fails, try tdb_expand. */
305 tdb_off_t alloc(struct tdb_context *tdb, size_t keylen, size_t datalen,
306 uint64_t hash, bool growing);
308 /* Put this record in a free list. */
309 int add_free_record(struct tdb_context *tdb, unsigned int zone_bits,
310 tdb_off_t off, tdb_len_t len_with_header);
312 /* Set up header for a used record. */
313 int set_header(struct tdb_context *tdb,
314 struct tdb_used_record *rec,
315 uint64_t keylen, uint64_t datalen,
316 uint64_t actuallen, uint64_t hash,
317 unsigned int zone_bits);
319 /* Used by tdb_check to verify. */
320 unsigned int size_to_bucket(unsigned int free_buckets, tdb_len_t data_len);
321 tdb_off_t bucket_off(tdb_off_t zone_off, tdb_off_t bucket);
324 /* Initialize tdb->methods. */
325 void tdb_io_init(struct tdb_context *tdb);
327 /* Convert endian of the buffer if required. */
328 void *tdb_convert(const struct tdb_context *tdb, void *buf, tdb_len_t size);
330 /* Unmap and try to map the tdb. */
331 void tdb_munmap(struct tdb_context *tdb);
332 void tdb_mmap(struct tdb_context *tdb);
334 /* Either make a copy into pad and return that, or return ptr into mmap.
335 * Converts endian (ie. will use pad in that case). */
336 void *tdb_get(struct tdb_context *tdb, tdb_off_t off, void *pad, size_t len);
338 /* Either alloc a copy, or give direct access. Release frees or noop. */
339 const void *tdb_access_read(struct tdb_context *tdb,
340 tdb_off_t off, tdb_len_t len, bool convert);
341 void tdb_access_release(struct tdb_context *tdb, const void *p);
343 /* Convenience routine to get an offset. */
344 tdb_off_t tdb_read_off(struct tdb_context *tdb, tdb_off_t off);
346 /* Write an offset at an offset. */
347 int tdb_write_off(struct tdb_context *tdb, tdb_off_t off, tdb_off_t val);
349 /* Clear an ondisk area. */
350 int zero_out(struct tdb_context *tdb, tdb_off_t off, tdb_len_t len);
352 /* Return a non-zero offset in this array, or num. */
353 tdb_off_t tdb_find_nonzero_off(struct tdb_context *tdb, tdb_off_t off,
356 /* Return a zero offset in this array, or num. */
357 tdb_off_t tdb_find_zero_off(struct tdb_context *tdb, tdb_off_t off,
360 /* Even on files, we can get partial writes due to signals. */
361 bool tdb_pwrite_all(int fd, const void *buf, size_t len, tdb_off_t off);
362 bool tdb_pread_all(int fd, void *buf, size_t len, tdb_off_t off);
363 bool tdb_read_all(int fd, void *buf, size_t len);
365 /* Allocate and make a copy of some offset. */
366 void *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len);
368 /* Writes a converted copy of a record. */
369 int tdb_write_convert(struct tdb_context *tdb, tdb_off_t off,
370 const void *rec, size_t len);
372 /* Reads record and converts it */
373 int tdb_read_convert(struct tdb_context *tdb, tdb_off_t off,
374 void *rec, size_t len);
377 uint64_t hash_record(struct tdb_context *tdb, tdb_off_t off);
380 void tdb_lock_init(struct tdb_context *tdb);
382 /* Lock/unlock a particular hash list. */
383 tdb_off_t tdb_lock_list(struct tdb_context *tdb, uint64_t hash,
384 int ltype, enum tdb_lock_flags waitflag);
385 int tdb_unlock_list(struct tdb_context *tdb, tdb_off_t list, int ltype);
387 /* Lock/unlock a particular free bucket. */
388 int tdb_lock_free_bucket(struct tdb_context *tdb, tdb_off_t b_off,
389 enum tdb_lock_flags waitflag);
390 void tdb_unlock_free_bucket(struct tdb_context *tdb, tdb_off_t b_off);
392 /* Do we have any locks? */
393 bool tdb_has_locks(struct tdb_context *tdb);
395 /* Lock entire database. */
396 int tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
397 enum tdb_lock_flags flags, bool upgradable);
398 int tdb_allrecord_unlock(struct tdb_context *tdb, int ltype);
400 /* Serialize db open. */
401 int tdb_lock_open(struct tdb_context *tdb);
402 void tdb_unlock_open(struct tdb_context *tdb);
404 /* Serialize db expand. */
405 int tdb_lock_expand(struct tdb_context *tdb, int ltype);
406 void tdb_unlock_expand(struct tdb_context *tdb, int ltype);
408 /* Expand the file. */
409 int tdb_expand(struct tdb_context *tdb, tdb_len_t klen, tdb_len_t dlen,
413 /* Low-level locking primitives. */
414 int tdb_nest_lock(struct tdb_context *tdb, tdb_off_t offset, int ltype,
415 enum tdb_lock_flags flags);
416 int tdb_nest_unlock(struct tdb_context *tdb, tdb_off_t offset, int ltype);
418 int tdb_munmap(struct tdb_context *tdb);
419 void tdb_mmap(struct tdb_context *tdb);
420 int tdb_lock(struct tdb_context *tdb, int list, int ltype);
421 int tdb_lock_nonblock(struct tdb_context *tdb, int list, int ltype);
422 bool tdb_have_locks(struct tdb_context *tdb);
423 int tdb_unlock(struct tdb_context *tdb, int list, int ltype);
424 int tdb_brlock(struct tdb_context *tdb,
425 int rw_type, tdb_off_t offset, size_t len,
426 enum tdb_lock_flags flags);
427 int tdb_brunlock(struct tdb_context *tdb,
428 int rw_type, tdb_off_t offset, size_t len);
429 bool tdb_have_extra_locks(struct tdb_context *tdb);
430 void tdb_release_extra_locks(struct tdb_context *tdb);
431 int tdb_transaction_lock(struct tdb_context *tdb, int ltype);
432 int tdb_transaction_unlock(struct tdb_context *tdb, int ltype);
433 int tdb_allrecord_lock(struct tdb_context *tdb, int ltype,
434 enum tdb_lock_flags flags, bool upgradable);
435 int tdb_allrecord_unlock(struct tdb_context *tdb, int ltype);
436 int tdb_allrecord_upgrade(struct tdb_context *tdb);
437 int tdb_write_lock_record(struct tdb_context *tdb, tdb_off_t off);
438 int tdb_write_unlock_record(struct tdb_context *tdb, tdb_off_t off);
439 int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
440 int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
441 int tdb_free(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec);
442 tdb_off_t tdb_allocate(struct tdb_context *tdb, tdb_len_t length, struct tdb_record *rec);
443 int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
444 int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d);
445 int tdb_lock_record(struct tdb_context *tdb, tdb_off_t off);
446 int tdb_unlock_record(struct tdb_context *tdb, tdb_off_t off);
447 bool tdb_needs_recovery(struct tdb_context *tdb);
448 int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec);
449 int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec);
450 int tdb_do_delete(struct tdb_context *tdb, tdb_off_t rec_ptr, struct tdb_record *rec);
451 unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len);
452 int tdb_parse_data(struct tdb_context *tdb, TDB_DATA key,
453 tdb_off_t offset, tdb_len_t len,
454 int (*parser)(TDB_DATA key, TDB_DATA data,
457 tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t hash, int locktype,
458 struct tdb_record *rec);
459 void tdb_io_init(struct tdb_context *tdb);
460 int tdb_expand(struct tdb_context *tdb, tdb_off_t size);
461 int tdb_rec_free_read(struct tdb_context *tdb, tdb_off_t off,
462 struct tdb_record *rec);
466 void tdb_trace(struct tdb_context *tdb, const char *op);
467 void tdb_trace_seqnum(struct tdb_context *tdb, uint32_t seqnum, const char *op);
468 void tdb_trace_open(struct tdb_context *tdb, const char *op,
469 unsigned hash_size, unsigned tdb_flags, unsigned open_flags);
470 void tdb_trace_ret(struct tdb_context *tdb, const char *op, int ret);
471 void tdb_trace_retrec(struct tdb_context *tdb, const char *op, TDB_DATA ret);
472 void tdb_trace_1rec(struct tdb_context *tdb, const char *op,
474 void tdb_trace_1rec_ret(struct tdb_context *tdb, const char *op,
475 TDB_DATA rec, int ret);
476 void tdb_trace_1rec_retrec(struct tdb_context *tdb, const char *op,
477 TDB_DATA rec, TDB_DATA ret);
478 void tdb_trace_2rec_flag_ret(struct tdb_context *tdb, const char *op,
479 TDB_DATA rec1, TDB_DATA rec2, unsigned flag,
481 void tdb_trace_2rec_retrec(struct tdb_context *tdb, const char *op,
482 TDB_DATA rec1, TDB_DATA rec2, TDB_DATA ret);
484 #define tdb_trace(tdb, op)
485 #define tdb_trace_seqnum(tdb, seqnum, op)
486 #define tdb_trace_open(tdb, op, hash_size, tdb_flags, open_flags)
487 #define tdb_trace_ret(tdb, op, ret)
488 #define tdb_trace_retrec(tdb, op, ret)
489 #define tdb_trace_1rec(tdb, op, rec)
490 #define tdb_trace_1rec_ret(tdb, op, rec, ret)
491 #define tdb_trace_1rec_retrec(tdb, op, rec, ret)
492 #define tdb_trace_2rec_flag_ret(tdb, op, rec1, rec2, flag, ret)
493 #define tdb_trace_2rec_retrec(tdb, op, rec1, rec2, ret)
494 #endif /* !TDB_TRACE */