lib/ntdb/private.h

   1 #ifndef NTDB_PRIVATE_H
   2 #define NTDB_PRIVATE_H
   3 /*
   4   Trivial Database 2: private types and prototypes
   5   Copyright (C) Rusty Russell 2010
   6
   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.
  11
  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.
  16
  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/>.
  19 */
  20
  21 #include "config.h"
  22 #ifndef HAVE_CCAN
  23 #error You need ccan to build ntdb!
  24 #endif
  25 #include "ntdb.h"
  26 #include <ccan/compiler/compiler.h>
  27 #include <ccan/likely/likely.h>
  28 #include <ccan/endian/endian.h>
  29
  30 #ifdef HAVE_LIBREPLACE
  31 #include "replace.h"
  32 #include "system/filesys.h"
  33 #include "system/time.h"
  34 #include "system/shmem.h"
  35 #include "system/select.h"
  36 #include "system/wait.h"
  37 #else
  38 #include <stdint.h>
  39 #include <stdbool.h>
  40 #include <stdlib.h>
  41 #include <stddef.h>
  42 #include <sys/time.h>
  43 #include <sys/mman.h>
  44 #include <unistd.h>
  45 #include <fcntl.h>
  46 #include <errno.h>
  47 #include <stdio.h>
  48 #include <utime.h>
  49 #include <unistd.h>
  50 #endif
  51 #include <assert.h>
  52
  53 #ifndef TEST_IT
  54 #define TEST_IT(cond)
  55 #endif
  56
  57 /* #define NTDB_TRACE 1 */
  58
  59 #ifndef __STRING
  60 #define __STRING(x)    #x
  61 #endif
  62
  63 #ifndef __STRINGSTRING
  64 #define __STRINGSTRING(x) __STRING(x)
  65 #endif
  66
  67 #ifndef __location__
  68 #define __location__ __FILE__ ":" __STRINGSTRING(__LINE__)
  69 #endif
  70
  71 typedef uint64_t ntdb_len_t;
  72 typedef uint64_t ntdb_off_t;
  73
  74 #define NTDB_MAGIC_FOOD "NTDB file\n"
  75 #define NTDB_VERSION ((uint64_t)(0x26011967 + 7))
  76 #define NTDB_USED_MAGIC ((uint64_t)0x1999)
  77 #define NTDB_HTABLE_MAGIC ((uint64_t)0x1888)
  78 #define NTDB_CHAIN_MAGIC ((uint64_t)0x1777)
  79 #define NTDB_FTABLE_MAGIC ((uint64_t)0x1666)
  80 #define NTDB_CAP_MAGIC ((uint64_t)0x1555)
  81 #define NTDB_FREE_MAGIC ((uint64_t)0xFE)
  82 #define NTDB_HASH_MAGIC (0xA1ABE11A01092008ULL)
  83 #define NTDB_RECOVERY_MAGIC (0xf53bc0e7ad124589ULL)
  84 #define NTDB_RECOVERY_INVALID_MAGIC (0x0ULL)
  85
  86 /* Capability bits. */
  87 #define NTDB_CAP_TYPE_MASK      0x1FFFFFFFFFFFFFFFULL
  88 #define NTDB_CAP_NOCHECK                0x8000000000000000ULL
  89 #define NTDB_CAP_NOWRITE                0x4000000000000000ULL
  90 #define NTDB_CAP_NOOPEN         0x2000000000000000ULL
  91
  92 #define NTDB_OFF_IS_ERR(off) unlikely(off >= (ntdb_off_t)(long)NTDB_ERR_LAST)
  93 #define NTDB_OFF_TO_ERR(off) ((enum NTDB_ERROR)(long)(off))
  94 #define NTDB_ERR_TO_OFF(ecode) ((ntdb_off_t)(long)(ecode))
  95
  96 /* Packing errors into pointers and v.v. */
  97 #define NTDB_PTR_IS_ERR(ptr)                                            \
  98         unlikely((unsigned long)(ptr) >= (unsigned long)NTDB_ERR_LAST)
  99 #define NTDB_PTR_ERR(p) ((enum NTDB_ERROR)(long)(p))
 100 #define NTDB_ERR_PTR(err) ((void *)(long)(err))
 101
 102 /* This doesn't really need to be pagesize, but we use it for similar
 103  * reasons. */
 104 #define NTDB_PGSIZE 16384
 105
 106 /* Common case of returning true, false or -ve error. */
 107 typedef int ntdb_bool_err;
 108
 109 /* Prevent others from opening the file. */
 110 #define NTDB_OPEN_LOCK 0
 111 /* Expanding file. */
 112 #define NTDB_EXPANSION_LOCK 2
 113 /* Doing a transaction. */
 114 #define NTDB_TRANSACTION_LOCK 8
 115 /* Hash chain locks. */
 116 #define NTDB_HASH_LOCK_START 64
 117
 118 /* Extend file by least 100 times larger than needed. */
 119 #define NTDB_EXTENSION_FACTOR 100
 120
 121 /* We steal this many upper bits, giving a maximum offset of 64 exabytes. */
 122 #define NTDB_OFF_UPPER_STEAL 8
 123
 124 /* And we use the lower bit, too. */
 125 #define NTDB_OFF_CHAIN_BIT      0
 126
 127 /* Hash table sits just after the header. */
 128 #define NTDB_HASH_OFFSET (sizeof(struct ntdb_header))
 129
 130 /* Additional features we understand.  Currently: none. */
 131 #define NTDB_FEATURE_MASK ((uint64_t)0)
 132
 133 /* The bit number where we store the extra hash bits. */
 134 /* Convenience mask to get actual offset. */
 135 #define NTDB_OFF_MASK                                                   \
 136         (((1ULL << (64 - NTDB_OFF_UPPER_STEAL)) - 1) - (1<<NTDB_OFF_CHAIN_BIT))
 137
 138 /* How many buckets in a free list: see size_to_bucket(). */
 139 #define NTDB_FREE_BUCKETS (64 - NTDB_OFF_UPPER_STEAL)
 140
 141 /* We have to be able to fit a free record here. */
 142 #define NTDB_MIN_DATA_LEN                                               \
 143         (sizeof(struct ntdb_free_record) - sizeof(struct ntdb_used_record))
 144
 145 /* Indicates this entry is not on an flist (can happen during coalescing) */
 146 #define NTDB_FTABLE_NONE ((1ULL << NTDB_OFF_UPPER_STEAL) - 1)
 147
 148 /* By default, hash is 64k bytes */
 149 #define NTDB_DEFAULT_HBITS 13
 150
 151 struct ntdb_used_record {
 152         /* For on-disk compatibility, we avoid bitfields:
 153            magic: 16,        (highest)
 154            key_len_bits: 5,
 155            extra_padding: 32
 156         */
 157         uint64_t magic_and_meta;
 158         /* The bottom key_len_bits*2 are key length, rest is data length. */
 159         uint64_t key_and_data_len;
 160 };
 161
 162 static inline unsigned rec_key_bits(const struct ntdb_used_record *r)
 163 {
 164         return ((r->magic_and_meta >> 43) & ((1 << 5)-1)) * 2;
 165 }
 166
 167 static inline uint64_t rec_key_length(const struct ntdb_used_record *r)
 168 {
 169         return r->key_and_data_len & ((1ULL << rec_key_bits(r)) - 1);
 170 }
 171
 172 static inline uint64_t rec_data_length(const struct ntdb_used_record *r)
 173 {
 174         return r->key_and_data_len >> rec_key_bits(r);
 175 }
 176
 177 static inline uint64_t rec_extra_padding(const struct ntdb_used_record *r)
 178 {
 179         return (r->magic_and_meta >> 11) & 0xFFFFFFFF;
 180 }
 181
 182 static inline uint16_t rec_magic(const struct ntdb_used_record *r)
 183 {
 184         return (r->magic_and_meta >> 48);
 185 }
 186
 187 struct ntdb_free_record {
 188         uint64_t magic_and_prev; /* NTDB_OFF_UPPER_STEAL bits magic, then prev */
 189         uint64_t ftable_and_len; /* Len not counting these two fields. */
 190         /* This is why the minimum record size is 8 bytes.  */
 191         uint64_t next;
 192 };
 193
 194 static inline uint64_t frec_prev(const struct ntdb_free_record *f)
 195 {
 196         return f->magic_and_prev & ((1ULL << (64 - NTDB_OFF_UPPER_STEAL)) - 1);
 197 }
 198
 199 static inline uint64_t frec_magic(const struct ntdb_free_record *f)
 200 {
 201         return f->magic_and_prev >> (64 - NTDB_OFF_UPPER_STEAL);
 202 }
 203
 204 static inline uint64_t frec_len(const struct ntdb_free_record *f)
 205 {
 206         return f->ftable_and_len & ((1ULL << (64 - NTDB_OFF_UPPER_STEAL))-1);
 207 }
 208
 209 static inline unsigned frec_ftable(const struct ntdb_free_record *f)
 210 {
 211         return f->ftable_and_len >> (64 - NTDB_OFF_UPPER_STEAL);
 212 }
 213
 214 struct ntdb_recovery_record {
 215         uint64_t magic;
 216         /* Length of record (add this header to get total length). */
 217         uint64_t max_len;
 218         /* Length used. */
 219         uint64_t len;
 220         /* Old length of file before transaction. */
 221         uint64_t eof;
 222 };
 223
 224 /* this is stored at the front of every database */
 225 struct ntdb_header {
 226         char magic_food[64]; /* for /etc/magic */
 227         /* FIXME: Make me 32 bit? */
 228         uint64_t version; /* version of the code */
 229         uint64_t hash_bits; /* bits for toplevel hash table. */
 230         uint64_t hash_test; /* result of hashing HASH_MAGIC. */
 231         uint64_t hash_seed; /* "random" seed written at creation time. */
 232         ntdb_off_t free_table; /* (First) free table. */
 233         ntdb_off_t recovery; /* Transaction recovery area. */
 234
 235         uint64_t features_used; /* Features all writers understand */
 236         uint64_t features_offered; /* Features offered */
 237
 238         uint64_t seqnum; /* Sequence number for NTDB_SEQNUM */
 239
 240         ntdb_off_t capabilities; /* Optional linked list of capabilities. */
 241         ntdb_off_t reserved[22];
 242
 243         /*
 244          * Hash table is next:
 245          *
 246          * struct ntdb_used_record htable_hdr;
 247          * ntdb_off_t htable[1 << hash_bits];
 248          */
 249 };
 250
 251 struct ntdb_freetable {
 252         struct ntdb_used_record hdr;
 253         ntdb_off_t next;
 254         ntdb_off_t buckets[NTDB_FREE_BUCKETS];
 255 };
 256
 257 struct ntdb_capability {
 258         struct ntdb_used_record hdr;
 259         ntdb_off_t type;
 260         ntdb_off_t next;
 261         /* ... */
 262 };
 263
 264 /* Information about a particular (locked) hash entry. */
 265 struct hash_info {
 266         /* Full hash value of entry. */
 267         uint32_t h;
 268         /* Start of hash table / chain. */
 269         ntdb_off_t table;
 270         /* Number of entries in this table/chain. */
 271         ntdb_off_t table_size;
 272         /* Bucket we (or an empty space) were found in. */
 273         ntdb_off_t bucket;
 274         /* Old value that was in that entry (if not found) */
 275         ntdb_off_t old_val;
 276 };
 277
 278 enum ntdb_lock_flags {
 279         /* WAIT == F_SETLKW, NOWAIT == F_SETLK */
 280         NTDB_LOCK_NOWAIT = 0,
 281         NTDB_LOCK_WAIT = 1,
 282         /* If set, don't log an error on failure. */
 283         NTDB_LOCK_PROBE = 2,
 284         /* If set, don't check for recovery (used by recovery code). */
 285         NTDB_LOCK_NOCHECK = 4,
 286 };
 287
 288 struct ntdb_lock {
 289         struct ntdb_context *owner;
 290         off_t off;
 291         uint32_t count;
 292         uint32_t ltype;
 293 };
 294
 295 /* This is only needed for ntdb_access_commit, but used everywhere to
 296  * simplify. */
 297 struct ntdb_access_hdr {
 298         struct ntdb_access_hdr *next;
 299         ntdb_off_t off;
 300         ntdb_len_t len;
 301         bool convert;
 302 };
 303
 304 /* mmaps we are keeping around because they are still direct accessed */
 305 struct ntdb_old_mmap {
 306         struct ntdb_old_mmap *next;
 307
 308         void *map_ptr;
 309         ntdb_len_t map_size;
 310 };
 311
 312 struct ntdb_file {
 313         /* How many are sharing us? */
 314         unsigned int refcnt;
 315
 316         /* Mmap (if any), or malloc (for NTDB_INTERNAL). */
 317         void *map_ptr;
 318
 319         /* How much space has been mapped (<= current file size) */
 320         ntdb_len_t map_size;
 321
 322         /* The file descriptor (-1 for NTDB_INTERNAL). */
 323         int fd;
 324
 325         /* How many are accessing directly? */
 326         unsigned int direct_count;
 327
 328         /* Old maps, still direct accessed. */
 329         struct ntdb_old_mmap *old_mmaps;
 330
 331         /* Lock information */
 332         pid_t locker;
 333         struct ntdb_lock allrecord_lock;
 334         size_t num_lockrecs;
 335         struct ntdb_lock *lockrecs;
 336
 337         /* Identity of this file. */
 338         dev_t device;
 339         ino_t inode;
 340 };
 341
 342 struct ntdb_methods {
 343         enum NTDB_ERROR (*tread)(struct ntdb_context *, ntdb_off_t, void *,
 344                                  ntdb_len_t);
 345         enum NTDB_ERROR (*twrite)(struct ntdb_context *, ntdb_off_t, const void *,
 346                                   ntdb_len_t);
 347         enum NTDB_ERROR (*oob)(struct ntdb_context *, ntdb_off_t, ntdb_len_t, bool);
 348         enum NTDB_ERROR (*expand_file)(struct ntdb_context *, ntdb_len_t);
 349         void *(*direct)(struct ntdb_context *, ntdb_off_t, size_t, bool);
 350         ntdb_off_t (*read_off)(struct ntdb_context *ntdb, ntdb_off_t off);
 351         enum NTDB_ERROR (*write_off)(struct ntdb_context *ntdb, ntdb_off_t off,
 352                                      ntdb_off_t val);
 353 };
 354
 355 /*
 356   internal prototypes
 357 */
 358 /* Get bits from a value. */
 359 static inline uint32_t bits_from(uint64_t val, unsigned start, unsigned num)
 360 {
 361         assert(num <= 32);
 362         return (val >> start) & ((1U << num) - 1);
 363 }
 364
 365
 366 /* hash.c: */
 367 uint32_t ntdb_jenkins_hash(const void *key, size_t length, uint32_t seed,
 368                            void *unused);
 369
 370 enum NTDB_ERROR first_in_hash(struct ntdb_context *ntdb,
 371                               struct hash_info *h,
 372                               NTDB_DATA *kbuf, size_t *dlen);
 373
 374 enum NTDB_ERROR next_in_hash(struct ntdb_context *ntdb,
 375                              struct hash_info *h,
 376                              NTDB_DATA *kbuf, size_t *dlen);
 377
 378 /* Hash random memory. */
 379 uint32_t ntdb_hash(struct ntdb_context *ntdb, const void *ptr, size_t len);
 380
 381 /* Find and lock a hash entry (or where it would be). */
 382 ntdb_off_t find_and_lock(struct ntdb_context *ntdb,
 383                          NTDB_DATA key,
 384                          int ltype,
 385                          struct hash_info *h,
 386                          struct ntdb_used_record *rec,
 387                          const char **rkey);
 388
 389 enum NTDB_ERROR replace_in_hash(struct ntdb_context *ntdb,
 390                                 const struct hash_info *h,
 391                                 ntdb_off_t new_off);
 392
 393 enum NTDB_ERROR add_to_hash(struct ntdb_context *ntdb,
 394                             const struct hash_info *h,
 395                             ntdb_off_t new_off);
 396
 397 enum NTDB_ERROR delete_from_hash(struct ntdb_context *ntdb,
 398                                  const struct hash_info *h);
 399
 400 /* For ntdb_check */
 401 bool is_subhash(ntdb_off_t val);
 402 enum NTDB_ERROR unknown_capability(struct ntdb_context *ntdb, const char *caller,
 403                                    ntdb_off_t type);
 404
 405 /* free.c: */
 406 enum NTDB_ERROR ntdb_ftable_init(struct ntdb_context *ntdb);
 407
 408 /* check.c needs these to iterate through free lists. */
 409 ntdb_off_t first_ftable(struct ntdb_context *ntdb);
 410 ntdb_off_t next_ftable(struct ntdb_context *ntdb, ntdb_off_t ftable);
 411
 412 /* This returns space or -ve error number. */
 413 ntdb_off_t alloc(struct ntdb_context *ntdb, size_t keylen, size_t datalen,
 414                  unsigned magic, bool growing);
 415
 416 /* Put this record in a free list. */
 417 enum NTDB_ERROR add_free_record(struct ntdb_context *ntdb,
 418                                 ntdb_off_t off, ntdb_len_t len_with_header,
 419                                 enum ntdb_lock_flags waitflag,
 420                                 bool coalesce_ok);
 421
 422 /* Set up header for a used/ftable/htable/chain/capability record. */
 423 enum NTDB_ERROR set_header(struct ntdb_context *ntdb,
 424                            struct ntdb_used_record *rec,
 425                            unsigned magic, uint64_t keylen, uint64_t datalen,
 426                            uint64_t actuallen);
 427
 428 /* Used by ntdb_check to verify. */
 429 unsigned int size_to_bucket(ntdb_len_t data_len);
 430 ntdb_off_t bucket_off(ntdb_off_t ftable_off, unsigned bucket);
 431
 432 /* Used by ntdb_summary */
 433 ntdb_off_t dead_space(struct ntdb_context *ntdb, ntdb_off_t off);
 434
 435 /* Adjust expansion, used by create_recovery_area */
 436 ntdb_off_t ntdb_expand_adjust(ntdb_off_t map_size, ntdb_off_t size);
 437
 438 /* io.c: */
 439 /* Initialize ntdb->methods. */
 440 void ntdb_io_init(struct ntdb_context *ntdb);
 441
 442 /* Convert endian of the buffer if required. */
 443 void *ntdb_convert(const struct ntdb_context *ntdb, void *buf, ntdb_len_t size);
 444
 445 /* Unmap and try to map the ntdb. */
 446 enum NTDB_ERROR ntdb_munmap(struct ntdb_context *ntdb);
 447 enum NTDB_ERROR ntdb_mmap(struct ntdb_context *ntdb);
 448
 449 /* Either alloc a copy, or give direct access.  Release frees or noop. */
 450 const void *ntdb_access_read(struct ntdb_context *ntdb,
 451                              ntdb_off_t off, ntdb_len_t len, bool convert);
 452 void *ntdb_access_write(struct ntdb_context *ntdb,
 453                         ntdb_off_t off, ntdb_len_t len, bool convert);
 454
 455 /* Release result of ntdb_access_read/write. */
 456 void ntdb_access_release(struct ntdb_context *ntdb, const void *p);
 457 /* Commit result of ntdb_acces_write. */
 458 enum NTDB_ERROR ntdb_access_commit(struct ntdb_context *ntdb, void *p);
 459
 460 /* Clear an ondisk area. */
 461 enum NTDB_ERROR zero_out(struct ntdb_context *ntdb, ntdb_off_t off, ntdb_len_t len);
 462
 463 /* Return a non-zero offset between >= start < end in this array (or end). */
 464 ntdb_off_t ntdb_find_nonzero_off(struct ntdb_context *ntdb,
 465                                  ntdb_off_t base,
 466                                  uint64_t start,
 467                                  uint64_t end);
 468
 469 /* Return a zero offset in this array, or num. */
 470 ntdb_off_t ntdb_find_zero_off(struct ntdb_context *ntdb, ntdb_off_t off,
 471                               uint64_t num);
 472
 473 /* Allocate and make a copy of some offset. */
 474 void *ntdb_alloc_read(struct ntdb_context *ntdb, ntdb_off_t offset, ntdb_len_t len);
 475
 476 /* Writes a converted copy of a record. */
 477 enum NTDB_ERROR ntdb_write_convert(struct ntdb_context *ntdb, ntdb_off_t off,
 478                                    const void *rec, size_t len);
 479
 480 /* Reads record and converts it */
 481 enum NTDB_ERROR ntdb_read_convert(struct ntdb_context *ntdb, ntdb_off_t off,
 482                                   void *rec, size_t len);
 483
 484 /* Bump the seqnum (caller checks for ntdb->flags & NTDB_SEQNUM) */
 485 void ntdb_inc_seqnum(struct ntdb_context *ntdb);
 486
 487 /* lock.c: */
 488 /* Print message because another ntdb owns a lock we want. */
 489 enum NTDB_ERROR owner_conflict(struct ntdb_context *ntdb, const char *call);
 490
 491 /* If we fork, we no longer really own locks. */
 492 bool check_lock_pid(struct ntdb_context *ntdb, const char *call, bool log);
 493
 494 /* Lock/unlock a hash bucket. */
 495 enum NTDB_ERROR ntdb_lock_hash(struct ntdb_context *ntdb,
 496                                unsigned int hbucket,
 497                                int ltype);
 498 enum NTDB_ERROR ntdb_unlock_hash(struct ntdb_context *ntdb,
 499                                  unsigned int hash, int ltype);
 500
 501 /* For closing the file. */
 502 void ntdb_lock_cleanup(struct ntdb_context *ntdb);
 503
 504 /* Lock/unlock a particular free bucket. */
 505 enum NTDB_ERROR ntdb_lock_free_bucket(struct ntdb_context *ntdb, ntdb_off_t b_off,
 506                                       enum ntdb_lock_flags waitflag);
 507 void ntdb_unlock_free_bucket(struct ntdb_context *ntdb, ntdb_off_t b_off);
 508
 509 /* Serialize transaction start. */
 510 enum NTDB_ERROR ntdb_transaction_lock(struct ntdb_context *ntdb, int ltype);
 511 void ntdb_transaction_unlock(struct ntdb_context *ntdb, int ltype);
 512
 513 /* Do we have any hash locks (ie. via ntdb_chainlock) ? */
 514 bool ntdb_has_hash_locks(struct ntdb_context *ntdb);
 515
 516 /* Lock entire database. */
 517 enum NTDB_ERROR ntdb_allrecord_lock(struct ntdb_context *ntdb, int ltype,
 518                                     enum ntdb_lock_flags flags, bool upgradable);
 519 void ntdb_allrecord_unlock(struct ntdb_context *ntdb, int ltype);
 520 enum NTDB_ERROR ntdb_allrecord_upgrade(struct ntdb_context *ntdb, off_t start);
 521
 522 /* Serialize db open. */
 523 enum NTDB_ERROR ntdb_lock_open(struct ntdb_context *ntdb,
 524                                int ltype, enum ntdb_lock_flags flags);
 525 void ntdb_unlock_open(struct ntdb_context *ntdb, int ltype);
 526 bool ntdb_has_open_lock(struct ntdb_context *ntdb);
 527
 528 /* Serialize db expand. */
 529 enum NTDB_ERROR ntdb_lock_expand(struct ntdb_context *ntdb, int ltype);
 530 void ntdb_unlock_expand(struct ntdb_context *ntdb, int ltype);
 531 bool ntdb_has_expansion_lock(struct ntdb_context *ntdb);
 532
 533 /* If it needs recovery, grab all the locks and do it. */
 534 enum NTDB_ERROR ntdb_lock_and_recover(struct ntdb_context *ntdb);
 535
 536 /* Default lock and unlock functions. */
 537 int ntdb_fcntl_lock(int fd, int rw, off_t off, off_t len, bool waitflag, void *);
 538 int ntdb_fcntl_unlock(int fd, int rw, off_t off, off_t len, void *);
 539
 540 /* transaction.c: */
 541 enum NTDB_ERROR ntdb_transaction_recover(struct ntdb_context *ntdb);
 542 ntdb_bool_err ntdb_needs_recovery(struct ntdb_context *ntdb);
 543
 544 struct ntdb_context {
 545         /* Single list of all TDBs, to detect multiple opens. */
 546         struct ntdb_context *next;
 547
 548         /* Filename of the database. */
 549         const char *name;
 550
 551         /* Logging function */
 552         void (*log_fn)(struct ntdb_context *ntdb,
 553                        enum ntdb_log_level level,
 554                        enum NTDB_ERROR ecode,
 555                        const char *message,
 556                        void *data);
 557         void *log_data;
 558
 559         /* Open flags passed to ntdb_open. */
 560         int open_flags;
 561
 562         /* low level (fnctl) lock functions. */
 563         int (*lock_fn)(int fd, int rw, off_t off, off_t len, bool w, void *);
 564         int (*unlock_fn)(int fd, int rw, off_t off, off_t len, void *);
 565         void *lock_data;
 566
 567         /* the ntdb flags passed to ntdb_open. */
 568         uint32_t flags;
 569
 570         /* Our statistics. */
 571         struct ntdb_attribute_stats stats;
 572
 573         /* The actual file information */
 574         struct ntdb_file *file;
 575
 576         /* Hash function. */
 577         uint32_t (*hash_fn)(const void *key, size_t len, uint32_t seed, void *);
 578         void *hash_data;
 579         uint32_t hash_seed;
 580         /* Bits in toplevel hash table. */
 581         unsigned int hash_bits;
 582
 583         /* Allocate and free functions. */
 584         void *(*alloc_fn)(const void *owner, size_t len, void *priv_data);
 585         void *(*expand_fn)(void *old, size_t newlen, void *priv_data);
 586         void (*free_fn)(void *old, void *priv_data);
 587         void *alloc_data;
 588
 589         /* Our open hook, if any. */
 590         enum NTDB_ERROR (*openhook)(int fd, void *data);
 591         void *openhook_data;
 592
 593         /* Set if we are in a transaction. */
 594         struct ntdb_transaction *transaction;
 595
 596         /* What free table are we using? */
 597         ntdb_off_t ftable_off;
 598         unsigned int ftable;
 599
 600         /* IO methods: changes for transactions. */
 601         const struct ntdb_methods *io;
 602
 603         /* Direct access information */
 604         struct ntdb_access_hdr *access;
 605 };
 606
 607 /* ntdb.c: */
 608 enum NTDB_ERROR COLD PRINTF_FMT(4, 5)
 609         ntdb_logerr(struct ntdb_context *ntdb,
 610                     enum NTDB_ERROR ecode,
 611                     enum ntdb_log_level level,
 612                     const char *fmt, ...);
 613
 614 static inline enum NTDB_ERROR ntdb_oob(struct ntdb_context *ntdb,
 615                                        ntdb_off_t off, ntdb_len_t len,
 616                                        bool probe)
 617 {
 618         if (likely(off + len >= off)
 619             && likely(off + len <= ntdb->file->map_size)
 620             && likely(!probe)) {
 621                     return NTDB_SUCCESS;
 622         }
 623         return ntdb->io->oob(ntdb, off, len, probe);
 624 }
 625
 626 /* Convenience routine to get an offset. */
 627 static inline ntdb_off_t ntdb_read_off(struct ntdb_context *ntdb,
 628                                        ntdb_off_t off)
 629 {
 630         return ntdb->io->read_off(ntdb, off);
 631 }
 632
 633 /* Write an offset at an offset. */
 634 static inline enum NTDB_ERROR ntdb_write_off(struct ntdb_context *ntdb,
 635                                              ntdb_off_t off,
 636                                ntdb_off_t val)
 637 {
 638         return ntdb->io->write_off(ntdb, off, val);
 639 }
 640
 641 #ifdef NTDB_TRACE
 642 void ntdb_trace(struct ntdb_context *ntdb, const char *op);
 643 void ntdb_trace_seqnum(struct ntdb_context *ntdb, uint32_t seqnum, const char *op);
 644 void ntdb_trace_open(struct ntdb_context *ntdb, const char *op,
 645                      unsigned hash_size, unsigned ntdb_flags, unsigned open_flags);
 646 void ntdb_trace_ret(struct ntdb_context *ntdb, const char *op, int ret);
 647 void ntdb_trace_retrec(struct ntdb_context *ntdb, const char *op, NTDB_DATA ret);
 648 void ntdb_trace_1rec(struct ntdb_context *ntdb, const char *op,
 649                      NTDB_DATA rec);
 650 void ntdb_trace_1rec_ret(struct ntdb_context *ntdb, const char *op,
 651                          NTDB_DATA rec, int ret);
 652 void ntdb_trace_1rec_retrec(struct ntdb_context *ntdb, const char *op,
 653                             NTDB_DATA rec, NTDB_DATA ret);
 654 void ntdb_trace_2rec_flag_ret(struct ntdb_context *ntdb, const char *op,
 655                               NTDB_DATA rec1, NTDB_DATA rec2, unsigned flag,
 656                               int ret);
 657 void ntdb_trace_2rec_retrec(struct ntdb_context *ntdb, const char *op,
 658                             NTDB_DATA rec1, NTDB_DATA rec2, NTDB_DATA ret);
 659 #else
 660 #define ntdb_trace(ntdb, op)
 661 #define ntdb_trace_seqnum(ntdb, seqnum, op)
 662 #define ntdb_trace_open(ntdb, op, hash_size, ntdb_flags, open_flags)
 663 #define ntdb_trace_ret(ntdb, op, ret)
 664 #define ntdb_trace_retrec(ntdb, op, ret)
 665 #define ntdb_trace_1rec(ntdb, op, rec)
 666 #define ntdb_trace_1rec_ret(ntdb, op, rec, ret)
 667 #define ntdb_trace_1rec_retrec(ntdb, op, rec, ret)
 668 #define ntdb_trace_2rec_flag_ret(ntdb, op, rec1, rec2, flag, ret)
 669 #define ntdb_trace_2rec_retrec(ntdb, op, rec1, rec2, ret)
 670 #endif /* !NTDB_TRACE */
 671
 672 #endif