33 #if (DEBUG +0) && defined(__GNUC__)
34 # define DPRINTF(fmt, ...) \
35 fprintf(stderr, "%s:%d: " fmt "\n", __func__, __LINE__, ##__VA_ARGS__)
37 # define DPRINTF(...) ((void) 0)
42 #define MDB_MAGIC 0xBEEFC0DE
44 #define MAXKEYSIZE 255
46 #define P_INVALID (~0UL)
48 #define F_ISSET(w, f) (((w) & (f)) == (f))
51 typedef uint16_t indx_t;
53 #define DEFAULT_READERS 126
54 #define DEFAULT_MAPSIZE 1048576
56 /* Lock descriptor stuff */
61 typedef struct MDB_rxbody {
66 #define CACHELINE 64 /* most CPUs. Itanium uses 128 */
69 typedef struct MDB_reader {
71 /* cache line alignment */
72 char pad[CACHELINE-sizeof(MDB_rxbody)];
77 uint32_t mt_version; \
78 pthread_mutex_t mt_mutex; \
80 uint32_t mt_numreaders
81 typedef struct MDB_txbody {
85 typedef struct MDB_txninfo {
87 char pad[CACHELINE-sizeof(MDB_txbody)];
88 pthread_mutex_t mt_wmutex;
89 char pad2[CACHELINE-sizeof(pthread_mutex_t)];
90 MDB_reader mt_readers[1];
93 /* Common header for all page types. Overflow pages
94 * occupy a number of contiguous pages with no
95 * headers on any page after the first.
97 typedef struct MDB_page { /* represents a page of storage */
98 pgno_t mp_pgno; /* page number */
99 #define P_BRANCH 0x01 /* branch page */
100 #define P_LEAF 0x02 /* leaf page */
101 #define P_OVERFLOW 0x04 /* overflow page */
102 #define P_META 0x08 /* meta page */
103 #define P_DIRTY 0x10 /* dirty page */
105 #define mp_lower mp_pb.pb.pb_lower
106 #define mp_upper mp_pb.pb.pb_upper
107 #define mp_pages mp_pb.pb_pages
110 indx_t pb_lower; /* lower bound of free space */
111 indx_t pb_upper; /* upper bound of free space */
113 uint32_t pb_pages; /* number of overflow pages */
115 indx_t mp_ptrs[1]; /* dynamic size */
118 #define PAGEHDRSZ ((unsigned) offsetof(MDB_page, mp_ptrs))
120 #define NUMKEYS(p) (((p)->mp_lower - PAGEHDRSZ) >> 1)
121 #define SIZELEFT(p) (indx_t)((p)->mp_upper - (p)->mp_lower)
122 #define PAGEFILL(env, p) (1000L * ((env)->me_meta.mm_stat.ms_psize - PAGEHDRSZ - SIZELEFT(p)) / \
123 ((env)->me_meta.mm_stat.ms_psize - PAGEHDRSZ))
124 #define IS_LEAF(p) F_ISSET((p)->mp_flags, P_LEAF)
125 #define IS_BRANCH(p) F_ISSET((p)->mp_flags, P_BRANCH)
126 #define IS_OVERFLOW(p) F_ISSET((p)->mp_flags, P_OVERFLOW)
128 #define OVPAGES(size, psize) (PAGEHDRSZ + size + psize - 1) / psize;
130 typedef struct MDB_meta { /* meta (footer) page content */
133 void *mm_address; /* address for fixed mapping */
134 size_t mm_mapsize; /* size of mmap region */
135 pgno_t mm_last_pg; /* last used page in file */
136 ulong mm_txnid; /* txnid that committed this page */
138 pgno_t mm_root; /* page number of root page */
141 typedef struct MDB_dhead { /* a dirty page */
142 SIMPLEQ_ENTRY(MDB_dpage) md_next; /* queue of dirty pages */
144 unsigned md_pi; /* parent index */
148 typedef struct MDB_dpage {
153 SIMPLEQ_HEAD(dirty_queue, MDB_dpage);
155 typedef struct MDB_oldpages {
156 struct MDB_oldpages *mo_next;
158 pgno_t mo_pages[1]; /* dynamic */
161 typedef struct MDB_pageparent {
167 static MDB_dpage *mdb_newpage(MDB_txn *txn, MDB_page *parent, unsigned int parent_idx, int num);
168 static int mdb_touch(MDB_txn *txn, MDB_pageparent *mp);
170 typedef struct MDB_ppage { /* ordered list of pages */
171 SLIST_ENTRY(MDB_ppage) mp_entry;
173 unsigned int mp_ki; /* cursor index on page */
175 SLIST_HEAD(page_stack, MDB_ppage);
177 #define CURSOR_EMPTY(c) SLIST_EMPTY(&(c)->mc_stack)
178 #define CURSOR_TOP(c) SLIST_FIRST(&(c)->mc_stack)
179 #define CURSOR_POP(c) SLIST_REMOVE_HEAD(&(c)->mc_stack, mp_entry)
180 #define CURSOR_PUSH(c,p) SLIST_INSERT_HEAD(&(c)->mc_stack, p, mp_entry)
185 struct page_stack mc_stack; /* stack of parent pages */
186 short mc_initialized; /* 1 if initialized */
187 short mc_eof; /* 1 if end is reached */
190 #define METAHASHLEN offsetof(MDB_meta, mm_hash)
191 #define METADATA(p) ((void *)((char *)p + PAGEHDRSZ))
193 typedef struct MDB_node {
194 #define mn_pgno mn_p.np_pgno
195 #define mn_dsize mn_p.np_dsize
197 pgno_t np_pgno; /* child page number */
198 uint32_t np_dsize; /* leaf data size */
200 unsigned int mn_flags:4;
201 unsigned int mn_ksize:12; /* key size */
202 #define F_BIGDATA 0x01 /* data put on overflow page */
207 pgno_t mt_root; /* current / new root page */
208 pgno_t mt_next_pgno; /* next unallocated page */
209 pgno_t mt_first_pgno;
213 pgno_t *mt_free_pgs; /* this is an IDL */
215 struct dirty_queue *dirty_queue; /* modified pages */
218 #define MDB_TXN_RDONLY 0x01 /* read-only transaction */
219 #define MDB_TXN_ERROR 0x02 /* an error has occurred */
220 unsigned int mt_flags;
223 /* Must be same as MDB_db, minus md_root/md_stat */
224 typedef struct MDB_db0 {
225 unsigned int md_flags;
226 MDB_cmp_func *md_cmp; /* user compare function */
227 MDB_rel_func *md_rel; /* user relocate function */
228 MDB_db *md_parent; /* parent tree */
233 unsigned int md_flags;
234 MDB_cmp_func *md_cmp; /* user compare function */
235 MDB_rel_func *md_rel; /* user relocate function */
236 MDB_db *md_parent; /* parent tree */
239 pgno_t md_root; /* page number of root page */
250 MDB_txninfo *me_txns;
251 MDB_db0 me_db; /* first DB, overlaps with meta */
253 MDB_txn *me_txn; /* current write transaction */
255 off_t me_size; /* current file size */
256 pthread_key_t me_txkey; /* thread-key for readers */
257 MDB_oldpages *me_pghead;
258 MDB_oldpages *me_pgtail;
261 #define NODESIZE offsetof(MDB_node, mn_data)
263 #define INDXSIZE(k) (NODESIZE + ((k) == NULL ? 0 : (k)->mv_size))
264 #define LEAFSIZE(k, d) (NODESIZE + (k)->mv_size + (d)->mv_size)
265 #define NODEPTR(p, i) ((MDB_node *)((char *)(p) + (p)->mp_ptrs[i]))
266 #define NODEKEY(node) (void *)((node)->mn_data)
267 #define NODEDATA(node) (void *)((char *)(node)->mn_data + (node)->mn_ksize)
268 #define NODEPGNO(node) ((node)->mn_pgno)
269 #define NODEDSZ(node) ((node)->mn_dsize)
271 #define MDB_COMMIT_PAGES 64 /* max number of pages to write in one commit */
272 #define MDB_MAXCACHE_DEF 1024 /* max number of pages to keep in cache */
274 static int mdb_search_page_root(MDB_db *db,
276 MDB_cursor *cursor, int modify,
277 MDB_pageparent *mpp);
278 static int mdb_search_page(MDB_db *db,
279 MDB_txn *txn, MDB_val *key,
280 MDB_cursor *cursor, int modify,
281 MDB_pageparent *mpp);
283 static int mdbenv_read_header(MDB_env *env);
284 static int mdb_check_meta_page(MDB_page *p);
285 static int mdbenv_read_meta(MDB_env *env);
286 static int mdbenv_write_meta(MDB_txn *txn);
287 static MDB_page *mdbenv_get_page(MDB_env *env, pgno_t pgno);
289 static MDB_node *mdb_search_node(MDB_db *db, MDB_page *mp,
290 MDB_val *key, int *exactp, unsigned int *kip);
291 static int mdb_add_node(MDB_db *bt, MDB_page *mp,
292 indx_t indx, MDB_val *key, MDB_val *data,
293 pgno_t pgno, uint8_t flags);
294 static void mdb_del_node(MDB_db *bt, MDB_page *mp,
296 static int mdb_read_data(MDB_db *bt, MDB_page *mp,
297 MDB_node *leaf, MDB_val *data);
299 static int mdb_rebalance(MDB_db *bt, MDB_pageparent *mp);
300 static int mdb_update_key(MDB_db *bt, MDB_page *mp,
301 indx_t indx, MDB_val *key);
302 static int mdb_move_node(MDB_db *bt,
303 MDB_pageparent *src, indx_t srcindx,
304 MDB_pageparent *dst, indx_t dstindx);
305 static int mdb_merge(MDB_db *bt, MDB_pageparent *src,
306 MDB_pageparent *dst);
307 static int mdb_split(MDB_db *bt, MDB_page **mpp,
308 unsigned int *newindxp, MDB_val *newkey,
309 MDB_val *newdata, pgno_t newpgno);
310 static MDB_dpage *mdbenv_new_page(MDB_env *env, uint32_t flags, int num);
312 static void cursor_pop_page(MDB_cursor *cursor);
313 static MDB_ppage *cursor_push_page(MDB_cursor *cursor,
316 static int mdb_set_key(MDB_db *bt, MDB_page *mp,
317 MDB_node *node, MDB_val *key);
318 static int mdb_sibling(MDB_cursor *cursor, int move_right);
319 static int mdb_cursor_next(MDB_cursor *cursor,
320 MDB_val *key, MDB_val *data);
321 static int mdb_cursor_set(MDB_cursor *cursor,
322 MDB_val *key, MDB_val *data, int *exactp);
323 static int mdb_cursor_first(MDB_cursor *cursor,
324 MDB_val *key, MDB_val *data);
326 static size_t mdb_leaf_size(MDB_db *bt, MDB_val *key,
328 static size_t mdb_branch_size(MDB_db *bt, MDB_val *key);
330 static pgno_t mdbenv_compact_tree(MDB_env *env, pgno_t pgno,
333 static int memncmp(const void *s1, size_t n1,
334 const void *s2, size_t n2);
335 static int memnrcmp(const void *s1, size_t n1,
336 const void *s2, size_t n2);
339 memncmp(const void *s1, size_t n1, const void *s2, size_t n2)
341 int diff, len_diff = -1;
344 len_diff = (n1 > n2);
347 diff = memcmp(s1, s2, n1);
348 return diff ? diff : len_diff;
352 memnrcmp(const void *s1, size_t n1, const void *s2, size_t n2)
354 const unsigned char *p1, *p2, *p1_lim;
361 p1 = (const unsigned char *)s1 + n1 - 1;
362 p2 = (const unsigned char *)s2 + n2 - 1;
364 for (p1_lim = (n1 <= n2 ? s1 : s2); *p1 == *p2; p1--, p2--) {
366 return (p1 != s1) ? (p1 != p2) : (p2 != s2) ? -1 : 0;
372 mdb_cmp(MDB_db *db, const MDB_val *a, const MDB_val *b)
374 return db->md_cmp(a, b);
378 _mdb_cmp(MDB_db *db, const MDB_val *key1, const MDB_val *key2)
380 if (F_ISSET(db->md_flags, MDB_REVERSEKEY))
381 return memnrcmp(key1->mv_data, key1->mv_size, key2->mv_data, key2->mv_size);
383 return memncmp((char *)key1->mv_data, key1->mv_size, key2->mv_data, key2->mv_size);
386 /* Allocate new page(s) for writing */
388 mdb_newpage(MDB_txn *txn, MDB_page *parent, unsigned int parent_idx, int num)
391 pgno_t pgno = P_INVALID;
393 if (txn->mt_env->me_pghead) {
394 ulong oldest = txn->mt_txnid - 2;
396 for (i=0; i<txn->mt_env->me_txns->mt_numreaders; i++) {
397 if (txn->mt_env->me_txns->mt_readers[i].mr_txnid < oldest)
398 oldest = txn->mt_env->me_txns->mt_readers[i].mr_txnid;
400 if (oldest > txn->mt_env->me_pghead->mo_txnid) {
401 MDB_oldpages *mop = txn->mt_env->me_pghead;
402 txn->mt_oldest = oldest;
407 /* peel pages off tail, so we only have to truncate the list */
408 pgno = MDB_IDL_LAST(mop->mo_pages);
409 if (MDB_IDL_IS_RANGE(mop->mo_pages)) {
411 if (mop->mo_pages[2] > mop->mo_pages[1])
412 mop->mo_pages[0] = 0;
416 if (MDB_IDL_IS_ZERO(mop->mo_pages)) {
417 txn->mt_env->me_pghead = mop->mo_next;
418 if (!txn->mt_env->me_pghead)
419 txn->mt_env->me_pgtail = NULL;
426 if ((dp = malloc(txn->mt_env->me_meta.mm_stat.ms_psize * num + sizeof(MDB_dhead))) == NULL)
429 dp->h.md_parent = parent;
430 dp->h.md_pi = parent_idx;
431 SIMPLEQ_INSERT_TAIL(txn->mt_u.dirty_queue, dp, h.md_next);
432 if (pgno == P_INVALID) {
433 dp->p.mp_pgno = txn->mt_next_pgno;
434 txn->mt_next_pgno += num;
436 dp->p.mp_pgno = pgno;
442 /* Touch a page: make it dirty and re-insert into tree with updated pgno.
445 mdb_touch(MDB_txn *txn, MDB_pageparent *pp)
447 MDB_page *mp = pp->mp_page;
452 if (!F_ISSET(mp->mp_flags, P_DIRTY)) {
454 if ((dp = mdb_newpage(txn, pp->mp_parent, pp->mp_pi, 1)) == NULL)
456 DPRINTF("touched page %lu -> %lu", mp->mp_pgno, dp->p.mp_pgno);
457 mdb_idl_insert(txn->mt_free_pgs, mp->mp_pgno);
458 pgno = dp->p.mp_pgno;
459 bcopy(mp, &dp->p, txn->mt_env->me_meta.mm_stat.ms_psize);
462 mp->mp_flags |= P_DIRTY;
464 /* Update the page number to new touched page. */
465 if (pp->mp_parent != NULL)
466 NODEPGNO(NODEPTR(pp->mp_parent, pp->mp_pi)) = mp->mp_pgno;
473 mdbenv_sync(MDB_env *env)
476 if (!F_ISSET(env->me_flags, MDB_NOSYNC)) {
477 if (fsync(env->me_fd))
484 mdb_txn_begin(MDB_env *env, int rdonly, MDB_txn **ret)
489 if ((txn = calloc(1, sizeof(*txn))) == NULL) {
490 DPRINTF("calloc: %s", strerror(errno));
495 txn->mt_flags |= MDB_TXN_RDONLY;
497 txn->mt_u.dirty_queue = calloc(1, sizeof(*txn->mt_u.dirty_queue));
498 if (txn->mt_u.dirty_queue == NULL) {
502 SIMPLEQ_INIT(txn->mt_u.dirty_queue);
504 pthread_mutex_lock(&env->me_txns->mt_wmutex);
505 env->me_txns->mt_txnid++;
506 txn->mt_free_pgs = malloc(MDB_IDL_UM_SIZEOF);
507 if (txn->mt_free_pgs == NULL) {
508 free(txn->mt_u.dirty_queue);
512 txn->mt_free_pgs[0] = 0;
514 txn->mt_txnid = env->me_txns->mt_txnid;
516 MDB_reader *r = pthread_getspecific(env->me_txkey);
519 pthread_mutex_lock(&env->me_txns->mt_mutex);
520 for (i=0; i<env->me_maxreaders; i++) {
521 if (env->me_txns->mt_readers[i].mr_pid == 0) {
522 env->me_txns->mt_readers[i].mr_pid = getpid();
523 env->me_txns->mt_readers[i].mr_tid = pthread_self();
524 pthread_setspecific(env->me_txkey, &env->me_txns->mt_readers[i]);
525 if (i >= env->me_txns->mt_numreaders)
526 env->me_txns->mt_numreaders = i+1;
530 pthread_mutex_unlock(&env->me_txns->mt_mutex);
531 if (i == env->me_maxreaders) {
535 r->mr_txnid = txn->mt_txnid;
536 txn->mt_u.reader = r;
543 if ((rc = mdbenv_read_meta(env)) != MDB_SUCCESS) {
548 txn->mt_next_pgno = env->me_meta.mm_last_pg+1;
549 txn->mt_first_pgno = txn->mt_next_pgno;
550 txn->mt_root = env->me_meta.mm_root;
551 DPRINTF("begin transaction %lu on mdbenv %p, root page %lu",
552 txn->mt_txnid, (void *) env, txn->mt_root);
559 mdb_txn_abort(MDB_txn *txn)
568 DPRINTF("abort transaction %lu on mdbenv %p, root page %lu",
569 txn->mt_txnid, (void *) env, txn->mt_root);
571 if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) {
572 txn->mt_u.reader->mr_txnid = 0;
574 /* Discard all dirty pages. Return any re-used pages
577 MDB_IDL_ZERO(txn->mt_free_pgs);
578 while (!SIMPLEQ_EMPTY(txn->mt_u.dirty_queue)) {
579 dp = SIMPLEQ_FIRST(txn->mt_u.dirty_queue);
580 SIMPLEQ_REMOVE_HEAD(txn->mt_u.dirty_queue, h.md_next);
581 if (dp->p.mp_pgno <= env->me_meta.mm_last_pg)
582 mdb_idl_insert(txn->mt_free_pgs, dp->p.mp_pgno);
585 /* put back to head of free list */
586 if (!MDB_IDL_IS_ZERO(txn->mt_free_pgs)) {
589 mop = malloc(sizeof(MDB_oldpages) + MDB_IDL_SIZEOF(txn->mt_free_pgs) - sizeof(pgno_t));
590 mop->mo_next = env->me_pghead;
591 mop->mo_txnid = txn->mt_oldest - 1;
592 if (!env->me_pghead) {
593 env->me_pgtail = mop;
595 env->me_pghead = mop;
596 bcopy(txn->mt_free_pgs, mop->mo_pages, MDB_IDL_SIZEOF(txn->mt_free_pgs));
599 free(txn->mt_free_pgs);
600 free(txn->mt_u.dirty_queue);
602 pthread_mutex_unlock(&env->me_txns->mt_wmutex);
609 mdb_txn_commit(MDB_txn *txn)
617 struct iovec iov[MDB_COMMIT_PAGES];
620 assert(txn->mt_env != NULL);
624 if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) {
625 DPRINTF("attempt to commit read-only transaction");
630 if (txn != env->me_txn) {
631 DPRINTF("attempt to commit unknown transaction");
636 if (F_ISSET(txn->mt_flags, MDB_TXN_ERROR)) {
637 DPRINTF("error flag is set, can't commit");
642 if (SIMPLEQ_EMPTY(txn->mt_u.dirty_queue))
645 DPRINTF("committing transaction %lu on mdbenv %p, root page %lu",
646 txn->mt_txnid, (void *) env, txn->mt_root);
648 /* Commit up to MDB_COMMIT_PAGES dirty pages to disk until done.
655 SIMPLEQ_FOREACH(dp, txn->mt_u.dirty_queue, h.md_next) {
656 if (dp->p.mp_pgno != next) {
657 lseek(env->me_fd, dp->p.mp_pgno * env->me_meta.mm_stat.ms_psize, SEEK_SET);
658 next = dp->p.mp_pgno;
662 DPRINTF("committing page %lu", dp->p.mp_pgno);
663 iov[n].iov_len = env->me_meta.mm_stat.ms_psize * dp->h.md_num;
664 iov[n].iov_base = &dp->p;
665 size += iov[n].iov_len;
666 next = dp->p.mp_pgno + dp->h.md_num;
667 /* clear dirty flag */
668 dp->p.mp_flags &= ~P_DIRTY;
669 if (++n >= MDB_COMMIT_PAGES) {
678 DPRINTF("committing %u dirty pages", n);
679 rc = writev(env->me_fd, iov, n);
683 DPRINTF("short write, filesystem full?");
685 DPRINTF("writev: %s", strerror(errno));
690 /* Drop the dirty pages.
692 while (!SIMPLEQ_EMPTY(txn->mt_u.dirty_queue)) {
693 dp = SIMPLEQ_FIRST(txn->mt_u.dirty_queue);
694 SIMPLEQ_REMOVE_HEAD(txn->mt_u.dirty_queue, h.md_next);
701 if ((n = mdbenv_sync(env)) != 0 ||
702 (n = mdbenv_write_meta(txn)) != MDB_SUCCESS ||
703 (n = mdbenv_sync(env)) != 0) {
709 /* add to tail of free list */
710 if (!MDB_IDL_IS_ZERO(txn->mt_free_pgs)) {
713 mop = malloc(sizeof(MDB_oldpages) + MDB_IDL_SIZEOF(txn->mt_free_pgs) - sizeof(pgno_t));
715 if (env->me_pghead) {
716 env->me_pgtail->mo_next = mop;
718 env->me_pghead = mop;
720 env->me_pgtail = mop;
721 bcopy(txn->mt_free_pgs, mop->mo_pages, MDB_IDL_SIZEOF(txn->mt_free_pgs));
722 mop->mo_txnid = txn->mt_txnid;
725 pthread_mutex_unlock(&env->me_txns->mt_wmutex);
726 free(txn->mt_free_pgs);
727 free(txn->mt_u.dirty_queue);
738 mdbenv_read_header(MDB_env *env)
747 /* We don't know the page size yet, so use a minimum value.
750 if ((rc = pread(env->me_fd, page, PAGESIZE, 0)) == 0) {
752 } else if (rc != PAGESIZE) {
755 DPRINTF("read: %s", strerror(errno));
759 p = (MDB_page *)page;
761 if (!F_ISSET(p->mp_flags, P_META)) {
762 DPRINTF("page %lu not a meta page", p->mp_pgno);
767 if (m->mm_magic != MDB_MAGIC) {
768 DPRINTF("meta has invalid magic");
772 if (m->mm_version != MDB_VERSION) {
773 DPRINTF("database is version %u, expected version %u",
774 m->mm_version, MDB_VERSION);
778 bcopy(m, &env->me_meta, sizeof(*m));
783 mdbenv_init_meta(MDB_env *env)
790 DPRINTF("writing new meta page");
791 psize = sysconf(_SC_PAGE_SIZE);
793 env->me_meta.mm_magic = MDB_MAGIC;
794 env->me_meta.mm_version = MDB_VERSION;
795 env->me_meta.mm_stat.ms_psize = psize;
796 env->me_meta.mm_stat.ms_flags = env->me_flags & 0xffff;
797 env->me_meta.mm_root = P_INVALID;
798 env->me_meta.mm_last_pg = 1;
800 p = calloc(2, psize);
802 p->mp_flags = P_META;
805 bcopy(&env->me_meta, meta, sizeof(*meta));
807 q = (MDB_page *)((char *)p + psize);
810 q->mp_flags = P_META;
813 bcopy(&env->me_meta, meta, sizeof(*meta));
815 rc = write(env->me_fd, p, psize * 2);
817 return (rc == psize * 2) ? MDB_SUCCESS : errno;
821 mdbenv_write_meta(MDB_txn *txn)
829 assert(txn->mt_env != NULL);
831 DPRINTF("writing meta page for root page %lu", txn->mt_root);
835 bcopy(&env->me_meta, &meta, sizeof(meta));
836 meta.mm_root = txn->mt_root;
837 meta.mm_last_pg = txn->mt_next_pgno - 1;
838 meta.mm_txnid = txn->mt_txnid;
841 if (!env->me_metatoggle)
842 off = env->me_meta.mm_stat.ms_psize;
845 lseek(env->me_fd, off, SEEK_SET);
846 rc = write(env->me_fd, &meta, sizeof(meta));
847 if (rc != sizeof(meta)) {
848 DPRINTF("write failed, disk error?");
855 /* Returns true if page p is a valid meta page, false otherwise.
858 mdb_check_meta_page(MDB_page *p)
860 if (!F_ISSET(p->mp_flags, P_META)) {
861 DPRINTF("page %lu not a meta page", p->mp_pgno);
869 mdbenv_read_meta(MDB_env *env)
877 if ((mp0 = mdbenv_get_page(env, 0)) == NULL ||
878 (mp1 = mdbenv_get_page(env, 1)) == NULL)
881 rc = mdb_check_meta_page(mp0);
884 rc = mdb_check_meta_page(mp1);
887 meta[0] = METADATA(mp0);
888 meta[1] = METADATA(mp1);
890 if (meta[0]->mm_txnid < meta[1]->mm_txnid)
893 if (meta[toggle]->mm_txnid > env->me_meta.mm_txnid) {
894 bcopy(meta[toggle], &env->me_meta, sizeof(env->me_meta));
895 env->me_metatoggle = toggle;
898 DPRINTF("Using meta page %d", toggle);
904 mdbenv_create(MDB_env **env)
908 e = calloc(1, sizeof(*e));
909 if (!e) return ENOMEM;
911 e->me_meta.mm_mapsize = DEFAULT_MAPSIZE;
912 e->me_maxreaders = DEFAULT_READERS;
919 mdbenv_set_mapsize(MDB_env *env, size_t size)
923 env->me_mapsize = env->me_meta.mm_mapsize = size;
928 mdbenv_set_maxreaders(MDB_env *env, int readers)
930 env->me_maxreaders = readers;
935 mdbenv_get_maxreaders(MDB_env *env, int *readers)
937 if (!env || !readers)
939 *readers = env->me_maxreaders;
944 mdbenv_open2(MDB_env *env, unsigned int flags)
948 env->me_flags = flags;
950 if ((i = mdbenv_read_header(env)) != 0) {
953 DPRINTF("new mdbenv");
957 if (!env->me_mapsize)
958 env->me_mapsize = env->me_meta.mm_mapsize;
961 if (env->me_meta.mm_address && (flags & MDB_FIXEDMAP))
963 env->me_map = mmap(env->me_meta.mm_address, env->me_mapsize, PROT_READ, i,
965 if (env->me_map == MAP_FAILED)
969 env->me_meta.mm_mapsize = env->me_mapsize;
970 if (flags & MDB_FIXEDMAP)
971 env->me_meta.mm_address = env->me_map;
972 i = mdbenv_init_meta(env);
973 if (i != MDB_SUCCESS) {
974 munmap(env->me_map, env->me_mapsize);
979 if ((i = mdbenv_read_meta(env)) != 0)
982 DPRINTF("opened database version %u, pagesize %u",
983 env->me_meta.mm_version, env->me_meta.mm_stat.ms_psize);
984 DPRINTF("depth: %u", env->me_meta.mm_stat.ms_depth);
985 DPRINTF("entries: %lu", env->me_meta.mm_stat.ms_entries);
986 DPRINTF("branch pages: %lu", env->me_meta.mm_stat.ms_branch_pages);
987 DPRINTF("leaf pages: %lu", env->me_meta.mm_stat.ms_leaf_pages);
988 DPRINTF("overflow pages: %lu", env->me_meta.mm_stat.ms_overflow_pages);
989 DPRINTF("root: %lu", env->me_meta.mm_root);
995 mdbenv_reader_dest(void *ptr)
997 MDB_reader *reader = ptr;
999 reader->mr_txnid = 0;
1005 mdbenv_open(MDB_env *env, const char *path, unsigned int flags, mode_t mode)
1007 int oflags, rc, shmid;
1011 if (F_ISSET(flags, MDB_RDONLY))
1014 oflags = O_RDWR | O_CREAT;
1016 if ((env->me_fd = open(path, oflags, mode)) == -1)
1019 env->me_shmkey = ftok(path, 'm');
1020 size = (env->me_maxreaders-1) * sizeof(MDB_reader) + sizeof(MDB_txninfo);
1021 shmid = shmget(env->me_shmkey, size, IPC_CREAT|IPC_EXCL|mode);
1023 if (errno == EEXIST) {
1024 shmid = shmget(env->me_shmkey, size, IPC_CREAT|mode);
1027 env->me_txns = shmat(shmid, NULL, 0);
1028 if (env->me_txns->mt_magic != MDB_MAGIC ||
1029 env->me_txns->mt_version != MDB_VERSION) {
1030 DPRINTF("invalid lock region %d", shmid);
1031 shmdt(env->me_txns);
1032 env->me_txns = NULL;
1039 pthread_mutexattr_t mattr;
1041 env->me_txns = shmat(shmid, NULL, 0);
1042 pthread_mutexattr_init(&mattr);
1043 pthread_mutexattr_setpshared(&mattr, PTHREAD_PROCESS_SHARED);
1044 pthread_mutex_init(&env->me_txns->mt_mutex, &mattr);
1045 pthread_mutex_init(&env->me_txns->mt_wmutex, &mattr);
1046 env->me_txns->mt_version = MDB_VERSION;
1047 env->me_txns->mt_magic = MDB_MAGIC;
1050 if ((rc = mdbenv_open2(env, flags)) != MDB_SUCCESS) {
1054 env->me_path = strdup(path);
1055 DPRINTF("opened dbenv %p", (void *) env);
1058 pthread_key_create(&env->me_txkey, mdbenv_reader_dest);
1064 mdbenv_close(MDB_env *env)
1072 munmap(env->me_map, env->me_mapsize);
1076 shmdt(env->me_txns);
1081 /* Search for key within a leaf page, using binary search.
1082 * Returns the smallest entry larger or equal to the key.
1083 * If exactp is non-null, stores whether the found entry was an exact match
1084 * in *exactp (1 or 0).
1085 * If kip is non-null, stores the index of the found entry in *kip.
1086 * If no entry larger or equal to the key is found, returns NULL.
1089 mdb_search_node(MDB_db *bt, MDB_page *mp, MDB_val *key,
1090 int *exactp, unsigned int *kip)
1098 DPRINTF("searching %u keys in %s page %lu",
1100 IS_LEAF(mp) ? "leaf" : "branch",
1103 assert(NUMKEYS(mp) > 0);
1105 bzero(&nodekey, sizeof(nodekey));
1107 low = IS_LEAF(mp) ? 0 : 1;
1108 high = NUMKEYS(mp) - 1;
1109 while (low <= high) {
1110 i = (low + high) >> 1;
1111 node = NODEPTR(mp, i);
1113 nodekey.mv_size = node->mn_ksize;
1114 nodekey.mv_data = NODEKEY(node);
1117 rc = bt->md_cmp(key, &nodekey);
1119 rc = _mdb_cmp(bt, key, &nodekey);
1122 DPRINTF("found leaf index %u [%.*s], rc = %i",
1123 i, (int)nodekey.mv_size, (char *)nodekey.mv_data, rc);
1125 DPRINTF("found branch index %u [%.*s -> %lu], rc = %i",
1126 i, (int)node->mn_ksize, (char *)NODEKEY(node),
1137 if (rc > 0) { /* Found entry is less than the key. */
1138 i++; /* Skip to get the smallest entry larger than key. */
1139 if (i >= NUMKEYS(mp))
1140 /* There is no entry larger or equal to the key. */
1144 *exactp = (rc == 0);
1145 if (kip) /* Store the key index if requested. */
1148 return NODEPTR(mp, i);
1152 cursor_pop_page(MDB_cursor *cursor)
1156 top = CURSOR_TOP(cursor);
1159 DPRINTF("popped page %lu off cursor %p", top->mp_page->mp_pgno, (void *) cursor);
1165 cursor_push_page(MDB_cursor *cursor, MDB_page *mp)
1169 DPRINTF("pushing page %lu on cursor %p", mp->mp_pgno, (void *) cursor);
1171 if ((ppage = calloc(1, sizeof(*ppage))) == NULL)
1173 ppage->mp_page = mp;
1174 CURSOR_PUSH(cursor, ppage);
1179 mdbenv_get_page(MDB_env *env, pgno_t pgno)
1182 MDB_txn *txn = env->me_txn;
1185 if (txn && !SIMPLEQ_EMPTY(txn->mt_u.dirty_queue)) {
1187 SIMPLEQ_FOREACH(dp, txn->mt_u.dirty_queue, h.md_next) {
1188 if (dp->p.mp_pgno == pgno) {
1196 p = (MDB_page *)(env->me_map + env->me_meta.mm_stat.ms_psize * pgno);
1202 mdb_search_page_root(MDB_db *bt, MDB_val *key,
1203 MDB_cursor *cursor, int modify, MDB_pageparent *mpp)
1205 MDB_page *mp = mpp->mp_page;
1208 if (cursor && cursor_push_page(cursor, mp) == NULL)
1211 while (IS_BRANCH(mp)) {
1215 DPRINTF("branch page %lu has %u keys", mp->mp_pgno, NUMKEYS(mp));
1216 assert(NUMKEYS(mp) > 1);
1217 DPRINTF("found index 0 to page %lu", NODEPGNO(NODEPTR(mp, 0)));
1219 if (key == NULL) /* Initialize cursor to first page. */
1223 node = mdb_search_node(bt, mp, key, &exact, &i);
1225 i = NUMKEYS(mp) - 1;
1233 DPRINTF("following index %u for key %.*s",
1234 i, (int)key->mv_size, (char *)key->mv_data);
1235 assert(i < NUMKEYS(mp));
1236 node = NODEPTR(mp, i);
1239 CURSOR_TOP(cursor)->mp_ki = i;
1241 mpp->mp_parent = mp;
1242 if ((mp = mdbenv_get_page(bt->md_env, NODEPGNO(node))) == NULL)
1247 if (cursor && cursor_push_page(cursor, mp) == NULL)
1251 MDB_dhead *dh = ((MDB_dhead *)mp)-1;
1252 if ((rc = mdb_touch(bt->md_env->me_txn, mpp)) != 0)
1254 dh = ((MDB_dhead *)mpp->mp_page)-1;
1255 dh->md_parent = mpp->mp_parent;
1256 dh->md_pi = mpp->mp_pi;
1263 DPRINTF("internal error, index points to a %02X page!?",
1268 DPRINTF("found leaf page %lu for key %.*s", mp->mp_pgno,
1269 key ? (int)key->mv_size : 0, key ? (char *)key->mv_data : NULL);
1274 /* Search for the page a given key should be in.
1275 * Stores a pointer to the found page in *mpp.
1276 * If key is NULL, search for the lowest page (used by mdb_cursor_first).
1277 * If cursor is non-null, pushes parent pages on the cursor stack.
1278 * If modify is true, visited pages are updated with new page numbers.
1281 mdb_search_page(MDB_db *db, MDB_txn *txn, MDB_val *key,
1282 MDB_cursor *cursor, int modify, MDB_pageparent *mpp)
1287 /* Can't modify pages outside a transaction. */
1288 if (txn == NULL && modify) {
1292 /* Choose which root page to start with. If a transaction is given
1293 * use the root page from the transaction, otherwise read the last
1294 * committed root page.
1297 if ((rc = mdbenv_read_meta(db->md_env)) != MDB_SUCCESS)
1299 root = db->md_env->me_meta.mm_root;
1300 } else if (F_ISSET(txn->mt_flags, MDB_TXN_ERROR)) {
1301 DPRINTF("transaction has failed, must abort");
1304 root = txn->mt_root;
1306 if (root == P_INVALID) { /* Tree is empty. */
1307 DPRINTF("tree is empty");
1311 if ((mpp->mp_page = mdbenv_get_page(db->md_env, root)) == NULL)
1314 DPRINTF("root page has flags 0x%X", mpp->mp_page->mp_flags);
1316 if (modify && !F_ISSET(mpp->mp_page->mp_flags, P_DIRTY)) {
1317 mpp->mp_parent = NULL;
1319 if ((rc = mdb_touch(txn, mpp)))
1321 txn->mt_root = mpp->mp_page->mp_pgno;
1324 return mdb_search_page_root(db, key, cursor, modify, mpp);
1328 mdb_read_data(MDB_db *db, MDB_page *mp, MDB_node *leaf,
1331 MDB_page *omp; /* overflow mpage */
1335 bzero(data, sizeof(*data));
1336 max = db->md_env->me_meta.mm_stat.ms_psize - PAGEHDRSZ;
1338 if (!F_ISSET(leaf->mn_flags, F_BIGDATA)) {
1339 data->mv_size = leaf->mn_dsize;
1340 data->mv_data = NODEDATA(leaf);
1344 /* Read overflow data.
1346 data->mv_size = leaf->mn_dsize;
1347 bcopy(NODEDATA(leaf), &pgno, sizeof(pgno));
1348 if ((omp = mdbenv_get_page(db->md_env, pgno)) == NULL) {
1349 DPRINTF("read overflow page %lu failed", pgno);
1352 data->mv_data = omp;
1358 mdb_get(MDB_db *db, MDB_txn *txn,
1359 MDB_val *key, MDB_val *data)
1367 DPRINTF("===> get key [%.*s]", (int)key->mv_size, (char *)key->mv_data);
1372 if (txn != NULL && db->md_env != txn->mt_env) {
1376 if (key->mv_size == 0 || key->mv_size > MAXKEYSIZE) {
1380 if ((rc = mdb_search_page(db, txn, key, NULL, 0, &mpp)) != MDB_SUCCESS)
1383 leaf = mdb_search_node(db, mpp.mp_page, key, &exact, NULL);
1385 rc = mdb_read_data(db, mpp.mp_page, leaf, data);
1394 mdb_sibling(MDB_cursor *cursor, int move_right)
1398 MDB_ppage *parent, *top;
1401 top = CURSOR_TOP(cursor);
1402 if ((parent = SLIST_NEXT(top, mp_entry)) == NULL) {
1403 return ENOENT; /* root has no siblings */
1406 DPRINTF("parent page is page %lu, index %u",
1407 parent->mp_page->mp_pgno, parent->mp_ki);
1409 cursor_pop_page(cursor);
1410 if (move_right ? (parent->mp_ki + 1 >= NUMKEYS(parent->mp_page))
1411 : (parent->mp_ki == 0)) {
1412 DPRINTF("no more keys left, moving to %s sibling",
1413 move_right ? "right" : "left");
1414 if ((rc = mdb_sibling(cursor, move_right)) != MDB_SUCCESS)
1416 parent = CURSOR_TOP(cursor);
1422 DPRINTF("just moving to %s index key %u",
1423 move_right ? "right" : "left", parent->mp_ki);
1425 assert(IS_BRANCH(parent->mp_page));
1427 indx = NODEPTR(parent->mp_page, parent->mp_ki);
1428 if ((mp = mdbenv_get_page(cursor->mc_db->md_env, indx->mn_pgno)) == NULL)
1431 mp->parent = parent->mp_page;
1432 mp->parent_index = parent->mp_ki;
1435 cursor_push_page(cursor, mp);
1441 mdb_set_key(MDB_db *bt, MDB_page *mp, MDB_node *node,
1447 key->mv_size = node->mn_ksize;
1448 key->mv_data = NODEKEY(node);
1454 mdb_cursor_next(MDB_cursor *cursor, MDB_val *key, MDB_val *data)
1460 if (cursor->mc_eof) {
1464 assert(cursor->mc_initialized);
1466 top = CURSOR_TOP(cursor);
1469 DPRINTF("cursor_next: top page is %lu in cursor %p", mp->mp_pgno, (void *) cursor);
1471 if (top->mp_ki + 1 >= NUMKEYS(mp)) {
1472 DPRINTF("=====> move to next sibling page");
1473 if (mdb_sibling(cursor, 1) != MDB_SUCCESS) {
1477 top = CURSOR_TOP(cursor);
1479 DPRINTF("next page is %lu, key index %u", mp->mp_pgno, top->mp_ki);
1483 DPRINTF("==> cursor points to page %lu with %u keys, key index %u",
1484 mp->mp_pgno, NUMKEYS(mp), top->mp_ki);
1486 assert(IS_LEAF(mp));
1487 leaf = NODEPTR(mp, top->mp_ki);
1489 if (data && mdb_read_data(cursor->mc_db, mp, leaf, data) != MDB_SUCCESS)
1492 return mdb_set_key(cursor->mc_db, mp, leaf, key);
1496 mdb_cursor_set(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
1506 assert(key->mv_size > 0);
1508 rc = mdb_search_page(cursor->mc_db, cursor->mc_txn, key, cursor, 0, &mpp);
1509 if (rc != MDB_SUCCESS)
1511 assert(IS_LEAF(mpp.mp_page));
1513 top = CURSOR_TOP(cursor);
1514 leaf = mdb_search_node(cursor->mc_db, mpp.mp_page, key, exactp, &top->mp_ki);
1515 if (exactp != NULL && !*exactp) {
1516 /* MDB_CURSOR_EXACT specified and not an exact match. */
1521 DPRINTF("===> inexact leaf not found, goto sibling");
1522 if ((rc = mdb_sibling(cursor, 1)) != MDB_SUCCESS)
1523 return rc; /* no entries matched */
1524 top = CURSOR_TOP(cursor);
1526 mpp.mp_page = top->mp_page;
1527 assert(IS_LEAF(mpp.mp_page));
1528 leaf = NODEPTR(mpp.mp_page, 0);
1531 cursor->mc_initialized = 1;
1534 if (data && (rc = mdb_read_data(cursor->mc_db, mpp.mp_page, leaf, data)) != MDB_SUCCESS)
1537 rc = mdb_set_key(cursor->mc_db, mpp.mp_page, leaf, key);
1538 if (rc == MDB_SUCCESS) {
1539 DPRINTF("==> cursor placed on key %.*s",
1540 (int)key->mv_size, (char *)key->mv_data);
1548 mdb_cursor_first(MDB_cursor *cursor, MDB_val *key, MDB_val *data)
1554 rc = mdb_search_page(cursor->mc_db, cursor->mc_txn, NULL, cursor, 0, &mpp);
1555 if (rc != MDB_SUCCESS)
1557 assert(IS_LEAF(mpp.mp_page));
1559 leaf = NODEPTR(mpp.mp_page, 0);
1560 cursor->mc_initialized = 1;
1563 if (data && (rc = mdb_read_data(cursor->mc_db, mpp.mp_page, leaf, data)) != MDB_SUCCESS)
1566 return mdb_set_key(cursor->mc_db, mpp.mp_page, leaf, key);
1570 mdb_cursor_get(MDB_cursor *cursor, MDB_val *key, MDB_val *data,
1580 case MDB_CURSOR_EXACT:
1581 while (CURSOR_TOP(cursor) != NULL)
1582 cursor_pop_page(cursor);
1583 if (key == NULL || key->mv_size == 0 || key->mv_size > MAXKEYSIZE) {
1585 } else if (op == MDB_CURSOR_EXACT)
1586 rc = mdb_cursor_set(cursor, key, data, &exact);
1588 rc = mdb_cursor_set(cursor, key, data, NULL);
1591 if (!cursor->mc_initialized)
1592 rc = mdb_cursor_first(cursor, key, data);
1594 rc = mdb_cursor_next(cursor, key, data);
1597 while (CURSOR_TOP(cursor) != NULL)
1598 cursor_pop_page(cursor);
1599 rc = mdb_cursor_first(cursor, key, data);
1602 DPRINTF("unhandled/unimplemented cursor operation %u", op);
1610 /* Allocate a page and initialize it
1613 mdbenv_new_page(MDB_env *env, uint32_t flags, int num)
1617 assert(env != NULL);
1618 assert(env->me_txn != NULL);
1620 if ((dp = mdb_newpage(env->me_txn, NULL, 0, num)) == NULL)
1622 DPRINTF("allocated new mpage %lu, page size %u",
1623 dp->p.mp_pgno, env->me_meta.mm_stat.ms_psize);
1624 dp->p.mp_flags = flags | P_DIRTY;
1625 dp->p.mp_lower = PAGEHDRSZ;
1626 dp->p.mp_upper = env->me_meta.mm_stat.ms_psize;
1628 if (IS_BRANCH(&dp->p))
1629 env->me_meta.mm_stat.ms_branch_pages++;
1630 else if (IS_LEAF(&dp->p))
1631 env->me_meta.mm_stat.ms_leaf_pages++;
1632 else if (IS_OVERFLOW(&dp->p)) {
1633 env->me_meta.mm_stat.ms_overflow_pages += num;
1634 dp->p.mp_pages = num;
1641 mdb_leaf_size(MDB_db *db, MDB_val *key, MDB_val *data)
1645 sz = LEAFSIZE(key, data);
1646 if (data->mv_size >= db->md_env->me_meta.mm_stat.ms_psize / MDB_MINKEYS) {
1647 /* put on overflow page */
1648 sz -= data->mv_size - sizeof(pgno_t);
1651 return sz + sizeof(indx_t);
1655 mdb_branch_size(MDB_db *db, MDB_val *key)
1660 if (sz >= db->md_env->me_meta.mm_stat.ms_psize / MDB_MINKEYS) {
1661 /* put on overflow page */
1662 /* not implemented */
1663 /* sz -= key->size - sizeof(pgno_t); */
1666 return sz + sizeof(indx_t);
1670 mdb_add_node(MDB_db *db, MDB_page *mp, indx_t indx,
1671 MDB_val *key, MDB_val *data, pgno_t pgno, uint8_t flags)
1674 size_t node_size = NODESIZE;
1677 MDB_dpage *ofp = NULL; /* overflow page */
1679 assert(mp->mp_upper >= mp->mp_lower);
1681 DPRINTF("add node [%.*s] to %s page %lu at index %i, key size %zu",
1682 key ? (int)key->mv_size : 0, key ? (char *)key->mv_data : NULL,
1683 IS_LEAF(mp) ? "leaf" : "branch",
1684 mp->mp_pgno, indx, key ? key->mv_size : 0);
1687 node_size += key->mv_size;
1691 if (F_ISSET(flags, F_BIGDATA)) {
1692 /* Data already on overflow page. */
1693 node_size += sizeof(pgno_t);
1694 } else if (data->mv_size >= db->md_env->me_meta.mm_stat.ms_psize / MDB_MINKEYS) {
1695 int ovpages = OVPAGES(data->mv_size, db->md_env->me_meta.mm_stat.ms_psize);
1696 /* Put data on overflow page. */
1697 DPRINTF("data size is %zu, put on overflow page",
1699 node_size += sizeof(pgno_t);
1700 if ((ofp = mdbenv_new_page(db->md_env, P_OVERFLOW, ovpages)) == NULL)
1702 DPRINTF("allocated overflow page %lu", ofp->p.mp_pgno);
1705 node_size += data->mv_size;
1709 if (node_size + sizeof(indx_t) > SIZELEFT(mp)) {
1710 DPRINTF("not enough room in page %lu, got %u ptrs",
1711 mp->mp_pgno, NUMKEYS(mp));
1712 DPRINTF("upper - lower = %u - %u = %u", mp->mp_upper, mp->mp_lower,
1713 mp->mp_upper - mp->mp_lower);
1714 DPRINTF("node size = %zu", node_size);
1718 /* Move higher pointers up one slot. */
1719 for (i = NUMKEYS(mp); i > indx; i--)
1720 mp->mp_ptrs[i] = mp->mp_ptrs[i - 1];
1722 /* Adjust free space offsets. */
1723 ofs = mp->mp_upper - node_size;
1724 assert(ofs >= mp->mp_lower + sizeof(indx_t));
1725 mp->mp_ptrs[indx] = ofs;
1727 mp->mp_lower += sizeof(indx_t);
1729 /* Write the node data. */
1730 node = NODEPTR(mp, indx);
1731 node->mn_ksize = (key == NULL) ? 0 : key->mv_size;
1732 node->mn_flags = flags;
1734 node->mn_dsize = data->mv_size;
1736 node->mn_pgno = pgno;
1739 bcopy(key->mv_data, NODEKEY(node), key->mv_size);
1744 if (F_ISSET(flags, F_BIGDATA))
1745 bcopy(data->mv_data, node->mn_data + key->mv_size,
1748 bcopy(data->mv_data, node->mn_data + key->mv_size,
1751 bcopy(&ofp->p.mp_pgno, node->mn_data + key->mv_size,
1753 bcopy(data->mv_data, METADATA(&ofp->p), data->mv_size);
1761 mdb_del_node(MDB_db *db, MDB_page *mp, indx_t indx)
1764 indx_t i, j, numkeys, ptr;
1768 DPRINTF("delete node %u on %s page %lu", indx,
1769 IS_LEAF(mp) ? "leaf" : "branch", mp->mp_pgno);
1770 assert(indx < NUMKEYS(mp));
1772 node = NODEPTR(mp, indx);
1773 sz = NODESIZE + node->mn_ksize;
1775 if (F_ISSET(node->mn_flags, F_BIGDATA))
1776 sz += sizeof(pgno_t);
1778 sz += NODEDSZ(node);
1781 ptr = mp->mp_ptrs[indx];
1782 numkeys = NUMKEYS(mp);
1783 for (i = j = 0; i < numkeys; i++) {
1785 mp->mp_ptrs[j] = mp->mp_ptrs[i];
1786 if (mp->mp_ptrs[i] < ptr)
1787 mp->mp_ptrs[j] += sz;
1792 base = (char *)mp + mp->mp_upper;
1793 bcopy(base, base + sz, ptr - mp->mp_upper);
1795 mp->mp_lower -= sizeof(indx_t);
1800 mdb_cursor_open(MDB_db *db, MDB_txn *txn, MDB_cursor **ret)
1804 if (db == NULL || ret == NULL)
1807 if (txn != NULL && db->md_env != txn->mt_env) {
1811 if ((cursor = calloc(1, sizeof(*cursor))) != NULL) {
1812 SLIST_INIT(&cursor->mc_stack);
1814 cursor->mc_txn = txn;
1823 mdb_cursor_close(MDB_cursor *cursor)
1825 if (cursor != NULL) {
1826 while (!CURSOR_EMPTY(cursor))
1827 cursor_pop_page(cursor);
1829 /* btree_close(cursor->bt); */
1835 mdb_update_key(MDB_db *db, MDB_page *mp, indx_t indx,
1838 indx_t ptr, i, numkeys;
1844 node = NODEPTR(mp, indx);
1845 ptr = mp->mp_ptrs[indx];
1846 DPRINTF("update key %u (ofs %u) [%.*s] to [%.*s] on page %lu",
1848 (int)node->mn_ksize, (char *)NODEKEY(node),
1849 (int)key->mv_size, (char *)key->mv_data,
1852 delta = key->mv_size - node->mn_ksize;
1854 if (delta > 0 && SIZELEFT(mp) < delta) {
1855 DPRINTF("OUCH! Not enough room, delta = %d", delta);
1859 numkeys = NUMKEYS(mp);
1860 for (i = 0; i < numkeys; i++) {
1861 if (mp->mp_ptrs[i] <= ptr)
1862 mp->mp_ptrs[i] -= delta;
1865 base = (char *)mp + mp->mp_upper;
1866 len = ptr - mp->mp_upper + NODESIZE;
1867 bcopy(base, base - delta, len);
1868 mp->mp_upper -= delta;
1870 node = NODEPTR(mp, indx);
1871 node->mn_ksize = key->mv_size;
1874 bcopy(key->mv_data, NODEKEY(node), key->mv_size);
1879 /* Move a node from src to dst.
1882 mdb_move_node(MDB_db *bt, MDB_pageparent *src, indx_t srcindx,
1883 MDB_pageparent *dst, indx_t dstindx)
1889 srcnode = NODEPTR(src->mp_page, srcindx);
1890 DPRINTF("moving %s node %u [%.*s] on page %lu to node %u on page %lu",
1891 IS_LEAF(src->mp_page) ? "leaf" : "branch",
1893 (int)srcnode->mn_ksize, (char *)NODEKEY(srcnode),
1894 src->mp_page->mp_pgno,
1895 dstindx, dst->mp_page->mp_pgno);
1897 /* Mark src and dst as dirty. */
1898 if ((rc = mdb_touch(bt->md_env->me_txn, src)) ||
1899 (rc = mdb_touch(bt->md_env->me_txn, dst)))
1902 /* Add the node to the destination page.
1904 key.mv_size = srcnode->mn_ksize;
1905 key.mv_data = NODEKEY(srcnode);
1906 data.mv_size = NODEDSZ(srcnode);
1907 data.mv_data = NODEDATA(srcnode);
1908 rc = mdb_add_node(bt, dst->mp_page, dstindx, &key, &data, NODEPGNO(srcnode),
1910 if (rc != MDB_SUCCESS)
1913 /* Delete the node from the source page.
1915 mdb_del_node(bt, src->mp_page, srcindx);
1917 /* Update the parent separators.
1919 if (srcindx == 0 && src->mp_pi != 0) {
1920 DPRINTF("update separator for source page %lu to [%.*s]",
1921 src->mp_page->mp_pgno, (int)key.mv_size, (char *)key.mv_data);
1922 if ((rc = mdb_update_key(bt, src->mp_parent, src->mp_pi,
1923 &key)) != MDB_SUCCESS)
1927 if (srcindx == 0 && IS_BRANCH(src->mp_page)) {
1929 nullkey.mv_size = 0;
1930 assert(mdb_update_key(bt, src->mp_page, 0, &nullkey) == MDB_SUCCESS);
1933 if (dstindx == 0 && dst->mp_pi != 0) {
1934 DPRINTF("update separator for destination page %lu to [%.*s]",
1935 dst->mp_page->mp_pgno, (int)key.mv_size, (char *)key.mv_data);
1936 if ((rc = mdb_update_key(bt, dst->mp_parent, dst->mp_pi,
1937 &key)) != MDB_SUCCESS)
1941 if (dstindx == 0 && IS_BRANCH(dst->mp_page)) {
1943 nullkey.mv_size = 0;
1944 assert(mdb_update_key(bt, dst->mp_page, 0, &nullkey) == MDB_SUCCESS);
1951 mdb_merge(MDB_db *bt, MDB_pageparent *src, MDB_pageparent *dst)
1960 DPRINTF("merging page %lu and %lu", src->mp_page->mp_pgno, dst->mp_page->mp_pgno);
1962 assert(src->mp_parent); /* can't merge root page */
1963 assert(dst->mp_parent);
1964 assert(bt->md_env->me_txn != NULL);
1966 /* Mark src and dst as dirty. */
1967 if ((rc = mdb_touch(bt->md_env->me_txn, src)) ||
1968 (rc = mdb_touch(bt->md_env->me_txn, dst)))
1971 /* Move all nodes from src to dst.
1973 for (i = 0; i < NUMKEYS(src->mp_page); i++) {
1974 srcnode = NODEPTR(src->mp_page, i);
1976 key.mv_size = srcnode->mn_ksize;
1977 key.mv_data = NODEKEY(srcnode);
1978 data.mv_size = NODEDSZ(srcnode);
1979 data.mv_data = NODEDATA(srcnode);
1980 rc = mdb_add_node(bt, dst->mp_page, NUMKEYS(dst->mp_page), &key,
1981 &data, NODEPGNO(srcnode), srcnode->mn_flags);
1982 if (rc != MDB_SUCCESS)
1986 DPRINTF("dst page %lu now has %u keys (%.1f%% filled)",
1987 dst->mp_page->mp_pgno, NUMKEYS(dst->mp_page), (float)PAGEFILL(bt->md_env, dst->mp_page) / 10);
1989 /* Unlink the src page from parent.
1991 mdb_del_node(bt, src->mp_parent, src->mp_pi);
1992 if (src->mp_pi == 0) {
1994 if ((rc = mdb_update_key(bt, src->mp_parent, 0, &key)) != MDB_SUCCESS)
1998 if (IS_LEAF(src->mp_page))
1999 bt->md_env->me_meta.mm_stat.ms_leaf_pages--;
2001 bt->md_env->me_meta.mm_stat.ms_branch_pages--;
2003 mpp.mp_page = src->mp_parent;
2004 dh = (MDB_dhead *)src->mp_parent;
2006 mpp.mp_parent = dh->md_parent;
2007 mpp.mp_pi = dh->md_pi;
2009 return mdb_rebalance(bt, &mpp);
2012 #define FILL_THRESHOLD 250
2015 mdb_rebalance(MDB_db *db, MDB_pageparent *mpp)
2020 indx_t si = 0, di = 0;
2023 assert(db->md_env->me_txn != NULL);
2024 assert(mpp != NULL);
2026 DPRINTF("rebalancing %s page %lu (has %u keys, %.1f%% full)",
2027 IS_LEAF(mpp->mp_page) ? "leaf" : "branch",
2028 mpp->mp_page->mp_pgno, NUMKEYS(mpp->mp_page), (float)PAGEFILL(db->md_env, mpp->mp_page) / 10);
2030 if (PAGEFILL(db->md_env, mpp->mp_page) >= FILL_THRESHOLD) {
2031 DPRINTF("no need to rebalance page %lu, above fill threshold",
2032 mpp->mp_page->mp_pgno);
2036 if (mpp->mp_parent == NULL) {
2037 if (NUMKEYS(mpp->mp_page) == 0) {
2038 DPRINTF("tree is completely empty");
2039 db->md_env->me_txn->mt_root = P_INVALID;
2040 db->md_env->me_meta.mm_stat.ms_depth--;
2041 db->md_env->me_meta.mm_stat.ms_leaf_pages--;
2042 } else if (IS_BRANCH(mpp->mp_page) && NUMKEYS(mpp->mp_page) == 1) {
2043 DPRINTF("collapsing root page!");
2044 db->md_env->me_txn->mt_root = NODEPGNO(NODEPTR(mpp->mp_page, 0));
2045 if ((root = mdbenv_get_page(db->md_env, db->md_env->me_txn->mt_root)) == NULL)
2047 db->md_env->me_meta.mm_stat.ms_depth--;
2048 db->md_env->me_meta.mm_stat.ms_branch_pages--;
2050 DPRINTF("root page doesn't need rebalancing");
2054 /* The parent (branch page) must have at least 2 pointers,
2055 * otherwise the tree is invalid.
2057 assert(NUMKEYS(mpp->mp_parent) > 1);
2059 /* Leaf page fill factor is below the threshold.
2060 * Try to move keys from left or right neighbor, or
2061 * merge with a neighbor page.
2066 if (mpp->mp_pi == 0) {
2067 /* We're the leftmost leaf in our parent.
2069 DPRINTF("reading right neighbor");
2070 node = NODEPTR(mpp->mp_parent, mpp->mp_pi + 1);
2071 if ((npp.mp_page = mdbenv_get_page(db->md_env, NODEPGNO(node))) == NULL)
2073 npp.mp_pi = mpp->mp_pi + 1;
2075 di = NUMKEYS(mpp->mp_page);
2077 /* There is at least one neighbor to the left.
2079 DPRINTF("reading left neighbor");
2080 node = NODEPTR(mpp->mp_parent, mpp->mp_pi - 1);
2081 if ((npp.mp_page = mdbenv_get_page(db->md_env, NODEPGNO(node))) == NULL)
2083 npp.mp_pi = mpp->mp_pi - 1;
2084 si = NUMKEYS(npp.mp_page) - 1;
2087 npp.mp_parent = mpp->mp_parent;
2089 DPRINTF("found neighbor page %lu (%u keys, %.1f%% full)",
2090 npp.mp_page->mp_pgno, NUMKEYS(npp.mp_page), (float)PAGEFILL(db->md_env, npp.mp_page) / 10);
2092 /* If the neighbor page is above threshold and has at least two
2093 * keys, move one key from it.
2095 * Otherwise we should try to merge them.
2097 if (PAGEFILL(db->md_env, npp.mp_page) >= FILL_THRESHOLD && NUMKEYS(npp.mp_page) >= 2)
2098 return mdb_move_node(db, &npp, si, mpp, di);
2099 else { /* FIXME: if (has_enough_room()) */
2100 if (mpp->mp_pi == 0)
2101 return mdb_merge(db, &npp, mpp);
2103 return mdb_merge(db, mpp, &npp);
2108 mdb_del(MDB_db *bt, MDB_txn *txn,
2109 MDB_val *key, MDB_val *data)
2116 DPRINTF("========> delete key %.*s", (int)key->mv_size, (char *)key->mv_data);
2118 assert(key != NULL);
2120 if (bt == NULL || txn == NULL)
2123 if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) {
2127 if (bt->md_env->me_txn != txn) {
2131 if (key->mv_size == 0 || key->mv_size > MAXKEYSIZE) {
2135 if ((rc = mdb_search_page(bt, txn, key, NULL, 1, &mpp)) != MDB_SUCCESS)
2138 leaf = mdb_search_node(bt, mpp.mp_page, key, &exact, &ki);
2139 if (leaf == NULL || !exact) {
2143 if (data && (rc = mdb_read_data(bt, NULL, leaf, data)) != MDB_SUCCESS)
2146 mdb_del_node(bt, mpp.mp_page, ki);
2147 /* add overflow pages to free list */
2148 if (F_ISSET(leaf->mn_flags, F_BIGDATA)) {
2152 bcopy(NODEDATA(leaf), &pg, sizeof(pg));
2153 ovpages = OVPAGES(NODEDSZ(leaf), bt->md_env->me_meta.mm_stat.ms_psize);
2154 for (i=0; i<ovpages; i++) {
2155 mdb_idl_insert(txn->mt_free_pgs, pg);
2159 bt->md_env->me_meta.mm_stat.ms_entries--;
2160 rc = mdb_rebalance(bt, &mpp);
2161 if (rc != MDB_SUCCESS)
2162 txn->mt_flags |= MDB_TXN_ERROR;
2167 /* Split page <*mpp>, and insert <key,(data|newpgno)> in either left or
2168 * right sibling, at index <*newindxp> (as if unsplit). Updates *mpp and
2169 * *newindxp with the actual values after split, ie if *mpp and *newindxp
2170 * refer to a node in the new right sibling page.
2173 mdb_split(MDB_db *bt, MDB_page **mpp, unsigned int *newindxp,
2174 MDB_val *newkey, MDB_val *newdata, pgno_t newpgno)
2177 int rc = MDB_SUCCESS, ins_new = 0;
2180 unsigned int i, j, split_indx;
2182 MDB_val sepkey, rkey, rdata;
2184 MDB_dpage *mdp, *rdp, *pdp;
2188 assert(bt->md_env != NULL);
2190 dh = ((MDB_dhead *)*mpp) - 1;
2191 mdp = (MDB_dpage *)dh;
2192 newindx = *newindxp;
2194 DPRINTF("-----> splitting %s page %lu and adding [%.*s] at index %i",
2195 IS_LEAF(&mdp->p) ? "leaf" : "branch", mdp->p.mp_pgno,
2196 (int)newkey->mv_size, (char *)newkey->mv_data, *newindxp);
2198 if (mdp->h.md_parent == NULL) {
2199 if ((pdp = mdbenv_new_page(bt->md_env, P_BRANCH, 1)) == NULL)
2202 mdp->h.md_parent = &pdp->p;
2203 bt->md_env->me_txn->mt_root = pdp->p.mp_pgno;
2204 DPRINTF("root split! new root = %lu", pdp->p.mp_pgno);
2205 bt->md_env->me_meta.mm_stat.ms_depth++;
2207 /* Add left (implicit) pointer. */
2208 if (mdb_add_node(bt, &pdp->p, 0, NULL, NULL,
2209 mdp->p.mp_pgno, 0) != MDB_SUCCESS)
2212 DPRINTF("parent branch page is %lu", mdp->h.md_parent->mp_pgno);
2215 /* Create a right sibling. */
2216 if ((rdp = mdbenv_new_page(bt->md_env, mdp->p.mp_flags, 1)) == NULL)
2218 rdp->h.md_parent = mdp->h.md_parent;
2219 rdp->h.md_pi = mdp->h.md_pi + 1;
2220 DPRINTF("new right sibling: page %lu", rdp->p.mp_pgno);
2222 /* Move half of the keys to the right sibling. */
2223 if ((copy = malloc(bt->md_env->me_meta.mm_stat.ms_psize)) == NULL)
2225 bcopy(&mdp->p, copy, bt->md_env->me_meta.mm_stat.ms_psize);
2226 bzero(&mdp->p.mp_ptrs, bt->md_env->me_meta.mm_stat.ms_psize - PAGEHDRSZ);
2227 mdp->p.mp_lower = PAGEHDRSZ;
2228 mdp->p.mp_upper = bt->md_env->me_meta.mm_stat.ms_psize;
2230 split_indx = NUMKEYS(copy) / 2 + 1;
2232 /* First find the separating key between the split pages.
2234 bzero(&sepkey, sizeof(sepkey));
2235 if (newindx == split_indx) {
2236 sepkey.mv_size = newkey->mv_size;
2237 sepkey.mv_data = newkey->mv_data;
2239 node = NODEPTR(copy, split_indx);
2240 sepkey.mv_size = node->mn_ksize;
2241 sepkey.mv_data = NODEKEY(node);
2244 DPRINTF("separator is [%.*s]", (int)sepkey.mv_size, (char *)sepkey.mv_data);
2246 /* Copy separator key to the parent.
2248 if (SIZELEFT(rdp->h.md_parent) < mdb_branch_size(bt, &sepkey)) {
2249 rc = mdb_split(bt, &rdp->h.md_parent, &rdp->h.md_pi,
2250 &sepkey, NULL, rdp->p.mp_pgno);
2252 /* Right page might now have changed parent.
2253 * Check if left page also changed parent.
2255 if (rdp->h.md_parent != mdp->h.md_parent &&
2256 mdp->h.md_pi >= NUMKEYS(mdp->h.md_parent)) {
2257 mdp->h.md_parent = rdp->h.md_parent;
2258 mdp->h.md_pi = rdp->h.md_pi - 1;
2261 rc = mdb_add_node(bt, rdp->h.md_parent, rdp->h.md_pi,
2262 &sepkey, NULL, rdp->p.mp_pgno, 0);
2264 if (rc != MDB_SUCCESS) {
2269 for (i = j = 0; i <= NUMKEYS(copy); j++) {
2270 if (i < split_indx) {
2271 /* Re-insert in left sibling. */
2274 /* Insert in right sibling. */
2275 if (i == split_indx)
2276 /* Reset insert index for right sibling. */
2277 j = (i == newindx && ins_new);
2281 if (i == newindx && !ins_new) {
2282 /* Insert the original entry that caused the split. */
2283 rkey.mv_data = newkey->mv_data;
2284 rkey.mv_size = newkey->mv_size;
2285 if (IS_LEAF(&mdp->p)) {
2286 rdata.mv_data = newdata->mv_data;
2287 rdata.mv_size = newdata->mv_size;
2294 /* Update page and index for the new key. */
2297 } else if (i == NUMKEYS(copy)) {
2300 node = NODEPTR(copy, i);
2301 rkey.mv_data = NODEKEY(node);
2302 rkey.mv_size = node->mn_ksize;
2303 if (IS_LEAF(&mdp->p)) {
2304 rdata.mv_data = NODEDATA(node);
2305 rdata.mv_size = node->mn_dsize;
2307 pgno = node->mn_pgno;
2308 flags = node->mn_flags;
2313 if (!IS_LEAF(&mdp->p) && j == 0) {
2314 /* First branch index doesn't need key data. */
2318 rc = mdb_add_node(bt, &pdp->p, j, &rkey, &rdata, pgno,flags);
2326 mdb_put(MDB_db *bt, MDB_txn *txn,
2327 MDB_val *key, MDB_val *data, unsigned int flags)
2329 int rc = MDB_SUCCESS, exact;
2334 assert(key != NULL);
2335 assert(data != NULL);
2337 if (bt == NULL || txn == NULL)
2340 if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) {
2344 if (bt->md_env->me_txn != txn) {
2348 if (key->mv_size == 0 || key->mv_size > MAXKEYSIZE) {
2352 DPRINTF("==> put key %.*s, size %zu, data size %zu",
2353 (int)key->mv_size, (char *)key->mv_data, key->mv_size, data->mv_size);
2355 rc = mdb_search_page(bt, txn, key, NULL, 1, &mpp);
2356 if (rc == MDB_SUCCESS) {
2357 leaf = mdb_search_node(bt, mpp.mp_page, key, &exact, &ki);
2358 if (leaf && exact) {
2359 if (F_ISSET(flags, MDB_NOOVERWRITE)) {
2360 DPRINTF("duplicate key %.*s",
2361 (int)key->mv_size, (char *)key->mv_data);
2364 mdb_del_node(bt, mpp.mp_page, ki);
2366 if (leaf == NULL) { /* append if not found */
2367 ki = NUMKEYS(mpp.mp_page);
2368 DPRINTF("appending key at index %i", ki);
2370 } else if (rc == ENOENT) {
2372 /* new file, just write a root leaf page */
2373 DPRINTF("allocating new root leaf page");
2374 if ((dp = mdbenv_new_page(bt->md_env, P_LEAF, 1)) == NULL) {
2377 mpp.mp_page = &dp->p;
2378 txn->mt_root = mpp.mp_page->mp_pgno;
2379 bt->md_env->me_meta.mm_stat.ms_depth++;
2385 assert(IS_LEAF(mpp.mp_page));
2386 DPRINTF("there are %u keys, should insert new key at index %i",
2387 NUMKEYS(mpp.mp_page), ki);
2389 if (SIZELEFT(mpp.mp_page) < mdb_leaf_size(bt, key, data)) {
2390 rc = mdb_split(bt, &mpp.mp_page, &ki, key, data, P_INVALID);
2392 /* There is room already in this leaf page. */
2393 rc = mdb_add_node(bt, mpp.mp_page, ki, key, data, 0, 0);
2396 if (rc != MDB_SUCCESS)
2397 txn->mt_flags |= MDB_TXN_ERROR;
2399 bt->md_env->me_meta.mm_stat.ms_entries++;
2406 mdbenv_get_flags(MDB_env *env, unsigned int *arg)
2411 *arg = env->me_flags;
2416 mdbenv_get_path(MDB_env *env, const char **arg)
2421 *arg = env->me_path;
2426 mdbenv_stat(MDB_env *env, MDB_stat **arg)
2428 if (env == NULL || arg == NULL)
2431 *arg = &env->me_meta.mm_stat;
2436 int mdb_open(MDB_env *env, MDB_txn *txn, const char *name, unsigned int flags, MDB_db **db)
2439 *db = (MDB_db *)&env->me_db;
2445 int mdb_stat(MDB_db *db, MDB_stat **arg)
2447 if (db == NULL || arg == NULL)
2450 *arg = &db->md_stat;
2455 void mdb_close(MDB_db *db)