1*0Sstevel@tonic-gate /*- 2*0Sstevel@tonic-gate * See the file LICENSE for redistribution information. 3*0Sstevel@tonic-gate * 4*0Sstevel@tonic-gate * Copyright (c) 1996, 1997, 1998 5*0Sstevel@tonic-gate * Sleepycat Software. All rights reserved. 6*0Sstevel@tonic-gate */ 7*0Sstevel@tonic-gate /* 8*0Sstevel@tonic-gate * Copyright (c) 1990, 1993, 1994, 1995, 1996 9*0Sstevel@tonic-gate * Keith Bostic. All rights reserved. 10*0Sstevel@tonic-gate */ 11*0Sstevel@tonic-gate /* 12*0Sstevel@tonic-gate * Copyright (c) 1990, 1993, 1994, 1995 13*0Sstevel@tonic-gate * The Regents of the University of California. All rights reserved. 14*0Sstevel@tonic-gate * 15*0Sstevel@tonic-gate * This code is derived from software contributed to Berkeley by 16*0Sstevel@tonic-gate * Mike Olson. 17*0Sstevel@tonic-gate * 18*0Sstevel@tonic-gate * Redistribution and use in source and binary forms, with or without 19*0Sstevel@tonic-gate * modification, are permitted provided that the following conditions 20*0Sstevel@tonic-gate * are met: 21*0Sstevel@tonic-gate * 1. Redistributions of source code must retain the above copyright 22*0Sstevel@tonic-gate * notice, this list of conditions and the following disclaimer. 23*0Sstevel@tonic-gate * 2. Redistributions in binary form must reproduce the above copyright 24*0Sstevel@tonic-gate * notice, this list of conditions and the following disclaimer in the 25*0Sstevel@tonic-gate * documentation and/or other materials provided with the distribution. 26*0Sstevel@tonic-gate * 3. All advertising materials mentioning features or use of this software 27*0Sstevel@tonic-gate * must display the following acknowledgement: 28*0Sstevel@tonic-gate * This product includes software developed by the University of 29*0Sstevel@tonic-gate * California, Berkeley and its contributors. 30*0Sstevel@tonic-gate * 4. Neither the name of the University nor the names of its contributors 31*0Sstevel@tonic-gate * may be used to endorse or promote products derived from this software 32*0Sstevel@tonic-gate * without specific prior written permission. 33*0Sstevel@tonic-gate * 34*0Sstevel@tonic-gate * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 35*0Sstevel@tonic-gate * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 36*0Sstevel@tonic-gate * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 37*0Sstevel@tonic-gate * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 38*0Sstevel@tonic-gate * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 39*0Sstevel@tonic-gate * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 40*0Sstevel@tonic-gate * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 41*0Sstevel@tonic-gate * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 42*0Sstevel@tonic-gate * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 43*0Sstevel@tonic-gate * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 44*0Sstevel@tonic-gate * SUCH DAMAGE. 45*0Sstevel@tonic-gate * 46*0Sstevel@tonic-gate * @(#)btree.h 10.26 (Sleepycat) 12/16/98 47*0Sstevel@tonic-gate */ 48*0Sstevel@tonic-gate 49*0Sstevel@tonic-gate /* Forward structure declarations. */ 50*0Sstevel@tonic-gate struct __btree; typedef struct __btree BTREE; 51*0Sstevel@tonic-gate struct __cursor; typedef struct __cursor CURSOR; 52*0Sstevel@tonic-gate struct __epg; typedef struct __epg EPG; 53*0Sstevel@tonic-gate struct __recno; typedef struct __recno RECNO; 54*0Sstevel@tonic-gate 55*0Sstevel@tonic-gate #define DEFMINKEYPAGE (2) 56*0Sstevel@tonic-gate 57*0Sstevel@tonic-gate #define ISINTERNAL(p) (TYPE(p) == P_IBTREE || TYPE(p) == P_IRECNO) 58*0Sstevel@tonic-gate #define ISLEAF(p) (TYPE(p) == P_LBTREE || TYPE(p) == P_LRECNO) 59*0Sstevel@tonic-gate 60*0Sstevel@tonic-gate /* 61*0Sstevel@tonic-gate * If doing transactions we have to hold the locks associated with a data item 62*0Sstevel@tonic-gate * from a page for the entire transaction. However, we don't have to hold the 63*0Sstevel@tonic-gate * locks associated with walking the tree. Distinguish between the two so that 64*0Sstevel@tonic-gate * we don't tie up the internal pages of the tree longer than necessary. 65*0Sstevel@tonic-gate */ 66*0Sstevel@tonic-gate #define __BT_LPUT(dbc, lock) \ 67*0Sstevel@tonic-gate (F_ISSET((dbc)->dbp, DB_AM_LOCKING) ? \ 68*0Sstevel@tonic-gate lock_put((dbc)->dbp->dbenv->lk_info, lock) : 0) 69*0Sstevel@tonic-gate #define __BT_TLPUT(dbc, lock) \ 70*0Sstevel@tonic-gate (F_ISSET((dbc)->dbp, DB_AM_LOCKING) && (dbc)->txn == NULL ? \ 71*0Sstevel@tonic-gate lock_put((dbc)->dbp->dbenv->lk_info, lock) : 0) 72*0Sstevel@tonic-gate 73*0Sstevel@tonic-gate /* 74*0Sstevel@tonic-gate * Flags to __bam_search() and __bam_rsearch(). 75*0Sstevel@tonic-gate * 76*0Sstevel@tonic-gate * Note, internal page searches must find the largest record less than key in 77*0Sstevel@tonic-gate * the tree so that descents work. Leaf page searches must find the smallest 78*0Sstevel@tonic-gate * record greater than key so that the returned index is the record's correct 79*0Sstevel@tonic-gate * position for insertion. 80*0Sstevel@tonic-gate * 81*0Sstevel@tonic-gate * The flags parameter to the search routines describes three aspects of the 82*0Sstevel@tonic-gate * search: the type of locking required (including if we're locking a pair of 83*0Sstevel@tonic-gate * pages), the item to return in the presence of duplicates and whether or not 84*0Sstevel@tonic-gate * to return deleted entries. To simplify both the mnemonic representation 85*0Sstevel@tonic-gate * and the code that checks for various cases, we construct a set of bitmasks. 86*0Sstevel@tonic-gate */ 87*0Sstevel@tonic-gate #define S_READ 0x00001 /* Read locks. */ 88*0Sstevel@tonic-gate #define S_WRITE 0x00002 /* Write locks. */ 89*0Sstevel@tonic-gate 90*0Sstevel@tonic-gate #define S_APPEND 0x00040 /* Append to the tree. */ 91*0Sstevel@tonic-gate #define S_DELNO 0x00080 /* Don't return deleted items. */ 92*0Sstevel@tonic-gate #define S_DUPFIRST 0x00100 /* Return first duplicate. */ 93*0Sstevel@tonic-gate #define S_DUPLAST 0x00200 /* Return last duplicate. */ 94*0Sstevel@tonic-gate #define S_EXACT 0x00400 /* Exact items only. */ 95*0Sstevel@tonic-gate #define S_PARENT 0x00800 /* Lock page pair. */ 96*0Sstevel@tonic-gate #define S_STACK 0x01000 /* Need a complete stack. */ 97*0Sstevel@tonic-gate #define S_PAST_EOF 0x02000 /* If doing insert search (or keyfirst 98*0Sstevel@tonic-gate * or keylast operations), or a split 99*0Sstevel@tonic-gate * on behalf of an insert, it's okay to 100*0Sstevel@tonic-gate * return an entry one past end-of-page. 101*0Sstevel@tonic-gate */ 102*0Sstevel@tonic-gate 103*0Sstevel@tonic-gate #define S_DELETE (S_WRITE | S_DUPFIRST | S_DELNO | S_EXACT | S_STACK) 104*0Sstevel@tonic-gate #define S_FIND (S_READ | S_DUPFIRST | S_DELNO) 105*0Sstevel@tonic-gate #define S_FIND_WR (S_WRITE | S_DUPFIRST | S_DELNO) 106*0Sstevel@tonic-gate #define S_INSERT (S_WRITE | S_DUPLAST | S_PAST_EOF | S_STACK) 107*0Sstevel@tonic-gate #define S_KEYFIRST (S_WRITE | S_DUPFIRST | S_PAST_EOF | S_STACK) 108*0Sstevel@tonic-gate #define S_KEYLAST (S_WRITE | S_DUPLAST | S_PAST_EOF | S_STACK) 109*0Sstevel@tonic-gate #define S_WRPAIR (S_WRITE | S_DUPLAST | S_PAST_EOF | S_PARENT) 110*0Sstevel@tonic-gate 111*0Sstevel@tonic-gate /* 112*0Sstevel@tonic-gate * Flags to __bam_iitem(). 113*0Sstevel@tonic-gate */ 114*0Sstevel@tonic-gate #define BI_DELETED 0x01 /* Key/data pair only placeholder. */ 115*0Sstevel@tonic-gate #define BI_DOINCR 0x02 /* Increment the record count. */ 116*0Sstevel@tonic-gate #define BI_NEWKEY 0x04 /* New key. */ 117*0Sstevel@tonic-gate 118*0Sstevel@tonic-gate /* 119*0Sstevel@tonic-gate * Various routines pass around page references. A page reference can be a 120*0Sstevel@tonic-gate * pointer to the page or a page number; for either, an indx can designate 121*0Sstevel@tonic-gate * an item on the page. 122*0Sstevel@tonic-gate */ 123*0Sstevel@tonic-gate struct __epg { 124*0Sstevel@tonic-gate PAGE *page; /* The page. */ 125*0Sstevel@tonic-gate db_indx_t indx; /* The index on the page. */ 126*0Sstevel@tonic-gate DB_LOCK lock; /* The page's lock. */ 127*0Sstevel@tonic-gate }; 128*0Sstevel@tonic-gate 129*0Sstevel@tonic-gate /* 130*0Sstevel@tonic-gate * We maintain a stack of the pages that we're locking in the tree. Btree's 131*0Sstevel@tonic-gate * (currently) only save two levels of the tree at a time, so the default 132*0Sstevel@tonic-gate * stack is always large enough. Recno trees have to lock the entire tree to 133*0Sstevel@tonic-gate * do inserts/deletes, however. Grow the stack as necessary. 134*0Sstevel@tonic-gate */ 135*0Sstevel@tonic-gate #define BT_STK_CLR(c) \ 136*0Sstevel@tonic-gate ((c)->csp = (c)->sp) 137*0Sstevel@tonic-gate 138*0Sstevel@tonic-gate #define BT_STK_ENTER(c, pagep, page_indx, lock, ret) do { \ 139*0Sstevel@tonic-gate if ((ret = \ 140*0Sstevel@tonic-gate (c)->csp == (c)->esp ? __bam_stkgrow(c) : 0) == 0) { \ 141*0Sstevel@tonic-gate (c)->csp->page = pagep; \ 142*0Sstevel@tonic-gate (c)->csp->indx = page_indx; \ 143*0Sstevel@tonic-gate (c)->csp->lock = lock; \ 144*0Sstevel@tonic-gate } \ 145*0Sstevel@tonic-gate } while (0) 146*0Sstevel@tonic-gate 147*0Sstevel@tonic-gate #define BT_STK_PUSH(c, pagep, page_indx, lock, ret) do { \ 148*0Sstevel@tonic-gate BT_STK_ENTER(c, pagep, page_indx, lock, ret); \ 149*0Sstevel@tonic-gate ++(c)->csp; \ 150*0Sstevel@tonic-gate } while (0) 151*0Sstevel@tonic-gate 152*0Sstevel@tonic-gate #define BT_STK_POP(c) \ 153*0Sstevel@tonic-gate ((c)->csp == (c)->stack ? NULL : --(c)->csp) 154*0Sstevel@tonic-gate 155*0Sstevel@tonic-gate /* 156*0Sstevel@tonic-gate * Arguments passed to __bam_ca_replace(). 157*0Sstevel@tonic-gate */ 158*0Sstevel@tonic-gate typedef enum { 159*0Sstevel@tonic-gate REPLACE_SETUP, 160*0Sstevel@tonic-gate REPLACE_SUCCESS, 161*0Sstevel@tonic-gate REPLACE_FAILED 162*0Sstevel@tonic-gate } ca_replace_arg; 163*0Sstevel@tonic-gate 164*0Sstevel@tonic-gate /* Arguments passed to __ram_ca(). */ 165*0Sstevel@tonic-gate typedef enum { 166*0Sstevel@tonic-gate CA_DELETE, 167*0Sstevel@tonic-gate CA_IAFTER, 168*0Sstevel@tonic-gate CA_IBEFORE 169*0Sstevel@tonic-gate } ca_recno_arg; 170*0Sstevel@tonic-gate 171*0Sstevel@tonic-gate #define RECNO_OOB 0 /* Illegal record number. */ 172*0Sstevel@tonic-gate 173*0Sstevel@tonic-gate /* Btree/Recno cursor. */ 174*0Sstevel@tonic-gate struct __cursor { 175*0Sstevel@tonic-gate DBC *dbc; /* Enclosing DBC. */ 176*0Sstevel@tonic-gate 177*0Sstevel@tonic-gate /* Per-thread information: shared by btree/recno. */ 178*0Sstevel@tonic-gate EPG *sp; /* Stack pointer. */ 179*0Sstevel@tonic-gate EPG *csp; /* Current stack entry. */ 180*0Sstevel@tonic-gate EPG *esp; /* End stack pointer. */ 181*0Sstevel@tonic-gate EPG stack[5]; 182*0Sstevel@tonic-gate 183*0Sstevel@tonic-gate /* Per-thread information: btree private. */ 184*0Sstevel@tonic-gate PAGE *page; /* Cursor page. */ 185*0Sstevel@tonic-gate 186*0Sstevel@tonic-gate db_pgno_t pgno; /* Page. */ 187*0Sstevel@tonic-gate db_indx_t indx; /* Page item ref'd by the cursor. */ 188*0Sstevel@tonic-gate 189*0Sstevel@tonic-gate db_pgno_t dpgno; /* Duplicate page. */ 190*0Sstevel@tonic-gate db_indx_t dindx; /* Page item ref'd by the cursor. */ 191*0Sstevel@tonic-gate 192*0Sstevel@tonic-gate DB_LOCK lock; /* Cursor read lock. */ 193*0Sstevel@tonic-gate db_lockmode_t mode; /* Lock mode. */ 194*0Sstevel@tonic-gate 195*0Sstevel@tonic-gate /* Per-thread information: recno private. */ 196*0Sstevel@tonic-gate db_recno_t recno; /* Current record number. */ 197*0Sstevel@tonic-gate 198*0Sstevel@tonic-gate /* 199*0Sstevel@tonic-gate * Btree: 200*0Sstevel@tonic-gate * We set a flag in the cursor structure if the underlying object has 201*0Sstevel@tonic-gate * been deleted. It's not strictly necessary, we could get the same 202*0Sstevel@tonic-gate * information by looking at the page itself. 203*0Sstevel@tonic-gate * 204*0Sstevel@tonic-gate * Recno: 205*0Sstevel@tonic-gate * When renumbering recno databases during deletes, cursors referencing 206*0Sstevel@tonic-gate * "deleted" records end up positioned between two records, and so must 207*0Sstevel@tonic-gate * be specially adjusted on the next operation. 208*0Sstevel@tonic-gate */ 209*0Sstevel@tonic-gate #define C_DELETED 0x0001 /* Record was deleted. */ 210*0Sstevel@tonic-gate u_int32_t flags; 211*0Sstevel@tonic-gate }; 212*0Sstevel@tonic-gate 213*0Sstevel@tonic-gate /* 214*0Sstevel@tonic-gate * The in-memory recno data structure. 215*0Sstevel@tonic-gate * 216*0Sstevel@tonic-gate * !!! 217*0Sstevel@tonic-gate * These fields are ignored as far as multi-threading is concerned. There 218*0Sstevel@tonic-gate * are no transaction semantics associated with backing files, nor is there 219*0Sstevel@tonic-gate * any thread protection. 220*0Sstevel@tonic-gate */ 221*0Sstevel@tonic-gate struct __recno { 222*0Sstevel@tonic-gate int re_delim; /* Variable-length delimiting byte. */ 223*0Sstevel@tonic-gate int re_pad; /* Fixed-length padding byte. */ 224*0Sstevel@tonic-gate u_int32_t re_len; /* Length for fixed-length records. */ 225*0Sstevel@tonic-gate 226*0Sstevel@tonic-gate char *re_source; /* Source file name. */ 227*0Sstevel@tonic-gate int re_fd; /* Source file descriptor */ 228*0Sstevel@tonic-gate db_recno_t re_last; /* Last record number read. */ 229*0Sstevel@tonic-gate void *re_cmap; /* Current point in mapped space. */ 230*0Sstevel@tonic-gate void *re_smap; /* Start of mapped space. */ 231*0Sstevel@tonic-gate void *re_emap; /* End of mapped space. */ 232*0Sstevel@tonic-gate size_t re_msize; /* Size of mapped region. */ 233*0Sstevel@tonic-gate /* Recno input function. */ 234*0Sstevel@tonic-gate int (*re_irec) __P((DBC *, db_recno_t)); 235*0Sstevel@tonic-gate 236*0Sstevel@tonic-gate #define RECNO_EOF 0x0001 /* EOF on backing source file. */ 237*0Sstevel@tonic-gate #define RECNO_MODIFIED 0x0002 /* Tree was modified. */ 238*0Sstevel@tonic-gate u_int32_t flags; 239*0Sstevel@tonic-gate }; 240*0Sstevel@tonic-gate 241*0Sstevel@tonic-gate /* 242*0Sstevel@tonic-gate * The in-memory, per-tree btree data structure. 243*0Sstevel@tonic-gate */ 244*0Sstevel@tonic-gate struct __btree { 245*0Sstevel@tonic-gate db_pgno_t bt_lpgno; /* Last insert location. */ 246*0Sstevel@tonic-gate 247*0Sstevel@tonic-gate db_indx_t bt_maxkey; /* Maximum keys per page. */ 248*0Sstevel@tonic-gate db_indx_t bt_minkey; /* Minimum keys per page. */ 249*0Sstevel@tonic-gate 250*0Sstevel@tonic-gate int (*bt_compare) /* Comparison function. */ 251*0Sstevel@tonic-gate __P((const DBT *, const DBT *)); 252*0Sstevel@tonic-gate size_t(*bt_prefix) /* Prefix function. */ 253*0Sstevel@tonic-gate __P((const DBT *, const DBT *)); 254*0Sstevel@tonic-gate 255*0Sstevel@tonic-gate db_indx_t bt_ovflsize; /* Maximum key/data on-page size. */ 256*0Sstevel@tonic-gate 257*0Sstevel@tonic-gate RECNO *recno; /* Private recno structure. */ 258*0Sstevel@tonic-gate }; 259*0Sstevel@tonic-gate 260*0Sstevel@tonic-gate #include "btree_auto.h" 261*0Sstevel@tonic-gate #include "btree_ext.h" 262*0Sstevel@tonic-gate #include "db_am.h" 263*0Sstevel@tonic-gate #include "common_ext.h" 264