1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1990, 1993, 1994 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Mike Olson. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 */ 34 35 #include <sys/types.h> 36 37 #include <errno.h> 38 #include <stdio.h> 39 #include <stdlib.h> 40 #include <string.h> 41 42 #include <db.h> 43 #include "btree.h" 44 45 static EPG *bt_fast(BTREE *, const DBT *, const DBT *, int *); 46 47 /* 48 * __BT_PUT -- Add a btree item to the tree. 49 * 50 * Parameters: 51 * dbp: pointer to access method 52 * key: key 53 * data: data 54 * flag: R_NOOVERWRITE, R_SETCURSOR, R_CURSOR 55 * 56 * Returns: 57 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 58 * tree and R_NOOVERWRITE specified. 59 */ 60 int 61 __bt_put(const DB *dbp, DBT *key, const DBT *data, u_int flags) 62 { 63 BTREE *t; 64 DBT tkey, tdata; 65 EPG *e; 66 PAGE *h; 67 indx_t idx, nxtindex; 68 pgno_t pg; 69 u_int32_t nbytes, tmp; 70 int dflags, exact, status; 71 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 72 73 t = dbp->internal; 74 75 /* Toss any page pinned across calls. */ 76 if (t->bt_pinned != NULL) { 77 mpool_put(t->bt_mp, t->bt_pinned, 0); 78 t->bt_pinned = NULL; 79 } 80 81 /* Check for change to a read-only tree. */ 82 if (F_ISSET(t, B_RDONLY)) { 83 errno = EPERM; 84 return (RET_ERROR); 85 } 86 87 switch (flags) { 88 case 0: 89 case R_NOOVERWRITE: 90 case R_SETCURSOR: 91 break; 92 case R_CURSOR: 93 /* 94 * If flags is R_CURSOR, put the cursor. Must already 95 * have started a scan and not have already deleted it. 96 */ 97 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 98 !F_ISSET(&t->bt_cursor, 99 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 100 break; 101 /* FALLTHROUGH */ 102 default: 103 errno = EINVAL; 104 return (RET_ERROR); 105 } 106 107 /* 108 * If the key/data pair won't fit on a page, store it on overflow 109 * pages. Only put the key on the overflow page if the pair are 110 * still too big after moving the data to an overflow page. 111 * 112 * XXX 113 * If the insert fails later on, the overflow pages aren't recovered. 114 */ 115 dflags = 0; 116 if (key->size + data->size > t->bt_ovflsize) { 117 if (key->size > t->bt_ovflsize) { 118 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 119 return (RET_ERROR); 120 tkey.data = kb; 121 tkey.size = NOVFLSIZE; 122 memmove(kb, &pg, sizeof(pgno_t)); 123 tmp = key->size; 124 memmove(kb + sizeof(pgno_t), 125 &tmp, sizeof(u_int32_t)); 126 dflags |= P_BIGKEY; 127 key = &tkey; 128 } 129 if (key->size + data->size > t->bt_ovflsize) { 130 if (__ovfl_put(t, data, &pg) == RET_ERROR) 131 return (RET_ERROR); 132 tdata.data = db; 133 tdata.size = NOVFLSIZE; 134 memmove(db, &pg, sizeof(pgno_t)); 135 tmp = data->size; 136 memmove(db + sizeof(pgno_t), 137 &tmp, sizeof(u_int32_t)); 138 dflags |= P_BIGDATA; 139 data = &tdata; 140 } 141 if (key->size + data->size > t->bt_ovflsize) 142 goto storekey; 143 } 144 145 /* Replace the cursor. */ 146 if (flags == R_CURSOR) { 147 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 148 return (RET_ERROR); 149 idx = t->bt_cursor.pg.index; 150 goto delete; 151 } 152 153 /* 154 * Find the key to delete, or, the location at which to insert. 155 * Bt_fast and __bt_search both pin the returned page. 156 */ 157 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 158 if ((e = __bt_search(t, key, &exact)) == NULL) 159 return (RET_ERROR); 160 h = e->page; 161 idx = e->index; 162 163 /* 164 * Add the key/data pair to the tree. If an identical key is already 165 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 166 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 167 * permitted) or an error is returned. 168 */ 169 switch (flags) { 170 case R_NOOVERWRITE: 171 if (!exact) 172 break; 173 mpool_put(t->bt_mp, h, 0); 174 return (RET_SPECIAL); 175 default: 176 if (!exact || !F_ISSET(t, B_NODUPS)) 177 break; 178 /* 179 * !!! 180 * Note, the delete may empty the page, so we need to put a 181 * new entry into the page immediately. 182 */ 183 delete: if (__bt_dleaf(t, key, h, idx) == RET_ERROR) { 184 mpool_put(t->bt_mp, h, 0); 185 return (RET_ERROR); 186 } 187 break; 188 } 189 190 /* 191 * If not enough room, or the user has put a ceiling on the number of 192 * keys permitted in the page, split the page. The split code will 193 * insert the key and data and unpin the current page. If inserting 194 * into the offset array, shift the pointers up. 195 */ 196 nbytes = NBLEAFDBT(key->size, data->size); 197 if ((u_int32_t)(h->upper - h->lower) < nbytes + sizeof(indx_t)) { 198 if ((status = __bt_split(t, h, key, 199 data, dflags, nbytes, idx)) != RET_SUCCESS) 200 return (status); 201 goto success; 202 } 203 204 if (idx < (nxtindex = NEXTINDEX(h))) 205 memmove(h->linp + idx + 1, h->linp + idx, 206 (nxtindex - idx) * sizeof(indx_t)); 207 h->lower += sizeof(indx_t); 208 209 h->linp[idx] = h->upper -= nbytes; 210 dest = (char *)h + h->upper; 211 WR_BLEAF(dest, key, data, dflags); 212 213 /* If the cursor is on this page, adjust it as necessary. */ 214 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 215 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 216 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx) 217 ++t->bt_cursor.pg.index; 218 219 if (t->bt_order == NOT) { 220 if (h->nextpg == P_INVALID) { 221 if (idx == NEXTINDEX(h) - 1) { 222 t->bt_order = FORWARD; 223 t->bt_last.index = idx; 224 t->bt_last.pgno = h->pgno; 225 } 226 } else if (h->prevpg == P_INVALID) { 227 if (idx == 0) { 228 t->bt_order = BACK; 229 t->bt_last.index = 0; 230 t->bt_last.pgno = h->pgno; 231 } 232 } 233 } 234 235 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 236 237 success: 238 if (flags == R_SETCURSOR) 239 __bt_setcur(t, e->page->pgno, e->index); 240 241 F_SET(t, B_MODIFIED); 242 return (RET_SUCCESS); 243 } 244 245 #ifdef STATISTICS 246 u_long bt_cache_hit, bt_cache_miss; 247 #endif 248 249 /* 250 * BT_FAST -- Do a quick check for sorted data. 251 * 252 * Parameters: 253 * t: tree 254 * key: key to insert 255 * 256 * Returns: 257 * EPG for new record or NULL if not found. 258 */ 259 static EPG * 260 bt_fast(BTREE *t, const DBT *key, const DBT *data, int *exactp) 261 { 262 PAGE *h; 263 u_int32_t nbytes; 264 int cmp; 265 266 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 267 t->bt_order = NOT; 268 return (NULL); 269 } 270 t->bt_cur.page = h; 271 t->bt_cur.index = t->bt_last.index; 272 273 /* 274 * If won't fit in this page or have too many keys in this page, 275 * have to search to get split stack. 276 */ 277 nbytes = NBLEAFDBT(key->size, data->size); 278 if ((u_int32_t)(h->upper - h->lower) < nbytes + sizeof(indx_t)) 279 goto miss; 280 281 if (t->bt_order == FORWARD) { 282 if (t->bt_cur.page->nextpg != P_INVALID) 283 goto miss; 284 if (t->bt_cur.index != NEXTINDEX(h) - 1) 285 goto miss; 286 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 287 goto miss; 288 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 289 } else { 290 if (t->bt_cur.page->prevpg != P_INVALID) 291 goto miss; 292 if (t->bt_cur.index != 0) 293 goto miss; 294 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 295 goto miss; 296 t->bt_last.index = 0; 297 } 298 *exactp = cmp == 0; 299 #ifdef STATISTICS 300 ++bt_cache_hit; 301 #endif 302 return (&t->bt_cur); 303 304 miss: 305 #ifdef STATISTICS 306 ++bt_cache_miss; 307 #endif 308 t->bt_order = NOT; 309 mpool_put(t->bt_mp, h, 0); 310 return (NULL); 311 } 312