1 /*- 2 * Copyright (c) 1990, 1993, 1994 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Mike Olson. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)bt_put.c 8.8 (Berkeley) 7/26/94 37 * $DragonFly: src/lib/libc/db/btree/bt_put.c,v 1.5 2005/03/16 06:24:36 joerg Exp $ 38 */ 39 40 #include <sys/types.h> 41 42 #include <errno.h> 43 #include <stdio.h> 44 #include <stdlib.h> 45 #include <string.h> 46 47 #include <db.h> 48 #include "btree.h" 49 50 static EPG *bt_fast (BTREE *, const DBT *, const DBT *, int *); 51 52 /* 53 * __BT_PUT -- Add a btree item to the tree. 54 * 55 * Parameters: 56 * dbp: pointer to access method 57 * key: key 58 * data: data 59 * flag: R_NOOVERWRITE 60 * 61 * Returns: 62 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 63 * tree and R_NOOVERWRITE specified. 64 */ 65 int 66 __bt_put(dbp, key, data, flags) 67 const DB *dbp; 68 DBT *key; 69 const DBT *data; 70 u_int flags; 71 { 72 BTREE *t; 73 DBT tkey, tdata; 74 EPG *e; 75 PAGE *h; 76 indx_t idx, nxtindex; 77 pgno_t pg; 78 u_int32_t nbytes; 79 int dflags, exact, status; 80 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 81 82 t = dbp->internal; 83 84 /* Toss any page pinned across calls. */ 85 if (t->bt_pinned != NULL) { 86 mpool_put(t->bt_mp, t->bt_pinned, 0); 87 t->bt_pinned = NULL; 88 } 89 90 /* Check for change to a read-only tree. */ 91 if (F_ISSET(t, B_RDONLY)) { 92 errno = EPERM; 93 return (RET_ERROR); 94 } 95 96 switch (flags) { 97 case 0: 98 case R_NOOVERWRITE: 99 break; 100 case R_CURSOR: 101 /* 102 * If flags is R_CURSOR, put the cursor. Must already 103 * have started a scan and not have already deleted it. 104 */ 105 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 106 !F_ISSET(&t->bt_cursor, 107 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 108 break; 109 /* FALLTHROUGH */ 110 default: 111 errno = EINVAL; 112 return (RET_ERROR); 113 } 114 115 /* 116 * If the key/data pair won't fit on a page, store it on overflow 117 * pages. Only put the key on the overflow page if the pair are 118 * still too big after moving the data to an overflow page. 119 * 120 * XXX 121 * If the insert fails later on, the overflow pages aren't recovered. 122 */ 123 dflags = 0; 124 if (key->size + data->size > t->bt_ovflsize) { 125 if (key->size > t->bt_ovflsize) { 126 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 127 return (RET_ERROR); 128 tkey.data = kb; 129 tkey.size = NOVFLSIZE; 130 memmove(kb, &pg, sizeof(pgno_t)); 131 memmove(kb + sizeof(pgno_t), 132 &key->size, sizeof(u_int32_t)); 133 dflags |= P_BIGKEY; 134 key = &tkey; 135 } 136 if (key->size + data->size > t->bt_ovflsize) { 137 if (__ovfl_put(t, data, &pg) == RET_ERROR) 138 return (RET_ERROR); 139 tdata.data = db; 140 tdata.size = NOVFLSIZE; 141 memmove(db, &pg, sizeof(pgno_t)); 142 memmove(db + sizeof(pgno_t), 143 &data->size, sizeof(u_int32_t)); 144 dflags |= P_BIGDATA; 145 data = &tdata; 146 } 147 if (key->size + data->size > t->bt_ovflsize) 148 goto storekey; 149 } 150 151 /* Replace the cursor. */ 152 if (flags == R_CURSOR) { 153 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 154 return (RET_ERROR); 155 idx = t->bt_cursor.pg.index; 156 goto delete; 157 } 158 159 /* 160 * Find the key to delete, or, the location at which to insert. 161 * Bt_fast and __bt_search both pin the returned page. 162 */ 163 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 164 if ((e = __bt_search(t, key, &exact)) == NULL) 165 return (RET_ERROR); 166 h = e->page; 167 idx = e->index; 168 169 /* 170 * Add the key/data pair to the tree. If an identical key is already 171 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 172 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 173 * permitted) or an error is returned. 174 */ 175 switch (flags) { 176 case R_NOOVERWRITE: 177 if (!exact) 178 break; 179 mpool_put(t->bt_mp, h, 0); 180 return (RET_SPECIAL); 181 default: 182 if (!exact || !F_ISSET(t, B_NODUPS)) 183 break; 184 /* 185 * !!! 186 * Note, the delete may empty the page, so we need to put a 187 * new entry into the page immediately. 188 */ 189 delete: if (__bt_dleaf(t, key, h, idx) == RET_ERROR) { 190 mpool_put(t->bt_mp, h, 0); 191 return (RET_ERROR); 192 } 193 break; 194 } 195 196 /* 197 * If not enough room, or the user has put a ceiling on the number of 198 * keys permitted in the page, split the page. The split code will 199 * insert the key and data and unpin the current page. If inserting 200 * into the offset array, shift the pointers up. 201 */ 202 nbytes = NBLEAFDBT(key->size, data->size); 203 if ((size_t)(h->upper - h->lower) < nbytes + sizeof(indx_t)) { 204 if ((status = __bt_split(t, h, key, 205 data, dflags, nbytes, idx)) != RET_SUCCESS) 206 return (status); 207 goto success; 208 } 209 210 if (idx < (nxtindex = NEXTINDEX(h))) 211 memmove(h->linp + idx + 1, h->linp + idx, 212 (nxtindex - idx) * sizeof(indx_t)); 213 h->lower += sizeof(indx_t); 214 215 h->linp[idx] = h->upper -= nbytes; 216 dest = (char *)h + h->upper; 217 WR_BLEAF(dest, key, data, dflags); 218 219 /* If the cursor is on this page, adjust it as necessary. */ 220 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 221 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 222 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx) 223 ++t->bt_cursor.pg.index; 224 225 if (t->bt_order == NOT) { 226 if (h->nextpg == P_INVALID) { 227 if (idx == NEXTINDEX(h) - 1) { 228 t->bt_order = FORWARD; 229 t->bt_last.index = idx; 230 t->bt_last.pgno = h->pgno; 231 } 232 } else if (h->prevpg == P_INVALID) { 233 if (idx == 0) { 234 t->bt_order = BACK; 235 t->bt_last.index = 0; 236 t->bt_last.pgno = h->pgno; 237 } 238 } 239 } 240 241 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 242 243 success: 244 if (flags == R_SETCURSOR) 245 __bt_setcur(t, e->page->pgno, e->index); 246 247 F_SET(t, B_MODIFIED); 248 return (RET_SUCCESS); 249 } 250 251 #ifdef STATISTICS 252 u_long bt_cache_hit, bt_cache_miss; 253 #endif 254 255 /* 256 * BT_FAST -- Do a quick check for sorted data. 257 * 258 * Parameters: 259 * t: tree 260 * key: key to insert 261 * 262 * Returns: 263 * EPG for new record or NULL if not found. 264 */ 265 static EPG * 266 bt_fast(t, key, data, exactp) 267 BTREE *t; 268 const DBT *key, *data; 269 int *exactp; 270 { 271 PAGE *h; 272 u_int32_t nbytes; 273 int cmp; 274 275 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 276 t->bt_order = NOT; 277 return (NULL); 278 } 279 t->bt_cur.page = h; 280 t->bt_cur.index = t->bt_last.index; 281 282 /* 283 * If won't fit in this page or have too many keys in this page, 284 * have to search to get split stack. 285 */ 286 nbytes = NBLEAFDBT(key->size, data->size); 287 if ((size_t)(h->upper - h->lower) < nbytes + sizeof(indx_t)) 288 goto miss; 289 290 if (t->bt_order == FORWARD) { 291 if (t->bt_cur.page->nextpg != P_INVALID) 292 goto miss; 293 if (t->bt_cur.index != NEXTINDEX(h) - 1) 294 goto miss; 295 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 296 goto miss; 297 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 298 } else { 299 if (t->bt_cur.page->prevpg != P_INVALID) 300 goto miss; 301 if (t->bt_cur.index != 0) 302 goto miss; 303 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 304 goto miss; 305 t->bt_last.index = 0; 306 } 307 *exactp = cmp == 0; 308 #ifdef STATISTICS 309 ++bt_cache_hit; 310 #endif 311 return (&t->bt_cur); 312 313 miss: 314 #ifdef STATISTICS 315 ++bt_cache_miss; 316 #endif 317 t->bt_order = NOT; 318 mpool_put(t->bt_mp, h, 0); 319 return (NULL); 320 } 321