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. Neither the name of the University nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 * 32 * @(#)bt_put.c 8.8 (Berkeley) 7/26/94 33 * $DragonFly: src/lib/libc/db/btree/bt_put.c,v 1.7 2005/09/19 09:20:37 asmodai Exp $ 34 */ 35 36 #include <sys/types.h> 37 38 #include <errno.h> 39 #include <stdio.h> 40 #include <stdlib.h> 41 #include <string.h> 42 43 #include <db.h> 44 #include "btree.h" 45 46 static EPG *bt_fast (BTREE *, const DBT *, const DBT *, int *); 47 48 /* 49 * __BT_PUT -- Add a btree item to the tree. 50 * 51 * Parameters: 52 * dbp: pointer to access method 53 * key: key 54 * data: data 55 * flag: R_NOOVERWRITE 56 * 57 * Returns: 58 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 59 * tree and R_NOOVERWRITE specified. 60 */ 61 int 62 __bt_put(dbp, key, data, flags) 63 const DB *dbp; 64 DBT *key; 65 const DBT *data; 66 u_int flags; 67 { 68 BTREE *t; 69 DBT tkey, tdata; 70 EPG *e = NULL; 71 PAGE *h; 72 indx_t idx, nxtindex; 73 pgno_t pg; 74 u_int32_t nbytes; 75 int dflags, exact, status; 76 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 77 78 t = dbp->internal; 79 80 /* Toss any page pinned across calls. */ 81 if (t->bt_pinned != NULL) { 82 mpool_put(t->bt_mp, t->bt_pinned, 0); 83 t->bt_pinned = NULL; 84 } 85 86 /* Check for change to a read-only tree. */ 87 if (F_ISSET(t, B_RDONLY)) { 88 errno = EPERM; 89 return (RET_ERROR); 90 } 91 92 switch (flags) { 93 case 0: 94 case R_NOOVERWRITE: 95 break; 96 case R_CURSOR: 97 /* 98 * If flags is R_CURSOR, put the cursor. Must already 99 * have started a scan and not have already deleted it. 100 */ 101 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 102 !F_ISSET(&t->bt_cursor, 103 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 104 break; 105 /* FALLTHROUGH */ 106 default: 107 errno = EINVAL; 108 return (RET_ERROR); 109 } 110 111 /* 112 * If the key/data pair won't fit on a page, store it on overflow 113 * pages. Only put the key on the overflow page if the pair are 114 * still too big after moving the data to an overflow page. 115 * 116 * XXX 117 * If the insert fails later on, the overflow pages aren't recovered. 118 */ 119 dflags = 0; 120 if (key->size + data->size > t->bt_ovflsize) { 121 if (key->size > t->bt_ovflsize) { 122 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 123 return (RET_ERROR); 124 tkey.data = kb; 125 tkey.size = NOVFLSIZE; 126 memmove(kb, &pg, sizeof(pgno_t)); 127 memmove(kb + sizeof(pgno_t), 128 &key->size, sizeof(u_int32_t)); 129 dflags |= P_BIGKEY; 130 key = &tkey; 131 } 132 if (key->size + data->size > t->bt_ovflsize) { 133 if (__ovfl_put(t, data, &pg) == RET_ERROR) 134 return (RET_ERROR); 135 tdata.data = db; 136 tdata.size = NOVFLSIZE; 137 memmove(db, &pg, sizeof(pgno_t)); 138 memmove(db + sizeof(pgno_t), 139 &data->size, sizeof(u_int32_t)); 140 dflags |= P_BIGDATA; 141 data = &tdata; 142 } 143 if (key->size + data->size > t->bt_ovflsize) 144 goto storekey; 145 } 146 147 /* Replace the cursor. */ 148 if (flags == R_CURSOR) { 149 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 150 return (RET_ERROR); 151 idx = t->bt_cursor.pg.index; 152 goto delete; 153 } 154 155 /* 156 * Find the key to delete, or, the location at which to insert. 157 * Bt_fast and __bt_search both pin the returned page. 158 */ 159 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 160 if ((e = __bt_search(t, key, &exact)) == NULL) 161 return (RET_ERROR); 162 h = e->page; 163 idx = e->index; 164 165 /* 166 * Add the key/data pair to the tree. If an identical key is already 167 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 168 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 169 * permitted) or an error is returned. 170 */ 171 switch (flags) { 172 case R_NOOVERWRITE: 173 if (!exact) 174 break; 175 mpool_put(t->bt_mp, h, 0); 176 return (RET_SPECIAL); 177 default: 178 if (!exact || !F_ISSET(t, B_NODUPS)) 179 break; 180 /* 181 * !!! 182 * Note, the delete may empty the page, so we need to put a 183 * new entry into the page immediately. 184 */ 185 delete: if (__bt_dleaf(t, key, h, idx) == RET_ERROR) { 186 mpool_put(t->bt_mp, h, 0); 187 return (RET_ERROR); 188 } 189 break; 190 } 191 192 /* 193 * If not enough room, or the user has put a ceiling on the number of 194 * keys permitted in the page, split the page. The split code will 195 * insert the key and data and unpin the current page. If inserting 196 * into the offset array, shift the pointers up. 197 */ 198 nbytes = NBLEAFDBT(key->size, data->size); 199 if ((size_t)(h->upper - h->lower) < nbytes + sizeof(indx_t)) { 200 if ((status = __bt_split(t, h, key, 201 data, dflags, nbytes, idx)) != RET_SUCCESS) 202 return (status); 203 goto success; 204 } 205 206 if (idx < (nxtindex = NEXTINDEX(h))) 207 memmove(h->linp + idx + 1, h->linp + idx, 208 (nxtindex - idx) * sizeof(indx_t)); 209 h->lower += sizeof(indx_t); 210 211 h->linp[idx] = h->upper -= nbytes; 212 dest = (char *)h + h->upper; 213 WR_BLEAF(dest, key, data, dflags); 214 215 /* If the cursor is on this page, adjust it as necessary. */ 216 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 217 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 218 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx) 219 ++t->bt_cursor.pg.index; 220 221 if (t->bt_order == NOT) { 222 if (h->nextpg == P_INVALID) { 223 if (idx == NEXTINDEX(h) - 1) { 224 t->bt_order = FORWARD; 225 t->bt_last.index = idx; 226 t->bt_last.pgno = h->pgno; 227 } 228 } else if (h->prevpg == P_INVALID) { 229 if (idx == 0) { 230 t->bt_order = BACK; 231 t->bt_last.index = 0; 232 t->bt_last.pgno = h->pgno; 233 } 234 } 235 } 236 237 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 238 239 success: 240 if (flags == R_SETCURSOR) 241 __bt_setcur(t, e->page->pgno, e->index); 242 243 F_SET(t, B_MODIFIED); 244 return (RET_SUCCESS); 245 } 246 247 #ifdef STATISTICS 248 u_long bt_cache_hit, bt_cache_miss; 249 #endif 250 251 /* 252 * BT_FAST -- Do a quick check for sorted data. 253 * 254 * Parameters: 255 * t: tree 256 * key: key to insert 257 * 258 * Returns: 259 * EPG for new record or NULL if not found. 260 */ 261 static EPG * 262 bt_fast(t, key, data, exactp) 263 BTREE *t; 264 const DBT *key, *data; 265 int *exactp; 266 { 267 PAGE *h; 268 u_int32_t nbytes; 269 int cmp; 270 271 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 272 t->bt_order = NOT; 273 return (NULL); 274 } 275 t->bt_cur.page = h; 276 t->bt_cur.index = t->bt_last.index; 277 278 /* 279 * If won't fit in this page or have too many keys in this page, 280 * have to search to get split stack. 281 */ 282 nbytes = NBLEAFDBT(key->size, data->size); 283 if ((size_t)(h->upper - h->lower) < nbytes + sizeof(indx_t)) 284 goto miss; 285 286 if (t->bt_order == FORWARD) { 287 if (t->bt_cur.page->nextpg != P_INVALID) 288 goto miss; 289 if (t->bt_cur.index != NEXTINDEX(h) - 1) 290 goto miss; 291 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 292 goto miss; 293 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 294 } else { 295 if (t->bt_cur.page->prevpg != P_INVALID) 296 goto miss; 297 if (t->bt_cur.index != 0) 298 goto miss; 299 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 300 goto miss; 301 t->bt_last.index = 0; 302 } 303 *exactp = cmp == 0; 304 #ifdef STATISTICS 305 ++bt_cache_hit; 306 #endif 307 return (&t->bt_cur); 308 309 miss: 310 #ifdef STATISTICS 311 ++bt_cache_miss; 312 #endif 313 t->bt_order = NOT; 314 mpool_put(t->bt_mp, h, 0); 315 return (NULL); 316 } 317