1 /* $NetBSD: bt_put.c,v 1.15 2007/02/03 23:46:09 christos Exp $ */ 2 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/cdefs.h> 36 #if defined(LIBC_SCCS) && !defined(lint) 37 #if 0 38 static char sccsid[] = "@(#)bt_put.c 8.8 (Berkeley) 7/26/94"; 39 #else 40 __RCSID("$NetBSD: bt_put.c,v 1.15 2007/02/03 23:46:09 christos Exp $"); 41 #endif 42 #endif /* LIBC_SCCS and not lint */ 43 44 #include "namespace.h" 45 #include <sys/types.h> 46 47 #include <assert.h> 48 #include <errno.h> 49 #include <stdio.h> 50 #include <stdlib.h> 51 #include <string.h> 52 53 #include <db.h> 54 #include "btree.h" 55 56 static EPG *bt_fast(BTREE *, const DBT *, const DBT *, int *); 57 58 /* 59 * __BT_PUT -- Add a btree item to the tree. 60 * 61 * Parameters: 62 * dbp: pointer to access method 63 * key: key 64 * data: data 65 * flag: R_NOOVERWRITE 66 * 67 * Returns: 68 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 69 * tree and R_NOOVERWRITE specified. 70 */ 71 int 72 __bt_put(const DB *dbp, DBT *key, const DBT *data, u_int flags) 73 { 74 BTREE *t; 75 DBT tkey, tdata; 76 EPG *e = NULL; /* pacify gcc */ 77 PAGE *h; 78 indx_t idx, nxtindex; 79 pgno_t pg; 80 u_int32_t nbytes, temp; 81 int dflags, exact, status; 82 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 83 84 t = dbp->internal; 85 86 /* Toss any page pinned across calls. */ 87 if (t->bt_pinned != NULL) { 88 mpool_put(t->bt_mp, t->bt_pinned, 0); 89 t->bt_pinned = NULL; 90 } 91 92 /* Check for change to a read-only tree. */ 93 if (F_ISSET(t, B_RDONLY)) { 94 errno = EPERM; 95 return (RET_ERROR); 96 } 97 98 switch (flags) { 99 case 0: 100 case R_NOOVERWRITE: 101 break; 102 case R_CURSOR: 103 /* 104 * If flags is R_CURSOR, put the cursor. Must already 105 * have started a scan and not have already deleted it. 106 */ 107 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 108 !F_ISSET(&t->bt_cursor, 109 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 110 break; 111 /* FALLTHROUGH */ 112 default: 113 errno = EINVAL; 114 return (RET_ERROR); 115 } 116 117 /* 118 * If the key/data pair won't fit on a page, store it on overflow 119 * pages. Only put the key on the overflow page if the pair are 120 * still too big after moving the data to an overflow page. 121 * 122 * XXX 123 * If the insert fails later on, the overflow pages aren't recovered. 124 */ 125 dflags = 0; 126 if (key->size + data->size > t->bt_ovflsize) { 127 if (key->size > t->bt_ovflsize) { 128 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 129 return (RET_ERROR); 130 tkey.data = kb; 131 tkey.size = NOVFLSIZE; 132 memmove(kb, &pg, sizeof(pgno_t)); 133 memmove(kb + sizeof(pgno_t), 134 &key->size, sizeof(u_int32_t)); 135 dflags |= P_BIGKEY; 136 key = &tkey; 137 } 138 if (key->size + data->size > t->bt_ovflsize) { 139 if (__ovfl_put(t, data, &pg) == RET_ERROR) 140 return (RET_ERROR); 141 tdata.data = db; 142 tdata.size = NOVFLSIZE; 143 memmove(db, &pg, sizeof(pgno_t)); 144 _DBFIT(data->size, u_int32_t); 145 temp = (u_int32_t)data->size; 146 (void)memmove(db + sizeof(pgno_t), 147 &temp, sizeof(u_int32_t)); 148 dflags |= P_BIGDATA; 149 data = &tdata; 150 } 151 if (key->size + data->size > t->bt_ovflsize) 152 goto storekey; 153 } 154 155 /* Replace the cursor. */ 156 if (flags == R_CURSOR) { 157 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 158 return (RET_ERROR); 159 idx = t->bt_cursor.pg.index; 160 goto delete; 161 } 162 163 /* 164 * Find the key to delete, or, the location at which to insert. 165 * Bt_fast and __bt_search both pin the returned page. 166 */ 167 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 168 if ((e = __bt_search(t, key, &exact)) == NULL) 169 return (RET_ERROR); 170 h = e->page; 171 idx = e->index; 172 173 /* 174 * Add the key/data pair to the tree. If an identical key is already 175 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 176 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 177 * permitted) or an error is returned. 178 */ 179 switch (flags) { 180 case R_NOOVERWRITE: 181 if (!exact) 182 break; 183 mpool_put(t->bt_mp, h, 0); 184 return (RET_SPECIAL); 185 default: 186 if (!exact || !F_ISSET(t, B_NODUPS)) 187 break; 188 /* 189 * !!! 190 * Note, the delete may empty the page, so we need to put a 191 * new entry into the page immediately. 192 */ 193 delete: if (__bt_dleaf(t, key, h, (u_int)idx) == RET_ERROR) { 194 mpool_put(t->bt_mp, h, 0); 195 return (RET_ERROR); 196 } 197 break; 198 } 199 200 /* 201 * If not enough room, or the user has put a ceiling on the number of 202 * keys permitted in the page, split the page. The split code will 203 * insert the key and data and unpin the current page. If inserting 204 * into the offset array, shift the pointers up. 205 */ 206 nbytes = NBLEAFDBT(key->size, data->size); 207 if (h->upper - h->lower < nbytes + sizeof(indx_t)) { 208 if ((status = __bt_split(t, h, key, 209 data, dflags, nbytes, (u_int)idx)) != RET_SUCCESS) 210 return (status); 211 goto success; 212 } 213 214 if (idx < (nxtindex = NEXTINDEX(h))) 215 memmove(h->linp + idx + 1, h->linp + idx, 216 (nxtindex - idx) * sizeof(indx_t)); 217 h->lower += sizeof(indx_t); 218 219 h->linp[idx] = h->upper -= nbytes; 220 dest = (char *)(void *)h + h->upper; 221 WR_BLEAF(dest, key, data, dflags); 222 223 /* If the cursor is on this page, adjust it as necessary. */ 224 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 225 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 226 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx) 227 ++t->bt_cursor.pg.index; 228 229 if (t->bt_order == NOT) { 230 if (h->nextpg == P_INVALID) { 231 if (idx == NEXTINDEX(h) - 1) { 232 t->bt_order = FORWARD; 233 t->bt_last.index = idx; 234 t->bt_last.pgno = h->pgno; 235 } 236 } else if (h->prevpg == P_INVALID) { 237 if (idx == 0) { 238 t->bt_order = BACK; 239 t->bt_last.index = 0; 240 t->bt_last.pgno = h->pgno; 241 } 242 } 243 } 244 245 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 246 247 success: 248 if (flags == R_SETCURSOR) 249 __bt_setcur(t, e->page->pgno, (u_int)e->index); 250 251 F_SET(t, B_MODIFIED); 252 return (RET_SUCCESS); 253 } 254 255 #ifdef STATISTICS 256 u_long bt_cache_hit, bt_cache_miss; 257 #endif 258 259 /* 260 * BT_FAST -- Do a quick check for sorted data. 261 * 262 * Parameters: 263 * t: tree 264 * key: key to insert 265 * 266 * Returns: 267 * EPG for new record or NULL if not found. 268 */ 269 static EPG * 270 bt_fast(BTREE *t, const DBT *key, const DBT *data, int *exactp) 271 { 272 PAGE *h; 273 u_int32_t nbytes; 274 int cmp; 275 276 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 277 t->bt_order = NOT; 278 return (NULL); 279 } 280 t->bt_cur.page = h; 281 t->bt_cur.index = t->bt_last.index; 282 283 /* 284 * If won't fit in this page or have too many keys in this page, 285 * have to search to get split stack. 286 */ 287 nbytes = NBLEAFDBT(key->size, data->size); 288 if (h->upper - h->lower < nbytes + sizeof(indx_t)) 289 goto miss; 290 291 if (t->bt_order == FORWARD) { 292 if (t->bt_cur.page->nextpg != P_INVALID) 293 goto miss; 294 if (t->bt_cur.index != NEXTINDEX(h) - 1) 295 goto miss; 296 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 297 goto miss; 298 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 299 } else { 300 if (t->bt_cur.page->prevpg != P_INVALID) 301 goto miss; 302 if (t->bt_cur.index != 0) 303 goto miss; 304 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 305 goto miss; 306 t->bt_last.index = 0; 307 } 308 *exactp = cmp == 0; 309 #ifdef STATISTICS 310 ++bt_cache_hit; 311 #endif 312 return (&t->bt_cur); 313 314 miss: 315 #ifdef STATISTICS 316 ++bt_cache_miss; 317 #endif 318 t->bt_order = NOT; 319 mpool_put(t->bt_mp, h, 0); 320 return (NULL); 321 } 322