1 /* $NetBSD: bt_put.c,v 1.19 2009/02/12 06:40:14 lukem 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 #if HAVE_NBTOOL_CONFIG_H 36 #include "nbtool_config.h" 37 #endif 38 39 #include <sys/cdefs.h> 40 #ifndef __minix 41 __RCSID("$NetBSD: bt_put.c,v 1.19 2009/02/12 06:40:14 lukem Exp $"); 42 #endif 43 44 #ifndef __minix 45 #include "namespace.h" 46 #endif 47 #include <sys/types.h> 48 49 #include <assert.h> 50 #include <errno.h> 51 #include <stdio.h> 52 #include <stdlib.h> 53 #include <string.h> 54 55 #include <db.h> 56 #include "btree.h" 57 58 static EPG *bt_fast(BTREE *, const DBT *, const DBT *, int *); 59 60 /* 61 * __BT_PUT -- Add a btree item to the tree. 62 * 63 * Parameters: 64 * dbp: pointer to access method 65 * key: key 66 * data: data 67 * flag: R_NOOVERWRITE 68 * 69 * Returns: 70 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 71 * tree and R_NOOVERWRITE specified. 72 */ 73 int 74 __bt_put(const DB *dbp, DBT *key, const DBT *data, u_int flags) 75 { 76 BTREE *t; 77 DBT tkey, tdata; 78 EPG *e = NULL; /* pacify gcc */ 79 PAGE *h; 80 indx_t idx, nxtindex; 81 pgno_t pg; 82 uint32_t nbytes, temp; 83 int dflags, exact, status; 84 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 85 86 t = dbp->internal; 87 88 /* Toss any page pinned across calls. */ 89 if (t->bt_pinned != NULL) { 90 mpool_put(t->bt_mp, t->bt_pinned, 0); 91 t->bt_pinned = NULL; 92 } 93 94 /* Check for change to a read-only tree. */ 95 if (F_ISSET(t, B_RDONLY)) { 96 errno = EPERM; 97 return (RET_ERROR); 98 } 99 100 switch (flags) { 101 case 0: 102 case R_NOOVERWRITE: 103 break; 104 case R_CURSOR: 105 /* 106 * If flags is R_CURSOR, put the cursor. Must already 107 * have started a scan and not have already deleted it. 108 */ 109 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 110 !F_ISSET(&t->bt_cursor, 111 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 112 break; 113 /* FALLTHROUGH */ 114 default: 115 errno = EINVAL; 116 return (RET_ERROR); 117 } 118 119 /* 120 * If the key/data pair won't fit on a page, store it on overflow 121 * pages. Only put the key on the overflow page if the pair are 122 * still too big after moving the data to an overflow page. 123 * 124 * XXX 125 * If the insert fails later on, the overflow pages aren't recovered. 126 */ 127 dflags = 0; 128 if (key->size + data->size > t->bt_ovflsize) { 129 if (key->size > t->bt_ovflsize) { 130 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 131 return (RET_ERROR); 132 tkey.data = kb; 133 tkey.size = NOVFLSIZE; 134 memmove(kb, &pg, sizeof(pgno_t)); 135 memmove(kb + sizeof(pgno_t), 136 &key->size, sizeof(uint32_t)); 137 dflags |= P_BIGKEY; 138 key = &tkey; 139 } 140 if (key->size + data->size > t->bt_ovflsize) { 141 if (__ovfl_put(t, data, &pg) == RET_ERROR) 142 return (RET_ERROR); 143 tdata.data = db; 144 tdata.size = NOVFLSIZE; 145 memmove(db, &pg, sizeof(pgno_t)); 146 _DBFIT(data->size, uint32_t); 147 temp = (uint32_t)data->size; 148 (void)memmove(db + sizeof(pgno_t), 149 &temp, sizeof(uint32_t)); 150 dflags |= P_BIGDATA; 151 data = &tdata; 152 } 153 if (key->size + data->size > t->bt_ovflsize) 154 goto storekey; 155 } 156 157 /* Replace the cursor. */ 158 if (flags == R_CURSOR) { 159 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 160 return (RET_ERROR); 161 idx = t->bt_cursor.pg.index; 162 goto delete; 163 } 164 165 /* 166 * Find the key to delete, or, the location at which to insert. 167 * Bt_fast and __bt_search both pin the returned page. 168 */ 169 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 170 if ((e = __bt_search(t, key, &exact)) == NULL) 171 return (RET_ERROR); 172 h = e->page; 173 idx = e->index; 174 175 /* 176 * Add the key/data pair to the tree. If an identical key is already 177 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 178 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 179 * permitted) or an error is returned. 180 */ 181 switch (flags) { 182 case R_NOOVERWRITE: 183 if (!exact) 184 break; 185 mpool_put(t->bt_mp, h, 0); 186 return (RET_SPECIAL); 187 default: 188 if (!exact || !F_ISSET(t, B_NODUPS)) 189 break; 190 /* 191 * !!! 192 * Note, the delete may empty the page, so we need to put a 193 * new entry into the page immediately. 194 */ 195 delete: if (__bt_dleaf(t, key, h, (u_int)idx) == RET_ERROR) { 196 mpool_put(t->bt_mp, h, 0); 197 return (RET_ERROR); 198 } 199 break; 200 } 201 202 /* 203 * If not enough room, or the user has put a ceiling on the number of 204 * keys permitted in the page, split the page. The split code will 205 * insert the key and data and unpin the current page. If inserting 206 * into the offset array, shift the pointers up. 207 */ 208 nbytes = NBLEAFDBT(key->size, data->size); 209 if ((uint32_t)h->upper - (uint32_t)h->lower < nbytes + sizeof(indx_t)) { 210 if ((status = __bt_split(t, h, key, 211 data, dflags, nbytes, (u_int)idx)) != RET_SUCCESS) 212 return (status); 213 goto success; 214 } 215 216 if (idx < (nxtindex = NEXTINDEX(h))) 217 memmove(h->linp + idx + 1, h->linp + idx, 218 (nxtindex - idx) * sizeof(indx_t)); 219 h->lower += sizeof(indx_t); 220 221 h->linp[idx] = h->upper -= nbytes; 222 dest = (char *)(void *)h + h->upper; 223 WR_BLEAF(dest, key, data, dflags); 224 225 /* If the cursor is on this page, adjust it as necessary. */ 226 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 227 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 228 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx) 229 ++t->bt_cursor.pg.index; 230 231 if (t->bt_order == NOT) { 232 if (h->nextpg == P_INVALID) { 233 if (idx == NEXTINDEX(h) - 1) { 234 t->bt_order = FORWARD; 235 t->bt_last.index = idx; 236 t->bt_last.pgno = h->pgno; 237 } 238 } else if (h->prevpg == P_INVALID) { 239 if (idx == 0) { 240 t->bt_order = BACK; 241 t->bt_last.index = 0; 242 t->bt_last.pgno = h->pgno; 243 } 244 } 245 } 246 247 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 248 249 success: 250 if (flags == R_SETCURSOR) 251 __bt_setcur(t, e->page->pgno, (u_int)e->index); 252 253 F_SET(t, B_MODIFIED); 254 return (RET_SUCCESS); 255 } 256 257 #ifdef STATISTICS 258 unsigned long bt_cache_hit, bt_cache_miss; 259 #endif 260 261 /* 262 * BT_FAST -- Do a quick check for sorted data. 263 * 264 * Parameters: 265 * t: tree 266 * key: key to insert 267 * 268 * Returns: 269 * EPG for new record or NULL if not found. 270 */ 271 static EPG * 272 bt_fast(BTREE *t, const DBT *key, const DBT *data, int *exactp) 273 { 274 PAGE *h; 275 uint32_t nbytes; 276 int cmp; 277 278 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 279 t->bt_order = NOT; 280 return (NULL); 281 } 282 t->bt_cur.page = h; 283 t->bt_cur.index = t->bt_last.index; 284 285 /* 286 * If won't fit in this page or have too many keys in this page, 287 * have to search to get split stack. 288 */ 289 nbytes = NBLEAFDBT(key->size, data->size); 290 if ((uint32_t)h->upper - (uint32_t)h->lower < nbytes + sizeof(indx_t)) 291 goto miss; 292 293 if (t->bt_order == FORWARD) { 294 if (t->bt_cur.page->nextpg != P_INVALID) 295 goto miss; 296 if (t->bt_cur.index != NEXTINDEX(h) - 1) 297 goto miss; 298 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 299 goto miss; 300 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 301 } else { 302 if (t->bt_cur.page->prevpg != P_INVALID) 303 goto miss; 304 if (t->bt_cur.index != 0) 305 goto miss; 306 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 307 goto miss; 308 t->bt_last.index = 0; 309 } 310 *exactp = cmp == 0; 311 #ifdef STATISTICS 312 ++bt_cache_hit; 313 #endif 314 return (&t->bt_cur); 315 316 miss: 317 #ifdef STATISTICS 318 ++bt_cache_miss; 319 #endif 320 t->bt_order = NOT; 321 mpool_put(t->bt_mp, h, 0); 322 return (NULL); 323 } 324