1 /* $NetBSD: bt_put.c,v 1.14 2003/12/30 21:20:16 martin 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 1.14 2003/12/30 21:20:16 martin Exp $"); 41 #endif 42 #endif /* LIBC_SCCS and not lint */ 43 44 #include "namespace.h" 45 #include <sys/types.h> 46 47 #include <errno.h> 48 #include <stdio.h> 49 #include <stdlib.h> 50 #include <string.h> 51 52 #include <db.h> 53 #include "btree.h" 54 55 static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *)); 56 57 /* 58 * __BT_PUT -- Add a btree item to the tree. 59 * 60 * Parameters: 61 * dbp: pointer to access method 62 * key: key 63 * data: data 64 * flag: R_NOOVERWRITE 65 * 66 * Returns: 67 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 68 * tree and R_NOOVERWRITE specified. 69 */ 70 int 71 __bt_put(dbp, key, data, flags) 72 const DB *dbp; 73 DBT *key; 74 const DBT *data; 75 u_int flags; 76 { 77 BTREE *t; 78 DBT tkey, tdata; 79 EPG *e = NULL; /* pacify gcc */ 80 PAGE *h; 81 indx_t idx, nxtindex; 82 pgno_t pg; 83 u_int32_t nbytes, temp; 84 int dflags, exact, status; 85 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 86 87 t = dbp->internal; 88 89 /* Toss any page pinned across calls. */ 90 if (t->bt_pinned != NULL) { 91 mpool_put(t->bt_mp, t->bt_pinned, 0); 92 t->bt_pinned = NULL; 93 } 94 95 /* Check for change to a read-only tree. */ 96 if (F_ISSET(t, B_RDONLY)) { 97 errno = EPERM; 98 return (RET_ERROR); 99 } 100 101 switch (flags) { 102 case 0: 103 case R_NOOVERWRITE: 104 break; 105 case R_CURSOR: 106 /* 107 * If flags is R_CURSOR, put the cursor. Must already 108 * have started a scan and not have already deleted it. 109 */ 110 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 111 !F_ISSET(&t->bt_cursor, 112 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 113 break; 114 /* FALLTHROUGH */ 115 default: 116 errno = EINVAL; 117 return (RET_ERROR); 118 } 119 120 /* 121 * If the key/data pair won't fit on a page, store it on overflow 122 * pages. Only put the key on the overflow page if the pair are 123 * still too big after moving the data to an overflow page. 124 * 125 * XXX 126 * If the insert fails later on, the overflow pages aren't recovered. 127 */ 128 dflags = 0; 129 if (key->size + data->size > t->bt_ovflsize) { 130 if (key->size > t->bt_ovflsize) { 131 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 132 return (RET_ERROR); 133 tkey.data = kb; 134 tkey.size = NOVFLSIZE; 135 memmove(kb, &pg, sizeof(pgno_t)); 136 memmove(kb + sizeof(pgno_t), 137 &key->size, sizeof(u_int32_t)); 138 dflags |= P_BIGKEY; 139 key = &tkey; 140 } 141 if (key->size + data->size > t->bt_ovflsize) { 142 if (__ovfl_put(t, data, &pg) == RET_ERROR) 143 return (RET_ERROR); 144 tdata.data = db; 145 tdata.size = NOVFLSIZE; 146 memmove(db, &pg, sizeof(pgno_t)); 147 temp = data->size; 148 memmove(db + sizeof(pgno_t), 149 &temp, sizeof(u_int32_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 (h->upper - 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 u_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(t, key, data, exactp) 273 BTREE *t; 274 const DBT *key, *data; 275 int *exactp; 276 { 277 PAGE *h; 278 u_int32_t nbytes; 279 int cmp; 280 281 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 282 t->bt_order = NOT; 283 return (NULL); 284 } 285 t->bt_cur.page = h; 286 t->bt_cur.index = t->bt_last.index; 287 288 /* 289 * If won't fit in this page or have too many keys in this page, 290 * have to search to get split stack. 291 */ 292 nbytes = NBLEAFDBT(key->size, data->size); 293 if (h->upper - h->lower < nbytes + sizeof(indx_t)) 294 goto miss; 295 296 if (t->bt_order == FORWARD) { 297 if (t->bt_cur.page->nextpg != P_INVALID) 298 goto miss; 299 if (t->bt_cur.index != NEXTINDEX(h) - 1) 300 goto miss; 301 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 302 goto miss; 303 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 304 } else { 305 if (t->bt_cur.page->prevpg != P_INVALID) 306 goto miss; 307 if (t->bt_cur.index != 0) 308 goto miss; 309 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 310 goto miss; 311 t->bt_last.index = 0; 312 } 313 *exactp = cmp == 0; 314 #ifdef STATISTICS 315 ++bt_cache_hit; 316 #endif 317 return (&t->bt_cur); 318 319 miss: 320 #ifdef STATISTICS 321 ++bt_cache_miss; 322 #endif 323 t->bt_order = NOT; 324 mpool_put(t->bt_mp, h, 0); 325 return (NULL); 326 } 327