1 /* $NetBSD: bt_put.c,v 1.12 1998/12/09 12:42:47 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. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 */ 38 39 #include <sys/cdefs.h> 40 #if defined(LIBC_SCCS) && !defined(lint) 41 #if 0 42 static char sccsid[] = "@(#)bt_put.c 8.8 (Berkeley) 7/26/94"; 43 #else 44 __RCSID("$NetBSD: bt_put.c,v 1.12 1998/12/09 12:42:47 christos Exp $"); 45 #endif 46 #endif /* LIBC_SCCS and not lint */ 47 48 #include "namespace.h" 49 #include <sys/types.h> 50 51 #include <errno.h> 52 #include <stdio.h> 53 #include <stdlib.h> 54 #include <string.h> 55 56 #include <db.h> 57 #include "btree.h" 58 59 static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *)); 60 61 /* 62 * __BT_PUT -- Add a btree item to the tree. 63 * 64 * Parameters: 65 * dbp: pointer to access method 66 * key: key 67 * data: data 68 * flag: R_NOOVERWRITE 69 * 70 * Returns: 71 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 72 * tree and R_NOOVERWRITE specified. 73 */ 74 int 75 __bt_put(dbp, key, data, flags) 76 const DB *dbp; 77 DBT *key; 78 const DBT *data; 79 u_int flags; 80 { 81 BTREE *t; 82 DBT tkey, tdata; 83 EPG *e = NULL; /* pacify gcc */ 84 PAGE *h; 85 indx_t idx, nxtindex; 86 pgno_t pg; 87 u_int32_t nbytes; 88 int dflags, exact, status; 89 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 90 91 t = dbp->internal; 92 93 /* Toss any page pinned across calls. */ 94 if (t->bt_pinned != NULL) { 95 mpool_put(t->bt_mp, t->bt_pinned, 0); 96 t->bt_pinned = NULL; 97 } 98 99 /* Check for change to a read-only tree. */ 100 if (F_ISSET(t, B_RDONLY)) { 101 errno = EPERM; 102 return (RET_ERROR); 103 } 104 105 switch (flags) { 106 case 0: 107 case R_NOOVERWRITE: 108 break; 109 case R_CURSOR: 110 /* 111 * If flags is R_CURSOR, put the cursor. Must already 112 * have started a scan and not have already deleted it. 113 */ 114 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 115 !F_ISSET(&t->bt_cursor, 116 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 117 break; 118 /* FALLTHROUGH */ 119 default: 120 errno = EINVAL; 121 return (RET_ERROR); 122 } 123 124 /* 125 * If the key/data pair won't fit on a page, store it on overflow 126 * pages. Only put the key on the overflow page if the pair are 127 * still too big after moving the data to an overflow page. 128 * 129 * XXX 130 * If the insert fails later on, the overflow pages aren't recovered. 131 */ 132 dflags = 0; 133 if (key->size + data->size > t->bt_ovflsize) { 134 if (key->size > t->bt_ovflsize) { 135 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 136 return (RET_ERROR); 137 tkey.data = kb; 138 tkey.size = NOVFLSIZE; 139 memmove(kb, &pg, sizeof(pgno_t)); 140 memmove(kb + sizeof(pgno_t), 141 &key->size, sizeof(u_int32_t)); 142 dflags |= P_BIGKEY; 143 key = &tkey; 144 } 145 if (key->size + data->size > t->bt_ovflsize) { 146 if (__ovfl_put(t, data, &pg) == RET_ERROR) 147 return (RET_ERROR); 148 tdata.data = db; 149 tdata.size = NOVFLSIZE; 150 memmove(db, &pg, sizeof(pgno_t)); 151 memmove(db + sizeof(pgno_t), 152 &data->size, sizeof(u_int32_t)); 153 dflags |= P_BIGDATA; 154 data = &tdata; 155 } 156 if (key->size + data->size > t->bt_ovflsize) 157 goto storekey; 158 } 159 160 /* Replace the cursor. */ 161 if (flags == R_CURSOR) { 162 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 163 return (RET_ERROR); 164 idx = t->bt_cursor.pg.index; 165 goto delete; 166 } 167 168 /* 169 * Find the key to delete, or, the location at which to insert. 170 * Bt_fast and __bt_search both pin the returned page. 171 */ 172 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 173 if ((e = __bt_search(t, key, &exact)) == NULL) 174 return (RET_ERROR); 175 h = e->page; 176 idx = e->index; 177 178 /* 179 * Add the key/data pair to the tree. If an identical key is already 180 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 181 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 182 * permitted) or an error is returned. 183 */ 184 switch (flags) { 185 case R_NOOVERWRITE: 186 if (!exact) 187 break; 188 mpool_put(t->bt_mp, h, 0); 189 return (RET_SPECIAL); 190 default: 191 if (!exact || !F_ISSET(t, B_NODUPS)) 192 break; 193 /* 194 * !!! 195 * Note, the delete may empty the page, so we need to put a 196 * new entry into the page immediately. 197 */ 198 delete: if (__bt_dleaf(t, key, h, (u_int)idx) == RET_ERROR) { 199 mpool_put(t->bt_mp, h, 0); 200 return (RET_ERROR); 201 } 202 break; 203 } 204 205 /* 206 * If not enough room, or the user has put a ceiling on the number of 207 * keys permitted in the page, split the page. The split code will 208 * insert the key and data and unpin the current page. If inserting 209 * into the offset array, shift the pointers up. 210 */ 211 nbytes = NBLEAFDBT(key->size, data->size); 212 if (h->upper - h->lower < nbytes + sizeof(indx_t)) { 213 if ((status = __bt_split(t, h, key, 214 data, dflags, nbytes, (u_int)idx)) != RET_SUCCESS) 215 return (status); 216 goto success; 217 } 218 219 if (idx < (nxtindex = NEXTINDEX(h))) 220 memmove(h->linp + idx + 1, h->linp + idx, 221 (nxtindex - idx) * sizeof(indx_t)); 222 h->lower += sizeof(indx_t); 223 224 h->linp[idx] = h->upper -= nbytes; 225 dest = (char *)(void *)h + h->upper; 226 WR_BLEAF(dest, key, data, dflags); 227 228 /* If the cursor is on this page, adjust it as necessary. */ 229 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 230 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 231 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx) 232 ++t->bt_cursor.pg.index; 233 234 if (t->bt_order == NOT) { 235 if (h->nextpg == P_INVALID) { 236 if (idx == NEXTINDEX(h) - 1) { 237 t->bt_order = FORWARD; 238 t->bt_last.index = idx; 239 t->bt_last.pgno = h->pgno; 240 } 241 } else if (h->prevpg == P_INVALID) { 242 if (idx == 0) { 243 t->bt_order = BACK; 244 t->bt_last.index = 0; 245 t->bt_last.pgno = h->pgno; 246 } 247 } 248 } 249 250 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 251 252 success: 253 if (flags == R_SETCURSOR) 254 __bt_setcur(t, e->page->pgno, (u_int)e->index); 255 256 F_SET(t, B_MODIFIED); 257 return (RET_SUCCESS); 258 } 259 260 #ifdef STATISTICS 261 u_long bt_cache_hit, bt_cache_miss; 262 #endif 263 264 /* 265 * BT_FAST -- Do a quick check for sorted data. 266 * 267 * Parameters: 268 * t: tree 269 * key: key to insert 270 * 271 * Returns: 272 * EPG for new record or NULL if not found. 273 */ 274 static EPG * 275 bt_fast(t, key, data, exactp) 276 BTREE *t; 277 const DBT *key, *data; 278 int *exactp; 279 { 280 PAGE *h; 281 u_int32_t nbytes; 282 int cmp; 283 284 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 285 t->bt_order = NOT; 286 return (NULL); 287 } 288 t->bt_cur.page = h; 289 t->bt_cur.index = t->bt_last.index; 290 291 /* 292 * If won't fit in this page or have too many keys in this page, 293 * have to search to get split stack. 294 */ 295 nbytes = NBLEAFDBT(key->size, data->size); 296 if (h->upper - h->lower < nbytes + sizeof(indx_t)) 297 goto miss; 298 299 if (t->bt_order == FORWARD) { 300 if (t->bt_cur.page->nextpg != P_INVALID) 301 goto miss; 302 if (t->bt_cur.index != NEXTINDEX(h) - 1) 303 goto miss; 304 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 305 goto miss; 306 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 307 } else { 308 if (t->bt_cur.page->prevpg != P_INVALID) 309 goto miss; 310 if (t->bt_cur.index != 0) 311 goto miss; 312 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 313 goto miss; 314 t->bt_last.index = 0; 315 } 316 *exactp = cmp == 0; 317 #ifdef STATISTICS 318 ++bt_cache_hit; 319 #endif 320 return (&t->bt_cur); 321 322 miss: 323 #ifdef STATISTICS 324 ++bt_cache_miss; 325 #endif 326 t->bt_order = NOT; 327 mpool_put(t->bt_mp, h, 0); 328 return (NULL); 329 } 330