1 /* $OpenBSD: bt_put.c,v 1.6 1999/02/15 05:11:23 millert 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 #if defined(LIBC_SCCS) && !defined(lint) 40 #if 0 41 static char sccsid[] = "@(#)bt_put.c 8.8 (Berkeley) 7/26/94"; 42 #else 43 static char rcsid[] = "$OpenBSD: bt_put.c,v 1.6 1999/02/15 05:11:23 millert Exp $"; 44 #endif 45 #endif /* LIBC_SCCS and not lint */ 46 47 #include <sys/types.h> 48 49 #include <errno.h> 50 #include <stdio.h> 51 #include <stdlib.h> 52 #include <string.h> 53 54 #include <db.h> 55 #include "btree.h" 56 57 static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *)); 58 59 /* 60 * __BT_PUT -- Add a btree item to the tree. 61 * 62 * Parameters: 63 * dbp: pointer to access method 64 * key: key 65 * data: data 66 * flag: R_NOOVERWRITE 67 * 68 * Returns: 69 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 70 * tree and R_NOOVERWRITE specified. 71 */ 72 int 73 __bt_put(dbp, key, data, flags) 74 const DB *dbp; 75 DBT *key; 76 const DBT *data; 77 u_int flags; 78 { 79 BTREE *t; 80 DBT tkey, tdata; 81 EPG *e; 82 PAGE *h; 83 indx_t index, nxtindex; 84 pgno_t pg; 85 u_int32_t nbytes; 86 int dflags, exact, status; 87 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 88 89 t = dbp->internal; 90 91 /* Toss any page pinned across calls. */ 92 if (t->bt_pinned != NULL) { 93 mpool_put(t->bt_mp, t->bt_pinned, 0); 94 t->bt_pinned = NULL; 95 } 96 97 /* Check for change to a read-only tree. */ 98 if (F_ISSET(t, B_RDONLY)) { 99 errno = EPERM; 100 return (RET_ERROR); 101 } 102 103 switch (flags) { 104 case 0: 105 case R_NOOVERWRITE: 106 break; 107 case R_CURSOR: 108 /* 109 * If flags is R_CURSOR, put the cursor. Must already 110 * have started a scan and not have already deleted it. 111 */ 112 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 113 !F_ISSET(&t->bt_cursor, 114 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 115 break; 116 /* FALLTHROUGH */ 117 default: 118 errno = EINVAL; 119 return (RET_ERROR); 120 } 121 122 /* 123 * If the key/data pair won't fit on a page, store it on overflow 124 * pages. Only put the key on the overflow page if the pair are 125 * still too big after moving the data to an overflow page. 126 * 127 * XXX 128 * If the insert fails later on, the overflow pages aren't recovered. 129 */ 130 dflags = 0; 131 if (key->size + data->size > t->bt_ovflsize) { 132 if (key->size > t->bt_ovflsize) { 133 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 134 return (RET_ERROR); 135 tkey.data = kb; 136 tkey.size = NOVFLSIZE; 137 memmove(kb, &pg, sizeof(pgno_t)); 138 memmove(kb + sizeof(pgno_t), 139 &key->size, sizeof(u_int32_t)); 140 dflags |= P_BIGKEY; 141 key = &tkey; 142 } 143 if (key->size + data->size > t->bt_ovflsize) { 144 if (__ovfl_put(t, data, &pg) == RET_ERROR) 145 return (RET_ERROR); 146 tdata.data = db; 147 tdata.size = NOVFLSIZE; 148 memmove(db, &pg, sizeof(pgno_t)); 149 memmove(db + sizeof(pgno_t), 150 &data->size, sizeof(u_int32_t)); 151 dflags |= P_BIGDATA; 152 data = &tdata; 153 } 154 if (key->size + data->size > t->bt_ovflsize) 155 goto storekey; 156 } 157 158 /* Replace the cursor. */ 159 if (flags == R_CURSOR) { 160 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 161 return (RET_ERROR); 162 index = t->bt_cursor.pg.index; 163 goto delete; 164 } 165 166 /* 167 * Find the key to delete, or, the location at which to insert. 168 * Bt_fast and __bt_search both pin the returned page. 169 */ 170 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 171 if ((e = __bt_search(t, key, &exact)) == NULL) 172 return (RET_ERROR); 173 h = e->page; 174 index = e->index; 175 176 /* 177 * Add the key/data pair to the tree. If an identical key is already 178 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 179 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 180 * permitted) or an error is returned. 181 */ 182 switch (flags) { 183 case R_NOOVERWRITE: 184 if (!exact) 185 break; 186 mpool_put(t->bt_mp, h, 0); 187 return (RET_SPECIAL); 188 default: 189 if (!exact || !F_ISSET(t, B_NODUPS)) 190 break; 191 /* 192 * !!! 193 * Note, the delete may empty the page, so we need to put a 194 * new entry into the page immediately. 195 */ 196 delete: if (__bt_dleaf(t, key, h, index) == RET_ERROR) { 197 mpool_put(t->bt_mp, h, 0); 198 return (RET_ERROR); 199 } 200 break; 201 } 202 203 /* 204 * If not enough room, or the user has put a ceiling on the number of 205 * keys permitted in the page, split the page. The split code will 206 * insert the key and data and unpin the current page. If inserting 207 * into the offset array, shift the pointers up. 208 */ 209 nbytes = NBLEAFDBT(key->size, data->size); 210 if (h->upper - h->lower < nbytes + sizeof(indx_t)) { 211 if ((status = __bt_split(t, h, key, 212 data, dflags, nbytes, index)) != RET_SUCCESS) 213 return (status); 214 goto success; 215 } 216 217 if (index < (nxtindex = NEXTINDEX(h))) 218 memmove(h->linp + index + 1, h->linp + index, 219 (nxtindex - index) * sizeof(indx_t)); 220 h->lower += sizeof(indx_t); 221 222 h->linp[index] = h->upper -= nbytes; 223 dest = (char *)h + h->upper; 224 WR_BLEAF(dest, key, data, dflags); 225 226 /* If the cursor is on this page, adjust it as necessary. */ 227 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 228 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 229 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= index) 230 ++t->bt_cursor.pg.index; 231 232 if (t->bt_order == NOT) { 233 if (h->nextpg == P_INVALID) { 234 if (index == NEXTINDEX(h) - 1) { 235 t->bt_order = FORWARD; 236 t->bt_last.index = index; 237 t->bt_last.pgno = h->pgno; 238 } 239 } else if (h->prevpg == P_INVALID) { 240 if (index == 0) { 241 t->bt_order = BACK; 242 t->bt_last.index = 0; 243 t->bt_last.pgno = h->pgno; 244 } 245 } 246 } 247 248 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 249 250 success: 251 if (flags == R_SETCURSOR) 252 __bt_setcur(t, e->page->pgno, e->index); 253 254 F_SET(t, B_MODIFIED); 255 return (RET_SUCCESS); 256 } 257 258 #ifdef STATISTICS 259 u_long bt_cache_hit, bt_cache_miss; 260 #endif 261 262 /* 263 * BT_FAST -- Do a quick check for sorted data. 264 * 265 * Parameters: 266 * t: tree 267 * key: key to insert 268 * 269 * Returns: 270 * EPG for new record or NULL if not found. 271 */ 272 static EPG * 273 bt_fast(t, key, data, exactp) 274 BTREE *t; 275 const DBT *key, *data; 276 int *exactp; 277 { 278 PAGE *h; 279 u_int32_t nbytes; 280 int cmp; 281 282 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 283 t->bt_order = NOT; 284 return (NULL); 285 } 286 t->bt_cur.page = h; 287 t->bt_cur.index = t->bt_last.index; 288 289 /* 290 * If won't fit in this page or have too many keys in this page, 291 * have to search to get split stack. 292 */ 293 nbytes = NBLEAFDBT(key->size, data->size); 294 if (h->upper - h->lower < nbytes + sizeof(indx_t)) 295 goto miss; 296 297 if (t->bt_order == FORWARD) { 298 if (t->bt_cur.page->nextpg != P_INVALID) 299 goto miss; 300 if (t->bt_cur.index != NEXTINDEX(h) - 1) 301 goto miss; 302 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 303 goto miss; 304 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 305 } else { 306 if (t->bt_cur.page->prevpg != P_INVALID) 307 goto miss; 308 if (t->bt_cur.index != 0) 309 goto miss; 310 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 311 goto miss; 312 t->bt_last.index = 0; 313 } 314 *exactp = cmp == 0; 315 #ifdef STATISTICS 316 ++bt_cache_hit; 317 #endif 318 return (&t->bt_cur); 319 320 miss: 321 #ifdef STATISTICS 322 ++bt_cache_miss; 323 #endif 324 t->bt_order = NOT; 325 mpool_put(t->bt_mp, h, 0); 326 return (NULL); 327 } 328