1 /*- 2 * Copyright (c) 1990, 1993, 1994 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Mike Olson. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. All advertising materials mentioning features or use of this software 17 * must display the following acknowledgement: 18 * This product includes software developed by the University of 19 * California, Berkeley and its contributors. 20 * 4. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 */ 36 37 #if defined(LIBC_SCCS) && !defined(lint) 38 static char sccsid[] = "@(#)bt_put.c 8.8 (Berkeley) 7/26/94"; 39 #endif /* LIBC_SCCS and not lint */ 40 #include <sys/cdefs.h> 41 __FBSDID("$FreeBSD$"); 42 43 #include <sys/types.h> 44 45 #include <errno.h> 46 #include <stdio.h> 47 #include <stdlib.h> 48 #include <string.h> 49 50 #include <db.h> 51 #include "btree.h" 52 53 static EPG *bt_fast(BTREE *, const DBT *, const DBT *, int *); 54 55 /* 56 * __BT_PUT -- Add a btree item to the tree. 57 * 58 * Parameters: 59 * dbp: pointer to access method 60 * key: key 61 * data: data 62 * flag: R_NOOVERWRITE 63 * 64 * Returns: 65 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 66 * tree and R_NOOVERWRITE specified. 67 */ 68 int 69 __bt_put(dbp, key, data, flags) 70 const DB *dbp; 71 DBT *key; 72 const DBT *data; 73 u_int flags; 74 { 75 BTREE *t; 76 DBT tkey, tdata; 77 EPG *e; 78 PAGE *h; 79 indx_t index, nxtindex; 80 pgno_t pg; 81 u_int32_t nbytes; 82 int dflags, exact, status; 83 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 84 85 t = dbp->internal; 86 87 /* Toss any page pinned across calls. */ 88 if (t->bt_pinned != NULL) { 89 mpool_put(t->bt_mp, t->bt_pinned, 0); 90 t->bt_pinned = NULL; 91 } 92 93 /* Check for change to a read-only tree. */ 94 if (F_ISSET(t, B_RDONLY)) { 95 errno = EPERM; 96 return (RET_ERROR); 97 } 98 99 switch (flags) { 100 case 0: 101 case R_NOOVERWRITE: 102 break; 103 case R_CURSOR: 104 /* 105 * If flags is R_CURSOR, put the cursor. Must already 106 * have started a scan and not have already deleted it. 107 */ 108 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 109 !F_ISSET(&t->bt_cursor, 110 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE)) 111 break; 112 /* FALLTHROUGH */ 113 default: 114 errno = EINVAL; 115 return (RET_ERROR); 116 } 117 118 /* 119 * If the key/data pair won't fit on a page, store it on overflow 120 * pages. Only put the key on the overflow page if the pair are 121 * still too big after moving the data to an overflow page. 122 * 123 * XXX 124 * If the insert fails later on, the overflow pages aren't recovered. 125 */ 126 dflags = 0; 127 if (key->size + data->size > t->bt_ovflsize) { 128 if (key->size > t->bt_ovflsize) { 129 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 130 return (RET_ERROR); 131 tkey.data = kb; 132 tkey.size = NOVFLSIZE; 133 memmove(kb, &pg, sizeof(pgno_t)); 134 memmove(kb + sizeof(pgno_t), 135 &key->size, sizeof(u_int32_t)); 136 dflags |= P_BIGKEY; 137 key = &tkey; 138 } 139 if (key->size + data->size > t->bt_ovflsize) { 140 if (__ovfl_put(t, data, &pg) == RET_ERROR) 141 return (RET_ERROR); 142 tdata.data = db; 143 tdata.size = NOVFLSIZE; 144 memmove(db, &pg, sizeof(pgno_t)); 145 memmove(db + sizeof(pgno_t), 146 &data->size, sizeof(u_int32_t)); 147 dflags |= P_BIGDATA; 148 data = &tdata; 149 } 150 if (key->size + data->size > t->bt_ovflsize) 151 goto storekey; 152 } 153 154 /* Replace the cursor. */ 155 if (flags == R_CURSOR) { 156 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL) 157 return (RET_ERROR); 158 index = t->bt_cursor.pg.index; 159 goto delete; 160 } 161 162 /* 163 * Find the key to delete, or, the location at which to insert. 164 * Bt_fast and __bt_search both pin the returned page. 165 */ 166 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 167 if ((e = __bt_search(t, key, &exact)) == NULL) 168 return (RET_ERROR); 169 h = e->page; 170 index = e->index; 171 172 /* 173 * Add the key/data pair to the tree. If an identical key is already 174 * in the tree, and R_NOOVERWRITE is set, an error is returned. If 175 * R_NOOVERWRITE is not set, the key is either added (if duplicates are 176 * permitted) or an error is returned. 177 */ 178 switch (flags) { 179 case R_NOOVERWRITE: 180 if (!exact) 181 break; 182 mpool_put(t->bt_mp, h, 0); 183 return (RET_SPECIAL); 184 default: 185 if (!exact || !F_ISSET(t, B_NODUPS)) 186 break; 187 /* 188 * !!! 189 * Note, the delete may empty the page, so we need to put a 190 * new entry into the page immediately. 191 */ 192 delete: if (__bt_dleaf(t, key, h, index) == RET_ERROR) { 193 mpool_put(t->bt_mp, h, 0); 194 return (RET_ERROR); 195 } 196 break; 197 } 198 199 /* 200 * If not enough room, or the user has put a ceiling on the number of 201 * keys permitted in the page, split the page. The split code will 202 * insert the key and data and unpin the current page. If inserting 203 * into the offset array, shift the pointers up. 204 */ 205 nbytes = NBLEAFDBT(key->size, data->size); 206 if (h->upper - h->lower < nbytes + sizeof(indx_t)) { 207 if ((status = __bt_split(t, h, key, 208 data, dflags, nbytes, index)) != RET_SUCCESS) 209 return (status); 210 goto success; 211 } 212 213 if (index < (nxtindex = NEXTINDEX(h))) 214 memmove(h->linp + index + 1, h->linp + index, 215 (nxtindex - index) * sizeof(indx_t)); 216 h->lower += sizeof(indx_t); 217 218 h->linp[index] = h->upper -= nbytes; 219 dest = (char *)h + h->upper; 220 WR_BLEAF(dest, key, data, dflags); 221 222 /* If the cursor is on this page, adjust it as necessary. */ 223 if (F_ISSET(&t->bt_cursor, CURS_INIT) && 224 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) && 225 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= index) 226 ++t->bt_cursor.pg.index; 227 228 if (t->bt_order == NOT) 229 if (h->nextpg == P_INVALID) { 230 if (index == NEXTINDEX(h) - 1) { 231 t->bt_order = FORWARD; 232 t->bt_last.index = index; 233 t->bt_last.pgno = h->pgno; 234 } 235 } else if (h->prevpg == P_INVALID) { 236 if (index == 0) { 237 t->bt_order = BACK; 238 t->bt_last.index = 0; 239 t->bt_last.pgno = h->pgno; 240 } 241 } 242 243 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 244 245 success: 246 if (flags == R_SETCURSOR) 247 __bt_setcur(t, e->page->pgno, e->index); 248 249 F_SET(t, B_MODIFIED); 250 return (RET_SUCCESS); 251 } 252 253 #ifdef STATISTICS 254 u_long bt_cache_hit, bt_cache_miss; 255 #endif 256 257 /* 258 * BT_FAST -- Do a quick check for sorted data. 259 * 260 * Parameters: 261 * t: tree 262 * key: key to insert 263 * 264 * Returns: 265 * EPG for new record or NULL if not found. 266 */ 267 static EPG * 268 bt_fast(t, key, data, exactp) 269 BTREE *t; 270 const DBT *key, *data; 271 int *exactp; 272 { 273 PAGE *h; 274 u_int32_t nbytes; 275 int cmp; 276 277 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 278 t->bt_order = NOT; 279 return (NULL); 280 } 281 t->bt_cur.page = h; 282 t->bt_cur.index = t->bt_last.index; 283 284 /* 285 * If won't fit in this page or have too many keys in this page, 286 * have to search to get split stack. 287 */ 288 nbytes = NBLEAFDBT(key->size, data->size); 289 if (h->upper - h->lower < nbytes + sizeof(indx_t)) 290 goto miss; 291 292 if (t->bt_order == FORWARD) { 293 if (t->bt_cur.page->nextpg != P_INVALID) 294 goto miss; 295 if (t->bt_cur.index != NEXTINDEX(h) - 1) 296 goto miss; 297 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0) 298 goto miss; 299 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index; 300 } else { 301 if (t->bt_cur.page->prevpg != P_INVALID) 302 goto miss; 303 if (t->bt_cur.index != 0) 304 goto miss; 305 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0) 306 goto miss; 307 t->bt_last.index = 0; 308 } 309 *exactp = cmp == 0; 310 #ifdef STATISTICS 311 ++bt_cache_hit; 312 #endif 313 return (&t->bt_cur); 314 315 miss: 316 #ifdef STATISTICS 317 ++bt_cache_miss; 318 #endif 319 t->bt_order = NOT; 320 mpool_put(t->bt_mp, h, 0); 321 return (NULL); 322 } 323