1 /*- 2 * Copyright (c) 1990, 1993 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 8.1 (Berkeley) 6/4/93"; 39 #endif /* LIBC_SCCS and not lint */ 40 41 #include <sys/types.h> 42 43 #include <errno.h> 44 #include <stdio.h> 45 #include <stdlib.h> 46 #include <string.h> 47 48 #include <db.h> 49 #include "btree.h" 50 51 static EPG *bt_fast __P((BTREE *, const DBT *, const DBT *, int *)); 52 53 /* 54 * __BT_PUT -- Add a btree item to the tree. 55 * 56 * Parameters: 57 * dbp: pointer to access method 58 * key: key 59 * data: data 60 * flag: R_NOOVERWRITE 61 * 62 * Returns: 63 * RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the 64 * tree and R_NOOVERWRITE specified. 65 */ 66 int 67 __bt_put(dbp, key, data, flags) 68 const DB *dbp; 69 DBT *key; 70 const DBT *data; 71 u_int flags; 72 { 73 BTREE *t; 74 DBT tkey, tdata; 75 EPG *e; 76 PAGE *h; 77 indx_t index, nxtindex; 78 pgno_t pg; 79 size_t nbytes; 80 int dflags, exact, status; 81 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE]; 82 83 t = dbp->internal; 84 85 switch (flags) { 86 case R_CURSOR: 87 if (!ISSET(t, B_SEQINIT)) 88 goto einval; 89 if (ISSET(t, B_DELCRSR)) 90 goto einval; 91 break; 92 case 0: 93 case R_NOOVERWRITE: 94 break; 95 default: 96 einval: errno = EINVAL; 97 return (RET_ERROR); 98 } 99 100 if (ISSET(t, B_RDONLY)) { 101 errno = EPERM; 102 return (RET_ERROR); 103 } 104 105 /* 106 * If the key/data won't fit on a page, store it on indirect pages. 107 * Only store the key on the overflow page if it's too big after the 108 * data is on an overflow page. 109 * 110 * XXX 111 * If the insert fails later on, these pages aren't recovered. 112 */ 113 dflags = 0; 114 if (key->size + data->size > t->bt_ovflsize) { 115 if (key->size > t->bt_ovflsize) { 116 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR) 117 return (RET_ERROR); 118 tkey.data = kb; 119 tkey.size = NOVFLSIZE; 120 memmove(kb, &pg, sizeof(pgno_t)); 121 memmove(kb + sizeof(pgno_t), 122 &key->size, sizeof(size_t)); 123 dflags |= P_BIGKEY; 124 key = &tkey; 125 } 126 if (key->size + data->size > t->bt_ovflsize) { 127 if (__ovfl_put(t, data, &pg) == RET_ERROR) 128 return (RET_ERROR); 129 tdata.data = db; 130 tdata.size = NOVFLSIZE; 131 memmove(db, &pg, sizeof(pgno_t)); 132 memmove(db + sizeof(pgno_t), 133 &data->size, sizeof(size_t)); 134 dflags |= P_BIGDATA; 135 data = &tdata; 136 } 137 if (key->size + data->size > t->bt_ovflsize) 138 goto storekey; 139 } 140 141 /* Replace the cursor. */ 142 if (flags == R_CURSOR) { 143 if ((h = mpool_get(t->bt_mp, t->bt_bcursor.pgno, 0)) == NULL) 144 return (RET_ERROR); 145 index = t->bt_bcursor.index; 146 goto delete; 147 } 148 149 /* 150 * Find the key to delete, or, the location at which to insert. Bt_fast 151 * and __bt_search pin the returned page. 152 */ 153 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL) 154 if ((e = __bt_search(t, key, &exact)) == NULL) 155 return (RET_ERROR); 156 h = e->page; 157 index = e->index; 158 159 /* 160 * Add the specified key/data pair to the tree. If an identical key 161 * is already in the tree, and R_NOOVERWRITE is set, an error is 162 * returned. If R_NOOVERWRITE is not set, the key is either added (if 163 * duplicates are permitted) or an error is returned. 164 * 165 * Pages are split as required. 166 */ 167 switch (flags) { 168 case R_NOOVERWRITE: 169 if (!exact) 170 break; 171 /* 172 * One special case is if the cursor references the record and 173 * it's been flagged for deletion. Then, we delete the record, 174 * leaving the cursor there -- this means that the inserted 175 * record will not be seen in a cursor scan. 176 */ 177 if (ISSET(t, B_DELCRSR) && t->bt_bcursor.pgno == h->pgno && 178 t->bt_bcursor.index == index) { 179 CLR(t, B_DELCRSR); 180 goto delete; 181 } 182 mpool_put(t->bt_mp, h, 0); 183 return (RET_SPECIAL); 184 default: 185 if (!exact || !ISSET(t, B_NODUPS)) 186 break; 187 delete: if (__bt_dleaf(t, h, index) == RET_ERROR) { 188 mpool_put(t->bt_mp, h, 0); 189 return (RET_ERROR); 190 } 191 break; 192 } 193 194 /* 195 * If not enough room, or the user has put a ceiling on the number of 196 * keys permitted in the page, split the page. The split code will 197 * insert the key and data and unpin the current page. If inserting 198 * into the offset array, shift the pointers up. 199 */ 200 nbytes = NBLEAFDBT(key->size, data->size); 201 if (h->upper - h->lower < nbytes + sizeof(indx_t)) { 202 if ((status = __bt_split(t, h, key, 203 data, dflags, nbytes, index)) != RET_SUCCESS) 204 return (status); 205 goto success; 206 } 207 208 if (index < (nxtindex = NEXTINDEX(h))) 209 memmove(h->linp + index + 1, h->linp + index, 210 (nxtindex - index) * sizeof(indx_t)); 211 h->lower += sizeof(indx_t); 212 213 h->linp[index] = h->upper -= nbytes; 214 dest = (char *)h + h->upper; 215 WR_BLEAF(dest, key, data, dflags); 216 217 if (t->bt_order == NOT) 218 if (h->nextpg == P_INVALID) { 219 if (index == NEXTINDEX(h) - 1) { 220 t->bt_order = FORWARD; 221 t->bt_last.index = index; 222 t->bt_last.pgno = h->pgno; 223 } 224 } else if (h->prevpg == P_INVALID) { 225 if (index == 0) { 226 t->bt_order = BACK; 227 t->bt_last.index = 0; 228 t->bt_last.pgno = h->pgno; 229 } 230 } 231 232 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 233 234 success: 235 if (flags == R_SETCURSOR) { 236 t->bt_bcursor.pgno = e->page->pgno; 237 t->bt_bcursor.index = e->index; 238 } 239 SET(t, B_MODIFIED); 240 return (RET_SUCCESS); 241 } 242 243 #ifdef STATISTICS 244 u_long bt_cache_hit, bt_cache_miss; 245 #endif 246 247 /* 248 * BT_FAST -- Do a quick check for sorted data. 249 * 250 * Parameters: 251 * t: tree 252 * key: key to insert 253 * 254 * Returns: 255 * EPG for new record or NULL if not found. 256 */ 257 static EPG * 258 bt_fast(t, key, data, exactp) 259 BTREE *t; 260 const DBT *key, *data; 261 int *exactp; 262 { 263 EPG e; 264 PAGE *h; 265 size_t nbytes; 266 int cmp; 267 268 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) { 269 t->bt_order = NOT; 270 return (NULL); 271 } 272 e.page = h; 273 e.index = t->bt_last.index; 274 275 /* 276 * If won't fit in this page or have too many keys in this page, have 277 * to search to get split stack. 278 */ 279 nbytes = NBLEAFDBT(key->size, data->size); 280 if (h->upper - h->lower < nbytes + sizeof(indx_t)) 281 goto miss; 282 283 if (t->bt_order == FORWARD) { 284 if (e.page->nextpg != P_INVALID) 285 goto miss; 286 if (e.index != NEXTINDEX(h) - 1) 287 goto miss; 288 if ((cmp = __bt_cmp(t, key, &e)) < 0) 289 goto miss; 290 t->bt_last.index = cmp ? ++e.index : e.index; 291 } else { 292 if (e.page->prevpg != P_INVALID) 293 goto miss; 294 if (e.index != 0) 295 goto miss; 296 if ((cmp = __bt_cmp(t, key, &e)) > 0) 297 goto miss; 298 t->bt_last.index = 0; 299 } 300 *exactp = cmp == 0; 301 #ifdef STATISTICS 302 ++bt_cache_hit; 303 #endif 304 return (&e); 305 306 miss: 307 #ifdef STATISTICS 308 ++bt_cache_miss; 309 #endif 310 t->bt_order = NOT; 311 mpool_put(t->bt_mp, h, 0); 312 return (NULL); 313 } 314