1*f14fb602SLionel Sambuc /* $NetBSD: bt_overflow.c,v 1.18 2012/03/13 21:13:32 christos Exp $ */ 22639ae9bSBen Gras 32639ae9bSBen Gras /*- 42639ae9bSBen Gras * Copyright (c) 1990, 1993, 1994 52639ae9bSBen Gras * The Regents of the University of California. All rights reserved. 62639ae9bSBen Gras * 72639ae9bSBen Gras * This code is derived from software contributed to Berkeley by 82639ae9bSBen Gras * Mike Olson. 92639ae9bSBen Gras * 102639ae9bSBen Gras * Redistribution and use in source and binary forms, with or without 112639ae9bSBen Gras * modification, are permitted provided that the following conditions 122639ae9bSBen Gras * are met: 132639ae9bSBen Gras * 1. Redistributions of source code must retain the above copyright 142639ae9bSBen Gras * notice, this list of conditions and the following disclaimer. 152639ae9bSBen Gras * 2. Redistributions in binary form must reproduce the above copyright 162639ae9bSBen Gras * notice, this list of conditions and the following disclaimer in the 172639ae9bSBen Gras * documentation and/or other materials provided with the distribution. 182639ae9bSBen Gras * 3. Neither the name of the University nor the names of its contributors 192639ae9bSBen Gras * may be used to endorse or promote products derived from this software 202639ae9bSBen Gras * without specific prior written permission. 212639ae9bSBen Gras * 222639ae9bSBen Gras * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 232639ae9bSBen Gras * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 242639ae9bSBen Gras * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 252639ae9bSBen Gras * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 262639ae9bSBen Gras * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 272639ae9bSBen Gras * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 282639ae9bSBen Gras * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 292639ae9bSBen Gras * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 302639ae9bSBen Gras * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 312639ae9bSBen Gras * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 322639ae9bSBen Gras * SUCH DAMAGE. 332639ae9bSBen Gras */ 342639ae9bSBen Gras 352639ae9bSBen Gras #if HAVE_NBTOOL_CONFIG_H 362639ae9bSBen Gras #include "nbtool_config.h" 372639ae9bSBen Gras #endif 382639ae9bSBen Gras 392639ae9bSBen Gras #include <sys/cdefs.h> 40*f14fb602SLionel Sambuc __RCSID("$NetBSD: bt_overflow.c,v 1.18 2012/03/13 21:13:32 christos Exp $"); 412639ae9bSBen Gras 422639ae9bSBen Gras #include "namespace.h" 432639ae9bSBen Gras #include <sys/param.h> 442639ae9bSBen Gras 452639ae9bSBen Gras #include <assert.h> 462639ae9bSBen Gras #include <stdio.h> 472639ae9bSBen Gras #include <stdlib.h> 482639ae9bSBen Gras #include <string.h> 492639ae9bSBen Gras 502639ae9bSBen Gras #include <db.h> 512639ae9bSBen Gras #include "btree.h" 522639ae9bSBen Gras 532639ae9bSBen Gras /* 542639ae9bSBen Gras * Big key/data code. 552639ae9bSBen Gras * 562639ae9bSBen Gras * Big key and data entries are stored on linked lists of pages. The initial 572639ae9bSBen Gras * reference is byte string stored with the key or data and is the page number 582639ae9bSBen Gras * and size. The actual record is stored in a chain of pages linked by the 592639ae9bSBen Gras * nextpg field of the PAGE header. 602639ae9bSBen Gras * 612639ae9bSBen Gras * The first page of the chain has a special property. If the record is used 622639ae9bSBen Gras * by an internal page, it cannot be deleted and the P_PRESERVE bit will be set 632639ae9bSBen Gras * in the header. 642639ae9bSBen Gras * 652639ae9bSBen Gras * XXX 662639ae9bSBen Gras * A single DBT is written to each chain, so a lot of space on the last page 672639ae9bSBen Gras * is wasted. This is a fairly major bug for some data sets. 682639ae9bSBen Gras */ 692639ae9bSBen Gras 702639ae9bSBen Gras /* 712639ae9bSBen Gras * __OVFL_GET -- Get an overflow key/data item. 722639ae9bSBen Gras * 732639ae9bSBen Gras * Parameters: 742639ae9bSBen Gras * t: tree 752639ae9bSBen Gras * p: pointer to { pgno_t, uint32_t } 762639ae9bSBen Gras * buf: storage address 772639ae9bSBen Gras * bufsz: storage size 782639ae9bSBen Gras * 792639ae9bSBen Gras * Returns: 802639ae9bSBen Gras * RET_ERROR, RET_SUCCESS 812639ae9bSBen Gras */ 822639ae9bSBen Gras int 832639ae9bSBen Gras __ovfl_get(BTREE *t, void *p, size_t *ssz, void **buf, size_t *bufsz) 842639ae9bSBen Gras { 852639ae9bSBen Gras PAGE *h; 862639ae9bSBen Gras pgno_t pg; 872639ae9bSBen Gras uint32_t sz, nb, plen; 882639ae9bSBen Gras size_t temp; 892639ae9bSBen Gras 90*f14fb602SLionel Sambuc memmove(&pg, p, sizeof(pg)); 912639ae9bSBen Gras memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(uint32_t)); 922639ae9bSBen Gras *ssz = sz; 932639ae9bSBen Gras 942639ae9bSBen Gras #ifdef DEBUG 952639ae9bSBen Gras if (pg == P_INVALID || sz == 0) 962639ae9bSBen Gras abort(); 972639ae9bSBen Gras #endif 982639ae9bSBen Gras /* Make the buffer bigger as necessary. */ 992639ae9bSBen Gras if (*bufsz < sz) { 100*f14fb602SLionel Sambuc *buf = (*buf == NULL ? malloc(sz) : realloc(*buf, sz)); 1012639ae9bSBen Gras if (*buf == NULL) 1022639ae9bSBen Gras return (RET_ERROR); 1032639ae9bSBen Gras *bufsz = sz; 1042639ae9bSBen Gras } 1052639ae9bSBen Gras 1062639ae9bSBen Gras /* 1072639ae9bSBen Gras * Step through the linked list of pages, copying the data on each one 1082639ae9bSBen Gras * into the buffer. Never copy more than the data's length. 1092639ae9bSBen Gras */ 1102639ae9bSBen Gras temp = t->bt_psize - BTDATAOFF; 1112639ae9bSBen Gras _DBFIT(temp, uint32_t); 1122639ae9bSBen Gras plen = (uint32_t)temp; 1132639ae9bSBen Gras for (p = *buf;; p = (char *)p + nb, pg = h->nextpg) { 1142639ae9bSBen Gras if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL) 1152639ae9bSBen Gras return (RET_ERROR); 1162639ae9bSBen Gras 1172639ae9bSBen Gras nb = MIN(sz, plen); 1182639ae9bSBen Gras memmove(p, (char *)(void *)h + BTDATAOFF, nb); 1192639ae9bSBen Gras mpool_put(t->bt_mp, h, 0); 1202639ae9bSBen Gras 1212639ae9bSBen Gras if ((sz -= nb) == 0) 1222639ae9bSBen Gras break; 1232639ae9bSBen Gras } 1242639ae9bSBen Gras return (RET_SUCCESS); 1252639ae9bSBen Gras } 1262639ae9bSBen Gras 1272639ae9bSBen Gras /* 1282639ae9bSBen Gras * __OVFL_PUT -- Store an overflow key/data item. 1292639ae9bSBen Gras * 1302639ae9bSBen Gras * Parameters: 1312639ae9bSBen Gras * t: tree 1322639ae9bSBen Gras * data: DBT to store 1332639ae9bSBen Gras * pgno: storage page number 1342639ae9bSBen Gras * 1352639ae9bSBen Gras * Returns: 1362639ae9bSBen Gras * RET_ERROR, RET_SUCCESS 1372639ae9bSBen Gras */ 1382639ae9bSBen Gras int 1392639ae9bSBen Gras __ovfl_put(BTREE *t, const DBT *dbt, pgno_t *pg) 1402639ae9bSBen Gras { 1412639ae9bSBen Gras PAGE *h, *last; 1422639ae9bSBen Gras void *p; 1432639ae9bSBen Gras pgno_t npg; 1442639ae9bSBen Gras uint32_t sz, nb, plen; 1452639ae9bSBen Gras size_t temp; 1462639ae9bSBen Gras 1472639ae9bSBen Gras /* 1482639ae9bSBen Gras * Allocate pages and copy the key/data record into them. Store the 1492639ae9bSBen Gras * number of the first page in the chain. 1502639ae9bSBen Gras */ 1512639ae9bSBen Gras temp = t->bt_psize - BTDATAOFF; 1522639ae9bSBen Gras _DBFIT(temp, uint32_t); 1532639ae9bSBen Gras plen = (uint32_t)temp; 1542639ae9bSBen Gras last = NULL; 1552639ae9bSBen Gras p = dbt->data; 1562639ae9bSBen Gras temp = dbt->size; 1572639ae9bSBen Gras _DBFIT(temp, uint32_t); 158*f14fb602SLionel Sambuc sz = (uint32_t)temp; 1592639ae9bSBen Gras for (;; p = (char *)p + plen, last = h) { 1602639ae9bSBen Gras if ((h = __bt_new(t, &npg)) == NULL) 1612639ae9bSBen Gras return (RET_ERROR); 1622639ae9bSBen Gras 1632639ae9bSBen Gras h->pgno = npg; 1642639ae9bSBen Gras h->nextpg = h->prevpg = P_INVALID; 1652639ae9bSBen Gras h->flags = P_OVERFLOW; 1662639ae9bSBen Gras h->lower = h->upper = 0; 1672639ae9bSBen Gras 1682639ae9bSBen Gras nb = MIN(sz, plen); 1692639ae9bSBen Gras (void)memmove((char *)(void *)h + BTDATAOFF, p, (size_t)nb); 1702639ae9bSBen Gras 1712639ae9bSBen Gras if (last) { 1722639ae9bSBen Gras last->nextpg = h->pgno; 1732639ae9bSBen Gras mpool_put(t->bt_mp, last, MPOOL_DIRTY); 1742639ae9bSBen Gras } else 1752639ae9bSBen Gras *pg = h->pgno; 1762639ae9bSBen Gras 1772639ae9bSBen Gras if ((sz -= nb) == 0) { 1782639ae9bSBen Gras mpool_put(t->bt_mp, h, MPOOL_DIRTY); 1792639ae9bSBen Gras break; 1802639ae9bSBen Gras } 1812639ae9bSBen Gras } 1822639ae9bSBen Gras return (RET_SUCCESS); 1832639ae9bSBen Gras } 1842639ae9bSBen Gras 1852639ae9bSBen Gras /* 1862639ae9bSBen Gras * __OVFL_DELETE -- Delete an overflow chain. 1872639ae9bSBen Gras * 1882639ae9bSBen Gras * Parameters: 1892639ae9bSBen Gras * t: tree 1902639ae9bSBen Gras * p: pointer to { pgno_t, uint32_t } 1912639ae9bSBen Gras * 1922639ae9bSBen Gras * Returns: 1932639ae9bSBen Gras * RET_ERROR, RET_SUCCESS 1942639ae9bSBen Gras */ 1952639ae9bSBen Gras int 1962639ae9bSBen Gras __ovfl_delete(BTREE *t, void *p) 1972639ae9bSBen Gras { 1982639ae9bSBen Gras PAGE *h; 1992639ae9bSBen Gras pgno_t pg; 2002639ae9bSBen Gras uint32_t sz, plen; 2012639ae9bSBen Gras size_t temp; 2022639ae9bSBen Gras 203*f14fb602SLionel Sambuc (void)memmove(&pg, p, sizeof(pg)); 2042639ae9bSBen Gras (void)memmove(&sz, (char *)p + sizeof(pgno_t), sizeof(uint32_t)); 2052639ae9bSBen Gras 2062639ae9bSBen Gras #ifdef DEBUG 2072639ae9bSBen Gras if (pg == P_INVALID || sz == 0) 2082639ae9bSBen Gras abort(); 2092639ae9bSBen Gras #endif 2102639ae9bSBen Gras if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL) 2112639ae9bSBen Gras return (RET_ERROR); 2122639ae9bSBen Gras 2132639ae9bSBen Gras /* Don't delete chains used by internal pages. */ 2142639ae9bSBen Gras if (h->flags & P_PRESERVE) { 2152639ae9bSBen Gras mpool_put(t->bt_mp, h, 0); 2162639ae9bSBen Gras return (RET_SUCCESS); 2172639ae9bSBen Gras } 2182639ae9bSBen Gras 2192639ae9bSBen Gras /* Step through the chain, calling the free routine for each page. */ 2202639ae9bSBen Gras temp = t->bt_psize - BTDATAOFF; 2212639ae9bSBen Gras _DBFIT(temp, uint32_t); 2222639ae9bSBen Gras plen = (uint32_t)temp; 2232639ae9bSBen Gras for (;; sz -= plen) { 2242639ae9bSBen Gras pg = h->nextpg; 2252639ae9bSBen Gras __bt_free(t, h); 2262639ae9bSBen Gras if (sz <= plen) 2272639ae9bSBen Gras break; 2282639ae9bSBen Gras if ((h = mpool_get(t->bt_mp, pg, 0)) == NULL) 2292639ae9bSBen Gras return (RET_ERROR); 2302639ae9bSBen Gras } 2312639ae9bSBen Gras return (RET_SUCCESS); 2322639ae9bSBen Gras } 233