db/btree/FreeBSD/bt_put.c

   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: src/lib/libc/db/btree/bt_put.c,v 1.4 2003/05/30 11:05:08 tmm Exp $");
  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, tmp;
  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                         tmp = key->size;
 135                         memmove(kb + sizeof(pgno_t),
 136                             &tmp, sizeof(u_int32_t));
 137                         dflags |= P_BIGKEY;
 138                         key = &tkey;
 139                 }
 140                 if (key->size + data->size > t->bt_ovflsize) {
 141                         if (__ovfl_put(t, data, &pg) == RET_ERROR)
 142                                 return (RET_ERROR);
 143                         tdata.data = db;
 144                         tdata.size = NOVFLSIZE;
 145                         memmove(db, &pg, sizeof(pgno_t));
 146                         tmp = data->size;
 147                         memmove(db + sizeof(pgno_t),
 148                             &tmp, sizeof(u_int32_t));
 149                         dflags |= P_BIGDATA;
 150                         data = &tdata;
 151                 }
 152                 if (key->size + data->size > t->bt_ovflsize)
 153                         goto storekey;
 154         }
 155
 156         /* Replace the cursor. */
 157         if (flags == R_CURSOR) {
 158                 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL)
 159                         return (RET_ERROR);
 160                 index = t->bt_cursor.pg.index;
 161                 goto delete;
 162         }
 163
 164         /*
 165          * Find the key to delete, or, the location at which to insert.
 166          * Bt_fast and __bt_search both pin the returned page.
 167          */
 168         if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
 169                 if ((e = __bt_search(t, key, &exact)) == NULL)
 170                         return (RET_ERROR);
 171         h = e->page;
 172         index = e->index;
 173
 174         /*
 175          * Add the key/data pair to the tree.  If an identical key is already
 176          * in the tree, and R_NOOVERWRITE is set, an error is returned.  If
 177          * R_NOOVERWRITE is not set, the key is either added (if duplicates are
 178          * permitted) or an error is returned.
 179          */
 180         switch (flags) {
 181         case R_NOOVERWRITE:
 182                 if (!exact)
 183                         break;
 184                 mpool_put(t->bt_mp, h, 0);
 185                 return (RET_SPECIAL);
 186         default:
 187                 if (!exact || !F_ISSET(t, B_NODUPS))
 188                         break;
 189                 /*
 190                  * !!!
 191                  * Note, the delete may empty the page, so we need to put a
 192                  * new entry into the page immediately.
 193                  */
 194 delete:         if (__bt_dleaf(t, key, h, index) == RET_ERROR) {
 195                         mpool_put(t->bt_mp, h, 0);
 196                         return (RET_ERROR);
 197                 }
 198                 break;
 199         }
 200
 201         /*
 202          * If not enough room, or the user has put a ceiling on the number of
 203          * keys permitted in the page, split the page.  The split code will
 204          * insert the key and data and unpin the current page.  If inserting
 205          * into the offset array, shift the pointers up.
 206          */
 207         nbytes = NBLEAFDBT(key->size, data->size);
 208         if (h->upper - h->lower < nbytes + sizeof(indx_t)) {
 209                 if ((status = __bt_split(t, h, key,
 210                     data, dflags, nbytes, index)) != RET_SUCCESS)
 211                         return (status);
 212                 goto success;
 213         }
 214
 215         if (index < (nxtindex = NEXTINDEX(h)))
 216                 memmove(h->linp + index + 1, h->linp + index,
 217                     (nxtindex - index) * sizeof(indx_t));
 218         h->lower += sizeof(indx_t);
 219
 220         h->linp[index] = h->upper -= nbytes;
 221         dest = (char *)h + h->upper;
 222         WR_BLEAF(dest, key, data, dflags);
 223
 224         /* If the cursor is on this page, adjust it as necessary. */
 225         if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
 226             !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
 227             t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= index)
 228                 ++t->bt_cursor.pg.index;
 229
 230         if (t->bt_order == NOT) {
 231                 if (h->nextpg == P_INVALID) {
 232                         if (index == NEXTINDEX(h) - 1) {
 233                                 t->bt_order = FORWARD;
 234                                 t->bt_last.index = index;
 235                                 t->bt_last.pgno = h->pgno;
 236                         }
 237                 } else if (h->prevpg == P_INVALID) {
 238                         if (index == 0) {
 239                                 t->bt_order = BACK;
 240                                 t->bt_last.index = 0;
 241                                 t->bt_last.pgno = h->pgno;
 242                         }
 243                 }
 244         }
 245
 246         mpool_put(t->bt_mp, h, MPOOL_DIRTY);
 247
 248 success:
 249         if (flags == R_SETCURSOR)
 250                 __bt_setcur(t, e->page->pgno, e->index);
 251
 252         F_SET(t, B_MODIFIED);
 253         return (RET_SUCCESS);
 254 }
 255
 256 #ifdef STATISTICS
 257 u_long bt_cache_hit, bt_cache_miss;
 258 #endif
 259
 260 /*
 261  * BT_FAST -- Do a quick check for sorted data.
 262  *
 263  * Parameters:
 264  *      t:      tree
 265  *      key:    key to insert
 266  *
 267  * Returns:
 268  *      EPG for new record or NULL if not found.
 269  */
 270 static EPG *
 271 bt_fast(t, key, data, exactp)
 272         BTREE *t;
 273         const DBT *key, *data;
 274         int *exactp;
 275 {
 276         PAGE *h;
 277         u_int32_t nbytes;
 278         int cmp;
 279
 280         if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
 281                 t->bt_order = NOT;
 282                 return (NULL);
 283         }
 284         t->bt_cur.page = h;
 285         t->bt_cur.index = t->bt_last.index;
 286
 287         /*
 288          * If won't fit in this page or have too many keys in this page,
 289          * have to search to get split stack.
 290          */
 291         nbytes = NBLEAFDBT(key->size, data->size);
 292         if (h->upper - h->lower < nbytes + sizeof(indx_t))
 293                 goto miss;
 294
 295         if (t->bt_order == FORWARD) {
 296                 if (t->bt_cur.page->nextpg != P_INVALID)
 297                         goto miss;
 298                 if (t->bt_cur.index != NEXTINDEX(h) - 1)
 299                         goto miss;
 300                 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0)
 301                         goto miss;
 302                 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index;
 303         } else {
 304                 if (t->bt_cur.page->prevpg != P_INVALID)
 305                         goto miss;
 306                 if (t->bt_cur.index != 0)
 307                         goto miss;
 308                 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0)
 309                         goto miss;
 310                 t->bt_last.index = 0;
 311         }
 312         *exactp = cmp == 0;
 313 #ifdef STATISTICS
 314         ++bt_cache_hit;
 315 #endif
 316         return (&t->bt_cur);
 317
 318 miss:
 319 #ifdef STATISTICS
 320         ++bt_cache_miss;
 321 #endif
 322         t->bt_order = NOT;
 323         mpool_put(t->bt_mp, h, 0);
 324         return (NULL);
 325 }