db2/db/db_overflow.c

   1 /*-
   2  * See the file LICENSE for redistribution information.
   3  *
   4  * Copyright (c) 1996, 1997
   5  *      Sleepycat Software.  All rights reserved.
   6  */
   7 /*
   8  * Copyright (c) 1990, 1993, 1994, 1995, 1996
   9  *      Keith Bostic.  All rights reserved.
  10  */
  11 /*
  12  * Copyright (c) 1990, 1993, 1994, 1995
  13  *      The Regents of the University of California.  All rights reserved.
  14  *
  15  * This code is derived from software contributed to Berkeley by
  16  * Mike Olson.
  17  *
  18  * Redistribution and use in source and binary forms, with or without
  19  * modification, are permitted provided that the following conditions
  20  * are met:
  21  * 1. Redistributions of source code must retain the above copyright
  22  *    notice, this list of conditions and the following disclaimer.
  23  * 2. Redistributions in binary form must reproduce the above copyright
  24  *    notice, this list of conditions and the following disclaimer in the
  25  *    documentation and/or other materials provided with the distribution.
  26  * 3. All advertising materials mentioning features or use of this software
  27  *    must display the following acknowledgement:
  28  *      This product includes software developed by the University of
  29  *      California, Berkeley and its contributors.
  30  * 4. Neither the name of the University nor the names of its contributors
  31  *    may be used to endorse or promote products derived from this software
  32  *    without specific prior written permission.
  33  *
  34  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  35  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  36  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  37  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  38  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  39  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  40  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  41  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  42  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  43  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  44  * SUCH DAMAGE.
  45  */
  46
  47 #include "config.h"
  48
  49 #ifndef lint
  50 static const char sccsid[] = "@(#)db_overflow.c 10.4 (Sleepycat) 7/2/97";
  51 #endif /* not lint */
  52
  53 #ifndef NO_SYSTEM_INCLUDES
  54 #include <sys/types.h>
  55
  56 #include <errno.h>
  57 #include <stdio.h>
  58 #include <stdlib.h>
  59 #include <string.h>
  60 #endif
  61
  62 #include "db_int.h"
  63 #include "db_page.h"
  64 #include "db_am.h"
  65 #include "common_ext.h"
  66
  67 /*
  68  * Big key/data code.
  69  *
  70  * Big key and data entries are stored on linked lists of pages.  The initial
  71  * reference is a structure with the total length of the item and the page
  72  * number where it begins.  Each entry in the linked list contains a pointer
  73  * to the next page of data, and so on.
  74  */
  75
  76 /*
  77  * __db_goff --
  78  *      Get an offpage item.
  79  *
  80  * PUBLIC: int __db_goff __P((DB *, DBT *,
  81  * PUBLIC:     u_int32_t, db_pgno_t, void **, u_int32_t *));
  82  */
  83 int
  84 __db_goff(dbp, dbt, tlen, pgno, bpp, bpsz)
  85         DB *dbp;
  86         DBT *dbt;
  87         u_int32_t tlen;
  88         db_pgno_t pgno;
  89         void **bpp;
  90         u_int32_t *bpsz;
  91 {
  92         PAGE *h;
  93         db_indx_t bytes;
  94         int ret;
  95         u_int32_t curoff, needed, start;
  96         u_int8_t *p, *src;
  97
  98         /*
  99          * Check if the buffer is big enough; if it is not and we are
 100          * allowed to malloc space, then we'll malloc it.  If we are
 101          * not (DB_DBT_USERMEM), then we'll set the dbt and return
 102          * appropriately.
 103          */
 104         if (F_ISSET(dbt, DB_DBT_PARTIAL)) {
 105                 start = dbt->doff;
 106                 needed = dbt->dlen;
 107         } else {
 108                 start = 0;
 109                 needed = tlen;
 110         }
 111
 112         /*
 113          * Allocate any necessary memory.
 114          *
 115          * XXX: Never allocate 0 bytes;
 116          */
 117         if (F_ISSET(dbt, DB_DBT_USERMEM)) {
 118                 if (needed > dbt->ulen) {
 119                         dbt->size = needed;
 120                         return (ENOMEM);
 121                 }
 122         } else if (F_ISSET(dbt, DB_DBT_MALLOC)) {
 123                 dbt->data = dbp->db_malloc == NULL ?
 124                     (void *)malloc(needed + 1) :
 125                     (void *)dbp->db_malloc(needed + 1);
 126                 if (dbt->data == NULL)
 127                         return (ENOMEM);
 128         } else if (*bpsz == 0 || *bpsz < needed) {
 129                 *bpp = (*bpp == NULL ?
 130                     (void *)malloc(needed + 1) :
 131                     (void *)realloc(*bpp, needed + 1));
 132                 if (*bpp == NULL)
 133                         return (ENOMEM);
 134                 *bpsz = needed + 1;
 135                 dbt->data = *bpp;
 136         } else
 137                 dbt->data = *bpp;
 138
 139         /*
 140          * Step through the linked list of pages, copying the data on each
 141          * one into the buffer.  Never copy more than the total data length.
 142          */
 143         dbt->size = needed;
 144         for (curoff = 0, p = dbt->data; pgno != P_INVALID && needed > 0;) {
 145                 if ((ret = memp_fget(dbp->mpf, &pgno, 0, &h)) != 0) {
 146                         (void)__db_pgerr(dbp, pgno);
 147                         return (ret);
 148                 }
 149                 /* Check if we need any bytes from this page. */
 150                 if (curoff + OV_LEN(h) >= start) {
 151                         src = (u_int8_t *)h + P_OVERHEAD;
 152                         bytes = OV_LEN(h);
 153                         if (start > curoff) {
 154                                 src += start - curoff;
 155                                 bytes -= start - curoff;
 156                         }
 157                         if (bytes > needed)
 158                                 bytes = needed;
 159                         memcpy(p, src, bytes);
 160                         p += bytes;
 161                         needed -= bytes;
 162                 }
 163                 curoff += OV_LEN(h);
 164                 pgno = h->next_pgno;
 165                 memp_fput(dbp->mpf, h, 0);
 166         }
 167         return (0);
 168 }
 169
 170 /*
 171  * __db_poff --
 172  *      Put an offpage item.
 173  *
 174  * PUBLIC: int __db_poff __P((DB *, const DBT *, db_pgno_t *,
 175  * PUBLIC:     int (*)(DB *, u_int32_t, PAGE **)));
 176  */
 177 int
 178 __db_poff(dbp, dbt, pgnop, newfunc)
 179         DB *dbp;
 180         const DBT *dbt;
 181         db_pgno_t *pgnop;
 182         int (*newfunc) __P((DB *, u_int32_t, PAGE **));
 183 {
 184         PAGE *pagep, *lastp;
 185         DB_LSN new_lsn, null_lsn;
 186         DBT tmp_dbt;
 187         db_indx_t pagespace;
 188         u_int32_t sz;
 189         u_int8_t *p;
 190         int ret;
 191
 192         /*
 193          * Allocate pages and copy the key/data item into them.  Calculate the
 194          * number of bytes we get for pages we fill completely with a single
 195          * item.
 196          */
 197         pagespace = P_MAXSPACE(dbp->pgsize);
 198
 199         lastp = NULL;
 200         for (p = dbt->data,
 201             sz = dbt->size; sz > 0; p += pagespace, sz -= pagespace) {
 202                 /*
 203                  * Reduce pagespace so we terminate the loop correctly and
 204                  * don't copy too much data.
 205                  */
 206                 if (sz < pagespace)
 207                         pagespace = sz;
 208
 209                 /*
 210                  * Allocate and initialize a new page and copy all or part of
 211                  * the item onto the page.  If sz is less than pagespace, we
 212                  * have a partial record.
 213                  */
 214                 if ((ret = newfunc(dbp, P_OVERFLOW, &pagep)) != 0)
 215                         return (ret);
 216                 if (DB_LOGGING(dbp)) {
 217                         tmp_dbt.data = p;
 218                         tmp_dbt.size = pagespace;
 219                         ZERO_LSN(null_lsn);
 220                         if ((ret = __db_big_log(dbp->dbenv->lg_info, dbp->txn,
 221                             &new_lsn, 0, DB_ADD_BIG, dbp->log_fileid,
 222                             PGNO(pagep), lastp ? PGNO(lastp) : PGNO_INVALID,
 223                             PGNO_INVALID, &tmp_dbt, &LSN(pagep),
 224                             lastp == NULL ? &null_lsn : &LSN(lastp),
 225                             &null_lsn)) != 0)
 226                                 return (ret);
 227
 228                         /* Move lsn onto page. */
 229                         if (lastp)
 230                                 LSN(lastp) = new_lsn;
 231                         LSN(pagep) = new_lsn;
 232                 }
 233
 234                 P_INIT(pagep, dbp->pgsize,
 235                     PGNO(pagep), PGNO_INVALID, PGNO_INVALID, 0, P_OVERFLOW);
 236                 OV_LEN(pagep) = pagespace;
 237                 OV_REF(pagep) = 1;
 238                 memcpy((u_int8_t *)pagep + P_OVERHEAD, p, pagespace);
 239
 240                 /*
 241                  * If this is the first entry, update the user's info.
 242                  * Otherwise, update the entry on the last page filled
 243                  * in and release that page.
 244                  */
 245                 if (lastp == NULL)
 246                         *pgnop = PGNO(pagep);
 247                 else {
 248                         lastp->next_pgno = PGNO(pagep);
 249                         pagep->prev_pgno = PGNO(lastp);
 250                         (void)memp_fput(dbp->mpf, lastp, DB_MPOOL_DIRTY);
 251                 }
 252                 lastp = pagep;
 253         }
 254         (void)memp_fput(dbp->mpf, lastp, DB_MPOOL_DIRTY);
 255         return (0);
 256 }
 257
 258 /*
 259  * __db_ioff --
 260  *      Increment the reference count on an overflow page.
 261  *
 262  * PUBLIC: int __db_ioff __P((DB *, db_pgno_t));
 263  */
 264 int
 265 __db_ioff(dbp, pgno)
 266         DB *dbp;
 267         db_pgno_t pgno;
 268 {
 269         PAGE *h;
 270         int ret;
 271
 272         if ((ret = memp_fget(dbp->mpf, &pgno, 0, &h)) != 0) {
 273                 (void)__db_pgerr(dbp, pgno);
 274                 return (ret);
 275         }
 276
 277         ++OV_REF(h);
 278         if (DB_LOGGING(dbp) && (ret = __db_ovref_log(dbp->dbenv->lg_info,
 279             dbp->txn, &LSN(h), 0, dbp->log_fileid, h->pgno, &LSN(h))) != 0)
 280                 return (ret);
 281
 282         (void)memp_fput(dbp->mpf, h, DB_MPOOL_DIRTY);
 283         return (0);
 284 }
 285
 286 /*
 287  * __db_doff --
 288  *      Delete an offpage chain of overflow pages.
 289  *
 290  * PUBLIC: int __db_doff __P((DB *, db_pgno_t, int (*)(DB *, PAGE *)));
 291  */
 292 int
 293 __db_doff(dbp, pgno, freefunc)
 294         DB *dbp;
 295         db_pgno_t pgno;
 296         int (*freefunc) __P((DB *, PAGE *));
 297 {
 298         PAGE *pagep;
 299         DB_LSN null_lsn;
 300         DBT tmp_dbt;
 301         int ret;
 302
 303         do {
 304                 if ((ret = memp_fget(dbp->mpf, &pgno, 0, &pagep)) != 0) {
 305                         (void)__db_pgerr(dbp, pgno);
 306                         return (ret);
 307                 }
 308
 309                 /*
 310                  * If it's an overflow page and it's referenced by more than
 311                  * one key/data item, decrement the reference count and return.
 312                  */
 313                 if (TYPE(pagep) == P_OVERFLOW && OV_REF(pagep) > 1) {
 314                         --OV_REF(pagep);
 315                         (void)memp_fput(dbp->mpf, pagep, DB_MPOOL_DIRTY);
 316                         return (0);
 317                 }
 318
 319                 if (DB_LOGGING(dbp)) {
 320                         tmp_dbt.data = (u_int8_t *)pagep + P_OVERHEAD;
 321                         tmp_dbt.size = OV_LEN(pagep);
 322                         ZERO_LSN(null_lsn);
 323                         if ((ret = __db_big_log(dbp->dbenv->lg_info, dbp->txn,
 324                             &LSN(pagep), 0, DB_REM_BIG, dbp->log_fileid,
 325                             PGNO(pagep), PREV_PGNO(pagep), NEXT_PGNO(pagep),
 326                             &tmp_dbt, &LSN(pagep), &null_lsn, &null_lsn)) != 0)
 327                                 return (ret);
 328                 }
 329                 pgno = pagep->next_pgno;
 330                 if ((ret = freefunc(dbp, pagep)) != 0)
 331                         return (ret);
 332         } while (pgno != PGNO_INVALID);
 333
 334         return (0);
 335 }
 336
 337 /*
 338  * __db_moff --
 339  *      Match on overflow pages.
 340  *
 341  * Given a starting page number and a key, return <0, 0, >0 to indicate if the
 342  * key on the page is less than, equal to or greater than the key specified.
 343  *
 344  * PUBLIC: int __db_moff __P((DB *, const DBT *, db_pgno_t));
 345  */
 346 int
 347 __db_moff(dbp, dbt, pgno)
 348         DB *dbp;
 349         const DBT *dbt;
 350         db_pgno_t pgno;
 351 {
 352         PAGE *pagep;
 353         u_int32_t cmp_bytes, key_left;
 354         int ret;
 355         u_int8_t *p1, *p2;
 356
 357         /* While there are both keys to compare. */
 358         for (ret = 0, p1 = dbt->data,
 359             key_left = dbt->size; key_left > 0 && pgno != PGNO_INVALID;) {
 360                 if (memp_fget(dbp->mpf, &pgno, 0, &pagep) != 0) {
 361                         (void)__db_pgerr(dbp, pgno);
 362                         return (0);     /* No system error return. */
 363                 }
 364
 365                 cmp_bytes = OV_LEN(pagep) < key_left ? OV_LEN(pagep) : key_left;
 366                 key_left -= cmp_bytes;
 367                 for (p2 =
 368                     (u_int8_t *)pagep + P_OVERHEAD; cmp_bytes-- > 0; ++p1, ++p2)
 369                         if (*p1 != *p2) {
 370                                 ret = (long)*p1 - (long)*p2;
 371                                 break;
 372                         }
 373                 pgno = NEXT_PGNO(pagep);
 374                 (void)memp_fput(dbp->mpf, pagep, 0);
 375                 if (ret != 0)
 376                         return (ret);
 377         }
 378         if (key_left > 0)               /* DBT is longer than page key. */
 379                 return (-1);
 380         if (pgno != PGNO_INVALID)       /* DBT is shorter than page key. */
 381                 return (1);
 382         return (0);
 383 }