db2/db/db_overflow.c

   1 /*-
   2  * See the file LICENSE for redistribution information.
   3  *
   4  * Copyright (c) 1996, 1997, 1998
   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 10.11 (Sleepycat) 5/7/98";
  51 #endif /* not lint */
  52
  53 #ifndef NO_SYSTEM_INCLUDES
  54 #include <sys/types.h>
  55
  56 #include <errno.h>
  57 #include <string.h>
  58 #endif
  59
  60 #include "db_int.h"
  61 #include "db_page.h"
  62 #include "db_am.h"
  63
  64 /*
  65  * Big key/data code.
  66  *
  67  * Big key and data entries are stored on linked lists of pages.  The initial
  68  * reference is a structure with the total length of the item and the page
  69  * number where it begins.  Each entry in the linked list contains a pointer
  70  * to the next page of data, and so on.
  71  */
  72
  73 /*
  74  * __db_goff --
  75  *      Get an offpage item.
  76  *
  77  * PUBLIC: int __db_goff __P((DB *, DBT *,
  78  * PUBLIC:     u_int32_t, db_pgno_t, void **, u_int32_t *));
  79  */
  80 int
  81 __db_goff(dbp, dbt, tlen, pgno, bpp, bpsz)
  82         DB *dbp;
  83         DBT *dbt;
  84         u_int32_t tlen;
  85         db_pgno_t pgno;
  86         void **bpp;
  87         u_int32_t *bpsz;
  88 {
  89         PAGE *h;
  90         db_indx_t bytes;
  91         u_int32_t curoff, needed, start;
  92         u_int8_t *p, *src;
  93         int ret;
  94
  95         /*
  96          * Check if the buffer is big enough; if it is not and we are
  97          * allowed to malloc space, then we'll malloc it.  If we are
  98          * not (DB_DBT_USERMEM), then we'll set the dbt and return
  99          * appropriately.
 100          */
 101         if (F_ISSET(dbt, DB_DBT_PARTIAL)) {
 102                 start = dbt->doff;
 103                 needed = dbt->dlen;
 104         } else {
 105                 start = 0;
 106                 needed = tlen;
 107         }
 108
 109         /*
 110          * Allocate any necessary memory.
 111          *
 112          * XXX: Never allocate 0 bytes;
 113          */
 114         if (F_ISSET(dbt, DB_DBT_USERMEM)) {
 115                 if (needed > dbt->ulen) {
 116                         dbt->size = needed;
 117                         return (ENOMEM);
 118                 }
 119         } else if (F_ISSET(dbt, DB_DBT_MALLOC)) {
 120                 dbt->data = dbp->db_malloc == NULL ?
 121                     (void *)__db_malloc(needed + 1) :
 122                     (void *)dbp->db_malloc(needed + 1);
 123                 if (dbt->data == NULL)
 124                         return (ENOMEM);
 125         } else if (*bpsz == 0 || *bpsz < needed) {
 126                 *bpp = (*bpp == NULL ?
 127                     (void *)__db_malloc(needed + 1) :
 128                     (void *)__db_realloc(*bpp, needed + 1));
 129                 if (*bpp == NULL)
 130                         return (ENOMEM);
 131                 *bpsz = needed + 1;
 132                 dbt->data = *bpp;
 133         } else
 134                 dbt->data = *bpp;
 135
 136         /*
 137          * Step through the linked list of pages, copying the data on each
 138          * one into the buffer.  Never copy more than the total data length.
 139          */
 140         dbt->size = needed;
 141         for (curoff = 0, p = dbt->data; pgno != P_INVALID && needed > 0;) {
 142                 if ((ret = memp_fget(dbp->mpf, &pgno, 0, &h)) != 0) {
 143                         (void)__db_pgerr(dbp, pgno);
 144                         return (ret);
 145                 }
 146                 /* Check if we need any bytes from this page. */
 147                 if (curoff + OV_LEN(h) >= start) {
 148                         src = (u_int8_t *)h + P_OVERHEAD;
 149                         bytes = OV_LEN(h);
 150                         if (start > curoff) {
 151                                 src += start - curoff;
 152                                 bytes -= start - curoff;
 153                         }
 154                         if (bytes > needed)
 155                                 bytes = needed;
 156                         memcpy(p, src, bytes);
 157                         p += bytes;
 158                         needed -= bytes;
 159                 }
 160                 curoff += OV_LEN(h);
 161                 pgno = h->next_pgno;
 162                 memp_fput(dbp->mpf, h, 0);
 163         }
 164         return (0);
 165 }
 166
 167 /*
 168  * __db_poff --
 169  *      Put an offpage item.
 170  *
 171  * PUBLIC: int __db_poff __P((DB *, const DBT *, db_pgno_t *,
 172  * PUBLIC:     int (*)(DB *, u_int32_t, PAGE **)));
 173  */
 174 int
 175 __db_poff(dbp, dbt, pgnop, newfunc)
 176         DB *dbp;
 177         const DBT *dbt;
 178         db_pgno_t *pgnop;
 179         int (*newfunc) __P((DB *, u_int32_t, PAGE **));
 180 {
 181         PAGE *pagep, *lastp;
 182         DB_LSN new_lsn, null_lsn;
 183         DBT tmp_dbt;
 184         db_indx_t pagespace;
 185         u_int32_t sz;
 186         u_int8_t *p;
 187         int ret;
 188
 189         /*
 190          * Allocate pages and copy the key/data item into them.  Calculate the
 191          * number of bytes we get for pages we fill completely with a single
 192          * item.
 193          */
 194         pagespace = P_MAXSPACE(dbp->pgsize);
 195
 196         lastp = NULL;
 197         for (p = dbt->data,
 198             sz = dbt->size; sz > 0; p += pagespace, sz -= pagespace) {
 199                 /*
 200                  * Reduce pagespace so we terminate the loop correctly and
 201                  * don't copy too much data.
 202                  */
 203                 if (sz < pagespace)
 204                         pagespace = sz;
 205
 206                 /*
 207                  * Allocate and initialize a new page and copy all or part of
 208                  * the item onto the page.  If sz is less than pagespace, we
 209                  * have a partial record.
 210                  */
 211                 if ((ret = newfunc(dbp, P_OVERFLOW, &pagep)) != 0)
 212                         return (ret);
 213                 if (DB_LOGGING(dbp)) {
 214                         tmp_dbt.data = p;
 215                         tmp_dbt.size = pagespace;
 216                         ZERO_LSN(null_lsn);
 217                         if ((ret = __db_big_log(dbp->dbenv->lg_info, dbp->txn,
 218                             &new_lsn, 0, DB_ADD_BIG, dbp->log_fileid,
 219                             PGNO(pagep), lastp ? PGNO(lastp) : PGNO_INVALID,
 220                             PGNO_INVALID, &tmp_dbt, &LSN(pagep),
 221                             lastp == NULL ? &null_lsn : &LSN(lastp),
 222                             &null_lsn)) != 0)
 223                                 return (ret);
 224
 225                         /* Move lsn onto page. */
 226                         if (lastp)
 227                                 LSN(lastp) = new_lsn;
 228                         LSN(pagep) = new_lsn;
 229                 }
 230
 231                 P_INIT(pagep, dbp->pgsize,
 232                     PGNO(pagep), PGNO_INVALID, PGNO_INVALID, 0, P_OVERFLOW);
 233                 OV_LEN(pagep) = pagespace;
 234                 OV_REF(pagep) = 1;
 235                 memcpy((u_int8_t *)pagep + P_OVERHEAD, p, pagespace);
 236
 237                 /*
 238                  * If this is the first entry, update the user's info.
 239                  * Otherwise, update the entry on the last page filled
 240                  * in and release that page.
 241                  */
 242                 if (lastp == NULL)
 243                         *pgnop = PGNO(pagep);
 244                 else {
 245                         lastp->next_pgno = PGNO(pagep);
 246                         pagep->prev_pgno = PGNO(lastp);
 247                         (void)memp_fput(dbp->mpf, lastp, DB_MPOOL_DIRTY);
 248                 }
 249                 lastp = pagep;
 250         }
 251         (void)memp_fput(dbp->mpf, lastp, DB_MPOOL_DIRTY);
 252         return (0);
 253 }
 254
 255 /*
 256  * __db_ovref --
 257  *      Increment/decrement the reference count on an overflow page.
 258  *
 259  * PUBLIC: int __db_ovref __P((DB *, db_pgno_t, int32_t));
 260  */
 261 int
 262 __db_ovref(dbp, pgno, adjust)
 263         DB *dbp;
 264         db_pgno_t pgno;
 265         int32_t adjust;
 266 {
 267         PAGE *h;
 268         int ret;
 269
 270         if ((ret = memp_fget(dbp->mpf, &pgno, 0, &h)) != 0) {
 271                 (void)__db_pgerr(dbp, pgno);
 272                 return (ret);
 273         }
 274
 275         if (DB_LOGGING(dbp))
 276                 if ((ret = __db_ovref_log(dbp->dbenv->lg_info, dbp->txn,
 277                     &LSN(h), 0, dbp->log_fileid, h->pgno, adjust,
 278                     &LSN(h))) != 0)
 279                         return (ret);
 280         OV_REF(h) += adjust;
 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                         (void)memp_fput(dbp->mpf, pagep, 0);
 315                         return (__db_ovref(dbp, pgno, -1));
 316                 }
 317
 318                 if (DB_LOGGING(dbp)) {
 319                         tmp_dbt.data = (u_int8_t *)pagep + P_OVERHEAD;
 320                         tmp_dbt.size = OV_LEN(pagep);
 321                         ZERO_LSN(null_lsn);
 322                         if ((ret = __db_big_log(dbp->dbenv->lg_info, dbp->txn,
 323                             &LSN(pagep), 0, DB_REM_BIG, dbp->log_fileid,
 324                             PGNO(pagep), PREV_PGNO(pagep), NEXT_PGNO(pagep),
 325                             &tmp_dbt, &LSN(pagep), &null_lsn, &null_lsn)) != 0)
 326                                 return (ret);
 327                 }
 328                 pgno = pagep->next_pgno;
 329                 if ((ret = freefunc(dbp, pagep)) != 0)
 330                         return (ret);
 331         } while (pgno != PGNO_INVALID);
 332
 333         return (0);
 334 }
 335
 336 /*
 337  * __db_moff --
 338  *      Match on overflow pages.
 339  *
 340  * Given a starting page number and a key, return <0, 0, >0 to indicate if the
 341  * key on the page is less than, equal to or greater than the key specified.
 342  *
 343  * PUBLIC: int __db_moff __P((DB *, const DBT *, db_pgno_t));
 344  */
 345 int
 346 __db_moff(dbp, dbt, pgno)
 347         DB *dbp;
 348         const DBT *dbt;
 349         db_pgno_t pgno;
 350 {
 351         PAGE *pagep;
 352         u_int32_t cmp_bytes, key_left;
 353         u_int8_t *p1, *p2;
 354         int ret;
 355
 356         /* While there are both keys to compare. */
 357         for (ret = 0, p1 = dbt->data,
 358             key_left = dbt->size; key_left > 0 && pgno != PGNO_INVALID;) {
 359                 if (memp_fget(dbp->mpf, &pgno, 0, &pagep) != 0) {
 360                         (void)__db_pgerr(dbp, pgno);
 361                         return (0);     /* No system error return. */
 362                 }
 363
 364                 cmp_bytes = OV_LEN(pagep) < key_left ? OV_LEN(pagep) : key_left;
 365                 key_left -= cmp_bytes;
 366                 for (p2 =
 367                     (u_int8_t *)pagep + P_OVERHEAD; cmp_bytes-- > 0; ++p1, ++p2)
 368                         if (*p1 != *p2) {
 369                                 ret = (long)*p1 - (long)*p2;
 370                                 break;
 371                         }
 372                 pgno = NEXT_PGNO(pagep);
 373                 (void)memp_fput(dbp->mpf, pagep, 0);
 374                 if (ret != 0)
 375                         return (ret);
 376         }
 377         if (key_left > 0)               /* DBT is longer than page key. */
 378                 return (-1);
 379         if (pgno != PGNO_INVALID)       /* DBT is shorter than page key. */
 380                 return (1);
 381         return (0);
 382 }