lib/libc/DB/mpool/mpool.c

   1 /*-
   2  * Copyright (c) 1990 The Regents of the University of California.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. All advertising materials mentioning features or use of this software
  14  *    must display the following acknowledgement:
  15  *      This product includes software developed by the University of
  16  *      California, Berkeley and its contributors.
  17  * 4. Neither the name of the University nor the names of its contributors
  18  *    may be used to endorse or promote products derived from this software
  19  *    without specific prior written permission.
  20  *
  21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  31  * SUCH DAMAGE.
  32  */
  33
  34 #if defined(LIBC_SCCS) && !defined(lint)
  35 static char sccsid[] = "@(#)mpool.c     5.5 (Berkeley) 2/19/93";
  36 #endif /* LIBC_SCCS and not lint */
  37
  38 #include <sys/param.h>
  39 #include <sys/stat.h>
  40
  41 #include <errno.h>
  42 #include <stdio.h>
  43 #include <stdlib.h>
  44 #include <string.h>
  45 #include <unistd.h>
  46
  47 #include <db.h>
  48 #define __MPOOLINTERFACE_PRIVATE
  49 #include "mpool.h"
  50
  51 static BKT *mpool_bkt __P((MPOOL *));
  52 static BKT *mpool_look __P((MPOOL *, pgno_t));
  53 static int  mpool_write __P((MPOOL *, BKT *));
  54 #ifdef DEBUG
  55 static void err __P((const char *fmt, ...));
  56 #endif
  57
  58 /*
  59  * MPOOL_OPEN -- initialize a memory pool.
  60  *
  61  * Parameters:
  62  *      key:            Shared buffer key.
  63  *      fd:             File descriptor.
  64  *      pagesize:       File page size.
  65  *      maxcache:       Max number of cached pages.
  66  *
  67  * Returns:
  68  *      MPOOL pointer, NULL on error.
  69  */
  70 MPOOL *
  71 mpool_open(key, fd, pagesize, maxcache)
  72         DBT *key;
  73         int fd;
  74         pgno_t pagesize, maxcache;
  75 {
  76         struct stat sb;
  77         MPOOL *mp;
  78         int entry;
  79
  80         if (fstat(fd, &sb))
  81                 return (NULL);
  82         /* XXX
  83          * We should only set st_size to 0 for pipes -- 4.4BSD has the fix so
  84          * that stat(2) returns true for ISSOCK on pipes.  Until then, this is
  85          * fairly close.
  86          */
  87         if (!S_ISREG(sb.st_mode)) {
  88                 errno = ESPIPE;
  89                 return (NULL);
  90         }
  91
  92         if ((mp = malloc(sizeof(MPOOL))) == NULL)
  93                 return (NULL);
  94         mp->free.cnext = mp->free.cprev = (BKT *)&mp->free;
  95         mp->lru.cnext = mp->lru.cprev = (BKT *)&mp->lru;
  96         for (entry = 0; entry < HASHSIZE; ++entry)
  97                 mp->hashtable[entry].hnext = mp->hashtable[entry].hprev =
  98                     mp->hashtable[entry].cnext = mp->hashtable[entry].cprev =
  99                     (BKT *)&mp->hashtable[entry];
 100         mp->curcache = 0;
 101         mp->maxcache = maxcache;
 102         mp->pagesize = pagesize;
 103         mp->npages = sb.st_size / pagesize;
 104         mp->fd = fd;
 105         mp->pgcookie = NULL;
 106         mp->pgin = mp->pgout = NULL;
 107
 108 #ifdef STATISTICS
 109         mp->cachehit = mp->cachemiss = mp->pagealloc = mp->pageflush =
 110             mp->pageget = mp->pagenew = mp->pageput = mp->pageread =
 111             mp->pagewrite = 0;
 112 #endif
 113         return (mp);
 114 }
 115
 116 /*
 117  * MPOOL_FILTER -- initialize input/output filters.
 118  *
 119  * Parameters:
 120  *      pgin:           Page in conversion routine.
 121  *      pgout:          Page out conversion routine.
 122  *      pgcookie:       Cookie for page in/out routines.
 123  */
 124 void
 125 mpool_filter(mp, pgin, pgout, pgcookie)
 126         MPOOL *mp;
 127         void (*pgin) __P((void *, pgno_t, void *));
 128         void (*pgout) __P((void *, pgno_t, void *));
 129         void *pgcookie;
 130 {
 131         mp->pgin = pgin;
 132         mp->pgout = pgout;
 133         mp->pgcookie = pgcookie;
 134 }
 135
 136 /*
 137  * MPOOL_NEW -- get a new page
 138  *
 139  * Parameters:
 140  *      mp:             mpool cookie
 141  *      pgnoadddr:      place to store new page number
 142  * Returns:
 143  *      RET_ERROR, RET_SUCCESS
 144  */
 145 void *
 146 mpool_new(mp, pgnoaddr)
 147         MPOOL *mp;
 148         pgno_t *pgnoaddr;
 149 {
 150         BKT *b;
 151         BKTHDR *hp;
 152
 153 #ifdef STATISTICS
 154         ++mp->pagenew;
 155 #endif
 156         /*
 157          * Get a BKT from the cache.  Assign a new page number, attach it to
 158          * the hash and lru chains and return.
 159          */
 160         if ((b = mpool_bkt(mp)) == NULL)
 161                 return (NULL);
 162         *pgnoaddr = b->pgno = mp->npages++;
 163         b->flags = MPOOL_PINNED;
 164         inshash(b, b->pgno);
 165         inschain(b, &mp->lru);
 166         return (b->page);
 167 }
 168
 169 /*
 170  * MPOOL_GET -- get a page from the pool
 171  *
 172  * Parameters:
 173  *      mp:     mpool cookie
 174  *      pgno:   page number
 175  *      flags:  not used
 176  *
 177  * Returns:
 178  *      RET_ERROR, RET_SUCCESS
 179  */
 180 void *
 181 mpool_get(mp, pgno, flags)
 182         MPOOL *mp;
 183         pgno_t pgno;
 184         u_int flags;            /* XXX not used? */
 185 {
 186         BKT *b;
 187         BKTHDR *hp;
 188         off_t off;
 189         int nr;
 190
 191         /*
 192          * If asking for a specific page that is already in the cache, find
 193          * it and return it.
 194          */
 195         if (b = mpool_look(mp, pgno)) {
 196 #ifdef STATISTICS
 197                 ++mp->pageget;
 198 #endif
 199 #ifdef DEBUG
 200                 if (b->flags & MPOOL_PINNED)
 201                         err("mpool_get: page %d already pinned", b->pgno);
 202 #endif
 203                 rmchain(b);
 204                 inschain(b, &mp->lru);
 205                 b->flags |= MPOOL_PINNED;
 206                 return (b->page);
 207         }
 208
 209         /* Not allowed to retrieve a non-existent page. */
 210         if (pgno >= mp->npages) {
 211                 errno = EINVAL;
 212                 return (NULL);
 213         }
 214
 215         /* Get a page from the cache. */
 216         if ((b = mpool_bkt(mp)) == NULL)
 217                 return (NULL);
 218         b->pgno = pgno;
 219         b->flags = MPOOL_PINNED;
 220
 221 #ifdef STATISTICS
 222         ++mp->pageread;
 223 #endif
 224         /* Read in the contents. */
 225         off = mp->pagesize * pgno;
 226         if (lseek(mp->fd, off, SEEK_SET) != off)
 227                 return (NULL);
 228         if ((nr = read(mp->fd, b->page, mp->pagesize)) != mp->pagesize) {
 229                 if (nr >= 0)
 230                         errno = EFTYPE;
 231                 return (NULL);
 232         }
 233         if (mp->pgin)
 234                 (mp->pgin)(mp->pgcookie, b->pgno, b->page);
 235
 236         inshash(b, b->pgno);
 237         inschain(b, &mp->lru);
 238 #ifdef STATISTICS
 239         ++mp->pageget;
 240 #endif
 241         return (b->page);
 242 }
 243
 244 /*
 245  * MPOOL_PUT -- return a page to the pool
 246  *
 247  * Parameters:
 248  *      mp:     mpool cookie
 249  *      page:   page pointer
 250  *      pgno:   page number
 251  *
 252  * Returns:
 253  *      RET_ERROR, RET_SUCCESS
 254  */
 255 int
 256 mpool_put(mp, page, flags)
 257         MPOOL *mp;
 258         void *page;
 259         u_int flags;
 260 {
 261         BKT *baddr;
 262 #ifdef DEBUG
 263         BKT *b;
 264 #endif
 265
 266 #ifdef STATISTICS
 267         ++mp->pageput;
 268 #endif
 269         baddr = (BKT *)((char *)page - sizeof(BKT));
 270 #ifdef DEBUG
 271         if (!(baddr->flags & MPOOL_PINNED))
 272                 err("mpool_put: page %d not pinned", b->pgno);
 273         for (b = mp->lru.cnext; b != (BKT *)&mp->lru; b = b->cnext) {
 274                 if (b == (BKT *)&mp->lru)
 275                         err("mpool_put: %0x: bad address", baddr);
 276                 if (b == baddr)
 277                         break;
 278         }
 279 #endif
 280         baddr->flags &= ~MPOOL_PINNED;
 281         baddr->flags |= flags & MPOOL_DIRTY;
 282         return (RET_SUCCESS);
 283 }
 284
 285 /*
 286  * MPOOL_CLOSE -- close the buffer pool
 287  *
 288  * Parameters:
 289  *      mp:     mpool cookie
 290  *
 291  * Returns:
 292  *      RET_ERROR, RET_SUCCESS
 293  */
 294 int
 295 mpool_close(mp)
 296         MPOOL *mp;
 297 {
 298         BKT *b, *next;
 299
 300         /* Free up any space allocated to the lru pages. */
 301         for (b = mp->lru.cprev; b != (BKT *)&mp->lru; b = next) {
 302                 next = b->cprev;
 303                 free(b);
 304         }
 305         free(mp);
 306         return (RET_SUCCESS);
 307 }
 308
 309 /*
 310  * MPOOL_SYNC -- sync the file to disk.
 311  *
 312  * Parameters:
 313  *      mp:     mpool cookie
 314  *
 315  * Returns:
 316  *      RET_ERROR, RET_SUCCESS
 317  */
 318 int
 319 mpool_sync(mp)
 320         MPOOL *mp;
 321 {
 322         BKT *b;
 323
 324         for (b = mp->lru.cprev; b != (BKT *)&mp->lru; b = b->cprev)
 325                 if (b->flags & MPOOL_DIRTY && mpool_write(mp, b) == RET_ERROR)
 326                         return (RET_ERROR);
 327         return (fsync(mp->fd) ? RET_ERROR : RET_SUCCESS);
 328 }
 329
 330 /*
 331  * MPOOL_BKT -- get/create a BKT from the cache
 332  *
 333  * Parameters:
 334  *      mp:     mpool cookie
 335  *
 336  * Returns:
 337  *      NULL on failure and a pointer to the BKT on success
 338  */
 339 static BKT *
 340 mpool_bkt(mp)
 341         MPOOL *mp;
 342 {
 343         BKT *b;
 344
 345         if (mp->curcache < mp->maxcache)
 346                 goto new;
 347
 348         /*
 349          * If the cache is maxxed out, search the lru list for a buffer we
 350          * can flush.  If we find one, write it if necessary and take it off
 351          * any lists.  If we don't find anything we grow the cache anyway.
 352          * The cache never shrinks.
 353          */
 354         for (b = mp->lru.cprev; b != (BKT *)&mp->lru; b = b->cprev)
 355                 if (!(b->flags & MPOOL_PINNED)) {
 356                         if (b->flags & MPOOL_DIRTY &&
 357                             mpool_write(mp, b) == RET_ERROR)
 358                                 return (NULL);
 359                         rmhash(b);
 360                         rmchain(b);
 361 #ifdef STATISTICS
 362                         ++mp->pageflush;
 363 #endif
 364 #ifdef DEBUG
 365                         {
 366                                 void *spage;
 367                                 spage = b->page;
 368                                 memset(b, 0xff, sizeof(BKT) + mp->pagesize);
 369                                 b->page = spage;
 370                         }
 371 #endif
 372                         return (b);
 373                 }
 374
 375 new:    if ((b = malloc(sizeof(BKT) + mp->pagesize)) == NULL)
 376                 return (NULL);
 377 #ifdef STATISTICS
 378         ++mp->pagealloc;
 379 #endif
 380 #ifdef DEBUG
 381         memset(b, 0xff, sizeof(BKT) + mp->pagesize);
 382 #endif
 383         b->page = (char *)b + sizeof(BKT);
 384         ++mp->curcache;
 385         return (b);
 386 }
 387
 388 /*
 389  * MPOOL_WRITE -- sync a page to disk
 390  *
 391  * Parameters:
 392  *      mp:     mpool cookie
 393  *
 394  * Returns:
 395  *      RET_ERROR, RET_SUCCESS
 396  */
 397 static int
 398 mpool_write(mp, b)
 399         MPOOL *mp;
 400         BKT *b;
 401 {
 402         off_t off;
 403
 404         if (mp->pgout)
 405                 (mp->pgout)(mp->pgcookie, b->pgno, b->page);
 406
 407 #ifdef STATISTICS
 408         ++mp->pagewrite;
 409 #endif
 410         off = mp->pagesize * b->pgno;
 411         if (lseek(mp->fd, off, SEEK_SET) != off)
 412                 return (RET_ERROR);
 413         if (write(mp->fd, b->page, mp->pagesize) != mp->pagesize)
 414                 return (RET_ERROR);
 415         b->flags &= ~MPOOL_DIRTY;
 416         return (RET_SUCCESS);
 417 }
 418
 419 /*
 420  * MPOOL_LOOK -- lookup a page
 421  *
 422  * Parameters:
 423  *      mp:     mpool cookie
 424  *      pgno:   page number
 425  *
 426  * Returns:
 427  *      NULL on failure and a pointer to the BKT on success
 428  */
 429 static BKT *
 430 mpool_look(mp, pgno)
 431         MPOOL *mp;
 432         pgno_t pgno;
 433 {
 434         register BKT *b;
 435         register BKTHDR *tb;
 436
 437         /* XXX
 438          * If find the buffer, put it first on the hash chain so can
 439          * find it again quickly.
 440          */
 441         tb = &mp->hashtable[HASHKEY(pgno)];
 442         for (b = tb->hnext; b != (BKT *)tb; b = b->hnext)
 443                 if (b->pgno == pgno) {
 444 #ifdef STATISTICS
 445                         ++mp->cachehit;
 446 #endif
 447                         return (b);
 448                 }
 449 #ifdef STATISTICS
 450         ++mp->cachemiss;
 451 #endif
 452         return (NULL);
 453 }
 454
 455 #ifdef STATISTICS
 456 /*
 457  * MPOOL_STAT -- cache statistics
 458  *
 459  * Parameters:
 460  *      mp:     mpool cookie
 461  */
 462 void
 463 mpool_stat(mp)
 464         MPOOL *mp;
 465 {
 466         BKT *b;
 467         int cnt;
 468         char *sep;
 469
 470         (void)fprintf(stderr, "%lu pages in the file\n", mp->npages);
 471         (void)fprintf(stderr,
 472             "page size %lu, cacheing %lu pages of %lu page max cache\n",
 473             mp->pagesize, mp->curcache, mp->maxcache);
 474         (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n",
 475             mp->pageput, mp->pageget, mp->pagenew);
 476         (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n",
 477             mp->pagealloc, mp->pageflush);
 478         if (mp->cachehit + mp->cachemiss)
 479                 (void)fprintf(stderr,
 480                     "%.0f%% cache hit rate (%lu hits, %lu misses)\n",
 481                     ((double)mp->cachehit / (mp->cachehit + mp->cachemiss))
 482                     * 100, mp->cachehit, mp->cachemiss);
 483         (void)fprintf(stderr, "%lu page reads, %lu page writes\n",
 484             mp->pageread, mp->pagewrite);
 485
 486         sep = "";
 487         cnt = 0;
 488         for (b = mp->lru.cnext; b != (BKT *)&mp->lru; b = b->cnext) {
 489                 (void)fprintf(stderr, "%s%d", sep, b->pgno);
 490                 if (b->flags & MPOOL_DIRTY)
 491                         (void)fprintf(stderr, "d");
 492                 if (b->flags & MPOOL_PINNED)
 493                         (void)fprintf(stderr, "P");
 494                 if (++cnt == 10) {
 495                         sep = "\n";
 496                         cnt = 0;
 497                 } else
 498                         sep = ", ";
 499
 500         }
 501         (void)fprintf(stderr, "\n");
 502 }
 503 #endif
 504
 505 #ifdef DEBUG
 506 #if __STDC__
 507 #include <stdarg.h>
 508 #else
 509 #include <varargs.h>
 510 #endif
 511
 512 static void
 513 #if __STDC__
 514 err(const char *fmt, ...)
 515 #else
 516 err(fmt, va_alist)
 517         char *fmt;
 518         va_dcl
 519 #endif
 520 {
 521         va_list ap;
 522 #if __STDC__
 523         va_start(ap, fmt);
 524 #else
 525         va_start(ap);
 526 #endif
 527         (void)vfprintf(stderr, fmt, ap);
 528         va_end(ap);
 529         (void)fprintf(stderr, "\n");
 530         abort();
 531         /* NOTREACHED */
 532 }
 533 #endif