db/mpool/mpool.c

   1 /*-
   2  * Copyright (c) 1990, 1993, 1994
   3  *      The Regents of the University of California.  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     8.5 (Berkeley) 7/26/94";
  36 #endif /* LIBC_SCCS and not lint */
  37
  38 #include <sys/param.h>
  39 #include <sys/queue.h>
  40 #include <sys/stat.h>
  41
  42 #include <errno.h>
  43 #include <stdio.h>
  44 #include <stdlib.h>
  45 #include <string.h>
  46 #include <unistd.h>
  47
  48 #include <db.h>
  49
  50 #define __MPOOLINTERFACE_PRIVATE
  51 #include <mpool.h>
  52
  53 static BKT *mpool_bkt __P((MPOOL *));
  54 static BKT *mpool_look __P((MPOOL *, pgno_t));
  55 static int  mpool_write __P((MPOOL *, BKT *));
  56
  57 /*
  58  * mpool_open --
  59  *      Initialize a memory pool.
  60  */
  61 MPOOL *
  62 mpool_open(key, fd, pagesize, maxcache)
  63         void *key;
  64         int fd;
  65         pgno_t pagesize, maxcache;
  66 {
  67         struct stat sb;
  68         MPOOL *mp;
  69         int entry;
  70
  71         /*
  72          * Get information about the file.
  73          *
  74          * XXX
  75          * We don't currently handle pipes, although we should.
  76          */
  77         if (fstat(fd, &sb))
  78                 return (NULL);
  79         if (!S_ISREG(sb.st_mode)) {
  80                 errno = ESPIPE;
  81                 return (NULL);
  82         }
  83
  84         /* Allocate and initialize the MPOOL cookie. */
  85         if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL)
  86                 return (NULL);
  87         CIRCLEQ_INIT(&mp->lqh);
  88         for (entry = 0; entry < HASHSIZE; ++entry)
  89                 CIRCLEQ_INIT(&mp->hqh[entry]);
  90         mp->maxcache = maxcache;
  91         mp->npages = sb.st_size / pagesize;
  92         mp->pagesize = pagesize;
  93         mp->fd = fd;
  94         return (mp);
  95 }
  96
  97 /*
  98  * mpool_filter --
  99  *      Initialize input/output filters.
 100  */
 101 void
 102 mpool_filter(mp, pgin, pgout, pgcookie)
 103         MPOOL *mp;
 104         void (*pgin) __P((void *, pgno_t, void *));
 105         void (*pgout) __P((void *, pgno_t, void *));
 106         void *pgcookie;
 107 {
 108         mp->pgin = pgin;
 109         mp->pgout = pgout;
 110         mp->pgcookie = pgcookie;
 111 }
 112
 113 /*
 114  * mpool_new --
 115  *      Get a new page of memory.
 116  */
 117 void *
 118 mpool_new(mp, pgnoaddr)
 119         MPOOL *mp;
 120         pgno_t *pgnoaddr;
 121 {
 122         struct _hqh *head;
 123         BKT *bp;
 124
 125         if (mp->npages == MAX_PAGE_NUMBER) {
 126                 (void)fprintf(stderr, "mpool_new: page allocation overflow.\n");
 127                 abort();
 128         }
 129 #ifdef STATISTICS
 130         ++mp->pagenew;
 131 #endif
 132         /*
 133          * Get a BKT from the cache.  Assign a new page number, attach
 134          * it to the head of the hash chain, the tail of the lru chain,
 135          * and return.
 136          */
 137         if ((bp = mpool_bkt(mp)) == NULL)
 138                 return (NULL);
 139         *pgnoaddr = bp->pgno = mp->npages++;
 140         bp->flags = MPOOL_PINNED;
 141
 142         head = &mp->hqh[HASHKEY(bp->pgno)];
 143         CIRCLEQ_INSERT_HEAD(head, bp, hq);
 144         CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q);
 145         return (bp->page);
 146 }
 147
 148 /*
 149  * mpool_get
 150  *      Get a page.
 151  */
 152 void *
 153 mpool_get(mp, pgno, flags)
 154         MPOOL *mp;
 155         pgno_t pgno;
 156         u_int flags;                            /* XXX not used? */
 157 {
 158         struct _hqh *head;
 159         BKT *bp;
 160         off_t off;
 161         int nr;
 162
 163         /* Check for attempt to retrieve a non-existent page. */
 164         if (pgno >= mp->npages) {
 165                 errno = EINVAL;
 166                 return (NULL);
 167         }
 168
 169 #ifdef STATISTICS
 170         ++mp->pageget;
 171 #endif
 172
 173         /* Check for a page that is cached. */
 174         if ((bp = mpool_look(mp, pgno)) != NULL) {
 175 #ifdef DEBUG
 176                 if (bp->flags & MPOOL_PINNED) {
 177                         (void)fprintf(stderr,
 178                             "mpool_get: page %d already pinned\n", bp->pgno);
 179                         abort();
 180                 }
 181 #endif
 182                 /*
 183                  * Move the page to the head of the hash chain and the tail
 184                  * of the lru chain.
 185                  */
 186                 head = &mp->hqh[HASHKEY(bp->pgno)];
 187                 CIRCLEQ_REMOVE(head, bp, hq);
 188                 CIRCLEQ_INSERT_HEAD(head, bp, hq);
 189                 CIRCLEQ_REMOVE(&mp->lqh, bp, q);
 190                 CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q);
 191
 192                 /* Return a pinned page. */
 193                 bp->flags |= MPOOL_PINNED;
 194                 return (bp->page);
 195         }
 196
 197         /* Get a page from the cache. */
 198         if ((bp = mpool_bkt(mp)) == NULL)
 199                 return (NULL);
 200
 201         /* Read in the contents. */
 202 #ifdef STATISTICS
 203         ++mp->pageread;
 204 #endif
 205         off = mp->pagesize * pgno;
 206         if (lseek(mp->fd, off, SEEK_SET) != off)
 207                 return (NULL);
 208         if ((u_long) (nr = read(mp->fd, bp->page, mp->pagesize))
 209             != mp->pagesize) {
 210                 if (nr >= 0)
 211                         errno = EFTYPE;
 212                 return (NULL);
 213         }
 214
 215         /* Set the page number, pin the page. */
 216         bp->pgno = pgno;
 217         bp->flags = MPOOL_PINNED;
 218
 219         /*
 220          * Add the page to the head of the hash chain and the tail
 221          * of the lru chain.
 222          */
 223         head = &mp->hqh[HASHKEY(bp->pgno)];
 224         CIRCLEQ_INSERT_HEAD(head, bp, hq);
 225         CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q);
 226
 227         /* Run through the user's filter. */
 228         if (mp->pgin != NULL)
 229                 (mp->pgin)(mp->pgcookie, bp->pgno, bp->page);
 230
 231         return (bp->page);
 232 }
 233
 234 /*
 235  * mpool_put
 236  *      Return a page.
 237  */
 238 int
 239 mpool_put(mp, page, flags)
 240         MPOOL *mp;
 241         void *page;
 242         u_int flags;
 243 {
 244         BKT *bp;
 245
 246 #ifdef STATISTICS
 247         ++mp->pageput;
 248 #endif
 249         bp = (BKT *)((char *)page - sizeof(BKT));
 250 #ifdef DEBUG
 251         if (!(bp->flags & MPOOL_PINNED)) {
 252                 (void)fprintf(stderr,
 253                     "mpool_put: page %d not pinned\n", bp->pgno);
 254                 abort();
 255         }
 256 #endif
 257         bp->flags &= ~MPOOL_PINNED;
 258         bp->flags |= flags & MPOOL_DIRTY;
 259         return (RET_SUCCESS);
 260 }
 261
 262 /*
 263  * mpool_close
 264  *      Close the buffer pool.
 265  */
 266 int
 267 mpool_close(mp)
 268         MPOOL *mp;
 269 {
 270         BKT *bp;
 271
 272         /* Free up any space allocated to the lru pages. */
 273         while ((bp = mp->lqh.cqh_first) != (void *)&mp->lqh) {
 274                 CIRCLEQ_REMOVE(&mp->lqh, mp->lqh.cqh_first, q);
 275                 free(bp);
 276         }
 277
 278         /* Free the MPOOL cookie. */
 279         free(mp);
 280         return (RET_SUCCESS);
 281 }
 282
 283 /*
 284  * mpool_sync
 285  *      Sync the pool to disk.
 286  */
 287 int
 288 mpool_sync(mp)
 289         MPOOL *mp;
 290 {
 291         BKT *bp;
 292
 293         /* Walk the lru chain, flushing any dirty pages to disk. */
 294         for (bp = mp->lqh.cqh_first;
 295             bp != (void *)&mp->lqh; bp = bp->q.cqe_next)
 296                 if (bp->flags & MPOOL_DIRTY &&
 297                     mpool_write(mp, bp) == RET_ERROR)
 298                         return (RET_ERROR);
 299
 300         /* Sync the file descriptor. */
 301         return (fsync(mp->fd) ? RET_ERROR : RET_SUCCESS);
 302 }
 303
 304 /*
 305  * mpool_bkt
 306  *      Get a page from the cache (or create one).
 307  */
 308 static BKT *
 309 mpool_bkt(mp)
 310         MPOOL *mp;
 311 {
 312         struct _hqh *head;
 313         BKT *bp;
 314
 315         /* If under the max cached, always create a new page. */
 316         if (mp->curcache < mp->maxcache)
 317                 goto new;
 318
 319         /*
 320          * If the cache is max'd out, walk the lru list for a buffer we
 321          * can flush.  If we find one, write it (if necessary) and take it
 322          * off any lists.  If we don't find anything we grow the cache anyway.
 323          * The cache never shrinks.
 324          */
 325         for (bp = mp->lqh.cqh_first;
 326             bp != (void *)&mp->lqh; bp = bp->q.cqe_next)
 327                 if (!(bp->flags & MPOOL_PINNED)) {
 328                         /* Flush if dirty. */
 329                         if (bp->flags & MPOOL_DIRTY &&
 330                             mpool_write(mp, bp) == RET_ERROR)
 331                                 return (NULL);
 332 #ifdef STATISTICS
 333                         ++mp->pageflush;
 334 #endif
 335                         /* Remove from the hash and lru queues. */
 336                         head = &mp->hqh[HASHKEY(bp->pgno)];
 337                         CIRCLEQ_REMOVE(head, bp, hq);
 338                         CIRCLEQ_REMOVE(&mp->lqh, bp, q);
 339 #ifdef DEBUG
 340                         { void *spage;
 341                                 spage = bp->page;
 342                                 memset(bp, 0xff, sizeof(BKT) + mp->pagesize);
 343                                 bp->page = spage;
 344                         }
 345 #endif
 346                         return (bp);
 347                 }
 348
 349 new:    if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL)
 350                 return (NULL);
 351 #ifdef STATISTICS
 352         ++mp->pagealloc;
 353 #endif
 354 #if defined(DEBUG) || defined(PURIFY)
 355         memset(bp, 0xff, sizeof(BKT) + mp->pagesize);
 356 #endif
 357         bp->page = (char *)bp + sizeof(BKT);
 358         ++mp->curcache;
 359         return (bp);
 360 }
 361
 362 /*
 363  * mpool_write
 364  *      Write a page to disk.
 365  */
 366 static int
 367 mpool_write(mp, bp)
 368         MPOOL *mp;
 369         BKT *bp;
 370 {
 371         off_t off;
 372
 373 #ifdef STATISTICS
 374         ++mp->pagewrite;
 375 #endif
 376
 377         /* Run through the user's filter. */
 378         if (mp->pgout)
 379                 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page);
 380
 381         off = mp->pagesize * bp->pgno;
 382         if (lseek(mp->fd, off, SEEK_SET) != off)
 383                 return (RET_ERROR);
 384         if ((u_long) write(mp->fd, bp->page, mp->pagesize) != mp->pagesize)
 385                 return (RET_ERROR);
 386
 387         bp->flags &= ~MPOOL_DIRTY;
 388         return (RET_SUCCESS);
 389 }
 390
 391 /*
 392  * mpool_look
 393  *      Lookup a page in the cache.
 394  */
 395 static BKT *
 396 mpool_look(mp, pgno)
 397         MPOOL *mp;
 398         pgno_t pgno;
 399 {
 400         struct _hqh *head;
 401         BKT *bp;
 402
 403         head = &mp->hqh[HASHKEY(pgno)];
 404         for (bp = head->cqh_first; bp != (void *)head; bp = bp->hq.cqe_next)
 405                 if (bp->pgno == pgno) {
 406 #ifdef STATISTICS
 407                         ++mp->cachehit;
 408 #endif
 409                         return (bp);
 410                 }
 411 #ifdef STATISTICS
 412         ++mp->cachemiss;
 413 #endif
 414         return (NULL);
 415 }
 416
 417 #ifdef STATISTICS
 418 /*
 419  * mpool_stat
 420  *      Print out cache statistics.
 421  */
 422 void
 423 mpool_stat(mp)
 424         MPOOL *mp;
 425 {
 426         BKT *bp;
 427         int cnt;
 428         char *sep;
 429
 430         (void)fprintf(stderr, "%lu pages in the file\n", mp->npages);
 431         (void)fprintf(stderr,
 432             "page size %lu, cacheing %lu pages of %lu page max cache\n",
 433             mp->pagesize, mp->curcache, mp->maxcache);
 434         (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n",
 435             mp->pageput, mp->pageget, mp->pagenew);
 436         (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n",
 437             mp->pagealloc, mp->pageflush);
 438         if (mp->cachehit + mp->cachemiss)
 439                 (void)fprintf(stderr,
 440                     "%.0f%% cache hit rate (%lu hits, %lu misses)\n",
 441                     ((double)mp->cachehit / (mp->cachehit + mp->cachemiss))
 442                     * 100, mp->cachehit, mp->cachemiss);
 443         (void)fprintf(stderr, "%lu page reads, %lu page writes\n",
 444             mp->pageread, mp->pagewrite);
 445
 446         sep = "";
 447         cnt = 0;
 448         for (bp = mp->lqh.cqh_first;
 449             bp != (void *)&mp->lqh; bp = bp->q.cqe_next) {
 450                 (void)fprintf(stderr, "%s%d", sep, bp->pgno);
 451                 if (bp->flags & MPOOL_DIRTY)
 452                         (void)fprintf(stderr, "d");
 453                 if (bp->flags & MPOOL_PINNED)
 454                         (void)fprintf(stderr, "P");
 455                 if (++cnt == 10) {
 456                         sep = "\n";
 457                         cnt = 0;
 458                 } else
 459                         sep = ", ";
 460
 461         }
 462         (void)fprintf(stderr, "\n");
 463 }
 464 #endif