db/mpool/mpool-fbsd.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 #include <sys/cdefs.h>
  38 __FBSDID("$FreeBSD: src/lib/libc/db/mpool/mpool.c,v 1.12 2004/09/10 05:41:41 kuriyama Exp $");
  39
  40 #include "namespace.h"
  41 #include <sys/param.h>
  42 #include <sys/queue.h>
  43 #include <sys/stat.h>
  44
  45 #include <errno.h>
  46 #include <stdio.h>
  47 #include <stdlib.h>
  48 #include <string.h>
  49 #include <unistd.h>
  50 #include "un-namespace.h"
  51
  52 #include <db.h>
  53
  54 #define __MPOOLINTERFACE_PRIVATE
  55 #include <mpool.h>
  56
  57 static BKT *mpool_bkt(MPOOL *);
  58 static BKT *mpool_look(MPOOL *, pgno_t);
  59 static int  mpool_write(MPOOL *, BKT *);
  60
  61 /*
  62  * mpool_open --
  63  *      Initialize a memory pool.
  64  */
  65 MPOOL *
  66 mpool_open(key, fd, pagesize, maxcache)
  67         void *key;
  68         int fd;
  69         pgno_t pagesize, maxcache;
  70 {
  71         struct stat sb;
  72         MPOOL *mp;
  73         int entry;
  74
  75         /*
  76          * Get information about the file.
  77          *
  78          * XXX
  79          * We don't currently handle pipes, although we should.
  80          */
  81         if (_fstat(fd, &sb))
  82                 return (NULL);
  83         if (!S_ISREG(sb.st_mode)) {
  84                 errno = ESPIPE;
  85                 return (NULL);
  86         }
  87
  88         /* Allocate and initialize the MPOOL cookie. */
  89         if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL)
  90                 return (NULL);
  91         TAILQ_INIT(&mp->lqh);
  92         for (entry = 0; entry < HASHSIZE; ++entry)
  93                 TAILQ_INIT(&mp->hqh[entry]);
  94         mp->maxcache = maxcache;
  95         mp->npages = sb.st_size / pagesize;
  96         mp->pagesize = pagesize;
  97         mp->fd = fd;
  98         return (mp);
  99 }
 100
 101 /*
 102  * mpool_filter --
 103  *      Initialize input/output filters.
 104  */
 105 void
 106 mpool_filter(mp, pgin, pgout, pgcookie)
 107         MPOOL *mp;
 108         void (*pgin)(void *, pgno_t, void *);
 109         void (*pgout)(void *, pgno_t, void *);
 110         void *pgcookie;
 111 {
 112         mp->pgin = pgin;
 113         mp->pgout = pgout;
 114         mp->pgcookie = pgcookie;
 115 }
 116
 117 /*
 118  * mpool_new --
 119  *      Get a new page of memory.
 120  */
 121 void *
 122 mpool_new(mp, pgnoaddr)
 123         MPOOL *mp;
 124         pgno_t *pgnoaddr;
 125 {
 126         struct _hqh *head;
 127         BKT *bp;
 128
 129         if (mp->npages == MAX_PAGE_NUMBER) {
 130                 (void)fprintf(stderr, "mpool_new: page allocation overflow.\n");
 131                 LIBC_ABORT("page allocation overflow");
 132         }
 133 #ifdef STATISTICS
 134         ++mp->pagenew;
 135 #endif
 136         /*
 137          * Get a BKT from the cache.  Assign a new page number, attach
 138          * it to the head of the hash chain, the tail of the lru chain,
 139          * and return.
 140          */
 141         if ((bp = mpool_bkt(mp)) == NULL)
 142                 return (NULL);
 143         *pgnoaddr = bp->pgno = mp->npages++;
 144         bp->flags = MPOOL_PINNED;
 145
 146         head = &mp->hqh[HASHKEY(bp->pgno)];
 147         TAILQ_INSERT_HEAD(head, bp, hq);
 148         TAILQ_INSERT_TAIL(&mp->lqh, bp, q);
 149         return (bp->page);
 150 }
 151
 152 /*
 153  * mpool_get
 154  *      Get a page.
 155  */
 156 void *
 157 mpool_get(mp, pgno, flags)
 158         MPOOL *mp;
 159         pgno_t pgno;
 160         u_int flags;                            /* XXX not used? */
 161 {
 162         struct _hqh *head;
 163         BKT *bp;
 164         off_t off;
 165         int nr;
 166
 167         /* Check for attempt to retrieve a non-existent page. */
 168         if (pgno >= mp->npages) {
 169                 errno = EINVAL;
 170                 return (NULL);
 171         }
 172
 173 #ifdef STATISTICS
 174         ++mp->pageget;
 175 #endif
 176
 177         /* Check for a page that is cached. */
 178         if ((bp = mpool_look(mp, pgno)) != NULL) {
 179 #ifdef DEBUG
 180                 if (bp->flags & MPOOL_PINNED) {
 181                         (void)fprintf(stderr,
 182                             "mpool_get: page %d already pinned\n", bp->pgno);
 183                         LIBC_ABORT("page %d already pinned", bp->pgno);
 184                 }
 185 #endif
 186                 /*
 187                  * Move the page to the head of the hash chain and the tail
 188                  * of the lru chain.
 189                  */
 190                 head = &mp->hqh[HASHKEY(bp->pgno)];
 191                 TAILQ_REMOVE(head, bp, hq);
 192                 TAILQ_INSERT_HEAD(head, bp, hq);
 193                 TAILQ_REMOVE(&mp->lqh, bp, q);
 194                 TAILQ_INSERT_TAIL(&mp->lqh, bp, q);
 195
 196                 /* Return a pinned page. */
 197                 bp->flags |= MPOOL_PINNED;
 198                 return (bp->page);
 199         }
 200
 201         /* Get a page from the cache. */
 202         if ((bp = mpool_bkt(mp)) == NULL)
 203                 return (NULL);
 204
 205         /* Read in the contents. */
 206 #ifdef STATISTICS
 207         ++mp->pageread;
 208 #endif
 209         off = mp->pagesize * pgno;
 210         nr = pread(mp->fd, bp->page, mp->pagesize, off);
 211         if (nr != mp->pagesize) {
 212                 if (nr >= 0)
 213                         errno = EFTYPE;
 214                 return (NULL);
 215         }
 216
 217         /* Set the page number, pin the page. */
 218         bp->pgno = pgno;
 219         bp->flags = MPOOL_PINNED;
 220
 221         /*
 222          * Add the page to the head of the hash chain and the tail
 223          * of the lru chain.
 224          */
 225         head = &mp->hqh[HASHKEY(bp->pgno)];
 226         TAILQ_INSERT_HEAD(head, bp, hq);
 227         TAILQ_INSERT_TAIL(&mp->lqh, bp, q);
 228
 229         /* Run through the user's filter. */
 230         if (mp->pgin != NULL)
 231                 (mp->pgin)(mp->pgcookie, bp->pgno, bp->page);
 232
 233         return (bp->page);
 234 }
 235
 236 /*
 237  * mpool_put
 238  *      Return a page.
 239  */
 240 int
 241 mpool_put(mp, page, flags)
 242         MPOOL *mp;
 243         void *page;
 244         u_int flags;
 245 {
 246         BKT *bp;
 247
 248 #ifdef STATISTICS
 249         ++mp->pageput;
 250 #endif
 251         bp = (BKT *)((char *)page - sizeof(BKT));
 252 #ifdef DEBUG
 253         if (!(bp->flags & MPOOL_PINNED)) {
 254                 (void)fprintf(stderr,
 255                     "mpool_put: page %d not pinned\n", bp->pgno);
 256                 LIBC_ABORT("page %d not pinned", bp->pgno);
 257         }
 258 #endif
 259         bp->flags &= ~MPOOL_PINNED;
 260         bp->flags |= flags & MPOOL_DIRTY;
 261         return (RET_SUCCESS);
 262 }
 263
 264 /*
 265  * mpool_close
 266  *      Close the buffer pool.
 267  */
 268 int
 269 mpool_close(mp)
 270         MPOOL *mp;
 271 {
 272         BKT *bp;
 273
 274         /* Free up any space allocated to the lru pages. */
 275         while (!TAILQ_EMPTY(&mp->lqh)) {
 276                 bp = TAILQ_FIRST(&mp->lqh);
 277                 TAILQ_REMOVE(&mp->lqh, bp, q);
 278                 free(bp);
 279         }
 280
 281         /* Free the MPOOL cookie. */
 282         free(mp);
 283         return (RET_SUCCESS);
 284 }
 285
 286 /*
 287  * mpool_sync
 288  *      Sync the pool to disk.
 289  */
 290 int
 291 mpool_sync(mp)
 292         MPOOL *mp;
 293 {
 294         BKT *bp;
 295
 296         /* Walk the lru chain, flushing any dirty pages to disk. */
 297         TAILQ_FOREACH(bp, &mp->lqh, q) {
 298                 if (bp->flags & MPOOL_DIRTY)
 299                         if (mpool_write(mp, bp) == RET_ERROR) {
 300                                 return (RET_ERROR);
 301                         } else {
 302                                 /* 4874757: Re-run through the user's pgin filter. */
 303                                 if (mp->pgin != NULL)
 304                                         (mp->pgin)(mp->pgcookie, bp->pgno, bp->page);
 305                         }
 306         }
 307
 308         /* Sync the file descriptor. */
 309         return (_fsync(mp->fd) ? RET_ERROR : RET_SUCCESS);
 310 }
 311
 312 /*
 313  * mpool_bkt
 314  *      Get a page from the cache (or create one).
 315  */
 316 static BKT *
 317 mpool_bkt(mp)
 318         MPOOL *mp;
 319 {
 320         struct _hqh *head;
 321         BKT *bp;
 322
 323         /* If under the max cached, always create a new page. */
 324         if (mp->curcache < mp->maxcache)
 325                 goto new;
 326
 327         /*
 328          * If the cache is max'd out, walk the lru list for a buffer we
 329          * can flush.  If we find one, write it (if necessary) and take it
 330          * off any lists.  If we don't find anything we grow the cache anyway.
 331          * The cache never shrinks.
 332          */
 333         TAILQ_FOREACH(bp, &mp->lqh, q)
 334                 if (!(bp->flags & MPOOL_PINNED)) {
 335                         /* Flush if dirty. */
 336                         if (bp->flags & MPOOL_DIRTY &&
 337                             mpool_write(mp, bp) == RET_ERROR)
 338                                 return (NULL);
 339 #ifdef STATISTICS
 340                         ++mp->pageflush;
 341 #endif
 342                         /* Remove from the hash and lru queues. */
 343                         head = &mp->hqh[HASHKEY(bp->pgno)];
 344                         TAILQ_REMOVE(head, bp, hq);
 345                         TAILQ_REMOVE(&mp->lqh, bp, q);
 346 #ifdef DEBUG
 347                         { void *spage;
 348                                 spage = bp->page;
 349                                 memset(bp, 0xff, sizeof(BKT) + mp->pagesize);
 350                                 bp->page = spage;
 351                         }
 352 #endif
 353                         return (bp);
 354                 }
 355
 356 new:    if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL)
 357                 return (NULL);
 358 #ifdef STATISTICS
 359         ++mp->pagealloc;
 360 #endif
 361 #if defined(DEBUG) || defined(PURIFY)
 362         memset(bp, 0xff, sizeof(BKT) + mp->pagesize);
 363 #endif
 364         bp->page = (char *)bp + sizeof(BKT);
 365         ++mp->curcache;
 366         return (bp);
 367 }
 368
 369 /*
 370  * mpool_write
 371  *      Write a page to disk.
 372  */
 373 static int
 374 mpool_write(mp, bp)
 375         MPOOL *mp;
 376         BKT *bp;
 377 {
 378         off_t off;
 379
 380 #ifdef STATISTICS
 381         ++mp->pagewrite;
 382 #endif
 383
 384         /* Run through the user's filter. */
 385         if (mp->pgout)
 386                 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page);
 387
 388         off = mp->pagesize * bp->pgno;
 389         if (pwrite(mp->fd, bp->page, mp->pagesize, off) != mp->pagesize)
 390                 return (RET_ERROR);
 391
 392         bp->flags &= ~MPOOL_DIRTY;
 393         return (RET_SUCCESS);
 394 }
 395
 396 /*
 397  * mpool_look
 398  *      Lookup a page in the cache.
 399  */
 400 static BKT *
 401 mpool_look(mp, pgno)
 402         MPOOL *mp;
 403         pgno_t pgno;
 404 {
 405         struct _hqh *head;
 406         BKT *bp;
 407
 408         head = &mp->hqh[HASHKEY(pgno)];
 409         TAILQ_FOREACH(bp, head, hq)
 410                 if (bp->pgno == pgno) {
 411 #ifdef STATISTICS
 412                         ++mp->cachehit;
 413 #endif
 414                         return (bp);
 415                 }
 416 #ifdef STATISTICS
 417         ++mp->cachemiss;
 418 #endif
 419         return (NULL);
 420 }
 421
 422 #ifdef STATISTICS
 423 /*
 424  * mpool_stat
 425  *      Print out cache statistics.
 426  */
 427 void
 428 mpool_stat(mp)
 429         MPOOL *mp;
 430 {
 431         BKT *bp;
 432         int cnt;
 433         char *sep;
 434
 435         (void)fprintf(stderr, "%u pages in the file\n", mp->npages);
 436         (void)fprintf(stderr,
 437             "page size %lu, cacheing %u pages of %u page max cache\n",
 438             mp->pagesize, mp->curcache, mp->maxcache);
 439         (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n",
 440             mp->pageput, mp->pageget, mp->pagenew);
 441         (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n",
 442             mp->pagealloc, mp->pageflush);
 443         if (mp->cachehit + mp->cachemiss)
 444                 (void)fprintf(stderr,
 445                     "%.0f%% cache hit rate (%lu hits, %lu misses)\n",
 446                     ((double)mp->cachehit / (mp->cachehit + mp->cachemiss))
 447                     * 100, mp->cachehit, mp->cachemiss);
 448         (void)fprintf(stderr, "%lu page reads, %lu page writes\n",
 449             mp->pageread, mp->pagewrite);
 450
 451         sep = "";
 452         cnt = 0;
 453         TAILQ_FOREACH(bp, &mp->lqh, q) {
 454                 (void)fprintf(stderr, "%s%d", sep, bp->pgno);
 455                 if (bp->flags & MPOOL_DIRTY)
 456                         (void)fprintf(stderr, "d");
 457                 if (bp->flags & MPOOL_PINNED)
 458                         (void)fprintf(stderr, "P");
 459                 if (++cnt == 10) {
 460                         sep = "\n";
 461                         cnt = 0;
 462                 } else
 463                         sep = ", ";
 464
 465         }
 466         (void)fprintf(stderr, "\n");
 467 }
 468 #endif