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 * $FreeBSD: src/lib/libc/db/mpool/mpool.c,v 1.5.2.1 2001/03/05 23:05:01 obrien Exp $ 34 * $DragonFly: src/lib/libc/db/mpool/mpool.c,v 1.4 2005/01/31 22:29:11 dillon Exp $ 35 * 36 * @(#)mpool.c 8.5 (Berkeley) 7/26/94 37 */ 38 39 #include "namespace.h" 40 #include <sys/param.h> 41 #include <sys/queue.h> 42 #include <sys/stat.h> 43 44 #include <errno.h> 45 #include <stdio.h> 46 #include <stdlib.h> 47 #include <string.h> 48 #include <unistd.h> 49 #include "un-namespace.h" 50 51 #include <db.h> 52 53 #define __MPOOLINTERFACE_PRIVATE 54 #include <mpool.h> 55 56 static BKT *mpool_bkt (MPOOL *); 57 static BKT *mpool_look (MPOOL *, pgno_t); 58 static int mpool_write (MPOOL *, BKT *); 59 60 /* 61 * mpool_open -- 62 * Initialize a memory pool. 63 */ 64 MPOOL * 65 mpool_open(key, fd, pagesize, maxcache) 66 void *key; 67 int fd; 68 pgno_t pagesize, maxcache; 69 { 70 struct stat sb; 71 MPOOL *mp; 72 int entry; 73 74 /* 75 * Get information about the file. 76 * 77 * XXX 78 * We don't currently handle pipes, although we should. 79 */ 80 if (_fstat(fd, &sb)) 81 return (NULL); 82 if (!S_ISREG(sb.st_mode)) { 83 errno = ESPIPE; 84 return (NULL); 85 } 86 87 /* Allocate and initialize the MPOOL cookie. */ 88 if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL) 89 return (NULL); 90 TAILQ_INIT(&mp->lqh); 91 for (entry = 0; entry < HASHSIZE; ++entry) 92 TAILQ_INIT(&mp->hqh[entry]); 93 mp->maxcache = maxcache; 94 mp->npages = sb.st_size / pagesize; 95 mp->pagesize = pagesize; 96 mp->fd = fd; 97 return (mp); 98 } 99 100 /* 101 * mpool_filter -- 102 * Initialize input/output filters. 103 */ 104 void 105 mpool_filter(mp, pgin, pgout, pgcookie) 106 MPOOL *mp; 107 void (*pgin) (void *, pgno_t, void *); 108 void (*pgout) (void *, pgno_t, void *); 109 void *pgcookie; 110 { 111 mp->pgin = pgin; 112 mp->pgout = pgout; 113 mp->pgcookie = pgcookie; 114 } 115 116 /* 117 * mpool_new -- 118 * Get a new page of memory. 119 */ 120 void * 121 mpool_new(mp, pgnoaddr) 122 MPOOL *mp; 123 pgno_t *pgnoaddr; 124 { 125 struct _hqh *head; 126 BKT *bp; 127 128 if (mp->npages == MAX_PAGE_NUMBER) { 129 (void)fprintf(stderr, "mpool_new: page allocation overflow.\n"); 130 abort(); 131 } 132 #ifdef STATISTICS 133 ++mp->pagenew; 134 #endif 135 /* 136 * Get a BKT from the cache. Assign a new page number, attach 137 * it to the head of the hash chain, the tail of the lru chain, 138 * and return. 139 */ 140 if ((bp = mpool_bkt(mp)) == NULL) 141 return (NULL); 142 *pgnoaddr = bp->pgno = mp->npages++; 143 bp->flags = MPOOL_PINNED; 144 145 head = &mp->hqh[HASHKEY(bp->pgno)]; 146 TAILQ_INSERT_HEAD(head, bp, hq); 147 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 148 return (bp->page); 149 } 150 151 /* 152 * mpool_get 153 * Get a page. 154 */ 155 void * 156 mpool_get(mp, pgno, flags) 157 MPOOL *mp; 158 pgno_t pgno; 159 u_int flags; /* XXX not used? */ 160 { 161 struct _hqh *head; 162 BKT *bp; 163 off_t off; 164 int nr; 165 166 /* Check for attempt to retrieve a non-existent page. */ 167 if (pgno >= mp->npages) { 168 errno = EINVAL; 169 return (NULL); 170 } 171 172 #ifdef STATISTICS 173 ++mp->pageget; 174 #endif 175 176 /* Check for a page that is cached. */ 177 if ((bp = mpool_look(mp, pgno)) != NULL) { 178 #ifdef DEBUG 179 if (bp->flags & MPOOL_PINNED) { 180 (void)fprintf(stderr, 181 "mpool_get: page %d already pinned\n", bp->pgno); 182 abort(); 183 } 184 #endif 185 /* 186 * Move the page to the head of the hash chain and the tail 187 * of the lru chain. 188 */ 189 head = &mp->hqh[HASHKEY(bp->pgno)]; 190 TAILQ_REMOVE(head, bp, hq); 191 TAILQ_INSERT_HEAD(head, bp, hq); 192 TAILQ_REMOVE(&mp->lqh, bp, q); 193 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 194 195 /* Return a pinned page. */ 196 bp->flags |= MPOOL_PINNED; 197 return (bp->page); 198 } 199 200 /* Get a page from the cache. */ 201 if ((bp = mpool_bkt(mp)) == NULL) 202 return (NULL); 203 204 /* Read in the contents. */ 205 #ifdef STATISTICS 206 ++mp->pageread; 207 #endif 208 off = mp->pagesize * pgno; 209 if (lseek(mp->fd, off, SEEK_SET) != off) 210 return (NULL); 211 if ((nr = _read(mp->fd, bp->page, mp->pagesize)) != 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 abort(); 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 mpool_write(mp, bp) == RET_ERROR) 300 return (RET_ERROR); 301 302 /* Sync the file descriptor. */ 303 return (_fsync(mp->fd) ? RET_ERROR : RET_SUCCESS); 304 } 305 306 /* 307 * mpool_bkt 308 * Get a page from the cache (or create one). 309 */ 310 static BKT * 311 mpool_bkt(mp) 312 MPOOL *mp; 313 { 314 struct _hqh *head; 315 BKT *bp; 316 317 /* If under the max cached, always create a new page. */ 318 if (mp->curcache < mp->maxcache) 319 goto new; 320 321 /* 322 * If the cache is max'd out, walk the lru list for a buffer we 323 * can flush. If we find one, write it (if necessary) and take it 324 * off any lists. If we don't find anything we grow the cache anyway. 325 * The cache never shrinks. 326 */ 327 TAILQ_FOREACH(bp, &mp->lqh, q) 328 if (!(bp->flags & MPOOL_PINNED)) { 329 /* Flush if dirty. */ 330 if (bp->flags & MPOOL_DIRTY && 331 mpool_write(mp, bp) == RET_ERROR) 332 return (NULL); 333 #ifdef STATISTICS 334 ++mp->pageflush; 335 #endif 336 /* Remove from the hash and lru queues. */ 337 head = &mp->hqh[HASHKEY(bp->pgno)]; 338 TAILQ_REMOVE(head, bp, hq); 339 TAILQ_REMOVE(&mp->lqh, bp, q); 340 #ifdef DEBUG 341 { void *spage; 342 spage = bp->page; 343 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 344 bp->page = spage; 345 } 346 #endif 347 return (bp); 348 } 349 350 new: if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL) 351 return (NULL); 352 #ifdef STATISTICS 353 ++mp->pagealloc; 354 #endif 355 #if defined(DEBUG) || defined(PURIFY) 356 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 357 #endif 358 bp->page = (char *)bp + sizeof(BKT); 359 ++mp->curcache; 360 return (bp); 361 } 362 363 /* 364 * mpool_write 365 * Write a page to disk. 366 */ 367 static int 368 mpool_write(mp, bp) 369 MPOOL *mp; 370 BKT *bp; 371 { 372 off_t off; 373 374 #ifdef STATISTICS 375 ++mp->pagewrite; 376 #endif 377 378 /* Run through the user's filter. */ 379 if (mp->pgout) 380 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page); 381 382 off = mp->pagesize * bp->pgno; 383 if (lseek(mp->fd, off, SEEK_SET) != off) 384 return (RET_ERROR); 385 if (_write(mp->fd, bp->page, mp->pagesize) != mp->pagesize) 386 return (RET_ERROR); 387 388 bp->flags &= ~MPOOL_DIRTY; 389 return (RET_SUCCESS); 390 } 391 392 /* 393 * mpool_look 394 * Lookup a page in the cache. 395 */ 396 static BKT * 397 mpool_look(mp, pgno) 398 MPOOL *mp; 399 pgno_t pgno; 400 { 401 struct _hqh *head; 402 BKT *bp; 403 404 head = &mp->hqh[HASHKEY(pgno)]; 405 TAILQ_FOREACH(bp, head, hq) 406 if (bp->pgno == pgno) { 407 #ifdef STATISTICS 408 ++mp->cachehit; 409 #endif 410 return (bp); 411 } 412 #ifdef STATISTICS 413 ++mp->cachemiss; 414 #endif 415 return (NULL); 416 } 417 418 #ifdef STATISTICS 419 /* 420 * mpool_stat 421 * Print out cache statistics. 422 */ 423 void 424 mpool_stat(mp) 425 MPOOL *mp; 426 { 427 BKT *bp; 428 int cnt; 429 char *sep; 430 431 (void)fprintf(stderr, "%lu pages in the file\n", mp->npages); 432 (void)fprintf(stderr, 433 "page size %lu, cacheing %lu pages of %lu page max cache\n", 434 mp->pagesize, mp->curcache, mp->maxcache); 435 (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 436 mp->pageput, mp->pageget, mp->pagenew); 437 (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n", 438 mp->pagealloc, mp->pageflush); 439 if (mp->cachehit + mp->cachemiss) 440 (void)fprintf(stderr, 441 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 442 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 443 * 100, mp->cachehit, mp->cachemiss); 444 (void)fprintf(stderr, "%lu page reads, %lu page writes\n", 445 mp->pageread, mp->pagewrite); 446 447 sep = ""; 448 cnt = 0; 449 TAILQ_FOREACH(bp, &mp->lqh, q) { 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 465