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 1.3 2003/11/12 20:21:23 eirikn Exp $ 35 * 36 * @(#)mpool.c 8.5 (Berkeley) 7/26/94 37 */ 38 39 #include <sys/param.h> 40 #include <sys/queue.h> 41 #include <sys/stat.h> 42 43 #include <errno.h> 44 #include <stdio.h> 45 #include <stdlib.h> 46 #include <string.h> 47 #include <unistd.h> 48 49 #include <db.h> 50 51 #define __MPOOLINTERFACE_PRIVATE 52 #include <mpool.h> 53 54 static BKT *mpool_bkt (MPOOL *); 55 static BKT *mpool_look (MPOOL *, pgno_t); 56 static int mpool_write (MPOOL *, BKT *); 57 58 /* 59 * mpool_open -- 60 * Initialize a memory pool. 61 */ 62 MPOOL * 63 mpool_open(key, fd, pagesize, maxcache) 64 void *key; 65 int fd; 66 pgno_t pagesize, maxcache; 67 { 68 struct stat sb; 69 MPOOL *mp; 70 int entry; 71 72 /* 73 * Get information about the file. 74 * 75 * XXX 76 * We don't currently handle pipes, although we should. 77 */ 78 if (fstat(fd, &sb)) 79 return (NULL); 80 if (!S_ISREG(sb.st_mode)) { 81 errno = ESPIPE; 82 return (NULL); 83 } 84 85 /* Allocate and initialize the MPOOL cookie. */ 86 if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL) 87 return (NULL); 88 TAILQ_INIT(&mp->lqh); 89 for (entry = 0; entry < HASHSIZE; ++entry) 90 TAILQ_INIT(&mp->hqh[entry]); 91 mp->maxcache = maxcache; 92 mp->npages = sb.st_size / pagesize; 93 mp->pagesize = pagesize; 94 mp->fd = fd; 95 return (mp); 96 } 97 98 /* 99 * mpool_filter -- 100 * Initialize input/output filters. 101 */ 102 void 103 mpool_filter(mp, pgin, pgout, pgcookie) 104 MPOOL *mp; 105 void (*pgin) (void *, pgno_t, void *); 106 void (*pgout) (void *, pgno_t, void *); 107 void *pgcookie; 108 { 109 mp->pgin = pgin; 110 mp->pgout = pgout; 111 mp->pgcookie = pgcookie; 112 } 113 114 /* 115 * mpool_new -- 116 * Get a new page of memory. 117 */ 118 void * 119 mpool_new(mp, pgnoaddr) 120 MPOOL *mp; 121 pgno_t *pgnoaddr; 122 { 123 struct _hqh *head; 124 BKT *bp; 125 126 if (mp->npages == MAX_PAGE_NUMBER) { 127 (void)fprintf(stderr, "mpool_new: page allocation overflow.\n"); 128 abort(); 129 } 130 #ifdef STATISTICS 131 ++mp->pagenew; 132 #endif 133 /* 134 * Get a BKT from the cache. Assign a new page number, attach 135 * it to the head of the hash chain, the tail of the lru chain, 136 * and return. 137 */ 138 if ((bp = mpool_bkt(mp)) == NULL) 139 return (NULL); 140 *pgnoaddr = bp->pgno = mp->npages++; 141 bp->flags = MPOOL_PINNED; 142 143 head = &mp->hqh[HASHKEY(bp->pgno)]; 144 TAILQ_INSERT_HEAD(head, bp, hq); 145 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 146 return (bp->page); 147 } 148 149 /* 150 * mpool_get 151 * Get a page. 152 */ 153 void * 154 mpool_get(mp, pgno, flags) 155 MPOOL *mp; 156 pgno_t pgno; 157 u_int flags; /* XXX not used? */ 158 { 159 struct _hqh *head; 160 BKT *bp; 161 off_t off; 162 int nr; 163 164 /* Check for attempt to retrieve a non-existent page. */ 165 if (pgno >= mp->npages) { 166 errno = EINVAL; 167 return (NULL); 168 } 169 170 #ifdef STATISTICS 171 ++mp->pageget; 172 #endif 173 174 /* Check for a page that is cached. */ 175 if ((bp = mpool_look(mp, pgno)) != NULL) { 176 #ifdef DEBUG 177 if (bp->flags & MPOOL_PINNED) { 178 (void)fprintf(stderr, 179 "mpool_get: page %d already pinned\n", bp->pgno); 180 abort(); 181 } 182 #endif 183 /* 184 * Move the page to the head of the hash chain and the tail 185 * of the lru chain. 186 */ 187 head = &mp->hqh[HASHKEY(bp->pgno)]; 188 TAILQ_REMOVE(head, bp, hq); 189 TAILQ_INSERT_HEAD(head, bp, hq); 190 TAILQ_REMOVE(&mp->lqh, bp, q); 191 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 192 193 /* Return a pinned page. */ 194 bp->flags |= MPOOL_PINNED; 195 return (bp->page); 196 } 197 198 /* Get a page from the cache. */ 199 if ((bp = mpool_bkt(mp)) == NULL) 200 return (NULL); 201 202 /* Read in the contents. */ 203 #ifdef STATISTICS 204 ++mp->pageread; 205 #endif 206 off = mp->pagesize * pgno; 207 if (lseek(mp->fd, off, SEEK_SET) != off) 208 return (NULL); 209 if ((nr = _read(mp->fd, bp->page, mp->pagesize)) != 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 TAILQ_INSERT_HEAD(head, bp, hq); 225 TAILQ_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 (!TAILQ_EMPTY(&mp->lqh)) { 274 bp = TAILQ_FIRST(&mp->lqh); 275 TAILQ_REMOVE(&mp->lqh, bp, q); 276 free(bp); 277 } 278 279 /* Free the MPOOL cookie. */ 280 free(mp); 281 return (RET_SUCCESS); 282 } 283 284 /* 285 * mpool_sync 286 * Sync the pool to disk. 287 */ 288 int 289 mpool_sync(mp) 290 MPOOL *mp; 291 { 292 BKT *bp; 293 294 /* Walk the lru chain, flushing any dirty pages to disk. */ 295 TAILQ_FOREACH(bp, &mp->lqh, q) 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 TAILQ_FOREACH(bp, &mp->lqh, q) 326 if (!(bp->flags & MPOOL_PINNED)) { 327 /* Flush if dirty. */ 328 if (bp->flags & MPOOL_DIRTY && 329 mpool_write(mp, bp) == RET_ERROR) 330 return (NULL); 331 #ifdef STATISTICS 332 ++mp->pageflush; 333 #endif 334 /* Remove from the hash and lru queues. */ 335 head = &mp->hqh[HASHKEY(bp->pgno)]; 336 TAILQ_REMOVE(head, bp, hq); 337 TAILQ_REMOVE(&mp->lqh, bp, q); 338 #ifdef DEBUG 339 { void *spage; 340 spage = bp->page; 341 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 342 bp->page = spage; 343 } 344 #endif 345 return (bp); 346 } 347 348 new: if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL) 349 return (NULL); 350 #ifdef STATISTICS 351 ++mp->pagealloc; 352 #endif 353 #if defined(DEBUG) || defined(PURIFY) 354 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 355 #endif 356 bp->page = (char *)bp + sizeof(BKT); 357 ++mp->curcache; 358 return (bp); 359 } 360 361 /* 362 * mpool_write 363 * Write a page to disk. 364 */ 365 static int 366 mpool_write(mp, bp) 367 MPOOL *mp; 368 BKT *bp; 369 { 370 off_t off; 371 372 #ifdef STATISTICS 373 ++mp->pagewrite; 374 #endif 375 376 /* Run through the user's filter. */ 377 if (mp->pgout) 378 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page); 379 380 off = mp->pagesize * bp->pgno; 381 if (lseek(mp->fd, off, SEEK_SET) != off) 382 return (RET_ERROR); 383 if (_write(mp->fd, bp->page, mp->pagesize) != mp->pagesize) 384 return (RET_ERROR); 385 386 bp->flags &= ~MPOOL_DIRTY; 387 return (RET_SUCCESS); 388 } 389 390 /* 391 * mpool_look 392 * Lookup a page in the cache. 393 */ 394 static BKT * 395 mpool_look(mp, pgno) 396 MPOOL *mp; 397 pgno_t pgno; 398 { 399 struct _hqh *head; 400 BKT *bp; 401 402 head = &mp->hqh[HASHKEY(pgno)]; 403 TAILQ_FOREACH(bp, head, hq) 404 if (bp->pgno == pgno) { 405 #ifdef STATISTICS 406 ++mp->cachehit; 407 #endif 408 return (bp); 409 } 410 #ifdef STATISTICS 411 ++mp->cachemiss; 412 #endif 413 return (NULL); 414 } 415 416 #ifdef STATISTICS 417 /* 418 * mpool_stat 419 * Print out cache statistics. 420 */ 421 void 422 mpool_stat(mp) 423 MPOOL *mp; 424 { 425 BKT *bp; 426 int cnt; 427 char *sep; 428 429 (void)fprintf(stderr, "%lu pages in the file\n", mp->npages); 430 (void)fprintf(stderr, 431 "page size %lu, cacheing %lu pages of %lu page max cache\n", 432 mp->pagesize, mp->curcache, mp->maxcache); 433 (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 434 mp->pageput, mp->pageget, mp->pagenew); 435 (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n", 436 mp->pagealloc, mp->pageflush); 437 if (mp->cachehit + mp->cachemiss) 438 (void)fprintf(stderr, 439 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 440 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 441 * 100, mp->cachehit, mp->cachemiss); 442 (void)fprintf(stderr, "%lu page reads, %lu page writes\n", 443 mp->pageread, mp->pagewrite); 444 445 sep = ""; 446 cnt = 0; 447 TAILQ_FOREACH(bp, &mp->lqh, q) { 448 (void)fprintf(stderr, "%s%d", sep, bp->pgno); 449 if (bp->flags & MPOOL_DIRTY) 450 (void)fprintf(stderr, "d"); 451 if (bp->flags & MPOOL_PINNED) 452 (void)fprintf(stderr, "P"); 453 if (++cnt == 10) { 454 sep = "\n"; 455 cnt = 0; 456 } else 457 sep = ", "; 458 459 } 460 (void)fprintf(stderr, "\n"); 461 } 462 #endif 463