1 /* $NetBSD: mpool.c,v 1.7 1997/07/13 18:52:09 christos Exp $ */ 2 3 /*- 4 * Copyright (c) 1990, 1993, 1994 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 */ 35 36 #include <sys/cdefs.h> 37 #if defined(LIBC_SCCS) && !defined(lint) 38 #if 0 39 static char sccsid[] = "@(#)mpool.c 8.5 (Berkeley) 7/26/94"; 40 #else 41 __RCSID("$NetBSD: mpool.c,v 1.7 1997/07/13 18:52:09 christos Exp $"); 42 #endif 43 #endif /* LIBC_SCCS and not lint */ 44 45 #include <sys/param.h> 46 #include <sys/queue.h> 47 #include <sys/stat.h> 48 49 #include <errno.h> 50 #include <stdio.h> 51 #include <stdlib.h> 52 #include <string.h> 53 #include <unistd.h> 54 55 #include <db.h> 56 57 #define __MPOOLINTERFACE_PRIVATE 58 #include <mpool.h> 59 60 static BKT *mpool_bkt __P((MPOOL *)); 61 static BKT *mpool_look __P((MPOOL *, pgno_t)); 62 static int mpool_write __P((MPOOL *, BKT *)); 63 64 /* 65 * mpool_open -- 66 * Initialize a memory pool. 67 */ 68 MPOOL * 69 mpool_open(key, fd, pagesize, maxcache) 70 void *key; 71 int fd; 72 pgno_t pagesize, maxcache; 73 { 74 struct stat sb; 75 MPOOL *mp; 76 int entry; 77 78 /* 79 * Get information about the file. 80 * 81 * XXX 82 * We don't currently handle pipes, although we should. 83 */ 84 if (fstat(fd, &sb)) 85 return (NULL); 86 if (!S_ISREG(sb.st_mode)) { 87 errno = ESPIPE; 88 return (NULL); 89 } 90 91 /* Allocate and initialize the MPOOL cookie. */ 92 if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL) 93 return (NULL); 94 CIRCLEQ_INIT(&mp->lqh); 95 for (entry = 0; entry < HASHSIZE; ++entry) 96 CIRCLEQ_INIT(&mp->hqh[entry]); 97 mp->maxcache = maxcache; 98 mp->npages = sb.st_size / pagesize; 99 mp->pagesize = pagesize; 100 mp->fd = fd; 101 return (mp); 102 } 103 104 /* 105 * mpool_filter -- 106 * Initialize input/output filters. 107 */ 108 void 109 mpool_filter(mp, pgin, pgout, pgcookie) 110 MPOOL *mp; 111 void (*pgin) __P((void *, pgno_t, void *)); 112 void (*pgout) __P((void *, pgno_t, void *)); 113 void *pgcookie; 114 { 115 mp->pgin = pgin; 116 mp->pgout = pgout; 117 mp->pgcookie = pgcookie; 118 } 119 120 /* 121 * mpool_new -- 122 * Get a new page of memory. 123 */ 124 void * 125 mpool_new(mp, pgnoaddr) 126 MPOOL *mp; 127 pgno_t *pgnoaddr; 128 { 129 struct _hqh *head; 130 BKT *bp; 131 132 if (mp->npages == MAX_PAGE_NUMBER) { 133 (void)fprintf(stderr, "mpool_new: page allocation overflow.\n"); 134 abort(); 135 } 136 #ifdef STATISTICS 137 ++mp->pagenew; 138 #endif 139 /* 140 * Get a BKT from the cache. Assign a new page number, attach 141 * it to the head of the hash chain, the tail of the lru chain, 142 * and return. 143 */ 144 if ((bp = mpool_bkt(mp)) == NULL) 145 return (NULL); 146 *pgnoaddr = bp->pgno = mp->npages++; 147 bp->flags = MPOOL_PINNED; 148 149 head = &mp->hqh[HASHKEY(bp->pgno)]; 150 CIRCLEQ_INSERT_HEAD(head, bp, hq); 151 CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q); 152 return (bp->page); 153 } 154 155 /* 156 * mpool_get 157 * Get a page. 158 */ 159 void * 160 mpool_get(mp, pgno, flags) 161 MPOOL *mp; 162 pgno_t pgno; 163 u_int flags; /* XXX not used? */ 164 { 165 struct _hqh *head; 166 BKT *bp; 167 off_t off; 168 int nr; 169 170 /* Check for attempt to retrieve a non-existent page. */ 171 if (pgno >= mp->npages) { 172 errno = EINVAL; 173 return (NULL); 174 } 175 176 #ifdef STATISTICS 177 ++mp->pageget; 178 #endif 179 180 /* Check for a page that is cached. */ 181 if ((bp = mpool_look(mp, pgno)) != NULL) { 182 #ifdef DEBUG 183 if (bp->flags & MPOOL_PINNED) { 184 (void)fprintf(stderr, 185 "mpool_get: page %d already pinned\n", bp->pgno); 186 abort(); 187 } 188 #endif 189 /* 190 * Move the page to the head of the hash chain and the tail 191 * of the lru chain. 192 */ 193 head = &mp->hqh[HASHKEY(bp->pgno)]; 194 CIRCLEQ_REMOVE(head, bp, hq); 195 CIRCLEQ_INSERT_HEAD(head, bp, hq); 196 CIRCLEQ_REMOVE(&mp->lqh, bp, q); 197 CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q); 198 199 /* Return a pinned page. */ 200 bp->flags |= MPOOL_PINNED; 201 return (bp->page); 202 } 203 204 /* Get a page from the cache. */ 205 if ((bp = mpool_bkt(mp)) == NULL) 206 return (NULL); 207 208 /* Read in the contents. */ 209 #ifdef STATISTICS 210 ++mp->pageread; 211 #endif 212 off = mp->pagesize * pgno; 213 if (lseek(mp->fd, off, SEEK_SET) != off) 214 return (NULL); 215 if ((nr = read(mp->fd, bp->page, mp->pagesize)) != mp->pagesize) { 216 if (nr >= 0) 217 errno = EFTYPE; 218 return (NULL); 219 } 220 221 /* Set the page number, pin the page. */ 222 bp->pgno = pgno; 223 bp->flags = MPOOL_PINNED; 224 225 /* 226 * Add the page to the head of the hash chain and the tail 227 * of the lru chain. 228 */ 229 head = &mp->hqh[HASHKEY(bp->pgno)]; 230 CIRCLEQ_INSERT_HEAD(head, bp, hq); 231 CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q); 232 233 /* Run through the user's filter. */ 234 if (mp->pgin != NULL) 235 (mp->pgin)(mp->pgcookie, bp->pgno, bp->page); 236 237 return (bp->page); 238 } 239 240 /* 241 * mpool_put 242 * Return a page. 243 */ 244 int 245 mpool_put(mp, page, flags) 246 MPOOL *mp; 247 void *page; 248 u_int flags; 249 { 250 BKT *bp; 251 252 #ifdef STATISTICS 253 ++mp->pageput; 254 #endif 255 bp = (BKT *)((char *)page - sizeof(BKT)); 256 #ifdef DEBUG 257 if (!(bp->flags & MPOOL_PINNED)) { 258 (void)fprintf(stderr, 259 "mpool_put: page %d not pinned\n", bp->pgno); 260 abort(); 261 } 262 #endif 263 bp->flags &= ~MPOOL_PINNED; 264 bp->flags |= flags & MPOOL_DIRTY; 265 return (RET_SUCCESS); 266 } 267 268 /* 269 * mpool_close 270 * Close the buffer pool. 271 */ 272 int 273 mpool_close(mp) 274 MPOOL *mp; 275 { 276 BKT *bp; 277 278 /* Free up any space allocated to the lru pages. */ 279 while ((bp = mp->lqh.cqh_first) != (void *)&mp->lqh) { 280 CIRCLEQ_REMOVE(&mp->lqh, mp->lqh.cqh_first, q); 281 free(bp); 282 } 283 284 /* Free the MPOOL cookie. */ 285 free(mp); 286 return (RET_SUCCESS); 287 } 288 289 /* 290 * mpool_sync 291 * Sync the pool to disk. 292 */ 293 int 294 mpool_sync(mp) 295 MPOOL *mp; 296 { 297 BKT *bp; 298 299 /* Walk the lru chain, flushing any dirty pages to disk. */ 300 for (bp = mp->lqh.cqh_first; 301 bp != (void *)&mp->lqh; bp = bp->q.cqe_next) 302 if (bp->flags & MPOOL_DIRTY && 303 mpool_write(mp, bp) == RET_ERROR) 304 return (RET_ERROR); 305 306 /* Sync the file descriptor. */ 307 return (fsync(mp->fd) ? RET_ERROR : RET_SUCCESS); 308 } 309 310 /* 311 * mpool_bkt 312 * Get a page from the cache (or create one). 313 */ 314 static BKT * 315 mpool_bkt(mp) 316 MPOOL *mp; 317 { 318 struct _hqh *head; 319 BKT *bp; 320 321 /* If under the max cached, always create a new page. */ 322 if (mp->curcache < mp->maxcache) 323 goto new; 324 325 /* 326 * If the cache is max'd out, walk the lru list for a buffer we 327 * can flush. If we find one, write it (if necessary) and take it 328 * off any lists. If we don't find anything we grow the cache anyway. 329 * The cache never shrinks. 330 */ 331 for (bp = mp->lqh.cqh_first; 332 bp != (void *)&mp->lqh; bp = bp->q.cqe_next) 333 if (!(bp->flags & MPOOL_PINNED)) { 334 /* Flush if dirty. */ 335 if (bp->flags & MPOOL_DIRTY && 336 mpool_write(mp, bp) == RET_ERROR) 337 return (NULL); 338 #ifdef STATISTICS 339 ++mp->pageflush; 340 #endif 341 /* Remove from the hash and lru queues. */ 342 head = &mp->hqh[HASHKEY(bp->pgno)]; 343 CIRCLEQ_REMOVE(head, bp, hq); 344 CIRCLEQ_REMOVE(&mp->lqh, bp, q); 345 #ifdef DEBUG 346 { void *spage; 347 spage = bp->page; 348 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 349 bp->page = spage; 350 } 351 #endif 352 return (bp); 353 } 354 355 new: if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL) 356 return (NULL); 357 #ifdef STATISTICS 358 ++mp->pagealloc; 359 #endif 360 #if defined(DEBUG) || defined(PURIFY) 361 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 362 #endif 363 bp->page = (char *)bp + sizeof(BKT); 364 ++mp->curcache; 365 return (bp); 366 } 367 368 /* 369 * mpool_write 370 * Write a page to disk. 371 */ 372 static int 373 mpool_write(mp, bp) 374 MPOOL *mp; 375 BKT *bp; 376 { 377 off_t off; 378 379 #ifdef STATISTICS 380 ++mp->pagewrite; 381 #endif 382 383 /* Run through the user's filter. */ 384 if (mp->pgout) 385 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page); 386 387 off = mp->pagesize * bp->pgno; 388 if (lseek(mp->fd, off, SEEK_SET) != off) 389 return (RET_ERROR); 390 if (write(mp->fd, bp->page, mp->pagesize) != mp->pagesize) 391 return (RET_ERROR); 392 393 bp->flags &= ~MPOOL_DIRTY; 394 return (RET_SUCCESS); 395 } 396 397 /* 398 * mpool_look 399 * Lookup a page in the cache. 400 */ 401 static BKT * 402 mpool_look(mp, pgno) 403 MPOOL *mp; 404 pgno_t pgno; 405 { 406 struct _hqh *head; 407 BKT *bp; 408 409 head = &mp->hqh[HASHKEY(pgno)]; 410 for (bp = head->cqh_first; bp != (void *)head; bp = bp->hq.cqe_next) 411 if (bp->pgno == pgno) { 412 #ifdef STATISTICS 413 ++mp->cachehit; 414 #endif 415 return (bp); 416 } 417 #ifdef STATISTICS 418 ++mp->cachemiss; 419 #endif 420 return (NULL); 421 } 422 423 #ifdef STATISTICS 424 /* 425 * mpool_stat 426 * Print out cache statistics. 427 */ 428 void 429 mpool_stat(mp) 430 MPOOL *mp; 431 { 432 BKT *bp; 433 int cnt; 434 char *sep; 435 436 (void)fprintf(stderr, "%lu pages in the file\n", mp->npages); 437 (void)fprintf(stderr, 438 "page size %lu, cacheing %lu pages of %lu page max cache\n", 439 mp->pagesize, mp->curcache, mp->maxcache); 440 (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 441 mp->pageput, mp->pageget, mp->pagenew); 442 (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n", 443 mp->pagealloc, mp->pageflush); 444 if (mp->cachehit + mp->cachemiss) 445 (void)fprintf(stderr, 446 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 447 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 448 * 100, mp->cachehit, mp->cachemiss); 449 (void)fprintf(stderr, "%lu page reads, %lu page writes\n", 450 mp->pageread, mp->pagewrite); 451 452 sep = ""; 453 cnt = 0; 454 for (bp = mp->lqh.cqh_first; 455 bp != (void *)&mp->lqh; bp = bp->q.cqe_next) { 456 (void)fprintf(stderr, "%s%d", sep, bp->pgno); 457 if (bp->flags & MPOOL_DIRTY) 458 (void)fprintf(stderr, "d"); 459 if (bp->flags & MPOOL_PINNED) 460 (void)fprintf(stderr, "P"); 461 if (++cnt == 10) { 462 sep = "\n"; 463 cnt = 0; 464 } else 465 sep = ", "; 466 467 } 468 (void)fprintf(stderr, "\n"); 469 } 470 #endif 471