1 /* $NetBSD: mpool.c,v 1.9 1998/06/30 21:30:52 thorpej 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 1.9 1998/06/30 21:30:52 thorpej Exp $"); 42 #endif 43 #endif /* LIBC_SCCS and not lint */ 44 45 #include "namespace.h" 46 #include <sys/param.h> 47 #include <sys/queue.h> 48 #include <sys/stat.h> 49 50 #include <errno.h> 51 #include <stdio.h> 52 #include <stdlib.h> 53 #include <string.h> 54 #include <unistd.h> 55 56 #include <db.h> 57 58 #define __MPOOLINTERFACE_PRIVATE 59 #include <mpool.h> 60 61 #ifdef __weak_alias 62 __weak_alias(mpool_close,_mpool_close); 63 __weak_alias(mpool_filter,_mpool_filter); 64 __weak_alias(mpool_get,_mpool_get); 65 __weak_alias(mpool_new,_mpool_new); 66 __weak_alias(mpool_open,_mpool_open); 67 __weak_alias(mpool_put,_mpool_put); 68 __weak_alias(mpool_sync,_mpool_sync); 69 #endif 70 71 static BKT *mpool_bkt __P((MPOOL *)); 72 static BKT *mpool_look __P((MPOOL *, pgno_t)); 73 static int mpool_write __P((MPOOL *, BKT *)); 74 75 /* 76 * mpool_open -- 77 * Initialize a memory pool. 78 */ 79 MPOOL * 80 mpool_open(key, fd, pagesize, maxcache) 81 void *key; 82 int fd; 83 pgno_t pagesize, maxcache; 84 { 85 struct stat sb; 86 MPOOL *mp; 87 int entry; 88 89 /* 90 * Get information about the file. 91 * 92 * XXX 93 * We don't currently handle pipes, although we should. 94 */ 95 if (fstat(fd, &sb)) 96 return (NULL); 97 if (!S_ISREG(sb.st_mode)) { 98 errno = ESPIPE; 99 return (NULL); 100 } 101 102 /* Allocate and initialize the MPOOL cookie. */ 103 if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL) 104 return (NULL); 105 CIRCLEQ_INIT(&mp->lqh); 106 for (entry = 0; entry < HASHSIZE; ++entry) 107 CIRCLEQ_INIT(&mp->hqh[entry]); 108 mp->maxcache = maxcache; 109 mp->npages = sb.st_size / pagesize; 110 mp->pagesize = pagesize; 111 mp->fd = fd; 112 return (mp); 113 } 114 115 /* 116 * mpool_filter -- 117 * Initialize input/output filters. 118 */ 119 void 120 mpool_filter(mp, pgin, pgout, pgcookie) 121 MPOOL *mp; 122 void (*pgin) __P((void *, pgno_t, void *)); 123 void (*pgout) __P((void *, pgno_t, void *)); 124 void *pgcookie; 125 { 126 mp->pgin = pgin; 127 mp->pgout = pgout; 128 mp->pgcookie = pgcookie; 129 } 130 131 /* 132 * mpool_new -- 133 * Get a new page of memory. 134 */ 135 void * 136 mpool_new(mp, pgnoaddr) 137 MPOOL *mp; 138 pgno_t *pgnoaddr; 139 { 140 struct _hqh *head; 141 BKT *bp; 142 143 if (mp->npages == MAX_PAGE_NUMBER) { 144 (void)fprintf(stderr, "mpool_new: page allocation overflow.\n"); 145 abort(); 146 } 147 #ifdef STATISTICS 148 ++mp->pagenew; 149 #endif 150 /* 151 * Get a BKT from the cache. Assign a new page number, attach 152 * it to the head of the hash chain, the tail of the lru chain, 153 * and return. 154 */ 155 if ((bp = mpool_bkt(mp)) == NULL) 156 return (NULL); 157 *pgnoaddr = bp->pgno = mp->npages++; 158 bp->flags = MPOOL_PINNED; 159 160 head = &mp->hqh[HASHKEY(bp->pgno)]; 161 CIRCLEQ_INSERT_HEAD(head, bp, hq); 162 CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q); 163 return (bp->page); 164 } 165 166 /* 167 * mpool_get 168 * Get a page. 169 */ 170 void * 171 mpool_get(mp, pgno, flags) 172 MPOOL *mp; 173 pgno_t pgno; 174 u_int flags; /* XXX not used? */ 175 { 176 struct _hqh *head; 177 BKT *bp; 178 off_t off; 179 int nr; 180 181 /* Check for attempt to retrieve a non-existent page. */ 182 if (pgno >= mp->npages) { 183 errno = EINVAL; 184 return (NULL); 185 } 186 187 #ifdef STATISTICS 188 ++mp->pageget; 189 #endif 190 191 /* Check for a page that is cached. */ 192 if ((bp = mpool_look(mp, pgno)) != NULL) { 193 #ifdef DEBUG 194 if (bp->flags & MPOOL_PINNED) { 195 (void)fprintf(stderr, 196 "mpool_get: page %d already pinned\n", bp->pgno); 197 abort(); 198 } 199 #endif 200 /* 201 * Move the page to the head of the hash chain and the tail 202 * of the lru chain. 203 */ 204 head = &mp->hqh[HASHKEY(bp->pgno)]; 205 CIRCLEQ_REMOVE(head, bp, hq); 206 CIRCLEQ_INSERT_HEAD(head, bp, hq); 207 CIRCLEQ_REMOVE(&mp->lqh, bp, q); 208 CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q); 209 210 /* Return a pinned page. */ 211 bp->flags |= MPOOL_PINNED; 212 return (bp->page); 213 } 214 215 /* Get a page from the cache. */ 216 if ((bp = mpool_bkt(mp)) == NULL) 217 return (NULL); 218 219 /* Read in the contents. */ 220 #ifdef STATISTICS 221 ++mp->pageread; 222 #endif 223 off = mp->pagesize * pgno; 224 if ((nr = pread(mp->fd, bp->page, mp->pagesize, off)) != mp->pagesize) { 225 if (nr >= 0) 226 errno = EFTYPE; 227 return (NULL); 228 } 229 230 /* Set the page number, pin the page. */ 231 bp->pgno = pgno; 232 bp->flags = MPOOL_PINNED; 233 234 /* 235 * Add the page to the head of the hash chain and the tail 236 * of the lru chain. 237 */ 238 head = &mp->hqh[HASHKEY(bp->pgno)]; 239 CIRCLEQ_INSERT_HEAD(head, bp, hq); 240 CIRCLEQ_INSERT_TAIL(&mp->lqh, bp, q); 241 242 /* Run through the user's filter. */ 243 if (mp->pgin != NULL) 244 (mp->pgin)(mp->pgcookie, bp->pgno, bp->page); 245 246 return (bp->page); 247 } 248 249 /* 250 * mpool_put 251 * Return a page. 252 */ 253 int 254 mpool_put(mp, page, flags) 255 MPOOL *mp; 256 void *page; 257 u_int flags; 258 { 259 BKT *bp; 260 261 #ifdef STATISTICS 262 ++mp->pageput; 263 #endif 264 bp = (BKT *)((char *)page - sizeof(BKT)); 265 #ifdef DEBUG 266 if (!(bp->flags & MPOOL_PINNED)) { 267 (void)fprintf(stderr, 268 "mpool_put: page %d not pinned\n", bp->pgno); 269 abort(); 270 } 271 #endif 272 bp->flags &= ~MPOOL_PINNED; 273 bp->flags |= flags & MPOOL_DIRTY; 274 return (RET_SUCCESS); 275 } 276 277 /* 278 * mpool_close 279 * Close the buffer pool. 280 */ 281 int 282 mpool_close(mp) 283 MPOOL *mp; 284 { 285 BKT *bp; 286 287 /* Free up any space allocated to the lru pages. */ 288 while ((bp = mp->lqh.cqh_first) != (void *)&mp->lqh) { 289 CIRCLEQ_REMOVE(&mp->lqh, mp->lqh.cqh_first, q); 290 free(bp); 291 } 292 293 /* Free the MPOOL cookie. */ 294 free(mp); 295 return (RET_SUCCESS); 296 } 297 298 /* 299 * mpool_sync 300 * Sync the pool to disk. 301 */ 302 int 303 mpool_sync(mp) 304 MPOOL *mp; 305 { 306 BKT *bp; 307 308 /* Walk the lru chain, flushing any dirty pages to disk. */ 309 for (bp = mp->lqh.cqh_first; 310 bp != (void *)&mp->lqh; bp = bp->q.cqe_next) 311 if (bp->flags & MPOOL_DIRTY && 312 mpool_write(mp, bp) == RET_ERROR) 313 return (RET_ERROR); 314 315 /* Sync the file descriptor. */ 316 return (fsync(mp->fd) ? RET_ERROR : RET_SUCCESS); 317 } 318 319 /* 320 * mpool_bkt 321 * Get a page from the cache (or create one). 322 */ 323 static BKT * 324 mpool_bkt(mp) 325 MPOOL *mp; 326 { 327 struct _hqh *head; 328 BKT *bp; 329 330 /* If under the max cached, always create a new page. */ 331 if (mp->curcache < mp->maxcache) 332 goto new; 333 334 /* 335 * If the cache is max'd out, walk the lru list for a buffer we 336 * can flush. If we find one, write it (if necessary) and take it 337 * off any lists. If we don't find anything we grow the cache anyway. 338 * The cache never shrinks. 339 */ 340 for (bp = mp->lqh.cqh_first; 341 bp != (void *)&mp->lqh; bp = bp->q.cqe_next) 342 if (!(bp->flags & MPOOL_PINNED)) { 343 /* Flush if dirty. */ 344 if (bp->flags & MPOOL_DIRTY && 345 mpool_write(mp, bp) == RET_ERROR) 346 return (NULL); 347 #ifdef STATISTICS 348 ++mp->pageflush; 349 #endif 350 /* Remove from the hash and lru queues. */ 351 head = &mp->hqh[HASHKEY(bp->pgno)]; 352 CIRCLEQ_REMOVE(head, bp, hq); 353 CIRCLEQ_REMOVE(&mp->lqh, bp, q); 354 #ifdef DEBUG 355 { void *spage; 356 spage = bp->page; 357 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 358 bp->page = spage; 359 } 360 #endif 361 return (bp); 362 } 363 364 new: if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL) 365 return (NULL); 366 #ifdef STATISTICS 367 ++mp->pagealloc; 368 #endif 369 #if defined(DEBUG) || defined(PURIFY) 370 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 371 #endif 372 bp->page = (char *)bp + sizeof(BKT); 373 ++mp->curcache; 374 return (bp); 375 } 376 377 /* 378 * mpool_write 379 * Write a page to disk. 380 */ 381 static int 382 mpool_write(mp, bp) 383 MPOOL *mp; 384 BKT *bp; 385 { 386 off_t off; 387 388 #ifdef STATISTICS 389 ++mp->pagewrite; 390 #endif 391 392 /* Run through the user's filter. */ 393 if (mp->pgout) 394 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page); 395 396 off = mp->pagesize * bp->pgno; 397 if (pwrite(mp->fd, bp->page, mp->pagesize, off) != mp->pagesize) 398 return (RET_ERROR); 399 400 bp->flags &= ~MPOOL_DIRTY; 401 return (RET_SUCCESS); 402 } 403 404 /* 405 * mpool_look 406 * Lookup a page in the cache. 407 */ 408 static BKT * 409 mpool_look(mp, pgno) 410 MPOOL *mp; 411 pgno_t pgno; 412 { 413 struct _hqh *head; 414 BKT *bp; 415 416 head = &mp->hqh[HASHKEY(pgno)]; 417 for (bp = head->cqh_first; bp != (void *)head; bp = bp->hq.cqe_next) 418 if (bp->pgno == pgno) { 419 #ifdef STATISTICS 420 ++mp->cachehit; 421 #endif 422 return (bp); 423 } 424 #ifdef STATISTICS 425 ++mp->cachemiss; 426 #endif 427 return (NULL); 428 } 429 430 #ifdef STATISTICS 431 /* 432 * mpool_stat 433 * Print out cache statistics. 434 */ 435 void 436 mpool_stat(mp) 437 MPOOL *mp; 438 { 439 BKT *bp; 440 int cnt; 441 char *sep; 442 443 (void)fprintf(stderr, "%lu pages in the file\n", mp->npages); 444 (void)fprintf(stderr, 445 "page size %lu, cacheing %lu pages of %lu page max cache\n", 446 mp->pagesize, mp->curcache, mp->maxcache); 447 (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 448 mp->pageput, mp->pageget, mp->pagenew); 449 (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n", 450 mp->pagealloc, mp->pageflush); 451 if (mp->cachehit + mp->cachemiss) 452 (void)fprintf(stderr, 453 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 454 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 455 * 100, mp->cachehit, mp->cachemiss); 456 (void)fprintf(stderr, "%lu page reads, %lu page writes\n", 457 mp->pageread, mp->pagewrite); 458 459 sep = ""; 460 cnt = 0; 461 for (bp = mp->lqh.cqh_first; 462 bp != (void *)&mp->lqh; bp = bp->q.cqe_next) { 463 (void)fprintf(stderr, "%s%d", sep, bp->pgno); 464 if (bp->flags & MPOOL_DIRTY) 465 (void)fprintf(stderr, "d"); 466 if (bp->flags & MPOOL_PINNED) 467 (void)fprintf(stderr, "P"); 468 if (++cnt == 10) { 469 sep = "\n"; 470 cnt = 0; 471 } else 472 sep = ", "; 473 474 } 475 (void)fprintf(stderr, "\n"); 476 } 477 #endif 478