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