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. 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 * $FreeBSD: src/lib/libc/db/mpool/mpool.c,v 1.5.2.1 2001/03/05 23:05:01 obrien Exp $ 30 * $DragonFly: src/lib/libc/db/mpool/mpool.c,v 1.5 2005/09/19 09:20:37 asmodai Exp $ 31 * 32 * @(#)mpool.c 8.5 (Berkeley) 7/26/94 33 */ 34 35 #include "namespace.h" 36 #include <sys/param.h> 37 #include <sys/queue.h> 38 #include <sys/stat.h> 39 40 #include <errno.h> 41 #include <stdio.h> 42 #include <stdlib.h> 43 #include <string.h> 44 #include <unistd.h> 45 #include "un-namespace.h" 46 47 #include <db.h> 48 49 #define __MPOOLINTERFACE_PRIVATE 50 #include <mpool.h> 51 52 static BKT *mpool_bkt (MPOOL *); 53 static BKT *mpool_look (MPOOL *, pgno_t); 54 static int mpool_write (MPOOL *, BKT *); 55 56 /* 57 * mpool_open -- 58 * Initialize a memory pool. 59 */ 60 MPOOL * 61 mpool_open(key, fd, pagesize, maxcache) 62 void *key; 63 int fd; 64 pgno_t pagesize, maxcache; 65 { 66 struct stat sb; 67 MPOOL *mp; 68 int entry; 69 70 /* 71 * Get information about the file. 72 * 73 * XXX 74 * We don't currently handle pipes, although we should. 75 */ 76 if (_fstat(fd, &sb)) 77 return (NULL); 78 if (!S_ISREG(sb.st_mode)) { 79 errno = ESPIPE; 80 return (NULL); 81 } 82 83 /* Allocate and initialize the MPOOL cookie. */ 84 if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL) 85 return (NULL); 86 TAILQ_INIT(&mp->lqh); 87 for (entry = 0; entry < HASHSIZE; ++entry) 88 TAILQ_INIT(&mp->hqh[entry]); 89 mp->maxcache = maxcache; 90 mp->npages = sb.st_size / pagesize; 91 mp->pagesize = pagesize; 92 mp->fd = fd; 93 return (mp); 94 } 95 96 /* 97 * mpool_filter -- 98 * Initialize input/output filters. 99 */ 100 void 101 mpool_filter(mp, pgin, pgout, pgcookie) 102 MPOOL *mp; 103 void (*pgin) (void *, pgno_t, void *); 104 void (*pgout) (void *, pgno_t, void *); 105 void *pgcookie; 106 { 107 mp->pgin = pgin; 108 mp->pgout = pgout; 109 mp->pgcookie = pgcookie; 110 } 111 112 /* 113 * mpool_new -- 114 * Get a new page of memory. 115 */ 116 void * 117 mpool_new(mp, pgnoaddr) 118 MPOOL *mp; 119 pgno_t *pgnoaddr; 120 { 121 struct _hqh *head; 122 BKT *bp; 123 124 if (mp->npages == MAX_PAGE_NUMBER) { 125 (void)fprintf(stderr, "mpool_new: page allocation overflow.\n"); 126 abort(); 127 } 128 #ifdef STATISTICS 129 ++mp->pagenew; 130 #endif 131 /* 132 * Get a BKT from the cache. Assign a new page number, attach 133 * it to the head of the hash chain, the tail of the lru chain, 134 * and return. 135 */ 136 if ((bp = mpool_bkt(mp)) == NULL) 137 return (NULL); 138 *pgnoaddr = bp->pgno = mp->npages++; 139 bp->flags = MPOOL_PINNED; 140 141 head = &mp->hqh[HASHKEY(bp->pgno)]; 142 TAILQ_INSERT_HEAD(head, bp, hq); 143 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 144 return (bp->page); 145 } 146 147 /* 148 * mpool_get 149 * Get a page. 150 */ 151 void * 152 mpool_get(mp, pgno, flags) 153 MPOOL *mp; 154 pgno_t pgno; 155 u_int flags; /* XXX not used? */ 156 { 157 struct _hqh *head; 158 BKT *bp; 159 off_t off; 160 int nr; 161 162 /* Check for attempt to retrieve a non-existent page. */ 163 if (pgno >= mp->npages) { 164 errno = EINVAL; 165 return (NULL); 166 } 167 168 #ifdef STATISTICS 169 ++mp->pageget; 170 #endif 171 172 /* Check for a page that is cached. */ 173 if ((bp = mpool_look(mp, pgno)) != NULL) { 174 #ifdef DEBUG 175 if (bp->flags & MPOOL_PINNED) { 176 (void)fprintf(stderr, 177 "mpool_get: page %d already pinned\n", bp->pgno); 178 abort(); 179 } 180 #endif 181 /* 182 * Move the page to the head of the hash chain and the tail 183 * of the lru chain. 184 */ 185 head = &mp->hqh[HASHKEY(bp->pgno)]; 186 TAILQ_REMOVE(head, bp, hq); 187 TAILQ_INSERT_HEAD(head, bp, hq); 188 TAILQ_REMOVE(&mp->lqh, bp, q); 189 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 190 191 /* Return a pinned page. */ 192 bp->flags |= MPOOL_PINNED; 193 return (bp->page); 194 } 195 196 /* Get a page from the cache. */ 197 if ((bp = mpool_bkt(mp)) == NULL) 198 return (NULL); 199 200 /* Read in the contents. */ 201 #ifdef STATISTICS 202 ++mp->pageread; 203 #endif 204 off = mp->pagesize * pgno; 205 if (lseek(mp->fd, off, SEEK_SET) != off) 206 return (NULL); 207 if ((nr = _read(mp->fd, bp->page, mp->pagesize)) != 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 (lseek(mp->fd, off, SEEK_SET) != off) 380 return (RET_ERROR); 381 if (_write(mp->fd, bp->page, mp->pagesize) != mp->pagesize) 382 return (RET_ERROR); 383 384 bp->flags &= ~MPOOL_DIRTY; 385 return (RET_SUCCESS); 386 } 387 388 /* 389 * mpool_look 390 * Lookup a page in the cache. 391 */ 392 static BKT * 393 mpool_look(mp, pgno) 394 MPOOL *mp; 395 pgno_t pgno; 396 { 397 struct _hqh *head; 398 BKT *bp; 399 400 head = &mp->hqh[HASHKEY(pgno)]; 401 TAILQ_FOREACH(bp, head, hq) 402 if (bp->pgno == pgno) { 403 #ifdef STATISTICS 404 ++mp->cachehit; 405 #endif 406 return (bp); 407 } 408 #ifdef STATISTICS 409 ++mp->cachemiss; 410 #endif 411 return (NULL); 412 } 413 414 #ifdef STATISTICS 415 /* 416 * mpool_stat 417 * Print out cache statistics. 418 */ 419 void 420 mpool_stat(mp) 421 MPOOL *mp; 422 { 423 BKT *bp; 424 int cnt; 425 char *sep; 426 427 (void)fprintf(stderr, "%lu pages in the file\n", mp->npages); 428 (void)fprintf(stderr, 429 "page size %lu, cacheing %lu pages of %lu page max cache\n", 430 mp->pagesize, mp->curcache, mp->maxcache); 431 (void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 432 mp->pageput, mp->pageget, mp->pagenew); 433 (void)fprintf(stderr, "%lu page allocs, %lu page flushes\n", 434 mp->pagealloc, mp->pageflush); 435 if (mp->cachehit + mp->cachemiss) 436 (void)fprintf(stderr, 437 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 438 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 439 * 100, mp->cachehit, mp->cachemiss); 440 (void)fprintf(stderr, "%lu page reads, %lu page writes\n", 441 mp->pageread, mp->pagewrite); 442 443 sep = ""; 444 cnt = 0; 445 TAILQ_FOREACH(bp, &mp->lqh, q) { 446 (void)fprintf(stderr, "%s%d", sep, bp->pgno); 447 if (bp->flags & MPOOL_DIRTY) 448 (void)fprintf(stderr, "d"); 449 if (bp->flags & MPOOL_PINNED) 450 (void)fprintf(stderr, "P"); 451 if (++cnt == 10) { 452 sep = "\n"; 453 cnt = 0; 454 } else 455 sep = ", "; 456 457 } 458 (void)fprintf(stderr, "\n"); 459 } 460 #endif 461