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 1.6 2005/11/12 23:01:55 swildner 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(void *key, int fd, pgno_t pagesize, pgno_t maxcache) 62 { 63 struct stat sb; 64 MPOOL *mp; 65 int entry; 66 67 /* 68 * Get information about the file. 69 * 70 * XXX 71 * We don't currently handle pipes, although we should. 72 */ 73 if (_fstat(fd, &sb)) 74 return (NULL); 75 if (!S_ISREG(sb.st_mode)) { 76 errno = ESPIPE; 77 return (NULL); 78 } 79 80 /* Allocate and initialize the MPOOL cookie. */ 81 if ((mp = (MPOOL *)calloc(1, sizeof(MPOOL))) == NULL) 82 return (NULL); 83 TAILQ_INIT(&mp->lqh); 84 for (entry = 0; entry < HASHSIZE; ++entry) 85 TAILQ_INIT(&mp->hqh[entry]); 86 mp->maxcache = maxcache; 87 mp->npages = sb.st_size / pagesize; 88 mp->pagesize = pagesize; 89 mp->fd = fd; 90 return (mp); 91 } 92 93 /* 94 * mpool_filter -- 95 * Initialize input/output filters. 96 */ 97 void 98 mpool_filter(MPOOL *mp, void (*pgin)(void *, pgno_t, void *), 99 void (*pgout)(void *, pgno_t, void *), void *pgcookie) 100 { 101 mp->pgin = pgin; 102 mp->pgout = pgout; 103 mp->pgcookie = pgcookie; 104 } 105 106 /* 107 * mpool_new -- 108 * Get a new page of memory. 109 */ 110 void * 111 mpool_new(MPOOL *mp, pgno_t *pgnoaddr) 112 { 113 struct _hqh *head; 114 BKT *bp; 115 116 if (mp->npages == MAX_PAGE_NUMBER) { 117 fprintf(stderr, "mpool_new: page allocation overflow.\n"); 118 abort(); 119 } 120 #ifdef STATISTICS 121 ++mp->pagenew; 122 #endif 123 /* 124 * Get a BKT from the cache. Assign a new page number, attach 125 * it to the head of the hash chain, the tail of the lru chain, 126 * and return. 127 */ 128 if ((bp = mpool_bkt(mp)) == NULL) 129 return (NULL); 130 *pgnoaddr = bp->pgno = mp->npages++; 131 bp->flags = MPOOL_PINNED; 132 133 head = &mp->hqh[HASHKEY(bp->pgno)]; 134 TAILQ_INSERT_HEAD(head, bp, hq); 135 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 136 return (bp->page); 137 } 138 139 /* 140 * mpool_get 141 * Get a page. 142 */ 143 void * 144 mpool_get(MPOOL *mp, 145 pgno_t pgno, 146 u_int flags) /* XXX not used? */ 147 { 148 struct _hqh *head; 149 BKT *bp; 150 off_t off; 151 int nr; 152 153 /* Check for attempt to retrieve a non-existent page. */ 154 if (pgno >= mp->npages) { 155 errno = EINVAL; 156 return (NULL); 157 } 158 159 #ifdef STATISTICS 160 ++mp->pageget; 161 #endif 162 163 /* Check for a page that is cached. */ 164 if ((bp = mpool_look(mp, pgno)) != NULL) { 165 #ifdef DEBUG 166 if (bp->flags & MPOOL_PINNED) { 167 fprintf(stderr, 168 "mpool_get: page %d already pinned\n", bp->pgno); 169 abort(); 170 } 171 #endif 172 /* 173 * Move the page to the head of the hash chain and the tail 174 * of the lru chain. 175 */ 176 head = &mp->hqh[HASHKEY(bp->pgno)]; 177 TAILQ_REMOVE(head, bp, hq); 178 TAILQ_INSERT_HEAD(head, bp, hq); 179 TAILQ_REMOVE(&mp->lqh, bp, q); 180 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 181 182 /* Return a pinned page. */ 183 bp->flags |= MPOOL_PINNED; 184 return (bp->page); 185 } 186 187 /* Get a page from the cache. */ 188 if ((bp = mpool_bkt(mp)) == NULL) 189 return (NULL); 190 191 /* Read in the contents. */ 192 #ifdef STATISTICS 193 ++mp->pageread; 194 #endif 195 off = mp->pagesize * pgno; 196 if (lseek(mp->fd, off, SEEK_SET) != off) 197 return (NULL); 198 if ((nr = _read(mp->fd, bp->page, mp->pagesize)) != mp->pagesize) { 199 if (nr >= 0) 200 errno = EFTYPE; 201 return (NULL); 202 } 203 204 /* Set the page number, pin the page. */ 205 bp->pgno = pgno; 206 bp->flags = MPOOL_PINNED; 207 208 /* 209 * Add the page to the head of the hash chain and the tail 210 * of the lru chain. 211 */ 212 head = &mp->hqh[HASHKEY(bp->pgno)]; 213 TAILQ_INSERT_HEAD(head, bp, hq); 214 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 215 216 /* Run through the user's filter. */ 217 if (mp->pgin != NULL) 218 (mp->pgin)(mp->pgcookie, bp->pgno, bp->page); 219 220 return (bp->page); 221 } 222 223 /* 224 * mpool_put 225 * Return a page. 226 */ 227 int 228 mpool_put(MPOOL *mp, void *page, u_int flags) 229 { 230 BKT *bp; 231 232 #ifdef STATISTICS 233 ++mp->pageput; 234 #endif 235 bp = (BKT *)((char *)page - sizeof(BKT)); 236 #ifdef DEBUG 237 if (!(bp->flags & MPOOL_PINNED)) { 238 fprintf(stderr, 239 "mpool_put: page %d not pinned\n", bp->pgno); 240 abort(); 241 } 242 #endif 243 bp->flags &= ~MPOOL_PINNED; 244 bp->flags |= flags & MPOOL_DIRTY; 245 return (RET_SUCCESS); 246 } 247 248 /* 249 * mpool_close 250 * Close the buffer pool. 251 */ 252 int 253 mpool_close(MPOOL *mp) 254 { 255 BKT *bp; 256 257 /* Free up any space allocated to the lru pages. */ 258 while (!TAILQ_EMPTY(&mp->lqh)) { 259 bp = TAILQ_FIRST(&mp->lqh); 260 TAILQ_REMOVE(&mp->lqh, bp, q); 261 free(bp); 262 } 263 264 /* Free the MPOOL cookie. */ 265 free(mp); 266 return (RET_SUCCESS); 267 } 268 269 /* 270 * mpool_sync 271 * Sync the pool to disk. 272 */ 273 int 274 mpool_sync(MPOOL *mp) 275 { 276 BKT *bp; 277 278 /* Walk the lru chain, flushing any dirty pages to disk. */ 279 TAILQ_FOREACH(bp, &mp->lqh, q) 280 if (bp->flags & MPOOL_DIRTY && 281 mpool_write(mp, bp) == RET_ERROR) 282 return (RET_ERROR); 283 284 /* Sync the file descriptor. */ 285 return (_fsync(mp->fd) ? RET_ERROR : RET_SUCCESS); 286 } 287 288 /* 289 * mpool_bkt 290 * Get a page from the cache (or create one). 291 */ 292 static BKT * 293 mpool_bkt(MPOOL *mp) 294 { 295 struct _hqh *head; 296 BKT *bp; 297 298 /* If under the max cached, always create a new page. */ 299 if (mp->curcache < mp->maxcache) 300 goto new; 301 302 /* 303 * If the cache is max'd out, walk the lru list for a buffer we 304 * can flush. If we find one, write it (if necessary) and take it 305 * off any lists. If we don't find anything we grow the cache anyway. 306 * The cache never shrinks. 307 */ 308 TAILQ_FOREACH(bp, &mp->lqh, q) 309 if (!(bp->flags & MPOOL_PINNED)) { 310 /* Flush if dirty. */ 311 if (bp->flags & MPOOL_DIRTY && 312 mpool_write(mp, bp) == RET_ERROR) 313 return (NULL); 314 #ifdef STATISTICS 315 ++mp->pageflush; 316 #endif 317 /* Remove from the hash and lru queues. */ 318 head = &mp->hqh[HASHKEY(bp->pgno)]; 319 TAILQ_REMOVE(head, bp, hq); 320 TAILQ_REMOVE(&mp->lqh, bp, q); 321 #ifdef DEBUG 322 { void *spage; 323 spage = bp->page; 324 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 325 bp->page = spage; 326 } 327 #endif 328 return (bp); 329 } 330 331 new: if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL) 332 return (NULL); 333 #ifdef STATISTICS 334 ++mp->pagealloc; 335 #endif 336 #if defined(DEBUG) || defined(PURIFY) 337 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 338 #endif 339 bp->page = (char *)bp + sizeof(BKT); 340 ++mp->curcache; 341 return (bp); 342 } 343 344 /* 345 * mpool_write 346 * Write a page to disk. 347 */ 348 static int 349 mpool_write(MPOOL *mp, BKT *bp) 350 { 351 off_t off; 352 353 #ifdef STATISTICS 354 ++mp->pagewrite; 355 #endif 356 357 /* Run through the user's filter. */ 358 if (mp->pgout) 359 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page); 360 361 off = mp->pagesize * bp->pgno; 362 if (lseek(mp->fd, off, SEEK_SET) != off) 363 return (RET_ERROR); 364 if (_write(mp->fd, bp->page, mp->pagesize) != mp->pagesize) 365 return (RET_ERROR); 366 367 bp->flags &= ~MPOOL_DIRTY; 368 return (RET_SUCCESS); 369 } 370 371 /* 372 * mpool_look 373 * Lookup a page in the cache. 374 */ 375 static BKT * 376 mpool_look(MPOOL *mp, pgno_t pgno) 377 { 378 struct _hqh *head; 379 BKT *bp; 380 381 head = &mp->hqh[HASHKEY(pgno)]; 382 TAILQ_FOREACH(bp, head, hq) 383 if (bp->pgno == pgno) { 384 #ifdef STATISTICS 385 ++mp->cachehit; 386 #endif 387 return (bp); 388 } 389 #ifdef STATISTICS 390 ++mp->cachemiss; 391 #endif 392 return (NULL); 393 } 394 395 #ifdef STATISTICS 396 /* 397 * mpool_stat 398 * Print out cache statistics. 399 */ 400 void 401 mpool_stat(MPOOL *mp) 402 { 403 BKT *bp; 404 int cnt; 405 char *sep; 406 407 fprintf(stderr, "%lu pages in the file\n", mp->npages); 408 fprintf(stderr, 409 "page size %lu, cacheing %lu pages of %lu page max cache\n", 410 mp->pagesize, mp->curcache, mp->maxcache); 411 fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 412 mp->pageput, mp->pageget, mp->pagenew); 413 fprintf(stderr, "%lu page allocs, %lu page flushes\n", 414 mp->pagealloc, mp->pageflush); 415 if (mp->cachehit + mp->cachemiss) 416 fprintf(stderr, 417 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 418 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 419 * 100, mp->cachehit, mp->cachemiss); 420 fprintf(stderr, "%lu page reads, %lu page writes\n", 421 mp->pageread, mp->pagewrite); 422 423 sep = ""; 424 cnt = 0; 425 TAILQ_FOREACH(bp, &mp->lqh, q) { 426 fprintf(stderr, "%s%d", sep, bp->pgno); 427 if (bp->flags & MPOOL_DIRTY) 428 fprintf(stderr, "d"); 429 if (bp->flags & MPOOL_PINNED) 430 fprintf(stderr, "P"); 431 if (++cnt == 10) { 432 sep = "\n"; 433 cnt = 0; 434 } else 435 sep = ", "; 436 437 } 438 fprintf(stderr, "\n"); 439 } 440 #endif 441