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 1.7 2005/11/19 20:46:32 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 __unused, 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, pgno_t pgno, u_int flags __unused) 145 { 146 struct _hqh *head; 147 BKT *bp; 148 off_t off; 149 int nr; 150 151 /* Check for attempt to retrieve a non-existent page. */ 152 if (pgno >= mp->npages) { 153 errno = EINVAL; 154 return (NULL); 155 } 156 157 #ifdef STATISTICS 158 ++mp->pageget; 159 #endif 160 161 /* Check for a page that is cached. */ 162 if ((bp = mpool_look(mp, pgno)) != NULL) { 163 #ifdef DEBUG 164 if (bp->flags & MPOOL_PINNED) { 165 fprintf(stderr, 166 "mpool_get: page %d already pinned\n", bp->pgno); 167 abort(); 168 } 169 #endif 170 /* 171 * Move the page to the head of the hash chain and the tail 172 * of the lru chain. 173 */ 174 head = &mp->hqh[HASHKEY(bp->pgno)]; 175 TAILQ_REMOVE(head, bp, hq); 176 TAILQ_INSERT_HEAD(head, bp, hq); 177 TAILQ_REMOVE(&mp->lqh, bp, q); 178 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 179 180 /* Return a pinned page. */ 181 bp->flags |= MPOOL_PINNED; 182 return (bp->page); 183 } 184 185 /* Get a page from the cache. */ 186 if ((bp = mpool_bkt(mp)) == NULL) 187 return (NULL); 188 189 /* Read in the contents. */ 190 #ifdef STATISTICS 191 ++mp->pageread; 192 #endif 193 off = mp->pagesize * pgno; 194 if (lseek(mp->fd, off, SEEK_SET) != off) 195 return (NULL); 196 if ((nr = _read(mp->fd, bp->page, mp->pagesize)) != mp->pagesize) { 197 if (nr >= 0) 198 errno = EFTYPE; 199 return (NULL); 200 } 201 202 /* Set the page number, pin the page. */ 203 bp->pgno = pgno; 204 bp->flags = MPOOL_PINNED; 205 206 /* 207 * Add the page to the head of the hash chain and the tail 208 * of the lru chain. 209 */ 210 head = &mp->hqh[HASHKEY(bp->pgno)]; 211 TAILQ_INSERT_HEAD(head, bp, hq); 212 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 213 214 /* Run through the user's filter. */ 215 if (mp->pgin != NULL) 216 (mp->pgin)(mp->pgcookie, bp->pgno, bp->page); 217 218 return (bp->page); 219 } 220 221 /* 222 * mpool_put 223 * Return a page. 224 */ 225 int 226 mpool_put(MPOOL *mp __unused, void *page, u_int flags) 227 { 228 BKT *bp; 229 230 #ifdef STATISTICS 231 ++mp->pageput; 232 #endif 233 bp = (BKT *)((char *)page - sizeof(BKT)); 234 #ifdef DEBUG 235 if (!(bp->flags & MPOOL_PINNED)) { 236 fprintf(stderr, 237 "mpool_put: page %d not pinned\n", bp->pgno); 238 abort(); 239 } 240 #endif 241 bp->flags &= ~MPOOL_PINNED; 242 bp->flags |= flags & MPOOL_DIRTY; 243 return (RET_SUCCESS); 244 } 245 246 /* 247 * mpool_close 248 * Close the buffer pool. 249 */ 250 int 251 mpool_close(MPOOL *mp) 252 { 253 BKT *bp; 254 255 /* Free up any space allocated to the lru pages. */ 256 while (!TAILQ_EMPTY(&mp->lqh)) { 257 bp = TAILQ_FIRST(&mp->lqh); 258 TAILQ_REMOVE(&mp->lqh, bp, q); 259 free(bp); 260 } 261 262 /* Free the MPOOL cookie. */ 263 free(mp); 264 return (RET_SUCCESS); 265 } 266 267 /* 268 * mpool_sync 269 * Sync the pool to disk. 270 */ 271 int 272 mpool_sync(MPOOL *mp) 273 { 274 BKT *bp; 275 276 /* Walk the lru chain, flushing any dirty pages to disk. */ 277 TAILQ_FOREACH(bp, &mp->lqh, q) 278 if (bp->flags & MPOOL_DIRTY && 279 mpool_write(mp, bp) == RET_ERROR) 280 return (RET_ERROR); 281 282 /* Sync the file descriptor. */ 283 return (_fsync(mp->fd) ? RET_ERROR : RET_SUCCESS); 284 } 285 286 /* 287 * mpool_bkt 288 * Get a page from the cache (or create one). 289 */ 290 static BKT * 291 mpool_bkt(MPOOL *mp) 292 { 293 struct _hqh *head; 294 BKT *bp; 295 296 /* If under the max cached, always create a new page. */ 297 if (mp->curcache < mp->maxcache) 298 goto new; 299 300 /* 301 * If the cache is max'd out, walk the lru list for a buffer we 302 * can flush. If we find one, write it (if necessary) and take it 303 * off any lists. If we don't find anything we grow the cache anyway. 304 * The cache never shrinks. 305 */ 306 TAILQ_FOREACH(bp, &mp->lqh, q) 307 if (!(bp->flags & MPOOL_PINNED)) { 308 /* Flush if dirty. */ 309 if (bp->flags & MPOOL_DIRTY && 310 mpool_write(mp, bp) == RET_ERROR) 311 return (NULL); 312 #ifdef STATISTICS 313 ++mp->pageflush; 314 #endif 315 /* Remove from the hash and lru queues. */ 316 head = &mp->hqh[HASHKEY(bp->pgno)]; 317 TAILQ_REMOVE(head, bp, hq); 318 TAILQ_REMOVE(&mp->lqh, bp, q); 319 #ifdef DEBUG 320 { void *spage; 321 spage = bp->page; 322 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 323 bp->page = spage; 324 } 325 #endif 326 return (bp); 327 } 328 329 new: if ((bp = (BKT *)malloc(sizeof(BKT) + mp->pagesize)) == NULL) 330 return (NULL); 331 #ifdef STATISTICS 332 ++mp->pagealloc; 333 #endif 334 #if defined(DEBUG) || defined(PURIFY) 335 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 336 #endif 337 bp->page = (char *)bp + sizeof(BKT); 338 ++mp->curcache; 339 return (bp); 340 } 341 342 /* 343 * mpool_write 344 * Write a page to disk. 345 */ 346 static int 347 mpool_write(MPOOL *mp, BKT *bp) 348 { 349 off_t off; 350 351 #ifdef STATISTICS 352 ++mp->pagewrite; 353 #endif 354 355 /* Run through the user's filter. */ 356 if (mp->pgout) 357 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page); 358 359 off = mp->pagesize * bp->pgno; 360 if (lseek(mp->fd, off, SEEK_SET) != off) 361 return (RET_ERROR); 362 if (_write(mp->fd, bp->page, mp->pagesize) != mp->pagesize) 363 return (RET_ERROR); 364 365 bp->flags &= ~MPOOL_DIRTY; 366 return (RET_SUCCESS); 367 } 368 369 /* 370 * mpool_look 371 * Lookup a page in the cache. 372 */ 373 static BKT * 374 mpool_look(MPOOL *mp, pgno_t pgno) 375 { 376 struct _hqh *head; 377 BKT *bp; 378 379 head = &mp->hqh[HASHKEY(pgno)]; 380 TAILQ_FOREACH(bp, head, hq) 381 if (bp->pgno == pgno) { 382 #ifdef STATISTICS 383 ++mp->cachehit; 384 #endif 385 return (bp); 386 } 387 #ifdef STATISTICS 388 ++mp->cachemiss; 389 #endif 390 return (NULL); 391 } 392 393 #ifdef STATISTICS 394 /* 395 * mpool_stat 396 * Print out cache statistics. 397 */ 398 void 399 mpool_stat(MPOOL *mp) 400 { 401 BKT *bp; 402 int cnt; 403 char *sep; 404 405 fprintf(stderr, "%lu pages in the file\n", mp->npages); 406 fprintf(stderr, 407 "page size %lu, cacheing %lu pages of %lu page max cache\n", 408 mp->pagesize, mp->curcache, mp->maxcache); 409 fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 410 mp->pageput, mp->pageget, mp->pagenew); 411 fprintf(stderr, "%lu page allocs, %lu page flushes\n", 412 mp->pagealloc, mp->pageflush); 413 if (mp->cachehit + mp->cachemiss) 414 fprintf(stderr, 415 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 416 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 417 * 100, mp->cachehit, mp->cachemiss); 418 fprintf(stderr, "%lu page reads, %lu page writes\n", 419 mp->pageread, mp->pagewrite); 420 421 sep = ""; 422 cnt = 0; 423 TAILQ_FOREACH(bp, &mp->lqh, q) { 424 fprintf(stderr, "%s%d", sep, bp->pgno); 425 if (bp->flags & MPOOL_DIRTY) 426 fprintf(stderr, "d"); 427 if (bp->flags & MPOOL_PINNED) 428 fprintf(stderr, "P"); 429 if (++cnt == 10) { 430 sep = "\n"; 431 cnt = 0; 432 } else 433 sep = ", "; 434 435 } 436 fprintf(stderr, "\n"); 437 } 438 #endif 439