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 nr = pread(mp->fd, bp->page, mp->pagesize, off); 195 if (nr != mp->pagesize) { 196 if (nr >= 0) 197 errno = EFTYPE; 198 return (NULL); 199 } 200 201 /* Set the page number, pin the page. */ 202 bp->pgno = pgno; 203 bp->flags = MPOOL_PINNED; 204 205 /* 206 * Add the page to the head of the hash chain and the tail 207 * of the lru chain. 208 */ 209 head = &mp->hqh[HASHKEY(bp->pgno)]; 210 TAILQ_INSERT_HEAD(head, bp, hq); 211 TAILQ_INSERT_TAIL(&mp->lqh, bp, q); 212 213 /* Run through the user's filter. */ 214 if (mp->pgin != NULL) 215 (mp->pgin)(mp->pgcookie, bp->pgno, bp->page); 216 217 return (bp->page); 218 } 219 220 /* 221 * mpool_put 222 * Return a page. 223 */ 224 int 225 mpool_put(MPOOL *mp __unused, void *page, u_int flags) 226 { 227 BKT *bp; 228 229 #ifdef STATISTICS 230 ++mp->pageput; 231 #endif 232 bp = (BKT *)((char *)page - sizeof(BKT)); 233 #ifdef DEBUG 234 if (!(bp->flags & MPOOL_PINNED)) { 235 fprintf(stderr, 236 "mpool_put: page %d not pinned\n", bp->pgno); 237 abort(); 238 } 239 #endif 240 bp->flags &= ~MPOOL_PINNED; 241 bp->flags |= flags & MPOOL_DIRTY; 242 return (RET_SUCCESS); 243 } 244 245 /* 246 * mpool_close 247 * Close the buffer pool. 248 */ 249 int 250 mpool_close(MPOOL *mp) 251 { 252 BKT *bp; 253 254 /* Free up any space allocated to the lru pages. */ 255 while (!TAILQ_EMPTY(&mp->lqh)) { 256 bp = TAILQ_FIRST(&mp->lqh); 257 TAILQ_REMOVE(&mp->lqh, bp, q); 258 free(bp); 259 } 260 261 /* Free the MPOOL cookie. */ 262 free(mp); 263 return (RET_SUCCESS); 264 } 265 266 /* 267 * mpool_sync 268 * Sync the pool to disk. 269 */ 270 int 271 mpool_sync(MPOOL *mp) 272 { 273 BKT *bp; 274 275 /* Walk the lru chain, flushing any dirty pages to disk. */ 276 TAILQ_FOREACH(bp, &mp->lqh, q) 277 if (bp->flags & MPOOL_DIRTY && 278 mpool_write(mp, bp) == RET_ERROR) 279 return (RET_ERROR); 280 281 /* Sync the file descriptor. */ 282 return (_fsync(mp->fd) ? RET_ERROR : RET_SUCCESS); 283 } 284 285 /* 286 * mpool_bkt 287 * Get a page from the cache (or create one). 288 */ 289 static BKT * 290 mpool_bkt(MPOOL *mp) 291 { 292 struct _hqh *head; 293 BKT *bp; 294 295 /* If under the max cached, always create a new page. */ 296 if (mp->curcache < mp->maxcache) 297 goto new; 298 299 /* 300 * If the cache is max'd out, walk the lru list for a buffer we 301 * can flush. If we find one, write it (if necessary) and take it 302 * off any lists. If we don't find anything we grow the cache anyway. 303 * The cache never shrinks. 304 */ 305 TAILQ_FOREACH(bp, &mp->lqh, q) 306 if (!(bp->flags & MPOOL_PINNED)) { 307 /* Flush if dirty. */ 308 if (bp->flags & MPOOL_DIRTY && 309 mpool_write(mp, bp) == RET_ERROR) 310 return (NULL); 311 #ifdef STATISTICS 312 ++mp->pageflush; 313 #endif 314 /* Remove from the hash and lru queues. */ 315 head = &mp->hqh[HASHKEY(bp->pgno)]; 316 TAILQ_REMOVE(head, bp, hq); 317 TAILQ_REMOVE(&mp->lqh, bp, q); 318 #ifdef DEBUG 319 { void *spage; 320 spage = bp->page; 321 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 322 bp->page = spage; 323 } 324 #endif 325 return (bp); 326 } 327 328 new: if ((bp = (BKT *)calloc(1, sizeof(BKT) + mp->pagesize)) == NULL) 329 return (NULL); 330 #ifdef STATISTICS 331 ++mp->pagealloc; 332 #endif 333 #if defined(DEBUG) || defined(PURIFY) 334 memset(bp, 0xff, sizeof(BKT) + mp->pagesize); 335 #endif 336 bp->page = (char *)bp + sizeof(BKT); 337 ++mp->curcache; 338 return (bp); 339 } 340 341 /* 342 * mpool_write 343 * Write a page to disk. 344 */ 345 static int 346 mpool_write(MPOOL *mp, BKT *bp) 347 { 348 off_t off; 349 350 #ifdef STATISTICS 351 ++mp->pagewrite; 352 #endif 353 354 /* Run through the user's filter. */ 355 if (mp->pgout) 356 (mp->pgout)(mp->pgcookie, bp->pgno, bp->page); 357 358 off = mp->pagesize * bp->pgno; 359 if (pwrite(mp->fd, bp->page, mp->pagesize, off) != mp->pagesize) 360 return (RET_ERROR); 361 362 bp->flags &= ~MPOOL_DIRTY; 363 return (RET_SUCCESS); 364 } 365 366 /* 367 * mpool_look 368 * Lookup a page in the cache. 369 */ 370 static BKT * 371 mpool_look(MPOOL *mp, pgno_t pgno) 372 { 373 struct _hqh *head; 374 BKT *bp; 375 376 head = &mp->hqh[HASHKEY(pgno)]; 377 TAILQ_FOREACH(bp, head, hq) 378 if (bp->pgno == pgno) { 379 #ifdef STATISTICS 380 ++mp->cachehit; 381 #endif 382 return (bp); 383 } 384 #ifdef STATISTICS 385 ++mp->cachemiss; 386 #endif 387 return (NULL); 388 } 389 390 #ifdef STATISTICS 391 /* 392 * mpool_stat 393 * Print out cache statistics. 394 */ 395 void 396 mpool_stat(MPOOL *mp) 397 { 398 BKT *bp; 399 int cnt; 400 char *sep; 401 402 fprintf(stderr, "%u pages in the file\n", mp->npages); 403 fprintf(stderr, 404 "page size %lu, caching %u pages of %u page max cache\n", 405 mp->pagesize, mp->curcache, mp->maxcache); 406 fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n", 407 mp->pageput, mp->pageget, mp->pagenew); 408 fprintf(stderr, "%lu page allocs, %lu page flushes\n", 409 mp->pagealloc, mp->pageflush); 410 if (mp->cachehit + mp->cachemiss) 411 fprintf(stderr, 412 "%.0f%% cache hit rate (%lu hits, %lu misses)\n", 413 ((double)mp->cachehit / (mp->cachehit + mp->cachemiss)) 414 * 100, mp->cachehit, mp->cachemiss); 415 fprintf(stderr, "%lu page reads, %lu page writes\n", 416 mp->pageread, mp->pagewrite); 417 418 sep = ""; 419 cnt = 0; 420 TAILQ_FOREACH(bp, &mp->lqh, q) { 421 fprintf(stderr, "%s%d", sep, bp->pgno); 422 if (bp->flags & MPOOL_DIRTY) 423 fprintf(stderr, "d"); 424 if (bp->flags & MPOOL_PINNED) 425 fprintf(stderr, "P"); 426 if (++cnt == 10) { 427 sep = "\n"; 428 cnt = 0; 429 } else 430 sep = ", "; 431 432 } 433 fprintf(stderr, "\n"); 434 } 435 #endif 436