1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1990, 1993, 1994
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 *    may be used to endorse or promote products derived from this software
17 *    without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32#if defined(LIBC_SCCS) && !defined(lint)
33static char sccsid[] = "@(#)mpool.c	8.7 (Berkeley) 11/2/95";
34#endif /* LIBC_SCCS and not lint */
35#include <sys/cdefs.h>
36__FBSDID("$FreeBSD$");
37
38#include "namespace.h"
39#include <sys/param.h>
40#include <sys/queue.h>
41#include <sys/stat.h>
42
43#include <errno.h>
44#include <stdio.h>
45#include <stdlib.h>
46#include <string.h>
47#include <unistd.h>
48#include "un-namespace.h"
49
50#include <db.h>
51
52#define	__MPOOLINTERFACE_PRIVATE
53#include <mpool.h>
54
55static BKT *mpool_bkt(MPOOL *);
56static BKT *mpool_look(MPOOL *, pgno_t);
57static int  mpool_write(MPOOL *, BKT *);
58
59/*
60 * mpool_open --
61 *	Initialize a memory pool.
62 */
63/* ARGSUSED */
64MPOOL *
65mpool_open(void *key, int fd, pgno_t pagesize, pgno_t 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 */
101void
102mpool_filter(MPOOL *mp, void (*pgin) (void *, pgno_t, void *),
103    void (*pgout) (void *, pgno_t, void *), void *pgcookie)
104{
105	mp->pgin = pgin;
106	mp->pgout = pgout;
107	mp->pgcookie = pgcookie;
108}
109
110/*
111 * mpool_new --
112 *	Get a new page of memory.
113 */
114void *
115mpool_new(MPOOL *mp, pgno_t *pgnoaddr, u_int flags)
116{
117	struct _hqh *head;
118	BKT *bp;
119
120	if (mp->npages == MAX_PAGE_NUMBER) {
121		(void)fprintf(stderr, "mpool_new: page allocation overflow.\n");
122		abort();
123	}
124#ifdef STATISTICS
125	++mp->pagenew;
126#endif
127	/*
128	 * Get a BKT from the cache.  Assign a new page number, attach
129	 * it to the head of the hash chain, the tail of the lru chain,
130	 * and return.
131	 */
132	if ((bp = mpool_bkt(mp)) == NULL)
133		return (NULL);
134	if (flags == MPOOL_PAGE_REQUEST) {
135		mp->npages++;
136		bp->pgno = *pgnoaddr;
137	} else
138		bp->pgno = *pgnoaddr = mp->npages++;
139
140	bp->flags = MPOOL_PINNED | MPOOL_INUSE;
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
148int
149mpool_delete(MPOOL *mp, void *page)
150{
151	struct _hqh *head;
152	BKT *bp;
153
154	bp = (BKT *)((char *)page - sizeof(BKT));
155
156#ifdef DEBUG
157	if (!(bp->flags & MPOOL_PINNED)) {
158		(void)fprintf(stderr,
159		    "mpool_delete: page %d not pinned\n", bp->pgno);
160		abort();
161	}
162#endif
163
164	/* Remove from the hash and lru queues. */
165	head = &mp->hqh[HASHKEY(bp->pgno)];
166	TAILQ_REMOVE(head, bp, hq);
167	TAILQ_REMOVE(&mp->lqh, bp, q);
168
169	free(bp);
170	mp->curcache--;
171	return (RET_SUCCESS);
172}
173
174/*
175 * mpool_get
176 *	Get a page.
177 */
178/* ARGSUSED */
179void *
180mpool_get(MPOOL *mp, pgno_t pgno,
181    u_int flags)		/* XXX not used? */
182{
183	struct _hqh *head;
184	BKT *bp;
185	off_t off;
186	int nr;
187
188#ifdef STATISTICS
189	++mp->pageget;
190#endif
191
192	/* Check for a page that is cached. */
193	if ((bp = mpool_look(mp, pgno)) != NULL) {
194#ifdef DEBUG
195		if (!(flags & MPOOL_IGNOREPIN) && bp->flags & MPOOL_PINNED) {
196			(void)fprintf(stderr,
197			    "mpool_get: page %d already pinned\n", bp->pgno);
198			abort();
199		}
200#endif
201		/*
202		 * Move the page to the head of the hash chain and the tail
203		 * of the lru chain.
204		 */
205		head = &mp->hqh[HASHKEY(bp->pgno)];
206		TAILQ_REMOVE(head, bp, hq);
207		TAILQ_INSERT_HEAD(head, bp, hq);
208		TAILQ_REMOVE(&mp->lqh, bp, q);
209		TAILQ_INSERT_TAIL(&mp->lqh, bp, q);
210
211		/* Return a pinned page. */
212		bp->flags |= MPOOL_PINNED;
213		return (bp->page);
214	}
215
216	/* Get a page from the cache. */
217	if ((bp = mpool_bkt(mp)) == NULL)
218		return (NULL);
219
220	/* Read in the contents. */
221	off = mp->pagesize * pgno;
222	if ((nr = pread(mp->fd, bp->page, mp->pagesize, off)) != (ssize_t)mp->pagesize) {
223		switch (nr) {
224		case -1:
225			/* errno is set for us by pread(). */
226			free(bp);
227			mp->curcache--;
228			return (NULL);
229		case 0:
230			/*
231			 * A zero-length read means you need to create a
232			 * new page.
233			 */
234			memset(bp->page, 0, mp->pagesize);
235			break;
236		default:
237			/* A partial read is definitely bad. */
238			free(bp);
239			mp->curcache--;
240			errno = EINVAL;
241			return (NULL);
242		}
243	}
244#ifdef STATISTICS
245	++mp->pageread;
246#endif
247
248	/* Set the page number, pin the page. */
249	bp->pgno = pgno;
250	if (!(flags & MPOOL_IGNOREPIN))
251		bp->flags = MPOOL_PINNED;
252	bp->flags |= MPOOL_INUSE;
253
254	/*
255	 * Add the page to the head of the hash chain and the tail
256	 * of the lru chain.
257	 */
258	head = &mp->hqh[HASHKEY(bp->pgno)];
259	TAILQ_INSERT_HEAD(head, bp, hq);
260	TAILQ_INSERT_TAIL(&mp->lqh, bp, q);
261
262	/* Run through the user's filter. */
263	if (mp->pgin != NULL)
264		(mp->pgin)(mp->pgcookie, bp->pgno, bp->page);
265
266	return (bp->page);
267}
268
269/*
270 * mpool_put
271 *	Return a page.
272 */
273/* ARGSUSED */
274int
275mpool_put(MPOOL *mp, void *page, u_int flags)
276{
277	BKT *bp;
278
279#ifdef STATISTICS
280	++mp->pageput;
281#endif
282	bp = (BKT *)((char *)page - sizeof(BKT));
283#ifdef DEBUG
284	if (!(bp->flags & MPOOL_PINNED)) {
285		(void)fprintf(stderr,
286		    "mpool_put: page %d not pinned\n", bp->pgno);
287		abort();
288	}
289#endif
290	bp->flags &= ~MPOOL_PINNED;
291	if (flags & MPOOL_DIRTY)
292		bp->flags |= flags & MPOOL_DIRTY;
293	return (RET_SUCCESS);
294}
295
296/*
297 * mpool_close
298 *	Close the buffer pool.
299 */
300int
301mpool_close(MPOOL *mp)
302{
303	BKT *bp;
304
305	/* Free up any space allocated to the lru pages. */
306	while (!TAILQ_EMPTY(&mp->lqh)) {
307		bp = TAILQ_FIRST(&mp->lqh);
308		TAILQ_REMOVE(&mp->lqh, bp, q);
309		free(bp);
310	}
311
312	/* Free the MPOOL cookie. */
313	free(mp);
314	return (RET_SUCCESS);
315}
316
317/*
318 * mpool_sync
319 *	Sync the pool to disk.
320 */
321int
322mpool_sync(MPOOL *mp)
323{
324	BKT *bp;
325
326	/* Walk the lru chain, flushing any dirty pages to disk. */
327	TAILQ_FOREACH(bp, &mp->lqh, q)
328		if (bp->flags & MPOOL_DIRTY &&
329		    mpool_write(mp, bp) == RET_ERROR)
330			return (RET_ERROR);
331
332	/* Sync the file descriptor. */
333	return (_fsync(mp->fd) ? RET_ERROR : RET_SUCCESS);
334}
335
336/*
337 * mpool_bkt
338 *	Get a page from the cache (or create one).
339 */
340static BKT *
341mpool_bkt(MPOOL *mp)
342{
343	struct _hqh *head;
344	BKT *bp;
345
346	/* If under the max cached, always create a new page. */
347	if (mp->curcache < mp->maxcache)
348		goto new;
349
350	/*
351	 * If the cache is max'd out, walk the lru list for a buffer we
352	 * can flush.  If we find one, write it (if necessary) and take it
353	 * off any lists.  If we don't find anything we grow the cache anyway.
354	 * The cache never shrinks.
355	 */
356	TAILQ_FOREACH(bp, &mp->lqh, q)
357		if (!(bp->flags & MPOOL_PINNED)) {
358			/* Flush if dirty. */
359			if (bp->flags & MPOOL_DIRTY &&
360			    mpool_write(mp, bp) == RET_ERROR)
361				return (NULL);
362#ifdef STATISTICS
363			++mp->pageflush;
364#endif
365			/* Remove from the hash and lru queues. */
366			head = &mp->hqh[HASHKEY(bp->pgno)];
367			TAILQ_REMOVE(head, bp, hq);
368			TAILQ_REMOVE(&mp->lqh, bp, q);
369#ifdef DEBUG
370			{ void *spage;
371				spage = bp->page;
372				memset(bp, 0xff, sizeof(BKT) + mp->pagesize);
373				bp->page = spage;
374			}
375#endif
376			bp->flags = 0;
377			return (bp);
378		}
379
380new:	if ((bp = (BKT *)calloc(1, sizeof(BKT) + mp->pagesize)) == NULL)
381		return (NULL);
382#ifdef STATISTICS
383	++mp->pagealloc;
384#endif
385	bp->page = (char *)bp + sizeof(BKT);
386	bp->flags = 0;
387	++mp->curcache;
388	return (bp);
389}
390
391/*
392 * mpool_write
393 *	Write a page to disk.
394 */
395static int
396mpool_write(MPOOL *mp, BKT *bp)
397{
398	off_t off;
399
400#ifdef STATISTICS
401	++mp->pagewrite;
402#endif
403
404	/* Run through the user's filter. */
405	if (mp->pgout)
406		(mp->pgout)(mp->pgcookie, bp->pgno, bp->page);
407
408	off = mp->pagesize * bp->pgno;
409	if (pwrite(mp->fd, bp->page, mp->pagesize, off) != (ssize_t)mp->pagesize)
410		return (RET_ERROR);
411
412	/*
413	 * Re-run through the input filter since this page may soon be
414	 * accessed via the cache, and whatever the user's output filter
415	 * did may screw things up if we don't let the input filter
416	 * restore the in-core copy.
417	 */
418	if (mp->pgin)
419		(mp->pgin)(mp->pgcookie, bp->pgno, bp->page);
420
421	bp->flags &= ~MPOOL_DIRTY;
422	return (RET_SUCCESS);
423}
424
425/*
426 * mpool_look
427 *	Lookup a page in the cache.
428 */
429static BKT *
430mpool_look(MPOOL *mp, pgno_t pgno)
431{
432	struct _hqh *head;
433	BKT *bp;
434
435	head = &mp->hqh[HASHKEY(pgno)];
436	TAILQ_FOREACH(bp, head, hq)
437		if ((bp->pgno == pgno) &&
438			((bp->flags & MPOOL_INUSE) == MPOOL_INUSE)) {
439#ifdef STATISTICS
440			++mp->cachehit;
441#endif
442			return (bp);
443		}
444#ifdef STATISTICS
445	++mp->cachemiss;
446#endif
447	return (NULL);
448}
449
450#ifdef STATISTICS
451/*
452 * mpool_stat
453 *	Print out cache statistics.
454 */
455void
456mpool_stat(MPOOL *mp)
457{
458	BKT *bp;
459	int cnt;
460	char *sep;
461
462	(void)fprintf(stderr, "%lu pages in the file\n", mp->npages);
463	(void)fprintf(stderr,
464	    "page size %lu, cacheing %lu pages of %lu page max cache\n",
465	    mp->pagesize, mp->curcache, mp->maxcache);
466	(void)fprintf(stderr, "%lu page puts, %lu page gets, %lu page new\n",
467	    mp->pageput, mp->pageget, mp->pagenew);
468	(void)fprintf(stderr, "%lu page allocs, %lu page flushes\n",
469	    mp->pagealloc, mp->pageflush);
470	if (mp->cachehit + mp->cachemiss)
471		(void)fprintf(stderr,
472		    "%.0f%% cache hit rate (%lu hits, %lu misses)\n",
473		    ((double)mp->cachehit / (mp->cachehit + mp->cachemiss))
474		    * 100, mp->cachehit, mp->cachemiss);
475	(void)fprintf(stderr, "%lu page reads, %lu page writes\n",
476	    mp->pageread, mp->pagewrite);
477
478	sep = "";
479	cnt = 0;
480	TAILQ_FOREACH(bp, &mp->lqh, q) {
481		(void)fprintf(stderr, "%s%d", sep, bp->pgno);
482		if (bp->flags & MPOOL_DIRTY)
483			(void)fprintf(stderr, "d");
484		if (bp->flags & MPOOL_PINNED)
485			(void)fprintf(stderr, "P");
486		if (++cnt == 10) {
487			sep = "\n";
488			cnt = 0;
489		} else
490			sep = ", ";
491
492	}
493	(void)fprintf(stderr, "\n");
494}
495#endif
496