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 * This code is derived from software contributed to Berkeley by
8 * Margo Seltzer.
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. Neither the name of the University nor the names of its contributors
19 *    may be used to endorse or promote products derived from this software
20 *    without specific prior written permission.
21 *
22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 * SUCH DAMAGE.
33 */
34
35#if defined(LIBC_SCCS) && !defined(lint)
36static char sccsid[] = "@(#)hash_buf.c	8.5 (Berkeley) 7/15/94";
37#endif /* LIBC_SCCS and not lint */
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD$");
40
41/*
42 * PACKAGE: hash
43 *
44 * DESCRIPTION:
45 *	Contains buffer management
46 *
47 * ROUTINES:
48 * External
49 *	__buf_init
50 *	__get_buf
51 *	__buf_free
52 *	__reclaim_buf
53 * Internal
54 *	newbuf
55 */
56
57#include <sys/param.h>
58
59#include <stddef.h>
60#include <stdio.h>
61#include <stdlib.h>
62#include <string.h>
63
64#ifdef DEBUG
65#include <assert.h>
66#endif
67
68#include <db.h>
69#include "hash.h"
70#include "page.h"
71#include "extern.h"
72
73static BUFHEAD *newbuf(HTAB *, u_int32_t, BUFHEAD *);
74
75/* Unlink B from its place in the lru */
76#define BUF_REMOVE(B) { \
77	(B)->prev->next = (B)->next; \
78	(B)->next->prev = (B)->prev; \
79}
80
81/* Insert B after P */
82#define BUF_INSERT(B, P) { \
83	(B)->next = (P)->next; \
84	(B)->prev = (P); \
85	(P)->next = (B); \
86	(B)->next->prev = (B); \
87}
88
89#define	MRU	hashp->bufhead.next
90#define	LRU	hashp->bufhead.prev
91
92#define MRU_INSERT(B)	BUF_INSERT((B), &hashp->bufhead)
93#define LRU_INSERT(B)	BUF_INSERT((B), LRU)
94
95/*
96 * We are looking for a buffer with address "addr".  If prev_bp is NULL, then
97 * address is a bucket index.  If prev_bp is not NULL, then it points to the
98 * page previous to an overflow page that we are trying to find.
99 *
100 * CAVEAT:  The buffer header accessed via prev_bp's ovfl field may no longer
101 * be valid.  Therefore, you must always verify that its address matches the
102 * address you are seeking.
103 */
104BUFHEAD *
105__get_buf(HTAB *hashp, u_int32_t addr,
106    BUFHEAD *prev_bp,	/* If prev_bp set, indicates a new overflow page. */
107    int newpage)
108{
109	BUFHEAD *bp;
110	u_int32_t is_disk_mask;
111	int is_disk, segment_ndx;
112	SEGMENT segp;
113
114	is_disk = 0;
115	is_disk_mask = 0;
116	if (prev_bp) {
117		bp = prev_bp->ovfl;
118		if (!bp || (bp->addr != addr))
119			bp = NULL;
120		if (!newpage)
121			is_disk = BUF_DISK;
122	} else {
123		/* Grab buffer out of directory */
124		segment_ndx = addr & (hashp->SGSIZE - 1);
125
126		/* valid segment ensured by __call_hash() */
127		segp = hashp->dir[addr >> hashp->SSHIFT];
128#ifdef DEBUG
129		assert(segp != NULL);
130#endif
131		bp = PTROF(segp[segment_ndx]);
132		is_disk_mask = ISDISK(segp[segment_ndx]);
133		is_disk = is_disk_mask || !hashp->new_file;
134	}
135
136	if (!bp) {
137		bp = newbuf(hashp, addr, prev_bp);
138		if (!bp ||
139		    __get_page(hashp, bp->page, addr, !prev_bp, is_disk, 0))
140			return (NULL);
141		if (!prev_bp)
142			segp[segment_ndx] =
143			    (BUFHEAD *)((intptr_t)bp | is_disk_mask);
144	} else {
145		BUF_REMOVE(bp);
146		MRU_INSERT(bp);
147	}
148	return (bp);
149}
150
151/*
152 * We need a buffer for this page. Either allocate one, or evict a resident
153 * one (if we have as many buffers as we're allowed) and put this one in.
154 *
155 * If newbuf finds an error (returning NULL), it also sets errno.
156 */
157static BUFHEAD *
158newbuf(HTAB *hashp, u_int32_t addr, BUFHEAD *prev_bp)
159{
160	BUFHEAD *bp;		/* The buffer we're going to use */
161	BUFHEAD *xbp;		/* Temp pointer */
162	BUFHEAD *next_xbp;
163	SEGMENT segp;
164	int segment_ndx;
165	u_int16_t oaddr, *shortp;
166
167	oaddr = 0;
168	bp = LRU;
169
170        /* It is bad to overwrite the page under the cursor. */
171        if (bp == hashp->cpage) {
172                BUF_REMOVE(bp);
173                MRU_INSERT(bp);
174                bp = LRU;
175        }
176
177	/* If prev_bp is part of bp overflow, create a new buffer. */
178	if (hashp->nbufs == 0 && prev_bp && bp->ovfl) {
179		BUFHEAD *ovfl;
180
181		for (ovfl = bp->ovfl; ovfl ; ovfl = ovfl->ovfl) {
182			if (ovfl == prev_bp) {
183				hashp->nbufs++;
184				break;
185			}
186		}
187	}
188
189	/*
190	 * If LRU buffer is pinned, the buffer pool is too small. We need to
191	 * allocate more buffers.
192	 */
193	if (hashp->nbufs || (bp->flags & BUF_PIN) || bp == hashp->cpage) {
194		/* Allocate a new one */
195		if ((bp = (BUFHEAD *)calloc(1, sizeof(BUFHEAD))) == NULL)
196			return (NULL);
197		if ((bp->page = (char *)calloc(1, hashp->BSIZE)) == NULL) {
198			free(bp);
199			return (NULL);
200		}
201		if (hashp->nbufs)
202			hashp->nbufs--;
203	} else {
204		/* Kick someone out */
205		BUF_REMOVE(bp);
206		/*
207		 * If this is an overflow page with addr 0, it's already been
208		 * flushed back in an overflow chain and initialized.
209		 */
210		if ((bp->addr != 0) || (bp->flags & BUF_BUCKET)) {
211			/*
212			 * Set oaddr before __put_page so that you get it
213			 * before bytes are swapped.
214			 */
215			shortp = (u_int16_t *)bp->page;
216			if (shortp[0])
217				oaddr = shortp[shortp[0] - 1];
218			if ((bp->flags & BUF_MOD) && __put_page(hashp, bp->page,
219			    bp->addr, (int)IS_BUCKET(bp->flags), 0))
220				return (NULL);
221			/*
222			 * Update the pointer to this page (i.e. invalidate it).
223			 *
224			 * If this is a new file (i.e. we created it at open
225			 * time), make sure that we mark pages which have been
226			 * written to disk so we retrieve them from disk later,
227			 * rather than allocating new pages.
228			 */
229			if (IS_BUCKET(bp->flags)) {
230				segment_ndx = bp->addr & (hashp->SGSIZE - 1);
231				segp = hashp->dir[bp->addr >> hashp->SSHIFT];
232#ifdef DEBUG
233				assert(segp != NULL);
234#endif
235
236				if (hashp->new_file &&
237				    ((bp->flags & BUF_MOD) ||
238				    ISDISK(segp[segment_ndx])))
239					segp[segment_ndx] = (BUFHEAD *)BUF_DISK;
240				else
241					segp[segment_ndx] = NULL;
242			}
243			/*
244			 * Since overflow pages can only be access by means of
245			 * their bucket, free overflow pages associated with
246			 * this bucket.
247			 */
248			for (xbp = bp; xbp->ovfl;) {
249				next_xbp = xbp->ovfl;
250				xbp->ovfl = NULL;
251				xbp = next_xbp;
252
253				/* Check that ovfl pointer is up date. */
254				if (IS_BUCKET(xbp->flags) ||
255				    (oaddr != xbp->addr))
256					break;
257
258				shortp = (u_int16_t *)xbp->page;
259				if (shortp[0])
260					/* set before __put_page */
261					oaddr = shortp[shortp[0] - 1];
262				if ((xbp->flags & BUF_MOD) && __put_page(hashp,
263				    xbp->page, xbp->addr, 0, 0))
264					return (NULL);
265				xbp->addr = 0;
266				xbp->flags = 0;
267				BUF_REMOVE(xbp);
268				LRU_INSERT(xbp);
269			}
270		}
271	}
272
273	/* Now assign this buffer */
274	bp->addr = addr;
275#ifdef DEBUG1
276	(void)fprintf(stderr, "NEWBUF1: %d->ovfl was %d is now %d\n",
277	    bp->addr, (bp->ovfl ? bp->ovfl->addr : 0), 0);
278#endif
279	bp->ovfl = NULL;
280	if (prev_bp) {
281		/*
282		 * If prev_bp is set, this is an overflow page, hook it in to
283		 * the buffer overflow links.
284		 */
285#ifdef DEBUG1
286		(void)fprintf(stderr, "NEWBUF2: %d->ovfl was %d is now %d\n",
287		    prev_bp->addr, (prev_bp->ovfl ? prev_bp->ovfl->addr : 0),
288		    (bp ? bp->addr : 0));
289#endif
290		prev_bp->ovfl = bp;
291		bp->flags = 0;
292	} else
293		bp->flags = BUF_BUCKET;
294	MRU_INSERT(bp);
295	return (bp);
296}
297
298void
299__buf_init(HTAB *hashp, int nbytes)
300{
301	BUFHEAD *bfp;
302	int npages;
303
304	bfp = &(hashp->bufhead);
305	npages = (nbytes + hashp->BSIZE - 1) >> hashp->BSHIFT;
306	npages = MAX(npages, MIN_BUFFERS);
307
308	hashp->nbufs = npages;
309	bfp->next = bfp;
310	bfp->prev = bfp;
311	/*
312	 * This space is calloc'd so these are already null.
313	 *
314	 * bfp->ovfl = NULL;
315	 * bfp->flags = 0;
316	 * bfp->page = NULL;
317	 * bfp->addr = 0;
318	 */
319}
320
321int
322__buf_free(HTAB *hashp, int do_free, int to_disk)
323{
324	BUFHEAD *bp;
325
326	/* Need to make sure that buffer manager has been initialized */
327	if (!LRU)
328		return (0);
329	for (bp = LRU; bp != &hashp->bufhead;) {
330		/* Check that the buffer is valid */
331		if (bp->addr || IS_BUCKET(bp->flags)) {
332			if (to_disk && (bp->flags & BUF_MOD) &&
333			    __put_page(hashp, bp->page,
334			    bp->addr, IS_BUCKET(bp->flags), 0))
335				return (-1);
336		}
337		/* Check if we are freeing stuff */
338		if (do_free) {
339			if (bp->page) {
340				(void)memset(bp->page, 0, hashp->BSIZE);
341				free(bp->page);
342			}
343			BUF_REMOVE(bp);
344			free(bp);
345			bp = LRU;
346		} else
347			bp = bp->prev;
348	}
349	return (0);
350}
351
352void
353__reclaim_buf(HTAB *hashp, BUFHEAD *bp)
354{
355	bp->ovfl = NULL;
356	bp->addr = 0;
357	bp->flags = 0;
358	BUF_REMOVE(bp);
359	LRU_INSERT(bp);
360}
361