bt_put.c revision 165903
1/*-
2 * Copyright (c) 1990, 1993, 1994
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * This code is derived from software contributed to Berkeley by
6 * Mike Olson.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 4. Neither the name of the University nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33#if defined(LIBC_SCCS) && !defined(lint)
34static char sccsid[] = "@(#)bt_put.c	8.8 (Berkeley) 7/26/94";
35#endif /* LIBC_SCCS and not lint */
36#include <sys/cdefs.h>
37__FBSDID("$FreeBSD: head/lib/libc/db/btree/bt_put.c 165903 2007-01-09 00:28:16Z imp $");
38
39#include <sys/types.h>
40
41#include <errno.h>
42#include <stdio.h>
43#include <stdlib.h>
44#include <string.h>
45
46#include <db.h>
47#include "btree.h"
48
49static EPG *bt_fast(BTREE *, const DBT *, const DBT *, int *);
50
51/*
52 * __BT_PUT -- Add a btree item to the tree.
53 *
54 * Parameters:
55 *	dbp:	pointer to access method
56 *	key:	key
57 *	data:	data
58 *	flag:	R_NOOVERWRITE
59 *
60 * Returns:
61 *	RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the
62 *	tree and R_NOOVERWRITE specified.
63 */
64int
65__bt_put(dbp, key, data, flags)
66	const DB *dbp;
67	DBT *key;
68	const DBT *data;
69	u_int flags;
70{
71	BTREE *t;
72	DBT tkey, tdata;
73	EPG *e;
74	PAGE *h;
75	indx_t index, nxtindex;
76	pgno_t pg;
77	u_int32_t nbytes, tmp;
78	int dflags, exact, status;
79	char *dest, db[NOVFLSIZE], kb[NOVFLSIZE];
80
81	t = dbp->internal;
82
83	/* Toss any page pinned across calls. */
84	if (t->bt_pinned != NULL) {
85		mpool_put(t->bt_mp, t->bt_pinned, 0);
86		t->bt_pinned = NULL;
87	}
88
89	/* Check for change to a read-only tree. */
90	if (F_ISSET(t, B_RDONLY)) {
91		errno = EPERM;
92		return (RET_ERROR);
93	}
94
95	switch (flags) {
96	case 0:
97	case R_NOOVERWRITE:
98		break;
99	case R_CURSOR:
100		/*
101		 * If flags is R_CURSOR, put the cursor.  Must already
102		 * have started a scan and not have already deleted it.
103		 */
104		if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
105		    !F_ISSET(&t->bt_cursor,
106		        CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE))
107			break;
108		/* FALLTHROUGH */
109	default:
110		errno = EINVAL;
111		return (RET_ERROR);
112	}
113
114	/*
115	 * If the key/data pair won't fit on a page, store it on overflow
116	 * pages.  Only put the key on the overflow page if the pair are
117	 * still too big after moving the data to an overflow page.
118	 *
119	 * XXX
120	 * If the insert fails later on, the overflow pages aren't recovered.
121	 */
122	dflags = 0;
123	if (key->size + data->size > t->bt_ovflsize) {
124		if (key->size > t->bt_ovflsize) {
125storekey:		if (__ovfl_put(t, key, &pg) == RET_ERROR)
126				return (RET_ERROR);
127			tkey.data = kb;
128			tkey.size = NOVFLSIZE;
129			memmove(kb, &pg, sizeof(pgno_t));
130			tmp = key->size;
131			memmove(kb + sizeof(pgno_t),
132			    &tmp, sizeof(u_int32_t));
133			dflags |= P_BIGKEY;
134			key = &tkey;
135		}
136		if (key->size + data->size > t->bt_ovflsize) {
137			if (__ovfl_put(t, data, &pg) == RET_ERROR)
138				return (RET_ERROR);
139			tdata.data = db;
140			tdata.size = NOVFLSIZE;
141			memmove(db, &pg, sizeof(pgno_t));
142			tmp = data->size;
143			memmove(db + sizeof(pgno_t),
144			    &tmp, sizeof(u_int32_t));
145			dflags |= P_BIGDATA;
146			data = &tdata;
147		}
148		if (key->size + data->size > t->bt_ovflsize)
149			goto storekey;
150	}
151
152	/* Replace the cursor. */
153	if (flags == R_CURSOR) {
154		if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL)
155			return (RET_ERROR);
156		index = t->bt_cursor.pg.index;
157		goto delete;
158	}
159
160	/*
161	 * Find the key to delete, or, the location at which to insert.
162	 * Bt_fast and __bt_search both pin the returned page.
163	 */
164	if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
165		if ((e = __bt_search(t, key, &exact)) == NULL)
166			return (RET_ERROR);
167	h = e->page;
168	index = e->index;
169
170	/*
171	 * Add the key/data pair to the tree.  If an identical key is already
172	 * in the tree, and R_NOOVERWRITE is set, an error is returned.  If
173	 * R_NOOVERWRITE is not set, the key is either added (if duplicates are
174	 * permitted) or an error is returned.
175	 */
176	switch (flags) {
177	case R_NOOVERWRITE:
178		if (!exact)
179			break;
180		mpool_put(t->bt_mp, h, 0);
181		return (RET_SPECIAL);
182	default:
183		if (!exact || !F_ISSET(t, B_NODUPS))
184			break;
185		/*
186		 * !!!
187		 * Note, the delete may empty the page, so we need to put a
188		 * new entry into the page immediately.
189		 */
190delete:		if (__bt_dleaf(t, key, h, index) == RET_ERROR) {
191			mpool_put(t->bt_mp, h, 0);
192			return (RET_ERROR);
193		}
194		break;
195	}
196
197	/*
198	 * If not enough room, or the user has put a ceiling on the number of
199	 * keys permitted in the page, split the page.  The split code will
200	 * insert the key and data and unpin the current page.  If inserting
201	 * into the offset array, shift the pointers up.
202	 */
203	nbytes = NBLEAFDBT(key->size, data->size);
204	if (h->upper - h->lower < nbytes + sizeof(indx_t)) {
205		if ((status = __bt_split(t, h, key,
206		    data, dflags, nbytes, index)) != RET_SUCCESS)
207			return (status);
208		goto success;
209	}
210
211	if (index < (nxtindex = NEXTINDEX(h)))
212		memmove(h->linp + index + 1, h->linp + index,
213		    (nxtindex - index) * sizeof(indx_t));
214	h->lower += sizeof(indx_t);
215
216	h->linp[index] = h->upper -= nbytes;
217	dest = (char *)h + h->upper;
218	WR_BLEAF(dest, key, data, dflags);
219
220	/* If the cursor is on this page, adjust it as necessary. */
221	if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
222	    !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
223	    t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= index)
224		++t->bt_cursor.pg.index;
225
226	if (t->bt_order == NOT) {
227		if (h->nextpg == P_INVALID) {
228			if (index == NEXTINDEX(h) - 1) {
229				t->bt_order = FORWARD;
230				t->bt_last.index = index;
231				t->bt_last.pgno = h->pgno;
232			}
233		} else if (h->prevpg == P_INVALID) {
234			if (index == 0) {
235				t->bt_order = BACK;
236				t->bt_last.index = 0;
237				t->bt_last.pgno = h->pgno;
238			}
239		}
240	}
241
242	mpool_put(t->bt_mp, h, MPOOL_DIRTY);
243
244success:
245	if (flags == R_SETCURSOR)
246		__bt_setcur(t, e->page->pgno, e->index);
247
248	F_SET(t, B_MODIFIED);
249	return (RET_SUCCESS);
250}
251
252#ifdef STATISTICS
253u_long bt_cache_hit, bt_cache_miss;
254#endif
255
256/*
257 * BT_FAST -- Do a quick check for sorted data.
258 *
259 * Parameters:
260 *	t:	tree
261 *	key:	key to insert
262 *
263 * Returns:
264 * 	EPG for new record or NULL if not found.
265 */
266static EPG *
267bt_fast(t, key, data, exactp)
268	BTREE *t;
269	const DBT *key, *data;
270	int *exactp;
271{
272	PAGE *h;
273	u_int32_t nbytes;
274	int cmp;
275
276	if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
277		t->bt_order = NOT;
278		return (NULL);
279	}
280	t->bt_cur.page = h;
281	t->bt_cur.index = t->bt_last.index;
282
283	/*
284	 * If won't fit in this page or have too many keys in this page,
285	 * have to search to get split stack.
286	 */
287	nbytes = NBLEAFDBT(key->size, data->size);
288	if (h->upper - h->lower < nbytes + sizeof(indx_t))
289		goto miss;
290
291	if (t->bt_order == FORWARD) {
292		if (t->bt_cur.page->nextpg != P_INVALID)
293			goto miss;
294		if (t->bt_cur.index != NEXTINDEX(h) - 1)
295			goto miss;
296		if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0)
297			goto miss;
298		t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index;
299	} else {
300		if (t->bt_cur.page->prevpg != P_INVALID)
301			goto miss;
302		if (t->bt_cur.index != 0)
303			goto miss;
304		if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0)
305			goto miss;
306		t->bt_last.index = 0;
307	}
308	*exactp = cmp == 0;
309#ifdef STATISTICS
310	++bt_cache_hit;
311#endif
312	return (&t->bt_cur);
313
314miss:
315#ifdef STATISTICS
316	++bt_cache_miss;
317#endif
318	t->bt_order = NOT;
319	mpool_put(t->bt_mp, h, 0);
320	return (NULL);
321}
322