bt_put.c revision 330897
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 * Mike Olson.
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 * 4. 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[] = "@(#)bt_put.c	8.8 (Berkeley) 7/26/94";
37#endif /* LIBC_SCCS and not lint */
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: stable/11/lib/libc/db/btree/bt_put.c 330897 2018-03-14 03:19:51Z eadler $");
40
41#include <sys/types.h>
42
43#include <errno.h>
44#include <stdio.h>
45#include <stdlib.h>
46#include <string.h>
47
48#include <db.h>
49#include "btree.h"
50
51static EPG *bt_fast(BTREE *, const DBT *, const DBT *, int *);
52
53/*
54 * __BT_PUT -- Add a btree item to the tree.
55 *
56 * Parameters:
57 *	dbp:	pointer to access method
58 *	key:	key
59 *	data:	data
60 *	flag:	R_NOOVERWRITE, R_SETCURSOR, R_CURSOR
61 *
62 * Returns:
63 *	RET_ERROR, RET_SUCCESS and RET_SPECIAL if the key is already in the
64 *	tree and R_NOOVERWRITE specified.
65 */
66int
67__bt_put(const DB *dbp, DBT *key, const DBT *data, u_int flags)
68{
69	BTREE *t;
70	DBT tkey, tdata;
71	EPG *e;
72	PAGE *h;
73	indx_t idx, nxtindex;
74	pgno_t pg;
75	u_int32_t nbytes, tmp;
76	int dflags, exact, status;
77	char *dest, db[NOVFLSIZE], kb[NOVFLSIZE];
78
79	t = dbp->internal;
80
81	/* Toss any page pinned across calls. */
82	if (t->bt_pinned != NULL) {
83		mpool_put(t->bt_mp, t->bt_pinned, 0);
84		t->bt_pinned = NULL;
85	}
86
87	/* Check for change to a read-only tree. */
88	if (F_ISSET(t, B_RDONLY)) {
89		errno = EPERM;
90		return (RET_ERROR);
91	}
92
93	switch (flags) {
94	case 0:
95	case R_NOOVERWRITE:
96	case R_SETCURSOR:
97		break;
98	case R_CURSOR:
99		/*
100		 * If flags is R_CURSOR, put the cursor.  Must already
101		 * have started a scan and not have already deleted it.
102		 */
103		if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
104		    !F_ISSET(&t->bt_cursor,
105			CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE))
106			break;
107		/* FALLTHROUGH */
108	default:
109		errno = EINVAL;
110		return (RET_ERROR);
111	}
112
113	/*
114	 * If the key/data pair won't fit on a page, store it on overflow
115	 * pages.  Only put the key on the overflow page if the pair are
116	 * still too big after moving the data to an overflow page.
117	 *
118	 * XXX
119	 * If the insert fails later on, the overflow pages aren't recovered.
120	 */
121	dflags = 0;
122	if (key->size + data->size > t->bt_ovflsize) {
123		if (key->size > t->bt_ovflsize) {
124storekey:		if (__ovfl_put(t, key, &pg) == RET_ERROR)
125				return (RET_ERROR);
126			tkey.data = kb;
127			tkey.size = NOVFLSIZE;
128			memmove(kb, &pg, sizeof(pgno_t));
129			tmp = key->size;
130			memmove(kb + sizeof(pgno_t),
131			    &tmp, sizeof(u_int32_t));
132			dflags |= P_BIGKEY;
133			key = &tkey;
134		}
135		if (key->size + data->size > t->bt_ovflsize) {
136			if (__ovfl_put(t, data, &pg) == RET_ERROR)
137				return (RET_ERROR);
138			tdata.data = db;
139			tdata.size = NOVFLSIZE;
140			memmove(db, &pg, sizeof(pgno_t));
141			tmp = data->size;
142			memmove(db + sizeof(pgno_t),
143			    &tmp, sizeof(u_int32_t));
144			dflags |= P_BIGDATA;
145			data = &tdata;
146		}
147		if (key->size + data->size > t->bt_ovflsize)
148			goto storekey;
149	}
150
151	/* Replace the cursor. */
152	if (flags == R_CURSOR) {
153		if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL)
154			return (RET_ERROR);
155		idx = t->bt_cursor.pg.index;
156		goto delete;
157	}
158
159	/*
160	 * Find the key to delete, or, the location at which to insert.
161	 * Bt_fast and __bt_search both pin the returned page.
162	 */
163	if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
164		if ((e = __bt_search(t, key, &exact)) == NULL)
165			return (RET_ERROR);
166	h = e->page;
167	idx = e->index;
168
169	/*
170	 * Add the key/data pair to the tree.  If an identical key is already
171	 * in the tree, and R_NOOVERWRITE is set, an error is returned.  If
172	 * R_NOOVERWRITE is not set, the key is either added (if duplicates are
173	 * permitted) or an error is returned.
174	 */
175	switch (flags) {
176	case R_NOOVERWRITE:
177		if (!exact)
178			break;
179		mpool_put(t->bt_mp, h, 0);
180		return (RET_SPECIAL);
181	default:
182		if (!exact || !F_ISSET(t, B_NODUPS))
183			break;
184		/*
185		 * !!!
186		 * Note, the delete may empty the page, so we need to put a
187		 * new entry into the page immediately.
188		 */
189delete:		if (__bt_dleaf(t, key, h, idx) == RET_ERROR) {
190			mpool_put(t->bt_mp, h, 0);
191			return (RET_ERROR);
192		}
193		break;
194	}
195
196	/*
197	 * If not enough room, or the user has put a ceiling on the number of
198	 * keys permitted in the page, split the page.  The split code will
199	 * insert the key and data and unpin the current page.  If inserting
200	 * into the offset array, shift the pointers up.
201	 */
202	nbytes = NBLEAFDBT(key->size, data->size);
203	if ((u_int32_t)(h->upper - h->lower) < nbytes + sizeof(indx_t)) {
204		if ((status = __bt_split(t, h, key,
205		    data, dflags, nbytes, idx)) != RET_SUCCESS)
206			return (status);
207		goto success;
208	}
209
210	if (idx < (nxtindex = NEXTINDEX(h)))
211		memmove(h->linp + idx + 1, h->linp + idx,
212		    (nxtindex - idx) * sizeof(indx_t));
213	h->lower += sizeof(indx_t);
214
215	h->linp[idx] = h->upper -= nbytes;
216	dest = (char *)h + h->upper;
217	WR_BLEAF(dest, key, data, dflags);
218
219	/* If the cursor is on this page, adjust it as necessary. */
220	if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
221	    !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
222	    t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx)
223		++t->bt_cursor.pg.index;
224
225	if (t->bt_order == NOT) {
226		if (h->nextpg == P_INVALID) {
227			if (idx == NEXTINDEX(h) - 1) {
228				t->bt_order = FORWARD;
229				t->bt_last.index = idx;
230				t->bt_last.pgno = h->pgno;
231			}
232		} else if (h->prevpg == P_INVALID) {
233			if (idx == 0) {
234				t->bt_order = BACK;
235				t->bt_last.index = 0;
236				t->bt_last.pgno = h->pgno;
237			}
238		}
239	}
240
241	mpool_put(t->bt_mp, h, MPOOL_DIRTY);
242
243success:
244	if (flags == R_SETCURSOR)
245		__bt_setcur(t, e->page->pgno, e->index);
246
247	F_SET(t, B_MODIFIED);
248	return (RET_SUCCESS);
249}
250
251#ifdef STATISTICS
252u_long bt_cache_hit, bt_cache_miss;
253#endif
254
255/*
256 * BT_FAST -- Do a quick check for sorted data.
257 *
258 * Parameters:
259 *	t:	tree
260 *	key:	key to insert
261 *
262 * Returns:
263 *	EPG for new record or NULL if not found.
264 */
265static EPG *
266bt_fast(BTREE *t, const DBT *key, const DBT *data, int *exactp)
267{
268	PAGE *h;
269	u_int32_t nbytes;
270	int cmp;
271
272	if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
273		t->bt_order = NOT;
274		return (NULL);
275	}
276	t->bt_cur.page = h;
277	t->bt_cur.index = t->bt_last.index;
278
279	/*
280	 * If won't fit in this page or have too many keys in this page,
281	 * have to search to get split stack.
282	 */
283	nbytes = NBLEAFDBT(key->size, data->size);
284	if ((u_int32_t)(h->upper - h->lower) < nbytes + sizeof(indx_t))
285		goto miss;
286
287	if (t->bt_order == FORWARD) {
288		if (t->bt_cur.page->nextpg != P_INVALID)
289			goto miss;
290		if (t->bt_cur.index != NEXTINDEX(h) - 1)
291			goto miss;
292		if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0)
293			goto miss;
294		t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index;
295	} else {
296		if (t->bt_cur.page->prevpg != P_INVALID)
297			goto miss;
298		if (t->bt_cur.index != 0)
299			goto miss;
300		if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0)
301			goto miss;
302		t->bt_last.index = 0;
303	}
304	*exactp = cmp == 0;
305#ifdef STATISTICS
306	++bt_cache_hit;
307#endif
308	return (&t->bt_cur);
309
310miss:
311#ifdef STATISTICS
312	++bt_cache_miss;
313#endif
314	t->bt_order = NOT;
315	mpool_put(t->bt_mp, h, 0);
316	return (NULL);
317}
318