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 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 */
36
37 #if defined(LIBC_SCCS) && !defined(lint)
38 static char sccsid[] = "@(#)bt_put.c 8.8 (Berkeley) 7/26/94";
39 #endif /* LIBC_SCCS and not lint */
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-int.h"
49 #include "btree.h"
50
51 static EPG *bt_fast __P((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
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 */
66 int
__bt_put(dbp,key,data,flags)67 __bt_put(dbp, key, data, flags)
68 const DB *dbp;
69 DBT *key;
70 const DBT *data;
71 u_int flags;
72 {
73 BTREE *t;
74 DBT tkey, tdata;
75 EPG *e = 0;
76 PAGE *h;
77 indx_t idx, nxtindex;
78 db_pgno_t pg;
79 u_int32_t nbytes;
80 int dflags, exact, status;
81 char *dest, db[NOVFLSIZE], kb[NOVFLSIZE];
82
83 t = dbp->internal;
84
85 /* Toss any page pinned across calls. */
86 if (t->bt_pinned != NULL) {
87 mpool_put(t->bt_mp, t->bt_pinned, 0);
88 t->bt_pinned = NULL;
89 }
90
91 /* Check for change to a read-only tree. */
92 if (F_ISSET(t, B_RDONLY)) {
93 errno = EPERM;
94 return (RET_ERROR);
95 }
96
97 switch (flags) {
98 case 0:
99 case R_NOOVERWRITE:
100 break;
101 case R_CURSOR:
102 /*
103 * If flags is R_CURSOR, put the cursor. Must already
104 * have started a scan and not have already deleted it.
105 */
106 if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
107 !F_ISSET(&t->bt_cursor,
108 CURS_ACQUIRE | CURS_AFTER | CURS_BEFORE))
109 break;
110 /* FALLTHROUGH */
111 default:
112 errno = EINVAL;
113 return (RET_ERROR);
114 }
115
116 /*
117 * If the key/data pair won't fit on a page, store it on overflow
118 * pages. Only put the key on the overflow page if the pair are
119 * still too big after moving the data to an overflow page.
120 *
121 * XXX
122 * If the insert fails later on, the overflow pages aren't recovered.
123 */
124 dflags = 0;
125 if (key->size + data->size > t->bt_ovflsize) {
126 if (key->size > t->bt_ovflsize) {
127 u_int32_t yuck_this_is_gross_code;
128 storekey: if (__ovfl_put(t, key, &pg) == RET_ERROR)
129 return (RET_ERROR);
130 tkey.data = kb;
131 tkey.size = NOVFLSIZE;
132 memmove(kb, &pg, sizeof(db_pgno_t));
133 yuck_this_is_gross_code = key->size;
134 if (yuck_this_is_gross_code != key->size)
135 abort ();
136 memmove(kb + sizeof(db_pgno_t),
137 &yuck_this_is_gross_code, sizeof(u_int32_t));
138 dflags |= P_BIGKEY;
139 key = &tkey;
140 }
141 if (key->size + data->size > t->bt_ovflsize) {
142 u_int32_t yuck_this_is_gross_code = data->size;
143 if (__ovfl_put(t, data, &pg) == RET_ERROR)
144 return (RET_ERROR);
145 tdata.data = db;
146 tdata.size = NOVFLSIZE;
147 memmove(db, &pg, sizeof(db_pgno_t));
148 if (yuck_this_is_gross_code != data->size)
149 abort ();
150 memmove(db + sizeof(db_pgno_t),
151 &yuck_this_is_gross_code, sizeof(u_int32_t));
152 dflags |= P_BIGDATA;
153 data = &tdata;
154 }
155 if (key->size + data->size > t->bt_ovflsize)
156 goto storekey;
157 }
158
159 /* Replace the cursor. */
160 if (flags == R_CURSOR) {
161 if ((h = mpool_get(t->bt_mp, t->bt_cursor.pg.pgno, 0)) == NULL)
162 return (RET_ERROR);
163 idx = t->bt_cursor.pg.index;
164 goto delete;
165 }
166
167 /*
168 * Find the key to delete, or, the location at which to insert.
169 * Bt_fast and __bt_search both pin the returned page.
170 */
171 if (t->bt_order == NOT || (e = bt_fast(t, key, data, &exact)) == NULL)
172 if ((e = __bt_search(t, key, &exact)) == NULL)
173 return (RET_ERROR);
174 h = e->page;
175 idx = e->index;
176
177 /*
178 * Add the key/data pair to the tree. If an identical key is already
179 * in the tree, and R_NOOVERWRITE is set, an error is returned. If
180 * R_NOOVERWRITE is not set, the key is either added (if duplicates are
181 * permitted) or an error is returned.
182 */
183 switch (flags) {
184 case R_NOOVERWRITE:
185 if (!exact)
186 break;
187 mpool_put(t->bt_mp, h, 0);
188 return (RET_SPECIAL);
189 default:
190 if (!exact || !F_ISSET(t, B_NODUPS))
191 break;
192 /*
193 * !!!
194 * Note, the delete may empty the page, so we need to put a
195 * new entry into the page immediately.
196 */
197 delete: if (__bt_dleaf(t, key, h, idx) == RET_ERROR) {
198 mpool_put(t->bt_mp, h, 0);
199 return (RET_ERROR);
200 }
201 break;
202 }
203
204 /*
205 * If not enough room, or the user has put a ceiling on the number of
206 * keys permitted in the page, split the page. The split code will
207 * insert the key and data and unpin the current page. If inserting
208 * into the offset array, shift the pointers up.
209 */
210 nbytes = NBLEAFDBT(key->size, data->size);
211 if (h->upper - h->lower < nbytes + sizeof(indx_t)) {
212 if ((status = __bt_split(t, h, key,
213 data, dflags, nbytes, idx)) != RET_SUCCESS)
214 return (status);
215 goto success;
216 }
217
218 if (idx < (nxtindex = NEXTINDEX(h)))
219 memmove(h->linp + idx + 1, h->linp + idx,
220 (nxtindex - idx) * sizeof(indx_t));
221 h->lower += sizeof(indx_t);
222
223 h->linp[idx] = h->upper -= nbytes;
224 dest = (char *)h + h->upper;
225 WR_BLEAF(dest, key, data, dflags);
226
227 /* If the cursor is on this page, adjust it as necessary. */
228 if (F_ISSET(&t->bt_cursor, CURS_INIT) &&
229 !F_ISSET(&t->bt_cursor, CURS_ACQUIRE) &&
230 t->bt_cursor.pg.pgno == h->pgno && t->bt_cursor.pg.index >= idx)
231 ++t->bt_cursor.pg.index;
232
233 if (t->bt_order == NOT) {
234 if (h->nextpg == P_INVALID) {
235 if (idx == NEXTINDEX(h) - 1) {
236 t->bt_order = FORWARD;
237 t->bt_last.index = idx;
238 t->bt_last.pgno = h->pgno;
239 }
240 } else if (h->prevpg == P_INVALID) {
241 if (idx == 0) {
242 t->bt_order = BACK;
243 t->bt_last.index = 0;
244 t->bt_last.pgno = h->pgno;
245 }
246 }
247 }
248
249 mpool_put(t->bt_mp, h, MPOOL_DIRTY);
250
251 success:
252 if (flags == R_SETCURSOR)
253 __bt_setcur(t, e->page->pgno, e->index);
254
255 F_SET(t, B_MODIFIED);
256 return (RET_SUCCESS);
257 }
258
259 #ifdef STATISTICS
260 u_long bt_cache_hit, bt_cache_miss;
261 #endif
262
263 /*
264 * BT_FAST -- Do a quick check for sorted data.
265 *
266 * Parameters:
267 * t: tree
268 * key: key to insert
269 *
270 * Returns:
271 * EPG for new record or NULL if not found.
272 */
273 static EPG *
bt_fast(t,key,data,exactp)274 bt_fast(t, key, data, exactp)
275 BTREE *t;
276 const DBT *key, *data;
277 int *exactp;
278 {
279 PAGE *h;
280 u_int32_t nbytes;
281 int cmp;
282
283 if ((h = mpool_get(t->bt_mp, t->bt_last.pgno, 0)) == NULL) {
284 t->bt_order = NOT;
285 return (NULL);
286 }
287 t->bt_cur.page = h;
288 t->bt_cur.index = t->bt_last.index;
289
290 /*
291 * If won't fit in this page or have too many keys in this page,
292 * have to search to get split stack.
293 */
294 nbytes = NBLEAFDBT(key->size, data->size);
295 if (h->upper - h->lower < nbytes + sizeof(indx_t))
296 goto miss;
297
298 if (t->bt_order == FORWARD) {
299 if (t->bt_cur.page->nextpg != P_INVALID)
300 goto miss;
301 if (t->bt_cur.index != NEXTINDEX(h) - 1)
302 goto miss;
303 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) < 0)
304 goto miss;
305 t->bt_last.index = cmp ? ++t->bt_cur.index : t->bt_cur.index;
306 } else {
307 if (t->bt_cur.page->prevpg != P_INVALID)
308 goto miss;
309 if (t->bt_cur.index != 0)
310 goto miss;
311 if ((cmp = __bt_cmp(t, key, &t->bt_cur)) > 0)
312 goto miss;
313 t->bt_last.index = 0;
314 }
315 *exactp = cmp == 0;
316 #ifdef STATISTICS
317 ++bt_cache_hit;
318 #endif
319 return (&t->bt_cur);
320
321 miss:
322 #ifdef STATISTICS
323 ++bt_cache_miss;
324 #endif
325 t->bt_order = NOT;
326 mpool_put(t->bt_mp, h, 0);
327 return (NULL);
328 }
329