1 /* 2 * Copyright (c) 1997-2000 by Sun Microsystems, Inc. 3 * All rights reserved. 4 */ 5 6 #ifndef _KRB5_BTREE_H 7 #define _KRB5_BTREE_H 8 9 #ifdef __cplusplus 10 extern "C" { 11 #endif 12 13 14 /*- 15 * Copyright (c) 1991, 1993, 1994 16 * The Regents of the University of California. All rights reserved. 17 * 18 * This code is derived from software contributed to Berkeley by 19 * Mike Olson. 20 * 21 * Redistribution and use in source and binary forms, with or without 22 * modification, are permitted provided that the following conditions 23 * are met: 24 * 1. Redistributions of source code must retain the above copyright 25 * notice, this list of conditions and the following disclaimer. 26 * 2. Redistributions in binary form must reproduce the above copyright 27 * notice, this list of conditions and the following disclaimer in the 28 * documentation and/or other materials provided with the distribution. 29 * 3. All advertising materials mentioning features or use of this software 30 * must display the following acknowledgement: 31 * This product includes software developed by the University of 32 * California, Berkeley and its contributors. 33 * 4. Neither the name of the University nor the names of its contributors 34 * may be used to endorse or promote products derived from this software 35 * without specific prior written permission. 36 * 37 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 38 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 39 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 40 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 41 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 42 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 43 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 44 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 45 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 46 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 47 * SUCH DAMAGE. 48 * 49 * @(#)btree.h 8.11 (Berkeley) 8/17/94 50 */ 51 52 /* Macros to set/clear/test flags. */ 53 #define F_SET(p, f) (p)->flags |= (f) 54 #define F_CLR(p, f) (p)->flags &= ~(f) 55 #define F_ISSET(p, f) ((p)->flags & (f)) 56 57 #include "mpool.h" 58 59 #define DEFMINKEYPAGE (2) /* Minimum keys per page */ 60 #define MINCACHE (5) /* Minimum cached pages */ 61 #define MINPSIZE (512) /* Minimum page size */ 62 63 /* 64 * Page 0 of a btree file contains a copy of the meta-data. This page is also 65 * used as an out-of-band page, i.e. page pointers that point to nowhere point 66 * to page 0. Page 1 is the root of the btree. 67 */ 68 #define P_INVALID 0 /* Invalid tree page number. */ 69 #define P_META 0 /* Tree metadata page number. */ 70 #define P_ROOT 1 /* Tree root page number. */ 71 72 /* 73 * There are five page layouts in the btree: btree internal pages (BINTERNAL), 74 * btree leaf pages (BLEAF), recno internal pages (RINTERNAL), recno leaf pages 75 * (RLEAF) and overflow pages. All five page types have a page header (PAGE). 76 * This implementation requires that values within structures NOT be padded. 77 * (ANSI C permits random padding.) If your compiler pads randomly you'll have 78 * to do some work to get this package to run. 79 */ 80 typedef struct _page { 81 db_pgno_t pgno; /* this page's page number */ 82 db_pgno_t prevpg; /* left sibling */ 83 db_pgno_t nextpg; /* right sibling */ 84 85 #define P_BINTERNAL 0x01 /* btree internal page */ 86 #define P_BLEAF 0x02 /* leaf page */ 87 #define P_OVERFLOW 0x04 /* overflow page */ 88 #define P_RINTERNAL 0x08 /* recno internal page */ 89 #define P_RLEAF 0x10 /* leaf page */ 90 #define P_TYPE 0x1f /* type mask */ 91 #define P_PRESERVE 0x20 /* never delete this chain of pages */ 92 u_int32_t flags; 93 94 indx_t lower; /* lower bound of free space on page */ 95 indx_t upper; /* upper bound of free space on page */ 96 indx_t linp[1]; /* indx_t-aligned VAR. LENGTH DATA */ 97 } PAGE; 98 99 /* First and next index. */ 100 #define BTDATAOFF \ 101 (sizeof(db_pgno_t) + sizeof(db_pgno_t) + sizeof(db_pgno_t) + \ 102 sizeof(u_int32_t) + sizeof(indx_t) + sizeof(indx_t)) 103 #define NEXTINDEX(p) (((p)->lower - BTDATAOFF) / sizeof(indx_t)) 104 105 /* 106 * For pages other than overflow pages, there is an array of offsets into the 107 * rest of the page immediately following the page header. Each offset is to 108 * an item which is unique to the type of page. The h_lower offset is just 109 * past the last filled-in index. The h_upper offset is the first item on the 110 * page. Offsets are from the beginning of the page. 111 * 112 * If an item is too big to store on a single page, a flag is set and the item 113 * is a { page, size } pair such that the page is the first page of an overflow 114 * chain with size bytes of item. Overflow pages are simply bytes without any 115 * external structure. 116 * 117 * The page number and size fields in the items are db_pgno_t-aligned so they can 118 * be manipulated without copying. (This presumes that 32 bit items can be 119 * manipulated on this system.) 120 */ 121 #define LALIGN(n) (((n) + sizeof(db_pgno_t) - 1) & ~(sizeof(db_pgno_t) - 1)) 122 #define NOVFLSIZE (sizeof(db_pgno_t) + sizeof(u_int32_t)) 123 124 /* 125 * For the btree internal pages, the item is a key. BINTERNALs are {key, pgno} 126 * pairs, such that the key compares less than or equal to all of the records 127 * on that page. For a tree without duplicate keys, an internal page with two 128 * consecutive keys, a and b, will have all records greater than or equal to a 129 * and less than b stored on the page associated with a. Duplicate keys are 130 * somewhat special and can cause duplicate internal and leaf page records and 131 * some minor modifications of the above rule. 132 */ 133 typedef struct _binternal { 134 u_int32_t ksize; /* key size */ 135 db_pgno_t pgno; /* page number stored on */ 136 #define P_BIGDATA 0x01 /* overflow data */ 137 #define P_BIGKEY 0x02 /* overflow key */ 138 u_char flags; 139 char bytes[1]; /* data */ 140 } BINTERNAL; 141 142 /* Get the page's BINTERNAL structure at index indx. */ 143 #define GETBINTERNAL(pg, indx) \ 144 ((BINTERNAL *)((char *)(pg) + (pg)->linp[indx])) 145 146 /* Get the number of bytes in the entry. */ 147 #define NBINTERNAL(len) \ 148 LALIGN(sizeof(u_int32_t) + sizeof(db_pgno_t) + sizeof(u_char) + (len)) 149 150 /* Copy a BINTERNAL entry to the page. */ 151 #define WR_BINTERNAL(p, size, pgno, flags) { \ 152 *(u_int32_t *)p = size; \ 153 p += sizeof(u_int32_t); \ 154 *(db_pgno_t *)p = pgno; \ 155 p += sizeof(db_pgno_t); \ 156 *(u_char *)p = flags; \ 157 p += sizeof(u_char); \ 158 } 159 160 /* 161 * For the recno internal pages, the item is a page number with the number of 162 * keys found on that page and below. 163 */ 164 typedef struct _rinternal { 165 recno_t nrecs; /* number of records */ 166 db_pgno_t pgno; /* page number stored below */ 167 } RINTERNAL; 168 169 /* Get the page's RINTERNAL structure at index indx. */ 170 #define GETRINTERNAL(pg, indx) \ 171 ((RINTERNAL *)((char *)(pg) + (pg)->linp[indx])) 172 173 /* Get the number of bytes in the entry. */ 174 #define NRINTERNAL \ 175 LALIGN(sizeof(recno_t) + sizeof(db_pgno_t)) 176 177 /* Copy a RINTERAL entry to the page. */ 178 #define WR_RINTERNAL(p, nrecs, pgno) { \ 179 *(recno_t *)p = nrecs; \ 180 p += sizeof(recno_t); \ 181 *(db_pgno_t *)p = pgno; \ 182 } 183 184 /* For the btree leaf pages, the item is a key and data pair. */ 185 typedef struct _bleaf { 186 u_int32_t ksize; /* size of key */ 187 u_int32_t dsize; /* size of data */ 188 u_char flags; /* P_BIGDATA, P_BIGKEY */ 189 char bytes[1]; /* data */ 190 } BLEAF; 191 192 /* Get the page's BLEAF structure at index indx. */ 193 #define GETBLEAF(pg, indx) \ 194 ((BLEAF *)((char *)(pg) + (pg)->linp[indx])) 195 196 /* Get the number of bytes in the entry. */ 197 #define NBLEAF(p) NBLEAFDBT((p)->ksize, (p)->dsize) 198 199 /* Get the number of bytes in the user's key/data pair. */ 200 #define NBLEAFDBT(ksize, dsize) \ 201 LALIGN(sizeof(u_int32_t) + sizeof(u_int32_t) + sizeof(u_char) + \ 202 (ksize) + (dsize)) 203 204 /* Copy a BLEAF entry to the page. */ 205 #define WR_BLEAF(p, key, data, flags) { \ 206 *(u_int32_t *)p = key->size; \ 207 p += sizeof(u_int32_t); \ 208 *(u_int32_t *)p = data->size; \ 209 p += sizeof(u_int32_t); \ 210 *(u_char *)p = flags; \ 211 p += sizeof(u_char); \ 212 memmove(p, key->data, key->size); \ 213 p += key->size; \ 214 memmove(p, data->data, data->size); \ 215 } 216 217 /* For the recno leaf pages, the item is a data entry. */ 218 typedef struct _rleaf { 219 u_int32_t dsize; /* size of data */ 220 u_char flags; /* P_BIGDATA */ 221 char bytes[1]; 222 } RLEAF; 223 224 /* Get the page's RLEAF structure at index indx. */ 225 #define GETRLEAF(pg, indx) \ 226 ((RLEAF *)((char *)(pg) + (pg)->linp[indx])) 227 228 /* Get the number of bytes in the entry. */ 229 #define NRLEAF(p) NRLEAFDBT((p)->dsize) 230 231 /* Get the number of bytes from the user's data. */ 232 #define NRLEAFDBT(dsize) \ 233 LALIGN(sizeof(u_int32_t) + sizeof(u_char) + (dsize)) 234 235 /* Copy a RLEAF entry to the page. */ 236 #define WR_RLEAF(p, data, flags) { \ 237 *(u_int32_t *)p = data->size; \ 238 p += sizeof(u_int32_t); \ 239 *(u_char *)p = flags; \ 240 p += sizeof(u_char); \ 241 memmove(p, data->data, data->size); \ 242 } 243 244 /* 245 * A record in the tree is either a pointer to a page and an index in the page 246 * or a page number and an index. These structures are used as a cursor, stack 247 * entry and search returns as well as to pass records to other routines. 248 * 249 * One comment about searches. Internal page searches must find the largest 250 * record less than key in the tree so that descents work. Leaf page searches 251 * must find the smallest record greater than key so that the returned index 252 * is the record's correct position for insertion. 253 */ 254 typedef struct _epgno { 255 db_pgno_t pgno; /* the page number */ 256 indx_t index; /* the index on the page */ 257 } EPGNO; 258 259 typedef struct _epg { 260 PAGE *page; /* the (pinned) page */ 261 indx_t index; /* the index on the page */ 262 } EPG; 263 264 /* 265 * About cursors. The cursor (and the page that contained the key/data pair 266 * that it referenced) can be deleted, which makes things a bit tricky. If 267 * there are no duplicates of the cursor key in the tree (i.e. B_NODUPS is set 268 * or there simply aren't any duplicates of the key) we copy the key that it 269 * referenced when it's deleted, and reacquire a new cursor key if the cursor 270 * is used again. If there are duplicates keys, we move to the next/previous 271 * key, and set a flag so that we know what happened. NOTE: if duplicate (to 272 * the cursor) keys are added to the tree during this process, it is undefined 273 * if they will be returned or not in a cursor scan. 274 * 275 * The flags determine the possible states of the cursor: 276 * 277 * CURS_INIT The cursor references *something*. 278 * CURS_ACQUIRE The cursor was deleted, and a key has been saved so that 279 * we can reacquire the right position in the tree. 280 * CURS_AFTER, CURS_BEFORE 281 * The cursor was deleted, and now references a key/data pair 282 * that has not yet been returned, either before or after the 283 * deleted key/data pair. 284 * XXX 285 * This structure is broken out so that we can eventually offer multiple 286 * cursors as part of the DB interface. 287 */ 288 typedef struct _cursor { 289 EPGNO pg; /* B: Saved tree reference. */ 290 DBT key; /* B: Saved key, or key.data == NULL. */ 291 recno_t rcursor; /* R: recno cursor (1-based) */ 292 293 #define CURS_ACQUIRE 0x01 /* B: Cursor needs to be reacquired. */ 294 #define CURS_AFTER 0x02 /* B: Unreturned cursor after key. */ 295 #define CURS_BEFORE 0x04 /* B: Unreturned cursor before key. */ 296 #define CURS_INIT 0x08 /* RB: Cursor initialized. */ 297 u_int8_t flags; 298 } CURSOR; 299 300 /* 301 * The metadata of the tree. The nrecs field is used only by the RECNO code. 302 * This is because the btree doesn't really need it and it requires that every 303 * put or delete call modify the metadata. 304 */ 305 typedef struct _btmeta { 306 u_int32_t magic; /* magic number */ 307 u_int32_t version; /* version */ 308 u_int32_t psize; /* page size */ 309 u_int32_t free; /* page number of first free page */ 310 u_int32_t nrecs; /* R: number of records */ 311 312 #define SAVEMETA (B_NODUPS | R_RECNO) 313 u_int32_t flags; /* bt_flags & SAVEMETA */ 314 } BTMETA; 315 316 /* The in-memory btree/recno data structure. */ 317 typedef struct _btree { 318 MPOOL *bt_mp; /* memory pool cookie */ 319 320 DB *bt_dbp; /* pointer to enclosing DB */ 321 322 EPG bt_cur; /* current (pinned) page */ 323 PAGE *bt_pinned; /* page pinned across calls */ 324 325 CURSOR bt_cursor; /* cursor */ 326 327 #define BT_PUSH(t, p, i) { \ 328 t->bt_sp->pgno = p; \ 329 t->bt_sp->index = i; \ 330 ++t->bt_sp; \ 331 } 332 #define BT_POP(t) (t->bt_sp == t->bt_stack ? NULL : --t->bt_sp) 333 #define BT_CLR(t) (t->bt_sp = t->bt_stack) 334 EPGNO bt_stack[50]; /* stack of parent pages */ 335 EPGNO *bt_sp; /* current stack pointer */ 336 337 DBT bt_rkey; /* returned key */ 338 DBT bt_rdata; /* returned data */ 339 340 int bt_fd; /* tree file descriptor */ 341 342 db_pgno_t bt_free; /* next free page */ 343 u_int32_t bt_psize; /* page size */ 344 indx_t bt_ovflsize; /* cut-off for key/data overflow */ 345 int bt_lorder; /* byte order */ 346 /* sorted order */ 347 enum { NOT, BACK, FORWARD } bt_order; 348 EPGNO bt_last; /* last insert */ 349 350 /* B: key comparison function */ 351 int (*bt_cmp) __P((const DBT *, const DBT *)); 352 /* B: prefix comparison function */ 353 size_t (*bt_pfx) __P((const DBT *, const DBT *)); 354 /* R: recno input function */ 355 int (*bt_irec) __P((struct _btree *, recno_t)); 356 357 FILE *bt_rfp; /* R: record FILE pointer */ 358 int bt_rfd; /* R: record file descriptor */ 359 360 caddr_t bt_cmap; /* R: current point in mapped space */ 361 caddr_t bt_smap; /* R: start of mapped space */ 362 caddr_t bt_emap; /* R: end of mapped space */ 363 size_t bt_msize; /* R: size of mapped region. */ 364 365 recno_t bt_nrecs; /* R: number of records */ 366 size_t bt_reclen; /* R: fixed record length */ 367 u_char bt_bval; /* R: delimiting byte/pad character */ 368 369 /* 370 * NB: 371 * B_NODUPS and R_RECNO are stored on disk, and may not be changed. 372 */ 373 #define B_INMEM 0x00001 /* in-memory tree */ 374 #define B_METADIRTY 0x00002 /* need to write metadata */ 375 #define B_MODIFIED 0x00004 /* tree modified */ 376 #define B_NEEDSWAP 0x00008 /* if byte order requires swapping */ 377 #define B_RDONLY 0x00010 /* read-only tree */ 378 379 #define B_NODUPS 0x00020 /* no duplicate keys permitted */ 380 #define R_RECNO 0x00080 /* record oriented tree */ 381 382 #define R_CLOSEFP 0x00040 /* opened a file pointer */ 383 #define R_EOF 0x00100 /* end of input file reached. */ 384 #define R_FIXLEN 0x00200 /* fixed length records */ 385 #define R_MEMMAPPED 0x00400 /* memory mapped file. */ 386 #define R_INMEM 0x00800 /* in-memory file */ 387 #define R_MODIFIED 0x01000 /* modified file */ 388 #define R_RDONLY 0x02000 /* read-only file */ 389 390 #define B_DB_LOCK 0x04000 /* DB_LOCK specified. */ 391 #define B_DB_SHMEM 0x08000 /* DB_SHMEM specified. */ 392 #define B_DB_TXN 0x10000 /* DB_TXN specified. */ 393 u_int32_t flags; 394 } BTREE; 395 396 #include "extern.h" 397 398 #ifdef __cplusplus 399 } 400 #endif 401 402 #endif /* !_KRB5_BTREE_H */ 403