1 /* $NetBSD: rec_open.c,v 1.17 2008/09/11 12:58:00 joerg Exp $ */ 2 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 * 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 HAVE_NBTOOL_CONFIG_H 36 #include "nbtool_config.h" 37 #endif 38 39 #include <sys/cdefs.h> 40 __RCSID("$NetBSD: rec_open.c,v 1.17 2008/09/11 12:58:00 joerg Exp $"); 41 42 #include "namespace.h" 43 #include <sys/types.h> 44 #include <sys/mman.h> 45 #include <sys/stat.h> 46 47 #include <assert.h> 48 #include <errno.h> 49 #include <fcntl.h> 50 #include <limits.h> 51 #include <stddef.h> 52 #include <stdio.h> 53 #include <unistd.h> 54 55 #include <db.h> 56 #include "recno.h" 57 58 DB * 59 __rec_open(const char *fname, int flags, mode_t mode, const RECNOINFO *openinfo, 60 int dflags) 61 { 62 BTREE *t; 63 BTREEINFO btopeninfo; 64 DB *dbp; 65 PAGE *h; 66 struct stat sb; 67 int rfd = -1; /* pacify gcc */ 68 int sverrno; 69 70 dbp = NULL; 71 /* Open the user's file -- if this fails, we're done. */ 72 if (fname != NULL) { 73 if ((rfd = open(fname, flags, mode)) == -1) 74 return (NULL); 75 if (fcntl(rfd, F_SETFD, FD_CLOEXEC) == -1) 76 goto err; 77 } 78 79 /* Create a btree in memory (backed by disk). */ 80 if (openinfo) { 81 if (openinfo->flags & ~(R_FIXEDLEN | R_NOKEY | R_SNAPSHOT)) 82 goto einval; 83 btopeninfo.flags = 0; 84 btopeninfo.cachesize = openinfo->cachesize; 85 btopeninfo.maxkeypage = 0; 86 btopeninfo.minkeypage = 0; 87 btopeninfo.psize = openinfo->psize; 88 btopeninfo.compare = NULL; 89 btopeninfo.prefix = NULL; 90 btopeninfo.lorder = openinfo->lorder; 91 dbp = __bt_open(openinfo->bfname, 92 O_RDWR, S_IRUSR | S_IWUSR, &btopeninfo, dflags); 93 } else 94 dbp = __bt_open(NULL, O_RDWR, S_IRUSR | S_IWUSR, NULL, dflags); 95 if (dbp == NULL) 96 goto err; 97 98 /* 99 * Some fields in the tree structure are recno specific. Fill them 100 * in and make the btree structure look like a recno structure. We 101 * don't change the bt_ovflsize value, it's close enough and slightly 102 * bigger. 103 */ 104 t = dbp->internal; 105 if (openinfo) { 106 if (openinfo->flags & R_FIXEDLEN) { 107 F_SET(t, R_FIXLEN); 108 t->bt_reclen = openinfo->reclen; 109 if (t->bt_reclen == 0) 110 goto einval; 111 } 112 t->bt_bval = openinfo->bval; 113 } else 114 t->bt_bval = '\n'; 115 116 F_SET(t, R_RECNO); 117 if (fname == NULL) 118 F_SET(t, R_EOF | R_INMEM); 119 else 120 t->bt_rfd = rfd; 121 122 if (fname != NULL) { 123 /* 124 * In 4.4BSD, stat(2) returns true for ISSOCK on pipes. 125 * Unfortunately, that's not portable, so we use lseek 126 * and check the errno values. 127 */ 128 errno = 0; 129 if (lseek(rfd, (off_t)0, SEEK_CUR) == -1 && errno == ESPIPE) { 130 switch (flags & O_ACCMODE) { 131 case O_RDONLY: 132 F_SET(t, R_RDONLY); 133 break; 134 default: 135 goto einval; 136 } 137 slow: if ((t->bt_rfp = fdopen(rfd, "r")) == NULL) 138 goto err; 139 F_SET(t, R_CLOSEFP); 140 t->bt_irec = 141 F_ISSET(t, R_FIXLEN) ? __rec_fpipe : __rec_vpipe; 142 } else { 143 switch (flags & O_ACCMODE) { 144 case O_RDONLY: 145 F_SET(t, R_RDONLY); 146 break; 147 case O_RDWR: 148 break; 149 default: 150 goto einval; 151 } 152 153 if (fstat(rfd, &sb)) 154 goto err; 155 /* 156 * Kluge -- we'd like to test to see if the file is too 157 * big to mmap. Since, we don't know what size or type 158 * off_t's or size_t's are, what the largest unsigned 159 * integral type is, or what random insanity the local 160 * C compiler will perpetrate, doing the comparison in 161 * a portable way is flatly impossible. Hope that mmap 162 * fails if the file is too large. 163 */ 164 if (sb.st_size == 0) 165 F_SET(t, R_EOF); 166 else { 167 #ifdef MMAP_NOT_AVAILABLE 168 /* 169 * XXX 170 * Mmap doesn't work correctly on many current 171 * systems. In particular, it can fail subtly, 172 * with cache coherency problems. Don't use it 173 * for now. 174 */ 175 t->bt_msize = sb.st_size; 176 if ((t->bt_smap = mmap(NULL, t->bt_msize, 177 PROT_READ, MAP_FILE | MAP_PRIVATE, rfd, 178 (off_t)0)) == (caddr_t)-1) 179 goto slow; 180 t->bt_cmap = t->bt_smap; 181 t->bt_emap = t->bt_smap + sb.st_size; 182 t->bt_irec = F_ISSET(t, R_FIXLEN) ? 183 __rec_fmap : __rec_vmap; 184 F_SET(t, R_MEMMAPPED); 185 #else 186 goto slow; 187 #endif 188 } 189 } 190 } 191 192 /* Use the recno routines. */ 193 dbp->close = __rec_close; 194 dbp->del = __rec_delete; 195 dbp->fd = __rec_fd; 196 dbp->get = __rec_get; 197 dbp->put = __rec_put; 198 dbp->seq = __rec_seq; 199 dbp->sync = __rec_sync; 200 201 /* If the root page was created, reset the flags. */ 202 if ((h = mpool_get(t->bt_mp, P_ROOT, 0)) == NULL) 203 goto err; 204 if ((h->flags & P_TYPE) == P_BLEAF) { 205 F_CLR(h, P_TYPE); 206 F_SET(h, P_RLEAF); 207 mpool_put(t->bt_mp, h, MPOOL_DIRTY); 208 } else 209 mpool_put(t->bt_mp, h, 0); 210 211 if (openinfo && openinfo->flags & R_SNAPSHOT && 212 !F_ISSET(t, R_EOF | R_INMEM) && 213 t->bt_irec(t, MAX_REC_NUMBER) == RET_ERROR) 214 goto err; 215 return (dbp); 216 217 einval: errno = EINVAL; 218 err: sverrno = errno; 219 if (dbp != NULL) 220 (void)__bt_close(dbp); 221 if (fname != NULL) 222 (void)close(rfd); 223 errno = sverrno; 224 return (NULL); 225 } 226 227 int 228 __rec_fd(const DB *dbp) 229 { 230 BTREE *t; 231 232 t = dbp->internal; 233 234 /* Toss any page pinned across calls. */ 235 if (t->bt_pinned != NULL) { 236 mpool_put(t->bt_mp, t->bt_pinned, 0); 237 t->bt_pinned = NULL; 238 } 239 240 /* In-memory database can't have a file descriptor. */ 241 if (F_ISSET(t, R_INMEM)) { 242 errno = ENOENT; 243 return (-1); 244 } 245 return (t->bt_rfd); 246 } 247