2 Unix SMB/CIFS implementation.
4 trivial database library
6 Copyright (C) Andrew Tridgell 1999-2005
7 Copyright (C) Paul `Rusty' Russell 2000
8 Copyright (C) Jeremy Allison 2000-2003
10 ** NOTE! The following LGPL license applies to the tdb
11 ** library. This does NOT imply that all of Samba is released
14 This library is free software; you can redistribute it and/or
15 modify it under the terms of the GNU Lesser General Public
16 License as published by the Free Software Foundation; either
17 version 3 of the License, or (at your option) any later version.
19 This library is distributed in the hope that it will be useful,
20 but WITHOUT ANY WARRANTY; without even the implied warranty of
21 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
22 Lesser General Public License for more details.
24 You should have received a copy of the GNU Lesser General Public
25 License along with this library; if not, see <http://www.gnu.org/licenses/>.
28 #include "tdb_private.h"
30 /* check for an out of bounds access - if it is out of bounds then
31 see if the database has been expanded by someone else and expand
33 note that "len" is the minimum length needed for the db
35 static int tdb_oob(struct tdb_context *tdb, tdb_off_t len, int probe)
38 if (len <= tdb->map_size)
40 if (tdb->flags & TDB_INTERNAL) {
42 /* Ensure ecode is set for log fn. */
43 tdb->ecode = TDB_ERR_IO;
44 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %d beyond internal malloc size %d\n",
45 (int)len, (int)tdb->map_size));
50 if (fstat(tdb->fd, &st) == -1) {
51 tdb->ecode = TDB_ERR_IO;
55 if (st.st_size < (size_t)len) {
57 /* Ensure ecode is set for log fn. */
58 tdb->ecode = TDB_ERR_IO;
59 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %d beyond eof at %d\n",
60 (int)len, (int)st.st_size));
65 /* Unmap, update size, remap */
66 if (tdb_munmap(tdb) == -1) {
67 tdb->ecode = TDB_ERR_IO;
70 tdb->map_size = st.st_size;
75 /* write a lump of data at a specified offset */
76 static int tdb_write(struct tdb_context *tdb, tdb_off_t off,
77 const void *buf, tdb_len_t len)
83 if (tdb->read_only || tdb->traverse_read) {
84 tdb->ecode = TDB_ERR_RDONLY;
88 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0)
92 memcpy(off + (char *)tdb->map_ptr, buf, len);
94 ssize_t written = pwrite(tdb->fd, buf, len, off);
95 if ((written != (ssize_t)len) && (written != -1)) {
97 tdb->ecode = TDB_ERR_IO;
98 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_write: wrote only "
99 "%d of %d bytes at %d, trying once more\n",
100 (int)written, len, off));
101 written = pwrite(tdb->fd, (const char *)buf+written,
106 /* Ensure ecode is set for log fn. */
107 tdb->ecode = TDB_ERR_IO;
108 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_write failed at %d "
109 "len=%d (%s)\n", off, len, strerror(errno)));
111 } else if (written != (ssize_t)len) {
112 tdb->ecode = TDB_ERR_IO;
113 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_write: failed to "
114 "write %d bytes at %d in two attempts\n",
122 /* Endian conversion: we only ever deal with 4 byte quantities */
123 void *tdb_convert(void *buf, uint32_t size)
125 uint32_t i, *p = (uint32_t *)buf;
126 for (i = 0; i < size / 4; i++)
127 p[i] = TDB_BYTEREV(p[i]);
132 /* read a lump of data at a specified offset, maybe convert */
133 static int tdb_read(struct tdb_context *tdb, tdb_off_t off, void *buf,
134 tdb_len_t len, int cv)
136 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0) {
141 memcpy(buf, off + (char *)tdb->map_ptr, len);
143 ssize_t ret = pread(tdb->fd, buf, len, off);
144 if (ret != (ssize_t)len) {
145 /* Ensure ecode is set for log fn. */
146 tdb->ecode = TDB_ERR_IO;
147 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_read failed at %d "
148 "len=%d ret=%d (%s) map_size=%d\n",
149 (int)off, (int)len, (int)ret, strerror(errno),
150 (int)tdb->map_size));
155 tdb_convert(buf, len);
163 do an unlocked scan of the hash table heads to find the next non-zero head. The value
164 will then be confirmed with the lock held
166 static void tdb_next_hash_chain(struct tdb_context *tdb, uint32_t *chain)
170 for (;h < tdb->header.hash_size;h++) {
171 if (0 != *(uint32_t *)(TDB_HASH_TOP(h) + (unsigned char *)tdb->map_ptr)) {
177 for (;h < tdb->header.hash_size;h++) {
178 if (tdb_ofs_read(tdb, TDB_HASH_TOP(h), &off) != 0 || off != 0) {
187 int tdb_munmap(struct tdb_context *tdb)
189 if (tdb->flags & TDB_INTERNAL)
196 ret = munmap(tdb->map_ptr, tdb->map_size);
205 void tdb_mmap(struct tdb_context *tdb)
207 if (tdb->flags & TDB_INTERNAL)
211 if (!(tdb->flags & TDB_NOMMAP)) {
212 tdb->map_ptr = mmap(NULL, tdb->map_size,
213 PROT_READ|(tdb->read_only? 0:PROT_WRITE),
214 MAP_SHARED, tdb->fd, 0);
217 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
220 if (tdb->map_ptr == MAP_FAILED) {
222 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_mmap failed for size %d (%s)\n",
223 tdb->map_size, strerror(errno)));
233 /* expand a file. we prefer to use ftruncate, as that is what posix
234 says to use for mmap expansion */
235 static int tdb_expand_file(struct tdb_context *tdb, tdb_off_t size, tdb_off_t addition)
239 if (tdb->read_only || tdb->traverse_read) {
240 tdb->ecode = TDB_ERR_RDONLY;
244 if (ftruncate(tdb->fd, size+addition) == -1) {
246 ssize_t written = pwrite(tdb->fd, &b, 1, (size+addition) - 1);
248 /* try once more, potentially revealing errno */
249 written = pwrite(tdb->fd, &b, 1, (size+addition) - 1);
252 /* again - give up, guessing errno */
256 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file to %d failed (%s)\n",
257 size+addition, strerror(errno)));
262 /* now fill the file with something. This ensures that the
263 file isn't sparse, which would be very bad if we ran out of
264 disk. This must be done with write, not via mmap */
265 memset(buf, TDB_PAD_BYTE, sizeof(buf));
267 size_t n = addition>sizeof(buf)?sizeof(buf):addition;
268 ssize_t written = pwrite(tdb->fd, buf, n, size);
270 /* prevent infinite loops: try _once_ more */
271 written = pwrite(tdb->fd, buf, n, size);
274 /* give up, trying to provide a useful errno */
275 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write "
276 "returned 0 twice: giving up!\n"));
279 } else if (written == -1) {
280 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write of "
281 "%d bytes failed (%s)\n", (int)n,
284 } else if (written != n) {
285 TDB_LOG((tdb, TDB_DEBUG_WARNING, "expand_file: wrote "
286 "only %d of %d bytes - retrying\n", (int)written,
296 /* expand the database at least size bytes by expanding the underlying
297 file and doing the mmap again if necessary */
298 int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
300 struct tdb_record rec;
301 tdb_off_t offset, new_size;
303 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
304 TDB_LOG((tdb, TDB_DEBUG_ERROR, "lock failed in tdb_expand\n"));
308 /* must know about any previous expansions by another process */
309 tdb->methods->tdb_oob(tdb, tdb->map_size + 1, 1);
311 /* always make room for at least 100 more records, and at
312 least 25% more space. Round the database up to a multiple
314 new_size = MAX(tdb->map_size + size*100, tdb->map_size * 1.25);
315 size = TDB_ALIGN(new_size, tdb->page_size) - tdb->map_size;
317 if (!(tdb->flags & TDB_INTERNAL))
321 * We must ensure the file is unmapped before doing this
322 * to ensure consistency with systems like OpenBSD where
323 * writes and mmaps are not consistent.
326 /* expand the file itself */
327 if (!(tdb->flags & TDB_INTERNAL)) {
328 if (tdb->methods->tdb_expand_file(tdb, tdb->map_size, size) != 0)
332 tdb->map_size += size;
334 if (tdb->flags & TDB_INTERNAL) {
335 char *new_map_ptr = (char *)realloc(tdb->map_ptr,
338 tdb->map_size -= size;
341 tdb->map_ptr = new_map_ptr;
344 * We must ensure the file is remapped before adding the space
345 * to ensure consistency with systems like OpenBSD where
346 * writes and mmaps are not consistent.
349 /* We're ok if the mmap fails as we'll fallback to read/write */
353 /* form a new freelist record */
354 memset(&rec,'\0',sizeof(rec));
355 rec.rec_len = size - sizeof(rec);
357 /* link it into the free list */
358 offset = tdb->map_size - size;
359 if (tdb_free(tdb, offset, &rec) == -1)
362 tdb_unlock(tdb, -1, F_WRLCK);
365 tdb_unlock(tdb, -1, F_WRLCK);
369 /* read/write a tdb_off_t */
370 int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
372 return tdb->methods->tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
375 int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
378 return tdb->methods->tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
382 /* read a lump of data, allocating the space for it */
383 unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len)
387 /* some systems don't like zero length malloc */
389 if (!(buf = (unsigned char *)malloc(len ? len : 1))) {
390 /* Ensure ecode is set for log fn. */
391 tdb->ecode = TDB_ERR_OOM;
392 TDB_LOG((tdb, TDB_DEBUG_ERROR,"tdb_alloc_read malloc failed len=%d (%s)\n",
393 len, strerror(errno)));
396 if (tdb->methods->tdb_read(tdb, offset, buf, len, 0) == -1) {
403 /* Give a piece of tdb data to a parser */
405 int tdb_parse_data(struct tdb_context *tdb, TDB_DATA key,
406 tdb_off_t offset, tdb_len_t len,
407 int (*parser)(TDB_DATA key, TDB_DATA data,
416 if ((tdb->transaction == NULL) && (tdb->map_ptr != NULL)) {
418 * Optimize by avoiding the malloc/memcpy/free, point the
419 * parser directly at the mmap area.
421 if (tdb->methods->tdb_oob(tdb, offset+len, 0) != 0) {
424 data.dptr = offset + (unsigned char *)tdb->map_ptr;
425 return parser(key, data, private_data);
428 if (!(data.dptr = tdb_alloc_read(tdb, offset, len))) {
432 result = parser(key, data, private_data);
437 /* read/write a record */
438 int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec)
440 if (tdb->methods->tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
442 if (TDB_BAD_MAGIC(rec)) {
443 /* Ensure ecode is set for log fn. */
444 tdb->ecode = TDB_ERR_CORRUPT;
445 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
448 return tdb->methods->tdb_oob(tdb, rec->next+sizeof(*rec), 0);
451 int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct tdb_record *rec)
453 struct tdb_record r = *rec;
454 return tdb->methods->tdb_write(tdb, offset, CONVERT(r), sizeof(r));
457 static const struct tdb_methods io_methods = {
467 initialise the default methods table
469 void tdb_io_init(struct tdb_context *tdb)
471 tdb->methods = &io_methods;