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 #define MAX(a,b) ((a) > (b) ? (a) : (b))
33 /* check for an out of bounds access - if it is out of bounds then
34 see if the database has been expanded by someone else and expand
36 note that "len" is the minimum length needed for the db
38 static int tdb_oob(struct tdb_context *tdb, tdb_off_t len, int probe)
41 if (len <= tdb->map_size)
43 if (tdb->flags & TDB_INTERNAL) {
45 /* Ensure ecode is set for log fn. */
46 tdb->ecode = TDB_ERR_IO;
47 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %d beyond internal malloc size %d\n",
48 (int)len, (int)tdb->map_size));
53 if (fstat(tdb->fd, &st) == -1) {
54 tdb->ecode = TDB_ERR_IO;
58 if (st.st_size < (size_t)len) {
60 /* Ensure ecode is set for log fn. */
61 tdb->ecode = TDB_ERR_IO;
62 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %d beyond eof at %d\n",
63 (int)len, (int)st.st_size));
68 /* Unmap, update size, remap */
69 if (tdb_munmap(tdb) == -1) {
70 tdb->ecode = TDB_ERR_IO;
73 tdb->map_size = st.st_size;
78 /* write a lump of data at a specified offset */
79 static int tdb_write(struct tdb_context *tdb, tdb_off_t off,
80 const void *buf, tdb_len_t len)
86 if (tdb->read_only || tdb->traverse_read) {
87 tdb->ecode = TDB_ERR_RDONLY;
91 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0)
95 memcpy(off + (char *)tdb->map_ptr, buf, len);
97 ssize_t written = pwrite(tdb->fd, buf, len, off);
98 if ((written != (ssize_t)len) && (written != -1)) {
100 tdb->ecode = TDB_ERR_IO;
101 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_write: wrote only "
102 "%d of %d bytes at %d, trying once more\n",
103 (int)written, len, off));
104 written = pwrite(tdb->fd, (const char *)buf+written,
109 /* Ensure ecode is set for log fn. */
110 tdb->ecode = TDB_ERR_IO;
111 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_write failed at %d "
112 "len=%d (%s)\n", off, len, strerror(errno)));
114 } else if (written != (ssize_t)len) {
115 tdb->ecode = TDB_ERR_IO;
116 TDB_LOG((tdb, TDB_DEBUG_FATAL, "tdb_write: failed to "
117 "write %d bytes at %d in two attempts\n",
125 /* Endian conversion: we only ever deal with 4 byte quantities */
126 void *tdb_convert(void *buf, uint32_t size)
128 uint32_t i, *p = (uint32_t *)buf;
129 for (i = 0; i < size / 4; i++)
130 p[i] = TDB_BYTEREV(p[i]);
135 /* read a lump of data at a specified offset, maybe convert */
136 static int tdb_read(struct tdb_context *tdb, tdb_off_t off, void *buf,
137 tdb_len_t len, int cv)
139 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0) {
144 memcpy(buf, off + (char *)tdb->map_ptr, len);
146 ssize_t ret = pread(tdb->fd, buf, len, off);
147 if (ret != (ssize_t)len) {
148 /* Ensure ecode is set for log fn. */
149 tdb->ecode = TDB_ERR_IO;
150 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_read failed at %d "
151 "len=%d ret=%d (%s) map_size=%d\n",
152 (int)off, (int)len, (int)ret, strerror(errno),
153 (int)tdb->map_size));
158 tdb_convert(buf, len);
166 do an unlocked scan of the hash table heads to find the next non-zero head. The value
167 will then be confirmed with the lock held
169 static void tdb_next_hash_chain(struct tdb_context *tdb, uint32_t *chain)
173 for (;h < tdb->header.hash_size;h++) {
174 if (0 != *(uint32_t *)(TDB_HASH_TOP(h) + (unsigned char *)tdb->map_ptr)) {
180 for (;h < tdb->header.hash_size;h++) {
181 if (tdb_ofs_read(tdb, TDB_HASH_TOP(h), &off) != 0 || off != 0) {
190 int tdb_munmap(struct tdb_context *tdb)
192 if (tdb->flags & TDB_INTERNAL)
197 int ret = munmap(tdb->map_ptr, tdb->map_size);
206 void tdb_mmap(struct tdb_context *tdb)
208 if (tdb->flags & TDB_INTERNAL)
212 if (!(tdb->flags & TDB_NOMMAP)) {
213 tdb->map_ptr = mmap(NULL, tdb->map_size,
214 PROT_READ|(tdb->read_only? 0:PROT_WRITE),
215 MAP_SHARED, tdb->fd, 0);
218 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
221 if (tdb->map_ptr == MAP_FAILED) {
223 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_mmap failed for size %d (%s)\n",
224 tdb->map_size, strerror(errno)));
234 /* expand a file. we prefer to use ftruncate, as that is what posix
235 says to use for mmap expansion */
236 static int tdb_expand_file(struct tdb_context *tdb, tdb_off_t size, tdb_off_t addition)
240 if (tdb->read_only || tdb->traverse_read) {
241 tdb->ecode = TDB_ERR_RDONLY;
245 if (ftruncate(tdb->fd, size+addition) == -1) {
247 ssize_t written = pwrite(tdb->fd, &b, 1, (size+addition) - 1);
249 /* try once more, potentially revealing errno */
250 written = pwrite(tdb->fd, &b, 1, (size+addition) - 1);
253 /* again - give up, guessing errno */
257 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file to %d failed (%s)\n",
258 size+addition, strerror(errno)));
263 /* now fill the file with something. This ensures that the
264 file isn't sparse, which would be very bad if we ran out of
265 disk. This must be done with write, not via mmap */
266 memset(buf, TDB_PAD_BYTE, sizeof(buf));
268 size_t n = addition>sizeof(buf)?sizeof(buf):addition;
269 ssize_t written = pwrite(tdb->fd, buf, n, size);
271 /* prevent infinite loops: try _once_ more */
272 written = pwrite(tdb->fd, buf, n, size);
275 /* give up, trying to provide a useful errno */
276 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write "
277 "returned 0 twice: giving up!\n"));
280 } else if (written == -1) {
281 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write of "
282 "%d bytes failed (%s)\n", (int)n,
285 } else if (written != n) {
286 TDB_LOG((tdb, TDB_DEBUG_WARNING, "expand_file: wrote "
287 "only %d of %d bytes - retrying\n", (int)written,
297 /* expand the database at least size bytes by expanding the underlying
298 file and doing the mmap again if necessary */
299 int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
301 struct list_struct rec;
302 tdb_off_t offset, new_size;
304 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
305 TDB_LOG((tdb, TDB_DEBUG_ERROR, "lock failed in tdb_expand\n"));
309 /* must know about any previous expansions by another process */
310 tdb->methods->tdb_oob(tdb, tdb->map_size + 1, 1);
312 /* always make room for at least 100 more records, and at
313 least 25% more space. Round the database up to a multiple
315 new_size = MAX(tdb->map_size + size*100, tdb->map_size * 1.25);
316 size = TDB_ALIGN(new_size, tdb->page_size) - tdb->map_size;
318 if (!(tdb->flags & TDB_INTERNAL))
322 * We must ensure the file is unmapped before doing this
323 * to ensure consistency with systems like OpenBSD where
324 * writes and mmaps are not consistent.
327 /* expand the file itself */
328 if (!(tdb->flags & TDB_INTERNAL)) {
329 if (tdb->methods->tdb_expand_file(tdb, tdb->map_size, size) != 0)
333 tdb->map_size += size;
335 if (tdb->flags & TDB_INTERNAL) {
336 char *new_map_ptr = (char *)realloc(tdb->map_ptr,
339 tdb->map_size -= size;
342 tdb->map_ptr = new_map_ptr;
345 * We must ensure the file is remapped before adding the space
346 * to ensure consistency with systems like OpenBSD where
347 * writes and mmaps are not consistent.
350 /* We're ok if the mmap fails as we'll fallback to read/write */
354 /* form a new freelist record */
355 memset(&rec,'\0',sizeof(rec));
356 rec.rec_len = size - sizeof(rec);
358 /* link it into the free list */
359 offset = tdb->map_size - size;
360 if (tdb_free(tdb, offset, &rec) == -1)
363 tdb_unlock(tdb, -1, F_WRLCK);
366 tdb_unlock(tdb, -1, F_WRLCK);
370 /* read/write a tdb_off_t */
371 int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
373 return tdb->methods->tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
376 int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
379 return tdb->methods->tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
383 /* read a lump of data, allocating the space for it */
384 unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len)
388 /* 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 list_struct *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 list_struct *rec)
453 struct list_struct 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;