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/>.
29 #include "tdb_private.h"
31 /* check for an out of bounds access - if it is out of bounds then
32 see if the database has been expanded by someone else and expand
34 note that "len" is the minimum length needed for the db
36 static int tdb_oob(struct tdb_context *tdb, tdb_off_t len, int probe)
39 if (len <= tdb->map_size)
41 if (tdb->flags & TDB_INTERNAL) {
43 /* Ensure ecode is set for log fn. */
44 tdb->ecode = TDB_ERR_IO;
45 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_oob len %d beyond internal malloc size %d\n",
46 (int)len, (int)tdb->map_size));
48 return TDB_ERRCODE(TDB_ERR_IO, -1);
51 if (fstat(tdb->fd, &st) == -1) {
52 return TDB_ERRCODE(TDB_ERR_IO, -1);
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));
62 return TDB_ERRCODE(TDB_ERR_IO, -1);
65 /* Unmap, update size, remap */
66 if (tdb_munmap(tdb) == -1)
67 return TDB_ERRCODE(TDB_ERR_IO, -1);
68 tdb->map_size = st.st_size;
73 /* write a lump of data at a specified offset */
74 static int tdb_write(struct tdb_context *tdb, tdb_off_t off,
75 const void *buf, tdb_len_t len)
81 if (tdb->read_only || tdb->traverse_read) {
82 tdb->ecode = TDB_ERR_RDONLY;
86 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0)
90 memcpy(off + (char *)tdb->map_ptr, buf, len);
91 } else if (pwrite(tdb->fd, buf, len, off) != (ssize_t)len) {
92 /* Ensure ecode is set for log fn. */
93 tdb->ecode = TDB_ERR_IO;
94 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_write failed at %d len=%d (%s)\n",
95 off, len, strerror(errno)));
96 return TDB_ERRCODE(TDB_ERR_IO, -1);
101 /* Endian conversion: we only ever deal with 4 byte quantities */
102 void *tdb_convert(void *buf, u32 size)
104 u32 i, *p = (u32 *)buf;
105 for (i = 0; i < size / 4; i++)
106 p[i] = TDB_BYTEREV(p[i]);
111 /* read a lump of data at a specified offset, maybe convert */
112 static int tdb_read(struct tdb_context *tdb, tdb_off_t off, void *buf,
113 tdb_len_t len, int cv)
115 if (tdb->methods->tdb_oob(tdb, off + len, 0) != 0) {
120 memcpy(buf, off + (char *)tdb->map_ptr, len);
122 ssize_t ret = pread(tdb->fd, buf, len, off);
123 if (ret != (ssize_t)len) {
124 /* Ensure ecode is set for log fn. */
125 tdb->ecode = TDB_ERR_IO;
126 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_read failed at %d "
127 "len=%d ret=%d (%s) map_size=%d\n",
128 (int)off, (int)len, (int)ret, strerror(errno),
129 (int)tdb->map_size));
130 return TDB_ERRCODE(TDB_ERR_IO, -1);
134 tdb_convert(buf, len);
142 do an unlocked scan of the hash table heads to find the next non-zero head. The value
143 will then be confirmed with the lock held
145 static void tdb_next_hash_chain(struct tdb_context *tdb, u32 *chain)
149 for (;h < tdb->header.hash_size;h++) {
150 if (0 != *(u32 *)(TDB_HASH_TOP(h) + (unsigned char *)tdb->map_ptr)) {
156 for (;h < tdb->header.hash_size;h++) {
157 if (tdb_ofs_read(tdb, TDB_HASH_TOP(h), &off) != 0 || off != 0) {
166 int tdb_munmap(struct tdb_context *tdb)
168 if (tdb->flags & TDB_INTERNAL)
173 int ret = munmap(tdb->map_ptr, tdb->map_size);
182 void tdb_mmap(struct tdb_context *tdb)
184 if (tdb->flags & TDB_INTERNAL)
188 if (!(tdb->flags & TDB_NOMMAP)) {
189 tdb->map_ptr = mmap(NULL, tdb->map_size,
190 PROT_READ|(tdb->read_only? 0:PROT_WRITE),
191 MAP_SHARED|MAP_FILE, tdb->fd, 0);
194 * NB. When mmap fails it returns MAP_FAILED *NOT* NULL !!!!
197 if (tdb->map_ptr == MAP_FAILED) {
199 TDB_LOG((tdb, TDB_DEBUG_WARNING, "tdb_mmap failed for size %d (%s)\n",
200 tdb->map_size, strerror(errno)));
210 /* expand a file. we prefer to use ftruncate, as that is what posix
211 says to use for mmap expansion */
212 static int tdb_expand_file(struct tdb_context *tdb, tdb_off_t size, tdb_off_t addition)
216 if (tdb->read_only || tdb->traverse_read) {
217 tdb->ecode = TDB_ERR_RDONLY;
221 if (ftruncate(tdb->fd, size+addition) == -1) {
223 ssize_t written = pwrite(tdb->fd, &b, 1, (size+addition) - 1);
225 /* try once more, potentially revealing errno */
226 written = pwrite(tdb->fd, &b, 1, (size+addition) - 1);
229 /* again - give up, guessing errno */
233 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file to %d failed (%s)\n",
234 size+addition, strerror(errno)));
239 /* now fill the file with something. This ensures that the
240 file isn't sparse, which would be very bad if we ran out of
241 disk. This must be done with write, not via mmap */
242 memset(buf, TDB_PAD_BYTE, sizeof(buf));
244 size_t n = addition>sizeof(buf)?sizeof(buf):addition;
245 ssize_t written = pwrite(tdb->fd, buf, n, size);
247 /* prevent infinite loops: try _once_ more */
248 written = pwrite(tdb->fd, buf, n, size);
251 /* give up, trying to provide a useful errno */
252 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write "
253 "returned 0 twice: giving up!\n"));
256 } else if (written == -1) {
257 TDB_LOG((tdb, TDB_DEBUG_FATAL, "expand_file write of "
258 "%d bytes failed (%s)\n", (int)n,
261 } else if (written != n) {
262 TDB_LOG((tdb, TDB_DEBUG_WARNING, "expand_file: wrote "
263 "only %d of %d bytes - retrying\n", (int)written,
273 /* expand the database at least size bytes by expanding the underlying
274 file and doing the mmap again if necessary */
275 int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
277 struct list_struct rec;
280 if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
281 TDB_LOG((tdb, TDB_DEBUG_ERROR, "lock failed in tdb_expand\n"));
285 /* must know about any previous expansions by another process */
286 tdb->methods->tdb_oob(tdb, tdb->map_size + 1, 1);
288 /* always make room for at least 10 more records, and round
289 the database up to a multiple of the page size */
290 size = TDB_ALIGN(tdb->map_size + size*10, tdb->page_size) - tdb->map_size;
292 if (!(tdb->flags & TDB_INTERNAL))
296 * We must ensure the file is unmapped before doing this
297 * to ensure consistency with systems like OpenBSD where
298 * writes and mmaps are not consistent.
301 /* expand the file itself */
302 if (!(tdb->flags & TDB_INTERNAL)) {
303 if (tdb->methods->tdb_expand_file(tdb, tdb->map_size, size) != 0)
307 tdb->map_size += size;
309 if (tdb->flags & TDB_INTERNAL) {
310 char *new_map_ptr = (char *)realloc(tdb->map_ptr,
313 tdb->map_size -= size;
316 tdb->map_ptr = new_map_ptr;
319 * We must ensure the file is remapped before adding the space
320 * to ensure consistency with systems like OpenBSD where
321 * writes and mmaps are not consistent.
324 /* We're ok if the mmap fails as we'll fallback to read/write */
328 /* form a new freelist record */
329 memset(&rec,'\0',sizeof(rec));
330 rec.rec_len = size - sizeof(rec);
332 /* link it into the free list */
333 offset = tdb->map_size - size;
334 if (tdb_free(tdb, offset, &rec) == -1)
337 tdb_unlock(tdb, -1, F_WRLCK);
340 tdb_unlock(tdb, -1, F_WRLCK);
344 /* read/write a tdb_off_t */
345 int tdb_ofs_read(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
347 return tdb->methods->tdb_read(tdb, offset, (char*)d, sizeof(*d), DOCONV());
350 int tdb_ofs_write(struct tdb_context *tdb, tdb_off_t offset, tdb_off_t *d)
353 return tdb->methods->tdb_write(tdb, offset, CONVERT(off), sizeof(*d));
357 /* read a lump of data, allocating the space for it */
358 unsigned char *tdb_alloc_read(struct tdb_context *tdb, tdb_off_t offset, tdb_len_t len)
362 /* some systems don't like zero length malloc */
367 if (!(buf = (unsigned char *)malloc(len))) {
368 /* Ensure ecode is set for log fn. */
369 tdb->ecode = TDB_ERR_OOM;
370 TDB_LOG((tdb, TDB_DEBUG_ERROR,"tdb_alloc_read malloc failed len=%d (%s)\n",
371 len, strerror(errno)));
372 return TDB_ERRCODE(TDB_ERR_OOM, buf);
374 if (tdb->methods->tdb_read(tdb, offset, buf, len, 0) == -1) {
381 /* Give a piece of tdb data to a parser */
383 int tdb_parse_data(struct tdb_context *tdb, TDB_DATA key,
384 tdb_off_t offset, tdb_len_t len,
385 int (*parser)(TDB_DATA key, TDB_DATA data,
394 if ((tdb->transaction == NULL) && (tdb->map_ptr != NULL)) {
396 * Optimize by avoiding the malloc/memcpy/free, point the
397 * parser directly at the mmap area.
399 if (tdb->methods->tdb_oob(tdb, offset+len, 0) != 0) {
402 data.dptr = offset + (unsigned char *)tdb->map_ptr;
403 return parser(key, data, private_data);
406 if (!(data.dptr = tdb_alloc_read(tdb, offset, len))) {
410 result = parser(key, data, private_data);
415 /* read/write a record */
416 int tdb_rec_read(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
418 if (tdb->methods->tdb_read(tdb, offset, rec, sizeof(*rec),DOCONV()) == -1)
420 if (TDB_BAD_MAGIC(rec)) {
421 /* Ensure ecode is set for log fn. */
422 tdb->ecode = TDB_ERR_CORRUPT;
423 TDB_LOG((tdb, TDB_DEBUG_FATAL,"tdb_rec_read bad magic 0x%x at offset=%d\n", rec->magic, offset));
424 return TDB_ERRCODE(TDB_ERR_CORRUPT, -1);
426 return tdb->methods->tdb_oob(tdb, rec->next+sizeof(*rec), 0);
429 int tdb_rec_write(struct tdb_context *tdb, tdb_off_t offset, struct list_struct *rec)
431 struct list_struct r = *rec;
432 return tdb->methods->tdb_write(tdb, offset, CONVERT(r), sizeof(r));
435 static const struct tdb_methods io_methods = {
445 initialise the default methods table
447 void tdb_io_init(struct tdb_context *tdb)
449 tdb->methods = &io_methods;