summaryrefslogtreecommitdiffstats
path: root/src/util/dict_lmdb.c
blob: bed20e0c450c1571b4fd33a6d389ffb753774fc8 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
/*++
/* NAME
/*	dict_lmdb 3
/* SUMMARY
/*	dictionary manager interface to OpenLDAP LMDB files
/* SYNOPSIS
/*	#include <dict_lmdb.h>
/*
/*	extern size_t dict_lmdb_map_size;
/*
/*	DEFINE_DICT_LMDB_MAP_SIZE;
/*
/*	DICT	*dict_lmdb_open(path, open_flags, dict_flags)
/*	const char *name;
/*	const char *path;
/*	int	open_flags;
/*	int	dict_flags;
/* DESCRIPTION
/*	dict_lmdb_open() opens the named LMDB database and makes
/*	it available via the generic interface described in
/*	dict_open(3).
/*
/*	The dict_lmdb_map_size variable specifies the initial
/*	database memory map size.  When a map becomes full its size
/*	is doubled, and other programs pick up the size change.
/*
/*	This variable cannot be exported via the dict(3) API and
/*	must therefore be defined in the calling program by invoking
/*	the DEFINE_DICT_LMDB_MAP_SIZE macro at the global level.
/* DIAGNOSTICS
/*	Fatal errors: cannot open file, file write error, out of
/*	memory.
/*
/*	If a jump buffer is specified with dict_setjmp(), then the LMDB
/*	client will call dict_longjmp() to return to that execution
/*	context after a recoverable error.
/* BUGS
/*	The on-the-fly map resize operations require no concurrent
/*	activity in the same database by other threads in the same
/*	memory address space.
/* SEE ALSO
/*	dict(3) generic dictionary manager
/* LICENSE
/* .ad
/* .fi
/*	The Secure Mailer license must be distributed with this software.
/* AUTHOR(S)
/*	Howard Chu
/*	Symas Corporation
/*
/*	Wietse Venema
/*	IBM T.J. Watson Research
/*	P.O. Box 704
/*	Yorktown Heights, NY 10598, USA
/*
/*	Wietse Venema
/*	Google, Inc.
/*	111 8th Avenue
/*	New York, NY 10011, USA
/*--*/

#include <sys_defs.h>

#ifdef HAS_LMDB

/* System library. */

#include <sys/stat.h>
#include <string.h>
#include <unistd.h>
#include <limits.h>

/* Utility library. */

#include <msg.h>
#include <mymalloc.h>
#include <htable.h>
#include <iostuff.h>
#include <vstring.h>
#include <myflock.h>
#include <stringops.h>
#include <slmdb.h>
#include <dict.h>
#include <dict_lmdb.h>
#include <warn_stat.h>

/* Application-specific. */

typedef struct {
    DICT    dict;			/* generic members */
    SLMDB   slmdb;			/* sane LMDB API */
    VSTRING *key_buf;			/* key buffer */
    VSTRING *val_buf;			/* value buffer */
} DICT_LMDB;

 /*
  * The LMDB database filename suffix happens to equal our DICT_TYPE_LMDB
  * prefix, but that doesn't mean it is kosher to use DICT_TYPE_LMDB where a
  * suffix is needed, so we define an explicit suffix here.
  */
#define DICT_LMDB_SUFFIX	"lmdb"

 /*
  * Make a safe string copy that is guaranteed to be null-terminated.
  */
#define SCOPY(buf, data, size) \
    vstring_str(vstring_strncpy(buf ? buf : (buf = vstring_alloc(10)), data, size))

 /*
  * Postfix writers recover from a "map full" error by increasing the memory
  * map size with a factor DICT_LMDB_SIZE_INCR (up to some limit) and
  * retrying the transaction.
  * 
  * Each dict(3) API call is retried no more than a few times. For bulk-mode
  * transactions the number of retries is proportional to the size of the
  * address space.
  * 
  * We do not expose these details to the Postfix user interface. The purpose of
  * Postfix is to solve problems, not punt them to the user.
  */
#define DICT_LMDB_SIZE_INCR	2	/* Increase size by 1 bit on retry */
#define DICT_LMDB_SIZE_MAX	SSIZE_T_MAX

#define DICT_LMDB_API_RETRY_LIMIT 2	/* Retries per dict(3) API call */
#define DICT_LMDB_BULK_RETRY_LIMIT \
	((int) (2 * sizeof(size_t) * CHAR_BIT))	/* Retries per bulk-mode
						 * transaction */

/* #define msg_verbose 1 */

/* dict_lmdb_lookup - find database entry */

static const char *dict_lmdb_lookup(DICT *dict, const char *name)
{
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) dict;
    MDB_val mdb_key;
    MDB_val mdb_value;
    const char *result = 0;
    int     status;
    ssize_t klen;

    dict->error = 0;
    klen = strlen(name);

    /*
     * Sanity check.
     */
    if ((dict->flags & (DICT_FLAG_TRY1NULL | DICT_FLAG_TRY0NULL)) == 0)
	msg_panic("dict_lmdb_lookup: no DICT_FLAG_TRY1NULL | DICT_FLAG_TRY0NULL flag");

    /*
     * Optionally fold the key.
     */
    if (dict->flags & DICT_FLAG_FOLD_FIX) {
	if (dict->fold_buf == 0)
	    dict->fold_buf = vstring_alloc(10);
	vstring_strcpy(dict->fold_buf, name);
	name = lowercase(vstring_str(dict->fold_buf));
    }

    /*
     * Acquire a shared lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
      && myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_SHARED) < 0)
	msg_fatal("%s: lock dictionary: %m", dict->name);

    /*
     * See if this LMDB file was written with one null byte appended to key
     * and value.
     */
    if (dict->flags & DICT_FLAG_TRY1NULL) {
	mdb_key.mv_data = (void *) name;
	mdb_key.mv_size = klen + 1;
	status = slmdb_get(&dict_lmdb->slmdb, &mdb_key, &mdb_value);
	if (status == 0) {
	    dict->flags &= ~DICT_FLAG_TRY0NULL;
	    result = SCOPY(dict_lmdb->val_buf, mdb_value.mv_data,
			   mdb_value.mv_size);
	} else if (status != MDB_NOTFOUND) {
	    msg_fatal("error reading %s:%s: %s",
		      dict_lmdb->dict.type, dict_lmdb->dict.name,
		      mdb_strerror(status));
	}
    }

    /*
     * See if this LMDB file was written with no null byte appended to key
     * and value.
     */
    if (result == 0 && (dict->flags & DICT_FLAG_TRY0NULL)) {
	mdb_key.mv_data = (void *) name;
	mdb_key.mv_size = klen;
	status = slmdb_get(&dict_lmdb->slmdb, &mdb_key, &mdb_value);
	if (status == 0) {
	    dict->flags &= ~DICT_FLAG_TRY1NULL;
	    result = SCOPY(dict_lmdb->val_buf, mdb_value.mv_data,
			   mdb_value.mv_size);
	} else if (status != MDB_NOTFOUND) {
	    msg_fatal("error reading %s:%s: %s",
		      dict_lmdb->dict.type, dict_lmdb->dict.name,
		      mdb_strerror(status));
	}
    }

    /*
     * Release the shared lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
	&& myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_NONE) < 0)
	msg_fatal("%s: unlock dictionary: %m", dict->name);

    return (result);
}

/* dict_lmdb_update - add or update database entry */

static int dict_lmdb_update(DICT *dict, const char *name, const char *value)
{
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) dict;
    MDB_val mdb_key;
    MDB_val mdb_value;
    int     status;

    dict->error = 0;

    /*
     * Sanity check.
     */
    if ((dict->flags & (DICT_FLAG_TRY1NULL | DICT_FLAG_TRY0NULL)) == 0)
	msg_panic("dict_lmdb_update: no DICT_FLAG_TRY1NULL | DICT_FLAG_TRY0NULL flag");

    /*
     * Optionally fold the key.
     */
    if (dict->flags & DICT_FLAG_FOLD_FIX) {
	if (dict->fold_buf == 0)
	    dict->fold_buf = vstring_alloc(10);
	vstring_strcpy(dict->fold_buf, name);
	name = lowercase(vstring_str(dict->fold_buf));
    }
    mdb_key.mv_data = (void *) name;
    mdb_value.mv_data = (void *) value;
    mdb_key.mv_size = strlen(name);
    mdb_value.mv_size = strlen(value);

    /*
     * If undecided about appending a null byte to key and value, choose a
     * default depending on the platform.
     */
    if ((dict->flags & DICT_FLAG_TRY1NULL)
	&& (dict->flags & DICT_FLAG_TRY0NULL)) {
#ifdef LMDB_NO_TRAILING_NULL
	dict->flags &= ~DICT_FLAG_TRY1NULL;
#else
	dict->flags &= ~DICT_FLAG_TRY0NULL;
#endif
    }

    /*
     * Optionally append a null byte to key and value.
     */
    if (dict->flags & DICT_FLAG_TRY1NULL) {
	mdb_key.mv_size++;
	mdb_value.mv_size++;
    }

    /*
     * Acquire an exclusive lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
    && myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_EXCLUSIVE) < 0)
	msg_fatal("%s: lock dictionary: %m", dict->name);

    /*
     * Do the update.
     */
    status = slmdb_put(&dict_lmdb->slmdb, &mdb_key, &mdb_value,
	       (dict->flags & DICT_FLAG_DUP_REPLACE) ? 0 : MDB_NOOVERWRITE);
    if (status != 0) {
	if (status == MDB_KEYEXIST) {
	    if (dict->flags & DICT_FLAG_DUP_IGNORE)
		 /* void */ ;
	    else if (dict->flags & DICT_FLAG_DUP_WARN)
		msg_warn("%s:%s: duplicate entry: \"%s\"",
			 dict_lmdb->dict.type, dict_lmdb->dict.name, name);
	    else
		msg_fatal("%s:%s: duplicate entry: \"%s\"",
			  dict_lmdb->dict.type, dict_lmdb->dict.name, name);
	} else {
	    msg_fatal("error updating %s:%s: %s",
		      dict_lmdb->dict.type, dict_lmdb->dict.name,
		      mdb_strerror(status));
	}
    }

    /*
     * Release the exclusive lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
	&& myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_NONE) < 0)
	msg_fatal("%s: unlock dictionary: %m", dict->name);

    return (status);
}

/* dict_lmdb_delete - delete one entry from the dictionary */

static int dict_lmdb_delete(DICT *dict, const char *name)
{
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) dict;
    MDB_val mdb_key;
    int     status = 1;
    ssize_t klen;

    dict->error = 0;
    klen = strlen(name);

    /*
     * Sanity check.
     */
    if ((dict->flags & (DICT_FLAG_TRY1NULL | DICT_FLAG_TRY0NULL)) == 0)
	msg_panic("dict_lmdb_delete: no DICT_FLAG_TRY1NULL | DICT_FLAG_TRY0NULL flag");

    /*
     * Optionally fold the key.
     */
    if (dict->flags & DICT_FLAG_FOLD_FIX) {
	if (dict->fold_buf == 0)
	    dict->fold_buf = vstring_alloc(10);
	vstring_strcpy(dict->fold_buf, name);
	name = lowercase(vstring_str(dict->fold_buf));
    }

    /*
     * Acquire an exclusive lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
    && myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_EXCLUSIVE) < 0)
	msg_fatal("%s: lock dictionary: %m", dict->name);

    /*
     * See if this LMDB file was written with one null byte appended to key
     * and value.
     */
    if (dict->flags & DICT_FLAG_TRY1NULL) {
	mdb_key.mv_data = (void *) name;
	mdb_key.mv_size = klen + 1;
	status = slmdb_del(&dict_lmdb->slmdb, &mdb_key);
	if (status != 0) {
	    if (status == MDB_NOTFOUND)
		status = 1;
	    else
		msg_fatal("error deleting from %s:%s: %s",
			  dict_lmdb->dict.type, dict_lmdb->dict.name,
			  mdb_strerror(status));
	} else {
	    dict->flags &= ~DICT_FLAG_TRY0NULL;	/* found */
	}
    }

    /*
     * See if this LMDB file was written with no null byte appended to key
     * and value.
     */
    if (status > 0 && (dict->flags & DICT_FLAG_TRY0NULL)) {
	mdb_key.mv_data = (void *) name;
	mdb_key.mv_size = klen;
	status = slmdb_del(&dict_lmdb->slmdb, &mdb_key);
	if (status != 0) {
	    if (status == MDB_NOTFOUND)
		status = 1;
	    else
		msg_fatal("error deleting from %s:%s: %s",
			  dict_lmdb->dict.type, dict_lmdb->dict.name,
			  mdb_strerror(status));
	} else {
	    dict->flags &= ~DICT_FLAG_TRY1NULL;	/* found */
	}
    }

    /*
     * Release the exclusive lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
	&& myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_NONE) < 0)
	msg_fatal("%s: unlock dictionary: %m", dict->name);

    return (status);
}

/* dict_lmdb_sequence - traverse the dictionary */

static int dict_lmdb_sequence(DICT *dict, int function,
			              const char **key, const char **value)
{
    const char *myname = "dict_lmdb_sequence";
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) dict;
    MDB_val mdb_key;
    MDB_val mdb_value;
    MDB_cursor_op op;
    int     status;

    dict->error = 0;

    /*
     * Determine the seek function.
     */
    switch (function) {
    case DICT_SEQ_FUN_FIRST:
	op = MDB_FIRST;
	break;
    case DICT_SEQ_FUN_NEXT:
	op = MDB_NEXT;
	break;
    default:
	msg_panic("%s: invalid function: %d", myname, function);
    }

    /*
     * Acquire a shared lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
      && myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_SHARED) < 0)
	msg_fatal("%s: lock dictionary: %m", dict->name);

    /*
     * Database lookup.
     */
    status = slmdb_cursor_get(&dict_lmdb->slmdb, &mdb_key, &mdb_value, op);

    switch (status) {

	/*
	 * Copy the key and value so they are guaranteed null terminated.
	 */
    case 0:
	*key = SCOPY(dict_lmdb->key_buf, mdb_key.mv_data, mdb_key.mv_size);
	if (mdb_value.mv_data != 0 && mdb_value.mv_size > 0)
	    *value = SCOPY(dict_lmdb->val_buf, mdb_value.mv_data,
			   mdb_value.mv_size);
	else
	    *value = "";			/* XXX */
	break;

	/*
	 * End-of-database.
	 */
    case MDB_NOTFOUND:
	status = 1;
	/* Not: mdb_cursor_close(). Wrong abstraction level. */
	break;

	/*
	 * Bust.
	 */
    default:
	msg_fatal("error seeking %s:%s: %s",
		  dict_lmdb->dict.type, dict_lmdb->dict.name,
		  mdb_strerror(status));
    }

    /*
     * Release the shared lock.
     */
    if ((dict->flags & DICT_FLAG_LOCK)
	&& myflock(dict->lock_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_NONE) < 0)
	msg_fatal("%s: unlock dictionary: %m", dict->name);

    return (status);
}

/* dict_lmdb_close - disassociate from data base */

static void dict_lmdb_close(DICT *dict)
{
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) dict;

    slmdb_close(&dict_lmdb->slmdb);
    if (dict_lmdb->key_buf)
	vstring_free(dict_lmdb->key_buf);
    if (dict_lmdb->val_buf)
	vstring_free(dict_lmdb->val_buf);
    if (dict->fold_buf)
	vstring_free(dict->fold_buf);
    dict_free(dict);
}

/* dict_lmdb_longjmp - repeat bulk transaction */

static void dict_lmdb_longjmp(void *context, int val)
{
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) context;

    dict_longjmp(&dict_lmdb->dict, val);
}

/* dict_lmdb_notify - debug logging */

static void dict_lmdb_notify(void *context, int error_code,...)
{
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) context;
    va_list ap;

    va_start(ap, error_code);
    switch (error_code) {
    case MDB_SUCCESS:
	msg_info("database %s:%s: using size limit %lu during open",
		 dict_lmdb->dict.type, dict_lmdb->dict.name,
		 (unsigned long) va_arg(ap, size_t));
	break;
    case MDB_MAP_FULL:
	msg_info("database %s:%s: using size limit %lu after MDB_MAP_FULL",
		 dict_lmdb->dict.type, dict_lmdb->dict.name,
		 (unsigned long) va_arg(ap, size_t));
	break;
    case MDB_MAP_RESIZED:
	msg_info("database %s:%s: using size limit %lu after MDB_MAP_RESIZED",
		 dict_lmdb->dict.type, dict_lmdb->dict.name,
		 (unsigned long) va_arg(ap, size_t));
	break;
    case MDB_READERS_FULL:
	msg_info("database %s:%s: pausing after MDB_READERS_FULL",
		 dict_lmdb->dict.type, dict_lmdb->dict.name);
	break;
    default:
	msg_warn("unknown MDB error code: %d", error_code);
	break;
    }
    va_end(ap);
}

/* dict_lmdb_assert - report LMDB internal assertion failure */

static void dict_lmdb_assert(void *context, const char *text)
{
    DICT_LMDB *dict_lmdb = (DICT_LMDB *) context;

    msg_fatal("%s:%s: internal error: %s",
	      dict_lmdb->dict.type, dict_lmdb->dict.name, text);
}

/* dict_lmdb_open - open LMDB data base */

DICT   *dict_lmdb_open(const char *path, int open_flags, int dict_flags)
{
    DICT_LMDB *dict_lmdb;
    DICT   *dict;
    struct stat st;
    SLMDB   slmdb;
    char   *mdb_path;
    int     mdb_flags, slmdb_flags, status;
    int     db_fd;

    /*
     * Let the optimizer worry about eliminating redundant code.
     */
#define DICT_LMDB_OPEN_RETURN(d) do { \
	DICT *__d = (d); \
	myfree(mdb_path); \
	return (__d); \
    } while (0)

    mdb_path = concatenate(path, "." DICT_TYPE_LMDB, (char *) 0);

    /*
     * Impedance adapters.
     */
    mdb_flags = MDB_NOSUBDIR | MDB_NOLOCK;
    if (open_flags == O_RDONLY)
	mdb_flags |= MDB_RDONLY;

    slmdb_flags = 0;
    if (dict_flags & DICT_FLAG_BULK_UPDATE)
	slmdb_flags |= SLMDB_FLAG_BULK;

    /*
     * Security violation.
     * 
     * By default, LMDB 0.9.9 writes uninitialized heap memory to a
     * world-readable database file, as chunks of up to 4096 bytes. This is a
     * huge memory disclosure vulnerability: memory content that a program
     * does not intend to share ends up in a world-readable file. The content
     * of uninitialized heap memory depends on program execution history.
     * That history includes code execution in other libraries that are
     * linked into the program.
     * 
     * This is a problem whenever the user who writes the database file differs
     * from the user who reads the database file. For example, a privileged
     * writer and an unprivileged reader. In the case of Postfix, the
     * postmap(1) and postalias(1) commands would leak uninitialized heap
     * memory, as chunks of up to 4096 bytes, from a root-privileged process
     * that writes to a database file, to unprivileged processes that read
     * from that database file.
     * 
     * As a workaround the postmap(1) and postalias(1) commands turn on
     * MDB_WRITEMAP which disables the use of malloc() in LMDB. However, that
     * does not address several disclosures of stack memory. We don't enable
     * this workaround for Postfix databases are maintained by Postfix daemon
     * processes, because those are accessible only by the postfix user.
     * 
     * LMDB 0.9.10 by default does not write uninitialized heap memory to file
     * (specify MDB_NOMEMINIT to revert that change). We use the MDB_WRITEMAP
     * workaround for older LMDB versions.
     */
#ifndef MDB_NOMEMINIT
    if (dict_flags & DICT_FLAG_BULK_UPDATE)	/* XXX Good enough */
	mdb_flags |= MDB_WRITEMAP;
#endif

    /*
     * Gracefully handle most database open errors.
     */
    if ((status = slmdb_init(&slmdb, dict_lmdb_map_size, DICT_LMDB_SIZE_INCR,
			     DICT_LMDB_SIZE_MAX)) != 0
	|| (status = slmdb_open(&slmdb, mdb_path, open_flags, mdb_flags,
				slmdb_flags)) != 0) {
	/* This leaks a little memory that would have been used otherwise. */
	dict = dict_surrogate(DICT_TYPE_LMDB, path, open_flags, dict_flags,
		    "open database %s: %s", mdb_path, mdb_strerror(status));
	DICT_LMDB_OPEN_RETURN(dict);
    }

    /*
     * XXX Persistent locking belongs in mkmap_lmdb.
     * 
     * We just need to acquire exclusive access momentarily. This establishes
     * that no readers are accessing old (obsoleted by copy-on-write) txn
     * snapshots, so we are free to reuse all eligible old pages. Downgrade
     * the lock right after acquiring it. This is sufficient to keep out
     * other writers until we are done.
     */
    db_fd = slmdb_fd(&slmdb);
    if (dict_flags & DICT_FLAG_BULK_UPDATE) {
	if (myflock(db_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_EXCLUSIVE) < 0)
	    msg_fatal("%s: lock dictionary: %m", mdb_path);
	if (myflock(db_fd, MYFLOCK_STYLE_FCNTL, MYFLOCK_OP_SHARED) < 0)
	    msg_fatal("%s: unlock dictionary: %m", mdb_path);
    }

    /*
     * Bundle up. From here on no more assignments to slmdb.
     */
    dict_lmdb = (DICT_LMDB *) dict_alloc(DICT_TYPE_LMDB, path, sizeof(*dict_lmdb));
    dict_lmdb->slmdb = slmdb;
    dict_lmdb->dict.lookup = dict_lmdb_lookup;
    dict_lmdb->dict.update = dict_lmdb_update;
    dict_lmdb->dict.delete = dict_lmdb_delete;
    dict_lmdb->dict.sequence = dict_lmdb_sequence;
    dict_lmdb->dict.close = dict_lmdb_close;

    if (fstat(db_fd, &st) < 0)
	msg_fatal("dict_lmdb_open: fstat: %m");
    dict_lmdb->dict.lock_fd = dict_lmdb->dict.stat_fd = db_fd;
    dict_lmdb->dict.lock_type = MYFLOCK_STYLE_FCNTL;
    dict_lmdb->dict.mtime = st.st_mtime;
    dict_lmdb->dict.owner.uid = st.st_uid;
    dict_lmdb->dict.owner.status = (st.st_uid != 0);

    dict_lmdb->key_buf = 0;
    dict_lmdb->val_buf = 0;

    /*
     * Warn if the source file is newer than the indexed file, except when
     * the source file changed only seconds ago.
     */
    if ((dict_flags & DICT_FLAG_LOCK) != 0
	&& stat(path, &st) == 0
	&& st.st_mtime > dict_lmdb->dict.mtime
	&& st.st_mtime < time((time_t *) 0) - 100)
	msg_warn("database %s is older than source file %s", mdb_path, path);

#define DICT_LMDB_IMPL_FLAGS	(DICT_FLAG_FIXED | DICT_FLAG_MULTI_WRITER)

    dict_lmdb->dict.flags = dict_flags | DICT_LMDB_IMPL_FLAGS;
    if ((dict_flags & (DICT_FLAG_TRY0NULL | DICT_FLAG_TRY1NULL)) == 0)
	dict_lmdb->dict.flags |= (DICT_FLAG_TRY0NULL | DICT_FLAG_TRY1NULL);
    if (dict_flags & DICT_FLAG_FOLD_FIX)
	dict_lmdb->dict.fold_buf = vstring_alloc(10);

    if (dict_flags & DICT_FLAG_BULK_UPDATE)
	dict_jmp_alloc(&dict_lmdb->dict);

    /*
     * The following requests return an error result only if we have serious
     * memory corruption problem.
     */
    if (slmdb_control(&dict_lmdb->slmdb,
		    CA_SLMDB_CTL_API_RETRY_LIMIT(DICT_LMDB_API_RETRY_LIMIT),
		  CA_SLMDB_CTL_BULK_RETRY_LIMIT(DICT_LMDB_BULK_RETRY_LIMIT),
		      CA_SLMDB_CTL_LONGJMP_FN(dict_lmdb_longjmp),
		      CA_SLMDB_CTL_NOTIFY_FN(msg_verbose ?
				    dict_lmdb_notify : (SLMDB_NOTIFY_FN) 0),
		      CA_SLMDB_CTL_ASSERT_FN(dict_lmdb_assert),
		      CA_SLMDB_CTL_CB_CONTEXT((void *) dict_lmdb),
		      CA_SLMDB_CTL_END) != 0)
	msg_panic("dict_lmdb_open: slmdb_control: %m");

    if (msg_verbose)
	dict_lmdb_notify((void *) dict_lmdb, MDB_SUCCESS,
			 slmdb_curr_limit(&dict_lmdb->slmdb));

    DICT_LMDB_OPEN_RETURN(DICT_DEBUG (&dict_lmdb->dict));
}

#endif