summaryrefslogtreecommitdiff
path: root/source3/lib/dbwrap_ctdb.c
blob: 0de61ceb6a8372be090f03a74c477ee743c67316 (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
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
/* 
   Unix SMB/CIFS implementation.
   Database interface wrapper around ctdbd
   Copyright (C) Volker Lendecke 2007
   
   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 3 of the License, or
   (at your option) any later version.
   
   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.
   
   You should have received a copy of the GNU General Public License
   along with this program.  If not, see <http://www.gnu.org/licenses/>.
*/

#include "includes.h"
#ifdef CLUSTER_SUPPORT
#include "ctdb.h"
#include "ctdb_private.h"
#include "ctdbd_conn.h"

struct db_ctdb_transaction_handle {
	struct db_ctdb_ctx *ctx;
	bool in_replay;
	/* we store the reads and writes done under a transaction one
	   list stores both reads and writes, the other just writes
	*/
	struct ctdb_marshall_buffer *m_all;
	struct ctdb_marshall_buffer *m_write;
};

struct db_ctdb_ctx {
	struct tdb_wrap *wtdb;
	uint32 db_id;
	struct db_ctdb_transaction_handle *transaction;
};

struct db_ctdb_rec {
	struct db_ctdb_ctx *ctdb_ctx;
	struct ctdb_ltdb_header header;
};

static struct db_record *fetch_locked_internal(struct db_ctdb_ctx *ctx,
					       TALLOC_CTX *mem_ctx,
					       TDB_DATA key,
					       bool persistent);

static NTSTATUS tdb_error_to_ntstatus(struct tdb_context *tdb)
{
	NTSTATUS status;
	enum TDB_ERROR tret = tdb_error(tdb);

	switch (tret) {
	case TDB_ERR_EXISTS:
		status = NT_STATUS_OBJECT_NAME_COLLISION;
		break;
	case TDB_ERR_NOEXIST:
		status = NT_STATUS_OBJECT_NAME_NOT_FOUND;
		break;
	default:
		status = NT_STATUS_INTERNAL_DB_CORRUPTION;
		break;
	}

	return status;
}



/*
  form a ctdb_rec_data record from a key/data pair
  
  note that header may be NULL. If not NULL then it is included in the data portion
  of the record
 */
static struct ctdb_rec_data *db_ctdb_marshall_record(TALLOC_CTX *mem_ctx, uint32_t reqid,	
						  TDB_DATA key, 
						  struct ctdb_ltdb_header *header,
						  TDB_DATA data)
{
	size_t length;
	struct ctdb_rec_data *d;

	length = offsetof(struct ctdb_rec_data, data) + key.dsize + 
		data.dsize + (header?sizeof(*header):0);
	d = (struct ctdb_rec_data *)talloc_size(mem_ctx, length);
	if (d == NULL) {
		return NULL;
	}
	d->length = length;
	d->reqid = reqid;
	d->keylen = key.dsize;
	memcpy(&d->data[0], key.dptr, key.dsize);
	if (header) {
		d->datalen = data.dsize + sizeof(*header);
		memcpy(&d->data[key.dsize], header, sizeof(*header));
		memcpy(&d->data[key.dsize+sizeof(*header)], data.dptr, data.dsize);
	} else {
		d->datalen = data.dsize;
		memcpy(&d->data[key.dsize], data.dptr, data.dsize);
	}
	return d;
}


/* helper function for marshalling multiple records */
static struct ctdb_marshall_buffer *db_ctdb_marshall_add(TALLOC_CTX *mem_ctx, 
					       struct ctdb_marshall_buffer *m,
					       uint64_t db_id,
					       uint32_t reqid,
					       TDB_DATA key,
					       struct ctdb_ltdb_header *header,
					       TDB_DATA data)
{
	struct ctdb_rec_data *r;
	size_t m_size, r_size;
	struct ctdb_marshall_buffer *m2;

	r = db_ctdb_marshall_record(mem_ctx, reqid, key, header, data);
	if (r == NULL) {
		talloc_free(m);
		return NULL;
	}

	if (m == NULL) {
		m = talloc_zero_size(mem_ctx, offsetof(struct ctdb_marshall_buffer, data));
		if (m == NULL) {
			return NULL;
		}
		m->db_id = db_id;
	}

	m_size = talloc_get_size(m);
	r_size = talloc_get_size(r);

	m2 = talloc_realloc_size(mem_ctx, m,  m_size + r_size);
	if (m2 == NULL) {
		talloc_free(m);
		return NULL;
	}

	memcpy(m_size + (uint8_t *)m2, r, r_size);

	talloc_free(r);

	m2->count++;

	return m2;
}

/* we've finished marshalling, return a data blob with the marshalled records */
static TDB_DATA db_ctdb_marshall_finish(struct ctdb_marshall_buffer *m)
{
	TDB_DATA data;
	data.dptr = (uint8_t *)m;
	data.dsize = talloc_get_size(m);
	return data;
}

/* 
   loop over a marshalling buffer 
   
     - pass r==NULL to start
     - loop the number of times indicated by m->count
*/
static struct ctdb_rec_data *db_ctdb_marshall_loop_next(struct ctdb_marshall_buffer *m, struct ctdb_rec_data *r,
						     uint32_t *reqid,
						     struct ctdb_ltdb_header *header,
						     TDB_DATA *key, TDB_DATA *data)
{
	if (r == NULL) {
		r = (struct ctdb_rec_data *)&m->data[0];
	} else {
		r = (struct ctdb_rec_data *)(r->length + (uint8_t *)r);
	}

	if (reqid != NULL) {
		*reqid = r->reqid;
	}
	
	if (key != NULL) {
		key->dptr   = &r->data[0];
		key->dsize  = r->keylen;
	}
	if (data != NULL) {
		data->dptr  = &r->data[r->keylen];
		data->dsize = r->datalen;
		if (header != NULL) {
			data->dptr += sizeof(*header);
			data->dsize -= sizeof(*header);
		}
	}

	if (header != NULL) {
		if (r->datalen < sizeof(*header)) {
			return NULL;
		}
		*header = *(struct ctdb_ltdb_header *)&r->data[r->keylen];
	}

	return r;
}



/* start a transaction on a database */
static int db_ctdb_transaction_destructor(struct db_ctdb_transaction_handle *h)
{
	tdb_transaction_cancel(h->ctx->wtdb->tdb);
	return 0;
}

/* start a transaction on a database */
static int db_ctdb_transaction_fetch_start(struct db_ctdb_transaction_handle *h)
{
	struct db_record *rh;
	TDB_DATA key;
	TALLOC_CTX *tmp_ctx;
	const char *keyname = CTDB_TRANSACTION_LOCK_KEY;
	int ret;
	struct db_ctdb_ctx *ctx = h->ctx;
	TDB_DATA data;

	key.dptr = discard_const(keyname);
	key.dsize = strlen(keyname);

again:
	tmp_ctx = talloc_new(h);

	rh = fetch_locked_internal(ctx, tmp_ctx, key, true);
	if (rh == NULL) {
		DEBUG(0,(__location__ " Failed to fetch_lock database\n"));		
		talloc_free(tmp_ctx);
		return -1;
	}
	talloc_free(rh);

	ret = tdb_transaction_start(ctx->wtdb->tdb);
	if (ret != 0) {
		DEBUG(0,(__location__ " Failed to start tdb transaction\n"));
		talloc_free(tmp_ctx);
		return -1;
	}

	data = tdb_fetch(ctx->wtdb->tdb, key);
	if ((data.dptr == NULL) ||
	    (data.dsize < sizeof(struct ctdb_ltdb_header)) ||
	    ((struct ctdb_ltdb_header *)data.dptr)->dmaster != get_my_vnn()) {
		SAFE_FREE(data.dptr);
		tdb_transaction_cancel(ctx->wtdb->tdb);
		talloc_free(tmp_ctx);
		goto again;
	}

	SAFE_FREE(data.dptr);
	talloc_free(tmp_ctx);

	return 0;
}


/* start a transaction on a database */
static int db_ctdb_transaction_start(struct db_context *db)
{
	struct db_ctdb_transaction_handle *h;
	int ret;
	struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
							struct db_ctdb_ctx);

	if (!db->persistent) {
		DEBUG(0,("transactions not supported on non-persistent database 0x%08x\n", 
			 ctx->db_id));
		return -1;
	}

	if (ctx->transaction) {
		DEBUG(0,("Nested transactions not supported on db 0x%08x\n", ctx->db_id));
		return -1;
	}

	h = talloc_zero(db, struct db_ctdb_transaction_handle);
	if (h == NULL) {
		DEBUG(0,(__location__ " oom for transaction handle\n"));		
		return -1;
	}

	h->ctx = ctx;

	ret = db_ctdb_transaction_fetch_start(h);
	if (ret != 0) {
		talloc_free(h);
		return -1;
	}

	talloc_set_destructor(h, db_ctdb_transaction_destructor);

	ctx->transaction = h;

	DEBUG(5,(__location__ " Started transaction on db 0x%08x\n", ctx->db_id));

	return 0;
}



/*
  fetch a record inside a transaction
 */
static int db_ctdb_transaction_fetch(struct db_ctdb_ctx *db, 
				     TALLOC_CTX *mem_ctx, 
				     TDB_DATA key, TDB_DATA *data)
{
	struct db_ctdb_transaction_handle *h = db->transaction;

	*data = tdb_fetch(h->ctx->wtdb->tdb, key);

	if (data->dptr != NULL) {
		uint8_t *oldptr = (uint8_t *)data->dptr;
		data->dsize -= sizeof(struct ctdb_ltdb_header);
		data->dptr = (uint8 *)
			talloc_memdup(
				mem_ctx, data->dptr+sizeof(struct ctdb_ltdb_header),
				data->dsize);
		SAFE_FREE(oldptr);
		if (data->dptr == NULL) {
			return -1;
		}
	}

	if (!h->in_replay) {
		h->m_all = db_ctdb_marshall_add(h, h->m_all, h->ctx->db_id, 1, key, NULL, *data);
		if (h->m_all == NULL) {
			DEBUG(0,(__location__ " Failed to add to marshalling record\n"));
			data->dsize = 0;
			talloc_free(data->dptr);
			return -1;
		}
	}

	return 0;
}


static NTSTATUS db_ctdb_store_transaction(struct db_record *rec, TDB_DATA data, int flag);
static NTSTATUS db_ctdb_delete_transaction(struct db_record *rec);

static struct db_record *db_ctdb_fetch_locked_transaction(struct db_ctdb_ctx *ctx,
							  TALLOC_CTX *mem_ctx,
							  TDB_DATA key)
{
	struct db_record *result;
	TDB_DATA ctdb_data;

	if (!(result = talloc(mem_ctx, struct db_record))) {
		DEBUG(0, ("talloc failed\n"));
		return NULL;
	}

	result->private_data = ctx->transaction;

	result->key.dsize = key.dsize;
	result->key.dptr = (uint8 *)talloc_memdup(result, key.dptr, key.dsize);
	if (result->key.dptr == NULL) {
		DEBUG(0, ("talloc failed\n"));
		TALLOC_FREE(result);
		return NULL;
	}

	result->store = db_ctdb_store_transaction;
	result->delete_rec = db_ctdb_delete_transaction;

	ctdb_data = tdb_fetch(ctx->wtdb->tdb, key);
	if (ctdb_data.dptr == NULL) {
		/* create the record */
		result->value = tdb_null;
		return result;
	}

	result->value.dsize = ctdb_data.dsize - sizeof(struct ctdb_ltdb_header);
	result->value.dptr = NULL;

	if ((result->value.dsize != 0)
	    && !(result->value.dptr = (uint8 *)talloc_memdup(
			 result, ctdb_data.dptr + sizeof(struct ctdb_ltdb_header),
			 result->value.dsize))) {
		DEBUG(0, ("talloc failed\n"));
		TALLOC_FREE(result);
	}

	SAFE_FREE(ctdb_data.dptr);

	return result;
}


/*
  stores a record inside a transaction
 */
static int db_ctdb_transaction_store(struct db_ctdb_transaction_handle *h, 
				     TDB_DATA key, TDB_DATA data)
{
	TALLOC_CTX *tmp_ctx = talloc_new(h);
	int ret;
	TDB_DATA rec;
	struct ctdb_ltdb_header header;

	/* we need the header so we can update the RSN */
	rec = tdb_fetch(h->ctx->wtdb->tdb, key);
	if (rec.dptr == NULL) {
		/* the record doesn't exist - create one with us as dmaster.
		   This is only safe because we are in a transaction and this
		   is a persistent database */
		ZERO_STRUCT(header);
		header.dmaster = get_my_vnn();
	} else {
		memcpy(&header, rec.dptr, sizeof(struct ctdb_ltdb_header));
		SAFE_FREE(rec.dptr);
	}

	header.rsn++;

	if (!h->in_replay) {
		h->m_all = db_ctdb_marshall_add(h, h->m_all, h->ctx->db_id, 0, key, NULL, data);
		if (h->m_all == NULL) {
			DEBUG(0,(__location__ " Failed to add to marshalling record\n"));
			talloc_free(tmp_ctx);
			return -1;
		}
		
		h->m_write = db_ctdb_marshall_add(h, h->m_write, h->ctx->db_id, 0, key, &header, data);
		if (h->m_write == NULL) {
			DEBUG(0,(__location__ " Failed to add to marshalling record\n"));
			talloc_free(tmp_ctx);
			return -1;
		}
	}
	
	rec.dptr = talloc_size(tmp_ctx, data.dsize + sizeof(struct ctdb_ltdb_header));
	if (rec.dptr == NULL) {
		DEBUG(0,(__location__ " Failed to alloc record\n"));
		talloc_free(tmp_ctx);
		return -1;
	}
	memcpy(rec.dptr, &header, sizeof(struct ctdb_ltdb_header));
	memcpy(sizeof(struct ctdb_ltdb_header) + (uint8_t *)rec.dptr, data.dptr, data.dsize);

	ret = tdb_store(h->ctx->wtdb->tdb, key, rec, TDB_REPLACE);

	talloc_free(tmp_ctx);
	
	return ret;
}


/* 
   a record store inside a transaction
 */
static NTSTATUS db_ctdb_store_transaction(struct db_record *rec, TDB_DATA data, int flag)
{
	struct db_ctdb_transaction_handle *h = talloc_get_type_abort(
		rec->private_data, struct db_ctdb_transaction_handle);
	int ret;

	ret = db_ctdb_transaction_store(h, rec->key, data);
	if (ret != 0) {
		return tdb_error_to_ntstatus(h->ctx->wtdb->tdb);
	}
	return NT_STATUS_OK;
}

/* 
   a record delete inside a transaction
 */
static NTSTATUS db_ctdb_delete_transaction(struct db_record *rec)
{
	struct db_ctdb_transaction_handle *h = talloc_get_type_abort(
		rec->private_data, struct db_ctdb_transaction_handle);
	int ret;

	ret = db_ctdb_transaction_store(h, rec->key, tdb_null);
	if (ret != 0) {
		return tdb_error_to_ntstatus(h->ctx->wtdb->tdb);
	}
	return NT_STATUS_OK;
}


/*
  replay a transaction
 */
static int ctdb_replay_transaction(struct db_ctdb_transaction_handle *h)
{
	int ret, i;
	struct ctdb_rec_data *rec = NULL;

	h->in_replay = true;

	ret = db_ctdb_transaction_fetch_start(h);
	if (ret != 0) {
		return ret;
	}

	for (i=0;i<h->m_all->count;i++) {
		TDB_DATA key, data;

		rec = db_ctdb_marshall_loop_next(h->m_all, rec, NULL, NULL, &key, &data);
		if (rec == NULL) {
			DEBUG(0, (__location__ " Out of records in ctdb_replay_transaction?\n"));
			goto failed;
		}

		if (rec->reqid == 0) {
			/* its a store */
			if (db_ctdb_transaction_store(h, key, data) != 0) {
				goto failed;
			}
		} else {
			TDB_DATA data2;
			TALLOC_CTX *tmp_ctx = talloc_new(h);

			if (db_ctdb_transaction_fetch(h->ctx, tmp_ctx, key, &data2) != 0) {
				talloc_free(tmp_ctx);
				goto failed;
			}
			if (data2.dsize != data.dsize ||
			    memcmp(data2.dptr, data.dptr, data.dsize) != 0) {
				/* the record has changed on us - we have to give up */
				talloc_free(tmp_ctx);
				goto failed;
			}
			talloc_free(tmp_ctx);
		}
	}
	
	return 0;

failed:
	tdb_transaction_cancel(h->ctx->wtdb->tdb);
	return -1;
}


/*
  commit a transaction
 */
static int db_ctdb_transaction_commit(struct db_context *db)
{
	struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
							struct db_ctdb_ctx);
	NTSTATUS rets;
	int ret;
	int status;
	struct db_ctdb_transaction_handle *h = ctx->transaction;

	if (h == NULL) {
		DEBUG(0,(__location__ " transaction commit with no open transaction on db 0x%08x\n", ctx->db_id));
		return -1;
	}

	DEBUG(5,(__location__ " Commit transaction on db 0x%08x\n", ctx->db_id));

	if (h->m_write == NULL) {
		/* no changes were made */
		talloc_free(h);
		ctx->transaction = NULL;
		return 0;
	}

	talloc_set_destructor(h, NULL);

	/* our commit strategy is quite complex.

	   - we first try to commit the changes to all other nodes

	   - if that works, then we commit locally and we are done

	   - if a commit on another node fails, then we need to cancel
	     the transaction, then restart the transaction (thus
	     opening a window of time for a pending recovery to
	     complete), then replay the transaction, checking all the
	     reads and writes (checking that reads give the same data,
	     and writes succeed). Then we retry the transaction to the
	     other nodes
	*/

again:
	/* tell ctdbd to commit to the other nodes */
	rets = ctdbd_control_local(messaging_ctdbd_connection(), 
				  CTDB_CONTROL_TRANS2_COMMIT, h->ctx->db_id, 0,
				  db_ctdb_marshall_finish(h->m_write), NULL, NULL, &status);
	if (!NT_STATUS_IS_OK(rets) || status != 0) {
		tdb_transaction_cancel(h->ctx->wtdb->tdb);
		sleep(1);
		if (ctdb_replay_transaction(h) != 0) {
			DEBUG(0,(__location__ " Failed to replay transaction\n"));
			ctdbd_control_local(messaging_ctdbd_connection(), CTDB_CONTROL_TRANS2_ERROR,
					    h->ctx->db_id, CTDB_CTRL_FLAG_NOREPLY, 
					    tdb_null, NULL, NULL, NULL);
			h->ctx->transaction = NULL;
			talloc_free(h);
			ctx->transaction = NULL;
			return -1;
		}
		goto again;
	}

	/* do the real commit locally */
	ret = tdb_transaction_commit(h->ctx->wtdb->tdb);
	if (ret != 0) {
		DEBUG(0,(__location__ " Failed to commit transaction\n"));
		ctdbd_control_local(messaging_ctdbd_connection(), CTDB_CONTROL_TRANS2_ERROR, h->ctx->db_id, 
				    CTDB_CTRL_FLAG_NOREPLY, tdb_null, NULL, NULL, NULL);
		h->ctx->transaction = NULL;
		talloc_free(h);
		return ret;
	}

	/* tell ctdbd that we are finished with our local commit */
	ctdbd_control_local(messaging_ctdbd_connection(), CTDB_CONTROL_TRANS2_FINISHED, 
			    h->ctx->db_id, CTDB_CTRL_FLAG_NOREPLY, 
			    tdb_null, NULL, NULL, NULL);
	h->ctx->transaction = NULL;
	talloc_free(h);
	return 0;
}


/*
  cancel a transaction
 */
static int db_ctdb_transaction_cancel(struct db_context *db)
{
	struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
							struct db_ctdb_ctx);
	struct db_ctdb_transaction_handle *h = ctx->transaction;

	if (h == NULL) {
		DEBUG(0,(__location__ " transaction cancel with no open transaction on db 0x%08x\n", ctx->db_id));
		return -1;
	}

	DEBUG(5,(__location__ " Cancel transaction on db 0x%08x\n", ctx->db_id));

	ctx->transaction = NULL;
	talloc_free(h);
	return 0;
}


static NTSTATUS db_ctdb_store(struct db_record *rec, TDB_DATA data, int flag)
{
	struct db_ctdb_rec *crec = talloc_get_type_abort(
		rec->private_data, struct db_ctdb_rec);
	TDB_DATA cdata;
	int ret;

	cdata.dsize = sizeof(crec->header) + data.dsize;

	if (!(cdata.dptr = SMB_MALLOC_ARRAY(uint8, cdata.dsize))) {
		return NT_STATUS_NO_MEMORY;
	}

	memcpy(cdata.dptr, &crec->header, sizeof(crec->header));
	memcpy(cdata.dptr + sizeof(crec->header), data.dptr, data.dsize);

	ret = tdb_store(crec->ctdb_ctx->wtdb->tdb, rec->key, cdata, TDB_REPLACE);

	SAFE_FREE(cdata.dptr);

	return (ret == 0) ? NT_STATUS_OK
			  : tdb_error_to_ntstatus(crec->ctdb_ctx->wtdb->tdb);
}


/* for persistent databases the store is a bit different. We have to
   ask the ctdb daemon to push the record to all nodes after the
   store */
static NTSTATUS db_ctdb_store_persistent(struct db_record *rec, TDB_DATA data, int flag)
{
	struct db_ctdb_rec *crec;
	struct db_record *record;
	TDB_DATA cdata;
	int ret;
	NTSTATUS status;
	uint32_t count;
	int max_retries = lp_parm_int(-1, "dbwrap ctdb", "max store retries", 5);

	for (count = 0, status = NT_STATUS_UNSUCCESSFUL, record = rec;
	     (count < max_retries) && !NT_STATUS_IS_OK(status);
	     count++)
	{
		if (count > 0) {
			/* retry */
			/*
			 * There is a hack here: We use rec as a memory
			 * context and re-use it as the record struct ptr.
			 * We don't free the record data allocated
			 * in each turn. So all gets freed when the caller
			 * releases the original record. This is because
			 * we don't get the record passed in by reference
			 * in the first place and the caller relies on
			 * having to free the record himself.
			 */
			record = fetch_locked_internal(crec->ctdb_ctx,
						       rec,
						       rec->key,
						       true /* persistent */);
			if (record == NULL) {
				DEBUG(5, ("fetch_locked_internal failed.\n"));
				status = NT_STATUS_NO_MEMORY;
				break;
			}
		}

		crec = talloc_get_type_abort(record->private_data,
					     struct db_ctdb_rec);

		cdata.dsize = sizeof(crec->header) + data.dsize;

		if (!(cdata.dptr = SMB_MALLOC_ARRAY(uint8, cdata.dsize))) {
			return NT_STATUS_NO_MEMORY;
		}

		crec->header.rsn++;

		memcpy(cdata.dptr, &crec->header, sizeof(crec->header));
		memcpy(cdata.dptr + sizeof(crec->header), data.dptr, data.dsize);

		status = ctdbd_start_persistent_update(
				messaging_ctdbd_connection(),
				crec->ctdb_ctx->db_id,
				rec->key,
				cdata);

		if (NT_STATUS_IS_OK(status)) {
			ret = tdb_store(crec->ctdb_ctx->wtdb->tdb, rec->key,
					cdata, TDB_REPLACE);
			status = (ret == 0)
					? NT_STATUS_OK
					: tdb_error_to_ntstatus(
						crec->ctdb_ctx->wtdb->tdb);
		}

		/*
		 * release the lock *now* in order to prevent deadlocks.
		 *
		 * There is a tradeoff: Usually, the record is still locked
		 * after db->store operation. This lock is usually released
		 * via the talloc destructor with the TALLOC_FREE to
		 * the record. So we have two choices:
		 *
		 * - Either re-lock the record after the call to persistent_store
		 *   or cancel_persistent update and this way not changing any
		 *   assumptions callers may have about the state, but possibly
		 *   introducing new race conditions.
		 *
		 * - Or don't lock the record again but just remove the
		 *   talloc_destructor. This is less racy but assumes that
		 *   the lock is always released via TALLOC_FREE of the record.
		 *
		 * I choose the first variant for now since it seems less racy.
		 * We can't guarantee that we succeed in getting the lock
		 * anyways. The only real danger here is that a caller
		 * performs multiple store operations after a fetch_locked()
		 * which is currently not the case.
		 */
		tdb_chainunlock(crec->ctdb_ctx->wtdb->tdb, rec->key);
		talloc_set_destructor(record, NULL);

		/* now tell ctdbd to update this record on all other nodes */
		if (NT_STATUS_IS_OK(status)) {
			status = ctdbd_persistent_store(
					messaging_ctdbd_connection(),
					crec->ctdb_ctx->db_id,
					rec->key,
					cdata);
		} else {
			ctdbd_cancel_persistent_update(
					messaging_ctdbd_connection(),
					crec->ctdb_ctx->db_id,
					rec->key,
					cdata);
			break;
		}

		SAFE_FREE(cdata.dptr);
	} /* retry-loop */

	if (!NT_STATUS_IS_OK(status)) {
		DEBUG(5, ("ctdbd_persistent_store failed after "
			  "%d retries with error %s - giving up.\n",
			  count, nt_errstr(status)));
	}

	SAFE_FREE(cdata.dptr);

	return status;
}

static NTSTATUS db_ctdb_delete(struct db_record *rec)
{
	TDB_DATA data;

	/*
	 * We have to store the header with empty data. TODO: Fix the
	 * tdb-level cleanup
	 */

	ZERO_STRUCT(data);

	return db_ctdb_store(rec, data, 0);

}

static NTSTATUS db_ctdb_delete_persistent(struct db_record *rec)
{
	TDB_DATA data;

	/*
	 * We have to store the header with empty data. TODO: Fix the
	 * tdb-level cleanup
	 */

	ZERO_STRUCT(data);

	return db_ctdb_store_persistent(rec, data, 0);

}

static int db_ctdb_record_destr(struct db_record* data)
{
	struct db_ctdb_rec *crec = talloc_get_type_abort(
		data->private_data, struct db_ctdb_rec);

	DEBUG(10, (DEBUGLEVEL > 10
		   ? "Unlocking db %u key %s\n"
		   : "Unlocking db %u key %.20s\n",
		   (int)crec->ctdb_ctx->db_id,
		   hex_encode(data, (unsigned char *)data->key.dptr,
			      data->key.dsize)));

	if (tdb_chainunlock(crec->ctdb_ctx->wtdb->tdb, data->key) != 0) {
		DEBUG(0, ("tdb_chainunlock failed\n"));
		return -1;
	}

	return 0;
}

static struct db_record *fetch_locked_internal(struct db_ctdb_ctx *ctx,
					       TALLOC_CTX *mem_ctx,
					       TDB_DATA key,
					       bool persistent)
{
	struct db_record *result;
	struct db_ctdb_rec *crec;
	NTSTATUS status;
	TDB_DATA ctdb_data;
	int migrate_attempts = 0;

	if (ctx->transaction != NULL) {
		return db_ctdb_fetch_locked_transaction(ctx, mem_ctx, key);
	}

	if (!(result = talloc(mem_ctx, struct db_record))) {
		DEBUG(0, ("talloc failed\n"));
		return NULL;
	}

	if (!(crec = TALLOC_ZERO_P(result, struct db_ctdb_rec))) {
		DEBUG(0, ("talloc failed\n"));
		TALLOC_FREE(result);
		return NULL;
	}

	result->private_data = (void *)crec;
	crec->ctdb_ctx = ctx;

	result->key.dsize = key.dsize;
	result->key.dptr = (uint8 *)talloc_memdup(result, key.dptr, key.dsize);
	if (result->key.dptr == NULL) {
		DEBUG(0, ("talloc failed\n"));
		TALLOC_FREE(result);
		return NULL;
	}

	/*
	 * Do a blocking lock on the record
	 */
again:

	if (DEBUGLEVEL >= 10) {
		char *keystr = hex_encode(result, key.dptr, key.dsize);
		DEBUG(10, (DEBUGLEVEL > 10
			   ? "Locking db %u key %s\n"
			   : "Locking db %u key %.20s\n",
			   (int)crec->ctdb_ctx->db_id, keystr));
		TALLOC_FREE(keystr);
	}
	
	if (tdb_chainlock(ctx->wtdb->tdb, key) != 0) {
		DEBUG(3, ("tdb_chainlock failed\n"));
		TALLOC_FREE(result);
		return NULL;
	}

	if (persistent) {
		result->store = db_ctdb_store_persistent;
		result->delete_rec = db_ctdb_delete_persistent;
	} else {
		result->store = db_ctdb_store;
		result->delete_rec = db_ctdb_delete;
	}
	talloc_set_destructor(result, db_ctdb_record_destr);

	ctdb_data = tdb_fetch(ctx->wtdb->tdb, key);

	/*
	 * See if we have a valid record and we are the dmaster. If so, we can
	 * take the shortcut and just return it.
	 */

	if ((ctdb_data.dptr == NULL) ||
	    (ctdb_data.dsize < sizeof(struct ctdb_ltdb_header)) ||
	    ((struct ctdb_ltdb_header *)ctdb_data.dptr)->dmaster != get_my_vnn()
#if 0
	    || (random() % 2 != 0)
#endif
) {
		SAFE_FREE(ctdb_data.dptr);
		tdb_chainunlock(ctx->wtdb->tdb, key);
		talloc_set_destructor(result, NULL);

		migrate_attempts += 1;

		DEBUG(10, ("ctdb_data.dptr = %p, dmaster = %u (%u)\n",
			   ctdb_data.dptr, ctdb_data.dptr ?
			   ((struct ctdb_ltdb_header *)ctdb_data.dptr)->dmaster : -1,
			   get_my_vnn()));

		status = ctdbd_migrate(messaging_ctdbd_connection(),ctx->db_id, key);
		if (!NT_STATUS_IS_OK(status)) {
			DEBUG(5, ("ctdb_migrate failed: %s\n",
				  nt_errstr(status)));
			TALLOC_FREE(result);
			return NULL;
		}
		/* now its migrated, try again */
		goto again;
	}

	if (migrate_attempts > 10) {
		DEBUG(0, ("db_ctdb_fetch_locked needed %d attempts\n",
			  migrate_attempts));
	}

	memcpy(&crec->header, ctdb_data.dptr, sizeof(crec->header));

	result->value.dsize = ctdb_data.dsize - sizeof(crec->header);
	result->value.dptr = NULL;

	if ((result->value.dsize != 0)
	    && !(result->value.dptr = (uint8 *)talloc_memdup(
			 result, ctdb_data.dptr + sizeof(crec->header),
			 result->value.dsize))) {
		DEBUG(0, ("talloc failed\n"));
		TALLOC_FREE(result);
	}

	SAFE_FREE(ctdb_data.dptr);

	return result;
}

static struct db_record *db_ctdb_fetch_locked(struct db_context *db,
					      TALLOC_CTX *mem_ctx,
					      TDB_DATA key)
{
	struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
							struct db_ctdb_ctx);

	return fetch_locked_internal(ctx, mem_ctx, key, db->persistent);
}

/*
  fetch (unlocked, no migration) operation on ctdb
 */
static int db_ctdb_fetch(struct db_context *db, TALLOC_CTX *mem_ctx,
			 TDB_DATA key, TDB_DATA *data)
{
	struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
							struct db_ctdb_ctx);
	NTSTATUS status;
	TDB_DATA ctdb_data;

	if (ctx->transaction) {
		return db_ctdb_transaction_fetch(ctx, mem_ctx, key, data);
	}

	/* try a direct fetch */
	ctdb_data = tdb_fetch(ctx->wtdb->tdb, key);

	/*
	 * See if we have a valid record and we are the dmaster. If so, we can
	 * take the shortcut and just return it.
	 * we bypass the dmaster check for persistent databases
	 */
	if ((ctdb_data.dptr != NULL) &&
	    (ctdb_data.dsize >= sizeof(struct ctdb_ltdb_header)) &&
	    (db->persistent ||
	     ((struct ctdb_ltdb_header *)ctdb_data.dptr)->dmaster == get_my_vnn())) {
		/* we are the dmaster - avoid the ctdb protocol op */

		data->dsize = ctdb_data.dsize - sizeof(struct ctdb_ltdb_header);
		if (data->dsize == 0) {
			SAFE_FREE(ctdb_data.dptr);
			data->dptr = NULL;
			return 0;
		}

		data->dptr = (uint8 *)talloc_memdup(
			mem_ctx, ctdb_data.dptr+sizeof(struct ctdb_ltdb_header),
			data->dsize);

		SAFE_FREE(ctdb_data.dptr);

		if (data->dptr == NULL) {
			return -1;
		}
		return 0;
	}

	SAFE_FREE(ctdb_data.dptr);

	/* we weren't able to get it locally - ask ctdb to fetch it for us */
	status = ctdbd_fetch(messaging_ctdbd_connection(),ctx->db_id, key, mem_ctx, data);
	if (!NT_STATUS_IS_OK(status)) {
		DEBUG(5, ("ctdbd_fetch failed: %s\n", nt_errstr(status)));
		return -1;
	}

	return 0;
}

struct traverse_state {
	struct db_context *db;
	int (*fn)(struct db_record *rec, void *private_data);
	void *private_data;
};

static void traverse_callback(TDB_DATA key, TDB_DATA data, void *private_data)
{
	struct traverse_state *state = (struct traverse_state *)private_data;
	struct db_record *rec;
	TALLOC_CTX *tmp_ctx = talloc_new(state->db);
	/* we have to give them a locked record to prevent races */
	rec = db_ctdb_fetch_locked(state->db, tmp_ctx, key);
	if (rec && rec->value.dsize > 0) {
		state->fn(rec, state->private_data);
	}
	talloc_free(tmp_ctx);
}

static int traverse_persistent_callback(TDB_CONTEXT *tdb, TDB_DATA kbuf, TDB_DATA dbuf,
					void *private_data)
{
	struct traverse_state *state = (struct traverse_state *)private_data;
	struct db_record *rec;
	TALLOC_CTX *tmp_ctx = talloc_new(state->db);
	int ret = 0;
	/* we have to give them a locked record to prevent races */
	rec = db_ctdb_fetch_locked(state->db, tmp_ctx, kbuf);
	if (rec && rec->value.dsize > 0) {
		ret = state->fn(rec, state->private_data);
	}
	talloc_free(tmp_ctx);
	return ret;
}

static int db_ctdb_traverse(struct db_context *db,
			    int (*fn)(struct db_record *rec,
				      void *private_data),
			    void *private_data)
{
        struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
                                                        struct db_ctdb_ctx);
	struct traverse_state state;

	state.db = db;
	state.fn = fn;
	state.private_data = private_data;

	if (db->persistent) {
		/* for persistent databases we don't need to do a ctdb traverse,
		   we can do a faster local traverse */
		return tdb_traverse(ctx->wtdb->tdb, traverse_persistent_callback, &state);
	}


	ctdbd_traverse(ctx->db_id, traverse_callback, &state);
	return 0;
}

static NTSTATUS db_ctdb_store_deny(struct db_record *rec, TDB_DATA data, int flag)
{
	return NT_STATUS_MEDIA_WRITE_PROTECTED;
}

static NTSTATUS db_ctdb_delete_deny(struct db_record *rec)
{
	return NT_STATUS_MEDIA_WRITE_PROTECTED;
}

static void traverse_read_callback(TDB_DATA key, TDB_DATA data, void *private_data)
{
	struct traverse_state *state = (struct traverse_state *)private_data;
	struct db_record rec;
	rec.key = key;
	rec.value = data;
	rec.store = db_ctdb_store_deny;
	rec.delete_rec = db_ctdb_delete_deny;
	rec.private_data = state->db;
	state->fn(&rec, state->private_data);
}

static int traverse_persistent_callback_read(TDB_CONTEXT *tdb, TDB_DATA kbuf, TDB_DATA dbuf,
					void *private_data)
{
	struct traverse_state *state = (struct traverse_state *)private_data;
	struct db_record rec;
	rec.key = kbuf;
	rec.value = dbuf;
	rec.store = db_ctdb_store_deny;
	rec.delete_rec = db_ctdb_delete_deny;
	rec.private_data = state->db;

	if (rec.value.dsize <= sizeof(struct ctdb_ltdb_header)) {
		/* a deleted record */
		return 0;
	}
	rec.value.dsize -= sizeof(struct ctdb_ltdb_header);
	rec.value.dptr += sizeof(struct ctdb_ltdb_header);

	return state->fn(&rec, state->private_data);
}

static int db_ctdb_traverse_read(struct db_context *db,
				 int (*fn)(struct db_record *rec,
					   void *private_data),
				 void *private_data)
{
        struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
                                                        struct db_ctdb_ctx);
	struct traverse_state state;

	state.db = db;
	state.fn = fn;
	state.private_data = private_data;

	if (db->persistent) {
		/* for persistent databases we don't need to do a ctdb traverse,
		   we can do a faster local traverse */
		return tdb_traverse_read(ctx->wtdb->tdb, traverse_persistent_callback_read, &state);
	}

	ctdbd_traverse(ctx->db_id, traverse_read_callback, &state);
	return 0;
}

static int db_ctdb_get_seqnum(struct db_context *db)
{
        struct db_ctdb_ctx *ctx = talloc_get_type_abort(db->private_data,
                                                        struct db_ctdb_ctx);
	return tdb_get_seqnum(ctx->wtdb->tdb);
}

struct db_context *db_open_ctdb(TALLOC_CTX *mem_ctx,
				const char *name,
				int hash_size, int tdb_flags,
				int open_flags, mode_t mode)
{
	struct db_context *result;
	struct db_ctdb_ctx *db_ctdb;
	char *db_path;

	if (!lp_clustering()) {
		DEBUG(10, ("Clustering disabled -- no ctdb\n"));
		return NULL;
	}

	if (!(result = TALLOC_ZERO_P(mem_ctx, struct db_context))) {
		DEBUG(0, ("talloc failed\n"));
		TALLOC_FREE(result);
		return NULL;
	}

	if (!(db_ctdb = TALLOC_P(result, struct db_ctdb_ctx))) {
		DEBUG(0, ("talloc failed\n"));
		TALLOC_FREE(result);
		return NULL;
	}

	db_ctdb->transaction = NULL;

	if (!NT_STATUS_IS_OK(ctdbd_db_attach(messaging_ctdbd_connection(),name, &db_ctdb->db_id, tdb_flags))) {
		DEBUG(0, ("ctdbd_db_attach failed for %s\n", name));
		TALLOC_FREE(result);
		return NULL;
	}

	db_path = ctdbd_dbpath(messaging_ctdbd_connection(), db_ctdb, db_ctdb->db_id);

	result->persistent = ((tdb_flags & TDB_CLEAR_IF_FIRST) == 0);

	/* only pass through specific flags */
	tdb_flags &= TDB_SEQNUM;

	/* honor permissions if user has specified O_CREAT */
	if (open_flags & O_CREAT) {
		chmod(db_path, mode);
	}

	db_ctdb->wtdb = tdb_wrap_open(db_ctdb, db_path, hash_size, tdb_flags, O_RDWR, 0);
	if (db_ctdb->wtdb == NULL) {
		DEBUG(0, ("Could not open tdb %s: %s\n", db_path, strerror(errno)));
		TALLOC_FREE(result);
		return NULL;
	}
	talloc_free(db_path);

	result->private_data = (void *)db_ctdb;
	result->fetch_locked = db_ctdb_fetch_locked;
	result->fetch = db_ctdb_fetch;
	result->traverse = db_ctdb_traverse;
	result->traverse_read = db_ctdb_traverse_read;
	result->get_seqnum = db_ctdb_get_seqnum;
	result->transaction_start = db_ctdb_transaction_start;
	result->transaction_commit = db_ctdb_transaction_commit;
	result->transaction_cancel = db_ctdb_transaction_cancel;

	DEBUG(3,("db_open_ctdb: opened database '%s' with dbid 0x%x\n",
		 name, db_ctdb->db_id));

	return result;
}
#endif