summaryrefslogtreecommitdiff
path: root/storage/innodb_plugin/row/row0uins.c
blob: 9f9c814f1a5ae01d6af4b5a6252774935a0389bb (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
/*****************************************************************************

Copyright (c) 1997, 2009, Innobase Oy. All Rights Reserved.

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; version 2 of the License.

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, write to the Free Software Foundation, Inc., 59 Temple
Place, Suite 330, Boston, MA 02111-1307 USA

*****************************************************************************/

/**************************************************//**
@file row/row0uins.c
Fresh insert undo

Created 2/25/1997 Heikki Tuuri
*******************************************************/

#include "row0uins.h"

#ifdef UNIV_NONINL
#include "row0uins.ic"
#endif

#include "dict0dict.h"
#include "dict0boot.h"
#include "dict0crea.h"
#include "trx0undo.h"
#include "trx0roll.h"
#include "btr0btr.h"
#include "mach0data.h"
#include "row0undo.h"
#include "row0vers.h"
#include "trx0trx.h"
#include "trx0rec.h"
#include "row0row.h"
#include "row0upd.h"
#include "que0que.h"
#include "ibuf0ibuf.h"
#include "log0log.h"

/***************************************************************//**
Removes a clustered index record. The pcur in node was positioned on the
record, now it is detached.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_ins_remove_clust_rec(
/*==========================*/
	undo_node_t*	node)	/*!< in: undo node */
{
	btr_cur_t*	btr_cur;
	ibool		success;
	ulint		err;
	ulint		n_tries		= 0;
	mtr_t		mtr;

	mtr_start(&mtr);

	success = btr_pcur_restore_position(BTR_MODIFY_LEAF, &(node->pcur),
					    &mtr);
	ut_a(success);

	if (ut_dulint_cmp(node->table->id, DICT_INDEXES_ID) == 0) {
		ut_ad(node->trx->dict_operation_lock_mode == RW_X_LATCH);

		/* Drop the index tree associated with the row in
		SYS_INDEXES table: */

		dict_drop_index_tree(btr_pcur_get_rec(&(node->pcur)), &mtr);

		mtr_commit(&mtr);

		mtr_start(&mtr);

		success = btr_pcur_restore_position(BTR_MODIFY_LEAF,
						    &(node->pcur), &mtr);
		ut_a(success);
	}

	btr_cur = btr_pcur_get_btr_cur(&(node->pcur));

	success = btr_cur_optimistic_delete(btr_cur, &mtr);

	btr_pcur_commit_specify_mtr(&(node->pcur), &mtr);

	if (success) {
		trx_undo_rec_release(node->trx, node->undo_no);

		return(DB_SUCCESS);
	}
retry:
	/* If did not succeed, try pessimistic descent to tree */
	mtr_start(&mtr);

	success = btr_pcur_restore_position(BTR_MODIFY_TREE,
					    &(node->pcur), &mtr);
	ut_a(success);

	btr_cur_pessimistic_delete(&err, FALSE, btr_cur,
				   trx_is_recv(node->trx)
				   ? RB_RECOVERY
				   : RB_NORMAL, &mtr);

	/* The delete operation may fail if we have little
	file space left: TODO: easiest to crash the database
	and restart with more file space */

	if (err == DB_OUT_OF_FILE_SPACE
	    && n_tries < BTR_CUR_RETRY_DELETE_N_TIMES) {

		btr_pcur_commit_specify_mtr(&(node->pcur), &mtr);

		n_tries++;

		os_thread_sleep(BTR_CUR_RETRY_SLEEP_TIME);

		goto retry;
	}

	btr_pcur_commit_specify_mtr(&(node->pcur), &mtr);

	trx_undo_rec_release(node->trx, node->undo_no);

	return(err);
}

/***************************************************************//**
Removes a secondary index entry if found.
@return	DB_SUCCESS, DB_FAIL, or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_ins_remove_sec_low(
/*========================*/
	ulint		mode,	/*!< in: BTR_MODIFY_LEAF or BTR_MODIFY_TREE,
				depending on whether we wish optimistic or
				pessimistic descent down the index tree */
	dict_index_t*	index,	/*!< in: index */
	dtuple_t*	entry)	/*!< in: index entry to remove */
{
	btr_pcur_t	pcur;
	btr_cur_t*	btr_cur;
	ibool		found;
	ibool		success;
	ulint		err;
	mtr_t		mtr;

	log_free_check();
	mtr_start(&mtr);

	found = row_search_index_entry(index, entry, mode, &pcur, &mtr);

	btr_cur = btr_pcur_get_btr_cur(&pcur);

	if (!found) {
		/* Not found */

		btr_pcur_close(&pcur);
		mtr_commit(&mtr);

		return(DB_SUCCESS);
	}

	if (mode == BTR_MODIFY_LEAF) {
		success = btr_cur_optimistic_delete(btr_cur, &mtr);

		if (success) {
			err = DB_SUCCESS;
		} else {
			err = DB_FAIL;
		}
	} else {
		ut_ad(mode == BTR_MODIFY_TREE);

		/* No need to distinguish RB_RECOVERY here, because we
		are deleting a secondary index record: the distinction
		between RB_NORMAL and RB_RECOVERY only matters when
		deleting a record that contains externally stored
		columns. */
		ut_ad(!dict_index_is_clust(index));
		btr_cur_pessimistic_delete(&err, FALSE, btr_cur,
					   RB_NORMAL, &mtr);
	}

	btr_pcur_close(&pcur);
	mtr_commit(&mtr);

	return(err);
}

/***************************************************************//**
Removes a secondary index entry from the index if found. Tries first
optimistic, then pessimistic descent down the tree.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
static
ulint
row_undo_ins_remove_sec(
/*====================*/
	dict_index_t*	index,	/*!< in: index */
	dtuple_t*	entry)	/*!< in: index entry to insert */
{
	ulint	err;
	ulint	n_tries	= 0;

	/* Try first optimistic descent to the B-tree */

	err = row_undo_ins_remove_sec_low(BTR_MODIFY_LEAF, index, entry);

	if (err == DB_SUCCESS) {

		return(err);
	}

	/* Try then pessimistic descent to the B-tree */
retry:
	err = row_undo_ins_remove_sec_low(BTR_MODIFY_TREE, index, entry);

	/* The delete operation may fail if we have little
	file space left: TODO: easiest to crash the database
	and restart with more file space */

	if (err != DB_SUCCESS && n_tries < BTR_CUR_RETRY_DELETE_N_TIMES) {

		n_tries++;

		os_thread_sleep(BTR_CUR_RETRY_SLEEP_TIME);

		goto retry;
	}

	return(err);
}

/***********************************************************//**
Parses the row reference and other info in a fresh insert undo record. */
static
void
row_undo_ins_parse_undo_rec(
/*========================*/
	undo_node_t*	node)	/*!< in/out: row undo node */
{
	dict_index_t*	clust_index;
	byte*		ptr;
	undo_no_t	undo_no;
	dulint		table_id;
	ulint		type;
	ulint		dummy;
	ibool		dummy_extern;

	ut_ad(node);

	ptr = trx_undo_rec_get_pars(node->undo_rec, &type, &dummy,
				    &dummy_extern, &undo_no, &table_id);
	ut_ad(type == TRX_UNDO_INSERT_REC);
	node->rec_type = type;

	node->update = NULL;
	node->table = dict_table_get_on_id(table_id, node->trx);

	/* Skip the UNDO if we can't find the table or the .ibd file. */
	if (UNIV_UNLIKELY(node->table == NULL)) {
	} else if (UNIV_UNLIKELY(node->table->ibd_file_missing)) {
		node->table = NULL;
	} else {
		clust_index = dict_table_get_first_index(node->table);

		if (clust_index != NULL) {
			ptr = trx_undo_rec_get_row_ref(
				ptr, clust_index, &node->ref, node->heap);
		} else {
			ut_print_timestamp(stderr);
			fprintf(stderr, "  InnoDB: table ");
			ut_print_name(stderr, node->trx, TRUE,
				      node->table->name);
			fprintf(stderr, " has no indexes, "
				"ignoring the table\n");

			node->table = NULL;
		}
	}
}

/***********************************************************//**
Undoes a fresh insert of a row to a table. A fresh insert means that
the same clustered index unique key did not have any record, even delete
marked, at the time of the insert.  InnoDB is eager in a rollback:
if it figures out that an index record will be removed in the purge
anyway, it will remove it in the rollback.
@return	DB_SUCCESS or DB_OUT_OF_FILE_SPACE */
UNIV_INTERN
ulint
row_undo_ins(
/*=========*/
	undo_node_t*	node)	/*!< in: row undo node */
{
	ut_ad(node);
	ut_ad(node->state == UNDO_NODE_INSERT);

	row_undo_ins_parse_undo_rec(node);

	if (!node->table || !row_undo_search_clust_to_pcur(node)) {
		trx_undo_rec_release(node->trx, node->undo_no);

		return(DB_SUCCESS);
	}

	/* Iterate over all the indexes and undo the insert.*/

	/* Skip the clustered index (the first index) */
	node->index = dict_table_get_next_index(
		dict_table_get_first_index(node->table));

	while (node->index != NULL) {
		dtuple_t*	entry;
		ulint		err;

		entry = row_build_index_entry(node->row, node->ext,
					      node->index, node->heap);
		if (UNIV_UNLIKELY(!entry)) {
			/* The database must have crashed after
			inserting a clustered index record but before
			writing all the externally stored columns of
			that record.  Because secondary index entries
			are inserted after the clustered index record,
			we may assume that the secondary index record
			does not exist.  However, this situation may
			only occur during the rollback of incomplete
			transactions. */
			ut_a(trx_is_recv(node->trx));
		} else {
			err = row_undo_ins_remove_sec(node->index, entry);

			if (err != DB_SUCCESS) {

				return(err);
			}
		}

		node->index = dict_table_get_next_index(node->index);
	}

	return(row_undo_ins_remove_clust_rec(node));
}