summaryrefslogtreecommitdiffstats
path: root/storage/maria/ma_locking.c
blob: 9084be1d29df1767d2e9ce55e0f58a682e3e7503 (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
/* Copyright (C) 2006 MySQL AB & MySQL Finland AB & TCX DataKonsult AB

   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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1335 USA */

/*
  Locking of Maria-tables.
  Must be first request before doing any furter calls to any Maria function.
  Is used to allow many process use the same non transactional Maria table
*/

#include "ma_ftdefs.h"

	/* lock table by F_UNLCK, F_RDLCK or F_WRLCK */

int maria_lock_database(MARIA_HA *info, int lock_type)
{
  int error;
  uint count;
  MARIA_SHARE *share= info->s;
  DBUG_ENTER("maria_lock_database");
  DBUG_PRINT("enter",("lock_type: %d  old lock %d  r_locks: %u  w_locks: %u "
                      "global_changed:  %d  open_count: %u  name: '%s'",
                      lock_type, info->lock_type, share->r_locks,
                      share->w_locks,
                      share->global_changed, share->state.open_count,
                      share->index_file_name.str));
  if (share->options & HA_OPTION_READ_ONLY_DATA ||
      info->lock_type == lock_type)
    DBUG_RETURN(0);
  if (lock_type == F_EXTRA_LCK)                 /* Used by TMP tables */
  {
    ++share->w_locks;
    ++share->tot_locks;
    info->lock_type= lock_type;
    DBUG_RETURN(0);
  }

  error=0;
  if (!info->intern_lock_locked)
    mysql_mutex_lock(&share->intern_lock);
  if (share->kfile.file >= 0)		/* May only be false on windows */
  {
    switch (lock_type) {
    case F_UNLCK:
      maria_ftparser_call_deinitializer(info);
      if (info->lock_type == F_RDLCK)
      {
	count= --share->r_locks;
        if (share->lock_restore_status)
          (*share->lock_restore_status)(info);
      }
      else
      {
	count= --share->w_locks;
        if (share->lock.update_status)
          _ma_update_status_with_lock(info);
      }
      --share->tot_locks;
      if (info->lock_type == F_WRLCK && !share->w_locks)
      {
        /* pages of transactional tables get flushed at Checkpoint */
        if (!share->base.born_transactional && !share->temporary &&
            _ma_flush_table_files(info,
                                  share->delay_key_write ? MARIA_FLUSH_DATA :
                                  MARIA_FLUSH_DATA | MARIA_FLUSH_INDEX,
                                  FLUSH_KEEP, FLUSH_KEEP))
          error= my_errno;
      }
      if (info->opt_flag & (READ_CACHE_USED | WRITE_CACHE_USED))
      {
	if (end_io_cache(&info->rec_cache))
	{
	  error= my_errno;
          _ma_set_fatal_error(info, error);
	}
      }
      if (!count)
      {
	DBUG_PRINT("info",("changed: %u  w_locks: %u",
			   (uint) share->changed, share->w_locks));
	if (share->changed && !share->w_locks)
	{
#ifdef HAVE_MMAP
          if ((share->mmaped_length !=
               share->state.state.data_file_length) &&
              (share->nonmmaped_inserts > MAX_NONMAPPED_INSERTS))
          {
            if (share->lock_key_trees)
              mysql_rwlock_wrlock(&share->mmap_lock);
            _ma_remap_file(info, share->state.state.data_file_length);
            share->nonmmaped_inserts= 0;
            if (share->lock_key_trees)
              mysql_rwlock_unlock(&share->mmap_lock);
          }
#endif
#ifdef MARIA_EXTERNAL_LOCKING
	  share->state.process= share->last_process=share->this_process;
	  share->state.unique=   info->last_unique=  info->this_unique;
	  share->state.update_count= info->last_loop= ++info->this_loop;
#endif
          /* transactional tables rather flush their state at Checkpoint */
          if (!share->base.born_transactional)
          {
            if (_ma_state_info_write_sub(share->kfile.file, &share->state,
                                         MA_STATE_INFO_WRITE_DONT_MOVE_OFFSET))
              error= my_errno;
            else
            {
              /* A value of 0 means below means "state flushed" */
              share->changed= 0;
            }
          }
	  if (maria_flush)
	  {
            if (_ma_sync_table_files(info))
	      error= my_errno;
	  }
	  else
	    share->not_flushed=1;
	  if (error)
            _ma_set_fatal_error(info, error);
	}
      }
      info->opt_flag&= ~(READ_CACHE_USED | WRITE_CACHE_USED);
      info->lock_type= F_UNLCK;
      break;
    case F_RDLCK:
      if (info->lock_type == F_WRLCK)
      {
        /*
          Change RW to READONLY

          mysqld does not turn write locks to read locks,
          so we're never here in mysqld.
        */
	share->w_locks--;
	share->r_locks++;
	info->lock_type=lock_type;
	break;
      }
#ifdef MARIA_EXTERNAL_LOCKING
      if (!share->r_locks && !share->w_locks)
      {
        /* note that a transactional table should not do this */
	if (_ma_state_info_read_dsk(share->kfile.file, &share->state))
	{
	  error=my_errno;
	  break;
	}
      }
#endif
      _ma_test_if_changed(info);
      share->r_locks++;
      share->tot_locks++;
      info->lock_type=lock_type;
      break;
    case F_WRLCK:
      if (info->lock_type == F_RDLCK)
      {						/* Change READONLY to RW */
	if (share->r_locks == 1)
	{
	  share->r_locks--;
	  share->w_locks++;
	  info->lock_type=lock_type;
	  break;
	}
      }
#ifdef MARIA_EXTERNAL_LOCKING
      if (!(share->options & HA_OPTION_READ_ONLY_DATA))
      {
	if (!share->w_locks)
	{
	  if (!share->r_locks)
	  {
            /*
              Note that transactional tables should not do this.
              If we enabled this code, we should make sure to skip it if
              born_transactional is true. We should not test
              now_transactional to decide if we can call
              _ma_state_info_read_dsk(), because it can temporarily be 0
              (TRUNCATE on a partitioned table) and thus it would make a state
              modification below without mutex, confusing a concurrent
              checkpoint running.
              Even if this code was enabled only for non-transactional tables:
              in scenario LOCK TABLE t1 WRITE; INSERT INTO t1; DELETE FROM t1;
              state on disk read by DELETE is obsolete as it was not flushed
              at the end of INSERT. MyISAM same. It however causes no issue as
              maria_delete_all_rows() calls _ma_reset_status() thus is not
              influenced by the obsolete read values.
            */
	    if (_ma_state_info_read_dsk(share->kfile.file, &share->state))
	    {
	      error=my_errno;
	      break;
	    }
	  }
	}
      }
#endif /* defined(MARIA_EXTERNAL_LOCKING) */
      _ma_test_if_changed(info);

      info->lock_type=lock_type;
      info->invalidator=share->invalidator;
      share->w_locks++;
      share->tot_locks++;
      break;
    default:
      DBUG_ASSERT(0);
      break;				/* Impossible */
    }
  }
#ifdef _WIN32
  else
  {
    /*
       Check for bad file descriptors if this table is part
       of a merge union. Failing to capture this may cause
       a crash on windows if the table is renamed and
       later on referenced by the merge table.
     */
    if( info->owned_by_merge && (info->s)->kfile.file < 0 )
    {
      error = HA_ERR_NO_SUCH_TABLE;
    }
  }
#endif
  if (!info->intern_lock_locked)
    mysql_mutex_unlock(&share->intern_lock);
  DBUG_RETURN(error);
} /* maria_lock_database */


/****************************************************************************
 ** functions to read / write the state
****************************************************************************/

int _ma_readinfo(register MARIA_HA *info __attribute__ ((unused)),
                 int lock_type __attribute__ ((unused)),
                 int check_keybuffer __attribute__ ((unused)))
{
#ifdef MARIA_EXTERNAL_LOCKING
  DBUG_ENTER("_ma_readinfo");

  if (info->lock_type == F_UNLCK)
  {
    MARIA_SHARE *share= info->s;
    if (!share->tot_locks)
    {
      /* should not be done for transactional tables */
      if (_ma_state_info_read_dsk(share->kfile.file, &share->state))
      {
        if (!my_errno)
          my_errno= HA_ERR_FILE_TOO_SHORT;
	DBUG_RETURN(1);
      }
    }
    if (check_keybuffer)
      VOID(_ma_test_if_changed(info));
    info->invalidator=share->invalidator;
  }
  else if (lock_type == F_WRLCK && info->lock_type == F_RDLCK)
  {
    my_errno=EACCES;				/* Not allowed to change */
    DBUG_RETURN(-1);				/* when have read_lock() */
  }
  DBUG_RETURN(0);
#else
  return 0;
#endif /* defined(MARIA_EXTERNAL_LOCKING) */
} /* _ma_readinfo */


/*
  Every isam-function that updates the isam-database MUST end with this
  request

  NOTES
    my_errno is not changed if this succeeds!
*/

int _ma_writeinfo(register MARIA_HA *info, uint operation)
{
  int error,olderror;
  MARIA_SHARE *share= info->s;
  DBUG_ENTER("_ma_writeinfo");
  DBUG_PRINT("info",("operation: %u  tot_locks: %u", operation,
		     share->tot_locks));

  error=0;
  if (share->tot_locks == 0 && !share->base.born_transactional)
  {
    /* transactional tables flush their state at Checkpoint */
    if (operation)
    {					/* Two threads can't be here */
      CRASH_IF_S3_TABLE(info->s);       /* S3 readonly doesn't come here */

      olderror= my_errno;               /* Remember last error */

#ifdef MARIA_EXTERNAL_LOCKING
      /*
        The following only makes sense if we want to be allow two different
        processes access the same table at the same time
      */
      share->state.process= share->last_process=   share->this_process;
      share->state.unique=  info->last_unique=	   info->this_unique;
      share->state.update_count= info->last_loop= ++info->this_loop;
#endif

      if ((error=
           _ma_state_info_write_sub(share->kfile.file,
                                    &share->state,
                                    MA_STATE_INFO_WRITE_DONT_MOVE_OFFSET)))
	olderror=my_errno;
#ifdef _WIN32
      if (maria_flush)
      {
	_commit(share->kfile.file);
	_commit(info->dfile.file);
      }
#endif
      my_errno=olderror;
    }
  }
  else if (operation)
    share->changed= 1;			/* Mark keyfile changed */
  DBUG_RETURN(error);
} /* _ma_writeinfo */


/*
  Test if an external process has changed the database
  (Should be called after readinfo)
*/

int _ma_test_if_changed(register MARIA_HA *info)
{
#ifdef MARIA_EXTERNAL_LOCKING
  MARIA_SHARE *share= info->s;
  if (share->state.process != share->last_process ||
      share->state.unique  != info->last_unique ||
      share->state.update_count != info->last_loop)
  {						/* Keyfile has changed */
    DBUG_PRINT("info",("index file changed"));
    if (share->state.process != share->this_process)
      VOID(flush_pagecache_blocks(share->pagecache, &share->kfile,
                                  FLUSH_RELEASE));
    share->last_process=share->state.process;
    info->last_unique=	share->state.unique;
    info->last_loop=	share->state.update_count;
    info->update|=	HA_STATE_WRITTEN;	/* Must use file on next */
    info->data_changed= 1;			/* For maria_is_changed */
    return 1;
  }
#endif
  return (!(info->update & HA_STATE_AKTIV) ||
	  (info->update & (HA_STATE_WRITTEN | HA_STATE_DELETED |
			   HA_STATE_KEY_CHANGED)));
} /* _ma_test_if_changed */


/*
  Put a mark in the .MAI file that someone is updating the table

  DOCUMENTATION
  state.open_count in the .MAI file is used the following way:
  - For the first change of the .MYI file in this process open_count is
    incremented by _ma_mark_file_changed(). (We have a write lock on the file
    when this happens)
  - In maria_close() it's decremented by _ma_decrement_open_count() if it
    was incremented in the same process.

  This mean that if we are the only process using the file, the open_count
  tells us if the MARIA file wasn't properly closed. (This is true if
  my_disable_locking is set).

  open_count is not maintained on disk for temporary tables.
*/

#define _MA_ALREADY_MARKED_FILE_CHANGED                                 \
  ((share->state.changed & STATE_CHANGED) && share->global_changed)

int _ma_mark_file_changed(register MARIA_SHARE *share)
{
  if (!share->base.born_transactional)
  {
    if (!_MA_ALREADY_MARKED_FILE_CHANGED)
    {
      int res= _ma_mark_file_changed_now(share);
      /*
        Ensure that STATE_NOT_ANALYZED is reset on table changes
      */
      share->state.changed|= (STATE_CHANGED | STATE_NOT_ANALYZED |
                              STATE_NOT_OPTIMIZED_KEYS);
      return res;
    }
  }
  else
  {
    /*
      For transactional tables, the table is marked changed when the first page
      is written. Here we just mark the state to be updated so that caller
      can do 'analyze table' and find that is has changed before any pages
      are written.
    */
    if (! test_all_bits(share->state.changed,
                        (STATE_CHANGED | STATE_NOT_ANALYZED |
                         STATE_NOT_OPTIMIZED_KEYS)))
    {
      mysql_mutex_lock(&share->intern_lock);
      share->state.changed|=(STATE_CHANGED | STATE_NOT_ANALYZED |
                            STATE_NOT_OPTIMIZED_KEYS);
      mysql_mutex_unlock(&share->intern_lock);
    }
  }
  return 0;
}

int _ma_mark_file_changed_now(register MARIA_SHARE *share)
{
  uchar buff[3];
  int error= 1;
  DBUG_ENTER("_ma_mark_file_changed_now");

  if (_MA_ALREADY_MARKED_FILE_CHANGED)
    DBUG_RETURN(0);
  mysql_mutex_lock(&share->intern_lock); /* recheck under mutex */
  if (! _MA_ALREADY_MARKED_FILE_CHANGED)
  {
    share->state.changed|=(STATE_CHANGED | STATE_NOT_ANALYZED |
                           STATE_NOT_OPTIMIZED_KEYS);
    if (!share->global_changed)
    {
      share->changed= share->global_changed= 1;
      share->state.open_count++;
    }
    /*
      Temp tables don't need an open_count as they are removed on crash.
      In theory transactional tables are fixed by log-based recovery, so don't
      need an open_count either, but if recovery has failed and logs have been
      removed (by maria-force-start-after-recovery-failures), we still need to
      detect dubious tables.
      If we didn't maintain open_count on disk for a table, after a crash
      we wouldn't know if it was closed at crash time (thus does not need a
      check) or not. So we would have to check all tables: overkill.
    */
    if (!share->temporary)
    {
      CRASH_IF_S3_TABLE(share);
      mi_int2store(buff,share->state.open_count);
      buff[2]=1;				/* Mark that it's changed */
      if (my_pwrite(share->kfile.file, buff, sizeof(buff),
                    sizeof(share->state.header) +
                    MARIA_FILE_OPEN_COUNT_OFFSET,
                    MYF(MY_NABP)))
        goto err;
    }
    /* Set uuid of file if not yet set (zerofilled file) */
    if (share->base.born_transactional &&
        !(share->state.org_changed & STATE_NOT_MOVABLE))
    {
      CRASH_IF_S3_TABLE(share);
      /* Lock table to current installation */
      if (_ma_set_uuid(share, 0) ||
          (share->state.create_rename_lsn == LSN_NEEDS_NEW_STATE_LSNS &&
           _ma_update_state_lsns_sub(share, LSN_IMPOSSIBLE,
                                     trnman_get_min_trid(),
                                     TRUE, TRUE)))
        goto err;
      share->state.changed|= STATE_NOT_MOVABLE;
      share->state.org_changed|= STATE_NOT_MOVABLE;
    }
  }
  error= 0;
err:
  mysql_mutex_unlock(&share->intern_lock);
  DBUG_RETURN(error);
#undef _MA_ALREADY_MARKED_FILE_CHANGED
}

/*
  Check that a region is all zero

  SYNOPSIS
    check_if_zero()
    pos		Start of memory to check
    length	length of memory region

  NOTES
    Used mainly to detect rows with wrong extent information
*/

my_bool _ma_check_if_zero(uchar *pos, size_t length)
{
  uchar *end;
  for (end= pos+ length; pos != end ; pos++)
    if (pos[0] != 0)
      return 1;
  return 0;
}

/*
  This is only called by close or by extra(HA_FLUSH) if the OS has the pwrite()
  call.  In these context the following code should be safe!
 */

int _ma_decrement_open_count(MARIA_HA *info, my_bool lock_tables)
{
  uchar buff[2];
  register MARIA_SHARE *share= info->s;
  int lock_error=0,write_error=0;
  DBUG_ENTER("_ma_decrement_open_count");

  if (share->global_changed)
  {
    uint old_lock=info->lock_type;
    share->global_changed=0;
    lock_error= (my_disable_locking || ! lock_tables ? 0 :
                 maria_lock_database(info, F_WRLCK));
    /* Its not fatal even if we couldn't get the lock ! */
    if (share->state.open_count > 0)
    {
      CRASH_IF_S3_TABLE(share);
      share->state.open_count--;
      share->changed= 1;                        /* We have to update state */
      /*
        For temporary tables that will just be deleted, we don't have
        to decrement state. For transactional tables the state will be
        updated in maria_close().
      */

      if (!share->temporary && !share->now_transactional)
      {
        mi_int2store(buff,share->state.open_count);
        write_error= (int) my_pwrite(share->kfile.file, buff, sizeof(buff),
                                     sizeof(share->state.header) +
                                     MARIA_FILE_OPEN_COUNT_OFFSET,
                                     MYF(MY_NABP));
      }
    }
    if (!lock_error && !my_disable_locking && lock_tables)
      lock_error=maria_lock_database(info,old_lock);
  }
  DBUG_RETURN(MY_TEST(lock_error || write_error));
}


/** @brief mark file as crashed */

void _ma_mark_file_crashed(MARIA_SHARE *share)
{
  uchar buff[2];
  DBUG_ENTER("_ma_mark_file_crashed");

  share->state.changed|= STATE_CRASHED;
  if (share->no_status_updates)
    DBUG_VOID_RETURN;                           /* Safety */

  mi_int2store(buff, share->state.changed);

  /*
    We can ignore the errors, as if the mark failed, there isn't anything
    else we can do;  The user should already have got an error that the
    table was crashed.
  */
  (void) my_pwrite(share->kfile.file, buff, sizeof(buff),
                   sizeof(share->state.header) +
                   MARIA_FILE_CHANGED_OFFSET,
                   MYF(MY_NABP));
  DBUG_VOID_RETURN;
}


/**
   @brief Set uuid of for a Maria file

   @fn _ma_set_uuid()
   @param share		Maria share
   @param reset_uuid    Instead of setting file to maria_uuid, set it to
			0 to mark it as movable
*/

my_bool _ma_set_uuid(MARIA_SHARE *share, my_bool reset_uuid)
{
  uchar buff[MY_UUID_SIZE], *uuid;

  uuid= maria_uuid;
  if (reset_uuid)
  {
    bzero(buff, sizeof(buff));
    uuid= buff;
  }
  CRASH_IF_S3_TABLE(share);
  return (my_bool) my_pwrite(share->kfile.file, uuid, MY_UUID_SIZE,
                             mi_uint2korr(share->state.header.base_pos),
                             MYF(MY_NABP));
}