aboutsummaryrefslogtreecommitdiff
path: root/fs/jbd/revoke.c
blob: 8ff5a7b89b9ae6ef852b86f62d32e1de6b0b8b7c (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
/*
 * linux/fs/jbd/revoke.c
 *
 * Written by Stephen C. Tweedie <sct@redhat.com>, 2000
 *
 * Copyright 2000 Red Hat corp --- All Rights Reserved
 *
 * This file is part of the Linux kernel and is made available under
 * the terms of the GNU General Public License, version 2, or at your
 * option, any later version, incorporated herein by reference.
 *
 * Journal revoke routines for the generic filesystem journaling code;
 * part of the ext2fs journaling system.
 *
 * Revoke is the mechanism used to prevent old log records for deleted
 * metadata from being replayed on top of newer data using the same
 * blocks.  The revoke mechanism is used in two separate places:
 *
 * + Commit: during commit we write the entire list of the current
 *   transaction's revoked blocks to the journal
 *
 * + Recovery: during recovery we record the transaction ID of all
 *   revoked blocks.  If there are multiple revoke records in the log
 *   for a single block, only the last one counts, and if there is a log
 *   entry for a block beyond the last revoke, then that log entry still
 *   gets replayed.
 *
 * We can get interactions between revokes and new log data within a
 * single transaction:
 *
 * Block is revoked and then journaled:
 *   The desired end result is the journaling of the new block, so we
 *   cancel the revoke before the transaction commits.
 *
 * Block is journaled and then revoked:
 *   The revoke must take precedence over the write of the block, so we
 *   need either to cancel the journal entry or to write the revoke
 *   later in the log than the log block.  In this case, we choose the
 *   latter: journaling a block cancels any revoke record for that block
 *   in the current transaction, so any revoke for that block in the
 *   transaction must have happened after the block was journaled and so
 *   the revoke must take precedence.
 *
 * Block is revoked and then written as data:
 *   The data write is allowed to succeed, but the revoke is _not_
 *   cancelled.  We still need to prevent old log records from
 *   overwriting the new data.  We don't even need to clear the revoke
 *   bit here.
 *
 * Revoke information on buffers is a tri-state value:
 *
 * RevokeValid clear:	no cached revoke status, need to look it up
 * RevokeValid set, Revoked clear:
 *			buffer has not been revoked, and cancel_revoke
 *			need do nothing.
 * RevokeValid set, Revoked set:
 *			buffer has been revoked.
 */

#ifndef __KERNEL__
#include "jfs_user.h"
#else
#include <linux/time.h>
#include <linux/fs.h>
#include <linux/jbd.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/list.h>
#include <linux/init.h>
#endif
#include <linux/log2.h>

static struct kmem_cache *revoke_record_cache;
static struct kmem_cache *revoke_table_cache;

/* Each revoke record represents one single revoked block.  During
   journal replay, this involves recording the transaction ID of the
   last transaction to revoke this block. */

struct jbd_revoke_record_s
{
	struct list_head  hash;
	tid_t		  sequence;	/* Used for recovery only */
	unsigned long	  blocknr;
};


/* The revoke table is just a simple hash table of revoke records. */
struct jbd_revoke_table_s
{
	/* It is conceivable that we might want a larger hash table
	 * for recovery.  Must be a power of two. */
	int		  hash_size;
	int		  hash_shift;
	struct list_head *hash_table;
};


#ifdef __KERNEL__
static void write_one_revoke_record(journal_t *, transaction_t *,
				    struct journal_head **, int *,
				    struct jbd_revoke_record_s *);
static void flush_descriptor(journal_t *, struct journal_head *, int);
#endif

/* Utility functions to maintain the revoke table */

/* Borrowed from buffer.c: this is a tried and tested block hash function */
static inline int hash(journal_t *journal, unsigned long block)
{
	struct jbd_revoke_table_s *table = journal->j_revoke;
	int hash_shift = table->hash_shift;

	return ((block << (hash_shift - 6)) ^
		(block >> 13) ^
		(block << (hash_shift - 12))) & (table->hash_size - 1);
}

static int insert_revoke_hash(journal_t *journal, unsigned long blocknr,
			      tid_t seq)
{
	struct list_head *hash_list;
	struct jbd_revoke_record_s *record;

repeat:
	record = kmem_cache_alloc(revoke_record_cache, GFP_NOFS);
	if (!record)
		goto oom;

	record->sequence = seq;
	record->blocknr = blocknr;
	hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];
	spin_lock(&journal->j_revoke_lock);
	list_add(&record->hash, hash_list);
	spin_unlock(&journal->j_revoke_lock);
	return 0;

oom:
	if (!journal_oom_retry)
		return -ENOMEM;
	jbd_debug(1, "ENOMEM in %s, retrying\n", __func__);
	yield();
	goto repeat;
}

/* Find a revoke record in the journal's hash table. */

static struct jbd_revoke_record_s *find_revoke_record(journal_t *journal,
						      unsigned long blocknr)
{
	struct list_head *hash_list;
	struct jbd_revoke_record_s *record;

	hash_list = &journal->j_revoke->hash_table[hash(journal, blocknr)];

	spin_lock(&journal->j_revoke_lock);
	record = (struct jbd_revoke_record_s *) hash_list->next;
	while (&(record->hash) != hash_list) {
		if (record->blocknr == blocknr) {
			spin_unlock(&journal->j_revoke_lock);
			return record;
		}
		record = (struct jbd_revoke_record_s *) record->hash.next;
	}
	spin_unlock(&journal->j_revoke_lock);
	return NULL;
}

int __init journal_init_revoke_caches(void)
{
	revoke_record_cache = kmem_cache_create("revoke_record",
					   sizeof(struct jbd_revoke_record_s),
					   0,
					   SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY,
					   NULL);
	if (!revoke_record_cache)
		return -ENOMEM;

	revoke_table_cache = kmem_cache_create("revoke_table",
					   sizeof(struct jbd_revoke_table_s),
					   0, SLAB_TEMPORARY, NULL);
	if (!revoke_table_cache) {
		kmem_cache_destroy(revoke_record_cache);
		revoke_record_cache = NULL;
		return -ENOMEM;
	}
	return 0;
}

void journal_destroy_revoke_caches(void)
{
	kmem_cache_destroy(revoke_record_cache);
	revoke_record_cache = NULL;
	kmem_cache_destroy(revoke_table_cache);
	revoke_table_cache = NULL;
}

static struct jbd_revoke_table_s *journal_init_revoke_table(int hash_size)
{
	int shift = 0;
	int tmp = hash_size;
	struct jbd_revoke_table_s *table;

	table = kmem_cache_alloc(revoke_table_cache, GFP_KERNEL);
	if (!table)
		goto out;

	while((tmp >>= 1UL) != 0UL)
		shift++;

	table->hash_size = hash_size;
	table->hash_shift = shift;
	table->hash_table =
		kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL);
	if (!table->hash_table) {
		kmem_cache_free(revoke_table_cache, table);
		table = NULL;
		goto out;
	}

	for (tmp = 0; tmp < hash_size; tmp++)
		INIT_LIST_HEAD(&table->hash_table[tmp]);

out:
	return table;
}

static void journal_destroy_revoke_table(struct jbd_revoke_table_s *table)
{
	int i;
	struct list_head *hash_list;

	for (i = 0; i < table->hash_size; i++) {
		hash_list = &table->hash_table[i];
		J_ASSERT(list_empty(hash_list));
	}

	kfree(table->hash_table);
	kmem_cache_free(revoke_table_cache, table);
}

/* Initialise the revoke table for a given journal to a given size. */
int journal_init_revoke(journal_t *journal, int hash_size)
{
	J_ASSERT(journal->j_revoke_table[0] == NULL);
	J_ASSERT(is_power_of_2(hash_size));

	journal->j_revoke_table[0] = journal_init_revoke_table(hash_size);
	if (!journal->j_revoke_table[0])
		goto fail0;

	journal->j_revoke_table[1] = journal_init_revoke_table(hash_size);
	if (!journal->j_revoke_table[1])
		goto fail1;

	journal->j_revoke = journal->j_revoke_table[1];

	spin_lock_init(&journal->j_revoke_lock);

	return 0;

fail1:
	journal_destroy_revoke_table(journal->j_revoke_table[0]);
fail0:
	return -ENOMEM;
}

/* Destroy a journal's revoke table.  The table must already be empty! */
void journal_destroy_revoke(journal_t *journal)
{
	journal->j_revoke = NULL;
	if (journal->j_revoke_table[0])
		journal_destroy_revoke_table(journal->j_revoke_table[0]);
	if (journal->j_revoke_table[1])
		journal_destroy_revoke_table(journal->j_revoke_table[1]);
}


#ifdef __KERNEL__

/*
 * journal_revoke: revoke a given buffer_head from the journal.  This
 * prevents the block from being replayed during recovery if we take a
 * crash after this current transaction commits.  Any subsequent
 * metadata writes of the buffer in this transaction cancel the
 * revoke.
 *
 * Note that this call may block --- it is up to the caller to make
 * sure that there are no further calls to journal_write_metadata
 * before the revoke is complete.  In ext3, this implies calling the
 * revoke before clearing the block bitmap when we are deleting
 * metadata.
 *
 * Revoke performs a journal_forget on any buffer_head passed in as a
 * parameter, but does _not_ forget the buffer_head if the bh was only
 * found implicitly.
 *
 * bh_in may not be a journalled buffer - it may have come off
 * the hash tables without an attached journal_head.
 *
 * If bh_in is non-zero, journal_revoke() will decrement its b_count
 * by one.
 */

int journal_revoke(handle_t *handle, unsigned long blocknr,
		   struct buffer_head *bh_in)
{
	struct buffer_head *bh = NULL;
	journal_t *journal;
	struct block_device *bdev;
	int err;

	might_sleep();
	if (bh_in)
		BUFFER_TRACE(bh_in, "enter");

	journal = handle->h_transaction->t_journal;
	if (!journal_set_features(journal, 0, 0, JFS_FEATURE_INCOMPAT_REVOKE)){
		J_ASSERT (!"Cannot set revoke feature!");
		return -EINVAL;
	}

	bdev = journal->j_fs_dev;
	bh = bh_in;

	if (!bh) {
		bh = __find_get_block(bdev, blocknr, journal->j_blocksize);
		if (bh)
			BUFFER_TRACE(bh, "found on hash");
	}
#ifdef JBD_EXPENSIVE_CHECKING
	else {
		struct buffer_head *bh2;

		/* If there is a different buffer_head lying around in
		 * memory anywhere... */
		bh2 = __find_get_block(bdev, blocknr, journal->j_blocksize);
		if (bh2) {
			/* ... and it has RevokeValid status... */
			if (bh2 != bh && buffer_revokevalid(bh2))
				/* ...then it better be revoked too,
				 * since it's illegal to create a revoke
				 * record against a buffer_head which is
				 * not marked revoked --- that would
				 * risk missing a subsequent revoke
				 * cancel. */
				J_ASSERT_BH(bh2, buffer_revoked(bh2));
			put_bh(bh2);
		}
	}
#endif

	/* We really ought not ever to revoke twice in a row without
           first having the revoke cancelled: it's illegal to free a
           block twice without allocating it in between! */
	if (bh) {
		if (!J_EXPECT_BH(bh, !buffer_revoked(bh),
				 "inconsistent data on disk")) {
			if (!bh_in)
				brelse(bh);
			return -EIO;
		}
		set_buffer_revoked(bh);
		set_buffer_revokevalid(bh);
		if (bh_in) {
			BUFFER_TRACE(bh_in, "call journal_forget");
			journal_forget(handle, bh_in);
		} else {
			BUFFER_TRACE(bh, "call brelse");
			__brelse(bh);
		}
	}

	jbd_debug(2, "insert revoke for block %lu, bh_in=%p\n", blocknr, bh_in);
	err = insert_revoke_hash(journal, blocknr,
				handle->h_transaction->t_tid);
	BUFFER_TRACE(bh_in, "exit");
	return err;
}

/*
 * Cancel an outstanding revoke.  For use only internally by the
 * journaling code (called from journal_get_write_access).
 *
 * We trust buffer_revoked() on the buffer if the buffer is already
 * being journaled: if there is no revoke pending on the buffer, then we
 * don't do anything here.
 *
 * This would break if it were possible for a buffer to be revoked and
 * discarded, and then reallocated within the same transaction.  In such
 * a case we would have lost the revoked bit, but when we arrived here
 * the second time we would still have a pending revoke to cancel.  So,
 * do not trust the Revoked bit on buffers unless RevokeValid is also
 * set.
 *
 * The caller must have the journal locked.
 */
int journal_cancel_revoke(handle_t *handle, struct journal_head *jh)
{
	struct jbd_revoke_record_s *record;
	journal_t *journal = handle->h_transaction->t_journal;
	int need_cancel;
	int did_revoke = 0;	/* akpm: debug */
	struct buffer_head *bh = jh2bh(jh);

	jbd_debug(4, "journal_head %p, cancelling revoke\n", jh);

	/* Is the existing Revoke bit valid?  If so, we trust it, and
	 * only perform the full cancel if the revoke bit is set.  If
	 * not, we can't trust the revoke bit, and we need to do the
	 * full search for a revoke record. */
	if (test_set_buffer_revokevalid(bh)) {
		need_cancel = test_clear_buffer_revoked(bh);
	} else {
		need_cancel = 1;
		clear_buffer_revoked(bh);
	}

	if (need_cancel) {
		record = find_revoke_record(journal, bh->b_blocknr);
		if (record) {
			jbd_debug(4, "cancelled existing revoke on "
				  "blocknr %llu\n", (unsigned long long)bh->b_blocknr);
			spin_lock(&journal->j_revoke_lock);
			list_del(&record->hash);
			spin_unlock(&journal->j_revoke_lock);
			kmem_cache_free(revoke_record_cache, record);
			did_revoke = 1;
		}
	}

#ifdef JBD_EXPENSIVE_CHECKING
	/* There better not be one left behind by now! */
	record = find_revoke_record(journal, bh->b_blocknr);
	J_ASSERT_JH(jh, record == NULL);
#endif

	/* Finally, have we just cleared revoke on an unhashed
	 * buffer_head?  If so, we'd better make sure we clear the
	 * revoked status on any hashed alias too, otherwise the revoke
	 * state machine will get very upset later on. */
	if (need_cancel) {
		struct buffer_head *bh2;
		bh2 = __find_get_block(bh->b_bdev, bh->b_blocknr, bh->b_size);
		if (bh2) {
			if (bh2 != bh)
				clear_buffer_revoked(bh2);
			__brelse(bh2);
		}
	}
	return did_revoke;
}

/* journal_switch_revoke table select j_revoke for next transaction
 * we do not want to suspend any processing until all revokes are
 * written -bzzz
 */
void journal_switch_revoke_table(journal_t *journal)
{
	int i;

	if (journal->j_revoke == journal->j_revoke_table[0])
		journal->j_revoke = journal->j_revoke_table[1];
	else
		journal->j_revoke = journal->j_revoke_table[0];

	for (i = 0; i < journal->j_revoke->hash_size; i++)
		INIT_LIST_HEAD(&journal->j_revoke->hash_table[i]);
}

/*
 * Write revoke records to the journal for all entries in the current
 * revoke hash, deleting the entries as we go.
 *
 * Called with the journal lock held.
 */

void journal_write_revoke_records(journal_t *journal,
				  transaction_t *transaction)
{
	struct journal_head *descriptor;
	struct jbd_revoke_record_s *record;
	struct jbd_revoke_table_s *revoke;
	struct list_head *hash_list;
	int i, offset, count;

	descriptor = NULL;
	offset = 0;
	count = 0;

	/* select revoke table for committing transaction */
	revoke = journal->j_revoke == journal->j_revoke_table[0] ?
		journal->j_revoke_table[1] : journal->j_revoke_table[0];

	for (i = 0; i < revoke->hash_size; i++) {
		hash_list = &revoke->hash_table[i];

		while (!list_empty(hash_list)) {
			record = (struct jbd_revoke_record_s *)
				hash_list->next;
			write_one_revoke_record(journal, transaction,
						&descriptor, &offset,
						record);
			count++;
			list_del(&record->hash);
			kmem_cache_free(revoke_record_cache, record);
		}
	}
	if (descriptor)
		flush_descriptor(journal, descriptor, offset);
	jbd_debug(1, "Wrote %d revoke records\n", count);
}

/*
 * Write out one revoke record.  We need to create a new descriptor
 * block if the old one is full or if we have not already created one.
 */

static void write_one_revoke_record(journal_t *journal,
				    transaction_t *transaction,
				    struct journal_head **descriptorp,
				    int *offsetp,
				    struct jbd_revoke_record_s *record)
{
	struct journal_head *descriptor;
	int offset;
	journal_header_t *header;

	/* If we are already aborting, this all becomes a noop.  We
           still need to go round the loop in
           journal_write_revoke_records in order to free all of the
           revoke records: only the IO to the journal is omitted. */
	if (is_journal_aborted(journal))
		return;

	descriptor = *descriptorp;
	offset = *offsetp;

	/* Make sure we have a descriptor with space left for the record */
	if (descriptor) {
		if (offset == journal->j_blocksize) {
			flush_descriptor(journal, descriptor, offset);
			descriptor = NULL;
		}
	}

	if (!descriptor) {
		descriptor = journal_get_descriptor_buffer(journal);
		if (!descriptor)
			return;
		header = (journal_header_t *) &jh2bh(descriptor)->b_data[0];
		header->h_magic     = cpu_to_be32(JFS_MAGIC_NUMBER);
		header->h_blocktype = cpu_to_be32(JFS_REVOKE_BLOCK);
		header->h_sequence  = cpu_to_be32(transaction->t_tid);

		/* Record it so that we can wait for IO completion later */
		JBUFFER_TRACE(descriptor, "file as BJ_LogCtl");
		journal_file_buffer(descriptor, transaction, BJ_LogCtl);

		offset = sizeof(journal_revoke_header_t);
		*descriptorp = descriptor;
	}

	* ((__be32 *)(&jh2bh(descriptor)->b_data[offset])) =
		cpu_to_be32(record->blocknr);
	offset += 4;
	*offsetp = offset;
}

/*
 * Flush a revoke descriptor out to the journal.  If we are aborting,
 * this is a noop; otherwise we are generating a buffer which needs to
 * be waited for during commit, so it has to go onto the appropriate
 * journal buffer list.
 */

static void flush_descriptor(journal_t *journal,
			     struct journal_head *descriptor,
			     int offset)
{
	journal_revoke_header_t *header;
	struct buffer_head *bh = jh2bh(descriptor);

	if (is_journal_aborted(journal)) {
		put_bh(bh);
		return;
	}

	header = (journal_revoke_header_t *) jh2bh(descriptor)->b_data;
	header->r_count = cpu_to_be32(offset);
	set_buffer_jwrite(bh);
	BUFFER_TRACE(bh, "write");
	set_buffer_dirty(bh);
	ll_rw_block(SWRITE, 1, &bh);
}
#endif

/*
 * Revoke support for recovery.
 *
 * Recovery needs to be able to:
 *
 *  record all revoke records, including the tid of the latest instance
 *  of each revoke in the journal
 *
 *  check whether a given block in a given transaction should be replayed
 *  (ie. has not been revoked by a revoke record in that or a subsequent
 *  transaction)
 *
 *  empty the revoke table after recovery.
 */

/*
 * First, setting revoke records.  We create a new revoke record for
 * every block ever revoked in the log as we scan it for recovery, and
 * we update the existing records if we find multiple revokes for a
 * single block.
 */

int journal_set_revoke(journal_t *journal,
		       unsigned long blocknr,
		       tid_t sequence)
{
	struct jbd_revoke_record_s *record;

	record = find_revoke_record(journal, blocknr);
	if (record) {
		/* If we have multiple occurrences, only record the
		 * latest sequence number in the hashed record */
		if (tid_gt(sequence, record->sequence))
			record->sequence = sequence;
		return 0;
	}
	return insert_revoke_hash(journal, blocknr, sequence);
}

/*
 * Test revoke records.  For a given block referenced in the log, has
 * that block been revoked?  A revoke record with a given transaction
 * sequence number revokes all blocks in that transaction and earlier
 * ones, but later transactions still need replayed.
 */

int journal_test_revoke(journal_t *journal,
			unsigned long blocknr,
			tid_t sequence)
{
	struct jbd_revoke_record_s *record;

	record = find_revoke_record(journal, blocknr);
	if (!record)
		return 0;
	if (tid_gt(sequence, record->sequence))
		return 0;
	return 1;
}

/*
 * Finally, once recovery is over, we need to clear the revoke table so
 * that it can be reused by the running filesystem.
 */

void journal_clear_revoke(journal_t *journal)
{
	int i;
	struct list_head *hash_list;
	struct jbd_revoke_record_s *record;
	struct jbd_revoke_table_s *revoke;

	revoke = journal->j_revoke;

	for (i = 0; i < revoke->hash_size; i++) {
		hash_list = &revoke->hash_table[i];
		while (!list_empty(hash_list)) {
			record = (struct jbd_revoke_record_s*) hash_list->next;
			list_del(&record->hash);
			kmem_cache_free(revoke_record_cache, record);
		}
	}
}