journal.c 121 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
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
/*
** Write ahead logging implementation copyright Chris Mason 2000
**
** The background commits make this code very interelated, and 
** overly complex.  I need to rethink things a bit....The major players:
**
** journal_begin -- call with the number of blocks you expect to log.  
**                  If the current transaction is too
** 		    old, it will block until the current transaction is 
** 		    finished, and then start a new one.
**		    Usually, your transaction will get joined in with 
**                  previous ones for speed.
**
** journal_join  -- same as journal_begin, but won't block on the current 
**                  transaction regardless of age.  Don't ever call
**                  this.  Ever.  There are only two places it should be 
**                  called from, and they are both inside this file.
**
** journal_mark_dirty -- adds blocks into this transaction.  clears any flags 
**                       that might make them get sent to disk
**                       and then marks them BH_JDirty.  Puts the buffer head 
**                       into the current transaction hash.  
**
** journal_end -- if the current transaction is batchable, it does nothing
**                   otherwise, it could do an async/synchronous commit, or
**                   a full flush of all log and real blocks in the 
**                   transaction.
**
** flush_old_commits -- if the current transaction is too old, it is ended and 
**                      commit blocks are sent to disk.  Forces commit blocks 
**                      to disk for all backgrounded commits that have been 
**                      around too long.
**		     -- Note, if you call this as an immediate flush from 
**		        from within kupdate, it will ignore the immediate flag
*/

#include <linux/config.h>
#include <asm/uaccess.h>
#include <asm/system.h>

#include <linux/time.h>
#include <asm/semaphore.h>

#include <linux/vmalloc.h>
#include <linux/reiserfs_fs.h>

#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/fcntl.h>
#include <linux/stat.h>
#include <linux/string.h>
#include <linux/smp_lock.h>
#include <linux/buffer_head.h>
#include <linux/workqueue.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>

/* gets a struct reiserfs_journal_list * from a list head */
#define JOURNAL_LIST_ENTRY(h) (list_entry((h), struct reiserfs_journal_list, \
                               j_list))
#define JOURNAL_WORK_ENTRY(h) (list_entry((h), struct reiserfs_journal_list, \
                               j_working_list))

/* the number of mounted filesystems.  This is used to decide when to
** start and kill the commit workqueue
*/
static int reiserfs_mounted_fs_count;

static struct workqueue_struct *commit_wq;

71
72
73
#define JOURNAL_TRANS_HALF 1018	/* must be correct to keep the desc and commit
				   structs at 4k */
#define BUFNR 64		/*read ahead */
Linus Torvalds's avatar
Linus Torvalds committed
74
75
76
77

/* cnode stat bits.  Move these into reiserfs_fs.h */

#define BLOCK_FREED 2		/* this block was freed, and can't be written.  */
78
#define BLOCK_FREED_HOLDER 3	/* this block was freed during this transaction, and can't be written */
Linus Torvalds's avatar
Linus Torvalds committed
79
80
81
82
83
84
85

#define BLOCK_NEEDS_FLUSH 4	/* used in flush_journal_list */
#define BLOCK_DIRTIED 5

/* journal list state bits */
#define LIST_TOUCHED 1
#define LIST_DIRTY   2
86
#define LIST_COMMIT_PENDING  4	/* someone will commit this list */
Linus Torvalds's avatar
Linus Torvalds committed
87
88
89
90

/* flags for do_journal_end */
#define FLUSH_ALL   1		/* flush commit and real blocks */
#define COMMIT_NOW  2		/* end and commit this transaction */
91
92
93
94
95
96
97
98
99
100
101
102
103
104
#define WAIT        4		/* wait for the log blocks to hit the disk */

static int do_journal_end(struct reiserfs_transaction_handle *,
			  struct super_block *, unsigned long nblocks,
			  int flags);
static int flush_journal_list(struct super_block *s,
			      struct reiserfs_journal_list *jl, int flushall);
static int flush_commit_list(struct super_block *s,
			     struct reiserfs_journal_list *jl, int flushall);
static int can_dirty(struct reiserfs_journal_cnode *cn);
static int journal_join(struct reiserfs_transaction_handle *th,
			struct super_block *p_s_sb, unsigned long nblocks);
static int release_journal_dev(struct super_block *super,
			       struct reiserfs_journal *journal);
Linus Torvalds's avatar
Linus Torvalds committed
105
static int dirty_one_transaction(struct super_block *s,
106
				 struct reiserfs_journal_list *jl);
Linus Torvalds's avatar
Linus Torvalds committed
107
108
109
110
111
static void flush_async_commits(void *p);
static void queue_log_writer(struct super_block *s);

/* values for join in do_journal_begin_r */
enum {
112
113
114
	JBEGIN_REG = 0,		/* regular journal begin */
	JBEGIN_JOIN = 1,	/* join the running transaction if at all possible */
	JBEGIN_ABORT = 2,	/* called from cleanup code, ignores aborted flag */
Linus Torvalds's avatar
Linus Torvalds committed
115
116
117
};

static int do_journal_begin_r(struct reiserfs_transaction_handle *th,
118
119
			      struct super_block *p_s_sb,
			      unsigned long nblocks, int join);
Linus Torvalds's avatar
Linus Torvalds committed
120

121
122
123
124
125
static void init_journal_hash(struct super_block *p_s_sb)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	memset(journal->j_hash_table, 0,
	       JOURNAL_HASH_SIZE * sizeof(struct reiserfs_journal_cnode *));
Linus Torvalds's avatar
Linus Torvalds committed
126
127
128
129
130
131
132
}

/*
** clears BH_Dirty and sticks the buffer on the clean list.  Called because I can't allow refile_buffer to
** make schedule happen after I've freed a block.  Look at remove_from_transaction and journal_mark_freed for
** more details.
*/
133
134
135
136
137
138
139
static int reiserfs_clean_and_file_buffer(struct buffer_head *bh)
{
	if (bh) {
		clear_buffer_dirty(bh);
		clear_buffer_journal_test(bh);
	}
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
140
141
142
143
}

static void disable_barrier(struct super_block *s)
{
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
	REISERFS_SB(s)->s_mount_opt &= ~(1 << REISERFS_BARRIER_FLUSH);
	printk("reiserfs: disabling flush barriers on %s\n",
	       reiserfs_bdevname(s));
}

static struct reiserfs_bitmap_node *allocate_bitmap_node(struct super_block
							 *p_s_sb)
{
	struct reiserfs_bitmap_node *bn;
	static int id;

	bn = reiserfs_kmalloc(sizeof(struct reiserfs_bitmap_node), GFP_NOFS,
			      p_s_sb);
	if (!bn) {
		return NULL;
	}
	bn->data = reiserfs_kmalloc(p_s_sb->s_blocksize, GFP_NOFS, p_s_sb);
	if (!bn->data) {
		reiserfs_kfree(bn, sizeof(struct reiserfs_bitmap_node), p_s_sb);
		return NULL;
	}
	bn->id = id++;
	memset(bn->data, 0, p_s_sb->s_blocksize);
	INIT_LIST_HEAD(&bn->list);
	return bn;
}

static struct reiserfs_bitmap_node *get_bitmap_node(struct super_block *p_s_sb)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_bitmap_node *bn = NULL;
	struct list_head *entry = journal->j_bitmap_nodes.next;

	journal->j_used_bitmap_nodes++;
      repeat:

	if (entry != &journal->j_bitmap_nodes) {
		bn = list_entry(entry, struct reiserfs_bitmap_node, list);
		list_del(entry);
		memset(bn->data, 0, p_s_sb->s_blocksize);
		journal->j_free_bitmap_nodes--;
		return bn;
	}
	bn = allocate_bitmap_node(p_s_sb);
	if (!bn) {
		yield();
		goto repeat;
	}
	return bn;
Linus Torvalds's avatar
Linus Torvalds committed
193
194
}
static inline void free_bitmap_node(struct super_block *p_s_sb,
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
				    struct reiserfs_bitmap_node *bn)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	journal->j_used_bitmap_nodes--;
	if (journal->j_free_bitmap_nodes > REISERFS_MAX_BITMAP_NODES) {
		reiserfs_kfree(bn->data, p_s_sb->s_blocksize, p_s_sb);
		reiserfs_kfree(bn, sizeof(struct reiserfs_bitmap_node), p_s_sb);
	} else {
		list_add(&bn->list, &journal->j_bitmap_nodes);
		journal->j_free_bitmap_nodes++;
	}
}

static void allocate_bitmap_nodes(struct super_block *p_s_sb)
{
	int i;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_bitmap_node *bn = NULL;
	for (i = 0; i < REISERFS_MIN_BITMAP_NODES; i++) {
		bn = allocate_bitmap_node(p_s_sb);
		if (bn) {
			list_add(&bn->list, &journal->j_bitmap_nodes);
			journal->j_free_bitmap_nodes++;
		} else {
			break;	// this is ok, we'll try again when more are needed 
		}
	}
Linus Torvalds's avatar
Linus Torvalds committed
222
223
224
}

static int set_bit_in_list_bitmap(struct super_block *p_s_sb, int block,
225
226
227
228
				  struct reiserfs_list_bitmap *jb)
{
	int bmap_nr = block / (p_s_sb->s_blocksize << 3);
	int bit_nr = block % (p_s_sb->s_blocksize << 3);
Linus Torvalds's avatar
Linus Torvalds committed
229

230
231
232
233
234
	if (!jb->bitmaps[bmap_nr]) {
		jb->bitmaps[bmap_nr] = get_bitmap_node(p_s_sb);
	}
	set_bit(bit_nr, (unsigned long *)jb->bitmaps[bmap_nr]->data);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
235
236
237
}

static void cleanup_bitmap_list(struct super_block *p_s_sb,
238
239
240
241
242
243
244
245
246
247
248
249
				struct reiserfs_list_bitmap *jb)
{
	int i;
	if (jb->bitmaps == NULL)
		return;

	for (i = 0; i < SB_BMAP_NR(p_s_sb); i++) {
		if (jb->bitmaps[i]) {
			free_bitmap_node(p_s_sb, jb->bitmaps[i]);
			jb->bitmaps[i] = NULL;
		}
	}
Linus Torvalds's avatar
Linus Torvalds committed
250
251
252
253
254
255
}

/*
** only call this on FS unmount.
*/
static int free_list_bitmaps(struct super_block *p_s_sb,
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
			     struct reiserfs_list_bitmap *jb_array)
{
	int i;
	struct reiserfs_list_bitmap *jb;
	for (i = 0; i < JOURNAL_NUM_BITMAPS; i++) {
		jb = jb_array + i;
		jb->journal_list = NULL;
		cleanup_bitmap_list(p_s_sb, jb);
		vfree(jb->bitmaps);
		jb->bitmaps = NULL;
	}
	return 0;
}

static int free_bitmap_nodes(struct super_block *p_s_sb)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct list_head *next = journal->j_bitmap_nodes.next;
	struct reiserfs_bitmap_node *bn;

	while (next != &journal->j_bitmap_nodes) {
		bn = list_entry(next, struct reiserfs_bitmap_node, list);
		list_del(next);
		reiserfs_kfree(bn->data, p_s_sb->s_blocksize, p_s_sb);
		reiserfs_kfree(bn, sizeof(struct reiserfs_bitmap_node), p_s_sb);
		next = journal->j_bitmap_nodes.next;
		journal->j_free_bitmap_nodes--;
	}

	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
286
287
288
289
290
291
292
}

/*
** get memory for JOURNAL_NUM_BITMAPS worth of bitmaps. 
** jb_array is the array to be filled in.
*/
int reiserfs_allocate_list_bitmaps(struct super_block *p_s_sb,
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
				   struct reiserfs_list_bitmap *jb_array,
				   int bmap_nr)
{
	int i;
	int failed = 0;
	struct reiserfs_list_bitmap *jb;
	int mem = bmap_nr * sizeof(struct reiserfs_bitmap_node *);

	for (i = 0; i < JOURNAL_NUM_BITMAPS; i++) {
		jb = jb_array + i;
		jb->journal_list = NULL;
		jb->bitmaps = vmalloc(mem);
		if (!jb->bitmaps) {
			reiserfs_warning(p_s_sb,
					 "clm-2000, unable to allocate bitmaps for journal lists");
			failed = 1;
			break;
		}
		memset(jb->bitmaps, 0, mem);
	}
	if (failed) {
		free_list_bitmaps(p_s_sb, jb_array);
		return -1;
	}
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
318
319
320
321
322
323
}

/*
** find an available list bitmap.  If you can't find one, flush a commit list 
** and try again
*/
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
static struct reiserfs_list_bitmap *get_list_bitmap(struct super_block *p_s_sb,
						    struct reiserfs_journal_list
						    *jl)
{
	int i, j;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_list_bitmap *jb = NULL;

	for (j = 0; j < (JOURNAL_NUM_BITMAPS * 3); j++) {
		i = journal->j_list_bitmap_index;
		journal->j_list_bitmap_index = (i + 1) % JOURNAL_NUM_BITMAPS;
		jb = journal->j_list_bitmap + i;
		if (journal->j_list_bitmap[i].journal_list) {
			flush_commit_list(p_s_sb,
					  journal->j_list_bitmap[i].
					  journal_list, 1);
			if (!journal->j_list_bitmap[i].journal_list) {
				break;
			}
		} else {
			break;
		}
	}
	if (jb->journal_list) {	/* double check to make sure if flushed correctly */
		return NULL;
	}
	jb->journal_list = jl;
	return jb;
Linus Torvalds's avatar
Linus Torvalds committed
352
353
354
355
356
357
358
}

/* 
** allocates a new chunk of X nodes, and links them all together as a list.
** Uses the cnode->next and cnode->prev pointers
** returns NULL on failure
*/
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
static struct reiserfs_journal_cnode *allocate_cnodes(int num_cnodes)
{
	struct reiserfs_journal_cnode *head;
	int i;
	if (num_cnodes <= 0) {
		return NULL;
	}
	head = vmalloc(num_cnodes * sizeof(struct reiserfs_journal_cnode));
	if (!head) {
		return NULL;
	}
	memset(head, 0, num_cnodes * sizeof(struct reiserfs_journal_cnode));
	head[0].prev = NULL;
	head[0].next = head + 1;
	for (i = 1; i < num_cnodes; i++) {
		head[i].prev = head + (i - 1);
		head[i].next = head + (i + 1);	/* if last one, overwrite it after the if */
	}
	head[num_cnodes - 1].next = NULL;
	return head;
Linus Torvalds's avatar
Linus Torvalds committed
379
380
381
382
383
}

/*
** pulls a cnode off the free list, or returns NULL on failure 
*/
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
static struct reiserfs_journal_cnode *get_cnode(struct super_block *p_s_sb)
{
	struct reiserfs_journal_cnode *cn;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);

	reiserfs_check_lock_depth(p_s_sb, "get_cnode");

	if (journal->j_cnode_free <= 0) {
		return NULL;
	}
	journal->j_cnode_used++;
	journal->j_cnode_free--;
	cn = journal->j_cnode_free_list;
	if (!cn) {
		return cn;
	}
	if (cn->next) {
		cn->next->prev = NULL;
	}
	journal->j_cnode_free_list = cn->next;
	memset(cn, 0, sizeof(struct reiserfs_journal_cnode));
	return cn;
Linus Torvalds's avatar
Linus Torvalds committed
406
407
408
409
410
}

/*
** returns a cnode to the free list 
*/
411
412
413
414
static void free_cnode(struct super_block *p_s_sb,
		       struct reiserfs_journal_cnode *cn)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
Linus Torvalds's avatar
Linus Torvalds committed
415

416
	reiserfs_check_lock_depth(p_s_sb, "free_cnode");
Linus Torvalds's avatar
Linus Torvalds committed
417

418
419
420
421
422
423
424
425
426
	journal->j_cnode_used--;
	journal->j_cnode_free++;
	/* memset(cn, 0, sizeof(struct reiserfs_journal_cnode)) ; */
	cn->next = journal->j_cnode_free_list;
	if (journal->j_cnode_free_list) {
		journal->j_cnode_free_list->prev = cn;
	}
	cn->prev = NULL;	/* not needed with the memset, but I might kill the memset, and forget to do this */
	journal->j_cnode_free_list = cn;
Linus Torvalds's avatar
Linus Torvalds committed
427
428
}

429
430
431
432
static void clear_prepared_bits(struct buffer_head *bh)
{
	clear_buffer_journal_prepared(bh);
	clear_buffer_journal_restore_dirty(bh);
Linus Torvalds's avatar
Linus Torvalds committed
433
434
435
436
437
}

/* utility function to force a BUG if it is called without the big
** kernel lock held.  caller is the string printed just before calling BUG()
*/
438
439
void reiserfs_check_lock_depth(struct super_block *sb, char *caller)
{
Linus Torvalds's avatar
Linus Torvalds committed
440
#ifdef CONFIG_SMP
441
442
443
444
	if (current->lock_depth < 0) {
		reiserfs_panic(sb, "%s called without kernel lock held",
			       caller);
	}
Linus Torvalds's avatar
Linus Torvalds committed
445
#else
446
	;
Linus Torvalds's avatar
Linus Torvalds committed
447
448
449
450
#endif
}

/* return a cnode with same dev, block number and size in table, or null if not found */
451
452
453
454
455
456
457
static inline struct reiserfs_journal_cnode *get_journal_hash_dev(struct
								  super_block
								  *sb,
								  struct
								  reiserfs_journal_cnode
								  **table,
								  long bl)
Linus Torvalds's avatar
Linus Torvalds committed
458
{
459
460
461
462
463
464
465
466
	struct reiserfs_journal_cnode *cn;
	cn = journal_hash(table, sb, bl);
	while (cn) {
		if (cn->blocknr == bl && cn->sb == sb)
			return cn;
		cn = cn->hnext;
	}
	return (struct reiserfs_journal_cnode *)0;
Linus Torvalds's avatar
Linus Torvalds committed
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
}

/*
** this actually means 'can this block be reallocated yet?'.  If you set search_all, a block can only be allocated
** if it is not in the current transaction, was not freed by the current transaction, and has no chance of ever
** being overwritten by a replay after crashing.
**
** If you don't set search_all, a block can only be allocated if it is not in the current transaction.  Since deleting
** a block removes it from the current transaction, this case should never happen.  If you don't set search_all, make
** sure you never write the block without logging it.
**
** next_zero_bit is a suggestion about the next block to try for find_forward.
** when bl is rejected because it is set in a journal list bitmap, we search
** for the next zero bit in the bitmap that rejected bl.  Then, we return that
** through next_zero_bit for find_forward to try.
**
** Just because we return something in next_zero_bit does not mean we won't
** reject it on the next call to reiserfs_in_journal
**
*/
int reiserfs_in_journal(struct super_block *p_s_sb,
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
			int bmap_nr, int bit_nr, int search_all,
			b_blocknr_t * next_zero_bit)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_journal_cnode *cn;
	struct reiserfs_list_bitmap *jb;
	int i;
	unsigned long bl;

	*next_zero_bit = 0;	/* always start this at zero. */

	PROC_INFO_INC(p_s_sb, journal.in_journal);
	/* If we aren't doing a search_all, this is a metablock, and it will be logged before use.
	 ** if we crash before the transaction that freed it commits,  this transaction won't
	 ** have committed either, and the block will never be written
	 */
	if (search_all) {
		for (i = 0; i < JOURNAL_NUM_BITMAPS; i++) {
			PROC_INFO_INC(p_s_sb, journal.in_journal_bitmap);
			jb = journal->j_list_bitmap + i;
			if (jb->journal_list && jb->bitmaps[bmap_nr] &&
			    test_bit(bit_nr,
				     (unsigned long *)jb->bitmaps[bmap_nr]->
				     data)) {
				*next_zero_bit =
				    find_next_zero_bit((unsigned long *)
						       (jb->bitmaps[bmap_nr]->
							data),
						       p_s_sb->s_blocksize << 3,
						       bit_nr + 1);
				return 1;
			}
		}
	}

	bl = bmap_nr * (p_s_sb->s_blocksize << 3) + bit_nr;
	/* is it in any old transactions? */
	if (search_all
	    && (cn =
		get_journal_hash_dev(p_s_sb, journal->j_list_hash_table, bl))) {
		return 1;
	}

	/* is it in the current transaction.  This should never happen */
	if ((cn = get_journal_hash_dev(p_s_sb, journal->j_hash_table, bl))) {
		BUG();
		return 1;
	}

	PROC_INFO_INC(p_s_sb, journal.in_journal_reusable);
	/* safe for reuse */
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
540
541
542
543
}

/* insert cn into table
*/
544
545
546
547
static inline void insert_journal_hash(struct reiserfs_journal_cnode **table,
				       struct reiserfs_journal_cnode *cn)
{
	struct reiserfs_journal_cnode *cn_orig;
Linus Torvalds's avatar
Linus Torvalds committed
548

549
550
551
552
553
554
555
	cn_orig = journal_hash(table, cn->sb, cn->blocknr);
	cn->hnext = cn_orig;
	cn->hprev = NULL;
	if (cn_orig) {
		cn_orig->hprev = cn;
	}
	journal_hash(table, cn->sb, cn->blocknr) = cn;
Linus Torvalds's avatar
Linus Torvalds committed
556
557
558
}

/* lock the current transaction */
559
static inline void lock_journal(struct super_block *p_s_sb)
560
561
562
{
	PROC_INFO_INC(p_s_sb, journal.lock_journal);
	down(&SB_JOURNAL(p_s_sb)->j_lock);
Linus Torvalds's avatar
Linus Torvalds committed
563
564
565
}

/* unlock the current transaction */
566
static inline void unlock_journal(struct super_block *p_s_sb)
567
568
{
	up(&SB_JOURNAL(p_s_sb)->j_lock);
Linus Torvalds's avatar
Linus Torvalds committed
569
570
571
572
}

static inline void get_journal_list(struct reiserfs_journal_list *jl)
{
573
	jl->j_refcount++;
Linus Torvalds's avatar
Linus Torvalds committed
574
575
576
}

static inline void put_journal_list(struct super_block *s,
577
				    struct reiserfs_journal_list *jl)
Linus Torvalds's avatar
Linus Torvalds committed
578
{
579
580
581
582
583
584
	if (jl->j_refcount < 1) {
		reiserfs_panic(s, "trans id %lu, refcount at %d",
			       jl->j_trans_id, jl->j_refcount);
	}
	if (--jl->j_refcount == 0)
		reiserfs_kfree(jl, sizeof(struct reiserfs_journal_list), s);
Linus Torvalds's avatar
Linus Torvalds committed
585
586
587
588
589
590
591
}

/*
** this used to be much more involved, and I'm keeping it just in case things get ugly again.
** it gets called by flush_commit_list, and cleans up any data stored about blocks freed during a
** transaction.
*/
592
593
594
static void cleanup_freed_for_journal_list(struct super_block *p_s_sb,
					   struct reiserfs_journal_list *jl)
{
Linus Torvalds's avatar
Linus Torvalds committed
595

596
597
598
599
600
601
	struct reiserfs_list_bitmap *jb = jl->j_list_bitmap;
	if (jb) {
		cleanup_bitmap_list(p_s_sb, jb);
	}
	jl->j_list_bitmap->journal_list = NULL;
	jl->j_list_bitmap = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
602
603
604
}

static int journal_list_still_alive(struct super_block *s,
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
				    unsigned long trans_id)
{
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	struct list_head *entry = &journal->j_journal_list;
	struct reiserfs_journal_list *jl;

	if (!list_empty(entry)) {
		jl = JOURNAL_LIST_ENTRY(entry->next);
		if (jl->j_trans_id <= trans_id) {
			return 1;
		}
	}
	return 0;
}

static void reiserfs_end_buffer_io_sync(struct buffer_head *bh, int uptodate)
{
	char b[BDEVNAME_SIZE];

	if (buffer_journaled(bh)) {
		reiserfs_warning(NULL,
				 "clm-2084: pinned buffer %lu:%s sent to disk",
				 bh->b_blocknr, bdevname(bh->b_bdev, b));
	}
	if (uptodate)
		set_buffer_uptodate(bh);
	else
		clear_buffer_uptodate(bh);
	unlock_buffer(bh);
	put_bh(bh);
}

static void reiserfs_end_ordered_io(struct buffer_head *bh, int uptodate)
{
	if (uptodate)
		set_buffer_uptodate(bh);
	else
		clear_buffer_uptodate(bh);
	unlock_buffer(bh);
	put_bh(bh);
}

static void submit_logged_buffer(struct buffer_head *bh)
{
	get_bh(bh);
	bh->b_end_io = reiserfs_end_buffer_io_sync;
	clear_buffer_journal_new(bh);
	clear_buffer_dirty(bh);
	if (!test_clear_buffer_journal_test(bh))
		BUG();
	if (!buffer_uptodate(bh))
		BUG();
	submit_bh(WRITE, bh);
}

static void submit_ordered_buffer(struct buffer_head *bh)
{
	get_bh(bh);
	bh->b_end_io = reiserfs_end_ordered_io;
	clear_buffer_dirty(bh);
	if (!buffer_uptodate(bh))
		BUG();
	submit_bh(WRITE, bh);
}

static int submit_barrier_buffer(struct buffer_head *bh)
{
	get_bh(bh);
	bh->b_end_io = reiserfs_end_ordered_io;
	clear_buffer_dirty(bh);
	if (!buffer_uptodate(bh))
		BUG();
	return submit_bh(WRITE_BARRIER, bh);
Linus Torvalds's avatar
Linus Torvalds committed
678
679
680
}

static void check_barrier_completion(struct super_block *s,
681
682
683
684
685
686
687
688
689
				     struct buffer_head *bh)
{
	if (buffer_eopnotsupp(bh)) {
		clear_buffer_eopnotsupp(bh);
		disable_barrier(s);
		set_buffer_uptodate(bh);
		set_buffer_dirty(bh);
		sync_dirty_buffer(bh);
	}
Linus Torvalds's avatar
Linus Torvalds committed
690
691
692
693
}

#define CHUNK_SIZE 32
struct buffer_chunk {
694
695
	struct buffer_head *bh[CHUNK_SIZE];
	int nr;
Linus Torvalds's avatar
Linus Torvalds committed
696
697
};

698
699
700
701
702
703
704
705
706
static void write_chunk(struct buffer_chunk *chunk)
{
	int i;
	get_fs_excl();
	for (i = 0; i < chunk->nr; i++) {
		submit_logged_buffer(chunk->bh[i]);
	}
	chunk->nr = 0;
	put_fs_excl();
Linus Torvalds's avatar
Linus Torvalds committed
707
708
}

709
710
711
712
713
714
715
716
717
static void write_ordered_chunk(struct buffer_chunk *chunk)
{
	int i;
	get_fs_excl();
	for (i = 0; i < chunk->nr; i++) {
		submit_ordered_buffer(chunk->bh[i]);
	}
	chunk->nr = 0;
	put_fs_excl();
Linus Torvalds's avatar
Linus Torvalds committed
718
719
720
}

static int add_to_chunk(struct buffer_chunk *chunk, struct buffer_head *bh,
721
			spinlock_t * lock, void (fn) (struct buffer_chunk *))
Linus Torvalds's avatar
Linus Torvalds committed
722
{
723
724
725
726
727
728
729
730
731
732
733
734
735
	int ret = 0;
	if (chunk->nr >= CHUNK_SIZE)
		BUG();
	chunk->bh[chunk->nr++] = bh;
	if (chunk->nr >= CHUNK_SIZE) {
		ret = 1;
		if (lock)
			spin_unlock(lock);
		fn(chunk);
		if (lock)
			spin_lock(lock);
	}
	return ret;
Linus Torvalds's avatar
Linus Torvalds committed
736
737
738
}

static atomic_t nr_reiserfs_jh = ATOMIC_INIT(0);
739
740
741
742
743
744
745
746
747
748
static struct reiserfs_jh *alloc_jh(void)
{
	struct reiserfs_jh *jh;
	while (1) {
		jh = kmalloc(sizeof(*jh), GFP_NOFS);
		if (jh) {
			atomic_inc(&nr_reiserfs_jh);
			return jh;
		}
		yield();
Linus Torvalds's avatar
Linus Torvalds committed
749
750
751
752
753
754
755
	}
}

/*
 * we want to free the jh when the buffer has been written
 * and waited on
 */
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
void reiserfs_free_jh(struct buffer_head *bh)
{
	struct reiserfs_jh *jh;

	jh = bh->b_private;
	if (jh) {
		bh->b_private = NULL;
		jh->bh = NULL;
		list_del_init(&jh->list);
		kfree(jh);
		if (atomic_read(&nr_reiserfs_jh) <= 0)
			BUG();
		atomic_dec(&nr_reiserfs_jh);
		put_bh(bh);
	}
Linus Torvalds's avatar
Linus Torvalds committed
771
772
773
}

static inline int __add_jh(struct reiserfs_journal *j, struct buffer_head *bh,
774
			   int tail)
Linus Torvalds's avatar
Linus Torvalds committed
775
{
776
	struct reiserfs_jh *jh;
Linus Torvalds's avatar
Linus Torvalds committed
777

778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
	if (bh->b_private) {
		spin_lock(&j->j_dirty_buffers_lock);
		if (!bh->b_private) {
			spin_unlock(&j->j_dirty_buffers_lock);
			goto no_jh;
		}
		jh = bh->b_private;
		list_del_init(&jh->list);
	} else {
	      no_jh:
		get_bh(bh);
		jh = alloc_jh();
		spin_lock(&j->j_dirty_buffers_lock);
		/* buffer must be locked for __add_jh, should be able to have
		 * two adds at the same time
		 */
		if (bh->b_private)
			BUG();
		jh->bh = bh;
		bh->b_private = jh;
Linus Torvalds's avatar
Linus Torvalds committed
798
	}
799
800
801
802
803
804
805
806
	jh->jl = j->j_current_jl;
	if (tail)
		list_add_tail(&jh->list, &jh->jl->j_tail_bh_list);
	else {
		list_add_tail(&jh->list, &jh->jl->j_bh_list);
	}
	spin_unlock(&j->j_dirty_buffers_lock);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
807
808
}

809
810
811
int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh)
{
	return __add_jh(SB_JOURNAL(inode->i_sb), bh, 1);
Linus Torvalds's avatar
Linus Torvalds committed
812
}
813
814
815
int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh)
{
	return __add_jh(SB_JOURNAL(inode->i_sb), bh, 0);
Linus Torvalds's avatar
Linus Torvalds committed
816
817
818
}

#define JH_ENTRY(l) list_entry((l), struct reiserfs_jh, list)
819
static int write_ordered_buffers(spinlock_t * lock,
Linus Torvalds's avatar
Linus Torvalds committed
820
				 struct reiserfs_journal *j,
821
				 struct reiserfs_journal_list *jl,
Linus Torvalds's avatar
Linus Torvalds committed
822
823
				 struct list_head *list)
{
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
	struct buffer_head *bh;
	struct reiserfs_jh *jh;
	int ret = j->j_errno;
	struct buffer_chunk chunk;
	struct list_head tmp;
	INIT_LIST_HEAD(&tmp);

	chunk.nr = 0;
	spin_lock(lock);
	while (!list_empty(list)) {
		jh = JH_ENTRY(list->next);
		bh = jh->bh;
		get_bh(bh);
		if (test_set_buffer_locked(bh)) {
			if (!buffer_dirty(bh)) {
				list_del_init(&jh->list);
				list_add(&jh->list, &tmp);
				goto loop_next;
			}
			spin_unlock(lock);
			if (chunk.nr)
				write_ordered_chunk(&chunk);
			wait_on_buffer(bh);
			cond_resched();
			spin_lock(lock);
			goto loop_next;
		}
		if (buffer_dirty(bh)) {
			list_del_init(&jh->list);
			list_add(&jh->list, &tmp);
			add_to_chunk(&chunk, bh, lock, write_ordered_chunk);
		} else {
			reiserfs_free_jh(bh);
			unlock_buffer(bh);
		}
	      loop_next:
		put_bh(bh);
		cond_resched_lock(lock);
	}
	if (chunk.nr) {
		spin_unlock(lock);
Linus Torvalds's avatar
Linus Torvalds committed
865
		write_ordered_chunk(&chunk);
866
		spin_lock(lock);
Linus Torvalds's avatar
Linus Torvalds committed
867
	}
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
	while (!list_empty(&tmp)) {
		jh = JH_ENTRY(tmp.prev);
		bh = jh->bh;
		get_bh(bh);
		reiserfs_free_jh(bh);

		if (buffer_locked(bh)) {
			spin_unlock(lock);
			wait_on_buffer(bh);
			spin_lock(lock);
		}
		if (!buffer_uptodate(bh)) {
			ret = -EIO;
		}
		put_bh(bh);
		cond_resched_lock(lock);
Linus Torvalds's avatar
Linus Torvalds committed
884
	}
885
886
887
	spin_unlock(lock);
	return ret;
}
Linus Torvalds's avatar
Linus Torvalds committed
888

889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
static int flush_older_commits(struct super_block *s,
			       struct reiserfs_journal_list *jl)
{
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	struct reiserfs_journal_list *other_jl;
	struct reiserfs_journal_list *first_jl;
	struct list_head *entry;
	unsigned long trans_id = jl->j_trans_id;
	unsigned long other_trans_id;
	unsigned long first_trans_id;

      find_first:
	/*
	 * first we walk backwards to find the oldest uncommitted transation
	 */
	first_jl = jl;
	entry = jl->j_list.prev;
	while (1) {
		other_jl = JOURNAL_LIST_ENTRY(entry);
		if (entry == &journal->j_journal_list ||
		    atomic_read(&other_jl->j_older_commits_done))
			break;
Linus Torvalds's avatar
Linus Torvalds committed
911

912
913
914
		first_jl = other_jl;
		entry = other_jl->j_list.prev;
	}
Linus Torvalds's avatar
Linus Torvalds committed
915

916
917
918
919
	/* if we didn't find any older uncommitted transactions, return now */
	if (first_jl == jl) {
		return 0;
	}
Linus Torvalds's avatar
Linus Torvalds committed
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
	first_trans_id = first_jl->j_trans_id;

	entry = &first_jl->j_list;
	while (1) {
		other_jl = JOURNAL_LIST_ENTRY(entry);
		other_trans_id = other_jl->j_trans_id;

		if (other_trans_id < trans_id) {
			if (atomic_read(&other_jl->j_commit_left) != 0) {
				flush_commit_list(s, other_jl, 0);

				/* list we were called with is gone, return */
				if (!journal_list_still_alive(s, trans_id))
					return 1;

				/* the one we just flushed is gone, this means all
				 * older lists are also gone, so first_jl is no longer
				 * valid either.  Go back to the beginning.
				 */
				if (!journal_list_still_alive
				    (s, other_trans_id)) {
					goto find_first;
				}
			}
			entry = entry->next;
			if (entry == &journal->j_journal_list)
				return 0;
		} else {
			return 0;
Linus Torvalds's avatar
Linus Torvalds committed
950
951
		}
	}
952
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
953
}
954
955
956
957
958
959
960
int reiserfs_async_progress_wait(struct super_block *s)
{
	DEFINE_WAIT(wait);
	struct reiserfs_journal *j = SB_JOURNAL(s);
	if (atomic_read(&j->j_async_throttle))
		blk_congestion_wait(WRITE, HZ / 10);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
961
962
963
964
965
966
967
968
969
}

/*
** if this journal list still has commit blocks unflushed, send them to disk.
**
** log areas must be flushed in order (transaction 2 can't commit before transaction 1)
** Before the commit block can by written, every other log block must be safely on disk
**
*/
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
static int flush_commit_list(struct super_block *s,
			     struct reiserfs_journal_list *jl, int flushall)
{
	int i;
	int bn;
	struct buffer_head *tbh = NULL;
	unsigned long trans_id = jl->j_trans_id;
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	int barrier = 0;
	int retval = 0;

	reiserfs_check_lock_depth(s, "flush_commit_list");

	if (atomic_read(&jl->j_older_commits_done)) {
		return 0;
	}

	get_fs_excl();

	/* before we can put our commit blocks on disk, we have to make sure everyone older than
	 ** us is on disk too
	 */
	BUG_ON(jl->j_len <= 0);
	BUG_ON(trans_id == journal->j_trans_id);

	get_journal_list(jl);
	if (flushall) {
		if (flush_older_commits(s, jl) == 1) {
			/* list disappeared during flush_older_commits.  return */
			goto put_jl;
		}
	}

	/* make sure nobody is trying to flush this one at the same time */
	down(&jl->j_commit_lock);
	if (!journal_list_still_alive(s, trans_id)) {
		up(&jl->j_commit_lock);
		goto put_jl;
	}
	BUG_ON(jl->j_trans_id == 0);

	/* this commit is done, exit */
	if (atomic_read(&(jl->j_commit_left)) <= 0) {
		if (flushall) {
			atomic_set(&(jl->j_older_commits_done), 1);
		}
		up(&jl->j_commit_lock);
		goto put_jl;
	}

	if (!list_empty(&jl->j_bh_list)) {
		unlock_kernel();
		write_ordered_buffers(&journal->j_dirty_buffers_lock,
				      journal, jl, &jl->j_bh_list);
		lock_kernel();
	}
	BUG_ON(!list_empty(&jl->j_bh_list));
	/*
	 * for the description block and all the log blocks, submit any buffers
	 * that haven't already reached the disk
	 */
	atomic_inc(&journal->j_async_throttle);
	for (i = 0; i < (jl->j_len + 1); i++) {
		bn = SB_ONDISK_JOURNAL_1st_BLOCK(s) + (jl->j_start + i) %
		    SB_ONDISK_JOURNAL_SIZE(s);
		tbh = journal_find_get_block(s, bn);
		if (buffer_dirty(tbh))	/* redundant, ll_rw_block() checks */
1037
			ll_rw_block(SWRITE, 1, &tbh);
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
		put_bh(tbh);
	}
	atomic_dec(&journal->j_async_throttle);

	/* wait on everything written so far before writing the commit
	 * if we are in barrier mode, send the commit down now
	 */
	barrier = reiserfs_barrier_flush(s);
	if (barrier) {
		int ret;
		lock_buffer(jl->j_commit_bh);
		ret = submit_barrier_buffer(jl->j_commit_bh);
		if (ret == -EOPNOTSUPP) {
			set_buffer_uptodate(jl->j_commit_bh);
			disable_barrier(s);
			barrier = 0;
		}
	}
	for (i = 0; i < (jl->j_len + 1); i++) {
		bn = SB_ONDISK_JOURNAL_1st_BLOCK(s) +
		    (jl->j_start + i) % SB_ONDISK_JOURNAL_SIZE(s);
		tbh = journal_find_get_block(s, bn);
		wait_on_buffer(tbh);
		// since we're using ll_rw_blk above, it might have skipped over
		// a locked buffer.  Double check here
		//
		if (buffer_dirty(tbh))	/* redundant, sync_dirty_buffer() checks */
			sync_dirty_buffer(tbh);
		if (unlikely(!buffer_uptodate(tbh))) {
Linus Torvalds's avatar
Linus Torvalds committed
1067
#ifdef CONFIG_REISERFS_CHECK
1068
			reiserfs_warning(s, "journal-601, buffer write failed");
Linus Torvalds's avatar
Linus Torvalds committed
1069
#endif
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
			retval = -EIO;
		}
		put_bh(tbh);	/* once for journal_find_get_block */
		put_bh(tbh);	/* once due to original getblk in do_journal_end */
		atomic_dec(&(jl->j_commit_left));
	}

	BUG_ON(atomic_read(&(jl->j_commit_left)) != 1);

	if (!barrier) {
		if (buffer_dirty(jl->j_commit_bh))
			BUG();
		mark_buffer_dirty(jl->j_commit_bh);
		sync_dirty_buffer(jl->j_commit_bh);
	} else
		wait_on_buffer(jl->j_commit_bh);

	check_barrier_completion(s, jl->j_commit_bh);

	/* If there was a write error in the journal - we can't commit this
	 * transaction - it will be invalid and, if successful, will just end
	 * up propogating the write error out to the filesystem. */
	if (unlikely(!buffer_uptodate(jl->j_commit_bh))) {
Linus Torvalds's avatar
Linus Torvalds committed
1093
#ifdef CONFIG_REISERFS_CHECK
1094
		reiserfs_warning(s, "journal-615: buffer write failed");
Linus Torvalds's avatar
Linus Torvalds committed
1095
#endif
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
		retval = -EIO;
	}
	bforget(jl->j_commit_bh);
	if (journal->j_last_commit_id != 0 &&
	    (jl->j_trans_id - journal->j_last_commit_id) != 1) {
		reiserfs_warning(s, "clm-2200: last commit %lu, current %lu",
				 journal->j_last_commit_id, jl->j_trans_id);
	}
	journal->j_last_commit_id = jl->j_trans_id;

	/* now, every commit block is on the disk.  It is safe to allow blocks freed during this transaction to be reallocated */
	cleanup_freed_for_journal_list(s, jl);

	retval = retval ? retval : journal->j_errno;

	/* mark the metadata dirty */
	if (!retval)
		dirty_one_transaction(s, jl);
	atomic_dec(&(jl->j_commit_left));

	if (flushall) {
		atomic_set(&(jl->j_older_commits_done), 1);
	}
	up(&jl->j_commit_lock);
      put_jl:
	put_journal_list(s, jl);

	if (retval)
		reiserfs_abort(s, retval, "Journal write error in %s",
			       __FUNCTION__);
	put_fs_excl();
	return retval;
Linus Torvalds's avatar
Linus Torvalds committed
1128
1129
1130
1131
1132
1133
}

/*
** flush_journal_list frequently needs to find a newer transaction for a given block.  This does that, or 
** returns NULL if it can't find anything 
*/
1134
1135
1136
1137
1138
1139
static struct reiserfs_journal_list *find_newer_jl_for_cn(struct
							  reiserfs_journal_cnode
							  *cn)
{
	struct super_block *sb = cn->sb;
	b_blocknr_t blocknr = cn->blocknr;
Linus Torvalds's avatar
Linus Torvalds committed
1140

1141
1142
1143
1144
1145
1146
1147
1148
	cn = cn->hprev;
	while (cn) {
		if (cn->sb == sb && cn->blocknr == blocknr && cn->jlist) {
			return cn->jlist;
		}
		cn = cn->hprev;
	}
	return NULL;
Linus Torvalds's avatar
Linus Torvalds committed
1149
1150
}

1151
1152
1153
1154
static void remove_journal_hash(struct super_block *,
				struct reiserfs_journal_cnode **,
				struct reiserfs_journal_list *, unsigned long,
				int);
Linus Torvalds's avatar
Linus Torvalds committed
1155
1156
1157
1158
1159
1160

/*
** once all the real blocks have been flushed, it is safe to remove them from the
** journal list for this transaction.  Aside from freeing the cnode, this also allows the
** block to be reallocated for data blocks if it had been deleted.
*/
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
static void remove_all_from_journal_list(struct super_block *p_s_sb,
					 struct reiserfs_journal_list *jl,
					 int debug)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_journal_cnode *cn, *last;
	cn = jl->j_realblock;

	/* which is better, to lock once around the whole loop, or
	 ** to lock for each call to remove_journal_hash?
	 */
	while (cn) {
		if (cn->blocknr != 0) {
			if (debug) {
				reiserfs_warning(p_s_sb,
						 "block %u, bh is %d, state %ld",
						 cn->blocknr, cn->bh ? 1 : 0,
						 cn->state);
			}
			cn->state = 0;
			remove_journal_hash(p_s_sb, journal->j_list_hash_table,
					    jl, cn->blocknr, 1);
		}
		last = cn;
		cn = cn->next;
		free_cnode(p_s_sb, last);
	}
	jl->j_realblock = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
1189
1190
1191
1192
1193
1194
1195
1196
1197
}

/*
** if this timestamp is greater than the timestamp we wrote last to the header block, write it to the header block.
** once this is done, I can safely say the log area for this transaction won't ever be replayed, and I can start
** releasing blocks in this transaction for reuse as data blocks.
** called by flush_journal_list, before it calls remove_all_from_journal_list
**
*/
1198
1199
1200
1201
1202
1203
static int _update_journal_header_block(struct super_block *p_s_sb,
					unsigned long offset,
					unsigned long trans_id)
{
	struct reiserfs_journal_header *jh;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
Linus Torvalds's avatar
Linus Torvalds committed
1204

1205
1206
	if (reiserfs_is_journal_aborted(journal))
		return -EIO;
Linus Torvalds's avatar
Linus Torvalds committed
1207

1208
1209
1210
1211
	if (trans_id >= journal->j_last_flush_trans_id) {
		if (buffer_locked((journal->j_header_bh))) {
			wait_on_buffer((journal->j_header_bh));
			if (unlikely(!buffer_uptodate(journal->j_header_bh))) {
Linus Torvalds's avatar
Linus Torvalds committed
1212
#ifdef CONFIG_REISERFS_CHECK
1213
1214
				reiserfs_warning(p_s_sb,
						 "journal-699: buffer write failed");
Linus Torvalds's avatar
Linus Torvalds committed
1215
#endif
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
1251
1252
1253
1254
1255
1256
				return -EIO;
			}
		}
		journal->j_last_flush_trans_id = trans_id;
		journal->j_first_unflushed_offset = offset;
		jh = (struct reiserfs_journal_header *)(journal->j_header_bh->
							b_data);
		jh->j_last_flush_trans_id = cpu_to_le32(trans_id);
		jh->j_first_unflushed_offset = cpu_to_le32(offset);
		jh->j_mount_id = cpu_to_le32(journal->j_mount_id);

		if (reiserfs_barrier_flush(p_s_sb)) {
			int ret;
			lock_buffer(journal->j_header_bh);
			ret = submit_barrier_buffer(journal->j_header_bh);
			if (ret == -EOPNOTSUPP) {
				set_buffer_uptodate(journal->j_header_bh);
				disable_barrier(p_s_sb);
				goto sync;
			}
			wait_on_buffer(journal->j_header_bh);
			check_barrier_completion(p_s_sb, journal->j_header_bh);
		} else {
		      sync:
			set_buffer_dirty(journal->j_header_bh);
			sync_dirty_buffer(journal->j_header_bh);
		}
		if (!buffer_uptodate(journal->j_header_bh)) {
			reiserfs_warning(p_s_sb,
					 "journal-837: IO error during journal replay");
			return -EIO;
		}
	}
	return 0;
}

static int update_journal_header_block(struct super_block *p_s_sb,
				       unsigned long offset,
				       unsigned long trans_id)
{
	return _update_journal_header_block(p_s_sb, offset, trans_id);
Linus Torvalds's avatar
Linus Torvalds committed
1257
}
1258

Linus Torvalds's avatar
Linus Torvalds committed
1259
1260
1261
1262
1263
/* 
** flush any and all journal lists older than you are 
** can only be called from flush_journal_list
*/
static int flush_older_journal_lists(struct super_block *p_s_sb,
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
				     struct reiserfs_journal_list *jl)
{
	struct list_head *entry;
	struct reiserfs_journal_list *other_jl;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	unsigned long trans_id = jl->j_trans_id;

	/* we know we are the only ones flushing things, no extra race
	 * protection is required.
	 */
      restart:
	entry = journal->j_journal_list.next;
	/* Did we wrap? */
	if (entry == &journal->j_journal_list)
		return 0;
	other_jl = JOURNAL_LIST_ENTRY(entry);
	if (other_jl->j_trans_id < trans_id) {
		BUG_ON(other_jl->j_refcount <= 0);
		/* do not flush all */
		flush_journal_list(p_s_sb, other_jl, 0);

		/* other_jl is now deleted from the list */
		goto restart;
	}
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
1289
1290
1291
}

static void del_from_work_list(struct super_block *s,
1292
1293
1294
1295
1296
1297
1298
			       struct reiserfs_journal_list *jl)
{
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	if (!list_empty(&jl->j_working_list)) {
		list_del_init(&jl->j_working_list);
		journal->j_num_work_lists--;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
}

/* flush a journal list, both commit and real blocks
**
** always set flushall to 1, unless you are calling from inside
** flush_journal_list
**
** IMPORTANT.  This can only be called while there are no journal writers, 
** and the journal is locked.  That means it can only be called from 
** do_journal_end, or by journal_release
*/
1310
1311
static int flush_journal_list(struct super_block *s,
			      struct reiserfs_journal_list *jl, int flushall)
Linus Torvalds's avatar
Linus Torvalds committed
1312
{
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
	struct reiserfs_journal_list *pjl;
	struct reiserfs_journal_cnode *cn, *last;
	int count;
	int was_jwait = 0;
	int was_dirty = 0;
	struct buffer_head *saved_bh;
	unsigned long j_len_saved = jl->j_len;
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	int err = 0;

	BUG_ON(j_len_saved <= 0);

	if (atomic_read(&journal->j_wcount) != 0) {
		reiserfs_warning(s,
				 "clm-2048: flush_journal_list called with wcount %d",
				 atomic_read(&journal->j_wcount));
	}
	BUG_ON(jl->j_trans_id == 0);
Linus Torvalds's avatar
Linus Torvalds committed
1331

1332
1333
1334
1335
1336
1337
	/* if flushall == 0, the lock is already held */
	if (flushall) {
		down(&journal->j_flush_sem);
	} else if (!down_trylock(&journal->j_flush_sem)) {
		BUG();
	}
Linus Torvalds's avatar
Linus Torvalds committed
1338

1339
1340
1341
1342
1343
1344
1345
	count = 0;
	if (j_len_saved > journal->j_trans_max) {
		reiserfs_panic(s,
			       "journal-715: flush_journal_list, length is %lu, trans id %lu\n",
			       j_len_saved, jl->j_trans_id);
		return 0;
	}
Linus Torvalds's avatar
Linus Torvalds committed
1346

1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
	get_fs_excl();

	/* if all the work is already done, get out of here */
	if (atomic_read(&(jl->j_nonzerolen)) <= 0 &&
	    atomic_read(&(jl->j_commit_left)) <= 0) {
		goto flush_older_and_return;
	}

	/* start by putting the commit list on disk.  This will also flush 
	 ** the commit lists of any olders transactions
	 */
	flush_commit_list(s, jl, 1);

	if (!(jl->j_state & LIST_DIRTY)
	    && !reiserfs_is_journal_aborted(journal))
		BUG();

	/* are we done now? */
	if (atomic_read(&(jl->j_nonzerolen)) <= 0 &&
	    atomic_read(&(jl->j_commit_left)) <= 0) {
		goto flush_older_and_return;
	}

	/* loop through each cnode, see if we need to write it, 
	 ** or wait on a more recent transaction, or just ignore it 
	 */
	if (atomic_read(&(journal->j_wcount)) != 0) {
		reiserfs_panic(s,
			       "journal-844: panic journal list is flushing, wcount is not 0\n");
	}
	cn = jl->j_realblock;
	while (cn) {
		was_jwait = 0;
		was_dirty = 0;
		saved_bh = NULL;
		/* blocknr of 0 is no longer in the hash, ignore it */
		if (cn->blocknr == 0) {
			goto free_cnode;
		}

		/* This transaction failed commit. Don't write out to the disk */
		if (!(jl->j_state & LIST_DIRTY))
			goto free_cnode;

		pjl = find_newer_jl_for_cn(cn);
		/* the order is important here.  We check pjl to make sure we
		 ** don't clear BH_JDirty_wait if we aren't the one writing this
		 ** block to disk
		 */
		if (!pjl && cn->bh) {
			saved_bh = cn->bh;

			/* we do this to make sure nobody releases the buffer while 
			 ** we are working with it 
			 */
			get_bh(saved_bh);

			if (buffer_journal_dirty(saved_bh)) {
				BUG_ON(!can_dirty(cn));
				was_jwait = 1;
				was_dirty = 1;
			} else if (can_dirty(cn)) {
				/* everything with !pjl && jwait should be writable */
				BUG();
			}
		}

		/* if someone has this block in a newer transaction, just make
		 ** sure they are commited, and don't try writing it to disk
		 */
		if (pjl) {
			if (atomic_read(&pjl->j_commit_left))
				flush_commit_list(s, pjl, 1);
			goto free_cnode;
		}

		/* bh == NULL when the block got to disk on its own, OR, 
		 ** the block got freed in a future transaction 
		 */
		if (saved_bh == NULL) {
			goto free_cnode;
		}

		/* this should never happen.  kupdate_one_transaction has this list
		 ** locked while it works, so we should never see a buffer here that
		 ** is not marked JDirty_wait
		 */
		if ((!was_jwait) && !buffer_locked(saved_bh)) {
			reiserfs_warning(s,
					 "journal-813: BAD! buffer %llu %cdirty %cjwait, "
					 "not in a newer tranasction",
					 (unsigned long long)saved_bh->
					 b_blocknr, was_dirty ? ' ' : '!',
					 was_jwait ? ' ' : '!');
		}
		if (was_dirty) {
			/* we inc again because saved_bh gets decremented at free_cnode */
			get_bh(saved_bh);
			set_bit(BLOCK_NEEDS_FLUSH, &cn->state);
			lock_buffer(saved_bh);
			BUG_ON(cn->blocknr != saved_bh->b_blocknr);
			if (buffer_dirty(saved_bh))
				submit_logged_buffer(saved_bh);
			else
				unlock_buffer(saved_bh);
			count++;
		} else {
			reiserfs_warning(s,
					 "clm-2082: Unable to flush buffer %llu in %s",
					 (unsigned long long)saved_bh->
					 b_blocknr, __FUNCTION__);
		}
	      free_cnode:
		last = cn;
		cn = cn->next;
		if (saved_bh) {
			/* we incremented this to keep others from taking the buffer head away */
			put_bh(saved_bh);
			if (atomic_read(&(saved_bh->b_count)) < 0) {
				reiserfs_warning(s,
						 "journal-945: saved_bh->b_count < 0");
			}
		}
	}
	if (count > 0) {
		cn = jl->j_realblock;
		while (cn) {
			if (test_bit(BLOCK_NEEDS_FLUSH, &cn->state)) {
				if (!cn->bh) {
					reiserfs_panic(s,
						       "journal-1011: cn->bh is NULL\n");
				}
				wait_on_buffer(cn->bh);
				if (!cn->bh) {
					reiserfs_panic(s,
						       "journal-1012: cn->bh is NULL\n");
				}
				if (unlikely(!buffer_uptodate(cn->bh))) {
#ifdef CONFIG_REISERFS_CHECK
					reiserfs_warning(s,
							 "journal-949: buffer write failed\n");
#endif
					err = -EIO;
				}
				/* note, we must clear the JDirty_wait bit after the up to date
				 ** check, otherwise we race against our flushpage routine
				 */
				BUG_ON(!test_clear_buffer_journal_dirty
				       (cn->bh));

				/* undo the inc from journal_mark_dirty */
				put_bh(cn->bh);
				brelse(cn->bh);
			}
			cn = cn->next;
		}
	}

	if (err)
		reiserfs_abort(s, -EIO,
			       "Write error while pushing transaction to disk in %s",
			       __FUNCTION__);
      flush_older_and_return:

	/* before we can update the journal header block, we _must_ flush all 
	 ** real blocks from all older transactions to disk.  This is because
	 ** once the header block is updated, this transaction will not be
	 ** replayed after a crash
	 */
	if (flushall) {
		flush_older_journal_lists(s, jl);
	}

	err = journal->j_errno;
	/* before we can remove everything from the hash tables for this 
	 ** transaction, we must make sure it can never be replayed
	 **
	 ** since we are only called from do_journal_end, we know for sure there
	 ** are no allocations going on while we are flushing journal lists.  So,
	 ** we only need to update the journal header block for the last list
	 ** being flushed
	 */
	if (!err && flushall) {
		err =
		    update_journal_header_block(s,
						(jl->j_start + jl->j_len +
						 2) % SB_ONDISK_JOURNAL_SIZE(s),
						jl->j_trans_id);
		if (err)
			reiserfs_abort(s, -EIO,
				       "Write error while updating journal header in %s",
				       __FUNCTION__);
	}
	remove_all_from_journal_list(s, jl, 0);
	list_del_init(&jl->j_list);
	journal->j_num_lists--;
	del_from_work_list(s, jl);

	if (journal->j_last_flush_id != 0 &&
	    (jl->j_trans_id - journal->j_last_flush_id) != 1) {
		reiserfs_warning(s, "clm-2201: last flush %lu, current %lu",
				 journal->j_last_flush_id, jl->j_trans_id);
	}
	journal->j_last_flush_id = jl->j_trans_id;

	/* not strictly required since we are freeing the list, but it should
	 * help find code using dead lists later on
	 */
	jl->j_len = 0;
	atomic_set(&(jl->j_nonzerolen), 0);
	jl->j_start = 0;
	jl->j_realblock = NULL;
	jl->j_commit_bh = NULL;
	jl->j_trans_id = 0;
	jl->j_state = 0;
	put_journal_list(s, jl);
	if (flushall)
		up(&journal->j_flush_sem);
	put_fs_excl();
	return err;
}

static int write_one_transaction(struct super_block *s,
				 struct reiserfs_journal_list *jl,
				 struct buffer_chunk *chunk)
{
	struct reiserfs_journal_cnode *cn;
	int ret = 0;

	jl->j_state |= LIST_TOUCHED;
	del_from_work_list(s, jl);
	if (jl->j_len == 0 || atomic_read(&jl->j_nonzerolen) == 0) {
		return 0;
	}

	cn = jl->j_realblock;
	while (cn) {
		/* if the blocknr == 0, this has been cleared from the hash,
		 ** skip it
		 */
		if (cn->blocknr == 0) {
			goto next;
		}
		if (cn->bh && can_dirty(cn) && buffer_dirty(cn->bh)) {
			struct buffer_head *tmp_bh;
			/* we can race against journal_mark_freed when we try
			 * to lock_buffer(cn->bh), so we have to inc the buffer
			 * count, and recheck things after locking
			 */
			tmp_bh = cn->bh;
			get_bh(tmp_bh);
			lock_buffer(tmp_bh);
			if (cn->bh && can_dirty(cn) && buffer_dirty(tmp_bh)) {
				if (!buffer_journal_dirty(tmp_bh) ||
				    buffer_journal_prepared(tmp_bh))
					BUG();
				add_to_chunk(chunk, tmp_bh, NULL, write_chunk);
				ret++;
			} else {
				/* note, cn->bh might be null now */
				unlock_buffer(tmp_bh);
			}
			put_bh(tmp_bh);
		}
	      next:
		cn = cn->next;
		cond_resched();
	}
	return ret;
}

/* used by flush_commit_list */
static int dirty_one_transaction(struct super_block *s,
				 struct reiserfs_journal_list *jl)
{
	struct reiserfs_journal_cnode *cn;
	struct reiserfs_journal_list *pjl;
	int ret = 0;

	jl->j_state |= LIST_DIRTY;
	cn = jl->j_realblock;
	while (cn) {
		/* look for a more recent transaction that logged this
		 ** buffer.  Only the most recent transaction with a buffer in
		 ** it is allowed to send that buffer to disk
		 */
		pjl = find_newer_jl_for_cn(cn);
		if (!pjl && cn->blocknr && cn->bh
		    && buffer_journal_dirty(cn->bh)) {
			BUG_ON(!can_dirty(cn));
			/* if the buffer is prepared, it will either be logged
			 * or restored.  If restored, we need to make sure
			 * it actually gets marked dirty
			 */
			clear_buffer_journal_new(cn->bh);
			if (buffer_journal_prepared(cn->bh)) {
				set_buffer_journal_restore_dirty(cn->bh);
			} else {
				set_buffer_journal_test(cn->bh);
				mark_buffer_dirty(cn->bh);
			}
		}
		cn = cn->next;
	}
	return ret;
}

static int kupdate_transactions(struct super_block *s,
				struct reiserfs_journal_list *jl,
				struct reiserfs_journal_list **next_jl,
				unsigned long *next_trans_id,
				int num_blocks, int num_trans)
{
	int ret = 0;
	int written = 0;
	int transactions_flushed = 0;
	unsigned long orig_trans_id = jl->j_trans_id;
	struct buffer_chunk chunk;
	struct list_head *entry;
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	chunk.nr = 0;

	down(&journal->j_flush_sem);
	if (!journal_list_still_alive(s, orig_trans_id)) {
		goto done;
	}

	/* we've got j_flush_sem held, nobody is going to delete any
	 * of these lists out from underneath us
	 */
	while ((num_trans && transactions_flushed < num_trans) ||
	       (!num_trans && written < num_blocks)) {

		if (jl->j_len == 0 || (jl->j_state & LIST_TOUCHED) ||
		    atomic_read(&jl->j_commit_left)
		    || !(jl->j_state & LIST_DIRTY)) {
			del_from_work_list(s, jl);
			break;
		}
		ret = write_one_transaction(s, jl, &chunk);

		if (ret < 0)
			goto done;
		transactions_flushed++;
		written += ret;
		entry = jl->j_list.next;

		/* did we wrap? */
		if (entry == &journal->j_journal_list) {
			break;
		}
		jl = JOURNAL_LIST_ENTRY(entry);

		/* don't bother with older transactions */
		if (jl->j_trans_id <= orig_trans_id)
			break;
	}
	if (chunk.nr) {
		write_chunk(&chunk);
	}

      done:
	up(&journal->j_flush_sem);
	return ret;
}

/* for o_sync and fsync heavy applications, they tend to use
** all the journa list slots with tiny transactions.  These
** trigger lots and lots of calls to update the header block, which
** adds seeks and slows things down.
**
** This function tries to clear out a large chunk of the journal lists
** at once, which makes everything faster since only the newest journal
Linus Torvalds's avatar
Linus Torvalds committed
1720
1721
1722
** list updates the header block
*/
static int flush_used_journal_lists(struct super_block *s,
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
				    struct reiserfs_journal_list *jl)
{
	unsigned long len = 0;
	unsigned long cur_len;
	int ret;
	int i;
	int limit = 256;
	struct reiserfs_journal_list *tjl;
	struct reiserfs_journal_list *flush_jl;
	unsigned long trans_id;
	struct reiserfs_journal *journal = SB_JOURNAL(s);

	flush_jl = tjl = jl;

	/* in data logging mode, try harder to flush a lot of blocks */
	if (reiserfs_data_log(s))
		limit = 1024;
	/* flush for 256 transactions or limit blocks, whichever comes first */
	for (i = 0; i < 256 && len < limit; i++) {
		if (atomic_read(&tjl->j_commit_left) ||
		    tjl->j_trans_id < jl->j_trans_id) {
			break;
		}
		cur_len = atomic_read(&tjl->j_nonzerolen);
		if (cur_len > 0) {
			tjl->j_state &= ~LIST_TOUCHED;
		}
		len += cur_len;
		flush_jl = tjl;
		if (tjl->j_list.next == &journal->j_journal_list)
			break;
		tjl = JOURNAL_LIST_ENTRY(tjl->j_list.next);
	}
	/* try to find a group of blocks we can flush across all the
	 ** transactions, but only bother if we've actually spanned
	 ** across multiple lists
	 */
	if (flush_jl != jl) {
		ret = kupdate_transactions(s, jl, &tjl, &trans_id, len, i);
	}
	flush_journal_list(s, flush_jl, 1);
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
1765
1766
1767
1768
1769
1770
1771
}

/*
** removes any nodes in table with name block and dev as bh.
** only touchs the hnext and hprev pointers.
*/
void remove_journal_hash(struct super_block *sb,
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
			 struct reiserfs_journal_cnode **table,
			 struct reiserfs_journal_list *jl,
			 unsigned long block, int remove_freed)
{
	struct reiserfs_journal_cnode *cur;
	struct reiserfs_journal_cnode **head;

	head = &(journal_hash(table, sb, block));
	if (!head) {
		return;
	}
	cur = *head;
	while (cur) {
		if (cur->blocknr == block && cur->sb == sb
		    && (jl == NULL || jl == cur->jlist)
		    && (!test_bit(BLOCK_FREED, &cur->state) || remove_freed)) {
			if (cur->hnext) {
				cur->hnext->hprev = cur->hprev;
			}
			if (cur->hprev) {
				cur->hprev->hnext = cur->hnext;
			} else {
				*head = cur->hnext;
			}
			cur->blocknr = 0;
			cur->sb = NULL;
			cur->state = 0;
			if (cur->bh && cur->jlist)	/* anybody who clears the cur->bh will also dec the nonzerolen */
				atomic_dec(&(cur->jlist->j_nonzerolen));
			cur->bh = NULL;
			cur->jlist = NULL;
		}
		cur = cur->hnext;
	}
}

static void free_journal_ram(struct super_block *p_s_sb)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	reiserfs_kfree(journal->j_current_jl,
		       sizeof(struct reiserfs_journal_list), p_s_sb);
	journal->j_num_lists--;

	vfree(journal->j_cnode_free_orig);
	free_list_bitmaps(p_s_sb, journal->j_list_bitmap);
	free_bitmap_nodes(p_s_sb);	/* must be after free_list_bitmaps */
	if (journal->j_header_bh) {
		brelse(journal->j_header_bh);
	}
	/* j_header_bh is on the journal dev, make sure not to release the journal
	 * dev until we brelse j_header_bh
	 */
	release_journal_dev(p_s_sb, journal);
	vfree(journal);
Linus Torvalds's avatar
Linus Torvalds committed
1826
1827
1828
1829
1830
1831
}

/*
** call on unmount.  Only set error to 1 if you haven't made your way out
** of read_super() yet.  Any other caller must keep error at 0.
*/
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
static int do_journal_release(struct reiserfs_transaction_handle *th,
			      struct super_block *p_s_sb, int error)
{
	struct reiserfs_transaction_handle myth;
	int flushed = 0;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);

	/* we only want to flush out transactions if we were called with error == 0
	 */
	if (!error && !(p_s_sb->s_flags & MS_RDONLY)) {
		/* end the current trans */
		BUG_ON(!th->t_trans_id);
		do_journal_end(th, p_s_sb, 10, FLUSH_ALL);

		/* make sure something gets logged to force our way into the flush code */
		if (!journal_join(&myth, p_s_sb, 1)) {
			reiserfs_prepare_for_journal(p_s_sb,
						     SB_BUFFER_WITH_SB(p_s_sb),
						     1);
			journal_mark_dirty(&myth, p_s_sb,
					   SB_BUFFER_WITH_SB(p_s_sb));
			do_journal_end(&myth, p_s_sb, 1, FLUSH_ALL);
			flushed = 1;
		}
	}

	/* this also catches errors during the do_journal_end above */
	if (!error && reiserfs_is_journal_aborted(journal)) {
		memset(&myth, 0, sizeof(myth));
		if (!journal_join_abort(&myth, p_s_sb, 1)) {
			reiserfs_prepare_for_journal(p_s_sb,
						     SB_BUFFER_WITH_SB(p_s_sb),
						     1);
			journal_mark_dirty(&myth, p_s_sb,
					   SB_BUFFER_WITH_SB(p_s_sb));
			do_journal_end(&myth, p_s_sb, 1, FLUSH_ALL);
		}
	}

	reiserfs_mounted_fs_count--;
	/* wait for all commits to finish */
	cancel_delayed_work(&SB_JOURNAL(p_s_sb)->j_work);
	flush_workqueue(commit_wq);
	if (!reiserfs_mounted_fs_count) {
		destroy_workqueue(commit_wq);
		commit_wq = NULL;
	}

	free_journal_ram(p_s_sb);

	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
1883
1884
1885
1886
1887
}

/*
** call on unmount.  flush all journal trans, release all alloc'd ram
*/
1888
1889
1890
1891
int journal_release(struct reiserfs_transaction_handle *th,
		    struct super_block *p_s_sb)
{
	return do_journal_release(th, p_s_sb, 0);
Linus Torvalds's avatar
Linus Torvalds committed
1892
}
1893

Linus Torvalds's avatar
Linus Torvalds committed
1894
1895
1896
/*
** only call from an error condition inside reiserfs_read_super!
*/
1897
1898
1899
1900
int journal_release_error(struct reiserfs_transaction_handle *th,
			  struct super_block *p_s_sb)
{
	return do_journal_release(th, p_s_sb, 1);
Linus Torvalds's avatar
Linus Torvalds committed
1901
1902
1903
}

/* compares description block with commit block.  returns 1 if they differ, 0 if they are the same */
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
static int journal_compare_desc_commit(struct super_block *p_s_sb,
				       struct reiserfs_journal_desc *desc,
				       struct reiserfs_journal_commit *commit)
{
	if (get_commit_trans_id(commit) != get_desc_trans_id(desc) ||
	    get_commit_trans_len(commit) != get_desc_trans_len(desc) ||
	    get_commit_trans_len(commit) > SB_JOURNAL(p_s_sb)->j_trans_max ||
	    get_commit_trans_len(commit) <= 0) {
		return 1;
	}
	return 0;
Linus Torvalds's avatar
Linus Torvalds committed
1915
}
1916

Linus Torvalds's avatar
Linus Torvalds committed
1917
1918
1919
1920
/* returns 0 if it did not find a description block  
** returns -1 if it found a corrupt commit block
** returns 1 if both desc and commit were valid 
*/
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
static int journal_transaction_is_valid(struct super_block *p_s_sb,
					struct buffer_head *d_bh,
					unsigned long *oldest_invalid_trans_id,
					unsigned long *newest_mount_id)
{
	struct reiserfs_journal_desc *desc;
	struct reiserfs_journal_commit *commit;
	struct buffer_head *c_bh;
	unsigned long offset;

	if (!d_bh)
		return 0;

	desc = (struct reiserfs_journal_desc *)d_bh->b_data;
	if (get_desc_trans_len(desc) > 0
	    && !memcmp(get_journal_desc_magic(d_bh), JOURNAL_DESC_MAGIC, 8)) {
		if (oldest_invalid_trans_id && *oldest_invalid_trans_id
		    && get_desc_trans_id(desc) > *oldest_invalid_trans_id) {
			reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
				       "journal-986: transaction "
				       "is valid returning because trans_id %d is greater than "
				       "oldest_invalid %lu",
				       get_desc_trans_id(desc),
				       *oldest_invalid_trans_id);
			return 0;
		}
		if (newest_mount_id
		    && *newest_mount_id > get_desc_mount_id(desc)) {
			reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
				       "journal-1087: transaction "
				       "is valid returning because mount_id %d is less than "
				       "newest_mount_id %lu",
				       get_desc_mount_id(desc),
				       *newest_mount_id);
			return -1;
		}
		if (get_desc_trans_len(desc) > SB_JOURNAL(p_s_sb)->j_trans_max) {
			reiserfs_warning(p_s_sb,
					 "journal-2018: Bad transaction length %d encountered, ignoring transaction",
					 get_desc_trans_len(desc));
			return -1;
		}
		offset = d_bh->b_blocknr - SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb);

		/* ok, we have a journal description block, lets see if the transaction was valid */
		c_bh =
		    journal_bread(p_s_sb,
				  SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
				  ((offset + get_desc_trans_len(desc) +
				    1) % SB_ONDISK_JOURNAL_SIZE(p_s_sb)));
		if (!c_bh)
			return 0;
		commit = (struct reiserfs_journal_commit *)c_bh->b_data;
		if (journal_compare_desc_commit(p_s_sb, desc, commit)) {
			reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
				       "journal_transaction_is_valid, commit offset %ld had bad "
				       "time %d or length %d",
				       c_bh->b_blocknr -
				       SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb),
				       get_commit_trans_id(commit),
				       get_commit_trans_len(commit));
			brelse(c_bh);
			if (oldest_invalid_trans_id) {
				*oldest_invalid_trans_id =
				    get_desc_trans_id(desc);
				reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
					       "journal-1004: "
					       "transaction_is_valid setting oldest invalid trans_id "
					       "to %d",
					       get_desc_trans_id(desc));
			}
			return -1;
		}
		brelse(c_bh);
		reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
			       "journal-1006: found valid "
			       "transaction start offset %llu, len %d id %d",
			       d_bh->b_blocknr -
			       SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb),
			       get_desc_trans_len(desc),
			       get_desc_trans_id(desc));
		return 1;
	} else {
		return 0;
	}
}

static void brelse_array(struct buffer_head **heads, int num)
{
	int i;
	for (i = 0; i < num; i++) {
		brelse(heads[i]);
	}
Linus Torvalds's avatar
Linus Torvalds committed
2014
2015
2016
2017
2018
2019
2020
}

/*
** given the start, and values for the oldest acceptable transactions,
** this either reads in a replays a transaction, or returns because the transaction
** is invalid, or too old.
*/
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
static int journal_read_transaction(struct super_block *p_s_sb,
				    unsigned long cur_dblock,
				    unsigned long oldest_start,
				    unsigned long oldest_trans_id,
				    unsigned long newest_mount_id)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_journal_desc *desc;
	struct reiserfs_journal_commit *commit;
	unsigned long trans_id = 0;
	struct buffer_head *c_bh;
	struct buffer_head *d_bh;
	struct buffer_head **log_blocks = NULL;
	struct buffer_head **real_blocks = NULL;
	unsigned long trans_offset;
	int i;
	int trans_half;

	d_bh = journal_bread(p_s_sb, cur_dblock);
	if (!d_bh)
		return 1;
	desc = (struct reiserfs_journal_desc *)d_bh->b_data;
	trans_offset = d_bh->b_blocknr - SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb);
	reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1037: "
		       "journal_read_transaction, offset %llu, len %d mount_id %d",
		       d_bh->b_blocknr - SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb),
		       get_desc_trans_len(desc), get_desc_mount_id(desc));
	if (get_desc_trans_id(desc) < oldest_trans_id) {
		reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1039: "
			       "journal_read_trans skipping because %lu is too old",
			       cur_dblock -
			       SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb));
		brelse(d_bh);
		return 1;
	}
	if (get_desc_mount_id(desc) != newest_mount_id) {
		reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1146: "
			       "journal_read_trans skipping because %d is != "
			       "newest_mount_id %lu", get_desc_mount_id(desc),
			       newest_mount_id);
		brelse(d_bh);
		return 1;
	}
	c_bh = journal_bread(p_s_sb, SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
			     ((trans_offset + get_desc_trans_len(desc) + 1) %
			      SB_ONDISK_JOURNAL_SIZE(p_s_sb)));
	if (!c_bh) {
		brelse(d_bh);
		return 1;
	}
	commit = (struct reiserfs_journal_commit *)c_bh->b_data;
	if (journal_compare_desc_commit(p_s_sb, desc, commit)) {
		reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
			       "journal_read_transaction, "
			       "commit offset %llu had bad time %d or length %d",
			       c_bh->b_blocknr -
			       SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb),
			       get_commit_trans_id(commit),
			       get_commit_trans_len(commit));
		brelse(c_bh);
		brelse(d_bh);
		return 1;
	}
	trans_id = get_desc_trans_id(desc);
	/* now we know we've got a good transaction, and it was inside the valid time ranges */
	log_blocks =
	    reiserfs_kmalloc(get_desc_trans_len(desc) *
			     sizeof(struct buffer_head *), GFP_NOFS, p_s_sb);
	real_blocks =
	    reiserfs_kmalloc(get_desc_trans_len(desc) *
			     sizeof(struct buffer_head *), GFP_NOFS, p_s_sb);
	if (!log_blocks || !real_blocks) {
		brelse(c_bh);
		brelse(d_bh);
		reiserfs_kfree(log_blocks,
			       get_desc_trans_len(desc) *
			       sizeof(struct buffer_head *), p_s_sb);
		reiserfs_kfree(real_blocks,
			       get_desc_trans_len(desc) *
			       sizeof(struct buffer_head *), p_s_sb);
		reiserfs_warning(p_s_sb,
				 "journal-1169: kmalloc failed, unable to mount FS");
		return -1;
	}
	/* get all the buffer heads */
	trans_half = journal_trans_half(p_s_sb->s_blocksize);
	for (i = 0; i < get_desc_trans_len(desc); i++) {
		log_blocks[i] =
		    journal_getblk(p_s_sb,
				   SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
				   (trans_offset + 1 +
				    i) % SB_ONDISK_JOURNAL_SIZE(p_s_sb));
		if (i < trans_half) {
			real_blocks[i] =
			    sb_getblk(p_s_sb,
				      le32_to_cpu(desc->j_realblock[i]));
		} else {
			real_blocks[i] =
			    sb_getblk(p_s_sb,
				      le32_to_cpu(commit->
						  j_realblock[i - trans_half]));
		}
		if (real_blocks[i]->b_blocknr > SB_BLOCK_COUNT(p_s_sb)) {
			reiserfs_warning(p_s_sb,
					 "journal-1207: REPLAY FAILURE fsck required! Block to replay is outside of filesystem");
			goto abort_replay;
		}
		/* make sure we don't try to replay onto log or reserved area */
		if (is_block_in_log_or_reserved_area
		    (p_s_sb, real_blocks[i]->b_blocknr)) {
			reiserfs_warning(p_s_sb,
					 "journal-1204: REPLAY FAILURE fsck required! Trying to replay onto a log block");
		      abort_replay:
			brelse_array(log_blocks, i);
			brelse_array(real_blocks, i);
			brelse(c_bh);
			brelse(d_bh);
			reiserfs_kfree(log_blocks,
				       get_desc_trans_len(desc) *
				       sizeof(struct buffer_head *), p_s_sb);
			reiserfs_kfree(real_blocks,
				       get_desc_trans_len(desc) *
				       sizeof(struct buffer_head *), p_s_sb);
			return -1;
		}
	}
	/* read in the log blocks, memcpy to the corresponding real block */
	ll_rw_block(READ, get_desc_trans_len(desc), log_blocks);
	for (i = 0; i < get_desc_trans_len(desc); i++) {
		wait_on_buffer(log_blocks[i]);
		if (!buffer_uptodate(log_blocks[i])) {
			reiserfs_warning(p_s_sb,
					 "journal-1212: REPLAY FAILURE fsck required! buffer write failed");
			brelse_array(log_blocks + i,
				     get_desc_trans_len(desc) - i);
			brelse_array(real_blocks, get_desc_trans_len(desc));
			brelse(c_bh);
			brelse(d_bh);
			reiserfs_kfree(log_blocks,
				       get_desc_trans_len(desc) *
				       sizeof(struct buffer_head *), p_s_sb);
			reiserfs_kfree(real_blocks,
				       get_desc_trans_len(desc) *
				       sizeof(struct buffer_head *), p_s_sb);
			return -1;
		}
		memcpy(real_blocks[i]->b_data, log_blocks[i]->b_data,
		       real_blocks[i]->b_size);
		set_buffer_uptodate(real_blocks[i]);