journal.c 124.1 KB
Newer Older
L
Linus Torvalds 已提交
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
/*
** 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 <asm/uaccess.h>
#include <asm/system.h>

#include <linux/time.h>
41
#include <linux/semaphore.h>
L
Linus Torvalds 已提交
42 43 44 45 46 47 48 49 50 51 52 53 54 55

#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>
56
#include <linux/backing-dev.h>
L
Linus Torvalds 已提交
57 58 59 60 61 62 63 64 65 66 67 68 69 70

/* 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 */
L
Linus Torvalds 已提交
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 */
L
Linus Torvalds 已提交
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 */
L
Linus Torvalds 已提交
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);
L
Linus Torvalds 已提交
105
static int dirty_one_transaction(struct super_block *s,
106
				 struct reiserfs_journal_list *jl);
D
David Howells 已提交
107
static void flush_async_commits(struct work_struct *work);
L
Linus Torvalds 已提交
108 109 110 111
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 */
L
Linus Torvalds 已提交
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);
L
Linus Torvalds 已提交
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 *));
L
Linus Torvalds 已提交
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;
L
Linus Torvalds 已提交
140 141 142 143
}

static void disable_barrier(struct super_block *s)
{
144 145 146 147 148 149 150 151 152 153 154
	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;

155
	bn = kmalloc(sizeof(struct reiserfs_bitmap_node), GFP_NOFS);
156 157 158
	if (!bn) {
		return NULL;
	}
159
	bn->data = kzalloc(p_s_sb->s_blocksize, GFP_NOFS);
160
	if (!bn->data) {
161
		kfree(bn);
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
		return NULL;
	}
	bn->id = id++;
	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;
L
Linus Torvalds 已提交
191 192
}
static inline void free_bitmap_node(struct super_block *p_s_sb,
193 194 195 196 197
				    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) {
198 199
		kfree(bn->data);
		kfree(bn);
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
	} 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 
		}
	}
L
Linus Torvalds 已提交
220 221
}

222 223
static int set_bit_in_list_bitmap(struct super_block *p_s_sb,
				  b_blocknr_t block,
224 225
				  struct reiserfs_list_bitmap *jb)
{
226 227
	unsigned int bmap_nr = block / (p_s_sb->s_blocksize << 3);
	unsigned int bit_nr = block % (p_s_sb->s_blocksize << 3);
L
Linus Torvalds 已提交
228

229 230 231 232 233
	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;
L
Linus Torvalds 已提交
234 235 236
}

static void cleanup_bitmap_list(struct super_block *p_s_sb,
237 238 239 240 241 242
				struct reiserfs_list_bitmap *jb)
{
	int i;
	if (jb->bitmaps == NULL)
		return;

243
	for (i = 0; i < reiserfs_bmap_count(p_s_sb); i++) {
244 245 246 247 248
		if (jb->bitmaps[i]) {
			free_bitmap_node(p_s_sb, jb->bitmaps[i]);
			jb->bitmaps[i] = NULL;
		}
	}
L
Linus Torvalds 已提交
249 250 251 252 253 254
}

/*
** only call this on FS unmount.
*/
static int free_list_bitmaps(struct super_block *p_s_sb,
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277
			     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);
278 279
		kfree(bn->data);
		kfree(bn);
280 281 282 283 284
		next = journal->j_bitmap_nodes.next;
		journal->j_free_bitmap_nodes--;
	}

	return 0;
L
Linus Torvalds 已提交
285 286 287 288 289 290 291
}

/*
** 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,
292
				   struct reiserfs_list_bitmap *jb_array,
293
				   unsigned int bmap_nr)
294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316
{
	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;
L
Linus Torvalds 已提交
317 318 319 320 321 322
}

/*
** find an available list bitmap.  If you can't find one, flush a commit list 
** and try again
*/
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
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;
L
Linus Torvalds 已提交
351 352 353 354 355 356 357
}

/* 
** 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
*/
358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
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;
L
Linus Torvalds 已提交
378 379 380 381 382
}

/*
** pulls a cnode off the free list, or returns NULL on failure 
*/
383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404
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;
L
Linus Torvalds 已提交
405 406 407 408 409
}

/*
** returns a cnode to the free list 
*/
410 411 412 413
static void free_cnode(struct super_block *p_s_sb,
		       struct reiserfs_journal_cnode *cn)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
L
Linus Torvalds 已提交
414

415
	reiserfs_check_lock_depth(p_s_sb, "free_cnode");
L
Linus Torvalds 已提交
416

417 418 419 420 421 422 423 424 425
	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;
L
Linus Torvalds 已提交
426 427
}

428 429 430 431
static void clear_prepared_bits(struct buffer_head *bh)
{
	clear_buffer_journal_prepared(bh);
	clear_buffer_journal_restore_dirty(bh);
L
Linus Torvalds 已提交
432 433 434 435 436
}

/* 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()
*/
437 438
void reiserfs_check_lock_depth(struct super_block *sb, char *caller)
{
L
Linus Torvalds 已提交
439
#ifdef CONFIG_SMP
440 441 442 443
	if (current->lock_depth < 0) {
		reiserfs_panic(sb, "%s called without kernel lock held",
			       caller);
	}
L
Linus Torvalds 已提交
444
#else
445
	;
L
Linus Torvalds 已提交
446 447 448 449
#endif
}

/* return a cnode with same dev, block number and size in table, or null if not found */
450 451 452 453 454 455 456
static inline struct reiserfs_journal_cnode *get_journal_hash_dev(struct
								  super_block
								  *sb,
								  struct
								  reiserfs_journal_cnode
								  **table,
								  long bl)
L
Linus Torvalds 已提交
457
{
458 459 460 461 462 463 464 465
	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;
L
Linus Torvalds 已提交
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486
}

/*
** 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,
487
			unsigned int bmap_nr, int bit_nr, int search_all,
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
			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;
L
Linus Torvalds 已提交
539 540 541 542
}

/* insert cn into table
*/
543 544 545 546
static inline void insert_journal_hash(struct reiserfs_journal_cnode **table,
				       struct reiserfs_journal_cnode *cn)
{
	struct reiserfs_journal_cnode *cn_orig;
L
Linus Torvalds 已提交
547

548 549 550 551 552 553 554
	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;
L
Linus Torvalds 已提交
555 556 557
}

/* lock the current transaction */
558
static inline void lock_journal(struct super_block *p_s_sb)
559 560 561
{
	PROC_INFO_INC(p_s_sb, journal.lock_journal);
	down(&SB_JOURNAL(p_s_sb)->j_lock);
L
Linus Torvalds 已提交
562 563 564
}

/* unlock the current transaction */
565
static inline void unlock_journal(struct super_block *p_s_sb)
566 567
{
	up(&SB_JOURNAL(p_s_sb)->j_lock);
L
Linus Torvalds 已提交
568 569 570 571
}

static inline void get_journal_list(struct reiserfs_journal_list *jl)
{
572
	jl->j_refcount++;
L
Linus Torvalds 已提交
573 574 575
}

static inline void put_journal_list(struct super_block *s,
576
				    struct reiserfs_journal_list *jl)
L
Linus Torvalds 已提交
577
{
578 579 580 581 582
	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)
583
		kfree(jl);
L
Linus Torvalds 已提交
584 585 586 587 588 589 590
}

/*
** 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.
*/
591 592 593
static void cleanup_freed_for_journal_list(struct super_block *p_s_sb,
					   struct reiserfs_journal_list *jl)
{
L
Linus Torvalds 已提交
594

595 596 597 598 599 600
	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;
L
Linus Torvalds 已提交
601 602 603
}

static int journal_list_still_alive(struct super_block *s,
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
				    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;
}

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
/*
 * If page->mapping was null, we failed to truncate this page for
 * some reason.  Most likely because it was truncated after being
 * logged via data=journal.
 *
 * This does a check to see if the buffer belongs to one of these
 * lost pages before doing the final put_bh.  If page->mapping was
 * null, it tries to free buffers on the page, which should make the
 * final page_cache_release drop the page from the lru.
 */
static void release_buffer_page(struct buffer_head *bh)
{
	struct page *page = bh->b_page;
	if (!page->mapping && !TestSetPageLocked(page)) {
		page_cache_get(page);
		put_bh(bh);
		if (!page->mapping)
			try_to_free_buffers(page);
		unlock_page(page);
		page_cache_release(page);
	} else {
		put_bh(bh);
	}
}

644 645 646 647 648 649 650 651 652 653 654 655 656
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);
657

658
	unlock_buffer(bh);
659
	release_buffer_page(bh);
660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702
}

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);
L
Linus Torvalds 已提交
703 704 705
}

static void check_barrier_completion(struct super_block *s,
706 707 708 709 710 711 712 713 714
				     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);
	}
L
Linus Torvalds 已提交
715 716 717 718
}

#define CHUNK_SIZE 32
struct buffer_chunk {
719 720
	struct buffer_head *bh[CHUNK_SIZE];
	int nr;
L
Linus Torvalds 已提交
721 722
};

723 724 725 726 727 728 729 730 731
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();
L
Linus Torvalds 已提交
732 733
}

734 735 736 737 738 739 740 741 742
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();
L
Linus Torvalds 已提交
743 744 745
}

static int add_to_chunk(struct buffer_chunk *chunk, struct buffer_head *bh,
746
			spinlock_t * lock, void (fn) (struct buffer_chunk *))
L
Linus Torvalds 已提交
747
{
748
	int ret = 0;
749
	BUG_ON(chunk->nr >= CHUNK_SIZE);
750 751 752 753 754 755 756 757 758 759
	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;
L
Linus Torvalds 已提交
760 761 762
}

static atomic_t nr_reiserfs_jh = ATOMIC_INIT(0);
763 764 765 766 767 768 769 770 771 772
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();
L
Linus Torvalds 已提交
773 774 775 776 777 778 779
	}
}

/*
 * we want to free the jh when the buffer has been written
 * and waited on
 */
780 781 782 783 784 785 786 787 788 789 790 791 792 793 794
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);
	}
L
Linus Torvalds 已提交
795 796 797
}

static inline int __add_jh(struct reiserfs_journal *j, struct buffer_head *bh,
798
			   int tail)
L
Linus Torvalds 已提交
799
{
800
	struct reiserfs_jh *jh;
L
Linus Torvalds 已提交
801

802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817
	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
		 */
818
		BUG_ON(bh->b_private);
819 820
		jh->bh = bh;
		bh->b_private = jh;
L
Linus Torvalds 已提交
821
	}
822 823 824 825 826 827 828 829
	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;
L
Linus Torvalds 已提交
830 831
}

832 833 834
int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh)
{
	return __add_jh(SB_JOURNAL(inode->i_sb), bh, 1);
L
Linus Torvalds 已提交
835
}
836 837 838
int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh)
{
	return __add_jh(SB_JOURNAL(inode->i_sb), bh, 0);
L
Linus Torvalds 已提交
839 840 841
}

#define JH_ENTRY(l) list_entry((l), struct reiserfs_jh, list)
842
static int write_ordered_buffers(spinlock_t * lock,
L
Linus Torvalds 已提交
843
				 struct reiserfs_journal *j,
844
				 struct reiserfs_journal_list *jl,
L
Linus Torvalds 已提交
845 846
				 struct list_head *list)
{
847 848 849 850 851 852 853 854 855 856 857 858 859 860 861
	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)) {
A
Akinobu Mita 已提交
862
				list_move(&jh->list, &tmp);
863 864 865 866 867 868 869 870 871 872
				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;
		}
873 874 875 876 877 878 879 880
		/* in theory, dirty non-uptodate buffers should never get here,
		 * but the upper layer io error paths still have a few quirks.
		 * Handle them here as gracefully as we can
		 */
		if (!buffer_uptodate(bh) && buffer_dirty(bh)) {
			clear_buffer_dirty(bh);
			ret = -EIO;
		}
881
		if (buffer_dirty(bh)) {
A
Akinobu Mita 已提交
882
			list_move(&jh->list, &tmp);
883 884 885 886 887 888 889 890 891 892 893
			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);
L
Linus Torvalds 已提交
894
		write_ordered_chunk(&chunk);
895
		spin_lock(lock);
L
Linus Torvalds 已提交
896
	}
897 898 899 900 901 902 903 904 905 906 907 908 909 910
	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;
		}
911 912 913 914 915 916 917 918 919 920 921 922 923
		/* ugly interaction with invalidatepage here.
		 * reiserfs_invalidate_page will pin any buffer that has a valid
		 * journal head from an older transaction.  If someone else sets
		 * our buffer dirty after we write it in the first loop, and
		 * then someone truncates the page away, nobody will ever write
		 * the buffer. We're safe if we write the page one last time
		 * after freeing the journal header.
		 */
		if (buffer_dirty(bh) && unlikely(bh->b_page->mapping == NULL)) {
			spin_unlock(lock);
			ll_rw_block(WRITE, 1, &bh);
			spin_lock(lock);
		}
924 925
		put_bh(bh);
		cond_resched_lock(lock);
L
Linus Torvalds 已提交
926
	}
927 928 929
	spin_unlock(lock);
	return ret;
}
L
Linus Torvalds 已提交
930

931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
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;
L
Linus Torvalds 已提交
953

954 955 956
		first_jl = other_jl;
		entry = other_jl->j_list.prev;
	}
L
Linus Torvalds 已提交
957

958 959 960 961
	/* if we didn't find any older uncommitted transactions, return now */
	if (first_jl == jl) {
		return 0;
	}
L
Linus Torvalds 已提交
962

963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991
	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;
L
Linus Torvalds 已提交
992 993
		}
	}
994
	return 0;
L
Linus Torvalds 已提交
995
}
A
Adrian Bunk 已提交
996 997

static int reiserfs_async_progress_wait(struct super_block *s)
998 999 1000 1001
{
	DEFINE_WAIT(wait);
	struct reiserfs_journal *j = SB_JOURNAL(s);
	if (atomic_read(&j->j_async_throttle))
1002
		congestion_wait(WRITE, HZ / 10);
1003
	return 0;
L
Linus Torvalds 已提交
1004 1005 1006 1007 1008 1009 1010 1011 1012
}

/*
** 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
**
*/
1013 1014 1015 1016
static int flush_commit_list(struct super_block *s,
			     struct reiserfs_journal_list *jl, int flushall)
{
	int i;
1017
	b_blocknr_t bn;
1018 1019 1020 1021 1022
	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;
1023
	int write_len;
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064

	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)) {
1065
		int ret;
1066
		unlock_kernel();
1067 1068 1069 1070
		ret = write_ordered_buffers(&journal->j_dirty_buffers_lock,
					    journal, jl, &jl->j_bh_list);
		if (ret < 0 && retval == 0)
			retval = ret;
1071 1072 1073 1074 1075
		lock_kernel();
	}
	BUG_ON(!list_empty(&jl->j_bh_list));
	/*
	 * for the description block and all the log blocks, submit any buffers
1076 1077 1078 1079
	 * that haven't already reached the disk.  Try to write at least 256
	 * log blocks. later on, we will only wait on blocks that correspond
	 * to this transaction, but while we're unplugging we might as well
	 * get a chunk of data on there.
1080 1081
	 */
	atomic_inc(&journal->j_async_throttle);
1082 1083 1084 1085
	write_len = jl->j_len + 1;
	if (write_len < 256)
		write_len = 256;
	for (i = 0 ; i < write_len ; i++) {
1086 1087 1088
		bn = SB_ONDISK_JOURNAL_1st_BLOCK(s) + (jl->j_start + i) %
		    SB_ONDISK_JOURNAL_SIZE(s);
		tbh = journal_find_get_block(s, bn);
1089 1090 1091 1092 1093
		if (tbh) {
			if (buffer_dirty(tbh))
			    ll_rw_block(WRITE, 1, &tbh) ;
			put_bh(tbh) ;
		}
1094 1095 1096
	}
	atomic_dec(&journal->j_async_throttle);

1097 1098 1099 1100
	/* We're skipping the commit if there's an error */
	if (retval || reiserfs_is_journal_aborted(journal))
		barrier = 0;

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
	/* 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))) {
L
Linus Torvalds 已提交
1126
#ifdef CONFIG_REISERFS_CHECK
1127
			reiserfs_warning(s, "journal-601, buffer write failed");
L
Linus Torvalds 已提交
1128
#endif
1129 1130 1131 1132 1133 1134 1135 1136 1137 1138
			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) {
1139 1140
		/* If there was a write error in the journal - we can't commit
		 * this transaction - it will be invalid and, if successful,
1141
		 * will just end up propagating the write error out to
1142 1143 1144 1145 1146 1147 1148
		 * the file system. */
		if (likely(!retval && !reiserfs_is_journal_aborted (journal))) {
			if (buffer_dirty(jl->j_commit_bh))
				BUG();
			mark_buffer_dirty(jl->j_commit_bh) ;
			sync_dirty_buffer(jl->j_commit_bh) ;
		}
1149 1150 1151 1152 1153 1154 1155
	} 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
1156
	 * up propagating the write error out to the filesystem. */
1157
	if (unlikely(!buffer_uptodate(jl->j_commit_bh))) {
L
Linus Torvalds 已提交
1158
#ifdef CONFIG_REISERFS_CHECK
1159
		reiserfs_warning(s, "journal-615: buffer write failed");
L
Linus Torvalds 已提交
1160
#endif
1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192
		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;
L
Linus Torvalds 已提交
1193 1194 1195 1196 1197 1198
}

/*
** 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 
*/
1199 1200 1201 1202 1203 1204
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;
L
Linus Torvalds 已提交
1205

1206 1207 1208 1209 1210 1211 1212 1213
	cn = cn->hprev;
	while (cn) {
		if (cn->sb == sb && cn->blocknr == blocknr && cn->jlist) {
			return cn->jlist;
		}
		cn = cn->hprev;
	}
	return NULL;
L
Linus Torvalds 已提交
1214 1215
}

1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230
static int newer_jl_done(struct reiserfs_journal_cnode *cn)
{
	struct super_block *sb = cn->sb;
	b_blocknr_t blocknr = cn->blocknr;

	cn = cn->hprev;
	while (cn) {
		if (cn->sb == sb && cn->blocknr == blocknr && cn->jlist &&
		    atomic_read(&cn->jlist->j_commit_left) != 0)
				    return 0;
		cn = cn->hprev;
	}
	return 1;
}

1231 1232 1233 1234
static void remove_journal_hash(struct super_block *,
				struct reiserfs_journal_cnode **,
				struct reiserfs_journal_list *, unsigned long,
				int);
L
Linus Torvalds 已提交
1235 1236 1237 1238 1239 1240

/*
** 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.
*/
1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268
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;
L
Linus Torvalds 已提交
1269 1270 1271 1272 1273 1274 1275 1276 1277
}

/*
** 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
**
*/
1278 1279 1280 1281 1282 1283
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);
L
Linus Torvalds 已提交
1284

1285 1286
	if (reiserfs_is_journal_aborted(journal))
		return -EIO;
L
Linus Torvalds 已提交
1287

1288 1289 1290 1291
	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))) {
L
Linus Torvalds 已提交
1292
#ifdef CONFIG_REISERFS_CHECK
1293 1294
				reiserfs_warning(p_s_sb,
						 "journal-699: buffer write failed");
L
Linus Torvalds 已提交
1295
#endif
1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336
				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);
L
Linus Torvalds 已提交
1337
}
1338

L
Linus Torvalds 已提交
1339 1340 1341 1342 1343
/* 
** 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,
1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368
				     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;
L
Linus Torvalds 已提交
1369 1370 1371
}

static void del_from_work_list(struct super_block *s,
1372 1373 1374 1375 1376 1377 1378
			       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--;
	}
L
Linus Torvalds 已提交
1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389
}

/* 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
*/
1390 1391
static int flush_journal_list(struct super_block *s,
			      struct reiserfs_journal_list *jl, int flushall)
L
Linus Torvalds 已提交
1392
{
1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410
	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);
L
Linus Torvalds 已提交
1411

1412 1413 1414 1415 1416 1417
	/* if flushall == 0, the lock is already held */
	if (flushall) {
		down(&journal->j_flush_sem);
	} else if (!down_trylock(&journal->j_flush_sem)) {
		BUG();
	}
L
Linus Torvalds 已提交
1418

1419 1420 1421 1422 1423 1424 1425
	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;
	}
L
Linus Torvalds 已提交
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
	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
M
Matt LaPlante 已提交
1495
		 ** sure they are committed, and don't try writing it to disk
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
		 */
		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));

1577
				/* drop one ref for us */
1578
				put_bh(cn->bh);
1579 1580
				/* drop one ref for journal_mark_dirty */
				release_buffer_page(cn->bh);
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
			}
			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;
}

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
static int test_transaction(struct super_block *s,
                            struct reiserfs_journal_list *jl)
{
	struct reiserfs_journal_cnode *cn;

	if (jl->j_len == 0 || atomic_read(&jl->j_nonzerolen) == 0)
		return 1;

	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 && !newer_jl_done(cn))
			return 0;
	      next:
		cn = cn->next;
		cond_resched();
	}
	return 0;
}

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 1720 1721 1722 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 1765 1766 1767 1768 1769 1770 1771 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
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
L
Linus Torvalds 已提交
1826 1827 1828
** list updates the header block
*/
static int flush_used_journal_lists(struct super_block *s,
1829 1830 1831 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
				    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;
L
Linus Torvalds 已提交
1871 1872 1873 1874 1875 1876 1877
}

/*
** 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,
1878 1879 1880 1881 1882 1883 1884 1885 1886 1887 1888 1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916
			 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);
1917
	kfree(journal->j_current_jl);
1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930
	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);
L
Linus Torvalds 已提交
1931 1932 1933 1934 1935 1936
}

/*
** 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.
*/
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
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;
L
Linus Torvalds 已提交
1988 1989 1990 1991 1992
}

/*
** call on unmount.  flush all journal trans, release all alloc'd ram
*/
1993 1994 1995 1996
int journal_release(struct reiserfs_transaction_handle *th,
		    struct super_block *p_s_sb)
{
	return do_journal_release(th, p_s_sb, 0);
L
Linus Torvalds 已提交
1997
}
1998

L
Linus Torvalds 已提交
1999 2000 2001
/*
** only call from an error condition inside reiserfs_read_super!
*/
2002 2003 2004 2005
int journal_release_error(struct reiserfs_transaction_handle *th,
			  struct super_block *p_s_sb)
{
	return do_journal_release(th, p_s_sb, 1);
L
Linus Torvalds 已提交
2006 2007 2008
}

/* compares description block with commit block.  returns 1 if they differ, 0 if they are the same */
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
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;
L
Linus Torvalds 已提交
2020
}
2021

L
Linus Torvalds 已提交
2022 2023 2024 2025
/* 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 
*/
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
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]);
	}
L
Linus Torvalds 已提交
2119 2120 2121 2122 2123 2124 2125
}

/*
** 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.
*/
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 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190
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 */
2191 2192 2193 2194
	log_blocks = kmalloc(get_desc_trans_len(desc) *
			     sizeof(struct buffer_head *), GFP_NOFS);
	real_blocks = kmalloc(get_desc_trans_len(desc) *
			      sizeof(struct buffer_head *), GFP_NOFS);
2195 2196 2197
	if (!log_blocks || !real_blocks) {
		brelse(c_bh);
		brelse(d_bh);
2198 2199
		kfree(log_blocks);
		kfree(real_blocks);
2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236
		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);
2237 2238
			kfree(log_blocks);
			kfree(real_blocks);
2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253
			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);
2254 2255
			kfree(log_blocks);
			kfree(real_blocks);
2256 2257 2258 2259 2260 2261 2262 2263 2264 2265
			return -1;
		}
		memcpy(real_blocks[i]->b_data, log_blocks[i]->b_data,
		       real_blocks[i]->b_size);
		set_buffer_uptodate(real_blocks[i]);
		brelse(log_blocks[i]);
	}
	/* flush out the real blocks */
	for (i = 0; i < get_desc_trans_len(desc); i++) {
		set_buffer_dirty(real_blocks[i]);
2266
		ll_rw_block(SWRITE, 1, real_blocks + i);
2267 2268 2269 2270 2271 2272 2273 2274 2275 2276
	}
	for (i = 0; i < get_desc_trans_len(desc); i++) {
		wait_on_buffer(real_blocks[i]);
		if (!buffer_uptodate(real_blocks[i])) {
			reiserfs_warning(p_s_sb,
					 "journal-1226: REPLAY FAILURE, fsck required! buffer write failed");
			brelse_array(real_blocks + i,
				     get_desc_trans_len(desc) - i);
			brelse(c_bh);
			brelse(d_bh);
2277 2278
			kfree(log_blocks);
			kfree(real_blocks);
2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294
			return -1;
		}
		brelse(real_blocks[i]);
	}
	cur_dblock =
	    SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
	    ((trans_offset + get_desc_trans_len(desc) +
	      2) % SB_ONDISK_JOURNAL_SIZE(p_s_sb));
	reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
		       "journal-1095: setting journal " "start to offset %ld",
		       cur_dblock - SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb));

	/* init starting values for the first transaction, in case this is the last transaction to be replayed. */
	journal->j_start = cur_dblock - SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb);
	journal->j_last_flush_trans_id = trans_id;
	journal->j_trans_id = trans_id + 1;
2295 2296 2297
	/* check for trans_id overflow */
	if (journal->j_trans_id == 0)
		journal->j_trans_id = 10;
2298 2299
	brelse(c_bh);
	brelse(d_bh);
2300 2301
	kfree(log_blocks);
	kfree(real_blocks);
2302
	return 0;
L
Linus Torvalds 已提交
2303 2304 2305 2306 2307 2308 2309 2310
}

/* This function reads blocks starting from block and to max_block of bufsize
   size (but no more than BUFNR blocks at a time). This proved to improve
   mounting speed on self-rebuilding raid5 arrays at least.
   Right now it is only used from journal code. But later we might use it
   from other places.
   Note: Do not use journal_getblk/sb_getblk functions here! */
2311 2312 2313
static struct buffer_head *reiserfs_breada(struct block_device *dev,
					   b_blocknr_t block, int bufsize,
					   b_blocknr_t max_block)
L
Linus Torvalds 已提交
2314
{
2315
	struct buffer_head *bhlist[BUFNR];
L
Linus Torvalds 已提交
2316
	unsigned int blocks = BUFNR;
2317
	struct buffer_head *bh;
L
Linus Torvalds 已提交
2318
	int i, j;
2319 2320 2321 2322 2323

	bh = __getblk(dev, block, bufsize);
	if (buffer_uptodate(bh))
		return (bh);

L
Linus Torvalds 已提交
2324 2325 2326 2327 2328 2329
	if (block + BUFNR > max_block) {
		blocks = max_block - block;
	}
	bhlist[0] = bh;
	j = 1;
	for (i = 1; i < blocks; i++) {
2330 2331 2332
		bh = __getblk(dev, block + i, bufsize);
		if (buffer_uptodate(bh)) {
			brelse(bh);
L
Linus Torvalds 已提交
2333
			break;
2334 2335
		} else
			bhlist[j++] = bh;
L
Linus Torvalds 已提交
2336
	}
2337 2338 2339
	ll_rw_block(READ, j, bhlist);
	for (i = 1; i < j; i++)
		brelse(bhlist[i]);
L
Linus Torvalds 已提交
2340
	bh = bhlist[0];
2341 2342
	wait_on_buffer(bh);
	if (buffer_uptodate(bh))
L
Linus Torvalds 已提交
2343
		return bh;
2344
	brelse(bh);
L
Linus Torvalds 已提交
2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356
	return NULL;
}

/*
** read and replay the log
** on a clean unmount, the journal header's next unflushed pointer will be to an invalid
** transaction.  This tests that before finding all the transactions in the log, which makes normal mount times fast.
**
** After a crash, this starts with the next unflushed transaction, and replays until it finds one too old, or invalid.
**
** On exit, it sets things up so the first transaction will work correctly.
*/
2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390
static int journal_read(struct super_block *p_s_sb)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_journal_desc *desc;
	unsigned long oldest_trans_id = 0;
	unsigned long oldest_invalid_trans_id = 0;
	time_t start;
	unsigned long oldest_start = 0;
	unsigned long cur_dblock = 0;
	unsigned long newest_mount_id = 9;
	struct buffer_head *d_bh;
	struct reiserfs_journal_header *jh;
	int valid_journal_header = 0;
	int replay_count = 0;
	int continue_replay = 1;
	int ret;
	char b[BDEVNAME_SIZE];

	cur_dblock = SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb);
	reiserfs_info(p_s_sb, "checking transaction log (%s)\n",
		      bdevname(journal->j_dev_bd, b));
	start = get_seconds();

	/* step 1, read in the journal header block.  Check the transaction it says 
	 ** is the first unflushed, and if that transaction is not valid, 
	 ** replay is done
	 */
	journal->j_header_bh = journal_bread(p_s_sb,
					     SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb)
					     + SB_ONDISK_JOURNAL_SIZE(p_s_sb));
	if (!journal->j_header_bh) {
		return 1;
	}
	jh = (struct reiserfs_journal_header *)(journal->j_header_bh->b_data);
2391
	if (le32_to_cpu(jh->j_first_unflushed_offset) <
2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521
	    SB_ONDISK_JOURNAL_SIZE(p_s_sb)
	    && le32_to_cpu(jh->j_last_flush_trans_id) > 0) {
		oldest_start =
		    SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
		    le32_to_cpu(jh->j_first_unflushed_offset);
		oldest_trans_id = le32_to_cpu(jh->j_last_flush_trans_id) + 1;
		newest_mount_id = le32_to_cpu(jh->j_mount_id);
		reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
			       "journal-1153: found in "
			       "header: first_unflushed_offset %d, last_flushed_trans_id "
			       "%lu", le32_to_cpu(jh->j_first_unflushed_offset),
			       le32_to_cpu(jh->j_last_flush_trans_id));
		valid_journal_header = 1;

		/* now, we try to read the first unflushed offset.  If it is not valid, 
		 ** there is nothing more we can do, and it makes no sense to read 
		 ** through the whole log.
		 */
		d_bh =
		    journal_bread(p_s_sb,
				  SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
				  le32_to_cpu(jh->j_first_unflushed_offset));
		ret = journal_transaction_is_valid(p_s_sb, d_bh, NULL, NULL);
		if (!ret) {
			continue_replay = 0;
		}
		brelse(d_bh);
		goto start_log_replay;
	}

	if (continue_replay && bdev_read_only(p_s_sb->s_bdev)) {
		reiserfs_warning(p_s_sb,
				 "clm-2076: device is readonly, unable to replay log");
		return -1;
	}

	/* ok, there are transactions that need to be replayed.  start with the first log block, find
	 ** all the valid transactions, and pick out the oldest.
	 */
	while (continue_replay
	       && cur_dblock <
	       (SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
		SB_ONDISK_JOURNAL_SIZE(p_s_sb))) {
		/* Note that it is required for blocksize of primary fs device and journal
		   device to be the same */
		d_bh =
		    reiserfs_breada(journal->j_dev_bd, cur_dblock,
				    p_s_sb->s_blocksize,
				    SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
				    SB_ONDISK_JOURNAL_SIZE(p_s_sb));
		ret =
		    journal_transaction_is_valid(p_s_sb, d_bh,
						 &oldest_invalid_trans_id,
						 &newest_mount_id);
		if (ret == 1) {
			desc = (struct reiserfs_journal_desc *)d_bh->b_data;
			if (oldest_start == 0) {	/* init all oldest_ values */
				oldest_trans_id = get_desc_trans_id(desc);
				oldest_start = d_bh->b_blocknr;
				newest_mount_id = get_desc_mount_id(desc);
				reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
					       "journal-1179: Setting "
					       "oldest_start to offset %llu, trans_id %lu",
					       oldest_start -
					       SB_ONDISK_JOURNAL_1st_BLOCK
					       (p_s_sb), oldest_trans_id);
			} else if (oldest_trans_id > get_desc_trans_id(desc)) {
				/* one we just read was older */
				oldest_trans_id = get_desc_trans_id(desc);
				oldest_start = d_bh->b_blocknr;
				reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
					       "journal-1180: Resetting "
					       "oldest_start to offset %lu, trans_id %lu",
					       oldest_start -
					       SB_ONDISK_JOURNAL_1st_BLOCK
					       (p_s_sb), oldest_trans_id);
			}
			if (newest_mount_id < get_desc_mount_id(desc)) {
				newest_mount_id = get_desc_mount_id(desc);
				reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
					       "journal-1299: Setting "
					       "newest_mount_id to %d",
					       get_desc_mount_id(desc));
			}
			cur_dblock += get_desc_trans_len(desc) + 2;
		} else {
			cur_dblock++;
		}
		brelse(d_bh);
	}

      start_log_replay:
	cur_dblock = oldest_start;
	if (oldest_trans_id) {
		reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
			       "journal-1206: Starting replay "
			       "from offset %llu, trans_id %lu",
			       cur_dblock - SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb),
			       oldest_trans_id);

	}
	replay_count = 0;
	while (continue_replay && oldest_trans_id > 0) {
		ret =
		    journal_read_transaction(p_s_sb, cur_dblock, oldest_start,
					     oldest_trans_id, newest_mount_id);
		if (ret < 0) {
			return ret;
		} else if (ret != 0) {
			break;
		}
		cur_dblock =
		    SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) + journal->j_start;
		replay_count++;
		if (cur_dblock == oldest_start)
			break;
	}

	if (oldest_trans_id == 0) {
		reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE,
			       "journal-1225: No valid " "transactions found");
	}
	/* j_start does not get set correctly if we don't replay any transactions.
	 ** if we had a valid journal_header, set j_start to the first unflushed transaction value,
	 ** copy the trans_id from the header
	 */
	if (valid_journal_header && replay_count == 0) {
		journal->j_start = le32_to_cpu(jh->j_first_unflushed_offset);
		journal->j_trans_id =
		    le32_to_cpu(jh->j_last_flush_trans_id) + 1;
2522 2523 2524
		/* check for trans_id overflow */
		if (journal->j_trans_id == 0)
			journal->j_trans_id = 10;
2525 2526 2527 2528 2529 2530
		journal->j_last_flush_trans_id =
		    le32_to_cpu(jh->j_last_flush_trans_id);
		journal->j_mount_id = le32_to_cpu(jh->j_mount_id) + 1;
	} else {
		journal->j_mount_id = newest_mount_id + 1;
	}
L
Linus Torvalds 已提交
2531
	reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1299: Setting "
2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547
		       "newest_mount_id to %lu", journal->j_mount_id);
	journal->j_first_unflushed_offset = journal->j_start;
	if (replay_count > 0) {
		reiserfs_info(p_s_sb,
			      "replayed %d transactions in %lu seconds\n",
			      replay_count, get_seconds() - start);
	}
	if (!bdev_read_only(p_s_sb->s_bdev) &&
	    _update_journal_header_block(p_s_sb, journal->j_start,
					 journal->j_last_flush_trans_id)) {
		/* replay failed, caller must call free_journal_ram and abort
		 ** the mount
		 */
		return -1;
	}
	return 0;
L
Linus Torvalds 已提交
2548 2549 2550 2551
}

static struct reiserfs_journal_list *alloc_journal_list(struct super_block *s)
{
2552
	struct reiserfs_journal_list *jl;
2553 2554
	jl = kzalloc(sizeof(struct reiserfs_journal_list),
		     GFP_NOFS | __GFP_NOFAIL);
2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596
	INIT_LIST_HEAD(&jl->j_list);
	INIT_LIST_HEAD(&jl->j_working_list);
	INIT_LIST_HEAD(&jl->j_tail_bh_list);
	INIT_LIST_HEAD(&jl->j_bh_list);
	sema_init(&jl->j_commit_lock, 1);
	SB_JOURNAL(s)->j_num_lists++;
	get_journal_list(jl);
	return jl;
}

static void journal_list_init(struct super_block *p_s_sb)
{
	SB_JOURNAL(p_s_sb)->j_current_jl = alloc_journal_list(p_s_sb);
}

static int release_journal_dev(struct super_block *super,
			       struct reiserfs_journal *journal)
{
	int result;

	result = 0;

	if (journal->j_dev_file != NULL) {
		result = filp_close(journal->j_dev_file, NULL);
		journal->j_dev_file = NULL;
		journal->j_dev_bd = NULL;
	} else if (journal->j_dev_bd != NULL) {
		result = blkdev_put(journal->j_dev_bd);
		journal->j_dev_bd = NULL;
	}

	if (result != 0) {
		reiserfs_warning(super,
				 "sh-457: release_journal_dev: Cannot release journal device: %i",
				 result);
	}
	return result;
}

static int journal_init_dev(struct super_block *super,
			    struct reiserfs_journal *journal,
			    const char *jdev_name)
L
Linus Torvalds 已提交
2597 2598 2599 2600 2601 2602 2603 2604
{
	int result;
	dev_t jdev;
	int blkdev_mode = FMODE_READ | FMODE_WRITE;
	char b[BDEVNAME_SIZE];

	result = 0;

2605 2606 2607 2608
	journal->j_dev_bd = NULL;
	journal->j_dev_file = NULL;
	jdev = SB_ONDISK_JOURNAL_DEVICE(super) ?
	    new_decode_dev(SB_ONDISK_JOURNAL_DEVICE(super)) : super->s_dev;
L
Linus Torvalds 已提交
2609 2610

	if (bdev_read_only(super->s_bdev))
2611
		blkdev_mode = FMODE_READ;
L
Linus Torvalds 已提交
2612 2613

	/* there is no "jdev" option and journal is on separate device */
2614
	if ((!jdev_name || !jdev_name[0])) {
L
Linus Torvalds 已提交
2615 2616 2617 2618
		journal->j_dev_bd = open_by_devnum(jdev, blkdev_mode);
		if (IS_ERR(journal->j_dev_bd)) {
			result = PTR_ERR(journal->j_dev_bd);
			journal->j_dev_bd = NULL;
2619 2620 2621
			reiserfs_warning(super, "sh-458: journal_init_dev: "
					 "cannot init journal device '%s': %i",
					 __bdevname(jdev, b), result);
L
Linus Torvalds 已提交
2622 2623 2624 2625 2626 2627
			return result;
		} else if (jdev != super->s_dev)
			set_blocksize(journal->j_dev_bd, super->s_blocksize);
		return 0;
	}

2628 2629
	journal->j_dev_file = filp_open(jdev_name, 0, 0);
	if (!IS_ERR(journal->j_dev_file)) {
L
Linus Torvalds 已提交
2630
		struct inode *jdev_inode = journal->j_dev_file->f_mapping->host;
2631
		if (!S_ISBLK(jdev_inode->i_mode)) {
2632
			reiserfs_warning(super, "journal_init_dev: '%s' is "
2633
					 "not a block device", jdev_name);
L
Linus Torvalds 已提交
2634
			result = -ENOTBLK;
2635 2636
			release_journal_dev(super, journal);
		} else {
L
Linus Torvalds 已提交
2637 2638 2639
			/* ok */
			journal->j_dev_bd = I_BDEV(jdev_inode);
			set_blocksize(journal->j_dev_bd, super->s_blocksize);
2640 2641
			reiserfs_info(super,
				      "journal_init_dev: journal device: %s\n",
2642
				      bdevname(journal->j_dev_bd, b));
L
Linus Torvalds 已提交
2643 2644
		}
	} else {
2645 2646 2647 2648 2649
		result = PTR_ERR(journal->j_dev_file);
		journal->j_dev_file = NULL;
		reiserfs_warning(super,
				 "journal_init_dev: Cannot open '%s': %i",
				 jdev_name, result);
L
Linus Torvalds 已提交
2650 2651 2652 2653
	}
	return result;
}

2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666 2667 2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708
/**
 * When creating/tuning a file system user can assign some
 * journal params within boundaries which depend on the ratio
 * blocksize/standard_blocksize.
 *
 * For blocks >= standard_blocksize transaction size should
 * be not less then JOURNAL_TRANS_MIN_DEFAULT, and not more
 * then JOURNAL_TRANS_MAX_DEFAULT.
 *
 * For blocks < standard_blocksize these boundaries should be
 * decreased proportionally.
 */
#define REISERFS_STANDARD_BLKSIZE (4096)

static int check_advise_trans_params(struct super_block *p_s_sb,
				     struct reiserfs_journal *journal)
{
        if (journal->j_trans_max) {
	        /* Non-default journal params.
		   Do sanity check for them. */
	        int ratio = 1;
		if (p_s_sb->s_blocksize < REISERFS_STANDARD_BLKSIZE)
		        ratio = REISERFS_STANDARD_BLKSIZE / p_s_sb->s_blocksize;

		if (journal->j_trans_max > JOURNAL_TRANS_MAX_DEFAULT / ratio ||
		    journal->j_trans_max < JOURNAL_TRANS_MIN_DEFAULT / ratio ||
		    SB_ONDISK_JOURNAL_SIZE(p_s_sb) / journal->j_trans_max <
		    JOURNAL_MIN_RATIO) {
		        reiserfs_warning(p_s_sb,
				 "sh-462: bad transaction max size (%u). FSCK?",
				 journal->j_trans_max);
			return 1;
		}
		if (journal->j_max_batch != (journal->j_trans_max) *
		        JOURNAL_MAX_BATCH_DEFAULT/JOURNAL_TRANS_MAX_DEFAULT) {
		        reiserfs_warning(p_s_sb,
				"sh-463: bad transaction max batch (%u). FSCK?",
				journal->j_max_batch);
			return 1;
		}
	} else {
		/* Default journal params.
                   The file system was created by old version
		   of mkreiserfs, so some fields contain zeros,
		   and we need to advise proper values for them */
	        if (p_s_sb->s_blocksize != REISERFS_STANDARD_BLKSIZE)
	                reiserfs_panic(p_s_sb, "sh-464: bad blocksize (%u)",
				       p_s_sb->s_blocksize);
		journal->j_trans_max = JOURNAL_TRANS_MAX_DEFAULT;
		journal->j_max_batch = JOURNAL_MAX_BATCH_DEFAULT;
		journal->j_max_commit_age = JOURNAL_MAX_COMMIT_AGE;
	}
	return 0;
}

L
Linus Torvalds 已提交
2709 2710 2711
/*
** must be called once on fs mount.  calls journal_read for you
*/
2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736
int journal_init(struct super_block *p_s_sb, const char *j_dev_name,
		 int old_format, unsigned int commit_max_age)
{
	int num_cnodes = SB_ONDISK_JOURNAL_SIZE(p_s_sb) * 2;
	struct buffer_head *bhjh;
	struct reiserfs_super_block *rs;
	struct reiserfs_journal_header *jh;
	struct reiserfs_journal *journal;
	struct reiserfs_journal_list *jl;
	char b[BDEVNAME_SIZE];

	journal = SB_JOURNAL(p_s_sb) = vmalloc(sizeof(struct reiserfs_journal));
	if (!journal) {
		reiserfs_warning(p_s_sb,
				 "journal-1256: unable to get memory for journal structure");
		return 1;
	}
	memset(journal, 0, sizeof(struct reiserfs_journal));
	INIT_LIST_HEAD(&journal->j_bitmap_nodes);
	INIT_LIST_HEAD(&journal->j_prealloc_list);
	INIT_LIST_HEAD(&journal->j_working_list);
	INIT_LIST_HEAD(&journal->j_journal_list);
	journal->j_persistent_trans = 0;
	if (reiserfs_allocate_list_bitmaps(p_s_sb,
					   journal->j_list_bitmap,
2737
					   reiserfs_bmap_count(p_s_sb)))
2738 2739 2740 2741 2742 2743 2744
		goto free_and_return;
	allocate_bitmap_nodes(p_s_sb);

	/* reserved for journal area support */
	SB_JOURNAL_1st_RESERVED_BLOCK(p_s_sb) = (old_format ?
						 REISERFS_OLD_DISK_OFFSET_IN_BYTES
						 / p_s_sb->s_blocksize +
2745
						 reiserfs_bmap_count(p_s_sb) +
2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803
						 1 :
						 REISERFS_DISK_OFFSET_IN_BYTES /
						 p_s_sb->s_blocksize + 2);

	/* Sanity check to see is the standard journal fitting withing first bitmap
	   (actual for small blocksizes) */
	if (!SB_ONDISK_JOURNAL_DEVICE(p_s_sb) &&
	    (SB_JOURNAL_1st_RESERVED_BLOCK(p_s_sb) +
	     SB_ONDISK_JOURNAL_SIZE(p_s_sb) > p_s_sb->s_blocksize * 8)) {
		reiserfs_warning(p_s_sb,
				 "journal-1393: journal does not fit for area "
				 "addressed by first of bitmap blocks. It starts at "
				 "%u and its size is %u. Block size %ld",
				 SB_JOURNAL_1st_RESERVED_BLOCK(p_s_sb),
				 SB_ONDISK_JOURNAL_SIZE(p_s_sb),
				 p_s_sb->s_blocksize);
		goto free_and_return;
	}

	if (journal_init_dev(p_s_sb, journal, j_dev_name) != 0) {
		reiserfs_warning(p_s_sb,
				 "sh-462: unable to initialize jornal device");
		goto free_and_return;
	}

	rs = SB_DISK_SUPER_BLOCK(p_s_sb);

	/* read journal header */
	bhjh = journal_bread(p_s_sb,
			     SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
			     SB_ONDISK_JOURNAL_SIZE(p_s_sb));
	if (!bhjh) {
		reiserfs_warning(p_s_sb,
				 "sh-459: unable to read journal header");
		goto free_and_return;
	}
	jh = (struct reiserfs_journal_header *)(bhjh->b_data);

	/* make sure that journal matches to the super block */
	if (is_reiserfs_jr(rs)
	    && (le32_to_cpu(jh->jh_journal.jp_journal_magic) !=
		sb_jp_journal_magic(rs))) {
		reiserfs_warning(p_s_sb,
				 "sh-460: journal header magic %x "
				 "(device %s) does not match to magic found in super "
				 "block %x", jh->jh_journal.jp_journal_magic,
				 bdevname(journal->j_dev_bd, b),
				 sb_jp_journal_magic(rs));
		brelse(bhjh);
		goto free_and_return;
	}

	journal->j_trans_max = le32_to_cpu(jh->jh_journal.jp_journal_trans_max);
	journal->j_max_batch = le32_to_cpu(jh->jh_journal.jp_journal_max_batch);
	journal->j_max_commit_age =
	    le32_to_cpu(jh->jh_journal.jp_journal_max_commit_age);
	journal->j_max_trans_age = JOURNAL_MAX_TRANS_AGE;

2804 2805
	if (check_advise_trans_params(p_s_sb, journal) != 0)
	        goto free_and_return;
2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848 2849 2850 2851 2852 2853 2854 2855 2856
	journal->j_default_max_commit_age = journal->j_max_commit_age;

	if (commit_max_age != 0) {
		journal->j_max_commit_age = commit_max_age;
		journal->j_max_trans_age = commit_max_age;
	}

	reiserfs_info(p_s_sb, "journal params: device %s, size %u, "
		      "journal first block %u, max trans len %u, max batch %u, "
		      "max commit age %u, max trans age %u\n",
		      bdevname(journal->j_dev_bd, b),
		      SB_ONDISK_JOURNAL_SIZE(p_s_sb),
		      SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb),
		      journal->j_trans_max,
		      journal->j_max_batch,
		      journal->j_max_commit_age, journal->j_max_trans_age);

	brelse(bhjh);

	journal->j_list_bitmap_index = 0;
	journal_list_init(p_s_sb);

	memset(journal->j_list_hash_table, 0,
	       JOURNAL_HASH_SIZE * sizeof(struct reiserfs_journal_cnode *));

	INIT_LIST_HEAD(&journal->j_dirty_buffers);
	spin_lock_init(&journal->j_dirty_buffers_lock);

	journal->j_start = 0;
	journal->j_len = 0;
	journal->j_len_alloc = 0;
	atomic_set(&(journal->j_wcount), 0);
	atomic_set(&(journal->j_async_throttle), 0);
	journal->j_bcount = 0;
	journal->j_trans_start_time = 0;
	journal->j_last = NULL;
	journal->j_first = NULL;
	init_waitqueue_head(&(journal->j_join_wait));
	sema_init(&journal->j_lock, 1);
	sema_init(&journal->j_flush_sem, 1);

	journal->j_trans_id = 10;
	journal->j_mount_id = 10;
	journal->j_state = 0;
	atomic_set(&(journal->j_jlock), 0);
	journal->j_cnode_free_list = allocate_cnodes(num_cnodes);
	journal->j_cnode_free_orig = journal->j_cnode_free_list;
	journal->j_cnode_free = journal->j_cnode_free_list ? num_cnodes : 0;
	journal->j_cnode_used = 0;
	journal->j_must_wait = 0;

2857 2858 2859 2860 2861 2862 2863 2864 2865
	if (journal->j_cnode_free == 0) {
        	reiserfs_warning(p_s_sb, "journal-2004: Journal cnode memory "
		                 "allocation failed (%ld bytes). Journal is "
		                 "too large for available memory. Usually "
		                 "this is due to a journal that is too large.",
		                 sizeof (struct reiserfs_journal_cnode) * num_cnodes);
        	goto free_and_return;
	}

2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882
	init_journal_hash(p_s_sb);
	jl = journal->j_current_jl;
	jl->j_list_bitmap = get_list_bitmap(p_s_sb, jl);
	if (!jl->j_list_bitmap) {
		reiserfs_warning(p_s_sb,
				 "journal-2005, get_list_bitmap failed for journal list 0");
		goto free_and_return;
	}
	if (journal_read(p_s_sb) < 0) {
		reiserfs_warning(p_s_sb, "Replay Failure, unable to mount");
		goto free_and_return;
	}

	reiserfs_mounted_fs_count++;
	if (reiserfs_mounted_fs_count <= 1)
		commit_wq = create_workqueue("reiserfs");

D
David Howells 已提交
2883 2884
	INIT_DELAYED_WORK(&journal->j_work, flush_async_commits);
	journal->j_work_sb = p_s_sb;
2885 2886 2887 2888
	return 0;
      free_and_return:
	free_journal_ram(p_s_sb);
	return 1;
L
Linus Torvalds 已提交
2889 2890 2891 2892 2893 2894 2895
}

/*
** test for a polite end of the current transaction.  Used by file_write, and should
** be used by delete to make sure they don't write more than can fit inside a single
** transaction
*/
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 2906 2907 2908 2909 2910 2911
int journal_transaction_should_end(struct reiserfs_transaction_handle *th,
				   int new_alloc)
{
	struct reiserfs_journal *journal = SB_JOURNAL(th->t_super);
	time_t now = get_seconds();
	/* cannot restart while nested */
	BUG_ON(!th->t_trans_id);
	if (th->t_refcount > 1)
		return 0;
	if (journal->j_must_wait > 0 ||
	    (journal->j_len_alloc + new_alloc) >= journal->j_max_batch ||
	    atomic_read(&(journal->j_jlock)) ||
	    (now - journal->j_trans_start_time) > journal->j_max_trans_age ||
	    journal->j_cnode_free < (journal->j_trans_max * 3)) {
		return 1;
	}
2912 2913 2914
	/* protected by the BKL here */
	journal->j_len_alloc += new_alloc;
	th->t_blocks_allocated += new_alloc ;
2915
	return 0;
L
Linus Torvalds 已提交
2916 2917 2918 2919 2920
}

/* this must be called inside a transaction, and requires the 
** kernel_lock to be held
*/
2921 2922 2923 2924 2925 2926 2927
void reiserfs_block_writes(struct reiserfs_transaction_handle *th)
{
	struct reiserfs_journal *journal = SB_JOURNAL(th->t_super);
	BUG_ON(!th->t_trans_id);
	journal->j_must_wait = 1;
	set_bit(J_WRITERS_BLOCKED, &journal->j_state);
	return;
L
Linus Torvalds 已提交
2928 2929 2930 2931 2932
}

/* this must be called without a transaction started, and does not
** require BKL
*/
2933 2934 2935 2936 2937
void reiserfs_allow_writes(struct super_block *s)
{
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	clear_bit(J_WRITERS_BLOCKED, &journal->j_state);
	wake_up(&journal->j_join_wait);
L
Linus Torvalds 已提交
2938 2939 2940 2941 2942
}

/* this must be called without a transaction started, and does not
** require BKL
*/
2943 2944 2945 2946 2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961
void reiserfs_wait_on_write_block(struct super_block *s)
{
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	wait_event(journal->j_join_wait,
		   !test_bit(J_WRITERS_BLOCKED, &journal->j_state));
}

static void queue_log_writer(struct super_block *s)
{
	wait_queue_t wait;
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	set_bit(J_WRITERS_QUEUED, &journal->j_state);

	/*
	 * we don't want to use wait_event here because
	 * we only want to wait once.
	 */
	init_waitqueue_entry(&wait, current);
	add_wait_queue(&journal->j_join_wait, &wait);
L
Linus Torvalds 已提交
2962
	set_current_state(TASK_UNINTERRUPTIBLE);
2963 2964
	if (test_bit(J_WRITERS_QUEUED, &journal->j_state))
		schedule();
2965
	__set_current_state(TASK_RUNNING);
2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980
	remove_wait_queue(&journal->j_join_wait, &wait);
}

static void wake_queued_writers(struct super_block *s)
{
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	if (test_and_clear_bit(J_WRITERS_QUEUED, &journal->j_state))
		wake_up(&journal->j_join_wait);
}

static void let_transaction_grow(struct super_block *sb, unsigned long trans_id)
{
	struct reiserfs_journal *journal = SB_JOURNAL(sb);
	unsigned long bcount = journal->j_bcount;
	while (1) {
2981
		schedule_timeout_uninterruptible(1);
2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992
		journal->j_current_jl->j_state |= LIST_COMMIT_PENDING;
		while ((atomic_read(&journal->j_wcount) > 0 ||
			atomic_read(&journal->j_jlock)) &&
		       journal->j_trans_id == trans_id) {
			queue_log_writer(sb);
		}
		if (journal->j_trans_id != trans_id)
			break;
		if (bcount == journal->j_bcount)
			break;
		bcount = journal->j_bcount;
L
Linus Torvalds 已提交
2993 2994 2995 2996 2997 2998 2999 3000 3001
	}
}

/* join == true if you must join an existing transaction.
** join == false if you can deal with waiting for others to finish
**
** this will block until the transaction is joinable.  send the number of blocks you
** expect to use in nblocks.
*/
3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013
static int do_journal_begin_r(struct reiserfs_transaction_handle *th,
			      struct super_block *p_s_sb, unsigned long nblocks,
			      int join)
{
	time_t now = get_seconds();
	int old_trans_id;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_transaction_handle myth;
	int sched_count = 0;
	int retval;

	reiserfs_check_lock_depth(p_s_sb, "journal_begin");
3014
	BUG_ON(nblocks > journal->j_trans_max);
3015 3016 3017 3018 3019 3020 3021 3022 3023 3024 3025 3026 3027 3028 3029 3030 3031 3032 3033 3034 3035 3036 3037 3038 3039 3040 3041 3042 3043 3044 3045 3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074 3075 3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087 3088 3089 3090 3091 3092 3093 3094 3095 3096 3097 3098 3099 3100 3101 3102 3103 3104 3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120 3121 3122 3123 3124 3125 3126 3127 3128 3129

	PROC_INFO_INC(p_s_sb, journal.journal_being);
	/* set here for journal_join */
	th->t_refcount = 1;
	th->t_super = p_s_sb;

      relock:
	lock_journal(p_s_sb);
	if (join != JBEGIN_ABORT && reiserfs_is_journal_aborted(journal)) {
		unlock_journal(p_s_sb);
		retval = journal->j_errno;
		goto out_fail;
	}
	journal->j_bcount++;

	if (test_bit(J_WRITERS_BLOCKED, &journal->j_state)) {
		unlock_journal(p_s_sb);
		reiserfs_wait_on_write_block(p_s_sb);
		PROC_INFO_INC(p_s_sb, journal.journal_relock_writers);
		goto relock;
	}
	now = get_seconds();

	/* if there is no room in the journal OR
	 ** if this transaction is too old, and we weren't called joinable, wait for it to finish before beginning 
	 ** we don't sleep if there aren't other writers
	 */

	if ((!join && journal->j_must_wait > 0) ||
	    (!join
	     && (journal->j_len_alloc + nblocks + 2) >= journal->j_max_batch)
	    || (!join && atomic_read(&journal->j_wcount) > 0
		&& journal->j_trans_start_time > 0
		&& (now - journal->j_trans_start_time) >
		journal->j_max_trans_age) || (!join
					      && atomic_read(&journal->j_jlock))
	    || (!join && journal->j_cnode_free < (journal->j_trans_max * 3))) {

		old_trans_id = journal->j_trans_id;
		unlock_journal(p_s_sb);	/* allow others to finish this transaction */

		if (!join && (journal->j_len_alloc + nblocks + 2) >=
		    journal->j_max_batch &&
		    ((journal->j_len + nblocks + 2) * 100) <
		    (journal->j_len_alloc * 75)) {
			if (atomic_read(&journal->j_wcount) > 10) {
				sched_count++;
				queue_log_writer(p_s_sb);
				goto relock;
			}
		}
		/* don't mess with joining the transaction if all we have to do is
		 * wait for someone else to do a commit
		 */
		if (atomic_read(&journal->j_jlock)) {
			while (journal->j_trans_id == old_trans_id &&
			       atomic_read(&journal->j_jlock)) {
				queue_log_writer(p_s_sb);
			}
			goto relock;
		}
		retval = journal_join(&myth, p_s_sb, 1);
		if (retval)
			goto out_fail;

		/* someone might have ended the transaction while we joined */
		if (old_trans_id != journal->j_trans_id) {
			retval = do_journal_end(&myth, p_s_sb, 1, 0);
		} else {
			retval = do_journal_end(&myth, p_s_sb, 1, COMMIT_NOW);
		}

		if (retval)
			goto out_fail;

		PROC_INFO_INC(p_s_sb, journal.journal_relock_wcount);
		goto relock;
	}
	/* we are the first writer, set trans_id */
	if (journal->j_trans_start_time == 0) {
		journal->j_trans_start_time = get_seconds();
	}
	atomic_inc(&(journal->j_wcount));
	journal->j_len_alloc += nblocks;
	th->t_blocks_logged = 0;
	th->t_blocks_allocated = nblocks;
	th->t_trans_id = journal->j_trans_id;
	unlock_journal(p_s_sb);
	INIT_LIST_HEAD(&th->t_list);
	get_fs_excl();
	return 0;

      out_fail:
	memset(th, 0, sizeof(*th));
	/* Re-set th->t_super, so we can properly keep track of how many
	 * persistent transactions there are. We need to do this so if this
	 * call is part of a failed restart_transaction, we can free it later */
	th->t_super = p_s_sb;
	return retval;
}

struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct
								    super_block
								    *s,
								    int nblocks)
{
	int ret;
	struct reiserfs_transaction_handle *th;

	/* if we're nesting into an existing transaction.  It will be
	 ** persistent on its own
	 */
	if (reiserfs_transaction_running(s)) {
		th = current->journal_info;
		th->t_refcount++;
3130 3131
		BUG_ON(th->t_refcount < 2);
		
3132 3133
		return th;
	}
3134
	th = kmalloc(sizeof(struct reiserfs_transaction_handle), GFP_NOFS);
3135 3136 3137 3138
	if (!th)
		return NULL;
	ret = journal_begin(th, s, nblocks);
	if (ret) {
3139
		kfree(th);
3140 3141 3142 3143 3144 3145 3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156
		return NULL;
	}

	SB_JOURNAL(s)->j_persistent_trans++;
	return th;
}

int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *th)
{
	struct super_block *s = th->t_super;
	int ret = 0;
	if (th->t_trans_id)
		ret = journal_end(th, th->t_super, th->t_blocks_allocated);
	else
		ret = -EIO;
	if (th->t_refcount == 0) {
		SB_JOURNAL(s)->j_persistent_trans--;
3157
		kfree(th);
3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170
	}
	return ret;
}

static int journal_join(struct reiserfs_transaction_handle *th,
			struct super_block *p_s_sb, unsigned long nblocks)
{
	struct reiserfs_transaction_handle *cur_th = current->journal_info;

	/* this keeps do_journal_end from NULLing out the current->journal_info
	 ** pointer
	 */
	th->t_handle_save = cur_th;
3171
	BUG_ON(cur_th && cur_th->t_refcount > 1);
3172 3173 3174 3175 3176 3177 3178 3179 3180 3181 3182 3183
	return do_journal_begin_r(th, p_s_sb, nblocks, JBEGIN_JOIN);
}

int journal_join_abort(struct reiserfs_transaction_handle *th,
		       struct super_block *p_s_sb, unsigned long nblocks)
{
	struct reiserfs_transaction_handle *cur_th = current->journal_info;

	/* this keeps do_journal_end from NULLing out the current->journal_info
	 ** pointer
	 */
	th->t_handle_save = cur_th;
3184
	BUG_ON(cur_th && cur_th->t_refcount > 1);
3185 3186 3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 3203 3204 3205 3206 3207 3208 3209 3210 3211 3212 3213 3214
	return do_journal_begin_r(th, p_s_sb, nblocks, JBEGIN_ABORT);
}

int journal_begin(struct reiserfs_transaction_handle *th,
		  struct super_block *p_s_sb, unsigned long nblocks)
{
	struct reiserfs_transaction_handle *cur_th = current->journal_info;
	int ret;

	th->t_handle_save = NULL;
	if (cur_th) {
		/* we are nesting into the current transaction */
		if (cur_th->t_super == p_s_sb) {
			BUG_ON(!cur_th->t_refcount);
			cur_th->t_refcount++;
			memcpy(th, cur_th, sizeof(*th));
			if (th->t_refcount <= 1)
				reiserfs_warning(p_s_sb,
						 "BAD: refcount <= 1, but journal_info != 0");
			return 0;
		} else {
			/* we've ended up with a handle from a different filesystem.
			 ** save it and restore on journal_end.  This should never
			 ** really happen...
			 */
			reiserfs_warning(p_s_sb,
					 "clm-2100: nesting info a different FS");
			th->t_handle_save = current->journal_info;
			current->journal_info = th;
		}
L
Linus Torvalds 已提交
3215
	} else {
3216 3217 3218
		current->journal_info = th;
	}
	ret = do_journal_begin_r(th, p_s_sb, nblocks, JBEGIN_REG);
3219
	BUG_ON(current->journal_info != th);
L
Linus Torvalds 已提交
3220

3221 3222 3223 3224 3225 3226 3227
	/* I guess this boils down to being the reciprocal of clm-2100 above.
	 * If do_journal_begin_r fails, we need to put it back, since journal_end
	 * won't be called to do it. */
	if (ret)
		current->journal_info = th->t_handle_save;
	else
		BUG_ON(!th->t_refcount);
L
Linus Torvalds 已提交
3228

3229
	return ret;
L
Linus Torvalds 已提交
3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240
}

/*
** puts bh into the current transaction.  If it was already there, reorders removes the
** old pointers from the hash, and puts new ones in (to make sure replay happen in the right order).
**
** if it was dirty, cleans and files onto the clean list.  I can't let it be dirty again until the
** transaction is committed.
** 
** if j_len, is bigger than j_len_alloc, it pushes j_len_alloc to 10 + j_len.
*/
3241 3242 3243 3244 3245 3246 3247 3248 3249 3250 3251 3252 3253 3254 3255 3256 3257 3258 3259 3260 3261 3262 3263 3264 3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363
int journal_mark_dirty(struct reiserfs_transaction_handle *th,
		       struct super_block *p_s_sb, struct buffer_head *bh)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_journal_cnode *cn = NULL;
	int count_already_incd = 0;
	int prepared = 0;
	BUG_ON(!th->t_trans_id);

	PROC_INFO_INC(p_s_sb, journal.mark_dirty);
	if (th->t_trans_id != journal->j_trans_id) {
		reiserfs_panic(th->t_super,
			       "journal-1577: handle trans id %ld != current trans id %ld\n",
			       th->t_trans_id, journal->j_trans_id);
	}

	p_s_sb->s_dirt = 1;

	prepared = test_clear_buffer_journal_prepared(bh);
	clear_buffer_journal_restore_dirty(bh);
	/* already in this transaction, we are done */
	if (buffer_journaled(bh)) {
		PROC_INFO_INC(p_s_sb, journal.mark_dirty_already);
		return 0;
	}

	/* this must be turned into a panic instead of a warning.  We can't allow
	 ** a dirty or journal_dirty or locked buffer to be logged, as some changes
	 ** could get to disk too early.  NOT GOOD.
	 */
	if (!prepared || buffer_dirty(bh)) {
		reiserfs_warning(p_s_sb, "journal-1777: buffer %llu bad state "
				 "%cPREPARED %cLOCKED %cDIRTY %cJDIRTY_WAIT",
				 (unsigned long long)bh->b_blocknr,
				 prepared ? ' ' : '!',
				 buffer_locked(bh) ? ' ' : '!',
				 buffer_dirty(bh) ? ' ' : '!',
				 buffer_journal_dirty(bh) ? ' ' : '!');
	}

	if (atomic_read(&(journal->j_wcount)) <= 0) {
		reiserfs_warning(p_s_sb,
				 "journal-1409: journal_mark_dirty returning because j_wcount was %d",
				 atomic_read(&(journal->j_wcount)));
		return 1;
	}
	/* this error means I've screwed up, and we've overflowed the transaction.  
	 ** Nothing can be done here, except make the FS readonly or panic.
	 */
	if (journal->j_len >= journal->j_trans_max) {
		reiserfs_panic(th->t_super,
			       "journal-1413: journal_mark_dirty: j_len (%lu) is too big\n",
			       journal->j_len);
	}

	if (buffer_journal_dirty(bh)) {
		count_already_incd = 1;
		PROC_INFO_INC(p_s_sb, journal.mark_dirty_notjournal);
		clear_buffer_journal_dirty(bh);
	}

	if (journal->j_len > journal->j_len_alloc) {
		journal->j_len_alloc = journal->j_len + JOURNAL_PER_BALANCE_CNT;
	}

	set_buffer_journaled(bh);

	/* now put this guy on the end */
	if (!cn) {
		cn = get_cnode(p_s_sb);
		if (!cn) {
			reiserfs_panic(p_s_sb, "get_cnode failed!\n");
		}

		if (th->t_blocks_logged == th->t_blocks_allocated) {
			th->t_blocks_allocated += JOURNAL_PER_BALANCE_CNT;
			journal->j_len_alloc += JOURNAL_PER_BALANCE_CNT;
		}
		th->t_blocks_logged++;
		journal->j_len++;

		cn->bh = bh;
		cn->blocknr = bh->b_blocknr;
		cn->sb = p_s_sb;
		cn->jlist = NULL;
		insert_journal_hash(journal->j_hash_table, cn);
		if (!count_already_incd) {
			get_bh(bh);
		}
	}
	cn->next = NULL;
	cn->prev = journal->j_last;
	cn->bh = bh;
	if (journal->j_last) {
		journal->j_last->next = cn;
		journal->j_last = cn;
	} else {
		journal->j_first = cn;
		journal->j_last = cn;
	}
	return 0;
}

int journal_end(struct reiserfs_transaction_handle *th,
		struct super_block *p_s_sb, unsigned long nblocks)
{
	if (!current->journal_info && th->t_refcount > 1)
		reiserfs_warning(p_s_sb, "REISER-NESTING: th NULL, refcount %d",
				 th->t_refcount);

	if (!th->t_trans_id) {
		WARN_ON(1);
		return -EIO;
	}

	th->t_refcount--;
	if (th->t_refcount > 0) {
		struct reiserfs_transaction_handle *cur_th =
		    current->journal_info;

		/* we aren't allowed to close a nested transaction on a different
		 ** filesystem from the one in the task struct
		 */
3364
		BUG_ON(cur_th->t_super != th->t_super);
3365 3366 3367 3368 3369 3370 3371 3372 3373

		if (th != cur_th) {
			memcpy(current->journal_info, th, sizeof(*th));
			th->t_trans_id = 0;
		}
		return 0;
	} else {
		return do_journal_end(th, p_s_sb, nblocks, 0);
	}
L
Linus Torvalds 已提交
3374 3375 3376 3377 3378 3379 3380 3381 3382
}

/* removes from the current transaction, relsing and descrementing any counters.  
** also files the removed buffer directly onto the clean list
**
** called by journal_mark_freed when a block has been deleted
**
** returns 1 if it cleaned and relsed the buffer. 0 otherwise
*/
3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405 3406 3407 3408 3409 3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422 3423 3424 3425 3426 3427
static int remove_from_transaction(struct super_block *p_s_sb,
				   b_blocknr_t blocknr, int already_cleaned)
{
	struct buffer_head *bh;
	struct reiserfs_journal_cnode *cn;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	int ret = 0;

	cn = get_journal_hash_dev(p_s_sb, journal->j_hash_table, blocknr);
	if (!cn || !cn->bh) {
		return ret;
	}
	bh = cn->bh;
	if (cn->prev) {
		cn->prev->next = cn->next;
	}
	if (cn->next) {
		cn->next->prev = cn->prev;
	}
	if (cn == journal->j_first) {
		journal->j_first = cn->next;
	}
	if (cn == journal->j_last) {
		journal->j_last = cn->prev;
	}
	if (bh)
		remove_journal_hash(p_s_sb, journal->j_hash_table, NULL,
				    bh->b_blocknr, 0);
	clear_buffer_journaled(bh);	/* don't log this one */

	if (!already_cleaned) {
		clear_buffer_journal_dirty(bh);
		clear_buffer_dirty(bh);
		clear_buffer_journal_test(bh);
		put_bh(bh);
		if (atomic_read(&(bh->b_count)) < 0) {
			reiserfs_warning(p_s_sb,
					 "journal-1752: remove from trans, b_count < 0");
		}
		ret = 1;
	}
	journal->j_len--;
	journal->j_len_alloc--;
	free_cnode(p_s_sb, cn);
	return ret;
L
Linus Torvalds 已提交
3428 3429 3430 3431
}

/*
** for any cnode in a journal list, it can only be dirtied of all the
M
Matt LaPlante 已提交
3432
** transactions that include it are committed to disk.
L
Linus Torvalds 已提交
3433 3434 3435 3436 3437 3438 3439
** this checks through each transaction, and returns 1 if you are allowed to dirty,
** and 0 if you aren't
**
** it is called by dirty_journal_list, which is called after flush_commit_list has gotten all the log
** blocks for a given transaction on disk
**
*/
3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468 3469 3470
static int can_dirty(struct reiserfs_journal_cnode *cn)
{
	struct super_block *sb = cn->sb;
	b_blocknr_t blocknr = cn->blocknr;
	struct reiserfs_journal_cnode *cur = cn->hprev;
	int can_dirty = 1;

	/* first test hprev.  These are all newer than cn, so any node here
	 ** with the same block number and dev means this node can't be sent
	 ** to disk right now.
	 */
	while (cur && can_dirty) {
		if (cur->jlist && cur->bh && cur->blocknr && cur->sb == sb &&
		    cur->blocknr == blocknr) {
			can_dirty = 0;
		}
		cur = cur->hprev;
	}
	/* then test hnext.  These are all older than cn.  As long as they
	 ** are committed to the log, it is safe to write cn to disk
	 */
	cur = cn->hnext;
	while (cur && can_dirty) {
		if (cur->jlist && cur->jlist->j_len > 0 &&
		    atomic_read(&(cur->jlist->j_commit_left)) > 0 && cur->bh &&
		    cur->blocknr && cur->sb == sb && cur->blocknr == blocknr) {
			can_dirty = 0;
		}
		cur = cur->hnext;
	}
	return can_dirty;
L
Linus Torvalds 已提交
3471 3472 3473
}

/* syncs the commit blocks, but does not force the real buffers to disk
M
Matt LaPlante 已提交
3474
** will wait until the current transaction is done/committed before returning 
L
Linus Torvalds 已提交
3475
*/
3476 3477 3478 3479
int journal_end_sync(struct reiserfs_transaction_handle *th,
		     struct super_block *p_s_sb, unsigned long nblocks)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
L
Linus Torvalds 已提交
3480

3481 3482
	BUG_ON(!th->t_trans_id);
	/* you can sync while nested, very, very bad */
3483
	BUG_ON(th->t_refcount > 1);
3484 3485 3486 3487 3488 3489
	if (journal->j_len == 0) {
		reiserfs_prepare_for_journal(p_s_sb, SB_BUFFER_WITH_SB(p_s_sb),
					     1);
		journal_mark_dirty(th, p_s_sb, SB_BUFFER_WITH_SB(p_s_sb));
	}
	return do_journal_end(th, p_s_sb, nblocks, COMMIT_NOW | WAIT);
L
Linus Torvalds 已提交
3490 3491 3492 3493 3494
}

/*
** writeback the pending async commits to disk
*/
D
David Howells 已提交
3495
static void flush_async_commits(struct work_struct *work)
3496
{
D
David Howells 已提交
3497 3498 3499
	struct reiserfs_journal *journal =
		container_of(work, struct reiserfs_journal, j_work.work);
	struct super_block *p_s_sb = journal->j_work_sb;
3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510
	struct reiserfs_journal_list *jl;
	struct list_head *entry;

	lock_kernel();
	if (!list_empty(&journal->j_journal_list)) {
		/* last entry is the youngest, commit it and you get everything */
		entry = journal->j_journal_list.prev;
		jl = JOURNAL_LIST_ENTRY(entry);
		flush_commit_list(p_s_sb, jl, 1);
	}
	unlock_kernel();
L
Linus Torvalds 已提交
3511 3512 3513 3514 3515 3516
}

/*
** flushes any old transactions to disk
** ends the current transaction if it is too old
*/
3517 3518 3519 3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551
int reiserfs_flush_old_commits(struct super_block *p_s_sb)
{
	time_t now;
	struct reiserfs_transaction_handle th;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);

	now = get_seconds();
	/* safety check so we don't flush while we are replaying the log during
	 * mount
	 */
	if (list_empty(&journal->j_journal_list)) {
		return 0;
	}

	/* check the current transaction.  If there are no writers, and it is
	 * too old, finish it, and force the commit blocks to disk
	 */
	if (atomic_read(&journal->j_wcount) <= 0 &&
	    journal->j_trans_start_time > 0 &&
	    journal->j_len > 0 &&
	    (now - journal->j_trans_start_time) > journal->j_max_trans_age) {
		if (!journal_join(&th, p_s_sb, 1)) {
			reiserfs_prepare_for_journal(p_s_sb,
						     SB_BUFFER_WITH_SB(p_s_sb),
						     1);
			journal_mark_dirty(&th, p_s_sb,
					   SB_BUFFER_WITH_SB(p_s_sb));

			/* we're only being called from kreiserfsd, it makes no sense to do
			 ** an async commit so that kreiserfsd can do it later
			 */
			do_journal_end(&th, p_s_sb, 1, COMMIT_NOW | WAIT);
		}
	}
	return p_s_sb->s_dirt;
L
Linus Torvalds 已提交
3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564
}

/*
** returns 0 if do_journal_end should return right away, returns 1 if do_journal_end should finish the commit
** 
** if the current transaction is too old, but still has writers, this will wait on j_join_wait until all 
** the writers are done.  By the time it wakes up, the transaction it was called has already ended, so it just
** flushes the commit list and returns 0.
**
** Won't batch when flush or commit_now is set.  Also won't batch when others are waiting on j_join_wait.
** 
** Note, we can't allow the journal_end to proceed while there are still writers in the log.
*/
3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580 3581 3582 3583 3584 3585 3586 3587 3588 3589 3590 3591 3592 3593
static int check_journal_end(struct reiserfs_transaction_handle *th,
			     struct super_block *p_s_sb, unsigned long nblocks,
			     int flags)
{

	time_t now;
	int flush = flags & FLUSH_ALL;
	int commit_now = flags & COMMIT_NOW;
	int wait_on_commit = flags & WAIT;
	struct reiserfs_journal_list *jl;
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);

	BUG_ON(!th->t_trans_id);

	if (th->t_trans_id != journal->j_trans_id) {
		reiserfs_panic(th->t_super,
			       "journal-1577: handle trans id %ld != current trans id %ld\n",
			       th->t_trans_id, journal->j_trans_id);
	}

	journal->j_len_alloc -= (th->t_blocks_allocated - th->t_blocks_logged);
	if (atomic_read(&(journal->j_wcount)) > 0) {	/* <= 0 is allowed.  unmounting might not call begin */
		atomic_dec(&(journal->j_wcount));
	}

	/* BUG, deal with case where j_len is 0, but people previously freed blocks need to be released 
	 ** will be dealt with by next transaction that actually writes something, but should be taken
	 ** care of in this trans
	 */
3594 3595
	BUG_ON(journal->j_len == 0);

3596 3597 3598 3599 3600 3601 3602 3603 3604 3605 3606 3607 3608 3609 3610 3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621 3622 3623 3624 3625 3626 3627 3628
	/* if wcount > 0, and we are called to with flush or commit_now,
	 ** we wait on j_join_wait.  We will wake up when the last writer has
	 ** finished the transaction, and started it on its way to the disk.
	 ** Then, we flush the commit or journal list, and just return 0 
	 ** because the rest of journal end was already done for this transaction.
	 */
	if (atomic_read(&(journal->j_wcount)) > 0) {
		if (flush || commit_now) {
			unsigned trans_id;

			jl = journal->j_current_jl;
			trans_id = jl->j_trans_id;
			if (wait_on_commit)
				jl->j_state |= LIST_COMMIT_PENDING;
			atomic_set(&(journal->j_jlock), 1);
			if (flush) {
				journal->j_next_full_flush = 1;
			}
			unlock_journal(p_s_sb);

			/* sleep while the current transaction is still j_jlocked */
			while (journal->j_trans_id == trans_id) {
				if (atomic_read(&journal->j_jlock)) {
					queue_log_writer(p_s_sb);
				} else {
					lock_journal(p_s_sb);
					if (journal->j_trans_id == trans_id) {
						atomic_set(&(journal->j_jlock),
							   1);
					}
					unlock_journal(p_s_sb);
				}
			}
3629 3630
			BUG_ON(journal->j_trans_id == trans_id);
			
3631 3632 3633 3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 3644 3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663 3664
			if (commit_now
			    && journal_list_still_alive(p_s_sb, trans_id)
			    && wait_on_commit) {
				flush_commit_list(p_s_sb, jl, 1);
			}
			return 0;
		}
		unlock_journal(p_s_sb);
		return 0;
	}

	/* deal with old transactions where we are the last writers */
	now = get_seconds();
	if ((now - journal->j_trans_start_time) > journal->j_max_trans_age) {
		commit_now = 1;
		journal->j_next_async_flush = 1;
	}
	/* don't batch when someone is waiting on j_join_wait */
	/* don't batch when syncing the commit or flushing the whole trans */
	if (!(journal->j_must_wait > 0) && !(atomic_read(&(journal->j_jlock)))
	    && !flush && !commit_now && (journal->j_len < journal->j_max_batch)
	    && journal->j_len_alloc < journal->j_max_batch
	    && journal->j_cnode_free > (journal->j_trans_max * 3)) {
		journal->j_bcount++;
		unlock_journal(p_s_sb);
		return 0;
	}

	if (journal->j_start > SB_ONDISK_JOURNAL_SIZE(p_s_sb)) {
		reiserfs_panic(p_s_sb,
			       "journal-003: journal_end: j_start (%ld) is too high\n",
			       journal->j_start);
	}
	return 1;
L
Linus Torvalds 已提交
3665 3666 3667 3668 3669 3670 3671 3672 3673 3674 3675 3676 3677 3678 3679 3680
}

/*
** Does all the work that makes deleting blocks safe.
** when deleting a block mark BH_JNew, just remove it from the current transaction, clean it's buffer_head and move on.
** 
** otherwise:
** set a bit for the block in the journal bitmap.  That will prevent it from being allocated for unformatted nodes
** before this transaction has finished.
**
** mark any cnodes for this block as BLOCK_FREED, and clear their bh pointers.  That will prevent any old transactions with
** this block from trying to flush to the real location.  Since we aren't removing the cnode from the journal_list_hash,
** the block can't be reallocated yet.
**
** Then remove it from the current transaction, decrementing any counters and filing it on the clean list.
*/
3681 3682 3683 3684 3685 3686 3687 3688 3689 3690 3691 3692 3693 3694 3695 3696 3697 3698 3699 3700 3701 3702 3703 3704 3705 3706 3707 3708 3709 3710 3711 3712 3713 3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754
int journal_mark_freed(struct reiserfs_transaction_handle *th,
		       struct super_block *p_s_sb, b_blocknr_t blocknr)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_journal_cnode *cn = NULL;
	struct buffer_head *bh = NULL;
	struct reiserfs_list_bitmap *jb = NULL;
	int cleaned = 0;
	BUG_ON(!th->t_trans_id);

	cn = get_journal_hash_dev(p_s_sb, journal->j_hash_table, blocknr);
	if (cn && cn->bh) {
		bh = cn->bh;
		get_bh(bh);
	}
	/* if it is journal new, we just remove it from this transaction */
	if (bh && buffer_journal_new(bh)) {
		clear_buffer_journal_new(bh);
		clear_prepared_bits(bh);
		reiserfs_clean_and_file_buffer(bh);
		cleaned = remove_from_transaction(p_s_sb, blocknr, cleaned);
	} else {
		/* set the bit for this block in the journal bitmap for this transaction */
		jb = journal->j_current_jl->j_list_bitmap;
		if (!jb) {
			reiserfs_panic(p_s_sb,
				       "journal-1702: journal_mark_freed, journal_list_bitmap is NULL\n");
		}
		set_bit_in_list_bitmap(p_s_sb, blocknr, jb);

		/* Note, the entire while loop is not allowed to schedule.  */

		if (bh) {
			clear_prepared_bits(bh);
			reiserfs_clean_and_file_buffer(bh);
		}
		cleaned = remove_from_transaction(p_s_sb, blocknr, cleaned);

		/* find all older transactions with this block, make sure they don't try to write it out */
		cn = get_journal_hash_dev(p_s_sb, journal->j_list_hash_table,
					  blocknr);
		while (cn) {
			if (p_s_sb == cn->sb && blocknr == cn->blocknr) {
				set_bit(BLOCK_FREED, &cn->state);
				if (cn->bh) {
					if (!cleaned) {
						/* remove_from_transaction will brelse the buffer if it was 
						 ** in the current trans
						 */
						clear_buffer_journal_dirty(cn->
									   bh);
						clear_buffer_dirty(cn->bh);
						clear_buffer_journal_test(cn->
									  bh);
						cleaned = 1;
						put_bh(cn->bh);
						if (atomic_read
						    (&(cn->bh->b_count)) < 0) {
							reiserfs_warning(p_s_sb,
									 "journal-2138: cn->bh->b_count < 0");
						}
					}
					if (cn->jlist) {	/* since we are clearing the bh, we MUST dec nonzerolen */
						atomic_dec(&
							   (cn->jlist->
							    j_nonzerolen));
					}
					cn->bh = NULL;
				}
			}
			cn = cn->hnext;
		}
	}

3755 3756
	if (bh)
		release_buffer_page(bh); /* get_hash grabs the buffer */
3757 3758 3759 3760 3761 3762 3763 3764
	return 0;
}

void reiserfs_update_inode_transaction(struct inode *inode)
{
	struct reiserfs_journal *journal = SB_JOURNAL(inode->i_sb);
	REISERFS_I(inode)->i_jl = journal->j_current_jl;
	REISERFS_I(inode)->i_trans_id = journal->j_trans_id;
L
Linus Torvalds 已提交
3765 3766 3767 3768 3769 3770 3771
}

/*
 * returns -1 on error, 0 if no commits/barriers were done and 1
 * if a transaction was actually committed and the barrier was done
 */
static int __commit_trans_jl(struct inode *inode, unsigned long id,
3772
			     struct reiserfs_journal_list *jl)
L
Linus Torvalds 已提交
3773
{
3774 3775 3776 3777 3778 3779 3780 3781 3782 3783 3784 3785 3786
	struct reiserfs_transaction_handle th;
	struct super_block *sb = inode->i_sb;
	struct reiserfs_journal *journal = SB_JOURNAL(sb);
	int ret = 0;

	/* is it from the current transaction, or from an unknown transaction? */
	if (id == journal->j_trans_id) {
		jl = journal->j_current_jl;
		/* try to let other writers come in and grow this transaction */
		let_transaction_grow(sb, id);
		if (journal->j_trans_id != id) {
			goto flush_commit_only;
		}
L
Linus Torvalds 已提交
3787

3788 3789 3790 3791 3792 3793 3794 3795 3796 3797 3798 3799
		ret = journal_begin(&th, sb, 1);
		if (ret)
			return ret;

		/* someone might have ended this transaction while we joined */
		if (journal->j_trans_id != id) {
			reiserfs_prepare_for_journal(sb, SB_BUFFER_WITH_SB(sb),
						     1);
			journal_mark_dirty(&th, sb, SB_BUFFER_WITH_SB(sb));
			ret = journal_end(&th, sb, 1);
			goto flush_commit_only;
		}
L
Linus Torvalds 已提交
3800

3801 3802 3803
		ret = journal_end_sync(&th, sb, 1);
		if (!ret)
			ret = 1;
L
Linus Torvalds 已提交
3804

3805 3806 3807 3808 3809 3810 3811 3812 3813 3814 3815 3816 3817 3818 3819 3820 3821 3822
	} else {
		/* this gets tricky, we have to make sure the journal list in
		 * the inode still exists.  We know the list is still around
		 * if we've got a larger transaction id than the oldest list
		 */
	      flush_commit_only:
		if (journal_list_still_alive(inode->i_sb, id)) {
			/*
			 * we only set ret to 1 when we know for sure
			 * the barrier hasn't been started yet on the commit
			 * block.
			 */
			if (atomic_read(&jl->j_commit_left) > 1)
				ret = 1;
			flush_commit_list(sb, jl, 1);
			if (journal->j_errno)
				ret = journal->j_errno;
		}
L
Linus Torvalds 已提交
3823
	}
3824 3825 3826
	/* otherwise the list is gone, and long since committed */
	return ret;
}
L
Linus Torvalds 已提交
3827

3828 3829 3830 3831
int reiserfs_commit_for_inode(struct inode *inode)
{
	unsigned long id = REISERFS_I(inode)->i_trans_id;
	struct reiserfs_journal_list *jl = REISERFS_I(inode)->i_jl;
L
Linus Torvalds 已提交
3832

3833 3834
	/* for the whole inode, assume unset id means it was
	 * changed in the current transaction.  More conservative
L
Linus Torvalds 已提交
3835
	 */
3836 3837 3838 3839 3840 3841 3842 3843 3844 3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864 3865 3866 3867
	if (!id || !jl) {
		reiserfs_update_inode_transaction(inode);
		id = REISERFS_I(inode)->i_trans_id;
		/* jl will be updated in __commit_trans_jl */
	}

	return __commit_trans_jl(inode, id, jl);
}

void reiserfs_restore_prepared_buffer(struct super_block *p_s_sb,
				      struct buffer_head *bh)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	PROC_INFO_INC(p_s_sb, journal.restore_prepared);
	if (!bh) {
		return;
	}
	if (test_clear_buffer_journal_restore_dirty(bh) &&
	    buffer_journal_dirty(bh)) {
		struct reiserfs_journal_cnode *cn;
		cn = get_journal_hash_dev(p_s_sb,
					  journal->j_list_hash_table,
					  bh->b_blocknr);
		if (cn && can_dirty(cn)) {
			set_buffer_journal_test(bh);
			mark_buffer_dirty(bh);
		}
	}
	clear_buffer_journal_prepared(bh);
}

extern struct tree_balance *cur_tb;
L
Linus Torvalds 已提交
3868 3869 3870 3871 3872 3873 3874 3875
/*
** before we can change a metadata block, we have to make sure it won't
** be written to disk while we are altering it.  So, we must:
** clean it
** wait on it.
** 
*/
int reiserfs_prepare_for_journal(struct super_block *p_s_sb,
3876 3877 3878 3879 3880 3881 3882 3883 3884 3885 3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904
				 struct buffer_head *bh, int wait)
{
	PROC_INFO_INC(p_s_sb, journal.prepare);

	if (test_set_buffer_locked(bh)) {
		if (!wait)
			return 0;
		lock_buffer(bh);
	}
	set_buffer_journal_prepared(bh);
	if (test_clear_buffer_dirty(bh) && buffer_journal_dirty(bh)) {
		clear_buffer_journal_test(bh);
		set_buffer_journal_restore_dirty(bh);
	}
	unlock_buffer(bh);
	return 1;
}

static void flush_old_journal_lists(struct super_block *s)
{
	struct reiserfs_journal *journal = SB_JOURNAL(s);
	struct reiserfs_journal_list *jl;
	struct list_head *entry;
	time_t now = get_seconds();

	while (!list_empty(&journal->j_journal_list)) {
		entry = journal->j_journal_list.next;
		jl = JOURNAL_LIST_ENTRY(entry);
		/* this check should always be run, to send old lists to disk */
3905 3906 3907
		if (jl->j_timestamp < (now - (JOURNAL_MAX_TRANS_AGE * 4)) &&
		    atomic_read(&jl->j_commit_left) == 0 &&
		    test_transaction(s, jl)) {
3908 3909 3910 3911
			flush_used_journal_lists(s, jl);
		} else {
			break;
		}
L
Linus Torvalds 已提交
3912 3913 3914 3915 3916 3917 3918 3919 3920 3921 3922 3923 3924
	}
}

/* 
** long and ugly.  If flush, will not return until all commit
** blocks and all real buffers in the trans are on disk.
** If no_async, won't return until all commit blocks are on disk.
**
** keep reading, there are comments as you go along
**
** If the journal is aborted, we just clean up. Things like flushing
** journal lists, etc just won't happen.
*/
3925 3926 3927 3928 3929 3930 3931 3932 3933 3934 3935 3936 3937 3938
static int do_journal_end(struct reiserfs_transaction_handle *th,
			  struct super_block *p_s_sb, unsigned long nblocks,
			  int flags)
{
	struct reiserfs_journal *journal = SB_JOURNAL(p_s_sb);
	struct reiserfs_journal_cnode *cn, *next, *jl_cn;
	struct reiserfs_journal_cnode *last_cn = NULL;
	struct reiserfs_journal_desc *desc;
	struct reiserfs_journal_commit *commit;
	struct buffer_head *c_bh;	/* commit bh */
	struct buffer_head *d_bh;	/* desc bh */
	int cur_write_start = 0;	/* start index of current log write */
	int old_start;
	int i;
3939 3940
	int flush;
	int wait_on_commit;
3941 3942 3943 3944 3945 3946 3947 3948 3949
	struct reiserfs_journal_list *jl, *temp_jl;
	struct list_head *entry, *safe;
	unsigned long jindex;
	unsigned long commit_trans_id;
	int trans_half;

	BUG_ON(th->t_refcount > 1);
	BUG_ON(!th->t_trans_id);

3950 3951 3952 3953 3954 3955 3956
	/* protect flush_older_commits from doing mistakes if the
           transaction ID counter gets overflowed.  */
	if (th->t_trans_id == ~0UL)
		flags |= FLUSH_ALL | COMMIT_NOW | WAIT;
	flush = flags & FLUSH_ALL;
	wait_on_commit = flags & WAIT;

3957 3958 3959 3960 3961 3962 3963 3964
	put_fs_excl();
	current->journal_info = th->t_handle_save;
	reiserfs_check_lock_depth(p_s_sb, "journal end");
	if (journal->j_len == 0) {
		reiserfs_prepare_for_journal(p_s_sb, SB_BUFFER_WITH_SB(p_s_sb),
					     1);
		journal_mark_dirty(th, p_s_sb, SB_BUFFER_WITH_SB(p_s_sb));
	}
L
Linus Torvalds 已提交
3965

3966 3967 3968 3969 3970 3971 3972 3973 3974 3975 3976 3977 3978 3979 3980 3981 3982 3983 3984 3985 3986 3987 3988 3989 3990 3991 3992 3993 3994 3995 3996 3997
	lock_journal(p_s_sb);
	if (journal->j_next_full_flush) {
		flags |= FLUSH_ALL;
		flush = 1;
	}
	if (journal->j_next_async_flush) {
		flags |= COMMIT_NOW | WAIT;
		wait_on_commit = 1;
	}

	/* check_journal_end locks the journal, and unlocks if it does not return 1 
	 ** it tells us if we should continue with the journal_end, or just return
	 */
	if (!check_journal_end(th, p_s_sb, nblocks, flags)) {
		p_s_sb->s_dirt = 1;
		wake_queued_writers(p_s_sb);
		reiserfs_async_progress_wait(p_s_sb);
		goto out;
	}

	/* check_journal_end might set these, check again */
	if (journal->j_next_full_flush) {
		flush = 1;
	}

	/*
	 ** j must wait means we have to flush the log blocks, and the real blocks for
	 ** this transaction
	 */
	if (journal->j_must_wait > 0) {
		flush = 1;
	}
L
Linus Torvalds 已提交
3998
#ifdef REISERFS_PREALLOCATE
3999 4000
	/* quota ops might need to nest, setup the journal_info pointer for them
	 * and raise the refcount so that it is > 0. */
4001
	current->journal_info = th;
4002
	th->t_refcount++;
4003 4004
	reiserfs_discard_all_prealloc(th);	/* it should not involve new blocks into
						 * the transaction */
4005
	th->t_refcount--;
4006
	current->journal_info = th->t_handle_save;
L
Linus Torvalds 已提交
4007
#endif
4008 4009 4010 4011 4012 4013 4014 4015 4016 4017 4018 4019 4020 4021 4022 4023 4024 4025 4026 4027 4028 4029 4030 4031 4032 4033 4034 4035 4036 4037 4038 4039 4040 4041 4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064 4065 4066 4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081 4082 4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104

	/* setup description block */
	d_bh =
	    journal_getblk(p_s_sb,
			   SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
			   journal->j_start);
	set_buffer_uptodate(d_bh);
	desc = (struct reiserfs_journal_desc *)(d_bh)->b_data;
	memset(d_bh->b_data, 0, d_bh->b_size);
	memcpy(get_journal_desc_magic(d_bh), JOURNAL_DESC_MAGIC, 8);
	set_desc_trans_id(desc, journal->j_trans_id);

	/* setup commit block.  Don't write (keep it clean too) this one until after everyone else is written */
	c_bh = journal_getblk(p_s_sb, SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
			      ((journal->j_start + journal->j_len +
				1) % SB_ONDISK_JOURNAL_SIZE(p_s_sb)));
	commit = (struct reiserfs_journal_commit *)c_bh->b_data;
	memset(c_bh->b_data, 0, c_bh->b_size);
	set_commit_trans_id(commit, journal->j_trans_id);
	set_buffer_uptodate(c_bh);

	/* init this journal list */
	jl = journal->j_current_jl;

	/* we lock the commit before doing anything because
	 * we want to make sure nobody tries to run flush_commit_list until
	 * the new transaction is fully setup, and we've already flushed the
	 * ordered bh list
	 */
	down(&jl->j_commit_lock);

	/* save the transaction id in case we need to commit it later */
	commit_trans_id = jl->j_trans_id;

	atomic_set(&jl->j_older_commits_done, 0);
	jl->j_trans_id = journal->j_trans_id;
	jl->j_timestamp = journal->j_trans_start_time;
	jl->j_commit_bh = c_bh;
	jl->j_start = journal->j_start;
	jl->j_len = journal->j_len;
	atomic_set(&jl->j_nonzerolen, journal->j_len);
	atomic_set(&jl->j_commit_left, journal->j_len + 2);
	jl->j_realblock = NULL;

	/* The ENTIRE FOR LOOP MUST not cause schedule to occur.
	 **  for each real block, add it to the journal list hash,
	 ** copy into real block index array in the commit or desc block
	 */
	trans_half = journal_trans_half(p_s_sb->s_blocksize);
	for (i = 0, cn = journal->j_first; cn; cn = cn->next, i++) {
		if (buffer_journaled(cn->bh)) {
			jl_cn = get_cnode(p_s_sb);
			if (!jl_cn) {
				reiserfs_panic(p_s_sb,
					       "journal-1676, get_cnode returned NULL\n");
			}
			if (i == 0) {
				jl->j_realblock = jl_cn;
			}
			jl_cn->prev = last_cn;
			jl_cn->next = NULL;
			if (last_cn) {
				last_cn->next = jl_cn;
			}
			last_cn = jl_cn;
			/* make sure the block we are trying to log is not a block 
			   of journal or reserved area */

			if (is_block_in_log_or_reserved_area
			    (p_s_sb, cn->bh->b_blocknr)) {
				reiserfs_panic(p_s_sb,
					       "journal-2332: Trying to log block %lu, which is a log block\n",
					       cn->bh->b_blocknr);
			}
			jl_cn->blocknr = cn->bh->b_blocknr;
			jl_cn->state = 0;
			jl_cn->sb = p_s_sb;
			jl_cn->bh = cn->bh;
			jl_cn->jlist = jl;
			insert_journal_hash(journal->j_list_hash_table, jl_cn);
			if (i < trans_half) {
				desc->j_realblock[i] =
				    cpu_to_le32(cn->bh->b_blocknr);
			} else {
				commit->j_realblock[i - trans_half] =
				    cpu_to_le32(cn->bh->b_blocknr);
			}
		} else {
			i--;
		}
	}
	set_desc_trans_len(desc, journal->j_len);
	set_desc_mount_id(desc, journal->j_mount_id);
	set_desc_trans_id(desc, journal->j_trans_id);
	set_commit_trans_len(commit, journal->j_len);

	/* special check in case all buffers in the journal were marked for not logging */
4105
	BUG_ON(journal->j_len == 0);
4106 4107 4108 4109 4110 4111 4112 4113 4114 4115 4116 4117 4118 4119 4120 4121 4122 4123 4124 4125 4126 4127 4128 4129 4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144 4145 4146 4147 4148 4149 4150 4151 4152 4153 4154 4155 4156 4157 4158 4159 4160 4161 4162 4163 4164 4165 4166 4167 4168 4169 4170 4171 4172 4173 4174 4175

	/* we're about to dirty all the log blocks, mark the description block
	 * dirty now too.  Don't mark the commit block dirty until all the
	 * others are on disk
	 */
	mark_buffer_dirty(d_bh);

	/* first data block is j_start + 1, so add one to cur_write_start wherever you use it */
	cur_write_start = journal->j_start;
	cn = journal->j_first;
	jindex = 1;		/* start at one so we don't get the desc again */
	while (cn) {
		clear_buffer_journal_new(cn->bh);
		/* copy all the real blocks into log area.  dirty log blocks */
		if (buffer_journaled(cn->bh)) {
			struct buffer_head *tmp_bh;
			char *addr;
			struct page *page;
			tmp_bh =
			    journal_getblk(p_s_sb,
					   SB_ONDISK_JOURNAL_1st_BLOCK(p_s_sb) +
					   ((cur_write_start +
					     jindex) %
					    SB_ONDISK_JOURNAL_SIZE(p_s_sb)));
			set_buffer_uptodate(tmp_bh);
			page = cn->bh->b_page;
			addr = kmap(page);
			memcpy(tmp_bh->b_data,
			       addr + offset_in_page(cn->bh->b_data),
			       cn->bh->b_size);
			kunmap(page);
			mark_buffer_dirty(tmp_bh);
			jindex++;
			set_buffer_journal_dirty(cn->bh);
			clear_buffer_journaled(cn->bh);
		} else {
			/* JDirty cleared sometime during transaction.  don't log this one */
			reiserfs_warning(p_s_sb,
					 "journal-2048: do_journal_end: BAD, buffer in journal hash, but not JDirty!");
			brelse(cn->bh);
		}
		next = cn->next;
		free_cnode(p_s_sb, cn);
		cn = next;
		cond_resched();
	}

	/* we are done  with both the c_bh and d_bh, but
	 ** c_bh must be written after all other commit blocks,
	 ** so we dirty/relse c_bh in flush_commit_list, with commit_left <= 1.
	 */

	journal->j_current_jl = alloc_journal_list(p_s_sb);

	/* now it is safe to insert this transaction on the main list */
	list_add_tail(&jl->j_list, &journal->j_journal_list);
	list_add_tail(&jl->j_working_list, &journal->j_working_list);
	journal->j_num_work_lists++;

	/* reset journal values for the next transaction */
	old_start = journal->j_start;
	journal->j_start =
	    (journal->j_start + journal->j_len +
	     2) % SB_ONDISK_JOURNAL_SIZE(p_s_sb);
	atomic_set(&(journal->j_wcount), 0);
	journal->j_bcount = 0;
	journal->j_last = NULL;
	journal->j_first = NULL;
	journal->j_len = 0;
	journal->j_trans_start_time = 0;
4176 4177 4178
	/* check for trans_id overflow */
	if (++journal->j_trans_id == 0)
		journal->j_trans_id = 10;
4179 4180 4181 4182 4183 4184 4185 4186 4187 4188 4189 4190 4191 4192 4193 4194 4195 4196 4197 4198 4199 4200 4201
	journal->j_current_jl->j_trans_id = journal->j_trans_id;
	journal->j_must_wait = 0;
	journal->j_len_alloc = 0;
	journal->j_next_full_flush = 0;
	journal->j_next_async_flush = 0;
	init_journal_hash(p_s_sb);

	// make sure reiserfs_add_jh sees the new current_jl before we
	// write out the tails
	smp_mb();

	/* tail conversion targets have to hit the disk before we end the
	 * transaction.  Otherwise a later transaction might repack the tail
	 * before this transaction commits, leaving the data block unflushed and
	 * clean, if we crash before the later transaction commits, the data block
	 * is lost.
	 */
	if (!list_empty(&jl->j_tail_bh_list)) {
		unlock_kernel();
		write_ordered_buffers(&journal->j_dirty_buffers_lock,
				      journal, jl, &jl->j_tail_bh_list);
		lock_kernel();
	}
4202
	BUG_ON(!list_empty(&jl->j_tail_bh_list));
4203 4204 4205 4206 4207 4208 4209 4210 4211 4212 4213 4214 4215 4216 4217 4218 4219 4220 4221 4222 4223 4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234 4235 4236 4237 4238 4239 4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255 4256 4257 4258 4259 4260 4261 4262 4263 4264 4265 4266 4267 4268 4269 4270 4271 4272 4273 4274 4275 4276 4277 4278 4279 4280 4281 4282 4283 4284 4285 4286 4287 4288 4289 4290 4291 4292 4293 4294 4295 4296 4297 4298
	up(&jl->j_commit_lock);

	/* honor the flush wishes from the caller, simple commits can
	 ** be done outside the journal lock, they are done below
	 **
	 ** if we don't flush the commit list right now, we put it into
	 ** the work queue so the people waiting on the async progress work
	 ** queue don't wait for this proc to flush journal lists and such.
	 */
	if (flush) {
		flush_commit_list(p_s_sb, jl, 1);
		flush_journal_list(p_s_sb, jl, 1);
	} else if (!(jl->j_state & LIST_COMMIT_PENDING))
		queue_delayed_work(commit_wq, &journal->j_work, HZ / 10);

	/* if the next transaction has any chance of wrapping, flush 
	 ** transactions that might get overwritten.  If any journal lists are very 
	 ** old flush them as well.  
	 */
      first_jl:
	list_for_each_safe(entry, safe, &journal->j_journal_list) {
		temp_jl = JOURNAL_LIST_ENTRY(entry);
		if (journal->j_start <= temp_jl->j_start) {
			if ((journal->j_start + journal->j_trans_max + 1) >=
			    temp_jl->j_start) {
				flush_used_journal_lists(p_s_sb, temp_jl);
				goto first_jl;
			} else if ((journal->j_start +
				    journal->j_trans_max + 1) <
				   SB_ONDISK_JOURNAL_SIZE(p_s_sb)) {
				/* if we don't cross into the next transaction and we don't
				 * wrap, there is no way we can overlap any later transactions
				 * break now
				 */
				break;
			}
		} else if ((journal->j_start +
			    journal->j_trans_max + 1) >
			   SB_ONDISK_JOURNAL_SIZE(p_s_sb)) {
			if (((journal->j_start + journal->j_trans_max + 1) %
			     SB_ONDISK_JOURNAL_SIZE(p_s_sb)) >=
			    temp_jl->j_start) {
				flush_used_journal_lists(p_s_sb, temp_jl);
				goto first_jl;
			} else {
				/* we don't overlap anything from out start to the end of the
				 * log, and our wrapped portion doesn't overlap anything at
				 * the start of the log.  We can break
				 */
				break;
			}
		}
	}
	flush_old_journal_lists(p_s_sb);

	journal->j_current_jl->j_list_bitmap =
	    get_list_bitmap(p_s_sb, journal->j_current_jl);

	if (!(journal->j_current_jl->j_list_bitmap)) {
		reiserfs_panic(p_s_sb,
			       "journal-1996: do_journal_end, could not get a list bitmap\n");
	}

	atomic_set(&(journal->j_jlock), 0);
	unlock_journal(p_s_sb);
	/* wake up any body waiting to join. */
	clear_bit(J_WRITERS_QUEUED, &journal->j_state);
	wake_up(&(journal->j_join_wait));

	if (!flush && wait_on_commit &&
	    journal_list_still_alive(p_s_sb, commit_trans_id)) {
		flush_commit_list(p_s_sb, jl, 1);
	}
      out:
	reiserfs_check_lock_depth(p_s_sb, "journal end2");

	memset(th, 0, sizeof(*th));
	/* Re-set th->t_super, so we can properly keep track of how many
	 * persistent transactions there are. We need to do this so if this
	 * call is part of a failed restart_transaction, we can free it later */
	th->t_super = p_s_sb;

	return journal->j_errno;
}

static void __reiserfs_journal_abort_hard(struct super_block *sb)
{
	struct reiserfs_journal *journal = SB_JOURNAL(sb);
	if (test_bit(J_ABORTED, &journal->j_state))
		return;

	printk(KERN_CRIT "REISERFS: Aborting journal for filesystem on %s\n",
	       reiserfs_bdevname(sb));

	sb->s_flags |= MS_RDONLY;
	set_bit(J_ABORTED, &journal->j_state);
L
Linus Torvalds 已提交
4299 4300

#ifdef CONFIG_REISERFS_CHECK
4301
	dump_stack();
L
Linus Torvalds 已提交
4302 4303 4304
#endif
}

4305
static void __reiserfs_journal_abort_soft(struct super_block *sb, int errno)
L
Linus Torvalds 已提交
4306
{
4307 4308 4309
	struct reiserfs_journal *journal = SB_JOURNAL(sb);
	if (test_bit(J_ABORTED, &journal->j_state))
		return;
L
Linus Torvalds 已提交
4310

4311 4312
	if (!journal->j_errno)
		journal->j_errno = errno;
L
Linus Torvalds 已提交
4313

4314
	__reiserfs_journal_abort_hard(sb);
L
Linus Torvalds 已提交
4315 4316
}

4317
void reiserfs_journal_abort(struct super_block *sb, int errno)
L
Linus Torvalds 已提交
4318
{
4319
	return __reiserfs_journal_abort_soft(sb, errno);
L
Linus Torvalds 已提交
4320
}