kernel-jbd.h revision 19007e51a548d1e0eff01025dde9f147a0beb26d
1/*
2 * linux/include/linux/jbd.h
3 *
4 * Written by Stephen C. Tweedie <sct@redhat.com>
5 *
6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
7 *
8 * This file is part of the Linux kernel and is made available under
9 * the terms of the GNU General Public License, version 2, or at your
10 * option, any later version, incorporated herein by reference.
11 *
12 * Definitions for transaction data structures for the buffer cache
13 * filesystem journaling support.
14 */
15
16#ifndef _LINUX_JBD_H
17#define _LINUX_JBD_H
18
19#if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE) || !defined(__KERNEL__)
20
21/* Allow this file to be included directly into e2fsprogs */
22#ifndef __KERNEL__
23#include "jfs_compat.h"
24#define JFS_DEBUG
25#define jfs_debug jbd_debug
26#else
27
28#include <linux/journal-head.h>
29#include <linux/stddef.h>
30#include <asm/semaphore.h>
31#endif
32
33#define journal_oom_retry 1
34
35#ifdef CONFIG_JBD_DEBUG
36/*
37 * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
38 * consistency checks.  By default we don't do this unless
39 * CONFIG_JBD_DEBUG is on.
40 */
41#define JBD_EXPENSIVE_CHECKING
42extern int journal_enable_debug;
43
44#define jbd_debug(n, f, a...)						\
45	do {								\
46		if ((n) <= journal_enable_debug) {			\
47			printk (KERN_DEBUG "(%s, %d): %s: ",		\
48				__FILE__, __LINE__, __FUNCTION__);	\
49		  	printk (f, ## a);				\
50		}							\
51	} while (0)
52#else
53#define jbd_debug(f, a...)	/**/
54#endif
55
56extern void * __jbd_kmalloc (char *where, size_t size, int flags, int retry);
57#define jbd_kmalloc(size, flags) \
58	__jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
59#define jbd_rep_kmalloc(size, flags) \
60	__jbd_kmalloc(__FUNCTION__, (size), (flags), 1)
61
62#define JFS_MIN_JOURNAL_BLOCKS 1024
63
64#ifdef __KERNEL__
65typedef struct handle_s		handle_t;	/* Atomic operation type */
66typedef struct journal_s	journal_t;	/* Journal control structure */
67#endif
68
69/*
70 * Internal structures used by the logging mechanism:
71 */
72
73#define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
74
75/*
76 * On-disk structures
77 */
78
79/*
80 * Descriptor block types:
81 */
82
83#define JFS_DESCRIPTOR_BLOCK	1
84#define JFS_COMMIT_BLOCK	2
85#define JFS_SUPERBLOCK_V1	3
86#define JFS_SUPERBLOCK_V2	4
87#define JFS_REVOKE_BLOCK	5
88
89/*
90 * Standard header for all descriptor blocks:
91 */
92typedef struct journal_header_s
93{
94	__u32		h_magic;
95	__u32		h_blocktype;
96	__u32		h_sequence;
97} journal_header_t;
98
99
100/*
101 * The block tag: used to describe a single buffer in the journal
102 */
103typedef struct journal_block_tag_s
104{
105	__u32		t_blocknr;	/* The on-disk block number */
106	__u32		t_flags;	/* See below */
107} journal_block_tag_t;
108
109/*
110 * The revoke descriptor: used on disk to describe a series of blocks to
111 * be revoked from the log
112 */
113typedef struct journal_revoke_header_s
114{
115	journal_header_t r_header;
116	int		 r_count;	/* Count of bytes used in the block */
117} journal_revoke_header_t;
118
119
120/* Definitions for the journal tag flags word: */
121#define JFS_FLAG_ESCAPE		1	/* on-disk block is escaped */
122#define JFS_FLAG_SAME_UUID	2	/* block has same uuid as previous */
123#define JFS_FLAG_DELETED	4	/* block deleted by this transaction */
124#define JFS_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
125
126
127/*
128 * The journal superblock.  All fields are in big-endian byte order.
129 */
130typedef struct journal_superblock_s
131{
132/* 0x0000 */
133	journal_header_t s_header;
134
135/* 0x000C */
136	/* Static information describing the journal */
137	__u32	s_blocksize;		/* journal device blocksize */
138	__u32	s_maxlen;		/* total blocks in journal file */
139	__u32	s_first;		/* first block of log information */
140
141/* 0x0018 */
142	/* Dynamic information describing the current state of the log */
143	__u32	s_sequence;		/* first commit ID expected in log */
144	__u32	s_start;		/* blocknr of start of log */
145
146/* 0x0020 */
147	/* Error value, as set by journal_abort(). */
148	__s32	s_errno;
149
150/* 0x0024 */
151	/* Remaining fields are only valid in a version-2 superblock */
152	__u32	s_feature_compat; 	/* compatible feature set */
153	__u32	s_feature_incompat; 	/* incompatible feature set */
154	__u32	s_feature_ro_compat; 	/* readonly-compatible feature set */
155/* 0x0030 */
156	__u8	s_uuid[16];		/* 128-bit uuid for journal */
157
158/* 0x0040 */
159	__u32	s_nr_users;		/* Nr of filesystems sharing log */
160
161	__u32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
162
163/* 0x0048 */
164	__u32	s_max_transaction;	/* Limit of journal blocks per trans.*/
165	__u32	s_max_trans_data;	/* Limit of data blocks per trans. */
166
167/* 0x0050 */
168	__u32	s_padding[44];
169
170/* 0x0100 */
171	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
172/* 0x0400 */
173} journal_superblock_t;
174
175#define JFS_HAS_COMPAT_FEATURE(j,mask)					\
176	((j)->j_format_version >= 2 &&					\
177	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
178#define JFS_HAS_RO_COMPAT_FEATURE(j,mask)				\
179	((j)->j_format_version >= 2 &&					\
180	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
181#define JFS_HAS_INCOMPAT_FEATURE(j,mask)				\
182	((j)->j_format_version >= 2 &&					\
183	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
184
185#define JFS_FEATURE_INCOMPAT_REVOKE	0x00000001
186
187/* Features known to this kernel version: */
188#define JFS_KNOWN_COMPAT_FEATURES	0
189#define JFS_KNOWN_ROCOMPAT_FEATURES	0
190#define JFS_KNOWN_INCOMPAT_FEATURES	JFS_FEATURE_INCOMPAT_REVOKE
191
192#ifdef __KERNEL__
193
194#include <linux/fs.h>
195#include <linux/sched.h>
196
197#define JBD_ASSERTIONS
198#ifdef JBD_ASSERTIONS
199#define J_ASSERT(assert)						\
200do {									\
201	if (!(assert)) {						\
202		printk (KERN_EMERG					\
203			"Assertion failure in %s() at %s:%d: \"%s\"\n",	\
204			__FUNCTION__, __FILE__, __LINE__, # assert);	\
205		BUG();							\
206	}								\
207} while (0)
208
209#if defined(CONFIG_BUFFER_DEBUG)
210void buffer_assertion_failure(struct buffer_head *bh);
211#define J_ASSERT_BH(bh, expr)						\
212	do {								\
213		if (!(expr))						\
214			buffer_assertion_failure(bh);			\
215		J_ASSERT(expr);						\
216	} while (0)
217#define J_ASSERT_JH(jh, expr)	J_ASSERT_BH(jh2bh(jh), expr)
218#else
219#define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
220#define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
221#endif
222
223#else
224#define J_ASSERT(assert)
225#endif		/* JBD_ASSERTIONS */
226
227enum jbd_state_bits {
228	BH_JWrite
229	  = BH_PrivateStart,	/* 1 if being written to log (@@@ DEBUGGING) */
230	BH_Freed,		/* 1 if buffer has been freed (truncated) */
231	BH_Revoked,		/* 1 if buffer has been revoked from the log */
232	BH_RevokeValid,		/* 1 if buffer revoked flag is valid */
233	BH_JBDDirty,		/* 1 if buffer is dirty but journaled */
234};
235
236/* Return true if the buffer is one which JBD is managing */
237static inline int buffer_jbd(struct buffer_head *bh)
238{
239	return __buffer_state(bh, JBD);
240}
241
242static inline struct buffer_head *jh2bh(struct journal_head *jh)
243{
244	return jh->b_bh;
245}
246
247static inline struct journal_head *bh2jh(struct buffer_head *bh)
248{
249	return bh->b_private;
250}
251
252struct jbd_revoke_table_s;
253
254/* The handle_t type represents a single atomic update being performed
255 * by some process.  All filesystem modifications made by the process go
256 * through this handle.  Recursive operations (such as quota operations)
257 * are gathered into a single update.
258 *
259 * The buffer credits field is used to account for journaled buffers
260 * being modified by the running process.  To ensure that there is
261 * enough log space for all outstanding operations, we need to limit the
262 * number of outstanding buffers possible at any time.  When the
263 * operation completes, any buffer credits not used are credited back to
264 * the transaction, so that at all times we know how many buffers the
265 * outstanding updates on a transaction might possibly touch. */
266
267struct handle_s
268{
269	/* Which compound transaction is this update a part of? */
270	transaction_t	      * h_transaction;
271
272	/* Number of remaining buffers we are allowed to dirty: */
273	int			h_buffer_credits;
274
275	/* Reference count on this handle */
276	int			h_ref;
277
278	/* Field for caller's use to track errors through large fs
279	   operations */
280	int			h_err;
281
282	/* Flags */
283	unsigned int	h_sync:		1;	/* sync-on-close */
284	unsigned int	h_jdata:	1;	/* force data journaling */
285	unsigned int	h_aborted:	1;	/* fatal error on handle */
286};
287
288
289/* The transaction_t type is the guts of the journaling mechanism.  It
290 * tracks a compound transaction through its various states:
291 *
292 * RUNNING:	accepting new updates
293 * LOCKED:	Updates still running but we don't accept new ones
294 * RUNDOWN:	Updates are tidying up but have finished requesting
295 *		new buffers to modify (state not used for now)
296 * FLUSH:       All updates complete, but we are still writing to disk
297 * COMMIT:      All data on disk, writing commit record
298 * FINISHED:	We still have to keep the transaction for checkpointing.
299 *
300 * The transaction keeps track of all of the buffers modified by a
301 * running transaction, and all of the buffers committed but not yet
302 * flushed to home for finished transactions.
303 */
304
305struct transaction_s
306{
307	/* Pointer to the journal for this transaction. */
308	journal_t *		t_journal;
309
310	/* Sequence number for this transaction */
311	tid_t			t_tid;
312
313	/* Transaction's current state */
314	enum {
315		T_RUNNING,
316		T_LOCKED,
317		T_RUNDOWN,
318		T_FLUSH,
319		T_COMMIT,
320		T_FINISHED
321	}			t_state;
322
323	/* Where in the log does this transaction's commit start? */
324	unsigned long		t_log_start;
325
326	/* Doubly-linked circular list of all inodes owned by this
327           transaction */	/* AKPM: unused */
328	struct inode *		t_ilist;
329
330	/* Number of buffers on the t_buffers list */
331	int			t_nr_buffers;
332
333	/* Doubly-linked circular list of all buffers reserved but not
334           yet modified by this transaction */
335	struct journal_head *	t_reserved_list;
336
337	/* Doubly-linked circular list of all metadata buffers owned by this
338           transaction */
339	struct journal_head *	t_buffers;
340
341	/*
342	 * Doubly-linked circular list of all data buffers still to be
343	 * flushed before this transaction can be committed.
344	 * Protected by journal_datalist_lock.
345	 */
346	struct journal_head *	t_sync_datalist;
347
348	/*
349	 * Doubly-linked circular list of all writepage data buffers
350	 * still to be written before this transaction can be committed.
351	 * Protected by journal_datalist_lock.
352	 */
353	struct journal_head *	t_async_datalist;
354
355	/* Doubly-linked circular list of all forget buffers (superceded
356           buffers which we can un-checkpoint once this transaction
357           commits) */
358	struct journal_head *	t_forget;
359
360	/*
361	 * Doubly-linked circular list of all buffers still to be
362	 * flushed before this transaction can be checkpointed.
363	 */
364	/* Protected by journal_datalist_lock */
365	struct journal_head *	t_checkpoint_list;
366
367	/* Doubly-linked circular list of temporary buffers currently
368           undergoing IO in the log */
369	struct journal_head *	t_iobuf_list;
370
371	/* Doubly-linked circular list of metadata buffers being
372           shadowed by log IO.  The IO buffers on the iobuf list and the
373           shadow buffers on this list match each other one for one at
374           all times. */
375	struct journal_head *	t_shadow_list;
376
377	/* Doubly-linked circular list of control buffers being written
378           to the log. */
379	struct journal_head *	t_log_list;
380
381	/* Number of outstanding updates running on this transaction */
382	int			t_updates;
383
384	/* Number of buffers reserved for use by all handles in this
385	 * transaction handle but not yet modified. */
386	int			t_outstanding_credits;
387
388	/*
389	 * Forward and backward links for the circular list of all
390	 * transactions awaiting checkpoint.
391	 */
392	/* Protected by journal_datalist_lock */
393	transaction_t		*t_cpnext, *t_cpprev;
394
395	/* When will the transaction expire (become due for commit), in
396	 * jiffies ? */
397	unsigned long		t_expires;
398
399	/* How many handles used this transaction? */
400	int t_handle_count;
401};
402
403
404/* The journal_t maintains all of the journaling state information for a
405 * single filesystem.  It is linked to from the fs superblock structure.
406 *
407 * We use the journal_t to keep track of all outstanding transaction
408 * activity on the filesystem, and to manage the state of the log
409 * writing process. */
410
411struct journal_s
412{
413	/* General journaling state flags */
414	unsigned long		j_flags;
415
416	/* Is there an outstanding uncleared error on the journal (from
417	 * a prior abort)? */
418	int			j_errno;
419
420	/* The superblock buffer */
421	struct buffer_head *	j_sb_buffer;
422	journal_superblock_t *	j_superblock;
423
424	/* Version of the superblock format */
425	int			j_format_version;
426
427	/* Number of processes waiting to create a barrier lock */
428	int			j_barrier_count;
429
430	/* The barrier lock itself */
431	struct semaphore	j_barrier;
432
433	/* Transactions: The current running transaction... */
434	transaction_t *		j_running_transaction;
435
436	/* ... the transaction we are pushing to disk ... */
437	transaction_t *		j_committing_transaction;
438
439	/* ... and a linked circular list of all transactions waiting
440	 * for checkpointing. */
441	/* Protected by journal_datalist_lock */
442	transaction_t *		j_checkpoint_transactions;
443
444	/* Wait queue for waiting for a locked transaction to start
445           committing, or for a barrier lock to be released */
446	wait_queue_head_t	j_wait_transaction_locked;
447
448	/* Wait queue for waiting for checkpointing to complete */
449	wait_queue_head_t	j_wait_logspace;
450
451	/* Wait queue for waiting for commit to complete */
452	wait_queue_head_t	j_wait_done_commit;
453
454	/* Wait queue to trigger checkpointing */
455	wait_queue_head_t	j_wait_checkpoint;
456
457	/* Wait queue to trigger commit */
458	wait_queue_head_t	j_wait_commit;
459
460	/* Wait queue to wait for updates to complete */
461	wait_queue_head_t	j_wait_updates;
462
463	/* Semaphore for locking against concurrent checkpoints */
464	struct semaphore 	j_checkpoint_sem;
465
466	/* The main journal lock, used by lock_journal() */
467	struct semaphore	j_sem;
468
469	/* Journal head: identifies the first unused block in the journal. */
470	unsigned long		j_head;
471
472	/* Journal tail: identifies the oldest still-used block in the
473	 * journal. */
474	unsigned long		j_tail;
475
476	/* Journal free: how many free blocks are there in the journal? */
477	unsigned long		j_free;
478
479	/* Journal start and end: the block numbers of the first usable
480	 * block and one beyond the last usable block in the journal. */
481	unsigned long		j_first, j_last;
482
483	/* Device, blocksize and starting block offset for the location
484	 * where we store the journal. */
485	kdev_t			j_dev;
486	int			j_blocksize;
487	unsigned int		j_blk_offset;
488
489	/* Device which holds the client fs.  For internal journal this
490	 * will be equal to j_dev. */
491	kdev_t			j_fs_dev;
492
493	/* Total maximum capacity of the journal region on disk. */
494	unsigned int		j_maxlen;
495
496	/* Optional inode where we store the journal.  If present, all
497	 * journal block numbers are mapped into this inode via
498	 * bmap(). */
499	struct inode *		j_inode;
500
501	/* Sequence number of the oldest transaction in the log */
502	tid_t			j_tail_sequence;
503	/* Sequence number of the next transaction to grant */
504	tid_t			j_transaction_sequence;
505	/* Sequence number of the most recently committed transaction */
506	tid_t			j_commit_sequence;
507	/* Sequence number of the most recent transaction wanting commit */
508	tid_t			j_commit_request;
509
510	/* Journal uuid: identifies the object (filesystem, LVM volume
511	 * etc) backed by this journal.  This will eventually be
512	 * replaced by an array of uuids, allowing us to index multiple
513	 * devices within a single journal and to perform atomic updates
514	 * across them.  */
515
516	__u8			j_uuid[16];
517
518	/* Pointer to the current commit thread for this journal */
519	struct task_struct *	j_task;
520
521	/* Maximum number of metadata buffers to allow in a single
522	 * compound commit transaction */
523	int			j_max_transaction_buffers;
524
525	/* What is the maximum transaction lifetime before we begin a
526	 * commit? */
527	unsigned long		j_commit_interval;
528
529	/* The timer used to wakeup the commit thread: */
530	struct timer_list *	j_commit_timer;
531	int			j_commit_timer_active;
532
533	/* Link all journals together - system-wide */
534	struct list_head	j_all_journals;
535
536	/* The revoke table: maintains the list of revoked blocks in the
537           current transaction. */
538	struct jbd_revoke_table_s *j_revoke;
539};
540
541/*
542 * Journal flag definitions
543 */
544#define JFS_UNMOUNT	0x001	/* Journal thread is being destroyed */
545#define JFS_ABORT	0x002	/* Journaling has been aborted for errors. */
546#define JFS_ACK_ERR	0x004	/* The errno in the sb has been acked */
547#define JFS_FLUSHED	0x008	/* The journal superblock has been flushed */
548#define JFS_LOADED	0x010	/* The journal superblock has been loaded */
549
550/*
551 * Function declarations for the journaling transaction and buffer
552 * management
553 */
554
555/* Filing buffers */
556extern void __journal_unfile_buffer(struct journal_head *);
557extern void journal_unfile_buffer(struct journal_head *);
558extern void __journal_refile_buffer(struct journal_head *);
559extern void journal_refile_buffer(struct journal_head *);
560extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
561extern void __journal_free_buffer(struct journal_head *bh);
562extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
563extern void __journal_clean_data_list(transaction_t *transaction);
564
565/* Log buffer allocation */
566extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
567extern unsigned long journal_next_log_block(journal_t *);
568
569/* Commit management */
570extern void journal_commit_transaction(journal_t *);
571
572/* Checkpoint list management */
573int __journal_clean_checkpoint_list(journal_t *journal);
574extern void journal_remove_checkpoint(struct journal_head *);
575extern void __journal_remove_checkpoint(struct journal_head *);
576extern void journal_insert_checkpoint(struct journal_head *, transaction_t *);
577extern void __journal_insert_checkpoint(struct journal_head *,transaction_t *);
578
579/* Buffer IO */
580extern int
581journal_write_metadata_buffer(transaction_t	  *transaction,
582			      struct journal_head  *jh_in,
583			      struct journal_head **jh_out,
584			      int		   blocknr);
585
586/* Transaction locking */
587extern void		__wait_on_journal (journal_t *);
588
589/*
590 * Journal locking.
591 *
592 * We need to lock the journal during transaction state changes so that
593 * nobody ever tries to take a handle on the running transaction while
594 * we are in the middle of moving it to the commit phase.
595 *
596 * Note that the locking is completely interrupt unsafe.  We never touch
597 * journal structures from interrupts.
598 *
599 * In 2.2, the BKL was required for lock_journal.  This is no longer
600 * the case.
601 */
602
603static inline void lock_journal(journal_t *journal)
604{
605	down(&journal->j_sem);
606}
607
608/* This returns zero if we acquired the semaphore */
609static inline int try_lock_journal(journal_t * journal)
610{
611	return down_trylock(&journal->j_sem);
612}
613
614static inline void unlock_journal(journal_t * journal)
615{
616	up(&journal->j_sem);
617}
618
619
620static inline handle_t *journal_current_handle(void)
621{
622	return current->journal_info;
623}
624
625/* The journaling code user interface:
626 *
627 * Create and destroy handles
628 * Register buffer modifications against the current transaction.
629 */
630
631extern handle_t *journal_start(journal_t *, int nblocks);
632extern handle_t *journal_try_start(journal_t *, int nblocks);
633extern int	 journal_restart (handle_t *, int nblocks);
634extern int	 journal_extend (handle_t *, int nblocks);
635extern int	 journal_get_write_access (handle_t *, struct buffer_head *);
636extern int	 journal_get_create_access (handle_t *, struct buffer_head *);
637extern int	 journal_get_undo_access (handle_t *, struct buffer_head *);
638extern int	 journal_dirty_data (handle_t *,
639				struct buffer_head *, int async);
640extern int	 journal_dirty_metadata (handle_t *, struct buffer_head *);
641extern void	 journal_release_buffer (handle_t *, struct buffer_head *);
642extern void	 journal_forget (handle_t *, struct buffer_head *);
643extern void	 journal_sync_buffer (struct buffer_head *);
644extern int	 journal_flushpage(journal_t *, struct page *, unsigned long);
645extern int	 journal_try_to_free_buffers(journal_t *, struct page *, int);
646extern int	 journal_stop(handle_t *);
647extern int	 journal_flush (journal_t *);
648
649extern void	 journal_lock_updates (journal_t *);
650extern void	 journal_unlock_updates (journal_t *);
651
652extern journal_t * journal_init_dev(kdev_t dev, kdev_t fs_dev,
653				int start, int len, int bsize);
654extern journal_t * journal_init_inode (struct inode *);
655extern int	   journal_update_format (journal_t *);
656extern int	   journal_check_used_features
657		   (journal_t *, unsigned long, unsigned long, unsigned long);
658extern int	   journal_check_available_features
659		   (journal_t *, unsigned long, unsigned long, unsigned long);
660extern int	   journal_set_features
661		   (journal_t *, unsigned long, unsigned long, unsigned long);
662extern int	   journal_create     (journal_t *);
663extern int	   journal_load       (journal_t *journal);
664extern void	   journal_destroy    (journal_t *);
665extern int	   journal_recover    (journal_t *journal);
666extern int	   journal_wipe       (journal_t *, int);
667extern int	   journal_skip_recovery (journal_t *);
668extern void	   journal_update_superblock (journal_t *, int);
669extern void	   __journal_abort      (journal_t *);
670extern void	   journal_abort      (journal_t *, int);
671extern int	   journal_errno      (journal_t *);
672extern void	   journal_ack_err    (journal_t *);
673extern int	   journal_clear_err  (journal_t *);
674extern unsigned long journal_bmap(journal_t *journal, unsigned long blocknr);
675extern int	    journal_force_commit(journal_t *journal);
676
677/*
678 * journal_head management
679 */
680extern struct journal_head
681		*journal_add_journal_head(struct buffer_head *bh);
682extern void	journal_remove_journal_head(struct buffer_head *bh);
683extern void	__journal_remove_journal_head(struct buffer_head *bh);
684extern void	journal_unlock_journal_head(struct journal_head *jh);
685
686/* Primary revoke support */
687#define JOURNAL_REVOKE_DEFAULT_HASH 256
688extern int	   journal_init_revoke(journal_t *, int);
689extern void	   journal_destroy_revoke_caches(void);
690extern int	   journal_init_revoke_caches(void);
691
692extern void	   journal_destroy_revoke(journal_t *);
693extern int	   journal_revoke (handle_t *,
694				unsigned long, struct buffer_head *);
695extern int	   journal_cancel_revoke(handle_t *, struct journal_head *);
696extern void	   journal_write_revoke_records(journal_t *, transaction_t *);
697
698/* Recovery revoke support */
699extern int	   journal_set_revoke(journal_t *, unsigned long, tid_t);
700extern int	   journal_test_revoke(journal_t *, unsigned long, tid_t);
701extern void	   journal_clear_revoke(journal_t *);
702extern void	   journal_brelse_array(struct buffer_head *b[], int n);
703
704/* The log thread user interface:
705 *
706 * Request space in the current transaction, and force transaction commit
707 * transitions on demand.
708 */
709
710extern int	log_space_left (journal_t *); /* Called with journal locked */
711extern tid_t	log_start_commit (journal_t *, transaction_t *);
712extern void	log_wait_commit (journal_t *, tid_t);
713extern int	log_do_checkpoint (journal_t *, int);
714
715extern void	log_wait_for_space(journal_t *, int nblocks);
716extern void	__journal_drop_transaction(journal_t *, transaction_t *);
717extern int	cleanup_journal_tail(journal_t *);
718
719/* Reduce journal memory usage by flushing */
720extern void shrink_journal_memory(void);
721
722/* Debugging code only: */
723
724#define jbd_ENOSYS() \
725do {								      \
726	printk (KERN_ERR "JBD unimplemented function " __FUNCTION__); \
727	current->state = TASK_UNINTERRUPTIBLE;			      \
728	schedule();						      \
729} while (1)
730
731/*
732 * is_journal_abort
733 *
734 * Simple test wrapper function to test the JFS_ABORT state flag.  This
735 * bit, when set, indicates that we have had a fatal error somewhere,
736 * either inside the journaling layer or indicated to us by the client
737 * (eg. ext3), and that we and should not commit any further
738 * transactions.
739 */
740
741static inline int is_journal_aborted(journal_t *journal)
742{
743	return journal->j_flags & JFS_ABORT;
744}
745
746static inline int is_handle_aborted(handle_t *handle)
747{
748	if (handle->h_aborted)
749		return 1;
750	return is_journal_aborted(handle->h_transaction->t_journal);
751}
752
753static inline void journal_abort_handle(handle_t *handle)
754{
755	handle->h_aborted = 1;
756}
757
758/* Not all architectures define BUG() */
759#ifndef BUG
760 #define BUG() do { \
761        printk("kernel BUG at %s:%d!\n", __FILE__, __LINE__); \
762	* ((char *) 0) = 0; \
763 } while (0)
764#endif /* BUG */
765
766#else
767
768extern int	   journal_recover    (journal_t *journal);
769extern int	   journal_skip_recovery (journal_t *);
770
771/* Primary revoke support */
772extern int	   journal_init_revoke(journal_t *, int);
773extern void	   journal_destroy_revoke_caches(void);
774extern int	   journal_init_revoke_caches(void);
775
776/* Recovery revoke support */
777extern int	   journal_set_revoke(journal_t *, unsigned long, tid_t);
778extern int	   journal_test_revoke(journal_t *, unsigned long, tid_t);
779extern void	   journal_clear_revoke(journal_t *);
780extern void	   journal_brelse_array(struct buffer_head *b[], int n);
781
782extern void	   journal_destroy_revoke(journal_t *);
783#endif /* __KERNEL__   */
784
785/* Comparison functions for transaction IDs: perform comparisons using
786 * modulo arithmetic so that they work over sequence number wraps. */
787
788static inline int tid_gt(tid_t x, tid_t y)
789{
790	int difference = (x - y);
791	return (difference > 0);
792}
793
794static inline int tid_geq(tid_t x, tid_t y)
795{
796	int difference = (x - y);
797	return (difference >= 0);
798}
799
800extern int journal_blocks_per_page(struct inode *inode);
801
802/*
803 * Definitions which augment the buffer_head layer
804 */
805
806/* journaling buffer types */
807#define BJ_None		0	/* Not journaled */
808#define BJ_SyncData	1	/* Normal data: flush before commit */
809#define BJ_AsyncData	2	/* writepage data: wait on it before commit */
810#define BJ_Metadata	3	/* Normal journaled metadata */
811#define BJ_Forget	4	/* Buffer superceded by this transaction */
812#define BJ_IO		5	/* Buffer is for temporary IO use */
813#define BJ_Shadow	6	/* Buffer contents being shadowed to the log */
814#define BJ_LogCtl	7	/* Buffer contains log descriptors */
815#define BJ_Reserved	8	/* Buffer is reserved for access by journal */
816#define BJ_Types	9
817
818extern int jbd_blocks_per_page(struct inode *inode);
819
820#ifdef __KERNEL__
821
822extern spinlock_t jh_splice_lock;
823/*
824 * Once `expr1' has been found true, take jh_splice_lock
825 * and then reevaluate everything.
826 */
827#define SPLICE_LOCK(expr1, expr2)				\
828	({							\
829		int ret = (expr1);				\
830		if (ret) {					\
831			spin_lock(&jh_splice_lock);		\
832			ret = (expr1) && (expr2);		\
833			spin_unlock(&jh_splice_lock);		\
834		}						\
835		ret;						\
836	})
837
838/*
839 * A number of buffer state predicates.  They test for
840 * buffer_jbd() because they are used in core kernel code.
841 *
842 * These will be racy on SMP unless we're *sure* that the
843 * buffer won't be detached from the journalling system
844 * in parallel.
845 */
846
847/* Return true if the buffer is on journal list `list' */
848static inline int buffer_jlist_eq(struct buffer_head *bh, int list)
849{
850	return SPLICE_LOCK(buffer_jbd(bh), bh2jh(bh)->b_jlist == list);
851}
852
853/* Return true if this bufer is dirty wrt the journal */
854static inline int buffer_jdirty(struct buffer_head *bh)
855{
856	return buffer_jbd(bh) && __buffer_state(bh, JBDDirty);
857}
858
859/* Return true if it's a data buffer which journalling is managing */
860static inline int buffer_jbd_data(struct buffer_head *bh)
861{
862	return SPLICE_LOCK(buffer_jbd(bh),
863			bh2jh(bh)->b_jlist == BJ_SyncData ||
864			bh2jh(bh)->b_jlist == BJ_AsyncData);
865}
866
867#ifdef CONFIG_SMP
868#define assert_spin_locked(lock)	J_ASSERT(spin_is_locked(lock))
869#else
870#define assert_spin_locked(lock)	do {} while(0)
871#endif
872
873#define buffer_trace_init(bh)	do {} while (0)
874#define print_buffer_fields(bh)	do {} while (0)
875#define print_buffer_trace(bh)	do {} while (0)
876#define BUFFER_TRACE(bh, info)	do {} while (0)
877#define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
878#define JBUFFER_TRACE(jh, info)	do {} while (0)
879
880#endif	/* __KERNEL__ */
881
882#endif	/* CONFIG_JBD || CONFIG_JBD_MODULE || !__KERNEL__ */
883
884/*
885 * Compatibility no-ops which allow the kernel to compile without CONFIG_JBD
886 * go here.
887 */
888
889#if defined(__KERNEL__) && !(defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE))
890
891#define J_ASSERT(expr)			do {} while (0)
892#define J_ASSERT_BH(bh, expr)		do {} while (0)
893#define buffer_jbd(bh)			0
894#define buffer_jlist_eq(bh, val)	0
895#define journal_buffer_journal_lru(bh)	0
896
897#endif	/* defined(__KERNEL__) && !defined(CONFIG_JBD) */
898#endif	/* _LINUX_JBD_H */
899