2 * linux/include/linux/jbd2.h
4 * Written by Stephen C. Tweedie <sct@redhat.com>
6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
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.
12 * Definitions for transaction data structures for the buffer cache
13 * filesystem journaling support.
19 /* Allow this file to be included directly into e2fsprogs */
21 #include "jfs_compat.h"
23 #define jfs_debug jbd_debug
26 #include <linux/types.h>
27 #include <linux/buffer_head.h>
28 #include <linux/journal-head.h>
29 #include <linux/stddef.h>
30 #include <linux/bit_spinlock.h>
31 #include <linux/mutex.h>
32 #include <linux/timer.h>
34 #include <asm/semaphore.h>
37 #define journal_oom_retry 1
40 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
41 * certain classes of error which can occur due to failed IOs. Under
42 * normal use we want ext4 to continue after such errors, because
43 * hardware _can_ fail, but for debugging purposes when running tests on
44 * known-good hardware we may want to trap these errors.
46 #undef JBD2_PARANOID_IOFAIL
49 * The default maximum commit age, in seconds.
51 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
53 #ifdef CONFIG_JBD2_DEBUG
55 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
56 * consistency checks. By default we don't do this unless
57 * CONFIG_JBD2_DEBUG is on.
59 #define JBD2_EXPENSIVE_CHECKING
60 extern u8 jbd2_journal_enable_debug;
62 #define jbd_debug(n, f, a...) \
64 if ((n) <= jbd2_journal_enable_debug) { \
65 printk (KERN_DEBUG "(%s, %d): %s: ", \
66 __FILE__, __LINE__, __FUNCTION__); \
71 #define jbd_debug(f, a...) /**/
74 static inline void *jbd2_alloc(size_t size, gfp_t flags)
76 return (void *)__get_free_pages(flags, get_order(size));
79 static inline void jbd2_free(void *ptr, size_t size)
81 free_pages((unsigned long)ptr, get_order(size));
84 #define JBD2_MIN_JOURNAL_BLOCKS 1024
89 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
91 * All filesystem modifications made by the process go
92 * through this handle. Recursive operations (such as quota operations)
93 * are gathered into a single update.
95 * The buffer credits field is used to account for journaled buffers
96 * being modified by the running process. To ensure that there is
97 * enough log space for all outstanding operations, we need to limit the
98 * number of outstanding buffers possible at any time. When the
99 * operation completes, any buffer credits not used are credited back to
100 * the transaction, so that at all times we know how many buffers the
101 * outstanding updates on a transaction might possibly touch.
103 * This is an opaque datatype.
105 typedef struct handle_s handle_t; /* Atomic operation type */
109 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
111 * journal_t is linked to from the fs superblock structure.
113 * We use the journal_t to keep track of all outstanding transaction
114 * activity on the filesystem, and to manage the state of the log
117 * This is an opaque datatype.
119 typedef struct journal_s journal_t; /* Journal control structure */
123 * Internal structures used by the logging mechanism:
126 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
133 * Descriptor block types:
136 #define JBD2_DESCRIPTOR_BLOCK 1
137 #define JBD2_COMMIT_BLOCK 2
138 #define JBD2_SUPERBLOCK_V1 3
139 #define JBD2_SUPERBLOCK_V2 4
140 #define JBD2_REVOKE_BLOCK 5
143 * Standard header for all descriptor blocks:
145 typedef struct journal_header_s
154 * The block tag: used to describe a single buffer in the journal.
155 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
156 * raw struct shouldn't be used for pointer math or sizeof() - use
157 * journal_tag_bytes(journal) instead to compute this.
159 typedef struct journal_block_tag_s
161 __be32 t_blocknr; /* The on-disk block number */
162 __be32 t_flags; /* See below */
163 __be32 t_blocknr_high; /* most-significant high 32bits. */
164 } journal_block_tag_t;
166 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
167 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
170 * The revoke descriptor: used on disk to describe a series of blocks to
171 * be revoked from the log
173 typedef struct jbd2_journal_revoke_header_s
175 journal_header_t r_header;
176 __be32 r_count; /* Count of bytes used in the block */
177 } jbd2_journal_revoke_header_t;
180 /* Definitions for the journal tag flags word: */
181 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
182 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
183 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
184 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
188 * The journal superblock. All fields are in big-endian byte order.
190 typedef struct journal_superblock_s
193 journal_header_t s_header;
196 /* Static information describing the journal */
197 __be32 s_blocksize; /* journal device blocksize */
198 __be32 s_maxlen; /* total blocks in journal file */
199 __be32 s_first; /* first block of log information */
202 /* Dynamic information describing the current state of the log */
203 __be32 s_sequence; /* first commit ID expected in log */
204 __be32 s_start; /* blocknr of start of log */
207 /* Error value, as set by jbd2_journal_abort(). */
211 /* Remaining fields are only valid in a version-2 superblock */
212 __be32 s_feature_compat; /* compatible feature set */
213 __be32 s_feature_incompat; /* incompatible feature set */
214 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
216 __u8 s_uuid[16]; /* 128-bit uuid for journal */
219 __be32 s_nr_users; /* Nr of filesystems sharing log */
221 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
224 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
225 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
231 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
233 } journal_superblock_t;
235 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
236 ((j)->j_format_version >= 2 && \
237 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
238 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
239 ((j)->j_format_version >= 2 && \
240 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
241 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
242 ((j)->j_format_version >= 2 && \
243 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
245 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
246 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
248 /* Features known to this kernel version: */
249 #define JBD2_KNOWN_COMPAT_FEATURES 0
250 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
251 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
252 JBD2_FEATURE_INCOMPAT_64BIT)
256 #include <linux/fs.h>
257 #include <linux/sched.h>
259 #define J_ASSERT(assert) BUG_ON(!(assert))
261 #if defined(CONFIG_BUFFER_DEBUG)
262 void buffer_assertion_failure(struct buffer_head *bh);
263 #define J_ASSERT_BH(bh, expr) \
266 buffer_assertion_failure(bh); \
269 #define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
271 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
272 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
275 #if defined(JBD2_PARANOID_IOFAIL)
276 #define J_EXPECT(expr, why...) J_ASSERT(expr)
277 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
278 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
280 #define __journal_expect(expr, why...) \
285 "EXT3-fs unexpected failure: %s;\n",# expr); \
286 printk(KERN_ERR why "\n"); \
290 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
291 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
292 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
295 enum jbd_state_bits {
296 BH_JBD /* Has an attached ext3 journal_head */
298 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
299 BH_Freed, /* Has been freed (truncated) */
300 BH_Revoked, /* Has been revoked from the log */
301 BH_RevokeValid, /* Revoked flag is valid */
302 BH_JBDDirty, /* Is dirty but journaled */
303 BH_State, /* Pins most journal_head state */
304 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
305 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
309 BUFFER_FNS(JWrite, jwrite)
310 BUFFER_FNS(JBDDirty, jbddirty)
311 TAS_BUFFER_FNS(JBDDirty, jbddirty)
312 BUFFER_FNS(Revoked, revoked)
313 TAS_BUFFER_FNS(Revoked, revoked)
314 BUFFER_FNS(RevokeValid, revokevalid)
315 TAS_BUFFER_FNS(RevokeValid, revokevalid)
316 BUFFER_FNS(Freed, freed)
318 static inline struct buffer_head *jh2bh(struct journal_head *jh)
323 static inline struct journal_head *bh2jh(struct buffer_head *bh)
325 return bh->b_private;
328 static inline void jbd_lock_bh_state(struct buffer_head *bh)
330 bit_spin_lock(BH_State, &bh->b_state);
333 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
335 return bit_spin_trylock(BH_State, &bh->b_state);
338 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
340 return bit_spin_is_locked(BH_State, &bh->b_state);
343 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
345 bit_spin_unlock(BH_State, &bh->b_state);
348 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
350 bit_spin_lock(BH_JournalHead, &bh->b_state);
353 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
355 bit_spin_unlock(BH_JournalHead, &bh->b_state);
358 struct jbd2_revoke_table_s;
361 * struct handle_s - The handle_s type is the concrete type associated with
363 * @h_transaction: Which compound transaction is this update a part of?
364 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
365 * @h_ref: Reference count on this handle
366 * @h_err: Field for caller's use to track errors through large fs operations
367 * @h_sync: flag for sync-on-close
368 * @h_jdata: flag to force data journaling
369 * @h_aborted: flag indicating fatal error on handle
372 /* Docbook can't yet cope with the bit fields, but will leave the documentation
373 * in so it can be fixed later.
378 /* Which compound transaction is this update a part of? */
379 transaction_t *h_transaction;
381 /* Number of remaining buffers we are allowed to dirty: */
382 int h_buffer_credits;
384 /* Reference count on this handle */
387 /* Field for caller's use to track errors through large fs */
391 /* Flags [no locking] */
392 unsigned int h_sync: 1; /* sync-on-close */
393 unsigned int h_jdata: 1; /* force data journaling */
394 unsigned int h_aborted: 1; /* fatal error on handle */
399 * Some stats for checkpoint phase
401 struct transaction_chp_stats_s {
402 unsigned long cs_chp_time;
403 unsigned long cs_forced_to_close;
404 unsigned long cs_written;
405 unsigned long cs_dropped;
408 /* The transaction_t type is the guts of the journaling mechanism. It
409 * tracks a compound transaction through its various states:
411 * RUNNING: accepting new updates
412 * LOCKED: Updates still running but we don't accept new ones
413 * RUNDOWN: Updates are tidying up but have finished requesting
414 * new buffers to modify (state not used for now)
415 * FLUSH: All updates complete, but we are still writing to disk
416 * COMMIT: All data on disk, writing commit record
417 * FINISHED: We still have to keep the transaction for checkpointing.
419 * The transaction keeps track of all of the buffers modified by a
420 * running transaction, and all of the buffers committed but not yet
421 * flushed to home for finished transactions.
428 * ->jbd_lock_bh_journal_head() (This is "innermost")
431 * ->jbd_lock_bh_state()
433 * jbd_lock_bh_state()
440 * ->j_list_lock (journal_unmap_buffer)
446 /* Pointer to the journal for this transaction. [no locking] */
447 journal_t *t_journal;
449 /* Sequence number for this transaction [no locking] */
453 * Transaction's current state
454 * [no locking - only kjournald2 alters this]
455 * [j_list_lock] guards transition of a transaction into T_FINISHED
456 * state and subsequent call of __jbd2_journal_drop_transaction()
457 * FIXME: needs barriers
458 * KLUDGE: [use j_state_lock]
470 * Where in the log does this transaction's commit start? [no locking]
472 unsigned long t_log_start;
474 /* Number of buffers on the t_buffers list [j_list_lock] */
478 * Doubly-linked circular list of all buffers reserved but not yet
479 * modified by this transaction [j_list_lock]
481 struct journal_head *t_reserved_list;
484 * Doubly-linked circular list of all buffers under writeout during
485 * commit [j_list_lock]
487 struct journal_head *t_locked_list;
490 * Doubly-linked circular list of all metadata buffers owned by this
491 * transaction [j_list_lock]
493 struct journal_head *t_buffers;
496 * Doubly-linked circular list of all data buffers still to be
497 * flushed before this transaction can be committed [j_list_lock]
499 struct journal_head *t_sync_datalist;
502 * Doubly-linked circular list of all forget buffers (superseded
503 * buffers which we can un-checkpoint once this transaction commits)
506 struct journal_head *t_forget;
509 * Doubly-linked circular list of all buffers still to be flushed before
510 * this transaction can be checkpointed. [j_list_lock]
512 struct journal_head *t_checkpoint_list;
515 * Doubly-linked circular list of all buffers submitted for IO while
516 * checkpointing. [j_list_lock]
518 struct journal_head *t_checkpoint_io_list;
521 * Doubly-linked circular list of temporary buffers currently undergoing
522 * IO in the log [j_list_lock]
524 struct journal_head *t_iobuf_list;
527 * Doubly-linked circular list of metadata buffers being shadowed by log
528 * IO. The IO buffers on the iobuf list and the shadow buffers on this
529 * list match each other one for one at all times. [j_list_lock]
531 struct journal_head *t_shadow_list;
534 * Doubly-linked circular list of control buffers being written to the
537 struct journal_head *t_log_list;
540 * Protects info related to handles
542 spinlock_t t_handle_lock;
545 * Longest time some handle had to wait for running transaction
547 unsigned long t_max_wait;
550 * When transaction started
552 unsigned long t_start;
555 * Checkpointing stats [j_checkpoint_sem]
557 struct transaction_chp_stats_s t_chp_stats;
560 * Number of outstanding updates running on this transaction
566 * Number of buffers reserved for use by all handles in this transaction
567 * handle but not yet modified. [t_handle_lock]
569 int t_outstanding_credits;
572 * Forward and backward links for the circular list of all transactions
573 * awaiting checkpoint. [j_list_lock]
575 transaction_t *t_cpnext, *t_cpprev;
578 * When will the transaction expire (become due for commit), in jiffies?
581 unsigned long t_expires;
584 * How many handles used this transaction? [t_handle_lock]
590 struct transaction_run_stats_s {
591 unsigned long rs_wait;
592 unsigned long rs_running;
593 unsigned long rs_locked;
594 unsigned long rs_flushing;
595 unsigned long rs_logging;
597 unsigned long rs_handle_count;
598 unsigned long rs_blocks;
599 unsigned long rs_blocks_logged;
602 struct transaction_stats_s {
604 unsigned long ts_tid;
606 struct transaction_run_stats_s run;
607 struct transaction_chp_stats_s chp;
611 #define JBD2_STATS_RUN 1
612 #define JBD2_STATS_CHECKPOINT 2
614 static inline unsigned long
615 jbd2_time_diff(unsigned long start, unsigned long end)
620 return end + (MAX_JIFFY_OFFSET - start);
624 * struct journal_s - The journal_s type is the concrete type associated with
626 * @j_flags: General journaling state flags
627 * @j_errno: Is there an outstanding uncleared error on the journal (from a
629 * @j_sb_buffer: First part of superblock buffer
630 * @j_superblock: Second part of superblock buffer
631 * @j_format_version: Version of the superblock format
632 * @j_state_lock: Protect the various scalars in the journal
633 * @j_barrier_count: Number of processes waiting to create a barrier lock
634 * @j_barrier: The barrier lock itself
635 * @j_running_transaction: The current running transaction..
636 * @j_committing_transaction: the transaction we are pushing to disk
637 * @j_checkpoint_transactions: a linked circular list of all transactions
638 * waiting for checkpointing
639 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
640 * to start committing, or for a barrier lock to be released
641 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
642 * @j_wait_done_commit: Wait queue for waiting for commit to complete
643 * @j_wait_checkpoint: Wait queue to trigger checkpointing
644 * @j_wait_commit: Wait queue to trigger commit
645 * @j_wait_updates: Wait queue to wait for updates to complete
646 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
647 * @j_head: Journal head - identifies the first unused block in the journal
648 * @j_tail: Journal tail - identifies the oldest still-used block in the
650 * @j_free: Journal free - how many free blocks are there in the journal?
651 * @j_first: The block number of the first usable block
652 * @j_last: The block number one beyond the last usable block
653 * @j_dev: Device where we store the journal
654 * @j_blocksize: blocksize for the location where we store the journal.
655 * @j_blk_offset: starting block offset for into the device where we store the
657 * @j_fs_dev: Device which holds the client fs. For internal journal this will
659 * @j_maxlen: Total maximum capacity of the journal region on disk.
660 * @j_list_lock: Protects the buffer lists and internal buffer state.
661 * @j_inode: Optional inode where we store the journal. If present, all journal
662 * block numbers are mapped into this inode via bmap().
663 * @j_tail_sequence: Sequence number of the oldest transaction in the log
664 * @j_transaction_sequence: Sequence number of the next transaction to grant
665 * @j_commit_sequence: Sequence number of the most recently committed
667 * @j_commit_request: Sequence number of the most recent transaction wanting
669 * @j_uuid: Uuid of client object.
670 * @j_task: Pointer to the current commit thread for this journal
671 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
672 * single compound commit transaction
673 * @j_commit_interval: What is the maximum transaction lifetime before we begin
675 * @j_commit_timer: The timer used to wakeup the commit thread
676 * @j_revoke_lock: Protect the revoke table
677 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
678 * current transaction.
679 * @j_revoke_table: alternate revoke tables for j_revoke
680 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
681 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
682 * number that will fit in j_blocksize
683 * @j_last_sync_writer: most recent pid which did a synchronous write
684 * @j_history: Buffer storing the transactions statistics history
685 * @j_history_max: Maximum number of transactions in the statistics history
686 * @j_history_cur: Current number of transactions in the statistics history
687 * @j_history_lock: Protect the transactions statistics history
688 * @j_proc_entry: procfs entry for the jbd statistics directory
689 * @j_stats: Overall statistics
690 * @j_private: An opaque pointer to fs-private information.
695 /* General journaling state flags [j_state_lock] */
696 unsigned long j_flags;
699 * Is there an outstanding uncleared error on the journal (from a prior
700 * abort)? [j_state_lock]
704 /* The superblock buffer */
705 struct buffer_head *j_sb_buffer;
706 journal_superblock_t *j_superblock;
708 /* Version of the superblock format */
709 int j_format_version;
712 * Protect the various scalars in the journal
714 spinlock_t j_state_lock;
717 * Number of processes waiting to create a barrier lock [j_state_lock]
721 /* The barrier lock itself */
722 struct mutex j_barrier;
725 * Transactions: The current running transaction...
726 * [j_state_lock] [caller holding open handle]
728 transaction_t *j_running_transaction;
731 * the transaction we are pushing to disk
732 * [j_state_lock] [caller holding open handle]
734 transaction_t *j_committing_transaction;
737 * ... and a linked circular list of all transactions waiting for
738 * checkpointing. [j_list_lock]
740 transaction_t *j_checkpoint_transactions;
743 * Wait queue for waiting for a locked transaction to start committing,
744 * or for a barrier lock to be released
746 wait_queue_head_t j_wait_transaction_locked;
748 /* Wait queue for waiting for checkpointing to complete */
749 wait_queue_head_t j_wait_logspace;
751 /* Wait queue for waiting for commit to complete */
752 wait_queue_head_t j_wait_done_commit;
754 /* Wait queue to trigger checkpointing */
755 wait_queue_head_t j_wait_checkpoint;
757 /* Wait queue to trigger commit */
758 wait_queue_head_t j_wait_commit;
760 /* Wait queue to wait for updates to complete */
761 wait_queue_head_t j_wait_updates;
763 /* Semaphore for locking against concurrent checkpoints */
764 struct mutex j_checkpoint_mutex;
767 * Journal head: identifies the first unused block in the journal.
770 unsigned long j_head;
773 * Journal tail: identifies the oldest still-used block in the journal.
776 unsigned long j_tail;
779 * Journal free: how many free blocks are there in the journal?
782 unsigned long j_free;
785 * Journal start and end: the block numbers of the first usable block
786 * and one beyond the last usable block in the journal. [j_state_lock]
788 unsigned long j_first;
789 unsigned long j_last;
792 * Device, blocksize and starting block offset for the location where we
795 struct block_device *j_dev;
797 unsigned long long j_blk_offset;
800 * Device which holds the client fs. For internal journal this will be
803 struct block_device *j_fs_dev;
805 /* Total maximum capacity of the journal region on disk. */
806 unsigned int j_maxlen;
809 * Protects the buffer lists and internal buffer state.
811 spinlock_t j_list_lock;
813 /* Optional inode where we store the journal. If present, all */
814 /* journal block numbers are mapped into this inode via */
816 struct inode *j_inode;
819 * Sequence number of the oldest transaction in the log [j_state_lock]
821 tid_t j_tail_sequence;
824 * Sequence number of the next transaction to grant [j_state_lock]
826 tid_t j_transaction_sequence;
829 * Sequence number of the most recently committed transaction
832 tid_t j_commit_sequence;
835 * Sequence number of the most recent transaction wanting commit
838 tid_t j_commit_request;
841 * Journal uuid: identifies the object (filesystem, LVM volume etc)
842 * backed by this journal. This will eventually be replaced by an array
843 * of uuids, allowing us to index multiple devices within a single
844 * journal and to perform atomic updates across them.
848 /* Pointer to the current commit thread for this journal */
849 struct task_struct *j_task;
852 * Maximum number of metadata buffers to allow in a single compound
855 int j_max_transaction_buffers;
858 * What is the maximum transaction lifetime before we begin a commit?
860 unsigned long j_commit_interval;
862 /* The timer used to wakeup the commit thread: */
863 struct timer_list j_commit_timer;
866 * The revoke table: maintains the list of revoked blocks in the
867 * current transaction. [j_revoke_lock]
869 spinlock_t j_revoke_lock;
870 struct jbd2_revoke_table_s *j_revoke;
871 struct jbd2_revoke_table_s *j_revoke_table[2];
874 * array of bhs for jbd2_journal_commit_transaction
876 struct buffer_head **j_wbuf;
879 pid_t j_last_sync_writer;
884 struct transaction_stats_s *j_history;
888 * Protect the transactions statistics history
890 spinlock_t j_history_lock;
891 struct proc_dir_entry *j_proc_entry;
892 struct transaction_stats_s j_stats;
895 * An opaque pointer to fs-private information. ext3 puts its
896 * superblock pointer here
902 * Journal flag definitions
904 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
905 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
906 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
907 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
908 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
909 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
912 * Function declarations for the journaling transaction and buffer
917 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
918 extern void __jbd2_journal_unfile_buffer(struct journal_head *);
919 extern void __jbd2_journal_refile_buffer(struct journal_head *);
920 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
921 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
922 extern void __journal_free_buffer(struct journal_head *bh);
923 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
924 extern void __journal_clean_data_list(transaction_t *transaction);
926 /* Log buffer allocation */
927 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
928 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
930 /* Commit management */
931 extern void jbd2_journal_commit_transaction(journal_t *);
933 /* Checkpoint list management */
934 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
935 int __jbd2_journal_remove_checkpoint(struct journal_head *);
936 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
940 jbd2_journal_write_metadata_buffer(transaction_t *transaction,
941 struct journal_head *jh_in,
942 struct journal_head **jh_out,
943 unsigned long long blocknr);
945 /* Transaction locking */
946 extern void __wait_on_journal (journal_t *);
951 * We need to lock the journal during transaction state changes so that nobody
952 * ever tries to take a handle on the running transaction while we are in the
953 * middle of moving it to the commit phase. j_state_lock does this.
955 * Note that the locking is completely interrupt unsafe. We never touch
956 * journal structures from interrupts.
959 static inline handle_t *journal_current_handle(void)
961 return current->journal_info;
964 /* The journaling code user interface:
966 * Create and destroy handles
967 * Register buffer modifications against the current transaction.
970 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
971 extern int jbd2_journal_restart (handle_t *, int nblocks);
972 extern int jbd2_journal_extend (handle_t *, int nblocks);
973 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
974 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
975 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
976 extern int jbd2_journal_dirty_data (handle_t *, struct buffer_head *);
977 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
978 extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
979 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
980 extern void journal_sync_buffer (struct buffer_head *);
981 extern void jbd2_journal_invalidatepage(journal_t *,
982 struct page *, unsigned long);
983 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
984 extern int jbd2_journal_stop(handle_t *);
985 extern int jbd2_journal_flush (journal_t *);
986 extern void jbd2_journal_lock_updates (journal_t *);
987 extern void jbd2_journal_unlock_updates (journal_t *);
989 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
990 struct block_device *fs_dev,
991 unsigned long long start, int len, int bsize);
992 extern journal_t * jbd2_journal_init_inode (struct inode *);
993 extern int jbd2_journal_update_format (journal_t *);
994 extern int jbd2_journal_check_used_features
995 (journal_t *, unsigned long, unsigned long, unsigned long);
996 extern int jbd2_journal_check_available_features
997 (journal_t *, unsigned long, unsigned long, unsigned long);
998 extern int jbd2_journal_set_features
999 (journal_t *, unsigned long, unsigned long, unsigned long);
1000 extern int jbd2_journal_create (journal_t *);
1001 extern int jbd2_journal_load (journal_t *journal);
1002 extern void jbd2_journal_destroy (journal_t *);
1003 extern int jbd2_journal_recover (journal_t *journal);
1004 extern int jbd2_journal_wipe (journal_t *, int);
1005 extern int jbd2_journal_skip_recovery (journal_t *);
1006 extern void jbd2_journal_update_superblock (journal_t *, int);
1007 extern void __jbd2_journal_abort_hard (journal_t *);
1008 extern void jbd2_journal_abort (journal_t *, int);
1009 extern int jbd2_journal_errno (journal_t *);
1010 extern void jbd2_journal_ack_err (journal_t *);
1011 extern int jbd2_journal_clear_err (journal_t *);
1012 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1013 extern int jbd2_journal_force_commit(journal_t *);
1016 * journal_head management
1018 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1019 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1020 void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1021 void jbd2_journal_put_journal_head(struct journal_head *jh);
1026 extern struct kmem_cache *jbd2_handle_cache;
1028 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1030 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
1033 static inline void jbd2_free_handle(handle_t *handle)
1035 kmem_cache_free(jbd2_handle_cache, handle);
1038 /* Primary revoke support */
1039 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1040 extern int jbd2_journal_init_revoke(journal_t *, int);
1041 extern void jbd2_journal_destroy_revoke_caches(void);
1042 extern int jbd2_journal_init_revoke_caches(void);
1044 extern void jbd2_journal_destroy_revoke(journal_t *);
1045 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1046 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1047 extern void jbd2_journal_write_revoke_records(journal_t *, transaction_t *);
1049 /* Recovery revoke support */
1050 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1051 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1052 extern void jbd2_journal_clear_revoke(journal_t *);
1053 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1056 * The log thread user interface:
1058 * Request space in the current transaction, and force transaction commit
1059 * transitions on demand.
1062 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1063 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1064 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1065 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1066 int jbd2_journal_force_commit_nested(journal_t *journal);
1067 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1068 int jbd2_log_do_checkpoint(journal_t *journal);
1070 void __jbd2_log_wait_for_space(journal_t *journal);
1071 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1072 extern int jbd2_cleanup_journal_tail(journal_t *);
1074 /* Debugging code only: */
1076 #define jbd_ENOSYS() \
1078 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1079 current->state = TASK_UNINTERRUPTIBLE; \
1086 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1087 * bit, when set, indicates that we have had a fatal error somewhere,
1088 * either inside the journaling layer or indicated to us by the client
1089 * (eg. ext3), and that we and should not commit any further
1093 static inline int is_journal_aborted(journal_t *journal)
1095 return journal->j_flags & JBD2_ABORT;
1098 static inline int is_handle_aborted(handle_t *handle)
1100 if (handle->h_aborted)
1102 return is_journal_aborted(handle->h_transaction->t_journal);
1105 static inline void jbd2_journal_abort_handle(handle_t *handle)
1107 handle->h_aborted = 1;
1110 #endif /* __KERNEL__ */
1112 /* Comparison functions for transaction IDs: perform comparisons using
1113 * modulo arithmetic so that they work over sequence number wraps. */
1115 static inline int tid_gt(tid_t x, tid_t y)
1117 int difference = (x - y);
1118 return (difference > 0);
1121 static inline int tid_geq(tid_t x, tid_t y)
1123 int difference = (x - y);
1124 return (difference >= 0);
1127 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1128 extern size_t journal_tag_bytes(journal_t *journal);
1131 * Return the minimum number of blocks which must be free in the journal
1132 * before a new transaction may be started. Must be called under j_state_lock.
1134 static inline int jbd_space_needed(journal_t *journal)
1136 int nblocks = journal->j_max_transaction_buffers;
1137 if (journal->j_committing_transaction)
1138 nblocks += journal->j_committing_transaction->
1139 t_outstanding_credits;
1144 * Definitions which augment the buffer_head layer
1147 /* journaling buffer types */
1148 #define BJ_None 0 /* Not journaled */
1149 #define BJ_SyncData 1 /* Normal data: flush before commit */
1150 #define BJ_Metadata 2 /* Normal journaled metadata */
1151 #define BJ_Forget 3 /* Buffer superseded by this transaction */
1152 #define BJ_IO 4 /* Buffer is for temporary IO use */
1153 #define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1154 #define BJ_LogCtl 6 /* Buffer contains log descriptors */
1155 #define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1156 #define BJ_Locked 8 /* Locked for I/O during commit */
1159 extern int jbd_blocks_per_page(struct inode *inode);
1163 #define buffer_trace_init(bh) do {} while (0)
1164 #define print_buffer_fields(bh) do {} while (0)
1165 #define print_buffer_trace(bh) do {} while (0)
1166 #define BUFFER_TRACE(bh, info) do {} while (0)
1167 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1168 #define JBUFFER_TRACE(jh, info) do {} while (0)
1170 #endif /* __KERNEL__ */
1172 #endif /* _LINUX_JBD2_H */