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
155 #define JBD2_CRC32_CHKSUM 1
156 #define JBD2_MD5_CHKSUM 2
157 #define JBD2_SHA1_CHKSUM 3
159 #define JBD2_CRC32_CHKSUM_SIZE 4
161 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
163 * Commit block header for storing transactional checksums:
165 struct commit_header {
169 unsigned char h_chksum_type;
170 unsigned char h_chksum_size;
171 unsigned char h_padding[2];
172 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
176 * The block tag: used to describe a single buffer in the journal.
177 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
178 * raw struct shouldn't be used for pointer math or sizeof() - use
179 * journal_tag_bytes(journal) instead to compute this.
181 typedef struct journal_block_tag_s
183 __be32 t_blocknr; /* The on-disk block number */
184 __be32 t_flags; /* See below */
185 __be32 t_blocknr_high; /* most-significant high 32bits. */
186 } journal_block_tag_t;
188 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
189 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
192 * The revoke descriptor: used on disk to describe a series of blocks to
193 * be revoked from the log
195 typedef struct jbd2_journal_revoke_header_s
197 journal_header_t r_header;
198 __be32 r_count; /* Count of bytes used in the block */
199 } jbd2_journal_revoke_header_t;
202 /* Definitions for the journal tag flags word: */
203 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
204 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
205 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
206 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
210 * The journal superblock. All fields are in big-endian byte order.
212 typedef struct journal_superblock_s
215 journal_header_t s_header;
218 /* Static information describing the journal */
219 __be32 s_blocksize; /* journal device blocksize */
220 __be32 s_maxlen; /* total blocks in journal file */
221 __be32 s_first; /* first block of log information */
224 /* Dynamic information describing the current state of the log */
225 __be32 s_sequence; /* first commit ID expected in log */
226 __be32 s_start; /* blocknr of start of log */
229 /* Error value, as set by jbd2_journal_abort(). */
233 /* Remaining fields are only valid in a version-2 superblock */
234 __be32 s_feature_compat; /* compatible feature set */
235 __be32 s_feature_incompat; /* incompatible feature set */
236 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
238 __u8 s_uuid[16]; /* 128-bit uuid for journal */
241 __be32 s_nr_users; /* Nr of filesystems sharing log */
243 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
246 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
247 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
253 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
255 } journal_superblock_t;
257 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
258 ((j)->j_format_version >= 2 && \
259 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
260 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
261 ((j)->j_format_version >= 2 && \
262 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
263 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
264 ((j)->j_format_version >= 2 && \
265 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
267 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
269 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
270 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
271 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
273 /* Features known to this kernel version: */
274 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
275 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
276 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
277 JBD2_FEATURE_INCOMPAT_64BIT | \
278 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)
282 #include <linux/fs.h>
283 #include <linux/sched.h>
285 #define J_ASSERT(assert) BUG_ON(!(assert))
287 #if defined(CONFIG_BUFFER_DEBUG)
288 void buffer_assertion_failure(struct buffer_head *bh);
289 #define J_ASSERT_BH(bh, expr) \
292 buffer_assertion_failure(bh); \
295 #define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
297 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
298 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
301 #if defined(JBD2_PARANOID_IOFAIL)
302 #define J_EXPECT(expr, why...) J_ASSERT(expr)
303 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
304 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
306 #define __journal_expect(expr, why...) \
311 "EXT3-fs unexpected failure: %s;\n",# expr); \
312 printk(KERN_ERR why "\n"); \
316 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
317 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
318 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
321 enum jbd_state_bits {
322 BH_JBD /* Has an attached ext3 journal_head */
324 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
325 BH_Freed, /* Has been freed (truncated) */
326 BH_Revoked, /* Has been revoked from the log */
327 BH_RevokeValid, /* Revoked flag is valid */
328 BH_JBDDirty, /* Is dirty but journaled */
329 BH_State, /* Pins most journal_head state */
330 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
331 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
335 BUFFER_FNS(JWrite, jwrite)
336 BUFFER_FNS(JBDDirty, jbddirty)
337 TAS_BUFFER_FNS(JBDDirty, jbddirty)
338 BUFFER_FNS(Revoked, revoked)
339 TAS_BUFFER_FNS(Revoked, revoked)
340 BUFFER_FNS(RevokeValid, revokevalid)
341 TAS_BUFFER_FNS(RevokeValid, revokevalid)
342 BUFFER_FNS(Freed, freed)
344 static inline struct buffer_head *jh2bh(struct journal_head *jh)
349 static inline struct journal_head *bh2jh(struct buffer_head *bh)
351 return bh->b_private;
354 static inline void jbd_lock_bh_state(struct buffer_head *bh)
356 bit_spin_lock(BH_State, &bh->b_state);
359 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
361 return bit_spin_trylock(BH_State, &bh->b_state);
364 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
366 return bit_spin_is_locked(BH_State, &bh->b_state);
369 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
371 bit_spin_unlock(BH_State, &bh->b_state);
374 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
376 bit_spin_lock(BH_JournalHead, &bh->b_state);
379 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
381 bit_spin_unlock(BH_JournalHead, &bh->b_state);
384 struct jbd2_revoke_table_s;
387 * struct handle_s - The handle_s type is the concrete type associated with
389 * @h_transaction: Which compound transaction is this update a part of?
390 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
391 * @h_ref: Reference count on this handle
392 * @h_err: Field for caller's use to track errors through large fs operations
393 * @h_sync: flag for sync-on-close
394 * @h_jdata: flag to force data journaling
395 * @h_aborted: flag indicating fatal error on handle
398 /* Docbook can't yet cope with the bit fields, but will leave the documentation
399 * in so it can be fixed later.
404 /* Which compound transaction is this update a part of? */
405 transaction_t *h_transaction;
407 /* Number of remaining buffers we are allowed to dirty: */
408 int h_buffer_credits;
410 /* Reference count on this handle */
413 /* Field for caller's use to track errors through large fs */
417 /* Flags [no locking] */
418 unsigned int h_sync: 1; /* sync-on-close */
419 unsigned int h_jdata: 1; /* force data journaling */
420 unsigned int h_aborted: 1; /* fatal error on handle */
422 #ifdef CONFIG_DEBUG_LOCK_ALLOC
423 struct lockdep_map h_lockdep_map;
429 * Some stats for checkpoint phase
431 struct transaction_chp_stats_s {
432 unsigned long cs_chp_time;
433 unsigned long cs_forced_to_close;
434 unsigned long cs_written;
435 unsigned long cs_dropped;
438 /* The transaction_t type is the guts of the journaling mechanism. It
439 * tracks a compound transaction through its various states:
441 * RUNNING: accepting new updates
442 * LOCKED: Updates still running but we don't accept new ones
443 * RUNDOWN: Updates are tidying up but have finished requesting
444 * new buffers to modify (state not used for now)
445 * FLUSH: All updates complete, but we are still writing to disk
446 * COMMIT: All data on disk, writing commit record
447 * FINISHED: We still have to keep the transaction for checkpointing.
449 * The transaction keeps track of all of the buffers modified by a
450 * running transaction, and all of the buffers committed but not yet
451 * flushed to home for finished transactions.
458 * ->jbd_lock_bh_journal_head() (This is "innermost")
461 * ->jbd_lock_bh_state()
463 * jbd_lock_bh_state()
470 * ->j_list_lock (journal_unmap_buffer)
476 /* Pointer to the journal for this transaction. [no locking] */
477 journal_t *t_journal;
479 /* Sequence number for this transaction [no locking] */
483 * Transaction's current state
484 * [no locking - only kjournald2 alters this]
485 * [j_list_lock] guards transition of a transaction into T_FINISHED
486 * state and subsequent call of __jbd2_journal_drop_transaction()
487 * FIXME: needs barriers
488 * KLUDGE: [use j_state_lock]
500 * Where in the log does this transaction's commit start? [no locking]
502 unsigned long t_log_start;
504 /* Number of buffers on the t_buffers list [j_list_lock] */
508 * Doubly-linked circular list of all buffers reserved but not yet
509 * modified by this transaction [j_list_lock]
511 struct journal_head *t_reserved_list;
514 * Doubly-linked circular list of all buffers under writeout during
515 * commit [j_list_lock]
517 struct journal_head *t_locked_list;
520 * Doubly-linked circular list of all metadata buffers owned by this
521 * transaction [j_list_lock]
523 struct journal_head *t_buffers;
526 * Doubly-linked circular list of all data buffers still to be
527 * flushed before this transaction can be committed [j_list_lock]
529 struct journal_head *t_sync_datalist;
532 * Doubly-linked circular list of all forget buffers (superseded
533 * buffers which we can un-checkpoint once this transaction commits)
536 struct journal_head *t_forget;
539 * Doubly-linked circular list of all buffers still to be flushed before
540 * this transaction can be checkpointed. [j_list_lock]
542 struct journal_head *t_checkpoint_list;
545 * Doubly-linked circular list of all buffers submitted for IO while
546 * checkpointing. [j_list_lock]
548 struct journal_head *t_checkpoint_io_list;
551 * Doubly-linked circular list of temporary buffers currently undergoing
552 * IO in the log [j_list_lock]
554 struct journal_head *t_iobuf_list;
557 * Doubly-linked circular list of metadata buffers being shadowed by log
558 * IO. The IO buffers on the iobuf list and the shadow buffers on this
559 * list match each other one for one at all times. [j_list_lock]
561 struct journal_head *t_shadow_list;
564 * Doubly-linked circular list of control buffers being written to the
567 struct journal_head *t_log_list;
570 * Protects info related to handles
572 spinlock_t t_handle_lock;
575 * Longest time some handle had to wait for running transaction
577 unsigned long t_max_wait;
580 * When transaction started
582 unsigned long t_start;
585 * Checkpointing stats [j_checkpoint_sem]
587 struct transaction_chp_stats_s t_chp_stats;
590 * Number of outstanding updates running on this transaction
596 * Number of buffers reserved for use by all handles in this transaction
597 * handle but not yet modified. [t_handle_lock]
599 int t_outstanding_credits;
602 * Forward and backward links for the circular list of all transactions
603 * awaiting checkpoint. [j_list_lock]
605 transaction_t *t_cpnext, *t_cpprev;
608 * When will the transaction expire (become due for commit), in jiffies?
611 unsigned long t_expires;
614 * How many handles used this transaction? [t_handle_lock]
620 struct transaction_run_stats_s {
621 unsigned long rs_wait;
622 unsigned long rs_running;
623 unsigned long rs_locked;
624 unsigned long rs_flushing;
625 unsigned long rs_logging;
627 unsigned long rs_handle_count;
628 unsigned long rs_blocks;
629 unsigned long rs_blocks_logged;
632 struct transaction_stats_s {
634 unsigned long ts_tid;
636 struct transaction_run_stats_s run;
637 struct transaction_chp_stats_s chp;
641 #define JBD2_STATS_RUN 1
642 #define JBD2_STATS_CHECKPOINT 2
644 static inline unsigned long
645 jbd2_time_diff(unsigned long start, unsigned long end)
650 return end + (MAX_JIFFY_OFFSET - start);
654 * struct journal_s - The journal_s type is the concrete type associated with
656 * @j_flags: General journaling state flags
657 * @j_errno: Is there an outstanding uncleared error on the journal (from a
659 * @j_sb_buffer: First part of superblock buffer
660 * @j_superblock: Second part of superblock buffer
661 * @j_format_version: Version of the superblock format
662 * @j_state_lock: Protect the various scalars in the journal
663 * @j_barrier_count: Number of processes waiting to create a barrier lock
664 * @j_barrier: The barrier lock itself
665 * @j_running_transaction: The current running transaction..
666 * @j_committing_transaction: the transaction we are pushing to disk
667 * @j_checkpoint_transactions: a linked circular list of all transactions
668 * waiting for checkpointing
669 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
670 * to start committing, or for a barrier lock to be released
671 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
672 * @j_wait_done_commit: Wait queue for waiting for commit to complete
673 * @j_wait_checkpoint: Wait queue to trigger checkpointing
674 * @j_wait_commit: Wait queue to trigger commit
675 * @j_wait_updates: Wait queue to wait for updates to complete
676 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
677 * @j_head: Journal head - identifies the first unused block in the journal
678 * @j_tail: Journal tail - identifies the oldest still-used block in the
680 * @j_free: Journal free - how many free blocks are there in the journal?
681 * @j_first: The block number of the first usable block
682 * @j_last: The block number one beyond the last usable block
683 * @j_dev: Device where we store the journal
684 * @j_blocksize: blocksize for the location where we store the journal.
685 * @j_blk_offset: starting block offset for into the device where we store the
687 * @j_fs_dev: Device which holds the client fs. For internal journal this will
689 * @j_maxlen: Total maximum capacity of the journal region on disk.
690 * @j_list_lock: Protects the buffer lists and internal buffer state.
691 * @j_inode: Optional inode where we store the journal. If present, all journal
692 * block numbers are mapped into this inode via bmap().
693 * @j_tail_sequence: Sequence number of the oldest transaction in the log
694 * @j_transaction_sequence: Sequence number of the next transaction to grant
695 * @j_commit_sequence: Sequence number of the most recently committed
697 * @j_commit_request: Sequence number of the most recent transaction wanting
699 * @j_uuid: Uuid of client object.
700 * @j_task: Pointer to the current commit thread for this journal
701 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
702 * single compound commit transaction
703 * @j_commit_interval: What is the maximum transaction lifetime before we begin
705 * @j_commit_timer: The timer used to wakeup the commit thread
706 * @j_revoke_lock: Protect the revoke table
707 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
708 * current transaction.
709 * @j_revoke_table: alternate revoke tables for j_revoke
710 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
711 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
712 * number that will fit in j_blocksize
713 * @j_last_sync_writer: most recent pid which did a synchronous write
714 * @j_history: Buffer storing the transactions statistics history
715 * @j_history_max: Maximum number of transactions in the statistics history
716 * @j_history_cur: Current number of transactions in the statistics history
717 * @j_history_lock: Protect the transactions statistics history
718 * @j_proc_entry: procfs entry for the jbd statistics directory
719 * @j_stats: Overall statistics
720 * @j_private: An opaque pointer to fs-private information.
725 /* General journaling state flags [j_state_lock] */
726 unsigned long j_flags;
729 * Is there an outstanding uncleared error on the journal (from a prior
730 * abort)? [j_state_lock]
734 /* The superblock buffer */
735 struct buffer_head *j_sb_buffer;
736 journal_superblock_t *j_superblock;
738 /* Version of the superblock format */
739 int j_format_version;
742 * Protect the various scalars in the journal
744 spinlock_t j_state_lock;
747 * Number of processes waiting to create a barrier lock [j_state_lock]
751 /* The barrier lock itself */
752 struct mutex j_barrier;
755 * Transactions: The current running transaction...
756 * [j_state_lock] [caller holding open handle]
758 transaction_t *j_running_transaction;
761 * the transaction we are pushing to disk
762 * [j_state_lock] [caller holding open handle]
764 transaction_t *j_committing_transaction;
767 * ... and a linked circular list of all transactions waiting for
768 * checkpointing. [j_list_lock]
770 transaction_t *j_checkpoint_transactions;
773 * Wait queue for waiting for a locked transaction to start committing,
774 * or for a barrier lock to be released
776 wait_queue_head_t j_wait_transaction_locked;
778 /* Wait queue for waiting for checkpointing to complete */
779 wait_queue_head_t j_wait_logspace;
781 /* Wait queue for waiting for commit to complete */
782 wait_queue_head_t j_wait_done_commit;
784 /* Wait queue to trigger checkpointing */
785 wait_queue_head_t j_wait_checkpoint;
787 /* Wait queue to trigger commit */
788 wait_queue_head_t j_wait_commit;
790 /* Wait queue to wait for updates to complete */
791 wait_queue_head_t j_wait_updates;
793 /* Semaphore for locking against concurrent checkpoints */
794 struct mutex j_checkpoint_mutex;
797 * Journal head: identifies the first unused block in the journal.
800 unsigned long j_head;
803 * Journal tail: identifies the oldest still-used block in the journal.
806 unsigned long j_tail;
809 * Journal free: how many free blocks are there in the journal?
812 unsigned long j_free;
815 * Journal start and end: the block numbers of the first usable block
816 * and one beyond the last usable block in the journal. [j_state_lock]
818 unsigned long j_first;
819 unsigned long j_last;
822 * Device, blocksize and starting block offset for the location where we
825 struct block_device *j_dev;
827 unsigned long long j_blk_offset;
830 * Device which holds the client fs. For internal journal this will be
833 struct block_device *j_fs_dev;
835 /* Total maximum capacity of the journal region on disk. */
836 unsigned int j_maxlen;
839 * Protects the buffer lists and internal buffer state.
841 spinlock_t j_list_lock;
843 /* Optional inode where we store the journal. If present, all */
844 /* journal block numbers are mapped into this inode via */
846 struct inode *j_inode;
849 * Sequence number of the oldest transaction in the log [j_state_lock]
851 tid_t j_tail_sequence;
854 * Sequence number of the next transaction to grant [j_state_lock]
856 tid_t j_transaction_sequence;
859 * Sequence number of the most recently committed transaction
862 tid_t j_commit_sequence;
865 * Sequence number of the most recent transaction wanting commit
868 tid_t j_commit_request;
871 * Journal uuid: identifies the object (filesystem, LVM volume etc)
872 * backed by this journal. This will eventually be replaced by an array
873 * of uuids, allowing us to index multiple devices within a single
874 * journal and to perform atomic updates across them.
878 /* Pointer to the current commit thread for this journal */
879 struct task_struct *j_task;
882 * Maximum number of metadata buffers to allow in a single compound
885 int j_max_transaction_buffers;
888 * What is the maximum transaction lifetime before we begin a commit?
890 unsigned long j_commit_interval;
892 /* The timer used to wakeup the commit thread: */
893 struct timer_list j_commit_timer;
896 * The revoke table: maintains the list of revoked blocks in the
897 * current transaction. [j_revoke_lock]
899 spinlock_t j_revoke_lock;
900 struct jbd2_revoke_table_s *j_revoke;
901 struct jbd2_revoke_table_s *j_revoke_table[2];
904 * array of bhs for jbd2_journal_commit_transaction
906 struct buffer_head **j_wbuf;
909 pid_t j_last_sync_writer;
914 struct transaction_stats_s *j_history;
918 * Protect the transactions statistics history
920 spinlock_t j_history_lock;
921 struct proc_dir_entry *j_proc_entry;
922 struct transaction_stats_s j_stats;
925 * An opaque pointer to fs-private information. ext3 puts its
926 * superblock pointer here
932 * Journal flag definitions
934 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
935 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
936 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
937 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
938 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
939 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
942 * Function declarations for the journaling transaction and buffer
947 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
948 extern void __jbd2_journal_unfile_buffer(struct journal_head *);
949 extern void __jbd2_journal_refile_buffer(struct journal_head *);
950 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
951 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
952 extern void __journal_free_buffer(struct journal_head *bh);
953 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
954 extern void __journal_clean_data_list(transaction_t *transaction);
956 /* Log buffer allocation */
957 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
958 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
960 /* Commit management */
961 extern void jbd2_journal_commit_transaction(journal_t *);
963 /* Checkpoint list management */
964 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
965 int __jbd2_journal_remove_checkpoint(struct journal_head *);
966 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
970 jbd2_journal_write_metadata_buffer(transaction_t *transaction,
971 struct journal_head *jh_in,
972 struct journal_head **jh_out,
973 unsigned long long blocknr);
975 /* Transaction locking */
976 extern void __wait_on_journal (journal_t *);
981 * We need to lock the journal during transaction state changes so that nobody
982 * ever tries to take a handle on the running transaction while we are in the
983 * middle of moving it to the commit phase. j_state_lock does this.
985 * Note that the locking is completely interrupt unsafe. We never touch
986 * journal structures from interrupts.
989 static inline handle_t *journal_current_handle(void)
991 return current->journal_info;
994 /* The journaling code user interface:
996 * Create and destroy handles
997 * Register buffer modifications against the current transaction.
1000 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1001 extern int jbd2_journal_restart (handle_t *, int nblocks);
1002 extern int jbd2_journal_extend (handle_t *, int nblocks);
1003 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1004 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1005 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1006 extern int jbd2_journal_dirty_data (handle_t *, struct buffer_head *);
1007 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1008 extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1009 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1010 extern void journal_sync_buffer (struct buffer_head *);
1011 extern void jbd2_journal_invalidatepage(journal_t *,
1012 struct page *, unsigned long);
1013 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1014 extern int jbd2_journal_stop(handle_t *);
1015 extern int jbd2_journal_flush (journal_t *);
1016 extern void jbd2_journal_lock_updates (journal_t *);
1017 extern void jbd2_journal_unlock_updates (journal_t *);
1019 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1020 struct block_device *fs_dev,
1021 unsigned long long start, int len, int bsize);
1022 extern journal_t * jbd2_journal_init_inode (struct inode *);
1023 extern int jbd2_journal_update_format (journal_t *);
1024 extern int jbd2_journal_check_used_features
1025 (journal_t *, unsigned long, unsigned long, unsigned long);
1026 extern int jbd2_journal_check_available_features
1027 (journal_t *, unsigned long, unsigned long, unsigned long);
1028 extern int jbd2_journal_set_features
1029 (journal_t *, unsigned long, unsigned long, unsigned long);
1030 extern void jbd2_journal_clear_features
1031 (journal_t *, unsigned long, unsigned long, unsigned long);
1032 extern int jbd2_journal_create (journal_t *);
1033 extern int jbd2_journal_load (journal_t *journal);
1034 extern void jbd2_journal_destroy (journal_t *);
1035 extern int jbd2_journal_recover (journal_t *journal);
1036 extern int jbd2_journal_wipe (journal_t *, int);
1037 extern int jbd2_journal_skip_recovery (journal_t *);
1038 extern void jbd2_journal_update_superblock (journal_t *, int);
1039 extern void __jbd2_journal_abort_hard (journal_t *);
1040 extern void jbd2_journal_abort (journal_t *, int);
1041 extern int jbd2_journal_errno (journal_t *);
1042 extern void jbd2_journal_ack_err (journal_t *);
1043 extern int jbd2_journal_clear_err (journal_t *);
1044 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1045 extern int jbd2_journal_force_commit(journal_t *);
1048 * journal_head management
1050 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1051 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1052 void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1053 void jbd2_journal_put_journal_head(struct journal_head *jh);
1058 extern struct kmem_cache *jbd2_handle_cache;
1060 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1062 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
1065 static inline void jbd2_free_handle(handle_t *handle)
1067 kmem_cache_free(jbd2_handle_cache, handle);
1070 /* Primary revoke support */
1071 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1072 extern int jbd2_journal_init_revoke(journal_t *, int);
1073 extern void jbd2_journal_destroy_revoke_caches(void);
1074 extern int jbd2_journal_init_revoke_caches(void);
1076 extern void jbd2_journal_destroy_revoke(journal_t *);
1077 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1078 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1079 extern void jbd2_journal_write_revoke_records(journal_t *, transaction_t *);
1081 /* Recovery revoke support */
1082 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1083 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1084 extern void jbd2_journal_clear_revoke(journal_t *);
1085 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1088 * The log thread user interface:
1090 * Request space in the current transaction, and force transaction commit
1091 * transitions on demand.
1094 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1095 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1096 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1097 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1098 int jbd2_journal_force_commit_nested(journal_t *journal);
1099 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1100 int jbd2_log_do_checkpoint(journal_t *journal);
1102 void __jbd2_log_wait_for_space(journal_t *journal);
1103 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1104 extern int jbd2_cleanup_journal_tail(journal_t *);
1106 /* Debugging code only: */
1108 #define jbd_ENOSYS() \
1110 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1111 current->state = TASK_UNINTERRUPTIBLE; \
1118 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1119 * bit, when set, indicates that we have had a fatal error somewhere,
1120 * either inside the journaling layer or indicated to us by the client
1121 * (eg. ext3), and that we and should not commit any further
1125 static inline int is_journal_aborted(journal_t *journal)
1127 return journal->j_flags & JBD2_ABORT;
1130 static inline int is_handle_aborted(handle_t *handle)
1132 if (handle->h_aborted)
1134 return is_journal_aborted(handle->h_transaction->t_journal);
1137 static inline void jbd2_journal_abort_handle(handle_t *handle)
1139 handle->h_aborted = 1;
1142 #endif /* __KERNEL__ */
1144 /* Comparison functions for transaction IDs: perform comparisons using
1145 * modulo arithmetic so that they work over sequence number wraps. */
1147 static inline int tid_gt(tid_t x, tid_t y)
1149 int difference = (x - y);
1150 return (difference > 0);
1153 static inline int tid_geq(tid_t x, tid_t y)
1155 int difference = (x - y);
1156 return (difference >= 0);
1159 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1160 extern size_t journal_tag_bytes(journal_t *journal);
1163 * Return the minimum number of blocks which must be free in the journal
1164 * before a new transaction may be started. Must be called under j_state_lock.
1166 static inline int jbd_space_needed(journal_t *journal)
1168 int nblocks = journal->j_max_transaction_buffers;
1169 if (journal->j_committing_transaction)
1170 nblocks += journal->j_committing_transaction->
1171 t_outstanding_credits;
1176 * Definitions which augment the buffer_head layer
1179 /* journaling buffer types */
1180 #define BJ_None 0 /* Not journaled */
1181 #define BJ_SyncData 1 /* Normal data: flush before commit */
1182 #define BJ_Metadata 2 /* Normal journaled metadata */
1183 #define BJ_Forget 3 /* Buffer superseded by this transaction */
1184 #define BJ_IO 4 /* Buffer is for temporary IO use */
1185 #define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1186 #define BJ_LogCtl 6 /* Buffer contains log descriptors */
1187 #define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1188 #define BJ_Locked 8 /* Locked for I/O during commit */
1191 extern int jbd_blocks_per_page(struct inode *inode);
1195 #define buffer_trace_init(bh) do {} while (0)
1196 #define print_buffer_fields(bh) do {} while (0)
1197 #define print_buffer_trace(bh) do {} while (0)
1198 #define BUFFER_TRACE(bh, info) do {} while (0)
1199 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1200 #define JBUFFER_TRACE(jh, info) do {} while (0)
1202 #endif /* __KERNEL__ */
1204 #endif /* _LINUX_JBD2_H */