2 * linux/include/linux/jbd.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 <asm/semaphore.h>
34 #define journal_oom_retry 1
37 * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
38 * certain classes of error which can occur due to failed IOs. Under
39 * normal use we want ext3 to continue after such errors, because
40 * hardware _can_ fail, but for debugging purposes when running tests on
41 * known-good hardware we may want to trap these errors.
43 #undef JBD_PARANOID_IOFAIL
46 * The default maximum commit age, in seconds.
48 #define JBD_DEFAULT_MAX_COMMIT_AGE 5
50 #ifdef CONFIG_JBD_DEBUG
52 * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
53 * consistency checks. By default we don't do this unless
54 * CONFIG_JBD_DEBUG is on.
56 #define JBD_EXPENSIVE_CHECKING
57 extern int journal_enable_debug;
59 #define jbd_debug(n, f, a...) \
61 if ((n) <= journal_enable_debug) { \
62 printk (KERN_DEBUG "(%s, %d): %s: ", \
63 __FILE__, __LINE__, __FUNCTION__); \
68 #define jbd_debug(f, a...) /**/
71 extern void * __jbd_kmalloc (const char *where, size_t size, gfp_t flags, int retry);
72 #define jbd_kmalloc(size, flags) \
73 __jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
74 #define jbd_rep_kmalloc(size, flags) \
75 __jbd_kmalloc(__FUNCTION__, (size), (flags), 1)
77 #define JFS_MIN_JOURNAL_BLOCKS 1024
82 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
84 * All filesystem modifications made by the process go
85 * through this handle. Recursive operations (such as quota operations)
86 * are gathered into a single update.
88 * The buffer credits field is used to account for journaled buffers
89 * being modified by the running process. To ensure that there is
90 * enough log space for all outstanding operations, we need to limit the
91 * number of outstanding buffers possible at any time. When the
92 * operation completes, any buffer credits not used are credited back to
93 * the transaction, so that at all times we know how many buffers the
94 * outstanding updates on a transaction might possibly touch.
96 * This is an opaque datatype.
98 typedef struct handle_s handle_t; /* Atomic operation type */
102 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
104 * journal_t is linked to from the fs superblock structure.
106 * We use the journal_t to keep track of all outstanding transaction
107 * activity on the filesystem, and to manage the state of the log
110 * This is an opaque datatype.
112 typedef struct journal_s journal_t; /* Journal control structure */
116 * Internal structures used by the logging mechanism:
119 #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
126 * Descriptor block types:
129 #define JFS_DESCRIPTOR_BLOCK 1
130 #define JFS_COMMIT_BLOCK 2
131 #define JFS_SUPERBLOCK_V1 3
132 #define JFS_SUPERBLOCK_V2 4
133 #define JFS_REVOKE_BLOCK 5
136 * Standard header for all descriptor blocks:
138 typedef struct journal_header_s
147 * The block tag: used to describe a single buffer in the journal
149 typedef struct journal_block_tag_s
151 __be32 t_blocknr; /* The on-disk block number */
152 __be32 t_flags; /* See below */
153 } journal_block_tag_t;
156 * The revoke descriptor: used on disk to describe a series of blocks to
157 * be revoked from the log
159 typedef struct journal_revoke_header_s
161 journal_header_t r_header;
162 __be32 r_count; /* Count of bytes used in the block */
163 } journal_revoke_header_t;
166 /* Definitions for the journal tag flags word: */
167 #define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */
168 #define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */
169 #define JFS_FLAG_DELETED 4 /* block deleted by this transaction */
170 #define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
174 * The journal superblock. All fields are in big-endian byte order.
176 typedef struct journal_superblock_s
179 journal_header_t s_header;
182 /* Static information describing the journal */
183 __be32 s_blocksize; /* journal device blocksize */
184 __be32 s_maxlen; /* total blocks in journal file */
185 __be32 s_first; /* first block of log information */
188 /* Dynamic information describing the current state of the log */
189 __be32 s_sequence; /* first commit ID expected in log */
190 __be32 s_start; /* blocknr of start of log */
193 /* Error value, as set by journal_abort(). */
197 /* Remaining fields are only valid in a version-2 superblock */
198 __be32 s_feature_compat; /* compatible feature set */
199 __be32 s_feature_incompat; /* incompatible feature set */
200 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
202 __u8 s_uuid[16]; /* 128-bit uuid for journal */
205 __be32 s_nr_users; /* Nr of filesystems sharing log */
207 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
210 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
211 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
217 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
219 } journal_superblock_t;
221 #define JFS_HAS_COMPAT_FEATURE(j,mask) \
222 ((j)->j_format_version >= 2 && \
223 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
224 #define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \
225 ((j)->j_format_version >= 2 && \
226 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
227 #define JFS_HAS_INCOMPAT_FEATURE(j,mask) \
228 ((j)->j_format_version >= 2 && \
229 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
231 #define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001
233 /* Features known to this kernel version: */
234 #define JFS_KNOWN_COMPAT_FEATURES 0
235 #define JFS_KNOWN_ROCOMPAT_FEATURES 0
236 #define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE
240 #include <linux/fs.h>
241 #include <linux/sched.h>
244 #define JBD_ASSERTIONS
245 #ifdef JBD_ASSERTIONS
246 #define J_ASSERT(assert) \
250 "Assertion failure in %s() at %s:%d: \"%s\"\n", \
251 __FUNCTION__, __FILE__, __LINE__, # assert); \
256 #if defined(CONFIG_BUFFER_DEBUG)
257 void buffer_assertion_failure(struct buffer_head *bh);
258 #define J_ASSERT_BH(bh, expr) \
261 buffer_assertion_failure(bh); \
264 #define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
266 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
267 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
271 #define J_ASSERT(assert) do { } while (0)
272 #endif /* JBD_ASSERTIONS */
274 #if defined(JBD_PARANOID_IOFAIL)
275 #define J_EXPECT(expr, why...) J_ASSERT(expr)
276 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
277 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
279 #define __journal_expect(expr, why...) \
284 "EXT3-fs unexpected failure: %s;\n",# expr); \
285 printk(KERN_ERR why "\n"); \
289 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
290 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
291 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
294 enum jbd_state_bits {
295 BH_JBD /* Has an attached ext3 journal_head */
297 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
298 BH_Freed, /* Has been freed (truncated) */
299 BH_Revoked, /* Has been revoked from the log */
300 BH_RevokeValid, /* Revoked flag is valid */
301 BH_JBDDirty, /* Is dirty but journaled */
302 BH_State, /* Pins most journal_head state */
303 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
304 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
308 BUFFER_FNS(JWrite, jwrite)
309 BUFFER_FNS(JBDDirty, jbddirty)
310 TAS_BUFFER_FNS(JBDDirty, jbddirty)
311 BUFFER_FNS(Revoked, revoked)
312 TAS_BUFFER_FNS(Revoked, revoked)
313 BUFFER_FNS(RevokeValid, revokevalid)
314 TAS_BUFFER_FNS(RevokeValid, revokevalid)
315 BUFFER_FNS(Freed, freed)
317 static inline struct buffer_head *jh2bh(struct journal_head *jh)
322 static inline struct journal_head *bh2jh(struct buffer_head *bh)
324 return bh->b_private;
327 static inline void jbd_lock_bh_state(struct buffer_head *bh)
329 bit_spin_lock(BH_State, &bh->b_state);
332 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
334 return bit_spin_trylock(BH_State, &bh->b_state);
337 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
339 return bit_spin_is_locked(BH_State, &bh->b_state);
342 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
344 bit_spin_unlock(BH_State, &bh->b_state);
347 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
349 bit_spin_lock(BH_JournalHead, &bh->b_state);
352 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
354 bit_spin_unlock(BH_JournalHead, &bh->b_state);
357 struct jbd_revoke_table_s;
360 * struct handle_s - The handle_s type is the concrete type associated with
362 * @h_transaction: Which compound transaction is this update a part of?
363 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
364 * @h_ref: Reference count on this handle
365 * @h_err: Field for caller's use to track errors through large fs operations
366 * @h_sync: flag for sync-on-close
367 * @h_jdata: flag to force data journaling
368 * @h_aborted: flag indicating fatal error on handle
371 /* Docbook can't yet cope with the bit fields, but will leave the documentation
372 * in so it can be fixed later.
377 /* Which compound transaction is this update a part of? */
378 transaction_t *h_transaction;
380 /* Number of remaining buffers we are allowed to dirty: */
381 int h_buffer_credits;
383 /* Reference count on this handle */
386 /* Field for caller's use to track errors through large fs */
390 /* Flags [no locking] */
391 unsigned int h_sync: 1; /* sync-on-close */
392 unsigned int h_jdata: 1; /* force data journaling */
393 unsigned int h_aborted: 1; /* fatal error on handle */
397 /* The transaction_t type is the guts of the journaling mechanism. It
398 * tracks a compound transaction through its various states:
400 * RUNNING: accepting new updates
401 * LOCKED: Updates still running but we don't accept new ones
402 * RUNDOWN: Updates are tidying up but have finished requesting
403 * new buffers to modify (state not used for now)
404 * FLUSH: All updates complete, but we are still writing to disk
405 * COMMIT: All data on disk, writing commit record
406 * FINISHED: We still have to keep the transaction for checkpointing.
408 * The transaction keeps track of all of the buffers modified by a
409 * running transaction, and all of the buffers committed but not yet
410 * flushed to home for finished transactions.
417 * ->jbd_lock_bh_journal_head() (This is "innermost")
420 * ->jbd_lock_bh_state()
422 * jbd_lock_bh_state()
429 * ->j_list_lock (journal_unmap_buffer)
435 /* Pointer to the journal for this transaction. [no locking] */
436 journal_t *t_journal;
438 /* Sequence number for this transaction [no locking] */
442 * Transaction's current state
443 * [no locking - only kjournald alters this]
444 * FIXME: needs barriers
445 * KLUDGE: [use j_state_lock]
457 * Where in the log does this transaction's commit start? [no locking]
459 unsigned long t_log_start;
461 /* Number of buffers on the t_buffers list [j_list_lock] */
465 * Doubly-linked circular list of all buffers reserved but not yet
466 * modified by this transaction [j_list_lock]
468 struct journal_head *t_reserved_list;
471 * Doubly-linked circular list of all buffers under writeout during
472 * commit [j_list_lock]
474 struct journal_head *t_locked_list;
477 * Doubly-linked circular list of all metadata buffers owned by this
478 * transaction [j_list_lock]
480 struct journal_head *t_buffers;
483 * Doubly-linked circular list of all data buffers still to be
484 * flushed before this transaction can be committed [j_list_lock]
486 struct journal_head *t_sync_datalist;
489 * Doubly-linked circular list of all forget buffers (superseded
490 * buffers which we can un-checkpoint once this transaction commits)
493 struct journal_head *t_forget;
496 * Doubly-linked circular list of all buffers still to be flushed before
497 * this transaction can be checkpointed. [j_list_lock]
499 struct journal_head *t_checkpoint_list;
502 * Doubly-linked circular list of all buffers submitted for IO while
503 * checkpointing. [j_list_lock]
505 struct journal_head *t_checkpoint_io_list;
508 * Doubly-linked circular list of temporary buffers currently undergoing
509 * IO in the log [j_list_lock]
511 struct journal_head *t_iobuf_list;
514 * Doubly-linked circular list of metadata buffers being shadowed by log
515 * IO. The IO buffers on the iobuf list and the shadow buffers on this
516 * list match each other one for one at all times. [j_list_lock]
518 struct journal_head *t_shadow_list;
521 * Doubly-linked circular list of control buffers being written to the
524 struct journal_head *t_log_list;
527 * Protects info related to handles
529 spinlock_t t_handle_lock;
532 * Number of outstanding updates running on this transaction
538 * Number of buffers reserved for use by all handles in this transaction
539 * handle but not yet modified. [t_handle_lock]
541 int t_outstanding_credits;
544 * Forward and backward links for the circular list of all transactions
545 * awaiting checkpoint. [j_list_lock]
547 transaction_t *t_cpnext, *t_cpprev;
550 * When will the transaction expire (become due for commit), in jiffies?
553 unsigned long t_expires;
556 * How many handles used this transaction? [t_handle_lock]
563 * struct journal_s - The journal_s type is the concrete type associated with
565 * @j_flags: General journaling state flags
566 * @j_errno: Is there an outstanding uncleared error on the journal (from a
568 * @j_sb_buffer: First part of superblock buffer
569 * @j_superblock: Second part of superblock buffer
570 * @j_format_version: Version of the superblock format
571 * @j_state_lock: Protect the various scalars in the journal
572 * @j_barrier_count: Number of processes waiting to create a barrier lock
573 * @j_barrier: The barrier lock itself
574 * @j_running_transaction: The current running transaction..
575 * @j_committing_transaction: the transaction we are pushing to disk
576 * @j_checkpoint_transactions: a linked circular list of all transactions
577 * waiting for checkpointing
578 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
579 * to start committing, or for a barrier lock to be released
580 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
581 * @j_wait_done_commit: Wait queue for waiting for commit to complete
582 * @j_wait_checkpoint: Wait queue to trigger checkpointing
583 * @j_wait_commit: Wait queue to trigger commit
584 * @j_wait_updates: Wait queue to wait for updates to complete
585 * @j_checkpoint_sem: Semaphore for locking against concurrent checkpoints
586 * @j_head: Journal head - identifies the first unused block in the journal
587 * @j_tail: Journal tail - identifies the oldest still-used block in the
589 * @j_free: Journal free - how many free blocks are there in the journal?
590 * @j_first: The block number of the first usable block
591 * @j_last: The block number one beyond the last usable block
592 * @j_dev: Device where we store the journal
593 * @j_blocksize: blocksize for the location where we store the journal.
594 * @j_blk_offset: starting block offset for into the device where we store the
596 * @j_fs_dev: Device which holds the client fs. For internal journal this will
598 * @j_maxlen: Total maximum capacity of the journal region on disk.
599 * @j_list_lock: Protects the buffer lists and internal buffer state.
600 * @j_inode: Optional inode where we store the journal. If present, all journal
601 * block numbers are mapped into this inode via bmap().
602 * @j_tail_sequence: Sequence number of the oldest transaction in the log
603 * @j_transaction_sequence: Sequence number of the next transaction to grant
604 * @j_commit_sequence: Sequence number of the most recently committed
606 * @j_commit_request: Sequence number of the most recent transaction wanting
608 * @j_uuid: Uuid of client object.
609 * @j_task: Pointer to the current commit thread for this journal
610 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
611 * single compound commit transaction
612 * @j_commit_interval: What is the maximum transaction lifetime before we begin
614 * @j_commit_timer: The timer used to wakeup the commit thread
615 * @j_revoke_lock: Protect the revoke table
616 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
617 * current transaction.
618 * @j_revoke_table: alternate revoke tables for j_revoke
619 * @j_wbuf: array of buffer_heads for journal_commit_transaction
620 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
621 * number that will fit in j_blocksize
622 * @j_last_sync_writer: most recent pid which did a synchronous write
623 * @j_private: An opaque pointer to fs-private information.
628 /* General journaling state flags [j_state_lock] */
629 unsigned long j_flags;
632 * Is there an outstanding uncleared error on the journal (from a prior
633 * abort)? [j_state_lock]
637 /* The superblock buffer */
638 struct buffer_head *j_sb_buffer;
639 journal_superblock_t *j_superblock;
641 /* Version of the superblock format */
642 int j_format_version;
645 * Protect the various scalars in the journal
647 spinlock_t j_state_lock;
650 * Number of processes waiting to create a barrier lock [j_state_lock]
654 /* The barrier lock itself */
655 struct semaphore j_barrier;
658 * Transactions: The current running transaction...
659 * [j_state_lock] [caller holding open handle]
661 transaction_t *j_running_transaction;
664 * the transaction we are pushing to disk
665 * [j_state_lock] [caller holding open handle]
667 transaction_t *j_committing_transaction;
670 * ... and a linked circular list of all transactions waiting for
671 * checkpointing. [j_list_lock]
673 transaction_t *j_checkpoint_transactions;
676 * Wait queue for waiting for a locked transaction to start committing,
677 * or for a barrier lock to be released
679 wait_queue_head_t j_wait_transaction_locked;
681 /* Wait queue for waiting for checkpointing to complete */
682 wait_queue_head_t j_wait_logspace;
684 /* Wait queue for waiting for commit to complete */
685 wait_queue_head_t j_wait_done_commit;
687 /* Wait queue to trigger checkpointing */
688 wait_queue_head_t j_wait_checkpoint;
690 /* Wait queue to trigger commit */
691 wait_queue_head_t j_wait_commit;
693 /* Wait queue to wait for updates to complete */
694 wait_queue_head_t j_wait_updates;
696 /* Semaphore for locking against concurrent checkpoints */
697 struct semaphore j_checkpoint_sem;
700 * Journal head: identifies the first unused block in the journal.
703 unsigned long j_head;
706 * Journal tail: identifies the oldest still-used block in the journal.
709 unsigned long j_tail;
712 * Journal free: how many free blocks are there in the journal?
715 unsigned long j_free;
718 * Journal start and end: the block numbers of the first usable block
719 * and one beyond the last usable block in the journal. [j_state_lock]
721 unsigned long j_first;
722 unsigned long j_last;
725 * Device, blocksize and starting block offset for the location where we
728 struct block_device *j_dev;
730 unsigned int j_blk_offset;
733 * Device which holds the client fs. For internal journal this will be
736 struct block_device *j_fs_dev;
738 /* Total maximum capacity of the journal region on disk. */
739 unsigned int j_maxlen;
742 * Protects the buffer lists and internal buffer state.
744 spinlock_t j_list_lock;
746 /* Optional inode where we store the journal. If present, all */
747 /* journal block numbers are mapped into this inode via */
749 struct inode *j_inode;
752 * Sequence number of the oldest transaction in the log [j_state_lock]
754 tid_t j_tail_sequence;
757 * Sequence number of the next transaction to grant [j_state_lock]
759 tid_t j_transaction_sequence;
762 * Sequence number of the most recently committed transaction
765 tid_t j_commit_sequence;
768 * Sequence number of the most recent transaction wanting commit
771 tid_t j_commit_request;
774 * Journal uuid: identifies the object (filesystem, LVM volume etc)
775 * backed by this journal. This will eventually be replaced by an array
776 * of uuids, allowing us to index multiple devices within a single
777 * journal and to perform atomic updates across them.
781 /* Pointer to the current commit thread for this journal */
782 struct task_struct *j_task;
785 * Maximum number of metadata buffers to allow in a single compound
788 int j_max_transaction_buffers;
791 * What is the maximum transaction lifetime before we begin a commit?
793 unsigned long j_commit_interval;
795 /* The timer used to wakeup the commit thread: */
796 struct timer_list *j_commit_timer;
799 * The revoke table: maintains the list of revoked blocks in the
800 * current transaction. [j_revoke_lock]
802 spinlock_t j_revoke_lock;
803 struct jbd_revoke_table_s *j_revoke;
804 struct jbd_revoke_table_s *j_revoke_table[2];
807 * array of bhs for journal_commit_transaction
809 struct buffer_head **j_wbuf;
812 pid_t j_last_sync_writer;
815 * An opaque pointer to fs-private information. ext3 puts its
816 * superblock pointer here
822 * Journal flag definitions
824 #define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
825 #define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
826 #define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
827 #define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
828 #define JFS_LOADED 0x010 /* The journal superblock has been loaded */
829 #define JFS_BARRIER 0x020 /* Use IDE barriers */
832 * Function declarations for the journaling transaction and buffer
837 extern void __journal_temp_unlink_buffer(struct journal_head *jh);
838 extern void journal_unfile_buffer(journal_t *, struct journal_head *);
839 extern void __journal_unfile_buffer(struct journal_head *);
840 extern void __journal_refile_buffer(struct journal_head *);
841 extern void journal_refile_buffer(journal_t *, struct journal_head *);
842 extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
843 extern void __journal_free_buffer(struct journal_head *bh);
844 extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
845 extern void __journal_clean_data_list(transaction_t *transaction);
847 /* Log buffer allocation */
848 extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
849 int journal_next_log_block(journal_t *, unsigned long *);
851 /* Commit management */
852 extern void journal_commit_transaction(journal_t *);
854 /* Checkpoint list management */
855 int __journal_clean_checkpoint_list(journal_t *journal);
856 int __journal_remove_checkpoint(struct journal_head *);
857 void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
861 journal_write_metadata_buffer(transaction_t *transaction,
862 struct journal_head *jh_in,
863 struct journal_head **jh_out,
866 /* Transaction locking */
867 extern void __wait_on_journal (journal_t *);
872 * We need to lock the journal during transaction state changes so that nobody
873 * ever tries to take a handle on the running transaction while we are in the
874 * middle of moving it to the commit phase. j_state_lock does this.
876 * Note that the locking is completely interrupt unsafe. We never touch
877 * journal structures from interrupts.
880 static inline handle_t *journal_current_handle(void)
882 return current->journal_info;
885 /* The journaling code user interface:
887 * Create and destroy handles
888 * Register buffer modifications against the current transaction.
891 extern handle_t *journal_start(journal_t *, int nblocks);
892 extern int journal_restart (handle_t *, int nblocks);
893 extern int journal_extend (handle_t *, int nblocks);
894 extern int journal_get_write_access(handle_t *, struct buffer_head *);
895 extern int journal_get_create_access (handle_t *, struct buffer_head *);
896 extern int journal_get_undo_access(handle_t *, struct buffer_head *);
897 extern int journal_dirty_data (handle_t *, struct buffer_head *);
898 extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
899 extern void journal_release_buffer (handle_t *, struct buffer_head *);
900 extern int journal_forget (handle_t *, struct buffer_head *);
901 extern void journal_sync_buffer (struct buffer_head *);
902 extern int journal_invalidatepage(journal_t *,
903 struct page *, unsigned long);
904 extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
905 extern int journal_stop(handle_t *);
906 extern int journal_flush (journal_t *);
907 extern void journal_lock_updates (journal_t *);
908 extern void journal_unlock_updates (journal_t *);
910 extern journal_t * journal_init_dev(struct block_device *bdev,
911 struct block_device *fs_dev,
912 int start, int len, int bsize);
913 extern journal_t * journal_init_inode (struct inode *);
914 extern int journal_update_format (journal_t *);
915 extern int journal_check_used_features
916 (journal_t *, unsigned long, unsigned long, unsigned long);
917 extern int journal_check_available_features
918 (journal_t *, unsigned long, unsigned long, unsigned long);
919 extern int journal_set_features
920 (journal_t *, unsigned long, unsigned long, unsigned long);
921 extern int journal_create (journal_t *);
922 extern int journal_load (journal_t *journal);
923 extern void journal_destroy (journal_t *);
924 extern int journal_recover (journal_t *journal);
925 extern int journal_wipe (journal_t *, int);
926 extern int journal_skip_recovery (journal_t *);
927 extern void journal_update_superblock (journal_t *, int);
928 extern void __journal_abort_hard (journal_t *);
929 extern void journal_abort (journal_t *, int);
930 extern int journal_errno (journal_t *);
931 extern void journal_ack_err (journal_t *);
932 extern int journal_clear_err (journal_t *);
933 extern int journal_bmap(journal_t *, unsigned long, unsigned long *);
934 extern int journal_force_commit(journal_t *);
937 * journal_head management
939 struct journal_head *journal_add_journal_head(struct buffer_head *bh);
940 struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
941 void journal_remove_journal_head(struct buffer_head *bh);
942 void journal_put_journal_head(struct journal_head *jh);
947 extern kmem_cache_t *jbd_handle_cache;
949 static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
951 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
954 static inline void jbd_free_handle(handle_t *handle)
956 kmem_cache_free(jbd_handle_cache, handle);
959 /* Primary revoke support */
960 #define JOURNAL_REVOKE_DEFAULT_HASH 256
961 extern int journal_init_revoke(journal_t *, int);
962 extern void journal_destroy_revoke_caches(void);
963 extern int journal_init_revoke_caches(void);
965 extern void journal_destroy_revoke(journal_t *);
966 extern int journal_revoke (handle_t *,
967 unsigned long, struct buffer_head *);
968 extern int journal_cancel_revoke(handle_t *, struct journal_head *);
969 extern void journal_write_revoke_records(journal_t *, transaction_t *);
971 /* Recovery revoke support */
972 extern int journal_set_revoke(journal_t *, unsigned long, tid_t);
973 extern int journal_test_revoke(journal_t *, unsigned long, tid_t);
974 extern void journal_clear_revoke(journal_t *);
975 extern void journal_brelse_array(struct buffer_head *b[], int n);
976 extern void journal_switch_revoke_table(journal_t *journal);
979 * The log thread user interface:
981 * Request space in the current transaction, and force transaction commit
982 * transitions on demand.
985 int __log_space_left(journal_t *); /* Called with journal locked */
986 int log_start_commit(journal_t *journal, tid_t tid);
987 int __log_start_commit(journal_t *journal, tid_t tid);
988 int journal_start_commit(journal_t *journal, tid_t *tid);
989 int journal_force_commit_nested(journal_t *journal);
990 int log_wait_commit(journal_t *journal, tid_t tid);
991 int log_do_checkpoint(journal_t *journal);
993 void __log_wait_for_space(journal_t *journal);
994 extern void __journal_drop_transaction(journal_t *, transaction_t *);
995 extern int cleanup_journal_tail(journal_t *);
997 /* Debugging code only: */
999 #define jbd_ENOSYS() \
1001 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1002 current->state = TASK_UNINTERRUPTIBLE; \
1009 * Simple test wrapper function to test the JFS_ABORT state flag. This
1010 * bit, when set, indicates that we have had a fatal error somewhere,
1011 * either inside the journaling layer or indicated to us by the client
1012 * (eg. ext3), and that we and should not commit any further
1016 static inline int is_journal_aborted(journal_t *journal)
1018 return journal->j_flags & JFS_ABORT;
1021 static inline int is_handle_aborted(handle_t *handle)
1023 if (handle->h_aborted)
1025 return is_journal_aborted(handle->h_transaction->t_journal);
1028 static inline void journal_abort_handle(handle_t *handle)
1030 handle->h_aborted = 1;
1033 #endif /* __KERNEL__ */
1035 /* Comparison functions for transaction IDs: perform comparisons using
1036 * modulo arithmetic so that they work over sequence number wraps. */
1038 static inline int tid_gt(tid_t x, tid_t y)
1040 int difference = (x - y);
1041 return (difference > 0);
1044 static inline int tid_geq(tid_t x, tid_t y)
1046 int difference = (x - y);
1047 return (difference >= 0);
1050 extern int journal_blocks_per_page(struct inode *inode);
1053 * Return the minimum number of blocks which must be free in the journal
1054 * before a new transaction may be started. Must be called under j_state_lock.
1056 static inline int jbd_space_needed(journal_t *journal)
1058 int nblocks = journal->j_max_transaction_buffers;
1059 if (journal->j_committing_transaction)
1060 nblocks += journal->j_committing_transaction->
1061 t_outstanding_credits;
1066 * Definitions which augment the buffer_head layer
1069 /* journaling buffer types */
1070 #define BJ_None 0 /* Not journaled */
1071 #define BJ_SyncData 1 /* Normal data: flush before commit */
1072 #define BJ_Metadata 2 /* Normal journaled metadata */
1073 #define BJ_Forget 3 /* Buffer superseded by this transaction */
1074 #define BJ_IO 4 /* Buffer is for temporary IO use */
1075 #define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1076 #define BJ_LogCtl 6 /* Buffer contains log descriptors */
1077 #define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1078 #define BJ_Locked 8 /* Locked for I/O during commit */
1081 extern int jbd_blocks_per_page(struct inode *inode);
1085 #define buffer_trace_init(bh) do {} while (0)
1086 #define print_buffer_fields(bh) do {} while (0)
1087 #define print_buffer_trace(bh) do {} while (0)
1088 #define BUFFER_TRACE(bh, info) do {} while (0)
1089 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1090 #define JBUFFER_TRACE(jh, info) do {} while (0)
1092 #endif /* __KERNEL__ */
1094 #endif /* _LINUX_JBD_H */