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>
35 #define journal_oom_retry 1
38 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
39 * certain classes of error which can occur due to failed IOs. Under
40 * normal use we want ext4 to continue after such errors, because
41 * hardware _can_ fail, but for debugging purposes when running tests on
42 * known-good hardware we may want to trap these errors.
44 #undef JBD2_PARANOID_IOFAIL
47 * The default maximum commit age, in seconds.
49 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
51 #ifdef CONFIG_JBD2_DEBUG
53 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
54 * consistency checks. By default we don't do this unless
55 * CONFIG_JBD2_DEBUG is on.
57 #define JBD2_EXPENSIVE_CHECKING
58 extern u8 jbd2_journal_enable_debug
;
60 #define jbd_debug(n, f, a...) \
62 if ((n) <= jbd2_journal_enable_debug) { \
63 printk (KERN_DEBUG "(%s, %d): %s: ", \
64 __FILE__, __LINE__, __func__); \
69 #define jbd_debug(f, a...) /**/
72 static inline void *jbd2_alloc(size_t size
, gfp_t flags
)
74 return (void *)__get_free_pages(flags
, get_order(size
));
77 static inline void jbd2_free(void *ptr
, size_t size
)
79 free_pages((unsigned long)ptr
, get_order(size
));
82 #define JBD2_MIN_JOURNAL_BLOCKS 1024
87 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
89 * All filesystem modifications made by the process go
90 * through this handle. Recursive operations (such as quota operations)
91 * are gathered into a single update.
93 * The buffer credits field is used to account for journaled buffers
94 * being modified by the running process. To ensure that there is
95 * enough log space for all outstanding operations, we need to limit the
96 * number of outstanding buffers possible at any time. When the
97 * operation completes, any buffer credits not used are credited back to
98 * the transaction, so that at all times we know how many buffers the
99 * outstanding updates on a transaction might possibly touch.
101 * This is an opaque datatype.
103 typedef struct handle_s handle_t
; /* Atomic operation type */
107 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
109 * journal_t is linked to from the fs superblock structure.
111 * We use the journal_t to keep track of all outstanding transaction
112 * activity on the filesystem, and to manage the state of the log
115 * This is an opaque datatype.
117 typedef struct journal_s journal_t
; /* Journal control structure */
121 * Internal structures used by the logging mechanism:
124 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
131 * Descriptor block types:
134 #define JBD2_DESCRIPTOR_BLOCK 1
135 #define JBD2_COMMIT_BLOCK 2
136 #define JBD2_SUPERBLOCK_V1 3
137 #define JBD2_SUPERBLOCK_V2 4
138 #define JBD2_REVOKE_BLOCK 5
141 * Standard header for all descriptor blocks:
143 typedef struct journal_header_s
153 #define JBD2_CRC32_CHKSUM 1
154 #define JBD2_MD5_CHKSUM 2
155 #define JBD2_SHA1_CHKSUM 3
157 #define JBD2_CRC32_CHKSUM_SIZE 4
159 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
161 * Commit block header for storing transactional checksums:
163 struct commit_header
{
167 unsigned char h_chksum_type
;
168 unsigned char h_chksum_size
;
169 unsigned char h_padding
[2];
170 __be32 h_chksum
[JBD2_CHECKSUM_BYTES
];
172 __be32 h_commit_nsec
;
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 "JBD2 unexpected failure: %s: %s;\n", \
313 printk(KERN_ERR why "\n"); \
317 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
318 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
319 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
322 enum jbd_state_bits
{
323 BH_JBD
/* Has an attached ext3 journal_head */
325 BH_JWrite
, /* Being written to log (@@@ DEBUGGING) */
326 BH_Freed
, /* Has been freed (truncated) */
327 BH_Revoked
, /* Has been revoked from the log */
328 BH_RevokeValid
, /* Revoked flag is valid */
329 BH_JBDDirty
, /* Is dirty but journaled */
330 BH_State
, /* Pins most journal_head state */
331 BH_JournalHead
, /* Pins bh->b_private and jh->b_bh */
332 BH_Unshadow
, /* Dummy bit, for BJ_Shadow wakeup filtering */
333 BH_JBDPrivateStart
, /* First bit available for private use by FS */
337 BUFFER_FNS(JWrite
, jwrite
)
338 BUFFER_FNS(JBDDirty
, jbddirty
)
339 TAS_BUFFER_FNS(JBDDirty
, jbddirty
)
340 BUFFER_FNS(Revoked
, revoked
)
341 TAS_BUFFER_FNS(Revoked
, revoked
)
342 BUFFER_FNS(RevokeValid
, revokevalid
)
343 TAS_BUFFER_FNS(RevokeValid
, revokevalid
)
344 BUFFER_FNS(Freed
, freed
)
346 static inline struct buffer_head
*jh2bh(struct journal_head
*jh
)
351 static inline struct journal_head
*bh2jh(struct buffer_head
*bh
)
353 return bh
->b_private
;
356 static inline void jbd_lock_bh_state(struct buffer_head
*bh
)
358 bit_spin_lock(BH_State
, &bh
->b_state
);
361 static inline int jbd_trylock_bh_state(struct buffer_head
*bh
)
363 return bit_spin_trylock(BH_State
, &bh
->b_state
);
366 static inline int jbd_is_locked_bh_state(struct buffer_head
*bh
)
368 return bit_spin_is_locked(BH_State
, &bh
->b_state
);
371 static inline void jbd_unlock_bh_state(struct buffer_head
*bh
)
373 bit_spin_unlock(BH_State
, &bh
->b_state
);
376 static inline void jbd_lock_bh_journal_head(struct buffer_head
*bh
)
378 bit_spin_lock(BH_JournalHead
, &bh
->b_state
);
381 static inline void jbd_unlock_bh_journal_head(struct buffer_head
*bh
)
383 bit_spin_unlock(BH_JournalHead
, &bh
->b_state
);
386 /* Flags in jbd_inode->i_flags */
387 #define __JI_COMMIT_RUNNING 0
388 /* Commit of the inode data in progress. We use this flag to protect us from
389 * concurrent deletion of inode. We cannot use reference to inode for this
390 * since we cannot afford doing last iput() on behalf of kjournald
392 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
395 * struct jbd_inode is the structure linking inodes in ordered mode
396 * present in a transaction so that we can sync them during commit.
399 /* Which transaction does this inode belong to? Either the running
400 * transaction or the committing one. [j_list_lock] */
401 transaction_t
*i_transaction
;
403 /* Pointer to the running transaction modifying inode's data in case
404 * there is already a committing transaction touching it. [j_list_lock] */
405 transaction_t
*i_next_transaction
;
407 /* List of inodes in the i_transaction [j_list_lock] */
408 struct list_head i_list
;
410 /* VFS inode this inode belongs to [constant during the lifetime
411 * of the structure] */
412 struct inode
*i_vfs_inode
;
414 /* Flags of inode [j_list_lock] */
415 unsigned int i_flags
;
418 struct jbd2_revoke_table_s
;
421 * struct handle_s - The handle_s type is the concrete type associated with
423 * @h_transaction: Which compound transaction is this update a part of?
424 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
425 * @h_ref: Reference count on this handle
426 * @h_err: Field for caller's use to track errors through large fs operations
427 * @h_sync: flag for sync-on-close
428 * @h_jdata: flag to force data journaling
429 * @h_aborted: flag indicating fatal error on handle
432 /* Docbook can't yet cope with the bit fields, but will leave the documentation
433 * in so it can be fixed later.
438 /* Which compound transaction is this update a part of? */
439 transaction_t
*h_transaction
;
441 /* Number of remaining buffers we are allowed to dirty: */
442 int h_buffer_credits
;
444 /* Reference count on this handle */
447 /* Field for caller's use to track errors through large fs */
451 /* Flags [no locking] */
452 unsigned int h_sync
: 1; /* sync-on-close */
453 unsigned int h_jdata
: 1; /* force data journaling */
454 unsigned int h_aborted
: 1; /* fatal error on handle */
456 #ifdef CONFIG_DEBUG_LOCK_ALLOC
457 struct lockdep_map h_lockdep_map
;
463 * Some stats for checkpoint phase
465 struct transaction_chp_stats_s
{
466 unsigned long cs_chp_time
;
467 unsigned long cs_forced_to_close
;
468 unsigned long cs_written
;
469 unsigned long cs_dropped
;
472 /* The transaction_t type is the guts of the journaling mechanism. It
473 * tracks a compound transaction through its various states:
475 * RUNNING: accepting new updates
476 * LOCKED: Updates still running but we don't accept new ones
477 * RUNDOWN: Updates are tidying up but have finished requesting
478 * new buffers to modify (state not used for now)
479 * FLUSH: All updates complete, but we are still writing to disk
480 * COMMIT: All data on disk, writing commit record
481 * FINISHED: We still have to keep the transaction for checkpointing.
483 * The transaction keeps track of all of the buffers modified by a
484 * running transaction, and all of the buffers committed but not yet
485 * flushed to home for finished transactions.
492 * ->jbd_lock_bh_journal_head() (This is "innermost")
495 * ->jbd_lock_bh_state()
497 * jbd_lock_bh_state()
504 * ->j_list_lock (journal_unmap_buffer)
510 /* Pointer to the journal for this transaction. [no locking] */
511 journal_t
*t_journal
;
513 /* Sequence number for this transaction [no locking] */
517 * Transaction's current state
518 * [no locking - only kjournald2 alters this]
519 * [j_list_lock] guards transition of a transaction into T_FINISHED
520 * state and subsequent call of __jbd2_journal_drop_transaction()
521 * FIXME: needs barriers
522 * KLUDGE: [use j_state_lock]
534 * Where in the log does this transaction's commit start? [no locking]
536 unsigned long t_log_start
;
538 /* Number of buffers on the t_buffers list [j_list_lock] */
542 * Doubly-linked circular list of all buffers reserved but not yet
543 * modified by this transaction [j_list_lock]
545 struct journal_head
*t_reserved_list
;
548 * Doubly-linked circular list of all metadata buffers owned by this
549 * transaction [j_list_lock]
551 struct journal_head
*t_buffers
;
554 * Doubly-linked circular list of all forget buffers (superseded
555 * buffers which we can un-checkpoint once this transaction commits)
558 struct journal_head
*t_forget
;
561 * Doubly-linked circular list of all buffers still to be flushed before
562 * this transaction can be checkpointed. [j_list_lock]
564 struct journal_head
*t_checkpoint_list
;
567 * Doubly-linked circular list of all buffers submitted for IO while
568 * checkpointing. [j_list_lock]
570 struct journal_head
*t_checkpoint_io_list
;
573 * Doubly-linked circular list of temporary buffers currently undergoing
574 * IO in the log [j_list_lock]
576 struct journal_head
*t_iobuf_list
;
579 * Doubly-linked circular list of metadata buffers being shadowed by log
580 * IO. The IO buffers on the iobuf list and the shadow buffers on this
581 * list match each other one for one at all times. [j_list_lock]
583 struct journal_head
*t_shadow_list
;
586 * Doubly-linked circular list of control buffers being written to the
589 struct journal_head
*t_log_list
;
592 * List of inodes whose data we've modified in data=ordered mode.
595 struct list_head t_inode_list
;
598 * Protects info related to handles
600 spinlock_t t_handle_lock
;
603 * Longest time some handle had to wait for running transaction
605 unsigned long t_max_wait
;
608 * When transaction started
610 unsigned long t_start
;
613 * Checkpointing stats [j_checkpoint_sem]
615 struct transaction_chp_stats_s t_chp_stats
;
618 * Number of outstanding updates running on this transaction
624 * Number of buffers reserved for use by all handles in this transaction
625 * handle but not yet modified. [t_handle_lock]
627 int t_outstanding_credits
;
630 * Forward and backward links for the circular list of all transactions
631 * awaiting checkpoint. [j_list_lock]
633 transaction_t
*t_cpnext
, *t_cpprev
;
636 * When will the transaction expire (become due for commit), in jiffies?
639 unsigned long t_expires
;
642 * How many handles used this transaction? [t_handle_lock]
647 * For use by the filesystem to store fs-specific data
648 * structures associated with the transaction
650 struct list_head t_private_list
;
653 struct transaction_run_stats_s
{
654 unsigned long rs_wait
;
655 unsigned long rs_running
;
656 unsigned long rs_locked
;
657 unsigned long rs_flushing
;
658 unsigned long rs_logging
;
660 unsigned long rs_handle_count
;
661 unsigned long rs_blocks
;
662 unsigned long rs_blocks_logged
;
665 struct transaction_stats_s
{
667 unsigned long ts_tid
;
669 struct transaction_run_stats_s run
;
670 struct transaction_chp_stats_s chp
;
674 #define JBD2_STATS_RUN 1
675 #define JBD2_STATS_CHECKPOINT 2
677 static inline unsigned long
678 jbd2_time_diff(unsigned long start
, unsigned long end
)
683 return end
+ (MAX_JIFFY_OFFSET
- start
);
687 * struct journal_s - The journal_s type is the concrete type associated with
689 * @j_flags: General journaling state flags
690 * @j_errno: Is there an outstanding uncleared error on the journal (from a
692 * @j_sb_buffer: First part of superblock buffer
693 * @j_superblock: Second part of superblock buffer
694 * @j_format_version: Version of the superblock format
695 * @j_state_lock: Protect the various scalars in the journal
696 * @j_barrier_count: Number of processes waiting to create a barrier lock
697 * @j_barrier: The barrier lock itself
698 * @j_running_transaction: The current running transaction..
699 * @j_committing_transaction: the transaction we are pushing to disk
700 * @j_checkpoint_transactions: a linked circular list of all transactions
701 * waiting for checkpointing
702 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
703 * to start committing, or for a barrier lock to be released
704 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
705 * @j_wait_done_commit: Wait queue for waiting for commit to complete
706 * @j_wait_checkpoint: Wait queue to trigger checkpointing
707 * @j_wait_commit: Wait queue to trigger commit
708 * @j_wait_updates: Wait queue to wait for updates to complete
709 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
710 * @j_head: Journal head - identifies the first unused block in the journal
711 * @j_tail: Journal tail - identifies the oldest still-used block in the
713 * @j_free: Journal free - how many free blocks are there in the journal?
714 * @j_first: The block number of the first usable block
715 * @j_last: The block number one beyond the last usable block
716 * @j_dev: Device where we store the journal
717 * @j_blocksize: blocksize for the location where we store the journal.
718 * @j_blk_offset: starting block offset for into the device where we store the
720 * @j_fs_dev: Device which holds the client fs. For internal journal this will
722 * @j_maxlen: Total maximum capacity of the journal region on disk.
723 * @j_list_lock: Protects the buffer lists and internal buffer state.
724 * @j_inode: Optional inode where we store the journal. If present, all journal
725 * block numbers are mapped into this inode via bmap().
726 * @j_tail_sequence: Sequence number of the oldest transaction in the log
727 * @j_transaction_sequence: Sequence number of the next transaction to grant
728 * @j_commit_sequence: Sequence number of the most recently committed
730 * @j_commit_request: Sequence number of the most recent transaction wanting
732 * @j_uuid: Uuid of client object.
733 * @j_task: Pointer to the current commit thread for this journal
734 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
735 * single compound commit transaction
736 * @j_commit_interval: What is the maximum transaction lifetime before we begin
738 * @j_commit_timer: The timer used to wakeup the commit thread
739 * @j_revoke_lock: Protect the revoke table
740 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
741 * current transaction.
742 * @j_revoke_table: alternate revoke tables for j_revoke
743 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
744 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
745 * number that will fit in j_blocksize
746 * @j_last_sync_writer: most recent pid which did a synchronous write
747 * @j_history: Buffer storing the transactions statistics history
748 * @j_history_max: Maximum number of transactions in the statistics history
749 * @j_history_cur: Current number of transactions in the statistics history
750 * @j_history_lock: Protect the transactions statistics history
751 * @j_proc_entry: procfs entry for the jbd statistics directory
752 * @j_stats: Overall statistics
753 * @j_private: An opaque pointer to fs-private information.
758 /* General journaling state flags [j_state_lock] */
759 unsigned long j_flags
;
762 * Is there an outstanding uncleared error on the journal (from a prior
763 * abort)? [j_state_lock]
767 /* The superblock buffer */
768 struct buffer_head
*j_sb_buffer
;
769 journal_superblock_t
*j_superblock
;
771 /* Version of the superblock format */
772 int j_format_version
;
775 * Protect the various scalars in the journal
777 spinlock_t j_state_lock
;
780 * Number of processes waiting to create a barrier lock [j_state_lock]
784 /* The barrier lock itself */
785 struct mutex j_barrier
;
788 * Transactions: The current running transaction...
789 * [j_state_lock] [caller holding open handle]
791 transaction_t
*j_running_transaction
;
794 * the transaction we are pushing to disk
795 * [j_state_lock] [caller holding open handle]
797 transaction_t
*j_committing_transaction
;
800 * ... and a linked circular list of all transactions waiting for
801 * checkpointing. [j_list_lock]
803 transaction_t
*j_checkpoint_transactions
;
806 * Wait queue for waiting for a locked transaction to start committing,
807 * or for a barrier lock to be released
809 wait_queue_head_t j_wait_transaction_locked
;
811 /* Wait queue for waiting for checkpointing to complete */
812 wait_queue_head_t j_wait_logspace
;
814 /* Wait queue for waiting for commit to complete */
815 wait_queue_head_t j_wait_done_commit
;
817 /* Wait queue to trigger checkpointing */
818 wait_queue_head_t j_wait_checkpoint
;
820 /* Wait queue to trigger commit */
821 wait_queue_head_t j_wait_commit
;
823 /* Wait queue to wait for updates to complete */
824 wait_queue_head_t j_wait_updates
;
826 /* Semaphore for locking against concurrent checkpoints */
827 struct mutex j_checkpoint_mutex
;
830 * Journal head: identifies the first unused block in the journal.
833 unsigned long j_head
;
836 * Journal tail: identifies the oldest still-used block in the journal.
839 unsigned long j_tail
;
842 * Journal free: how many free blocks are there in the journal?
845 unsigned long j_free
;
848 * Journal start and end: the block numbers of the first usable block
849 * and one beyond the last usable block in the journal. [j_state_lock]
851 unsigned long j_first
;
852 unsigned long j_last
;
855 * Device, blocksize and starting block offset for the location where we
858 struct block_device
*j_dev
;
860 unsigned long long j_blk_offset
;
861 char j_devname
[BDEVNAME_SIZE
+24];
864 * Device which holds the client fs. For internal journal this will be
867 struct block_device
*j_fs_dev
;
869 /* Total maximum capacity of the journal region on disk. */
870 unsigned int j_maxlen
;
873 * Protects the buffer lists and internal buffer state.
875 spinlock_t j_list_lock
;
877 /* Optional inode where we store the journal. If present, all */
878 /* journal block numbers are mapped into this inode via */
880 struct inode
*j_inode
;
883 * Sequence number of the oldest transaction in the log [j_state_lock]
885 tid_t j_tail_sequence
;
888 * Sequence number of the next transaction to grant [j_state_lock]
890 tid_t j_transaction_sequence
;
893 * Sequence number of the most recently committed transaction
896 tid_t j_commit_sequence
;
899 * Sequence number of the most recent transaction wanting commit
902 tid_t j_commit_request
;
905 * Journal uuid: identifies the object (filesystem, LVM volume etc)
906 * backed by this journal. This will eventually be replaced by an array
907 * of uuids, allowing us to index multiple devices within a single
908 * journal and to perform atomic updates across them.
912 /* Pointer to the current commit thread for this journal */
913 struct task_struct
*j_task
;
916 * Maximum number of metadata buffers to allow in a single compound
919 int j_max_transaction_buffers
;
922 * What is the maximum transaction lifetime before we begin a commit?
924 unsigned long j_commit_interval
;
926 /* The timer used to wakeup the commit thread: */
927 struct timer_list j_commit_timer
;
930 * The revoke table: maintains the list of revoked blocks in the
931 * current transaction. [j_revoke_lock]
933 spinlock_t j_revoke_lock
;
934 struct jbd2_revoke_table_s
*j_revoke
;
935 struct jbd2_revoke_table_s
*j_revoke_table
[2];
938 * array of bhs for jbd2_journal_commit_transaction
940 struct buffer_head
**j_wbuf
;
943 pid_t j_last_sync_writer
;
945 /* This function is called when a transaction is closed */
946 void (*j_commit_callback
)(journal_t
*,
952 struct transaction_stats_s
*j_history
;
956 * Protect the transactions statistics history
958 spinlock_t j_history_lock
;
959 struct proc_dir_entry
*j_proc_entry
;
960 struct transaction_stats_s j_stats
;
962 /* Failed journal commit ID */
963 unsigned int j_failed_commit
;
966 * An opaque pointer to fs-private information. ext3 puts its
967 * superblock pointer here
973 * Journal flag definitions
975 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
976 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
977 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
978 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
979 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
980 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
981 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
982 * data write error in ordered
986 * Function declarations for the journaling transaction and buffer
991 extern void jbd2_journal_unfile_buffer(journal_t
*, struct journal_head
*);
992 extern void __jbd2_journal_unfile_buffer(struct journal_head
*);
993 extern void __jbd2_journal_refile_buffer(struct journal_head
*);
994 extern void jbd2_journal_refile_buffer(journal_t
*, struct journal_head
*);
995 extern void __jbd2_journal_file_buffer(struct journal_head
*, transaction_t
*, int);
996 extern void __journal_free_buffer(struct journal_head
*bh
);
997 extern void jbd2_journal_file_buffer(struct journal_head
*, transaction_t
*, int);
998 extern void __journal_clean_data_list(transaction_t
*transaction
);
1000 /* Log buffer allocation */
1001 extern struct journal_head
* jbd2_journal_get_descriptor_buffer(journal_t
*);
1002 int jbd2_journal_next_log_block(journal_t
*, unsigned long long *);
1004 /* Commit management */
1005 extern void jbd2_journal_commit_transaction(journal_t
*);
1007 /* Checkpoint list management */
1008 int __jbd2_journal_clean_checkpoint_list(journal_t
*journal
);
1009 int __jbd2_journal_remove_checkpoint(struct journal_head
*);
1010 void __jbd2_journal_insert_checkpoint(struct journal_head
*, transaction_t
*);
1014 jbd2_journal_write_metadata_buffer(transaction_t
*transaction
,
1015 struct journal_head
*jh_in
,
1016 struct journal_head
**jh_out
,
1017 unsigned long long blocknr
);
1019 /* Transaction locking */
1020 extern void __wait_on_journal (journal_t
*);
1025 * We need to lock the journal during transaction state changes so that nobody
1026 * ever tries to take a handle on the running transaction while we are in the
1027 * middle of moving it to the commit phase. j_state_lock does this.
1029 * Note that the locking is completely interrupt unsafe. We never touch
1030 * journal structures from interrupts.
1033 static inline handle_t
*journal_current_handle(void)
1035 return current
->journal_info
;
1038 /* The journaling code user interface:
1040 * Create and destroy handles
1041 * Register buffer modifications against the current transaction.
1044 extern handle_t
*jbd2_journal_start(journal_t
*, int nblocks
);
1045 extern int jbd2_journal_restart (handle_t
*, int nblocks
);
1046 extern int jbd2_journal_extend (handle_t
*, int nblocks
);
1047 extern int jbd2_journal_get_write_access(handle_t
*, struct buffer_head
*);
1048 extern int jbd2_journal_get_create_access (handle_t
*, struct buffer_head
*);
1049 extern int jbd2_journal_get_undo_access(handle_t
*, struct buffer_head
*);
1050 extern int jbd2_journal_dirty_metadata (handle_t
*, struct buffer_head
*);
1051 extern void jbd2_journal_release_buffer (handle_t
*, struct buffer_head
*);
1052 extern int jbd2_journal_forget (handle_t
*, struct buffer_head
*);
1053 extern void journal_sync_buffer (struct buffer_head
*);
1054 extern void jbd2_journal_invalidatepage(journal_t
*,
1055 struct page
*, unsigned long);
1056 extern int jbd2_journal_try_to_free_buffers(journal_t
*, struct page
*, gfp_t
);
1057 extern int jbd2_journal_stop(handle_t
*);
1058 extern int jbd2_journal_flush (journal_t
*);
1059 extern void jbd2_journal_lock_updates (journal_t
*);
1060 extern void jbd2_journal_unlock_updates (journal_t
*);
1062 extern journal_t
* jbd2_journal_init_dev(struct block_device
*bdev
,
1063 struct block_device
*fs_dev
,
1064 unsigned long long start
, int len
, int bsize
);
1065 extern journal_t
* jbd2_journal_init_inode (struct inode
*);
1066 extern int jbd2_journal_update_format (journal_t
*);
1067 extern int jbd2_journal_check_used_features
1068 (journal_t
*, unsigned long, unsigned long, unsigned long);
1069 extern int jbd2_journal_check_available_features
1070 (journal_t
*, unsigned long, unsigned long, unsigned long);
1071 extern int jbd2_journal_set_features
1072 (journal_t
*, unsigned long, unsigned long, unsigned long);
1073 extern void jbd2_journal_clear_features
1074 (journal_t
*, unsigned long, unsigned long, unsigned long);
1075 extern int jbd2_journal_create (journal_t
*);
1076 extern int jbd2_journal_load (journal_t
*journal
);
1077 extern int jbd2_journal_destroy (journal_t
*);
1078 extern int jbd2_journal_recover (journal_t
*journal
);
1079 extern int jbd2_journal_wipe (journal_t
*, int);
1080 extern int jbd2_journal_skip_recovery (journal_t
*);
1081 extern void jbd2_journal_update_superblock (journal_t
*, int);
1082 extern void __jbd2_journal_abort_hard (journal_t
*);
1083 extern void jbd2_journal_abort (journal_t
*, int);
1084 extern int jbd2_journal_errno (journal_t
*);
1085 extern void jbd2_journal_ack_err (journal_t
*);
1086 extern int jbd2_journal_clear_err (journal_t
*);
1087 extern int jbd2_journal_bmap(journal_t
*, unsigned long, unsigned long long *);
1088 extern int jbd2_journal_force_commit(journal_t
*);
1089 extern int jbd2_journal_file_inode(handle_t
*handle
, struct jbd2_inode
*inode
);
1090 extern int jbd2_journal_begin_ordered_truncate(journal_t
*journal
,
1091 struct jbd2_inode
*inode
, loff_t new_size
);
1092 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode
*jinode
, struct inode
*inode
);
1093 extern void jbd2_journal_release_jbd_inode(journal_t
*journal
, struct jbd2_inode
*jinode
);
1096 * journal_head management
1098 struct journal_head
*jbd2_journal_add_journal_head(struct buffer_head
*bh
);
1099 struct journal_head
*jbd2_journal_grab_journal_head(struct buffer_head
*bh
);
1100 void jbd2_journal_remove_journal_head(struct buffer_head
*bh
);
1101 void jbd2_journal_put_journal_head(struct journal_head
*jh
);
1106 extern struct kmem_cache
*jbd2_handle_cache
;
1108 static inline handle_t
*jbd2_alloc_handle(gfp_t gfp_flags
)
1110 return kmem_cache_alloc(jbd2_handle_cache
, gfp_flags
);
1113 static inline void jbd2_free_handle(handle_t
*handle
)
1115 kmem_cache_free(jbd2_handle_cache
, handle
);
1118 /* Primary revoke support */
1119 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1120 extern int jbd2_journal_init_revoke(journal_t
*, int);
1121 extern void jbd2_journal_destroy_revoke_caches(void);
1122 extern int jbd2_journal_init_revoke_caches(void);
1124 extern void jbd2_journal_destroy_revoke(journal_t
*);
1125 extern int jbd2_journal_revoke (handle_t
*, unsigned long long, struct buffer_head
*);
1126 extern int jbd2_journal_cancel_revoke(handle_t
*, struct journal_head
*);
1127 extern void jbd2_journal_write_revoke_records(journal_t
*, transaction_t
*);
1129 /* Recovery revoke support */
1130 extern int jbd2_journal_set_revoke(journal_t
*, unsigned long long, tid_t
);
1131 extern int jbd2_journal_test_revoke(journal_t
*, unsigned long long, tid_t
);
1132 extern void jbd2_journal_clear_revoke(journal_t
*);
1133 extern void jbd2_journal_switch_revoke_table(journal_t
*journal
);
1136 * The log thread user interface:
1138 * Request space in the current transaction, and force transaction commit
1139 * transitions on demand.
1142 int __jbd2_log_space_left(journal_t
*); /* Called with journal locked */
1143 int jbd2_log_start_commit(journal_t
*journal
, tid_t tid
);
1144 int __jbd2_log_start_commit(journal_t
*journal
, tid_t tid
);
1145 int jbd2_journal_start_commit(journal_t
*journal
, tid_t
*tid
);
1146 int jbd2_journal_force_commit_nested(journal_t
*journal
);
1147 int jbd2_log_wait_commit(journal_t
*journal
, tid_t tid
);
1148 int jbd2_log_do_checkpoint(journal_t
*journal
);
1150 void __jbd2_log_wait_for_space(journal_t
*journal
);
1151 extern void __jbd2_journal_drop_transaction(journal_t
*, transaction_t
*);
1152 extern int jbd2_cleanup_journal_tail(journal_t
*);
1154 /* Debugging code only: */
1156 #define jbd_ENOSYS() \
1158 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1159 current->state = TASK_UNINTERRUPTIBLE; \
1166 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1167 * bit, when set, indicates that we have had a fatal error somewhere,
1168 * either inside the journaling layer or indicated to us by the client
1169 * (eg. ext3), and that we and should not commit any further
1173 static inline int is_journal_aborted(journal_t
*journal
)
1175 return journal
->j_flags
& JBD2_ABORT
;
1178 static inline int is_handle_aborted(handle_t
*handle
)
1180 if (handle
->h_aborted
)
1182 return is_journal_aborted(handle
->h_transaction
->t_journal
);
1185 static inline void jbd2_journal_abort_handle(handle_t
*handle
)
1187 handle
->h_aborted
= 1;
1190 #endif /* __KERNEL__ */
1192 /* Comparison functions for transaction IDs: perform comparisons using
1193 * modulo arithmetic so that they work over sequence number wraps. */
1195 static inline int tid_gt(tid_t x
, tid_t y
)
1197 int difference
= (x
- y
);
1198 return (difference
> 0);
1201 static inline int tid_geq(tid_t x
, tid_t y
)
1203 int difference
= (x
- y
);
1204 return (difference
>= 0);
1207 extern int jbd2_journal_blocks_per_page(struct inode
*inode
);
1208 extern size_t journal_tag_bytes(journal_t
*journal
);
1211 * Return the minimum number of blocks which must be free in the journal
1212 * before a new transaction may be started. Must be called under j_state_lock.
1214 static inline int jbd_space_needed(journal_t
*journal
)
1216 int nblocks
= journal
->j_max_transaction_buffers
;
1217 if (journal
->j_committing_transaction
)
1218 nblocks
+= journal
->j_committing_transaction
->
1219 t_outstanding_credits
;
1224 * Definitions which augment the buffer_head layer
1227 /* journaling buffer types */
1228 #define BJ_None 0 /* Not journaled */
1229 #define BJ_Metadata 1 /* Normal journaled metadata */
1230 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1231 #define BJ_IO 3 /* Buffer is for temporary IO use */
1232 #define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1233 #define BJ_LogCtl 5 /* Buffer contains log descriptors */
1234 #define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1237 extern int jbd_blocks_per_page(struct inode
*inode
);
1241 #define buffer_trace_init(bh) do {} while (0)
1242 #define print_buffer_fields(bh) do {} while (0)
1243 #define print_buffer_trace(bh) do {} while (0)
1244 #define BUFFER_TRACE(bh, info) do {} while (0)
1245 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1246 #define JBUFFER_TRACE(jh, info) do {} while (0)
1248 #endif /* __KERNEL__ */
1250 #endif /* _LINUX_JBD2_H */