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>
33 #include <linux/slab.h>
36 #define journal_oom_retry 1
39 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
40 * certain classes of error which can occur due to failed IOs. Under
41 * normal use we want ext4 to continue after such errors, because
42 * hardware _can_ fail, but for debugging purposes when running tests on
43 * known-good hardware we may want to trap these errors.
45 #undef JBD2_PARANOID_IOFAIL
48 * The default maximum commit age, in seconds.
50 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
52 #ifdef CONFIG_JBD2_DEBUG
54 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
55 * consistency checks. By default we don't do this unless
56 * CONFIG_JBD2_DEBUG is on.
58 #define JBD2_EXPENSIVE_CHECKING
59 extern u8 jbd2_journal_enable_debug;
61 #define jbd_debug(n, f, a...) \
63 if ((n) <= jbd2_journal_enable_debug) { \
64 printk (KERN_DEBUG "(%s, %d): %s: ", \
65 __FILE__, __LINE__, __func__); \
70 #define jbd_debug(f, a...) /**/
73 extern void *jbd2_alloc(size_t size, gfp_t flags);
74 extern void jbd2_free(void *ptr, size_t size);
76 #define JBD2_MIN_JOURNAL_BLOCKS 1024
81 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
83 * All filesystem modifications made by the process go
84 * through this handle. Recursive operations (such as quota operations)
85 * are gathered into a single update.
87 * The buffer credits field is used to account for journaled buffers
88 * being modified by the running process. To ensure that there is
89 * enough log space for all outstanding operations, we need to limit the
90 * number of outstanding buffers possible at any time. When the
91 * operation completes, any buffer credits not used are credited back to
92 * the transaction, so that at all times we know how many buffers the
93 * outstanding updates on a transaction might possibly touch.
95 * This is an opaque datatype.
97 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
101 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
103 * journal_t is linked to from the fs superblock structure.
105 * We use the journal_t to keep track of all outstanding transaction
106 * activity on the filesystem, and to manage the state of the log
109 * This is an opaque datatype.
111 typedef struct journal_s journal_t; /* Journal control structure */
115 * Internal structures used by the logging mechanism:
118 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
125 * Descriptor block types:
128 #define JBD2_DESCRIPTOR_BLOCK 1
129 #define JBD2_COMMIT_BLOCK 2
130 #define JBD2_SUPERBLOCK_V1 3
131 #define JBD2_SUPERBLOCK_V2 4
132 #define JBD2_REVOKE_BLOCK 5
135 * Standard header for all descriptor blocks:
137 typedef struct journal_header_s
147 #define JBD2_CRC32_CHKSUM 1
148 #define JBD2_MD5_CHKSUM 2
149 #define JBD2_SHA1_CHKSUM 3
151 #define JBD2_CRC32_CHKSUM_SIZE 4
153 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
155 * Commit block header for storing transactional checksums:
157 struct commit_header {
161 unsigned char h_chksum_type;
162 unsigned char h_chksum_size;
163 unsigned char h_padding[2];
164 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
166 __be32 h_commit_nsec;
170 * The block tag: used to describe a single buffer in the journal.
171 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
172 * raw struct shouldn't be used for pointer math or sizeof() - use
173 * journal_tag_bytes(journal) instead to compute this.
175 typedef struct journal_block_tag_s
177 __be32 t_blocknr; /* The on-disk block number */
178 __be32 t_flags; /* See below */
179 __be32 t_blocknr_high; /* most-significant high 32bits. */
180 } journal_block_tag_t;
182 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
183 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
186 * The revoke descriptor: used on disk to describe a series of blocks to
187 * be revoked from the log
189 typedef struct jbd2_journal_revoke_header_s
191 journal_header_t r_header;
192 __be32 r_count; /* Count of bytes used in the block */
193 } jbd2_journal_revoke_header_t;
196 /* Definitions for the journal tag flags word: */
197 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
198 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
199 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
200 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
204 * The journal superblock. All fields are in big-endian byte order.
206 typedef struct journal_superblock_s
209 journal_header_t s_header;
212 /* Static information describing the journal */
213 __be32 s_blocksize; /* journal device blocksize */
214 __be32 s_maxlen; /* total blocks in journal file */
215 __be32 s_first; /* first block of log information */
218 /* Dynamic information describing the current state of the log */
219 __be32 s_sequence; /* first commit ID expected in log */
220 __be32 s_start; /* blocknr of start of log */
223 /* Error value, as set by jbd2_journal_abort(). */
227 /* Remaining fields are only valid in a version-2 superblock */
228 __be32 s_feature_compat; /* compatible feature set */
229 __be32 s_feature_incompat; /* incompatible feature set */
230 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
232 __u8 s_uuid[16]; /* 128-bit uuid for journal */
235 __be32 s_nr_users; /* Nr of filesystems sharing log */
237 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
240 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
241 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
247 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
249 } journal_superblock_t;
251 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
252 ((j)->j_format_version >= 2 && \
253 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
254 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
255 ((j)->j_format_version >= 2 && \
256 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
257 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
258 ((j)->j_format_version >= 2 && \
259 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
261 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
263 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
264 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
265 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
267 /* Features known to this kernel version: */
268 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
269 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
270 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
271 JBD2_FEATURE_INCOMPAT_64BIT | \
272 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)
276 #include <linux/fs.h>
277 #include <linux/sched.h>
279 #define J_ASSERT(assert) BUG_ON(!(assert))
281 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
282 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
284 #if defined(JBD2_PARANOID_IOFAIL)
285 #define J_EXPECT(expr, why...) J_ASSERT(expr)
286 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
287 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
289 #define __journal_expect(expr, why...) \
294 "JBD2 unexpected failure: %s: %s;\n", \
296 printk(KERN_ERR why "\n"); \
300 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
301 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
302 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
305 enum jbd_state_bits {
306 BH_JBD /* Has an attached ext3 journal_head */
308 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
309 BH_Freed, /* Has been freed (truncated) */
310 BH_Revoked, /* Has been revoked from the log */
311 BH_RevokeValid, /* Revoked flag is valid */
312 BH_JBDDirty, /* Is dirty but journaled */
313 BH_State, /* Pins most journal_head state */
314 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
315 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
316 BH_JBDPrivateStart, /* First bit available for private use by FS */
320 BUFFER_FNS(JWrite, jwrite)
321 BUFFER_FNS(JBDDirty, jbddirty)
322 TAS_BUFFER_FNS(JBDDirty, jbddirty)
323 BUFFER_FNS(Revoked, revoked)
324 TAS_BUFFER_FNS(Revoked, revoked)
325 BUFFER_FNS(RevokeValid, revokevalid)
326 TAS_BUFFER_FNS(RevokeValid, revokevalid)
327 BUFFER_FNS(Freed, freed)
329 static inline struct buffer_head *jh2bh(struct journal_head *jh)
334 static inline struct journal_head *bh2jh(struct buffer_head *bh)
336 return bh->b_private;
339 static inline void jbd_lock_bh_state(struct buffer_head *bh)
341 bit_spin_lock(BH_State, &bh->b_state);
344 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
346 return bit_spin_trylock(BH_State, &bh->b_state);
349 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
351 return bit_spin_is_locked(BH_State, &bh->b_state);
354 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
356 bit_spin_unlock(BH_State, &bh->b_state);
359 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
361 bit_spin_lock(BH_JournalHead, &bh->b_state);
364 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
366 bit_spin_unlock(BH_JournalHead, &bh->b_state);
369 /* Flags in jbd_inode->i_flags */
370 #define __JI_COMMIT_RUNNING 0
371 /* Commit of the inode data in progress. We use this flag to protect us from
372 * concurrent deletion of inode. We cannot use reference to inode for this
373 * since we cannot afford doing last iput() on behalf of kjournald
375 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
378 * struct jbd_inode is the structure linking inodes in ordered mode
379 * present in a transaction so that we can sync them during commit.
382 /* Which transaction does this inode belong to? Either the running
383 * transaction or the committing one. [j_list_lock] */
384 transaction_t *i_transaction;
386 /* Pointer to the running transaction modifying inode's data in case
387 * there is already a committing transaction touching it. [j_list_lock] */
388 transaction_t *i_next_transaction;
390 /* List of inodes in the i_transaction [j_list_lock] */
391 struct list_head i_list;
393 /* VFS inode this inode belongs to [constant during the lifetime
394 * of the structure] */
395 struct inode *i_vfs_inode;
397 /* Flags of inode [j_list_lock] */
398 unsigned long i_flags;
401 struct jbd2_revoke_table_s;
404 * struct handle_s - The handle_s type is the concrete type associated with
406 * @h_transaction: Which compound transaction is this update a part of?
407 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
408 * @h_ref: Reference count on this handle
409 * @h_err: Field for caller's use to track errors through large fs operations
410 * @h_sync: flag for sync-on-close
411 * @h_jdata: flag to force data journaling
412 * @h_aborted: flag indicating fatal error on handle
415 /* Docbook can't yet cope with the bit fields, but will leave the documentation
416 * in so it can be fixed later.
419 struct jbd2_journal_handle
421 /* Which compound transaction is this update a part of? */
422 transaction_t *h_transaction;
424 /* Number of remaining buffers we are allowed to dirty: */
425 int h_buffer_credits;
427 /* Reference count on this handle */
430 /* Field for caller's use to track errors through large fs */
434 /* Flags [no locking] */
435 unsigned int h_sync:1; /* sync-on-close */
436 unsigned int h_jdata:1; /* force data journaling */
437 unsigned int h_aborted:1; /* fatal error on handle */
438 unsigned int h_cowing:1; /* COWing block to snapshot */
440 /* Number of buffers requested by user:
441 * (before adding the COW credits factor) */
442 unsigned int h_base_credits:14;
444 /* Number of buffers the user is allowed to dirty:
445 * (counts only buffers dirtied when !h_cowing) */
446 unsigned int h_user_credits:14;
449 #ifdef CONFIG_DEBUG_LOCK_ALLOC
450 struct lockdep_map h_lockdep_map;
453 #ifdef CONFIG_JBD2_DEBUG
454 /* COW debugging counters: */
455 unsigned int h_cow_moved; /* blocks moved to snapshot */
456 unsigned int h_cow_copied; /* blocks copied to snapshot */
457 unsigned int h_cow_ok_jh; /* blocks already COWed during current
459 unsigned int h_cow_ok_bitmap; /* blocks not set in COW bitmap */
460 unsigned int h_cow_ok_mapped;/* blocks already mapped in snapshot */
461 unsigned int h_cow_bitmaps; /* COW bitmaps created */
462 unsigned int h_cow_excluded; /* blocks set in exclude bitmap */
468 * Some stats for checkpoint phase
470 struct transaction_chp_stats_s {
471 unsigned long cs_chp_time;
472 __u32 cs_forced_to_close;
477 /* The transaction_t type is the guts of the journaling mechanism. It
478 * tracks a compound transaction through its various states:
480 * RUNNING: accepting new updates
481 * LOCKED: Updates still running but we don't accept new ones
482 * RUNDOWN: Updates are tidying up but have finished requesting
483 * new buffers to modify (state not used for now)
484 * FLUSH: All updates complete, but we are still writing to disk
485 * COMMIT: All data on disk, writing commit record
486 * FINISHED: We still have to keep the transaction for checkpointing.
488 * The transaction keeps track of all of the buffers modified by a
489 * running transaction, and all of the buffers committed but not yet
490 * flushed to home for finished transactions.
497 * ->jbd_lock_bh_journal_head() (This is "innermost")
500 * ->jbd_lock_bh_state()
502 * jbd_lock_bh_state()
509 * ->j_list_lock (journal_unmap_buffer)
515 /* Pointer to the journal for this transaction. [no locking] */
516 journal_t *t_journal;
518 /* Sequence number for this transaction [no locking] */
522 * Transaction's current state
523 * [no locking - only kjournald2 alters this]
524 * [j_list_lock] guards transition of a transaction into T_FINISHED
525 * state and subsequent call of __jbd2_journal_drop_transaction()
526 * FIXME: needs barriers
527 * KLUDGE: [use j_state_lock]
539 * Where in the log does this transaction's commit start? [no locking]
541 unsigned long t_log_start;
543 /* Number of buffers on the t_buffers list [j_list_lock] */
547 * Doubly-linked circular list of all buffers reserved but not yet
548 * modified by this transaction [j_list_lock]
550 struct journal_head *t_reserved_list;
553 * Doubly-linked circular list of all metadata buffers owned by this
554 * transaction [j_list_lock]
556 struct journal_head *t_buffers;
559 * Doubly-linked circular list of all forget buffers (superseded
560 * buffers which we can un-checkpoint once this transaction commits)
563 struct journal_head *t_forget;
566 * Doubly-linked circular list of all buffers still to be flushed before
567 * this transaction can be checkpointed. [j_list_lock]
569 struct journal_head *t_checkpoint_list;
572 * Doubly-linked circular list of all buffers submitted for IO while
573 * checkpointing. [j_list_lock]
575 struct journal_head *t_checkpoint_io_list;
578 * Doubly-linked circular list of temporary buffers currently undergoing
579 * IO in the log [j_list_lock]
581 struct journal_head *t_iobuf_list;
584 * Doubly-linked circular list of metadata buffers being shadowed by log
585 * IO. The IO buffers on the iobuf list and the shadow buffers on this
586 * list match each other one for one at all times. [j_list_lock]
588 struct journal_head *t_shadow_list;
591 * Doubly-linked circular list of control buffers being written to the
594 struct journal_head *t_log_list;
597 * List of inodes whose data we've modified in data=ordered mode.
600 struct list_head t_inode_list;
603 * Protects info related to handles
605 spinlock_t t_handle_lock;
608 * Longest time some handle had to wait for running transaction
610 unsigned long t_max_wait;
613 * When transaction started
615 unsigned long t_start;
618 * Checkpointing stats [j_checkpoint_sem]
620 struct transaction_chp_stats_s t_chp_stats;
623 * Number of outstanding updates running on this transaction
629 * Number of buffers reserved for use by all handles in this transaction
630 * handle but not yet modified. [t_handle_lock]
632 atomic_t t_outstanding_credits;
635 * Forward and backward links for the circular list of all transactions
636 * awaiting checkpoint. [j_list_lock]
638 transaction_t *t_cpnext, *t_cpprev;
641 * When will the transaction expire (become due for commit), in jiffies?
644 unsigned long t_expires;
647 * When this transaction started, in nanoseconds [no locking]
649 ktime_t t_start_time;
652 * How many handles used this transaction? [t_handle_lock]
654 atomic_t t_handle_count;
657 * This transaction is being forced and some process is
658 * waiting for it to finish.
660 unsigned int t_synchronous_commit:1;
661 unsigned int t_flushed_data_blocks:1;
664 * For use by the filesystem to store fs-specific data
665 * structures associated with the transaction
667 struct list_head t_private_list;
670 struct transaction_run_stats_s {
671 unsigned long rs_wait;
672 unsigned long rs_running;
673 unsigned long rs_locked;
674 unsigned long rs_flushing;
675 unsigned long rs_logging;
677 __u32 rs_handle_count;
679 __u32 rs_blocks_logged;
682 struct transaction_stats_s {
683 unsigned long ts_tid;
684 struct transaction_run_stats_s run;
687 static inline unsigned long
688 jbd2_time_diff(unsigned long start, unsigned long end)
693 return end + (MAX_JIFFY_OFFSET - start);
696 #define JBD2_NR_BATCH 64
699 * struct journal_s - The journal_s type is the concrete type associated with
701 * @j_flags: General journaling state flags
702 * @j_errno: Is there an outstanding uncleared error on the journal (from a
704 * @j_sb_buffer: First part of superblock buffer
705 * @j_superblock: Second part of superblock buffer
706 * @j_format_version: Version of the superblock format
707 * @j_state_lock: Protect the various scalars in the journal
708 * @j_barrier_count: Number of processes waiting to create a barrier lock
709 * @j_barrier: The barrier lock itself
710 * @j_running_transaction: The current running transaction..
711 * @j_committing_transaction: the transaction we are pushing to disk
712 * @j_checkpoint_transactions: a linked circular list of all transactions
713 * waiting for checkpointing
714 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
715 * to start committing, or for a barrier lock to be released
716 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
717 * @j_wait_done_commit: Wait queue for waiting for commit to complete
718 * @j_wait_checkpoint: Wait queue to trigger checkpointing
719 * @j_wait_commit: Wait queue to trigger commit
720 * @j_wait_updates: Wait queue to wait for updates to complete
721 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
722 * @j_head: Journal head - identifies the first unused block in the journal
723 * @j_tail: Journal tail - identifies the oldest still-used block in the
725 * @j_free: Journal free - how many free blocks are there in the journal?
726 * @j_first: The block number of the first usable block
727 * @j_last: The block number one beyond the last usable block
728 * @j_dev: Device where we store the journal
729 * @j_blocksize: blocksize for the location where we store the journal.
730 * @j_blk_offset: starting block offset for into the device where we store the
732 * @j_fs_dev: Device which holds the client fs. For internal journal this will
734 * @j_maxlen: Total maximum capacity of the journal region on disk.
735 * @j_list_lock: Protects the buffer lists and internal buffer state.
736 * @j_inode: Optional inode where we store the journal. If present, all journal
737 * block numbers are mapped into this inode via bmap().
738 * @j_tail_sequence: Sequence number of the oldest transaction in the log
739 * @j_transaction_sequence: Sequence number of the next transaction to grant
740 * @j_commit_sequence: Sequence number of the most recently committed
742 * @j_commit_request: Sequence number of the most recent transaction wanting
744 * @j_uuid: Uuid of client object.
745 * @j_task: Pointer to the current commit thread for this journal
746 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
747 * single compound commit transaction
748 * @j_commit_interval: What is the maximum transaction lifetime before we begin
750 * @j_commit_timer: The timer used to wakeup the commit thread
751 * @j_revoke_lock: Protect the revoke table
752 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
753 * current transaction.
754 * @j_revoke_table: alternate revoke tables for j_revoke
755 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
756 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
757 * number that will fit in j_blocksize
758 * @j_last_sync_writer: most recent pid which did a synchronous write
759 * @j_history: Buffer storing the transactions statistics history
760 * @j_history_max: Maximum number of transactions in the statistics history
761 * @j_history_cur: Current number of transactions in the statistics history
762 * @j_history_lock: Protect the transactions statistics history
763 * @j_proc_entry: procfs entry for the jbd statistics directory
764 * @j_stats: Overall statistics
765 * @j_private: An opaque pointer to fs-private information.
770 /* General journaling state flags [j_state_lock] */
771 unsigned long j_flags;
774 * Is there an outstanding uncleared error on the journal (from a prior
775 * abort)? [j_state_lock]
779 /* The superblock buffer */
780 struct buffer_head *j_sb_buffer;
781 journal_superblock_t *j_superblock;
783 /* Version of the superblock format */
784 int j_format_version;
787 * Protect the various scalars in the journal
789 rwlock_t j_state_lock;
792 * Number of processes waiting to create a barrier lock [j_state_lock]
796 /* The barrier lock itself */
797 struct mutex j_barrier;
800 * Transactions: The current running transaction...
801 * [j_state_lock] [caller holding open handle]
803 transaction_t *j_running_transaction;
806 * the transaction we are pushing to disk
807 * [j_state_lock] [caller holding open handle]
809 transaction_t *j_committing_transaction;
812 * ... and a linked circular list of all transactions waiting for
813 * checkpointing. [j_list_lock]
815 transaction_t *j_checkpoint_transactions;
818 * Wait queue for waiting for a locked transaction to start committing,
819 * or for a barrier lock to be released
821 wait_queue_head_t j_wait_transaction_locked;
823 /* Wait queue for waiting for checkpointing to complete */
824 wait_queue_head_t j_wait_logspace;
826 /* Wait queue for waiting for commit to complete */
827 wait_queue_head_t j_wait_done_commit;
829 /* Wait queue to trigger checkpointing */
830 wait_queue_head_t j_wait_checkpoint;
832 /* Wait queue to trigger commit */
833 wait_queue_head_t j_wait_commit;
835 /* Wait queue to wait for updates to complete */
836 wait_queue_head_t j_wait_updates;
838 /* Semaphore for locking against concurrent checkpoints */
839 struct mutex j_checkpoint_mutex;
842 * List of buffer heads used by the checkpoint routine. This
843 * was moved from jbd2_log_do_checkpoint() to reduce stack
844 * usage. Access to this array is controlled by the
845 * j_checkpoint_mutex. [j_checkpoint_mutex]
847 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
850 * Journal head: identifies the first unused block in the journal.
853 unsigned long j_head;
856 * Journal tail: identifies the oldest still-used block in the journal.
859 unsigned long j_tail;
862 * Journal free: how many free blocks are there in the journal?
865 unsigned long j_free;
868 * Journal start and end: the block numbers of the first usable block
869 * and one beyond the last usable block in the journal. [j_state_lock]
871 unsigned long j_first;
872 unsigned long j_last;
875 * Device, blocksize and starting block offset for the location where we
878 struct block_device *j_dev;
880 unsigned long long j_blk_offset;
881 char j_devname[BDEVNAME_SIZE+24];
884 * Device which holds the client fs. For internal journal this will be
887 struct block_device *j_fs_dev;
889 /* Total maximum capacity of the journal region on disk. */
890 unsigned int j_maxlen;
893 * Protects the buffer lists and internal buffer state.
895 spinlock_t j_list_lock;
897 /* Optional inode where we store the journal. If present, all */
898 /* journal block numbers are mapped into this inode via */
900 struct inode *j_inode;
903 * Sequence number of the oldest transaction in the log [j_state_lock]
905 tid_t j_tail_sequence;
908 * Sequence number of the next transaction to grant [j_state_lock]
910 tid_t j_transaction_sequence;
913 * Sequence number of the most recently committed transaction
916 tid_t j_commit_sequence;
919 * Sequence number of the most recent transaction wanting commit
922 tid_t j_commit_request;
925 * Journal uuid: identifies the object (filesystem, LVM volume etc)
926 * backed by this journal. This will eventually be replaced by an array
927 * of uuids, allowing us to index multiple devices within a single
928 * journal and to perform atomic updates across them.
932 /* Pointer to the current commit thread for this journal */
933 struct task_struct *j_task;
936 * Maximum number of metadata buffers to allow in a single compound
939 int j_max_transaction_buffers;
942 * What is the maximum transaction lifetime before we begin a commit?
944 unsigned long j_commit_interval;
946 /* The timer used to wakeup the commit thread: */
947 struct timer_list j_commit_timer;
950 * The revoke table: maintains the list of revoked blocks in the
951 * current transaction. [j_revoke_lock]
953 spinlock_t j_revoke_lock;
954 struct jbd2_revoke_table_s *j_revoke;
955 struct jbd2_revoke_table_s *j_revoke_table[2];
958 * array of bhs for jbd2_journal_commit_transaction
960 struct buffer_head **j_wbuf;
964 * this is the pid of hte last person to run a synchronous operation
965 * through the journal
967 pid_t j_last_sync_writer;
970 * the average amount of time in nanoseconds it takes to commit a
971 * transaction to disk. [j_state_lock]
973 u64 j_average_commit_time;
976 * minimum and maximum times that we should wait for
977 * additional filesystem operations to get batched into a
978 * synchronous handle in microseconds
980 u32 j_min_batch_time;
981 u32 j_max_batch_time;
983 /* This function is called when a transaction is closed */
984 void (*j_commit_callback)(journal_t *,
990 spinlock_t j_history_lock;
991 struct proc_dir_entry *j_proc_entry;
992 struct transaction_stats_s j_stats;
994 /* Failed journal commit ID */
995 unsigned int j_failed_commit;
998 * An opaque pointer to fs-private information. ext3 puts its
999 * superblock pointer here
1005 * Journal flag definitions
1007 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1008 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1009 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1010 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1011 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1012 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
1013 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1014 * data write error in ordered
1018 * Function declarations for the journaling transaction and buffer
1022 /* Filing buffers */
1023 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1024 extern void __jbd2_journal_unfile_buffer(struct journal_head *);
1025 extern void __jbd2_journal_refile_buffer(struct journal_head *);
1026 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1027 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1028 extern void __journal_free_buffer(struct journal_head *bh);
1029 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1030 extern void __journal_clean_data_list(transaction_t *transaction);
1032 /* Log buffer allocation */
1033 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
1034 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1036 /* Commit management */
1037 extern void jbd2_journal_commit_transaction(journal_t *);
1039 /* Checkpoint list management */
1040 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1041 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1042 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1049 struct jbd2_buffer_trigger_type {
1051 * Fired a the moment data to write to the journal are known to be
1052 * stable - so either at the moment b_frozen_data is created or just
1053 * before a buffer is written to the journal. mapped_data is a mapped
1054 * buffer that is the frozen data for commit.
1056 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1057 struct buffer_head *bh, void *mapped_data,
1061 * Fired during journal abort for dirty buffers that will not be
1064 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1065 struct buffer_head *bh);
1068 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1070 struct jbd2_buffer_trigger_type *triggers);
1071 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1072 struct jbd2_buffer_trigger_type *triggers);
1076 jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1077 struct journal_head *jh_in,
1078 struct journal_head **jh_out,
1079 unsigned long long blocknr);
1081 /* Transaction locking */
1082 extern void __wait_on_journal (journal_t *);
1087 * We need to lock the journal during transaction state changes so that nobody
1088 * ever tries to take a handle on the running transaction while we are in the
1089 * middle of moving it to the commit phase. j_state_lock does this.
1091 * Note that the locking is completely interrupt unsafe. We never touch
1092 * journal structures from interrupts.
1095 static inline handle_t *journal_current_handle(void)
1097 return current->journal_info;
1100 /* The journaling code user interface:
1102 * Create and destroy handles
1103 * Register buffer modifications against the current transaction.
1106 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1107 extern handle_t *jbd2__journal_start(journal_t *, int nblocks, int gfp_mask);
1108 extern int jbd2_journal_restart(handle_t *, int nblocks);
1109 extern int jbd2__journal_restart(handle_t *, int nblocks, int gfp_mask);
1110 extern int jbd2_journal_extend (handle_t *, int nblocks);
1111 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1112 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1113 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1114 void jbd2_journal_set_triggers(struct buffer_head *,
1115 struct jbd2_buffer_trigger_type *type);
1116 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1117 extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1118 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1119 extern void journal_sync_buffer (struct buffer_head *);
1120 extern void jbd2_journal_invalidatepage(journal_t *,
1121 struct page *, unsigned long);
1122 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1123 extern int jbd2_journal_stop(handle_t *);
1124 extern int jbd2_journal_flush (journal_t *);
1125 extern void jbd2_journal_lock_updates (journal_t *);
1126 extern void jbd2_journal_unlock_updates (journal_t *);
1128 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1129 struct block_device *fs_dev,
1130 unsigned long long start, int len, int bsize);
1131 extern journal_t * jbd2_journal_init_inode (struct inode *);
1132 extern int jbd2_journal_update_format (journal_t *);
1133 extern int jbd2_journal_check_used_features
1134 (journal_t *, unsigned long, unsigned long, unsigned long);
1135 extern int jbd2_journal_check_available_features
1136 (journal_t *, unsigned long, unsigned long, unsigned long);
1137 extern int jbd2_journal_set_features
1138 (journal_t *, unsigned long, unsigned long, unsigned long);
1139 extern void jbd2_journal_clear_features
1140 (journal_t *, unsigned long, unsigned long, unsigned long);
1141 extern int jbd2_journal_load (journal_t *journal);
1142 extern int jbd2_journal_destroy (journal_t *);
1143 extern int jbd2_journal_recover (journal_t *journal);
1144 extern int jbd2_journal_wipe (journal_t *, int);
1145 extern int jbd2_journal_skip_recovery (journal_t *);
1146 extern void jbd2_journal_update_superblock (journal_t *, int);
1147 extern void __jbd2_journal_abort_hard (journal_t *);
1148 extern void jbd2_journal_abort (journal_t *, int);
1149 extern int jbd2_journal_errno (journal_t *);
1150 extern void jbd2_journal_ack_err (journal_t *);
1151 extern int jbd2_journal_clear_err (journal_t *);
1152 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1153 extern int jbd2_journal_force_commit(journal_t *);
1154 extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1155 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1156 struct jbd2_inode *inode, loff_t new_size);
1157 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1158 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1161 * journal_head management
1163 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1164 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1165 void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1166 void jbd2_journal_put_journal_head(struct journal_head *jh);
1171 extern struct kmem_cache *jbd2_handle_cache;
1173 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1175 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
1178 static inline void jbd2_free_handle(handle_t *handle)
1180 kmem_cache_free(jbd2_handle_cache, handle);
1184 * jbd2_inode management (optional, for those file systems that want to use
1185 * dynamically allocated jbd2_inode structures)
1187 extern struct kmem_cache *jbd2_inode_cache;
1189 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1191 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1194 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1196 kmem_cache_free(jbd2_inode_cache, jinode);
1199 /* Primary revoke support */
1200 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1201 extern int jbd2_journal_init_revoke(journal_t *, int);
1202 extern void jbd2_journal_destroy_revoke_caches(void);
1203 extern int jbd2_journal_init_revoke_caches(void);
1205 extern void jbd2_journal_destroy_revoke(journal_t *);
1206 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1207 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1208 extern void jbd2_journal_write_revoke_records(journal_t *,
1209 transaction_t *, int);
1211 /* Recovery revoke support */
1212 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1213 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1214 extern void jbd2_journal_clear_revoke(journal_t *);
1215 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1218 * The log thread user interface:
1220 * Request space in the current transaction, and force transaction commit
1221 * transitions on demand.
1224 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1225 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1226 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1227 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1228 int jbd2_journal_force_commit_nested(journal_t *journal);
1229 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1230 int jbd2_log_do_checkpoint(journal_t *journal);
1232 void __jbd2_log_wait_for_space(journal_t *journal);
1233 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1234 extern int jbd2_cleanup_journal_tail(journal_t *);
1236 /* Debugging code only: */
1238 #define jbd_ENOSYS() \
1240 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1241 current->state = TASK_UNINTERRUPTIBLE; \
1248 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1249 * bit, when set, indicates that we have had a fatal error somewhere,
1250 * either inside the journaling layer or indicated to us by the client
1251 * (eg. ext3), and that we and should not commit any further
1255 static inline int is_journal_aborted(journal_t *journal)
1257 return journal->j_flags & JBD2_ABORT;
1260 static inline int is_handle_aborted(handle_t *handle)
1262 if (handle->h_aborted)
1264 return is_journal_aborted(handle->h_transaction->t_journal);
1267 static inline void jbd2_journal_abort_handle(handle_t *handle)
1269 handle->h_aborted = 1;
1272 #endif /* __KERNEL__ */
1274 /* Comparison functions for transaction IDs: perform comparisons using
1275 * modulo arithmetic so that they work over sequence number wraps. */
1277 static inline int tid_gt(tid_t x, tid_t y)
1279 int difference = (x - y);
1280 return (difference > 0);
1283 static inline int tid_geq(tid_t x, tid_t y)
1285 int difference = (x - y);
1286 return (difference >= 0);
1289 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1290 extern size_t journal_tag_bytes(journal_t *journal);
1293 * Return the minimum number of blocks which must be free in the journal
1294 * before a new transaction may be started. Must be called under j_state_lock.
1296 static inline int jbd_space_needed(journal_t *journal)
1298 int nblocks = journal->j_max_transaction_buffers;
1299 if (journal->j_committing_transaction)
1300 nblocks += atomic_read(&journal->j_committing_transaction->
1301 t_outstanding_credits);
1306 * Definitions which augment the buffer_head layer
1309 /* journaling buffer types */
1310 #define BJ_None 0 /* Not journaled */
1311 #define BJ_Metadata 1 /* Normal journaled metadata */
1312 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1313 #define BJ_IO 3 /* Buffer is for temporary IO use */
1314 #define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1315 #define BJ_LogCtl 5 /* Buffer contains log descriptors */
1316 #define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1319 extern int jbd_blocks_per_page(struct inode *inode);
1323 #define buffer_trace_init(bh) do {} while (0)
1324 #define print_buffer_fields(bh) do {} while (0)
1325 #define print_buffer_trace(bh) do {} while (0)
1326 #define BUFFER_TRACE(bh, info) do {} while (0)
1327 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1328 #define JBUFFER_TRACE(jh, info) do {} while (0)
1331 * jbd2_dev_to_name is a utility function used by the jbd2 and ext4
1332 * tracing infrastructure to map a dev_t to a device name.
1334 extern const char *jbd2_dev_to_name(dev_t device);
1336 #endif /* __KERNEL__ */
1338 #endif /* _LINUX_JBD2_H */