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"
25 #include <linux/types.h>
26 #include <linux/buffer_head.h>
27 #include <linux/journal-head.h>
28 #include <linux/stddef.h>
29 #include <linux/bit_spinlock.h>
30 #include <linux/mutex.h>
31 #include <linux/timer.h>
32 #include <linux/slab.h>
33 #include <crypto/hash.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 ushort 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
150 #define JBD2_CRC32C_CHKSUM 4
152 #define JBD2_CRC32_CHKSUM_SIZE 4
154 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
156 * Commit block header for storing transactional checksums:
158 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
159 * fields are used to store a checksum of the descriptor and data blocks.
161 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
162 * field is used to store crc32c(uuid+commit_block). Each journal metadata
163 * block gets its own checksum, and data block checksums are stored in
164 * journal_block_tag (in the descriptor). The other h_chksum* fields are
167 * Checksum v1 and v2 are mutually exclusive features.
169 struct commit_header {
173 unsigned char h_chksum_type;
174 unsigned char h_chksum_size;
175 unsigned char h_padding[2];
176 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
178 __be32 h_commit_nsec;
182 * The block tag: used to describe a single buffer in the journal.
183 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
184 * raw struct shouldn't be used for pointer math or sizeof() - use
185 * journal_tag_bytes(journal) instead to compute this.
187 typedef struct journal_block_tag_s
189 __be32 t_blocknr; /* The on-disk block number */
190 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
191 __be16 t_flags; /* See below */
192 __be32 t_blocknr_high; /* most-significant high 32bits. */
193 } journal_block_tag_t;
195 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
196 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
198 /* Tail of descriptor block, for checksumming */
199 struct jbd2_journal_block_tail {
200 __be32 t_checksum; /* crc32c(uuid+descr_block) */
204 * The revoke descriptor: used on disk to describe a series of blocks to
205 * be revoked from the log
207 typedef struct jbd2_journal_revoke_header_s
209 journal_header_t r_header;
210 __be32 r_count; /* Count of bytes used in the block */
211 } jbd2_journal_revoke_header_t;
213 /* Tail of revoke block, for checksumming */
214 struct jbd2_journal_revoke_tail {
215 __be32 r_checksum; /* crc32c(uuid+revoke_block) */
218 /* Definitions for the journal tag flags word: */
219 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
220 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
221 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
222 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
226 * The journal superblock. All fields are in big-endian byte order.
228 typedef struct journal_superblock_s
231 journal_header_t s_header;
234 /* Static information describing the journal */
235 __be32 s_blocksize; /* journal device blocksize */
236 __be32 s_maxlen; /* total blocks in journal file */
237 __be32 s_first; /* first block of log information */
240 /* Dynamic information describing the current state of the log */
241 __be32 s_sequence; /* first commit ID expected in log */
242 __be32 s_start; /* blocknr of start of log */
245 /* Error value, as set by jbd2_journal_abort(). */
249 /* Remaining fields are only valid in a version-2 superblock */
250 __be32 s_feature_compat; /* compatible feature set */
251 __be32 s_feature_incompat; /* incompatible feature set */
252 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
254 __u8 s_uuid[16]; /* 128-bit uuid for journal */
257 __be32 s_nr_users; /* Nr of filesystems sharing log */
259 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
262 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
263 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
266 __u8 s_checksum_type; /* checksum type */
269 __be32 s_checksum; /* crc32c(superblock) */
272 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
274 } journal_superblock_t;
276 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
277 ((j)->j_format_version >= 2 && \
278 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
279 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
280 ((j)->j_format_version >= 2 && \
281 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
282 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
283 ((j)->j_format_version >= 2 && \
284 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
286 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
288 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
289 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
290 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
291 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
293 /* Features known to this kernel version: */
294 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
295 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
296 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
297 JBD2_FEATURE_INCOMPAT_64BIT | \
298 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
299 JBD2_FEATURE_INCOMPAT_CSUM_V2)
303 #include <linux/fs.h>
304 #include <linux/sched.h>
305 #include <linux/jbd_common.h>
307 #define J_ASSERT(assert) BUG_ON(!(assert))
309 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
310 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
312 #if defined(JBD2_PARANOID_IOFAIL)
313 #define J_EXPECT(expr, why...) J_ASSERT(expr)
314 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
315 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
317 #define __journal_expect(expr, why...) \
322 "JBD2 unexpected failure: %s: %s;\n", \
324 printk(KERN_ERR why "\n"); \
328 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
329 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
330 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
333 /* Flags in jbd_inode->i_flags */
334 #define __JI_COMMIT_RUNNING 0
335 /* Commit of the inode data in progress. We use this flag to protect us from
336 * concurrent deletion of inode. We cannot use reference to inode for this
337 * since we cannot afford doing last iput() on behalf of kjournald
339 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
342 * struct jbd_inode is the structure linking inodes in ordered mode
343 * present in a transaction so that we can sync them during commit.
346 /* Which transaction does this inode belong to? Either the running
347 * transaction or the committing one. [j_list_lock] */
348 transaction_t *i_transaction;
350 /* Pointer to the running transaction modifying inode's data in case
351 * there is already a committing transaction touching it. [j_list_lock] */
352 transaction_t *i_next_transaction;
354 /* List of inodes in the i_transaction [j_list_lock] */
355 struct list_head i_list;
357 /* VFS inode this inode belongs to [constant during the lifetime
358 * of the structure] */
359 struct inode *i_vfs_inode;
361 /* Flags of inode [j_list_lock] */
362 unsigned long i_flags;
365 struct jbd2_revoke_table_s;
368 * struct handle_s - The handle_s type is the concrete type associated with
370 * @h_transaction: Which compound transaction is this update a part of?
371 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
372 * @h_ref: Reference count on this handle
373 * @h_err: Field for caller's use to track errors through large fs operations
374 * @h_sync: flag for sync-on-close
375 * @h_jdata: flag to force data journaling
376 * @h_aborted: flag indicating fatal error on handle
379 /* Docbook can't yet cope with the bit fields, but will leave the documentation
380 * in so it can be fixed later.
383 struct jbd2_journal_handle
385 /* Which compound transaction is this update a part of? */
386 transaction_t *h_transaction;
388 /* Number of remaining buffers we are allowed to dirty: */
389 int h_buffer_credits;
391 /* Reference count on this handle */
394 /* Field for caller's use to track errors through large fs */
398 /* Flags [no locking] */
399 unsigned int h_sync: 1; /* sync-on-close */
400 unsigned int h_jdata: 1; /* force data journaling */
401 unsigned int h_aborted: 1; /* fatal error on handle */
402 unsigned int h_type: 8; /* for handle statistics */
403 unsigned int h_line_no: 16; /* for handle statistics */
405 unsigned long h_start_jiffies;
406 unsigned int h_requested_credits;
408 #ifdef CONFIG_DEBUG_LOCK_ALLOC
409 struct lockdep_map h_lockdep_map;
415 * Some stats for checkpoint phase
417 struct transaction_chp_stats_s {
418 unsigned long cs_chp_time;
419 __u32 cs_forced_to_close;
424 /* The transaction_t type is the guts of the journaling mechanism. It
425 * tracks a compound transaction through its various states:
427 * RUNNING: accepting new updates
428 * LOCKED: Updates still running but we don't accept new ones
429 * RUNDOWN: Updates are tidying up but have finished requesting
430 * new buffers to modify (state not used for now)
431 * FLUSH: All updates complete, but we are still writing to disk
432 * COMMIT: All data on disk, writing commit record
433 * FINISHED: We still have to keep the transaction for checkpointing.
435 * The transaction keeps track of all of the buffers modified by a
436 * running transaction, and all of the buffers committed but not yet
437 * flushed to home for finished transactions.
444 * ->jbd_lock_bh_journal_head() (This is "innermost")
447 * ->jbd_lock_bh_state()
449 * jbd_lock_bh_state()
456 * ->j_list_lock (journal_unmap_buffer)
462 /* Pointer to the journal for this transaction. [no locking] */
463 journal_t *t_journal;
465 /* Sequence number for this transaction [no locking] */
469 * Transaction's current state
470 * [no locking - only kjournald2 alters this]
471 * [j_list_lock] guards transition of a transaction into T_FINISHED
472 * state and subsequent call of __jbd2_journal_drop_transaction()
473 * FIXME: needs barriers
474 * KLUDGE: [use j_state_lock]
488 * Where in the log does this transaction's commit start? [no locking]
490 unsigned long t_log_start;
492 /* Number of buffers on the t_buffers list [j_list_lock] */
496 * Doubly-linked circular list of all buffers reserved but not yet
497 * modified by this transaction [j_list_lock]
499 struct journal_head *t_reserved_list;
502 * Doubly-linked circular list of all metadata buffers owned by this
503 * transaction [j_list_lock]
505 struct journal_head *t_buffers;
508 * Doubly-linked circular list of all forget buffers (superseded
509 * buffers which we can un-checkpoint once this transaction commits)
512 struct journal_head *t_forget;
515 * Doubly-linked circular list of all buffers still to be flushed before
516 * this transaction can be checkpointed. [j_list_lock]
518 struct journal_head *t_checkpoint_list;
521 * Doubly-linked circular list of all buffers submitted for IO while
522 * checkpointing. [j_list_lock]
524 struct journal_head *t_checkpoint_io_list;
527 * Doubly-linked circular list of temporary buffers currently undergoing
528 * IO in the log [j_list_lock]
530 struct journal_head *t_iobuf_list;
533 * Doubly-linked circular list of metadata buffers being shadowed by log
534 * IO. The IO buffers on the iobuf list and the shadow buffers on this
535 * list match each other one for one at all times. [j_list_lock]
537 struct journal_head *t_shadow_list;
540 * Doubly-linked circular list of control buffers being written to the
543 struct journal_head *t_log_list;
546 * List of inodes whose data we've modified in data=ordered mode.
549 struct list_head t_inode_list;
552 * Protects info related to handles
554 spinlock_t t_handle_lock;
557 * Longest time some handle had to wait for running transaction
559 unsigned long t_max_wait;
562 * When transaction started
564 unsigned long t_start;
567 * When commit was requested
569 unsigned long t_requested;
572 * Checkpointing stats [j_checkpoint_sem]
574 struct transaction_chp_stats_s t_chp_stats;
577 * Number of outstanding updates running on this transaction
583 * Number of buffers reserved for use by all handles in this transaction
584 * handle but not yet modified. [t_handle_lock]
586 atomic_t t_outstanding_credits;
589 * Forward and backward links for the circular list of all transactions
590 * awaiting checkpoint. [j_list_lock]
592 transaction_t *t_cpnext, *t_cpprev;
595 * When will the transaction expire (become due for commit), in jiffies?
598 unsigned long t_expires;
601 * When this transaction started, in nanoseconds [no locking]
603 ktime_t t_start_time;
606 * How many handles used this transaction? [t_handle_lock]
608 atomic_t t_handle_count;
611 * This transaction is being forced and some process is
612 * waiting for it to finish.
614 unsigned int t_synchronous_commit:1;
616 /* Disk flush needs to be sent to fs partition [no locking] */
617 int t_need_data_flush;
620 * For use by the filesystem to store fs-specific data
621 * structures associated with the transaction
623 struct list_head t_private_list;
626 struct transaction_run_stats_s {
627 unsigned long rs_wait;
628 unsigned long rs_request_delay;
629 unsigned long rs_running;
630 unsigned long rs_locked;
631 unsigned long rs_flushing;
632 unsigned long rs_logging;
634 __u32 rs_handle_count;
636 __u32 rs_blocks_logged;
639 struct transaction_stats_s {
640 unsigned long ts_tid;
641 unsigned long ts_requested;
642 struct transaction_run_stats_s run;
645 static inline unsigned long
646 jbd2_time_diff(unsigned long start, unsigned long end)
651 return end + (MAX_JIFFY_OFFSET - start);
654 #define JBD2_NR_BATCH 64
657 * struct journal_s - The journal_s type is the concrete type associated with
659 * @j_flags: General journaling state flags
660 * @j_errno: Is there an outstanding uncleared error on the journal (from a
662 * @j_sb_buffer: First part of superblock buffer
663 * @j_superblock: Second part of superblock buffer
664 * @j_format_version: Version of the superblock format
665 * @j_state_lock: Protect the various scalars in the journal
666 * @j_barrier_count: Number of processes waiting to create a barrier lock
667 * @j_barrier: The barrier lock itself
668 * @j_running_transaction: The current running transaction..
669 * @j_committing_transaction: the transaction we are pushing to disk
670 * @j_checkpoint_transactions: a linked circular list of all transactions
671 * waiting for checkpointing
672 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
673 * to start committing, or for a barrier lock to be released
674 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
675 * @j_wait_done_commit: Wait queue for waiting for commit to complete
676 * @j_wait_checkpoint: Wait queue to trigger checkpointing
677 * @j_wait_commit: Wait queue to trigger commit
678 * @j_wait_updates: Wait queue to wait for updates to complete
679 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
680 * @j_head: Journal head - identifies the first unused block in the journal
681 * @j_tail: Journal tail - identifies the oldest still-used block in the
683 * @j_free: Journal free - how many free blocks are there in the journal?
684 * @j_first: The block number of the first usable block
685 * @j_last: The block number one beyond the last usable block
686 * @j_dev: Device where we store the journal
687 * @j_blocksize: blocksize for the location where we store the journal.
688 * @j_blk_offset: starting block offset for into the device where we store the
690 * @j_fs_dev: Device which holds the client fs. For internal journal this will
692 * @j_maxlen: Total maximum capacity of the journal region on disk.
693 * @j_list_lock: Protects the buffer lists and internal buffer state.
694 * @j_inode: Optional inode where we store the journal. If present, all journal
695 * block numbers are mapped into this inode via bmap().
696 * @j_tail_sequence: Sequence number of the oldest transaction in the log
697 * @j_transaction_sequence: Sequence number of the next transaction to grant
698 * @j_commit_sequence: Sequence number of the most recently committed
700 * @j_commit_request: Sequence number of the most recent transaction wanting
702 * @j_uuid: Uuid of client object.
703 * @j_task: Pointer to the current commit thread for this journal
704 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
705 * single compound commit transaction
706 * @j_commit_interval: What is the maximum transaction lifetime before we begin
708 * @j_commit_timer: The timer used to wakeup the commit thread
709 * @j_revoke_lock: Protect the revoke table
710 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
711 * current transaction.
712 * @j_revoke_table: alternate revoke tables for j_revoke
713 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
714 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
715 * number that will fit in j_blocksize
716 * @j_last_sync_writer: most recent pid which did a synchronous write
717 * @j_history: Buffer storing the transactions statistics history
718 * @j_history_max: Maximum number of transactions in the statistics history
719 * @j_history_cur: Current number of transactions in the statistics history
720 * @j_history_lock: Protect the transactions statistics history
721 * @j_proc_entry: procfs entry for the jbd statistics directory
722 * @j_stats: Overall statistics
723 * @j_private: An opaque pointer to fs-private information.
728 /* General journaling state flags [j_state_lock] */
729 unsigned long j_flags;
732 * Is there an outstanding uncleared error on the journal (from a prior
733 * abort)? [j_state_lock]
737 /* The superblock buffer */
738 struct buffer_head *j_sb_buffer;
739 journal_superblock_t *j_superblock;
741 /* Version of the superblock format */
742 int j_format_version;
745 * Protect the various scalars in the journal
747 rwlock_t j_state_lock;
750 * Number of processes waiting to create a barrier lock [j_state_lock]
754 /* The barrier lock itself */
755 struct mutex j_barrier;
758 * Transactions: The current running transaction...
759 * [j_state_lock] [caller holding open handle]
761 transaction_t *j_running_transaction;
764 * the transaction we are pushing to disk
765 * [j_state_lock] [caller holding open handle]
767 transaction_t *j_committing_transaction;
770 * ... and a linked circular list of all transactions waiting for
771 * checkpointing. [j_list_lock]
773 transaction_t *j_checkpoint_transactions;
776 * Wait queue for waiting for a locked transaction to start committing,
777 * or for a barrier lock to be released
779 wait_queue_head_t j_wait_transaction_locked;
781 /* Wait queue for waiting for checkpointing to complete */
782 wait_queue_head_t j_wait_logspace;
784 /* Wait queue for waiting for commit to complete */
785 wait_queue_head_t j_wait_done_commit;
787 /* Wait queue to trigger checkpointing */
788 wait_queue_head_t j_wait_checkpoint;
790 /* Wait queue to trigger commit */
791 wait_queue_head_t j_wait_commit;
793 /* Wait queue to wait for updates to complete */
794 wait_queue_head_t j_wait_updates;
796 /* Semaphore for locking against concurrent checkpoints */
797 struct mutex j_checkpoint_mutex;
800 * List of buffer heads used by the checkpoint routine. This
801 * was moved from jbd2_log_do_checkpoint() to reduce stack
802 * usage. Access to this array is controlled by the
803 * j_checkpoint_mutex. [j_checkpoint_mutex]
805 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
808 * Journal head: identifies the first unused block in the journal.
811 unsigned long j_head;
814 * Journal tail: identifies the oldest still-used block in the journal.
817 unsigned long j_tail;
820 * Journal free: how many free blocks are there in the journal?
823 unsigned long j_free;
826 * Journal start and end: the block numbers of the first usable block
827 * and one beyond the last usable block in the journal. [j_state_lock]
829 unsigned long j_first;
830 unsigned long j_last;
833 * Device, blocksize and starting block offset for the location where we
836 struct block_device *j_dev;
838 unsigned long long j_blk_offset;
839 char j_devname[BDEVNAME_SIZE+24];
842 * Device which holds the client fs. For internal journal this will be
845 struct block_device *j_fs_dev;
847 /* Total maximum capacity of the journal region on disk. */
848 unsigned int j_maxlen;
851 * Protects the buffer lists and internal buffer state.
853 spinlock_t j_list_lock;
855 /* Optional inode where we store the journal. If present, all */
856 /* journal block numbers are mapped into this inode via */
858 struct inode *j_inode;
861 * Sequence number of the oldest transaction in the log [j_state_lock]
863 tid_t j_tail_sequence;
866 * Sequence number of the next transaction to grant [j_state_lock]
868 tid_t j_transaction_sequence;
871 * Sequence number of the most recently committed transaction
874 tid_t j_commit_sequence;
877 * Sequence number of the most recent transaction wanting commit
880 tid_t j_commit_request;
883 * Journal uuid: identifies the object (filesystem, LVM volume etc)
884 * backed by this journal. This will eventually be replaced by an array
885 * of uuids, allowing us to index multiple devices within a single
886 * journal and to perform atomic updates across them.
890 /* Pointer to the current commit thread for this journal */
891 struct task_struct *j_task;
894 * Maximum number of metadata buffers to allow in a single compound
897 int j_max_transaction_buffers;
900 * What is the maximum transaction lifetime before we begin a commit?
902 unsigned long j_commit_interval;
904 /* The timer used to wakeup the commit thread: */
905 struct timer_list j_commit_timer;
908 * The revoke table: maintains the list of revoked blocks in the
909 * current transaction. [j_revoke_lock]
911 spinlock_t j_revoke_lock;
912 struct jbd2_revoke_table_s *j_revoke;
913 struct jbd2_revoke_table_s *j_revoke_table[2];
916 * array of bhs for jbd2_journal_commit_transaction
918 struct buffer_head **j_wbuf;
922 * this is the pid of hte last person to run a synchronous operation
923 * through the journal
925 pid_t j_last_sync_writer;
928 * the average amount of time in nanoseconds it takes to commit a
929 * transaction to disk. [j_state_lock]
931 u64 j_average_commit_time;
934 * minimum and maximum times that we should wait for
935 * additional filesystem operations to get batched into a
936 * synchronous handle in microseconds
938 u32 j_min_batch_time;
939 u32 j_max_batch_time;
941 /* This function is called when a transaction is closed */
942 void (*j_commit_callback)(journal_t *,
948 spinlock_t j_history_lock;
949 struct proc_dir_entry *j_proc_entry;
950 struct transaction_stats_s j_stats;
952 /* Failed journal commit ID */
953 unsigned int j_failed_commit;
956 * An opaque pointer to fs-private information. ext3 puts its
957 * superblock pointer here
961 /* Reference to checksum algorithm driver via cryptoapi */
962 struct crypto_shash *j_chksum_driver;
964 /* Precomputed journal UUID checksum for seeding other checksums */
969 * Journal flag definitions
971 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
972 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
973 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
974 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
975 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
976 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
977 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
978 * data write error in ordered
982 * Function declarations for the journaling transaction and buffer
987 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
988 extern void __jbd2_journal_refile_buffer(struct journal_head *);
989 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
990 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
991 extern void __journal_free_buffer(struct journal_head *bh);
992 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
993 extern void __journal_clean_data_list(transaction_t *transaction);
995 /* Log buffer allocation */
996 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
997 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
998 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
999 unsigned long *block);
1000 void __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1001 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1003 /* Commit management */
1004 extern void jbd2_journal_commit_transaction(journal_t *);
1006 /* Checkpoint list management */
1007 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1008 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1009 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1016 struct jbd2_buffer_trigger_type {
1018 * Fired a the moment data to write to the journal are known to be
1019 * stable - so either at the moment b_frozen_data is created or just
1020 * before a buffer is written to the journal. mapped_data is a mapped
1021 * buffer that is the frozen data for commit.
1023 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1024 struct buffer_head *bh, void *mapped_data,
1028 * Fired during journal abort for dirty buffers that will not be
1031 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1032 struct buffer_head *bh);
1035 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1037 struct jbd2_buffer_trigger_type *triggers);
1038 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1039 struct jbd2_buffer_trigger_type *triggers);
1043 jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1044 struct journal_head *jh_in,
1045 struct journal_head **jh_out,
1046 unsigned long long blocknr);
1048 /* Transaction locking */
1049 extern void __wait_on_journal (journal_t *);
1051 /* Transaction cache support */
1052 extern void jbd2_journal_destroy_transaction_cache(void);
1053 extern int jbd2_journal_init_transaction_cache(void);
1054 extern void jbd2_journal_free_transaction(transaction_t *);
1059 * We need to lock the journal during transaction state changes so that nobody
1060 * ever tries to take a handle on the running transaction while we are in the
1061 * middle of moving it to the commit phase. j_state_lock does this.
1063 * Note that the locking is completely interrupt unsafe. We never touch
1064 * journal structures from interrupts.
1067 static inline handle_t *journal_current_handle(void)
1069 return current->journal_info;
1072 /* The journaling code user interface:
1074 * Create and destroy handles
1075 * Register buffer modifications against the current transaction.
1078 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1079 extern handle_t *jbd2__journal_start(journal_t *, int nblocks, gfp_t gfp_mask,
1080 unsigned int type, unsigned int line_no);
1081 extern int jbd2_journal_restart(handle_t *, int nblocks);
1082 extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
1083 extern int jbd2_journal_extend (handle_t *, int nblocks);
1084 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1085 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1086 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1087 void jbd2_journal_set_triggers(struct buffer_head *,
1088 struct jbd2_buffer_trigger_type *type);
1089 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1090 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1091 extern void journal_sync_buffer (struct buffer_head *);
1092 extern int jbd2_journal_invalidatepage(journal_t *,
1093 struct page *, unsigned long);
1094 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1095 extern int jbd2_journal_stop(handle_t *);
1096 extern int jbd2_journal_flush (journal_t *);
1097 extern void jbd2_journal_lock_updates (journal_t *);
1098 extern void jbd2_journal_unlock_updates (journal_t *);
1100 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1101 struct block_device *fs_dev,
1102 unsigned long long start, int len, int bsize);
1103 extern journal_t * jbd2_journal_init_inode (struct inode *);
1104 extern int jbd2_journal_update_format (journal_t *);
1105 extern int jbd2_journal_check_used_features
1106 (journal_t *, unsigned long, unsigned long, unsigned long);
1107 extern int jbd2_journal_check_available_features
1108 (journal_t *, unsigned long, unsigned long, unsigned long);
1109 extern int jbd2_journal_set_features
1110 (journal_t *, unsigned long, unsigned long, unsigned long);
1111 extern void jbd2_journal_clear_features
1112 (journal_t *, unsigned long, unsigned long, unsigned long);
1113 extern int jbd2_journal_load (journal_t *journal);
1114 extern int jbd2_journal_destroy (journal_t *);
1115 extern int jbd2_journal_recover (journal_t *journal);
1116 extern int jbd2_journal_wipe (journal_t *, int);
1117 extern int jbd2_journal_skip_recovery (journal_t *);
1118 extern void jbd2_journal_update_sb_errno(journal_t *);
1119 extern void jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
1120 unsigned long, int);
1121 extern void __jbd2_journal_abort_hard (journal_t *);
1122 extern void jbd2_journal_abort (journal_t *, int);
1123 extern int jbd2_journal_errno (journal_t *);
1124 extern void jbd2_journal_ack_err (journal_t *);
1125 extern int jbd2_journal_clear_err (journal_t *);
1126 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1127 extern int jbd2_journal_force_commit(journal_t *);
1128 extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1129 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1130 struct jbd2_inode *inode, loff_t new_size);
1131 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1132 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1135 * journal_head management
1137 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1138 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1139 void jbd2_journal_put_journal_head(struct journal_head *jh);
1144 extern struct kmem_cache *jbd2_handle_cache;
1146 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1148 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1151 static inline void jbd2_free_handle(handle_t *handle)
1153 kmem_cache_free(jbd2_handle_cache, handle);
1157 * jbd2_inode management (optional, for those file systems that want to use
1158 * dynamically allocated jbd2_inode structures)
1160 extern struct kmem_cache *jbd2_inode_cache;
1162 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1164 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1167 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1169 kmem_cache_free(jbd2_inode_cache, jinode);
1172 /* Primary revoke support */
1173 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1174 extern int jbd2_journal_init_revoke(journal_t *, int);
1175 extern void jbd2_journal_destroy_revoke_caches(void);
1176 extern int jbd2_journal_init_revoke_caches(void);
1178 extern void jbd2_journal_destroy_revoke(journal_t *);
1179 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1180 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1181 extern void jbd2_journal_write_revoke_records(journal_t *,
1182 transaction_t *, int);
1184 /* Recovery revoke support */
1185 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1186 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1187 extern void jbd2_journal_clear_revoke(journal_t *);
1188 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1189 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
1192 * The log thread user interface:
1194 * Request space in the current transaction, and force transaction commit
1195 * transitions on demand.
1198 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1199 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1200 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1201 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1202 int jbd2_journal_force_commit_nested(journal_t *journal);
1203 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1204 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1205 int jbd2_log_do_checkpoint(journal_t *journal);
1206 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1208 void __jbd2_log_wait_for_space(journal_t *journal);
1209 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1210 extern int jbd2_cleanup_journal_tail(journal_t *);
1212 /* Debugging code only: */
1214 #define jbd_ENOSYS() \
1216 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1217 current->state = TASK_UNINTERRUPTIBLE; \
1224 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1225 * bit, when set, indicates that we have had a fatal error somewhere,
1226 * either inside the journaling layer or indicated to us by the client
1227 * (eg. ext3), and that we and should not commit any further
1231 static inline int is_journal_aborted(journal_t *journal)
1233 return journal->j_flags & JBD2_ABORT;
1236 static inline int is_handle_aborted(handle_t *handle)
1238 if (handle->h_aborted)
1240 return is_journal_aborted(handle->h_transaction->t_journal);
1243 static inline void jbd2_journal_abort_handle(handle_t *handle)
1245 handle->h_aborted = 1;
1248 #endif /* __KERNEL__ */
1250 /* Comparison functions for transaction IDs: perform comparisons using
1251 * modulo arithmetic so that they work over sequence number wraps. */
1253 static inline int tid_gt(tid_t x, tid_t y)
1255 int difference = (x - y);
1256 return (difference > 0);
1259 static inline int tid_geq(tid_t x, tid_t y)
1261 int difference = (x - y);
1262 return (difference >= 0);
1265 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1266 extern size_t journal_tag_bytes(journal_t *journal);
1269 * Return the minimum number of blocks which must be free in the journal
1270 * before a new transaction may be started. Must be called under j_state_lock.
1272 static inline int jbd_space_needed(journal_t *journal)
1274 int nblocks = journal->j_max_transaction_buffers;
1275 if (journal->j_committing_transaction)
1276 nblocks += atomic_read(&journal->j_committing_transaction->
1277 t_outstanding_credits);
1282 * Definitions which augment the buffer_head layer
1285 /* journaling buffer types */
1286 #define BJ_None 0 /* Not journaled */
1287 #define BJ_Metadata 1 /* Normal journaled metadata */
1288 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1289 #define BJ_IO 3 /* Buffer is for temporary IO use */
1290 #define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1291 #define BJ_LogCtl 5 /* Buffer contains log descriptors */
1292 #define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1295 extern int jbd_blocks_per_page(struct inode *inode);
1297 /* JBD uses a CRC32 checksum */
1298 #define JBD_MAX_CHECKSUM_SIZE 4
1300 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1301 const void *address, unsigned int length)
1304 struct shash_desc shash;
1305 char ctx[JBD_MAX_CHECKSUM_SIZE];
1309 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1310 JBD_MAX_CHECKSUM_SIZE);
1312 desc.shash.tfm = journal->j_chksum_driver;
1313 desc.shash.flags = 0;
1314 *(u32 *)desc.ctx = crc;
1316 err = crypto_shash_update(&desc.shash, address, length);
1319 return *(u32 *)desc.ctx;
1324 #define buffer_trace_init(bh) do {} while (0)
1325 #define print_buffer_fields(bh) do {} while (0)
1326 #define print_buffer_trace(bh) do {} while (0)
1327 #define BUFFER_TRACE(bh, info) do {} while (0)
1328 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1329 #define JBUFFER_TRACE(jh, info) do {} while (0)
1331 #endif /* __KERNEL__ */
1333 #endif /* _LINUX_JBD2_H */