1 /* SPDX-License-Identifier: GPL-2.0-or-later */
3 * linux/include/linux/jbd2.h
5 * Written by Stephen C. Tweedie <sct@redhat.com>
7 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
9 * Definitions for transaction data structures for the buffer cache
10 * filesystem journaling support.
16 /* Allow this file to be included directly into e2fsprogs */
18 #include "jfs_compat.h"
22 #include <linux/types.h>
23 #include <linux/buffer_head.h>
24 #include <linux/journal-head.h>
25 #include <linux/stddef.h>
26 #include <linux/mutex.h>
27 #include <linux/timer.h>
28 #include <linux/slab.h>
29 #include <linux/bit_spinlock.h>
30 #include <linux/blkdev.h>
31 #include <crypto/hash.h>
34 #define journal_oom_retry 1
37 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
38 * certain classes of error which can occur due to failed IOs. Under
39 * normal use we want ext4 to continue after such errors, because
40 * hardware _can_ fail, but for debugging purposes when running tests on
41 * known-good hardware we may want to trap these errors.
43 #undef JBD2_PARANOID_IOFAIL
46 * The default maximum commit age, in seconds.
48 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
50 #ifdef CONFIG_JBD2_DEBUG
52 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
53 * consistency checks. By default we don't do this unless
54 * CONFIG_JBD2_DEBUG is on.
56 #define JBD2_EXPENSIVE_CHECKING
57 extern ushort jbd2_journal_enable_debug;
58 void __jbd2_debug(int level, const char *file, const char *func,
59 unsigned int line, const char *fmt, ...);
61 #define jbd_debug(n, fmt, a...) \
62 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
64 #define jbd_debug(n, fmt, a...) no_printk(fmt, ##a)
67 extern void *jbd2_alloc(size_t size, gfp_t flags);
68 extern void jbd2_free(void *ptr, size_t size);
70 #define JBD2_MIN_JOURNAL_BLOCKS 1024
71 #define JBD2_DEFAULT_FAST_COMMIT_BLOCKS 256
76 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
78 * All filesystem modifications made by the process go
79 * through this handle. Recursive operations (such as quota operations)
80 * are gathered into a single update.
82 * The buffer credits field is used to account for journaled buffers
83 * being modified by the running process. To ensure that there is
84 * enough log space for all outstanding operations, we need to limit the
85 * number of outstanding buffers possible at any time. When the
86 * operation completes, any buffer credits not used are credited back to
87 * the transaction, so that at all times we know how many buffers the
88 * outstanding updates on a transaction might possibly touch.
90 * This is an opaque datatype.
92 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
96 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
98 * journal_t is linked to from the fs superblock structure.
100 * We use the journal_t to keep track of all outstanding transaction
101 * activity on the filesystem, and to manage the state of the log
104 * This is an opaque datatype.
106 typedef struct journal_s journal_t; /* Journal control structure */
110 * Internal structures used by the logging mechanism:
113 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
120 * Descriptor block types:
123 #define JBD2_DESCRIPTOR_BLOCK 1
124 #define JBD2_COMMIT_BLOCK 2
125 #define JBD2_SUPERBLOCK_V1 3
126 #define JBD2_SUPERBLOCK_V2 4
127 #define JBD2_REVOKE_BLOCK 5
130 * Standard header for all descriptor blocks:
132 typedef struct journal_header_s
142 #define JBD2_CRC32_CHKSUM 1
143 #define JBD2_MD5_CHKSUM 2
144 #define JBD2_SHA1_CHKSUM 3
145 #define JBD2_CRC32C_CHKSUM 4
147 #define JBD2_CRC32_CHKSUM_SIZE 4
149 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
151 * Commit block header for storing transactional checksums:
153 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
154 * fields are used to store a checksum of the descriptor and data blocks.
156 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
157 * field is used to store crc32c(uuid+commit_block). Each journal metadata
158 * block gets its own checksum, and data block checksums are stored in
159 * journal_block_tag (in the descriptor). The other h_chksum* fields are
162 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
163 * journal_block_tag3_t to store a full 32-bit checksum. Everything else
166 * Checksum v1, v2, and v3 are mutually exclusive features.
168 struct commit_header {
172 unsigned char h_chksum_type;
173 unsigned char h_chksum_size;
174 unsigned char h_padding[2];
175 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
177 __be32 h_commit_nsec;
181 * The block tag: used to describe a single buffer in the journal.
182 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
183 * raw struct shouldn't be used for pointer math or sizeof() - use
184 * journal_tag_bytes(journal) instead to compute this.
186 typedef struct journal_block_tag3_s
188 __be32 t_blocknr; /* The on-disk block number */
189 __be32 t_flags; /* See below */
190 __be32 t_blocknr_high; /* most-significant high 32bits. */
191 __be32 t_checksum; /* crc32c(uuid+seq+block) */
192 } journal_block_tag3_t;
194 typedef struct journal_block_tag_s
196 __be32 t_blocknr; /* The on-disk block number */
197 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
198 __be16 t_flags; /* See below */
199 __be32 t_blocknr_high; /* most-significant high 32bits. */
200 } journal_block_tag_t;
202 /* Tail of descriptor or revoke block, for checksumming */
203 struct jbd2_journal_block_tail {
204 __be32 t_checksum; /* crc32c(uuid+descr_block) */
208 * The revoke descriptor: used on disk to describe a series of blocks to
209 * be revoked from the log
211 typedef struct jbd2_journal_revoke_header_s
213 journal_header_t r_header;
214 __be32 r_count; /* Count of bytes used in the block */
215 } jbd2_journal_revoke_header_t;
217 /* Definitions for the journal tag flags word: */
218 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
219 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
220 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
221 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
225 * The journal superblock. All fields are in big-endian byte order.
227 typedef struct journal_superblock_s
230 journal_header_t s_header;
233 /* Static information describing the journal */
234 __be32 s_blocksize; /* journal device blocksize */
235 __be32 s_maxlen; /* total blocks in journal file */
236 __be32 s_first; /* first block of log information */
239 /* Dynamic information describing the current state of the log */
240 __be32 s_sequence; /* first commit ID expected in log */
241 __be32 s_start; /* blocknr of start of log */
244 /* Error value, as set by jbd2_journal_abort(). */
248 /* Remaining fields are only valid in a version-2 superblock */
249 __be32 s_feature_compat; /* compatible feature set */
250 __be32 s_feature_incompat; /* incompatible feature set */
251 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
253 __u8 s_uuid[16]; /* 128-bit uuid for journal */
256 __be32 s_nr_users; /* Nr of filesystems sharing log */
258 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
261 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
262 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
265 __u8 s_checksum_type; /* checksum type */
268 __be32 s_num_fc_blks; /* Number of fast commit blocks */
271 __be32 s_checksum; /* crc32c(superblock) */
274 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
276 } journal_superblock_t;
278 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
279 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
280 ((j)->j_format_version >= 2 && \
281 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
282 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
283 ((j)->j_format_version >= 2 && \
284 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
285 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
286 ((j)->j_format_version >= 2 && \
287 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
289 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
291 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
292 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
293 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
294 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
295 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010
296 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT 0x00000020
298 /* See "journal feature predicate functions" below */
300 /* Features known to this kernel version: */
301 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
302 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
303 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
304 JBD2_FEATURE_INCOMPAT_64BIT | \
305 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
306 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
307 JBD2_FEATURE_INCOMPAT_CSUM_V3 | \
308 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)
312 #include <linux/fs.h>
313 #include <linux/sched.h>
315 enum jbd_state_bits {
316 BH_JBD /* Has an attached ext3 journal_head */
318 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
319 BH_Freed, /* Has been freed (truncated) */
320 BH_Revoked, /* Has been revoked from the log */
321 BH_RevokeValid, /* Revoked flag is valid */
322 BH_JBDDirty, /* Is dirty but journaled */
323 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
324 BH_Shadow, /* IO on shadow buffer is running */
325 BH_Verified, /* Metadata block has been verified ok */
326 BH_JBDPrivateStart, /* First bit available for private use by FS */
330 BUFFER_FNS(JWrite, jwrite)
331 BUFFER_FNS(JBDDirty, jbddirty)
332 TAS_BUFFER_FNS(JBDDirty, jbddirty)
333 BUFFER_FNS(Revoked, revoked)
334 TAS_BUFFER_FNS(Revoked, revoked)
335 BUFFER_FNS(RevokeValid, revokevalid)
336 TAS_BUFFER_FNS(RevokeValid, revokevalid)
337 BUFFER_FNS(Freed, freed)
338 BUFFER_FNS(Shadow, shadow)
339 BUFFER_FNS(Verified, verified)
341 static inline struct buffer_head *jh2bh(struct journal_head *jh)
346 static inline struct journal_head *bh2jh(struct buffer_head *bh)
348 return bh->b_private;
351 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
353 bit_spin_lock(BH_JournalHead, &bh->b_state);
356 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
358 bit_spin_unlock(BH_JournalHead, &bh->b_state);
361 #define J_ASSERT(assert) BUG_ON(!(assert))
363 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
364 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
366 #if defined(JBD2_PARANOID_IOFAIL)
367 #define J_EXPECT(expr, why...) J_ASSERT(expr)
368 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
369 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
371 #define __journal_expect(expr, why...) \
376 "JBD2 unexpected failure: %s: %s;\n", \
378 printk(KERN_ERR why "\n"); \
382 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
383 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
384 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
387 /* Flags in jbd_inode->i_flags */
388 #define __JI_COMMIT_RUNNING 0
389 #define __JI_WRITE_DATA 1
390 #define __JI_WAIT_DATA 2
393 * Commit of the inode data in progress. We use this flag to protect us from
394 * concurrent deletion of inode. We cannot use reference to inode for this
395 * since we cannot afford doing last iput() on behalf of kjournald
397 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
398 /* Write allocated dirty buffers in this inode before commit */
399 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
400 /* Wait for outstanding data writes for this inode before commit */
401 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
404 * struct jbd2_inode - The jbd_inode type is the structure linking inodes in
405 * ordered mode present in a transaction so that we can sync them during commit.
411 * Which transaction does this inode belong to? Either the running
412 * transaction or the committing one. [j_list_lock]
414 transaction_t *i_transaction;
417 * @i_next_transaction:
419 * Pointer to the running transaction modifying inode's data in case
420 * there is already a committing transaction touching it. [j_list_lock]
422 transaction_t *i_next_transaction;
425 * @i_list: List of inodes in the i_transaction [j_list_lock]
427 struct list_head i_list;
432 * VFS inode this inode belongs to [constant for lifetime of structure]
434 struct inode *i_vfs_inode;
437 * @i_flags: Flags of inode [j_list_lock]
439 unsigned long i_flags;
444 * Offset in bytes where the dirty range for this inode starts.
447 loff_t i_dirty_start;
452 * Inclusive offset in bytes where the dirty range for this inode
453 * ends. [j_list_lock]
458 struct jbd2_revoke_table_s;
461 * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete
462 * type associated with handle_t.
463 * @h_transaction: Which compound transaction is this update a part of?
464 * @h_journal: Which journal handle belongs to - used iff h_reserved set.
465 * @h_rsv_handle: Handle reserved for finishing the logical operation.
466 * @h_total_credits: Number of remaining buffers we are allowed to add to
467 * journal. These are dirty buffers and revoke descriptor blocks.
468 * @h_revoke_credits: Number of remaining revoke records available for handle
469 * @h_ref: Reference count on this handle.
470 * @h_err: Field for caller's use to track errors through large fs operations.
471 * @h_sync: Flag for sync-on-close.
472 * @h_jdata: Flag to force data journaling.
473 * @h_reserved: Flag for handle for reserved credits.
474 * @h_aborted: Flag indicating fatal error on handle.
475 * @h_type: For handle statistics.
476 * @h_line_no: For handle statistics.
477 * @h_start_jiffies: Handle Start time.
478 * @h_requested_credits: Holds @h_total_credits after handle is started.
479 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
480 * @saved_alloc_context: Saved context while transaction is open.
483 /* Docbook can't yet cope with the bit fields, but will leave the documentation
484 * in so it can be fixed later.
487 struct jbd2_journal_handle
490 transaction_t *h_transaction;
491 /* Which journal handle belongs to - used iff h_reserved set */
492 journal_t *h_journal;
495 handle_t *h_rsv_handle;
497 int h_revoke_credits;
498 int h_revoke_credits_requested;
502 /* Flags [no locking] */
503 unsigned int h_sync: 1;
504 unsigned int h_jdata: 1;
505 unsigned int h_reserved: 1;
506 unsigned int h_aborted: 1;
507 unsigned int h_type: 8;
508 unsigned int h_line_no: 16;
510 unsigned long h_start_jiffies;
511 unsigned int h_requested_credits;
513 unsigned int saved_alloc_context;
518 * Some stats for checkpoint phase
520 struct transaction_chp_stats_s {
521 unsigned long cs_chp_time;
522 __u32 cs_forced_to_close;
527 /* The transaction_t type is the guts of the journaling mechanism. It
528 * tracks a compound transaction through its various states:
530 * RUNNING: accepting new updates
531 * LOCKED: Updates still running but we don't accept new ones
532 * RUNDOWN: Updates are tidying up but have finished requesting
533 * new buffers to modify (state not used for now)
534 * FLUSH: All updates complete, but we are still writing to disk
535 * COMMIT: All data on disk, writing commit record
536 * FINISHED: We still have to keep the transaction for checkpointing.
538 * The transaction keeps track of all of the buffers modified by a
539 * running transaction, and all of the buffers committed but not yet
540 * flushed to home for finished transactions.
541 * (Locking Documentation improved by LockDoc)
548 * ->jbd_lock_bh_journal_head() (This is "innermost")
560 * ->j_list_lock (journal_unmap_buffer)
566 /* Pointer to the journal for this transaction. [no locking] */
567 journal_t *t_journal;
569 /* Sequence number for this transaction [no locking] */
573 * Transaction's current state
574 * [no locking - only kjournald2 alters this]
575 * [j_list_lock] guards transition of a transaction into T_FINISHED
576 * state and subsequent call of __jbd2_journal_drop_transaction()
577 * FIXME: needs barriers
578 * KLUDGE: [use j_state_lock]
593 * Where in the log does this transaction's commit start? [no locking]
595 unsigned long t_log_start;
598 * Number of buffers on the t_buffers list [j_list_lock, no locks
599 * needed for jbd2 thread]
604 * Doubly-linked circular list of all buffers reserved but not yet
605 * modified by this transaction [j_list_lock, no locks needed fo
608 struct journal_head *t_reserved_list;
611 * Doubly-linked circular list of all metadata buffers owned by this
612 * transaction [j_list_lock, no locks needed for jbd2 thread]
614 struct journal_head *t_buffers;
617 * Doubly-linked circular list of all forget buffers (superseded
618 * buffers which we can un-checkpoint once this transaction commits)
621 struct journal_head *t_forget;
624 * Doubly-linked circular list of all buffers still to be flushed before
625 * this transaction can be checkpointed. [j_list_lock]
627 struct journal_head *t_checkpoint_list;
630 * Doubly-linked circular list of metadata buffers being
631 * shadowed by log IO. The IO buffers on the iobuf list and
632 * the shadow buffers on this list match each other one for
633 * one at all times. [j_list_lock, no locks needed for jbd2
636 struct journal_head *t_shadow_list;
639 * List of inodes associated with the transaction; e.g., ext4 uses
640 * this to track inodes in data=ordered and data=journal mode that
641 * need special handling on transaction commit; also used by ocfs2.
644 struct list_head t_inode_list;
647 * Protects info related to handles
649 spinlock_t t_handle_lock;
652 * Longest time some handle had to wait for running transaction
654 unsigned long t_max_wait;
657 * When transaction started
659 unsigned long t_start;
662 * When commit was requested [j_state_lock]
664 unsigned long t_requested;
667 * Checkpointing stats [j_list_lock]
669 struct transaction_chp_stats_s t_chp_stats;
672 * Number of outstanding updates running on this transaction
678 * Number of blocks reserved for this transaction in the journal.
679 * This is including all credits reserved when starting transaction
680 * handles as well as all journal descriptor blocks needed for this
681 * transaction. [none]
683 atomic_t t_outstanding_credits;
686 * Number of revoke records for this transaction added by already
687 * stopped handles. [none]
689 atomic_t t_outstanding_revokes;
692 * How many handles used this transaction? [none]
694 atomic_t t_handle_count;
697 * Forward and backward links for the circular list of all transactions
698 * awaiting checkpoint. [j_list_lock]
700 transaction_t *t_cpnext, *t_cpprev;
703 * When will the transaction expire (become due for commit), in jiffies?
706 unsigned long t_expires;
709 * When this transaction started, in nanoseconds [no locking]
711 ktime_t t_start_time;
714 * This transaction is being forced and some process is
715 * waiting for it to finish.
717 unsigned int t_synchronous_commit:1;
719 /* Disk flush needs to be sent to fs partition [no locking] */
720 int t_need_data_flush;
723 * For use by the filesystem to store fs-specific data
724 * structures associated with the transaction
726 struct list_head t_private_list;
729 struct transaction_run_stats_s {
730 unsigned long rs_wait;
731 unsigned long rs_request_delay;
732 unsigned long rs_running;
733 unsigned long rs_locked;
734 unsigned long rs_flushing;
735 unsigned long rs_logging;
737 __u32 rs_handle_count;
739 __u32 rs_blocks_logged;
742 struct transaction_stats_s {
743 unsigned long ts_tid;
744 unsigned long ts_requested;
745 struct transaction_run_stats_s run;
748 static inline unsigned long
749 jbd2_time_diff(unsigned long start, unsigned long end)
754 return end + (MAX_JIFFY_OFFSET - start);
757 #define JBD2_NR_BATCH 64
759 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
761 #define JBD2_FC_REPLAY_STOP 0
762 #define JBD2_FC_REPLAY_CONTINUE 1
765 * struct journal_s - The journal_s type is the concrete type associated with
771 * @j_flags: General journaling state flags [j_state_lock,
772 * no lock for quick racy checks]
774 unsigned long j_flags;
777 * @j_atomic_flags: Atomic journaling state flags.
779 unsigned long j_atomic_flags;
784 * Is there an outstanding uncleared error on the journal (from a prior
785 * abort)? [j_state_lock]
790 * @j_abort_mutex: Lock the whole aborting procedure.
792 struct mutex j_abort_mutex;
795 * @j_sb_buffer: The first part of the superblock buffer.
797 struct buffer_head *j_sb_buffer;
800 * @j_superblock: The second part of the superblock buffer.
802 journal_superblock_t *j_superblock;
805 * @j_format_version: Version of the superblock format.
807 int j_format_version;
810 * @j_state_lock: Protect the various scalars in the journal.
812 rwlock_t j_state_lock;
817 * Number of processes waiting to create a barrier lock [j_state_lock,
818 * no lock for quick racy checks]
823 * @j_barrier: The barrier lock itself.
825 struct mutex j_barrier;
828 * @j_running_transaction:
830 * Transactions: The current running transaction...
831 * [j_state_lock, no lock for quick racy checks] [caller holding
834 transaction_t *j_running_transaction;
837 * @j_committing_transaction:
839 * the transaction we are pushing to disk
840 * [j_state_lock] [caller holding open handle]
842 transaction_t *j_committing_transaction;
845 * @j_checkpoint_transactions:
847 * ... and a linked circular list of all transactions waiting for
848 * checkpointing. [j_list_lock]
850 transaction_t *j_checkpoint_transactions;
853 * @j_wait_transaction_locked:
855 * Wait queue for waiting for a locked transaction to start committing,
856 * or for a barrier lock to be released.
858 wait_queue_head_t j_wait_transaction_locked;
861 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
863 wait_queue_head_t j_wait_done_commit;
866 * @j_wait_commit: Wait queue to trigger commit.
868 wait_queue_head_t j_wait_commit;
871 * @j_wait_updates: Wait queue to wait for updates to complete.
873 wait_queue_head_t j_wait_updates;
878 * Wait queue to wait for reserved buffer credits to drop.
880 wait_queue_head_t j_wait_reserved;
885 * Wait queue to wait for completion of async fast commits.
887 wait_queue_head_t j_fc_wait;
890 * @j_checkpoint_mutex:
892 * Semaphore for locking against concurrent checkpoints.
894 struct mutex j_checkpoint_mutex;
899 * List of buffer heads used by the checkpoint routine. This
900 * was moved from jbd2_log_do_checkpoint() to reduce stack
901 * usage. Access to this array is controlled by the
902 * @j_checkpoint_mutex. [j_checkpoint_mutex]
904 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
909 * Journal head shrinker, reclaim buffer's journal head which
910 * has been written back.
912 struct shrinker j_shrinker;
915 * @j_checkpoint_jh_count:
917 * Number of journal buffers on the checkpoint list. [j_list_lock]
919 struct percpu_counter j_checkpoint_jh_count;
922 * @j_shrink_transaction:
924 * Record next transaction will shrink on the checkpoint list.
927 transaction_t *j_shrink_transaction;
932 * Journal head: identifies the first unused block in the journal.
935 unsigned long j_head;
940 * Journal tail: identifies the oldest still-used block in the journal.
943 unsigned long j_tail;
948 * Journal free: how many free blocks are there in the journal?
951 unsigned long j_free;
956 * The block number of the first usable block in the journal
959 unsigned long j_first;
964 * The block number one beyond the last usable block in the journal
967 unsigned long j_last;
972 * The block number of the first fast commit block in the journal
975 unsigned long j_fc_first;
980 * Number of fast commit blocks currently allocated. Accessed only
981 * during fast commit. Currently only process can do fast commit, so
982 * this field is not protected by any lock.
984 unsigned long j_fc_off;
989 * The block number one beyond the last fast commit block in the journal
992 unsigned long j_fc_last;
995 * @j_dev: Device where we store the journal.
997 struct block_device *j_dev;
1000 * @j_blocksize: Block size for the location where we store the journal.
1007 * Starting block offset into the device where we store the journal.
1009 unsigned long long j_blk_offset;
1012 * @j_devname: Journal device name.
1014 char j_devname[BDEVNAME_SIZE+24];
1019 * Device which holds the client fs. For internal journal this will be
1022 struct block_device *j_fs_dev;
1025 * @j_total_len: Total maximum capacity of the journal region on disk.
1027 unsigned int j_total_len;
1030 * @j_reserved_credits:
1032 * Number of buffers reserved from the running transaction.
1034 atomic_t j_reserved_credits;
1037 * @j_list_lock: Protects the buffer lists and internal buffer state.
1039 spinlock_t j_list_lock;
1044 * Optional inode where we store the journal. If present, all
1045 * journal block numbers are mapped into this inode via bmap().
1047 struct inode *j_inode;
1052 * Sequence number of the oldest transaction in the log [j_state_lock]
1054 tid_t j_tail_sequence;
1057 * @j_transaction_sequence:
1059 * Sequence number of the next transaction to grant [j_state_lock]
1061 tid_t j_transaction_sequence;
1064 * @j_commit_sequence:
1066 * Sequence number of the most recently committed transaction
1067 * [j_state_lock, no lock for quick racy checks]
1069 tid_t j_commit_sequence;
1072 * @j_commit_request:
1074 * Sequence number of the most recent transaction wanting commit
1075 * [j_state_lock, no lock for quick racy checks]
1077 tid_t j_commit_request;
1082 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1083 * backed by this journal. This will eventually be replaced by an array
1084 * of uuids, allowing us to index multiple devices within a single
1085 * journal and to perform atomic updates across them.
1090 * @j_task: Pointer to the current commit thread for this journal.
1092 struct task_struct *j_task;
1095 * @j_max_transaction_buffers:
1097 * Maximum number of metadata buffers to allow in a single compound
1098 * commit transaction.
1100 int j_max_transaction_buffers;
1103 * @j_revoke_records_per_block:
1105 * Number of revoke records that fit in one descriptor block.
1107 int j_revoke_records_per_block;
1110 * @j_commit_interval:
1112 * What is the maximum transaction lifetime before we begin a commit?
1114 unsigned long j_commit_interval;
1117 * @j_commit_timer: The timer used to wakeup the commit thread.
1119 struct timer_list j_commit_timer;
1122 * @j_revoke_lock: Protect the revoke table.
1124 spinlock_t j_revoke_lock;
1129 * The revoke table - maintains the list of revoked blocks in the
1130 * current transaction.
1132 struct jbd2_revoke_table_s *j_revoke;
1135 * @j_revoke_table: Alternate revoke tables for j_revoke.
1137 struct jbd2_revoke_table_s *j_revoke_table[2];
1140 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1142 struct buffer_head **j_wbuf;
1145 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only
1146 * during a fast commit. Currently only process can do fast commit, so
1147 * this field is not protected by any lock.
1149 struct buffer_head **j_fc_wbuf;
1154 * Size of @j_wbuf array.
1161 * Size of @j_fc_wbuf array.
1166 * @j_last_sync_writer:
1168 * The pid of the last person to run a synchronous operation
1169 * through the journal.
1171 pid_t j_last_sync_writer;
1174 * @j_average_commit_time:
1176 * The average amount of time in nanoseconds it takes to commit a
1177 * transaction to disk. [j_state_lock]
1179 u64 j_average_commit_time;
1182 * @j_min_batch_time:
1184 * Minimum time that we should wait for additional filesystem operations
1185 * to get batched into a synchronous handle in microseconds.
1187 u32 j_min_batch_time;
1190 * @j_max_batch_time:
1192 * Maximum time that we should wait for additional filesystem operations
1193 * to get batched into a synchronous handle in microseconds.
1195 u32 j_max_batch_time;
1198 * @j_commit_callback:
1200 * This function is called when a transaction is closed.
1202 void (*j_commit_callback)(journal_t *,
1206 * @j_submit_inode_data_buffers:
1208 * This function is called for all inodes associated with the
1209 * committing transaction marked with JI_WRITE_DATA flag
1210 * before we start to write out the transaction to the journal.
1212 int (*j_submit_inode_data_buffers)
1213 (struct jbd2_inode *);
1216 * @j_finish_inode_data_buffers:
1218 * This function is called for all inodes associated with the
1219 * committing transaction marked with JI_WAIT_DATA flag
1220 * after we have written the transaction to the journal
1221 * but before we write out the commit block.
1223 int (*j_finish_inode_data_buffers)
1224 (struct jbd2_inode *);
1227 * Journal statistics
1231 * @j_history_lock: Protect the transactions statistics history.
1233 spinlock_t j_history_lock;
1236 * @j_proc_entry: procfs entry for the jbd statistics directory.
1238 struct proc_dir_entry *j_proc_entry;
1241 * @j_stats: Overall statistics.
1243 struct transaction_stats_s j_stats;
1246 * @j_failed_commit: Failed journal commit ID.
1248 unsigned int j_failed_commit;
1253 * An opaque pointer to fs-private information. ext3 puts its
1254 * superblock pointer here.
1261 * Reference to checksum algorithm driver via cryptoapi.
1263 struct crypto_shash *j_chksum_driver;
1268 * Precomputed journal UUID checksum for seeding other checksums.
1272 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1274 * @j_trans_commit_map:
1276 * Lockdep entity to track transaction commit dependencies. Handles
1277 * hold this "lock" for read, when we wait for commit, we acquire the
1278 * "lock" for writing. This matches the properties of jbd2 journalling
1279 * where the running transaction has to wait for all handles to be
1280 * dropped to commit that transaction and also acquiring a handle may
1281 * require transaction commit to finish.
1283 struct lockdep_map j_trans_commit_map;
1287 * @j_fc_cleanup_callback:
1289 * Clean-up after fast commit or full commit. JBD2 calls this function
1290 * after every commit operation.
1292 void (*j_fc_cleanup_callback)(struct journal_s *journal, int full, tid_t tid);
1295 * @j_fc_replay_callback:
1297 * File-system specific function that performs replay of a fast
1298 * commit. JBD2 calls this function for each fast commit block found in
1299 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1300 * to indicate that the block was processed correctly and more fast
1301 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1302 * indicates the end of replay (no more blocks remaining). A negative
1303 * return value indicates error.
1305 int (*j_fc_replay_callback)(struct journal_s *journal,
1306 struct buffer_head *bh,
1307 enum passtype pass, int off,
1308 tid_t expected_commit_id);
1311 #define jbd2_might_wait_for_commit(j) \
1313 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1314 rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1317 /* journal feature predicate functions */
1318 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1319 static inline bool jbd2_has_feature_##name(journal_t *j) \
1321 return ((j)->j_format_version >= 2 && \
1322 ((j)->j_superblock->s_feature_compat & \
1323 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1325 static inline void jbd2_set_feature_##name(journal_t *j) \
1327 (j)->j_superblock->s_feature_compat |= \
1328 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1330 static inline void jbd2_clear_feature_##name(journal_t *j) \
1332 (j)->j_superblock->s_feature_compat &= \
1333 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1336 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1337 static inline bool jbd2_has_feature_##name(journal_t *j) \
1339 return ((j)->j_format_version >= 2 && \
1340 ((j)->j_superblock->s_feature_ro_compat & \
1341 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1343 static inline void jbd2_set_feature_##name(journal_t *j) \
1345 (j)->j_superblock->s_feature_ro_compat |= \
1346 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1348 static inline void jbd2_clear_feature_##name(journal_t *j) \
1350 (j)->j_superblock->s_feature_ro_compat &= \
1351 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1354 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1355 static inline bool jbd2_has_feature_##name(journal_t *j) \
1357 return ((j)->j_format_version >= 2 && \
1358 ((j)->j_superblock->s_feature_incompat & \
1359 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1361 static inline void jbd2_set_feature_##name(journal_t *j) \
1363 (j)->j_superblock->s_feature_incompat |= \
1364 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1366 static inline void jbd2_clear_feature_##name(journal_t *j) \
1368 (j)->j_superblock->s_feature_incompat &= \
1369 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1372 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM)
1374 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE)
1375 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
1376 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT)
1377 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2)
1378 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3)
1379 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit, FAST_COMMIT)
1382 * Journal flag definitions
1384 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1385 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1386 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1387 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1388 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1389 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
1390 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1391 * data write error in ordered
1393 #define JBD2_FAST_COMMIT_ONGOING 0x100 /* Fast commit is ongoing */
1394 #define JBD2_FULL_COMMIT_ONGOING 0x200 /* Full commit is ongoing */
1395 #define JBD2_JOURNAL_FLUSH_DISCARD 0x0001
1396 #define JBD2_JOURNAL_FLUSH_ZEROOUT 0x0002
1397 #define JBD2_JOURNAL_FLUSH_VALID (JBD2_JOURNAL_FLUSH_DISCARD | \
1398 JBD2_JOURNAL_FLUSH_ZEROOUT)
1401 * Journal atomic flag definitions
1403 #define JBD2_CHECKPOINT_IO_ERROR 0x001 /* Detect io error while writing
1404 * buffer back to disk */
1407 * Function declarations for the journaling transaction and buffer
1411 /* Filing buffers */
1412 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1413 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1414 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1415 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1416 extern void __journal_free_buffer(struct journal_head *bh);
1417 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1418 extern void __journal_clean_data_list(transaction_t *transaction);
1419 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1421 list_add_tail(&bh->b_assoc_buffers, head);
1423 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1425 list_del_init(&bh->b_assoc_buffers);
1428 /* Log buffer allocation */
1429 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1430 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1431 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1432 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1433 unsigned long *block);
1434 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1435 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1437 /* Commit management */
1438 extern void jbd2_journal_commit_transaction(journal_t *);
1440 /* Checkpoint list management */
1441 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1442 unsigned long jbd2_journal_shrink_checkpoint_list(journal_t *journal, unsigned long *nr_to_scan);
1443 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1444 int jbd2_journal_try_remove_checkpoint(struct journal_head *jh);
1445 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1446 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1453 struct jbd2_buffer_trigger_type {
1455 * Fired a the moment data to write to the journal are known to be
1456 * stable - so either at the moment b_frozen_data is created or just
1457 * before a buffer is written to the journal. mapped_data is a mapped
1458 * buffer that is the frozen data for commit.
1460 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1461 struct buffer_head *bh, void *mapped_data,
1465 * Fired during journal abort for dirty buffers that will not be
1468 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1469 struct buffer_head *bh);
1472 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1474 struct jbd2_buffer_trigger_type *triggers);
1475 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1476 struct jbd2_buffer_trigger_type *triggers);
1479 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1480 struct journal_head *jh_in,
1481 struct buffer_head **bh_out,
1484 /* Transaction locking */
1485 extern void __wait_on_journal (journal_t *);
1487 /* Transaction cache support */
1488 extern void jbd2_journal_destroy_transaction_cache(void);
1489 extern int __init jbd2_journal_init_transaction_cache(void);
1490 extern void jbd2_journal_free_transaction(transaction_t *);
1495 * We need to lock the journal during transaction state changes so that nobody
1496 * ever tries to take a handle on the running transaction while we are in the
1497 * middle of moving it to the commit phase. j_state_lock does this.
1499 * Note that the locking is completely interrupt unsafe. We never touch
1500 * journal structures from interrupts.
1503 static inline handle_t *journal_current_handle(void)
1505 return current->journal_info;
1508 /* The journaling code user interface:
1510 * Create and destroy handles
1511 * Register buffer modifications against the current transaction.
1514 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1515 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1516 int revoke_records, gfp_t gfp_mask,
1517 unsigned int type, unsigned int line_no);
1518 extern int jbd2_journal_restart(handle_t *, int nblocks);
1519 extern int jbd2__journal_restart(handle_t *, int nblocks,
1520 int revoke_records, gfp_t gfp_mask);
1521 extern int jbd2_journal_start_reserved(handle_t *handle,
1522 unsigned int type, unsigned int line_no);
1523 extern void jbd2_journal_free_reserved(handle_t *handle);
1524 extern int jbd2_journal_extend(handle_t *handle, int nblocks,
1525 int revoke_records);
1526 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1527 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1528 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1529 void jbd2_journal_set_triggers(struct buffer_head *,
1530 struct jbd2_buffer_trigger_type *type);
1531 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1532 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1533 extern int jbd2_journal_invalidatepage(journal_t *,
1534 struct page *, unsigned int, unsigned int);
1535 extern int jbd2_journal_try_to_free_buffers(journal_t *journal, struct page *page);
1536 extern int jbd2_journal_stop(handle_t *);
1537 extern int jbd2_journal_flush(journal_t *journal, unsigned int flags);
1538 extern void jbd2_journal_lock_updates (journal_t *);
1539 extern void jbd2_journal_unlock_updates (journal_t *);
1541 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1542 struct block_device *fs_dev,
1543 unsigned long long start, int len, int bsize);
1544 extern journal_t * jbd2_journal_init_inode (struct inode *);
1545 extern int jbd2_journal_update_format (journal_t *);
1546 extern int jbd2_journal_check_used_features
1547 (journal_t *, unsigned long, unsigned long, unsigned long);
1548 extern int jbd2_journal_check_available_features
1549 (journal_t *, unsigned long, unsigned long, unsigned long);
1550 extern int jbd2_journal_set_features
1551 (journal_t *, unsigned long, unsigned long, unsigned long);
1552 extern void jbd2_journal_clear_features
1553 (journal_t *, unsigned long, unsigned long, unsigned long);
1554 extern int jbd2_journal_load (journal_t *journal);
1555 extern int jbd2_journal_destroy (journal_t *);
1556 extern int jbd2_journal_recover (journal_t *journal);
1557 extern int jbd2_journal_wipe (journal_t *, int);
1558 extern int jbd2_journal_skip_recovery (journal_t *);
1559 extern void jbd2_journal_update_sb_errno(journal_t *);
1560 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
1561 unsigned long, int);
1562 extern void jbd2_journal_abort (journal_t *, int);
1563 extern int jbd2_journal_errno (journal_t *);
1564 extern void jbd2_journal_ack_err (journal_t *);
1565 extern int jbd2_journal_clear_err (journal_t *);
1566 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1567 extern int jbd2_journal_force_commit(journal_t *);
1568 extern int jbd2_journal_force_commit_nested(journal_t *);
1569 extern int jbd2_journal_inode_ranged_write(handle_t *handle,
1570 struct jbd2_inode *inode, loff_t start_byte,
1572 extern int jbd2_journal_inode_ranged_wait(handle_t *handle,
1573 struct jbd2_inode *inode, loff_t start_byte,
1575 extern int jbd2_journal_submit_inode_data_buffers(
1576 struct jbd2_inode *jinode);
1577 extern int jbd2_journal_finish_inode_data_buffers(
1578 struct jbd2_inode *jinode);
1579 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1580 struct jbd2_inode *inode, loff_t new_size);
1581 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1582 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1585 * journal_head management
1587 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1588 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1589 void jbd2_journal_put_journal_head(struct journal_head *jh);
1594 extern struct kmem_cache *jbd2_handle_cache;
1596 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1598 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1601 static inline void jbd2_free_handle(handle_t *handle)
1603 kmem_cache_free(jbd2_handle_cache, handle);
1607 * jbd2_inode management (optional, for those file systems that want to use
1608 * dynamically allocated jbd2_inode structures)
1610 extern struct kmem_cache *jbd2_inode_cache;
1612 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1614 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1617 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1619 kmem_cache_free(jbd2_inode_cache, jinode);
1622 /* Primary revoke support */
1623 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1624 extern int jbd2_journal_init_revoke(journal_t *, int);
1625 extern void jbd2_journal_destroy_revoke_record_cache(void);
1626 extern void jbd2_journal_destroy_revoke_table_cache(void);
1627 extern int __init jbd2_journal_init_revoke_record_cache(void);
1628 extern int __init jbd2_journal_init_revoke_table_cache(void);
1630 extern void jbd2_journal_destroy_revoke(journal_t *);
1631 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1632 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1633 extern void jbd2_journal_write_revoke_records(transaction_t *transaction,
1634 struct list_head *log_bufs);
1636 /* Recovery revoke support */
1637 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1638 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1639 extern void jbd2_journal_clear_revoke(journal_t *);
1640 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1641 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
1644 * The log thread user interface:
1646 * Request space in the current transaction, and force transaction commit
1647 * transitions on demand.
1650 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1651 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1652 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1653 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1654 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1655 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1656 int jbd2_log_do_checkpoint(journal_t *journal);
1657 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1659 void __jbd2_log_wait_for_space(journal_t *journal);
1660 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1661 extern int jbd2_cleanup_journal_tail(journal_t *);
1663 /* Fast commit related APIs */
1664 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1665 int jbd2_fc_end_commit(journal_t *journal);
1666 int jbd2_fc_end_commit_fallback(journal_t *journal);
1667 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1668 int jbd2_submit_inode_data(struct jbd2_inode *jinode);
1669 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1670 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1671 int jbd2_fc_release_bufs(journal_t *journal);
1673 static inline int jbd2_journal_get_max_txn_bufs(journal_t *journal)
1675 return (journal->j_total_len - journal->j_fc_wbufsize) / 4;
1681 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1682 * bit, when set, indicates that we have had a fatal error somewhere,
1683 * either inside the journaling layer or indicated to us by the client
1684 * (eg. ext3), and that we and should not commit any further
1688 static inline int is_journal_aborted(journal_t *journal)
1690 return journal->j_flags & JBD2_ABORT;
1693 static inline int is_handle_aborted(handle_t *handle)
1695 if (handle->h_aborted || !handle->h_transaction)
1697 return is_journal_aborted(handle->h_transaction->t_journal);
1700 static inline void jbd2_journal_abort_handle(handle_t *handle)
1702 handle->h_aborted = 1;
1705 #endif /* __KERNEL__ */
1707 /* Comparison functions for transaction IDs: perform comparisons using
1708 * modulo arithmetic so that they work over sequence number wraps. */
1710 static inline int tid_gt(tid_t x, tid_t y)
1712 int difference = (x - y);
1713 return (difference > 0);
1716 static inline int tid_geq(tid_t x, tid_t y)
1718 int difference = (x - y);
1719 return (difference >= 0);
1722 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1723 extern size_t journal_tag_bytes(journal_t *journal);
1725 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1727 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1730 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1732 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1733 journal->j_chksum_driver == NULL);
1735 return journal->j_chksum_driver != NULL;
1738 static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb)
1740 int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks);
1742 return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS;
1746 * Return number of free blocks in the log. Must be called under j_state_lock.
1748 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1750 /* Allow for rounding errors */
1751 long free = journal->j_free - 32;
1753 if (journal->j_committing_transaction) {
1754 free -= atomic_read(&journal->
1755 j_committing_transaction->t_outstanding_credits);
1757 return max_t(long, free, 0);
1761 * Definitions which augment the buffer_head layer
1764 /* journaling buffer types */
1765 #define BJ_None 0 /* Not journaled */
1766 #define BJ_Metadata 1 /* Normal journaled metadata */
1767 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1768 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
1769 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1772 extern int jbd_blocks_per_page(struct inode *inode);
1774 /* JBD uses a CRC32 checksum */
1775 #define JBD_MAX_CHECKSUM_SIZE 4
1777 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1778 const void *address, unsigned int length)
1781 struct shash_desc shash;
1782 char ctx[JBD_MAX_CHECKSUM_SIZE];
1786 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1787 JBD_MAX_CHECKSUM_SIZE);
1789 desc.shash.tfm = journal->j_chksum_driver;
1790 *(u32 *)desc.ctx = crc;
1792 err = crypto_shash_update(&desc.shash, address, length);
1795 return *(u32 *)desc.ctx;
1798 /* Return most recent uncommitted transaction */
1799 static inline tid_t jbd2_get_latest_transaction(journal_t *journal)
1803 read_lock(&journal->j_state_lock);
1804 tid = journal->j_commit_request;
1805 if (journal->j_running_transaction)
1806 tid = journal->j_running_transaction->t_tid;
1807 read_unlock(&journal->j_state_lock);
1811 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1815 if (!handle->h_reserved)
1816 journal = handle->h_transaction->t_journal;
1818 journal = handle->h_journal;
1820 return handle->h_total_credits -
1821 DIV_ROUND_UP(handle->h_revoke_credits_requested,
1822 journal->j_revoke_records_per_block);
1827 #define buffer_trace_init(bh) do {} while (0)
1828 #define print_buffer_fields(bh) do {} while (0)
1829 #define print_buffer_trace(bh) do {} while (0)
1830 #define BUFFER_TRACE(bh, info) do {} while (0)
1831 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1832 #define JBUFFER_TRACE(jh, info) do {} while (0)
1834 #endif /* __KERNEL__ */
1836 #define EFSBADCRC EBADMSG /* Bad CRC detected */
1837 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
1839 #endif /* _LINUX_JBD2_H */