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/mutex.h>
30 #include <linux/timer.h>
31 #include <linux/slab.h>
32 #include <crypto/hash.h>
35 #define journal_oom_retry 1
38 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
39 * certain classes of error which can occur due to failed IOs. Under
40 * normal use we want ext4 to continue after such errors, because
41 * hardware _can_ fail, but for debugging purposes when running tests on
42 * known-good hardware we may want to trap these errors.
44 #undef JBD2_PARANOID_IOFAIL
47 * The default maximum commit age, in seconds.
49 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
51 #ifdef CONFIG_JBD2_DEBUG
53 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
54 * consistency checks. By default we don't do this unless
55 * CONFIG_JBD2_DEBUG is on.
57 #define JBD2_EXPENSIVE_CHECKING
58 extern ushort jbd2_journal_enable_debug
;
59 void __jbd2_debug(int level
, const char *file
, const char *func
,
60 unsigned int line
, const char *fmt
, ...);
62 #define jbd_debug(n, fmt, a...) \
63 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
65 #define jbd_debug(n, fmt, a...) /**/
68 extern void *jbd2_alloc(size_t size
, gfp_t flags
);
69 extern void jbd2_free(void *ptr
, size_t size
);
71 #define JBD2_MIN_JOURNAL_BLOCKS 1024
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 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 /* Tail of revoke block, for checksumming */
218 struct jbd2_journal_revoke_tail
{
219 __be32 r_checksum
; /* crc32c(uuid+revoke_block) */
222 /* Definitions for the journal tag flags word: */
223 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
224 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
225 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
226 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
230 * The journal superblock. All fields are in big-endian byte order.
232 typedef struct journal_superblock_s
235 journal_header_t s_header
;
238 /* Static information describing the journal */
239 __be32 s_blocksize
; /* journal device blocksize */
240 __be32 s_maxlen
; /* total blocks in journal file */
241 __be32 s_first
; /* first block of log information */
244 /* Dynamic information describing the current state of the log */
245 __be32 s_sequence
; /* first commit ID expected in log */
246 __be32 s_start
; /* blocknr of start of log */
249 /* Error value, as set by jbd2_journal_abort(). */
253 /* Remaining fields are only valid in a version-2 superblock */
254 __be32 s_feature_compat
; /* compatible feature set */
255 __be32 s_feature_incompat
; /* incompatible feature set */
256 __be32 s_feature_ro_compat
; /* readonly-compatible feature set */
258 __u8 s_uuid
[16]; /* 128-bit uuid for journal */
261 __be32 s_nr_users
; /* Nr of filesystems sharing log */
263 __be32 s_dynsuper
; /* Blocknr of dynamic superblock copy*/
266 __be32 s_max_transaction
; /* Limit of journal blocks per trans.*/
267 __be32 s_max_trans_data
; /* Limit of data blocks per trans. */
270 __u8 s_checksum_type
; /* checksum type */
273 __be32 s_checksum
; /* crc32c(superblock) */
276 __u8 s_users
[16*48]; /* ids of all fs'es sharing the log */
278 } journal_superblock_t
;
280 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
281 ((j)->j_format_version >= 2 && \
282 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
283 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
284 ((j)->j_format_version >= 2 && \
285 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
286 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
287 ((j)->j_format_version >= 2 && \
288 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
290 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
292 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
293 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
294 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
295 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
296 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010
298 /* Features known to this kernel version: */
299 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
300 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
301 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
302 JBD2_FEATURE_INCOMPAT_64BIT | \
303 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
304 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
305 JBD2_FEATURE_INCOMPAT_CSUM_V3)
309 #include <linux/fs.h>
310 #include <linux/sched.h>
312 enum jbd_state_bits
{
313 BH_JBD
/* Has an attached ext3 journal_head */
315 BH_JWrite
, /* Being written to log (@@@ DEBUGGING) */
316 BH_Freed
, /* Has been freed (truncated) */
317 BH_Revoked
, /* Has been revoked from the log */
318 BH_RevokeValid
, /* Revoked flag is valid */
319 BH_JBDDirty
, /* Is dirty but journaled */
320 BH_State
, /* Pins most journal_head state */
321 BH_JournalHead
, /* Pins bh->b_private and jh->b_bh */
322 BH_Shadow
, /* IO on shadow buffer is running */
323 BH_Verified
, /* Metadata block has been verified ok */
324 BH_JBDPrivateStart
, /* First bit available for private use by FS */
328 BUFFER_FNS(JWrite
, jwrite
)
329 BUFFER_FNS(JBDDirty
, jbddirty
)
330 TAS_BUFFER_FNS(JBDDirty
, jbddirty
)
331 BUFFER_FNS(Revoked
, revoked
)
332 TAS_BUFFER_FNS(Revoked
, revoked
)
333 BUFFER_FNS(RevokeValid
, revokevalid
)
334 TAS_BUFFER_FNS(RevokeValid
, revokevalid
)
335 BUFFER_FNS(Freed
, freed
)
336 BUFFER_FNS(Shadow
, shadow
)
337 BUFFER_FNS(Verified
, verified
)
339 #include <linux/jbd_common.h>
341 #define J_ASSERT(assert) BUG_ON(!(assert))
343 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
344 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
346 #if defined(JBD2_PARANOID_IOFAIL)
347 #define J_EXPECT(expr, why...) J_ASSERT(expr)
348 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
349 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
351 #define __journal_expect(expr, why...) \
356 "JBD2 unexpected failure: %s: %s;\n", \
358 printk(KERN_ERR why "\n"); \
362 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
363 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
364 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
367 /* Flags in jbd_inode->i_flags */
368 #define __JI_COMMIT_RUNNING 0
369 /* Commit of the inode data in progress. We use this flag to protect us from
370 * concurrent deletion of inode. We cannot use reference to inode for this
371 * since we cannot afford doing last iput() on behalf of kjournald
373 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
376 * struct jbd_inode is the structure linking inodes in ordered mode
377 * present in a transaction so that we can sync them during commit.
380 /* Which transaction does this inode belong to? Either the running
381 * transaction or the committing one. [j_list_lock] */
382 transaction_t
*i_transaction
;
384 /* Pointer to the running transaction modifying inode's data in case
385 * there is already a committing transaction touching it. [j_list_lock] */
386 transaction_t
*i_next_transaction
;
388 /* List of inodes in the i_transaction [j_list_lock] */
389 struct list_head i_list
;
391 /* VFS inode this inode belongs to [constant during the lifetime
392 * of the structure] */
393 struct inode
*i_vfs_inode
;
395 /* Flags of inode [j_list_lock] */
396 unsigned long i_flags
;
399 struct jbd2_revoke_table_s
;
402 * struct handle_s - The handle_s type is the concrete type associated with
404 * @h_transaction: Which compound transaction is this update a part of?
405 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
406 * @h_ref: Reference count on this handle
407 * @h_err: Field for caller's use to track errors through large fs operations
408 * @h_sync: flag for sync-on-close
409 * @h_jdata: flag to force data journaling
410 * @h_aborted: flag indicating fatal error on handle
413 /* Docbook can't yet cope with the bit fields, but will leave the documentation
414 * in so it can be fixed later.
417 struct jbd2_journal_handle
420 /* Which compound transaction is this update a part of? */
421 transaction_t
*h_transaction
;
422 /* Which journal handle belongs to - used iff h_reserved set */
423 journal_t
*h_journal
;
426 /* Handle reserved for finishing the logical operation */
427 handle_t
*h_rsv_handle
;
429 /* Number of remaining buffers we are allowed to dirty: */
430 int h_buffer_credits
;
432 /* Reference count on this handle */
435 /* Field for caller's use to track errors through large fs */
439 /* Flags [no locking] */
440 unsigned int h_sync
: 1; /* sync-on-close */
441 unsigned int h_jdata
: 1; /* force data journaling */
442 unsigned int h_reserved
: 1; /* handle with reserved credits */
443 unsigned int h_aborted
: 1; /* fatal error on handle */
444 unsigned int h_type
: 8; /* for handle statistics */
445 unsigned int h_line_no
: 16; /* for handle statistics */
447 unsigned long h_start_jiffies
;
448 unsigned int h_requested_credits
;
450 #ifdef CONFIG_DEBUG_LOCK_ALLOC
451 struct lockdep_map h_lockdep_map
;
457 * Some stats for checkpoint phase
459 struct transaction_chp_stats_s
{
460 unsigned long cs_chp_time
;
461 __u32 cs_forced_to_close
;
466 /* The transaction_t type is the guts of the journaling mechanism. It
467 * tracks a compound transaction through its various states:
469 * RUNNING: accepting new updates
470 * LOCKED: Updates still running but we don't accept new ones
471 * RUNDOWN: Updates are tidying up but have finished requesting
472 * new buffers to modify (state not used for now)
473 * FLUSH: All updates complete, but we are still writing to disk
474 * COMMIT: All data on disk, writing commit record
475 * FINISHED: We still have to keep the transaction for checkpointing.
477 * The transaction keeps track of all of the buffers modified by a
478 * running transaction, and all of the buffers committed but not yet
479 * flushed to home for finished transactions.
486 * ->jbd_lock_bh_journal_head() (This is "innermost")
489 * ->jbd_lock_bh_state()
491 * jbd_lock_bh_state()
498 * ->j_list_lock (journal_unmap_buffer)
504 /* Pointer to the journal for this transaction. [no locking] */
505 journal_t
*t_journal
;
507 /* Sequence number for this transaction [no locking] */
511 * Transaction's current state
512 * [no locking - only kjournald2 alters this]
513 * [j_list_lock] guards transition of a transaction into T_FINISHED
514 * state and subsequent call of __jbd2_journal_drop_transaction()
515 * FIXME: needs barriers
516 * KLUDGE: [use j_state_lock]
530 * Where in the log does this transaction's commit start? [no locking]
532 unsigned long t_log_start
;
534 /* Number of buffers on the t_buffers list [j_list_lock] */
538 * Doubly-linked circular list of all buffers reserved but not yet
539 * modified by this transaction [j_list_lock]
541 struct journal_head
*t_reserved_list
;
544 * Doubly-linked circular list of all metadata buffers owned by this
545 * transaction [j_list_lock]
547 struct journal_head
*t_buffers
;
550 * Doubly-linked circular list of all forget buffers (superseded
551 * buffers which we can un-checkpoint once this transaction commits)
554 struct journal_head
*t_forget
;
557 * Doubly-linked circular list of all buffers still to be flushed before
558 * this transaction can be checkpointed. [j_list_lock]
560 struct journal_head
*t_checkpoint_list
;
563 * Doubly-linked circular list of all buffers submitted for IO while
564 * checkpointing. [j_list_lock]
566 struct journal_head
*t_checkpoint_io_list
;
569 * Doubly-linked circular list of metadata buffers being shadowed by log
570 * IO. The IO buffers on the iobuf list and the shadow buffers on this
571 * list match each other one for one at all times. [j_list_lock]
573 struct journal_head
*t_shadow_list
;
576 * List of inodes whose data we've modified in data=ordered mode.
579 struct list_head t_inode_list
;
582 * Protects info related to handles
584 spinlock_t t_handle_lock
;
587 * Longest time some handle had to wait for running transaction
589 unsigned long t_max_wait
;
592 * When transaction started
594 unsigned long t_start
;
597 * When commit was requested
599 unsigned long t_requested
;
602 * Checkpointing stats [j_checkpoint_sem]
604 struct transaction_chp_stats_s t_chp_stats
;
607 * Number of outstanding updates running on this transaction
613 * Number of buffers reserved for use by all handles in this transaction
614 * handle but not yet modified. [t_handle_lock]
616 atomic_t t_outstanding_credits
;
619 * Forward and backward links for the circular list of all transactions
620 * awaiting checkpoint. [j_list_lock]
622 transaction_t
*t_cpnext
, *t_cpprev
;
625 * When will the transaction expire (become due for commit), in jiffies?
628 unsigned long t_expires
;
631 * When this transaction started, in nanoseconds [no locking]
633 ktime_t t_start_time
;
636 * How many handles used this transaction? [t_handle_lock]
638 atomic_t t_handle_count
;
641 * This transaction is being forced and some process is
642 * waiting for it to finish.
644 unsigned int t_synchronous_commit
:1;
646 /* Disk flush needs to be sent to fs partition [no locking] */
647 int t_need_data_flush
;
650 * For use by the filesystem to store fs-specific data
651 * structures associated with the transaction
653 struct list_head t_private_list
;
656 struct transaction_run_stats_s
{
657 unsigned long rs_wait
;
658 unsigned long rs_request_delay
;
659 unsigned long rs_running
;
660 unsigned long rs_locked
;
661 unsigned long rs_flushing
;
662 unsigned long rs_logging
;
664 __u32 rs_handle_count
;
666 __u32 rs_blocks_logged
;
669 struct transaction_stats_s
{
670 unsigned long ts_tid
;
671 unsigned long ts_requested
;
672 struct transaction_run_stats_s run
;
675 static inline unsigned long
676 jbd2_time_diff(unsigned long start
, unsigned long end
)
681 return end
+ (MAX_JIFFY_OFFSET
- start
);
684 #define JBD2_NR_BATCH 64
687 * struct journal_s - The journal_s type is the concrete type associated with
689 * @j_flags: General journaling state flags
690 * @j_errno: Is there an outstanding uncleared error on the journal (from a
692 * @j_sb_buffer: First part of superblock buffer
693 * @j_superblock: Second part of superblock buffer
694 * @j_format_version: Version of the superblock format
695 * @j_state_lock: Protect the various scalars in the journal
696 * @j_barrier_count: Number of processes waiting to create a barrier lock
697 * @j_barrier: The barrier lock itself
698 * @j_running_transaction: The current running transaction..
699 * @j_committing_transaction: the transaction we are pushing to disk
700 * @j_checkpoint_transactions: a linked circular list of all transactions
701 * waiting for checkpointing
702 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
703 * to start committing, or for a barrier lock to be released
704 * @j_wait_done_commit: Wait queue for waiting for commit to complete
705 * @j_wait_commit: Wait queue to trigger commit
706 * @j_wait_updates: Wait queue to wait for updates to complete
707 * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop
708 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
709 * @j_head: Journal head - identifies the first unused block in the journal
710 * @j_tail: Journal tail - identifies the oldest still-used block in the
712 * @j_free: Journal free - how many free blocks are there in the journal?
713 * @j_first: The block number of the first usable block
714 * @j_last: The block number one beyond the last usable block
715 * @j_dev: Device where we store the journal
716 * @j_blocksize: blocksize for the location where we store the journal.
717 * @j_blk_offset: starting block offset for into the device where we store the
719 * @j_fs_dev: Device which holds the client fs. For internal journal this will
721 * @j_reserved_credits: Number of buffers reserved from the running transaction
722 * @j_maxlen: Total maximum capacity of the journal region on disk.
723 * @j_list_lock: Protects the buffer lists and internal buffer state.
724 * @j_inode: Optional inode where we store the journal. If present, all journal
725 * block numbers are mapped into this inode via bmap().
726 * @j_tail_sequence: Sequence number of the oldest transaction in the log
727 * @j_transaction_sequence: Sequence number of the next transaction to grant
728 * @j_commit_sequence: Sequence number of the most recently committed
730 * @j_commit_request: Sequence number of the most recent transaction wanting
732 * @j_uuid: Uuid of client object.
733 * @j_task: Pointer to the current commit thread for this journal
734 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
735 * single compound commit transaction
736 * @j_commit_interval: What is the maximum transaction lifetime before we begin
738 * @j_commit_timer: The timer used to wakeup the commit thread
739 * @j_revoke_lock: Protect the revoke table
740 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
741 * current transaction.
742 * @j_revoke_table: alternate revoke tables for j_revoke
743 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
744 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
745 * number that will fit in j_blocksize
746 * @j_last_sync_writer: most recent pid which did a synchronous write
747 * @j_history: Buffer storing the transactions statistics history
748 * @j_history_max: Maximum number of transactions in the statistics history
749 * @j_history_cur: Current number of transactions in the statistics history
750 * @j_history_lock: Protect the transactions statistics history
751 * @j_proc_entry: procfs entry for the jbd statistics directory
752 * @j_stats: Overall statistics
753 * @j_private: An opaque pointer to fs-private information.
758 /* General journaling state flags [j_state_lock] */
759 unsigned long j_flags
;
762 * Is there an outstanding uncleared error on the journal (from a prior
763 * abort)? [j_state_lock]
767 /* The superblock buffer */
768 struct buffer_head
*j_sb_buffer
;
769 journal_superblock_t
*j_superblock
;
771 /* Version of the superblock format */
772 int j_format_version
;
775 * Protect the various scalars in the journal
777 rwlock_t j_state_lock
;
780 * Number of processes waiting to create a barrier lock [j_state_lock]
784 /* The barrier lock itself */
785 struct mutex j_barrier
;
788 * Transactions: The current running transaction...
789 * [j_state_lock] [caller holding open handle]
791 transaction_t
*j_running_transaction
;
794 * the transaction we are pushing to disk
795 * [j_state_lock] [caller holding open handle]
797 transaction_t
*j_committing_transaction
;
800 * ... and a linked circular list of all transactions waiting for
801 * checkpointing. [j_list_lock]
803 transaction_t
*j_checkpoint_transactions
;
806 * Wait queue for waiting for a locked transaction to start committing,
807 * or for a barrier lock to be released
809 wait_queue_head_t j_wait_transaction_locked
;
811 /* Wait queue for waiting for commit to complete */
812 wait_queue_head_t j_wait_done_commit
;
814 /* Wait queue to trigger commit */
815 wait_queue_head_t j_wait_commit
;
817 /* Wait queue to wait for updates to complete */
818 wait_queue_head_t j_wait_updates
;
820 /* Wait queue to wait for reserved buffer credits to drop */
821 wait_queue_head_t j_wait_reserved
;
823 /* Semaphore for locking against concurrent checkpoints */
824 struct mutex j_checkpoint_mutex
;
827 * List of buffer heads used by the checkpoint routine. This
828 * was moved from jbd2_log_do_checkpoint() to reduce stack
829 * usage. Access to this array is controlled by the
830 * j_checkpoint_mutex. [j_checkpoint_mutex]
832 struct buffer_head
*j_chkpt_bhs
[JBD2_NR_BATCH
];
835 * Journal head: identifies the first unused block in the journal.
838 unsigned long j_head
;
841 * Journal tail: identifies the oldest still-used block in the journal.
844 unsigned long j_tail
;
847 * Journal free: how many free blocks are there in the journal?
850 unsigned long j_free
;
853 * Journal start and end: the block numbers of the first usable block
854 * and one beyond the last usable block in the journal. [j_state_lock]
856 unsigned long j_first
;
857 unsigned long j_last
;
860 * Device, blocksize and starting block offset for the location where we
863 struct block_device
*j_dev
;
865 unsigned long long j_blk_offset
;
866 char j_devname
[BDEVNAME_SIZE
+24];
869 * Device which holds the client fs. For internal journal this will be
872 struct block_device
*j_fs_dev
;
874 /* Total maximum capacity of the journal region on disk. */
875 unsigned int j_maxlen
;
877 /* Number of buffers reserved from the running transaction */
878 atomic_t j_reserved_credits
;
881 * Protects the buffer lists and internal buffer state.
883 spinlock_t j_list_lock
;
885 /* Optional inode where we store the journal. If present, all */
886 /* journal block numbers are mapped into this inode via */
888 struct inode
*j_inode
;
891 * Sequence number of the oldest transaction in the log [j_state_lock]
893 tid_t j_tail_sequence
;
896 * Sequence number of the next transaction to grant [j_state_lock]
898 tid_t j_transaction_sequence
;
901 * Sequence number of the most recently committed transaction
904 tid_t j_commit_sequence
;
907 * Sequence number of the most recent transaction wanting commit
910 tid_t j_commit_request
;
913 * Journal uuid: identifies the object (filesystem, LVM volume etc)
914 * backed by this journal. This will eventually be replaced by an array
915 * of uuids, allowing us to index multiple devices within a single
916 * journal and to perform atomic updates across them.
920 /* Pointer to the current commit thread for this journal */
921 struct task_struct
*j_task
;
924 * Maximum number of metadata buffers to allow in a single compound
927 int j_max_transaction_buffers
;
930 * What is the maximum transaction lifetime before we begin a commit?
932 unsigned long j_commit_interval
;
934 /* The timer used to wakeup the commit thread: */
935 struct timer_list j_commit_timer
;
938 * The revoke table: maintains the list of revoked blocks in the
939 * current transaction. [j_revoke_lock]
941 spinlock_t j_revoke_lock
;
942 struct jbd2_revoke_table_s
*j_revoke
;
943 struct jbd2_revoke_table_s
*j_revoke_table
[2];
946 * array of bhs for jbd2_journal_commit_transaction
948 struct buffer_head
**j_wbuf
;
952 * this is the pid of hte last person to run a synchronous operation
953 * through the journal
955 pid_t j_last_sync_writer
;
958 * the average amount of time in nanoseconds it takes to commit a
959 * transaction to disk. [j_state_lock]
961 u64 j_average_commit_time
;
964 * minimum and maximum times that we should wait for
965 * additional filesystem operations to get batched into a
966 * synchronous handle in microseconds
968 u32 j_min_batch_time
;
969 u32 j_max_batch_time
;
971 /* This function is called when a transaction is closed */
972 void (*j_commit_callback
)(journal_t
*,
978 spinlock_t j_history_lock
;
979 struct proc_dir_entry
*j_proc_entry
;
980 struct transaction_stats_s j_stats
;
982 /* Failed journal commit ID */
983 unsigned int j_failed_commit
;
986 * An opaque pointer to fs-private information. ext3 puts its
987 * superblock pointer here
991 /* Reference to checksum algorithm driver via cryptoapi */
992 struct crypto_shash
*j_chksum_driver
;
994 /* Precomputed journal UUID checksum for seeding other checksums */
999 * Journal flag definitions
1001 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1002 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1003 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1004 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1005 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1006 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
1007 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1008 * data write error in ordered
1012 * Function declarations for the journaling transaction and buffer
1016 /* Filing buffers */
1017 extern void jbd2_journal_unfile_buffer(journal_t
*, struct journal_head
*);
1018 extern void __jbd2_journal_refile_buffer(struct journal_head
*);
1019 extern void jbd2_journal_refile_buffer(journal_t
*, struct journal_head
*);
1020 extern void __jbd2_journal_file_buffer(struct journal_head
*, transaction_t
*, int);
1021 extern void __journal_free_buffer(struct journal_head
*bh
);
1022 extern void jbd2_journal_file_buffer(struct journal_head
*, transaction_t
*, int);
1023 extern void __journal_clean_data_list(transaction_t
*transaction
);
1024 static inline void jbd2_file_log_bh(struct list_head
*head
, struct buffer_head
*bh
)
1026 list_add_tail(&bh
->b_assoc_buffers
, head
);
1028 static inline void jbd2_unfile_log_bh(struct buffer_head
*bh
)
1030 list_del_init(&bh
->b_assoc_buffers
);
1033 /* Log buffer allocation */
1034 struct buffer_head
*jbd2_journal_get_descriptor_buffer(journal_t
*journal
);
1035 int jbd2_journal_next_log_block(journal_t
*, unsigned long long *);
1036 int jbd2_journal_get_log_tail(journal_t
*journal
, tid_t
*tid
,
1037 unsigned long *block
);
1038 void __jbd2_update_log_tail(journal_t
*journal
, tid_t tid
, unsigned long block
);
1039 void jbd2_update_log_tail(journal_t
*journal
, tid_t tid
, unsigned long block
);
1041 /* Commit management */
1042 extern void jbd2_journal_commit_transaction(journal_t
*);
1044 /* Checkpoint list management */
1045 int __jbd2_journal_clean_checkpoint_list(journal_t
*journal
);
1046 int __jbd2_journal_remove_checkpoint(struct journal_head
*);
1047 void __jbd2_journal_insert_checkpoint(struct journal_head
*, transaction_t
*);
1054 struct jbd2_buffer_trigger_type
{
1056 * Fired a the moment data to write to the journal are known to be
1057 * stable - so either at the moment b_frozen_data is created or just
1058 * before a buffer is written to the journal. mapped_data is a mapped
1059 * buffer that is the frozen data for commit.
1061 void (*t_frozen
)(struct jbd2_buffer_trigger_type
*type
,
1062 struct buffer_head
*bh
, void *mapped_data
,
1066 * Fired during journal abort for dirty buffers that will not be
1069 void (*t_abort
)(struct jbd2_buffer_trigger_type
*type
,
1070 struct buffer_head
*bh
);
1073 extern void jbd2_buffer_frozen_trigger(struct journal_head
*jh
,
1075 struct jbd2_buffer_trigger_type
*triggers
);
1076 extern void jbd2_buffer_abort_trigger(struct journal_head
*jh
,
1077 struct jbd2_buffer_trigger_type
*triggers
);
1080 extern int jbd2_journal_write_metadata_buffer(transaction_t
*transaction
,
1081 struct journal_head
*jh_in
,
1082 struct buffer_head
**bh_out
,
1085 /* Transaction locking */
1086 extern void __wait_on_journal (journal_t
*);
1088 /* Transaction cache support */
1089 extern void jbd2_journal_destroy_transaction_cache(void);
1090 extern int jbd2_journal_init_transaction_cache(void);
1091 extern void jbd2_journal_free_transaction(transaction_t
*);
1096 * We need to lock the journal during transaction state changes so that nobody
1097 * ever tries to take a handle on the running transaction while we are in the
1098 * middle of moving it to the commit phase. j_state_lock does this.
1100 * Note that the locking is completely interrupt unsafe. We never touch
1101 * journal structures from interrupts.
1104 static inline handle_t
*journal_current_handle(void)
1106 return current
->journal_info
;
1109 /* The journaling code user interface:
1111 * Create and destroy handles
1112 * Register buffer modifications against the current transaction.
1115 extern handle_t
*jbd2_journal_start(journal_t
*, int nblocks
);
1116 extern handle_t
*jbd2__journal_start(journal_t
*, int blocks
, int rsv_blocks
,
1117 gfp_t gfp_mask
, unsigned int type
,
1118 unsigned int line_no
);
1119 extern int jbd2_journal_restart(handle_t
*, int nblocks
);
1120 extern int jbd2__journal_restart(handle_t
*, int nblocks
, gfp_t gfp_mask
);
1121 extern int jbd2_journal_start_reserved(handle_t
*handle
,
1122 unsigned int type
, unsigned int line_no
);
1123 extern void jbd2_journal_free_reserved(handle_t
*handle
);
1124 extern int jbd2_journal_extend (handle_t
*, int nblocks
);
1125 extern int jbd2_journal_get_write_access(handle_t
*, struct buffer_head
*);
1126 extern int jbd2_journal_get_create_access (handle_t
*, struct buffer_head
*);
1127 extern int jbd2_journal_get_undo_access(handle_t
*, struct buffer_head
*);
1128 void jbd2_journal_set_triggers(struct buffer_head
*,
1129 struct jbd2_buffer_trigger_type
*type
);
1130 extern int jbd2_journal_dirty_metadata (handle_t
*, struct buffer_head
*);
1131 extern int jbd2_journal_forget (handle_t
*, struct buffer_head
*);
1132 extern void journal_sync_buffer (struct buffer_head
*);
1133 extern int jbd2_journal_invalidatepage(journal_t
*,
1134 struct page
*, unsigned int, unsigned int);
1135 extern int jbd2_journal_try_to_free_buffers(journal_t
*, struct page
*, gfp_t
);
1136 extern int jbd2_journal_stop(handle_t
*);
1137 extern int jbd2_journal_flush (journal_t
*);
1138 extern void jbd2_journal_lock_updates (journal_t
*);
1139 extern void jbd2_journal_unlock_updates (journal_t
*);
1141 extern journal_t
* jbd2_journal_init_dev(struct block_device
*bdev
,
1142 struct block_device
*fs_dev
,
1143 unsigned long long start
, int len
, int bsize
);
1144 extern journal_t
* jbd2_journal_init_inode (struct inode
*);
1145 extern int jbd2_journal_update_format (journal_t
*);
1146 extern int jbd2_journal_check_used_features
1147 (journal_t
*, unsigned long, unsigned long, unsigned long);
1148 extern int jbd2_journal_check_available_features
1149 (journal_t
*, unsigned long, unsigned long, unsigned long);
1150 extern int jbd2_journal_set_features
1151 (journal_t
*, unsigned long, unsigned long, unsigned long);
1152 extern void jbd2_journal_clear_features
1153 (journal_t
*, unsigned long, unsigned long, unsigned long);
1154 extern int jbd2_journal_load (journal_t
*journal
);
1155 extern int jbd2_journal_destroy (journal_t
*);
1156 extern int jbd2_journal_recover (journal_t
*journal
);
1157 extern int jbd2_journal_wipe (journal_t
*, int);
1158 extern int jbd2_journal_skip_recovery (journal_t
*);
1159 extern void jbd2_journal_update_sb_errno(journal_t
*);
1160 extern void jbd2_journal_update_sb_log_tail (journal_t
*, tid_t
,
1161 unsigned long, int);
1162 extern void __jbd2_journal_abort_hard (journal_t
*);
1163 extern void jbd2_journal_abort (journal_t
*, int);
1164 extern int jbd2_journal_errno (journal_t
*);
1165 extern void jbd2_journal_ack_err (journal_t
*);
1166 extern int jbd2_journal_clear_err (journal_t
*);
1167 extern int jbd2_journal_bmap(journal_t
*, unsigned long, unsigned long long *);
1168 extern int jbd2_journal_force_commit(journal_t
*);
1169 extern int jbd2_journal_force_commit_nested(journal_t
*);
1170 extern int jbd2_journal_file_inode(handle_t
*handle
, struct jbd2_inode
*inode
);
1171 extern int jbd2_journal_begin_ordered_truncate(journal_t
*journal
,
1172 struct jbd2_inode
*inode
, loff_t new_size
);
1173 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode
*jinode
, struct inode
*inode
);
1174 extern void jbd2_journal_release_jbd_inode(journal_t
*journal
, struct jbd2_inode
*jinode
);
1177 * journal_head management
1179 struct journal_head
*jbd2_journal_add_journal_head(struct buffer_head
*bh
);
1180 struct journal_head
*jbd2_journal_grab_journal_head(struct buffer_head
*bh
);
1181 void jbd2_journal_put_journal_head(struct journal_head
*jh
);
1186 extern struct kmem_cache
*jbd2_handle_cache
;
1188 static inline handle_t
*jbd2_alloc_handle(gfp_t gfp_flags
)
1190 return kmem_cache_zalloc(jbd2_handle_cache
, gfp_flags
);
1193 static inline void jbd2_free_handle(handle_t
*handle
)
1195 kmem_cache_free(jbd2_handle_cache
, handle
);
1199 * jbd2_inode management (optional, for those file systems that want to use
1200 * dynamically allocated jbd2_inode structures)
1202 extern struct kmem_cache
*jbd2_inode_cache
;
1204 static inline struct jbd2_inode
*jbd2_alloc_inode(gfp_t gfp_flags
)
1206 return kmem_cache_alloc(jbd2_inode_cache
, gfp_flags
);
1209 static inline void jbd2_free_inode(struct jbd2_inode
*jinode
)
1211 kmem_cache_free(jbd2_inode_cache
, jinode
);
1214 /* Primary revoke support */
1215 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1216 extern int jbd2_journal_init_revoke(journal_t
*, int);
1217 extern void jbd2_journal_destroy_revoke_caches(void);
1218 extern int jbd2_journal_init_revoke_caches(void);
1220 extern void jbd2_journal_destroy_revoke(journal_t
*);
1221 extern int jbd2_journal_revoke (handle_t
*, unsigned long long, struct buffer_head
*);
1222 extern int jbd2_journal_cancel_revoke(handle_t
*, struct journal_head
*);
1223 extern void jbd2_journal_write_revoke_records(journal_t
*journal
,
1224 transaction_t
*transaction
,
1225 struct list_head
*log_bufs
,
1228 /* Recovery revoke support */
1229 extern int jbd2_journal_set_revoke(journal_t
*, unsigned long long, tid_t
);
1230 extern int jbd2_journal_test_revoke(journal_t
*, unsigned long long, tid_t
);
1231 extern void jbd2_journal_clear_revoke(journal_t
*);
1232 extern void jbd2_journal_switch_revoke_table(journal_t
*journal
);
1233 extern void jbd2_clear_buffer_revoked_flags(journal_t
*journal
);
1236 * The log thread user interface:
1238 * Request space in the current transaction, and force transaction commit
1239 * transitions on demand.
1242 int jbd2_log_start_commit(journal_t
*journal
, tid_t tid
);
1243 int __jbd2_log_start_commit(journal_t
*journal
, tid_t tid
);
1244 int jbd2_journal_start_commit(journal_t
*journal
, tid_t
*tid
);
1245 int jbd2_log_wait_commit(journal_t
*journal
, tid_t tid
);
1246 int jbd2_complete_transaction(journal_t
*journal
, tid_t tid
);
1247 int jbd2_log_do_checkpoint(journal_t
*journal
);
1248 int jbd2_trans_will_send_data_barrier(journal_t
*journal
, tid_t tid
);
1250 void __jbd2_log_wait_for_space(journal_t
*journal
);
1251 extern void __jbd2_journal_drop_transaction(journal_t
*, transaction_t
*);
1252 extern int jbd2_cleanup_journal_tail(journal_t
*);
1254 /* Debugging code only: */
1256 #define jbd_ENOSYS() \
1258 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1259 current->state = TASK_UNINTERRUPTIBLE; \
1266 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1267 * bit, when set, indicates that we have had a fatal error somewhere,
1268 * either inside the journaling layer or indicated to us by the client
1269 * (eg. ext3), and that we and should not commit any further
1273 static inline int is_journal_aborted(journal_t
*journal
)
1275 return journal
->j_flags
& JBD2_ABORT
;
1278 static inline int is_handle_aborted(handle_t
*handle
)
1280 if (handle
->h_aborted
|| !handle
->h_transaction
)
1282 return is_journal_aborted(handle
->h_transaction
->t_journal
);
1285 static inline void jbd2_journal_abort_handle(handle_t
*handle
)
1287 handle
->h_aborted
= 1;
1290 #endif /* __KERNEL__ */
1292 /* Comparison functions for transaction IDs: perform comparisons using
1293 * modulo arithmetic so that they work over sequence number wraps. */
1295 static inline int tid_gt(tid_t x
, tid_t y
)
1297 int difference
= (x
- y
);
1298 return (difference
> 0);
1301 static inline int tid_geq(tid_t x
, tid_t y
)
1303 int difference
= (x
- y
);
1304 return (difference
>= 0);
1307 extern int jbd2_journal_blocks_per_page(struct inode
*inode
);
1308 extern size_t journal_tag_bytes(journal_t
*journal
);
1310 static inline int jbd2_journal_has_csum_v2or3(journal_t
*journal
)
1312 if (JBD2_HAS_INCOMPAT_FEATURE(journal
, JBD2_FEATURE_INCOMPAT_CSUM_V2
) ||
1313 JBD2_HAS_INCOMPAT_FEATURE(journal
, JBD2_FEATURE_INCOMPAT_CSUM_V3
))
1320 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
1321 * transaction control blocks.
1323 #define JBD2_CONTROL_BLOCKS_SHIFT 5
1326 * Return the minimum number of blocks which must be free in the journal
1327 * before a new transaction may be started. Must be called under j_state_lock.
1329 static inline int jbd2_space_needed(journal_t
*journal
)
1331 int nblocks
= journal
->j_max_transaction_buffers
;
1332 return nblocks
+ (nblocks
>> JBD2_CONTROL_BLOCKS_SHIFT
);
1336 * Return number of free blocks in the log. Must be called under j_state_lock.
1338 static inline unsigned long jbd2_log_space_left(journal_t
*journal
)
1340 /* Allow for rounding errors */
1341 unsigned long free
= journal
->j_free
- 32;
1343 if (journal
->j_committing_transaction
) {
1344 unsigned long committing
= atomic_read(&journal
->
1345 j_committing_transaction
->t_outstanding_credits
);
1347 /* Transaction + control blocks */
1348 free
-= committing
+ (committing
>> JBD2_CONTROL_BLOCKS_SHIFT
);
1354 * Definitions which augment the buffer_head layer
1357 /* journaling buffer types */
1358 #define BJ_None 0 /* Not journaled */
1359 #define BJ_Metadata 1 /* Normal journaled metadata */
1360 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1361 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
1362 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1365 extern int jbd_blocks_per_page(struct inode
*inode
);
1367 /* JBD uses a CRC32 checksum */
1368 #define JBD_MAX_CHECKSUM_SIZE 4
1370 static inline u32
jbd2_chksum(journal_t
*journal
, u32 crc
,
1371 const void *address
, unsigned int length
)
1374 struct shash_desc shash
;
1375 char ctx
[JBD_MAX_CHECKSUM_SIZE
];
1379 BUG_ON(crypto_shash_descsize(journal
->j_chksum_driver
) >
1380 JBD_MAX_CHECKSUM_SIZE
);
1382 desc
.shash
.tfm
= journal
->j_chksum_driver
;
1383 desc
.shash
.flags
= 0;
1384 *(u32
*)desc
.ctx
= crc
;
1386 err
= crypto_shash_update(&desc
.shash
, address
, length
);
1389 return *(u32
*)desc
.ctx
;
1392 /* Return most recent uncommitted transaction */
1393 static inline tid_t
jbd2_get_latest_transaction(journal_t
*journal
)
1397 read_lock(&journal
->j_state_lock
);
1398 tid
= journal
->j_commit_request
;
1399 if (journal
->j_running_transaction
)
1400 tid
= journal
->j_running_transaction
->t_tid
;
1401 read_unlock(&journal
->j_state_lock
);
1407 #define buffer_trace_init(bh) do {} while (0)
1408 #define print_buffer_fields(bh) do {} while (0)
1409 #define print_buffer_trace(bh) do {} while (0)
1410 #define BUFFER_TRACE(bh, info) do {} while (0)
1411 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1412 #define JBUFFER_TRACE(jh, info) do {} while (0)
1414 #endif /* __KERNEL__ */
1416 #endif /* _LINUX_JBD2_H */