2 * linux/include/linux/jbd2.h
4 * Written by Stephen C. Tweedie <sct@redhat.com>
6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
8 * This file is part of the Linux kernel and is made available under
9 * the terms of the GNU General Public License, version 2, or at your
10 * option, any later version, incorporated herein by reference.
12 * Definitions for transaction data structures for the buffer cache
13 * filesystem journaling support.
19 /* Allow this file to be included directly into e2fsprogs */
21 #include "jfs_compat.h"
25 #include <linux/types.h>
26 #include <linux/buffer_head.h>
27 #include <linux/journal-head.h>
28 #include <linux/stddef.h>
29 #include <linux/bit_spinlock.h>
30 #include <linux/mutex.h>
31 #include <linux/timer.h>
32 #include <linux/slab.h>
33 #include <crypto/hash.h>
36 #define journal_oom_retry 1
39 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
40 * certain classes of error which can occur due to failed IOs. Under
41 * normal use we want ext4 to continue after such errors, because
42 * hardware _can_ fail, but for debugging purposes when running tests on
43 * known-good hardware we may want to trap these errors.
45 #undef JBD2_PARANOID_IOFAIL
48 * The default maximum commit age, in seconds.
50 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
52 #ifdef CONFIG_JBD2_DEBUG
54 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
55 * consistency checks. By default we don't do this unless
56 * CONFIG_JBD2_DEBUG is on.
58 #define JBD2_EXPENSIVE_CHECKING
59 extern ushort jbd2_journal_enable_debug
;
61 #define jbd_debug(n, f, a...) \
63 if ((n) <= jbd2_journal_enable_debug) { \
64 printk (KERN_DEBUG "(%s, %d): %s: ", \
65 __FILE__, __LINE__, __func__); \
70 #define jbd_debug(f, a...) /**/
73 extern void *jbd2_alloc(size_t size
, gfp_t flags
);
74 extern void jbd2_free(void *ptr
, size_t size
);
76 #define JBD2_MIN_JOURNAL_BLOCKS 1024
81 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
83 * All filesystem modifications made by the process go
84 * through this handle. Recursive operations (such as quota operations)
85 * are gathered into a single update.
87 * The buffer credits field is used to account for journaled buffers
88 * being modified by the running process. To ensure that there is
89 * enough log space for all outstanding operations, we need to limit the
90 * number of outstanding buffers possible at any time. When the
91 * operation completes, any buffer credits not used are credited back to
92 * the transaction, so that at all times we know how many buffers the
93 * outstanding updates on a transaction might possibly touch.
95 * This is an opaque datatype.
97 typedef struct jbd2_journal_handle handle_t
; /* Atomic operation type */
101 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
103 * journal_t is linked to from the fs superblock structure.
105 * We use the journal_t to keep track of all outstanding transaction
106 * activity on the filesystem, and to manage the state of the log
109 * This is an opaque datatype.
111 typedef struct journal_s journal_t
; /* Journal control structure */
115 * Internal structures used by the logging mechanism:
118 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
125 * Descriptor block types:
128 #define JBD2_DESCRIPTOR_BLOCK 1
129 #define JBD2_COMMIT_BLOCK 2
130 #define JBD2_SUPERBLOCK_V1 3
131 #define JBD2_SUPERBLOCK_V2 4
132 #define JBD2_REVOKE_BLOCK 5
135 * Standard header for all descriptor blocks:
137 typedef struct journal_header_s
147 #define JBD2_CRC32_CHKSUM 1
148 #define JBD2_MD5_CHKSUM 2
149 #define JBD2_SHA1_CHKSUM 3
150 #define JBD2_CRC32C_CHKSUM 4
152 #define JBD2_CRC32_CHKSUM_SIZE 4
154 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
156 * Commit block header for storing transactional checksums:
158 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
159 * fields are used to store a checksum of the descriptor and data blocks.
161 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
162 * field is used to store crc32c(uuid+commit_block). Each journal metadata
163 * block gets its own checksum, and data block checksums are stored in
164 * journal_block_tag (in the descriptor). The other h_chksum* fields are
167 * Checksum v1 and v2 are mutually exclusive features.
169 struct commit_header
{
173 unsigned char h_chksum_type
;
174 unsigned char h_chksum_size
;
175 unsigned char h_padding
[2];
176 __be32 h_chksum
[JBD2_CHECKSUM_BYTES
];
178 __be32 h_commit_nsec
;
182 * The block tag: used to describe a single buffer in the journal.
183 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
184 * raw struct shouldn't be used for pointer math or sizeof() - use
185 * journal_tag_bytes(journal) instead to compute this.
187 typedef struct journal_block_tag_s
189 __be32 t_blocknr
; /* The on-disk block number */
190 __be16 t_checksum
; /* truncated crc32c(uuid+seq+block) */
191 __be16 t_flags
; /* See below */
192 __be32 t_blocknr_high
; /* most-significant high 32bits. */
193 } journal_block_tag_t
;
195 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
196 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
198 /* Tail of descriptor block, for checksumming */
199 struct jbd2_journal_block_tail
{
200 __be32 t_checksum
; /* crc32c(uuid+descr_block) */
204 * The revoke descriptor: used on disk to describe a series of blocks to
205 * be revoked from the log
207 typedef struct jbd2_journal_revoke_header_s
209 journal_header_t r_header
;
210 __be32 r_count
; /* Count of bytes used in the block */
211 } jbd2_journal_revoke_header_t
;
213 /* Tail of revoke block, for checksumming */
214 struct jbd2_journal_revoke_tail
{
215 __be32 r_checksum
; /* crc32c(uuid+revoke_block) */
218 /* Definitions for the journal tag flags word: */
219 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
220 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
221 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
222 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
226 * The journal superblock. All fields are in big-endian byte order.
228 typedef struct journal_superblock_s
231 journal_header_t s_header
;
234 /* Static information describing the journal */
235 __be32 s_blocksize
; /* journal device blocksize */
236 __be32 s_maxlen
; /* total blocks in journal file */
237 __be32 s_first
; /* first block of log information */
240 /* Dynamic information describing the current state of the log */
241 __be32 s_sequence
; /* first commit ID expected in log */
242 __be32 s_start
; /* blocknr of start of log */
245 /* Error value, as set by jbd2_journal_abort(). */
249 /* Remaining fields are only valid in a version-2 superblock */
250 __be32 s_feature_compat
; /* compatible feature set */
251 __be32 s_feature_incompat
; /* incompatible feature set */
252 __be32 s_feature_ro_compat
; /* readonly-compatible feature set */
254 __u8 s_uuid
[16]; /* 128-bit uuid for journal */
257 __be32 s_nr_users
; /* Nr of filesystems sharing log */
259 __be32 s_dynsuper
; /* Blocknr of dynamic superblock copy*/
262 __be32 s_max_transaction
; /* Limit of journal blocks per trans.*/
263 __be32 s_max_trans_data
; /* Limit of data blocks per trans. */
266 __u8 s_checksum_type
; /* checksum type */
269 __be32 s_checksum
; /* crc32c(superblock) */
272 __u8 s_users
[16*48]; /* ids of all fs'es sharing the log */
274 } journal_superblock_t
;
276 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
277 ((j)->j_format_version >= 2 && \
278 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
279 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
280 ((j)->j_format_version >= 2 && \
281 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
282 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
283 ((j)->j_format_version >= 2 && \
284 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
286 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
288 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
289 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
290 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
291 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
293 /* Features known to this kernel version: */
294 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
295 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
296 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
297 JBD2_FEATURE_INCOMPAT_64BIT | \
298 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
299 JBD2_FEATURE_INCOMPAT_CSUM_V2)
303 #include <linux/fs.h>
304 #include <linux/sched.h>
306 enum jbd_state_bits
{
307 BH_JBD
/* Has an attached ext3 journal_head */
309 BH_JWrite
, /* Being written to log (@@@ DEBUGGING) */
310 BH_Freed
, /* Has been freed (truncated) */
311 BH_Revoked
, /* Has been revoked from the log */
312 BH_RevokeValid
, /* Revoked flag is valid */
313 BH_JBDDirty
, /* Is dirty but journaled */
314 BH_State
, /* Pins most journal_head state */
315 BH_JournalHead
, /* Pins bh->b_private and jh->b_bh */
316 BH_Shadow
, /* IO on shadow buffer is running */
317 BH_Verified
, /* Metadata block has been verified ok */
318 BH_JBDPrivateStart
, /* First bit available for private use by FS */
322 BUFFER_FNS(JWrite
, jwrite
)
323 BUFFER_FNS(JBDDirty
, jbddirty
)
324 TAS_BUFFER_FNS(JBDDirty
, jbddirty
)
325 BUFFER_FNS(Revoked
, revoked
)
326 TAS_BUFFER_FNS(Revoked
, revoked
)
327 BUFFER_FNS(RevokeValid
, revokevalid
)
328 TAS_BUFFER_FNS(RevokeValid
, revokevalid
)
329 BUFFER_FNS(Freed
, freed
)
330 BUFFER_FNS(Shadow
, shadow
)
331 BUFFER_FNS(Verified
, verified
)
333 #include <linux/jbd_common.h>
335 #define J_ASSERT(assert) BUG_ON(!(assert))
337 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
338 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
340 #if defined(JBD2_PARANOID_IOFAIL)
341 #define J_EXPECT(expr, why...) J_ASSERT(expr)
342 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
343 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
345 #define __journal_expect(expr, why...) \
350 "JBD2 unexpected failure: %s: %s;\n", \
352 printk(KERN_ERR why "\n"); \
356 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
357 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
358 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
361 /* Flags in jbd_inode->i_flags */
362 #define __JI_COMMIT_RUNNING 0
363 /* Commit of the inode data in progress. We use this flag to protect us from
364 * concurrent deletion of inode. We cannot use reference to inode for this
365 * since we cannot afford doing last iput() on behalf of kjournald
367 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
370 * struct jbd_inode is the structure linking inodes in ordered mode
371 * present in a transaction so that we can sync them during commit.
374 /* Which transaction does this inode belong to? Either the running
375 * transaction or the committing one. [j_list_lock] */
376 transaction_t
*i_transaction
;
378 /* Pointer to the running transaction modifying inode's data in case
379 * there is already a committing transaction touching it. [j_list_lock] */
380 transaction_t
*i_next_transaction
;
382 /* List of inodes in the i_transaction [j_list_lock] */
383 struct list_head i_list
;
385 /* VFS inode this inode belongs to [constant during the lifetime
386 * of the structure] */
387 struct inode
*i_vfs_inode
;
389 /* Flags of inode [j_list_lock] */
390 unsigned long i_flags
;
393 struct jbd2_revoke_table_s
;
396 * struct handle_s - The handle_s type is the concrete type associated with
398 * @h_transaction: Which compound transaction is this update a part of?
399 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
400 * @h_ref: Reference count on this handle
401 * @h_err: Field for caller's use to track errors through large fs operations
402 * @h_sync: flag for sync-on-close
403 * @h_jdata: flag to force data journaling
404 * @h_aborted: flag indicating fatal error on handle
407 /* Docbook can't yet cope with the bit fields, but will leave the documentation
408 * in so it can be fixed later.
411 struct jbd2_journal_handle
414 /* Which compound transaction is this update a part of? */
415 transaction_t
*h_transaction
;
416 /* Which journal handle belongs to - used iff h_reserved set */
417 journal_t
*h_journal
;
420 /* Handle reserved for finishing the logical operation */
421 handle_t
*h_rsv_handle
;
423 /* Number of remaining buffers we are allowed to dirty: */
424 int h_buffer_credits
;
426 /* Reference count on this handle */
429 /* Field for caller's use to track errors through large fs */
433 /* Flags [no locking] */
434 unsigned int h_sync
: 1; /* sync-on-close */
435 unsigned int h_jdata
: 1; /* force data journaling */
436 unsigned int h_reserved
: 1; /* handle with reserved credits */
437 unsigned int h_aborted
: 1; /* fatal error on handle */
438 unsigned int h_type
: 8; /* for handle statistics */
439 unsigned int h_line_no
: 16; /* for handle statistics */
441 unsigned long h_start_jiffies
;
442 unsigned int h_requested_credits
;
444 #ifdef CONFIG_DEBUG_LOCK_ALLOC
445 struct lockdep_map h_lockdep_map
;
451 * Some stats for checkpoint phase
453 struct transaction_chp_stats_s
{
454 unsigned long cs_chp_time
;
455 __u32 cs_forced_to_close
;
460 /* The transaction_t type is the guts of the journaling mechanism. It
461 * tracks a compound transaction through its various states:
463 * RUNNING: accepting new updates
464 * LOCKED: Updates still running but we don't accept new ones
465 * RUNDOWN: Updates are tidying up but have finished requesting
466 * new buffers to modify (state not used for now)
467 * FLUSH: All updates complete, but we are still writing to disk
468 * COMMIT: All data on disk, writing commit record
469 * FINISHED: We still have to keep the transaction for checkpointing.
471 * The transaction keeps track of all of the buffers modified by a
472 * running transaction, and all of the buffers committed but not yet
473 * flushed to home for finished transactions.
480 * ->jbd_lock_bh_journal_head() (This is "innermost")
483 * ->jbd_lock_bh_state()
485 * jbd_lock_bh_state()
492 * ->j_list_lock (journal_unmap_buffer)
498 /* Pointer to the journal for this transaction. [no locking] */
499 journal_t
*t_journal
;
501 /* Sequence number for this transaction [no locking] */
505 * Transaction's current state
506 * [no locking - only kjournald2 alters this]
507 * [j_list_lock] guards transition of a transaction into T_FINISHED
508 * state and subsequent call of __jbd2_journal_drop_transaction()
509 * FIXME: needs barriers
510 * KLUDGE: [use j_state_lock]
524 * Where in the log does this transaction's commit start? [no locking]
526 unsigned long t_log_start
;
528 /* Number of buffers on the t_buffers list [j_list_lock] */
532 * Doubly-linked circular list of all buffers reserved but not yet
533 * modified by this transaction [j_list_lock]
535 struct journal_head
*t_reserved_list
;
538 * Doubly-linked circular list of all metadata buffers owned by this
539 * transaction [j_list_lock]
541 struct journal_head
*t_buffers
;
544 * Doubly-linked circular list of all forget buffers (superseded
545 * buffers which we can un-checkpoint once this transaction commits)
548 struct journal_head
*t_forget
;
551 * Doubly-linked circular list of all buffers still to be flushed before
552 * this transaction can be checkpointed. [j_list_lock]
554 struct journal_head
*t_checkpoint_list
;
557 * Doubly-linked circular list of all buffers submitted for IO while
558 * checkpointing. [j_list_lock]
560 struct journal_head
*t_checkpoint_io_list
;
563 * Doubly-linked circular list of metadata buffers being shadowed by log
564 * IO. The IO buffers on the iobuf list and the shadow buffers on this
565 * list match each other one for one at all times. [j_list_lock]
567 struct journal_head
*t_shadow_list
;
570 * List of inodes whose data we've modified in data=ordered mode.
573 struct list_head t_inode_list
;
576 * Protects info related to handles
578 spinlock_t t_handle_lock
;
581 * Longest time some handle had to wait for running transaction
583 unsigned long t_max_wait
;
586 * When transaction started
588 unsigned long t_start
;
591 * When commit was requested
593 unsigned long t_requested
;
596 * Checkpointing stats [j_checkpoint_sem]
598 struct transaction_chp_stats_s t_chp_stats
;
601 * Number of outstanding updates running on this transaction
607 * Number of buffers reserved for use by all handles in this transaction
608 * handle but not yet modified. [t_handle_lock]
610 atomic_t t_outstanding_credits
;
613 * Forward and backward links for the circular list of all transactions
614 * awaiting checkpoint. [j_list_lock]
616 transaction_t
*t_cpnext
, *t_cpprev
;
619 * When will the transaction expire (become due for commit), in jiffies?
622 unsigned long t_expires
;
625 * When this transaction started, in nanoseconds [no locking]
627 ktime_t t_start_time
;
630 * How many handles used this transaction? [t_handle_lock]
632 atomic_t t_handle_count
;
635 * This transaction is being forced and some process is
636 * waiting for it to finish.
638 unsigned int t_synchronous_commit
:1;
640 /* Disk flush needs to be sent to fs partition [no locking] */
641 int t_need_data_flush
;
644 * For use by the filesystem to store fs-specific data
645 * structures associated with the transaction
647 struct list_head t_private_list
;
650 struct transaction_run_stats_s
{
651 unsigned long rs_wait
;
652 unsigned long rs_request_delay
;
653 unsigned long rs_running
;
654 unsigned long rs_locked
;
655 unsigned long rs_flushing
;
656 unsigned long rs_logging
;
658 __u32 rs_handle_count
;
660 __u32 rs_blocks_logged
;
663 struct transaction_stats_s
{
664 unsigned long ts_tid
;
665 unsigned long ts_requested
;
666 struct transaction_run_stats_s run
;
669 static inline unsigned long
670 jbd2_time_diff(unsigned long start
, unsigned long end
)
675 return end
+ (MAX_JIFFY_OFFSET
- start
);
678 #define JBD2_NR_BATCH 64
681 * struct journal_s - The journal_s type is the concrete type associated with
683 * @j_flags: General journaling state flags
684 * @j_errno: Is there an outstanding uncleared error on the journal (from a
686 * @j_sb_buffer: First part of superblock buffer
687 * @j_superblock: Second part of superblock buffer
688 * @j_format_version: Version of the superblock format
689 * @j_state_lock: Protect the various scalars in the journal
690 * @j_barrier_count: Number of processes waiting to create a barrier lock
691 * @j_barrier: The barrier lock itself
692 * @j_running_transaction: The current running transaction..
693 * @j_committing_transaction: the transaction we are pushing to disk
694 * @j_checkpoint_transactions: a linked circular list of all transactions
695 * waiting for checkpointing
696 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
697 * to start committing, or for a barrier lock to be released
698 * @j_wait_done_commit: Wait queue for waiting for commit to complete
699 * @j_wait_commit: Wait queue to trigger commit
700 * @j_wait_updates: Wait queue to wait for updates to complete
701 * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop
702 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
703 * @j_head: Journal head - identifies the first unused block in the journal
704 * @j_tail: Journal tail - identifies the oldest still-used block in the
706 * @j_free: Journal free - how many free blocks are there in the journal?
707 * @j_first: The block number of the first usable block
708 * @j_last: The block number one beyond the last usable block
709 * @j_dev: Device where we store the journal
710 * @j_blocksize: blocksize for the location where we store the journal.
711 * @j_blk_offset: starting block offset for into the device where we store the
713 * @j_fs_dev: Device which holds the client fs. For internal journal this will
715 * @j_reserved_credits: Number of buffers reserved from the running transaction
716 * @j_maxlen: Total maximum capacity of the journal region on disk.
717 * @j_list_lock: Protects the buffer lists and internal buffer state.
718 * @j_inode: Optional inode where we store the journal. If present, all journal
719 * block numbers are mapped into this inode via bmap().
720 * @j_tail_sequence: Sequence number of the oldest transaction in the log
721 * @j_transaction_sequence: Sequence number of the next transaction to grant
722 * @j_commit_sequence: Sequence number of the most recently committed
724 * @j_commit_request: Sequence number of the most recent transaction wanting
726 * @j_uuid: Uuid of client object.
727 * @j_task: Pointer to the current commit thread for this journal
728 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
729 * single compound commit transaction
730 * @j_commit_interval: What is the maximum transaction lifetime before we begin
732 * @j_commit_timer: The timer used to wakeup the commit thread
733 * @j_revoke_lock: Protect the revoke table
734 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
735 * current transaction.
736 * @j_revoke_table: alternate revoke tables for j_revoke
737 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
738 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
739 * number that will fit in j_blocksize
740 * @j_last_sync_writer: most recent pid which did a synchronous write
741 * @j_history: Buffer storing the transactions statistics history
742 * @j_history_max: Maximum number of transactions in the statistics history
743 * @j_history_cur: Current number of transactions in the statistics history
744 * @j_history_lock: Protect the transactions statistics history
745 * @j_proc_entry: procfs entry for the jbd statistics directory
746 * @j_stats: Overall statistics
747 * @j_private: An opaque pointer to fs-private information.
752 /* General journaling state flags [j_state_lock] */
753 unsigned long j_flags
;
756 * Is there an outstanding uncleared error on the journal (from a prior
757 * abort)? [j_state_lock]
761 /* The superblock buffer */
762 struct buffer_head
*j_sb_buffer
;
763 journal_superblock_t
*j_superblock
;
765 /* Version of the superblock format */
766 int j_format_version
;
769 * Protect the various scalars in the journal
771 rwlock_t j_state_lock
;
774 * Number of processes waiting to create a barrier lock [j_state_lock]
778 /* The barrier lock itself */
779 struct mutex j_barrier
;
782 * Transactions: The current running transaction...
783 * [j_state_lock] [caller holding open handle]
785 transaction_t
*j_running_transaction
;
788 * the transaction we are pushing to disk
789 * [j_state_lock] [caller holding open handle]
791 transaction_t
*j_committing_transaction
;
794 * ... and a linked circular list of all transactions waiting for
795 * checkpointing. [j_list_lock]
797 transaction_t
*j_checkpoint_transactions
;
800 * Wait queue for waiting for a locked transaction to start committing,
801 * or for a barrier lock to be released
803 wait_queue_head_t j_wait_transaction_locked
;
805 /* Wait queue for waiting for commit to complete */
806 wait_queue_head_t j_wait_done_commit
;
808 /* Wait queue to trigger commit */
809 wait_queue_head_t j_wait_commit
;
811 /* Wait queue to wait for updates to complete */
812 wait_queue_head_t j_wait_updates
;
814 /* Wait queue to wait for reserved buffer credits to drop */
815 wait_queue_head_t j_wait_reserved
;
817 /* Semaphore for locking against concurrent checkpoints */
818 struct mutex j_checkpoint_mutex
;
821 * List of buffer heads used by the checkpoint routine. This
822 * was moved from jbd2_log_do_checkpoint() to reduce stack
823 * usage. Access to this array is controlled by the
824 * j_checkpoint_mutex. [j_checkpoint_mutex]
826 struct buffer_head
*j_chkpt_bhs
[JBD2_NR_BATCH
];
829 * Journal head: identifies the first unused block in the journal.
832 unsigned long j_head
;
835 * Journal tail: identifies the oldest still-used block in the journal.
838 unsigned long j_tail
;
841 * Journal free: how many free blocks are there in the journal?
844 unsigned long j_free
;
847 * Journal start and end: the block numbers of the first usable block
848 * and one beyond the last usable block in the journal. [j_state_lock]
850 unsigned long j_first
;
851 unsigned long j_last
;
854 * Device, blocksize and starting block offset for the location where we
857 struct block_device
*j_dev
;
859 unsigned long long j_blk_offset
;
860 char j_devname
[BDEVNAME_SIZE
+24];
863 * Device which holds the client fs. For internal journal this will be
866 struct block_device
*j_fs_dev
;
868 /* Total maximum capacity of the journal region on disk. */
869 unsigned int j_maxlen
;
871 /* Number of buffers reserved from the running transaction */
872 atomic_t j_reserved_credits
;
875 * Protects the buffer lists and internal buffer state.
877 spinlock_t j_list_lock
;
879 /* Optional inode where we store the journal. If present, all */
880 /* journal block numbers are mapped into this inode via */
882 struct inode
*j_inode
;
885 * Sequence number of the oldest transaction in the log [j_state_lock]
887 tid_t j_tail_sequence
;
890 * Sequence number of the next transaction to grant [j_state_lock]
892 tid_t j_transaction_sequence
;
895 * Sequence number of the most recently committed transaction
898 tid_t j_commit_sequence
;
901 * Sequence number of the most recent transaction wanting commit
904 tid_t j_commit_request
;
907 * Journal uuid: identifies the object (filesystem, LVM volume etc)
908 * backed by this journal. This will eventually be replaced by an array
909 * of uuids, allowing us to index multiple devices within a single
910 * journal and to perform atomic updates across them.
914 /* Pointer to the current commit thread for this journal */
915 struct task_struct
*j_task
;
918 * Maximum number of metadata buffers to allow in a single compound
921 int j_max_transaction_buffers
;
924 * What is the maximum transaction lifetime before we begin a commit?
926 unsigned long j_commit_interval
;
928 /* The timer used to wakeup the commit thread: */
929 struct timer_list j_commit_timer
;
932 * The revoke table: maintains the list of revoked blocks in the
933 * current transaction. [j_revoke_lock]
935 spinlock_t j_revoke_lock
;
936 struct jbd2_revoke_table_s
*j_revoke
;
937 struct jbd2_revoke_table_s
*j_revoke_table
[2];
940 * array of bhs for jbd2_journal_commit_transaction
942 struct buffer_head
**j_wbuf
;
946 * this is the pid of hte last person to run a synchronous operation
947 * through the journal
949 pid_t j_last_sync_writer
;
952 * the average amount of time in nanoseconds it takes to commit a
953 * transaction to disk. [j_state_lock]
955 u64 j_average_commit_time
;
958 * minimum and maximum times that we should wait for
959 * additional filesystem operations to get batched into a
960 * synchronous handle in microseconds
962 u32 j_min_batch_time
;
963 u32 j_max_batch_time
;
965 /* This function is called when a transaction is closed */
966 void (*j_commit_callback
)(journal_t
*,
972 spinlock_t j_history_lock
;
973 struct proc_dir_entry
*j_proc_entry
;
974 struct transaction_stats_s j_stats
;
976 /* Failed journal commit ID */
977 unsigned int j_failed_commit
;
980 * An opaque pointer to fs-private information. ext3 puts its
981 * superblock pointer here
985 /* Reference to checksum algorithm driver via cryptoapi */
986 struct crypto_shash
*j_chksum_driver
;
988 /* Precomputed journal UUID checksum for seeding other checksums */
993 * Journal flag definitions
995 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
996 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
997 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
998 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
999 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1000 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
1001 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1002 * data write error in ordered
1006 * Function declarations for the journaling transaction and buffer
1010 /* Filing buffers */
1011 extern void jbd2_journal_unfile_buffer(journal_t
*, struct journal_head
*);
1012 extern void __jbd2_journal_refile_buffer(struct journal_head
*);
1013 extern void jbd2_journal_refile_buffer(journal_t
*, struct journal_head
*);
1014 extern void __jbd2_journal_file_buffer(struct journal_head
*, transaction_t
*, int);
1015 extern void __journal_free_buffer(struct journal_head
*bh
);
1016 extern void jbd2_journal_file_buffer(struct journal_head
*, transaction_t
*, int);
1017 extern void __journal_clean_data_list(transaction_t
*transaction
);
1018 static inline void jbd2_file_log_bh(struct list_head
*head
, struct buffer_head
*bh
)
1020 list_add_tail(&bh
->b_assoc_buffers
, head
);
1022 static inline void jbd2_unfile_log_bh(struct buffer_head
*bh
)
1024 list_del_init(&bh
->b_assoc_buffers
);
1027 /* Log buffer allocation */
1028 struct buffer_head
*jbd2_journal_get_descriptor_buffer(journal_t
*journal
);
1029 int jbd2_journal_next_log_block(journal_t
*, unsigned long long *);
1030 int jbd2_journal_get_log_tail(journal_t
*journal
, tid_t
*tid
,
1031 unsigned long *block
);
1032 void __jbd2_update_log_tail(journal_t
*journal
, tid_t tid
, unsigned long block
);
1033 void jbd2_update_log_tail(journal_t
*journal
, tid_t tid
, unsigned long block
);
1035 /* Commit management */
1036 extern void jbd2_journal_commit_transaction(journal_t
*);
1038 /* Checkpoint list management */
1039 int __jbd2_journal_clean_checkpoint_list(journal_t
*journal
);
1040 int __jbd2_journal_remove_checkpoint(struct journal_head
*);
1041 void __jbd2_journal_insert_checkpoint(struct journal_head
*, transaction_t
*);
1048 struct jbd2_buffer_trigger_type
{
1050 * Fired a the moment data to write to the journal are known to be
1051 * stable - so either at the moment b_frozen_data is created or just
1052 * before a buffer is written to the journal. mapped_data is a mapped
1053 * buffer that is the frozen data for commit.
1055 void (*t_frozen
)(struct jbd2_buffer_trigger_type
*type
,
1056 struct buffer_head
*bh
, void *mapped_data
,
1060 * Fired during journal abort for dirty buffers that will not be
1063 void (*t_abort
)(struct jbd2_buffer_trigger_type
*type
,
1064 struct buffer_head
*bh
);
1067 extern void jbd2_buffer_frozen_trigger(struct journal_head
*jh
,
1069 struct jbd2_buffer_trigger_type
*triggers
);
1070 extern void jbd2_buffer_abort_trigger(struct journal_head
*jh
,
1071 struct jbd2_buffer_trigger_type
*triggers
);
1074 extern int jbd2_journal_write_metadata_buffer(transaction_t
*transaction
,
1075 struct journal_head
*jh_in
,
1076 struct buffer_head
**bh_out
,
1079 /* Transaction locking */
1080 extern void __wait_on_journal (journal_t
*);
1082 /* Transaction cache support */
1083 extern void jbd2_journal_destroy_transaction_cache(void);
1084 extern int jbd2_journal_init_transaction_cache(void);
1085 extern void jbd2_journal_free_transaction(transaction_t
*);
1090 * We need to lock the journal during transaction state changes so that nobody
1091 * ever tries to take a handle on the running transaction while we are in the
1092 * middle of moving it to the commit phase. j_state_lock does this.
1094 * Note that the locking is completely interrupt unsafe. We never touch
1095 * journal structures from interrupts.
1098 static inline handle_t
*journal_current_handle(void)
1100 return current
->journal_info
;
1103 /* The journaling code user interface:
1105 * Create and destroy handles
1106 * Register buffer modifications against the current transaction.
1109 extern handle_t
*jbd2_journal_start(journal_t
*, int nblocks
);
1110 extern handle_t
*jbd2__journal_start(journal_t
*, int blocks
, int rsv_blocks
,
1111 gfp_t gfp_mask
, unsigned int type
,
1112 unsigned int line_no
);
1113 extern int jbd2_journal_restart(handle_t
*, int nblocks
);
1114 extern int jbd2__journal_restart(handle_t
*, int nblocks
, gfp_t gfp_mask
);
1115 extern int jbd2_journal_start_reserved(handle_t
*handle
,
1116 unsigned int type
, unsigned int line_no
);
1117 extern void jbd2_journal_free_reserved(handle_t
*handle
);
1118 extern int jbd2_journal_extend (handle_t
*, int nblocks
);
1119 extern int jbd2_journal_get_write_access(handle_t
*, struct buffer_head
*);
1120 extern int jbd2_journal_get_create_access (handle_t
*, struct buffer_head
*);
1121 extern int jbd2_journal_get_undo_access(handle_t
*, struct buffer_head
*);
1122 void jbd2_journal_set_triggers(struct buffer_head
*,
1123 struct jbd2_buffer_trigger_type
*type
);
1124 extern int jbd2_journal_dirty_metadata (handle_t
*, struct buffer_head
*);
1125 extern int jbd2_journal_forget (handle_t
*, struct buffer_head
*);
1126 extern void journal_sync_buffer (struct buffer_head
*);
1127 extern int jbd2_journal_invalidatepage(journal_t
*,
1128 struct page
*, unsigned int, unsigned int);
1129 extern int jbd2_journal_try_to_free_buffers(journal_t
*, struct page
*, gfp_t
);
1130 extern int jbd2_journal_stop(handle_t
*);
1131 extern int jbd2_journal_flush (journal_t
*);
1132 extern void jbd2_journal_lock_updates (journal_t
*);
1133 extern void jbd2_journal_unlock_updates (journal_t
*);
1135 extern journal_t
* jbd2_journal_init_dev(struct block_device
*bdev
,
1136 struct block_device
*fs_dev
,
1137 unsigned long long start
, int len
, int bsize
);
1138 extern journal_t
* jbd2_journal_init_inode (struct inode
*);
1139 extern int jbd2_journal_update_format (journal_t
*);
1140 extern int jbd2_journal_check_used_features
1141 (journal_t
*, unsigned long, unsigned long, unsigned long);
1142 extern int jbd2_journal_check_available_features
1143 (journal_t
*, unsigned long, unsigned long, unsigned long);
1144 extern int jbd2_journal_set_features
1145 (journal_t
*, unsigned long, unsigned long, unsigned long);
1146 extern void jbd2_journal_clear_features
1147 (journal_t
*, unsigned long, unsigned long, unsigned long);
1148 extern int jbd2_journal_load (journal_t
*journal
);
1149 extern int jbd2_journal_destroy (journal_t
*);
1150 extern int jbd2_journal_recover (journal_t
*journal
);
1151 extern int jbd2_journal_wipe (journal_t
*, int);
1152 extern int jbd2_journal_skip_recovery (journal_t
*);
1153 extern void jbd2_journal_update_sb_errno(journal_t
*);
1154 extern void jbd2_journal_update_sb_log_tail (journal_t
*, tid_t
,
1155 unsigned long, int);
1156 extern void __jbd2_journal_abort_hard (journal_t
*);
1157 extern void jbd2_journal_abort (journal_t
*, int);
1158 extern int jbd2_journal_errno (journal_t
*);
1159 extern void jbd2_journal_ack_err (journal_t
*);
1160 extern int jbd2_journal_clear_err (journal_t
*);
1161 extern int jbd2_journal_bmap(journal_t
*, unsigned long, unsigned long long *);
1162 extern int jbd2_journal_force_commit(journal_t
*);
1163 extern int jbd2_journal_force_commit_nested(journal_t
*);
1164 extern int jbd2_journal_file_inode(handle_t
*handle
, struct jbd2_inode
*inode
);
1165 extern int jbd2_journal_begin_ordered_truncate(journal_t
*journal
,
1166 struct jbd2_inode
*inode
, loff_t new_size
);
1167 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode
*jinode
, struct inode
*inode
);
1168 extern void jbd2_journal_release_jbd_inode(journal_t
*journal
, struct jbd2_inode
*jinode
);
1171 * journal_head management
1173 struct journal_head
*jbd2_journal_add_journal_head(struct buffer_head
*bh
);
1174 struct journal_head
*jbd2_journal_grab_journal_head(struct buffer_head
*bh
);
1175 void jbd2_journal_put_journal_head(struct journal_head
*jh
);
1180 extern struct kmem_cache
*jbd2_handle_cache
;
1182 static inline handle_t
*jbd2_alloc_handle(gfp_t gfp_flags
)
1184 return kmem_cache_zalloc(jbd2_handle_cache
, gfp_flags
);
1187 static inline void jbd2_free_handle(handle_t
*handle
)
1189 kmem_cache_free(jbd2_handle_cache
, handle
);
1193 * jbd2_inode management (optional, for those file systems that want to use
1194 * dynamically allocated jbd2_inode structures)
1196 extern struct kmem_cache
*jbd2_inode_cache
;
1198 static inline struct jbd2_inode
*jbd2_alloc_inode(gfp_t gfp_flags
)
1200 return kmem_cache_alloc(jbd2_inode_cache
, gfp_flags
);
1203 static inline void jbd2_free_inode(struct jbd2_inode
*jinode
)
1205 kmem_cache_free(jbd2_inode_cache
, jinode
);
1208 /* Primary revoke support */
1209 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1210 extern int jbd2_journal_init_revoke(journal_t
*, int);
1211 extern void jbd2_journal_destroy_revoke_caches(void);
1212 extern int jbd2_journal_init_revoke_caches(void);
1214 extern void jbd2_journal_destroy_revoke(journal_t
*);
1215 extern int jbd2_journal_revoke (handle_t
*, unsigned long long, struct buffer_head
*);
1216 extern int jbd2_journal_cancel_revoke(handle_t
*, struct journal_head
*);
1217 extern void jbd2_journal_write_revoke_records(journal_t
*journal
,
1218 transaction_t
*transaction
,
1219 struct list_head
*log_bufs
,
1222 /* Recovery revoke support */
1223 extern int jbd2_journal_set_revoke(journal_t
*, unsigned long long, tid_t
);
1224 extern int jbd2_journal_test_revoke(journal_t
*, unsigned long long, tid_t
);
1225 extern void jbd2_journal_clear_revoke(journal_t
*);
1226 extern void jbd2_journal_switch_revoke_table(journal_t
*journal
);
1227 extern void jbd2_clear_buffer_revoked_flags(journal_t
*journal
);
1230 * The log thread user interface:
1232 * Request space in the current transaction, and force transaction commit
1233 * transitions on demand.
1236 int jbd2_log_start_commit(journal_t
*journal
, tid_t tid
);
1237 int __jbd2_log_start_commit(journal_t
*journal
, tid_t tid
);
1238 int jbd2_journal_start_commit(journal_t
*journal
, tid_t
*tid
);
1239 int jbd2_log_wait_commit(journal_t
*journal
, tid_t tid
);
1240 int jbd2_complete_transaction(journal_t
*journal
, tid_t tid
);
1241 int jbd2_log_do_checkpoint(journal_t
*journal
);
1242 int jbd2_trans_will_send_data_barrier(journal_t
*journal
, tid_t tid
);
1244 void __jbd2_log_wait_for_space(journal_t
*journal
);
1245 extern void __jbd2_journal_drop_transaction(journal_t
*, transaction_t
*);
1246 extern int jbd2_cleanup_journal_tail(journal_t
*);
1248 /* Debugging code only: */
1250 #define jbd_ENOSYS() \
1252 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1253 current->state = TASK_UNINTERRUPTIBLE; \
1260 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1261 * bit, when set, indicates that we have had a fatal error somewhere,
1262 * either inside the journaling layer or indicated to us by the client
1263 * (eg. ext3), and that we and should not commit any further
1267 static inline int is_journal_aborted(journal_t
*journal
)
1269 return journal
->j_flags
& JBD2_ABORT
;
1272 static inline int is_handle_aborted(handle_t
*handle
)
1274 if (handle
->h_aborted
)
1276 return is_journal_aborted(handle
->h_transaction
->t_journal
);
1279 static inline void jbd2_journal_abort_handle(handle_t
*handle
)
1281 handle
->h_aborted
= 1;
1284 #endif /* __KERNEL__ */
1286 /* Comparison functions for transaction IDs: perform comparisons using
1287 * modulo arithmetic so that they work over sequence number wraps. */
1289 static inline int tid_gt(tid_t x
, tid_t y
)
1291 int difference
= (x
- y
);
1292 return (difference
> 0);
1295 static inline int tid_geq(tid_t x
, tid_t y
)
1297 int difference
= (x
- y
);
1298 return (difference
>= 0);
1301 extern int jbd2_journal_blocks_per_page(struct inode
*inode
);
1302 extern size_t journal_tag_bytes(journal_t
*journal
);
1305 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
1306 * transaction control blocks.
1308 #define JBD2_CONTROL_BLOCKS_SHIFT 5
1311 * Return the minimum number of blocks which must be free in the journal
1312 * before a new transaction may be started. Must be called under j_state_lock.
1314 static inline int jbd2_space_needed(journal_t
*journal
)
1316 int nblocks
= journal
->j_max_transaction_buffers
;
1317 return nblocks
+ (nblocks
>> JBD2_CONTROL_BLOCKS_SHIFT
);
1321 * Return number of free blocks in the log. Must be called under j_state_lock.
1323 static inline unsigned long jbd2_log_space_left(journal_t
*journal
)
1325 /* Allow for rounding errors */
1326 unsigned long free
= journal
->j_free
- 32;
1328 if (journal
->j_committing_transaction
) {
1329 unsigned long committing
= atomic_read(&journal
->
1330 j_committing_transaction
->t_outstanding_credits
);
1332 /* Transaction + control blocks */
1333 free
-= committing
+ (committing
>> JBD2_CONTROL_BLOCKS_SHIFT
);
1339 * Definitions which augment the buffer_head layer
1342 /* journaling buffer types */
1343 #define BJ_None 0 /* Not journaled */
1344 #define BJ_Metadata 1 /* Normal journaled metadata */
1345 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1346 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
1347 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1350 extern int jbd_blocks_per_page(struct inode
*inode
);
1352 /* JBD uses a CRC32 checksum */
1353 #define JBD_MAX_CHECKSUM_SIZE 4
1355 static inline u32
jbd2_chksum(journal_t
*journal
, u32 crc
,
1356 const void *address
, unsigned int length
)
1359 struct shash_desc shash
;
1360 char ctx
[JBD_MAX_CHECKSUM_SIZE
];
1364 BUG_ON(crypto_shash_descsize(journal
->j_chksum_driver
) >
1365 JBD_MAX_CHECKSUM_SIZE
);
1367 desc
.shash
.tfm
= journal
->j_chksum_driver
;
1368 desc
.shash
.flags
= 0;
1369 *(u32
*)desc
.ctx
= crc
;
1371 err
= crypto_shash_update(&desc
.shash
, address
, length
);
1374 return *(u32
*)desc
.ctx
;
1379 #define buffer_trace_init(bh) do {} while (0)
1380 #define print_buffer_fields(bh) do {} while (0)
1381 #define print_buffer_trace(bh) do {} while (0)
1382 #define BUFFER_TRACE(bh, info) do {} while (0)
1383 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1384 #define JBUFFER_TRACE(jh, info) do {} while (0)
1386 #endif /* __KERNEL__ */
1388 #endif /* _LINUX_JBD2_H */