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 * Checksum v1 and v2 are mutually exclusive features.
164 struct commit_header {
168 unsigned char h_chksum_type;
169 unsigned char h_chksum_size;
170 unsigned char h_padding[2];
171 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
173 __be32 h_commit_nsec;
177 * The block tag: used to describe a single buffer in the journal.
178 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
179 * raw struct shouldn't be used for pointer math or sizeof() - use
180 * journal_tag_bytes(journal) instead to compute this.
182 typedef struct journal_block_tag_s
184 __be32 t_blocknr; /* The on-disk block number */
185 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
186 __be16 t_flags; /* See below */
187 __be32 t_blocknr_high; /* most-significant high 32bits. */
188 } journal_block_tag_t;
190 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
191 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
193 /* Tail of descriptor block, for checksumming */
194 struct jbd2_journal_block_tail {
195 __be32 t_checksum; /* crc32c(uuid+descr_block) */
199 * The revoke descriptor: used on disk to describe a series of blocks to
200 * be revoked from the log
202 typedef struct jbd2_journal_revoke_header_s
204 journal_header_t r_header;
205 __be32 r_count; /* Count of bytes used in the block */
206 } jbd2_journal_revoke_header_t;
208 /* Tail of revoke block, for checksumming */
209 struct jbd2_journal_revoke_tail {
210 __be32 r_checksum; /* crc32c(uuid+revoke_block) */
213 /* Definitions for the journal tag flags word: */
214 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
215 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
216 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
217 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
221 * The journal superblock. All fields are in big-endian byte order.
223 typedef struct journal_superblock_s
226 journal_header_t s_header;
229 /* Static information describing the journal */
230 __be32 s_blocksize; /* journal device blocksize */
231 __be32 s_maxlen; /* total blocks in journal file */
232 __be32 s_first; /* first block of log information */
235 /* Dynamic information describing the current state of the log */
236 __be32 s_sequence; /* first commit ID expected in log */
237 __be32 s_start; /* blocknr of start of log */
240 /* Error value, as set by jbd2_journal_abort(). */
244 /* Remaining fields are only valid in a version-2 superblock */
245 __be32 s_feature_compat; /* compatible feature set */
246 __be32 s_feature_incompat; /* incompatible feature set */
247 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
249 __u8 s_uuid[16]; /* 128-bit uuid for journal */
252 __be32 s_nr_users; /* Nr of filesystems sharing log */
254 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
257 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
258 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
261 __u8 s_checksum_type; /* checksum type */
264 __be32 s_checksum; /* crc32c(superblock) */
267 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
269 } journal_superblock_t;
271 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
272 ((j)->j_format_version >= 2 && \
273 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
274 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
275 ((j)->j_format_version >= 2 && \
276 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
277 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
278 ((j)->j_format_version >= 2 && \
279 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
281 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
283 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
284 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
285 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
286 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
288 /* Features known to this kernel version: */
289 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
290 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
291 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
292 JBD2_FEATURE_INCOMPAT_64BIT | \
293 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
294 JBD2_FEATURE_INCOMPAT_CSUM_V2)
298 #include <linux/fs.h>
299 #include <linux/sched.h>
301 enum jbd_state_bits {
302 BH_JBD /* Has an attached ext3 journal_head */
304 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
305 BH_Freed, /* Has been freed (truncated) */
306 BH_Revoked, /* Has been revoked from the log */
307 BH_RevokeValid, /* Revoked flag is valid */
308 BH_JBDDirty, /* Is dirty but journaled */
309 BH_State, /* Pins most journal_head state */
310 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
311 BH_Shadow, /* IO on shadow buffer is running */
312 BH_Verified, /* Metadata block has been verified ok */
313 BH_JBDPrivateStart, /* First bit available for private use by FS */
317 BUFFER_FNS(JWrite, jwrite)
318 BUFFER_FNS(JBDDirty, jbddirty)
319 TAS_BUFFER_FNS(JBDDirty, jbddirty)
320 BUFFER_FNS(Revoked, revoked)
321 TAS_BUFFER_FNS(Revoked, revoked)
322 BUFFER_FNS(RevokeValid, revokevalid)
323 TAS_BUFFER_FNS(RevokeValid, revokevalid)
324 BUFFER_FNS(Freed, freed)
325 BUFFER_FNS(Shadow, shadow)
326 BUFFER_FNS(Verified, verified)
328 #include <linux/jbd_common.h>
330 #define J_ASSERT(assert) BUG_ON(!(assert))
332 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
333 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
335 #if defined(JBD2_PARANOID_IOFAIL)
336 #define J_EXPECT(expr, why...) J_ASSERT(expr)
337 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
338 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
340 #define __journal_expect(expr, why...) \
345 "JBD2 unexpected failure: %s: %s;\n", \
347 printk(KERN_ERR why "\n"); \
351 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
352 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
353 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
356 /* Flags in jbd_inode->i_flags */
357 #define __JI_COMMIT_RUNNING 0
358 /* Commit of the inode data in progress. We use this flag to protect us from
359 * concurrent deletion of inode. We cannot use reference to inode for this
360 * since we cannot afford doing last iput() on behalf of kjournald
362 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
365 * struct jbd_inode is the structure linking inodes in ordered mode
366 * present in a transaction so that we can sync them during commit.
369 /* Which transaction does this inode belong to? Either the running
370 * transaction or the committing one. [j_list_lock] */
371 transaction_t *i_transaction;
373 /* Pointer to the running transaction modifying inode's data in case
374 * there is already a committing transaction touching it. [j_list_lock] */
375 transaction_t *i_next_transaction;
377 /* List of inodes in the i_transaction [j_list_lock] */
378 struct list_head i_list;
380 /* VFS inode this inode belongs to [constant during the lifetime
381 * of the structure] */
382 struct inode *i_vfs_inode;
384 /* Flags of inode [j_list_lock] */
385 unsigned long i_flags;
388 struct jbd2_revoke_table_s;
391 * struct handle_s - The handle_s type is the concrete type associated with
393 * @h_transaction: Which compound transaction is this update a part of?
394 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
395 * @h_ref: Reference count on this handle
396 * @h_err: Field for caller's use to track errors through large fs operations
397 * @h_sync: flag for sync-on-close
398 * @h_jdata: flag to force data journaling
399 * @h_aborted: flag indicating fatal error on handle
402 /* Docbook can't yet cope with the bit fields, but will leave the documentation
403 * in so it can be fixed later.
406 struct jbd2_journal_handle
409 /* Which compound transaction is this update a part of? */
410 transaction_t *h_transaction;
411 /* Which journal handle belongs to - used iff h_reserved set */
412 journal_t *h_journal;
415 /* Handle reserved for finishing the logical operation */
416 handle_t *h_rsv_handle;
418 /* Number of remaining buffers we are allowed to dirty: */
419 int h_buffer_credits;
421 /* Reference count on this handle */
424 /* Field for caller's use to track errors through large fs */
428 /* Flags [no locking] */
429 unsigned int h_sync: 1; /* sync-on-close */
430 unsigned int h_jdata: 1; /* force data journaling */
431 unsigned int h_reserved: 1; /* handle with reserved credits */
432 unsigned int h_aborted: 1; /* fatal error on handle */
433 unsigned int h_type: 8; /* for handle statistics */
434 unsigned int h_line_no: 16; /* for handle statistics */
436 unsigned long h_start_jiffies;
437 unsigned int h_requested_credits;
439 #ifdef CONFIG_DEBUG_LOCK_ALLOC
440 struct lockdep_map h_lockdep_map;
446 * Some stats for checkpoint phase
448 struct transaction_chp_stats_s {
449 unsigned long cs_chp_time;
450 __u32 cs_forced_to_close;
455 /* The transaction_t type is the guts of the journaling mechanism. It
456 * tracks a compound transaction through its various states:
458 * RUNNING: accepting new updates
459 * LOCKED: Updates still running but we don't accept new ones
460 * RUNDOWN: Updates are tidying up but have finished requesting
461 * new buffers to modify (state not used for now)
462 * FLUSH: All updates complete, but we are still writing to disk
463 * COMMIT: All data on disk, writing commit record
464 * FINISHED: We still have to keep the transaction for checkpointing.
466 * The transaction keeps track of all of the buffers modified by a
467 * running transaction, and all of the buffers committed but not yet
468 * flushed to home for finished transactions.
475 * ->jbd_lock_bh_journal_head() (This is "innermost")
478 * ->jbd_lock_bh_state()
480 * jbd_lock_bh_state()
487 * ->j_list_lock (journal_unmap_buffer)
493 /* Pointer to the journal for this transaction. [no locking] */
494 journal_t *t_journal;
496 /* Sequence number for this transaction [no locking] */
500 * Transaction's current state
501 * [no locking - only kjournald2 alters this]
502 * [j_list_lock] guards transition of a transaction into T_FINISHED
503 * state and subsequent call of __jbd2_journal_drop_transaction()
504 * FIXME: needs barriers
505 * KLUDGE: [use j_state_lock]
519 * Where in the log does this transaction's commit start? [no locking]
521 unsigned long t_log_start;
523 /* Number of buffers on the t_buffers list [j_list_lock] */
527 * Doubly-linked circular list of all buffers reserved but not yet
528 * modified by this transaction [j_list_lock]
530 struct journal_head *t_reserved_list;
533 * Doubly-linked circular list of all metadata buffers owned by this
534 * transaction [j_list_lock]
536 struct journal_head *t_buffers;
539 * Doubly-linked circular list of all forget buffers (superseded
540 * buffers which we can un-checkpoint once this transaction commits)
543 struct journal_head *t_forget;
546 * Doubly-linked circular list of all buffers still to be flushed before
547 * this transaction can be checkpointed. [j_list_lock]
549 struct journal_head *t_checkpoint_list;
552 * Doubly-linked circular list of all buffers submitted for IO while
553 * checkpointing. [j_list_lock]
555 struct journal_head *t_checkpoint_io_list;
558 * Doubly-linked circular list of metadata buffers being shadowed by log
559 * IO. The IO buffers on the iobuf list and the shadow buffers on this
560 * list match each other one for one at all times. [j_list_lock]
562 struct journal_head *t_shadow_list;
565 * List of inodes whose data we've modified in data=ordered mode.
568 struct list_head t_inode_list;
571 * Protects info related to handles
573 spinlock_t t_handle_lock;
576 * Longest time some handle had to wait for running transaction
578 unsigned long t_max_wait;
581 * When transaction started
583 unsigned long t_start;
586 * When commit was requested
588 unsigned long t_requested;
591 * Checkpointing stats [j_checkpoint_sem]
593 struct transaction_chp_stats_s t_chp_stats;
596 * Number of outstanding updates running on this transaction
602 * Number of buffers reserved for use by all handles in this transaction
603 * handle but not yet modified. [t_handle_lock]
605 atomic_t t_outstanding_credits;
608 * Forward and backward links for the circular list of all transactions
609 * awaiting checkpoint. [j_list_lock]
611 transaction_t *t_cpnext, *t_cpprev;
614 * When will the transaction expire (become due for commit), in jiffies?
617 unsigned long t_expires;
620 * When this transaction started, in nanoseconds [no locking]
622 ktime_t t_start_time;
625 * How many handles used this transaction? [t_handle_lock]
627 atomic_t t_handle_count;
630 * This transaction is being forced and some process is
631 * waiting for it to finish.
633 unsigned int t_synchronous_commit:1;
635 /* Disk flush needs to be sent to fs partition [no locking] */
636 int t_need_data_flush;
639 * For use by the filesystem to store fs-specific data
640 * structures associated with the transaction
642 struct list_head t_private_list;
645 struct transaction_run_stats_s {
646 unsigned long rs_wait;
647 unsigned long rs_request_delay;
648 unsigned long rs_running;
649 unsigned long rs_locked;
650 unsigned long rs_flushing;
651 unsigned long rs_logging;
653 __u32 rs_handle_count;
655 __u32 rs_blocks_logged;
658 struct transaction_stats_s {
659 unsigned long ts_tid;
660 unsigned long ts_requested;
661 struct transaction_run_stats_s run;
664 static inline unsigned long
665 jbd2_time_diff(unsigned long start, unsigned long end)
670 return end + (MAX_JIFFY_OFFSET - start);
673 #define JBD2_NR_BATCH 64
676 * struct journal_s - The journal_s type is the concrete type associated with
678 * @j_flags: General journaling state flags
679 * @j_errno: Is there an outstanding uncleared error on the journal (from a
681 * @j_sb_buffer: First part of superblock buffer
682 * @j_superblock: Second part of superblock buffer
683 * @j_format_version: Version of the superblock format
684 * @j_state_lock: Protect the various scalars in the journal
685 * @j_barrier_count: Number of processes waiting to create a barrier lock
686 * @j_barrier: The barrier lock itself
687 * @j_running_transaction: The current running transaction..
688 * @j_committing_transaction: the transaction we are pushing to disk
689 * @j_checkpoint_transactions: a linked circular list of all transactions
690 * waiting for checkpointing
691 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
692 * to start committing, or for a barrier lock to be released
693 * @j_wait_done_commit: Wait queue for waiting for commit to complete
694 * @j_wait_commit: Wait queue to trigger commit
695 * @j_wait_updates: Wait queue to wait for updates to complete
696 * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop
697 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
698 * @j_head: Journal head - identifies the first unused block in the journal
699 * @j_tail: Journal tail - identifies the oldest still-used block in the
701 * @j_free: Journal free - how many free blocks are there in the journal?
702 * @j_first: The block number of the first usable block
703 * @j_last: The block number one beyond the last usable block
704 * @j_dev: Device where we store the journal
705 * @j_blocksize: blocksize for the location where we store the journal.
706 * @j_blk_offset: starting block offset for into the device where we store the
708 * @j_fs_dev: Device which holds the client fs. For internal journal this will
710 * @j_reserved_credits: Number of buffers reserved from the running transaction
711 * @j_maxlen: Total maximum capacity of the journal region on disk.
712 * @j_list_lock: Protects the buffer lists and internal buffer state.
713 * @j_inode: Optional inode where we store the journal. If present, all journal
714 * block numbers are mapped into this inode via bmap().
715 * @j_tail_sequence: Sequence number of the oldest transaction in the log
716 * @j_transaction_sequence: Sequence number of the next transaction to grant
717 * @j_commit_sequence: Sequence number of the most recently committed
719 * @j_commit_request: Sequence number of the most recent transaction wanting
721 * @j_uuid: Uuid of client object.
722 * @j_task: Pointer to the current commit thread for this journal
723 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
724 * single compound commit transaction
725 * @j_commit_interval: What is the maximum transaction lifetime before we begin
727 * @j_commit_timer: The timer used to wakeup the commit thread
728 * @j_revoke_lock: Protect the revoke table
729 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
730 * current transaction.
731 * @j_revoke_table: alternate revoke tables for j_revoke
732 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
733 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
734 * number that will fit in j_blocksize
735 * @j_last_sync_writer: most recent pid which did a synchronous write
736 * @j_history: Buffer storing the transactions statistics history
737 * @j_history_max: Maximum number of transactions in the statistics history
738 * @j_history_cur: Current number of transactions in the statistics history
739 * @j_history_lock: Protect the transactions statistics history
740 * @j_proc_entry: procfs entry for the jbd statistics directory
741 * @j_stats: Overall statistics
742 * @j_private: An opaque pointer to fs-private information.
747 /* General journaling state flags [j_state_lock] */
748 unsigned long j_flags;
751 * Is there an outstanding uncleared error on the journal (from a prior
752 * abort)? [j_state_lock]
756 /* The superblock buffer */
757 struct buffer_head *j_sb_buffer;
758 journal_superblock_t *j_superblock;
760 /* Version of the superblock format */
761 int j_format_version;
764 * Protect the various scalars in the journal
766 rwlock_t j_state_lock;
769 * Number of processes waiting to create a barrier lock [j_state_lock]
773 /* The barrier lock itself */
774 struct mutex j_barrier;
777 * Transactions: The current running transaction...
778 * [j_state_lock] [caller holding open handle]
780 transaction_t *j_running_transaction;
783 * the transaction we are pushing to disk
784 * [j_state_lock] [caller holding open handle]
786 transaction_t *j_committing_transaction;
789 * ... and a linked circular list of all transactions waiting for
790 * checkpointing. [j_list_lock]
792 transaction_t *j_checkpoint_transactions;
795 * Wait queue for waiting for a locked transaction to start committing,
796 * or for a barrier lock to be released
798 wait_queue_head_t j_wait_transaction_locked;
800 /* Wait queue for waiting for commit to complete */
801 wait_queue_head_t j_wait_done_commit;
803 /* Wait queue to trigger commit */
804 wait_queue_head_t j_wait_commit;
806 /* Wait queue to wait for updates to complete */
807 wait_queue_head_t j_wait_updates;
809 /* Wait queue to wait for reserved buffer credits to drop */
810 wait_queue_head_t j_wait_reserved;
812 /* Semaphore for locking against concurrent checkpoints */
813 struct mutex j_checkpoint_mutex;
816 * List of buffer heads used by the checkpoint routine. This
817 * was moved from jbd2_log_do_checkpoint() to reduce stack
818 * usage. Access to this array is controlled by the
819 * j_checkpoint_mutex. [j_checkpoint_mutex]
821 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
824 * Journal head: identifies the first unused block in the journal.
827 unsigned long j_head;
830 * Journal tail: identifies the oldest still-used block in the journal.
833 unsigned long j_tail;
836 * Journal free: how many free blocks are there in the journal?
839 unsigned long j_free;
842 * Journal start and end: the block numbers of the first usable block
843 * and one beyond the last usable block in the journal. [j_state_lock]
845 unsigned long j_first;
846 unsigned long j_last;
849 * Device, blocksize and starting block offset for the location where we
852 struct block_device *j_dev;
854 unsigned long long j_blk_offset;
855 char j_devname[BDEVNAME_SIZE+24];
858 * Device which holds the client fs. For internal journal this will be
861 struct block_device *j_fs_dev;
863 /* Total maximum capacity of the journal region on disk. */
864 unsigned int j_maxlen;
866 /* Number of buffers reserved from the running transaction */
867 atomic_t j_reserved_credits;
870 * Protects the buffer lists and internal buffer state.
872 spinlock_t j_list_lock;
874 /* Optional inode where we store the journal. If present, all */
875 /* journal block numbers are mapped into this inode via */
877 struct inode *j_inode;
880 * Sequence number of the oldest transaction in the log [j_state_lock]
882 tid_t j_tail_sequence;
885 * Sequence number of the next transaction to grant [j_state_lock]
887 tid_t j_transaction_sequence;
890 * Sequence number of the most recently committed transaction
893 tid_t j_commit_sequence;
896 * Sequence number of the most recent transaction wanting commit
899 tid_t j_commit_request;
902 * Journal uuid: identifies the object (filesystem, LVM volume etc)
903 * backed by this journal. This will eventually be replaced by an array
904 * of uuids, allowing us to index multiple devices within a single
905 * journal and to perform atomic updates across them.
909 /* Pointer to the current commit thread for this journal */
910 struct task_struct *j_task;
913 * Maximum number of metadata buffers to allow in a single compound
916 int j_max_transaction_buffers;
919 * What is the maximum transaction lifetime before we begin a commit?
921 unsigned long j_commit_interval;
923 /* The timer used to wakeup the commit thread: */
924 struct timer_list j_commit_timer;
927 * The revoke table: maintains the list of revoked blocks in the
928 * current transaction. [j_revoke_lock]
930 spinlock_t j_revoke_lock;
931 struct jbd2_revoke_table_s *j_revoke;
932 struct jbd2_revoke_table_s *j_revoke_table[2];
935 * array of bhs for jbd2_journal_commit_transaction
937 struct buffer_head **j_wbuf;
941 * this is the pid of hte last person to run a synchronous operation
942 * through the journal
944 pid_t j_last_sync_writer;
947 * the average amount of time in nanoseconds it takes to commit a
948 * transaction to disk. [j_state_lock]
950 u64 j_average_commit_time;
953 * minimum and maximum times that we should wait for
954 * additional filesystem operations to get batched into a
955 * synchronous handle in microseconds
957 u32 j_min_batch_time;
958 u32 j_max_batch_time;
960 /* This function is called when a transaction is closed */
961 void (*j_commit_callback)(journal_t *,
967 spinlock_t j_history_lock;
968 struct proc_dir_entry *j_proc_entry;
969 struct transaction_stats_s j_stats;
971 /* Failed journal commit ID */
972 unsigned int j_failed_commit;
975 * An opaque pointer to fs-private information. ext3 puts its
976 * superblock pointer here
980 /* Reference to checksum algorithm driver via cryptoapi */
981 struct crypto_shash *j_chksum_driver;
983 /* Precomputed journal UUID checksum for seeding other checksums */
988 * Journal flag definitions
990 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
991 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
992 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
993 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
994 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
995 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
996 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
997 * data write error in ordered
1001 * Function declarations for the journaling transaction and buffer
1005 /* Filing buffers */
1006 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1007 extern void __jbd2_journal_refile_buffer(struct journal_head *);
1008 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1009 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1010 extern void __journal_free_buffer(struct journal_head *bh);
1011 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1012 extern void __journal_clean_data_list(transaction_t *transaction);
1013 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1015 list_add_tail(&bh->b_assoc_buffers, head);
1017 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1019 list_del_init(&bh->b_assoc_buffers);
1022 /* Log buffer allocation */
1023 struct buffer_head *jbd2_journal_get_descriptor_buffer(journal_t *journal);
1024 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1025 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1026 unsigned long *block);
1027 void __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1028 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1030 /* Commit management */
1031 extern void jbd2_journal_commit_transaction(journal_t *);
1033 /* Checkpoint list management */
1034 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1035 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1036 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1043 struct jbd2_buffer_trigger_type {
1045 * Fired a the moment data to write to the journal are known to be
1046 * stable - so either at the moment b_frozen_data is created or just
1047 * before a buffer is written to the journal. mapped_data is a mapped
1048 * buffer that is the frozen data for commit.
1050 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1051 struct buffer_head *bh, void *mapped_data,
1055 * Fired during journal abort for dirty buffers that will not be
1058 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1059 struct buffer_head *bh);
1062 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1064 struct jbd2_buffer_trigger_type *triggers);
1065 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1066 struct jbd2_buffer_trigger_type *triggers);
1069 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1070 struct journal_head *jh_in,
1071 struct buffer_head **bh_out,
1074 /* Transaction locking */
1075 extern void __wait_on_journal (journal_t *);
1077 /* Transaction cache support */
1078 extern void jbd2_journal_destroy_transaction_cache(void);
1079 extern int jbd2_journal_init_transaction_cache(void);
1080 extern void jbd2_journal_free_transaction(transaction_t *);
1085 * We need to lock the journal during transaction state changes so that nobody
1086 * ever tries to take a handle on the running transaction while we are in the
1087 * middle of moving it to the commit phase. j_state_lock does this.
1089 * Note that the locking is completely interrupt unsafe. We never touch
1090 * journal structures from interrupts.
1093 static inline handle_t *journal_current_handle(void)
1095 return current->journal_info;
1098 /* The journaling code user interface:
1100 * Create and destroy handles
1101 * Register buffer modifications against the current transaction.
1104 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1105 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1106 gfp_t gfp_mask, unsigned int type,
1107 unsigned int line_no);
1108 extern int jbd2_journal_restart(handle_t *, int nblocks);
1109 extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
1110 extern int jbd2_journal_start_reserved(handle_t *handle,
1111 unsigned int type, unsigned int line_no);
1112 extern void jbd2_journal_free_reserved(handle_t *handle);
1113 extern int jbd2_journal_extend (handle_t *, int nblocks);
1114 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1115 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1116 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1117 void jbd2_journal_set_triggers(struct buffer_head *,
1118 struct jbd2_buffer_trigger_type *type);
1119 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1120 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1121 extern void journal_sync_buffer (struct buffer_head *);
1122 extern int jbd2_journal_invalidatepage(journal_t *,
1123 struct page *, unsigned int, unsigned int);
1124 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1125 extern int jbd2_journal_stop(handle_t *);
1126 extern int jbd2_journal_flush (journal_t *);
1127 extern void jbd2_journal_lock_updates (journal_t *);
1128 extern void jbd2_journal_unlock_updates (journal_t *);
1130 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1131 struct block_device *fs_dev,
1132 unsigned long long start, int len, int bsize);
1133 extern journal_t * jbd2_journal_init_inode (struct inode *);
1134 extern int jbd2_journal_update_format (journal_t *);
1135 extern int jbd2_journal_check_used_features
1136 (journal_t *, unsigned long, unsigned long, unsigned long);
1137 extern int jbd2_journal_check_available_features
1138 (journal_t *, unsigned long, unsigned long, unsigned long);
1139 extern int jbd2_journal_set_features
1140 (journal_t *, unsigned long, unsigned long, unsigned long);
1141 extern void jbd2_journal_clear_features
1142 (journal_t *, unsigned long, unsigned long, unsigned long);
1143 extern int jbd2_journal_load (journal_t *journal);
1144 extern int jbd2_journal_destroy (journal_t *);
1145 extern int jbd2_journal_recover (journal_t *journal);
1146 extern int jbd2_journal_wipe (journal_t *, int);
1147 extern int jbd2_journal_skip_recovery (journal_t *);
1148 extern void jbd2_journal_update_sb_errno(journal_t *);
1149 extern void jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
1150 unsigned long, int);
1151 extern void __jbd2_journal_abort_hard (journal_t *);
1152 extern void jbd2_journal_abort (journal_t *, int);
1153 extern int jbd2_journal_errno (journal_t *);
1154 extern void jbd2_journal_ack_err (journal_t *);
1155 extern int jbd2_journal_clear_err (journal_t *);
1156 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1157 extern int jbd2_journal_force_commit(journal_t *);
1158 extern int jbd2_journal_force_commit_nested(journal_t *);
1159 extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1160 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1161 struct jbd2_inode *inode, loff_t new_size);
1162 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1163 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1166 * journal_head management
1168 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1169 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1170 void jbd2_journal_put_journal_head(struct journal_head *jh);
1175 extern struct kmem_cache *jbd2_handle_cache;
1177 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1179 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1182 static inline void jbd2_free_handle(handle_t *handle)
1184 kmem_cache_free(jbd2_handle_cache, handle);
1188 * jbd2_inode management (optional, for those file systems that want to use
1189 * dynamically allocated jbd2_inode structures)
1191 extern struct kmem_cache *jbd2_inode_cache;
1193 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1195 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1198 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1200 kmem_cache_free(jbd2_inode_cache, jinode);
1203 /* Primary revoke support */
1204 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1205 extern int jbd2_journal_init_revoke(journal_t *, int);
1206 extern void jbd2_journal_destroy_revoke_caches(void);
1207 extern int jbd2_journal_init_revoke_caches(void);
1209 extern void jbd2_journal_destroy_revoke(journal_t *);
1210 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1211 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1212 extern void jbd2_journal_write_revoke_records(journal_t *journal,
1213 transaction_t *transaction,
1214 struct list_head *log_bufs,
1217 /* Recovery revoke support */
1218 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1219 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1220 extern void jbd2_journal_clear_revoke(journal_t *);
1221 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1222 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
1225 * The log thread user interface:
1227 * Request space in the current transaction, and force transaction commit
1228 * transitions on demand.
1231 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1232 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1233 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1234 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1235 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1236 int jbd2_log_do_checkpoint(journal_t *journal);
1237 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1239 void __jbd2_log_wait_for_space(journal_t *journal);
1240 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1241 extern int jbd2_cleanup_journal_tail(journal_t *);
1243 /* Debugging code only: */
1245 #define jbd_ENOSYS() \
1247 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1248 current->state = TASK_UNINTERRUPTIBLE; \
1255 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1256 * bit, when set, indicates that we have had a fatal error somewhere,
1257 * either inside the journaling layer or indicated to us by the client
1258 * (eg. ext3), and that we and should not commit any further
1262 static inline int is_journal_aborted(journal_t *journal)
1264 return journal->j_flags & JBD2_ABORT;
1267 static inline int is_handle_aborted(handle_t *handle)
1269 if (handle->h_aborted || !handle->h_transaction)
1271 return is_journal_aborted(handle->h_transaction->t_journal);
1274 static inline void jbd2_journal_abort_handle(handle_t *handle)
1276 handle->h_aborted = 1;
1279 #endif /* __KERNEL__ */
1281 /* Comparison functions for transaction IDs: perform comparisons using
1282 * modulo arithmetic so that they work over sequence number wraps. */
1284 static inline int tid_gt(tid_t x, tid_t y)
1286 int difference = (x - y);
1287 return (difference > 0);
1290 static inline int tid_geq(tid_t x, tid_t y)
1292 int difference = (x - y);
1293 return (difference >= 0);
1296 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1297 extern size_t journal_tag_bytes(journal_t *journal);
1300 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
1301 * transaction control blocks.
1303 #define JBD2_CONTROL_BLOCKS_SHIFT 5
1306 * Return the minimum number of blocks which must be free in the journal
1307 * before a new transaction may be started. Must be called under j_state_lock.
1309 static inline int jbd2_space_needed(journal_t *journal)
1311 int nblocks = journal->j_max_transaction_buffers;
1312 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT);
1316 * Return number of free blocks in the log. Must be called under j_state_lock.
1318 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1320 /* Allow for rounding errors */
1321 unsigned long free = journal->j_free - 32;
1323 if (journal->j_committing_transaction) {
1324 unsigned long committing = atomic_read(&journal->
1325 j_committing_transaction->t_outstanding_credits);
1327 /* Transaction + control blocks */
1328 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT);
1334 * Definitions which augment the buffer_head layer
1337 /* journaling buffer types */
1338 #define BJ_None 0 /* Not journaled */
1339 #define BJ_Metadata 1 /* Normal journaled metadata */
1340 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1341 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
1342 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1345 extern int jbd_blocks_per_page(struct inode *inode);
1347 /* JBD uses a CRC32 checksum */
1348 #define JBD_MAX_CHECKSUM_SIZE 4
1350 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1351 const void *address, unsigned int length)
1354 struct shash_desc shash;
1355 char ctx[JBD_MAX_CHECKSUM_SIZE];
1359 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1360 JBD_MAX_CHECKSUM_SIZE);
1362 desc.shash.tfm = journal->j_chksum_driver;
1363 desc.shash.flags = 0;
1364 *(u32 *)desc.ctx = crc;
1366 err = crypto_shash_update(&desc.shash, address, length);
1369 return *(u32 *)desc.ctx;
1372 /* Return most recent uncommitted transaction */
1373 static inline tid_t jbd2_get_latest_transaction(journal_t *journal)
1377 read_lock(&journal->j_state_lock);
1378 tid = journal->j_commit_request;
1379 if (journal->j_running_transaction)
1380 tid = journal->j_running_transaction->t_tid;
1381 read_unlock(&journal->j_state_lock);
1387 #define buffer_trace_init(bh) do {} while (0)
1388 #define print_buffer_fields(bh) do {} while (0)
1389 #define print_buffer_trace(bh) do {} while (0)
1390 #define BUFFER_TRACE(bh, info) do {} while (0)
1391 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1392 #define JBUFFER_TRACE(jh, info) do {} while (0)
1394 #endif /* __KERNEL__ */
1396 #endif /* _LINUX_JBD2_H */