writeback: split writeback_inodes_wb
[linux-2.6/next.git] / include / linux / jbd2.h
bloba4d2e9f7088ada70d8b1357f418c32cd09a5bf1a
1 /*
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.
16 #ifndef _LINUX_JBD2_H
17 #define _LINUX_JBD2_H
19 /* Allow this file to be included directly into e2fsprogs */
20 #ifndef __KERNEL__
21 #include "jfs_compat.h"
22 #define JBD2_DEBUG
23 #define jfs_debug jbd_debug
24 #else
26 #include <linux/types.h>
27 #include <linux/buffer_head.h>
28 #include <linux/journal-head.h>
29 #include <linux/stddef.h>
30 #include <linux/bit_spinlock.h>
31 #include <linux/mutex.h>
32 #include <linux/timer.h>
33 #include <linux/slab.h>
34 #endif
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 u8 jbd2_journal_enable_debug;
61 #define jbd_debug(n, f, a...) \
62 do { \
63 if ((n) <= jbd2_journal_enable_debug) { \
64 printk (KERN_DEBUG "(%s, %d): %s: ", \
65 __FILE__, __LINE__, __func__); \
66 printk (f, ## a); \
67 } \
68 } while (0)
69 #else
70 #define jbd_debug(f, a...) /**/
71 #endif
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
78 #ifdef __KERNEL__
80 /**
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.
96 **/
97 typedef struct handle_s 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
107 * writing process.
109 * This is an opaque datatype.
111 typedef struct journal_s journal_t; /* Journal control structure */
112 #endif
115 * Internal structures used by the logging mechanism:
118 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
121 * On-disk structures
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
139 __be32 h_magic;
140 __be32 h_blocktype;
141 __be32 h_sequence;
142 } journal_header_t;
145 * Checksum types.
147 #define JBD2_CRC32_CHKSUM 1
148 #define JBD2_MD5_CHKSUM 2
149 #define JBD2_SHA1_CHKSUM 3
151 #define JBD2_CRC32_CHKSUM_SIZE 4
153 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
155 * Commit block header for storing transactional checksums:
157 struct commit_header {
158 __be32 h_magic;
159 __be32 h_blocktype;
160 __be32 h_sequence;
161 unsigned char h_chksum_type;
162 unsigned char h_chksum_size;
163 unsigned char h_padding[2];
164 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
165 __be64 h_commit_sec;
166 __be32 h_commit_nsec;
170 * The block tag: used to describe a single buffer in the journal.
171 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
172 * raw struct shouldn't be used for pointer math or sizeof() - use
173 * journal_tag_bytes(journal) instead to compute this.
175 typedef struct journal_block_tag_s
177 __be32 t_blocknr; /* The on-disk block number */
178 __be32 t_flags; /* See below */
179 __be32 t_blocknr_high; /* most-significant high 32bits. */
180 } journal_block_tag_t;
182 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
183 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
186 * The revoke descriptor: used on disk to describe a series of blocks to
187 * be revoked from the log
189 typedef struct jbd2_journal_revoke_header_s
191 journal_header_t r_header;
192 __be32 r_count; /* Count of bytes used in the block */
193 } jbd2_journal_revoke_header_t;
196 /* Definitions for the journal tag flags word: */
197 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
198 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
199 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
200 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
204 * The journal superblock. All fields are in big-endian byte order.
206 typedef struct journal_superblock_s
208 /* 0x0000 */
209 journal_header_t s_header;
211 /* 0x000C */
212 /* Static information describing the journal */
213 __be32 s_blocksize; /* journal device blocksize */
214 __be32 s_maxlen; /* total blocks in journal file */
215 __be32 s_first; /* first block of log information */
217 /* 0x0018 */
218 /* Dynamic information describing the current state of the log */
219 __be32 s_sequence; /* first commit ID expected in log */
220 __be32 s_start; /* blocknr of start of log */
222 /* 0x0020 */
223 /* Error value, as set by jbd2_journal_abort(). */
224 __be32 s_errno;
226 /* 0x0024 */
227 /* Remaining fields are only valid in a version-2 superblock */
228 __be32 s_feature_compat; /* compatible feature set */
229 __be32 s_feature_incompat; /* incompatible feature set */
230 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
231 /* 0x0030 */
232 __u8 s_uuid[16]; /* 128-bit uuid for journal */
234 /* 0x0040 */
235 __be32 s_nr_users; /* Nr of filesystems sharing log */
237 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
239 /* 0x0048 */
240 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
241 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
243 /* 0x0050 */
244 __u32 s_padding[44];
246 /* 0x0100 */
247 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
248 /* 0x0400 */
249 } journal_superblock_t;
251 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
252 ((j)->j_format_version >= 2 && \
253 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
254 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
255 ((j)->j_format_version >= 2 && \
256 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
257 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
258 ((j)->j_format_version >= 2 && \
259 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
261 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
263 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
264 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
265 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
267 /* Features known to this kernel version: */
268 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
269 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
270 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
271 JBD2_FEATURE_INCOMPAT_64BIT | \
272 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)
274 #ifdef __KERNEL__
276 #include <linux/fs.h>
277 #include <linux/sched.h>
279 #define J_ASSERT(assert) BUG_ON(!(assert))
281 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
282 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
284 #if defined(JBD2_PARANOID_IOFAIL)
285 #define J_EXPECT(expr, why...) J_ASSERT(expr)
286 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
287 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
288 #else
289 #define __journal_expect(expr, why...) \
290 ({ \
291 int val = (expr); \
292 if (!val) { \
293 printk(KERN_ERR \
294 "JBD2 unexpected failure: %s: %s;\n", \
295 __func__, #expr); \
296 printk(KERN_ERR why "\n"); \
298 val; \
300 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
301 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
302 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
303 #endif
305 enum jbd_state_bits {
306 BH_JBD /* Has an attached ext3 journal_head */
307 = BH_PrivateStart,
308 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
309 BH_Freed, /* Has been freed (truncated) */
310 BH_Revoked, /* Has been revoked from the log */
311 BH_RevokeValid, /* Revoked flag is valid */
312 BH_JBDDirty, /* Is dirty but journaled */
313 BH_State, /* Pins most journal_head state */
314 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
315 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
316 BH_JBDPrivateStart, /* First bit available for private use by FS */
319 BUFFER_FNS(JBD, jbd)
320 BUFFER_FNS(JWrite, jwrite)
321 BUFFER_FNS(JBDDirty, jbddirty)
322 TAS_BUFFER_FNS(JBDDirty, jbddirty)
323 BUFFER_FNS(Revoked, revoked)
324 TAS_BUFFER_FNS(Revoked, revoked)
325 BUFFER_FNS(RevokeValid, revokevalid)
326 TAS_BUFFER_FNS(RevokeValid, revokevalid)
327 BUFFER_FNS(Freed, freed)
329 static inline struct buffer_head *jh2bh(struct journal_head *jh)
331 return jh->b_bh;
334 static inline struct journal_head *bh2jh(struct buffer_head *bh)
336 return bh->b_private;
339 static inline void jbd_lock_bh_state(struct buffer_head *bh)
341 bit_spin_lock(BH_State, &bh->b_state);
344 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
346 return bit_spin_trylock(BH_State, &bh->b_state);
349 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
351 return bit_spin_is_locked(BH_State, &bh->b_state);
354 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
356 bit_spin_unlock(BH_State, &bh->b_state);
359 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
361 bit_spin_lock(BH_JournalHead, &bh->b_state);
364 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
366 bit_spin_unlock(BH_JournalHead, &bh->b_state);
369 /* Flags in jbd_inode->i_flags */
370 #define __JI_COMMIT_RUNNING 0
371 /* Commit of the inode data in progress. We use this flag to protect us from
372 * concurrent deletion of inode. We cannot use reference to inode for this
373 * since we cannot afford doing last iput() on behalf of kjournald
375 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
378 * struct jbd_inode is the structure linking inodes in ordered mode
379 * present in a transaction so that we can sync them during commit.
381 struct jbd2_inode {
382 /* Which transaction does this inode belong to? Either the running
383 * transaction or the committing one. [j_list_lock] */
384 transaction_t *i_transaction;
386 /* Pointer to the running transaction modifying inode's data in case
387 * there is already a committing transaction touching it. [j_list_lock] */
388 transaction_t *i_next_transaction;
390 /* List of inodes in the i_transaction [j_list_lock] */
391 struct list_head i_list;
393 /* VFS inode this inode belongs to [constant during the lifetime
394 * of the structure] */
395 struct inode *i_vfs_inode;
397 /* Flags of inode [j_list_lock] */
398 unsigned int i_flags;
401 struct jbd2_revoke_table_s;
404 * struct handle_s - The handle_s type is the concrete type associated with
405 * handle_t.
406 * @h_transaction: Which compound transaction is this update a part of?
407 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
408 * @h_ref: Reference count on this handle
409 * @h_err: Field for caller's use to track errors through large fs operations
410 * @h_sync: flag for sync-on-close
411 * @h_jdata: flag to force data journaling
412 * @h_aborted: flag indicating fatal error on handle
415 /* Docbook can't yet cope with the bit fields, but will leave the documentation
416 * in so it can be fixed later.
419 struct handle_s
421 /* Which compound transaction is this update a part of? */
422 transaction_t *h_transaction;
424 /* Number of remaining buffers we are allowed to dirty: */
425 int h_buffer_credits;
427 /* Reference count on this handle */
428 int h_ref;
430 /* Field for caller's use to track errors through large fs */
431 /* operations */
432 int h_err;
434 /* Flags [no locking] */
435 unsigned int h_sync: 1; /* sync-on-close */
436 unsigned int h_jdata: 1; /* force data journaling */
437 unsigned int h_aborted: 1; /* fatal error on handle */
439 #ifdef CONFIG_DEBUG_LOCK_ALLOC
440 struct lockdep_map h_lockdep_map;
441 #endif
446 * Some stats for checkpoint phase
448 struct transaction_chp_stats_s {
449 unsigned long cs_chp_time;
450 __u32 cs_forced_to_close;
451 __u32 cs_written;
452 __u32 cs_dropped;
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.
472 * Lock ranking:
474 * j_list_lock
475 * ->jbd_lock_bh_journal_head() (This is "innermost")
477 * j_state_lock
478 * ->jbd_lock_bh_state()
480 * jbd_lock_bh_state()
481 * ->j_list_lock
483 * j_state_lock
484 * ->t_handle_lock
486 * j_state_lock
487 * ->j_list_lock (journal_unmap_buffer)
491 struct transaction_s
493 /* Pointer to the journal for this transaction. [no locking] */
494 journal_t *t_journal;
496 /* Sequence number for this transaction [no locking] */
497 tid_t t_tid;
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]
507 enum {
508 T_RUNNING,
509 T_LOCKED,
510 T_RUNDOWN,
511 T_FLUSH,
512 T_COMMIT,
513 T_FINISHED
514 } t_state;
517 * Where in the log does this transaction's commit start? [no locking]
519 unsigned long t_log_start;
521 /* Number of buffers on the t_buffers list [j_list_lock] */
522 int t_nr_buffers;
525 * Doubly-linked circular list of all buffers reserved but not yet
526 * modified by this transaction [j_list_lock]
528 struct journal_head *t_reserved_list;
531 * Doubly-linked circular list of all metadata buffers owned by this
532 * transaction [j_list_lock]
534 struct journal_head *t_buffers;
537 * Doubly-linked circular list of all forget buffers (superseded
538 * buffers which we can un-checkpoint once this transaction commits)
539 * [j_list_lock]
541 struct journal_head *t_forget;
544 * Doubly-linked circular list of all buffers still to be flushed before
545 * this transaction can be checkpointed. [j_list_lock]
547 struct journal_head *t_checkpoint_list;
550 * Doubly-linked circular list of all buffers submitted for IO while
551 * checkpointing. [j_list_lock]
553 struct journal_head *t_checkpoint_io_list;
556 * Doubly-linked circular list of temporary buffers currently undergoing
557 * IO in the log [j_list_lock]
559 struct journal_head *t_iobuf_list;
562 * Doubly-linked circular list of metadata buffers being shadowed by log
563 * IO. The IO buffers on the iobuf list and the shadow buffers on this
564 * list match each other one for one at all times. [j_list_lock]
566 struct journal_head *t_shadow_list;
569 * Doubly-linked circular list of control buffers being written to the
570 * log. [j_list_lock]
572 struct journal_head *t_log_list;
575 * List of inodes whose data we've modified in data=ordered mode.
576 * [j_list_lock]
578 struct list_head t_inode_list;
581 * Protects info related to handles
583 spinlock_t t_handle_lock;
586 * Longest time some handle had to wait for running transaction
588 unsigned long t_max_wait;
591 * When transaction started
593 unsigned long t_start;
596 * Checkpointing stats [j_checkpoint_sem]
598 struct transaction_chp_stats_s t_chp_stats;
601 * Number of outstanding updates running on this transaction
602 * [t_handle_lock]
604 int t_updates;
607 * Number of buffers reserved for use by all handles in this transaction
608 * handle but not yet modified. [t_handle_lock]
610 int 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?
620 * [no locking]
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 int 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;
639 unsigned int t_flushed_data_blocks:1;
642 * For use by the filesystem to store fs-specific data
643 * structures associated with the transaction
645 struct list_head t_private_list;
648 struct transaction_run_stats_s {
649 unsigned long rs_wait;
650 unsigned long rs_running;
651 unsigned long rs_locked;
652 unsigned long rs_flushing;
653 unsigned long rs_logging;
655 __u32 rs_handle_count;
656 __u32 rs_blocks;
657 __u32 rs_blocks_logged;
660 struct transaction_stats_s {
661 unsigned long ts_tid;
662 struct transaction_run_stats_s run;
665 static inline unsigned long
666 jbd2_time_diff(unsigned long start, unsigned long end)
668 if (end >= start)
669 return end - start;
671 return end + (MAX_JIFFY_OFFSET - start);
674 #define JBD2_NR_BATCH 64
677 * struct journal_s - The journal_s type is the concrete type associated with
678 * journal_t.
679 * @j_flags: General journaling state flags
680 * @j_errno: Is there an outstanding uncleared error on the journal (from a
681 * prior abort)?
682 * @j_sb_buffer: First part of superblock buffer
683 * @j_superblock: Second part of superblock buffer
684 * @j_format_version: Version of the superblock format
685 * @j_state_lock: Protect the various scalars in the journal
686 * @j_barrier_count: Number of processes waiting to create a barrier lock
687 * @j_barrier: The barrier lock itself
688 * @j_running_transaction: The current running transaction..
689 * @j_committing_transaction: the transaction we are pushing to disk
690 * @j_checkpoint_transactions: a linked circular list of all transactions
691 * waiting for checkpointing
692 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
693 * to start committing, or for a barrier lock to be released
694 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
695 * @j_wait_done_commit: Wait queue for waiting for commit to complete
696 * @j_wait_checkpoint: Wait queue to trigger checkpointing
697 * @j_wait_commit: Wait queue to trigger commit
698 * @j_wait_updates: Wait queue to wait for updates to complete
699 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
700 * @j_head: Journal head - identifies the first unused block in the journal
701 * @j_tail: Journal tail - identifies the oldest still-used block in the
702 * journal.
703 * @j_free: Journal free - how many free blocks are there in the journal?
704 * @j_first: The block number of the first usable block
705 * @j_last: The block number one beyond the last usable block
706 * @j_dev: Device where we store the journal
707 * @j_blocksize: blocksize for the location where we store the journal.
708 * @j_blk_offset: starting block offset for into the device where we store the
709 * journal
710 * @j_fs_dev: Device which holds the client fs. For internal journal this will
711 * be equal to j_dev
712 * @j_maxlen: Total maximum capacity of the journal region on disk.
713 * @j_list_lock: Protects the buffer lists and internal buffer state.
714 * @j_inode: Optional inode where we store the journal. If present, all journal
715 * block numbers are mapped into this inode via bmap().
716 * @j_tail_sequence: Sequence number of the oldest transaction in the log
717 * @j_transaction_sequence: Sequence number of the next transaction to grant
718 * @j_commit_sequence: Sequence number of the most recently committed
719 * transaction
720 * @j_commit_request: Sequence number of the most recent transaction wanting
721 * commit
722 * @j_uuid: Uuid of client object.
723 * @j_task: Pointer to the current commit thread for this journal
724 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
725 * single compound commit transaction
726 * @j_commit_interval: What is the maximum transaction lifetime before we begin
727 * a commit?
728 * @j_commit_timer: The timer used to wakeup the commit thread
729 * @j_revoke_lock: Protect the revoke table
730 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
731 * current transaction.
732 * @j_revoke_table: alternate revoke tables for j_revoke
733 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
734 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
735 * number that will fit in j_blocksize
736 * @j_last_sync_writer: most recent pid which did a synchronous write
737 * @j_history: Buffer storing the transactions statistics history
738 * @j_history_max: Maximum number of transactions in the statistics history
739 * @j_history_cur: Current number of transactions in the statistics history
740 * @j_history_lock: Protect the transactions statistics history
741 * @j_proc_entry: procfs entry for the jbd statistics directory
742 * @j_stats: Overall statistics
743 * @j_private: An opaque pointer to fs-private information.
746 struct journal_s
748 /* General journaling state flags [j_state_lock] */
749 unsigned long j_flags;
752 * Is there an outstanding uncleared error on the journal (from a prior
753 * abort)? [j_state_lock]
755 int j_errno;
757 /* The superblock buffer */
758 struct buffer_head *j_sb_buffer;
759 journal_superblock_t *j_superblock;
761 /* Version of the superblock format */
762 int j_format_version;
765 * Protect the various scalars in the journal
767 spinlock_t j_state_lock;
770 * Number of processes waiting to create a barrier lock [j_state_lock]
772 int j_barrier_count;
774 /* The barrier lock itself */
775 struct mutex j_barrier;
778 * Transactions: The current running transaction...
779 * [j_state_lock] [caller holding open handle]
781 transaction_t *j_running_transaction;
784 * the transaction we are pushing to disk
785 * [j_state_lock] [caller holding open handle]
787 transaction_t *j_committing_transaction;
790 * ... and a linked circular list of all transactions waiting for
791 * checkpointing. [j_list_lock]
793 transaction_t *j_checkpoint_transactions;
796 * Wait queue for waiting for a locked transaction to start committing,
797 * or for a barrier lock to be released
799 wait_queue_head_t j_wait_transaction_locked;
801 /* Wait queue for waiting for checkpointing to complete */
802 wait_queue_head_t j_wait_logspace;
804 /* Wait queue for waiting for commit to complete */
805 wait_queue_head_t j_wait_done_commit;
807 /* Wait queue to trigger checkpointing */
808 wait_queue_head_t j_wait_checkpoint;
810 /* Wait queue to trigger commit */
811 wait_queue_head_t j_wait_commit;
813 /* Wait queue to wait for updates to complete */
814 wait_queue_head_t j_wait_updates;
816 /* Semaphore for locking against concurrent checkpoints */
817 struct mutex j_checkpoint_mutex;
820 * List of buffer heads used by the checkpoint routine. This
821 * was moved from jbd2_log_do_checkpoint() to reduce stack
822 * usage. Access to this array is controlled by the
823 * j_checkpoint_mutex. [j_checkpoint_mutex]
825 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
828 * Journal head: identifies the first unused block in the journal.
829 * [j_state_lock]
831 unsigned long j_head;
834 * Journal tail: identifies the oldest still-used block in the journal.
835 * [j_state_lock]
837 unsigned long j_tail;
840 * Journal free: how many free blocks are there in the journal?
841 * [j_state_lock]
843 unsigned long j_free;
846 * Journal start and end: the block numbers of the first usable block
847 * and one beyond the last usable block in the journal. [j_state_lock]
849 unsigned long j_first;
850 unsigned long j_last;
853 * Device, blocksize and starting block offset for the location where we
854 * store the journal.
856 struct block_device *j_dev;
857 int j_blocksize;
858 unsigned long long j_blk_offset;
859 char j_devname[BDEVNAME_SIZE+24];
862 * Device which holds the client fs. For internal journal this will be
863 * equal to j_dev.
865 struct block_device *j_fs_dev;
867 /* Total maximum capacity of the journal region on disk. */
868 unsigned int j_maxlen;
871 * Protects the buffer lists and internal buffer state.
873 spinlock_t j_list_lock;
875 /* Optional inode where we store the journal. If present, all */
876 /* journal block numbers are mapped into this inode via */
877 /* bmap(). */
878 struct inode *j_inode;
881 * Sequence number of the oldest transaction in the log [j_state_lock]
883 tid_t j_tail_sequence;
886 * Sequence number of the next transaction to grant [j_state_lock]
888 tid_t j_transaction_sequence;
891 * Sequence number of the most recently committed transaction
892 * [j_state_lock].
894 tid_t j_commit_sequence;
897 * Sequence number of the most recent transaction wanting commit
898 * [j_state_lock]
900 tid_t j_commit_request;
903 * Journal uuid: identifies the object (filesystem, LVM volume etc)
904 * backed by this journal. This will eventually be replaced by an array
905 * of uuids, allowing us to index multiple devices within a single
906 * journal and to perform atomic updates across them.
908 __u8 j_uuid[16];
910 /* Pointer to the current commit thread for this journal */
911 struct task_struct *j_task;
914 * Maximum number of metadata buffers to allow in a single compound
915 * commit transaction
917 int j_max_transaction_buffers;
920 * What is the maximum transaction lifetime before we begin a commit?
922 unsigned long j_commit_interval;
924 /* The timer used to wakeup the commit thread: */
925 struct timer_list j_commit_timer;
928 * The revoke table: maintains the list of revoked blocks in the
929 * current transaction. [j_revoke_lock]
931 spinlock_t j_revoke_lock;
932 struct jbd2_revoke_table_s *j_revoke;
933 struct jbd2_revoke_table_s *j_revoke_table[2];
936 * array of bhs for jbd2_journal_commit_transaction
938 struct buffer_head **j_wbuf;
939 int j_wbufsize;
942 * this is the pid of hte last person to run a synchronous operation
943 * through the journal
945 pid_t j_last_sync_writer;
948 * the average amount of time in nanoseconds it takes to commit a
949 * transaction to disk. [j_state_lock]
951 u64 j_average_commit_time;
954 * minimum and maximum times that we should wait for
955 * additional filesystem operations to get batched into a
956 * synchronous handle in microseconds
958 u32 j_min_batch_time;
959 u32 j_max_batch_time;
961 /* This function is called when a transaction is closed */
962 void (*j_commit_callback)(journal_t *,
963 transaction_t *);
966 * Journal statistics
968 spinlock_t j_history_lock;
969 struct proc_dir_entry *j_proc_entry;
970 struct transaction_stats_s j_stats;
972 /* Failed journal commit ID */
973 unsigned int j_failed_commit;
976 * An opaque pointer to fs-private information. ext3 puts its
977 * superblock pointer here
979 void *j_private;
983 * Journal flag definitions
985 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
986 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
987 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
988 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
989 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
990 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
991 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
992 * data write error in ordered
993 * mode */
996 * Function declarations for the journaling transaction and buffer
997 * management
1000 /* Filing buffers */
1001 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1002 extern void __jbd2_journal_unfile_buffer(struct journal_head *);
1003 extern void __jbd2_journal_refile_buffer(struct journal_head *);
1004 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1005 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1006 extern void __journal_free_buffer(struct journal_head *bh);
1007 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1008 extern void __journal_clean_data_list(transaction_t *transaction);
1010 /* Log buffer allocation */
1011 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
1012 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1014 /* Commit management */
1015 extern void jbd2_journal_commit_transaction(journal_t *);
1017 /* Checkpoint list management */
1018 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1019 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1020 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1024 * Triggers
1027 struct jbd2_buffer_trigger_type {
1029 * Fired just before a buffer is written to the journal.
1030 * mapped_data is a mapped buffer that is the frozen data for
1031 * commit.
1033 void (*t_commit)(struct jbd2_buffer_trigger_type *type,
1034 struct buffer_head *bh, void *mapped_data,
1035 size_t size);
1038 * Fired during journal abort for dirty buffers that will not be
1039 * committed.
1041 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1042 struct buffer_head *bh);
1045 extern void jbd2_buffer_commit_trigger(struct journal_head *jh,
1046 void *mapped_data,
1047 struct jbd2_buffer_trigger_type *triggers);
1048 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1049 struct jbd2_buffer_trigger_type *triggers);
1051 /* Buffer IO */
1052 extern int
1053 jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1054 struct journal_head *jh_in,
1055 struct journal_head **jh_out,
1056 unsigned long long blocknr);
1058 /* Transaction locking */
1059 extern void __wait_on_journal (journal_t *);
1062 * Journal locking.
1064 * We need to lock the journal during transaction state changes so that nobody
1065 * ever tries to take a handle on the running transaction while we are in the
1066 * middle of moving it to the commit phase. j_state_lock does this.
1068 * Note that the locking is completely interrupt unsafe. We never touch
1069 * journal structures from interrupts.
1072 static inline handle_t *journal_current_handle(void)
1074 return current->journal_info;
1077 /* The journaling code user interface:
1079 * Create and destroy handles
1080 * Register buffer modifications against the current transaction.
1083 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1084 extern int jbd2_journal_restart (handle_t *, int nblocks);
1085 extern int jbd2_journal_extend (handle_t *, int nblocks);
1086 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1087 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1088 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1089 void jbd2_journal_set_triggers(struct buffer_head *,
1090 struct jbd2_buffer_trigger_type *type);
1091 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1092 extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1093 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1094 extern void journal_sync_buffer (struct buffer_head *);
1095 extern void jbd2_journal_invalidatepage(journal_t *,
1096 struct page *, unsigned long);
1097 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1098 extern int jbd2_journal_stop(handle_t *);
1099 extern int jbd2_journal_flush (journal_t *);
1100 extern void jbd2_journal_lock_updates (journal_t *);
1101 extern void jbd2_journal_unlock_updates (journal_t *);
1103 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1104 struct block_device *fs_dev,
1105 unsigned long long start, int len, int bsize);
1106 extern journal_t * jbd2_journal_init_inode (struct inode *);
1107 extern int jbd2_journal_update_format (journal_t *);
1108 extern int jbd2_journal_check_used_features
1109 (journal_t *, unsigned long, unsigned long, unsigned long);
1110 extern int jbd2_journal_check_available_features
1111 (journal_t *, unsigned long, unsigned long, unsigned long);
1112 extern int jbd2_journal_set_features
1113 (journal_t *, unsigned long, unsigned long, unsigned long);
1114 extern void jbd2_journal_clear_features
1115 (journal_t *, unsigned long, unsigned long, unsigned long);
1116 extern int jbd2_journal_load (journal_t *journal);
1117 extern int jbd2_journal_destroy (journal_t *);
1118 extern int jbd2_journal_recover (journal_t *journal);
1119 extern int jbd2_journal_wipe (journal_t *, int);
1120 extern int jbd2_journal_skip_recovery (journal_t *);
1121 extern void jbd2_journal_update_superblock (journal_t *, int);
1122 extern void __jbd2_journal_abort_hard (journal_t *);
1123 extern void jbd2_journal_abort (journal_t *, int);
1124 extern int jbd2_journal_errno (journal_t *);
1125 extern void jbd2_journal_ack_err (journal_t *);
1126 extern int jbd2_journal_clear_err (journal_t *);
1127 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1128 extern int jbd2_journal_force_commit(journal_t *);
1129 extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1130 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1131 struct jbd2_inode *inode, loff_t new_size);
1132 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1133 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1136 * journal_head management
1138 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1139 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1140 void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1141 void jbd2_journal_put_journal_head(struct journal_head *jh);
1144 * handle management
1146 extern struct kmem_cache *jbd2_handle_cache;
1148 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1150 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
1153 static inline void jbd2_free_handle(handle_t *handle)
1155 kmem_cache_free(jbd2_handle_cache, handle);
1158 /* Primary revoke support */
1159 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1160 extern int jbd2_journal_init_revoke(journal_t *, int);
1161 extern void jbd2_journal_destroy_revoke_caches(void);
1162 extern int jbd2_journal_init_revoke_caches(void);
1164 extern void jbd2_journal_destroy_revoke(journal_t *);
1165 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1166 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1167 extern void jbd2_journal_write_revoke_records(journal_t *,
1168 transaction_t *, int);
1170 /* Recovery revoke support */
1171 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1172 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1173 extern void jbd2_journal_clear_revoke(journal_t *);
1174 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1177 * The log thread user interface:
1179 * Request space in the current transaction, and force transaction commit
1180 * transitions on demand.
1183 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1184 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1185 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1186 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1187 int jbd2_journal_force_commit_nested(journal_t *journal);
1188 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1189 int jbd2_log_do_checkpoint(journal_t *journal);
1191 void __jbd2_log_wait_for_space(journal_t *journal);
1192 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1193 extern int jbd2_cleanup_journal_tail(journal_t *);
1195 /* Debugging code only: */
1197 #define jbd_ENOSYS() \
1198 do { \
1199 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1200 current->state = TASK_UNINTERRUPTIBLE; \
1201 schedule(); \
1202 } while (1)
1205 * is_journal_abort
1207 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1208 * bit, when set, indicates that we have had a fatal error somewhere,
1209 * either inside the journaling layer or indicated to us by the client
1210 * (eg. ext3), and that we and should not commit any further
1211 * transactions.
1214 static inline int is_journal_aborted(journal_t *journal)
1216 return journal->j_flags & JBD2_ABORT;
1219 static inline int is_handle_aborted(handle_t *handle)
1221 if (handle->h_aborted)
1222 return 1;
1223 return is_journal_aborted(handle->h_transaction->t_journal);
1226 static inline void jbd2_journal_abort_handle(handle_t *handle)
1228 handle->h_aborted = 1;
1231 #endif /* __KERNEL__ */
1233 /* Comparison functions for transaction IDs: perform comparisons using
1234 * modulo arithmetic so that they work over sequence number wraps. */
1236 static inline int tid_gt(tid_t x, tid_t y)
1238 int difference = (x - y);
1239 return (difference > 0);
1242 static inline int tid_geq(tid_t x, tid_t y)
1244 int difference = (x - y);
1245 return (difference >= 0);
1248 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1249 extern size_t journal_tag_bytes(journal_t *journal);
1252 * Return the minimum number of blocks which must be free in the journal
1253 * before a new transaction may be started. Must be called under j_state_lock.
1255 static inline int jbd_space_needed(journal_t *journal)
1257 int nblocks = journal->j_max_transaction_buffers;
1258 if (journal->j_committing_transaction)
1259 nblocks += journal->j_committing_transaction->
1260 t_outstanding_credits;
1261 return nblocks;
1265 * Definitions which augment the buffer_head layer
1268 /* journaling buffer types */
1269 #define BJ_None 0 /* Not journaled */
1270 #define BJ_Metadata 1 /* Normal journaled metadata */
1271 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1272 #define BJ_IO 3 /* Buffer is for temporary IO use */
1273 #define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1274 #define BJ_LogCtl 5 /* Buffer contains log descriptors */
1275 #define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1276 #define BJ_Types 7
1278 extern int jbd_blocks_per_page(struct inode *inode);
1280 #ifdef __KERNEL__
1282 #define buffer_trace_init(bh) do {} while (0)
1283 #define print_buffer_fields(bh) do {} while (0)
1284 #define print_buffer_trace(bh) do {} while (0)
1285 #define BUFFER_TRACE(bh, info) do {} while (0)
1286 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1287 #define JBUFFER_TRACE(jh, info) do {} while (0)
1290 * jbd2_dev_to_name is a utility function used by the jbd2 and ext4
1291 * tracing infrastructure to map a dev_t to a device name.
1293 extern const char *jbd2_dev_to_name(dev_t device);
1295 #endif /* __KERNEL__ */
1297 #endif /* _LINUX_JBD2_H */