mfd: Add a data argument to the WM8350 IRQ free function
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / include / linux / jbd2.h
blob1ec876358180ecd72ec00f06de81af9d6e4bba71
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 #endif
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 u8 jbd2_journal_enable_debug;
60 #define jbd_debug(n, f, a...) \
61 do { \
62 if ((n) <= jbd2_journal_enable_debug) { \
63 printk (KERN_DEBUG "(%s, %d): %s: ", \
64 __FILE__, __LINE__, __func__); \
65 printk (f, ## a); \
66 } \
67 } while (0)
68 #else
69 #define jbd_debug(f, a...) /**/
70 #endif
72 extern void *jbd2_alloc(size_t size, gfp_t flags);
73 extern void jbd2_free(void *ptr, size_t size);
75 #define JBD2_MIN_JOURNAL_BLOCKS 1024
77 #ifdef __KERNEL__
79 /**
80 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
82 * All filesystem modifications made by the process go
83 * through this handle. Recursive operations (such as quota operations)
84 * are gathered into a single update.
86 * The buffer credits field is used to account for journaled buffers
87 * being modified by the running process. To ensure that there is
88 * enough log space for all outstanding operations, we need to limit the
89 * number of outstanding buffers possible at any time. When the
90 * operation completes, any buffer credits not used are credited back to
91 * the transaction, so that at all times we know how many buffers the
92 * outstanding updates on a transaction might possibly touch.
94 * This is an opaque datatype.
95 **/
96 typedef struct handle_s handle_t; /* Atomic operation type */
99 /**
100 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
102 * journal_t is linked to from the fs superblock structure.
104 * We use the journal_t to keep track of all outstanding transaction
105 * activity on the filesystem, and to manage the state of the log
106 * writing process.
108 * This is an opaque datatype.
110 typedef struct journal_s journal_t; /* Journal control structure */
111 #endif
114 * Internal structures used by the logging mechanism:
117 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
120 * On-disk structures
124 * Descriptor block types:
127 #define JBD2_DESCRIPTOR_BLOCK 1
128 #define JBD2_COMMIT_BLOCK 2
129 #define JBD2_SUPERBLOCK_V1 3
130 #define JBD2_SUPERBLOCK_V2 4
131 #define JBD2_REVOKE_BLOCK 5
134 * Standard header for all descriptor blocks:
136 typedef struct journal_header_s
138 __be32 h_magic;
139 __be32 h_blocktype;
140 __be32 h_sequence;
141 } journal_header_t;
144 * Checksum types.
146 #define JBD2_CRC32_CHKSUM 1
147 #define JBD2_MD5_CHKSUM 2
148 #define JBD2_SHA1_CHKSUM 3
150 #define JBD2_CRC32_CHKSUM_SIZE 4
152 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
154 * Commit block header for storing transactional checksums:
156 struct commit_header {
157 __be32 h_magic;
158 __be32 h_blocktype;
159 __be32 h_sequence;
160 unsigned char h_chksum_type;
161 unsigned char h_chksum_size;
162 unsigned char h_padding[2];
163 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
164 __be64 h_commit_sec;
165 __be32 h_commit_nsec;
169 * The block tag: used to describe a single buffer in the journal.
170 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
171 * raw struct shouldn't be used for pointer math or sizeof() - use
172 * journal_tag_bytes(journal) instead to compute this.
174 typedef struct journal_block_tag_s
176 __be32 t_blocknr; /* The on-disk block number */
177 __be32 t_flags; /* See below */
178 __be32 t_blocknr_high; /* most-significant high 32bits. */
179 } journal_block_tag_t;
181 #define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
182 #define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
185 * The revoke descriptor: used on disk to describe a series of blocks to
186 * be revoked from the log
188 typedef struct jbd2_journal_revoke_header_s
190 journal_header_t r_header;
191 __be32 r_count; /* Count of bytes used in the block */
192 } jbd2_journal_revoke_header_t;
195 /* Definitions for the journal tag flags word: */
196 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
197 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
198 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
199 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
203 * The journal superblock. All fields are in big-endian byte order.
205 typedef struct journal_superblock_s
207 /* 0x0000 */
208 journal_header_t s_header;
210 /* 0x000C */
211 /* Static information describing the journal */
212 __be32 s_blocksize; /* journal device blocksize */
213 __be32 s_maxlen; /* total blocks in journal file */
214 __be32 s_first; /* first block of log information */
216 /* 0x0018 */
217 /* Dynamic information describing the current state of the log */
218 __be32 s_sequence; /* first commit ID expected in log */
219 __be32 s_start; /* blocknr of start of log */
221 /* 0x0020 */
222 /* Error value, as set by jbd2_journal_abort(). */
223 __be32 s_errno;
225 /* 0x0024 */
226 /* Remaining fields are only valid in a version-2 superblock */
227 __be32 s_feature_compat; /* compatible feature set */
228 __be32 s_feature_incompat; /* incompatible feature set */
229 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
230 /* 0x0030 */
231 __u8 s_uuid[16]; /* 128-bit uuid for journal */
233 /* 0x0040 */
234 __be32 s_nr_users; /* Nr of filesystems sharing log */
236 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
238 /* 0x0048 */
239 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
240 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
242 /* 0x0050 */
243 __u32 s_padding[44];
245 /* 0x0100 */
246 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
247 /* 0x0400 */
248 } journal_superblock_t;
250 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
251 ((j)->j_format_version >= 2 && \
252 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
253 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
254 ((j)->j_format_version >= 2 && \
255 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
256 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
257 ((j)->j_format_version >= 2 && \
258 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
260 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
262 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
263 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
264 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
266 /* Features known to this kernel version: */
267 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
268 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
269 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
270 JBD2_FEATURE_INCOMPAT_64BIT | \
271 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)
273 #ifdef __KERNEL__
275 #include <linux/fs.h>
276 #include <linux/sched.h>
278 #define J_ASSERT(assert) BUG_ON(!(assert))
280 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
281 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
283 #if defined(JBD2_PARANOID_IOFAIL)
284 #define J_EXPECT(expr, why...) J_ASSERT(expr)
285 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
286 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
287 #else
288 #define __journal_expect(expr, why...) \
289 ({ \
290 int val = (expr); \
291 if (!val) { \
292 printk(KERN_ERR \
293 "JBD2 unexpected failure: %s: %s;\n", \
294 __func__, #expr); \
295 printk(KERN_ERR why "\n"); \
297 val; \
299 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
300 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
301 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
302 #endif
304 enum jbd_state_bits {
305 BH_JBD /* Has an attached ext3 journal_head */
306 = BH_PrivateStart,
307 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
308 BH_Freed, /* Has been freed (truncated) */
309 BH_Revoked, /* Has been revoked from the log */
310 BH_RevokeValid, /* Revoked flag is valid */
311 BH_JBDDirty, /* Is dirty but journaled */
312 BH_State, /* Pins most journal_head state */
313 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
314 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
315 BH_JBDPrivateStart, /* First bit available for private use by FS */
318 BUFFER_FNS(JBD, jbd)
319 BUFFER_FNS(JWrite, jwrite)
320 BUFFER_FNS(JBDDirty, jbddirty)
321 TAS_BUFFER_FNS(JBDDirty, jbddirty)
322 BUFFER_FNS(Revoked, revoked)
323 TAS_BUFFER_FNS(Revoked, revoked)
324 BUFFER_FNS(RevokeValid, revokevalid)
325 TAS_BUFFER_FNS(RevokeValid, revokevalid)
326 BUFFER_FNS(Freed, freed)
328 static inline struct buffer_head *jh2bh(struct journal_head *jh)
330 return jh->b_bh;
333 static inline struct journal_head *bh2jh(struct buffer_head *bh)
335 return bh->b_private;
338 static inline void jbd_lock_bh_state(struct buffer_head *bh)
340 bit_spin_lock(BH_State, &bh->b_state);
343 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
345 return bit_spin_trylock(BH_State, &bh->b_state);
348 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
350 return bit_spin_is_locked(BH_State, &bh->b_state);
353 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
355 bit_spin_unlock(BH_State, &bh->b_state);
358 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
360 bit_spin_lock(BH_JournalHead, &bh->b_state);
363 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
365 bit_spin_unlock(BH_JournalHead, &bh->b_state);
368 /* Flags in jbd_inode->i_flags */
369 #define __JI_COMMIT_RUNNING 0
370 /* Commit of the inode data in progress. We use this flag to protect us from
371 * concurrent deletion of inode. We cannot use reference to inode for this
372 * since we cannot afford doing last iput() on behalf of kjournald
374 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
377 * struct jbd_inode is the structure linking inodes in ordered mode
378 * present in a transaction so that we can sync them during commit.
380 struct jbd2_inode {
381 /* Which transaction does this inode belong to? Either the running
382 * transaction or the committing one. [j_list_lock] */
383 transaction_t *i_transaction;
385 /* Pointer to the running transaction modifying inode's data in case
386 * there is already a committing transaction touching it. [j_list_lock] */
387 transaction_t *i_next_transaction;
389 /* List of inodes in the i_transaction [j_list_lock] */
390 struct list_head i_list;
392 /* VFS inode this inode belongs to [constant during the lifetime
393 * of the structure] */
394 struct inode *i_vfs_inode;
396 /* Flags of inode [j_list_lock] */
397 unsigned int i_flags;
400 struct jbd2_revoke_table_s;
403 * struct handle_s - The handle_s type is the concrete type associated with
404 * handle_t.
405 * @h_transaction: Which compound transaction is this update a part of?
406 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
407 * @h_ref: Reference count on this handle
408 * @h_err: Field for caller's use to track errors through large fs operations
409 * @h_sync: flag for sync-on-close
410 * @h_jdata: flag to force data journaling
411 * @h_aborted: flag indicating fatal error on handle
414 /* Docbook can't yet cope with the bit fields, but will leave the documentation
415 * in so it can be fixed later.
418 struct handle_s
420 /* Which compound transaction is this update a part of? */
421 transaction_t *h_transaction;
423 /* Number of remaining buffers we are allowed to dirty: */
424 int h_buffer_credits;
426 /* Reference count on this handle */
427 int h_ref;
429 /* Field for caller's use to track errors through large fs */
430 /* operations */
431 int h_err;
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_aborted: 1; /* fatal error on handle */
438 #ifdef CONFIG_DEBUG_LOCK_ALLOC
439 struct lockdep_map h_lockdep_map;
440 #endif
445 * Some stats for checkpoint phase
447 struct transaction_chp_stats_s {
448 unsigned long cs_chp_time;
449 __u32 cs_forced_to_close;
450 __u32 cs_written;
451 __u32 cs_dropped;
454 /* The transaction_t type is the guts of the journaling mechanism. It
455 * tracks a compound transaction through its various states:
457 * RUNNING: accepting new updates
458 * LOCKED: Updates still running but we don't accept new ones
459 * RUNDOWN: Updates are tidying up but have finished requesting
460 * new buffers to modify (state not used for now)
461 * FLUSH: All updates complete, but we are still writing to disk
462 * COMMIT: All data on disk, writing commit record
463 * FINISHED: We still have to keep the transaction for checkpointing.
465 * The transaction keeps track of all of the buffers modified by a
466 * running transaction, and all of the buffers committed but not yet
467 * flushed to home for finished transactions.
471 * Lock ranking:
473 * j_list_lock
474 * ->jbd_lock_bh_journal_head() (This is "innermost")
476 * j_state_lock
477 * ->jbd_lock_bh_state()
479 * jbd_lock_bh_state()
480 * ->j_list_lock
482 * j_state_lock
483 * ->t_handle_lock
485 * j_state_lock
486 * ->j_list_lock (journal_unmap_buffer)
490 struct transaction_s
492 /* Pointer to the journal for this transaction. [no locking] */
493 journal_t *t_journal;
495 /* Sequence number for this transaction [no locking] */
496 tid_t t_tid;
499 * Transaction's current state
500 * [no locking - only kjournald2 alters this]
501 * [j_list_lock] guards transition of a transaction into T_FINISHED
502 * state and subsequent call of __jbd2_journal_drop_transaction()
503 * FIXME: needs barriers
504 * KLUDGE: [use j_state_lock]
506 enum {
507 T_RUNNING,
508 T_LOCKED,
509 T_RUNDOWN,
510 T_FLUSH,
511 T_COMMIT,
512 T_FINISHED
513 } t_state;
516 * Where in the log does this transaction's commit start? [no locking]
518 unsigned long t_log_start;
520 /* Number of buffers on the t_buffers list [j_list_lock] */
521 int t_nr_buffers;
524 * Doubly-linked circular list of all buffers reserved but not yet
525 * modified by this transaction [j_list_lock]
527 struct journal_head *t_reserved_list;
530 * Doubly-linked circular list of all metadata buffers owned by this
531 * transaction [j_list_lock]
533 struct journal_head *t_buffers;
536 * Doubly-linked circular list of all forget buffers (superseded
537 * buffers which we can un-checkpoint once this transaction commits)
538 * [j_list_lock]
540 struct journal_head *t_forget;
543 * Doubly-linked circular list of all buffers still to be flushed before
544 * this transaction can be checkpointed. [j_list_lock]
546 struct journal_head *t_checkpoint_list;
549 * Doubly-linked circular list of all buffers submitted for IO while
550 * checkpointing. [j_list_lock]
552 struct journal_head *t_checkpoint_io_list;
555 * Doubly-linked circular list of temporary buffers currently undergoing
556 * IO in the log [j_list_lock]
558 struct journal_head *t_iobuf_list;
561 * Doubly-linked circular list of metadata buffers being shadowed by log
562 * IO. The IO buffers on the iobuf list and the shadow buffers on this
563 * list match each other one for one at all times. [j_list_lock]
565 struct journal_head *t_shadow_list;
568 * Doubly-linked circular list of control buffers being written to the
569 * log. [j_list_lock]
571 struct journal_head *t_log_list;
574 * List of inodes whose data we've modified in data=ordered mode.
575 * [j_list_lock]
577 struct list_head t_inode_list;
580 * Protects info related to handles
582 spinlock_t t_handle_lock;
585 * Longest time some handle had to wait for running transaction
587 unsigned long t_max_wait;
590 * When transaction started
592 unsigned long t_start;
595 * Checkpointing stats [j_checkpoint_sem]
597 struct transaction_chp_stats_s t_chp_stats;
600 * Number of outstanding updates running on this transaction
601 * [t_handle_lock]
603 int t_updates;
606 * Number of buffers reserved for use by all handles in this transaction
607 * handle but not yet modified. [t_handle_lock]
609 int t_outstanding_credits;
612 * Forward and backward links for the circular list of all transactions
613 * awaiting checkpoint. [j_list_lock]
615 transaction_t *t_cpnext, *t_cpprev;
618 * When will the transaction expire (become due for commit), in jiffies?
619 * [no locking]
621 unsigned long t_expires;
624 * When this transaction started, in nanoseconds [no locking]
626 ktime_t t_start_time;
629 * How many handles used this transaction? [t_handle_lock]
631 int t_handle_count;
634 * This transaction is being forced and some process is
635 * waiting for it to finish.
637 unsigned int t_synchronous_commit:1;
638 unsigned int t_flushed_data_blocks:1;
641 * For use by the filesystem to store fs-specific data
642 * structures associated with the transaction
644 struct list_head t_private_list;
647 struct transaction_run_stats_s {
648 unsigned long rs_wait;
649 unsigned long rs_running;
650 unsigned long rs_locked;
651 unsigned long rs_flushing;
652 unsigned long rs_logging;
654 __u32 rs_handle_count;
655 __u32 rs_blocks;
656 __u32 rs_blocks_logged;
659 struct transaction_stats_s {
660 unsigned long ts_tid;
661 struct transaction_run_stats_s run;
664 static inline unsigned long
665 jbd2_time_diff(unsigned long start, unsigned long end)
667 if (end >= start)
668 return end - start;
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
677 * journal_t.
678 * @j_flags: General journaling state flags
679 * @j_errno: Is there an outstanding uncleared error on the journal (from a
680 * prior abort)?
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_logspace: Wait queue for waiting for checkpointing to complete
694 * @j_wait_done_commit: Wait queue for waiting for commit to complete
695 * @j_wait_checkpoint: Wait queue to trigger checkpointing
696 * @j_wait_commit: Wait queue to trigger commit
697 * @j_wait_updates: Wait queue to wait for updates to complete
698 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
699 * @j_head: Journal head - identifies the first unused block in the journal
700 * @j_tail: Journal tail - identifies the oldest still-used block in the
701 * journal.
702 * @j_free: Journal free - how many free blocks are there in the journal?
703 * @j_first: The block number of the first usable block
704 * @j_last: The block number one beyond the last usable block
705 * @j_dev: Device where we store the journal
706 * @j_blocksize: blocksize for the location where we store the journal.
707 * @j_blk_offset: starting block offset for into the device where we store the
708 * journal
709 * @j_fs_dev: Device which holds the client fs. For internal journal this will
710 * be equal to j_dev
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
718 * transaction
719 * @j_commit_request: Sequence number of the most recent transaction wanting
720 * commit
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
726 * a commit?
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.
745 struct journal_s
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]
754 int j_errno;
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 spinlock_t j_state_lock;
769 * Number of processes waiting to create a barrier lock [j_state_lock]
771 int j_barrier_count;
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 checkpointing to complete */
801 wait_queue_head_t j_wait_logspace;
803 /* Wait queue for waiting for commit to complete */
804 wait_queue_head_t j_wait_done_commit;
806 /* Wait queue to trigger checkpointing */
807 wait_queue_head_t j_wait_checkpoint;
809 /* Wait queue to trigger commit */
810 wait_queue_head_t j_wait_commit;
812 /* Wait queue to wait for updates to complete */
813 wait_queue_head_t j_wait_updates;
815 /* Semaphore for locking against concurrent checkpoints */
816 struct mutex j_checkpoint_mutex;
819 * List of buffer heads used by the checkpoint routine. This
820 * was moved from jbd2_log_do_checkpoint() to reduce stack
821 * usage. Access to this array is controlled by the
822 * j_checkpoint_mutex. [j_checkpoint_mutex]
824 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
827 * Journal head: identifies the first unused block in the journal.
828 * [j_state_lock]
830 unsigned long j_head;
833 * Journal tail: identifies the oldest still-used block in the journal.
834 * [j_state_lock]
836 unsigned long j_tail;
839 * Journal free: how many free blocks are there in the journal?
840 * [j_state_lock]
842 unsigned long j_free;
845 * Journal start and end: the block numbers of the first usable block
846 * and one beyond the last usable block in the journal. [j_state_lock]
848 unsigned long j_first;
849 unsigned long j_last;
852 * Device, blocksize and starting block offset for the location where we
853 * store the journal.
855 struct block_device *j_dev;
856 int j_blocksize;
857 unsigned long long j_blk_offset;
858 char j_devname[BDEVNAME_SIZE+24];
861 * Device which holds the client fs. For internal journal this will be
862 * equal to j_dev.
864 struct block_device *j_fs_dev;
866 /* Total maximum capacity of the journal region on disk. */
867 unsigned int j_maxlen;
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 */
876 /* bmap(). */
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
891 * [j_state_lock].
893 tid_t j_commit_sequence;
896 * Sequence number of the most recent transaction wanting commit
897 * [j_state_lock]
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.
907 __u8 j_uuid[16];
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
914 * commit transaction
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;
938 int j_wbufsize;
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 *,
962 transaction_t *);
965 * Journal statistics
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
978 void *j_private;
982 * Journal flag definitions
984 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
985 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
986 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
987 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
988 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
989 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
990 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
991 * data write error in ordered
992 * mode */
995 * Function declarations for the journaling transaction and buffer
996 * management
999 /* Filing buffers */
1000 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1001 extern void __jbd2_journal_unfile_buffer(struct journal_head *);
1002 extern void __jbd2_journal_refile_buffer(struct journal_head *);
1003 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1004 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1005 extern void __journal_free_buffer(struct journal_head *bh);
1006 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1007 extern void __journal_clean_data_list(transaction_t *transaction);
1009 /* Log buffer allocation */
1010 extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
1011 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1013 /* Commit management */
1014 extern void jbd2_journal_commit_transaction(journal_t *);
1016 /* Checkpoint list management */
1017 int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1018 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1019 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1023 * Triggers
1026 struct jbd2_buffer_trigger_type {
1028 * Fired just before a buffer is written to the journal.
1029 * mapped_data is a mapped buffer that is the frozen data for
1030 * commit.
1032 void (*t_commit)(struct jbd2_buffer_trigger_type *type,
1033 struct buffer_head *bh, void *mapped_data,
1034 size_t size);
1037 * Fired during journal abort for dirty buffers that will not be
1038 * committed.
1040 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1041 struct buffer_head *bh);
1044 extern void jbd2_buffer_commit_trigger(struct journal_head *jh,
1045 void *mapped_data,
1046 struct jbd2_buffer_trigger_type *triggers);
1047 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1048 struct jbd2_buffer_trigger_type *triggers);
1050 /* Buffer IO */
1051 extern int
1052 jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1053 struct journal_head *jh_in,
1054 struct journal_head **jh_out,
1055 unsigned long long blocknr);
1057 /* Transaction locking */
1058 extern void __wait_on_journal (journal_t *);
1061 * Journal locking.
1063 * We need to lock the journal during transaction state changes so that nobody
1064 * ever tries to take a handle on the running transaction while we are in the
1065 * middle of moving it to the commit phase. j_state_lock does this.
1067 * Note that the locking is completely interrupt unsafe. We never touch
1068 * journal structures from interrupts.
1071 static inline handle_t *journal_current_handle(void)
1073 return current->journal_info;
1076 /* The journaling code user interface:
1078 * Create and destroy handles
1079 * Register buffer modifications against the current transaction.
1082 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1083 extern int jbd2_journal_restart (handle_t *, int nblocks);
1084 extern int jbd2_journal_extend (handle_t *, int nblocks);
1085 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1086 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1087 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1088 void jbd2_journal_set_triggers(struct buffer_head *,
1089 struct jbd2_buffer_trigger_type *type);
1090 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1091 extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1092 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1093 extern void journal_sync_buffer (struct buffer_head *);
1094 extern void jbd2_journal_invalidatepage(journal_t *,
1095 struct page *, unsigned long);
1096 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1097 extern int jbd2_journal_stop(handle_t *);
1098 extern int jbd2_journal_flush (journal_t *);
1099 extern void jbd2_journal_lock_updates (journal_t *);
1100 extern void jbd2_journal_unlock_updates (journal_t *);
1102 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1103 struct block_device *fs_dev,
1104 unsigned long long start, int len, int bsize);
1105 extern journal_t * jbd2_journal_init_inode (struct inode *);
1106 extern int jbd2_journal_update_format (journal_t *);
1107 extern int jbd2_journal_check_used_features
1108 (journal_t *, unsigned long, unsigned long, unsigned long);
1109 extern int jbd2_journal_check_available_features
1110 (journal_t *, unsigned long, unsigned long, unsigned long);
1111 extern int jbd2_journal_set_features
1112 (journal_t *, unsigned long, unsigned long, unsigned long);
1113 extern void jbd2_journal_clear_features
1114 (journal_t *, unsigned long, unsigned long, unsigned long);
1115 extern int jbd2_journal_load (journal_t *journal);
1116 extern int jbd2_journal_destroy (journal_t *);
1117 extern int jbd2_journal_recover (journal_t *journal);
1118 extern int jbd2_journal_wipe (journal_t *, int);
1119 extern int jbd2_journal_skip_recovery (journal_t *);
1120 extern void jbd2_journal_update_superblock (journal_t *, int);
1121 extern void __jbd2_journal_abort_hard (journal_t *);
1122 extern void jbd2_journal_abort (journal_t *, int);
1123 extern int jbd2_journal_errno (journal_t *);
1124 extern void jbd2_journal_ack_err (journal_t *);
1125 extern int jbd2_journal_clear_err (journal_t *);
1126 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1127 extern int jbd2_journal_force_commit(journal_t *);
1128 extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1129 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1130 struct jbd2_inode *inode, loff_t new_size);
1131 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1132 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1135 * journal_head management
1137 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1138 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1139 void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1140 void jbd2_journal_put_journal_head(struct journal_head *jh);
1143 * handle management
1145 extern struct kmem_cache *jbd2_handle_cache;
1147 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1149 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
1152 static inline void jbd2_free_handle(handle_t *handle)
1154 kmem_cache_free(jbd2_handle_cache, handle);
1157 /* Primary revoke support */
1158 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1159 extern int jbd2_journal_init_revoke(journal_t *, int);
1160 extern void jbd2_journal_destroy_revoke_caches(void);
1161 extern int jbd2_journal_init_revoke_caches(void);
1163 extern void jbd2_journal_destroy_revoke(journal_t *);
1164 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1165 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1166 extern void jbd2_journal_write_revoke_records(journal_t *,
1167 transaction_t *, int);
1169 /* Recovery revoke support */
1170 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1171 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1172 extern void jbd2_journal_clear_revoke(journal_t *);
1173 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1176 * The log thread user interface:
1178 * Request space in the current transaction, and force transaction commit
1179 * transitions on demand.
1182 int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1183 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1184 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1185 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1186 int jbd2_journal_force_commit_nested(journal_t *journal);
1187 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1188 int jbd2_log_do_checkpoint(journal_t *journal);
1190 void __jbd2_log_wait_for_space(journal_t *journal);
1191 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1192 extern int jbd2_cleanup_journal_tail(journal_t *);
1194 /* Debugging code only: */
1196 #define jbd_ENOSYS() \
1197 do { \
1198 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
1199 current->state = TASK_UNINTERRUPTIBLE; \
1200 schedule(); \
1201 } while (1)
1204 * is_journal_abort
1206 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1207 * bit, when set, indicates that we have had a fatal error somewhere,
1208 * either inside the journaling layer or indicated to us by the client
1209 * (eg. ext3), and that we and should not commit any further
1210 * transactions.
1213 static inline int is_journal_aborted(journal_t *journal)
1215 return journal->j_flags & JBD2_ABORT;
1218 static inline int is_handle_aborted(handle_t *handle)
1220 if (handle->h_aborted)
1221 return 1;
1222 return is_journal_aborted(handle->h_transaction->t_journal);
1225 static inline void jbd2_journal_abort_handle(handle_t *handle)
1227 handle->h_aborted = 1;
1230 #endif /* __KERNEL__ */
1232 /* Comparison functions for transaction IDs: perform comparisons using
1233 * modulo arithmetic so that they work over sequence number wraps. */
1235 static inline int tid_gt(tid_t x, tid_t y)
1237 int difference = (x - y);
1238 return (difference > 0);
1241 static inline int tid_geq(tid_t x, tid_t y)
1243 int difference = (x - y);
1244 return (difference >= 0);
1247 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1248 extern size_t journal_tag_bytes(journal_t *journal);
1251 * Return the minimum number of blocks which must be free in the journal
1252 * before a new transaction may be started. Must be called under j_state_lock.
1254 static inline int jbd_space_needed(journal_t *journal)
1256 int nblocks = journal->j_max_transaction_buffers;
1257 if (journal->j_committing_transaction)
1258 nblocks += journal->j_committing_transaction->
1259 t_outstanding_credits;
1260 return nblocks;
1264 * Definitions which augment the buffer_head layer
1267 /* journaling buffer types */
1268 #define BJ_None 0 /* Not journaled */
1269 #define BJ_Metadata 1 /* Normal journaled metadata */
1270 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1271 #define BJ_IO 3 /* Buffer is for temporary IO use */
1272 #define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1273 #define BJ_LogCtl 5 /* Buffer contains log descriptors */
1274 #define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1275 #define BJ_Types 7
1277 extern int jbd_blocks_per_page(struct inode *inode);
1279 #ifdef __KERNEL__
1281 #define buffer_trace_init(bh) do {} while (0)
1282 #define print_buffer_fields(bh) do {} while (0)
1283 #define print_buffer_trace(bh) do {} while (0)
1284 #define BUFFER_TRACE(bh, info) do {} while (0)
1285 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1286 #define JBUFFER_TRACE(jh, info) do {} while (0)
1289 * jbd2_dev_to_name is a utility function used by the jbd2 and ext4
1290 * tracing infrastructure to map a dev_t to a device name.
1292 extern const char *jbd2_dev_to_name(dev_t device);
1294 #endif /* __KERNEL__ */
1296 #endif /* _LINUX_JBD2_H */