blob: 41ee79962bb26c7dfc26dd7027de7abf59de3736 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/include/linux/jbd.h
3 *
4 * Written by Stephen C. Tweedie <sct@redhat.com>
5 *
6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
7 *
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.
11 *
12 * Definitions for transaction data structures for the buffer cache
13 * filesystem journaling support.
14 */
15
16#ifndef _LINUX_JBD_H
17#define _LINUX_JBD_H
18
Linus Torvalds1da177e2005-04-16 15:20:36 -070019/* Allow this file to be included directly into e2fsprogs */
20#ifndef __KERNEL__
21#include "jfs_compat.h"
22#define JFS_DEBUG
23#define jfs_debug jbd_debug
24#else
25
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -080026#include <linux/types.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/buffer_head.h>
28#include <linux/journal-head.h>
29#include <linux/stddef.h>
Ingo Molnarfb1c8f92005-09-10 00:25:56 -070030#include <linux/bit_spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <asm/semaphore.h>
32#endif
33
34#define journal_oom_retry 1
35
36/*
37 * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
38 * certain classes of error which can occur due to failed IOs. Under
39 * normal use we want ext3 to continue after such errors, because
40 * hardware _can_ fail, but for debugging purposes when running tests on
41 * known-good hardware we may want to trap these errors.
42 */
43#undef JBD_PARANOID_IOFAIL
44
45/*
46 * The default maximum commit age, in seconds.
47 */
48#define JBD_DEFAULT_MAX_COMMIT_AGE 5
49
50#ifdef CONFIG_JBD_DEBUG
51/*
52 * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
53 * consistency checks. By default we don't do this unless
54 * CONFIG_JBD_DEBUG is on.
55 */
56#define JBD_EXPENSIVE_CHECKING
57extern int journal_enable_debug;
58
59#define jbd_debug(n, f, a...) \
60 do { \
61 if ((n) <= journal_enable_debug) { \
62 printk (KERN_DEBUG "(%s, %d): %s: ", \
63 __FILE__, __LINE__, __FUNCTION__); \
64 printk (f, ## a); \
65 } \
66 } while (0)
67#else
68#define jbd_debug(f, a...) /**/
69#endif
70
Al Viro27496a82005-10-21 03:20:48 -040071extern void * __jbd_kmalloc (const char *where, size_t size, gfp_t flags, int retry);
Linus Torvalds1da177e2005-04-16 15:20:36 -070072#define jbd_kmalloc(size, flags) \
73 __jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
74#define jbd_rep_kmalloc(size, flags) \
75 __jbd_kmalloc(__FUNCTION__, (size), (flags), 1)
76
77#define JFS_MIN_JOURNAL_BLOCKS 1024
78
79#ifdef __KERNEL__
80
81/**
82 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
83 *
84 * All filesystem modifications made by the process go
85 * through this handle. Recursive operations (such as quota operations)
86 * are gathered into a single update.
87 *
88 * The buffer credits field is used to account for journaled buffers
89 * being modified by the running process. To ensure that there is
90 * enough log space for all outstanding operations, we need to limit the
91 * number of outstanding buffers possible at any time. When the
92 * operation completes, any buffer credits not used are credited back to
93 * the transaction, so that at all times we know how many buffers the
94 * outstanding updates on a transaction might possibly touch.
95 *
96 * This is an opaque datatype.
97 **/
98typedef struct handle_s handle_t; /* Atomic operation type */
99
100
101/**
102 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
103 *
104 * journal_t is linked to from the fs superblock structure.
105 *
106 * We use the journal_t to keep track of all outstanding transaction
107 * activity on the filesystem, and to manage the state of the log
108 * writing process.
109 *
110 * This is an opaque datatype.
111 **/
112typedef struct journal_s journal_t; /* Journal control structure */
113#endif
114
115/*
116 * Internal structures used by the logging mechanism:
117 */
118
119#define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
120
121/*
122 * On-disk structures
123 */
124
125/*
126 * Descriptor block types:
127 */
128
129#define JFS_DESCRIPTOR_BLOCK 1
130#define JFS_COMMIT_BLOCK 2
131#define JFS_SUPERBLOCK_V1 3
132#define JFS_SUPERBLOCK_V2 4
133#define JFS_REVOKE_BLOCK 5
134
135/*
136 * Standard header for all descriptor blocks:
137 */
138typedef struct journal_header_s
139{
140 __be32 h_magic;
141 __be32 h_blocktype;
142 __be32 h_sequence;
143} journal_header_t;
144
145
146/*
147 * The block tag: used to describe a single buffer in the journal
148 */
149typedef struct journal_block_tag_s
150{
151 __be32 t_blocknr; /* The on-disk block number */
152 __be32 t_flags; /* See below */
153} journal_block_tag_t;
154
155/*
156 * The revoke descriptor: used on disk to describe a series of blocks to
157 * be revoked from the log
158 */
159typedef struct journal_revoke_header_s
160{
161 journal_header_t r_header;
162 __be32 r_count; /* Count of bytes used in the block */
163} journal_revoke_header_t;
164
165
166/* Definitions for the journal tag flags word: */
167#define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */
168#define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */
169#define JFS_FLAG_DELETED 4 /* block deleted by this transaction */
170#define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
171
172
173/*
174 * The journal superblock. All fields are in big-endian byte order.
175 */
176typedef struct journal_superblock_s
177{
178/* 0x0000 */
179 journal_header_t s_header;
180
181/* 0x000C */
182 /* Static information describing the journal */
183 __be32 s_blocksize; /* journal device blocksize */
184 __be32 s_maxlen; /* total blocks in journal file */
185 __be32 s_first; /* first block of log information */
186
187/* 0x0018 */
188 /* Dynamic information describing the current state of the log */
189 __be32 s_sequence; /* first commit ID expected in log */
190 __be32 s_start; /* blocknr of start of log */
191
192/* 0x0020 */
193 /* Error value, as set by journal_abort(). */
194 __be32 s_errno;
195
196/* 0x0024 */
197 /* Remaining fields are only valid in a version-2 superblock */
198 __be32 s_feature_compat; /* compatible feature set */
199 __be32 s_feature_incompat; /* incompatible feature set */
200 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
201/* 0x0030 */
202 __u8 s_uuid[16]; /* 128-bit uuid for journal */
203
204/* 0x0040 */
205 __be32 s_nr_users; /* Nr of filesystems sharing log */
206
207 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
208
209/* 0x0048 */
210 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
211 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
212
213/* 0x0050 */
214 __u32 s_padding[44];
215
216/* 0x0100 */
217 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
218/* 0x0400 */
219} journal_superblock_t;
220
221#define JFS_HAS_COMPAT_FEATURE(j,mask) \
222 ((j)->j_format_version >= 2 && \
223 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
224#define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \
225 ((j)->j_format_version >= 2 && \
226 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
227#define JFS_HAS_INCOMPAT_FEATURE(j,mask) \
228 ((j)->j_format_version >= 2 && \
229 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
230
231#define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001
232
233/* Features known to this kernel version: */
234#define JFS_KNOWN_COMPAT_FEATURES 0
235#define JFS_KNOWN_ROCOMPAT_FEATURES 0
236#define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE
237
238#ifdef __KERNEL__
239
240#include <linux/fs.h>
241#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242
243#define JBD_ASSERTIONS
244#ifdef JBD_ASSERTIONS
245#define J_ASSERT(assert) \
246do { \
247 if (!(assert)) { \
248 printk (KERN_EMERG \
249 "Assertion failure in %s() at %s:%d: \"%s\"\n", \
250 __FUNCTION__, __FILE__, __LINE__, # assert); \
251 BUG(); \
252 } \
253} while (0)
254
255#if defined(CONFIG_BUFFER_DEBUG)
256void buffer_assertion_failure(struct buffer_head *bh);
257#define J_ASSERT_BH(bh, expr) \
258 do { \
259 if (!(expr)) \
260 buffer_assertion_failure(bh); \
261 J_ASSERT(expr); \
262 } while (0)
263#define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
264#else
265#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
266#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
267#endif
268
269#else
270#define J_ASSERT(assert) do { } while (0)
271#endif /* JBD_ASSERTIONS */
272
273#if defined(JBD_PARANOID_IOFAIL)
274#define J_EXPECT(expr, why...) J_ASSERT(expr)
275#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
276#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
277#else
278#define __journal_expect(expr, why...) \
279 ({ \
280 int val = (expr); \
281 if (!val) { \
282 printk(KERN_ERR \
283 "EXT3-fs unexpected failure: %s;\n",# expr); \
284 printk(KERN_ERR why "\n"); \
285 } \
286 val; \
287 })
288#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
289#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
290#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
291#endif
292
293enum jbd_state_bits {
294 BH_JBD /* Has an attached ext3 journal_head */
295 = BH_PrivateStart,
296 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
297 BH_Freed, /* Has been freed (truncated) */
298 BH_Revoked, /* Has been revoked from the log */
299 BH_RevokeValid, /* Revoked flag is valid */
300 BH_JBDDirty, /* Is dirty but journaled */
301 BH_State, /* Pins most journal_head state */
302 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
303 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
304};
305
306BUFFER_FNS(JBD, jbd)
307BUFFER_FNS(JWrite, jwrite)
308BUFFER_FNS(JBDDirty, jbddirty)
309TAS_BUFFER_FNS(JBDDirty, jbddirty)
310BUFFER_FNS(Revoked, revoked)
311TAS_BUFFER_FNS(Revoked, revoked)
312BUFFER_FNS(RevokeValid, revokevalid)
313TAS_BUFFER_FNS(RevokeValid, revokevalid)
314BUFFER_FNS(Freed, freed)
315
316static inline struct buffer_head *jh2bh(struct journal_head *jh)
317{
318 return jh->b_bh;
319}
320
321static inline struct journal_head *bh2jh(struct buffer_head *bh)
322{
323 return bh->b_private;
324}
325
326static inline void jbd_lock_bh_state(struct buffer_head *bh)
327{
328 bit_spin_lock(BH_State, &bh->b_state);
329}
330
331static inline int jbd_trylock_bh_state(struct buffer_head *bh)
332{
333 return bit_spin_trylock(BH_State, &bh->b_state);
334}
335
336static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
337{
338 return bit_spin_is_locked(BH_State, &bh->b_state);
339}
340
341static inline void jbd_unlock_bh_state(struct buffer_head *bh)
342{
343 bit_spin_unlock(BH_State, &bh->b_state);
344}
345
346static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
347{
348 bit_spin_lock(BH_JournalHead, &bh->b_state);
349}
350
351static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
352{
353 bit_spin_unlock(BH_JournalHead, &bh->b_state);
354}
355
356struct jbd_revoke_table_s;
357
358/**
359 * struct handle_s - The handle_s type is the concrete type associated with
360 * handle_t.
361 * @h_transaction: Which compound transaction is this update a part of?
362 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
363 * @h_ref: Reference count on this handle
364 * @h_err: Field for caller's use to track errors through large fs operations
365 * @h_sync: flag for sync-on-close
366 * @h_jdata: flag to force data journaling
367 * @h_aborted: flag indicating fatal error on handle
368 **/
369
370/* Docbook can't yet cope with the bit fields, but will leave the documentation
371 * in so it can be fixed later.
372 */
373
374struct handle_s
375{
376 /* Which compound transaction is this update a part of? */
377 transaction_t *h_transaction;
378
379 /* Number of remaining buffers we are allowed to dirty: */
380 int h_buffer_credits;
381
382 /* Reference count on this handle */
383 int h_ref;
384
385 /* Field for caller's use to track errors through large fs */
386 /* operations */
387 int h_err;
388
389 /* Flags [no locking] */
390 unsigned int h_sync: 1; /* sync-on-close */
391 unsigned int h_jdata: 1; /* force data journaling */
392 unsigned int h_aborted: 1; /* fatal error on handle */
393};
394
395
396/* The transaction_t type is the guts of the journaling mechanism. It
397 * tracks a compound transaction through its various states:
398 *
399 * RUNNING: accepting new updates
400 * LOCKED: Updates still running but we don't accept new ones
401 * RUNDOWN: Updates are tidying up but have finished requesting
402 * new buffers to modify (state not used for now)
403 * FLUSH: All updates complete, but we are still writing to disk
404 * COMMIT: All data on disk, writing commit record
405 * FINISHED: We still have to keep the transaction for checkpointing.
406 *
407 * The transaction keeps track of all of the buffers modified by a
408 * running transaction, and all of the buffers committed but not yet
409 * flushed to home for finished transactions.
410 */
411
412/*
413 * Lock ranking:
414 *
415 * j_list_lock
416 * ->jbd_lock_bh_journal_head() (This is "innermost")
417 *
418 * j_state_lock
419 * ->jbd_lock_bh_state()
420 *
421 * jbd_lock_bh_state()
422 * ->j_list_lock
423 *
424 * j_state_lock
425 * ->t_handle_lock
426 *
427 * j_state_lock
428 * ->j_list_lock (journal_unmap_buffer)
429 *
430 */
431
432struct transaction_s
433{
434 /* Pointer to the journal for this transaction. [no locking] */
435 journal_t *t_journal;
436
437 /* Sequence number for this transaction [no locking] */
438 tid_t t_tid;
439
440 /*
441 * Transaction's current state
442 * [no locking - only kjournald alters this]
443 * FIXME: needs barriers
444 * KLUDGE: [use j_state_lock]
445 */
446 enum {
447 T_RUNNING,
448 T_LOCKED,
449 T_RUNDOWN,
450 T_FLUSH,
451 T_COMMIT,
452 T_FINISHED
453 } t_state;
454
455 /*
456 * Where in the log does this transaction's commit start? [no locking]
457 */
458 unsigned long t_log_start;
459
460 /* Number of buffers on the t_buffers list [j_list_lock] */
461 int t_nr_buffers;
462
463 /*
464 * Doubly-linked circular list of all buffers reserved but not yet
465 * modified by this transaction [j_list_lock]
466 */
467 struct journal_head *t_reserved_list;
468
469 /*
470 * Doubly-linked circular list of all buffers under writeout during
471 * commit [j_list_lock]
472 */
473 struct journal_head *t_locked_list;
474
475 /*
476 * Doubly-linked circular list of all metadata buffers owned by this
477 * transaction [j_list_lock]
478 */
479 struct journal_head *t_buffers;
480
481 /*
482 * Doubly-linked circular list of all data buffers still to be
483 * flushed before this transaction can be committed [j_list_lock]
484 */
485 struct journal_head *t_sync_datalist;
486
487 /*
488 * Doubly-linked circular list of all forget buffers (superseded
489 * buffers which we can un-checkpoint once this transaction commits)
490 * [j_list_lock]
491 */
492 struct journal_head *t_forget;
493
494 /*
495 * Doubly-linked circular list of all buffers still to be flushed before
496 * this transaction can be checkpointed. [j_list_lock]
497 */
498 struct journal_head *t_checkpoint_list;
499
500 /*
501 * Doubly-linked circular list of temporary buffers currently undergoing
502 * IO in the log [j_list_lock]
503 */
504 struct journal_head *t_iobuf_list;
505
506 /*
507 * Doubly-linked circular list of metadata buffers being shadowed by log
508 * IO. The IO buffers on the iobuf list and the shadow buffers on this
509 * list match each other one for one at all times. [j_list_lock]
510 */
511 struct journal_head *t_shadow_list;
512
513 /*
514 * Doubly-linked circular list of control buffers being written to the
515 * log. [j_list_lock]
516 */
517 struct journal_head *t_log_list;
518
519 /*
520 * Protects info related to handles
521 */
522 spinlock_t t_handle_lock;
523
524 /*
525 * Number of outstanding updates running on this transaction
526 * [t_handle_lock]
527 */
528 int t_updates;
529
530 /*
531 * Number of buffers reserved for use by all handles in this transaction
532 * handle but not yet modified. [t_handle_lock]
533 */
534 int t_outstanding_credits;
535
536 /*
537 * Forward and backward links for the circular list of all transactions
538 * awaiting checkpoint. [j_list_lock]
539 */
540 transaction_t *t_cpnext, *t_cpprev;
541
542 /*
543 * When will the transaction expire (become due for commit), in jiffies?
544 * [no locking]
545 */
546 unsigned long t_expires;
547
548 /*
549 * How many handles used this transaction? [t_handle_lock]
550 */
551 int t_handle_count;
552
553};
554
555/**
556 * struct journal_s - The journal_s type is the concrete type associated with
557 * journal_t.
558 * @j_flags: General journaling state flags
559 * @j_errno: Is there an outstanding uncleared error on the journal (from a
560 * prior abort)?
561 * @j_sb_buffer: First part of superblock buffer
562 * @j_superblock: Second part of superblock buffer
563 * @j_format_version: Version of the superblock format
564 * @j_state_lock: Protect the various scalars in the journal
565 * @j_barrier_count: Number of processes waiting to create a barrier lock
566 * @j_barrier: The barrier lock itself
567 * @j_running_transaction: The current running transaction..
568 * @j_committing_transaction: the transaction we are pushing to disk
569 * @j_checkpoint_transactions: a linked circular list of all transactions
570 * waiting for checkpointing
571 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
572 * to start committing, or for a barrier lock to be released
573 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
574 * @j_wait_done_commit: Wait queue for waiting for commit to complete
575 * @j_wait_checkpoint: Wait queue to trigger checkpointing
576 * @j_wait_commit: Wait queue to trigger commit
577 * @j_wait_updates: Wait queue to wait for updates to complete
578 * @j_checkpoint_sem: Semaphore for locking against concurrent checkpoints
579 * @j_head: Journal head - identifies the first unused block in the journal
580 * @j_tail: Journal tail - identifies the oldest still-used block in the
581 * journal.
582 * @j_free: Journal free - how many free blocks are there in the journal?
583 * @j_first: The block number of the first usable block
584 * @j_last: The block number one beyond the last usable block
585 * @j_dev: Device where we store the journal
586 * @j_blocksize: blocksize for the location where we store the journal.
587 * @j_blk_offset: starting block offset for into the device where we store the
588 * journal
589 * @j_fs_dev: Device which holds the client fs. For internal journal this will
590 * be equal to j_dev
591 * @j_maxlen: Total maximum capacity of the journal region on disk.
592 * @j_list_lock: Protects the buffer lists and internal buffer state.
593 * @j_inode: Optional inode where we store the journal. If present, all journal
594 * block numbers are mapped into this inode via bmap().
595 * @j_tail_sequence: Sequence number of the oldest transaction in the log
596 * @j_transaction_sequence: Sequence number of the next transaction to grant
597 * @j_commit_sequence: Sequence number of the most recently committed
598 * transaction
599 * @j_commit_request: Sequence number of the most recent transaction wanting
600 * commit
601 * @j_uuid: Uuid of client object.
602 * @j_task: Pointer to the current commit thread for this journal
603 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
604 * single compound commit transaction
605 * @j_commit_interval: What is the maximum transaction lifetime before we begin
606 * a commit?
607 * @j_commit_timer: The timer used to wakeup the commit thread
608 * @j_revoke_lock: Protect the revoke table
609 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
610 * current transaction.
611 * @j_revoke_table: alternate revoke tables for j_revoke
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800612 * @j_wbuf: array of buffer_heads for journal_commit_transaction
613 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
614 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800615 * @j_last_sync_writer: most recent pid which did a synchronous write
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 * @j_private: An opaque pointer to fs-private information.
617 */
618
619struct journal_s
620{
621 /* General journaling state flags [j_state_lock] */
622 unsigned long j_flags;
623
624 /*
625 * Is there an outstanding uncleared error on the journal (from a prior
626 * abort)? [j_state_lock]
627 */
628 int j_errno;
629
630 /* The superblock buffer */
631 struct buffer_head *j_sb_buffer;
632 journal_superblock_t *j_superblock;
633
634 /* Version of the superblock format */
635 int j_format_version;
636
637 /*
638 * Protect the various scalars in the journal
639 */
640 spinlock_t j_state_lock;
641
642 /*
643 * Number of processes waiting to create a barrier lock [j_state_lock]
644 */
645 int j_barrier_count;
646
647 /* The barrier lock itself */
648 struct semaphore j_barrier;
649
650 /*
651 * Transactions: The current running transaction...
652 * [j_state_lock] [caller holding open handle]
653 */
654 transaction_t *j_running_transaction;
655
656 /*
657 * the transaction we are pushing to disk
658 * [j_state_lock] [caller holding open handle]
659 */
660 transaction_t *j_committing_transaction;
661
662 /*
663 * ... and a linked circular list of all transactions waiting for
664 * checkpointing. [j_list_lock]
665 */
666 transaction_t *j_checkpoint_transactions;
667
668 /*
669 * Wait queue for waiting for a locked transaction to start committing,
670 * or for a barrier lock to be released
671 */
672 wait_queue_head_t j_wait_transaction_locked;
673
674 /* Wait queue for waiting for checkpointing to complete */
675 wait_queue_head_t j_wait_logspace;
676
677 /* Wait queue for waiting for commit to complete */
678 wait_queue_head_t j_wait_done_commit;
679
680 /* Wait queue to trigger checkpointing */
681 wait_queue_head_t j_wait_checkpoint;
682
683 /* Wait queue to trigger commit */
684 wait_queue_head_t j_wait_commit;
685
686 /* Wait queue to wait for updates to complete */
687 wait_queue_head_t j_wait_updates;
688
689 /* Semaphore for locking against concurrent checkpoints */
690 struct semaphore j_checkpoint_sem;
691
692 /*
693 * Journal head: identifies the first unused block in the journal.
694 * [j_state_lock]
695 */
696 unsigned long j_head;
697
698 /*
699 * Journal tail: identifies the oldest still-used block in the journal.
700 * [j_state_lock]
701 */
702 unsigned long j_tail;
703
704 /*
705 * Journal free: how many free blocks are there in the journal?
706 * [j_state_lock]
707 */
708 unsigned long j_free;
709
710 /*
711 * Journal start and end: the block numbers of the first usable block
712 * and one beyond the last usable block in the journal. [j_state_lock]
713 */
714 unsigned long j_first;
715 unsigned long j_last;
716
717 /*
718 * Device, blocksize and starting block offset for the location where we
719 * store the journal.
720 */
721 struct block_device *j_dev;
722 int j_blocksize;
723 unsigned int j_blk_offset;
724
725 /*
726 * Device which holds the client fs. For internal journal this will be
727 * equal to j_dev.
728 */
729 struct block_device *j_fs_dev;
730
731 /* Total maximum capacity of the journal region on disk. */
732 unsigned int j_maxlen;
733
734 /*
735 * Protects the buffer lists and internal buffer state.
736 */
737 spinlock_t j_list_lock;
738
739 /* Optional inode where we store the journal. If present, all */
740 /* journal block numbers are mapped into this inode via */
741 /* bmap(). */
742 struct inode *j_inode;
743
744 /*
745 * Sequence number of the oldest transaction in the log [j_state_lock]
746 */
747 tid_t j_tail_sequence;
748
749 /*
750 * Sequence number of the next transaction to grant [j_state_lock]
751 */
752 tid_t j_transaction_sequence;
753
754 /*
755 * Sequence number of the most recently committed transaction
756 * [j_state_lock].
757 */
758 tid_t j_commit_sequence;
759
760 /*
761 * Sequence number of the most recent transaction wanting commit
762 * [j_state_lock]
763 */
764 tid_t j_commit_request;
765
766 /*
767 * Journal uuid: identifies the object (filesystem, LVM volume etc)
768 * backed by this journal. This will eventually be replaced by an array
769 * of uuids, allowing us to index multiple devices within a single
770 * journal and to perform atomic updates across them.
771 */
772 __u8 j_uuid[16];
773
774 /* Pointer to the current commit thread for this journal */
775 struct task_struct *j_task;
776
777 /*
778 * Maximum number of metadata buffers to allow in a single compound
779 * commit transaction
780 */
781 int j_max_transaction_buffers;
782
783 /*
784 * What is the maximum transaction lifetime before we begin a commit?
785 */
786 unsigned long j_commit_interval;
787
788 /* The timer used to wakeup the commit thread: */
789 struct timer_list *j_commit_timer;
790
791 /*
792 * The revoke table: maintains the list of revoked blocks in the
793 * current transaction. [j_revoke_lock]
794 */
795 spinlock_t j_revoke_lock;
796 struct jbd_revoke_table_s *j_revoke;
797 struct jbd_revoke_table_s *j_revoke_table[2];
798
799 /*
800 * array of bhs for journal_commit_transaction
801 */
802 struct buffer_head **j_wbuf;
803 int j_wbufsize;
804
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800805 pid_t j_last_sync_writer;
806
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807 /*
808 * An opaque pointer to fs-private information. ext3 puts its
809 * superblock pointer here
810 */
811 void *j_private;
812};
813
814/*
815 * Journal flag definitions
816 */
817#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
818#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
819#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
820#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
821#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
822#define JFS_BARRIER 0x020 /* Use IDE barriers */
823
824/*
825 * Function declarations for the journaling transaction and buffer
826 * management
827 */
828
829/* Filing buffers */
830extern void __journal_temp_unlink_buffer(struct journal_head *jh);
831extern void journal_unfile_buffer(journal_t *, struct journal_head *);
832extern void __journal_unfile_buffer(struct journal_head *);
833extern void __journal_refile_buffer(struct journal_head *);
834extern void journal_refile_buffer(journal_t *, struct journal_head *);
835extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
836extern void __journal_free_buffer(struct journal_head *bh);
837extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
838extern void __journal_clean_data_list(transaction_t *transaction);
839
840/* Log buffer allocation */
841extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
842int journal_next_log_block(journal_t *, unsigned long *);
843
844/* Commit management */
845extern void journal_commit_transaction(journal_t *);
846
847/* Checkpoint list management */
848int __journal_clean_checkpoint_list(journal_t *journal);
Mark Fasheh7c8903f2006-02-14 13:53:03 -0800849void __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
851
852/* Buffer IO */
853extern int
854journal_write_metadata_buffer(transaction_t *transaction,
855 struct journal_head *jh_in,
856 struct journal_head **jh_out,
857 int blocknr);
858
859/* Transaction locking */
860extern void __wait_on_journal (journal_t *);
861
862/*
863 * Journal locking.
864 *
865 * We need to lock the journal during transaction state changes so that nobody
866 * ever tries to take a handle on the running transaction while we are in the
867 * middle of moving it to the commit phase. j_state_lock does this.
868 *
869 * Note that the locking is completely interrupt unsafe. We never touch
870 * journal structures from interrupts.
871 */
872
873static inline handle_t *journal_current_handle(void)
874{
875 return current->journal_info;
876}
877
878/* The journaling code user interface:
879 *
880 * Create and destroy handles
881 * Register buffer modifications against the current transaction.
882 */
883
884extern handle_t *journal_start(journal_t *, int nblocks);
885extern int journal_restart (handle_t *, int nblocks);
886extern int journal_extend (handle_t *, int nblocks);
887extern int journal_get_write_access(handle_t *, struct buffer_head *);
888extern int journal_get_create_access (handle_t *, struct buffer_head *);
889extern int journal_get_undo_access(handle_t *, struct buffer_head *);
890extern int journal_dirty_data (handle_t *, struct buffer_head *);
891extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
892extern void journal_release_buffer (handle_t *, struct buffer_head *);
893extern int journal_forget (handle_t *, struct buffer_head *);
894extern void journal_sync_buffer (struct buffer_head *);
895extern int journal_invalidatepage(journal_t *,
896 struct page *, unsigned long);
Al Viro27496a82005-10-21 03:20:48 -0400897extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898extern int journal_stop(handle_t *);
899extern int journal_flush (journal_t *);
900extern void journal_lock_updates (journal_t *);
901extern void journal_unlock_updates (journal_t *);
902
903extern journal_t * journal_init_dev(struct block_device *bdev,
904 struct block_device *fs_dev,
905 int start, int len, int bsize);
906extern journal_t * journal_init_inode (struct inode *);
907extern int journal_update_format (journal_t *);
908extern int journal_check_used_features
909 (journal_t *, unsigned long, unsigned long, unsigned long);
910extern int journal_check_available_features
911 (journal_t *, unsigned long, unsigned long, unsigned long);
912extern int journal_set_features
913 (journal_t *, unsigned long, unsigned long, unsigned long);
914extern int journal_create (journal_t *);
915extern int journal_load (journal_t *journal);
916extern void journal_destroy (journal_t *);
917extern int journal_recover (journal_t *journal);
918extern int journal_wipe (journal_t *, int);
919extern int journal_skip_recovery (journal_t *);
920extern void journal_update_superblock (journal_t *, int);
921extern void __journal_abort_hard (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922extern void journal_abort (journal_t *, int);
923extern int journal_errno (journal_t *);
924extern void journal_ack_err (journal_t *);
925extern int journal_clear_err (journal_t *);
926extern int journal_bmap(journal_t *, unsigned long, unsigned long *);
927extern int journal_force_commit(journal_t *);
928
929/*
930 * journal_head management
931 */
932struct journal_head *journal_add_journal_head(struct buffer_head *bh);
933struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
934void journal_remove_journal_head(struct buffer_head *bh);
935void journal_put_journal_head(struct journal_head *jh);
936
937/*
938 * handle management
939 */
940extern kmem_cache_t *jbd_handle_cache;
941
Al Virodd0fc662005-10-07 07:46:04 +0100942static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943{
944 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
945}
946
947static inline void jbd_free_handle(handle_t *handle)
948{
949 kmem_cache_free(jbd_handle_cache, handle);
950}
951
952/* Primary revoke support */
953#define JOURNAL_REVOKE_DEFAULT_HASH 256
954extern int journal_init_revoke(journal_t *, int);
955extern void journal_destroy_revoke_caches(void);
956extern int journal_init_revoke_caches(void);
957
958extern void journal_destroy_revoke(journal_t *);
959extern int journal_revoke (handle_t *,
960 unsigned long, struct buffer_head *);
961extern int journal_cancel_revoke(handle_t *, struct journal_head *);
962extern void journal_write_revoke_records(journal_t *, transaction_t *);
963
964/* Recovery revoke support */
965extern int journal_set_revoke(journal_t *, unsigned long, tid_t);
966extern int journal_test_revoke(journal_t *, unsigned long, tid_t);
967extern void journal_clear_revoke(journal_t *);
968extern void journal_brelse_array(struct buffer_head *b[], int n);
969extern void journal_switch_revoke_table(journal_t *journal);
970
971/*
972 * The log thread user interface:
973 *
974 * Request space in the current transaction, and force transaction commit
975 * transitions on demand.
976 */
977
978int __log_space_left(journal_t *); /* Called with journal locked */
979int log_start_commit(journal_t *journal, tid_t tid);
980int __log_start_commit(journal_t *journal, tid_t tid);
981int journal_start_commit(journal_t *journal, tid_t *tid);
982int journal_force_commit_nested(journal_t *journal);
983int log_wait_commit(journal_t *journal, tid_t tid);
984int log_do_checkpoint(journal_t *journal);
985
986void __log_wait_for_space(journal_t *journal);
987extern void __journal_drop_transaction(journal_t *, transaction_t *);
988extern int cleanup_journal_tail(journal_t *);
989
990/* Debugging code only: */
991
992#define jbd_ENOSYS() \
993do { \
994 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
995 current->state = TASK_UNINTERRUPTIBLE; \
996 schedule(); \
997} while (1)
998
999/*
1000 * is_journal_abort
1001 *
1002 * Simple test wrapper function to test the JFS_ABORT state flag. This
1003 * bit, when set, indicates that we have had a fatal error somewhere,
1004 * either inside the journaling layer or indicated to us by the client
1005 * (eg. ext3), and that we and should not commit any further
1006 * transactions.
1007 */
1008
1009static inline int is_journal_aborted(journal_t *journal)
1010{
1011 return journal->j_flags & JFS_ABORT;
1012}
1013
1014static inline int is_handle_aborted(handle_t *handle)
1015{
1016 if (handle->h_aborted)
1017 return 1;
1018 return is_journal_aborted(handle->h_transaction->t_journal);
1019}
1020
1021static inline void journal_abort_handle(handle_t *handle)
1022{
1023 handle->h_aborted = 1;
1024}
1025
1026#endif /* __KERNEL__ */
1027
1028/* Comparison functions for transaction IDs: perform comparisons using
1029 * modulo arithmetic so that they work over sequence number wraps. */
1030
1031static inline int tid_gt(tid_t x, tid_t y)
1032{
1033 int difference = (x - y);
1034 return (difference > 0);
1035}
1036
1037static inline int tid_geq(tid_t x, tid_t y)
1038{
1039 int difference = (x - y);
1040 return (difference >= 0);
1041}
1042
1043extern int journal_blocks_per_page(struct inode *inode);
1044
1045/*
1046 * Return the minimum number of blocks which must be free in the journal
1047 * before a new transaction may be started. Must be called under j_state_lock.
1048 */
1049static inline int jbd_space_needed(journal_t *journal)
1050{
1051 int nblocks = journal->j_max_transaction_buffers;
1052 if (journal->j_committing_transaction)
1053 nblocks += journal->j_committing_transaction->
1054 t_outstanding_credits;
1055 return nblocks;
1056}
1057
1058/*
1059 * Definitions which augment the buffer_head layer
1060 */
1061
1062/* journaling buffer types */
1063#define BJ_None 0 /* Not journaled */
1064#define BJ_SyncData 1 /* Normal data: flush before commit */
1065#define BJ_Metadata 2 /* Normal journaled metadata */
1066#define BJ_Forget 3 /* Buffer superseded by this transaction */
1067#define BJ_IO 4 /* Buffer is for temporary IO use */
1068#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1069#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1070#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1071#define BJ_Locked 8 /* Locked for I/O during commit */
1072#define BJ_Types 9
1073
1074extern int jbd_blocks_per_page(struct inode *inode);
1075
1076#ifdef __KERNEL__
1077
1078#define buffer_trace_init(bh) do {} while (0)
1079#define print_buffer_fields(bh) do {} while (0)
1080#define print_buffer_trace(bh) do {} while (0)
1081#define BUFFER_TRACE(bh, info) do {} while (0)
1082#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1083#define JBUFFER_TRACE(jh, info) do {} while (0)
1084
1085#endif /* __KERNEL__ */
1086
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087#endif /* _LINUX_JBD_H */