aboutsummaryrefslogtreecommitdiffstats
path: root/include/linux/jbd.h
diff options
context:
space:
mode:
Diffstat (limited to 'include/linux/jbd.h')
-rw-r--r--include/linux/jbd.h1098
1 files changed, 1098 insertions, 0 deletions
diff --git a/include/linux/jbd.h b/include/linux/jbd.h
new file mode 100644
index 000000000000..593407e865b1
--- /dev/null
+++ b/include/linux/jbd.h
@@ -0,0 +1,1098 @@
1/*
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
19#if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE) || !defined(__KERNEL__)
20
21/* Allow this file to be included directly into e2fsprogs */
22#ifndef __KERNEL__
23#include "jfs_compat.h"
24#define JFS_DEBUG
25#define jfs_debug jbd_debug
26#else
27
28#include <linux/buffer_head.h>
29#include <linux/journal-head.h>
30#include <linux/stddef.h>
31#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
71extern void * __jbd_kmalloc (const char *where, size_t size, int flags, int retry);
72#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>
242#include <asm/bug.h>
243
244#define JBD_ASSERTIONS
245#ifdef JBD_ASSERTIONS
246#define J_ASSERT(assert) \
247do { \
248 if (!(assert)) { \
249 printk (KERN_EMERG \
250 "Assertion failure in %s() at %s:%d: \"%s\"\n", \
251 __FUNCTION__, __FILE__, __LINE__, # assert); \
252 BUG(); \
253 } \
254} while (0)
255
256#if defined(CONFIG_BUFFER_DEBUG)
257void buffer_assertion_failure(struct buffer_head *bh);
258#define J_ASSERT_BH(bh, expr) \
259 do { \
260 if (!(expr)) \
261 buffer_assertion_failure(bh); \
262 J_ASSERT(expr); \
263 } while (0)
264#define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
265#else
266#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
267#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
268#endif
269
270#else
271#define J_ASSERT(assert) do { } while (0)
272#endif /* JBD_ASSERTIONS */
273
274#if defined(JBD_PARANOID_IOFAIL)
275#define J_EXPECT(expr, why...) J_ASSERT(expr)
276#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
277#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
278#else
279#define __journal_expect(expr, why...) \
280 ({ \
281 int val = (expr); \
282 if (!val) { \
283 printk(KERN_ERR \
284 "EXT3-fs unexpected failure: %s;\n",# expr); \
285 printk(KERN_ERR why "\n"); \
286 } \
287 val; \
288 })
289#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
290#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
291#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
292#endif
293
294enum jbd_state_bits {
295 BH_JBD /* Has an attached ext3 journal_head */
296 = BH_PrivateStart,
297 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
298 BH_Freed, /* Has been freed (truncated) */
299 BH_Revoked, /* Has been revoked from the log */
300 BH_RevokeValid, /* Revoked flag is valid */
301 BH_JBDDirty, /* Is dirty but journaled */
302 BH_State, /* Pins most journal_head state */
303 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
304 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
305};
306
307BUFFER_FNS(JBD, jbd)
308BUFFER_FNS(JWrite, jwrite)
309BUFFER_FNS(JBDDirty, jbddirty)
310TAS_BUFFER_FNS(JBDDirty, jbddirty)
311BUFFER_FNS(Revoked, revoked)
312TAS_BUFFER_FNS(Revoked, revoked)
313BUFFER_FNS(RevokeValid, revokevalid)
314TAS_BUFFER_FNS(RevokeValid, revokevalid)
315BUFFER_FNS(Freed, freed)
316
317static inline struct buffer_head *jh2bh(struct journal_head *jh)
318{
319 return jh->b_bh;
320}
321
322static inline struct journal_head *bh2jh(struct buffer_head *bh)
323{
324 return bh->b_private;
325}
326
327static inline void jbd_lock_bh_state(struct buffer_head *bh)
328{
329 bit_spin_lock(BH_State, &bh->b_state);
330}
331
332static inline int jbd_trylock_bh_state(struct buffer_head *bh)
333{
334 return bit_spin_trylock(BH_State, &bh->b_state);
335}
336
337static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
338{
339 return bit_spin_is_locked(BH_State, &bh->b_state);
340}
341
342static inline void jbd_unlock_bh_state(struct buffer_head *bh)
343{
344 bit_spin_unlock(BH_State, &bh->b_state);
345}
346
347static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
348{
349 bit_spin_lock(BH_JournalHead, &bh->b_state);
350}
351
352static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
353{
354 bit_spin_unlock(BH_JournalHead, &bh->b_state);
355}
356
357struct jbd_revoke_table_s;
358
359/**
360 * struct handle_s - The handle_s type is the concrete type associated with
361 * handle_t.
362 * @h_transaction: Which compound transaction is this update a part of?
363 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
364 * @h_ref: Reference count on this handle
365 * @h_err: Field for caller's use to track errors through large fs operations
366 * @h_sync: flag for sync-on-close
367 * @h_jdata: flag to force data journaling
368 * @h_aborted: flag indicating fatal error on handle
369 **/
370
371/* Docbook can't yet cope with the bit fields, but will leave the documentation
372 * in so it can be fixed later.
373 */
374
375struct handle_s
376{
377 /* Which compound transaction is this update a part of? */
378 transaction_t *h_transaction;
379
380 /* Number of remaining buffers we are allowed to dirty: */
381 int h_buffer_credits;
382
383 /* Reference count on this handle */
384 int h_ref;
385
386 /* Field for caller's use to track errors through large fs */
387 /* operations */
388 int h_err;
389
390 /* Flags [no locking] */
391 unsigned int h_sync: 1; /* sync-on-close */
392 unsigned int h_jdata: 1; /* force data journaling */
393 unsigned int h_aborted: 1; /* fatal error on handle */
394};
395
396
397/* The transaction_t type is the guts of the journaling mechanism. It
398 * tracks a compound transaction through its various states:
399 *
400 * RUNNING: accepting new updates
401 * LOCKED: Updates still running but we don't accept new ones
402 * RUNDOWN: Updates are tidying up but have finished requesting
403 * new buffers to modify (state not used for now)
404 * FLUSH: All updates complete, but we are still writing to disk
405 * COMMIT: All data on disk, writing commit record
406 * FINISHED: We still have to keep the transaction for checkpointing.
407 *
408 * The transaction keeps track of all of the buffers modified by a
409 * running transaction, and all of the buffers committed but not yet
410 * flushed to home for finished transactions.
411 */
412
413/*
414 * Lock ranking:
415 *
416 * j_list_lock
417 * ->jbd_lock_bh_journal_head() (This is "innermost")
418 *
419 * j_state_lock
420 * ->jbd_lock_bh_state()
421 *
422 * jbd_lock_bh_state()
423 * ->j_list_lock
424 *
425 * j_state_lock
426 * ->t_handle_lock
427 *
428 * j_state_lock
429 * ->j_list_lock (journal_unmap_buffer)
430 *
431 */
432
433struct transaction_s
434{
435 /* Pointer to the journal for this transaction. [no locking] */
436 journal_t *t_journal;
437
438 /* Sequence number for this transaction [no locking] */
439 tid_t t_tid;
440
441 /*
442 * Transaction's current state
443 * [no locking - only kjournald alters this]
444 * FIXME: needs barriers
445 * KLUDGE: [use j_state_lock]
446 */
447 enum {
448 T_RUNNING,
449 T_LOCKED,
450 T_RUNDOWN,
451 T_FLUSH,
452 T_COMMIT,
453 T_FINISHED
454 } t_state;
455
456 /*
457 * Where in the log does this transaction's commit start? [no locking]
458 */
459 unsigned long t_log_start;
460
461 /* Number of buffers on the t_buffers list [j_list_lock] */
462 int t_nr_buffers;
463
464 /*
465 * Doubly-linked circular list of all buffers reserved but not yet
466 * modified by this transaction [j_list_lock]
467 */
468 struct journal_head *t_reserved_list;
469
470 /*
471 * Doubly-linked circular list of all buffers under writeout during
472 * commit [j_list_lock]
473 */
474 struct journal_head *t_locked_list;
475
476 /*
477 * Doubly-linked circular list of all metadata buffers owned by this
478 * transaction [j_list_lock]
479 */
480 struct journal_head *t_buffers;
481
482 /*
483 * Doubly-linked circular list of all data buffers still to be
484 * flushed before this transaction can be committed [j_list_lock]
485 */
486 struct journal_head *t_sync_datalist;
487
488 /*
489 * Doubly-linked circular list of all forget buffers (superseded
490 * buffers which we can un-checkpoint once this transaction commits)
491 * [j_list_lock]
492 */
493 struct journal_head *t_forget;
494
495 /*
496 * Doubly-linked circular list of all buffers still to be flushed before
497 * this transaction can be checkpointed. [j_list_lock]
498 */
499 struct journal_head *t_checkpoint_list;
500
501 /*
502 * Doubly-linked circular list of temporary buffers currently undergoing
503 * IO in the log [j_list_lock]
504 */
505 struct journal_head *t_iobuf_list;
506
507 /*
508 * Doubly-linked circular list of metadata buffers being shadowed by log
509 * IO. The IO buffers on the iobuf list and the shadow buffers on this
510 * list match each other one for one at all times. [j_list_lock]
511 */
512 struct journal_head *t_shadow_list;
513
514 /*
515 * Doubly-linked circular list of control buffers being written to the
516 * log. [j_list_lock]
517 */
518 struct journal_head *t_log_list;
519
520 /*
521 * Protects info related to handles
522 */
523 spinlock_t t_handle_lock;
524
525 /*
526 * Number of outstanding updates running on this transaction
527 * [t_handle_lock]
528 */
529 int t_updates;
530
531 /*
532 * Number of buffers reserved for use by all handles in this transaction
533 * handle but not yet modified. [t_handle_lock]
534 */
535 int t_outstanding_credits;
536
537 /*
538 * Forward and backward links for the circular list of all transactions
539 * awaiting checkpoint. [j_list_lock]
540 */
541 transaction_t *t_cpnext, *t_cpprev;
542
543 /*
544 * When will the transaction expire (become due for commit), in jiffies?
545 * [no locking]
546 */
547 unsigned long t_expires;
548
549 /*
550 * How many handles used this transaction? [t_handle_lock]
551 */
552 int t_handle_count;
553
554};
555
556/**
557 * struct journal_s - The journal_s type is the concrete type associated with
558 * journal_t.
559 * @j_flags: General journaling state flags
560 * @j_errno: Is there an outstanding uncleared error on the journal (from a
561 * prior abort)?
562 * @j_sb_buffer: First part of superblock buffer
563 * @j_superblock: Second part of superblock buffer
564 * @j_format_version: Version of the superblock format
565 * @j_state_lock: Protect the various scalars in the journal
566 * @j_barrier_count: Number of processes waiting to create a barrier lock
567 * @j_barrier: The barrier lock itself
568 * @j_running_transaction: The current running transaction..
569 * @j_committing_transaction: the transaction we are pushing to disk
570 * @j_checkpoint_transactions: a linked circular list of all transactions
571 * waiting for checkpointing
572 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
573 * to start committing, or for a barrier lock to be released
574 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
575 * @j_wait_done_commit: Wait queue for waiting for commit to complete
576 * @j_wait_checkpoint: Wait queue to trigger checkpointing
577 * @j_wait_commit: Wait queue to trigger commit
578 * @j_wait_updates: Wait queue to wait for updates to complete
579 * @j_checkpoint_sem: Semaphore for locking against concurrent checkpoints
580 * @j_head: Journal head - identifies the first unused block in the journal
581 * @j_tail: Journal tail - identifies the oldest still-used block in the
582 * journal.
583 * @j_free: Journal free - how many free blocks are there in the journal?
584 * @j_first: The block number of the first usable block
585 * @j_last: The block number one beyond the last usable block
586 * @j_dev: Device where we store the journal
587 * @j_blocksize: blocksize for the location where we store the journal.
588 * @j_blk_offset: starting block offset for into the device where we store the
589 * journal
590 * @j_fs_dev: Device which holds the client fs. For internal journal this will
591 * be equal to j_dev
592 * @j_maxlen: Total maximum capacity of the journal region on disk.
593 * @j_list_lock: Protects the buffer lists and internal buffer state.
594 * @j_inode: Optional inode where we store the journal. If present, all journal
595 * block numbers are mapped into this inode via bmap().
596 * @j_tail_sequence: Sequence number of the oldest transaction in the log
597 * @j_transaction_sequence: Sequence number of the next transaction to grant
598 * @j_commit_sequence: Sequence number of the most recently committed
599 * transaction
600 * @j_commit_request: Sequence number of the most recent transaction wanting
601 * commit
602 * @j_uuid: Uuid of client object.
603 * @j_task: Pointer to the current commit thread for this journal
604 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
605 * single compound commit transaction
606 * @j_commit_interval: What is the maximum transaction lifetime before we begin
607 * a commit?
608 * @j_commit_timer: The timer used to wakeup the commit thread
609 * @j_revoke_lock: Protect the revoke table
610 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
611 * current transaction.
612 * @j_revoke_table: alternate revoke tables for j_revoke
613 * @j_private: An opaque pointer to fs-private information.
614 */
615
616struct journal_s
617{
618 /* General journaling state flags [j_state_lock] */
619 unsigned long j_flags;
620
621 /*
622 * Is there an outstanding uncleared error on the journal (from a prior
623 * abort)? [j_state_lock]
624 */
625 int j_errno;
626
627 /* The superblock buffer */
628 struct buffer_head *j_sb_buffer;
629 journal_superblock_t *j_superblock;
630
631 /* Version of the superblock format */
632 int j_format_version;
633
634 /*
635 * Protect the various scalars in the journal
636 */
637 spinlock_t j_state_lock;
638
639 /*
640 * Number of processes waiting to create a barrier lock [j_state_lock]
641 */
642 int j_barrier_count;
643
644 /* The barrier lock itself */
645 struct semaphore j_barrier;
646
647 /*
648 * Transactions: The current running transaction...
649 * [j_state_lock] [caller holding open handle]
650 */
651 transaction_t *j_running_transaction;
652
653 /*
654 * the transaction we are pushing to disk
655 * [j_state_lock] [caller holding open handle]
656 */
657 transaction_t *j_committing_transaction;
658
659 /*
660 * ... and a linked circular list of all transactions waiting for
661 * checkpointing. [j_list_lock]
662 */
663 transaction_t *j_checkpoint_transactions;
664
665 /*
666 * Wait queue for waiting for a locked transaction to start committing,
667 * or for a barrier lock to be released
668 */
669 wait_queue_head_t j_wait_transaction_locked;
670
671 /* Wait queue for waiting for checkpointing to complete */
672 wait_queue_head_t j_wait_logspace;
673
674 /* Wait queue for waiting for commit to complete */
675 wait_queue_head_t j_wait_done_commit;
676
677 /* Wait queue to trigger checkpointing */
678 wait_queue_head_t j_wait_checkpoint;
679
680 /* Wait queue to trigger commit */
681 wait_queue_head_t j_wait_commit;
682
683 /* Wait queue to wait for updates to complete */
684 wait_queue_head_t j_wait_updates;
685
686 /* Semaphore for locking against concurrent checkpoints */
687 struct semaphore j_checkpoint_sem;
688
689 /*
690 * Journal head: identifies the first unused block in the journal.
691 * [j_state_lock]
692 */
693 unsigned long j_head;
694
695 /*
696 * Journal tail: identifies the oldest still-used block in the journal.
697 * [j_state_lock]
698 */
699 unsigned long j_tail;
700
701 /*
702 * Journal free: how many free blocks are there in the journal?
703 * [j_state_lock]
704 */
705 unsigned long j_free;
706
707 /*
708 * Journal start and end: the block numbers of the first usable block
709 * and one beyond the last usable block in the journal. [j_state_lock]
710 */
711 unsigned long j_first;
712 unsigned long j_last;
713
714 /*
715 * Device, blocksize and starting block offset for the location where we
716 * store the journal.
717 */
718 struct block_device *j_dev;
719 int j_blocksize;
720 unsigned int j_blk_offset;
721
722 /*
723 * Device which holds the client fs. For internal journal this will be
724 * equal to j_dev.
725 */
726 struct block_device *j_fs_dev;
727
728 /* Total maximum capacity of the journal region on disk. */
729 unsigned int j_maxlen;
730
731 /*
732 * Protects the buffer lists and internal buffer state.
733 */
734 spinlock_t j_list_lock;
735
736 /* Optional inode where we store the journal. If present, all */
737 /* journal block numbers are mapped into this inode via */
738 /* bmap(). */
739 struct inode *j_inode;
740
741 /*
742 * Sequence number of the oldest transaction in the log [j_state_lock]
743 */
744 tid_t j_tail_sequence;
745
746 /*
747 * Sequence number of the next transaction to grant [j_state_lock]
748 */
749 tid_t j_transaction_sequence;
750
751 /*
752 * Sequence number of the most recently committed transaction
753 * [j_state_lock].
754 */
755 tid_t j_commit_sequence;
756
757 /*
758 * Sequence number of the most recent transaction wanting commit
759 * [j_state_lock]
760 */
761 tid_t j_commit_request;
762
763 /*
764 * Journal uuid: identifies the object (filesystem, LVM volume etc)
765 * backed by this journal. This will eventually be replaced by an array
766 * of uuids, allowing us to index multiple devices within a single
767 * journal and to perform atomic updates across them.
768 */
769 __u8 j_uuid[16];
770
771 /* Pointer to the current commit thread for this journal */
772 struct task_struct *j_task;
773
774 /*
775 * Maximum number of metadata buffers to allow in a single compound
776 * commit transaction
777 */
778 int j_max_transaction_buffers;
779
780 /*
781 * What is the maximum transaction lifetime before we begin a commit?
782 */
783 unsigned long j_commit_interval;
784
785 /* The timer used to wakeup the commit thread: */
786 struct timer_list *j_commit_timer;
787
788 /*
789 * The revoke table: maintains the list of revoked blocks in the
790 * current transaction. [j_revoke_lock]
791 */
792 spinlock_t j_revoke_lock;
793 struct jbd_revoke_table_s *j_revoke;
794 struct jbd_revoke_table_s *j_revoke_table[2];
795
796 /*
797 * array of bhs for journal_commit_transaction
798 */
799 struct buffer_head **j_wbuf;
800 int j_wbufsize;
801
802 /*
803 * An opaque pointer to fs-private information. ext3 puts its
804 * superblock pointer here
805 */
806 void *j_private;
807};
808
809/*
810 * Journal flag definitions
811 */
812#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
813#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
814#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
815#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
816#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
817#define JFS_BARRIER 0x020 /* Use IDE barriers */
818
819/*
820 * Function declarations for the journaling transaction and buffer
821 * management
822 */
823
824/* Filing buffers */
825extern void __journal_temp_unlink_buffer(struct journal_head *jh);
826extern void journal_unfile_buffer(journal_t *, struct journal_head *);
827extern void __journal_unfile_buffer(struct journal_head *);
828extern void __journal_refile_buffer(struct journal_head *);
829extern void journal_refile_buffer(journal_t *, struct journal_head *);
830extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
831extern void __journal_free_buffer(struct journal_head *bh);
832extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
833extern void __journal_clean_data_list(transaction_t *transaction);
834
835/* Log buffer allocation */
836extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
837int journal_next_log_block(journal_t *, unsigned long *);
838
839/* Commit management */
840extern void journal_commit_transaction(journal_t *);
841
842/* Checkpoint list management */
843int __journal_clean_checkpoint_list(journal_t *journal);
844void __journal_remove_checkpoint(struct journal_head *);
845void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
846
847/* Buffer IO */
848extern int
849journal_write_metadata_buffer(transaction_t *transaction,
850 struct journal_head *jh_in,
851 struct journal_head **jh_out,
852 int blocknr);
853
854/* Transaction locking */
855extern void __wait_on_journal (journal_t *);
856
857/*
858 * Journal locking.
859 *
860 * We need to lock the journal during transaction state changes so that nobody
861 * ever tries to take a handle on the running transaction while we are in the
862 * middle of moving it to the commit phase. j_state_lock does this.
863 *
864 * Note that the locking is completely interrupt unsafe. We never touch
865 * journal structures from interrupts.
866 */
867
868static inline handle_t *journal_current_handle(void)
869{
870 return current->journal_info;
871}
872
873/* The journaling code user interface:
874 *
875 * Create and destroy handles
876 * Register buffer modifications against the current transaction.
877 */
878
879extern handle_t *journal_start(journal_t *, int nblocks);
880extern int journal_restart (handle_t *, int nblocks);
881extern int journal_extend (handle_t *, int nblocks);
882extern int journal_get_write_access(handle_t *, struct buffer_head *);
883extern int journal_get_create_access (handle_t *, struct buffer_head *);
884extern int journal_get_undo_access(handle_t *, struct buffer_head *);
885extern int journal_dirty_data (handle_t *, struct buffer_head *);
886extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
887extern void journal_release_buffer (handle_t *, struct buffer_head *);
888extern int journal_forget (handle_t *, struct buffer_head *);
889extern void journal_sync_buffer (struct buffer_head *);
890extern int journal_invalidatepage(journal_t *,
891 struct page *, unsigned long);
892extern int journal_try_to_free_buffers(journal_t *, struct page *, int);
893extern int journal_stop(handle_t *);
894extern int journal_flush (journal_t *);
895extern void journal_lock_updates (journal_t *);
896extern void journal_unlock_updates (journal_t *);
897
898extern journal_t * journal_init_dev(struct block_device *bdev,
899 struct block_device *fs_dev,
900 int start, int len, int bsize);
901extern journal_t * journal_init_inode (struct inode *);
902extern int journal_update_format (journal_t *);
903extern int journal_check_used_features
904 (journal_t *, unsigned long, unsigned long, unsigned long);
905extern int journal_check_available_features
906 (journal_t *, unsigned long, unsigned long, unsigned long);
907extern int journal_set_features
908 (journal_t *, unsigned long, unsigned long, unsigned long);
909extern int journal_create (journal_t *);
910extern int journal_load (journal_t *journal);
911extern void journal_destroy (journal_t *);
912extern int journal_recover (journal_t *journal);
913extern int journal_wipe (journal_t *, int);
914extern int journal_skip_recovery (journal_t *);
915extern void journal_update_superblock (journal_t *, int);
916extern void __journal_abort_hard (journal_t *);
917extern void __journal_abort_soft (journal_t *, int);
918extern void journal_abort (journal_t *, int);
919extern int journal_errno (journal_t *);
920extern void journal_ack_err (journal_t *);
921extern int journal_clear_err (journal_t *);
922extern int journal_bmap(journal_t *, unsigned long, unsigned long *);
923extern int journal_force_commit(journal_t *);
924
925/*
926 * journal_head management
927 */
928struct journal_head *journal_add_journal_head(struct buffer_head *bh);
929struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
930void journal_remove_journal_head(struct buffer_head *bh);
931void journal_put_journal_head(struct journal_head *jh);
932
933/*
934 * handle management
935 */
936extern kmem_cache_t *jbd_handle_cache;
937
938static inline handle_t *jbd_alloc_handle(unsigned int __nocast gfp_flags)
939{
940 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
941}
942
943static inline void jbd_free_handle(handle_t *handle)
944{
945 kmem_cache_free(jbd_handle_cache, handle);
946}
947
948/* Primary revoke support */
949#define JOURNAL_REVOKE_DEFAULT_HASH 256
950extern int journal_init_revoke(journal_t *, int);
951extern void journal_destroy_revoke_caches(void);
952extern int journal_init_revoke_caches(void);
953
954extern void journal_destroy_revoke(journal_t *);
955extern int journal_revoke (handle_t *,
956 unsigned long, struct buffer_head *);
957extern int journal_cancel_revoke(handle_t *, struct journal_head *);
958extern void journal_write_revoke_records(journal_t *, transaction_t *);
959
960/* Recovery revoke support */
961extern int journal_set_revoke(journal_t *, unsigned long, tid_t);
962extern int journal_test_revoke(journal_t *, unsigned long, tid_t);
963extern void journal_clear_revoke(journal_t *);
964extern void journal_brelse_array(struct buffer_head *b[], int n);
965extern void journal_switch_revoke_table(journal_t *journal);
966
967/*
968 * The log thread user interface:
969 *
970 * Request space in the current transaction, and force transaction commit
971 * transitions on demand.
972 */
973
974int __log_space_left(journal_t *); /* Called with journal locked */
975int log_start_commit(journal_t *journal, tid_t tid);
976int __log_start_commit(journal_t *journal, tid_t tid);
977int journal_start_commit(journal_t *journal, tid_t *tid);
978int journal_force_commit_nested(journal_t *journal);
979int log_wait_commit(journal_t *journal, tid_t tid);
980int log_do_checkpoint(journal_t *journal);
981
982void __log_wait_for_space(journal_t *journal);
983extern void __journal_drop_transaction(journal_t *, transaction_t *);
984extern int cleanup_journal_tail(journal_t *);
985
986/* Debugging code only: */
987
988#define jbd_ENOSYS() \
989do { \
990 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
991 current->state = TASK_UNINTERRUPTIBLE; \
992 schedule(); \
993} while (1)
994
995/*
996 * is_journal_abort
997 *
998 * Simple test wrapper function to test the JFS_ABORT state flag. This
999 * bit, when set, indicates that we have had a fatal error somewhere,
1000 * either inside the journaling layer or indicated to us by the client
1001 * (eg. ext3), and that we and should not commit any further
1002 * transactions.
1003 */
1004
1005static inline int is_journal_aborted(journal_t *journal)
1006{
1007 return journal->j_flags & JFS_ABORT;
1008}
1009
1010static inline int is_handle_aborted(handle_t *handle)
1011{
1012 if (handle->h_aborted)
1013 return 1;
1014 return is_journal_aborted(handle->h_transaction->t_journal);
1015}
1016
1017static inline void journal_abort_handle(handle_t *handle)
1018{
1019 handle->h_aborted = 1;
1020}
1021
1022#endif /* __KERNEL__ */
1023
1024/* Comparison functions for transaction IDs: perform comparisons using
1025 * modulo arithmetic so that they work over sequence number wraps. */
1026
1027static inline int tid_gt(tid_t x, tid_t y)
1028{
1029 int difference = (x - y);
1030 return (difference > 0);
1031}
1032
1033static inline int tid_geq(tid_t x, tid_t y)
1034{
1035 int difference = (x - y);
1036 return (difference >= 0);
1037}
1038
1039extern int journal_blocks_per_page(struct inode *inode);
1040
1041/*
1042 * Return the minimum number of blocks which must be free in the journal
1043 * before a new transaction may be started. Must be called under j_state_lock.
1044 */
1045static inline int jbd_space_needed(journal_t *journal)
1046{
1047 int nblocks = journal->j_max_transaction_buffers;
1048 if (journal->j_committing_transaction)
1049 nblocks += journal->j_committing_transaction->
1050 t_outstanding_credits;
1051 return nblocks;
1052}
1053
1054/*
1055 * Definitions which augment the buffer_head layer
1056 */
1057
1058/* journaling buffer types */
1059#define BJ_None 0 /* Not journaled */
1060#define BJ_SyncData 1 /* Normal data: flush before commit */
1061#define BJ_Metadata 2 /* Normal journaled metadata */
1062#define BJ_Forget 3 /* Buffer superseded by this transaction */
1063#define BJ_IO 4 /* Buffer is for temporary IO use */
1064#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1065#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1066#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1067#define BJ_Locked 8 /* Locked for I/O during commit */
1068#define BJ_Types 9
1069
1070extern int jbd_blocks_per_page(struct inode *inode);
1071
1072#ifdef __KERNEL__
1073
1074#define buffer_trace_init(bh) do {} while (0)
1075#define print_buffer_fields(bh) do {} while (0)
1076#define print_buffer_trace(bh) do {} while (0)
1077#define BUFFER_TRACE(bh, info) do {} while (0)
1078#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1079#define JBUFFER_TRACE(jh, info) do {} while (0)
1080
1081#endif /* __KERNEL__ */
1082
1083#endif /* CONFIG_JBD || CONFIG_JBD_MODULE || !__KERNEL__ */
1084
1085/*
1086 * Compatibility no-ops which allow the kernel to compile without CONFIG_JBD
1087 * go here.
1088 */
1089
1090#if defined(__KERNEL__) && !(defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE))
1091
1092#define J_ASSERT(expr) do {} while (0)
1093#define J_ASSERT_BH(bh, expr) do {} while (0)
1094#define buffer_jbd(bh) 0
1095#define journal_buffer_journal_lru(bh) 0
1096
1097#endif /* defined(__KERNEL__) && !defined(CONFIG_JBD) */
1098#endif /* _LINUX_JBD_H */