blob: 700a93b79189b52584d0f9208503fd912fc9de73 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/include/linux/jbd.h
Mingming Caoae6ddcc2006-09-27 01:49:27 -07003 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 * 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>
Arjan van de Ven2c68ee72006-03-23 03:00:35 -080031#include <linux/mutex.h>
Andrew Mortone3df1892006-03-25 03:06:53 -080032#include <linux/timer.h>
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +020033#include <linux/lockdep.h>
Andrew Mortone3df1892006-03-25 03:06:53 -080034
Linus Torvalds1da177e2005-04-16 15:20:36 -070035#include <asm/semaphore.h>
36#endif
37
38#define journal_oom_retry 1
39
40/*
41 * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
42 * certain classes of error which can occur due to failed IOs. Under
43 * normal use we want ext3 to continue after such errors, because
44 * hardware _can_ fail, but for debugging purposes when running tests on
45 * known-good hardware we may want to trap these errors.
46 */
47#undef JBD_PARANOID_IOFAIL
48
49/*
50 * The default maximum commit age, in seconds.
51 */
52#define JBD_DEFAULT_MAX_COMMIT_AGE 5
53
54#ifdef CONFIG_JBD_DEBUG
55/*
56 * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
57 * consistency checks. By default we don't do this unless
58 * CONFIG_JBD_DEBUG is on.
59 */
60#define JBD_EXPENSIVE_CHECKING
61extern int journal_enable_debug;
62
63#define jbd_debug(n, f, a...) \
64 do { \
65 if ((n) <= journal_enable_debug) { \
66 printk (KERN_DEBUG "(%s, %d): %s: ", \
67 __FILE__, __LINE__, __FUNCTION__); \
Dave Kleikampe9ad5622006-09-27 01:49:35 -070068 printk (f, ## a); \
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 } \
70 } while (0)
71#else
72#define jbd_debug(f, a...) /**/
73#endif
74
Al Viro27496a82005-10-21 03:20:48 -040075extern void * __jbd_kmalloc (const char *where, size_t size, gfp_t flags, int retry);
Badari Pulavartyea817392006-08-27 01:23:52 -070076extern void * jbd_slab_alloc(size_t size, gfp_t flags);
77extern void jbd_slab_free(void *ptr, size_t size);
78
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#define jbd_kmalloc(size, flags) \
80 __jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
81#define jbd_rep_kmalloc(size, flags) \
82 __jbd_kmalloc(__FUNCTION__, (size), (flags), 1)
83
84#define JFS_MIN_JOURNAL_BLOCKS 1024
85
86#ifdef __KERNEL__
87
88/**
89 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
90 *
91 * All filesystem modifications made by the process go
92 * through this handle. Recursive operations (such as quota operations)
93 * are gathered into a single update.
94 *
95 * The buffer credits field is used to account for journaled buffers
96 * being modified by the running process. To ensure that there is
97 * enough log space for all outstanding operations, we need to limit the
98 * number of outstanding buffers possible at any time. When the
99 * operation completes, any buffer credits not used are credited back to
100 * the transaction, so that at all times we know how many buffers the
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700101 * outstanding updates on a transaction might possibly touch.
102 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103 * This is an opaque datatype.
104 **/
105typedef struct handle_s handle_t; /* Atomic operation type */
106
107
108/**
109 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
110 *
111 * journal_t is linked to from the fs superblock structure.
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700112 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113 * We use the journal_t to keep track of all outstanding transaction
114 * activity on the filesystem, and to manage the state of the log
115 * writing process.
116 *
117 * This is an opaque datatype.
118 **/
119typedef struct journal_s journal_t; /* Journal control structure */
120#endif
121
122/*
123 * Internal structures used by the logging mechanism:
124 */
125
126#define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
127
128/*
129 * On-disk structures
130 */
131
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700132/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133 * Descriptor block types:
134 */
135
136#define JFS_DESCRIPTOR_BLOCK 1
137#define JFS_COMMIT_BLOCK 2
138#define JFS_SUPERBLOCK_V1 3
139#define JFS_SUPERBLOCK_V2 4
140#define JFS_REVOKE_BLOCK 5
141
142/*
143 * Standard header for all descriptor blocks:
144 */
145typedef struct journal_header_s
146{
147 __be32 h_magic;
148 __be32 h_blocktype;
149 __be32 h_sequence;
150} journal_header_t;
151
152
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700153/*
154 * The block tag: used to describe a single buffer in the journal
Linus Torvalds1da177e2005-04-16 15:20:36 -0700155 */
156typedef struct journal_block_tag_s
157{
158 __be32 t_blocknr; /* The on-disk block number */
159 __be32 t_flags; /* See below */
160} journal_block_tag_t;
161
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700162/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 * The revoke descriptor: used on disk to describe a series of blocks to
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700164 * be revoked from the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165 */
166typedef struct journal_revoke_header_s
167{
168 journal_header_t r_header;
169 __be32 r_count; /* Count of bytes used in the block */
170} journal_revoke_header_t;
171
172
173/* Definitions for the journal tag flags word: */
174#define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */
175#define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */
176#define JFS_FLAG_DELETED 4 /* block deleted by this transaction */
177#define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
178
179
180/*
181 * The journal superblock. All fields are in big-endian byte order.
182 */
183typedef struct journal_superblock_s
184{
185/* 0x0000 */
186 journal_header_t s_header;
187
188/* 0x000C */
189 /* Static information describing the journal */
190 __be32 s_blocksize; /* journal device blocksize */
191 __be32 s_maxlen; /* total blocks in journal file */
192 __be32 s_first; /* first block of log information */
193
194/* 0x0018 */
195 /* Dynamic information describing the current state of the log */
196 __be32 s_sequence; /* first commit ID expected in log */
197 __be32 s_start; /* blocknr of start of log */
198
199/* 0x0020 */
200 /* Error value, as set by journal_abort(). */
201 __be32 s_errno;
202
203/* 0x0024 */
204 /* Remaining fields are only valid in a version-2 superblock */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700205 __be32 s_feature_compat; /* compatible feature set */
206 __be32 s_feature_incompat; /* incompatible feature set */
207 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208/* 0x0030 */
209 __u8 s_uuid[16]; /* 128-bit uuid for journal */
210
211/* 0x0040 */
212 __be32 s_nr_users; /* Nr of filesystems sharing log */
213
214 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
215
216/* 0x0048 */
217 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
218 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
219
220/* 0x0050 */
221 __u32 s_padding[44];
222
223/* 0x0100 */
224 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
225/* 0x0400 */
226} journal_superblock_t;
227
228#define JFS_HAS_COMPAT_FEATURE(j,mask) \
229 ((j)->j_format_version >= 2 && \
230 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
231#define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \
232 ((j)->j_format_version >= 2 && \
233 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
234#define JFS_HAS_INCOMPAT_FEATURE(j,mask) \
235 ((j)->j_format_version >= 2 && \
236 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
237
238#define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001
239
240/* Features known to this kernel version: */
241#define JFS_KNOWN_COMPAT_FEATURES 0
242#define JFS_KNOWN_ROCOMPAT_FEATURES 0
243#define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE
244
245#ifdef __KERNEL__
246
247#include <linux/fs.h>
248#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
250#define JBD_ASSERTIONS
251#ifdef JBD_ASSERTIONS
252#define J_ASSERT(assert) \
253do { \
254 if (!(assert)) { \
255 printk (KERN_EMERG \
256 "Assertion failure in %s() at %s:%d: \"%s\"\n", \
257 __FUNCTION__, __FILE__, __LINE__, # assert); \
258 BUG(); \
259 } \
260} while (0)
261
262#if defined(CONFIG_BUFFER_DEBUG)
263void buffer_assertion_failure(struct buffer_head *bh);
264#define J_ASSERT_BH(bh, expr) \
265 do { \
266 if (!(expr)) \
267 buffer_assertion_failure(bh); \
268 J_ASSERT(expr); \
269 } while (0)
270#define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
271#else
272#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
273#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
274#endif
275
276#else
277#define J_ASSERT(assert) do { } while (0)
278#endif /* JBD_ASSERTIONS */
279
280#if defined(JBD_PARANOID_IOFAIL)
281#define J_EXPECT(expr, why...) J_ASSERT(expr)
282#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
283#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
284#else
285#define __journal_expect(expr, why...) \
286 ({ \
287 int val = (expr); \
288 if (!val) { \
289 printk(KERN_ERR \
290 "EXT3-fs unexpected failure: %s;\n",# expr); \
291 printk(KERN_ERR why "\n"); \
292 } \
293 val; \
294 })
295#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
296#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
297#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
298#endif
299
300enum jbd_state_bits {
301 BH_JBD /* Has an attached ext3 journal_head */
302 = BH_PrivateStart,
303 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
304 BH_Freed, /* Has been freed (truncated) */
305 BH_Revoked, /* Has been revoked from the log */
306 BH_RevokeValid, /* Revoked flag is valid */
307 BH_JBDDirty, /* Is dirty but journaled */
308 BH_State, /* Pins most journal_head state */
309 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
310 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
311};
312
313BUFFER_FNS(JBD, jbd)
314BUFFER_FNS(JWrite, jwrite)
315BUFFER_FNS(JBDDirty, jbddirty)
316TAS_BUFFER_FNS(JBDDirty, jbddirty)
317BUFFER_FNS(Revoked, revoked)
318TAS_BUFFER_FNS(Revoked, revoked)
319BUFFER_FNS(RevokeValid, revokevalid)
320TAS_BUFFER_FNS(RevokeValid, revokevalid)
321BUFFER_FNS(Freed, freed)
322
323static inline struct buffer_head *jh2bh(struct journal_head *jh)
324{
325 return jh->b_bh;
326}
327
328static inline struct journal_head *bh2jh(struct buffer_head *bh)
329{
330 return bh->b_private;
331}
332
333static inline void jbd_lock_bh_state(struct buffer_head *bh)
334{
335 bit_spin_lock(BH_State, &bh->b_state);
336}
337
338static inline int jbd_trylock_bh_state(struct buffer_head *bh)
339{
340 return bit_spin_trylock(BH_State, &bh->b_state);
341}
342
343static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
344{
345 return bit_spin_is_locked(BH_State, &bh->b_state);
346}
347
348static inline void jbd_unlock_bh_state(struct buffer_head *bh)
349{
350 bit_spin_unlock(BH_State, &bh->b_state);
351}
352
353static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
354{
355 bit_spin_lock(BH_JournalHead, &bh->b_state);
356}
357
358static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
359{
360 bit_spin_unlock(BH_JournalHead, &bh->b_state);
361}
362
363struct jbd_revoke_table_s;
364
365/**
366 * struct handle_s - The handle_s type is the concrete type associated with
367 * handle_t.
368 * @h_transaction: Which compound transaction is this update a part of?
369 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
370 * @h_ref: Reference count on this handle
371 * @h_err: Field for caller's use to track errors through large fs operations
372 * @h_sync: flag for sync-on-close
373 * @h_jdata: flag to force data journaling
374 * @h_aborted: flag indicating fatal error on handle
375 **/
376
377/* Docbook can't yet cope with the bit fields, but will leave the documentation
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700378 * in so it can be fixed later.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 */
380
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700381struct handle_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382{
383 /* Which compound transaction is this update a part of? */
384 transaction_t *h_transaction;
385
386 /* Number of remaining buffers we are allowed to dirty: */
387 int h_buffer_credits;
388
389 /* Reference count on this handle */
390 int h_ref;
391
392 /* Field for caller's use to track errors through large fs */
393 /* operations */
394 int h_err;
395
396 /* Flags [no locking] */
397 unsigned int h_sync: 1; /* sync-on-close */
398 unsigned int h_jdata: 1; /* force data journaling */
399 unsigned int h_aborted: 1; /* fatal error on handle */
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +0200400
401#ifdef CONFIG_DEBUG_LOCK_ALLOC
402 struct lockdep_map h_lockdep_map;
403#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404};
405
406
407/* The transaction_t type is the guts of the journaling mechanism. It
408 * tracks a compound transaction through its various states:
409 *
410 * RUNNING: accepting new updates
411 * LOCKED: Updates still running but we don't accept new ones
412 * RUNDOWN: Updates are tidying up but have finished requesting
413 * new buffers to modify (state not used for now)
414 * FLUSH: All updates complete, but we are still writing to disk
415 * COMMIT: All data on disk, writing commit record
416 * FINISHED: We still have to keep the transaction for checkpointing.
417 *
418 * The transaction keeps track of all of the buffers modified by a
419 * running transaction, and all of the buffers committed but not yet
420 * flushed to home for finished transactions.
421 */
422
423/*
424 * Lock ranking:
425 *
426 * j_list_lock
427 * ->jbd_lock_bh_journal_head() (This is "innermost")
428 *
429 * j_state_lock
430 * ->jbd_lock_bh_state()
431 *
432 * jbd_lock_bh_state()
433 * ->j_list_lock
434 *
435 * j_state_lock
436 * ->t_handle_lock
437 *
438 * j_state_lock
439 * ->j_list_lock (journal_unmap_buffer)
440 *
441 */
442
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700443struct transaction_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444{
445 /* Pointer to the journal for this transaction. [no locking] */
446 journal_t *t_journal;
447
448 /* Sequence number for this transaction [no locking] */
449 tid_t t_tid;
450
451 /*
452 * Transaction's current state
453 * [no locking - only kjournald alters this]
454 * FIXME: needs barriers
455 * KLUDGE: [use j_state_lock]
456 */
457 enum {
458 T_RUNNING,
459 T_LOCKED,
460 T_RUNDOWN,
461 T_FLUSH,
462 T_COMMIT,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700463 T_FINISHED
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 } t_state;
465
466 /*
467 * Where in the log does this transaction's commit start? [no locking]
468 */
469 unsigned long t_log_start;
470
471 /* Number of buffers on the t_buffers list [j_list_lock] */
472 int t_nr_buffers;
473
474 /*
475 * Doubly-linked circular list of all buffers reserved but not yet
476 * modified by this transaction [j_list_lock]
477 */
478 struct journal_head *t_reserved_list;
479
480 /*
481 * Doubly-linked circular list of all buffers under writeout during
482 * commit [j_list_lock]
483 */
484 struct journal_head *t_locked_list;
485
486 /*
487 * Doubly-linked circular list of all metadata buffers owned by this
488 * transaction [j_list_lock]
489 */
490 struct journal_head *t_buffers;
491
492 /*
493 * Doubly-linked circular list of all data buffers still to be
494 * flushed before this transaction can be committed [j_list_lock]
495 */
496 struct journal_head *t_sync_datalist;
497
498 /*
499 * Doubly-linked circular list of all forget buffers (superseded
500 * buffers which we can un-checkpoint once this transaction commits)
501 * [j_list_lock]
502 */
503 struct journal_head *t_forget;
504
505 /*
506 * Doubly-linked circular list of all buffers still to be flushed before
507 * this transaction can be checkpointed. [j_list_lock]
508 */
509 struct journal_head *t_checkpoint_list;
510
511 /*
Jan Kara78ce89c2006-06-23 02:06:05 -0700512 * Doubly-linked circular list of all buffers submitted for IO while
513 * checkpointing. [j_list_lock]
514 */
515 struct journal_head *t_checkpoint_io_list;
516
517 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 * Doubly-linked circular list of temporary buffers currently undergoing
519 * IO in the log [j_list_lock]
520 */
521 struct journal_head *t_iobuf_list;
522
523 /*
524 * Doubly-linked circular list of metadata buffers being shadowed by log
525 * IO. The IO buffers on the iobuf list and the shadow buffers on this
526 * list match each other one for one at all times. [j_list_lock]
527 */
528 struct journal_head *t_shadow_list;
529
530 /*
531 * Doubly-linked circular list of control buffers being written to the
532 * log. [j_list_lock]
533 */
534 struct journal_head *t_log_list;
535
536 /*
537 * Protects info related to handles
538 */
539 spinlock_t t_handle_lock;
540
541 /*
542 * Number of outstanding updates running on this transaction
543 * [t_handle_lock]
544 */
545 int t_updates;
546
547 /*
548 * Number of buffers reserved for use by all handles in this transaction
549 * handle but not yet modified. [t_handle_lock]
550 */
551 int t_outstanding_credits;
552
553 /*
554 * Forward and backward links for the circular list of all transactions
555 * awaiting checkpoint. [j_list_lock]
556 */
557 transaction_t *t_cpnext, *t_cpprev;
558
559 /*
560 * When will the transaction expire (become due for commit), in jiffies?
561 * [no locking]
562 */
563 unsigned long t_expires;
564
565 /*
566 * How many handles used this transaction? [t_handle_lock]
567 */
568 int t_handle_count;
569
570};
571
572/**
573 * struct journal_s - The journal_s type is the concrete type associated with
574 * journal_t.
575 * @j_flags: General journaling state flags
576 * @j_errno: Is there an outstanding uncleared error on the journal (from a
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700577 * prior abort)?
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 * @j_sb_buffer: First part of superblock buffer
579 * @j_superblock: Second part of superblock buffer
580 * @j_format_version: Version of the superblock format
581 * @j_state_lock: Protect the various scalars in the journal
582 * @j_barrier_count: Number of processes waiting to create a barrier lock
583 * @j_barrier: The barrier lock itself
584 * @j_running_transaction: The current running transaction..
585 * @j_committing_transaction: the transaction we are pushing to disk
586 * @j_checkpoint_transactions: a linked circular list of all transactions
587 * waiting for checkpointing
588 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
589 * to start committing, or for a barrier lock to be released
590 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700591 * @j_wait_done_commit: Wait queue for waiting for commit to complete
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 * @j_wait_checkpoint: Wait queue to trigger checkpointing
593 * @j_wait_commit: Wait queue to trigger commit
594 * @j_wait_updates: Wait queue to wait for updates to complete
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800595 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 * @j_head: Journal head - identifies the first unused block in the journal
597 * @j_tail: Journal tail - identifies the oldest still-used block in the
598 * journal.
599 * @j_free: Journal free - how many free blocks are there in the journal?
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700600 * @j_first: The block number of the first usable block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 * @j_last: The block number one beyond the last usable block
602 * @j_dev: Device where we store the journal
603 * @j_blocksize: blocksize for the location where we store the journal.
604 * @j_blk_offset: starting block offset for into the device where we store the
605 * journal
606 * @j_fs_dev: Device which holds the client fs. For internal journal this will
607 * be equal to j_dev
608 * @j_maxlen: Total maximum capacity of the journal region on disk.
609 * @j_list_lock: Protects the buffer lists and internal buffer state.
610 * @j_inode: Optional inode where we store the journal. If present, all journal
611 * block numbers are mapped into this inode via bmap().
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700612 * @j_tail_sequence: Sequence number of the oldest transaction in the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 * @j_transaction_sequence: Sequence number of the next transaction to grant
614 * @j_commit_sequence: Sequence number of the most recently committed
615 * transaction
616 * @j_commit_request: Sequence number of the most recent transaction wanting
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700617 * commit
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 * @j_uuid: Uuid of client object.
619 * @j_task: Pointer to the current commit thread for this journal
620 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
621 * single compound commit transaction
622 * @j_commit_interval: What is the maximum transaction lifetime before we begin
623 * a commit?
624 * @j_commit_timer: The timer used to wakeup the commit thread
625 * @j_revoke_lock: Protect the revoke table
626 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
627 * current transaction.
628 * @j_revoke_table: alternate revoke tables for j_revoke
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800629 * @j_wbuf: array of buffer_heads for journal_commit_transaction
630 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
631 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800632 * @j_last_sync_writer: most recent pid which did a synchronous write
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 * @j_private: An opaque pointer to fs-private information.
634 */
635
636struct journal_s
637{
638 /* General journaling state flags [j_state_lock] */
639 unsigned long j_flags;
640
641 /*
642 * Is there an outstanding uncleared error on the journal (from a prior
643 * abort)? [j_state_lock]
644 */
645 int j_errno;
646
647 /* The superblock buffer */
648 struct buffer_head *j_sb_buffer;
649 journal_superblock_t *j_superblock;
650
651 /* Version of the superblock format */
652 int j_format_version;
653
654 /*
655 * Protect the various scalars in the journal
656 */
657 spinlock_t j_state_lock;
658
659 /*
660 * Number of processes waiting to create a barrier lock [j_state_lock]
661 */
662 int j_barrier_count;
663
664 /* The barrier lock itself */
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800665 struct mutex j_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666
667 /*
668 * Transactions: The current running transaction...
669 * [j_state_lock] [caller holding open handle]
670 */
671 transaction_t *j_running_transaction;
672
673 /*
674 * the transaction we are pushing to disk
675 * [j_state_lock] [caller holding open handle]
676 */
677 transaction_t *j_committing_transaction;
678
679 /*
680 * ... and a linked circular list of all transactions waiting for
681 * checkpointing. [j_list_lock]
682 */
683 transaction_t *j_checkpoint_transactions;
684
685 /*
686 * Wait queue for waiting for a locked transaction to start committing,
687 * or for a barrier lock to be released
688 */
689 wait_queue_head_t j_wait_transaction_locked;
690
691 /* Wait queue for waiting for checkpointing to complete */
692 wait_queue_head_t j_wait_logspace;
693
694 /* Wait queue for waiting for commit to complete */
695 wait_queue_head_t j_wait_done_commit;
696
697 /* Wait queue to trigger checkpointing */
698 wait_queue_head_t j_wait_checkpoint;
699
700 /* Wait queue to trigger commit */
701 wait_queue_head_t j_wait_commit;
702
703 /* Wait queue to wait for updates to complete */
704 wait_queue_head_t j_wait_updates;
705
706 /* Semaphore for locking against concurrent checkpoints */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700707 struct mutex j_checkpoint_mutex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
709 /*
710 * Journal head: identifies the first unused block in the journal.
711 * [j_state_lock]
712 */
713 unsigned long j_head;
714
715 /*
716 * Journal tail: identifies the oldest still-used block in the journal.
717 * [j_state_lock]
718 */
719 unsigned long j_tail;
720
721 /*
722 * Journal free: how many free blocks are there in the journal?
723 * [j_state_lock]
724 */
725 unsigned long j_free;
726
727 /*
728 * Journal start and end: the block numbers of the first usable block
729 * and one beyond the last usable block in the journal. [j_state_lock]
730 */
731 unsigned long j_first;
732 unsigned long j_last;
733
734 /*
735 * Device, blocksize and starting block offset for the location where we
736 * store the journal.
737 */
738 struct block_device *j_dev;
739 int j_blocksize;
Eric Sandeen37ed3222006-09-27 01:49:31 -0700740 unsigned long j_blk_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
742 /*
743 * Device which holds the client fs. For internal journal this will be
744 * equal to j_dev.
745 */
746 struct block_device *j_fs_dev;
747
748 /* Total maximum capacity of the journal region on disk. */
749 unsigned int j_maxlen;
750
751 /*
752 * Protects the buffer lists and internal buffer state.
753 */
754 spinlock_t j_list_lock;
755
756 /* Optional inode where we store the journal. If present, all */
757 /* journal block numbers are mapped into this inode via */
758 /* bmap(). */
759 struct inode *j_inode;
760
761 /*
762 * Sequence number of the oldest transaction in the log [j_state_lock]
763 */
764 tid_t j_tail_sequence;
765
766 /*
767 * Sequence number of the next transaction to grant [j_state_lock]
768 */
769 tid_t j_transaction_sequence;
770
771 /*
772 * Sequence number of the most recently committed transaction
773 * [j_state_lock].
774 */
775 tid_t j_commit_sequence;
776
777 /*
778 * Sequence number of the most recent transaction wanting commit
779 * [j_state_lock]
780 */
781 tid_t j_commit_request;
782
783 /*
784 * Journal uuid: identifies the object (filesystem, LVM volume etc)
785 * backed by this journal. This will eventually be replaced by an array
786 * of uuids, allowing us to index multiple devices within a single
787 * journal and to perform atomic updates across them.
788 */
789 __u8 j_uuid[16];
790
791 /* Pointer to the current commit thread for this journal */
792 struct task_struct *j_task;
793
794 /*
795 * Maximum number of metadata buffers to allow in a single compound
796 * commit transaction
797 */
798 int j_max_transaction_buffers;
799
800 /*
801 * What is the maximum transaction lifetime before we begin a commit?
802 */
803 unsigned long j_commit_interval;
804
805 /* The timer used to wakeup the commit thread: */
Andrew Mortone3df1892006-03-25 03:06:53 -0800806 struct timer_list j_commit_timer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700807
808 /*
809 * The revoke table: maintains the list of revoked blocks in the
810 * current transaction. [j_revoke_lock]
811 */
812 spinlock_t j_revoke_lock;
813 struct jbd_revoke_table_s *j_revoke;
814 struct jbd_revoke_table_s *j_revoke_table[2];
815
816 /*
817 * array of bhs for journal_commit_transaction
818 */
819 struct buffer_head **j_wbuf;
820 int j_wbufsize;
821
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800822 pid_t j_last_sync_writer;
823
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 /*
825 * An opaque pointer to fs-private information. ext3 puts its
826 * superblock pointer here
827 */
828 void *j_private;
829};
830
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700831/*
832 * Journal flag definitions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833 */
834#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
835#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
836#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
837#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
838#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
839#define JFS_BARRIER 0x020 /* Use IDE barriers */
840
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700841/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 * Function declarations for the journaling transaction and buffer
843 * management
844 */
845
846/* Filing buffers */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847extern void journal_unfile_buffer(journal_t *, struct journal_head *);
848extern void __journal_unfile_buffer(struct journal_head *);
849extern void __journal_refile_buffer(struct journal_head *);
850extern void journal_refile_buffer(journal_t *, struct journal_head *);
851extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
852extern void __journal_free_buffer(struct journal_head *bh);
853extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
854extern void __journal_clean_data_list(transaction_t *transaction);
855
856/* Log buffer allocation */
857extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
858int journal_next_log_block(journal_t *, unsigned long *);
859
860/* Commit management */
861extern void journal_commit_transaction(journal_t *);
862
863/* Checkpoint list management */
864int __journal_clean_checkpoint_list(journal_t *journal);
Jan Kara78ce89c2006-06-23 02:06:05 -0700865int __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
867
868/* Buffer IO */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700869extern int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870journal_write_metadata_buffer(transaction_t *transaction,
871 struct journal_head *jh_in,
872 struct journal_head **jh_out,
Eric Sandeen37ed3222006-09-27 01:49:31 -0700873 unsigned long blocknr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874
875/* Transaction locking */
876extern void __wait_on_journal (journal_t *);
877
878/*
879 * Journal locking.
880 *
881 * We need to lock the journal during transaction state changes so that nobody
882 * ever tries to take a handle on the running transaction while we are in the
883 * middle of moving it to the commit phase. j_state_lock does this.
884 *
885 * Note that the locking is completely interrupt unsafe. We never touch
886 * journal structures from interrupts.
887 */
888
889static inline handle_t *journal_current_handle(void)
890{
891 return current->journal_info;
892}
893
894/* The journaling code user interface:
895 *
896 * Create and destroy handles
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700897 * Register buffer modifications against the current transaction.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 */
899
900extern handle_t *journal_start(journal_t *, int nblocks);
901extern int journal_restart (handle_t *, int nblocks);
902extern int journal_extend (handle_t *, int nblocks);
903extern int journal_get_write_access(handle_t *, struct buffer_head *);
904extern int journal_get_create_access (handle_t *, struct buffer_head *);
905extern int journal_get_undo_access(handle_t *, struct buffer_head *);
906extern int journal_dirty_data (handle_t *, struct buffer_head *);
907extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
908extern void journal_release_buffer (handle_t *, struct buffer_head *);
909extern int journal_forget (handle_t *, struct buffer_head *);
910extern void journal_sync_buffer (struct buffer_head *);
NeilBrown2ff28e22006-03-26 01:37:18 -0800911extern void journal_invalidatepage(journal_t *,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 struct page *, unsigned long);
Al Viro27496a82005-10-21 03:20:48 -0400913extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914extern int journal_stop(handle_t *);
915extern int journal_flush (journal_t *);
916extern void journal_lock_updates (journal_t *);
917extern void journal_unlock_updates (journal_t *);
918
919extern journal_t * journal_init_dev(struct block_device *bdev,
920 struct block_device *fs_dev,
921 int start, int len, int bsize);
922extern journal_t * journal_init_inode (struct inode *);
923extern int journal_update_format (journal_t *);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700924extern int journal_check_used_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700926extern int journal_check_available_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700928extern int journal_set_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929 (journal_t *, unsigned long, unsigned long, unsigned long);
930extern int journal_create (journal_t *);
931extern int journal_load (journal_t *journal);
932extern void journal_destroy (journal_t *);
933extern int journal_recover (journal_t *journal);
934extern int journal_wipe (journal_t *, int);
935extern int journal_skip_recovery (journal_t *);
936extern void journal_update_superblock (journal_t *, int);
937extern void __journal_abort_hard (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938extern void journal_abort (journal_t *, int);
939extern int journal_errno (journal_t *);
940extern void journal_ack_err (journal_t *);
941extern int journal_clear_err (journal_t *);
942extern int journal_bmap(journal_t *, unsigned long, unsigned long *);
943extern int journal_force_commit(journal_t *);
944
945/*
946 * journal_head management
947 */
948struct journal_head *journal_add_journal_head(struct buffer_head *bh);
949struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
950void journal_remove_journal_head(struct buffer_head *bh);
951void journal_put_journal_head(struct journal_head *jh);
952
953/*
954 * handle management
955 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800956extern struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957
Al Virodd0fc662005-10-07 07:46:04 +0100958static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959{
960 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
961}
962
963static inline void jbd_free_handle(handle_t *handle)
964{
965 kmem_cache_free(jbd_handle_cache, handle);
966}
967
968/* Primary revoke support */
969#define JOURNAL_REVOKE_DEFAULT_HASH 256
970extern int journal_init_revoke(journal_t *, int);
971extern void journal_destroy_revoke_caches(void);
972extern int journal_init_revoke_caches(void);
973
974extern void journal_destroy_revoke(journal_t *);
975extern int journal_revoke (handle_t *,
976 unsigned long, struct buffer_head *);
977extern int journal_cancel_revoke(handle_t *, struct journal_head *);
978extern void journal_write_revoke_records(journal_t *, transaction_t *);
979
980/* Recovery revoke support */
981extern int journal_set_revoke(journal_t *, unsigned long, tid_t);
982extern int journal_test_revoke(journal_t *, unsigned long, tid_t);
983extern void journal_clear_revoke(journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984extern void journal_switch_revoke_table(journal_t *journal);
985
986/*
987 * The log thread user interface:
988 *
989 * Request space in the current transaction, and force transaction commit
990 * transitions on demand.
991 */
992
993int __log_space_left(journal_t *); /* Called with journal locked */
994int log_start_commit(journal_t *journal, tid_t tid);
995int __log_start_commit(journal_t *journal, tid_t tid);
996int journal_start_commit(journal_t *journal, tid_t *tid);
997int journal_force_commit_nested(journal_t *journal);
998int log_wait_commit(journal_t *journal, tid_t tid);
999int log_do_checkpoint(journal_t *journal);
1000
1001void __log_wait_for_space(journal_t *journal);
1002extern void __journal_drop_transaction(journal_t *, transaction_t *);
1003extern int cleanup_journal_tail(journal_t *);
1004
1005/* Debugging code only: */
1006
1007#define jbd_ENOSYS() \
1008do { \
1009 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1010 current->state = TASK_UNINTERRUPTIBLE; \
1011 schedule(); \
1012} while (1)
1013
1014/*
1015 * is_journal_abort
1016 *
1017 * Simple test wrapper function to test the JFS_ABORT state flag. This
1018 * bit, when set, indicates that we have had a fatal error somewhere,
1019 * either inside the journaling layer or indicated to us by the client
1020 * (eg. ext3), and that we and should not commit any further
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001021 * transactions.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 */
1023
1024static inline int is_journal_aborted(journal_t *journal)
1025{
1026 return journal->j_flags & JFS_ABORT;
1027}
1028
1029static inline int is_handle_aborted(handle_t *handle)
1030{
1031 if (handle->h_aborted)
1032 return 1;
1033 return is_journal_aborted(handle->h_transaction->t_journal);
1034}
1035
1036static inline void journal_abort_handle(handle_t *handle)
1037{
1038 handle->h_aborted = 1;
1039}
1040
1041#endif /* __KERNEL__ */
1042
1043/* Comparison functions for transaction IDs: perform comparisons using
1044 * modulo arithmetic so that they work over sequence number wraps. */
1045
1046static inline int tid_gt(tid_t x, tid_t y)
1047{
1048 int difference = (x - y);
1049 return (difference > 0);
1050}
1051
1052static inline int tid_geq(tid_t x, tid_t y)
1053{
1054 int difference = (x - y);
1055 return (difference >= 0);
1056}
1057
1058extern int journal_blocks_per_page(struct inode *inode);
1059
1060/*
1061 * Return the minimum number of blocks which must be free in the journal
1062 * before a new transaction may be started. Must be called under j_state_lock.
1063 */
1064static inline int jbd_space_needed(journal_t *journal)
1065{
1066 int nblocks = journal->j_max_transaction_buffers;
1067 if (journal->j_committing_transaction)
1068 nblocks += journal->j_committing_transaction->
1069 t_outstanding_credits;
1070 return nblocks;
1071}
1072
1073/*
1074 * Definitions which augment the buffer_head layer
1075 */
1076
1077/* journaling buffer types */
1078#define BJ_None 0 /* Not journaled */
1079#define BJ_SyncData 1 /* Normal data: flush before commit */
1080#define BJ_Metadata 2 /* Normal journaled metadata */
1081#define BJ_Forget 3 /* Buffer superseded by this transaction */
1082#define BJ_IO 4 /* Buffer is for temporary IO use */
1083#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1084#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1085#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1086#define BJ_Locked 8 /* Locked for I/O during commit */
1087#define BJ_Types 9
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001088
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089extern int jbd_blocks_per_page(struct inode *inode);
1090
1091#ifdef __KERNEL__
1092
1093#define buffer_trace_init(bh) do {} while (0)
1094#define print_buffer_fields(bh) do {} while (0)
1095#define print_buffer_trace(bh) do {} while (0)
1096#define BUFFER_TRACE(bh, info) do {} while (0)
1097#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1098#define JBUFFER_TRACE(jh, info) do {} while (0)
1099
1100#endif /* __KERNEL__ */
1101
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102#endif /* _LINUX_JBD_H */