blob: f3aa59cb675d84cc92308bf038661826679cf5f0 [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#define journal_oom_retry 1
36
37/*
Kazuo Moriwaka0cbfdc82009-03-03 13:01:23 +010038 * Define JBD_PARANOID_IOFAIL to cause a kernel BUG() if ext3 finds
Linus Torvalds1da177e2005-04-16 15:20:36 -070039 * certain classes of error which can occur due to failed IOs. Under
40 * normal use we want ext3 to continue after such errors, because
41 * hardware _can_ fail, but for debugging purposes when running tests on
42 * known-good hardware we may want to trap these errors.
43 */
44#undef JBD_PARANOID_IOFAIL
45
46/*
47 * The default maximum commit age, in seconds.
48 */
49#define JBD_DEFAULT_MAX_COMMIT_AGE 5
50
51#ifdef CONFIG_JBD_DEBUG
52/*
53 * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
54 * consistency checks. By default we don't do this unless
55 * CONFIG_JBD_DEBUG is on.
56 */
57#define JBD_EXPENSIVE_CHECKING
Jose R. Santosc2a91592007-10-18 23:39:22 -070058extern u8 journal_enable_debug;
Linus Torvalds1da177e2005-04-16 15:20:36 -070059
60#define jbd_debug(n, f, a...) \
61 do { \
62 if ((n) <= journal_enable_debug) { \
63 printk (KERN_DEBUG "(%s, %d): %s: ", \
Harvey Harrisond5c003b2008-10-15 22:01:24 -070064 __FILE__, __LINE__, __func__); \
Dave Kleikampe9ad5622006-09-27 01:49:35 -070065 printk (f, ## a); \
Linus Torvalds1da177e2005-04-16 15:20:36 -070066 } \
67 } while (0)
68#else
69#define jbd_debug(f, a...) /**/
70#endif
71
Mingming Caoc089d492007-10-16 18:38:25 -040072static inline void *jbd_alloc(size_t size, gfp_t flags)
73{
74 return (void *)__get_free_pages(flags, get_order(size));
75}
76
77static inline void jbd_free(void *ptr, size_t size)
78{
79 free_pages((unsigned long)ptr, get_order(size));
80};
81
Linus Torvalds1da177e2005-04-16 15:20:36 -070082#define JFS_MIN_JOURNAL_BLOCKS 1024
83
Linus Torvalds1da177e2005-04-16 15:20:36 -070084
85/**
86 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
87 *
88 * All filesystem modifications made by the process go
89 * through this handle. Recursive operations (such as quota operations)
90 * are gathered into a single update.
91 *
92 * The buffer credits field is used to account for journaled buffers
93 * being modified by the running process. To ensure that there is
94 * enough log space for all outstanding operations, we need to limit the
95 * number of outstanding buffers possible at any time. When the
96 * operation completes, any buffer credits not used are credited back to
97 * the transaction, so that at all times we know how many buffers the
Mingming Caoae6ddcc2006-09-27 01:49:27 -070098 * outstanding updates on a transaction might possibly touch.
99 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 * This is an opaque datatype.
101 **/
102typedef struct handle_s handle_t; /* Atomic operation type */
103
104
105/**
106 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
107 *
108 * journal_t is linked to from the fs superblock structure.
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700109 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110 * We use the journal_t to keep track of all outstanding transaction
111 * activity on the filesystem, and to manage the state of the log
112 * writing process.
113 *
114 * This is an opaque datatype.
115 **/
116typedef struct journal_s journal_t; /* Journal control structure */
117#endif
118
119/*
120 * Internal structures used by the logging mechanism:
121 */
122
123#define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
124
125/*
126 * On-disk structures
127 */
128
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700129/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130 * Descriptor block types:
131 */
132
133#define JFS_DESCRIPTOR_BLOCK 1
134#define JFS_COMMIT_BLOCK 2
135#define JFS_SUPERBLOCK_V1 3
136#define JFS_SUPERBLOCK_V2 4
137#define JFS_REVOKE_BLOCK 5
138
139/*
140 * Standard header for all descriptor blocks:
141 */
142typedef struct journal_header_s
143{
144 __be32 h_magic;
145 __be32 h_blocktype;
146 __be32 h_sequence;
147} journal_header_t;
148
149
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700150/*
151 * The block tag: used to describe a single buffer in the journal
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 */
153typedef struct journal_block_tag_s
154{
155 __be32 t_blocknr; /* The on-disk block number */
156 __be32 t_flags; /* See below */
157} journal_block_tag_t;
158
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700159/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700160 * The revoke descriptor: used on disk to describe a series of blocks to
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700161 * be revoked from the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 */
163typedef struct journal_revoke_header_s
164{
165 journal_header_t r_header;
166 __be32 r_count; /* Count of bytes used in the block */
167} journal_revoke_header_t;
168
169
170/* Definitions for the journal tag flags word: */
171#define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */
172#define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */
173#define JFS_FLAG_DELETED 4 /* block deleted by this transaction */
174#define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
175
176
177/*
178 * The journal superblock. All fields are in big-endian byte order.
179 */
180typedef struct journal_superblock_s
181{
182/* 0x0000 */
183 journal_header_t s_header;
184
185/* 0x000C */
186 /* Static information describing the journal */
187 __be32 s_blocksize; /* journal device blocksize */
188 __be32 s_maxlen; /* total blocks in journal file */
189 __be32 s_first; /* first block of log information */
190
191/* 0x0018 */
192 /* Dynamic information describing the current state of the log */
193 __be32 s_sequence; /* first commit ID expected in log */
194 __be32 s_start; /* blocknr of start of log */
195
196/* 0x0020 */
197 /* Error value, as set by journal_abort(). */
198 __be32 s_errno;
199
200/* 0x0024 */
201 /* Remaining fields are only valid in a version-2 superblock */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700202 __be32 s_feature_compat; /* compatible feature set */
203 __be32 s_feature_incompat; /* incompatible feature set */
204 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700205/* 0x0030 */
206 __u8 s_uuid[16]; /* 128-bit uuid for journal */
207
208/* 0x0040 */
209 __be32 s_nr_users; /* Nr of filesystems sharing log */
210
211 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
212
213/* 0x0048 */
214 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
215 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
216
217/* 0x0050 */
218 __u32 s_padding[44];
219
220/* 0x0100 */
221 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
222/* 0x0400 */
223} journal_superblock_t;
224
225#define JFS_HAS_COMPAT_FEATURE(j,mask) \
226 ((j)->j_format_version >= 2 && \
227 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
228#define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \
229 ((j)->j_format_version >= 2 && \
230 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
231#define JFS_HAS_INCOMPAT_FEATURE(j,mask) \
232 ((j)->j_format_version >= 2 && \
233 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
234
235#define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001
236
237/* Features known to this kernel version: */
238#define JFS_KNOWN_COMPAT_FEATURES 0
239#define JFS_KNOWN_ROCOMPAT_FEATURES 0
240#define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE
241
242#ifdef __KERNEL__
243
244#include <linux/fs.h>
245#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
Chris Snook1c099242007-10-18 23:39:21 -0700247#define J_ASSERT(assert) BUG_ON(!(assert))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
250#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252#if defined(JBD_PARANOID_IOFAIL)
253#define J_EXPECT(expr, why...) J_ASSERT(expr)
254#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
255#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
256#else
257#define __journal_expect(expr, why...) \
258 ({ \
259 int val = (expr); \
260 if (!val) { \
261 printk(KERN_ERR \
262 "EXT3-fs unexpected failure: %s;\n",# expr); \
263 printk(KERN_ERR why "\n"); \
264 } \
265 val; \
266 })
267#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
268#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
269#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
270#endif
271
272enum jbd_state_bits {
273 BH_JBD /* Has an attached ext3 journal_head */
274 = BH_PrivateStart,
275 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
276 BH_Freed, /* Has been freed (truncated) */
277 BH_Revoked, /* Has been revoked from the log */
278 BH_RevokeValid, /* Revoked flag is valid */
279 BH_JBDDirty, /* Is dirty but journaled */
280 BH_State, /* Pins most journal_head state */
281 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
282 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
283};
284
285BUFFER_FNS(JBD, jbd)
286BUFFER_FNS(JWrite, jwrite)
287BUFFER_FNS(JBDDirty, jbddirty)
288TAS_BUFFER_FNS(JBDDirty, jbddirty)
289BUFFER_FNS(Revoked, revoked)
290TAS_BUFFER_FNS(Revoked, revoked)
291BUFFER_FNS(RevokeValid, revokevalid)
292TAS_BUFFER_FNS(RevokeValid, revokevalid)
293BUFFER_FNS(Freed, freed)
294
295static inline struct buffer_head *jh2bh(struct journal_head *jh)
296{
297 return jh->b_bh;
298}
299
300static inline struct journal_head *bh2jh(struct buffer_head *bh)
301{
302 return bh->b_private;
303}
304
305static inline void jbd_lock_bh_state(struct buffer_head *bh)
306{
307 bit_spin_lock(BH_State, &bh->b_state);
308}
309
310static inline int jbd_trylock_bh_state(struct buffer_head *bh)
311{
312 return bit_spin_trylock(BH_State, &bh->b_state);
313}
314
315static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
316{
317 return bit_spin_is_locked(BH_State, &bh->b_state);
318}
319
320static inline void jbd_unlock_bh_state(struct buffer_head *bh)
321{
322 bit_spin_unlock(BH_State, &bh->b_state);
323}
324
325static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
326{
327 bit_spin_lock(BH_JournalHead, &bh->b_state);
328}
329
330static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
331{
332 bit_spin_unlock(BH_JournalHead, &bh->b_state);
333}
334
335struct jbd_revoke_table_s;
336
337/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700338 * struct handle_s - this is the concrete type associated with handle_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339 * @h_transaction: Which compound transaction is this update a part of?
340 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
341 * @h_ref: Reference count on this handle
342 * @h_err: Field for caller's use to track errors through large fs operations
343 * @h_sync: flag for sync-on-close
344 * @h_jdata: flag to force data journaling
345 * @h_aborted: flag indicating fatal error on handle
Randy Dunlape6716b82007-10-15 17:30:19 -0700346 * @h_lockdep_map: lockdep info for debugging lock problems
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700348struct handle_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349{
350 /* Which compound transaction is this update a part of? */
351 transaction_t *h_transaction;
352
353 /* Number of remaining buffers we are allowed to dirty: */
354 int h_buffer_credits;
355
356 /* Reference count on this handle */
357 int h_ref;
358
359 /* Field for caller's use to track errors through large fs */
360 /* operations */
361 int h_err;
362
363 /* Flags [no locking] */
364 unsigned int h_sync: 1; /* sync-on-close */
365 unsigned int h_jdata: 1; /* force data journaling */
366 unsigned int h_aborted: 1; /* fatal error on handle */
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +0200367
368#ifdef CONFIG_DEBUG_LOCK_ALLOC
369 struct lockdep_map h_lockdep_map;
370#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371};
372
373
374/* The transaction_t type is the guts of the journaling mechanism. It
375 * tracks a compound transaction through its various states:
376 *
377 * RUNNING: accepting new updates
378 * LOCKED: Updates still running but we don't accept new ones
379 * RUNDOWN: Updates are tidying up but have finished requesting
380 * new buffers to modify (state not used for now)
381 * FLUSH: All updates complete, but we are still writing to disk
382 * COMMIT: All data on disk, writing commit record
383 * FINISHED: We still have to keep the transaction for checkpointing.
384 *
385 * The transaction keeps track of all of the buffers modified by a
386 * running transaction, and all of the buffers committed but not yet
387 * flushed to home for finished transactions.
388 */
389
390/*
391 * Lock ranking:
392 *
393 * j_list_lock
394 * ->jbd_lock_bh_journal_head() (This is "innermost")
395 *
396 * j_state_lock
397 * ->jbd_lock_bh_state()
398 *
399 * jbd_lock_bh_state()
400 * ->j_list_lock
401 *
402 * j_state_lock
403 * ->t_handle_lock
404 *
405 * j_state_lock
406 * ->j_list_lock (journal_unmap_buffer)
407 *
408 */
409
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700410struct transaction_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411{
412 /* Pointer to the journal for this transaction. [no locking] */
413 journal_t *t_journal;
414
415 /* Sequence number for this transaction [no locking] */
416 tid_t t_tid;
417
418 /*
419 * Transaction's current state
420 * [no locking - only kjournald alters this]
Jan Karad4beaf42007-12-04 23:45:27 -0800421 * [j_list_lock] guards transition of a transaction into T_FINISHED
422 * state and subsequent call of __journal_drop_transaction()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 * FIXME: needs barriers
424 * KLUDGE: [use j_state_lock]
425 */
426 enum {
427 T_RUNNING,
428 T_LOCKED,
429 T_RUNDOWN,
430 T_FLUSH,
431 T_COMMIT,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700432 T_FINISHED
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 } t_state;
434
435 /*
436 * Where in the log does this transaction's commit start? [no locking]
437 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200438 unsigned int t_log_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439
440 /* Number of buffers on the t_buffers list [j_list_lock] */
441 int t_nr_buffers;
442
443 /*
444 * Doubly-linked circular list of all buffers reserved but not yet
445 * modified by this transaction [j_list_lock]
446 */
447 struct journal_head *t_reserved_list;
448
449 /*
450 * Doubly-linked circular list of all buffers under writeout during
451 * commit [j_list_lock]
452 */
453 struct journal_head *t_locked_list;
454
455 /*
456 * Doubly-linked circular list of all metadata buffers owned by this
457 * transaction [j_list_lock]
458 */
459 struct journal_head *t_buffers;
460
461 /*
462 * Doubly-linked circular list of all data buffers still to be
463 * flushed before this transaction can be committed [j_list_lock]
464 */
465 struct journal_head *t_sync_datalist;
466
467 /*
468 * Doubly-linked circular list of all forget buffers (superseded
469 * buffers which we can un-checkpoint once this transaction commits)
470 * [j_list_lock]
471 */
472 struct journal_head *t_forget;
473
474 /*
475 * Doubly-linked circular list of all buffers still to be flushed before
476 * this transaction can be checkpointed. [j_list_lock]
477 */
478 struct journal_head *t_checkpoint_list;
479
480 /*
Jan Kara78ce89c2006-06-23 02:06:05 -0700481 * Doubly-linked circular list of all buffers submitted for IO while
482 * checkpointing. [j_list_lock]
483 */
484 struct journal_head *t_checkpoint_io_list;
485
486 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 * Doubly-linked circular list of temporary buffers currently undergoing
488 * IO in the log [j_list_lock]
489 */
490 struct journal_head *t_iobuf_list;
491
492 /*
493 * Doubly-linked circular list of metadata buffers being shadowed by log
494 * IO. The IO buffers on the iobuf list and the shadow buffers on this
495 * list match each other one for one at all times. [j_list_lock]
496 */
497 struct journal_head *t_shadow_list;
498
499 /*
500 * Doubly-linked circular list of control buffers being written to the
501 * log. [j_list_lock]
502 */
503 struct journal_head *t_log_list;
504
505 /*
506 * Protects info related to handles
507 */
508 spinlock_t t_handle_lock;
509
510 /*
511 * Number of outstanding updates running on this transaction
512 * [t_handle_lock]
513 */
514 int t_updates;
515
516 /*
517 * Number of buffers reserved for use by all handles in this transaction
518 * handle but not yet modified. [t_handle_lock]
519 */
520 int t_outstanding_credits;
521
522 /*
523 * Forward and backward links for the circular list of all transactions
524 * awaiting checkpoint. [j_list_lock]
525 */
526 transaction_t *t_cpnext, *t_cpprev;
527
528 /*
529 * When will the transaction expire (become due for commit), in jiffies?
530 * [no locking]
531 */
532 unsigned long t_expires;
533
534 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800535 * When this transaction started, in nanoseconds [no locking]
536 */
537 ktime_t t_start_time;
538
539 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 * How many handles used this transaction? [t_handle_lock]
541 */
542 int t_handle_count;
543
Theodore Ts'o512a0042009-03-27 22:14:27 -0400544 /*
545 * This transaction is being forced and some process is
546 * waiting for it to finish.
547 */
H Hartley Sweeten4fefce92009-09-22 16:48:02 -0700548 unsigned int t_synchronous_commit:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549};
550
551/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700552 * struct journal_s - this is the concrete type associated with journal_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 * @j_flags: General journaling state flags
554 * @j_errno: Is there an outstanding uncleared error on the journal (from a
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700555 * prior abort)?
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 * @j_sb_buffer: First part of superblock buffer
557 * @j_superblock: Second part of superblock buffer
558 * @j_format_version: Version of the superblock format
559 * @j_state_lock: Protect the various scalars in the journal
560 * @j_barrier_count: Number of processes waiting to create a barrier lock
561 * @j_barrier: The barrier lock itself
562 * @j_running_transaction: The current running transaction..
563 * @j_committing_transaction: the transaction we are pushing to disk
564 * @j_checkpoint_transactions: a linked circular list of all transactions
565 * waiting for checkpointing
566 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
567 * to start committing, or for a barrier lock to be released
568 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700569 * @j_wait_done_commit: Wait queue for waiting for commit to complete
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570 * @j_wait_checkpoint: Wait queue to trigger checkpointing
571 * @j_wait_commit: Wait queue to trigger commit
572 * @j_wait_updates: Wait queue to wait for updates to complete
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800573 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574 * @j_head: Journal head - identifies the first unused block in the journal
575 * @j_tail: Journal tail - identifies the oldest still-used block in the
576 * journal.
577 * @j_free: Journal free - how many free blocks are there in the journal?
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700578 * @j_first: The block number of the first usable block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 * @j_last: The block number one beyond the last usable block
580 * @j_dev: Device where we store the journal
581 * @j_blocksize: blocksize for the location where we store the journal.
582 * @j_blk_offset: starting block offset for into the device where we store the
583 * journal
584 * @j_fs_dev: Device which holds the client fs. For internal journal this will
585 * be equal to j_dev
586 * @j_maxlen: Total maximum capacity of the journal region on disk.
587 * @j_list_lock: Protects the buffer lists and internal buffer state.
588 * @j_inode: Optional inode where we store the journal. If present, all journal
589 * block numbers are mapped into this inode via bmap().
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700590 * @j_tail_sequence: Sequence number of the oldest transaction in the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 * @j_transaction_sequence: Sequence number of the next transaction to grant
592 * @j_commit_sequence: Sequence number of the most recently committed
593 * transaction
594 * @j_commit_request: Sequence number of the most recent transaction wanting
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700595 * commit
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 * @j_uuid: Uuid of client object.
597 * @j_task: Pointer to the current commit thread for this journal
598 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
599 * single compound commit transaction
600 * @j_commit_interval: What is the maximum transaction lifetime before we begin
601 * a commit?
602 * @j_commit_timer: The timer used to wakeup the commit thread
603 * @j_revoke_lock: Protect the revoke table
604 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
605 * current transaction.
606 * @j_revoke_table: alternate revoke tables for j_revoke
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800607 * @j_wbuf: array of buffer_heads for journal_commit_transaction
608 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
609 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800610 * @j_last_sync_writer: most recent pid which did a synchronous write
Randy Dunlap3eabdb72009-01-15 13:51:01 -0800611 * @j_average_commit_time: the average amount of time in nanoseconds it
612 * takes to commit a transaction to the disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 * @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 */
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800645 struct mutex j_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
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 */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700687 struct mutex j_checkpoint_mutex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
689 /*
690 * Journal head: identifies the first unused block in the journal.
691 * [j_state_lock]
692 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200693 unsigned int j_head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
695 /*
696 * Journal tail: identifies the oldest still-used block in the journal.
697 * [j_state_lock]
698 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200699 unsigned int j_tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700700
701 /*
702 * Journal free: how many free blocks are there in the journal?
703 * [j_state_lock]
704 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200705 unsigned int j_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
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 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200711 unsigned int j_first;
712 unsigned int j_last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713
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;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200720 unsigned int j_blk_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721
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: */
Andrew Mortone3df1892006-03-25 03:06:53 -0800786 struct timer_list j_commit_timer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
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
Josef Bacikf420d4d2009-01-07 18:07:24 -0800802 /*
803 * this is the pid of the last person to run a synchronous operation
804 * through the journal.
805 */
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800806 pid_t j_last_sync_writer;
807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800809 * the average amount of time in nanoseconds it takes to commit a
810 * transaction to the disk. [j_state_lock]
811 */
812 u64 j_average_commit_time;
813
814 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 * An opaque pointer to fs-private information. ext3 puts its
816 * superblock pointer here
817 */
818 void *j_private;
819};
820
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700821/*
822 * Journal flag definitions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 */
824#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
825#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
826#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
827#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
828#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
829#define JFS_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700830#define JFS_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
831 * data write error in ordered
832 * mode */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700834/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 * Function declarations for the journaling transaction and buffer
836 * management
837 */
838
839/* Filing buffers */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840extern void journal_unfile_buffer(journal_t *, struct journal_head *);
841extern void __journal_unfile_buffer(struct journal_head *);
842extern void __journal_refile_buffer(struct journal_head *);
843extern void journal_refile_buffer(journal_t *, struct journal_head *);
844extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
845extern void __journal_free_buffer(struct journal_head *bh);
846extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
847extern void __journal_clean_data_list(transaction_t *transaction);
848
849/* Log buffer allocation */
850extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200851int journal_next_log_block(journal_t *, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852
853/* Commit management */
854extern void journal_commit_transaction(journal_t *);
855
856/* Checkpoint list management */
857int __journal_clean_checkpoint_list(journal_t *journal);
Jan Kara78ce89c2006-06-23 02:06:05 -0700858int __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
860
861/* Buffer IO */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700862extern int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863journal_write_metadata_buffer(transaction_t *transaction,
864 struct journal_head *jh_in,
865 struct journal_head **jh_out,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200866 unsigned int blocknr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
868/* Transaction locking */
869extern void __wait_on_journal (journal_t *);
870
871/*
872 * Journal locking.
873 *
874 * We need to lock the journal during transaction state changes so that nobody
875 * ever tries to take a handle on the running transaction while we are in the
876 * middle of moving it to the commit phase. j_state_lock does this.
877 *
878 * Note that the locking is completely interrupt unsafe. We never touch
879 * journal structures from interrupts.
880 */
881
882static inline handle_t *journal_current_handle(void)
883{
884 return current->journal_info;
885}
886
887/* The journaling code user interface:
888 *
889 * Create and destroy handles
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700890 * Register buffer modifications against the current transaction.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 */
892
893extern handle_t *journal_start(journal_t *, int nblocks);
894extern int journal_restart (handle_t *, int nblocks);
895extern int journal_extend (handle_t *, int nblocks);
896extern int journal_get_write_access(handle_t *, struct buffer_head *);
897extern int journal_get_create_access (handle_t *, struct buffer_head *);
898extern int journal_get_undo_access(handle_t *, struct buffer_head *);
899extern int journal_dirty_data (handle_t *, struct buffer_head *);
900extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
901extern void journal_release_buffer (handle_t *, struct buffer_head *);
902extern int journal_forget (handle_t *, struct buffer_head *);
903extern void journal_sync_buffer (struct buffer_head *);
NeilBrown2ff28e22006-03-26 01:37:18 -0800904extern void journal_invalidatepage(journal_t *,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 struct page *, unsigned long);
Al Viro27496a82005-10-21 03:20:48 -0400906extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907extern int journal_stop(handle_t *);
908extern int journal_flush (journal_t *);
909extern void journal_lock_updates (journal_t *);
910extern void journal_unlock_updates (journal_t *);
911
912extern journal_t * journal_init_dev(struct block_device *bdev,
913 struct block_device *fs_dev,
914 int start, int len, int bsize);
915extern journal_t * journal_init_inode (struct inode *);
916extern int journal_update_format (journal_t *);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700917extern int journal_check_used_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700919extern int journal_check_available_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700921extern int journal_set_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 (journal_t *, unsigned long, unsigned long, unsigned long);
923extern int journal_create (journal_t *);
924extern int journal_load (journal_t *journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -0700925extern int journal_destroy (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926extern int journal_recover (journal_t *journal);
927extern int journal_wipe (journal_t *, int);
928extern int journal_skip_recovery (journal_t *);
929extern void journal_update_superblock (journal_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930extern void journal_abort (journal_t *, int);
931extern int journal_errno (journal_t *);
932extern void journal_ack_err (journal_t *);
933extern int journal_clear_err (journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200934extern int journal_bmap(journal_t *, unsigned int, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935extern int journal_force_commit(journal_t *);
936
937/*
938 * journal_head management
939 */
940struct journal_head *journal_add_journal_head(struct buffer_head *bh);
941struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
942void journal_remove_journal_head(struct buffer_head *bh);
943void journal_put_journal_head(struct journal_head *jh);
944
945/*
946 * handle management
947 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800948extern struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949
Al Virodd0fc662005-10-07 07:46:04 +0100950static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951{
952 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
953}
954
955static inline void jbd_free_handle(handle_t *handle)
956{
957 kmem_cache_free(jbd_handle_cache, handle);
958}
959
960/* Primary revoke support */
961#define JOURNAL_REVOKE_DEFAULT_HASH 256
962extern int journal_init_revoke(journal_t *, int);
963extern void journal_destroy_revoke_caches(void);
964extern int journal_init_revoke_caches(void);
965
966extern void journal_destroy_revoke(journal_t *);
967extern int journal_revoke (handle_t *,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200968 unsigned int, struct buffer_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969extern int journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o38d726d2009-04-14 10:10:47 -0400970extern void journal_write_revoke_records(journal_t *,
971 transaction_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
973/* Recovery revoke support */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200974extern int journal_set_revoke(journal_t *, unsigned int, tid_t);
975extern int journal_test_revoke(journal_t *, unsigned int, tid_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976extern void journal_clear_revoke(journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977extern void journal_switch_revoke_table(journal_t *journal);
978
979/*
980 * The log thread user interface:
981 *
982 * Request space in the current transaction, and force transaction commit
983 * transitions on demand.
984 */
985
986int __log_space_left(journal_t *); /* Called with journal locked */
987int log_start_commit(journal_t *journal, tid_t tid);
988int __log_start_commit(journal_t *journal, tid_t tid);
989int journal_start_commit(journal_t *journal, tid_t *tid);
990int journal_force_commit_nested(journal_t *journal);
991int log_wait_commit(journal_t *journal, tid_t tid);
992int log_do_checkpoint(journal_t *journal);
993
994void __log_wait_for_space(journal_t *journal);
995extern void __journal_drop_transaction(journal_t *, transaction_t *);
996extern int cleanup_journal_tail(journal_t *);
997
998/* Debugging code only: */
999
1000#define jbd_ENOSYS() \
1001do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -07001002 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 current->state = TASK_UNINTERRUPTIBLE; \
1004 schedule(); \
1005} while (1)
1006
1007/*
1008 * is_journal_abort
1009 *
1010 * Simple test wrapper function to test the JFS_ABORT state flag. This
1011 * bit, when set, indicates that we have had a fatal error somewhere,
1012 * either inside the journaling layer or indicated to us by the client
1013 * (eg. ext3), and that we and should not commit any further
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001014 * transactions.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001015 */
1016
1017static inline int is_journal_aborted(journal_t *journal)
1018{
1019 return journal->j_flags & JFS_ABORT;
1020}
1021
1022static inline int is_handle_aborted(handle_t *handle)
1023{
1024 if (handle->h_aborted)
1025 return 1;
1026 return is_journal_aborted(handle->h_transaction->t_journal);
1027}
1028
1029static inline void journal_abort_handle(handle_t *handle)
1030{
1031 handle->h_aborted = 1;
1032}
1033
1034#endif /* __KERNEL__ */
1035
1036/* Comparison functions for transaction IDs: perform comparisons using
1037 * modulo arithmetic so that they work over sequence number wraps. */
1038
1039static inline int tid_gt(tid_t x, tid_t y)
1040{
1041 int difference = (x - y);
1042 return (difference > 0);
1043}
1044
1045static inline int tid_geq(tid_t x, tid_t y)
1046{
1047 int difference = (x - y);
1048 return (difference >= 0);
1049}
1050
1051extern int journal_blocks_per_page(struct inode *inode);
1052
1053/*
1054 * Return the minimum number of blocks which must be free in the journal
1055 * before a new transaction may be started. Must be called under j_state_lock.
1056 */
1057static inline int jbd_space_needed(journal_t *journal)
1058{
1059 int nblocks = journal->j_max_transaction_buffers;
1060 if (journal->j_committing_transaction)
1061 nblocks += journal->j_committing_transaction->
1062 t_outstanding_credits;
1063 return nblocks;
1064}
1065
1066/*
1067 * Definitions which augment the buffer_head layer
1068 */
1069
1070/* journaling buffer types */
1071#define BJ_None 0 /* Not journaled */
1072#define BJ_SyncData 1 /* Normal data: flush before commit */
1073#define BJ_Metadata 2 /* Normal journaled metadata */
1074#define BJ_Forget 3 /* Buffer superseded by this transaction */
1075#define BJ_IO 4 /* Buffer is for temporary IO use */
1076#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1077#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1078#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1079#define BJ_Locked 8 /* Locked for I/O during commit */
1080#define BJ_Types 9
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001081
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082extern int jbd_blocks_per_page(struct inode *inode);
1083
1084#ifdef __KERNEL__
1085
1086#define buffer_trace_init(bh) do {} while (0)
1087#define print_buffer_fields(bh) do {} while (0)
1088#define print_buffer_trace(bh) do {} while (0)
1089#define BUFFER_TRACE(bh, info) do {} while (0)
1090#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1091#define JBUFFER_TRACE(jh, info) do {} while (0)
1092
1093#endif /* __KERNEL__ */
1094
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095#endif /* _LINUX_JBD_H */