blob: 64246dce5663079b5f69afe9eabc05efc3e1229e [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/*
38 * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
39 * 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
249#if defined(CONFIG_BUFFER_DEBUG)
250void buffer_assertion_failure(struct buffer_head *bh);
251#define J_ASSERT_BH(bh, expr) \
252 do { \
253 if (!(expr)) \
254 buffer_assertion_failure(bh); \
255 J_ASSERT(expr); \
256 } while (0)
257#define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
258#else
259#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
260#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
261#endif
262
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263#if defined(JBD_PARANOID_IOFAIL)
264#define J_EXPECT(expr, why...) J_ASSERT(expr)
265#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
266#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
267#else
268#define __journal_expect(expr, why...) \
269 ({ \
270 int val = (expr); \
271 if (!val) { \
272 printk(KERN_ERR \
273 "EXT3-fs unexpected failure: %s;\n",# expr); \
274 printk(KERN_ERR why "\n"); \
275 } \
276 val; \
277 })
278#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
279#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
280#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
281#endif
282
283enum jbd_state_bits {
284 BH_JBD /* Has an attached ext3 journal_head */
285 = BH_PrivateStart,
286 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
287 BH_Freed, /* Has been freed (truncated) */
288 BH_Revoked, /* Has been revoked from the log */
289 BH_RevokeValid, /* Revoked flag is valid */
290 BH_JBDDirty, /* Is dirty but journaled */
291 BH_State, /* Pins most journal_head state */
292 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
293 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
294};
295
296BUFFER_FNS(JBD, jbd)
297BUFFER_FNS(JWrite, jwrite)
298BUFFER_FNS(JBDDirty, jbddirty)
299TAS_BUFFER_FNS(JBDDirty, jbddirty)
300BUFFER_FNS(Revoked, revoked)
301TAS_BUFFER_FNS(Revoked, revoked)
302BUFFER_FNS(RevokeValid, revokevalid)
303TAS_BUFFER_FNS(RevokeValid, revokevalid)
304BUFFER_FNS(Freed, freed)
305
306static inline struct buffer_head *jh2bh(struct journal_head *jh)
307{
308 return jh->b_bh;
309}
310
311static inline struct journal_head *bh2jh(struct buffer_head *bh)
312{
313 return bh->b_private;
314}
315
316static inline void jbd_lock_bh_state(struct buffer_head *bh)
317{
318 bit_spin_lock(BH_State, &bh->b_state);
319}
320
321static inline int jbd_trylock_bh_state(struct buffer_head *bh)
322{
323 return bit_spin_trylock(BH_State, &bh->b_state);
324}
325
326static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
327{
328 return bit_spin_is_locked(BH_State, &bh->b_state);
329}
330
331static inline void jbd_unlock_bh_state(struct buffer_head *bh)
332{
333 bit_spin_unlock(BH_State, &bh->b_state);
334}
335
336static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
337{
338 bit_spin_lock(BH_JournalHead, &bh->b_state);
339}
340
341static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
342{
343 bit_spin_unlock(BH_JournalHead, &bh->b_state);
344}
345
346struct jbd_revoke_table_s;
347
348/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700349 * struct handle_s - this is the concrete type associated with handle_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 * @h_transaction: Which compound transaction is this update a part of?
351 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
352 * @h_ref: Reference count on this handle
353 * @h_err: Field for caller's use to track errors through large fs operations
354 * @h_sync: flag for sync-on-close
355 * @h_jdata: flag to force data journaling
356 * @h_aborted: flag indicating fatal error on handle
Randy Dunlape6716b82007-10-15 17:30:19 -0700357 * @h_lockdep_map: lockdep info for debugging lock problems
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700359struct handle_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360{
361 /* Which compound transaction is this update a part of? */
362 transaction_t *h_transaction;
363
364 /* Number of remaining buffers we are allowed to dirty: */
365 int h_buffer_credits;
366
367 /* Reference count on this handle */
368 int h_ref;
369
370 /* Field for caller's use to track errors through large fs */
371 /* operations */
372 int h_err;
373
374 /* Flags [no locking] */
375 unsigned int h_sync: 1; /* sync-on-close */
376 unsigned int h_jdata: 1; /* force data journaling */
377 unsigned int h_aborted: 1; /* fatal error on handle */
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +0200378
379#ifdef CONFIG_DEBUG_LOCK_ALLOC
380 struct lockdep_map h_lockdep_map;
381#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382};
383
384
385/* The transaction_t type is the guts of the journaling mechanism. It
386 * tracks a compound transaction through its various states:
387 *
388 * RUNNING: accepting new updates
389 * LOCKED: Updates still running but we don't accept new ones
390 * RUNDOWN: Updates are tidying up but have finished requesting
391 * new buffers to modify (state not used for now)
392 * FLUSH: All updates complete, but we are still writing to disk
393 * COMMIT: All data on disk, writing commit record
394 * FINISHED: We still have to keep the transaction for checkpointing.
395 *
396 * The transaction keeps track of all of the buffers modified by a
397 * running transaction, and all of the buffers committed but not yet
398 * flushed to home for finished transactions.
399 */
400
401/*
402 * Lock ranking:
403 *
404 * j_list_lock
405 * ->jbd_lock_bh_journal_head() (This is "innermost")
406 *
407 * j_state_lock
408 * ->jbd_lock_bh_state()
409 *
410 * jbd_lock_bh_state()
411 * ->j_list_lock
412 *
413 * j_state_lock
414 * ->t_handle_lock
415 *
416 * j_state_lock
417 * ->j_list_lock (journal_unmap_buffer)
418 *
419 */
420
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700421struct transaction_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422{
423 /* Pointer to the journal for this transaction. [no locking] */
424 journal_t *t_journal;
425
426 /* Sequence number for this transaction [no locking] */
427 tid_t t_tid;
428
429 /*
430 * Transaction's current state
431 * [no locking - only kjournald alters this]
Jan Karad4beaf42007-12-04 23:45:27 -0800432 * [j_list_lock] guards transition of a transaction into T_FINISHED
433 * state and subsequent call of __journal_drop_transaction()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 * FIXME: needs barriers
435 * KLUDGE: [use j_state_lock]
436 */
437 enum {
438 T_RUNNING,
439 T_LOCKED,
440 T_RUNDOWN,
441 T_FLUSH,
442 T_COMMIT,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700443 T_FINISHED
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 } t_state;
445
446 /*
447 * Where in the log does this transaction's commit start? [no locking]
448 */
449 unsigned long t_log_start;
450
451 /* Number of buffers on the t_buffers list [j_list_lock] */
452 int t_nr_buffers;
453
454 /*
455 * Doubly-linked circular list of all buffers reserved but not yet
456 * modified by this transaction [j_list_lock]
457 */
458 struct journal_head *t_reserved_list;
459
460 /*
461 * Doubly-linked circular list of all buffers under writeout during
462 * commit [j_list_lock]
463 */
464 struct journal_head *t_locked_list;
465
466 /*
467 * Doubly-linked circular list of all metadata buffers owned by this
468 * transaction [j_list_lock]
469 */
470 struct journal_head *t_buffers;
471
472 /*
473 * Doubly-linked circular list of all data buffers still to be
474 * flushed before this transaction can be committed [j_list_lock]
475 */
476 struct journal_head *t_sync_datalist;
477
478 /*
479 * Doubly-linked circular list of all forget buffers (superseded
480 * buffers which we can un-checkpoint once this transaction commits)
481 * [j_list_lock]
482 */
483 struct journal_head *t_forget;
484
485 /*
486 * Doubly-linked circular list of all buffers still to be flushed before
487 * this transaction can be checkpointed. [j_list_lock]
488 */
489 struct journal_head *t_checkpoint_list;
490
491 /*
Jan Kara78ce89c2006-06-23 02:06:05 -0700492 * Doubly-linked circular list of all buffers submitted for IO while
493 * checkpointing. [j_list_lock]
494 */
495 struct journal_head *t_checkpoint_io_list;
496
497 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 * Doubly-linked circular list of temporary buffers currently undergoing
499 * IO in the log [j_list_lock]
500 */
501 struct journal_head *t_iobuf_list;
502
503 /*
504 * Doubly-linked circular list of metadata buffers being shadowed by log
505 * IO. The IO buffers on the iobuf list and the shadow buffers on this
506 * list match each other one for one at all times. [j_list_lock]
507 */
508 struct journal_head *t_shadow_list;
509
510 /*
511 * Doubly-linked circular list of control buffers being written to the
512 * log. [j_list_lock]
513 */
514 struct journal_head *t_log_list;
515
516 /*
517 * Protects info related to handles
518 */
519 spinlock_t t_handle_lock;
520
521 /*
522 * Number of outstanding updates running on this transaction
523 * [t_handle_lock]
524 */
525 int t_updates;
526
527 /*
528 * Number of buffers reserved for use by all handles in this transaction
529 * handle but not yet modified. [t_handle_lock]
530 */
531 int t_outstanding_credits;
532
533 /*
534 * Forward and backward links for the circular list of all transactions
535 * awaiting checkpoint. [j_list_lock]
536 */
537 transaction_t *t_cpnext, *t_cpprev;
538
539 /*
540 * When will the transaction expire (become due for commit), in jiffies?
541 * [no locking]
542 */
543 unsigned long t_expires;
544
545 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800546 * When this transaction started, in nanoseconds [no locking]
547 */
548 ktime_t t_start_time;
549
550 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 * How many handles used this transaction? [t_handle_lock]
552 */
553 int t_handle_count;
554
555};
556
557/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700558 * struct journal_s - this is the concrete type associated with journal_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 * @j_flags: General journaling state flags
560 * @j_errno: Is there an outstanding uncleared error on the journal (from a
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700561 * prior abort)?
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562 * @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
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700575 * @j_wait_done_commit: Wait queue for waiting for commit to complete
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 * @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
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800579 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 * @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?
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700584 * @j_first: The block number of the first usable block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 * @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().
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700596 * @j_tail_sequence: Sequence number of the oldest transaction in the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 * @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
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700601 * commit
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 * @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
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800613 * @j_wbuf: array of buffer_heads for journal_commit_transaction
614 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
615 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800616 * @j_last_sync_writer: most recent pid which did a synchronous write
Randy Dunlap3eabdb72009-01-15 13:51:01 -0800617 * @j_average_commit_time: the average amount of time in nanoseconds it
618 * takes to commit a transaction to the disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 * @j_private: An opaque pointer to fs-private information.
620 */
621
622struct journal_s
623{
624 /* General journaling state flags [j_state_lock] */
625 unsigned long j_flags;
626
627 /*
628 * Is there an outstanding uncleared error on the journal (from a prior
629 * abort)? [j_state_lock]
630 */
631 int j_errno;
632
633 /* The superblock buffer */
634 struct buffer_head *j_sb_buffer;
635 journal_superblock_t *j_superblock;
636
637 /* Version of the superblock format */
638 int j_format_version;
639
640 /*
641 * Protect the various scalars in the journal
642 */
643 spinlock_t j_state_lock;
644
645 /*
646 * Number of processes waiting to create a barrier lock [j_state_lock]
647 */
648 int j_barrier_count;
649
650 /* The barrier lock itself */
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800651 struct mutex j_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
653 /*
654 * Transactions: The current running transaction...
655 * [j_state_lock] [caller holding open handle]
656 */
657 transaction_t *j_running_transaction;
658
659 /*
660 * the transaction we are pushing to disk
661 * [j_state_lock] [caller holding open handle]
662 */
663 transaction_t *j_committing_transaction;
664
665 /*
666 * ... and a linked circular list of all transactions waiting for
667 * checkpointing. [j_list_lock]
668 */
669 transaction_t *j_checkpoint_transactions;
670
671 /*
672 * Wait queue for waiting for a locked transaction to start committing,
673 * or for a barrier lock to be released
674 */
675 wait_queue_head_t j_wait_transaction_locked;
676
677 /* Wait queue for waiting for checkpointing to complete */
678 wait_queue_head_t j_wait_logspace;
679
680 /* Wait queue for waiting for commit to complete */
681 wait_queue_head_t j_wait_done_commit;
682
683 /* Wait queue to trigger checkpointing */
684 wait_queue_head_t j_wait_checkpoint;
685
686 /* Wait queue to trigger commit */
687 wait_queue_head_t j_wait_commit;
688
689 /* Wait queue to wait for updates to complete */
690 wait_queue_head_t j_wait_updates;
691
692 /* Semaphore for locking against concurrent checkpoints */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700693 struct mutex j_checkpoint_mutex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
695 /*
696 * Journal head: identifies the first unused block in the journal.
697 * [j_state_lock]
698 */
699 unsigned long j_head;
700
701 /*
702 * Journal tail: identifies the oldest still-used block in the journal.
703 * [j_state_lock]
704 */
705 unsigned long j_tail;
706
707 /*
708 * Journal free: how many free blocks are there in the journal?
709 * [j_state_lock]
710 */
711 unsigned long j_free;
712
713 /*
714 * Journal start and end: the block numbers of the first usable block
715 * and one beyond the last usable block in the journal. [j_state_lock]
716 */
717 unsigned long j_first;
718 unsigned long j_last;
719
720 /*
721 * Device, blocksize and starting block offset for the location where we
722 * store the journal.
723 */
724 struct block_device *j_dev;
725 int j_blocksize;
Eric Sandeen37ed3222006-09-27 01:49:31 -0700726 unsigned long j_blk_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
728 /*
729 * Device which holds the client fs. For internal journal this will be
730 * equal to j_dev.
731 */
732 struct block_device *j_fs_dev;
733
734 /* Total maximum capacity of the journal region on disk. */
735 unsigned int j_maxlen;
736
737 /*
738 * Protects the buffer lists and internal buffer state.
739 */
740 spinlock_t j_list_lock;
741
742 /* Optional inode where we store the journal. If present, all */
743 /* journal block numbers are mapped into this inode via */
744 /* bmap(). */
745 struct inode *j_inode;
746
747 /*
748 * Sequence number of the oldest transaction in the log [j_state_lock]
749 */
750 tid_t j_tail_sequence;
751
752 /*
753 * Sequence number of the next transaction to grant [j_state_lock]
754 */
755 tid_t j_transaction_sequence;
756
757 /*
758 * Sequence number of the most recently committed transaction
759 * [j_state_lock].
760 */
761 tid_t j_commit_sequence;
762
763 /*
764 * Sequence number of the most recent transaction wanting commit
765 * [j_state_lock]
766 */
767 tid_t j_commit_request;
768
769 /*
770 * Journal uuid: identifies the object (filesystem, LVM volume etc)
771 * backed by this journal. This will eventually be replaced by an array
772 * of uuids, allowing us to index multiple devices within a single
773 * journal and to perform atomic updates across them.
774 */
775 __u8 j_uuid[16];
776
777 /* Pointer to the current commit thread for this journal */
778 struct task_struct *j_task;
779
780 /*
781 * Maximum number of metadata buffers to allow in a single compound
782 * commit transaction
783 */
784 int j_max_transaction_buffers;
785
786 /*
787 * What is the maximum transaction lifetime before we begin a commit?
788 */
789 unsigned long j_commit_interval;
790
791 /* The timer used to wakeup the commit thread: */
Andrew Mortone3df1892006-03-25 03:06:53 -0800792 struct timer_list j_commit_timer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793
794 /*
795 * The revoke table: maintains the list of revoked blocks in the
796 * current transaction. [j_revoke_lock]
797 */
798 spinlock_t j_revoke_lock;
799 struct jbd_revoke_table_s *j_revoke;
800 struct jbd_revoke_table_s *j_revoke_table[2];
801
802 /*
803 * array of bhs for journal_commit_transaction
804 */
805 struct buffer_head **j_wbuf;
806 int j_wbufsize;
807
Josef Bacikf420d4d2009-01-07 18:07:24 -0800808 /*
809 * this is the pid of the last person to run a synchronous operation
810 * through the journal.
811 */
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800812 pid_t j_last_sync_writer;
813
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800815 * the average amount of time in nanoseconds it takes to commit a
816 * transaction to the disk. [j_state_lock]
817 */
818 u64 j_average_commit_time;
819
820 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 * An opaque pointer to fs-private information. ext3 puts its
822 * superblock pointer here
823 */
824 void *j_private;
825};
826
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700827/*
828 * Journal flag definitions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829 */
830#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
831#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
832#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
833#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
834#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
835#define JFS_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700836#define JFS_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
837 * data write error in ordered
838 * mode */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700840/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841 * Function declarations for the journaling transaction and buffer
842 * management
843 */
844
845/* Filing buffers */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846extern void journal_unfile_buffer(journal_t *, struct journal_head *);
847extern void __journal_unfile_buffer(struct journal_head *);
848extern void __journal_refile_buffer(struct journal_head *);
849extern void journal_refile_buffer(journal_t *, struct journal_head *);
850extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
851extern void __journal_free_buffer(struct journal_head *bh);
852extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
853extern void __journal_clean_data_list(transaction_t *transaction);
854
855/* Log buffer allocation */
856extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
857int journal_next_log_block(journal_t *, unsigned long *);
858
859/* Commit management */
860extern void journal_commit_transaction(journal_t *);
861
862/* Checkpoint list management */
863int __journal_clean_checkpoint_list(journal_t *journal);
Jan Kara78ce89c2006-06-23 02:06:05 -0700864int __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
866
867/* Buffer IO */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700868extern int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869journal_write_metadata_buffer(transaction_t *transaction,
870 struct journal_head *jh_in,
871 struct journal_head **jh_out,
Eric Sandeen37ed3222006-09-27 01:49:31 -0700872 unsigned long blocknr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873
874/* Transaction locking */
875extern void __wait_on_journal (journal_t *);
876
877/*
878 * Journal locking.
879 *
880 * We need to lock the journal during transaction state changes so that nobody
881 * ever tries to take a handle on the running transaction while we are in the
882 * middle of moving it to the commit phase. j_state_lock does this.
883 *
884 * Note that the locking is completely interrupt unsafe. We never touch
885 * journal structures from interrupts.
886 */
887
888static inline handle_t *journal_current_handle(void)
889{
890 return current->journal_info;
891}
892
893/* The journaling code user interface:
894 *
895 * Create and destroy handles
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700896 * Register buffer modifications against the current transaction.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 */
898
899extern handle_t *journal_start(journal_t *, int nblocks);
900extern int journal_restart (handle_t *, int nblocks);
901extern int journal_extend (handle_t *, int nblocks);
902extern int journal_get_write_access(handle_t *, struct buffer_head *);
903extern int journal_get_create_access (handle_t *, struct buffer_head *);
904extern int journal_get_undo_access(handle_t *, struct buffer_head *);
905extern int journal_dirty_data (handle_t *, struct buffer_head *);
906extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
907extern void journal_release_buffer (handle_t *, struct buffer_head *);
908extern int journal_forget (handle_t *, struct buffer_head *);
909extern void journal_sync_buffer (struct buffer_head *);
NeilBrown2ff28e22006-03-26 01:37:18 -0800910extern void journal_invalidatepage(journal_t *,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 struct page *, unsigned long);
Al Viro27496a82005-10-21 03:20:48 -0400912extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913extern int journal_stop(handle_t *);
914extern int journal_flush (journal_t *);
915extern void journal_lock_updates (journal_t *);
916extern void journal_unlock_updates (journal_t *);
917
918extern journal_t * journal_init_dev(struct block_device *bdev,
919 struct block_device *fs_dev,
920 int start, int len, int bsize);
921extern journal_t * journal_init_inode (struct inode *);
922extern int journal_update_format (journal_t *);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700923extern int journal_check_used_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700925extern int journal_check_available_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700927extern int journal_set_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928 (journal_t *, unsigned long, unsigned long, unsigned long);
929extern int journal_create (journal_t *);
930extern int journal_load (journal_t *journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -0700931extern int journal_destroy (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932extern int journal_recover (journal_t *journal);
933extern int journal_wipe (journal_t *, int);
934extern int journal_skip_recovery (journal_t *);
935extern void journal_update_superblock (journal_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936extern void journal_abort (journal_t *, int);
937extern int journal_errno (journal_t *);
938extern void journal_ack_err (journal_t *);
939extern int journal_clear_err (journal_t *);
940extern int journal_bmap(journal_t *, unsigned long, unsigned long *);
941extern int journal_force_commit(journal_t *);
942
943/*
944 * journal_head management
945 */
946struct journal_head *journal_add_journal_head(struct buffer_head *bh);
947struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
948void journal_remove_journal_head(struct buffer_head *bh);
949void journal_put_journal_head(struct journal_head *jh);
950
951/*
952 * handle management
953 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800954extern struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955
Al Virodd0fc662005-10-07 07:46:04 +0100956static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957{
958 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
959}
960
961static inline void jbd_free_handle(handle_t *handle)
962{
963 kmem_cache_free(jbd_handle_cache, handle);
964}
965
966/* Primary revoke support */
967#define JOURNAL_REVOKE_DEFAULT_HASH 256
968extern int journal_init_revoke(journal_t *, int);
969extern void journal_destroy_revoke_caches(void);
970extern int journal_init_revoke_caches(void);
971
972extern void journal_destroy_revoke(journal_t *);
973extern int journal_revoke (handle_t *,
974 unsigned long, struct buffer_head *);
975extern int journal_cancel_revoke(handle_t *, struct journal_head *);
976extern void journal_write_revoke_records(journal_t *, transaction_t *);
977
978/* Recovery revoke support */
979extern int journal_set_revoke(journal_t *, unsigned long, tid_t);
980extern int journal_test_revoke(journal_t *, unsigned long, tid_t);
981extern void journal_clear_revoke(journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982extern void journal_switch_revoke_table(journal_t *journal);
983
984/*
985 * The log thread user interface:
986 *
987 * Request space in the current transaction, and force transaction commit
988 * transitions on demand.
989 */
990
991int __log_space_left(journal_t *); /* Called with journal locked */
992int log_start_commit(journal_t *journal, tid_t tid);
993int __log_start_commit(journal_t *journal, tid_t tid);
994int journal_start_commit(journal_t *journal, tid_t *tid);
995int journal_force_commit_nested(journal_t *journal);
996int log_wait_commit(journal_t *journal, tid_t tid);
997int log_do_checkpoint(journal_t *journal);
998
999void __log_wait_for_space(journal_t *journal);
1000extern void __journal_drop_transaction(journal_t *, transaction_t *);
1001extern int cleanup_journal_tail(journal_t *);
1002
1003/* Debugging code only: */
1004
1005#define jbd_ENOSYS() \
1006do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -07001007 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008 current->state = TASK_UNINTERRUPTIBLE; \
1009 schedule(); \
1010} while (1)
1011
1012/*
1013 * is_journal_abort
1014 *
1015 * Simple test wrapper function to test the JFS_ABORT state flag. This
1016 * bit, when set, indicates that we have had a fatal error somewhere,
1017 * either inside the journaling layer or indicated to us by the client
1018 * (eg. ext3), and that we and should not commit any further
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001019 * transactions.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 */
1021
1022static inline int is_journal_aborted(journal_t *journal)
1023{
1024 return journal->j_flags & JFS_ABORT;
1025}
1026
1027static inline int is_handle_aborted(handle_t *handle)
1028{
1029 if (handle->h_aborted)
1030 return 1;
1031 return is_journal_aborted(handle->h_transaction->t_journal);
1032}
1033
1034static inline void journal_abort_handle(handle_t *handle)
1035{
1036 handle->h_aborted = 1;
1037}
1038
1039#endif /* __KERNEL__ */
1040
1041/* Comparison functions for transaction IDs: perform comparisons using
1042 * modulo arithmetic so that they work over sequence number wraps. */
1043
1044static inline int tid_gt(tid_t x, tid_t y)
1045{
1046 int difference = (x - y);
1047 return (difference > 0);
1048}
1049
1050static inline int tid_geq(tid_t x, tid_t y)
1051{
1052 int difference = (x - y);
1053 return (difference >= 0);
1054}
1055
1056extern int journal_blocks_per_page(struct inode *inode);
1057
1058/*
1059 * Return the minimum number of blocks which must be free in the journal
1060 * before a new transaction may be started. Must be called under j_state_lock.
1061 */
1062static inline int jbd_space_needed(journal_t *journal)
1063{
1064 int nblocks = journal->j_max_transaction_buffers;
1065 if (journal->j_committing_transaction)
1066 nblocks += journal->j_committing_transaction->
1067 t_outstanding_credits;
1068 return nblocks;
1069}
1070
1071/*
1072 * Definitions which augment the buffer_head layer
1073 */
1074
1075/* journaling buffer types */
1076#define BJ_None 0 /* Not journaled */
1077#define BJ_SyncData 1 /* Normal data: flush before commit */
1078#define BJ_Metadata 2 /* Normal journaled metadata */
1079#define BJ_Forget 3 /* Buffer superseded by this transaction */
1080#define BJ_IO 4 /* Buffer is for temporary IO use */
1081#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1082#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1083#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1084#define BJ_Locked 8 /* Locked for I/O during commit */
1085#define BJ_Types 9
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001086
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087extern int jbd_blocks_per_page(struct inode *inode);
1088
1089#ifdef __KERNEL__
1090
1091#define buffer_trace_init(bh) do {} while (0)
1092#define print_buffer_fields(bh) do {} while (0)
1093#define print_buffer_trace(bh) do {} while (0)
1094#define BUFFER_TRACE(bh, info) do {} while (0)
1095#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1096#define JBUFFER_TRACE(jh, info) do {} while (0)
1097
1098#endif /* __KERNEL__ */
1099
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100#endif /* _LINUX_JBD_H */