blob: 516a2a27e87a3ce4c7ed3eeeedb12219aa28faec [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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090034#include <linux/slab.h>
Andrew Mortone3df1892006-03-25 03:06:53 -080035
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#define journal_oom_retry 1
37
38/*
Kazuo Moriwaka0cbfdc82009-03-03 13:01:23 +010039 * Define JBD_PARANOID_IOFAIL to cause a kernel BUG() if ext3 finds
Linus Torvalds1da177e2005-04-16 15:20:36 -070040 * certain classes of error which can occur due to failed IOs. Under
41 * normal use we want ext3 to continue after such errors, because
42 * hardware _can_ fail, but for debugging purposes when running tests on
43 * known-good hardware we may want to trap these errors.
44 */
45#undef JBD_PARANOID_IOFAIL
46
47/*
48 * The default maximum commit age, in seconds.
49 */
50#define JBD_DEFAULT_MAX_COMMIT_AGE 5
51
52#ifdef CONFIG_JBD_DEBUG
53/*
54 * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
55 * consistency checks. By default we don't do this unless
56 * CONFIG_JBD_DEBUG is on.
57 */
58#define JBD_EXPENSIVE_CHECKING
Jose R. Santosc2a91592007-10-18 23:39:22 -070059extern u8 journal_enable_debug;
Linus Torvalds1da177e2005-04-16 15:20:36 -070060
61#define jbd_debug(n, f, a...) \
62 do { \
63 if ((n) <= journal_enable_debug) { \
64 printk (KERN_DEBUG "(%s, %d): %s: ", \
Harvey Harrisond5c003b2008-10-15 22:01:24 -070065 __FILE__, __LINE__, __func__); \
Dave Kleikampe9ad5622006-09-27 01:49:35 -070066 printk (f, ## a); \
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 } \
68 } while (0)
69#else
70#define jbd_debug(f, a...) /**/
71#endif
72
Mingming Caoc089d492007-10-16 18:38:25 -040073static inline void *jbd_alloc(size_t size, gfp_t flags)
74{
75 return (void *)__get_free_pages(flags, get_order(size));
76}
77
78static inline void jbd_free(void *ptr, size_t size)
79{
80 free_pages((unsigned long)ptr, get_order(size));
81};
82
Linus Torvalds1da177e2005-04-16 15:20:36 -070083#define JFS_MIN_JOURNAL_BLOCKS 1024
84
Linus Torvalds1da177e2005-04-16 15:20:36 -070085
86/**
87 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
88 *
89 * All filesystem modifications made by the process go
90 * through this handle. Recursive operations (such as quota operations)
91 * are gathered into a single update.
92 *
93 * The buffer credits field is used to account for journaled buffers
94 * being modified by the running process. To ensure that there is
95 * enough log space for all outstanding operations, we need to limit the
96 * number of outstanding buffers possible at any time. When the
97 * operation completes, any buffer credits not used are credited back to
98 * the transaction, so that at all times we know how many buffers the
Mingming Caoae6ddcc2006-09-27 01:49:27 -070099 * outstanding updates on a transaction might possibly touch.
100 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101 * This is an opaque datatype.
102 **/
103typedef struct handle_s handle_t; /* Atomic operation type */
104
105
106/**
107 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
108 *
109 * journal_t is linked to from the fs superblock structure.
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700110 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 * We use the journal_t to keep track of all outstanding transaction
112 * activity on the filesystem, and to manage the state of the log
113 * writing process.
114 *
115 * This is an opaque datatype.
116 **/
117typedef struct journal_s journal_t; /* Journal control structure */
118#endif
119
120/*
121 * Internal structures used by the logging mechanism:
122 */
123
124#define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
125
126/*
127 * On-disk structures
128 */
129
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700130/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131 * Descriptor block types:
132 */
133
134#define JFS_DESCRIPTOR_BLOCK 1
135#define JFS_COMMIT_BLOCK 2
136#define JFS_SUPERBLOCK_V1 3
137#define JFS_SUPERBLOCK_V2 4
138#define JFS_REVOKE_BLOCK 5
139
140/*
141 * Standard header for all descriptor blocks:
142 */
143typedef struct journal_header_s
144{
145 __be32 h_magic;
146 __be32 h_blocktype;
147 __be32 h_sequence;
148} journal_header_t;
149
150
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700151/*
152 * The block tag: used to describe a single buffer in the journal
Linus Torvalds1da177e2005-04-16 15:20:36 -0700153 */
154typedef struct journal_block_tag_s
155{
156 __be32 t_blocknr; /* The on-disk block number */
157 __be32 t_flags; /* See below */
158} journal_block_tag_t;
159
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700160/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161 * The revoke descriptor: used on disk to describe a series of blocks to
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700162 * be revoked from the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 */
164typedef struct journal_revoke_header_s
165{
166 journal_header_t r_header;
167 __be32 r_count; /* Count of bytes used in the block */
168} journal_revoke_header_t;
169
170
171/* Definitions for the journal tag flags word: */
172#define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */
173#define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */
174#define JFS_FLAG_DELETED 4 /* block deleted by this transaction */
175#define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
176
177
178/*
179 * The journal superblock. All fields are in big-endian byte order.
180 */
181typedef struct journal_superblock_s
182{
183/* 0x0000 */
184 journal_header_t s_header;
185
186/* 0x000C */
187 /* Static information describing the journal */
188 __be32 s_blocksize; /* journal device blocksize */
189 __be32 s_maxlen; /* total blocks in journal file */
190 __be32 s_first; /* first block of log information */
191
192/* 0x0018 */
193 /* Dynamic information describing the current state of the log */
194 __be32 s_sequence; /* first commit ID expected in log */
195 __be32 s_start; /* blocknr of start of log */
196
197/* 0x0020 */
198 /* Error value, as set by journal_abort(). */
199 __be32 s_errno;
200
201/* 0x0024 */
202 /* Remaining fields are only valid in a version-2 superblock */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700203 __be32 s_feature_compat; /* compatible feature set */
204 __be32 s_feature_incompat; /* incompatible feature set */
205 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206/* 0x0030 */
207 __u8 s_uuid[16]; /* 128-bit uuid for journal */
208
209/* 0x0040 */
210 __be32 s_nr_users; /* Nr of filesystems sharing log */
211
212 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
213
214/* 0x0048 */
215 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
216 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
217
218/* 0x0050 */
219 __u32 s_padding[44];
220
221/* 0x0100 */
222 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
223/* 0x0400 */
224} journal_superblock_t;
225
226#define JFS_HAS_COMPAT_FEATURE(j,mask) \
227 ((j)->j_format_version >= 2 && \
228 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
229#define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \
230 ((j)->j_format_version >= 2 && \
231 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
232#define JFS_HAS_INCOMPAT_FEATURE(j,mask) \
233 ((j)->j_format_version >= 2 && \
234 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
235
236#define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001
237
238/* Features known to this kernel version: */
239#define JFS_KNOWN_COMPAT_FEATURES 0
240#define JFS_KNOWN_ROCOMPAT_FEATURES 0
241#define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE
242
243#ifdef __KERNEL__
244
245#include <linux/fs.h>
246#include <linux/sched.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247
Chris Snook1c099242007-10-18 23:39:21 -0700248#define J_ASSERT(assert) BUG_ON(!(assert))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
251#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253#if defined(JBD_PARANOID_IOFAIL)
254#define J_EXPECT(expr, why...) J_ASSERT(expr)
255#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
256#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
257#else
258#define __journal_expect(expr, why...) \
259 ({ \
260 int val = (expr); \
261 if (!val) { \
262 printk(KERN_ERR \
263 "EXT3-fs unexpected failure: %s;\n",# expr); \
264 printk(KERN_ERR why "\n"); \
265 } \
266 val; \
267 })
268#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
269#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
270#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
271#endif
272
273enum jbd_state_bits {
274 BH_JBD /* Has an attached ext3 journal_head */
275 = BH_PrivateStart,
276 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
277 BH_Freed, /* Has been freed (truncated) */
278 BH_Revoked, /* Has been revoked from the log */
279 BH_RevokeValid, /* Revoked flag is valid */
280 BH_JBDDirty, /* Is dirty but journaled */
281 BH_State, /* Pins most journal_head state */
282 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
283 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
284};
285
286BUFFER_FNS(JBD, jbd)
287BUFFER_FNS(JWrite, jwrite)
288BUFFER_FNS(JBDDirty, jbddirty)
289TAS_BUFFER_FNS(JBDDirty, jbddirty)
290BUFFER_FNS(Revoked, revoked)
291TAS_BUFFER_FNS(Revoked, revoked)
292BUFFER_FNS(RevokeValid, revokevalid)
293TAS_BUFFER_FNS(RevokeValid, revokevalid)
294BUFFER_FNS(Freed, freed)
295
296static inline struct buffer_head *jh2bh(struct journal_head *jh)
297{
298 return jh->b_bh;
299}
300
301static inline struct journal_head *bh2jh(struct buffer_head *bh)
302{
303 return bh->b_private;
304}
305
306static inline void jbd_lock_bh_state(struct buffer_head *bh)
307{
308 bit_spin_lock(BH_State, &bh->b_state);
309}
310
311static inline int jbd_trylock_bh_state(struct buffer_head *bh)
312{
313 return bit_spin_trylock(BH_State, &bh->b_state);
314}
315
316static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
317{
318 return bit_spin_is_locked(BH_State, &bh->b_state);
319}
320
321static inline void jbd_unlock_bh_state(struct buffer_head *bh)
322{
323 bit_spin_unlock(BH_State, &bh->b_state);
324}
325
326static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
327{
328 bit_spin_lock(BH_JournalHead, &bh->b_state);
329}
330
331static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
332{
333 bit_spin_unlock(BH_JournalHead, &bh->b_state);
334}
335
336struct jbd_revoke_table_s;
337
338/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700339 * struct handle_s - this is the concrete type associated with handle_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 * @h_transaction: Which compound transaction is this update a part of?
341 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
342 * @h_ref: Reference count on this handle
343 * @h_err: Field for caller's use to track errors through large fs operations
344 * @h_sync: flag for sync-on-close
345 * @h_jdata: flag to force data journaling
346 * @h_aborted: flag indicating fatal error on handle
Randy Dunlape6716b82007-10-15 17:30:19 -0700347 * @h_lockdep_map: lockdep info for debugging lock problems
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700349struct handle_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350{
351 /* Which compound transaction is this update a part of? */
352 transaction_t *h_transaction;
353
354 /* Number of remaining buffers we are allowed to dirty: */
355 int h_buffer_credits;
356
357 /* Reference count on this handle */
358 int h_ref;
359
360 /* Field for caller's use to track errors through large fs */
361 /* operations */
362 int h_err;
363
364 /* Flags [no locking] */
365 unsigned int h_sync: 1; /* sync-on-close */
366 unsigned int h_jdata: 1; /* force data journaling */
367 unsigned int h_aborted: 1; /* fatal error on handle */
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +0200368
369#ifdef CONFIG_DEBUG_LOCK_ALLOC
370 struct lockdep_map h_lockdep_map;
371#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372};
373
374
375/* The transaction_t type is the guts of the journaling mechanism. It
376 * tracks a compound transaction through its various states:
377 *
378 * RUNNING: accepting new updates
379 * LOCKED: Updates still running but we don't accept new ones
380 * RUNDOWN: Updates are tidying up but have finished requesting
381 * new buffers to modify (state not used for now)
382 * FLUSH: All updates complete, but we are still writing to disk
383 * COMMIT: All data on disk, writing commit record
384 * FINISHED: We still have to keep the transaction for checkpointing.
385 *
386 * The transaction keeps track of all of the buffers modified by a
387 * running transaction, and all of the buffers committed but not yet
388 * flushed to home for finished transactions.
389 */
390
391/*
392 * Lock ranking:
393 *
394 * j_list_lock
395 * ->jbd_lock_bh_journal_head() (This is "innermost")
396 *
397 * j_state_lock
398 * ->jbd_lock_bh_state()
399 *
400 * jbd_lock_bh_state()
401 * ->j_list_lock
402 *
403 * j_state_lock
404 * ->t_handle_lock
405 *
406 * j_state_lock
407 * ->j_list_lock (journal_unmap_buffer)
408 *
409 */
410
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700411struct transaction_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412{
413 /* Pointer to the journal for this transaction. [no locking] */
414 journal_t *t_journal;
415
416 /* Sequence number for this transaction [no locking] */
417 tid_t t_tid;
418
419 /*
420 * Transaction's current state
421 * [no locking - only kjournald alters this]
Jan Karad4beaf42007-12-04 23:45:27 -0800422 * [j_list_lock] guards transition of a transaction into T_FINISHED
423 * state and subsequent call of __journal_drop_transaction()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 * FIXME: needs barriers
425 * KLUDGE: [use j_state_lock]
426 */
427 enum {
428 T_RUNNING,
429 T_LOCKED,
430 T_RUNDOWN,
431 T_FLUSH,
432 T_COMMIT,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700433 T_FINISHED
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 } t_state;
435
436 /*
437 * Where in the log does this transaction's commit start? [no locking]
438 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200439 unsigned int t_log_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
441 /* Number of buffers on the t_buffers list [j_list_lock] */
442 int t_nr_buffers;
443
444 /*
445 * Doubly-linked circular list of all buffers reserved but not yet
446 * modified by this transaction [j_list_lock]
447 */
448 struct journal_head *t_reserved_list;
449
450 /*
451 * Doubly-linked circular list of all buffers under writeout during
452 * commit [j_list_lock]
453 */
454 struct journal_head *t_locked_list;
455
456 /*
457 * Doubly-linked circular list of all metadata buffers owned by this
458 * transaction [j_list_lock]
459 */
460 struct journal_head *t_buffers;
461
462 /*
463 * Doubly-linked circular list of all data buffers still to be
464 * flushed before this transaction can be committed [j_list_lock]
465 */
466 struct journal_head *t_sync_datalist;
467
468 /*
469 * Doubly-linked circular list of all forget buffers (superseded
470 * buffers which we can un-checkpoint once this transaction commits)
471 * [j_list_lock]
472 */
473 struct journal_head *t_forget;
474
475 /*
476 * Doubly-linked circular list of all buffers still to be flushed before
477 * this transaction can be checkpointed. [j_list_lock]
478 */
479 struct journal_head *t_checkpoint_list;
480
481 /*
Jan Kara78ce89c2006-06-23 02:06:05 -0700482 * Doubly-linked circular list of all buffers submitted for IO while
483 * checkpointing. [j_list_lock]
484 */
485 struct journal_head *t_checkpoint_io_list;
486
487 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 * Doubly-linked circular list of temporary buffers currently undergoing
489 * IO in the log [j_list_lock]
490 */
491 struct journal_head *t_iobuf_list;
492
493 /*
494 * Doubly-linked circular list of metadata buffers being shadowed by log
495 * IO. The IO buffers on the iobuf list and the shadow buffers on this
496 * list match each other one for one at all times. [j_list_lock]
497 */
498 struct journal_head *t_shadow_list;
499
500 /*
501 * Doubly-linked circular list of control buffers being written to the
502 * log. [j_list_lock]
503 */
504 struct journal_head *t_log_list;
505
506 /*
507 * Protects info related to handles
508 */
509 spinlock_t t_handle_lock;
510
511 /*
512 * Number of outstanding updates running on this transaction
513 * [t_handle_lock]
514 */
515 int t_updates;
516
517 /*
518 * Number of buffers reserved for use by all handles in this transaction
519 * handle but not yet modified. [t_handle_lock]
520 */
521 int t_outstanding_credits;
522
523 /*
524 * Forward and backward links for the circular list of all transactions
525 * awaiting checkpoint. [j_list_lock]
526 */
527 transaction_t *t_cpnext, *t_cpprev;
528
529 /*
530 * When will the transaction expire (become due for commit), in jiffies?
531 * [no locking]
532 */
533 unsigned long t_expires;
534
535 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800536 * When this transaction started, in nanoseconds [no locking]
537 */
538 ktime_t t_start_time;
539
540 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541 * How many handles used this transaction? [t_handle_lock]
542 */
543 int t_handle_count;
544
Theodore Ts'o512a0042009-03-27 22:14:27 -0400545 /*
546 * This transaction is being forced and some process is
547 * waiting for it to finish.
548 */
H Hartley Sweeten4fefce92009-09-22 16:48:02 -0700549 unsigned int t_synchronous_commit:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550};
551
552/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700553 * struct journal_s - this is the concrete type associated with journal_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 * @j_flags: General journaling state flags
555 * @j_errno: Is there an outstanding uncleared error on the journal (from a
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700556 * prior abort)?
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 * @j_sb_buffer: First part of superblock buffer
558 * @j_superblock: Second part of superblock buffer
559 * @j_format_version: Version of the superblock format
560 * @j_state_lock: Protect the various scalars in the journal
561 * @j_barrier_count: Number of processes waiting to create a barrier lock
562 * @j_barrier: The barrier lock itself
563 * @j_running_transaction: The current running transaction..
564 * @j_committing_transaction: the transaction we are pushing to disk
565 * @j_checkpoint_transactions: a linked circular list of all transactions
566 * waiting for checkpointing
567 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
568 * to start committing, or for a barrier lock to be released
569 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700570 * @j_wait_done_commit: Wait queue for waiting for commit to complete
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 * @j_wait_checkpoint: Wait queue to trigger checkpointing
572 * @j_wait_commit: Wait queue to trigger commit
573 * @j_wait_updates: Wait queue to wait for updates to complete
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800574 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575 * @j_head: Journal head - identifies the first unused block in the journal
576 * @j_tail: Journal tail - identifies the oldest still-used block in the
577 * journal.
578 * @j_free: Journal free - how many free blocks are there in the journal?
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700579 * @j_first: The block number of the first usable block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 * @j_last: The block number one beyond the last usable block
581 * @j_dev: Device where we store the journal
582 * @j_blocksize: blocksize for the location where we store the journal.
583 * @j_blk_offset: starting block offset for into the device where we store the
584 * journal
585 * @j_fs_dev: Device which holds the client fs. For internal journal this will
586 * be equal to j_dev
587 * @j_maxlen: Total maximum capacity of the journal region on disk.
588 * @j_list_lock: Protects the buffer lists and internal buffer state.
589 * @j_inode: Optional inode where we store the journal. If present, all journal
590 * block numbers are mapped into this inode via bmap().
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700591 * @j_tail_sequence: Sequence number of the oldest transaction in the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592 * @j_transaction_sequence: Sequence number of the next transaction to grant
593 * @j_commit_sequence: Sequence number of the most recently committed
594 * transaction
595 * @j_commit_request: Sequence number of the most recent transaction wanting
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700596 * commit
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597 * @j_uuid: Uuid of client object.
598 * @j_task: Pointer to the current commit thread for this journal
599 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
600 * single compound commit transaction
601 * @j_commit_interval: What is the maximum transaction lifetime before we begin
602 * a commit?
603 * @j_commit_timer: The timer used to wakeup the commit thread
604 * @j_revoke_lock: Protect the revoke table
605 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
606 * current transaction.
607 * @j_revoke_table: alternate revoke tables for j_revoke
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800608 * @j_wbuf: array of buffer_heads for journal_commit_transaction
609 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
610 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800611 * @j_last_sync_writer: most recent pid which did a synchronous write
Randy Dunlap3eabdb72009-01-15 13:51:01 -0800612 * @j_average_commit_time: the average amount of time in nanoseconds it
613 * takes to commit a transaction to the disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614 * @j_private: An opaque pointer to fs-private information.
615 */
616
617struct journal_s
618{
619 /* General journaling state flags [j_state_lock] */
620 unsigned long j_flags;
621
622 /*
623 * Is there an outstanding uncleared error on the journal (from a prior
624 * abort)? [j_state_lock]
625 */
626 int j_errno;
627
628 /* The superblock buffer */
629 struct buffer_head *j_sb_buffer;
630 journal_superblock_t *j_superblock;
631
632 /* Version of the superblock format */
633 int j_format_version;
634
635 /*
636 * Protect the various scalars in the journal
637 */
638 spinlock_t j_state_lock;
639
640 /*
641 * Number of processes waiting to create a barrier lock [j_state_lock]
642 */
643 int j_barrier_count;
644
645 /* The barrier lock itself */
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800646 struct mutex j_barrier;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700647
648 /*
649 * Transactions: The current running transaction...
650 * [j_state_lock] [caller holding open handle]
651 */
652 transaction_t *j_running_transaction;
653
654 /*
655 * the transaction we are pushing to disk
656 * [j_state_lock] [caller holding open handle]
657 */
658 transaction_t *j_committing_transaction;
659
660 /*
661 * ... and a linked circular list of all transactions waiting for
662 * checkpointing. [j_list_lock]
663 */
664 transaction_t *j_checkpoint_transactions;
665
666 /*
667 * Wait queue for waiting for a locked transaction to start committing,
668 * or for a barrier lock to be released
669 */
670 wait_queue_head_t j_wait_transaction_locked;
671
672 /* Wait queue for waiting for checkpointing to complete */
673 wait_queue_head_t j_wait_logspace;
674
675 /* Wait queue for waiting for commit to complete */
676 wait_queue_head_t j_wait_done_commit;
677
678 /* Wait queue to trigger checkpointing */
679 wait_queue_head_t j_wait_checkpoint;
680
681 /* Wait queue to trigger commit */
682 wait_queue_head_t j_wait_commit;
683
684 /* Wait queue to wait for updates to complete */
685 wait_queue_head_t j_wait_updates;
686
687 /* Semaphore for locking against concurrent checkpoints */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700688 struct mutex j_checkpoint_mutex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
690 /*
691 * Journal head: identifies the first unused block in the journal.
692 * [j_state_lock]
693 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200694 unsigned int j_head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695
696 /*
697 * Journal tail: identifies the oldest still-used block in the journal.
698 * [j_state_lock]
699 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200700 unsigned int j_tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
702 /*
703 * Journal free: how many free blocks are there in the journal?
704 * [j_state_lock]
705 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200706 unsigned int j_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
708 /*
709 * Journal start and end: the block numbers of the first usable block
710 * and one beyond the last usable block in the journal. [j_state_lock]
711 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200712 unsigned int j_first;
713 unsigned int j_last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700714
715 /*
716 * Device, blocksize and starting block offset for the location where we
717 * store the journal.
718 */
719 struct block_device *j_dev;
720 int j_blocksize;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200721 unsigned int j_blk_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722
723 /*
724 * Device which holds the client fs. For internal journal this will be
725 * equal to j_dev.
726 */
727 struct block_device *j_fs_dev;
728
729 /* Total maximum capacity of the journal region on disk. */
730 unsigned int j_maxlen;
731
732 /*
733 * Protects the buffer lists and internal buffer state.
734 */
735 spinlock_t j_list_lock;
736
737 /* Optional inode where we store the journal. If present, all */
738 /* journal block numbers are mapped into this inode via */
739 /* bmap(). */
740 struct inode *j_inode;
741
742 /*
743 * Sequence number of the oldest transaction in the log [j_state_lock]
744 */
745 tid_t j_tail_sequence;
746
747 /*
748 * Sequence number of the next transaction to grant [j_state_lock]
749 */
750 tid_t j_transaction_sequence;
751
752 /*
753 * Sequence number of the most recently committed transaction
754 * [j_state_lock].
755 */
756 tid_t j_commit_sequence;
757
758 /*
759 * Sequence number of the most recent transaction wanting commit
760 * [j_state_lock]
761 */
762 tid_t j_commit_request;
763
764 /*
765 * Journal uuid: identifies the object (filesystem, LVM volume etc)
766 * backed by this journal. This will eventually be replaced by an array
767 * of uuids, allowing us to index multiple devices within a single
768 * journal and to perform atomic updates across them.
769 */
770 __u8 j_uuid[16];
771
772 /* Pointer to the current commit thread for this journal */
773 struct task_struct *j_task;
774
775 /*
776 * Maximum number of metadata buffers to allow in a single compound
777 * commit transaction
778 */
779 int j_max_transaction_buffers;
780
781 /*
782 * What is the maximum transaction lifetime before we begin a commit?
783 */
784 unsigned long j_commit_interval;
785
786 /* The timer used to wakeup the commit thread: */
Andrew Mortone3df1892006-03-25 03:06:53 -0800787 struct timer_list j_commit_timer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788
789 /*
790 * The revoke table: maintains the list of revoked blocks in the
791 * current transaction. [j_revoke_lock]
792 */
793 spinlock_t j_revoke_lock;
794 struct jbd_revoke_table_s *j_revoke;
795 struct jbd_revoke_table_s *j_revoke_table[2];
796
797 /*
798 * array of bhs for journal_commit_transaction
799 */
800 struct buffer_head **j_wbuf;
801 int j_wbufsize;
802
Josef Bacikf420d4d2009-01-07 18:07:24 -0800803 /*
804 * this is the pid of the last person to run a synchronous operation
805 * through the journal.
806 */
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800807 pid_t j_last_sync_writer;
808
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800810 * the average amount of time in nanoseconds it takes to commit a
811 * transaction to the disk. [j_state_lock]
812 */
813 u64 j_average_commit_time;
814
815 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 * An opaque pointer to fs-private information. ext3 puts its
817 * superblock pointer here
818 */
819 void *j_private;
820};
821
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700822/*
823 * Journal flag definitions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824 */
825#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
826#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
827#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
828#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
829#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
830#define JFS_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700831#define JFS_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
832 * data write error in ordered
833 * mode */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700835/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 * Function declarations for the journaling transaction and buffer
837 * management
838 */
839
840/* Filing buffers */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700841extern void journal_unfile_buffer(journal_t *, struct journal_head *);
842extern void __journal_unfile_buffer(struct journal_head *);
843extern void __journal_refile_buffer(struct journal_head *);
844extern void journal_refile_buffer(journal_t *, struct journal_head *);
845extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
846extern void __journal_free_buffer(struct journal_head *bh);
847extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
848extern void __journal_clean_data_list(transaction_t *transaction);
849
850/* Log buffer allocation */
851extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200852int journal_next_log_block(journal_t *, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
854/* Commit management */
855extern void journal_commit_transaction(journal_t *);
856
857/* Checkpoint list management */
858int __journal_clean_checkpoint_list(journal_t *journal);
Jan Kara78ce89c2006-06-23 02:06:05 -0700859int __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
861
862/* Buffer IO */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700863extern int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864journal_write_metadata_buffer(transaction_t *transaction,
865 struct journal_head *jh_in,
866 struct journal_head **jh_out,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200867 unsigned int blocknr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
869/* Transaction locking */
870extern void __wait_on_journal (journal_t *);
871
872/*
873 * Journal locking.
874 *
875 * We need to lock the journal during transaction state changes so that nobody
876 * ever tries to take a handle on the running transaction while we are in the
877 * middle of moving it to the commit phase. j_state_lock does this.
878 *
879 * Note that the locking is completely interrupt unsafe. We never touch
880 * journal structures from interrupts.
881 */
882
883static inline handle_t *journal_current_handle(void)
884{
885 return current->journal_info;
886}
887
888/* The journaling code user interface:
889 *
890 * Create and destroy handles
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700891 * Register buffer modifications against the current transaction.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 */
893
894extern handle_t *journal_start(journal_t *, int nblocks);
895extern int journal_restart (handle_t *, int nblocks);
896extern int journal_extend (handle_t *, int nblocks);
897extern int journal_get_write_access(handle_t *, struct buffer_head *);
898extern int journal_get_create_access (handle_t *, struct buffer_head *);
899extern int journal_get_undo_access(handle_t *, struct buffer_head *);
900extern int journal_dirty_data (handle_t *, struct buffer_head *);
901extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
902extern void journal_release_buffer (handle_t *, struct buffer_head *);
903extern int journal_forget (handle_t *, struct buffer_head *);
904extern void journal_sync_buffer (struct buffer_head *);
NeilBrown2ff28e22006-03-26 01:37:18 -0800905extern void journal_invalidatepage(journal_t *,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 struct page *, unsigned long);
Al Viro27496a82005-10-21 03:20:48 -0400907extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908extern int journal_stop(handle_t *);
909extern int journal_flush (journal_t *);
910extern void journal_lock_updates (journal_t *);
911extern void journal_unlock_updates (journal_t *);
912
913extern journal_t * journal_init_dev(struct block_device *bdev,
914 struct block_device *fs_dev,
915 int start, int len, int bsize);
916extern journal_t * journal_init_inode (struct inode *);
917extern int journal_update_format (journal_t *);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700918extern int journal_check_used_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700920extern int journal_check_available_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700922extern int journal_set_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 (journal_t *, unsigned long, unsigned long, unsigned long);
924extern int journal_create (journal_t *);
925extern int journal_load (journal_t *journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -0700926extern int journal_destroy (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927extern int journal_recover (journal_t *journal);
928extern int journal_wipe (journal_t *, int);
929extern int journal_skip_recovery (journal_t *);
930extern void journal_update_superblock (journal_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931extern void journal_abort (journal_t *, int);
932extern int journal_errno (journal_t *);
933extern void journal_ack_err (journal_t *);
934extern int journal_clear_err (journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200935extern int journal_bmap(journal_t *, unsigned int, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936extern int journal_force_commit(journal_t *);
937
938/*
939 * journal_head management
940 */
941struct journal_head *journal_add_journal_head(struct buffer_head *bh);
942struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
943void journal_remove_journal_head(struct buffer_head *bh);
944void journal_put_journal_head(struct journal_head *jh);
945
946/*
947 * handle management
948 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800949extern struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950
Al Virodd0fc662005-10-07 07:46:04 +0100951static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700952{
953 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
954}
955
956static inline void jbd_free_handle(handle_t *handle)
957{
958 kmem_cache_free(jbd_handle_cache, handle);
959}
960
961/* Primary revoke support */
962#define JOURNAL_REVOKE_DEFAULT_HASH 256
963extern int journal_init_revoke(journal_t *, int);
964extern void journal_destroy_revoke_caches(void);
965extern int journal_init_revoke_caches(void);
966
967extern void journal_destroy_revoke(journal_t *);
968extern int journal_revoke (handle_t *,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200969 unsigned int, struct buffer_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970extern int journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o38d726d2009-04-14 10:10:47 -0400971extern void journal_write_revoke_records(journal_t *,
972 transaction_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973
974/* Recovery revoke support */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200975extern int journal_set_revoke(journal_t *, unsigned int, tid_t);
976extern int journal_test_revoke(journal_t *, unsigned int, tid_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977extern void journal_clear_revoke(journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978extern void journal_switch_revoke_table(journal_t *journal);
979
980/*
981 * The log thread user interface:
982 *
983 * Request space in the current transaction, and force transaction commit
984 * transitions on demand.
985 */
986
987int __log_space_left(journal_t *); /* Called with journal locked */
988int log_start_commit(journal_t *journal, tid_t tid);
989int __log_start_commit(journal_t *journal, tid_t tid);
990int journal_start_commit(journal_t *journal, tid_t *tid);
991int journal_force_commit_nested(journal_t *journal);
992int log_wait_commit(journal_t *journal, tid_t tid);
993int log_do_checkpoint(journal_t *journal);
994
995void __log_wait_for_space(journal_t *journal);
996extern void __journal_drop_transaction(journal_t *, transaction_t *);
997extern int cleanup_journal_tail(journal_t *);
998
999/* Debugging code only: */
1000
1001#define jbd_ENOSYS() \
1002do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -07001003 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004 current->state = TASK_UNINTERRUPTIBLE; \
1005 schedule(); \
1006} while (1)
1007
1008/*
1009 * is_journal_abort
1010 *
1011 * Simple test wrapper function to test the JFS_ABORT state flag. This
1012 * bit, when set, indicates that we have had a fatal error somewhere,
1013 * either inside the journaling layer or indicated to us by the client
1014 * (eg. ext3), and that we and should not commit any further
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001015 * transactions.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016 */
1017
1018static inline int is_journal_aborted(journal_t *journal)
1019{
1020 return journal->j_flags & JFS_ABORT;
1021}
1022
1023static inline int is_handle_aborted(handle_t *handle)
1024{
1025 if (handle->h_aborted)
1026 return 1;
1027 return is_journal_aborted(handle->h_transaction->t_journal);
1028}
1029
1030static inline void journal_abort_handle(handle_t *handle)
1031{
1032 handle->h_aborted = 1;
1033}
1034
1035#endif /* __KERNEL__ */
1036
1037/* Comparison functions for transaction IDs: perform comparisons using
1038 * modulo arithmetic so that they work over sequence number wraps. */
1039
1040static inline int tid_gt(tid_t x, tid_t y)
1041{
1042 int difference = (x - y);
1043 return (difference > 0);
1044}
1045
1046static inline int tid_geq(tid_t x, tid_t y)
1047{
1048 int difference = (x - y);
1049 return (difference >= 0);
1050}
1051
1052extern int journal_blocks_per_page(struct inode *inode);
1053
1054/*
1055 * Return the minimum number of blocks which must be free in the journal
1056 * before a new transaction may be started. Must be called under j_state_lock.
1057 */
1058static inline int jbd_space_needed(journal_t *journal)
1059{
1060 int nblocks = journal->j_max_transaction_buffers;
1061 if (journal->j_committing_transaction)
1062 nblocks += journal->j_committing_transaction->
1063 t_outstanding_credits;
1064 return nblocks;
1065}
1066
1067/*
1068 * Definitions which augment the buffer_head layer
1069 */
1070
1071/* journaling buffer types */
1072#define BJ_None 0 /* Not journaled */
1073#define BJ_SyncData 1 /* Normal data: flush before commit */
1074#define BJ_Metadata 2 /* Normal journaled metadata */
1075#define BJ_Forget 3 /* Buffer superseded by this transaction */
1076#define BJ_IO 4 /* Buffer is for temporary IO use */
1077#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1078#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1079#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1080#define BJ_Locked 8 /* Locked for I/O during commit */
1081#define BJ_Types 9
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001082
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083extern int jbd_blocks_per_page(struct inode *inode);
1084
1085#ifdef __KERNEL__
1086
1087#define buffer_trace_init(bh) do {} while (0)
1088#define print_buffer_fields(bh) do {} while (0)
1089#define print_buffer_trace(bh) do {} while (0)
1090#define BUFFER_TRACE(bh, info) do {} while (0)
1091#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1092#define JBUFFER_TRACE(jh, info) do {} while (0)
1093
1094#endif /* __KERNEL__ */
1095
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096#endif /* _LINUX_JBD_H */