blob: d32615280be92baeb3fd4cdfa2d05aa4e1528080 [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>
Arjan van de Ven2c68ee72006-03-23 03:00:35 -080030#include <linux/mutex.h>
Andrew Mortone3df1892006-03-25 03:06:53 -080031#include <linux/timer.h>
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +020032#include <linux/lockdep.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090033#include <linux/slab.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
Paul Gortmakere99a03c2013-08-01 15:01:08 -040060void __jbd_debug(int level, const char *file, const char *func,
61 unsigned int line, const char *fmt, ...);
62
63#define jbd_debug(n, fmt, a...) \
64 __jbd_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
Linus Torvalds1da177e2005-04-16 15:20:36 -070065#else
Paul Gortmakere99a03c2013-08-01 15:01:08 -040066#define jbd_debug(n, fmt, a...) /**/
Linus Torvalds1da177e2005-04-16 15:20:36 -070067#endif
68
Mingming Caoc089d492007-10-16 18:38:25 -040069static inline void *jbd_alloc(size_t size, gfp_t flags)
70{
71 return (void *)__get_free_pages(flags, get_order(size));
72}
73
74static inline void jbd_free(void *ptr, size_t size)
75{
76 free_pages((unsigned long)ptr, get_order(size));
Nathaniel Yazdani75b92222013-07-03 13:52:21 -070077}
Mingming Caoc089d492007-10-16 18:38:25 -040078
Linus Torvalds1da177e2005-04-16 15:20:36 -070079#define JFS_MIN_JOURNAL_BLOCKS 1024
80
Linus Torvalds1da177e2005-04-16 15:20:36 -070081
82/**
83 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
84 *
85 * All filesystem modifications made by the process go
86 * through this handle. Recursive operations (such as quota operations)
87 * are gathered into a single update.
88 *
89 * The buffer credits field is used to account for journaled buffers
90 * being modified by the running process. To ensure that there is
91 * enough log space for all outstanding operations, we need to limit the
92 * number of outstanding buffers possible at any time. When the
93 * operation completes, any buffer credits not used are credited back to
94 * the transaction, so that at all times we know how many buffers the
Mingming Caoae6ddcc2006-09-27 01:49:27 -070095 * outstanding updates on a transaction might possibly touch.
96 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 * This is an opaque datatype.
98 **/
99typedef struct handle_s handle_t; /* Atomic operation type */
100
101
102/**
103 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
104 *
105 * journal_t is linked to from the fs superblock structure.
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700106 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700107 * We use the journal_t to keep track of all outstanding transaction
108 * activity on the filesystem, and to manage the state of the log
109 * writing process.
110 *
111 * This is an opaque datatype.
112 **/
113typedef struct journal_s journal_t; /* Journal control structure */
114#endif
115
116/*
117 * Internal structures used by the logging mechanism:
118 */
119
120#define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
121
122/*
123 * On-disk structures
124 */
125
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700126/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700127 * Descriptor block types:
128 */
129
130#define JFS_DESCRIPTOR_BLOCK 1
131#define JFS_COMMIT_BLOCK 2
132#define JFS_SUPERBLOCK_V1 3
133#define JFS_SUPERBLOCK_V2 4
134#define JFS_REVOKE_BLOCK 5
135
136/*
137 * Standard header for all descriptor blocks:
138 */
139typedef struct journal_header_s
140{
141 __be32 h_magic;
142 __be32 h_blocktype;
143 __be32 h_sequence;
144} journal_header_t;
145
146
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700147/*
148 * The block tag: used to describe a single buffer in the journal
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149 */
150typedef struct journal_block_tag_s
151{
152 __be32 t_blocknr; /* The on-disk block number */
153 __be32 t_flags; /* See below */
154} journal_block_tag_t;
155
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700156/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 * The revoke descriptor: used on disk to describe a series of blocks to
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700158 * be revoked from the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159 */
160typedef struct journal_revoke_header_s
161{
162 journal_header_t r_header;
163 __be32 r_count; /* Count of bytes used in the block */
164} journal_revoke_header_t;
165
166
167/* Definitions for the journal tag flags word: */
168#define JFS_FLAG_ESCAPE 1 /* on-disk block is escaped */
169#define JFS_FLAG_SAME_UUID 2 /* block has same uuid as previous */
170#define JFS_FLAG_DELETED 4 /* block deleted by this transaction */
171#define JFS_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
172
173
174/*
175 * The journal superblock. All fields are in big-endian byte order.
176 */
177typedef struct journal_superblock_s
178{
179/* 0x0000 */
180 journal_header_t s_header;
181
182/* 0x000C */
183 /* Static information describing the journal */
184 __be32 s_blocksize; /* journal device blocksize */
185 __be32 s_maxlen; /* total blocks in journal file */
186 __be32 s_first; /* first block of log information */
187
188/* 0x0018 */
189 /* Dynamic information describing the current state of the log */
190 __be32 s_sequence; /* first commit ID expected in log */
191 __be32 s_start; /* blocknr of start of log */
192
193/* 0x0020 */
194 /* Error value, as set by journal_abort(). */
195 __be32 s_errno;
196
197/* 0x0024 */
198 /* Remaining fields are only valid in a version-2 superblock */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700199 __be32 s_feature_compat; /* compatible feature set */
200 __be32 s_feature_incompat; /* incompatible feature set */
201 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700202/* 0x0030 */
203 __u8 s_uuid[16]; /* 128-bit uuid for journal */
204
205/* 0x0040 */
206 __be32 s_nr_users; /* Nr of filesystems sharing log */
207
208 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
209
210/* 0x0048 */
211 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
212 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
213
214/* 0x0050 */
215 __u32 s_padding[44];
216
217/* 0x0100 */
218 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
219/* 0x0400 */
220} journal_superblock_t;
221
222#define JFS_HAS_COMPAT_FEATURE(j,mask) \
223 ((j)->j_format_version >= 2 && \
224 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
225#define JFS_HAS_RO_COMPAT_FEATURE(j,mask) \
226 ((j)->j_format_version >= 2 && \
227 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
228#define JFS_HAS_INCOMPAT_FEATURE(j,mask) \
229 ((j)->j_format_version >= 2 && \
230 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
231
232#define JFS_FEATURE_INCOMPAT_REVOKE 0x00000001
233
234/* Features known to this kernel version: */
235#define JFS_KNOWN_COMPAT_FEATURES 0
236#define JFS_KNOWN_ROCOMPAT_FEATURES 0
237#define JFS_KNOWN_INCOMPAT_FEATURES JFS_FEATURE_INCOMPAT_REVOKE
238
239#ifdef __KERNEL__
240
241#include <linux/fs.h>
242#include <linux/sched.h>
Jan Karab34090e2013-06-04 12:08:56 -0400243
244enum jbd_state_bits {
245 BH_JBD /* Has an attached ext3 journal_head */
246 = BH_PrivateStart,
247 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
248 BH_Freed, /* Has been freed (truncated) */
249 BH_Revoked, /* Has been revoked from the log */
250 BH_RevokeValid, /* Revoked flag is valid */
251 BH_JBDDirty, /* Is dirty but journaled */
252 BH_State, /* Pins most journal_head state */
253 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
254 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
255 BH_JBDPrivateStart, /* First bit available for private use by FS */
256};
257
258BUFFER_FNS(JBD, jbd)
259BUFFER_FNS(JWrite, jwrite)
260BUFFER_FNS(JBDDirty, jbddirty)
261TAS_BUFFER_FNS(JBDDirty, jbddirty)
262BUFFER_FNS(Revoked, revoked)
263TAS_BUFFER_FNS(Revoked, revoked)
264BUFFER_FNS(RevokeValid, revokevalid)
265TAS_BUFFER_FNS(RevokeValid, revokevalid)
266BUFFER_FNS(Freed, freed)
267
Thomas Gleixner44606672011-10-27 04:38:18 -0400268#include <linux/jbd_common.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
Chris Snook1c099242007-10-18 23:39:21 -0700270#define J_ASSERT(assert) BUG_ON(!(assert))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
273#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275#if defined(JBD_PARANOID_IOFAIL)
276#define J_EXPECT(expr, why...) J_ASSERT(expr)
277#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
278#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
279#else
280#define __journal_expect(expr, why...) \
281 ({ \
282 int val = (expr); \
283 if (!val) { \
284 printk(KERN_ERR \
285 "EXT3-fs unexpected failure: %s;\n",# expr); \
286 printk(KERN_ERR why "\n"); \
287 } \
288 val; \
289 })
290#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
291#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
292#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
293#endif
294
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295struct jbd_revoke_table_s;
296
297/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700298 * struct handle_s - this is the concrete type associated with handle_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 * @h_transaction: Which compound transaction is this update a part of?
300 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
301 * @h_ref: Reference count on this handle
302 * @h_err: Field for caller's use to track errors through large fs operations
303 * @h_sync: flag for sync-on-close
304 * @h_jdata: flag to force data journaling
305 * @h_aborted: flag indicating fatal error on handle
Randy Dunlape6716b82007-10-15 17:30:19 -0700306 * @h_lockdep_map: lockdep info for debugging lock problems
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700308struct handle_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309{
310 /* Which compound transaction is this update a part of? */
311 transaction_t *h_transaction;
312
313 /* Number of remaining buffers we are allowed to dirty: */
314 int h_buffer_credits;
315
316 /* Reference count on this handle */
317 int h_ref;
318
319 /* Field for caller's use to track errors through large fs */
320 /* operations */
321 int h_err;
322
323 /* Flags [no locking] */
324 unsigned int h_sync: 1; /* sync-on-close */
325 unsigned int h_jdata: 1; /* force data journaling */
326 unsigned int h_aborted: 1; /* fatal error on handle */
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +0200327
328#ifdef CONFIG_DEBUG_LOCK_ALLOC
329 struct lockdep_map h_lockdep_map;
330#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331};
332
333
334/* The transaction_t type is the guts of the journaling mechanism. It
335 * tracks a compound transaction through its various states:
336 *
337 * RUNNING: accepting new updates
338 * LOCKED: Updates still running but we don't accept new ones
339 * RUNDOWN: Updates are tidying up but have finished requesting
340 * new buffers to modify (state not used for now)
341 * FLUSH: All updates complete, but we are still writing to disk
342 * COMMIT: All data on disk, writing commit record
343 * FINISHED: We still have to keep the transaction for checkpointing.
344 *
345 * The transaction keeps track of all of the buffers modified by a
346 * running transaction, and all of the buffers committed but not yet
347 * flushed to home for finished transactions.
348 */
349
350/*
351 * Lock ranking:
352 *
353 * j_list_lock
354 * ->jbd_lock_bh_journal_head() (This is "innermost")
355 *
356 * j_state_lock
357 * ->jbd_lock_bh_state()
358 *
359 * jbd_lock_bh_state()
360 * ->j_list_lock
361 *
362 * j_state_lock
363 * ->t_handle_lock
364 *
365 * j_state_lock
366 * ->j_list_lock (journal_unmap_buffer)
367 *
368 */
369
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700370struct transaction_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371{
372 /* Pointer to the journal for this transaction. [no locking] */
373 journal_t *t_journal;
374
375 /* Sequence number for this transaction [no locking] */
376 tid_t t_tid;
377
378 /*
379 * Transaction's current state
380 * [no locking - only kjournald alters this]
Jan Karad4beaf42007-12-04 23:45:27 -0800381 * [j_list_lock] guards transition of a transaction into T_FINISHED
382 * state and subsequent call of __journal_drop_transaction()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 * FIXME: needs barriers
384 * KLUDGE: [use j_state_lock]
385 */
386 enum {
387 T_RUNNING,
388 T_LOCKED,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 T_FLUSH,
390 T_COMMIT,
Jan Kara03f4d802010-04-15 22:16:24 +0200391 T_COMMIT_RECORD,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700392 T_FINISHED
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 } t_state;
394
395 /*
396 * Where in the log does this transaction's commit start? [no locking]
397 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200398 unsigned int t_log_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
400 /* Number of buffers on the t_buffers list [j_list_lock] */
401 int t_nr_buffers;
402
403 /*
404 * Doubly-linked circular list of all buffers reserved but not yet
405 * modified by this transaction [j_list_lock]
406 */
407 struct journal_head *t_reserved_list;
408
409 /*
410 * Doubly-linked circular list of all buffers under writeout during
411 * commit [j_list_lock]
412 */
413 struct journal_head *t_locked_list;
414
415 /*
416 * Doubly-linked circular list of all metadata buffers owned by this
417 * transaction [j_list_lock]
418 */
419 struct journal_head *t_buffers;
420
421 /*
422 * Doubly-linked circular list of all data buffers still to be
423 * flushed before this transaction can be committed [j_list_lock]
424 */
425 struct journal_head *t_sync_datalist;
426
427 /*
428 * Doubly-linked circular list of all forget buffers (superseded
429 * buffers which we can un-checkpoint once this transaction commits)
430 * [j_list_lock]
431 */
432 struct journal_head *t_forget;
433
434 /*
435 * Doubly-linked circular list of all buffers still to be flushed before
436 * this transaction can be checkpointed. [j_list_lock]
437 */
438 struct journal_head *t_checkpoint_list;
439
440 /*
Jan Kara78ce89c2006-06-23 02:06:05 -0700441 * Doubly-linked circular list of all buffers submitted for IO while
442 * checkpointing. [j_list_lock]
443 */
444 struct journal_head *t_checkpoint_io_list;
445
446 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 * Doubly-linked circular list of temporary buffers currently undergoing
448 * IO in the log [j_list_lock]
449 */
450 struct journal_head *t_iobuf_list;
451
452 /*
453 * Doubly-linked circular list of metadata buffers being shadowed by log
454 * IO. The IO buffers on the iobuf list and the shadow buffers on this
455 * list match each other one for one at all times. [j_list_lock]
456 */
457 struct journal_head *t_shadow_list;
458
459 /*
460 * Doubly-linked circular list of control buffers being written to the
461 * log. [j_list_lock]
462 */
463 struct journal_head *t_log_list;
464
465 /*
466 * Protects info related to handles
467 */
468 spinlock_t t_handle_lock;
469
470 /*
471 * Number of outstanding updates running on this transaction
472 * [t_handle_lock]
473 */
474 int t_updates;
475
476 /*
477 * Number of buffers reserved for use by all handles in this transaction
478 * handle but not yet modified. [t_handle_lock]
479 */
480 int t_outstanding_credits;
481
482 /*
483 * Forward and backward links for the circular list of all transactions
484 * awaiting checkpoint. [j_list_lock]
485 */
486 transaction_t *t_cpnext, *t_cpprev;
487
488 /*
489 * When will the transaction expire (become due for commit), in jiffies?
490 * [no locking]
491 */
492 unsigned long t_expires;
493
494 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800495 * When this transaction started, in nanoseconds [no locking]
496 */
497 ktime_t t_start_time;
498
499 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 * How many handles used this transaction? [t_handle_lock]
501 */
502 int t_handle_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503};
504
505/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700506 * struct journal_s - this is the concrete type associated with journal_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 * @j_flags: General journaling state flags
508 * @j_errno: Is there an outstanding uncleared error on the journal (from a
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700509 * prior abort)?
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 * @j_sb_buffer: First part of superblock buffer
511 * @j_superblock: Second part of superblock buffer
512 * @j_format_version: Version of the superblock format
513 * @j_state_lock: Protect the various scalars in the journal
514 * @j_barrier_count: Number of processes waiting to create a barrier lock
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 * @j_running_transaction: The current running transaction..
516 * @j_committing_transaction: the transaction we are pushing to disk
517 * @j_checkpoint_transactions: a linked circular list of all transactions
518 * waiting for checkpointing
519 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
520 * to start committing, or for a barrier lock to be released
521 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700522 * @j_wait_done_commit: Wait queue for waiting for commit to complete
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 * @j_wait_checkpoint: Wait queue to trigger checkpointing
524 * @j_wait_commit: Wait queue to trigger commit
525 * @j_wait_updates: Wait queue to wait for updates to complete
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800526 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 * @j_head: Journal head - identifies the first unused block in the journal
528 * @j_tail: Journal tail - identifies the oldest still-used block in the
529 * journal.
530 * @j_free: Journal free - how many free blocks are there in the journal?
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700531 * @j_first: The block number of the first usable block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532 * @j_last: The block number one beyond the last usable block
533 * @j_dev: Device where we store the journal
534 * @j_blocksize: blocksize for the location where we store the journal.
535 * @j_blk_offset: starting block offset for into the device where we store the
536 * journal
537 * @j_fs_dev: Device which holds the client fs. For internal journal this will
538 * be equal to j_dev
539 * @j_maxlen: Total maximum capacity of the journal region on disk.
540 * @j_list_lock: Protects the buffer lists and internal buffer state.
541 * @j_inode: Optional inode where we store the journal. If present, all journal
542 * block numbers are mapped into this inode via bmap().
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700543 * @j_tail_sequence: Sequence number of the oldest transaction in the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 * @j_transaction_sequence: Sequence number of the next transaction to grant
545 * @j_commit_sequence: Sequence number of the most recently committed
546 * transaction
547 * @j_commit_request: Sequence number of the most recent transaction wanting
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700548 * commit
Jan Kara2db938b2011-02-21 17:25:37 +0100549 * @j_commit_waited: Sequence number of the most recent transaction someone
550 * is waiting for to commit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 * @j_uuid: Uuid of client object.
552 * @j_task: Pointer to the current commit thread for this journal
553 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
554 * single compound commit transaction
555 * @j_commit_interval: What is the maximum transaction lifetime before we begin
556 * a commit?
557 * @j_commit_timer: The timer used to wakeup the commit thread
558 * @j_revoke_lock: Protect the revoke table
559 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
560 * current transaction.
561 * @j_revoke_table: alternate revoke tables for j_revoke
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800562 * @j_wbuf: array of buffer_heads for journal_commit_transaction
563 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
564 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800565 * @j_last_sync_writer: most recent pid which did a synchronous write
Randy Dunlap3eabdb72009-01-15 13:51:01 -0800566 * @j_average_commit_time: the average amount of time in nanoseconds it
567 * takes to commit a transaction to the disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 * @j_private: An opaque pointer to fs-private information.
569 */
570
571struct journal_s
572{
573 /* General journaling state flags [j_state_lock] */
574 unsigned long j_flags;
575
576 /*
577 * Is there an outstanding uncleared error on the journal (from a prior
578 * abort)? [j_state_lock]
579 */
580 int j_errno;
581
582 /* The superblock buffer */
583 struct buffer_head *j_sb_buffer;
584 journal_superblock_t *j_superblock;
585
586 /* Version of the superblock format */
587 int j_format_version;
588
589 /*
590 * Protect the various scalars in the journal
591 */
592 spinlock_t j_state_lock;
593
594 /*
595 * Number of processes waiting to create a barrier lock [j_state_lock]
596 */
597 int j_barrier_count;
598
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 /*
600 * Transactions: The current running transaction...
601 * [j_state_lock] [caller holding open handle]
602 */
603 transaction_t *j_running_transaction;
604
605 /*
606 * the transaction we are pushing to disk
607 * [j_state_lock] [caller holding open handle]
608 */
609 transaction_t *j_committing_transaction;
610
611 /*
612 * ... and a linked circular list of all transactions waiting for
613 * checkpointing. [j_list_lock]
614 */
615 transaction_t *j_checkpoint_transactions;
616
617 /*
618 * Wait queue for waiting for a locked transaction to start committing,
619 * or for a barrier lock to be released
620 */
621 wait_queue_head_t j_wait_transaction_locked;
622
623 /* Wait queue for waiting for checkpointing to complete */
624 wait_queue_head_t j_wait_logspace;
625
626 /* Wait queue for waiting for commit to complete */
627 wait_queue_head_t j_wait_done_commit;
628
629 /* Wait queue to trigger checkpointing */
630 wait_queue_head_t j_wait_checkpoint;
631
632 /* Wait queue to trigger commit */
633 wait_queue_head_t j_wait_commit;
634
635 /* Wait queue to wait for updates to complete */
636 wait_queue_head_t j_wait_updates;
637
638 /* Semaphore for locking against concurrent checkpoints */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700639 struct mutex j_checkpoint_mutex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640
641 /*
642 * Journal head: identifies the first unused block in the journal.
643 * [j_state_lock]
644 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200645 unsigned int j_head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646
647 /*
648 * Journal tail: identifies the oldest still-used block in the journal.
649 * [j_state_lock]
650 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200651 unsigned int j_tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
653 /*
654 * Journal free: how many free blocks are there in the journal?
655 * [j_state_lock]
656 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200657 unsigned int j_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659 /*
660 * Journal start and end: the block numbers of the first usable block
661 * and one beyond the last usable block in the journal. [j_state_lock]
662 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200663 unsigned int j_first;
664 unsigned int j_last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
666 /*
667 * Device, blocksize and starting block offset for the location where we
668 * store the journal.
669 */
670 struct block_device *j_dev;
671 int j_blocksize;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200672 unsigned int j_blk_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673
674 /*
675 * Device which holds the client fs. For internal journal this will be
676 * equal to j_dev.
677 */
678 struct block_device *j_fs_dev;
679
680 /* Total maximum capacity of the journal region on disk. */
681 unsigned int j_maxlen;
682
683 /*
684 * Protects the buffer lists and internal buffer state.
685 */
686 spinlock_t j_list_lock;
687
688 /* Optional inode where we store the journal. If present, all */
689 /* journal block numbers are mapped into this inode via */
690 /* bmap(). */
691 struct inode *j_inode;
692
693 /*
694 * Sequence number of the oldest transaction in the log [j_state_lock]
695 */
696 tid_t j_tail_sequence;
697
698 /*
699 * Sequence number of the next transaction to grant [j_state_lock]
700 */
701 tid_t j_transaction_sequence;
702
703 /*
704 * Sequence number of the most recently committed transaction
705 * [j_state_lock].
706 */
707 tid_t j_commit_sequence;
708
709 /*
710 * Sequence number of the most recent transaction wanting commit
711 * [j_state_lock]
712 */
713 tid_t j_commit_request;
714
715 /*
Jan Kara2db938b2011-02-21 17:25:37 +0100716 * Sequence number of the most recent transaction someone is waiting
717 * for to commit.
718 * [j_state_lock]
719 */
720 tid_t j_commit_waited;
721
722 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 * Journal uuid: identifies the object (filesystem, LVM volume etc)
724 * backed by this journal. This will eventually be replaced by an array
725 * of uuids, allowing us to index multiple devices within a single
726 * journal and to perform atomic updates across them.
727 */
728 __u8 j_uuid[16];
729
730 /* Pointer to the current commit thread for this journal */
731 struct task_struct *j_task;
732
733 /*
734 * Maximum number of metadata buffers to allow in a single compound
735 * commit transaction
736 */
737 int j_max_transaction_buffers;
738
739 /*
740 * What is the maximum transaction lifetime before we begin a commit?
741 */
742 unsigned long j_commit_interval;
743
744 /* The timer used to wakeup the commit thread: */
Andrew Mortone3df1892006-03-25 03:06:53 -0800745 struct timer_list j_commit_timer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
747 /*
748 * The revoke table: maintains the list of revoked blocks in the
749 * current transaction. [j_revoke_lock]
750 */
751 spinlock_t j_revoke_lock;
752 struct jbd_revoke_table_s *j_revoke;
753 struct jbd_revoke_table_s *j_revoke_table[2];
754
755 /*
756 * array of bhs for journal_commit_transaction
757 */
758 struct buffer_head **j_wbuf;
759 int j_wbufsize;
760
Josef Bacikf420d4d2009-01-07 18:07:24 -0800761 /*
762 * this is the pid of the last person to run a synchronous operation
763 * through the journal.
764 */
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800765 pid_t j_last_sync_writer;
766
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800768 * the average amount of time in nanoseconds it takes to commit a
769 * transaction to the disk. [j_state_lock]
770 */
771 u64 j_average_commit_time;
772
773 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 * An opaque pointer to fs-private information. ext3 puts its
775 * superblock pointer here
776 */
777 void *j_private;
778};
779
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700780/*
781 * Journal flag definitions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 */
783#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
784#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
785#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
786#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
787#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
788#define JFS_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700789#define JFS_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
790 * data write error in ordered
791 * mode */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700793/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 * Function declarations for the journaling transaction and buffer
795 * management
796 */
797
798/* Filing buffers */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799extern void journal_unfile_buffer(journal_t *, struct journal_head *);
800extern void __journal_unfile_buffer(struct journal_head *);
801extern void __journal_refile_buffer(struct journal_head *);
802extern void journal_refile_buffer(journal_t *, struct journal_head *);
803extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
804extern void __journal_free_buffer(struct journal_head *bh);
805extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
806extern void __journal_clean_data_list(transaction_t *transaction);
807
808/* Log buffer allocation */
809extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200810int journal_next_log_block(journal_t *, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811
812/* Commit management */
813extern void journal_commit_transaction(journal_t *);
814
815/* Checkpoint list management */
816int __journal_clean_checkpoint_list(journal_t *journal);
Jan Kara78ce89c2006-06-23 02:06:05 -0700817int __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
819
820/* Buffer IO */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700821extern int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822journal_write_metadata_buffer(transaction_t *transaction,
823 struct journal_head *jh_in,
824 struct journal_head **jh_out,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200825 unsigned int blocknr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
827/* Transaction locking */
828extern void __wait_on_journal (journal_t *);
829
830/*
831 * Journal locking.
832 *
833 * We need to lock the journal during transaction state changes so that nobody
834 * ever tries to take a handle on the running transaction while we are in the
835 * middle of moving it to the commit phase. j_state_lock does this.
836 *
837 * Note that the locking is completely interrupt unsafe. We never touch
838 * journal structures from interrupts.
839 */
840
841static inline handle_t *journal_current_handle(void)
842{
843 return current->journal_info;
844}
845
846/* The journaling code user interface:
847 *
848 * Create and destroy handles
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700849 * Register buffer modifications against the current transaction.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 */
851
852extern handle_t *journal_start(journal_t *, int nblocks);
853extern int journal_restart (handle_t *, int nblocks);
854extern int journal_extend (handle_t *, int nblocks);
855extern int journal_get_write_access(handle_t *, struct buffer_head *);
856extern int journal_get_create_access (handle_t *, struct buffer_head *);
857extern int journal_get_undo_access(handle_t *, struct buffer_head *);
858extern int journal_dirty_data (handle_t *, struct buffer_head *);
859extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
860extern void journal_release_buffer (handle_t *, struct buffer_head *);
861extern int journal_forget (handle_t *, struct buffer_head *);
862extern void journal_sync_buffer (struct buffer_head *);
NeilBrown2ff28e22006-03-26 01:37:18 -0800863extern void journal_invalidatepage(journal_t *,
Lukas Czernerd8c89002013-05-21 23:26:36 -0400864 struct page *, unsigned int, unsigned int);
Al Viro27496a82005-10-21 03:20:48 -0400865extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866extern int journal_stop(handle_t *);
867extern int journal_flush (journal_t *);
868extern void journal_lock_updates (journal_t *);
869extern void journal_unlock_updates (journal_t *);
870
871extern journal_t * journal_init_dev(struct block_device *bdev,
872 struct block_device *fs_dev,
873 int start, int len, int bsize);
874extern journal_t * journal_init_inode (struct inode *);
875extern int journal_update_format (journal_t *);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700876extern int journal_check_used_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700878extern int journal_check_available_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700880extern int journal_set_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881 (journal_t *, unsigned long, unsigned long, unsigned long);
882extern int journal_create (journal_t *);
883extern int journal_load (journal_t *journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -0700884extern int journal_destroy (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885extern int journal_recover (journal_t *journal);
886extern int journal_wipe (journal_t *, int);
887extern int journal_skip_recovery (journal_t *);
Jan Karafd2cbd42012-04-07 11:05:19 +0200888extern void journal_update_sb_log_tail (journal_t *, tid_t, unsigned int,
889 int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890extern void journal_abort (journal_t *, int);
891extern int journal_errno (journal_t *);
892extern void journal_ack_err (journal_t *);
893extern int journal_clear_err (journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200894extern int journal_bmap(journal_t *, unsigned int, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895extern int journal_force_commit(journal_t *);
896
897/*
898 * journal_head management
899 */
900struct journal_head *journal_add_journal_head(struct buffer_head *bh);
901struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902void journal_put_journal_head(struct journal_head *jh);
903
904/*
905 * handle management
906 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800907extern struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908
Al Virodd0fc662005-10-07 07:46:04 +0100909static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910{
Zheng Liue162b2f2013-04-12 02:24:06 +0000911 return kmem_cache_zalloc(jbd_handle_cache, gfp_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912}
913
914static inline void jbd_free_handle(handle_t *handle)
915{
916 kmem_cache_free(jbd_handle_cache, handle);
917}
918
919/* Primary revoke support */
920#define JOURNAL_REVOKE_DEFAULT_HASH 256
921extern int journal_init_revoke(journal_t *, int);
922extern void journal_destroy_revoke_caches(void);
923extern int journal_init_revoke_caches(void);
924
925extern void journal_destroy_revoke(journal_t *);
926extern int journal_revoke (handle_t *,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200927 unsigned int, struct buffer_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928extern int journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o38d726d2009-04-14 10:10:47 -0400929extern void journal_write_revoke_records(journal_t *,
930 transaction_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931
932/* Recovery revoke support */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200933extern int journal_set_revoke(journal_t *, unsigned int, tid_t);
934extern int journal_test_revoke(journal_t *, unsigned int, tid_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700935extern void journal_clear_revoke(journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936extern void journal_switch_revoke_table(journal_t *journal);
Yongqiang Yang8c111b32011-11-19 17:34:29 +0800937extern void journal_clear_buffer_revoked_flags(journal_t *journal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938
939/*
940 * The log thread user interface:
941 *
942 * Request space in the current transaction, and force transaction commit
943 * transitions on demand.
944 */
945
946int __log_space_left(journal_t *); /* Called with journal locked */
947int log_start_commit(journal_t *journal, tid_t tid);
948int __log_start_commit(journal_t *journal, tid_t tid);
949int journal_start_commit(journal_t *journal, tid_t *tid);
950int journal_force_commit_nested(journal_t *journal);
951int log_wait_commit(journal_t *journal, tid_t tid);
952int log_do_checkpoint(journal_t *journal);
Jan Kara03f4d802010-04-15 22:16:24 +0200953int journal_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700954
955void __log_wait_for_space(journal_t *journal);
956extern void __journal_drop_transaction(journal_t *, transaction_t *);
957extern int cleanup_journal_tail(journal_t *);
958
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959/*
960 * is_journal_abort
961 *
962 * Simple test wrapper function to test the JFS_ABORT state flag. This
963 * bit, when set, indicates that we have had a fatal error somewhere,
964 * either inside the journaling layer or indicated to us by the client
965 * (eg. ext3), and that we and should not commit any further
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700966 * transactions.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 */
968
969static inline int is_journal_aborted(journal_t *journal)
970{
971 return journal->j_flags & JFS_ABORT;
972}
973
974static inline int is_handle_aborted(handle_t *handle)
975{
976 if (handle->h_aborted)
977 return 1;
978 return is_journal_aborted(handle->h_transaction->t_journal);
979}
980
981static inline void journal_abort_handle(handle_t *handle)
982{
983 handle->h_aborted = 1;
984}
985
986#endif /* __KERNEL__ */
987
988/* Comparison functions for transaction IDs: perform comparisons using
989 * modulo arithmetic so that they work over sequence number wraps. */
990
991static inline int tid_gt(tid_t x, tid_t y)
992{
993 int difference = (x - y);
994 return (difference > 0);
995}
996
997static inline int tid_geq(tid_t x, tid_t y)
998{
999 int difference = (x - y);
1000 return (difference >= 0);
1001}
1002
1003extern int journal_blocks_per_page(struct inode *inode);
1004
1005/*
1006 * Return the minimum number of blocks which must be free in the journal
1007 * before a new transaction may be started. Must be called under j_state_lock.
1008 */
1009static inline int jbd_space_needed(journal_t *journal)
1010{
1011 int nblocks = journal->j_max_transaction_buffers;
1012 if (journal->j_committing_transaction)
1013 nblocks += journal->j_committing_transaction->
1014 t_outstanding_credits;
1015 return nblocks;
1016}
1017
1018/*
1019 * Definitions which augment the buffer_head layer
1020 */
1021
1022/* journaling buffer types */
1023#define BJ_None 0 /* Not journaled */
1024#define BJ_SyncData 1 /* Normal data: flush before commit */
1025#define BJ_Metadata 2 /* Normal journaled metadata */
1026#define BJ_Forget 3 /* Buffer superseded by this transaction */
1027#define BJ_IO 4 /* Buffer is for temporary IO use */
1028#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1029#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1030#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1031#define BJ_Locked 8 /* Locked for I/O during commit */
1032#define BJ_Types 9
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001033
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034extern int jbd_blocks_per_page(struct inode *inode);
1035
1036#ifdef __KERNEL__
1037
1038#define buffer_trace_init(bh) do {} while (0)
1039#define print_buffer_fields(bh) do {} while (0)
1040#define print_buffer_trace(bh) do {} while (0)
1041#define BUFFER_TRACE(bh, info) do {} while (0)
1042#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1043#define JBUFFER_TRACE(jh, info) do {} while (0)
1044
1045#endif /* __KERNEL__ */
1046
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047#endif /* _LINUX_JBD_H */