blob: 8685d1be12c71a2d28d8be80b3bbe9ae507e1ee4 [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
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>
Jan Karab34090e2013-06-04 12:08:56 -0400246
247enum jbd_state_bits {
248 BH_JBD /* Has an attached ext3 journal_head */
249 = BH_PrivateStart,
250 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
251 BH_Freed, /* Has been freed (truncated) */
252 BH_Revoked, /* Has been revoked from the log */
253 BH_RevokeValid, /* Revoked flag is valid */
254 BH_JBDDirty, /* Is dirty but journaled */
255 BH_State, /* Pins most journal_head state */
256 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
257 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
258 BH_JBDPrivateStart, /* First bit available for private use by FS */
259};
260
261BUFFER_FNS(JBD, jbd)
262BUFFER_FNS(JWrite, jwrite)
263BUFFER_FNS(JBDDirty, jbddirty)
264TAS_BUFFER_FNS(JBDDirty, jbddirty)
265BUFFER_FNS(Revoked, revoked)
266TAS_BUFFER_FNS(Revoked, revoked)
267BUFFER_FNS(RevokeValid, revokevalid)
268TAS_BUFFER_FNS(RevokeValid, revokevalid)
269BUFFER_FNS(Freed, freed)
270
Thomas Gleixner44606672011-10-27 04:38:18 -0400271#include <linux/jbd_common.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272
Chris Snook1c099242007-10-18 23:39:21 -0700273#define J_ASSERT(assert) BUG_ON(!(assert))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
276#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278#if defined(JBD_PARANOID_IOFAIL)
279#define J_EXPECT(expr, why...) J_ASSERT(expr)
280#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
281#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
282#else
283#define __journal_expect(expr, why...) \
284 ({ \
285 int val = (expr); \
286 if (!val) { \
287 printk(KERN_ERR \
288 "EXT3-fs unexpected failure: %s;\n",# expr); \
289 printk(KERN_ERR why "\n"); \
290 } \
291 val; \
292 })
293#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
294#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
295#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
296#endif
297
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298struct jbd_revoke_table_s;
299
300/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700301 * struct handle_s - this is the concrete type associated with handle_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302 * @h_transaction: Which compound transaction is this update a part of?
303 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
304 * @h_ref: Reference count on this handle
305 * @h_err: Field for caller's use to track errors through large fs operations
306 * @h_sync: flag for sync-on-close
307 * @h_jdata: flag to force data journaling
308 * @h_aborted: flag indicating fatal error on handle
Randy Dunlape6716b82007-10-15 17:30:19 -0700309 * @h_lockdep_map: lockdep info for debugging lock problems
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310 */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700311struct handle_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312{
313 /* Which compound transaction is this update a part of? */
314 transaction_t *h_transaction;
315
316 /* Number of remaining buffers we are allowed to dirty: */
317 int h_buffer_credits;
318
319 /* Reference count on this handle */
320 int h_ref;
321
322 /* Field for caller's use to track errors through large fs */
323 /* operations */
324 int h_err;
325
326 /* Flags [no locking] */
327 unsigned int h_sync: 1; /* sync-on-close */
328 unsigned int h_jdata: 1; /* force data journaling */
329 unsigned int h_aborted: 1; /* fatal error on handle */
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +0200330
331#ifdef CONFIG_DEBUG_LOCK_ALLOC
332 struct lockdep_map h_lockdep_map;
333#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334};
335
336
337/* The transaction_t type is the guts of the journaling mechanism. It
338 * tracks a compound transaction through its various states:
339 *
340 * RUNNING: accepting new updates
341 * LOCKED: Updates still running but we don't accept new ones
342 * RUNDOWN: Updates are tidying up but have finished requesting
343 * new buffers to modify (state not used for now)
344 * FLUSH: All updates complete, but we are still writing to disk
345 * COMMIT: All data on disk, writing commit record
346 * FINISHED: We still have to keep the transaction for checkpointing.
347 *
348 * The transaction keeps track of all of the buffers modified by a
349 * running transaction, and all of the buffers committed but not yet
350 * flushed to home for finished transactions.
351 */
352
353/*
354 * Lock ranking:
355 *
356 * j_list_lock
357 * ->jbd_lock_bh_journal_head() (This is "innermost")
358 *
359 * j_state_lock
360 * ->jbd_lock_bh_state()
361 *
362 * jbd_lock_bh_state()
363 * ->j_list_lock
364 *
365 * j_state_lock
366 * ->t_handle_lock
367 *
368 * j_state_lock
369 * ->j_list_lock (journal_unmap_buffer)
370 *
371 */
372
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700373struct transaction_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374{
375 /* Pointer to the journal for this transaction. [no locking] */
376 journal_t *t_journal;
377
378 /* Sequence number for this transaction [no locking] */
379 tid_t t_tid;
380
381 /*
382 * Transaction's current state
383 * [no locking - only kjournald alters this]
Jan Karad4beaf42007-12-04 23:45:27 -0800384 * [j_list_lock] guards transition of a transaction into T_FINISHED
385 * state and subsequent call of __journal_drop_transaction()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 * FIXME: needs barriers
387 * KLUDGE: [use j_state_lock]
388 */
389 enum {
390 T_RUNNING,
391 T_LOCKED,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700392 T_FLUSH,
393 T_COMMIT,
Jan Kara03f4d802010-04-15 22:16:24 +0200394 T_COMMIT_RECORD,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700395 T_FINISHED
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 } t_state;
397
398 /*
399 * Where in the log does this transaction's commit start? [no locking]
400 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200401 unsigned int t_log_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402
403 /* Number of buffers on the t_buffers list [j_list_lock] */
404 int t_nr_buffers;
405
406 /*
407 * Doubly-linked circular list of all buffers reserved but not yet
408 * modified by this transaction [j_list_lock]
409 */
410 struct journal_head *t_reserved_list;
411
412 /*
413 * Doubly-linked circular list of all buffers under writeout during
414 * commit [j_list_lock]
415 */
416 struct journal_head *t_locked_list;
417
418 /*
419 * Doubly-linked circular list of all metadata buffers owned by this
420 * transaction [j_list_lock]
421 */
422 struct journal_head *t_buffers;
423
424 /*
425 * Doubly-linked circular list of all data buffers still to be
426 * flushed before this transaction can be committed [j_list_lock]
427 */
428 struct journal_head *t_sync_datalist;
429
430 /*
431 * Doubly-linked circular list of all forget buffers (superseded
432 * buffers which we can un-checkpoint once this transaction commits)
433 * [j_list_lock]
434 */
435 struct journal_head *t_forget;
436
437 /*
438 * Doubly-linked circular list of all buffers still to be flushed before
439 * this transaction can be checkpointed. [j_list_lock]
440 */
441 struct journal_head *t_checkpoint_list;
442
443 /*
Jan Kara78ce89c2006-06-23 02:06:05 -0700444 * Doubly-linked circular list of all buffers submitted for IO while
445 * checkpointing. [j_list_lock]
446 */
447 struct journal_head *t_checkpoint_io_list;
448
449 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 * Doubly-linked circular list of temporary buffers currently undergoing
451 * IO in the log [j_list_lock]
452 */
453 struct journal_head *t_iobuf_list;
454
455 /*
456 * Doubly-linked circular list of metadata buffers being shadowed by log
457 * IO. The IO buffers on the iobuf list and the shadow buffers on this
458 * list match each other one for one at all times. [j_list_lock]
459 */
460 struct journal_head *t_shadow_list;
461
462 /*
463 * Doubly-linked circular list of control buffers being written to the
464 * log. [j_list_lock]
465 */
466 struct journal_head *t_log_list;
467
468 /*
469 * Protects info related to handles
470 */
471 spinlock_t t_handle_lock;
472
473 /*
474 * Number of outstanding updates running on this transaction
475 * [t_handle_lock]
476 */
477 int t_updates;
478
479 /*
480 * Number of buffers reserved for use by all handles in this transaction
481 * handle but not yet modified. [t_handle_lock]
482 */
483 int t_outstanding_credits;
484
485 /*
486 * Forward and backward links for the circular list of all transactions
487 * awaiting checkpoint. [j_list_lock]
488 */
489 transaction_t *t_cpnext, *t_cpprev;
490
491 /*
492 * When will the transaction expire (become due for commit), in jiffies?
493 * [no locking]
494 */
495 unsigned long t_expires;
496
497 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800498 * When this transaction started, in nanoseconds [no locking]
499 */
500 ktime_t t_start_time;
501
502 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 * How many handles used this transaction? [t_handle_lock]
504 */
505 int t_handle_count;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506};
507
508/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700509 * struct journal_s - this is the concrete type associated with journal_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 * @j_flags: General journaling state flags
511 * @j_errno: Is there an outstanding uncleared error on the journal (from a
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700512 * prior abort)?
Linus Torvalds1da177e2005-04-16 15:20:36 -0700513 * @j_sb_buffer: First part of superblock buffer
514 * @j_superblock: Second part of superblock buffer
515 * @j_format_version: Version of the superblock format
516 * @j_state_lock: Protect the various scalars in the journal
517 * @j_barrier_count: Number of processes waiting to create a barrier lock
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 * @j_running_transaction: The current running transaction..
519 * @j_committing_transaction: the transaction we are pushing to disk
520 * @j_checkpoint_transactions: a linked circular list of all transactions
521 * waiting for checkpointing
522 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
523 * to start committing, or for a barrier lock to be released
524 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700525 * @j_wait_done_commit: Wait queue for waiting for commit to complete
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 * @j_wait_checkpoint: Wait queue to trigger checkpointing
527 * @j_wait_commit: Wait queue to trigger commit
528 * @j_wait_updates: Wait queue to wait for updates to complete
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800529 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 * @j_head: Journal head - identifies the first unused block in the journal
531 * @j_tail: Journal tail - identifies the oldest still-used block in the
532 * journal.
533 * @j_free: Journal free - how many free blocks are there in the journal?
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700534 * @j_first: The block number of the first usable block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 * @j_last: The block number one beyond the last usable block
536 * @j_dev: Device where we store the journal
537 * @j_blocksize: blocksize for the location where we store the journal.
538 * @j_blk_offset: starting block offset for into the device where we store the
539 * journal
540 * @j_fs_dev: Device which holds the client fs. For internal journal this will
541 * be equal to j_dev
542 * @j_maxlen: Total maximum capacity of the journal region on disk.
543 * @j_list_lock: Protects the buffer lists and internal buffer state.
544 * @j_inode: Optional inode where we store the journal. If present, all journal
545 * block numbers are mapped into this inode via bmap().
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700546 * @j_tail_sequence: Sequence number of the oldest transaction in the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 * @j_transaction_sequence: Sequence number of the next transaction to grant
548 * @j_commit_sequence: Sequence number of the most recently committed
549 * transaction
550 * @j_commit_request: Sequence number of the most recent transaction wanting
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700551 * commit
Jan Kara2db938b2011-02-21 17:25:37 +0100552 * @j_commit_waited: Sequence number of the most recent transaction someone
553 * is waiting for to commit.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554 * @j_uuid: Uuid of client object.
555 * @j_task: Pointer to the current commit thread for this journal
556 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
557 * single compound commit transaction
558 * @j_commit_interval: What is the maximum transaction lifetime before we begin
559 * a commit?
560 * @j_commit_timer: The timer used to wakeup the commit thread
561 * @j_revoke_lock: Protect the revoke table
562 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
563 * current transaction.
564 * @j_revoke_table: alternate revoke tables for j_revoke
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800565 * @j_wbuf: array of buffer_heads for journal_commit_transaction
566 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
567 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800568 * @j_last_sync_writer: most recent pid which did a synchronous write
Randy Dunlap3eabdb72009-01-15 13:51:01 -0800569 * @j_average_commit_time: the average amount of time in nanoseconds it
570 * takes to commit a transaction to the disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 * @j_private: An opaque pointer to fs-private information.
572 */
573
574struct journal_s
575{
576 /* General journaling state flags [j_state_lock] */
577 unsigned long j_flags;
578
579 /*
580 * Is there an outstanding uncleared error on the journal (from a prior
581 * abort)? [j_state_lock]
582 */
583 int j_errno;
584
585 /* The superblock buffer */
586 struct buffer_head *j_sb_buffer;
587 journal_superblock_t *j_superblock;
588
589 /* Version of the superblock format */
590 int j_format_version;
591
592 /*
593 * Protect the various scalars in the journal
594 */
595 spinlock_t j_state_lock;
596
597 /*
598 * Number of processes waiting to create a barrier lock [j_state_lock]
599 */
600 int j_barrier_count;
601
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 /*
603 * Transactions: The current running transaction...
604 * [j_state_lock] [caller holding open handle]
605 */
606 transaction_t *j_running_transaction;
607
608 /*
609 * the transaction we are pushing to disk
610 * [j_state_lock] [caller holding open handle]
611 */
612 transaction_t *j_committing_transaction;
613
614 /*
615 * ... and a linked circular list of all transactions waiting for
616 * checkpointing. [j_list_lock]
617 */
618 transaction_t *j_checkpoint_transactions;
619
620 /*
621 * Wait queue for waiting for a locked transaction to start committing,
622 * or for a barrier lock to be released
623 */
624 wait_queue_head_t j_wait_transaction_locked;
625
626 /* Wait queue for waiting for checkpointing to complete */
627 wait_queue_head_t j_wait_logspace;
628
629 /* Wait queue for waiting for commit to complete */
630 wait_queue_head_t j_wait_done_commit;
631
632 /* Wait queue to trigger checkpointing */
633 wait_queue_head_t j_wait_checkpoint;
634
635 /* Wait queue to trigger commit */
636 wait_queue_head_t j_wait_commit;
637
638 /* Wait queue to wait for updates to complete */
639 wait_queue_head_t j_wait_updates;
640
641 /* Semaphore for locking against concurrent checkpoints */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700642 struct mutex j_checkpoint_mutex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643
644 /*
645 * Journal head: identifies the first unused block in the journal.
646 * [j_state_lock]
647 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200648 unsigned int j_head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
650 /*
651 * Journal tail: identifies the oldest still-used block in the journal.
652 * [j_state_lock]
653 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200654 unsigned int j_tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
656 /*
657 * Journal free: how many free blocks are there in the journal?
658 * [j_state_lock]
659 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200660 unsigned int j_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661
662 /*
663 * Journal start and end: the block numbers of the first usable block
664 * and one beyond the last usable block in the journal. [j_state_lock]
665 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200666 unsigned int j_first;
667 unsigned int j_last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
669 /*
670 * Device, blocksize and starting block offset for the location where we
671 * store the journal.
672 */
673 struct block_device *j_dev;
674 int j_blocksize;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200675 unsigned int j_blk_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
677 /*
678 * Device which holds the client fs. For internal journal this will be
679 * equal to j_dev.
680 */
681 struct block_device *j_fs_dev;
682
683 /* Total maximum capacity of the journal region on disk. */
684 unsigned int j_maxlen;
685
686 /*
687 * Protects the buffer lists and internal buffer state.
688 */
689 spinlock_t j_list_lock;
690
691 /* Optional inode where we store the journal. If present, all */
692 /* journal block numbers are mapped into this inode via */
693 /* bmap(). */
694 struct inode *j_inode;
695
696 /*
697 * Sequence number of the oldest transaction in the log [j_state_lock]
698 */
699 tid_t j_tail_sequence;
700
701 /*
702 * Sequence number of the next transaction to grant [j_state_lock]
703 */
704 tid_t j_transaction_sequence;
705
706 /*
707 * Sequence number of the most recently committed transaction
708 * [j_state_lock].
709 */
710 tid_t j_commit_sequence;
711
712 /*
713 * Sequence number of the most recent transaction wanting commit
714 * [j_state_lock]
715 */
716 tid_t j_commit_request;
717
718 /*
Jan Kara2db938b2011-02-21 17:25:37 +0100719 * Sequence number of the most recent transaction someone is waiting
720 * for to commit.
721 * [j_state_lock]
722 */
723 tid_t j_commit_waited;
724
725 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700726 * Journal uuid: identifies the object (filesystem, LVM volume etc)
727 * backed by this journal. This will eventually be replaced by an array
728 * of uuids, allowing us to index multiple devices within a single
729 * journal and to perform atomic updates across them.
730 */
731 __u8 j_uuid[16];
732
733 /* Pointer to the current commit thread for this journal */
734 struct task_struct *j_task;
735
736 /*
737 * Maximum number of metadata buffers to allow in a single compound
738 * commit transaction
739 */
740 int j_max_transaction_buffers;
741
742 /*
743 * What is the maximum transaction lifetime before we begin a commit?
744 */
745 unsigned long j_commit_interval;
746
747 /* The timer used to wakeup the commit thread: */
Andrew Mortone3df1892006-03-25 03:06:53 -0800748 struct timer_list j_commit_timer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
750 /*
751 * The revoke table: maintains the list of revoked blocks in the
752 * current transaction. [j_revoke_lock]
753 */
754 spinlock_t j_revoke_lock;
755 struct jbd_revoke_table_s *j_revoke;
756 struct jbd_revoke_table_s *j_revoke_table[2];
757
758 /*
759 * array of bhs for journal_commit_transaction
760 */
761 struct buffer_head **j_wbuf;
762 int j_wbufsize;
763
Josef Bacikf420d4d2009-01-07 18:07:24 -0800764 /*
765 * this is the pid of the last person to run a synchronous operation
766 * through the journal.
767 */
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800768 pid_t j_last_sync_writer;
769
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800771 * the average amount of time in nanoseconds it takes to commit a
772 * transaction to the disk. [j_state_lock]
773 */
774 u64 j_average_commit_time;
775
776 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 * An opaque pointer to fs-private information. ext3 puts its
778 * superblock pointer here
779 */
780 void *j_private;
781};
782
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700783/*
784 * Journal flag definitions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785 */
786#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
787#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
788#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
789#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
790#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
791#define JFS_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700792#define JFS_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
793 * data write error in ordered
794 * mode */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700795
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700796/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797 * Function declarations for the journaling transaction and buffer
798 * management
799 */
800
801/* Filing buffers */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802extern void journal_unfile_buffer(journal_t *, struct journal_head *);
803extern void __journal_unfile_buffer(struct journal_head *);
804extern void __journal_refile_buffer(struct journal_head *);
805extern void journal_refile_buffer(journal_t *, struct journal_head *);
806extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
807extern void __journal_free_buffer(struct journal_head *bh);
808extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
809extern void __journal_clean_data_list(transaction_t *transaction);
810
811/* Log buffer allocation */
812extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200813int journal_next_log_block(journal_t *, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
815/* Commit management */
816extern void journal_commit_transaction(journal_t *);
817
818/* Checkpoint list management */
819int __journal_clean_checkpoint_list(journal_t *journal);
Jan Kara78ce89c2006-06-23 02:06:05 -0700820int __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
822
823/* Buffer IO */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700824extern int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825journal_write_metadata_buffer(transaction_t *transaction,
826 struct journal_head *jh_in,
827 struct journal_head **jh_out,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200828 unsigned int blocknr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700829
830/* Transaction locking */
831extern void __wait_on_journal (journal_t *);
832
833/*
834 * Journal locking.
835 *
836 * We need to lock the journal during transaction state changes so that nobody
837 * ever tries to take a handle on the running transaction while we are in the
838 * middle of moving it to the commit phase. j_state_lock does this.
839 *
840 * Note that the locking is completely interrupt unsafe. We never touch
841 * journal structures from interrupts.
842 */
843
844static inline handle_t *journal_current_handle(void)
845{
846 return current->journal_info;
847}
848
849/* The journaling code user interface:
850 *
851 * Create and destroy handles
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700852 * Register buffer modifications against the current transaction.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 */
854
855extern handle_t *journal_start(journal_t *, int nblocks);
856extern int journal_restart (handle_t *, int nblocks);
857extern int journal_extend (handle_t *, int nblocks);
858extern int journal_get_write_access(handle_t *, struct buffer_head *);
859extern int journal_get_create_access (handle_t *, struct buffer_head *);
860extern int journal_get_undo_access(handle_t *, struct buffer_head *);
861extern int journal_dirty_data (handle_t *, struct buffer_head *);
862extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
863extern void journal_release_buffer (handle_t *, struct buffer_head *);
864extern int journal_forget (handle_t *, struct buffer_head *);
865extern void journal_sync_buffer (struct buffer_head *);
NeilBrown2ff28e22006-03-26 01:37:18 -0800866extern void journal_invalidatepage(journal_t *,
Lukas Czernerd8c89002013-05-21 23:26:36 -0400867 struct page *, unsigned int, unsigned int);
Al Viro27496a82005-10-21 03:20:48 -0400868extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869extern int journal_stop(handle_t *);
870extern int journal_flush (journal_t *);
871extern void journal_lock_updates (journal_t *);
872extern void journal_unlock_updates (journal_t *);
873
874extern journal_t * journal_init_dev(struct block_device *bdev,
875 struct block_device *fs_dev,
876 int start, int len, int bsize);
877extern journal_t * journal_init_inode (struct inode *);
878extern int journal_update_format (journal_t *);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700879extern int journal_check_used_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700881extern int journal_check_available_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700883extern int journal_set_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 (journal_t *, unsigned long, unsigned long, unsigned long);
885extern int journal_create (journal_t *);
886extern int journal_load (journal_t *journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -0700887extern int journal_destroy (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888extern int journal_recover (journal_t *journal);
889extern int journal_wipe (journal_t *, int);
890extern int journal_skip_recovery (journal_t *);
Jan Karafd2cbd42012-04-07 11:05:19 +0200891extern void journal_update_sb_log_tail (journal_t *, tid_t, unsigned int,
892 int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893extern void journal_abort (journal_t *, int);
894extern int journal_errno (journal_t *);
895extern void journal_ack_err (journal_t *);
896extern int journal_clear_err (journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200897extern int journal_bmap(journal_t *, unsigned int, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898extern int journal_force_commit(journal_t *);
899
900/*
901 * journal_head management
902 */
903struct journal_head *journal_add_journal_head(struct buffer_head *bh);
904struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905void journal_put_journal_head(struct journal_head *jh);
906
907/*
908 * handle management
909 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800910extern struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911
Al Virodd0fc662005-10-07 07:46:04 +0100912static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913{
Zheng Liue162b2f2013-04-12 02:24:06 +0000914 return kmem_cache_zalloc(jbd_handle_cache, gfp_flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916
917static inline void jbd_free_handle(handle_t *handle)
918{
919 kmem_cache_free(jbd_handle_cache, handle);
920}
921
922/* Primary revoke support */
923#define JOURNAL_REVOKE_DEFAULT_HASH 256
924extern int journal_init_revoke(journal_t *, int);
925extern void journal_destroy_revoke_caches(void);
926extern int journal_init_revoke_caches(void);
927
928extern void journal_destroy_revoke(journal_t *);
929extern int journal_revoke (handle_t *,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200930 unsigned int, struct buffer_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700931extern int journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o38d726d2009-04-14 10:10:47 -0400932extern void journal_write_revoke_records(journal_t *,
933 transaction_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934
935/* Recovery revoke support */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200936extern int journal_set_revoke(journal_t *, unsigned int, tid_t);
937extern int journal_test_revoke(journal_t *, unsigned int, tid_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938extern void journal_clear_revoke(journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939extern void journal_switch_revoke_table(journal_t *journal);
Yongqiang Yang8c111b32011-11-19 17:34:29 +0800940extern void journal_clear_buffer_revoked_flags(journal_t *journal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700941
942/*
943 * The log thread user interface:
944 *
945 * Request space in the current transaction, and force transaction commit
946 * transitions on demand.
947 */
948
949int __log_space_left(journal_t *); /* Called with journal locked */
950int log_start_commit(journal_t *journal, tid_t tid);
951int __log_start_commit(journal_t *journal, tid_t tid);
952int journal_start_commit(journal_t *journal, tid_t *tid);
953int journal_force_commit_nested(journal_t *journal);
954int log_wait_commit(journal_t *journal, tid_t tid);
955int log_do_checkpoint(journal_t *journal);
Jan Kara03f4d802010-04-15 22:16:24 +0200956int journal_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957
958void __log_wait_for_space(journal_t *journal);
959extern void __journal_drop_transaction(journal_t *, transaction_t *);
960extern int cleanup_journal_tail(journal_t *);
961
962/* Debugging code only: */
963
964#define jbd_ENOSYS() \
965do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -0700966 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700967 current->state = TASK_UNINTERRUPTIBLE; \
968 schedule(); \
969} while (1)
970
971/*
972 * is_journal_abort
973 *
974 * Simple test wrapper function to test the JFS_ABORT state flag. This
975 * bit, when set, indicates that we have had a fatal error somewhere,
976 * either inside the journaling layer or indicated to us by the client
977 * (eg. ext3), and that we and should not commit any further
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700978 * transactions.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 */
980
981static inline int is_journal_aborted(journal_t *journal)
982{
983 return journal->j_flags & JFS_ABORT;
984}
985
986static inline int is_handle_aborted(handle_t *handle)
987{
988 if (handle->h_aborted)
989 return 1;
990 return is_journal_aborted(handle->h_transaction->t_journal);
991}
992
993static inline void journal_abort_handle(handle_t *handle)
994{
995 handle->h_aborted = 1;
996}
997
998#endif /* __KERNEL__ */
999
1000/* Comparison functions for transaction IDs: perform comparisons using
1001 * modulo arithmetic so that they work over sequence number wraps. */
1002
1003static inline int tid_gt(tid_t x, tid_t y)
1004{
1005 int difference = (x - y);
1006 return (difference > 0);
1007}
1008
1009static inline int tid_geq(tid_t x, tid_t y)
1010{
1011 int difference = (x - y);
1012 return (difference >= 0);
1013}
1014
1015extern int journal_blocks_per_page(struct inode *inode);
1016
1017/*
1018 * Return the minimum number of blocks which must be free in the journal
1019 * before a new transaction may be started. Must be called under j_state_lock.
1020 */
1021static inline int jbd_space_needed(journal_t *journal)
1022{
1023 int nblocks = journal->j_max_transaction_buffers;
1024 if (journal->j_committing_transaction)
1025 nblocks += journal->j_committing_transaction->
1026 t_outstanding_credits;
1027 return nblocks;
1028}
1029
1030/*
1031 * Definitions which augment the buffer_head layer
1032 */
1033
1034/* journaling buffer types */
1035#define BJ_None 0 /* Not journaled */
1036#define BJ_SyncData 1 /* Normal data: flush before commit */
1037#define BJ_Metadata 2 /* Normal journaled metadata */
1038#define BJ_Forget 3 /* Buffer superseded by this transaction */
1039#define BJ_IO 4 /* Buffer is for temporary IO use */
1040#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1041#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1042#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1043#define BJ_Locked 8 /* Locked for I/O during commit */
1044#define BJ_Types 9
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001045
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046extern int jbd_blocks_per_page(struct inode *inode);
1047
1048#ifdef __KERNEL__
1049
1050#define buffer_trace_init(bh) do {} while (0)
1051#define print_buffer_fields(bh) do {} while (0)
1052#define print_buffer_trace(bh) do {} while (0)
1053#define BUFFER_TRACE(bh, info) do {} while (0)
1054#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1055#define JBUFFER_TRACE(jh, info) do {} while (0)
1056
1057#endif /* __KERNEL__ */
1058
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059#endif /* _LINUX_JBD_H */