blob: d211732b9e99c8a6d95bff50b56505c651285a70 [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>
Thomas Gleixner44606672011-10-27 04:38:18 -0400247#include <linux/jbd_common.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
Chris Snook1c099242007-10-18 23:39:21 -0700249#define J_ASSERT(assert) BUG_ON(!(assert))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
252#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254#if defined(JBD_PARANOID_IOFAIL)
255#define J_EXPECT(expr, why...) J_ASSERT(expr)
256#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
257#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
258#else
259#define __journal_expect(expr, why...) \
260 ({ \
261 int val = (expr); \
262 if (!val) { \
263 printk(KERN_ERR \
264 "EXT3-fs unexpected failure: %s;\n",# expr); \
265 printk(KERN_ERR why "\n"); \
266 } \
267 val; \
268 })
269#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
270#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
271#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
272#endif
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274struct jbd_revoke_table_s;
275
276/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700277 * struct handle_s - this is the concrete type associated with handle_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 * @h_transaction: Which compound transaction is this update a part of?
279 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
280 * @h_ref: Reference count on this handle
281 * @h_err: Field for caller's use to track errors through large fs operations
282 * @h_sync: flag for sync-on-close
283 * @h_jdata: flag to force data journaling
284 * @h_aborted: flag indicating fatal error on handle
Randy Dunlape6716b82007-10-15 17:30:19 -0700285 * @h_lockdep_map: lockdep info for debugging lock problems
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700287struct handle_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288{
289 /* Which compound transaction is this update a part of? */
290 transaction_t *h_transaction;
291
292 /* Number of remaining buffers we are allowed to dirty: */
293 int h_buffer_credits;
294
295 /* Reference count on this handle */
296 int h_ref;
297
298 /* Field for caller's use to track errors through large fs */
299 /* operations */
300 int h_err;
301
302 /* Flags [no locking] */
303 unsigned int h_sync: 1; /* sync-on-close */
304 unsigned int h_jdata: 1; /* force data journaling */
305 unsigned int h_aborted: 1; /* fatal error on handle */
Peter Zijlstra34a3d1e2007-10-11 22:11:12 +0200306
307#ifdef CONFIG_DEBUG_LOCK_ALLOC
308 struct lockdep_map h_lockdep_map;
309#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310};
311
312
313/* The transaction_t type is the guts of the journaling mechanism. It
314 * tracks a compound transaction through its various states:
315 *
316 * RUNNING: accepting new updates
317 * LOCKED: Updates still running but we don't accept new ones
318 * RUNDOWN: Updates are tidying up but have finished requesting
319 * new buffers to modify (state not used for now)
320 * FLUSH: All updates complete, but we are still writing to disk
321 * COMMIT: All data on disk, writing commit record
322 * FINISHED: We still have to keep the transaction for checkpointing.
323 *
324 * The transaction keeps track of all of the buffers modified by a
325 * running transaction, and all of the buffers committed but not yet
326 * flushed to home for finished transactions.
327 */
328
329/*
330 * Lock ranking:
331 *
332 * j_list_lock
333 * ->jbd_lock_bh_journal_head() (This is "innermost")
334 *
335 * j_state_lock
336 * ->jbd_lock_bh_state()
337 *
338 * jbd_lock_bh_state()
339 * ->j_list_lock
340 *
341 * j_state_lock
342 * ->t_handle_lock
343 *
344 * j_state_lock
345 * ->j_list_lock (journal_unmap_buffer)
346 *
347 */
348
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700349struct transaction_s
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350{
351 /* Pointer to the journal for this transaction. [no locking] */
352 journal_t *t_journal;
353
354 /* Sequence number for this transaction [no locking] */
355 tid_t t_tid;
356
357 /*
358 * Transaction's current state
359 * [no locking - only kjournald alters this]
Jan Karad4beaf42007-12-04 23:45:27 -0800360 * [j_list_lock] guards transition of a transaction into T_FINISHED
361 * state and subsequent call of __journal_drop_transaction()
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 * FIXME: needs barriers
363 * KLUDGE: [use j_state_lock]
364 */
365 enum {
366 T_RUNNING,
367 T_LOCKED,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 T_FLUSH,
369 T_COMMIT,
Jan Kara03f4d802010-04-15 22:16:24 +0200370 T_COMMIT_RECORD,
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700371 T_FINISHED
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 } t_state;
373
374 /*
375 * Where in the log does this transaction's commit start? [no locking]
376 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200377 unsigned int t_log_start;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378
379 /* Number of buffers on the t_buffers list [j_list_lock] */
380 int t_nr_buffers;
381
382 /*
383 * Doubly-linked circular list of all buffers reserved but not yet
384 * modified by this transaction [j_list_lock]
385 */
386 struct journal_head *t_reserved_list;
387
388 /*
389 * Doubly-linked circular list of all buffers under writeout during
390 * commit [j_list_lock]
391 */
392 struct journal_head *t_locked_list;
393
394 /*
395 * Doubly-linked circular list of all metadata buffers owned by this
396 * transaction [j_list_lock]
397 */
398 struct journal_head *t_buffers;
399
400 /*
401 * Doubly-linked circular list of all data buffers still to be
402 * flushed before this transaction can be committed [j_list_lock]
403 */
404 struct journal_head *t_sync_datalist;
405
406 /*
407 * Doubly-linked circular list of all forget buffers (superseded
408 * buffers which we can un-checkpoint once this transaction commits)
409 * [j_list_lock]
410 */
411 struct journal_head *t_forget;
412
413 /*
414 * Doubly-linked circular list of all buffers still to be flushed before
415 * this transaction can be checkpointed. [j_list_lock]
416 */
417 struct journal_head *t_checkpoint_list;
418
419 /*
Jan Kara78ce89c2006-06-23 02:06:05 -0700420 * Doubly-linked circular list of all buffers submitted for IO while
421 * checkpointing. [j_list_lock]
422 */
423 struct journal_head *t_checkpoint_io_list;
424
425 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 * Doubly-linked circular list of temporary buffers currently undergoing
427 * IO in the log [j_list_lock]
428 */
429 struct journal_head *t_iobuf_list;
430
431 /*
432 * Doubly-linked circular list of metadata buffers being shadowed by log
433 * IO. The IO buffers on the iobuf list and the shadow buffers on this
434 * list match each other one for one at all times. [j_list_lock]
435 */
436 struct journal_head *t_shadow_list;
437
438 /*
439 * Doubly-linked circular list of control buffers being written to the
440 * log. [j_list_lock]
441 */
442 struct journal_head *t_log_list;
443
444 /*
445 * Protects info related to handles
446 */
447 spinlock_t t_handle_lock;
448
449 /*
450 * Number of outstanding updates running on this transaction
451 * [t_handle_lock]
452 */
453 int t_updates;
454
455 /*
456 * Number of buffers reserved for use by all handles in this transaction
457 * handle but not yet modified. [t_handle_lock]
458 */
459 int t_outstanding_credits;
460
461 /*
462 * Forward and backward links for the circular list of all transactions
463 * awaiting checkpoint. [j_list_lock]
464 */
465 transaction_t *t_cpnext, *t_cpprev;
466
467 /*
468 * When will the transaction expire (become due for commit), in jiffies?
469 * [no locking]
470 */
471 unsigned long t_expires;
472
473 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800474 * When this transaction started, in nanoseconds [no locking]
475 */
476 ktime_t t_start_time;
477
478 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 * How many handles used this transaction? [t_handle_lock]
480 */
481 int t_handle_count;
482
Theodore Ts'o512a0042009-03-27 22:14:27 -0400483 /*
484 * This transaction is being forced and some process is
485 * waiting for it to finish.
486 */
H Hartley Sweeten4fefce92009-09-22 16:48:02 -0700487 unsigned int t_synchronous_commit:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488};
489
490/**
Randy Dunlapa6b91912008-03-19 17:01:00 -0700491 * struct journal_s - this is the concrete type associated with journal_t.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 * @j_flags: General journaling state flags
493 * @j_errno: Is there an outstanding uncleared error on the journal (from a
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700494 * prior abort)?
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 * @j_sb_buffer: First part of superblock buffer
496 * @j_superblock: Second part of superblock buffer
497 * @j_format_version: Version of the superblock format
498 * @j_state_lock: Protect the various scalars in the journal
499 * @j_barrier_count: Number of processes waiting to create a barrier lock
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 * @j_running_transaction: The current running transaction..
501 * @j_committing_transaction: the transaction we are pushing to disk
502 * @j_checkpoint_transactions: a linked circular list of all transactions
503 * waiting for checkpointing
504 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
505 * to start committing, or for a barrier lock to be released
506 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700507 * @j_wait_done_commit: Wait queue for waiting for commit to complete
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 * @j_wait_checkpoint: Wait queue to trigger checkpointing
509 * @j_wait_commit: Wait queue to trigger commit
510 * @j_wait_updates: Wait queue to wait for updates to complete
Arjan van de Ven2c68ee72006-03-23 03:00:35 -0800511 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 * @j_head: Journal head - identifies the first unused block in the journal
513 * @j_tail: Journal tail - identifies the oldest still-used block in the
514 * journal.
515 * @j_free: Journal free - how many free blocks are there in the journal?
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700516 * @j_first: The block number of the first usable block
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 * @j_last: The block number one beyond the last usable block
518 * @j_dev: Device where we store the journal
519 * @j_blocksize: blocksize for the location where we store the journal.
520 * @j_blk_offset: starting block offset for into the device where we store the
521 * journal
522 * @j_fs_dev: Device which holds the client fs. For internal journal this will
523 * be equal to j_dev
524 * @j_maxlen: Total maximum capacity of the journal region on disk.
525 * @j_list_lock: Protects the buffer lists and internal buffer state.
526 * @j_inode: Optional inode where we store the journal. If present, all journal
527 * block numbers are mapped into this inode via bmap().
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700528 * @j_tail_sequence: Sequence number of the oldest transaction in the log
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 * @j_transaction_sequence: Sequence number of the next transaction to grant
530 * @j_commit_sequence: Sequence number of the most recently committed
531 * transaction
532 * @j_commit_request: Sequence number of the most recent transaction wanting
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700533 * commit
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534 * @j_uuid: Uuid of client object.
535 * @j_task: Pointer to the current commit thread for this journal
536 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
537 * single compound commit transaction
538 * @j_commit_interval: What is the maximum transaction lifetime before we begin
539 * a commit?
540 * @j_commit_timer: The timer used to wakeup the commit thread
541 * @j_revoke_lock: Protect the revoke table
542 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
543 * current transaction.
544 * @j_revoke_table: alternate revoke tables for j_revoke
Randy Dunlap6c8bec62005-11-07 01:01:04 -0800545 * @j_wbuf: array of buffer_heads for journal_commit_transaction
546 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
547 * number that will fit in j_blocksize
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800548 * @j_last_sync_writer: most recent pid which did a synchronous write
Randy Dunlap3eabdb72009-01-15 13:51:01 -0800549 * @j_average_commit_time: the average amount of time in nanoseconds it
550 * takes to commit a transaction to the disk.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551 * @j_private: An opaque pointer to fs-private information.
552 */
553
554struct journal_s
555{
556 /* General journaling state flags [j_state_lock] */
557 unsigned long j_flags;
558
559 /*
560 * Is there an outstanding uncleared error on the journal (from a prior
561 * abort)? [j_state_lock]
562 */
563 int j_errno;
564
565 /* The superblock buffer */
566 struct buffer_head *j_sb_buffer;
567 journal_superblock_t *j_superblock;
568
569 /* Version of the superblock format */
570 int j_format_version;
571
572 /*
573 * Protect the various scalars in the journal
574 */
575 spinlock_t j_state_lock;
576
577 /*
578 * Number of processes waiting to create a barrier lock [j_state_lock]
579 */
580 int j_barrier_count;
581
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 /*
583 * Transactions: The current running transaction...
584 * [j_state_lock] [caller holding open handle]
585 */
586 transaction_t *j_running_transaction;
587
588 /*
589 * the transaction we are pushing to disk
590 * [j_state_lock] [caller holding open handle]
591 */
592 transaction_t *j_committing_transaction;
593
594 /*
595 * ... and a linked circular list of all transactions waiting for
596 * checkpointing. [j_list_lock]
597 */
598 transaction_t *j_checkpoint_transactions;
599
600 /*
601 * Wait queue for waiting for a locked transaction to start committing,
602 * or for a barrier lock to be released
603 */
604 wait_queue_head_t j_wait_transaction_locked;
605
606 /* Wait queue for waiting for checkpointing to complete */
607 wait_queue_head_t j_wait_logspace;
608
609 /* Wait queue for waiting for commit to complete */
610 wait_queue_head_t j_wait_done_commit;
611
612 /* Wait queue to trigger checkpointing */
613 wait_queue_head_t j_wait_checkpoint;
614
615 /* Wait queue to trigger commit */
616 wait_queue_head_t j_wait_commit;
617
618 /* Wait queue to wait for updates to complete */
619 wait_queue_head_t j_wait_updates;
620
621 /* Semaphore for locking against concurrent checkpoints */
Dave Kleikampe9ad5622006-09-27 01:49:35 -0700622 struct mutex j_checkpoint_mutex;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623
624 /*
625 * Journal head: identifies the first unused block in the journal.
626 * [j_state_lock]
627 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200628 unsigned int j_head;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
630 /*
631 * Journal tail: identifies the oldest still-used block in the journal.
632 * [j_state_lock]
633 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200634 unsigned int j_tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635
636 /*
637 * Journal free: how many free blocks are there in the journal?
638 * [j_state_lock]
639 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200640 unsigned int j_free;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641
642 /*
643 * Journal start and end: the block numbers of the first usable block
644 * and one beyond the last usable block in the journal. [j_state_lock]
645 */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200646 unsigned int j_first;
647 unsigned int j_last;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648
649 /*
650 * Device, blocksize and starting block offset for the location where we
651 * store the journal.
652 */
653 struct block_device *j_dev;
654 int j_blocksize;
Jan Kara9c28cbc2009-08-03 19:21:00 +0200655 unsigned int j_blk_offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700656
657 /*
658 * Device which holds the client fs. For internal journal this will be
659 * equal to j_dev.
660 */
661 struct block_device *j_fs_dev;
662
663 /* Total maximum capacity of the journal region on disk. */
664 unsigned int j_maxlen;
665
666 /*
667 * Protects the buffer lists and internal buffer state.
668 */
669 spinlock_t j_list_lock;
670
671 /* Optional inode where we store the journal. If present, all */
672 /* journal block numbers are mapped into this inode via */
673 /* bmap(). */
674 struct inode *j_inode;
675
676 /*
677 * Sequence number of the oldest transaction in the log [j_state_lock]
678 */
679 tid_t j_tail_sequence;
680
681 /*
682 * Sequence number of the next transaction to grant [j_state_lock]
683 */
684 tid_t j_transaction_sequence;
685
686 /*
687 * Sequence number of the most recently committed transaction
688 * [j_state_lock].
689 */
690 tid_t j_commit_sequence;
691
692 /*
693 * Sequence number of the most recent transaction wanting commit
694 * [j_state_lock]
695 */
696 tid_t j_commit_request;
697
698 /*
699 * Journal uuid: identifies the object (filesystem, LVM volume etc)
700 * backed by this journal. This will eventually be replaced by an array
701 * of uuids, allowing us to index multiple devices within a single
702 * journal and to perform atomic updates across them.
703 */
704 __u8 j_uuid[16];
705
706 /* Pointer to the current commit thread for this journal */
707 struct task_struct *j_task;
708
709 /*
710 * Maximum number of metadata buffers to allow in a single compound
711 * commit transaction
712 */
713 int j_max_transaction_buffers;
714
715 /*
716 * What is the maximum transaction lifetime before we begin a commit?
717 */
718 unsigned long j_commit_interval;
719
720 /* The timer used to wakeup the commit thread: */
Andrew Mortone3df1892006-03-25 03:06:53 -0800721 struct timer_list j_commit_timer;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722
723 /*
724 * The revoke table: maintains the list of revoked blocks in the
725 * current transaction. [j_revoke_lock]
726 */
727 spinlock_t j_revoke_lock;
728 struct jbd_revoke_table_s *j_revoke;
729 struct jbd_revoke_table_s *j_revoke_table[2];
730
731 /*
732 * array of bhs for journal_commit_transaction
733 */
734 struct buffer_head **j_wbuf;
735 int j_wbufsize;
736
Josef Bacikf420d4d2009-01-07 18:07:24 -0800737 /*
738 * this is the pid of the last person to run a synchronous operation
739 * through the journal.
740 */
Andrew Mortonfe1dcbc2006-02-04 23:27:54 -0800741 pid_t j_last_sync_writer;
742
Linus Torvalds1da177e2005-04-16 15:20:36 -0700743 /*
Josef Bacikf420d4d2009-01-07 18:07:24 -0800744 * the average amount of time in nanoseconds it takes to commit a
745 * transaction to the disk. [j_state_lock]
746 */
747 u64 j_average_commit_time;
748
749 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 * An opaque pointer to fs-private information. ext3 puts its
751 * superblock pointer here
752 */
753 void *j_private;
754};
755
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700756/*
757 * Journal flag definitions
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 */
759#define JFS_UNMOUNT 0x001 /* Journal thread is being destroyed */
760#define JFS_ABORT 0x002 /* Journaling has been aborted for errors. */
761#define JFS_ACK_ERR 0x004 /* The errno in the sb has been acked */
762#define JFS_FLUSHED 0x008 /* The journal superblock has been flushed */
763#define JFS_LOADED 0x010 /* The journal superblock has been loaded */
764#define JFS_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai0e4fb5e2008-10-18 20:27:57 -0700765#define JFS_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
766 * data write error in ordered
767 * mode */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700769/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770 * Function declarations for the journaling transaction and buffer
771 * management
772 */
773
774/* Filing buffers */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700775extern void journal_unfile_buffer(journal_t *, struct journal_head *);
776extern void __journal_unfile_buffer(struct journal_head *);
777extern void __journal_refile_buffer(struct journal_head *);
778extern void journal_refile_buffer(journal_t *, struct journal_head *);
779extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
780extern void __journal_free_buffer(struct journal_head *bh);
781extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
782extern void __journal_clean_data_list(transaction_t *transaction);
783
784/* Log buffer allocation */
785extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200786int journal_next_log_block(journal_t *, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787
788/* Commit management */
789extern void journal_commit_transaction(journal_t *);
790
791/* Checkpoint list management */
792int __journal_clean_checkpoint_list(journal_t *journal);
Jan Kara78ce89c2006-06-23 02:06:05 -0700793int __journal_remove_checkpoint(struct journal_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
795
796/* Buffer IO */
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700797extern int
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798journal_write_metadata_buffer(transaction_t *transaction,
799 struct journal_head *jh_in,
800 struct journal_head **jh_out,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200801 unsigned int blocknr);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802
803/* Transaction locking */
804extern void __wait_on_journal (journal_t *);
805
806/*
807 * Journal locking.
808 *
809 * We need to lock the journal during transaction state changes so that nobody
810 * ever tries to take a handle on the running transaction while we are in the
811 * middle of moving it to the commit phase. j_state_lock does this.
812 *
813 * Note that the locking is completely interrupt unsafe. We never touch
814 * journal structures from interrupts.
815 */
816
817static inline handle_t *journal_current_handle(void)
818{
819 return current->journal_info;
820}
821
822/* The journaling code user interface:
823 *
824 * Create and destroy handles
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700825 * Register buffer modifications against the current transaction.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826 */
827
828extern handle_t *journal_start(journal_t *, int nblocks);
829extern int journal_restart (handle_t *, int nblocks);
830extern int journal_extend (handle_t *, int nblocks);
831extern int journal_get_write_access(handle_t *, struct buffer_head *);
832extern int journal_get_create_access (handle_t *, struct buffer_head *);
833extern int journal_get_undo_access(handle_t *, struct buffer_head *);
834extern int journal_dirty_data (handle_t *, struct buffer_head *);
835extern int journal_dirty_metadata (handle_t *, struct buffer_head *);
836extern void journal_release_buffer (handle_t *, struct buffer_head *);
837extern int journal_forget (handle_t *, struct buffer_head *);
838extern void journal_sync_buffer (struct buffer_head *);
NeilBrown2ff28e22006-03-26 01:37:18 -0800839extern void journal_invalidatepage(journal_t *,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840 struct page *, unsigned long);
Al Viro27496a82005-10-21 03:20:48 -0400841extern int journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842extern int journal_stop(handle_t *);
843extern int journal_flush (journal_t *);
844extern void journal_lock_updates (journal_t *);
845extern void journal_unlock_updates (journal_t *);
846
847extern journal_t * journal_init_dev(struct block_device *bdev,
848 struct block_device *fs_dev,
849 int start, int len, int bsize);
850extern journal_t * journal_init_inode (struct inode *);
851extern int journal_update_format (journal_t *);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700852extern int journal_check_used_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700854extern int journal_check_available_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700856extern int journal_set_features
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 (journal_t *, unsigned long, unsigned long, unsigned long);
858extern int journal_create (journal_t *);
859extern int journal_load (journal_t *journal);
Hidehiro Kawai4afe9782008-10-22 14:15:00 -0700860extern int journal_destroy (journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861extern int journal_recover (journal_t *journal);
862extern int journal_wipe (journal_t *, int);
863extern int journal_skip_recovery (journal_t *);
864extern void journal_update_superblock (journal_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865extern void journal_abort (journal_t *, int);
866extern int journal_errno (journal_t *);
867extern void journal_ack_err (journal_t *);
868extern int journal_clear_err (journal_t *);
Jan Kara9c28cbc2009-08-03 19:21:00 +0200869extern int journal_bmap(journal_t *, unsigned int, unsigned int *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870extern int journal_force_commit(journal_t *);
871
872/*
873 * journal_head management
874 */
875struct journal_head *journal_add_journal_head(struct buffer_head *bh);
876struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877void journal_put_journal_head(struct journal_head *jh);
878
879/*
880 * handle management
881 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800882extern struct kmem_cache *jbd_handle_cache;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
Al Virodd0fc662005-10-07 07:46:04 +0100884static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885{
886 return kmem_cache_alloc(jbd_handle_cache, gfp_flags);
887}
888
889static inline void jbd_free_handle(handle_t *handle)
890{
891 kmem_cache_free(jbd_handle_cache, handle);
892}
893
894/* Primary revoke support */
895#define JOURNAL_REVOKE_DEFAULT_HASH 256
896extern int journal_init_revoke(journal_t *, int);
897extern void journal_destroy_revoke_caches(void);
898extern int journal_init_revoke_caches(void);
899
900extern void journal_destroy_revoke(journal_t *);
901extern int journal_revoke (handle_t *,
Jan Kara9c28cbc2009-08-03 19:21:00 +0200902 unsigned int, struct buffer_head *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903extern int journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o38d726d2009-04-14 10:10:47 -0400904extern void journal_write_revoke_records(journal_t *,
905 transaction_t *, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
907/* Recovery revoke support */
Jan Kara9c28cbc2009-08-03 19:21:00 +0200908extern int journal_set_revoke(journal_t *, unsigned int, tid_t);
909extern int journal_test_revoke(journal_t *, unsigned int, tid_t);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910extern void journal_clear_revoke(journal_t *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911extern void journal_switch_revoke_table(journal_t *journal);
Yongqiang Yang8c111b32011-11-19 17:34:29 +0800912extern void journal_clear_buffer_revoked_flags(journal_t *journal);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913
914/*
915 * The log thread user interface:
916 *
917 * Request space in the current transaction, and force transaction commit
918 * transitions on demand.
919 */
920
921int __log_space_left(journal_t *); /* Called with journal locked */
922int log_start_commit(journal_t *journal, tid_t tid);
923int __log_start_commit(journal_t *journal, tid_t tid);
924int journal_start_commit(journal_t *journal, tid_t *tid);
925int journal_force_commit_nested(journal_t *journal);
926int log_wait_commit(journal_t *journal, tid_t tid);
927int log_do_checkpoint(journal_t *journal);
Jan Kara03f4d802010-04-15 22:16:24 +0200928int journal_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929
930void __log_wait_for_space(journal_t *journal);
931extern void __journal_drop_transaction(journal_t *, transaction_t *);
932extern int cleanup_journal_tail(journal_t *);
933
934/* Debugging code only: */
935
936#define jbd_ENOSYS() \
937do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -0700938 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 current->state = TASK_UNINTERRUPTIBLE; \
940 schedule(); \
941} while (1)
942
943/*
944 * is_journal_abort
945 *
946 * Simple test wrapper function to test the JFS_ABORT state flag. This
947 * bit, when set, indicates that we have had a fatal error somewhere,
948 * either inside the journaling layer or indicated to us by the client
949 * (eg. ext3), and that we and should not commit any further
Mingming Caoae6ddcc2006-09-27 01:49:27 -0700950 * transactions.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951 */
952
953static inline int is_journal_aborted(journal_t *journal)
954{
955 return journal->j_flags & JFS_ABORT;
956}
957
958static inline int is_handle_aborted(handle_t *handle)
959{
960 if (handle->h_aborted)
961 return 1;
962 return is_journal_aborted(handle->h_transaction->t_journal);
963}
964
965static inline void journal_abort_handle(handle_t *handle)
966{
967 handle->h_aborted = 1;
968}
969
970#endif /* __KERNEL__ */
971
972/* Comparison functions for transaction IDs: perform comparisons using
973 * modulo arithmetic so that they work over sequence number wraps. */
974
975static inline int tid_gt(tid_t x, tid_t y)
976{
977 int difference = (x - y);
978 return (difference > 0);
979}
980
981static inline int tid_geq(tid_t x, tid_t y)
982{
983 int difference = (x - y);
984 return (difference >= 0);
985}
986
987extern int journal_blocks_per_page(struct inode *inode);
988
989/*
990 * Return the minimum number of blocks which must be free in the journal
991 * before a new transaction may be started. Must be called under j_state_lock.
992 */
993static inline int jbd_space_needed(journal_t *journal)
994{
995 int nblocks = journal->j_max_transaction_buffers;
996 if (journal->j_committing_transaction)
997 nblocks += journal->j_committing_transaction->
998 t_outstanding_credits;
999 return nblocks;
1000}
1001
1002/*
1003 * Definitions which augment the buffer_head layer
1004 */
1005
1006/* journaling buffer types */
1007#define BJ_None 0 /* Not journaled */
1008#define BJ_SyncData 1 /* Normal data: flush before commit */
1009#define BJ_Metadata 2 /* Normal journaled metadata */
1010#define BJ_Forget 3 /* Buffer superseded by this transaction */
1011#define BJ_IO 4 /* Buffer is for temporary IO use */
1012#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1013#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1014#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1015#define BJ_Locked 8 /* Locked for I/O during commit */
1016#define BJ_Types 9
Mingming Caoae6ddcc2006-09-27 01:49:27 -07001017
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018extern int jbd_blocks_per_page(struct inode *inode);
1019
1020#ifdef __KERNEL__
1021
1022#define buffer_trace_init(bh) do {} while (0)
1023#define print_buffer_fields(bh) do {} while (0)
1024#define print_buffer_trace(bh) do {} while (0)
1025#define BUFFER_TRACE(bh, info) do {} while (0)
1026#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1027#define JBUFFER_TRACE(jh, info) do {} while (0)
1028
1029#endif /* __KERNEL__ */
1030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031#endif /* _LINUX_JBD_H */