blob: 0e0fedd2039a3866a51fe093fb6792ef527c1804 [file] [log] [blame]
Dave Kleikamp470decc2006-10-11 01:20:57 -07001/*
Mingming Caof7f4bcc2006-10-11 01:20:59 -07002 * linux/include/linux/jbd2.h
Dave Kleikamp470decc2006-10-11 01:20:57 -07003 *
4 * 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
19/* Allow this file to be included directly into e2fsprogs */
20#ifndef __KERNEL__
21#include "jfs_compat.h"
Mingming Caof7f4bcc2006-10-11 01:20:59 -070022#define JBD2_DEBUG
Dave Kleikamp470decc2006-10-11 01:20:57 -070023#define jfs_debug jbd_debug
24#else
25
26#include <linux/types.h>
27#include <linux/buffer_head.h>
28#include <linux/journal-head.h>
29#include <linux/stddef.h>
30#include <linux/bit_spinlock.h>
31#include <linux/mutex.h>
32#include <linux/timer.h>
33
34#include <asm/semaphore.h>
35#endif
36
37#define journal_oom_retry 1
38
39/*
40 * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
41 * certain classes of error which can occur due to failed IOs. Under
42 * normal use we want ext3 to continue after such errors, because
43 * hardware _can_ fail, but for debugging purposes when running tests on
44 * known-good hardware we may want to trap these errors.
45 */
46#undef JBD_PARANOID_IOFAIL
47
48/*
49 * The default maximum commit age, in seconds.
50 */
51#define JBD_DEFAULT_MAX_COMMIT_AGE 5
52
53#ifdef CONFIG_JBD_DEBUG
54/*
55 * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
56 * consistency checks. By default we don't do this unless
57 * CONFIG_JBD_DEBUG is on.
58 */
59#define JBD_EXPENSIVE_CHECKING
Mingming Caof7f4bcc2006-10-11 01:20:59 -070060extern int jbd2_journal_enable_debug;
Dave Kleikamp470decc2006-10-11 01:20:57 -070061
62#define jbd_debug(n, f, a...) \
63 do { \
Mingming Caof7f4bcc2006-10-11 01:20:59 -070064 if ((n) <= jbd2_journal_enable_debug) { \
Dave Kleikamp470decc2006-10-11 01:20:57 -070065 printk (KERN_DEBUG "(%s, %d): %s: ", \
66 __FILE__, __LINE__, __FUNCTION__); \
67 printk (f, ## a); \
68 } \
69 } while (0)
70#else
71#define jbd_debug(f, a...) /**/
72#endif
73
Mingming Caof7f4bcc2006-10-11 01:20:59 -070074extern void * __jbd2_kmalloc (const char *where, size_t size, gfp_t flags, int retry);
75extern void * jbd2_slab_alloc(size_t size, gfp_t flags);
76extern void jbd2_slab_free(void *ptr, size_t size);
Dave Kleikamp470decc2006-10-11 01:20:57 -070077
78#define jbd_kmalloc(size, flags) \
Mingming Caof7f4bcc2006-10-11 01:20:59 -070079 __jbd2_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
Dave Kleikamp470decc2006-10-11 01:20:57 -070080#define jbd_rep_kmalloc(size, flags) \
Mingming Caof7f4bcc2006-10-11 01:20:59 -070081 __jbd2_kmalloc(__FUNCTION__, (size), (flags), 1)
Dave Kleikamp470decc2006-10-11 01:20:57 -070082
Mingming Caof7f4bcc2006-10-11 01:20:59 -070083#define JBD2_MIN_JOURNAL_BLOCKS 1024
Dave Kleikamp470decc2006-10-11 01:20:57 -070084
85#ifdef __KERNEL__
86
87/**
88 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
89 *
90 * All filesystem modifications made by the process go
91 * through this handle. Recursive operations (such as quota operations)
92 * are gathered into a single update.
93 *
94 * The buffer credits field is used to account for journaled buffers
95 * being modified by the running process. To ensure that there is
96 * enough log space for all outstanding operations, we need to limit the
97 * number of outstanding buffers possible at any time. When the
98 * operation completes, any buffer credits not used are credited back to
99 * the transaction, so that at all times we know how many buffers the
100 * outstanding updates on a transaction might possibly touch.
101 *
102 * This is an opaque datatype.
103 **/
104typedef struct handle_s handle_t; /* Atomic operation type */
105
106
107/**
108 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
109 *
110 * journal_t is linked to from the fs superblock structure.
111 *
112 * We use the journal_t to keep track of all outstanding transaction
113 * activity on the filesystem, and to manage the state of the log
114 * writing process.
115 *
116 * This is an opaque datatype.
117 **/
118typedef struct journal_s journal_t; /* Journal control structure */
119#endif
120
121/*
122 * Internal structures used by the logging mechanism:
123 */
124
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700125#define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700126
127/*
128 * On-disk structures
129 */
130
131/*
132 * Descriptor block types:
133 */
134
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700135#define JBD2_DESCRIPTOR_BLOCK 1
136#define JBD2_COMMIT_BLOCK 2
137#define JBD2_SUPERBLOCK_V1 3
138#define JBD2_SUPERBLOCK_V2 4
139#define JBD2_REVOKE_BLOCK 5
Dave Kleikamp470decc2006-10-11 01:20:57 -0700140
141/*
142 * Standard header for all descriptor blocks:
143 */
144typedef struct journal_header_s
145{
146 __be32 h_magic;
147 __be32 h_blocktype;
148 __be32 h_sequence;
149} journal_header_t;
150
151
152/*
Zach Brownb517bea2006-10-11 01:21:08 -0700153 * The block tag: used to describe a single buffer in the journal.
154 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
155 * raw struct shouldn't be used for pointer math or sizeof() - use
156 * journal_tag_bytes(journal) instead to compute this.
Dave Kleikamp470decc2006-10-11 01:20:57 -0700157 */
158typedef struct journal_block_tag_s
159{
160 __be32 t_blocknr; /* The on-disk block number */
161 __be32 t_flags; /* See below */
Zach Brownb517bea2006-10-11 01:21:08 -0700162 __be32 t_blocknr_high; /* most-significant high 32bits. */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700163} journal_block_tag_t;
164
Zach Brownb517bea2006-10-11 01:21:08 -0700165#define JBD_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
166#define JBD_TAG_SIZE64 (sizeof(journal_block_tag_t))
167
Dave Kleikamp470decc2006-10-11 01:20:57 -0700168/*
169 * The revoke descriptor: used on disk to describe a series of blocks to
170 * be revoked from the log
171 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700172typedef struct jbd2_journal_revoke_header_s
Dave Kleikamp470decc2006-10-11 01:20:57 -0700173{
174 journal_header_t r_header;
175 __be32 r_count; /* Count of bytes used in the block */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700176} jbd2_journal_revoke_header_t;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700177
178
179/* Definitions for the journal tag flags word: */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700180#define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
181#define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
182#define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
183#define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700184
185
186/*
187 * The journal superblock. All fields are in big-endian byte order.
188 */
189typedef struct journal_superblock_s
190{
191/* 0x0000 */
192 journal_header_t s_header;
193
194/* 0x000C */
195 /* Static information describing the journal */
196 __be32 s_blocksize; /* journal device blocksize */
197 __be32 s_maxlen; /* total blocks in journal file */
198 __be32 s_first; /* first block of log information */
199
200/* 0x0018 */
201 /* Dynamic information describing the current state of the log */
202 __be32 s_sequence; /* first commit ID expected in log */
203 __be32 s_start; /* blocknr of start of log */
204
205/* 0x0020 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700206 /* Error value, as set by jbd2_journal_abort(). */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700207 __be32 s_errno;
208
209/* 0x0024 */
210 /* Remaining fields are only valid in a version-2 superblock */
211 __be32 s_feature_compat; /* compatible feature set */
212 __be32 s_feature_incompat; /* incompatible feature set */
213 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
214/* 0x0030 */
215 __u8 s_uuid[16]; /* 128-bit uuid for journal */
216
217/* 0x0040 */
218 __be32 s_nr_users; /* Nr of filesystems sharing log */
219
220 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
221
222/* 0x0048 */
223 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
224 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
225
226/* 0x0050 */
227 __u32 s_padding[44];
228
229/* 0x0100 */
230 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
231/* 0x0400 */
232} journal_superblock_t;
233
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700234#define JBD2_HAS_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700235 ((j)->j_format_version >= 2 && \
236 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700237#define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700238 ((j)->j_format_version >= 2 && \
239 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700240#define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700241 ((j)->j_format_version >= 2 && \
242 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
243
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700244#define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
Zach Brownb517bea2006-10-11 01:21:08 -0700245#define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
Dave Kleikamp470decc2006-10-11 01:20:57 -0700246
247/* Features known to this kernel version: */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700248#define JBD2_KNOWN_COMPAT_FEATURES 0
249#define JBD2_KNOWN_ROCOMPAT_FEATURES 0
Zach Brownb517bea2006-10-11 01:21:08 -0700250#define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
251 JBD2_FEATURE_INCOMPAT_64BIT)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700252
253#ifdef __KERNEL__
254
255#include <linux/fs.h>
256#include <linux/sched.h>
257
258#define JBD_ASSERTIONS
259#ifdef JBD_ASSERTIONS
260#define J_ASSERT(assert) \
261do { \
262 if (!(assert)) { \
263 printk (KERN_EMERG \
264 "Assertion failure in %s() at %s:%d: \"%s\"\n", \
265 __FUNCTION__, __FILE__, __LINE__, # assert); \
266 BUG(); \
267 } \
268} while (0)
269
270#if defined(CONFIG_BUFFER_DEBUG)
271void buffer_assertion_failure(struct buffer_head *bh);
272#define J_ASSERT_BH(bh, expr) \
273 do { \
274 if (!(expr)) \
275 buffer_assertion_failure(bh); \
276 J_ASSERT(expr); \
277 } while (0)
278#define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
279#else
280#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
281#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
282#endif
283
284#else
285#define J_ASSERT(assert) do { } while (0)
286#endif /* JBD_ASSERTIONS */
287
288#if defined(JBD_PARANOID_IOFAIL)
289#define J_EXPECT(expr, why...) J_ASSERT(expr)
290#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
291#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
292#else
293#define __journal_expect(expr, why...) \
294 ({ \
295 int val = (expr); \
296 if (!val) { \
297 printk(KERN_ERR \
298 "EXT3-fs unexpected failure: %s;\n",# expr); \
299 printk(KERN_ERR why "\n"); \
300 } \
301 val; \
302 })
303#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
304#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
305#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
306#endif
307
308enum jbd_state_bits {
309 BH_JBD /* Has an attached ext3 journal_head */
310 = BH_PrivateStart,
311 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
312 BH_Freed, /* Has been freed (truncated) */
313 BH_Revoked, /* Has been revoked from the log */
314 BH_RevokeValid, /* Revoked flag is valid */
315 BH_JBDDirty, /* Is dirty but journaled */
316 BH_State, /* Pins most journal_head state */
317 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
318 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
319};
320
321BUFFER_FNS(JBD, jbd)
322BUFFER_FNS(JWrite, jwrite)
323BUFFER_FNS(JBDDirty, jbddirty)
324TAS_BUFFER_FNS(JBDDirty, jbddirty)
325BUFFER_FNS(Revoked, revoked)
326TAS_BUFFER_FNS(Revoked, revoked)
327BUFFER_FNS(RevokeValid, revokevalid)
328TAS_BUFFER_FNS(RevokeValid, revokevalid)
329BUFFER_FNS(Freed, freed)
330
331static inline struct buffer_head *jh2bh(struct journal_head *jh)
332{
333 return jh->b_bh;
334}
335
336static inline struct journal_head *bh2jh(struct buffer_head *bh)
337{
338 return bh->b_private;
339}
340
341static inline void jbd_lock_bh_state(struct buffer_head *bh)
342{
343 bit_spin_lock(BH_State, &bh->b_state);
344}
345
346static inline int jbd_trylock_bh_state(struct buffer_head *bh)
347{
348 return bit_spin_trylock(BH_State, &bh->b_state);
349}
350
351static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
352{
353 return bit_spin_is_locked(BH_State, &bh->b_state);
354}
355
356static inline void jbd_unlock_bh_state(struct buffer_head *bh)
357{
358 bit_spin_unlock(BH_State, &bh->b_state);
359}
360
361static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
362{
363 bit_spin_lock(BH_JournalHead, &bh->b_state);
364}
365
366static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
367{
368 bit_spin_unlock(BH_JournalHead, &bh->b_state);
369}
370
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700371struct jbd2_revoke_table_s;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700372
373/**
374 * struct handle_s - The handle_s type is the concrete type associated with
375 * handle_t.
376 * @h_transaction: Which compound transaction is this update a part of?
377 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
378 * @h_ref: Reference count on this handle
379 * @h_err: Field for caller's use to track errors through large fs operations
380 * @h_sync: flag for sync-on-close
381 * @h_jdata: flag to force data journaling
382 * @h_aborted: flag indicating fatal error on handle
383 **/
384
385/* Docbook can't yet cope with the bit fields, but will leave the documentation
386 * in so it can be fixed later.
387 */
388
389struct handle_s
390{
391 /* Which compound transaction is this update a part of? */
392 transaction_t *h_transaction;
393
394 /* Number of remaining buffers we are allowed to dirty: */
395 int h_buffer_credits;
396
397 /* Reference count on this handle */
398 int h_ref;
399
400 /* Field for caller's use to track errors through large fs */
401 /* operations */
402 int h_err;
403
404 /* Flags [no locking] */
405 unsigned int h_sync: 1; /* sync-on-close */
406 unsigned int h_jdata: 1; /* force data journaling */
407 unsigned int h_aborted: 1; /* fatal error on handle */
408};
409
410
411/* The transaction_t type is the guts of the journaling mechanism. It
412 * tracks a compound transaction through its various states:
413 *
414 * RUNNING: accepting new updates
415 * LOCKED: Updates still running but we don't accept new ones
416 * RUNDOWN: Updates are tidying up but have finished requesting
417 * new buffers to modify (state not used for now)
418 * FLUSH: All updates complete, but we are still writing to disk
419 * COMMIT: All data on disk, writing commit record
420 * FINISHED: We still have to keep the transaction for checkpointing.
421 *
422 * The transaction keeps track of all of the buffers modified by a
423 * running transaction, and all of the buffers committed but not yet
424 * flushed to home for finished transactions.
425 */
426
427/*
428 * Lock ranking:
429 *
430 * j_list_lock
431 * ->jbd_lock_bh_journal_head() (This is "innermost")
432 *
433 * j_state_lock
434 * ->jbd_lock_bh_state()
435 *
436 * jbd_lock_bh_state()
437 * ->j_list_lock
438 *
439 * j_state_lock
440 * ->t_handle_lock
441 *
442 * j_state_lock
443 * ->j_list_lock (journal_unmap_buffer)
444 *
445 */
446
447struct transaction_s
448{
449 /* Pointer to the journal for this transaction. [no locking] */
450 journal_t *t_journal;
451
452 /* Sequence number for this transaction [no locking] */
453 tid_t t_tid;
454
455 /*
456 * Transaction's current state
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700457 * [no locking - only kjournald2 alters this]
Dave Kleikamp470decc2006-10-11 01:20:57 -0700458 * FIXME: needs barriers
459 * KLUDGE: [use j_state_lock]
460 */
461 enum {
462 T_RUNNING,
463 T_LOCKED,
464 T_RUNDOWN,
465 T_FLUSH,
466 T_COMMIT,
467 T_FINISHED
468 } t_state;
469
470 /*
471 * Where in the log does this transaction's commit start? [no locking]
472 */
473 unsigned long t_log_start;
474
475 /* Number of buffers on the t_buffers list [j_list_lock] */
476 int t_nr_buffers;
477
478 /*
479 * Doubly-linked circular list of all buffers reserved but not yet
480 * modified by this transaction [j_list_lock]
481 */
482 struct journal_head *t_reserved_list;
483
484 /*
485 * Doubly-linked circular list of all buffers under writeout during
486 * commit [j_list_lock]
487 */
488 struct journal_head *t_locked_list;
489
490 /*
491 * Doubly-linked circular list of all metadata buffers owned by this
492 * transaction [j_list_lock]
493 */
494 struct journal_head *t_buffers;
495
496 /*
497 * Doubly-linked circular list of all data buffers still to be
498 * flushed before this transaction can be committed [j_list_lock]
499 */
500 struct journal_head *t_sync_datalist;
501
502 /*
503 * Doubly-linked circular list of all forget buffers (superseded
504 * buffers which we can un-checkpoint once this transaction commits)
505 * [j_list_lock]
506 */
507 struct journal_head *t_forget;
508
509 /*
510 * Doubly-linked circular list of all buffers still to be flushed before
511 * this transaction can be checkpointed. [j_list_lock]
512 */
513 struct journal_head *t_checkpoint_list;
514
515 /*
516 * Doubly-linked circular list of all buffers submitted for IO while
517 * checkpointing. [j_list_lock]
518 */
519 struct journal_head *t_checkpoint_io_list;
520
521 /*
522 * Doubly-linked circular list of temporary buffers currently undergoing
523 * IO in the log [j_list_lock]
524 */
525 struct journal_head *t_iobuf_list;
526
527 /*
528 * Doubly-linked circular list of metadata buffers being shadowed by log
529 * IO. The IO buffers on the iobuf list and the shadow buffers on this
530 * list match each other one for one at all times. [j_list_lock]
531 */
532 struct journal_head *t_shadow_list;
533
534 /*
535 * Doubly-linked circular list of control buffers being written to the
536 * log. [j_list_lock]
537 */
538 struct journal_head *t_log_list;
539
540 /*
541 * Protects info related to handles
542 */
543 spinlock_t t_handle_lock;
544
545 /*
546 * Number of outstanding updates running on this transaction
547 * [t_handle_lock]
548 */
549 int t_updates;
550
551 /*
552 * Number of buffers reserved for use by all handles in this transaction
553 * handle but not yet modified. [t_handle_lock]
554 */
555 int t_outstanding_credits;
556
557 /*
558 * Forward and backward links for the circular list of all transactions
559 * awaiting checkpoint. [j_list_lock]
560 */
561 transaction_t *t_cpnext, *t_cpprev;
562
563 /*
564 * When will the transaction expire (become due for commit), in jiffies?
565 * [no locking]
566 */
567 unsigned long t_expires;
568
569 /*
570 * How many handles used this transaction? [t_handle_lock]
571 */
572 int t_handle_count;
573
574};
575
576/**
577 * struct journal_s - The journal_s type is the concrete type associated with
578 * journal_t.
579 * @j_flags: General journaling state flags
580 * @j_errno: Is there an outstanding uncleared error on the journal (from a
581 * prior abort)?
582 * @j_sb_buffer: First part of superblock buffer
583 * @j_superblock: Second part of superblock buffer
584 * @j_format_version: Version of the superblock format
585 * @j_state_lock: Protect the various scalars in the journal
586 * @j_barrier_count: Number of processes waiting to create a barrier lock
587 * @j_barrier: The barrier lock itself
588 * @j_running_transaction: The current running transaction..
589 * @j_committing_transaction: the transaction we are pushing to disk
590 * @j_checkpoint_transactions: a linked circular list of all transactions
591 * waiting for checkpointing
592 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
593 * to start committing, or for a barrier lock to be released
594 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
595 * @j_wait_done_commit: Wait queue for waiting for commit to complete
596 * @j_wait_checkpoint: Wait queue to trigger checkpointing
597 * @j_wait_commit: Wait queue to trigger commit
598 * @j_wait_updates: Wait queue to wait for updates to complete
599 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
600 * @j_head: Journal head - identifies the first unused block in the journal
601 * @j_tail: Journal tail - identifies the oldest still-used block in the
602 * journal.
603 * @j_free: Journal free - how many free blocks are there in the journal?
604 * @j_first: The block number of the first usable block
605 * @j_last: The block number one beyond the last usable block
606 * @j_dev: Device where we store the journal
607 * @j_blocksize: blocksize for the location where we store the journal.
608 * @j_blk_offset: starting block offset for into the device where we store the
609 * journal
610 * @j_fs_dev: Device which holds the client fs. For internal journal this will
611 * be equal to j_dev
612 * @j_maxlen: Total maximum capacity of the journal region on disk.
613 * @j_list_lock: Protects the buffer lists and internal buffer state.
614 * @j_inode: Optional inode where we store the journal. If present, all journal
615 * block numbers are mapped into this inode via bmap().
616 * @j_tail_sequence: Sequence number of the oldest transaction in the log
617 * @j_transaction_sequence: Sequence number of the next transaction to grant
618 * @j_commit_sequence: Sequence number of the most recently committed
619 * transaction
620 * @j_commit_request: Sequence number of the most recent transaction wanting
621 * commit
622 * @j_uuid: Uuid of client object.
623 * @j_task: Pointer to the current commit thread for this journal
624 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
625 * single compound commit transaction
626 * @j_commit_interval: What is the maximum transaction lifetime before we begin
627 * a commit?
628 * @j_commit_timer: The timer used to wakeup the commit thread
629 * @j_revoke_lock: Protect the revoke table
630 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
631 * current transaction.
632 * @j_revoke_table: alternate revoke tables for j_revoke
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700633 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700634 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
635 * number that will fit in j_blocksize
636 * @j_last_sync_writer: most recent pid which did a synchronous write
637 * @j_private: An opaque pointer to fs-private information.
638 */
639
640struct journal_s
641{
642 /* General journaling state flags [j_state_lock] */
643 unsigned long j_flags;
644
645 /*
646 * Is there an outstanding uncleared error on the journal (from a prior
647 * abort)? [j_state_lock]
648 */
649 int j_errno;
650
651 /* The superblock buffer */
652 struct buffer_head *j_sb_buffer;
653 journal_superblock_t *j_superblock;
654
655 /* Version of the superblock format */
656 int j_format_version;
657
658 /*
659 * Protect the various scalars in the journal
660 */
661 spinlock_t j_state_lock;
662
663 /*
664 * Number of processes waiting to create a barrier lock [j_state_lock]
665 */
666 int j_barrier_count;
667
668 /* The barrier lock itself */
669 struct mutex j_barrier;
670
671 /*
672 * Transactions: The current running transaction...
673 * [j_state_lock] [caller holding open handle]
674 */
675 transaction_t *j_running_transaction;
676
677 /*
678 * the transaction we are pushing to disk
679 * [j_state_lock] [caller holding open handle]
680 */
681 transaction_t *j_committing_transaction;
682
683 /*
684 * ... and a linked circular list of all transactions waiting for
685 * checkpointing. [j_list_lock]
686 */
687 transaction_t *j_checkpoint_transactions;
688
689 /*
690 * Wait queue for waiting for a locked transaction to start committing,
691 * or for a barrier lock to be released
692 */
693 wait_queue_head_t j_wait_transaction_locked;
694
695 /* Wait queue for waiting for checkpointing to complete */
696 wait_queue_head_t j_wait_logspace;
697
698 /* Wait queue for waiting for commit to complete */
699 wait_queue_head_t j_wait_done_commit;
700
701 /* Wait queue to trigger checkpointing */
702 wait_queue_head_t j_wait_checkpoint;
703
704 /* Wait queue to trigger commit */
705 wait_queue_head_t j_wait_commit;
706
707 /* Wait queue to wait for updates to complete */
708 wait_queue_head_t j_wait_updates;
709
710 /* Semaphore for locking against concurrent checkpoints */
711 struct mutex j_checkpoint_mutex;
712
713 /*
714 * Journal head: identifies the first unused block in the journal.
715 * [j_state_lock]
716 */
717 unsigned long j_head;
718
719 /*
720 * Journal tail: identifies the oldest still-used block in the journal.
721 * [j_state_lock]
722 */
723 unsigned long j_tail;
724
725 /*
726 * Journal free: how many free blocks are there in the journal?
727 * [j_state_lock]
728 */
729 unsigned long j_free;
730
731 /*
732 * Journal start and end: the block numbers of the first usable block
733 * and one beyond the last usable block in the journal. [j_state_lock]
734 */
735 unsigned long j_first;
736 unsigned long j_last;
737
738 /*
739 * Device, blocksize and starting block offset for the location where we
740 * store the journal.
741 */
742 struct block_device *j_dev;
743 int j_blocksize;
Mingming Cao18eba7a2006-10-11 01:21:13 -0700744 unsigned long long j_blk_offset;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700745
746 /*
747 * Device which holds the client fs. For internal journal this will be
748 * equal to j_dev.
749 */
750 struct block_device *j_fs_dev;
751
752 /* Total maximum capacity of the journal region on disk. */
753 unsigned int j_maxlen;
754
755 /*
756 * Protects the buffer lists and internal buffer state.
757 */
758 spinlock_t j_list_lock;
759
760 /* Optional inode where we store the journal. If present, all */
761 /* journal block numbers are mapped into this inode via */
762 /* bmap(). */
763 struct inode *j_inode;
764
765 /*
766 * Sequence number of the oldest transaction in the log [j_state_lock]
767 */
768 tid_t j_tail_sequence;
769
770 /*
771 * Sequence number of the next transaction to grant [j_state_lock]
772 */
773 tid_t j_transaction_sequence;
774
775 /*
776 * Sequence number of the most recently committed transaction
777 * [j_state_lock].
778 */
779 tid_t j_commit_sequence;
780
781 /*
782 * Sequence number of the most recent transaction wanting commit
783 * [j_state_lock]
784 */
785 tid_t j_commit_request;
786
787 /*
788 * Journal uuid: identifies the object (filesystem, LVM volume etc)
789 * backed by this journal. This will eventually be replaced by an array
790 * of uuids, allowing us to index multiple devices within a single
791 * journal and to perform atomic updates across them.
792 */
793 __u8 j_uuid[16];
794
795 /* Pointer to the current commit thread for this journal */
796 struct task_struct *j_task;
797
798 /*
799 * Maximum number of metadata buffers to allow in a single compound
800 * commit transaction
801 */
802 int j_max_transaction_buffers;
803
804 /*
805 * What is the maximum transaction lifetime before we begin a commit?
806 */
807 unsigned long j_commit_interval;
808
809 /* The timer used to wakeup the commit thread: */
810 struct timer_list j_commit_timer;
811
812 /*
813 * The revoke table: maintains the list of revoked blocks in the
814 * current transaction. [j_revoke_lock]
815 */
816 spinlock_t j_revoke_lock;
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700817 struct jbd2_revoke_table_s *j_revoke;
818 struct jbd2_revoke_table_s *j_revoke_table[2];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700819
820 /*
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700821 * array of bhs for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700822 */
823 struct buffer_head **j_wbuf;
824 int j_wbufsize;
825
826 pid_t j_last_sync_writer;
827
828 /*
829 * An opaque pointer to fs-private information. ext3 puts its
830 * superblock pointer here
831 */
832 void *j_private;
833};
834
835/*
836 * Journal flag definitions
837 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700838#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
839#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
840#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
841#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
842#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
843#define JBD2_BARRIER 0x020 /* Use IDE barriers */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700844
845/*
846 * Function declarations for the journaling transaction and buffer
847 * management
848 */
849
850/* Filing buffers */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700851extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
852extern void __jbd2_journal_unfile_buffer(struct journal_head *);
853extern void __jbd2_journal_refile_buffer(struct journal_head *);
854extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
855extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700856extern void __journal_free_buffer(struct journal_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700857extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700858extern void __journal_clean_data_list(transaction_t *transaction);
859
860/* Log buffer allocation */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700861extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -0700862int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700863
864/* Commit management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700865extern void jbd2_journal_commit_transaction(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700866
867/* Checkpoint list management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700868int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
869int __jbd2_journal_remove_checkpoint(struct journal_head *);
870void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700871
872/* Buffer IO */
873extern int
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700874jbd2_journal_write_metadata_buffer(transaction_t *transaction,
Dave Kleikamp470decc2006-10-11 01:20:57 -0700875 struct journal_head *jh_in,
876 struct journal_head **jh_out,
Mingming Cao18eba7a2006-10-11 01:21:13 -0700877 unsigned long long blocknr);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700878
879/* Transaction locking */
880extern void __wait_on_journal (journal_t *);
881
882/*
883 * Journal locking.
884 *
885 * We need to lock the journal during transaction state changes so that nobody
886 * ever tries to take a handle on the running transaction while we are in the
887 * middle of moving it to the commit phase. j_state_lock does this.
888 *
889 * Note that the locking is completely interrupt unsafe. We never touch
890 * journal structures from interrupts.
891 */
892
893static inline handle_t *journal_current_handle(void)
894{
895 return current->journal_info;
896}
897
898/* The journaling code user interface:
899 *
900 * Create and destroy handles
901 * Register buffer modifications against the current transaction.
902 */
903
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700904extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
905extern int jbd2_journal_restart (handle_t *, int nblocks);
906extern int jbd2_journal_extend (handle_t *, int nblocks);
907extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
908extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
909extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
910extern int jbd2_journal_dirty_data (handle_t *, struct buffer_head *);
911extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
912extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
913extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700914extern void journal_sync_buffer (struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700915extern void jbd2_journal_invalidatepage(journal_t *,
Dave Kleikamp470decc2006-10-11 01:20:57 -0700916 struct page *, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700917extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
918extern int jbd2_journal_stop(handle_t *);
919extern int jbd2_journal_flush (journal_t *);
920extern void jbd2_journal_lock_updates (journal_t *);
921extern void jbd2_journal_unlock_updates (journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700922
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700923extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
Dave Kleikamp470decc2006-10-11 01:20:57 -0700924 struct block_device *fs_dev,
Mingming Cao18eba7a2006-10-11 01:21:13 -0700925 unsigned long long start, int len, int bsize);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700926extern journal_t * jbd2_journal_init_inode (struct inode *);
927extern int jbd2_journal_update_format (journal_t *);
928extern int jbd2_journal_check_used_features
Dave Kleikamp470decc2006-10-11 01:20:57 -0700929 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700930extern int jbd2_journal_check_available_features
Dave Kleikamp470decc2006-10-11 01:20:57 -0700931 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700932extern int jbd2_journal_set_features
Dave Kleikamp470decc2006-10-11 01:20:57 -0700933 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700934extern int jbd2_journal_create (journal_t *);
935extern int jbd2_journal_load (journal_t *journal);
936extern void jbd2_journal_destroy (journal_t *);
937extern int jbd2_journal_recover (journal_t *journal);
938extern int jbd2_journal_wipe (journal_t *, int);
939extern int jbd2_journal_skip_recovery (journal_t *);
940extern void jbd2_journal_update_superblock (journal_t *, int);
941extern void __jbd2_journal_abort_hard (journal_t *);
942extern void jbd2_journal_abort (journal_t *, int);
943extern int jbd2_journal_errno (journal_t *);
944extern void jbd2_journal_ack_err (journal_t *);
945extern int jbd2_journal_clear_err (journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -0700946extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700947extern int jbd2_journal_force_commit(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700948
949/*
950 * journal_head management
951 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700952struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
953struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
954void jbd2_journal_remove_journal_head(struct buffer_head *bh);
955void jbd2_journal_put_journal_head(struct journal_head *jh);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700956
957/*
958 * handle management
959 */
Christoph Lametere18b8902006-12-06 20:33:20 -0800960extern struct kmem_cache *jbd2_handle_cache;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700961
962static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
963{
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700964 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700965}
966
967static inline void jbd_free_handle(handle_t *handle)
968{
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700969 kmem_cache_free(jbd2_handle_cache, handle);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700970}
971
972/* Primary revoke support */
973#define JOURNAL_REVOKE_DEFAULT_HASH 256
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700974extern int jbd2_journal_init_revoke(journal_t *, int);
975extern void jbd2_journal_destroy_revoke_caches(void);
976extern int jbd2_journal_init_revoke_caches(void);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700977
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700978extern void jbd2_journal_destroy_revoke(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -0700979extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700980extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
981extern void jbd2_journal_write_revoke_records(journal_t *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700982
983/* Recovery revoke support */
Mingming Cao18eba7a2006-10-11 01:21:13 -0700984extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
985extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700986extern void jbd2_journal_clear_revoke(journal_t *);
987extern void jbd2_journal_switch_revoke_table(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700988
989/*
990 * The log thread user interface:
991 *
992 * Request space in the current transaction, and force transaction commit
993 * transitions on demand.
994 */
995
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700996int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
997int jbd2_log_start_commit(journal_t *journal, tid_t tid);
998int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
999int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1000int jbd2_journal_force_commit_nested(journal_t *journal);
1001int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1002int jbd2_log_do_checkpoint(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001003
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001004void __jbd2_log_wait_for_space(journal_t *journal);
1005extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1006extern int jbd2_cleanup_journal_tail(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001007
1008/* Debugging code only: */
1009
1010#define jbd_ENOSYS() \
1011do { \
1012 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1013 current->state = TASK_UNINTERRUPTIBLE; \
1014 schedule(); \
1015} while (1)
1016
1017/*
1018 * is_journal_abort
1019 *
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001020 * Simple test wrapper function to test the JBD2_ABORT state flag. This
Dave Kleikamp470decc2006-10-11 01:20:57 -07001021 * bit, when set, indicates that we have had a fatal error somewhere,
1022 * either inside the journaling layer or indicated to us by the client
1023 * (eg. ext3), and that we and should not commit any further
1024 * transactions.
1025 */
1026
1027static inline int is_journal_aborted(journal_t *journal)
1028{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001029 return journal->j_flags & JBD2_ABORT;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001030}
1031
1032static inline int is_handle_aborted(handle_t *handle)
1033{
1034 if (handle->h_aborted)
1035 return 1;
1036 return is_journal_aborted(handle->h_transaction->t_journal);
1037}
1038
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001039static inline void jbd2_journal_abort_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001040{
1041 handle->h_aborted = 1;
1042}
1043
1044#endif /* __KERNEL__ */
1045
1046/* Comparison functions for transaction IDs: perform comparisons using
1047 * modulo arithmetic so that they work over sequence number wraps. */
1048
1049static inline int tid_gt(tid_t x, tid_t y)
1050{
1051 int difference = (x - y);
1052 return (difference > 0);
1053}
1054
1055static inline int tid_geq(tid_t x, tid_t y)
1056{
1057 int difference = (x - y);
1058 return (difference >= 0);
1059}
1060
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001061extern int jbd2_journal_blocks_per_page(struct inode *inode);
Zach Brownb517bea2006-10-11 01:21:08 -07001062extern size_t journal_tag_bytes(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001063
1064/*
1065 * Return the minimum number of blocks which must be free in the journal
1066 * before a new transaction may be started. Must be called under j_state_lock.
1067 */
1068static inline int jbd_space_needed(journal_t *journal)
1069{
1070 int nblocks = journal->j_max_transaction_buffers;
1071 if (journal->j_committing_transaction)
1072 nblocks += journal->j_committing_transaction->
1073 t_outstanding_credits;
1074 return nblocks;
1075}
1076
1077/*
1078 * Definitions which augment the buffer_head layer
1079 */
1080
1081/* journaling buffer types */
1082#define BJ_None 0 /* Not journaled */
1083#define BJ_SyncData 1 /* Normal data: flush before commit */
1084#define BJ_Metadata 2 /* Normal journaled metadata */
1085#define BJ_Forget 3 /* Buffer superseded by this transaction */
1086#define BJ_IO 4 /* Buffer is for temporary IO use */
1087#define BJ_Shadow 5 /* Buffer contents being shadowed to the log */
1088#define BJ_LogCtl 6 /* Buffer contains log descriptors */
1089#define BJ_Reserved 7 /* Buffer is reserved for access by journal */
1090#define BJ_Locked 8 /* Locked for I/O during commit */
1091#define BJ_Types 9
1092
1093extern int jbd_blocks_per_page(struct inode *inode);
1094
1095#ifdef __KERNEL__
1096
1097#define buffer_trace_init(bh) do {} while (0)
1098#define print_buffer_fields(bh) do {} while (0)
1099#define print_buffer_trace(bh) do {} while (0)
1100#define BUFFER_TRACE(bh, info) do {} while (0)
1101#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1102#define JBUFFER_TRACE(jh, info) do {} while (0)
1103
1104#endif /* __KERNEL__ */
1105
1106#endif /* _LINUX_JBD_H */