blob: a4d2e9f7088ada70d8b1357f418c32cd09a5bf1a [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
Mingming Caocd02ff02007-10-16 18:38:25 -040016#ifndef _LINUX_JBD2_H
17#define _LINUX_JBD2_H
Dave Kleikamp470decc2006-10-11 01:20:57 -070018
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>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090033#include <linux/slab.h>
Dave Kleikamp470decc2006-10-11 01:20:57 -070034#endif
35
36#define journal_oom_retry 1
37
38/*
Mingming Caocd02ff02007-10-16 18:38:25 -040039 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
Dave Kleikamp470decc2006-10-11 01:20:57 -070040 * certain classes of error which can occur due to failed IOs. Under
Mingming Caocd02ff02007-10-16 18:38:25 -040041 * normal use we want ext4 to continue after such errors, because
Dave Kleikamp470decc2006-10-11 01:20:57 -070042 * hardware _can_ fail, but for debugging purposes when running tests on
43 * known-good hardware we may want to trap these errors.
44 */
Mingming Caocd02ff02007-10-16 18:38:25 -040045#undef JBD2_PARANOID_IOFAIL
Dave Kleikamp470decc2006-10-11 01:20:57 -070046
47/*
48 * The default maximum commit age, in seconds.
49 */
Mingming Caocd02ff02007-10-16 18:38:25 -040050#define JBD2_DEFAULT_MAX_COMMIT_AGE 5
Dave Kleikamp470decc2006-10-11 01:20:57 -070051
Jose R. Santose23291b2007-07-18 08:57:06 -040052#ifdef CONFIG_JBD2_DEBUG
Dave Kleikamp470decc2006-10-11 01:20:57 -070053/*
Mingming Caocd02ff02007-10-16 18:38:25 -040054 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
Dave Kleikamp470decc2006-10-11 01:20:57 -070055 * consistency checks. By default we don't do this unless
Jose R. Santose23291b2007-07-18 08:57:06 -040056 * CONFIG_JBD2_DEBUG is on.
Dave Kleikamp470decc2006-10-11 01:20:57 -070057 */
Mingming Caocd02ff02007-10-16 18:38:25 -040058#define JBD2_EXPENSIVE_CHECKING
Jose R. Santos0f49d5d2007-07-18 08:50:18 -040059extern u8 jbd2_journal_enable_debug;
Dave Kleikamp470decc2006-10-11 01:20:57 -070060
61#define jbd_debug(n, f, a...) \
62 do { \
Mingming Caof7f4bcc2006-10-11 01:20:59 -070063 if ((n) <= jbd2_journal_enable_debug) { \
Dave Kleikamp470decc2006-10-11 01:20:57 -070064 printk (KERN_DEBUG "(%s, %d): %s: ", \
Harvey Harrisond5c003b2008-10-15 22:01:24 -070065 __FILE__, __LINE__, __func__); \
Dave Kleikamp470decc2006-10-11 01:20:57 -070066 printk (f, ## a); \
67 } \
68 } while (0)
69#else
70#define jbd_debug(f, a...) /**/
71#endif
72
Theodore Ts'od2eecb02009-12-07 10:36:20 -050073extern void *jbd2_alloc(size_t size, gfp_t flags);
74extern void jbd2_free(void *ptr, size_t size);
Mingming Caoaf1e76d2007-10-16 18:38:25 -040075
Mingming Caof7f4bcc2006-10-11 01:20:59 -070076#define JBD2_MIN_JOURNAL_BLOCKS 1024
Dave Kleikamp470decc2006-10-11 01:20:57 -070077
78#ifdef __KERNEL__
79
80/**
81 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
82 *
83 * All filesystem modifications made by the process go
84 * through this handle. Recursive operations (such as quota operations)
85 * are gathered into a single update.
86 *
87 * The buffer credits field is used to account for journaled buffers
88 * being modified by the running process. To ensure that there is
89 * enough log space for all outstanding operations, we need to limit the
90 * number of outstanding buffers possible at any time. When the
91 * operation completes, any buffer credits not used are credited back to
92 * the transaction, so that at all times we know how many buffers the
93 * outstanding updates on a transaction might possibly touch.
94 *
95 * This is an opaque datatype.
96 **/
97typedef struct handle_s handle_t; /* Atomic operation type */
98
99
100/**
101 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
102 *
103 * journal_t is linked to from the fs superblock structure.
104 *
105 * We use the journal_t to keep track of all outstanding transaction
106 * activity on the filesystem, and to manage the state of the log
107 * writing process.
108 *
109 * This is an opaque datatype.
110 **/
111typedef struct journal_s journal_t; /* Journal control structure */
112#endif
113
114/*
115 * Internal structures used by the logging mechanism:
116 */
117
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700118#define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700119
120/*
121 * On-disk structures
122 */
123
124/*
125 * Descriptor block types:
126 */
127
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700128#define JBD2_DESCRIPTOR_BLOCK 1
129#define JBD2_COMMIT_BLOCK 2
130#define JBD2_SUPERBLOCK_V1 3
131#define JBD2_SUPERBLOCK_V2 4
132#define JBD2_REVOKE_BLOCK 5
Dave Kleikamp470decc2006-10-11 01:20:57 -0700133
134/*
135 * Standard header for all descriptor blocks:
136 */
137typedef struct journal_header_s
138{
139 __be32 h_magic;
140 __be32 h_blocktype;
141 __be32 h_sequence;
142} journal_header_t;
143
Girish Shilamkar818d2762008-01-28 23:58:27 -0500144/*
145 * Checksum types.
146 */
147#define JBD2_CRC32_CHKSUM 1
148#define JBD2_MD5_CHKSUM 2
149#define JBD2_SHA1_CHKSUM 3
150
151#define JBD2_CRC32_CHKSUM_SIZE 4
152
153#define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
154/*
155 * Commit block header for storing transactional checksums:
156 */
157struct commit_header {
158 __be32 h_magic;
159 __be32 h_blocktype;
160 __be32 h_sequence;
161 unsigned char h_chksum_type;
162 unsigned char h_chksum_size;
163 unsigned char h_padding[2];
164 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
Theodore Ts'o736603a2008-07-11 19:27:31 -0400165 __be64 h_commit_sec;
166 __be32 h_commit_nsec;
Girish Shilamkar818d2762008-01-28 23:58:27 -0500167};
Dave Kleikamp470decc2006-10-11 01:20:57 -0700168
169/*
Zach Brownb517bea2006-10-11 01:21:08 -0700170 * The block tag: used to describe a single buffer in the journal.
171 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
172 * raw struct shouldn't be used for pointer math or sizeof() - use
173 * journal_tag_bytes(journal) instead to compute this.
Dave Kleikamp470decc2006-10-11 01:20:57 -0700174 */
175typedef struct journal_block_tag_s
176{
177 __be32 t_blocknr; /* The on-disk block number */
178 __be32 t_flags; /* See below */
Zach Brownb517bea2006-10-11 01:21:08 -0700179 __be32 t_blocknr_high; /* most-significant high 32bits. */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700180} journal_block_tag_t;
181
Mingming Caocd02ff02007-10-16 18:38:25 -0400182#define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
183#define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
Zach Brownb517bea2006-10-11 01:21:08 -0700184
Dave Kleikamp470decc2006-10-11 01:20:57 -0700185/*
186 * The revoke descriptor: used on disk to describe a series of blocks to
187 * be revoked from the log
188 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700189typedef struct jbd2_journal_revoke_header_s
Dave Kleikamp470decc2006-10-11 01:20:57 -0700190{
191 journal_header_t r_header;
192 __be32 r_count; /* Count of bytes used in the block */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700193} jbd2_journal_revoke_header_t;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700194
195
196/* Definitions for the journal tag flags word: */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700197#define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
198#define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
199#define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
200#define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700201
202
203/*
204 * The journal superblock. All fields are in big-endian byte order.
205 */
206typedef struct journal_superblock_s
207{
208/* 0x0000 */
209 journal_header_t s_header;
210
211/* 0x000C */
212 /* Static information describing the journal */
213 __be32 s_blocksize; /* journal device blocksize */
214 __be32 s_maxlen; /* total blocks in journal file */
215 __be32 s_first; /* first block of log information */
216
217/* 0x0018 */
218 /* Dynamic information describing the current state of the log */
219 __be32 s_sequence; /* first commit ID expected in log */
220 __be32 s_start; /* blocknr of start of log */
221
222/* 0x0020 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700223 /* Error value, as set by jbd2_journal_abort(). */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700224 __be32 s_errno;
225
226/* 0x0024 */
227 /* Remaining fields are only valid in a version-2 superblock */
228 __be32 s_feature_compat; /* compatible feature set */
229 __be32 s_feature_incompat; /* incompatible feature set */
230 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
231/* 0x0030 */
232 __u8 s_uuid[16]; /* 128-bit uuid for journal */
233
234/* 0x0040 */
235 __be32 s_nr_users; /* Nr of filesystems sharing log */
236
237 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
238
239/* 0x0048 */
240 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
241 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
242
243/* 0x0050 */
244 __u32 s_padding[44];
245
246/* 0x0100 */
247 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
248/* 0x0400 */
249} journal_superblock_t;
250
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700251#define JBD2_HAS_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700252 ((j)->j_format_version >= 2 && \
253 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700254#define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700255 ((j)->j_format_version >= 2 && \
256 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700257#define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700258 ((j)->j_format_version >= 2 && \
259 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
260
Girish Shilamkar818d2762008-01-28 23:58:27 -0500261#define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
262
263#define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
264#define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
265#define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
Dave Kleikamp470decc2006-10-11 01:20:57 -0700266
267/* Features known to this kernel version: */
Girish Shilamkar818d2762008-01-28 23:58:27 -0500268#define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700269#define JBD2_KNOWN_ROCOMPAT_FEATURES 0
Zach Brownb517bea2006-10-11 01:21:08 -0700270#define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
Girish Shilamkar818d2762008-01-28 23:58:27 -0500271 JBD2_FEATURE_INCOMPAT_64BIT | \
272 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700273
274#ifdef __KERNEL__
275
276#include <linux/fs.h>
277#include <linux/sched.h>
278
Chris Snook36df53f2008-01-28 23:58:27 -0500279#define J_ASSERT(assert) BUG_ON(!(assert))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700280
Dave Kleikamp470decc2006-10-11 01:20:57 -0700281#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
282#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700283
Mingming Caocd02ff02007-10-16 18:38:25 -0400284#if defined(JBD2_PARANOID_IOFAIL)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700285#define J_EXPECT(expr, why...) J_ASSERT(expr)
286#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
287#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
288#else
289#define __journal_expect(expr, why...) \
290 ({ \
291 int val = (expr); \
292 if (!val) { \
293 printk(KERN_ERR \
Theodore Ts'o08ec8c32009-01-16 11:57:00 -0500294 "JBD2 unexpected failure: %s: %s;\n", \
295 __func__, #expr); \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700296 printk(KERN_ERR why "\n"); \
297 } \
298 val; \
299 })
300#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
301#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
302#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
303#endif
304
305enum jbd_state_bits {
306 BH_JBD /* Has an attached ext3 journal_head */
307 = BH_PrivateStart,
308 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
309 BH_Freed, /* Has been freed (truncated) */
310 BH_Revoked, /* Has been revoked from the log */
311 BH_RevokeValid, /* Revoked flag is valid */
312 BH_JBDDirty, /* Is dirty but journaled */
313 BH_State, /* Pins most journal_head state */
314 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
315 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
Mark Fashehe97fcd92008-11-18 17:15:24 -0800316 BH_JBDPrivateStart, /* First bit available for private use by FS */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700317};
318
319BUFFER_FNS(JBD, jbd)
320BUFFER_FNS(JWrite, jwrite)
321BUFFER_FNS(JBDDirty, jbddirty)
322TAS_BUFFER_FNS(JBDDirty, jbddirty)
323BUFFER_FNS(Revoked, revoked)
324TAS_BUFFER_FNS(Revoked, revoked)
325BUFFER_FNS(RevokeValid, revokevalid)
326TAS_BUFFER_FNS(RevokeValid, revokevalid)
327BUFFER_FNS(Freed, freed)
328
329static inline struct buffer_head *jh2bh(struct journal_head *jh)
330{
331 return jh->b_bh;
332}
333
334static inline struct journal_head *bh2jh(struct buffer_head *bh)
335{
336 return bh->b_private;
337}
338
339static inline void jbd_lock_bh_state(struct buffer_head *bh)
340{
341 bit_spin_lock(BH_State, &bh->b_state);
342}
343
344static inline int jbd_trylock_bh_state(struct buffer_head *bh)
345{
346 return bit_spin_trylock(BH_State, &bh->b_state);
347}
348
349static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
350{
351 return bit_spin_is_locked(BH_State, &bh->b_state);
352}
353
354static inline void jbd_unlock_bh_state(struct buffer_head *bh)
355{
356 bit_spin_unlock(BH_State, &bh->b_state);
357}
358
359static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
360{
361 bit_spin_lock(BH_JournalHead, &bh->b_state);
362}
363
364static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
365{
366 bit_spin_unlock(BH_JournalHead, &bh->b_state);
367}
368
Jan Karac851ed52008-07-11 19:27:31 -0400369/* Flags in jbd_inode->i_flags */
370#define __JI_COMMIT_RUNNING 0
371/* Commit of the inode data in progress. We use this flag to protect us from
372 * concurrent deletion of inode. We cannot use reference to inode for this
373 * since we cannot afford doing last iput() on behalf of kjournald
374 */
375#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
376
377/**
378 * struct jbd_inode is the structure linking inodes in ordered mode
379 * present in a transaction so that we can sync them during commit.
380 */
381struct jbd2_inode {
382 /* Which transaction does this inode belong to? Either the running
383 * transaction or the committing one. [j_list_lock] */
384 transaction_t *i_transaction;
385
386 /* Pointer to the running transaction modifying inode's data in case
387 * there is already a committing transaction touching it. [j_list_lock] */
388 transaction_t *i_next_transaction;
389
390 /* List of inodes in the i_transaction [j_list_lock] */
391 struct list_head i_list;
392
393 /* VFS inode this inode belongs to [constant during the lifetime
394 * of the structure] */
395 struct inode *i_vfs_inode;
396
397 /* Flags of inode [j_list_lock] */
398 unsigned int i_flags;
399};
400
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700401struct jbd2_revoke_table_s;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700402
403/**
404 * struct handle_s - The handle_s type is the concrete type associated with
405 * handle_t.
406 * @h_transaction: Which compound transaction is this update a part of?
407 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
408 * @h_ref: Reference count on this handle
409 * @h_err: Field for caller's use to track errors through large fs operations
410 * @h_sync: flag for sync-on-close
411 * @h_jdata: flag to force data journaling
412 * @h_aborted: flag indicating fatal error on handle
413 **/
414
415/* Docbook can't yet cope with the bit fields, but will leave the documentation
416 * in so it can be fixed later.
417 */
418
419struct handle_s
420{
421 /* Which compound transaction is this update a part of? */
422 transaction_t *h_transaction;
423
424 /* Number of remaining buffers we are allowed to dirty: */
425 int h_buffer_credits;
426
427 /* Reference count on this handle */
428 int h_ref;
429
430 /* Field for caller's use to track errors through large fs */
431 /* operations */
432 int h_err;
433
434 /* Flags [no locking] */
435 unsigned int h_sync: 1; /* sync-on-close */
436 unsigned int h_jdata: 1; /* force data journaling */
437 unsigned int h_aborted: 1; /* fatal error on handle */
Mingming Cao7b751062008-01-28 23:58:27 -0500438
439#ifdef CONFIG_DEBUG_LOCK_ALLOC
440 struct lockdep_map h_lockdep_map;
441#endif
Dave Kleikamp470decc2006-10-11 01:20:57 -0700442};
443
444
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500445/*
446 * Some stats for checkpoint phase
447 */
448struct transaction_chp_stats_s {
449 unsigned long cs_chp_time;
Theodore Ts'obf699322009-09-30 00:32:06 -0400450 __u32 cs_forced_to_close;
451 __u32 cs_written;
452 __u32 cs_dropped;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500453};
454
Dave Kleikamp470decc2006-10-11 01:20:57 -0700455/* The transaction_t type is the guts of the journaling mechanism. It
456 * tracks a compound transaction through its various states:
457 *
458 * RUNNING: accepting new updates
459 * LOCKED: Updates still running but we don't accept new ones
460 * RUNDOWN: Updates are tidying up but have finished requesting
461 * new buffers to modify (state not used for now)
462 * FLUSH: All updates complete, but we are still writing to disk
463 * COMMIT: All data on disk, writing commit record
464 * FINISHED: We still have to keep the transaction for checkpointing.
465 *
466 * The transaction keeps track of all of the buffers modified by a
467 * running transaction, and all of the buffers committed but not yet
468 * flushed to home for finished transactions.
469 */
470
471/*
472 * Lock ranking:
473 *
474 * j_list_lock
475 * ->jbd_lock_bh_journal_head() (This is "innermost")
476 *
477 * j_state_lock
478 * ->jbd_lock_bh_state()
479 *
480 * jbd_lock_bh_state()
481 * ->j_list_lock
482 *
483 * j_state_lock
484 * ->t_handle_lock
485 *
486 * j_state_lock
487 * ->j_list_lock (journal_unmap_buffer)
488 *
489 */
490
491struct transaction_s
492{
493 /* Pointer to the journal for this transaction. [no locking] */
494 journal_t *t_journal;
495
496 /* Sequence number for this transaction [no locking] */
497 tid_t t_tid;
498
499 /*
500 * Transaction's current state
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700501 * [no locking - only kjournald2 alters this]
Jan Karaf5a7a6b2008-01-28 23:58:27 -0500502 * [j_list_lock] guards transition of a transaction into T_FINISHED
503 * state and subsequent call of __jbd2_journal_drop_transaction()
Dave Kleikamp470decc2006-10-11 01:20:57 -0700504 * FIXME: needs barriers
505 * KLUDGE: [use j_state_lock]
506 */
507 enum {
508 T_RUNNING,
509 T_LOCKED,
510 T_RUNDOWN,
511 T_FLUSH,
512 T_COMMIT,
513 T_FINISHED
514 } t_state;
515
516 /*
517 * Where in the log does this transaction's commit start? [no locking]
518 */
519 unsigned long t_log_start;
520
521 /* Number of buffers on the t_buffers list [j_list_lock] */
522 int t_nr_buffers;
523
524 /*
525 * Doubly-linked circular list of all buffers reserved but not yet
526 * modified by this transaction [j_list_lock]
527 */
528 struct journal_head *t_reserved_list;
529
530 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700531 * Doubly-linked circular list of all metadata buffers owned by this
532 * transaction [j_list_lock]
533 */
534 struct journal_head *t_buffers;
535
536 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700537 * Doubly-linked circular list of all forget buffers (superseded
538 * buffers which we can un-checkpoint once this transaction commits)
539 * [j_list_lock]
540 */
541 struct journal_head *t_forget;
542
543 /*
544 * Doubly-linked circular list of all buffers still to be flushed before
545 * this transaction can be checkpointed. [j_list_lock]
546 */
547 struct journal_head *t_checkpoint_list;
548
549 /*
550 * Doubly-linked circular list of all buffers submitted for IO while
551 * checkpointing. [j_list_lock]
552 */
553 struct journal_head *t_checkpoint_io_list;
554
555 /*
556 * Doubly-linked circular list of temporary buffers currently undergoing
557 * IO in the log [j_list_lock]
558 */
559 struct journal_head *t_iobuf_list;
560
561 /*
562 * Doubly-linked circular list of metadata buffers being shadowed by log
563 * IO. The IO buffers on the iobuf list and the shadow buffers on this
564 * list match each other one for one at all times. [j_list_lock]
565 */
566 struct journal_head *t_shadow_list;
567
568 /*
569 * Doubly-linked circular list of control buffers being written to the
570 * log. [j_list_lock]
571 */
572 struct journal_head *t_log_list;
573
574 /*
Jan Karac851ed52008-07-11 19:27:31 -0400575 * List of inodes whose data we've modified in data=ordered mode.
576 * [j_list_lock]
577 */
578 struct list_head t_inode_list;
579
580 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700581 * Protects info related to handles
582 */
583 spinlock_t t_handle_lock;
584
585 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500586 * Longest time some handle had to wait for running transaction
587 */
588 unsigned long t_max_wait;
589
590 /*
591 * When transaction started
592 */
593 unsigned long t_start;
594
595 /*
596 * Checkpointing stats [j_checkpoint_sem]
597 */
598 struct transaction_chp_stats_s t_chp_stats;
599
600 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700601 * Number of outstanding updates running on this transaction
602 * [t_handle_lock]
603 */
604 int t_updates;
605
606 /*
607 * Number of buffers reserved for use by all handles in this transaction
608 * handle but not yet modified. [t_handle_lock]
609 */
610 int t_outstanding_credits;
611
612 /*
613 * Forward and backward links for the circular list of all transactions
614 * awaiting checkpoint. [j_list_lock]
615 */
616 transaction_t *t_cpnext, *t_cpprev;
617
618 /*
619 * When will the transaction expire (become due for commit), in jiffies?
620 * [no locking]
621 */
622 unsigned long t_expires;
623
624 /*
Josef Bacike07f7182008-11-26 01:14:26 -0500625 * When this transaction started, in nanoseconds [no locking]
626 */
627 ktime_t t_start_time;
628
629 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700630 * How many handles used this transaction? [t_handle_lock]
631 */
632 int t_handle_count;
633
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400634 /*
Theodore Ts'o70585482009-03-25 23:35:46 -0400635 * This transaction is being forced and some process is
636 * waiting for it to finish.
637 */
H Hartley Sweeten0ccff1a2009-08-17 22:38:04 -0400638 unsigned int t_synchronous_commit:1;
Theodore Ts'occ3e1be2009-12-23 06:52:08 -0500639 unsigned int t_flushed_data_blocks:1;
Theodore Ts'o70585482009-03-25 23:35:46 -0400640
641 /*
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400642 * For use by the filesystem to store fs-specific data
643 * structures associated with the transaction
644 */
645 struct list_head t_private_list;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700646};
647
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500648struct transaction_run_stats_s {
649 unsigned long rs_wait;
650 unsigned long rs_running;
651 unsigned long rs_locked;
652 unsigned long rs_flushing;
653 unsigned long rs_logging;
654
Theodore Ts'obf699322009-09-30 00:32:06 -0400655 __u32 rs_handle_count;
656 __u32 rs_blocks;
657 __u32 rs_blocks_logged;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500658};
659
660struct transaction_stats_s {
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500661 unsigned long ts_tid;
Theodore Ts'obf699322009-09-30 00:32:06 -0400662 struct transaction_run_stats_s run;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500663};
664
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500665static inline unsigned long
666jbd2_time_diff(unsigned long start, unsigned long end)
667{
668 if (end >= start)
669 return end - start;
670
671 return end + (MAX_JIFFY_OFFSET - start);
672}
673
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500674#define JBD2_NR_BATCH 64
675
Dave Kleikamp470decc2006-10-11 01:20:57 -0700676/**
677 * struct journal_s - The journal_s type is the concrete type associated with
678 * journal_t.
679 * @j_flags: General journaling state flags
680 * @j_errno: Is there an outstanding uncleared error on the journal (from a
681 * prior abort)?
682 * @j_sb_buffer: First part of superblock buffer
683 * @j_superblock: Second part of superblock buffer
684 * @j_format_version: Version of the superblock format
685 * @j_state_lock: Protect the various scalars in the journal
686 * @j_barrier_count: Number of processes waiting to create a barrier lock
687 * @j_barrier: The barrier lock itself
688 * @j_running_transaction: The current running transaction..
689 * @j_committing_transaction: the transaction we are pushing to disk
690 * @j_checkpoint_transactions: a linked circular list of all transactions
691 * waiting for checkpointing
692 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
693 * to start committing, or for a barrier lock to be released
694 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
695 * @j_wait_done_commit: Wait queue for waiting for commit to complete
696 * @j_wait_checkpoint: Wait queue to trigger checkpointing
697 * @j_wait_commit: Wait queue to trigger commit
698 * @j_wait_updates: Wait queue to wait for updates to complete
699 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
700 * @j_head: Journal head - identifies the first unused block in the journal
701 * @j_tail: Journal tail - identifies the oldest still-used block in the
702 * journal.
703 * @j_free: Journal free - how many free blocks are there in the journal?
704 * @j_first: The block number of the first usable block
705 * @j_last: The block number one beyond the last usable block
706 * @j_dev: Device where we store the journal
707 * @j_blocksize: blocksize for the location where we store the journal.
708 * @j_blk_offset: starting block offset for into the device where we store the
709 * journal
710 * @j_fs_dev: Device which holds the client fs. For internal journal this will
711 * be equal to j_dev
712 * @j_maxlen: Total maximum capacity of the journal region on disk.
713 * @j_list_lock: Protects the buffer lists and internal buffer state.
714 * @j_inode: Optional inode where we store the journal. If present, all journal
715 * block numbers are mapped into this inode via bmap().
716 * @j_tail_sequence: Sequence number of the oldest transaction in the log
717 * @j_transaction_sequence: Sequence number of the next transaction to grant
718 * @j_commit_sequence: Sequence number of the most recently committed
719 * transaction
720 * @j_commit_request: Sequence number of the most recent transaction wanting
721 * commit
722 * @j_uuid: Uuid of client object.
723 * @j_task: Pointer to the current commit thread for this journal
724 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
725 * single compound commit transaction
726 * @j_commit_interval: What is the maximum transaction lifetime before we begin
727 * a commit?
728 * @j_commit_timer: The timer used to wakeup the commit thread
729 * @j_revoke_lock: Protect the revoke table
730 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
731 * current transaction.
732 * @j_revoke_table: alternate revoke tables for j_revoke
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700733 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700734 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
735 * number that will fit in j_blocksize
736 * @j_last_sync_writer: most recent pid which did a synchronous write
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500737 * @j_history: Buffer storing the transactions statistics history
738 * @j_history_max: Maximum number of transactions in the statistics history
739 * @j_history_cur: Current number of transactions in the statistics history
740 * @j_history_lock: Protect the transactions statistics history
741 * @j_proc_entry: procfs entry for the jbd statistics directory
742 * @j_stats: Overall statistics
Dave Kleikamp470decc2006-10-11 01:20:57 -0700743 * @j_private: An opaque pointer to fs-private information.
744 */
745
746struct journal_s
747{
748 /* General journaling state flags [j_state_lock] */
749 unsigned long j_flags;
750
751 /*
752 * Is there an outstanding uncleared error on the journal (from a prior
753 * abort)? [j_state_lock]
754 */
755 int j_errno;
756
757 /* The superblock buffer */
758 struct buffer_head *j_sb_buffer;
759 journal_superblock_t *j_superblock;
760
761 /* Version of the superblock format */
762 int j_format_version;
763
764 /*
765 * Protect the various scalars in the journal
766 */
767 spinlock_t j_state_lock;
768
769 /*
770 * Number of processes waiting to create a barrier lock [j_state_lock]
771 */
772 int j_barrier_count;
773
774 /* The barrier lock itself */
775 struct mutex j_barrier;
776
777 /*
778 * Transactions: The current running transaction...
779 * [j_state_lock] [caller holding open handle]
780 */
781 transaction_t *j_running_transaction;
782
783 /*
784 * the transaction we are pushing to disk
785 * [j_state_lock] [caller holding open handle]
786 */
787 transaction_t *j_committing_transaction;
788
789 /*
790 * ... and a linked circular list of all transactions waiting for
791 * checkpointing. [j_list_lock]
792 */
793 transaction_t *j_checkpoint_transactions;
794
795 /*
796 * Wait queue for waiting for a locked transaction to start committing,
797 * or for a barrier lock to be released
798 */
799 wait_queue_head_t j_wait_transaction_locked;
800
801 /* Wait queue for waiting for checkpointing to complete */
802 wait_queue_head_t j_wait_logspace;
803
804 /* Wait queue for waiting for commit to complete */
805 wait_queue_head_t j_wait_done_commit;
806
807 /* Wait queue to trigger checkpointing */
808 wait_queue_head_t j_wait_checkpoint;
809
810 /* Wait queue to trigger commit */
811 wait_queue_head_t j_wait_commit;
812
813 /* Wait queue to wait for updates to complete */
814 wait_queue_head_t j_wait_updates;
815
816 /* Semaphore for locking against concurrent checkpoints */
817 struct mutex j_checkpoint_mutex;
818
819 /*
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500820 * List of buffer heads used by the checkpoint routine. This
821 * was moved from jbd2_log_do_checkpoint() to reduce stack
822 * usage. Access to this array is controlled by the
823 * j_checkpoint_mutex. [j_checkpoint_mutex]
824 */
825 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
826
827 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700828 * Journal head: identifies the first unused block in the journal.
829 * [j_state_lock]
830 */
831 unsigned long j_head;
832
833 /*
834 * Journal tail: identifies the oldest still-used block in the journal.
835 * [j_state_lock]
836 */
837 unsigned long j_tail;
838
839 /*
840 * Journal free: how many free blocks are there in the journal?
841 * [j_state_lock]
842 */
843 unsigned long j_free;
844
845 /*
846 * Journal start and end: the block numbers of the first usable block
847 * and one beyond the last usable block in the journal. [j_state_lock]
848 */
849 unsigned long j_first;
850 unsigned long j_last;
851
852 /*
853 * Device, blocksize and starting block offset for the location where we
854 * store the journal.
855 */
856 struct block_device *j_dev;
857 int j_blocksize;
Theodore Ts'o05496762008-09-16 14:36:17 -0400858 unsigned long long j_blk_offset;
859 char j_devname[BDEVNAME_SIZE+24];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700860
861 /*
862 * Device which holds the client fs. For internal journal this will be
863 * equal to j_dev.
864 */
865 struct block_device *j_fs_dev;
866
867 /* Total maximum capacity of the journal region on disk. */
868 unsigned int j_maxlen;
869
870 /*
871 * Protects the buffer lists and internal buffer state.
872 */
873 spinlock_t j_list_lock;
874
875 /* Optional inode where we store the journal. If present, all */
876 /* journal block numbers are mapped into this inode via */
877 /* bmap(). */
878 struct inode *j_inode;
879
880 /*
881 * Sequence number of the oldest transaction in the log [j_state_lock]
882 */
883 tid_t j_tail_sequence;
884
885 /*
886 * Sequence number of the next transaction to grant [j_state_lock]
887 */
888 tid_t j_transaction_sequence;
889
890 /*
891 * Sequence number of the most recently committed transaction
892 * [j_state_lock].
893 */
894 tid_t j_commit_sequence;
895
896 /*
897 * Sequence number of the most recent transaction wanting commit
898 * [j_state_lock]
899 */
900 tid_t j_commit_request;
901
902 /*
903 * Journal uuid: identifies the object (filesystem, LVM volume etc)
904 * backed by this journal. This will eventually be replaced by an array
905 * of uuids, allowing us to index multiple devices within a single
906 * journal and to perform atomic updates across them.
907 */
908 __u8 j_uuid[16];
909
910 /* Pointer to the current commit thread for this journal */
911 struct task_struct *j_task;
912
913 /*
914 * Maximum number of metadata buffers to allow in a single compound
915 * commit transaction
916 */
917 int j_max_transaction_buffers;
918
919 /*
920 * What is the maximum transaction lifetime before we begin a commit?
921 */
922 unsigned long j_commit_interval;
923
924 /* The timer used to wakeup the commit thread: */
925 struct timer_list j_commit_timer;
926
927 /*
928 * The revoke table: maintains the list of revoked blocks in the
929 * current transaction. [j_revoke_lock]
930 */
931 spinlock_t j_revoke_lock;
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700932 struct jbd2_revoke_table_s *j_revoke;
933 struct jbd2_revoke_table_s *j_revoke_table[2];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700934
935 /*
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700936 * array of bhs for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700937 */
938 struct buffer_head **j_wbuf;
939 int j_wbufsize;
940
Josef Bacike07f7182008-11-26 01:14:26 -0500941 /*
942 * this is the pid of hte last person to run a synchronous operation
943 * through the journal
944 */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700945 pid_t j_last_sync_writer;
946
Josef Bacike07f7182008-11-26 01:14:26 -0500947 /*
948 * the average amount of time in nanoseconds it takes to commit a
949 * transaction to disk. [j_state_lock]
950 */
951 u64 j_average_commit_time;
952
Theodore Ts'o30773842009-01-03 20:27:38 -0500953 /*
954 * minimum and maximum times that we should wait for
955 * additional filesystem operations to get batched into a
956 * synchronous handle in microseconds
957 */
958 u32 j_min_batch_time;
959 u32 j_max_batch_time;
960
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400961 /* This function is called when a transaction is closed */
962 void (*j_commit_callback)(journal_t *,
963 transaction_t *);
964
Dave Kleikamp470decc2006-10-11 01:20:57 -0700965 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500966 * Journal statistics
967 */
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500968 spinlock_t j_history_lock;
969 struct proc_dir_entry *j_proc_entry;
970 struct transaction_stats_s j_stats;
971
Theodore Ts'o624080e2008-06-06 17:50:40 -0400972 /* Failed journal commit ID */
973 unsigned int j_failed_commit;
974
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500975 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700976 * An opaque pointer to fs-private information. ext3 puts its
977 * superblock pointer here
978 */
979 void *j_private;
980};
981
982/*
983 * Journal flag definitions
984 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700985#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
986#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
987#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
988#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
989#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
990#define JBD2_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai5bf56832008-10-10 22:12:43 -0400991#define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
992 * data write error in ordered
993 * mode */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700994
995/*
996 * Function declarations for the journaling transaction and buffer
997 * management
998 */
999
1000/* Filing buffers */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001001extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1002extern void __jbd2_journal_unfile_buffer(struct journal_head *);
1003extern void __jbd2_journal_refile_buffer(struct journal_head *);
1004extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1005extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001006extern void __journal_free_buffer(struct journal_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001007extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001008extern void __journal_clean_data_list(transaction_t *transaction);
1009
1010/* Log buffer allocation */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001011extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001012int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001013
1014/* Commit management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001015extern void jbd2_journal_commit_transaction(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001016
1017/* Checkpoint list management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001018int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1019int __jbd2_journal_remove_checkpoint(struct journal_head *);
1020void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001021
Joel Beckere06c8222008-09-11 15:35:47 -07001022
1023/*
1024 * Triggers
1025 */
1026
1027struct jbd2_buffer_trigger_type {
1028 /*
1029 * Fired just before a buffer is written to the journal.
1030 * mapped_data is a mapped buffer that is the frozen data for
1031 * commit.
1032 */
1033 void (*t_commit)(struct jbd2_buffer_trigger_type *type,
1034 struct buffer_head *bh, void *mapped_data,
1035 size_t size);
1036
1037 /*
1038 * Fired during journal abort for dirty buffers that will not be
1039 * committed.
1040 */
1041 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1042 struct buffer_head *bh);
1043};
1044
1045extern void jbd2_buffer_commit_trigger(struct journal_head *jh,
1046 void *mapped_data,
1047 struct jbd2_buffer_trigger_type *triggers);
1048extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1049 struct jbd2_buffer_trigger_type *triggers);
1050
Dave Kleikamp470decc2006-10-11 01:20:57 -07001051/* Buffer IO */
1052extern int
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001053jbd2_journal_write_metadata_buffer(transaction_t *transaction,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001054 struct journal_head *jh_in,
1055 struct journal_head **jh_out,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001056 unsigned long long blocknr);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001057
1058/* Transaction locking */
1059extern void __wait_on_journal (journal_t *);
1060
1061/*
1062 * Journal locking.
1063 *
1064 * We need to lock the journal during transaction state changes so that nobody
1065 * ever tries to take a handle on the running transaction while we are in the
1066 * middle of moving it to the commit phase. j_state_lock does this.
1067 *
1068 * Note that the locking is completely interrupt unsafe. We never touch
1069 * journal structures from interrupts.
1070 */
1071
1072static inline handle_t *journal_current_handle(void)
1073{
1074 return current->journal_info;
1075}
1076
1077/* The journaling code user interface:
1078 *
1079 * Create and destroy handles
1080 * Register buffer modifications against the current transaction.
1081 */
1082
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001083extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1084extern int jbd2_journal_restart (handle_t *, int nblocks);
1085extern int jbd2_journal_extend (handle_t *, int nblocks);
1086extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1087extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1088extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
Joel Beckere06c8222008-09-11 15:35:47 -07001089void jbd2_journal_set_triggers(struct buffer_head *,
1090 struct jbd2_buffer_trigger_type *type);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001091extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1092extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1093extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001094extern void journal_sync_buffer (struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001095extern void jbd2_journal_invalidatepage(journal_t *,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001096 struct page *, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001097extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1098extern int jbd2_journal_stop(handle_t *);
1099extern int jbd2_journal_flush (journal_t *);
1100extern void jbd2_journal_lock_updates (journal_t *);
1101extern void jbd2_journal_unlock_updates (journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001102
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001103extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001104 struct block_device *fs_dev,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001105 unsigned long long start, int len, int bsize);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001106extern journal_t * jbd2_journal_init_inode (struct inode *);
1107extern int jbd2_journal_update_format (journal_t *);
1108extern int jbd2_journal_check_used_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001109 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001110extern int jbd2_journal_check_available_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001111 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001112extern int jbd2_journal_set_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001113 (journal_t *, unsigned long, unsigned long, unsigned long);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001114extern void jbd2_journal_clear_features
1115 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001116extern int jbd2_journal_load (journal_t *journal);
Hidehiro Kawai44519fa2008-10-10 20:29:13 -04001117extern int jbd2_journal_destroy (journal_t *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001118extern int jbd2_journal_recover (journal_t *journal);
1119extern int jbd2_journal_wipe (journal_t *, int);
1120extern int jbd2_journal_skip_recovery (journal_t *);
1121extern void jbd2_journal_update_superblock (journal_t *, int);
1122extern void __jbd2_journal_abort_hard (journal_t *);
1123extern void jbd2_journal_abort (journal_t *, int);
1124extern int jbd2_journal_errno (journal_t *);
1125extern void jbd2_journal_ack_err (journal_t *);
1126extern int jbd2_journal_clear_err (journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001127extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001128extern int jbd2_journal_force_commit(journal_t *);
Jan Karac851ed52008-07-11 19:27:31 -04001129extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
Jan Kara7f5aa212009-02-10 11:15:34 -05001130extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1131 struct jbd2_inode *inode, loff_t new_size);
Jan Karac851ed52008-07-11 19:27:31 -04001132extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1133extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001134
1135/*
1136 * journal_head management
1137 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001138struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1139struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1140void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1141void jbd2_journal_put_journal_head(struct journal_head *jh);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001142
1143/*
1144 * handle management
1145 */
Christoph Lametere18b8902006-12-06 20:33:20 -08001146extern struct kmem_cache *jbd2_handle_cache;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001147
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001148static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001149{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001150 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001151}
1152
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001153static inline void jbd2_free_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001154{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001155 kmem_cache_free(jbd2_handle_cache, handle);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001156}
1157
1158/* Primary revoke support */
1159#define JOURNAL_REVOKE_DEFAULT_HASH 256
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001160extern int jbd2_journal_init_revoke(journal_t *, int);
1161extern void jbd2_journal_destroy_revoke_caches(void);
1162extern int jbd2_journal_init_revoke_caches(void);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001163
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001164extern void jbd2_journal_destroy_revoke(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001165extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001166extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o67c457a2009-04-14 07:50:56 -04001167extern void jbd2_journal_write_revoke_records(journal_t *,
1168 transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001169
1170/* Recovery revoke support */
Mingming Cao18eba7a2006-10-11 01:21:13 -07001171extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1172extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001173extern void jbd2_journal_clear_revoke(journal_t *);
1174extern void jbd2_journal_switch_revoke_table(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001175
1176/*
1177 * The log thread user interface:
1178 *
1179 * Request space in the current transaction, and force transaction commit
1180 * transitions on demand.
1181 */
1182
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001183int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1184int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1185int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1186int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1187int jbd2_journal_force_commit_nested(journal_t *journal);
1188int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1189int jbd2_log_do_checkpoint(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001190
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001191void __jbd2_log_wait_for_space(journal_t *journal);
Aneesh Kumar K.Vfb684072008-11-06 17:50:21 -05001192extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1193extern int jbd2_cleanup_journal_tail(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001194
1195/* Debugging code only: */
1196
1197#define jbd_ENOSYS() \
1198do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -07001199 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Dave Kleikamp470decc2006-10-11 01:20:57 -07001200 current->state = TASK_UNINTERRUPTIBLE; \
1201 schedule(); \
1202} while (1)
1203
1204/*
1205 * is_journal_abort
1206 *
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001207 * Simple test wrapper function to test the JBD2_ABORT state flag. This
Dave Kleikamp470decc2006-10-11 01:20:57 -07001208 * bit, when set, indicates that we have had a fatal error somewhere,
1209 * either inside the journaling layer or indicated to us by the client
1210 * (eg. ext3), and that we and should not commit any further
1211 * transactions.
1212 */
1213
1214static inline int is_journal_aborted(journal_t *journal)
1215{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001216 return journal->j_flags & JBD2_ABORT;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001217}
1218
1219static inline int is_handle_aborted(handle_t *handle)
1220{
1221 if (handle->h_aborted)
1222 return 1;
1223 return is_journal_aborted(handle->h_transaction->t_journal);
1224}
1225
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001226static inline void jbd2_journal_abort_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001227{
1228 handle->h_aborted = 1;
1229}
1230
1231#endif /* __KERNEL__ */
1232
1233/* Comparison functions for transaction IDs: perform comparisons using
1234 * modulo arithmetic so that they work over sequence number wraps. */
1235
1236static inline int tid_gt(tid_t x, tid_t y)
1237{
1238 int difference = (x - y);
1239 return (difference > 0);
1240}
1241
1242static inline int tid_geq(tid_t x, tid_t y)
1243{
1244 int difference = (x - y);
1245 return (difference >= 0);
1246}
1247
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001248extern int jbd2_journal_blocks_per_page(struct inode *inode);
Zach Brownb517bea2006-10-11 01:21:08 -07001249extern size_t journal_tag_bytes(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001250
1251/*
1252 * Return the minimum number of blocks which must be free in the journal
1253 * before a new transaction may be started. Must be called under j_state_lock.
1254 */
1255static inline int jbd_space_needed(journal_t *journal)
1256{
1257 int nblocks = journal->j_max_transaction_buffers;
1258 if (journal->j_committing_transaction)
1259 nblocks += journal->j_committing_transaction->
1260 t_outstanding_credits;
1261 return nblocks;
1262}
1263
1264/*
1265 * Definitions which augment the buffer_head layer
1266 */
1267
1268/* journaling buffer types */
1269#define BJ_None 0 /* Not journaled */
Jan Kara87c89c22008-07-11 19:27:31 -04001270#define BJ_Metadata 1 /* Normal journaled metadata */
1271#define BJ_Forget 2 /* Buffer superseded by this transaction */
1272#define BJ_IO 3 /* Buffer is for temporary IO use */
1273#define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1274#define BJ_LogCtl 5 /* Buffer contains log descriptors */
1275#define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1276#define BJ_Types 7
Dave Kleikamp470decc2006-10-11 01:20:57 -07001277
1278extern int jbd_blocks_per_page(struct inode *inode);
1279
1280#ifdef __KERNEL__
1281
1282#define buffer_trace_init(bh) do {} while (0)
1283#define print_buffer_fields(bh) do {} while (0)
1284#define print_buffer_trace(bh) do {} while (0)
1285#define BUFFER_TRACE(bh, info) do {} while (0)
1286#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1287#define JBUFFER_TRACE(jh, info) do {} while (0)
1288
Theodore Ts'o879c5e62009-06-17 11:47:48 -04001289/*
1290 * jbd2_dev_to_name is a utility function used by the jbd2 and ext4
1291 * tracing infrastructure to map a dev_t to a device name.
1292 */
1293extern const char *jbd2_dev_to_name(dev_t device);
1294
Dave Kleikamp470decc2006-10-11 01:20:57 -07001295#endif /* __KERNEL__ */
1296
Mingming Caocd02ff02007-10-16 18:38:25 -04001297#endif /* _LINUX_JBD2_H */