blob: b45109c61fba90451a0396f6c845e7c55964ccb3 [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>
Dave Kleikamp470decc2006-10-11 01:20:57 -070033#endif
34
35#define journal_oom_retry 1
36
37/*
Mingming Caocd02ff02007-10-16 18:38:25 -040038 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
Dave Kleikamp470decc2006-10-11 01:20:57 -070039 * certain classes of error which can occur due to failed IOs. Under
Mingming Caocd02ff02007-10-16 18:38:25 -040040 * normal use we want ext4 to continue after such errors, because
Dave Kleikamp470decc2006-10-11 01:20:57 -070041 * hardware _can_ fail, but for debugging purposes when running tests on
42 * known-good hardware we may want to trap these errors.
43 */
Mingming Caocd02ff02007-10-16 18:38:25 -040044#undef JBD2_PARANOID_IOFAIL
Dave Kleikamp470decc2006-10-11 01:20:57 -070045
46/*
47 * The default maximum commit age, in seconds.
48 */
Mingming Caocd02ff02007-10-16 18:38:25 -040049#define JBD2_DEFAULT_MAX_COMMIT_AGE 5
Dave Kleikamp470decc2006-10-11 01:20:57 -070050
Jose R. Santose23291b2007-07-18 08:57:06 -040051#ifdef CONFIG_JBD2_DEBUG
Dave Kleikamp470decc2006-10-11 01:20:57 -070052/*
Mingming Caocd02ff02007-10-16 18:38:25 -040053 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
Dave Kleikamp470decc2006-10-11 01:20:57 -070054 * consistency checks. By default we don't do this unless
Jose R. Santose23291b2007-07-18 08:57:06 -040055 * CONFIG_JBD2_DEBUG is on.
Dave Kleikamp470decc2006-10-11 01:20:57 -070056 */
Mingming Caocd02ff02007-10-16 18:38:25 -040057#define JBD2_EXPENSIVE_CHECKING
Jose R. Santos0f49d5d2007-07-18 08:50:18 -040058extern u8 jbd2_journal_enable_debug;
Dave Kleikamp470decc2006-10-11 01:20:57 -070059
60#define jbd_debug(n, f, a...) \
61 do { \
Mingming Caof7f4bcc2006-10-11 01:20:59 -070062 if ((n) <= jbd2_journal_enable_debug) { \
Dave Kleikamp470decc2006-10-11 01:20:57 -070063 printk (KERN_DEBUG "(%s, %d): %s: ", \
Harvey Harrisond5c003b2008-10-15 22:01:24 -070064 __FILE__, __LINE__, __func__); \
Dave Kleikamp470decc2006-10-11 01:20:57 -070065 printk (f, ## a); \
66 } \
67 } while (0)
68#else
69#define jbd_debug(f, a...) /**/
70#endif
71
Mingming Caoaf1e76d2007-10-16 18:38:25 -040072static inline void *jbd2_alloc(size_t size, gfp_t flags)
73{
74 return (void *)__get_free_pages(flags, get_order(size));
75}
76
77static inline void jbd2_free(void *ptr, size_t size)
78{
79 free_pages((unsigned long)ptr, get_order(size));
80};
81
Mingming Caof7f4bcc2006-10-11 01:20:59 -070082#define JBD2_MIN_JOURNAL_BLOCKS 1024
Dave Kleikamp470decc2006-10-11 01:20:57 -070083
84#ifdef __KERNEL__
85
86/**
87 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
88 *
89 * All filesystem modifications made by the process go
90 * through this handle. Recursive operations (such as quota operations)
91 * are gathered into a single update.
92 *
93 * The buffer credits field is used to account for journaled buffers
94 * being modified by the running process. To ensure that there is
95 * enough log space for all outstanding operations, we need to limit the
96 * number of outstanding buffers possible at any time. When the
97 * operation completes, any buffer credits not used are credited back to
98 * the transaction, so that at all times we know how many buffers the
99 * outstanding updates on a transaction might possibly touch.
100 *
101 * This is an opaque datatype.
102 **/
103typedef struct handle_s handle_t; /* Atomic operation type */
104
105
106/**
107 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
108 *
109 * journal_t is linked to from the fs superblock structure.
110 *
111 * We use the journal_t to keep track of all outstanding transaction
112 * activity on the filesystem, and to manage the state of the log
113 * writing process.
114 *
115 * This is an opaque datatype.
116 **/
117typedef struct journal_s journal_t; /* Journal control structure */
118#endif
119
120/*
121 * Internal structures used by the logging mechanism:
122 */
123
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700124#define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700125
126/*
127 * On-disk structures
128 */
129
130/*
131 * Descriptor block types:
132 */
133
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700134#define JBD2_DESCRIPTOR_BLOCK 1
135#define JBD2_COMMIT_BLOCK 2
136#define JBD2_SUPERBLOCK_V1 3
137#define JBD2_SUPERBLOCK_V2 4
138#define JBD2_REVOKE_BLOCK 5
Dave Kleikamp470decc2006-10-11 01:20:57 -0700139
140/*
141 * Standard header for all descriptor blocks:
142 */
143typedef struct journal_header_s
144{
145 __be32 h_magic;
146 __be32 h_blocktype;
147 __be32 h_sequence;
148} journal_header_t;
149
Girish Shilamkar818d2762008-01-28 23:58:27 -0500150/*
151 * Checksum types.
152 */
153#define JBD2_CRC32_CHKSUM 1
154#define JBD2_MD5_CHKSUM 2
155#define JBD2_SHA1_CHKSUM 3
156
157#define JBD2_CRC32_CHKSUM_SIZE 4
158
159#define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
160/*
161 * Commit block header for storing transactional checksums:
162 */
163struct commit_header {
164 __be32 h_magic;
165 __be32 h_blocktype;
166 __be32 h_sequence;
167 unsigned char h_chksum_type;
168 unsigned char h_chksum_size;
169 unsigned char h_padding[2];
170 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
Theodore Ts'o736603a2008-07-11 19:27:31 -0400171 __be64 h_commit_sec;
172 __be32 h_commit_nsec;
Girish Shilamkar818d2762008-01-28 23:58:27 -0500173};
Dave Kleikamp470decc2006-10-11 01:20:57 -0700174
175/*
Zach Brownb517bea2006-10-11 01:21:08 -0700176 * The block tag: used to describe a single buffer in the journal.
177 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
178 * raw struct shouldn't be used for pointer math or sizeof() - use
179 * journal_tag_bytes(journal) instead to compute this.
Dave Kleikamp470decc2006-10-11 01:20:57 -0700180 */
181typedef struct journal_block_tag_s
182{
183 __be32 t_blocknr; /* The on-disk block number */
184 __be32 t_flags; /* See below */
Zach Brownb517bea2006-10-11 01:21:08 -0700185 __be32 t_blocknr_high; /* most-significant high 32bits. */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700186} journal_block_tag_t;
187
Mingming Caocd02ff02007-10-16 18:38:25 -0400188#define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
189#define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
Zach Brownb517bea2006-10-11 01:21:08 -0700190
Dave Kleikamp470decc2006-10-11 01:20:57 -0700191/*
192 * The revoke descriptor: used on disk to describe a series of blocks to
193 * be revoked from the log
194 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700195typedef struct jbd2_journal_revoke_header_s
Dave Kleikamp470decc2006-10-11 01:20:57 -0700196{
197 journal_header_t r_header;
198 __be32 r_count; /* Count of bytes used in the block */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700199} jbd2_journal_revoke_header_t;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700200
201
202/* Definitions for the journal tag flags word: */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700203#define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
204#define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
205#define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
206#define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700207
208
209/*
210 * The journal superblock. All fields are in big-endian byte order.
211 */
212typedef struct journal_superblock_s
213{
214/* 0x0000 */
215 journal_header_t s_header;
216
217/* 0x000C */
218 /* Static information describing the journal */
219 __be32 s_blocksize; /* journal device blocksize */
220 __be32 s_maxlen; /* total blocks in journal file */
221 __be32 s_first; /* first block of log information */
222
223/* 0x0018 */
224 /* Dynamic information describing the current state of the log */
225 __be32 s_sequence; /* first commit ID expected in log */
226 __be32 s_start; /* blocknr of start of log */
227
228/* 0x0020 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700229 /* Error value, as set by jbd2_journal_abort(). */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700230 __be32 s_errno;
231
232/* 0x0024 */
233 /* Remaining fields are only valid in a version-2 superblock */
234 __be32 s_feature_compat; /* compatible feature set */
235 __be32 s_feature_incompat; /* incompatible feature set */
236 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
237/* 0x0030 */
238 __u8 s_uuid[16]; /* 128-bit uuid for journal */
239
240/* 0x0040 */
241 __be32 s_nr_users; /* Nr of filesystems sharing log */
242
243 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
244
245/* 0x0048 */
246 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
247 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
248
249/* 0x0050 */
250 __u32 s_padding[44];
251
252/* 0x0100 */
253 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
254/* 0x0400 */
255} journal_superblock_t;
256
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700257#define JBD2_HAS_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700258 ((j)->j_format_version >= 2 && \
259 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700260#define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700261 ((j)->j_format_version >= 2 && \
262 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700263#define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700264 ((j)->j_format_version >= 2 && \
265 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
266
Girish Shilamkar818d2762008-01-28 23:58:27 -0500267#define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
268
269#define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
270#define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
271#define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
Dave Kleikamp470decc2006-10-11 01:20:57 -0700272
273/* Features known to this kernel version: */
Girish Shilamkar818d2762008-01-28 23:58:27 -0500274#define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700275#define JBD2_KNOWN_ROCOMPAT_FEATURES 0
Zach Brownb517bea2006-10-11 01:21:08 -0700276#define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
Girish Shilamkar818d2762008-01-28 23:58:27 -0500277 JBD2_FEATURE_INCOMPAT_64BIT | \
278 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700279
280#ifdef __KERNEL__
281
282#include <linux/fs.h>
283#include <linux/sched.h>
284
Chris Snook36df53f2008-01-28 23:58:27 -0500285#define J_ASSERT(assert) BUG_ON(!(assert))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700286
287#if defined(CONFIG_BUFFER_DEBUG)
288void buffer_assertion_failure(struct buffer_head *bh);
289#define J_ASSERT_BH(bh, expr) \
290 do { \
291 if (!(expr)) \
292 buffer_assertion_failure(bh); \
293 J_ASSERT(expr); \
294 } while (0)
295#define J_ASSERT_JH(jh, expr) J_ASSERT_BH(jh2bh(jh), expr)
296#else
297#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
298#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
299#endif
300
Mingming Caocd02ff02007-10-16 18:38:25 -0400301#if defined(JBD2_PARANOID_IOFAIL)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700302#define J_EXPECT(expr, why...) J_ASSERT(expr)
303#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
304#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
305#else
306#define __journal_expect(expr, why...) \
307 ({ \
308 int val = (expr); \
309 if (!val) { \
310 printk(KERN_ERR \
311 "EXT3-fs unexpected failure: %s;\n",# expr); \
312 printk(KERN_ERR why "\n"); \
313 } \
314 val; \
315 })
316#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
317#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
318#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
319#endif
320
321enum jbd_state_bits {
322 BH_JBD /* Has an attached ext3 journal_head */
323 = BH_PrivateStart,
324 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
325 BH_Freed, /* Has been freed (truncated) */
326 BH_Revoked, /* Has been revoked from the log */
327 BH_RevokeValid, /* Revoked flag is valid */
328 BH_JBDDirty, /* Is dirty but journaled */
329 BH_State, /* Pins most journal_head state */
330 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
331 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
Mark Fashehe97fcd92008-11-18 17:15:24 -0800332 BH_JBDPrivateStart, /* First bit available for private use by FS */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700333};
334
335BUFFER_FNS(JBD, jbd)
336BUFFER_FNS(JWrite, jwrite)
337BUFFER_FNS(JBDDirty, jbddirty)
338TAS_BUFFER_FNS(JBDDirty, jbddirty)
339BUFFER_FNS(Revoked, revoked)
340TAS_BUFFER_FNS(Revoked, revoked)
341BUFFER_FNS(RevokeValid, revokevalid)
342TAS_BUFFER_FNS(RevokeValid, revokevalid)
343BUFFER_FNS(Freed, freed)
344
345static inline struct buffer_head *jh2bh(struct journal_head *jh)
346{
347 return jh->b_bh;
348}
349
350static inline struct journal_head *bh2jh(struct buffer_head *bh)
351{
352 return bh->b_private;
353}
354
355static inline void jbd_lock_bh_state(struct buffer_head *bh)
356{
357 bit_spin_lock(BH_State, &bh->b_state);
358}
359
360static inline int jbd_trylock_bh_state(struct buffer_head *bh)
361{
362 return bit_spin_trylock(BH_State, &bh->b_state);
363}
364
365static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
366{
367 return bit_spin_is_locked(BH_State, &bh->b_state);
368}
369
370static inline void jbd_unlock_bh_state(struct buffer_head *bh)
371{
372 bit_spin_unlock(BH_State, &bh->b_state);
373}
374
375static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
376{
377 bit_spin_lock(BH_JournalHead, &bh->b_state);
378}
379
380static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
381{
382 bit_spin_unlock(BH_JournalHead, &bh->b_state);
383}
384
Jan Karac851ed52008-07-11 19:27:31 -0400385/* Flags in jbd_inode->i_flags */
386#define __JI_COMMIT_RUNNING 0
387/* Commit of the inode data in progress. We use this flag to protect us from
388 * concurrent deletion of inode. We cannot use reference to inode for this
389 * since we cannot afford doing last iput() on behalf of kjournald
390 */
391#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
392
393/**
394 * struct jbd_inode is the structure linking inodes in ordered mode
395 * present in a transaction so that we can sync them during commit.
396 */
397struct jbd2_inode {
398 /* Which transaction does this inode belong to? Either the running
399 * transaction or the committing one. [j_list_lock] */
400 transaction_t *i_transaction;
401
402 /* Pointer to the running transaction modifying inode's data in case
403 * there is already a committing transaction touching it. [j_list_lock] */
404 transaction_t *i_next_transaction;
405
406 /* List of inodes in the i_transaction [j_list_lock] */
407 struct list_head i_list;
408
409 /* VFS inode this inode belongs to [constant during the lifetime
410 * of the structure] */
411 struct inode *i_vfs_inode;
412
413 /* Flags of inode [j_list_lock] */
414 unsigned int i_flags;
415};
416
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700417struct jbd2_revoke_table_s;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700418
419/**
420 * struct handle_s - The handle_s type is the concrete type associated with
421 * handle_t.
422 * @h_transaction: Which compound transaction is this update a part of?
423 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
424 * @h_ref: Reference count on this handle
425 * @h_err: Field for caller's use to track errors through large fs operations
426 * @h_sync: flag for sync-on-close
427 * @h_jdata: flag to force data journaling
428 * @h_aborted: flag indicating fatal error on handle
429 **/
430
431/* Docbook can't yet cope with the bit fields, but will leave the documentation
432 * in so it can be fixed later.
433 */
434
435struct handle_s
436{
437 /* Which compound transaction is this update a part of? */
438 transaction_t *h_transaction;
439
440 /* Number of remaining buffers we are allowed to dirty: */
441 int h_buffer_credits;
442
443 /* Reference count on this handle */
444 int h_ref;
445
446 /* Field for caller's use to track errors through large fs */
447 /* operations */
448 int h_err;
449
450 /* Flags [no locking] */
451 unsigned int h_sync: 1; /* sync-on-close */
452 unsigned int h_jdata: 1; /* force data journaling */
453 unsigned int h_aborted: 1; /* fatal error on handle */
Mingming Cao7b751062008-01-28 23:58:27 -0500454
455#ifdef CONFIG_DEBUG_LOCK_ALLOC
456 struct lockdep_map h_lockdep_map;
457#endif
Dave Kleikamp470decc2006-10-11 01:20:57 -0700458};
459
460
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500461/*
462 * Some stats for checkpoint phase
463 */
464struct transaction_chp_stats_s {
465 unsigned long cs_chp_time;
466 unsigned long cs_forced_to_close;
467 unsigned long cs_written;
468 unsigned long cs_dropped;
469};
470
Dave Kleikamp470decc2006-10-11 01:20:57 -0700471/* The transaction_t type is the guts of the journaling mechanism. It
472 * tracks a compound transaction through its various states:
473 *
474 * RUNNING: accepting new updates
475 * LOCKED: Updates still running but we don't accept new ones
476 * RUNDOWN: Updates are tidying up but have finished requesting
477 * new buffers to modify (state not used for now)
478 * FLUSH: All updates complete, but we are still writing to disk
479 * COMMIT: All data on disk, writing commit record
480 * FINISHED: We still have to keep the transaction for checkpointing.
481 *
482 * The transaction keeps track of all of the buffers modified by a
483 * running transaction, and all of the buffers committed but not yet
484 * flushed to home for finished transactions.
485 */
486
487/*
488 * Lock ranking:
489 *
490 * j_list_lock
491 * ->jbd_lock_bh_journal_head() (This is "innermost")
492 *
493 * j_state_lock
494 * ->jbd_lock_bh_state()
495 *
496 * jbd_lock_bh_state()
497 * ->j_list_lock
498 *
499 * j_state_lock
500 * ->t_handle_lock
501 *
502 * j_state_lock
503 * ->j_list_lock (journal_unmap_buffer)
504 *
505 */
506
507struct transaction_s
508{
509 /* Pointer to the journal for this transaction. [no locking] */
510 journal_t *t_journal;
511
512 /* Sequence number for this transaction [no locking] */
513 tid_t t_tid;
514
515 /*
516 * Transaction's current state
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700517 * [no locking - only kjournald2 alters this]
Jan Karaf5a7a6b2008-01-28 23:58:27 -0500518 * [j_list_lock] guards transition of a transaction into T_FINISHED
519 * state and subsequent call of __jbd2_journal_drop_transaction()
Dave Kleikamp470decc2006-10-11 01:20:57 -0700520 * FIXME: needs barriers
521 * KLUDGE: [use j_state_lock]
522 */
523 enum {
524 T_RUNNING,
525 T_LOCKED,
526 T_RUNDOWN,
527 T_FLUSH,
528 T_COMMIT,
529 T_FINISHED
530 } t_state;
531
532 /*
533 * Where in the log does this transaction's commit start? [no locking]
534 */
535 unsigned long t_log_start;
536
537 /* Number of buffers on the t_buffers list [j_list_lock] */
538 int t_nr_buffers;
539
540 /*
541 * Doubly-linked circular list of all buffers reserved but not yet
542 * modified by this transaction [j_list_lock]
543 */
544 struct journal_head *t_reserved_list;
545
546 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700547 * Doubly-linked circular list of all metadata buffers owned by this
548 * transaction [j_list_lock]
549 */
550 struct journal_head *t_buffers;
551
552 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700553 * Doubly-linked circular list of all forget buffers (superseded
554 * buffers which we can un-checkpoint once this transaction commits)
555 * [j_list_lock]
556 */
557 struct journal_head *t_forget;
558
559 /*
560 * Doubly-linked circular list of all buffers still to be flushed before
561 * this transaction can be checkpointed. [j_list_lock]
562 */
563 struct journal_head *t_checkpoint_list;
564
565 /*
566 * Doubly-linked circular list of all buffers submitted for IO while
567 * checkpointing. [j_list_lock]
568 */
569 struct journal_head *t_checkpoint_io_list;
570
571 /*
572 * Doubly-linked circular list of temporary buffers currently undergoing
573 * IO in the log [j_list_lock]
574 */
575 struct journal_head *t_iobuf_list;
576
577 /*
578 * Doubly-linked circular list of metadata buffers being shadowed by log
579 * IO. The IO buffers on the iobuf list and the shadow buffers on this
580 * list match each other one for one at all times. [j_list_lock]
581 */
582 struct journal_head *t_shadow_list;
583
584 /*
585 * Doubly-linked circular list of control buffers being written to the
586 * log. [j_list_lock]
587 */
588 struct journal_head *t_log_list;
589
590 /*
Jan Karac851ed52008-07-11 19:27:31 -0400591 * List of inodes whose data we've modified in data=ordered mode.
592 * [j_list_lock]
593 */
594 struct list_head t_inode_list;
595
596 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700597 * Protects info related to handles
598 */
599 spinlock_t t_handle_lock;
600
601 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500602 * Longest time some handle had to wait for running transaction
603 */
604 unsigned long t_max_wait;
605
606 /*
607 * When transaction started
608 */
609 unsigned long t_start;
610
611 /*
612 * Checkpointing stats [j_checkpoint_sem]
613 */
614 struct transaction_chp_stats_s t_chp_stats;
615
616 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700617 * Number of outstanding updates running on this transaction
618 * [t_handle_lock]
619 */
620 int t_updates;
621
622 /*
623 * Number of buffers reserved for use by all handles in this transaction
624 * handle but not yet modified. [t_handle_lock]
625 */
626 int t_outstanding_credits;
627
628 /*
629 * Forward and backward links for the circular list of all transactions
630 * awaiting checkpoint. [j_list_lock]
631 */
632 transaction_t *t_cpnext, *t_cpprev;
633
634 /*
635 * When will the transaction expire (become due for commit), in jiffies?
636 * [no locking]
637 */
638 unsigned long t_expires;
639
640 /*
Josef Bacike07f7182008-11-26 01:14:26 -0500641 * When this transaction started, in nanoseconds [no locking]
642 */
643 ktime_t t_start_time;
644
645 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700646 * How many handles used this transaction? [t_handle_lock]
647 */
648 int t_handle_count;
649
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400650 /*
651 * For use by the filesystem to store fs-specific data
652 * structures associated with the transaction
653 */
654 struct list_head t_private_list;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700655};
656
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500657struct transaction_run_stats_s {
658 unsigned long rs_wait;
659 unsigned long rs_running;
660 unsigned long rs_locked;
661 unsigned long rs_flushing;
662 unsigned long rs_logging;
663
664 unsigned long rs_handle_count;
665 unsigned long rs_blocks;
666 unsigned long rs_blocks_logged;
667};
668
669struct transaction_stats_s {
670 int ts_type;
671 unsigned long ts_tid;
672 union {
673 struct transaction_run_stats_s run;
674 struct transaction_chp_stats_s chp;
675 } u;
676};
677
678#define JBD2_STATS_RUN 1
679#define JBD2_STATS_CHECKPOINT 2
680
681static inline unsigned long
682jbd2_time_diff(unsigned long start, unsigned long end)
683{
684 if (end >= start)
685 return end - start;
686
687 return end + (MAX_JIFFY_OFFSET - start);
688}
689
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500690#define JBD2_NR_BATCH 64
691
Dave Kleikamp470decc2006-10-11 01:20:57 -0700692/**
693 * struct journal_s - The journal_s type is the concrete type associated with
694 * journal_t.
695 * @j_flags: General journaling state flags
696 * @j_errno: Is there an outstanding uncleared error on the journal (from a
697 * prior abort)?
698 * @j_sb_buffer: First part of superblock buffer
699 * @j_superblock: Second part of superblock buffer
700 * @j_format_version: Version of the superblock format
701 * @j_state_lock: Protect the various scalars in the journal
702 * @j_barrier_count: Number of processes waiting to create a barrier lock
703 * @j_barrier: The barrier lock itself
704 * @j_running_transaction: The current running transaction..
705 * @j_committing_transaction: the transaction we are pushing to disk
706 * @j_checkpoint_transactions: a linked circular list of all transactions
707 * waiting for checkpointing
708 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
709 * to start committing, or for a barrier lock to be released
710 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
711 * @j_wait_done_commit: Wait queue for waiting for commit to complete
712 * @j_wait_checkpoint: Wait queue to trigger checkpointing
713 * @j_wait_commit: Wait queue to trigger commit
714 * @j_wait_updates: Wait queue to wait for updates to complete
715 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
716 * @j_head: Journal head - identifies the first unused block in the journal
717 * @j_tail: Journal tail - identifies the oldest still-used block in the
718 * journal.
719 * @j_free: Journal free - how many free blocks are there in the journal?
720 * @j_first: The block number of the first usable block
721 * @j_last: The block number one beyond the last usable block
722 * @j_dev: Device where we store the journal
723 * @j_blocksize: blocksize for the location where we store the journal.
724 * @j_blk_offset: starting block offset for into the device where we store the
725 * journal
726 * @j_fs_dev: Device which holds the client fs. For internal journal this will
727 * be equal to j_dev
728 * @j_maxlen: Total maximum capacity of the journal region on disk.
729 * @j_list_lock: Protects the buffer lists and internal buffer state.
730 * @j_inode: Optional inode where we store the journal. If present, all journal
731 * block numbers are mapped into this inode via bmap().
732 * @j_tail_sequence: Sequence number of the oldest transaction in the log
733 * @j_transaction_sequence: Sequence number of the next transaction to grant
734 * @j_commit_sequence: Sequence number of the most recently committed
735 * transaction
736 * @j_commit_request: Sequence number of the most recent transaction wanting
737 * commit
738 * @j_uuid: Uuid of client object.
739 * @j_task: Pointer to the current commit thread for this journal
740 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
741 * single compound commit transaction
742 * @j_commit_interval: What is the maximum transaction lifetime before we begin
743 * a commit?
744 * @j_commit_timer: The timer used to wakeup the commit thread
745 * @j_revoke_lock: Protect the revoke table
746 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
747 * current transaction.
748 * @j_revoke_table: alternate revoke tables for j_revoke
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700749 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700750 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
751 * number that will fit in j_blocksize
752 * @j_last_sync_writer: most recent pid which did a synchronous write
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500753 * @j_history: Buffer storing the transactions statistics history
754 * @j_history_max: Maximum number of transactions in the statistics history
755 * @j_history_cur: Current number of transactions in the statistics history
756 * @j_history_lock: Protect the transactions statistics history
757 * @j_proc_entry: procfs entry for the jbd statistics directory
758 * @j_stats: Overall statistics
Dave Kleikamp470decc2006-10-11 01:20:57 -0700759 * @j_private: An opaque pointer to fs-private information.
760 */
761
762struct journal_s
763{
764 /* General journaling state flags [j_state_lock] */
765 unsigned long j_flags;
766
767 /*
768 * Is there an outstanding uncleared error on the journal (from a prior
769 * abort)? [j_state_lock]
770 */
771 int j_errno;
772
773 /* The superblock buffer */
774 struct buffer_head *j_sb_buffer;
775 journal_superblock_t *j_superblock;
776
777 /* Version of the superblock format */
778 int j_format_version;
779
780 /*
781 * Protect the various scalars in the journal
782 */
783 spinlock_t j_state_lock;
784
785 /*
786 * Number of processes waiting to create a barrier lock [j_state_lock]
787 */
788 int j_barrier_count;
789
790 /* The barrier lock itself */
791 struct mutex j_barrier;
792
793 /*
794 * Transactions: The current running transaction...
795 * [j_state_lock] [caller holding open handle]
796 */
797 transaction_t *j_running_transaction;
798
799 /*
800 * the transaction we are pushing to disk
801 * [j_state_lock] [caller holding open handle]
802 */
803 transaction_t *j_committing_transaction;
804
805 /*
806 * ... and a linked circular list of all transactions waiting for
807 * checkpointing. [j_list_lock]
808 */
809 transaction_t *j_checkpoint_transactions;
810
811 /*
812 * Wait queue for waiting for a locked transaction to start committing,
813 * or for a barrier lock to be released
814 */
815 wait_queue_head_t j_wait_transaction_locked;
816
817 /* Wait queue for waiting for checkpointing to complete */
818 wait_queue_head_t j_wait_logspace;
819
820 /* Wait queue for waiting for commit to complete */
821 wait_queue_head_t j_wait_done_commit;
822
823 /* Wait queue to trigger checkpointing */
824 wait_queue_head_t j_wait_checkpoint;
825
826 /* Wait queue to trigger commit */
827 wait_queue_head_t j_wait_commit;
828
829 /* Wait queue to wait for updates to complete */
830 wait_queue_head_t j_wait_updates;
831
832 /* Semaphore for locking against concurrent checkpoints */
833 struct mutex j_checkpoint_mutex;
834
835 /*
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500836 * List of buffer heads used by the checkpoint routine. This
837 * was moved from jbd2_log_do_checkpoint() to reduce stack
838 * usage. Access to this array is controlled by the
839 * j_checkpoint_mutex. [j_checkpoint_mutex]
840 */
841 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
842
843 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700844 * Journal head: identifies the first unused block in the journal.
845 * [j_state_lock]
846 */
847 unsigned long j_head;
848
849 /*
850 * Journal tail: identifies the oldest still-used block in the journal.
851 * [j_state_lock]
852 */
853 unsigned long j_tail;
854
855 /*
856 * Journal free: how many free blocks are there in the journal?
857 * [j_state_lock]
858 */
859 unsigned long j_free;
860
861 /*
862 * Journal start and end: the block numbers of the first usable block
863 * and one beyond the last usable block in the journal. [j_state_lock]
864 */
865 unsigned long j_first;
866 unsigned long j_last;
867
868 /*
869 * Device, blocksize and starting block offset for the location where we
870 * store the journal.
871 */
872 struct block_device *j_dev;
873 int j_blocksize;
Theodore Ts'o05496762008-09-16 14:36:17 -0400874 unsigned long long j_blk_offset;
875 char j_devname[BDEVNAME_SIZE+24];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700876
877 /*
878 * Device which holds the client fs. For internal journal this will be
879 * equal to j_dev.
880 */
881 struct block_device *j_fs_dev;
882
883 /* Total maximum capacity of the journal region on disk. */
884 unsigned int j_maxlen;
885
886 /*
887 * Protects the buffer lists and internal buffer state.
888 */
889 spinlock_t j_list_lock;
890
891 /* Optional inode where we store the journal. If present, all */
892 /* journal block numbers are mapped into this inode via */
893 /* bmap(). */
894 struct inode *j_inode;
895
896 /*
897 * Sequence number of the oldest transaction in the log [j_state_lock]
898 */
899 tid_t j_tail_sequence;
900
901 /*
902 * Sequence number of the next transaction to grant [j_state_lock]
903 */
904 tid_t j_transaction_sequence;
905
906 /*
907 * Sequence number of the most recently committed transaction
908 * [j_state_lock].
909 */
910 tid_t j_commit_sequence;
911
912 /*
913 * Sequence number of the most recent transaction wanting commit
914 * [j_state_lock]
915 */
916 tid_t j_commit_request;
917
918 /*
919 * Journal uuid: identifies the object (filesystem, LVM volume etc)
920 * backed by this journal. This will eventually be replaced by an array
921 * of uuids, allowing us to index multiple devices within a single
922 * journal and to perform atomic updates across them.
923 */
924 __u8 j_uuid[16];
925
926 /* Pointer to the current commit thread for this journal */
927 struct task_struct *j_task;
928
929 /*
930 * Maximum number of metadata buffers to allow in a single compound
931 * commit transaction
932 */
933 int j_max_transaction_buffers;
934
935 /*
936 * What is the maximum transaction lifetime before we begin a commit?
937 */
938 unsigned long j_commit_interval;
939
940 /* The timer used to wakeup the commit thread: */
941 struct timer_list j_commit_timer;
942
943 /*
944 * The revoke table: maintains the list of revoked blocks in the
945 * current transaction. [j_revoke_lock]
946 */
947 spinlock_t j_revoke_lock;
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700948 struct jbd2_revoke_table_s *j_revoke;
949 struct jbd2_revoke_table_s *j_revoke_table[2];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700950
951 /*
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700952 * array of bhs for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700953 */
954 struct buffer_head **j_wbuf;
955 int j_wbufsize;
956
Josef Bacike07f7182008-11-26 01:14:26 -0500957 /*
958 * this is the pid of hte last person to run a synchronous operation
959 * through the journal
960 */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700961 pid_t j_last_sync_writer;
962
Josef Bacike07f7182008-11-26 01:14:26 -0500963 /*
964 * the average amount of time in nanoseconds it takes to commit a
965 * transaction to disk. [j_state_lock]
966 */
967 u64 j_average_commit_time;
968
Theodore Ts'o30773842009-01-03 20:27:38 -0500969 /*
970 * minimum and maximum times that we should wait for
971 * additional filesystem operations to get batched into a
972 * synchronous handle in microseconds
973 */
974 u32 j_min_batch_time;
975 u32 j_max_batch_time;
976
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400977 /* This function is called when a transaction is closed */
978 void (*j_commit_callback)(journal_t *,
979 transaction_t *);
980
Dave Kleikamp470decc2006-10-11 01:20:57 -0700981 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500982 * Journal statistics
983 */
984 struct transaction_stats_s *j_history;
985 int j_history_max;
986 int j_history_cur;
987 /*
988 * Protect the transactions statistics history
989 */
990 spinlock_t j_history_lock;
991 struct proc_dir_entry *j_proc_entry;
992 struct transaction_stats_s j_stats;
993
Theodore Ts'o624080e2008-06-06 17:50:40 -0400994 /* Failed journal commit ID */
995 unsigned int j_failed_commit;
996
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500997 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700998 * An opaque pointer to fs-private information. ext3 puts its
999 * superblock pointer here
1000 */
1001 void *j_private;
1002};
1003
1004/*
1005 * Journal flag definitions
1006 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001007#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1008#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1009#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1010#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1011#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1012#define JBD2_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001013#define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1014 * data write error in ordered
1015 * mode */
Dave Kleikamp470decc2006-10-11 01:20:57 -07001016
1017/*
1018 * Function declarations for the journaling transaction and buffer
1019 * management
1020 */
1021
1022/* Filing buffers */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001023extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1024extern void __jbd2_journal_unfile_buffer(struct journal_head *);
1025extern void __jbd2_journal_refile_buffer(struct journal_head *);
1026extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1027extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001028extern void __journal_free_buffer(struct journal_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001029extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001030extern void __journal_clean_data_list(transaction_t *transaction);
1031
1032/* Log buffer allocation */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001033extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001034int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001035
1036/* Commit management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001037extern void jbd2_journal_commit_transaction(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001038
1039/* Checkpoint list management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001040int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1041int __jbd2_journal_remove_checkpoint(struct journal_head *);
1042void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001043
Joel Beckere06c8222008-09-11 15:35:47 -07001044
1045/*
1046 * Triggers
1047 */
1048
1049struct jbd2_buffer_trigger_type {
1050 /*
1051 * Fired just before a buffer is written to the journal.
1052 * mapped_data is a mapped buffer that is the frozen data for
1053 * commit.
1054 */
1055 void (*t_commit)(struct jbd2_buffer_trigger_type *type,
1056 struct buffer_head *bh, void *mapped_data,
1057 size_t size);
1058
1059 /*
1060 * Fired during journal abort for dirty buffers that will not be
1061 * committed.
1062 */
1063 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1064 struct buffer_head *bh);
1065};
1066
1067extern void jbd2_buffer_commit_trigger(struct journal_head *jh,
1068 void *mapped_data,
1069 struct jbd2_buffer_trigger_type *triggers);
1070extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1071 struct jbd2_buffer_trigger_type *triggers);
1072
Dave Kleikamp470decc2006-10-11 01:20:57 -07001073/* Buffer IO */
1074extern int
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001075jbd2_journal_write_metadata_buffer(transaction_t *transaction,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001076 struct journal_head *jh_in,
1077 struct journal_head **jh_out,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001078 unsigned long long blocknr);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001079
1080/* Transaction locking */
1081extern void __wait_on_journal (journal_t *);
1082
1083/*
1084 * Journal locking.
1085 *
1086 * We need to lock the journal during transaction state changes so that nobody
1087 * ever tries to take a handle on the running transaction while we are in the
1088 * middle of moving it to the commit phase. j_state_lock does this.
1089 *
1090 * Note that the locking is completely interrupt unsafe. We never touch
1091 * journal structures from interrupts.
1092 */
1093
1094static inline handle_t *journal_current_handle(void)
1095{
1096 return current->journal_info;
1097}
1098
1099/* The journaling code user interface:
1100 *
1101 * Create and destroy handles
1102 * Register buffer modifications against the current transaction.
1103 */
1104
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001105extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1106extern int jbd2_journal_restart (handle_t *, int nblocks);
1107extern int jbd2_journal_extend (handle_t *, int nblocks);
1108extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1109extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1110extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
Joel Beckere06c8222008-09-11 15:35:47 -07001111void jbd2_journal_set_triggers(struct buffer_head *,
1112 struct jbd2_buffer_trigger_type *type);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001113extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1114extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1115extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001116extern void journal_sync_buffer (struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001117extern void jbd2_journal_invalidatepage(journal_t *,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001118 struct page *, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001119extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1120extern int jbd2_journal_stop(handle_t *);
1121extern int jbd2_journal_flush (journal_t *);
1122extern void jbd2_journal_lock_updates (journal_t *);
1123extern void jbd2_journal_unlock_updates (journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001124
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001125extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001126 struct block_device *fs_dev,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001127 unsigned long long start, int len, int bsize);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001128extern journal_t * jbd2_journal_init_inode (struct inode *);
1129extern int jbd2_journal_update_format (journal_t *);
1130extern int jbd2_journal_check_used_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001131 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001132extern int jbd2_journal_check_available_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001133 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001134extern int jbd2_journal_set_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001135 (journal_t *, unsigned long, unsigned long, unsigned long);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001136extern void jbd2_journal_clear_features
1137 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001138extern int jbd2_journal_load (journal_t *journal);
Hidehiro Kawai44519fa2008-10-10 20:29:13 -04001139extern int jbd2_journal_destroy (journal_t *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001140extern int jbd2_journal_recover (journal_t *journal);
1141extern int jbd2_journal_wipe (journal_t *, int);
1142extern int jbd2_journal_skip_recovery (journal_t *);
1143extern void jbd2_journal_update_superblock (journal_t *, int);
1144extern void __jbd2_journal_abort_hard (journal_t *);
1145extern void jbd2_journal_abort (journal_t *, int);
1146extern int jbd2_journal_errno (journal_t *);
1147extern void jbd2_journal_ack_err (journal_t *);
1148extern int jbd2_journal_clear_err (journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001149extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001150extern int jbd2_journal_force_commit(journal_t *);
Jan Karac851ed52008-07-11 19:27:31 -04001151extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1152extern int jbd2_journal_begin_ordered_truncate(struct jbd2_inode *inode, loff_t new_size);
1153extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1154extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001155
1156/*
1157 * journal_head management
1158 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001159struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1160struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1161void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1162void jbd2_journal_put_journal_head(struct journal_head *jh);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001163
1164/*
1165 * handle management
1166 */
Christoph Lametere18b8902006-12-06 20:33:20 -08001167extern struct kmem_cache *jbd2_handle_cache;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001168
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001169static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001170{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001171 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001172}
1173
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001174static inline void jbd2_free_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001175{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001176 kmem_cache_free(jbd2_handle_cache, handle);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001177}
1178
1179/* Primary revoke support */
1180#define JOURNAL_REVOKE_DEFAULT_HASH 256
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001181extern int jbd2_journal_init_revoke(journal_t *, int);
1182extern void jbd2_journal_destroy_revoke_caches(void);
1183extern int jbd2_journal_init_revoke_caches(void);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001184
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001185extern void jbd2_journal_destroy_revoke(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001186extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001187extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1188extern void jbd2_journal_write_revoke_records(journal_t *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001189
1190/* Recovery revoke support */
Mingming Cao18eba7a2006-10-11 01:21:13 -07001191extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1192extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001193extern void jbd2_journal_clear_revoke(journal_t *);
1194extern void jbd2_journal_switch_revoke_table(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001195
1196/*
1197 * The log thread user interface:
1198 *
1199 * Request space in the current transaction, and force transaction commit
1200 * transitions on demand.
1201 */
1202
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001203int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1204int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1205int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1206int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1207int jbd2_journal_force_commit_nested(journal_t *journal);
1208int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1209int jbd2_log_do_checkpoint(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001210
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001211void __jbd2_log_wait_for_space(journal_t *journal);
Aneesh Kumar K.Vfb684072008-11-06 17:50:21 -05001212extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1213extern int jbd2_cleanup_journal_tail(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001214
1215/* Debugging code only: */
1216
1217#define jbd_ENOSYS() \
1218do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -07001219 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Dave Kleikamp470decc2006-10-11 01:20:57 -07001220 current->state = TASK_UNINTERRUPTIBLE; \
1221 schedule(); \
1222} while (1)
1223
1224/*
1225 * is_journal_abort
1226 *
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001227 * Simple test wrapper function to test the JBD2_ABORT state flag. This
Dave Kleikamp470decc2006-10-11 01:20:57 -07001228 * bit, when set, indicates that we have had a fatal error somewhere,
1229 * either inside the journaling layer or indicated to us by the client
1230 * (eg. ext3), and that we and should not commit any further
1231 * transactions.
1232 */
1233
1234static inline int is_journal_aborted(journal_t *journal)
1235{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001236 return journal->j_flags & JBD2_ABORT;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001237}
1238
1239static inline int is_handle_aborted(handle_t *handle)
1240{
1241 if (handle->h_aborted)
1242 return 1;
1243 return is_journal_aborted(handle->h_transaction->t_journal);
1244}
1245
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001246static inline void jbd2_journal_abort_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001247{
1248 handle->h_aborted = 1;
1249}
1250
1251#endif /* __KERNEL__ */
1252
1253/* Comparison functions for transaction IDs: perform comparisons using
1254 * modulo arithmetic so that they work over sequence number wraps. */
1255
1256static inline int tid_gt(tid_t x, tid_t y)
1257{
1258 int difference = (x - y);
1259 return (difference > 0);
1260}
1261
1262static inline int tid_geq(tid_t x, tid_t y)
1263{
1264 int difference = (x - y);
1265 return (difference >= 0);
1266}
1267
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001268extern int jbd2_journal_blocks_per_page(struct inode *inode);
Zach Brownb517bea2006-10-11 01:21:08 -07001269extern size_t journal_tag_bytes(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001270
1271/*
1272 * Return the minimum number of blocks which must be free in the journal
1273 * before a new transaction may be started. Must be called under j_state_lock.
1274 */
1275static inline int jbd_space_needed(journal_t *journal)
1276{
1277 int nblocks = journal->j_max_transaction_buffers;
1278 if (journal->j_committing_transaction)
1279 nblocks += journal->j_committing_transaction->
1280 t_outstanding_credits;
1281 return nblocks;
1282}
1283
1284/*
1285 * Definitions which augment the buffer_head layer
1286 */
1287
1288/* journaling buffer types */
1289#define BJ_None 0 /* Not journaled */
Jan Kara87c89c22008-07-11 19:27:31 -04001290#define BJ_Metadata 1 /* Normal journaled metadata */
1291#define BJ_Forget 2 /* Buffer superseded by this transaction */
1292#define BJ_IO 3 /* Buffer is for temporary IO use */
1293#define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1294#define BJ_LogCtl 5 /* Buffer contains log descriptors */
1295#define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1296#define BJ_Types 7
Dave Kleikamp470decc2006-10-11 01:20:57 -07001297
1298extern int jbd_blocks_per_page(struct inode *inode);
1299
1300#ifdef __KERNEL__
1301
1302#define buffer_trace_init(bh) do {} while (0)
1303#define print_buffer_fields(bh) do {} while (0)
1304#define print_buffer_trace(bh) do {} while (0)
1305#define BUFFER_TRACE(bh, info) do {} while (0)
1306#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1307#define JBUFFER_TRACE(jh, info) do {} while (0)
1308
1309#endif /* __KERNEL__ */
1310
Mingming Caocd02ff02007-10-16 18:38:25 -04001311#endif /* _LINUX_JBD2_H */