blob: 3dd2090070989863af4e87f9cc25b21900578bd9 [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: ", \
64 __FILE__, __LINE__, __FUNCTION__); \
65 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 */
332};
333
334BUFFER_FNS(JBD, jbd)
335BUFFER_FNS(JWrite, jwrite)
336BUFFER_FNS(JBDDirty, jbddirty)
337TAS_BUFFER_FNS(JBDDirty, jbddirty)
338BUFFER_FNS(Revoked, revoked)
339TAS_BUFFER_FNS(Revoked, revoked)
340BUFFER_FNS(RevokeValid, revokevalid)
341TAS_BUFFER_FNS(RevokeValid, revokevalid)
342BUFFER_FNS(Freed, freed)
343
344static inline struct buffer_head *jh2bh(struct journal_head *jh)
345{
346 return jh->b_bh;
347}
348
349static inline struct journal_head *bh2jh(struct buffer_head *bh)
350{
351 return bh->b_private;
352}
353
354static inline void jbd_lock_bh_state(struct buffer_head *bh)
355{
356 bit_spin_lock(BH_State, &bh->b_state);
357}
358
359static inline int jbd_trylock_bh_state(struct buffer_head *bh)
360{
361 return bit_spin_trylock(BH_State, &bh->b_state);
362}
363
364static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
365{
366 return bit_spin_is_locked(BH_State, &bh->b_state);
367}
368
369static inline void jbd_unlock_bh_state(struct buffer_head *bh)
370{
371 bit_spin_unlock(BH_State, &bh->b_state);
372}
373
374static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
375{
376 bit_spin_lock(BH_JournalHead, &bh->b_state);
377}
378
379static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
380{
381 bit_spin_unlock(BH_JournalHead, &bh->b_state);
382}
383
Jan Karac851ed52008-07-11 19:27:31 -0400384/* Flags in jbd_inode->i_flags */
385#define __JI_COMMIT_RUNNING 0
386/* Commit of the inode data in progress. We use this flag to protect us from
387 * concurrent deletion of inode. We cannot use reference to inode for this
388 * since we cannot afford doing last iput() on behalf of kjournald
389 */
390#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
391
392/**
393 * struct jbd_inode is the structure linking inodes in ordered mode
394 * present in a transaction so that we can sync them during commit.
395 */
396struct jbd2_inode {
397 /* Which transaction does this inode belong to? Either the running
398 * transaction or the committing one. [j_list_lock] */
399 transaction_t *i_transaction;
400
401 /* Pointer to the running transaction modifying inode's data in case
402 * there is already a committing transaction touching it. [j_list_lock] */
403 transaction_t *i_next_transaction;
404
405 /* List of inodes in the i_transaction [j_list_lock] */
406 struct list_head i_list;
407
408 /* VFS inode this inode belongs to [constant during the lifetime
409 * of the structure] */
410 struct inode *i_vfs_inode;
411
412 /* Flags of inode [j_list_lock] */
413 unsigned int i_flags;
414};
415
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700416struct jbd2_revoke_table_s;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700417
418/**
419 * struct handle_s - The handle_s type is the concrete type associated with
420 * handle_t.
421 * @h_transaction: Which compound transaction is this update a part of?
422 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
423 * @h_ref: Reference count on this handle
424 * @h_err: Field for caller's use to track errors through large fs operations
425 * @h_sync: flag for sync-on-close
426 * @h_jdata: flag to force data journaling
427 * @h_aborted: flag indicating fatal error on handle
428 **/
429
430/* Docbook can't yet cope with the bit fields, but will leave the documentation
431 * in so it can be fixed later.
432 */
433
434struct handle_s
435{
436 /* Which compound transaction is this update a part of? */
437 transaction_t *h_transaction;
438
439 /* Number of remaining buffers we are allowed to dirty: */
440 int h_buffer_credits;
441
442 /* Reference count on this handle */
443 int h_ref;
444
445 /* Field for caller's use to track errors through large fs */
446 /* operations */
447 int h_err;
448
449 /* Flags [no locking] */
450 unsigned int h_sync: 1; /* sync-on-close */
451 unsigned int h_jdata: 1; /* force data journaling */
452 unsigned int h_aborted: 1; /* fatal error on handle */
Mingming Cao7b751062008-01-28 23:58:27 -0500453
454#ifdef CONFIG_DEBUG_LOCK_ALLOC
455 struct lockdep_map h_lockdep_map;
456#endif
Dave Kleikamp470decc2006-10-11 01:20:57 -0700457};
458
459
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500460/*
461 * Some stats for checkpoint phase
462 */
463struct transaction_chp_stats_s {
464 unsigned long cs_chp_time;
465 unsigned long cs_forced_to_close;
466 unsigned long cs_written;
467 unsigned long cs_dropped;
468};
469
Dave Kleikamp470decc2006-10-11 01:20:57 -0700470/* The transaction_t type is the guts of the journaling mechanism. It
471 * tracks a compound transaction through its various states:
472 *
473 * RUNNING: accepting new updates
474 * LOCKED: Updates still running but we don't accept new ones
475 * RUNDOWN: Updates are tidying up but have finished requesting
476 * new buffers to modify (state not used for now)
477 * FLUSH: All updates complete, but we are still writing to disk
478 * COMMIT: All data on disk, writing commit record
479 * FINISHED: We still have to keep the transaction for checkpointing.
480 *
481 * The transaction keeps track of all of the buffers modified by a
482 * running transaction, and all of the buffers committed but not yet
483 * flushed to home for finished transactions.
484 */
485
486/*
487 * Lock ranking:
488 *
489 * j_list_lock
490 * ->jbd_lock_bh_journal_head() (This is "innermost")
491 *
492 * j_state_lock
493 * ->jbd_lock_bh_state()
494 *
495 * jbd_lock_bh_state()
496 * ->j_list_lock
497 *
498 * j_state_lock
499 * ->t_handle_lock
500 *
501 * j_state_lock
502 * ->j_list_lock (journal_unmap_buffer)
503 *
504 */
505
506struct transaction_s
507{
508 /* Pointer to the journal for this transaction. [no locking] */
509 journal_t *t_journal;
510
511 /* Sequence number for this transaction [no locking] */
512 tid_t t_tid;
513
514 /*
515 * Transaction's current state
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700516 * [no locking - only kjournald2 alters this]
Jan Karaf5a7a6b2008-01-28 23:58:27 -0500517 * [j_list_lock] guards transition of a transaction into T_FINISHED
518 * state and subsequent call of __jbd2_journal_drop_transaction()
Dave Kleikamp470decc2006-10-11 01:20:57 -0700519 * FIXME: needs barriers
520 * KLUDGE: [use j_state_lock]
521 */
522 enum {
523 T_RUNNING,
524 T_LOCKED,
525 T_RUNDOWN,
526 T_FLUSH,
527 T_COMMIT,
528 T_FINISHED
529 } t_state;
530
531 /*
532 * Where in the log does this transaction's commit start? [no locking]
533 */
534 unsigned long t_log_start;
535
536 /* Number of buffers on the t_buffers list [j_list_lock] */
537 int t_nr_buffers;
538
539 /*
540 * Doubly-linked circular list of all buffers reserved but not yet
541 * modified by this transaction [j_list_lock]
542 */
543 struct journal_head *t_reserved_list;
544
545 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700546 * Doubly-linked circular list of all metadata buffers owned by this
547 * transaction [j_list_lock]
548 */
549 struct journal_head *t_buffers;
550
551 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700552 * Doubly-linked circular list of all forget buffers (superseded
553 * buffers which we can un-checkpoint once this transaction commits)
554 * [j_list_lock]
555 */
556 struct journal_head *t_forget;
557
558 /*
559 * Doubly-linked circular list of all buffers still to be flushed before
560 * this transaction can be checkpointed. [j_list_lock]
561 */
562 struct journal_head *t_checkpoint_list;
563
564 /*
565 * Doubly-linked circular list of all buffers submitted for IO while
566 * checkpointing. [j_list_lock]
567 */
568 struct journal_head *t_checkpoint_io_list;
569
570 /*
571 * Doubly-linked circular list of temporary buffers currently undergoing
572 * IO in the log [j_list_lock]
573 */
574 struct journal_head *t_iobuf_list;
575
576 /*
577 * Doubly-linked circular list of metadata buffers being shadowed by log
578 * IO. The IO buffers on the iobuf list and the shadow buffers on this
579 * list match each other one for one at all times. [j_list_lock]
580 */
581 struct journal_head *t_shadow_list;
582
583 /*
584 * Doubly-linked circular list of control buffers being written to the
585 * log. [j_list_lock]
586 */
587 struct journal_head *t_log_list;
588
589 /*
Jan Karac851ed52008-07-11 19:27:31 -0400590 * List of inodes whose data we've modified in data=ordered mode.
591 * [j_list_lock]
592 */
593 struct list_head t_inode_list;
594
595 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700596 * Protects info related to handles
597 */
598 spinlock_t t_handle_lock;
599
600 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500601 * Longest time some handle had to wait for running transaction
602 */
603 unsigned long t_max_wait;
604
605 /*
606 * When transaction started
607 */
608 unsigned long t_start;
609
610 /*
611 * Checkpointing stats [j_checkpoint_sem]
612 */
613 struct transaction_chp_stats_s t_chp_stats;
614
615 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700616 * Number of outstanding updates running on this transaction
617 * [t_handle_lock]
618 */
619 int t_updates;
620
621 /*
622 * Number of buffers reserved for use by all handles in this transaction
623 * handle but not yet modified. [t_handle_lock]
624 */
625 int t_outstanding_credits;
626
627 /*
628 * Forward and backward links for the circular list of all transactions
629 * awaiting checkpoint. [j_list_lock]
630 */
631 transaction_t *t_cpnext, *t_cpprev;
632
633 /*
634 * When will the transaction expire (become due for commit), in jiffies?
635 * [no locking]
636 */
637 unsigned long t_expires;
638
639 /*
640 * How many handles used this transaction? [t_handle_lock]
641 */
642 int t_handle_count;
643
644};
645
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500646struct transaction_run_stats_s {
647 unsigned long rs_wait;
648 unsigned long rs_running;
649 unsigned long rs_locked;
650 unsigned long rs_flushing;
651 unsigned long rs_logging;
652
653 unsigned long rs_handle_count;
654 unsigned long rs_blocks;
655 unsigned long rs_blocks_logged;
656};
657
658struct transaction_stats_s {
659 int ts_type;
660 unsigned long ts_tid;
661 union {
662 struct transaction_run_stats_s run;
663 struct transaction_chp_stats_s chp;
664 } u;
665};
666
667#define JBD2_STATS_RUN 1
668#define JBD2_STATS_CHECKPOINT 2
669
670static inline unsigned long
671jbd2_time_diff(unsigned long start, unsigned long end)
672{
673 if (end >= start)
674 return end - start;
675
676 return end + (MAX_JIFFY_OFFSET - start);
677}
678
Dave Kleikamp470decc2006-10-11 01:20:57 -0700679/**
680 * struct journal_s - The journal_s type is the concrete type associated with
681 * journal_t.
682 * @j_flags: General journaling state flags
683 * @j_errno: Is there an outstanding uncleared error on the journal (from a
684 * prior abort)?
685 * @j_sb_buffer: First part of superblock buffer
686 * @j_superblock: Second part of superblock buffer
687 * @j_format_version: Version of the superblock format
688 * @j_state_lock: Protect the various scalars in the journal
689 * @j_barrier_count: Number of processes waiting to create a barrier lock
690 * @j_barrier: The barrier lock itself
691 * @j_running_transaction: The current running transaction..
692 * @j_committing_transaction: the transaction we are pushing to disk
693 * @j_checkpoint_transactions: a linked circular list of all transactions
694 * waiting for checkpointing
695 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
696 * to start committing, or for a barrier lock to be released
697 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
698 * @j_wait_done_commit: Wait queue for waiting for commit to complete
699 * @j_wait_checkpoint: Wait queue to trigger checkpointing
700 * @j_wait_commit: Wait queue to trigger commit
701 * @j_wait_updates: Wait queue to wait for updates to complete
702 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
703 * @j_head: Journal head - identifies the first unused block in the journal
704 * @j_tail: Journal tail - identifies the oldest still-used block in the
705 * journal.
706 * @j_free: Journal free - how many free blocks are there in the journal?
707 * @j_first: The block number of the first usable block
708 * @j_last: The block number one beyond the last usable block
709 * @j_dev: Device where we store the journal
710 * @j_blocksize: blocksize for the location where we store the journal.
711 * @j_blk_offset: starting block offset for into the device where we store the
712 * journal
713 * @j_fs_dev: Device which holds the client fs. For internal journal this will
714 * be equal to j_dev
715 * @j_maxlen: Total maximum capacity of the journal region on disk.
716 * @j_list_lock: Protects the buffer lists and internal buffer state.
717 * @j_inode: Optional inode where we store the journal. If present, all journal
718 * block numbers are mapped into this inode via bmap().
719 * @j_tail_sequence: Sequence number of the oldest transaction in the log
720 * @j_transaction_sequence: Sequence number of the next transaction to grant
721 * @j_commit_sequence: Sequence number of the most recently committed
722 * transaction
723 * @j_commit_request: Sequence number of the most recent transaction wanting
724 * commit
725 * @j_uuid: Uuid of client object.
726 * @j_task: Pointer to the current commit thread for this journal
727 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
728 * single compound commit transaction
729 * @j_commit_interval: What is the maximum transaction lifetime before we begin
730 * a commit?
731 * @j_commit_timer: The timer used to wakeup the commit thread
732 * @j_revoke_lock: Protect the revoke table
733 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
734 * current transaction.
735 * @j_revoke_table: alternate revoke tables for j_revoke
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700736 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700737 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
738 * number that will fit in j_blocksize
739 * @j_last_sync_writer: most recent pid which did a synchronous write
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500740 * @j_history: Buffer storing the transactions statistics history
741 * @j_history_max: Maximum number of transactions in the statistics history
742 * @j_history_cur: Current number of transactions in the statistics history
743 * @j_history_lock: Protect the transactions statistics history
744 * @j_proc_entry: procfs entry for the jbd statistics directory
745 * @j_stats: Overall statistics
Dave Kleikamp470decc2006-10-11 01:20:57 -0700746 * @j_private: An opaque pointer to fs-private information.
747 */
748
749struct journal_s
750{
751 /* General journaling state flags [j_state_lock] */
752 unsigned long j_flags;
753
754 /*
755 * Is there an outstanding uncleared error on the journal (from a prior
756 * abort)? [j_state_lock]
757 */
758 int j_errno;
759
760 /* The superblock buffer */
761 struct buffer_head *j_sb_buffer;
762 journal_superblock_t *j_superblock;
763
764 /* Version of the superblock format */
765 int j_format_version;
766
767 /*
768 * Protect the various scalars in the journal
769 */
770 spinlock_t j_state_lock;
771
772 /*
773 * Number of processes waiting to create a barrier lock [j_state_lock]
774 */
775 int j_barrier_count;
776
777 /* The barrier lock itself */
778 struct mutex j_barrier;
779
780 /*
781 * Transactions: The current running transaction...
782 * [j_state_lock] [caller holding open handle]
783 */
784 transaction_t *j_running_transaction;
785
786 /*
787 * the transaction we are pushing to disk
788 * [j_state_lock] [caller holding open handle]
789 */
790 transaction_t *j_committing_transaction;
791
792 /*
793 * ... and a linked circular list of all transactions waiting for
794 * checkpointing. [j_list_lock]
795 */
796 transaction_t *j_checkpoint_transactions;
797
798 /*
799 * Wait queue for waiting for a locked transaction to start committing,
800 * or for a barrier lock to be released
801 */
802 wait_queue_head_t j_wait_transaction_locked;
803
804 /* Wait queue for waiting for checkpointing to complete */
805 wait_queue_head_t j_wait_logspace;
806
807 /* Wait queue for waiting for commit to complete */
808 wait_queue_head_t j_wait_done_commit;
809
810 /* Wait queue to trigger checkpointing */
811 wait_queue_head_t j_wait_checkpoint;
812
813 /* Wait queue to trigger commit */
814 wait_queue_head_t j_wait_commit;
815
816 /* Wait queue to wait for updates to complete */
817 wait_queue_head_t j_wait_updates;
818
819 /* Semaphore for locking against concurrent checkpoints */
820 struct mutex j_checkpoint_mutex;
821
822 /*
823 * Journal head: identifies the first unused block in the journal.
824 * [j_state_lock]
825 */
826 unsigned long j_head;
827
828 /*
829 * Journal tail: identifies the oldest still-used block in the journal.
830 * [j_state_lock]
831 */
832 unsigned long j_tail;
833
834 /*
835 * Journal free: how many free blocks are there in the journal?
836 * [j_state_lock]
837 */
838 unsigned long j_free;
839
840 /*
841 * Journal start and end: the block numbers of the first usable block
842 * and one beyond the last usable block in the journal. [j_state_lock]
843 */
844 unsigned long j_first;
845 unsigned long j_last;
846
847 /*
848 * Device, blocksize and starting block offset for the location where we
849 * store the journal.
850 */
851 struct block_device *j_dev;
852 int j_blocksize;
Mingming Cao18eba7a2006-10-11 01:21:13 -0700853 unsigned long long j_blk_offset;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700854
855 /*
856 * Device which holds the client fs. For internal journal this will be
857 * equal to j_dev.
858 */
859 struct block_device *j_fs_dev;
860
861 /* Total maximum capacity of the journal region on disk. */
862 unsigned int j_maxlen;
863
864 /*
865 * Protects the buffer lists and internal buffer state.
866 */
867 spinlock_t j_list_lock;
868
869 /* Optional inode where we store the journal. If present, all */
870 /* journal block numbers are mapped into this inode via */
871 /* bmap(). */
872 struct inode *j_inode;
873
874 /*
875 * Sequence number of the oldest transaction in the log [j_state_lock]
876 */
877 tid_t j_tail_sequence;
878
879 /*
880 * Sequence number of the next transaction to grant [j_state_lock]
881 */
882 tid_t j_transaction_sequence;
883
884 /*
885 * Sequence number of the most recently committed transaction
886 * [j_state_lock].
887 */
888 tid_t j_commit_sequence;
889
890 /*
891 * Sequence number of the most recent transaction wanting commit
892 * [j_state_lock]
893 */
894 tid_t j_commit_request;
895
896 /*
897 * Journal uuid: identifies the object (filesystem, LVM volume etc)
898 * backed by this journal. This will eventually be replaced by an array
899 * of uuids, allowing us to index multiple devices within a single
900 * journal and to perform atomic updates across them.
901 */
902 __u8 j_uuid[16];
903
904 /* Pointer to the current commit thread for this journal */
905 struct task_struct *j_task;
906
907 /*
908 * Maximum number of metadata buffers to allow in a single compound
909 * commit transaction
910 */
911 int j_max_transaction_buffers;
912
913 /*
914 * What is the maximum transaction lifetime before we begin a commit?
915 */
916 unsigned long j_commit_interval;
917
918 /* The timer used to wakeup the commit thread: */
919 struct timer_list j_commit_timer;
920
921 /*
922 * The revoke table: maintains the list of revoked blocks in the
923 * current transaction. [j_revoke_lock]
924 */
925 spinlock_t j_revoke_lock;
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700926 struct jbd2_revoke_table_s *j_revoke;
927 struct jbd2_revoke_table_s *j_revoke_table[2];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700928
929 /*
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700930 * array of bhs for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700931 */
932 struct buffer_head **j_wbuf;
933 int j_wbufsize;
934
935 pid_t j_last_sync_writer;
936
937 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500938 * Journal statistics
939 */
940 struct transaction_stats_s *j_history;
941 int j_history_max;
942 int j_history_cur;
943 /*
944 * Protect the transactions statistics history
945 */
946 spinlock_t j_history_lock;
947 struct proc_dir_entry *j_proc_entry;
948 struct transaction_stats_s j_stats;
949
Theodore Ts'o624080e2008-06-06 17:50:40 -0400950 /* Failed journal commit ID */
951 unsigned int j_failed_commit;
952
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500953 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700954 * An opaque pointer to fs-private information. ext3 puts its
955 * superblock pointer here
956 */
957 void *j_private;
958};
959
960/*
961 * Journal flag definitions
962 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700963#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
964#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
965#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
966#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
967#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
968#define JBD2_BARRIER 0x020 /* Use IDE barriers */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700969
970/*
971 * Function declarations for the journaling transaction and buffer
972 * management
973 */
974
975/* Filing buffers */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700976extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
977extern void __jbd2_journal_unfile_buffer(struct journal_head *);
978extern void __jbd2_journal_refile_buffer(struct journal_head *);
979extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
980extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700981extern void __journal_free_buffer(struct journal_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700982extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700983extern void __journal_clean_data_list(transaction_t *transaction);
984
985/* Log buffer allocation */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700986extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -0700987int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700988
989/* Commit management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700990extern void jbd2_journal_commit_transaction(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700991
992/* Checkpoint list management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700993int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
994int __jbd2_journal_remove_checkpoint(struct journal_head *);
995void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700996
997/* Buffer IO */
998extern int
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700999jbd2_journal_write_metadata_buffer(transaction_t *transaction,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001000 struct journal_head *jh_in,
1001 struct journal_head **jh_out,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001002 unsigned long long blocknr);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001003
1004/* Transaction locking */
1005extern void __wait_on_journal (journal_t *);
1006
1007/*
1008 * Journal locking.
1009 *
1010 * We need to lock the journal during transaction state changes so that nobody
1011 * ever tries to take a handle on the running transaction while we are in the
1012 * middle of moving it to the commit phase. j_state_lock does this.
1013 *
1014 * Note that the locking is completely interrupt unsafe. We never touch
1015 * journal structures from interrupts.
1016 */
1017
1018static inline handle_t *journal_current_handle(void)
1019{
1020 return current->journal_info;
1021}
1022
1023/* The journaling code user interface:
1024 *
1025 * Create and destroy handles
1026 * Register buffer modifications against the current transaction.
1027 */
1028
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001029extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1030extern int jbd2_journal_restart (handle_t *, int nblocks);
1031extern int jbd2_journal_extend (handle_t *, int nblocks);
1032extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1033extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1034extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001035extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1036extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1037extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001038extern void journal_sync_buffer (struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001039extern void jbd2_journal_invalidatepage(journal_t *,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001040 struct page *, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001041extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1042extern int jbd2_journal_stop(handle_t *);
1043extern int jbd2_journal_flush (journal_t *);
1044extern void jbd2_journal_lock_updates (journal_t *);
1045extern void jbd2_journal_unlock_updates (journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001046
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001047extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001048 struct block_device *fs_dev,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001049 unsigned long long start, int len, int bsize);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001050extern journal_t * jbd2_journal_init_inode (struct inode *);
1051extern int jbd2_journal_update_format (journal_t *);
1052extern int jbd2_journal_check_used_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001053 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001054extern int jbd2_journal_check_available_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001055 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001056extern int jbd2_journal_set_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001057 (journal_t *, unsigned long, unsigned long, unsigned long);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001058extern void jbd2_journal_clear_features
1059 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001060extern int jbd2_journal_create (journal_t *);
1061extern int jbd2_journal_load (journal_t *journal);
1062extern void jbd2_journal_destroy (journal_t *);
1063extern int jbd2_journal_recover (journal_t *journal);
1064extern int jbd2_journal_wipe (journal_t *, int);
1065extern int jbd2_journal_skip_recovery (journal_t *);
1066extern void jbd2_journal_update_superblock (journal_t *, int);
1067extern void __jbd2_journal_abort_hard (journal_t *);
1068extern void jbd2_journal_abort (journal_t *, int);
1069extern int jbd2_journal_errno (journal_t *);
1070extern void jbd2_journal_ack_err (journal_t *);
1071extern int jbd2_journal_clear_err (journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001072extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001073extern int jbd2_journal_force_commit(journal_t *);
Jan Karac851ed52008-07-11 19:27:31 -04001074extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1075extern int jbd2_journal_begin_ordered_truncate(struct jbd2_inode *inode, loff_t new_size);
1076extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1077extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001078
1079/*
1080 * journal_head management
1081 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001082struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1083struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1084void jbd2_journal_remove_journal_head(struct buffer_head *bh);
1085void jbd2_journal_put_journal_head(struct journal_head *jh);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001086
1087/*
1088 * handle management
1089 */
Christoph Lametere18b8902006-12-06 20:33:20 -08001090extern struct kmem_cache *jbd2_handle_cache;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001091
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001092static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001093{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001094 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001095}
1096
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001097static inline void jbd2_free_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001098{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001099 kmem_cache_free(jbd2_handle_cache, handle);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001100}
1101
1102/* Primary revoke support */
1103#define JOURNAL_REVOKE_DEFAULT_HASH 256
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001104extern int jbd2_journal_init_revoke(journal_t *, int);
1105extern void jbd2_journal_destroy_revoke_caches(void);
1106extern int jbd2_journal_init_revoke_caches(void);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001107
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001108extern void jbd2_journal_destroy_revoke(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001109extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001110extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1111extern void jbd2_journal_write_revoke_records(journal_t *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001112
1113/* Recovery revoke support */
Mingming Cao18eba7a2006-10-11 01:21:13 -07001114extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1115extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001116extern void jbd2_journal_clear_revoke(journal_t *);
1117extern void jbd2_journal_switch_revoke_table(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001118
1119/*
1120 * The log thread user interface:
1121 *
1122 * Request space in the current transaction, and force transaction commit
1123 * transitions on demand.
1124 */
1125
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001126int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1127int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1128int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1129int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1130int jbd2_journal_force_commit_nested(journal_t *journal);
1131int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1132int jbd2_log_do_checkpoint(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001133
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001134void __jbd2_log_wait_for_space(journal_t *journal);
1135extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1136extern int jbd2_cleanup_journal_tail(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001137
1138/* Debugging code only: */
1139
1140#define jbd_ENOSYS() \
1141do { \
1142 printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1143 current->state = TASK_UNINTERRUPTIBLE; \
1144 schedule(); \
1145} while (1)
1146
1147/*
1148 * is_journal_abort
1149 *
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001150 * Simple test wrapper function to test the JBD2_ABORT state flag. This
Dave Kleikamp470decc2006-10-11 01:20:57 -07001151 * bit, when set, indicates that we have had a fatal error somewhere,
1152 * either inside the journaling layer or indicated to us by the client
1153 * (eg. ext3), and that we and should not commit any further
1154 * transactions.
1155 */
1156
1157static inline int is_journal_aborted(journal_t *journal)
1158{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001159 return journal->j_flags & JBD2_ABORT;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001160}
1161
1162static inline int is_handle_aborted(handle_t *handle)
1163{
1164 if (handle->h_aborted)
1165 return 1;
1166 return is_journal_aborted(handle->h_transaction->t_journal);
1167}
1168
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001169static inline void jbd2_journal_abort_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001170{
1171 handle->h_aborted = 1;
1172}
1173
1174#endif /* __KERNEL__ */
1175
1176/* Comparison functions for transaction IDs: perform comparisons using
1177 * modulo arithmetic so that they work over sequence number wraps. */
1178
1179static inline int tid_gt(tid_t x, tid_t y)
1180{
1181 int difference = (x - y);
1182 return (difference > 0);
1183}
1184
1185static inline int tid_geq(tid_t x, tid_t y)
1186{
1187 int difference = (x - y);
1188 return (difference >= 0);
1189}
1190
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001191extern int jbd2_journal_blocks_per_page(struct inode *inode);
Zach Brownb517bea2006-10-11 01:21:08 -07001192extern size_t journal_tag_bytes(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001193
1194/*
1195 * Return the minimum number of blocks which must be free in the journal
1196 * before a new transaction may be started. Must be called under j_state_lock.
1197 */
1198static inline int jbd_space_needed(journal_t *journal)
1199{
1200 int nblocks = journal->j_max_transaction_buffers;
1201 if (journal->j_committing_transaction)
1202 nblocks += journal->j_committing_transaction->
1203 t_outstanding_credits;
1204 return nblocks;
1205}
1206
1207/*
1208 * Definitions which augment the buffer_head layer
1209 */
1210
1211/* journaling buffer types */
1212#define BJ_None 0 /* Not journaled */
Jan Kara87c89c22008-07-11 19:27:31 -04001213#define BJ_Metadata 1 /* Normal journaled metadata */
1214#define BJ_Forget 2 /* Buffer superseded by this transaction */
1215#define BJ_IO 3 /* Buffer is for temporary IO use */
1216#define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1217#define BJ_LogCtl 5 /* Buffer contains log descriptors */
1218#define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1219#define BJ_Types 7
Dave Kleikamp470decc2006-10-11 01:20:57 -07001220
1221extern int jbd_blocks_per_page(struct inode *inode);
1222
1223#ifdef __KERNEL__
1224
1225#define buffer_trace_init(bh) do {} while (0)
1226#define print_buffer_fields(bh) do {} while (0)
1227#define print_buffer_trace(bh) do {} while (0)
1228#define BUFFER_TRACE(bh, info) do {} while (0)
1229#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1230#define JBUFFER_TRACE(jh, info) do {} while (0)
1231
1232#endif /* __KERNEL__ */
1233
Mingming Caocd02ff02007-10-16 18:38:25 -04001234#endif /* _LINUX_JBD2_H */