blob: 38f307b8c3342c4bef86ac3366c785696df54cce [file] [log] [blame]
Dave Kleikamp470decc2006-10-11 01:20:57 -07001/*
Mingming Caof7f4bcc2006-10-11 01:20:59 -07002 * linux/include/linux/jbd2.h
Dave Kleikamp470decc2006-10-11 01:20:57 -07003 *
4 * Written by Stephen C. Tweedie <sct@redhat.com>
5 *
6 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
7 *
8 * This file is part of the Linux kernel and is made available under
9 * the terms of the GNU General Public License, version 2, or at your
10 * option, any later version, incorporated herein by reference.
11 *
12 * Definitions for transaction data structures for the buffer cache
13 * filesystem journaling support.
14 */
15
Mingming Caocd02ff02007-10-16 18:38:25 -040016#ifndef _LINUX_JBD2_H
17#define _LINUX_JBD2_H
Dave Kleikamp470decc2006-10-11 01:20:57 -070018
19/* Allow this file to be included directly into e2fsprogs */
20#ifndef __KERNEL__
21#include "jfs_compat.h"
Mingming Caof7f4bcc2006-10-11 01:20:59 -070022#define JBD2_DEBUG
Dave Kleikamp470decc2006-10-11 01:20:57 -070023#define jfs_debug jbd_debug
24#else
25
26#include <linux/types.h>
27#include <linux/buffer_head.h>
28#include <linux/journal-head.h>
29#include <linux/stddef.h>
30#include <linux/bit_spinlock.h>
31#include <linux/mutex.h>
32#include <linux/timer.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090033#include <linux/slab.h>
Dave Kleikamp470decc2006-10-11 01:20:57 -070034#endif
35
36#define journal_oom_retry 1
37
38/*
Mingming Caocd02ff02007-10-16 18:38:25 -040039 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
Dave Kleikamp470decc2006-10-11 01:20:57 -070040 * certain classes of error which can occur due to failed IOs. Under
Mingming Caocd02ff02007-10-16 18:38:25 -040041 * normal use we want ext4 to continue after such errors, because
Dave Kleikamp470decc2006-10-11 01:20:57 -070042 * hardware _can_ fail, but for debugging purposes when running tests on
43 * known-good hardware we may want to trap these errors.
44 */
Mingming Caocd02ff02007-10-16 18:38:25 -040045#undef JBD2_PARANOID_IOFAIL
Dave Kleikamp470decc2006-10-11 01:20:57 -070046
47/*
48 * The default maximum commit age, in seconds.
49 */
Mingming Caocd02ff02007-10-16 18:38:25 -040050#define JBD2_DEFAULT_MAX_COMMIT_AGE 5
Dave Kleikamp470decc2006-10-11 01:20:57 -070051
Jose R. Santose23291b2007-07-18 08:57:06 -040052#ifdef CONFIG_JBD2_DEBUG
Dave Kleikamp470decc2006-10-11 01:20:57 -070053/*
Mingming Caocd02ff02007-10-16 18:38:25 -040054 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
Dave Kleikamp470decc2006-10-11 01:20:57 -070055 * consistency checks. By default we don't do this unless
Jose R. Santose23291b2007-07-18 08:57:06 -040056 * CONFIG_JBD2_DEBUG is on.
Dave Kleikamp470decc2006-10-11 01:20:57 -070057 */
Mingming Caocd02ff02007-10-16 18:38:25 -040058#define JBD2_EXPENSIVE_CHECKING
Jose R. Santos0f49d5d2007-07-18 08:50:18 -040059extern u8 jbd2_journal_enable_debug;
Dave Kleikamp470decc2006-10-11 01:20:57 -070060
61#define jbd_debug(n, f, a...) \
62 do { \
Mingming Caof7f4bcc2006-10-11 01:20:59 -070063 if ((n) <= jbd2_journal_enable_debug) { \
Dave Kleikamp470decc2006-10-11 01:20:57 -070064 printk (KERN_DEBUG "(%s, %d): %s: ", \
Harvey Harrisond5c003b2008-10-15 22:01:24 -070065 __FILE__, __LINE__, __func__); \
Dave Kleikamp470decc2006-10-11 01:20:57 -070066 printk (f, ## a); \
67 } \
68 } while (0)
69#else
70#define jbd_debug(f, a...) /**/
71#endif
72
Theodore Ts'od2eecb02009-12-07 10:36:20 -050073extern void *jbd2_alloc(size_t size, gfp_t flags);
74extern void jbd2_free(void *ptr, size_t size);
Mingming Caoaf1e76d2007-10-16 18:38:25 -040075
Mingming Caof7f4bcc2006-10-11 01:20:59 -070076#define JBD2_MIN_JOURNAL_BLOCKS 1024
Dave Kleikamp470decc2006-10-11 01:20:57 -070077
78#ifdef __KERNEL__
79
80/**
81 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
82 *
83 * All filesystem modifications made by the process go
84 * through this handle. Recursive operations (such as quota operations)
85 * are gathered into a single update.
86 *
87 * The buffer credits field is used to account for journaled buffers
88 * being modified by the running process. To ensure that there is
89 * enough log space for all outstanding operations, we need to limit the
90 * number of outstanding buffers possible at any time. When the
91 * operation completes, any buffer credits not used are credited back to
92 * the transaction, so that at all times we know how many buffers the
93 * outstanding updates on a transaction might possibly touch.
94 *
95 * This is an opaque datatype.
96 **/
Theodore Ts'o8aefcd52011-01-10 12:29:43 -050097typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
Dave Kleikamp470decc2006-10-11 01:20:57 -070098
99
100/**
101 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
102 *
103 * journal_t is linked to from the fs superblock structure.
104 *
105 * We use the journal_t to keep track of all outstanding transaction
106 * activity on the filesystem, and to manage the state of the log
107 * writing process.
108 *
109 * This is an opaque datatype.
110 **/
111typedef struct journal_s journal_t; /* Journal control structure */
112#endif
113
114/*
115 * Internal structures used by the logging mechanism:
116 */
117
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700118#define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700119
120/*
121 * On-disk structures
122 */
123
124/*
125 * Descriptor block types:
126 */
127
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700128#define JBD2_DESCRIPTOR_BLOCK 1
129#define JBD2_COMMIT_BLOCK 2
130#define JBD2_SUPERBLOCK_V1 3
131#define JBD2_SUPERBLOCK_V2 4
132#define JBD2_REVOKE_BLOCK 5
Dave Kleikamp470decc2006-10-11 01:20:57 -0700133
134/*
135 * Standard header for all descriptor blocks:
136 */
137typedef struct journal_header_s
138{
139 __be32 h_magic;
140 __be32 h_blocktype;
141 __be32 h_sequence;
142} journal_header_t;
143
Girish Shilamkar818d2762008-01-28 23:58:27 -0500144/*
145 * Checksum types.
146 */
147#define JBD2_CRC32_CHKSUM 1
148#define JBD2_MD5_CHKSUM 2
149#define JBD2_SHA1_CHKSUM 3
150
151#define JBD2_CRC32_CHKSUM_SIZE 4
152
153#define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
154/*
155 * Commit block header for storing transactional checksums:
156 */
157struct commit_header {
158 __be32 h_magic;
159 __be32 h_blocktype;
160 __be32 h_sequence;
161 unsigned char h_chksum_type;
162 unsigned char h_chksum_size;
163 unsigned char h_padding[2];
164 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
Theodore Ts'o736603a2008-07-11 19:27:31 -0400165 __be64 h_commit_sec;
166 __be32 h_commit_nsec;
Girish Shilamkar818d2762008-01-28 23:58:27 -0500167};
Dave Kleikamp470decc2006-10-11 01:20:57 -0700168
169/*
Zach Brownb517bea2006-10-11 01:21:08 -0700170 * The block tag: used to describe a single buffer in the journal.
171 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
172 * raw struct shouldn't be used for pointer math or sizeof() - use
173 * journal_tag_bytes(journal) instead to compute this.
Dave Kleikamp470decc2006-10-11 01:20:57 -0700174 */
175typedef struct journal_block_tag_s
176{
177 __be32 t_blocknr; /* The on-disk block number */
178 __be32 t_flags; /* See below */
Zach Brownb517bea2006-10-11 01:21:08 -0700179 __be32 t_blocknr_high; /* most-significant high 32bits. */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700180} journal_block_tag_t;
181
Mingming Caocd02ff02007-10-16 18:38:25 -0400182#define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
183#define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
Zach Brownb517bea2006-10-11 01:21:08 -0700184
Dave Kleikamp470decc2006-10-11 01:20:57 -0700185/*
186 * The revoke descriptor: used on disk to describe a series of blocks to
187 * be revoked from the log
188 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700189typedef struct jbd2_journal_revoke_header_s
Dave Kleikamp470decc2006-10-11 01:20:57 -0700190{
191 journal_header_t r_header;
192 __be32 r_count; /* Count of bytes used in the block */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700193} jbd2_journal_revoke_header_t;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700194
195
196/* Definitions for the journal tag flags word: */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700197#define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
198#define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
199#define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
200#define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700201
202
203/*
204 * The journal superblock. All fields are in big-endian byte order.
205 */
206typedef struct journal_superblock_s
207{
208/* 0x0000 */
209 journal_header_t s_header;
210
211/* 0x000C */
212 /* Static information describing the journal */
213 __be32 s_blocksize; /* journal device blocksize */
214 __be32 s_maxlen; /* total blocks in journal file */
215 __be32 s_first; /* first block of log information */
216
217/* 0x0018 */
218 /* Dynamic information describing the current state of the log */
219 __be32 s_sequence; /* first commit ID expected in log */
220 __be32 s_start; /* blocknr of start of log */
221
222/* 0x0020 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700223 /* Error value, as set by jbd2_journal_abort(). */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700224 __be32 s_errno;
225
226/* 0x0024 */
227 /* Remaining fields are only valid in a version-2 superblock */
228 __be32 s_feature_compat; /* compatible feature set */
229 __be32 s_feature_incompat; /* incompatible feature set */
230 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
231/* 0x0030 */
232 __u8 s_uuid[16]; /* 128-bit uuid for journal */
233
234/* 0x0040 */
235 __be32 s_nr_users; /* Nr of filesystems sharing log */
236
237 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
238
239/* 0x0048 */
240 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
241 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
242
243/* 0x0050 */
244 __u32 s_padding[44];
245
246/* 0x0100 */
247 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
248/* 0x0400 */
249} journal_superblock_t;
250
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700251#define JBD2_HAS_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700252 ((j)->j_format_version >= 2 && \
253 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700254#define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700255 ((j)->j_format_version >= 2 && \
256 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700257#define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700258 ((j)->j_format_version >= 2 && \
259 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
260
Girish Shilamkar818d2762008-01-28 23:58:27 -0500261#define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
262
263#define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
264#define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
265#define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
Dave Kleikamp470decc2006-10-11 01:20:57 -0700266
267/* Features known to this kernel version: */
Girish Shilamkar818d2762008-01-28 23:58:27 -0500268#define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700269#define JBD2_KNOWN_ROCOMPAT_FEATURES 0
Zach Brownb517bea2006-10-11 01:21:08 -0700270#define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
Girish Shilamkar818d2762008-01-28 23:58:27 -0500271 JBD2_FEATURE_INCOMPAT_64BIT | \
272 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700273
274#ifdef __KERNEL__
275
276#include <linux/fs.h>
277#include <linux/sched.h>
278
Chris Snook36df53f2008-01-28 23:58:27 -0500279#define J_ASSERT(assert) BUG_ON(!(assert))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700280
Dave Kleikamp470decc2006-10-11 01:20:57 -0700281#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
282#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700283
Mingming Caocd02ff02007-10-16 18:38:25 -0400284#if defined(JBD2_PARANOID_IOFAIL)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700285#define J_EXPECT(expr, why...) J_ASSERT(expr)
286#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
287#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
288#else
289#define __journal_expect(expr, why...) \
290 ({ \
291 int val = (expr); \
292 if (!val) { \
293 printk(KERN_ERR \
Theodore Ts'o08ec8c32009-01-16 11:57:00 -0500294 "JBD2 unexpected failure: %s: %s;\n", \
295 __func__, #expr); \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700296 printk(KERN_ERR why "\n"); \
297 } \
298 val; \
299 })
300#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
301#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
302#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
303#endif
304
305enum jbd_state_bits {
306 BH_JBD /* Has an attached ext3 journal_head */
307 = BH_PrivateStart,
308 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
309 BH_Freed, /* Has been freed (truncated) */
310 BH_Revoked, /* Has been revoked from the log */
311 BH_RevokeValid, /* Revoked flag is valid */
312 BH_JBDDirty, /* Is dirty but journaled */
313 BH_State, /* Pins most journal_head state */
314 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
315 BH_Unshadow, /* Dummy bit, for BJ_Shadow wakeup filtering */
Mark Fashehe97fcd92008-11-18 17:15:24 -0800316 BH_JBDPrivateStart, /* First bit available for private use by FS */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700317};
318
319BUFFER_FNS(JBD, jbd)
320BUFFER_FNS(JWrite, jwrite)
321BUFFER_FNS(JBDDirty, jbddirty)
322TAS_BUFFER_FNS(JBDDirty, jbddirty)
323BUFFER_FNS(Revoked, revoked)
324TAS_BUFFER_FNS(Revoked, revoked)
325BUFFER_FNS(RevokeValid, revokevalid)
326TAS_BUFFER_FNS(RevokeValid, revokevalid)
327BUFFER_FNS(Freed, freed)
328
329static inline struct buffer_head *jh2bh(struct journal_head *jh)
330{
331 return jh->b_bh;
332}
333
334static inline struct journal_head *bh2jh(struct buffer_head *bh)
335{
336 return bh->b_private;
337}
338
339static inline void jbd_lock_bh_state(struct buffer_head *bh)
340{
341 bit_spin_lock(BH_State, &bh->b_state);
342}
343
344static inline int jbd_trylock_bh_state(struct buffer_head *bh)
345{
346 return bit_spin_trylock(BH_State, &bh->b_state);
347}
348
349static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
350{
351 return bit_spin_is_locked(BH_State, &bh->b_state);
352}
353
354static inline void jbd_unlock_bh_state(struct buffer_head *bh)
355{
356 bit_spin_unlock(BH_State, &bh->b_state);
357}
358
359static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
360{
361 bit_spin_lock(BH_JournalHead, &bh->b_state);
362}
363
364static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
365{
366 bit_spin_unlock(BH_JournalHead, &bh->b_state);
367}
368
Jan Karac851ed52008-07-11 19:27:31 -0400369/* Flags in jbd_inode->i_flags */
370#define __JI_COMMIT_RUNNING 0
371/* Commit of the inode data in progress. We use this flag to protect us from
372 * concurrent deletion of inode. We cannot use reference to inode for this
373 * since we cannot afford doing last iput() on behalf of kjournald
374 */
375#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
376
377/**
378 * struct jbd_inode is the structure linking inodes in ordered mode
379 * present in a transaction so that we can sync them during commit.
380 */
381struct jbd2_inode {
382 /* Which transaction does this inode belong to? Either the running
383 * transaction or the committing one. [j_list_lock] */
384 transaction_t *i_transaction;
385
386 /* Pointer to the running transaction modifying inode's data in case
387 * there is already a committing transaction touching it. [j_list_lock] */
388 transaction_t *i_next_transaction;
389
390 /* List of inodes in the i_transaction [j_list_lock] */
391 struct list_head i_list;
392
393 /* VFS inode this inode belongs to [constant during the lifetime
394 * of the structure] */
395 struct inode *i_vfs_inode;
396
397 /* Flags of inode [j_list_lock] */
Brian King39e3ac22010-10-27 21:25:12 -0400398 unsigned long i_flags;
Jan Karac851ed52008-07-11 19:27:31 -0400399};
400
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700401struct jbd2_revoke_table_s;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700402
403/**
404 * struct handle_s - The handle_s type is the concrete type associated with
405 * handle_t.
406 * @h_transaction: Which compound transaction is this update a part of?
407 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
408 * @h_ref: Reference count on this handle
409 * @h_err: Field for caller's use to track errors through large fs operations
410 * @h_sync: flag for sync-on-close
411 * @h_jdata: flag to force data journaling
412 * @h_aborted: flag indicating fatal error on handle
413 **/
414
415/* Docbook can't yet cope with the bit fields, but will leave the documentation
416 * in so it can be fixed later.
417 */
418
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500419struct jbd2_journal_handle
Dave Kleikamp470decc2006-10-11 01:20:57 -0700420{
421 /* Which compound transaction is this update a part of? */
422 transaction_t *h_transaction;
423
424 /* Number of remaining buffers we are allowed to dirty: */
425 int h_buffer_credits;
426
427 /* Reference count on this handle */
428 int h_ref;
429
430 /* Field for caller's use to track errors through large fs */
431 /* operations */
432 int h_err;
433
434 /* Flags [no locking] */
Amir Goldstein93737452011-03-20 21:13:43 -0400435 unsigned int h_sync:1; /* sync-on-close */
436 unsigned int h_jdata:1; /* force data journaling */
437 unsigned int h_aborted:1; /* fatal error on handle */
438 unsigned int h_cowing:1; /* COWing block to snapshot */
439
440 /* Number of buffers requested by user:
441 * (before adding the COW credits factor) */
442 unsigned int h_base_credits:14;
443
444 /* Number of buffers the user is allowed to dirty:
445 * (counts only buffers dirtied when !h_cowing) */
446 unsigned int h_user_credits:14;
447
Mingming Cao7b751062008-01-28 23:58:27 -0500448
449#ifdef CONFIG_DEBUG_LOCK_ALLOC
450 struct lockdep_map h_lockdep_map;
451#endif
Amir Goldstein93737452011-03-20 21:13:43 -0400452
453#ifdef CONFIG_JBD2_DEBUG
454 /* COW debugging counters: */
455 unsigned int h_cow_moved; /* blocks moved to snapshot */
456 unsigned int h_cow_copied; /* blocks copied to snapshot */
457 unsigned int h_cow_ok_jh; /* blocks already COWed during current
458 transaction */
459 unsigned int h_cow_ok_bitmap; /* blocks not set in COW bitmap */
460 unsigned int h_cow_ok_mapped;/* blocks already mapped in snapshot */
461 unsigned int h_cow_bitmaps; /* COW bitmaps created */
462 unsigned int h_cow_excluded; /* blocks set in exclude bitmap */
463#endif
Dave Kleikamp470decc2006-10-11 01:20:57 -0700464};
465
466
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500467/*
468 * Some stats for checkpoint phase
469 */
470struct transaction_chp_stats_s {
471 unsigned long cs_chp_time;
Theodore Ts'obf699322009-09-30 00:32:06 -0400472 __u32 cs_forced_to_close;
473 __u32 cs_written;
474 __u32 cs_dropped;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500475};
476
Dave Kleikamp470decc2006-10-11 01:20:57 -0700477/* The transaction_t type is the guts of the journaling mechanism. It
478 * tracks a compound transaction through its various states:
479 *
480 * RUNNING: accepting new updates
481 * LOCKED: Updates still running but we don't accept new ones
482 * RUNDOWN: Updates are tidying up but have finished requesting
483 * new buffers to modify (state not used for now)
484 * FLUSH: All updates complete, but we are still writing to disk
485 * COMMIT: All data on disk, writing commit record
486 * FINISHED: We still have to keep the transaction for checkpointing.
487 *
488 * The transaction keeps track of all of the buffers modified by a
489 * running transaction, and all of the buffers committed but not yet
490 * flushed to home for finished transactions.
491 */
492
493/*
494 * Lock ranking:
495 *
496 * j_list_lock
497 * ->jbd_lock_bh_journal_head() (This is "innermost")
498 *
499 * j_state_lock
500 * ->jbd_lock_bh_state()
501 *
502 * jbd_lock_bh_state()
503 * ->j_list_lock
504 *
505 * j_state_lock
506 * ->t_handle_lock
507 *
508 * j_state_lock
509 * ->j_list_lock (journal_unmap_buffer)
510 *
511 */
512
513struct transaction_s
514{
515 /* Pointer to the journal for this transaction. [no locking] */
516 journal_t *t_journal;
517
518 /* Sequence number for this transaction [no locking] */
519 tid_t t_tid;
520
521 /*
522 * Transaction's current state
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700523 * [no locking - only kjournald2 alters this]
Jan Karaf5a7a6b2008-01-28 23:58:27 -0500524 * [j_list_lock] guards transition of a transaction into T_FINISHED
525 * state and subsequent call of __jbd2_journal_drop_transaction()
Dave Kleikamp470decc2006-10-11 01:20:57 -0700526 * FIXME: needs barriers
527 * KLUDGE: [use j_state_lock]
528 */
529 enum {
530 T_RUNNING,
531 T_LOCKED,
Dave Kleikamp470decc2006-10-11 01:20:57 -0700532 T_FLUSH,
533 T_COMMIT,
Jan Karabbd2be32011-05-24 11:59:18 -0400534 T_COMMIT_DFLUSH,
535 T_COMMIT_JFLUSH,
Dave Kleikamp470decc2006-10-11 01:20:57 -0700536 T_FINISHED
537 } t_state;
538
539 /*
540 * Where in the log does this transaction's commit start? [no locking]
541 */
542 unsigned long t_log_start;
543
544 /* Number of buffers on the t_buffers list [j_list_lock] */
545 int t_nr_buffers;
546
547 /*
548 * Doubly-linked circular list of all buffers reserved but not yet
549 * modified by this transaction [j_list_lock]
550 */
551 struct journal_head *t_reserved_list;
552
553 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700554 * Doubly-linked circular list of all metadata buffers owned by this
555 * transaction [j_list_lock]
556 */
557 struct journal_head *t_buffers;
558
559 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700560 * Doubly-linked circular list of all forget buffers (superseded
561 * buffers which we can un-checkpoint once this transaction commits)
562 * [j_list_lock]
563 */
564 struct journal_head *t_forget;
565
566 /*
567 * Doubly-linked circular list of all buffers still to be flushed before
568 * this transaction can be checkpointed. [j_list_lock]
569 */
570 struct journal_head *t_checkpoint_list;
571
572 /*
573 * Doubly-linked circular list of all buffers submitted for IO while
574 * checkpointing. [j_list_lock]
575 */
576 struct journal_head *t_checkpoint_io_list;
577
578 /*
579 * Doubly-linked circular list of temporary buffers currently undergoing
580 * IO in the log [j_list_lock]
581 */
582 struct journal_head *t_iobuf_list;
583
584 /*
585 * Doubly-linked circular list of metadata buffers being shadowed by log
586 * IO. The IO buffers on the iobuf list and the shadow buffers on this
587 * list match each other one for one at all times. [j_list_lock]
588 */
589 struct journal_head *t_shadow_list;
590
591 /*
592 * Doubly-linked circular list of control buffers being written to the
593 * log. [j_list_lock]
594 */
595 struct journal_head *t_log_list;
596
597 /*
Jan Karac851ed52008-07-11 19:27:31 -0400598 * List of inodes whose data we've modified in data=ordered mode.
599 * [j_list_lock]
600 */
601 struct list_head t_inode_list;
602
603 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700604 * Protects info related to handles
605 */
606 spinlock_t t_handle_lock;
607
608 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500609 * Longest time some handle had to wait for running transaction
610 */
611 unsigned long t_max_wait;
612
613 /*
614 * When transaction started
615 */
616 unsigned long t_start;
617
618 /*
619 * Checkpointing stats [j_checkpoint_sem]
620 */
621 struct transaction_chp_stats_s t_chp_stats;
622
623 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700624 * Number of outstanding updates running on this transaction
625 * [t_handle_lock]
626 */
Theodore Ts'oa51dca92010-08-02 08:43:25 -0400627 atomic_t t_updates;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700628
629 /*
630 * Number of buffers reserved for use by all handles in this transaction
631 * handle but not yet modified. [t_handle_lock]
632 */
Theodore Ts'oa51dca92010-08-02 08:43:25 -0400633 atomic_t t_outstanding_credits;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700634
635 /*
636 * Forward and backward links for the circular list of all transactions
637 * awaiting checkpoint. [j_list_lock]
638 */
639 transaction_t *t_cpnext, *t_cpprev;
640
641 /*
642 * When will the transaction expire (become due for commit), in jiffies?
643 * [no locking]
644 */
645 unsigned long t_expires;
646
647 /*
Josef Bacike07f7182008-11-26 01:14:26 -0500648 * When this transaction started, in nanoseconds [no locking]
649 */
650 ktime_t t_start_time;
651
652 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700653 * How many handles used this transaction? [t_handle_lock]
654 */
Theodore Ts'o8dd42042010-08-03 21:38:29 -0400655 atomic_t t_handle_count;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700656
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400657 /*
Theodore Ts'o70585482009-03-25 23:35:46 -0400658 * This transaction is being forced and some process is
659 * waiting for it to finish.
660 */
H Hartley Sweeten0ccff1a2009-08-17 22:38:04 -0400661 unsigned int t_synchronous_commit:1;
Jan Kara81be12c2011-05-24 11:52:40 -0400662
663 /* Disk flush needs to be sent to fs partition [no locking] */
664 int t_need_data_flush;
Theodore Ts'o70585482009-03-25 23:35:46 -0400665
666 /*
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400667 * For use by the filesystem to store fs-specific data
668 * structures associated with the transaction
669 */
670 struct list_head t_private_list;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700671};
672
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500673struct transaction_run_stats_s {
674 unsigned long rs_wait;
675 unsigned long rs_running;
676 unsigned long rs_locked;
677 unsigned long rs_flushing;
678 unsigned long rs_logging;
679
Theodore Ts'obf699322009-09-30 00:32:06 -0400680 __u32 rs_handle_count;
681 __u32 rs_blocks;
682 __u32 rs_blocks_logged;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500683};
684
685struct transaction_stats_s {
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500686 unsigned long ts_tid;
Theodore Ts'obf699322009-09-30 00:32:06 -0400687 struct transaction_run_stats_s run;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500688};
689
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500690static inline unsigned long
691jbd2_time_diff(unsigned long start, unsigned long end)
692{
693 if (end >= start)
694 return end - start;
695
696 return end + (MAX_JIFFY_OFFSET - start);
697}
698
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500699#define JBD2_NR_BATCH 64
700
Dave Kleikamp470decc2006-10-11 01:20:57 -0700701/**
702 * struct journal_s - The journal_s type is the concrete type associated with
703 * journal_t.
704 * @j_flags: General journaling state flags
705 * @j_errno: Is there an outstanding uncleared error on the journal (from a
706 * prior abort)?
707 * @j_sb_buffer: First part of superblock buffer
708 * @j_superblock: Second part of superblock buffer
709 * @j_format_version: Version of the superblock format
710 * @j_state_lock: Protect the various scalars in the journal
711 * @j_barrier_count: Number of processes waiting to create a barrier lock
712 * @j_barrier: The barrier lock itself
713 * @j_running_transaction: The current running transaction..
714 * @j_committing_transaction: the transaction we are pushing to disk
715 * @j_checkpoint_transactions: a linked circular list of all transactions
716 * waiting for checkpointing
717 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
718 * to start committing, or for a barrier lock to be released
719 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
720 * @j_wait_done_commit: Wait queue for waiting for commit to complete
721 * @j_wait_checkpoint: Wait queue to trigger checkpointing
722 * @j_wait_commit: Wait queue to trigger commit
723 * @j_wait_updates: Wait queue to wait for updates to complete
724 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
725 * @j_head: Journal head - identifies the first unused block in the journal
726 * @j_tail: Journal tail - identifies the oldest still-used block in the
727 * journal.
728 * @j_free: Journal free - how many free blocks are there in the journal?
729 * @j_first: The block number of the first usable block
730 * @j_last: The block number one beyond the last usable block
731 * @j_dev: Device where we store the journal
732 * @j_blocksize: blocksize for the location where we store the journal.
733 * @j_blk_offset: starting block offset for into the device where we store the
734 * journal
735 * @j_fs_dev: Device which holds the client fs. For internal journal this will
736 * be equal to j_dev
737 * @j_maxlen: Total maximum capacity of the journal region on disk.
738 * @j_list_lock: Protects the buffer lists and internal buffer state.
739 * @j_inode: Optional inode where we store the journal. If present, all journal
740 * block numbers are mapped into this inode via bmap().
741 * @j_tail_sequence: Sequence number of the oldest transaction in the log
742 * @j_transaction_sequence: Sequence number of the next transaction to grant
743 * @j_commit_sequence: Sequence number of the most recently committed
744 * transaction
745 * @j_commit_request: Sequence number of the most recent transaction wanting
746 * commit
747 * @j_uuid: Uuid of client object.
748 * @j_task: Pointer to the current commit thread for this journal
749 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
750 * single compound commit transaction
751 * @j_commit_interval: What is the maximum transaction lifetime before we begin
752 * a commit?
753 * @j_commit_timer: The timer used to wakeup the commit thread
754 * @j_revoke_lock: Protect the revoke table
755 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
756 * current transaction.
757 * @j_revoke_table: alternate revoke tables for j_revoke
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700758 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700759 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
760 * number that will fit in j_blocksize
761 * @j_last_sync_writer: most recent pid which did a synchronous write
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500762 * @j_history: Buffer storing the transactions statistics history
763 * @j_history_max: Maximum number of transactions in the statistics history
764 * @j_history_cur: Current number of transactions in the statistics history
765 * @j_history_lock: Protect the transactions statistics history
766 * @j_proc_entry: procfs entry for the jbd statistics directory
767 * @j_stats: Overall statistics
Dave Kleikamp470decc2006-10-11 01:20:57 -0700768 * @j_private: An opaque pointer to fs-private information.
769 */
770
771struct journal_s
772{
773 /* General journaling state flags [j_state_lock] */
774 unsigned long j_flags;
775
776 /*
777 * Is there an outstanding uncleared error on the journal (from a prior
778 * abort)? [j_state_lock]
779 */
780 int j_errno;
781
782 /* The superblock buffer */
783 struct buffer_head *j_sb_buffer;
784 journal_superblock_t *j_superblock;
785
786 /* Version of the superblock format */
787 int j_format_version;
788
789 /*
790 * Protect the various scalars in the journal
791 */
Theodore Ts'oa931da62010-08-03 21:35:12 -0400792 rwlock_t j_state_lock;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700793
794 /*
795 * Number of processes waiting to create a barrier lock [j_state_lock]
796 */
797 int j_barrier_count;
798
799 /* The barrier lock itself */
800 struct mutex j_barrier;
801
802 /*
803 * Transactions: The current running transaction...
804 * [j_state_lock] [caller holding open handle]
805 */
806 transaction_t *j_running_transaction;
807
808 /*
809 * the transaction we are pushing to disk
810 * [j_state_lock] [caller holding open handle]
811 */
812 transaction_t *j_committing_transaction;
813
814 /*
815 * ... and a linked circular list of all transactions waiting for
816 * checkpointing. [j_list_lock]
817 */
818 transaction_t *j_checkpoint_transactions;
819
820 /*
821 * Wait queue for waiting for a locked transaction to start committing,
822 * or for a barrier lock to be released
823 */
824 wait_queue_head_t j_wait_transaction_locked;
825
826 /* Wait queue for waiting for checkpointing to complete */
827 wait_queue_head_t j_wait_logspace;
828
829 /* Wait queue for waiting for commit to complete */
830 wait_queue_head_t j_wait_done_commit;
831
832 /* Wait queue to trigger checkpointing */
833 wait_queue_head_t j_wait_checkpoint;
834
835 /* Wait queue to trigger commit */
836 wait_queue_head_t j_wait_commit;
837
838 /* Wait queue to wait for updates to complete */
839 wait_queue_head_t j_wait_updates;
840
841 /* Semaphore for locking against concurrent checkpoints */
842 struct mutex j_checkpoint_mutex;
843
844 /*
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500845 * List of buffer heads used by the checkpoint routine. This
846 * was moved from jbd2_log_do_checkpoint() to reduce stack
847 * usage. Access to this array is controlled by the
848 * j_checkpoint_mutex. [j_checkpoint_mutex]
849 */
850 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
851
852 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700853 * Journal head: identifies the first unused block in the journal.
854 * [j_state_lock]
855 */
856 unsigned long j_head;
857
858 /*
859 * Journal tail: identifies the oldest still-used block in the journal.
860 * [j_state_lock]
861 */
862 unsigned long j_tail;
863
864 /*
865 * Journal free: how many free blocks are there in the journal?
866 * [j_state_lock]
867 */
868 unsigned long j_free;
869
870 /*
871 * Journal start and end: the block numbers of the first usable block
872 * and one beyond the last usable block in the journal. [j_state_lock]
873 */
874 unsigned long j_first;
875 unsigned long j_last;
876
877 /*
878 * Device, blocksize and starting block offset for the location where we
879 * store the journal.
880 */
881 struct block_device *j_dev;
882 int j_blocksize;
Theodore Ts'o05496762008-09-16 14:36:17 -0400883 unsigned long long j_blk_offset;
884 char j_devname[BDEVNAME_SIZE+24];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700885
886 /*
887 * Device which holds the client fs. For internal journal this will be
888 * equal to j_dev.
889 */
890 struct block_device *j_fs_dev;
891
892 /* Total maximum capacity of the journal region on disk. */
893 unsigned int j_maxlen;
894
895 /*
896 * Protects the buffer lists and internal buffer state.
897 */
898 spinlock_t j_list_lock;
899
900 /* Optional inode where we store the journal. If present, all */
901 /* journal block numbers are mapped into this inode via */
902 /* bmap(). */
903 struct inode *j_inode;
904
905 /*
906 * Sequence number of the oldest transaction in the log [j_state_lock]
907 */
908 tid_t j_tail_sequence;
909
910 /*
911 * Sequence number of the next transaction to grant [j_state_lock]
912 */
913 tid_t j_transaction_sequence;
914
915 /*
916 * Sequence number of the most recently committed transaction
917 * [j_state_lock].
918 */
919 tid_t j_commit_sequence;
920
921 /*
922 * Sequence number of the most recent transaction wanting commit
923 * [j_state_lock]
924 */
925 tid_t j_commit_request;
926
927 /*
928 * Journal uuid: identifies the object (filesystem, LVM volume etc)
929 * backed by this journal. This will eventually be replaced by an array
930 * of uuids, allowing us to index multiple devices within a single
931 * journal and to perform atomic updates across them.
932 */
933 __u8 j_uuid[16];
934
935 /* Pointer to the current commit thread for this journal */
936 struct task_struct *j_task;
937
938 /*
939 * Maximum number of metadata buffers to allow in a single compound
940 * commit transaction
941 */
942 int j_max_transaction_buffers;
943
944 /*
945 * What is the maximum transaction lifetime before we begin a commit?
946 */
947 unsigned long j_commit_interval;
948
949 /* The timer used to wakeup the commit thread: */
950 struct timer_list j_commit_timer;
951
952 /*
953 * The revoke table: maintains the list of revoked blocks in the
954 * current transaction. [j_revoke_lock]
955 */
956 spinlock_t j_revoke_lock;
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700957 struct jbd2_revoke_table_s *j_revoke;
958 struct jbd2_revoke_table_s *j_revoke_table[2];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700959
960 /*
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700961 * array of bhs for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700962 */
963 struct buffer_head **j_wbuf;
964 int j_wbufsize;
965
Josef Bacike07f7182008-11-26 01:14:26 -0500966 /*
967 * this is the pid of hte last person to run a synchronous operation
968 * through the journal
969 */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700970 pid_t j_last_sync_writer;
971
Josef Bacike07f7182008-11-26 01:14:26 -0500972 /*
973 * the average amount of time in nanoseconds it takes to commit a
974 * transaction to disk. [j_state_lock]
975 */
976 u64 j_average_commit_time;
977
Theodore Ts'o30773842009-01-03 20:27:38 -0500978 /*
979 * minimum and maximum times that we should wait for
980 * additional filesystem operations to get batched into a
981 * synchronous handle in microseconds
982 */
983 u32 j_min_batch_time;
984 u32 j_max_batch_time;
985
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400986 /* This function is called when a transaction is closed */
987 void (*j_commit_callback)(journal_t *,
988 transaction_t *);
989
Dave Kleikamp470decc2006-10-11 01:20:57 -0700990 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500991 * Journal statistics
992 */
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500993 spinlock_t j_history_lock;
994 struct proc_dir_entry *j_proc_entry;
995 struct transaction_stats_s j_stats;
996
Theodore Ts'o624080e2008-06-06 17:50:40 -0400997 /* Failed journal commit ID */
998 unsigned int j_failed_commit;
999
Johann Lombardi8e85fb32008-01-28 23:58:27 -05001000 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -07001001 * An opaque pointer to fs-private information. ext3 puts its
1002 * superblock pointer here
1003 */
1004 void *j_private;
1005};
1006
1007/*
1008 * Journal flag definitions
1009 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001010#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1011#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1012#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1013#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1014#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1015#define JBD2_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai5bf56832008-10-10 22:12:43 -04001016#define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1017 * data write error in ordered
1018 * mode */
Dave Kleikamp470decc2006-10-11 01:20:57 -07001019
1020/*
1021 * Function declarations for the journaling transaction and buffer
1022 * management
1023 */
1024
1025/* Filing buffers */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001026extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001027extern void __jbd2_journal_refile_buffer(struct journal_head *);
1028extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1029extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001030extern void __journal_free_buffer(struct journal_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001031extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001032extern void __journal_clean_data_list(transaction_t *transaction);
1033
1034/* Log buffer allocation */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001035extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001036int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001037
1038/* Commit management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001039extern void jbd2_journal_commit_transaction(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001040
1041/* Checkpoint list management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001042int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1043int __jbd2_journal_remove_checkpoint(struct journal_head *);
1044void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001045
Joel Beckere06c8222008-09-11 15:35:47 -07001046
1047/*
1048 * Triggers
1049 */
1050
1051struct jbd2_buffer_trigger_type {
1052 /*
Jan Kara13ceef02010-07-14 07:56:33 +02001053 * Fired a the moment data to write to the journal are known to be
1054 * stable - so either at the moment b_frozen_data is created or just
1055 * before a buffer is written to the journal. mapped_data is a mapped
1056 * buffer that is the frozen data for commit.
Joel Beckere06c8222008-09-11 15:35:47 -07001057 */
Jan Kara13ceef02010-07-14 07:56:33 +02001058 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
Joel Beckere06c8222008-09-11 15:35:47 -07001059 struct buffer_head *bh, void *mapped_data,
1060 size_t size);
1061
1062 /*
1063 * Fired during journal abort for dirty buffers that will not be
1064 * committed.
1065 */
1066 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1067 struct buffer_head *bh);
1068};
1069
Jan Kara13ceef02010-07-14 07:56:33 +02001070extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
Joel Beckere06c8222008-09-11 15:35:47 -07001071 void *mapped_data,
1072 struct jbd2_buffer_trigger_type *triggers);
1073extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1074 struct jbd2_buffer_trigger_type *triggers);
1075
Dave Kleikamp470decc2006-10-11 01:20:57 -07001076/* Buffer IO */
1077extern int
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001078jbd2_journal_write_metadata_buffer(transaction_t *transaction,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001079 struct journal_head *jh_in,
1080 struct journal_head **jh_out,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001081 unsigned long long blocknr);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001082
1083/* Transaction locking */
1084extern void __wait_on_journal (journal_t *);
1085
1086/*
1087 * Journal locking.
1088 *
1089 * We need to lock the journal during transaction state changes so that nobody
1090 * ever tries to take a handle on the running transaction while we are in the
1091 * middle of moving it to the commit phase. j_state_lock does this.
1092 *
1093 * Note that the locking is completely interrupt unsafe. We never touch
1094 * journal structures from interrupts.
1095 */
1096
1097static inline handle_t *journal_current_handle(void)
1098{
1099 return current->journal_info;
1100}
1101
1102/* The journaling code user interface:
1103 *
1104 * Create and destroy handles
1105 * Register buffer modifications against the current transaction.
1106 */
1107
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001108extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
Theodore Ts'o47def822010-07-27 11:56:05 -04001109extern handle_t *jbd2__journal_start(journal_t *, int nblocks, int gfp_mask);
1110extern int jbd2_journal_restart(handle_t *, int nblocks);
1111extern int jbd2__journal_restart(handle_t *, int nblocks, int gfp_mask);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001112extern int jbd2_journal_extend (handle_t *, int nblocks);
1113extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1114extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1115extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
Joel Beckere06c8222008-09-11 15:35:47 -07001116void jbd2_journal_set_triggers(struct buffer_head *,
1117 struct jbd2_buffer_trigger_type *type);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001118extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1119extern void jbd2_journal_release_buffer (handle_t *, struct buffer_head *);
1120extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001121extern void journal_sync_buffer (struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001122extern void jbd2_journal_invalidatepage(journal_t *,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001123 struct page *, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001124extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1125extern int jbd2_journal_stop(handle_t *);
1126extern int jbd2_journal_flush (journal_t *);
1127extern void jbd2_journal_lock_updates (journal_t *);
1128extern void jbd2_journal_unlock_updates (journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001129
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001130extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001131 struct block_device *fs_dev,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001132 unsigned long long start, int len, int bsize);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001133extern journal_t * jbd2_journal_init_inode (struct inode *);
1134extern int jbd2_journal_update_format (journal_t *);
1135extern int jbd2_journal_check_used_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001136 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001137extern int jbd2_journal_check_available_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001138 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001139extern int jbd2_journal_set_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001140 (journal_t *, unsigned long, unsigned long, unsigned long);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001141extern void jbd2_journal_clear_features
1142 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001143extern int jbd2_journal_load (journal_t *journal);
Hidehiro Kawai44519fa2008-10-10 20:29:13 -04001144extern int jbd2_journal_destroy (journal_t *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001145extern int jbd2_journal_recover (journal_t *journal);
1146extern int jbd2_journal_wipe (journal_t *, int);
1147extern int jbd2_journal_skip_recovery (journal_t *);
1148extern void jbd2_journal_update_superblock (journal_t *, int);
1149extern void __jbd2_journal_abort_hard (journal_t *);
1150extern void jbd2_journal_abort (journal_t *, int);
1151extern int jbd2_journal_errno (journal_t *);
1152extern void jbd2_journal_ack_err (journal_t *);
1153extern int jbd2_journal_clear_err (journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001154extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001155extern int jbd2_journal_force_commit(journal_t *);
Jan Karac851ed52008-07-11 19:27:31 -04001156extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
Jan Kara7f5aa212009-02-10 11:15:34 -05001157extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1158 struct jbd2_inode *inode, loff_t new_size);
Jan Karac851ed52008-07-11 19:27:31 -04001159extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1160extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001161
1162/*
1163 * journal_head management
1164 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001165struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1166struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001167void jbd2_journal_put_journal_head(struct journal_head *jh);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001168
1169/*
1170 * handle management
1171 */
Christoph Lametere18b8902006-12-06 20:33:20 -08001172extern struct kmem_cache *jbd2_handle_cache;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001173
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001174static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001175{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001176 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001177}
1178
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001179static inline void jbd2_free_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001180{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001181 kmem_cache_free(jbd2_handle_cache, handle);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001182}
1183
Theodore Ts'o8aefcd52011-01-10 12:29:43 -05001184/*
1185 * jbd2_inode management (optional, for those file systems that want to use
1186 * dynamically allocated jbd2_inode structures)
1187 */
1188extern struct kmem_cache *jbd2_inode_cache;
1189
1190static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1191{
1192 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1193}
1194
1195static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1196{
1197 kmem_cache_free(jbd2_inode_cache, jinode);
1198}
1199
Dave Kleikamp470decc2006-10-11 01:20:57 -07001200/* Primary revoke support */
1201#define JOURNAL_REVOKE_DEFAULT_HASH 256
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001202extern int jbd2_journal_init_revoke(journal_t *, int);
1203extern void jbd2_journal_destroy_revoke_caches(void);
1204extern int jbd2_journal_init_revoke_caches(void);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001205
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001206extern void jbd2_journal_destroy_revoke(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001207extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001208extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o67c457a2009-04-14 07:50:56 -04001209extern void jbd2_journal_write_revoke_records(journal_t *,
1210 transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001211
1212/* Recovery revoke support */
Mingming Cao18eba7a2006-10-11 01:21:13 -07001213extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1214extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001215extern void jbd2_journal_clear_revoke(journal_t *);
1216extern void jbd2_journal_switch_revoke_table(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001217
1218/*
1219 * The log thread user interface:
1220 *
1221 * Request space in the current transaction, and force transaction commit
1222 * transitions on demand.
1223 */
1224
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001225int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1226int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1227int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1228int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1229int jbd2_journal_force_commit_nested(journal_t *journal);
1230int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1231int jbd2_log_do_checkpoint(journal_t *journal);
Jan Karabbd2be32011-05-24 11:59:18 -04001232int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001233
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001234void __jbd2_log_wait_for_space(journal_t *journal);
Aneesh Kumar K.Vfb684072008-11-06 17:50:21 -05001235extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1236extern int jbd2_cleanup_journal_tail(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001237
1238/* Debugging code only: */
1239
1240#define jbd_ENOSYS() \
1241do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -07001242 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Dave Kleikamp470decc2006-10-11 01:20:57 -07001243 current->state = TASK_UNINTERRUPTIBLE; \
1244 schedule(); \
1245} while (1)
1246
1247/*
1248 * is_journal_abort
1249 *
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001250 * Simple test wrapper function to test the JBD2_ABORT state flag. This
Dave Kleikamp470decc2006-10-11 01:20:57 -07001251 * bit, when set, indicates that we have had a fatal error somewhere,
1252 * either inside the journaling layer or indicated to us by the client
1253 * (eg. ext3), and that we and should not commit any further
1254 * transactions.
1255 */
1256
1257static inline int is_journal_aborted(journal_t *journal)
1258{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001259 return journal->j_flags & JBD2_ABORT;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001260}
1261
1262static inline int is_handle_aborted(handle_t *handle)
1263{
1264 if (handle->h_aborted)
1265 return 1;
1266 return is_journal_aborted(handle->h_transaction->t_journal);
1267}
1268
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001269static inline void jbd2_journal_abort_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001270{
1271 handle->h_aborted = 1;
1272}
1273
1274#endif /* __KERNEL__ */
1275
1276/* Comparison functions for transaction IDs: perform comparisons using
1277 * modulo arithmetic so that they work over sequence number wraps. */
1278
1279static inline int tid_gt(tid_t x, tid_t y)
1280{
1281 int difference = (x - y);
1282 return (difference > 0);
1283}
1284
1285static inline int tid_geq(tid_t x, tid_t y)
1286{
1287 int difference = (x - y);
1288 return (difference >= 0);
1289}
1290
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001291extern int jbd2_journal_blocks_per_page(struct inode *inode);
Zach Brownb517bea2006-10-11 01:21:08 -07001292extern size_t journal_tag_bytes(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001293
1294/*
1295 * Return the minimum number of blocks which must be free in the journal
1296 * before a new transaction may be started. Must be called under j_state_lock.
1297 */
1298static inline int jbd_space_needed(journal_t *journal)
1299{
1300 int nblocks = journal->j_max_transaction_buffers;
1301 if (journal->j_committing_transaction)
Theodore Ts'oa51dca92010-08-02 08:43:25 -04001302 nblocks += atomic_read(&journal->j_committing_transaction->
1303 t_outstanding_credits);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001304 return nblocks;
1305}
1306
1307/*
1308 * Definitions which augment the buffer_head layer
1309 */
1310
1311/* journaling buffer types */
1312#define BJ_None 0 /* Not journaled */
Jan Kara87c89c22008-07-11 19:27:31 -04001313#define BJ_Metadata 1 /* Normal journaled metadata */
1314#define BJ_Forget 2 /* Buffer superseded by this transaction */
1315#define BJ_IO 3 /* Buffer is for temporary IO use */
1316#define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1317#define BJ_LogCtl 5 /* Buffer contains log descriptors */
1318#define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1319#define BJ_Types 7
Dave Kleikamp470decc2006-10-11 01:20:57 -07001320
1321extern int jbd_blocks_per_page(struct inode *inode);
1322
1323#ifdef __KERNEL__
1324
1325#define buffer_trace_init(bh) do {} while (0)
1326#define print_buffer_fields(bh) do {} while (0)
1327#define print_buffer_trace(bh) do {} while (0)
1328#define BUFFER_TRACE(bh, info) do {} while (0)
1329#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1330#define JBUFFER_TRACE(jh, info) do {} while (0)
1331
1332#endif /* __KERNEL__ */
1333
Mingming Caocd02ff02007-10-16 18:38:25 -04001334#endif /* _LINUX_JBD2_H */