blob: 50e5a5e6a71214535b8cce3d823862da4f7b930f [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#else
24
25#include <linux/types.h>
26#include <linux/buffer_head.h>
27#include <linux/journal-head.h>
28#include <linux/stddef.h>
29#include <linux/bit_spinlock.h>
30#include <linux/mutex.h>
31#include <linux/timer.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090032#include <linux/slab.h>
Darrick J. Wong01b5adc2012-05-27 07:50:56 -040033#include <crypto/hash.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
Theodore Ts'ob6e96d02013-02-09 16:29:20 -050059extern ushort 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
Darrick J. Wong8f888ef2012-05-22 22:43:41 -0400150#define JBD2_CRC32C_CHKSUM 4
Girish Shilamkar818d2762008-01-28 23:58:27 -0500151
152#define JBD2_CRC32_CHKSUM_SIZE 4
153
154#define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
155/*
156 * Commit block header for storing transactional checksums:
Darrick J. Wong8f888ef2012-05-22 22:43:41 -0400157 *
158 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
159 * fields are used to store a checksum of the descriptor and data blocks.
160 *
161 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
162 * field is used to store crc32c(uuid+commit_block). Each journal metadata
163 * block gets its own checksum, and data block checksums are stored in
164 * journal_block_tag (in the descriptor). The other h_chksum* fields are
165 * not used.
166 *
167 * Checksum v1 and v2 are mutually exclusive features.
Girish Shilamkar818d2762008-01-28 23:58:27 -0500168 */
169struct commit_header {
170 __be32 h_magic;
171 __be32 h_blocktype;
172 __be32 h_sequence;
173 unsigned char h_chksum_type;
174 unsigned char h_chksum_size;
175 unsigned char h_padding[2];
176 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
Theodore Ts'o736603a2008-07-11 19:27:31 -0400177 __be64 h_commit_sec;
178 __be32 h_commit_nsec;
Girish Shilamkar818d2762008-01-28 23:58:27 -0500179};
Dave Kleikamp470decc2006-10-11 01:20:57 -0700180
181/*
Zach Brownb517bea2006-10-11 01:21:08 -0700182 * The block tag: used to describe a single buffer in the journal.
183 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
184 * raw struct shouldn't be used for pointer math or sizeof() - use
185 * journal_tag_bytes(journal) instead to compute this.
Dave Kleikamp470decc2006-10-11 01:20:57 -0700186 */
187typedef struct journal_block_tag_s
188{
189 __be32 t_blocknr; /* The on-disk block number */
Darrick J. Wong8f888ef2012-05-22 22:43:41 -0400190 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
191 __be16 t_flags; /* See below */
Zach Brownb517bea2006-10-11 01:21:08 -0700192 __be32 t_blocknr_high; /* most-significant high 32bits. */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700193} journal_block_tag_t;
194
Mingming Caocd02ff02007-10-16 18:38:25 -0400195#define JBD2_TAG_SIZE32 (offsetof(journal_block_tag_t, t_blocknr_high))
196#define JBD2_TAG_SIZE64 (sizeof(journal_block_tag_t))
Zach Brownb517bea2006-10-11 01:21:08 -0700197
Darrick J. Wong8f888ef2012-05-22 22:43:41 -0400198/* Tail of descriptor block, for checksumming */
199struct jbd2_journal_block_tail {
200 __be32 t_checksum; /* crc32c(uuid+descr_block) */
201};
202
Dave Kleikamp470decc2006-10-11 01:20:57 -0700203/*
204 * The revoke descriptor: used on disk to describe a series of blocks to
205 * be revoked from the log
206 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700207typedef struct jbd2_journal_revoke_header_s
Dave Kleikamp470decc2006-10-11 01:20:57 -0700208{
209 journal_header_t r_header;
210 __be32 r_count; /* Count of bytes used in the block */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700211} jbd2_journal_revoke_header_t;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700212
Darrick J. Wong8f888ef2012-05-22 22:43:41 -0400213/* Tail of revoke block, for checksumming */
214struct jbd2_journal_revoke_tail {
215 __be32 r_checksum; /* crc32c(uuid+revoke_block) */
216};
Dave Kleikamp470decc2006-10-11 01:20:57 -0700217
218/* Definitions for the journal tag flags word: */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700219#define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
220#define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
221#define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
222#define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700223
224
225/*
226 * The journal superblock. All fields are in big-endian byte order.
227 */
228typedef struct journal_superblock_s
229{
230/* 0x0000 */
231 journal_header_t s_header;
232
233/* 0x000C */
234 /* Static information describing the journal */
235 __be32 s_blocksize; /* journal device blocksize */
236 __be32 s_maxlen; /* total blocks in journal file */
237 __be32 s_first; /* first block of log information */
238
239/* 0x0018 */
240 /* Dynamic information describing the current state of the log */
241 __be32 s_sequence; /* first commit ID expected in log */
242 __be32 s_start; /* blocknr of start of log */
243
244/* 0x0020 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700245 /* Error value, as set by jbd2_journal_abort(). */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700246 __be32 s_errno;
247
248/* 0x0024 */
249 /* Remaining fields are only valid in a version-2 superblock */
250 __be32 s_feature_compat; /* compatible feature set */
251 __be32 s_feature_incompat; /* incompatible feature set */
252 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
253/* 0x0030 */
254 __u8 s_uuid[16]; /* 128-bit uuid for journal */
255
256/* 0x0040 */
257 __be32 s_nr_users; /* Nr of filesystems sharing log */
258
259 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
260
261/* 0x0048 */
262 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
263 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
264
265/* 0x0050 */
Darrick J. Wong8f888ef2012-05-22 22:43:41 -0400266 __u8 s_checksum_type; /* checksum type */
267 __u8 s_padding2[3];
268 __u32 s_padding[42];
269 __be32 s_checksum; /* crc32c(superblock) */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700270
271/* 0x0100 */
272 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
273/* 0x0400 */
274} journal_superblock_t;
275
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700276#define JBD2_HAS_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700277 ((j)->j_format_version >= 2 && \
278 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700279#define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700280 ((j)->j_format_version >= 2 && \
281 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700282#define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700283 ((j)->j_format_version >= 2 && \
284 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
285
Girish Shilamkar818d2762008-01-28 23:58:27 -0500286#define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
287
288#define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
289#define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
290#define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
Darrick J. Wong8f888ef2012-05-22 22:43:41 -0400291#define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
Dave Kleikamp470decc2006-10-11 01:20:57 -0700292
293/* Features known to this kernel version: */
Girish Shilamkar818d2762008-01-28 23:58:27 -0500294#define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700295#define JBD2_KNOWN_ROCOMPAT_FEATURES 0
Zach Brownb517bea2006-10-11 01:21:08 -0700296#define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
Girish Shilamkar818d2762008-01-28 23:58:27 -0500297 JBD2_FEATURE_INCOMPAT_64BIT | \
Darrick J. Wonge93376c2012-05-27 08:12:42 -0400298 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
299 JBD2_FEATURE_INCOMPAT_CSUM_V2)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700300
301#ifdef __KERNEL__
302
303#include <linux/fs.h>
304#include <linux/sched.h>
Thomas Gleixner44606672011-10-27 04:38:18 -0400305#include <linux/jbd_common.h>
Dave Kleikamp470decc2006-10-11 01:20:57 -0700306
Chris Snook36df53f2008-01-28 23:58:27 -0500307#define J_ASSERT(assert) BUG_ON(!(assert))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700308
Dave Kleikamp470decc2006-10-11 01:20:57 -0700309#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
310#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700311
Mingming Caocd02ff02007-10-16 18:38:25 -0400312#if defined(JBD2_PARANOID_IOFAIL)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700313#define J_EXPECT(expr, why...) J_ASSERT(expr)
314#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
315#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
316#else
317#define __journal_expect(expr, why...) \
318 ({ \
319 int val = (expr); \
320 if (!val) { \
321 printk(KERN_ERR \
Theodore Ts'o08ec8c32009-01-16 11:57:00 -0500322 "JBD2 unexpected failure: %s: %s;\n", \
323 __func__, #expr); \
Dave Kleikamp470decc2006-10-11 01:20:57 -0700324 printk(KERN_ERR why "\n"); \
325 } \
326 val; \
327 })
328#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
329#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
330#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
331#endif
332
Jan Karac851ed52008-07-11 19:27:31 -0400333/* Flags in jbd_inode->i_flags */
334#define __JI_COMMIT_RUNNING 0
335/* Commit of the inode data in progress. We use this flag to protect us from
336 * concurrent deletion of inode. We cannot use reference to inode for this
337 * since we cannot afford doing last iput() on behalf of kjournald
338 */
339#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
340
341/**
342 * struct jbd_inode is the structure linking inodes in ordered mode
343 * present in a transaction so that we can sync them during commit.
344 */
345struct jbd2_inode {
346 /* Which transaction does this inode belong to? Either the running
347 * transaction or the committing one. [j_list_lock] */
348 transaction_t *i_transaction;
349
350 /* Pointer to the running transaction modifying inode's data in case
351 * there is already a committing transaction touching it. [j_list_lock] */
352 transaction_t *i_next_transaction;
353
354 /* List of inodes in the i_transaction [j_list_lock] */
355 struct list_head i_list;
356
357 /* VFS inode this inode belongs to [constant during the lifetime
358 * of the structure] */
359 struct inode *i_vfs_inode;
360
361 /* Flags of inode [j_list_lock] */
Brian King39e3ac22010-10-27 21:25:12 -0400362 unsigned long i_flags;
Jan Karac851ed52008-07-11 19:27:31 -0400363};
364
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700365struct jbd2_revoke_table_s;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700366
367/**
368 * struct handle_s - The handle_s type is the concrete type associated with
369 * handle_t.
370 * @h_transaction: Which compound transaction is this update a part of?
371 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
372 * @h_ref: Reference count on this handle
373 * @h_err: Field for caller's use to track errors through large fs operations
374 * @h_sync: flag for sync-on-close
375 * @h_jdata: flag to force data journaling
376 * @h_aborted: flag indicating fatal error on handle
377 **/
378
379/* Docbook can't yet cope with the bit fields, but will leave the documentation
380 * in so it can be fixed later.
381 */
382
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500383struct jbd2_journal_handle
Dave Kleikamp470decc2006-10-11 01:20:57 -0700384{
385 /* Which compound transaction is this update a part of? */
386 transaction_t *h_transaction;
387
388 /* Number of remaining buffers we are allowed to dirty: */
389 int h_buffer_credits;
390
391 /* Reference count on this handle */
392 int h_ref;
393
394 /* Field for caller's use to track errors through large fs */
395 /* operations */
396 int h_err;
397
398 /* Flags [no locking] */
Theodore Ts'o078d5032013-02-07 00:02:15 -0500399 unsigned int h_sync: 1; /* sync-on-close */
400 unsigned int h_jdata: 1; /* force data journaling */
401 unsigned int h_aborted: 1; /* fatal error on handle */
Theodore Ts'o343d9c22013-02-08 13:00:22 -0500402 unsigned int h_type: 8; /* for handle statistics */
403 unsigned int h_line_no: 16; /* for handle statistics */
404
405 unsigned long h_start_jiffies;
406 unsigned int h_requested_credits;
Mingming Cao7b751062008-01-28 23:58:27 -0500407
408#ifdef CONFIG_DEBUG_LOCK_ALLOC
409 struct lockdep_map h_lockdep_map;
410#endif
Dave Kleikamp470decc2006-10-11 01:20:57 -0700411};
412
413
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500414/*
415 * Some stats for checkpoint phase
416 */
417struct transaction_chp_stats_s {
418 unsigned long cs_chp_time;
Theodore Ts'obf699322009-09-30 00:32:06 -0400419 __u32 cs_forced_to_close;
420 __u32 cs_written;
421 __u32 cs_dropped;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500422};
423
Dave Kleikamp470decc2006-10-11 01:20:57 -0700424/* The transaction_t type is the guts of the journaling mechanism. It
425 * tracks a compound transaction through its various states:
426 *
427 * RUNNING: accepting new updates
428 * LOCKED: Updates still running but we don't accept new ones
429 * RUNDOWN: Updates are tidying up but have finished requesting
430 * new buffers to modify (state not used for now)
431 * FLUSH: All updates complete, but we are still writing to disk
432 * COMMIT: All data on disk, writing commit record
433 * FINISHED: We still have to keep the transaction for checkpointing.
434 *
435 * The transaction keeps track of all of the buffers modified by a
436 * running transaction, and all of the buffers committed but not yet
437 * flushed to home for finished transactions.
438 */
439
440/*
441 * Lock ranking:
442 *
443 * j_list_lock
444 * ->jbd_lock_bh_journal_head() (This is "innermost")
445 *
446 * j_state_lock
447 * ->jbd_lock_bh_state()
448 *
449 * jbd_lock_bh_state()
450 * ->j_list_lock
451 *
452 * j_state_lock
453 * ->t_handle_lock
454 *
455 * j_state_lock
456 * ->j_list_lock (journal_unmap_buffer)
457 *
458 */
459
460struct transaction_s
461{
462 /* Pointer to the journal for this transaction. [no locking] */
463 journal_t *t_journal;
464
465 /* Sequence number for this transaction [no locking] */
466 tid_t t_tid;
467
468 /*
469 * Transaction's current state
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700470 * [no locking - only kjournald2 alters this]
Jan Karaf5a7a6b2008-01-28 23:58:27 -0500471 * [j_list_lock] guards transition of a transaction into T_FINISHED
472 * state and subsequent call of __jbd2_journal_drop_transaction()
Dave Kleikamp470decc2006-10-11 01:20:57 -0700473 * FIXME: needs barriers
474 * KLUDGE: [use j_state_lock]
475 */
476 enum {
477 T_RUNNING,
478 T_LOCKED,
Dave Kleikamp470decc2006-10-11 01:20:57 -0700479 T_FLUSH,
480 T_COMMIT,
Jan Karabbd2be32011-05-24 11:59:18 -0400481 T_COMMIT_DFLUSH,
482 T_COMMIT_JFLUSH,
Dave Kleikamp470decc2006-10-11 01:20:57 -0700483 T_FINISHED
484 } t_state;
485
486 /*
487 * Where in the log does this transaction's commit start? [no locking]
488 */
489 unsigned long t_log_start;
490
491 /* Number of buffers on the t_buffers list [j_list_lock] */
492 int t_nr_buffers;
493
494 /*
495 * Doubly-linked circular list of all buffers reserved but not yet
496 * modified by this transaction [j_list_lock]
497 */
498 struct journal_head *t_reserved_list;
499
500 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700501 * Doubly-linked circular list of all metadata buffers owned by this
502 * transaction [j_list_lock]
503 */
504 struct journal_head *t_buffers;
505
506 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700507 * Doubly-linked circular list of all forget buffers (superseded
508 * buffers which we can un-checkpoint once this transaction commits)
509 * [j_list_lock]
510 */
511 struct journal_head *t_forget;
512
513 /*
514 * Doubly-linked circular list of all buffers still to be flushed before
515 * this transaction can be checkpointed. [j_list_lock]
516 */
517 struct journal_head *t_checkpoint_list;
518
519 /*
520 * Doubly-linked circular list of all buffers submitted for IO while
521 * checkpointing. [j_list_lock]
522 */
523 struct journal_head *t_checkpoint_io_list;
524
525 /*
526 * Doubly-linked circular list of temporary buffers currently undergoing
527 * IO in the log [j_list_lock]
528 */
529 struct journal_head *t_iobuf_list;
530
531 /*
532 * Doubly-linked circular list of metadata buffers being shadowed by log
533 * IO. The IO buffers on the iobuf list and the shadow buffers on this
534 * list match each other one for one at all times. [j_list_lock]
535 */
536 struct journal_head *t_shadow_list;
537
538 /*
539 * Doubly-linked circular list of control buffers being written to the
540 * log. [j_list_lock]
541 */
542 struct journal_head *t_log_list;
543
544 /*
Jan Karac851ed52008-07-11 19:27:31 -0400545 * List of inodes whose data we've modified in data=ordered mode.
546 * [j_list_lock]
547 */
548 struct list_head t_inode_list;
549
550 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700551 * Protects info related to handles
552 */
553 spinlock_t t_handle_lock;
554
555 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500556 * Longest time some handle had to wait for running transaction
557 */
558 unsigned long t_max_wait;
559
560 /*
561 * When transaction started
562 */
563 unsigned long t_start;
564
565 /*
Theodore Ts'o9fff24a2013-02-06 22:30:23 -0500566 * When commit was requested
567 */
568 unsigned long t_requested;
569
570 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500571 * Checkpointing stats [j_checkpoint_sem]
572 */
573 struct transaction_chp_stats_s t_chp_stats;
574
575 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700576 * Number of outstanding updates running on this transaction
577 * [t_handle_lock]
578 */
Theodore Ts'oa51dca92010-08-02 08:43:25 -0400579 atomic_t t_updates;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700580
581 /*
582 * Number of buffers reserved for use by all handles in this transaction
583 * handle but not yet modified. [t_handle_lock]
584 */
Theodore Ts'oa51dca92010-08-02 08:43:25 -0400585 atomic_t t_outstanding_credits;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700586
587 /*
588 * Forward and backward links for the circular list of all transactions
589 * awaiting checkpoint. [j_list_lock]
590 */
591 transaction_t *t_cpnext, *t_cpprev;
592
593 /*
594 * When will the transaction expire (become due for commit), in jiffies?
595 * [no locking]
596 */
597 unsigned long t_expires;
598
599 /*
Josef Bacike07f7182008-11-26 01:14:26 -0500600 * When this transaction started, in nanoseconds [no locking]
601 */
602 ktime_t t_start_time;
603
604 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700605 * How many handles used this transaction? [t_handle_lock]
606 */
Theodore Ts'o8dd42042010-08-03 21:38:29 -0400607 atomic_t t_handle_count;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700608
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400609 /*
Theodore Ts'o70585482009-03-25 23:35:46 -0400610 * This transaction is being forced and some process is
611 * waiting for it to finish.
612 */
H Hartley Sweeten0ccff1a2009-08-17 22:38:04 -0400613 unsigned int t_synchronous_commit:1;
Jan Kara81be12c2011-05-24 11:52:40 -0400614
615 /* Disk flush needs to be sent to fs partition [no locking] */
616 int t_need_data_flush;
Theodore Ts'o70585482009-03-25 23:35:46 -0400617
618 /*
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400619 * For use by the filesystem to store fs-specific data
620 * structures associated with the transaction
621 */
622 struct list_head t_private_list;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700623};
624
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500625struct transaction_run_stats_s {
626 unsigned long rs_wait;
Theodore Ts'o9fff24a2013-02-06 22:30:23 -0500627 unsigned long rs_request_delay;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500628 unsigned long rs_running;
629 unsigned long rs_locked;
630 unsigned long rs_flushing;
631 unsigned long rs_logging;
632
Theodore Ts'obf699322009-09-30 00:32:06 -0400633 __u32 rs_handle_count;
634 __u32 rs_blocks;
635 __u32 rs_blocks_logged;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500636};
637
638struct transaction_stats_s {
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500639 unsigned long ts_tid;
Theodore Ts'o9fff24a2013-02-06 22:30:23 -0500640 unsigned long ts_requested;
Theodore Ts'obf699322009-09-30 00:32:06 -0400641 struct transaction_run_stats_s run;
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500642};
643
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500644static inline unsigned long
645jbd2_time_diff(unsigned long start, unsigned long end)
646{
647 if (end >= start)
648 return end - start;
649
650 return end + (MAX_JIFFY_OFFSET - start);
651}
652
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500653#define JBD2_NR_BATCH 64
654
Dave Kleikamp470decc2006-10-11 01:20:57 -0700655/**
656 * struct journal_s - The journal_s type is the concrete type associated with
657 * journal_t.
658 * @j_flags: General journaling state flags
659 * @j_errno: Is there an outstanding uncleared error on the journal (from a
660 * prior abort)?
661 * @j_sb_buffer: First part of superblock buffer
662 * @j_superblock: Second part of superblock buffer
663 * @j_format_version: Version of the superblock format
664 * @j_state_lock: Protect the various scalars in the journal
665 * @j_barrier_count: Number of processes waiting to create a barrier lock
666 * @j_barrier: The barrier lock itself
667 * @j_running_transaction: The current running transaction..
668 * @j_committing_transaction: the transaction we are pushing to disk
669 * @j_checkpoint_transactions: a linked circular list of all transactions
670 * waiting for checkpointing
671 * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
672 * to start committing, or for a barrier lock to be released
673 * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
674 * @j_wait_done_commit: Wait queue for waiting for commit to complete
675 * @j_wait_checkpoint: Wait queue to trigger checkpointing
676 * @j_wait_commit: Wait queue to trigger commit
677 * @j_wait_updates: Wait queue to wait for updates to complete
678 * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
679 * @j_head: Journal head - identifies the first unused block in the journal
680 * @j_tail: Journal tail - identifies the oldest still-used block in the
681 * journal.
682 * @j_free: Journal free - how many free blocks are there in the journal?
683 * @j_first: The block number of the first usable block
684 * @j_last: The block number one beyond the last usable block
685 * @j_dev: Device where we store the journal
686 * @j_blocksize: blocksize for the location where we store the journal.
687 * @j_blk_offset: starting block offset for into the device where we store the
688 * journal
689 * @j_fs_dev: Device which holds the client fs. For internal journal this will
690 * be equal to j_dev
691 * @j_maxlen: Total maximum capacity of the journal region on disk.
692 * @j_list_lock: Protects the buffer lists and internal buffer state.
693 * @j_inode: Optional inode where we store the journal. If present, all journal
694 * block numbers are mapped into this inode via bmap().
695 * @j_tail_sequence: Sequence number of the oldest transaction in the log
696 * @j_transaction_sequence: Sequence number of the next transaction to grant
697 * @j_commit_sequence: Sequence number of the most recently committed
698 * transaction
699 * @j_commit_request: Sequence number of the most recent transaction wanting
700 * commit
701 * @j_uuid: Uuid of client object.
702 * @j_task: Pointer to the current commit thread for this journal
703 * @j_max_transaction_buffers: Maximum number of metadata buffers to allow in a
704 * single compound commit transaction
705 * @j_commit_interval: What is the maximum transaction lifetime before we begin
706 * a commit?
707 * @j_commit_timer: The timer used to wakeup the commit thread
708 * @j_revoke_lock: Protect the revoke table
709 * @j_revoke: The revoke table - maintains the list of revoked blocks in the
710 * current transaction.
711 * @j_revoke_table: alternate revoke tables for j_revoke
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700712 * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700713 * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
714 * number that will fit in j_blocksize
715 * @j_last_sync_writer: most recent pid which did a synchronous write
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500716 * @j_history: Buffer storing the transactions statistics history
717 * @j_history_max: Maximum number of transactions in the statistics history
718 * @j_history_cur: Current number of transactions in the statistics history
719 * @j_history_lock: Protect the transactions statistics history
720 * @j_proc_entry: procfs entry for the jbd statistics directory
721 * @j_stats: Overall statistics
Dave Kleikamp470decc2006-10-11 01:20:57 -0700722 * @j_private: An opaque pointer to fs-private information.
723 */
724
725struct journal_s
726{
727 /* General journaling state flags [j_state_lock] */
728 unsigned long j_flags;
729
730 /*
731 * Is there an outstanding uncleared error on the journal (from a prior
732 * abort)? [j_state_lock]
733 */
734 int j_errno;
735
736 /* The superblock buffer */
737 struct buffer_head *j_sb_buffer;
738 journal_superblock_t *j_superblock;
739
740 /* Version of the superblock format */
741 int j_format_version;
742
743 /*
744 * Protect the various scalars in the journal
745 */
Theodore Ts'oa931da62010-08-03 21:35:12 -0400746 rwlock_t j_state_lock;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700747
748 /*
749 * Number of processes waiting to create a barrier lock [j_state_lock]
750 */
751 int j_barrier_count;
752
753 /* The barrier lock itself */
754 struct mutex j_barrier;
755
756 /*
757 * Transactions: The current running transaction...
758 * [j_state_lock] [caller holding open handle]
759 */
760 transaction_t *j_running_transaction;
761
762 /*
763 * the transaction we are pushing to disk
764 * [j_state_lock] [caller holding open handle]
765 */
766 transaction_t *j_committing_transaction;
767
768 /*
769 * ... and a linked circular list of all transactions waiting for
770 * checkpointing. [j_list_lock]
771 */
772 transaction_t *j_checkpoint_transactions;
773
774 /*
775 * Wait queue for waiting for a locked transaction to start committing,
776 * or for a barrier lock to be released
777 */
778 wait_queue_head_t j_wait_transaction_locked;
779
780 /* Wait queue for waiting for checkpointing to complete */
781 wait_queue_head_t j_wait_logspace;
782
783 /* Wait queue for waiting for commit to complete */
784 wait_queue_head_t j_wait_done_commit;
785
786 /* Wait queue to trigger checkpointing */
787 wait_queue_head_t j_wait_checkpoint;
788
789 /* Wait queue to trigger commit */
790 wait_queue_head_t j_wait_commit;
791
792 /* Wait queue to wait for updates to complete */
793 wait_queue_head_t j_wait_updates;
794
795 /* Semaphore for locking against concurrent checkpoints */
796 struct mutex j_checkpoint_mutex;
797
798 /*
Theodore Ts'o1a0d3782008-11-05 00:09:22 -0500799 * List of buffer heads used by the checkpoint routine. This
800 * was moved from jbd2_log_do_checkpoint() to reduce stack
801 * usage. Access to this array is controlled by the
802 * j_checkpoint_mutex. [j_checkpoint_mutex]
803 */
804 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
805
806 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700807 * Journal head: identifies the first unused block in the journal.
808 * [j_state_lock]
809 */
810 unsigned long j_head;
811
812 /*
813 * Journal tail: identifies the oldest still-used block in the journal.
814 * [j_state_lock]
815 */
816 unsigned long j_tail;
817
818 /*
819 * Journal free: how many free blocks are there in the journal?
820 * [j_state_lock]
821 */
822 unsigned long j_free;
823
824 /*
825 * Journal start and end: the block numbers of the first usable block
826 * and one beyond the last usable block in the journal. [j_state_lock]
827 */
828 unsigned long j_first;
829 unsigned long j_last;
830
831 /*
832 * Device, blocksize and starting block offset for the location where we
833 * store the journal.
834 */
835 struct block_device *j_dev;
836 int j_blocksize;
Theodore Ts'o05496762008-09-16 14:36:17 -0400837 unsigned long long j_blk_offset;
838 char j_devname[BDEVNAME_SIZE+24];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700839
840 /*
841 * Device which holds the client fs. For internal journal this will be
842 * equal to j_dev.
843 */
844 struct block_device *j_fs_dev;
845
846 /* Total maximum capacity of the journal region on disk. */
847 unsigned int j_maxlen;
848
849 /*
850 * Protects the buffer lists and internal buffer state.
851 */
852 spinlock_t j_list_lock;
853
854 /* Optional inode where we store the journal. If present, all */
855 /* journal block numbers are mapped into this inode via */
856 /* bmap(). */
857 struct inode *j_inode;
858
859 /*
860 * Sequence number of the oldest transaction in the log [j_state_lock]
861 */
862 tid_t j_tail_sequence;
863
864 /*
865 * Sequence number of the next transaction to grant [j_state_lock]
866 */
867 tid_t j_transaction_sequence;
868
869 /*
870 * Sequence number of the most recently committed transaction
871 * [j_state_lock].
872 */
873 tid_t j_commit_sequence;
874
875 /*
876 * Sequence number of the most recent transaction wanting commit
877 * [j_state_lock]
878 */
879 tid_t j_commit_request;
880
881 /*
882 * Journal uuid: identifies the object (filesystem, LVM volume etc)
883 * backed by this journal. This will eventually be replaced by an array
884 * of uuids, allowing us to index multiple devices within a single
885 * journal and to perform atomic updates across them.
886 */
887 __u8 j_uuid[16];
888
889 /* Pointer to the current commit thread for this journal */
890 struct task_struct *j_task;
891
892 /*
893 * Maximum number of metadata buffers to allow in a single compound
894 * commit transaction
895 */
896 int j_max_transaction_buffers;
897
898 /*
899 * What is the maximum transaction lifetime before we begin a commit?
900 */
901 unsigned long j_commit_interval;
902
903 /* The timer used to wakeup the commit thread: */
904 struct timer_list j_commit_timer;
905
906 /*
907 * The revoke table: maintains the list of revoked blocks in the
908 * current transaction. [j_revoke_lock]
909 */
910 spinlock_t j_revoke_lock;
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700911 struct jbd2_revoke_table_s *j_revoke;
912 struct jbd2_revoke_table_s *j_revoke_table[2];
Dave Kleikamp470decc2006-10-11 01:20:57 -0700913
914 /*
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700915 * array of bhs for jbd2_journal_commit_transaction
Dave Kleikamp470decc2006-10-11 01:20:57 -0700916 */
917 struct buffer_head **j_wbuf;
918 int j_wbufsize;
919
Josef Bacike07f7182008-11-26 01:14:26 -0500920 /*
921 * this is the pid of hte last person to run a synchronous operation
922 * through the journal
923 */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700924 pid_t j_last_sync_writer;
925
Josef Bacike07f7182008-11-26 01:14:26 -0500926 /*
927 * the average amount of time in nanoseconds it takes to commit a
928 * transaction to disk. [j_state_lock]
929 */
930 u64 j_average_commit_time;
931
Theodore Ts'o30773842009-01-03 20:27:38 -0500932 /*
933 * minimum and maximum times that we should wait for
934 * additional filesystem operations to get batched into a
935 * synchronous handle in microseconds
936 */
937 u32 j_min_batch_time;
938 u32 j_max_batch_time;
939
Theodore Ts'o3e624fc2008-10-16 20:00:24 -0400940 /* This function is called when a transaction is closed */
941 void (*j_commit_callback)(journal_t *,
942 transaction_t *);
943
Dave Kleikamp470decc2006-10-11 01:20:57 -0700944 /*
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500945 * Journal statistics
946 */
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500947 spinlock_t j_history_lock;
948 struct proc_dir_entry *j_proc_entry;
949 struct transaction_stats_s j_stats;
950
Theodore Ts'o624080e2008-06-06 17:50:40 -0400951 /* Failed journal commit ID */
952 unsigned int j_failed_commit;
953
Johann Lombardi8e85fb32008-01-28 23:58:27 -0500954 /*
Dave Kleikamp470decc2006-10-11 01:20:57 -0700955 * An opaque pointer to fs-private information. ext3 puts its
956 * superblock pointer here
957 */
958 void *j_private;
Darrick J. Wong01b5adc2012-05-27 07:50:56 -0400959
960 /* Reference to checksum algorithm driver via cryptoapi */
961 struct crypto_shash *j_chksum_driver;
Darrick J. Wong4fd5ea42012-05-27 08:08:22 -0400962
963 /* Precomputed journal UUID checksum for seeding other checksums */
964 __u32 j_csum_seed;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700965};
966
967/*
968 * Journal flag definitions
969 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700970#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
971#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
972#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
973#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
974#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
975#define JBD2_BARRIER 0x020 /* Use IDE barriers */
Hidehiro Kawai5bf56832008-10-10 22:12:43 -0400976#define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
977 * data write error in ordered
978 * mode */
Dave Kleikamp470decc2006-10-11 01:20:57 -0700979
980/*
981 * Function declarations for the journaling transaction and buffer
982 * management
983 */
984
985/* Filing buffers */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700986extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700987extern void __jbd2_journal_refile_buffer(struct journal_head *);
988extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
989extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700990extern void __journal_free_buffer(struct journal_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700991extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700992extern void __journal_clean_data_list(transaction_t *transaction);
993
994/* Log buffer allocation */
Mingming Caof7f4bcc2006-10-11 01:20:59 -0700995extern struct journal_head * jbd2_journal_get_descriptor_buffer(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -0700996int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
Jan Kara79feb522012-03-13 22:22:54 -0400997int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
998 unsigned long *block);
999void __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
Jan Kara33395782012-03-13 22:45:38 -04001000void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001001
1002/* Commit management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001003extern void jbd2_journal_commit_transaction(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001004
1005/* Checkpoint list management */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001006int __jbd2_journal_clean_checkpoint_list(journal_t *journal);
1007int __jbd2_journal_remove_checkpoint(struct journal_head *);
1008void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001009
Joel Beckere06c8222008-09-11 15:35:47 -07001010
1011/*
1012 * Triggers
1013 */
1014
1015struct jbd2_buffer_trigger_type {
1016 /*
Jan Kara13ceef02010-07-14 07:56:33 +02001017 * Fired a the moment data to write to the journal are known to be
1018 * stable - so either at the moment b_frozen_data is created or just
1019 * before a buffer is written to the journal. mapped_data is a mapped
1020 * buffer that is the frozen data for commit.
Joel Beckere06c8222008-09-11 15:35:47 -07001021 */
Jan Kara13ceef02010-07-14 07:56:33 +02001022 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
Joel Beckere06c8222008-09-11 15:35:47 -07001023 struct buffer_head *bh, void *mapped_data,
1024 size_t size);
1025
1026 /*
1027 * Fired during journal abort for dirty buffers that will not be
1028 * committed.
1029 */
1030 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1031 struct buffer_head *bh);
1032};
1033
Jan Kara13ceef02010-07-14 07:56:33 +02001034extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
Joel Beckere06c8222008-09-11 15:35:47 -07001035 void *mapped_data,
1036 struct jbd2_buffer_trigger_type *triggers);
1037extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1038 struct jbd2_buffer_trigger_type *triggers);
1039
Dave Kleikamp470decc2006-10-11 01:20:57 -07001040/* Buffer IO */
1041extern int
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001042jbd2_journal_write_metadata_buffer(transaction_t *transaction,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001043 struct journal_head *jh_in,
1044 struct journal_head **jh_out,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001045 unsigned long long blocknr);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001046
1047/* Transaction locking */
1048extern void __wait_on_journal (journal_t *);
1049
Yongqiang Yang0c2022e2012-02-20 17:53:02 -05001050/* Transaction cache support */
1051extern void jbd2_journal_destroy_transaction_cache(void);
1052extern int jbd2_journal_init_transaction_cache(void);
1053extern void jbd2_journal_free_transaction(transaction_t *);
1054
Dave Kleikamp470decc2006-10-11 01:20:57 -07001055/*
1056 * Journal locking.
1057 *
1058 * We need to lock the journal during transaction state changes so that nobody
1059 * ever tries to take a handle on the running transaction while we are in the
1060 * middle of moving it to the commit phase. j_state_lock does this.
1061 *
1062 * Note that the locking is completely interrupt unsafe. We never touch
1063 * journal structures from interrupts.
1064 */
1065
1066static inline handle_t *journal_current_handle(void)
1067{
1068 return current->journal_info;
1069}
1070
1071/* The journaling code user interface:
1072 *
1073 * Create and destroy handles
1074 * Register buffer modifications against the current transaction.
1075 */
1076
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001077extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
Theodore Ts'o343d9c22013-02-08 13:00:22 -05001078extern handle_t *jbd2__journal_start(journal_t *, int nblocks, gfp_t gfp_mask,
1079 unsigned int type, unsigned int line_no);
Theodore Ts'o47def822010-07-27 11:56:05 -04001080extern int jbd2_journal_restart(handle_t *, int nblocks);
Dan Carpenterd2159fb2011-09-04 10:20:14 -04001081extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001082extern int jbd2_journal_extend (handle_t *, int nblocks);
1083extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1084extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1085extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
Joel Beckere06c8222008-09-11 15:35:47 -07001086void jbd2_journal_set_triggers(struct buffer_head *,
1087 struct jbd2_buffer_trigger_type *type);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001088extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001089extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001090extern void journal_sync_buffer (struct buffer_head *);
Jan Kara53e87262012-12-25 13:29:52 -05001091extern int jbd2_journal_invalidatepage(journal_t *,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001092 struct page *, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001093extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1094extern int jbd2_journal_stop(handle_t *);
1095extern int jbd2_journal_flush (journal_t *);
1096extern void jbd2_journal_lock_updates (journal_t *);
1097extern void jbd2_journal_unlock_updates (journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001098
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001099extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
Dave Kleikamp470decc2006-10-11 01:20:57 -07001100 struct block_device *fs_dev,
Mingming Cao18eba7a2006-10-11 01:21:13 -07001101 unsigned long long start, int len, int bsize);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001102extern journal_t * jbd2_journal_init_inode (struct inode *);
1103extern int jbd2_journal_update_format (journal_t *);
1104extern int jbd2_journal_check_used_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001105 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001106extern int jbd2_journal_check_available_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001107 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001108extern int jbd2_journal_set_features
Dave Kleikamp470decc2006-10-11 01:20:57 -07001109 (journal_t *, unsigned long, unsigned long, unsigned long);
Girish Shilamkar818d2762008-01-28 23:58:27 -05001110extern void jbd2_journal_clear_features
1111 (journal_t *, unsigned long, unsigned long, unsigned long);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001112extern int jbd2_journal_load (journal_t *journal);
Hidehiro Kawai44519fa2008-10-10 20:29:13 -04001113extern int jbd2_journal_destroy (journal_t *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001114extern int jbd2_journal_recover (journal_t *journal);
1115extern int jbd2_journal_wipe (journal_t *, int);
1116extern int jbd2_journal_skip_recovery (journal_t *);
Theodore Ts'od796c522012-08-05 19:04:57 -04001117extern void jbd2_journal_update_sb_errno(journal_t *);
Jan Kara79feb522012-03-13 22:22:54 -04001118extern void jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
1119 unsigned long, int);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001120extern void __jbd2_journal_abort_hard (journal_t *);
1121extern void jbd2_journal_abort (journal_t *, int);
1122extern int jbd2_journal_errno (journal_t *);
1123extern void jbd2_journal_ack_err (journal_t *);
1124extern int jbd2_journal_clear_err (journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001125extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001126extern int jbd2_journal_force_commit(journal_t *);
Jan Karac851ed52008-07-11 19:27:31 -04001127extern int jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
Jan Kara7f5aa212009-02-10 11:15:34 -05001128extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1129 struct jbd2_inode *inode, loff_t new_size);
Jan Karac851ed52008-07-11 19:27:31 -04001130extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1131extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001132
1133/*
1134 * journal_head management
1135 */
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001136struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1137struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001138void jbd2_journal_put_journal_head(struct journal_head *jh);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001139
1140/*
1141 * handle management
1142 */
Christoph Lametere18b8902006-12-06 20:33:20 -08001143extern struct kmem_cache *jbd2_handle_cache;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001144
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001145static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001146{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001147 return kmem_cache_alloc(jbd2_handle_cache, gfp_flags);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001148}
1149
Mingming Caoaf1e76d2007-10-16 18:38:25 -04001150static inline void jbd2_free_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001151{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001152 kmem_cache_free(jbd2_handle_cache, handle);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001153}
1154
Theodore Ts'o8aefcd52011-01-10 12:29:43 -05001155/*
1156 * jbd2_inode management (optional, for those file systems that want to use
1157 * dynamically allocated jbd2_inode structures)
1158 */
1159extern struct kmem_cache *jbd2_inode_cache;
1160
1161static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1162{
1163 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1164}
1165
1166static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1167{
1168 kmem_cache_free(jbd2_inode_cache, jinode);
1169}
1170
Dave Kleikamp470decc2006-10-11 01:20:57 -07001171/* Primary revoke support */
1172#define JOURNAL_REVOKE_DEFAULT_HASH 256
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001173extern int jbd2_journal_init_revoke(journal_t *, int);
1174extern void jbd2_journal_destroy_revoke_caches(void);
1175extern int jbd2_journal_init_revoke_caches(void);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001176
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001177extern void jbd2_journal_destroy_revoke(journal_t *);
Mingming Cao18eba7a2006-10-11 01:21:13 -07001178extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001179extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
Theodore Ts'o67c457a2009-04-14 07:50:56 -04001180extern void jbd2_journal_write_revoke_records(journal_t *,
1181 transaction_t *, int);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001182
1183/* Recovery revoke support */
Mingming Cao18eba7a2006-10-11 01:21:13 -07001184extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1185extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001186extern void jbd2_journal_clear_revoke(journal_t *);
1187extern void jbd2_journal_switch_revoke_table(journal_t *journal);
Yongqiang Yang1ba37262011-12-28 17:46:46 -05001188extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001189
1190/*
1191 * The log thread user interface:
1192 *
1193 * Request space in the current transaction, and force transaction commit
1194 * transitions on demand.
1195 */
1196
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001197int __jbd2_log_space_left(journal_t *); /* Called with journal locked */
1198int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1199int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1200int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1201int jbd2_journal_force_commit_nested(journal_t *journal);
1202int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1203int jbd2_log_do_checkpoint(journal_t *journal);
Jan Karabbd2be32011-05-24 11:59:18 -04001204int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001205
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001206void __jbd2_log_wait_for_space(journal_t *journal);
Aneesh Kumar K.Vfb684072008-11-06 17:50:21 -05001207extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1208extern int jbd2_cleanup_journal_tail(journal_t *);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001209
1210/* Debugging code only: */
1211
1212#define jbd_ENOSYS() \
1213do { \
Harvey Harrisond5c003b2008-10-15 22:01:24 -07001214 printk (KERN_ERR "JBD unimplemented function %s\n", __func__); \
Dave Kleikamp470decc2006-10-11 01:20:57 -07001215 current->state = TASK_UNINTERRUPTIBLE; \
1216 schedule(); \
1217} while (1)
1218
1219/*
1220 * is_journal_abort
1221 *
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001222 * Simple test wrapper function to test the JBD2_ABORT state flag. This
Dave Kleikamp470decc2006-10-11 01:20:57 -07001223 * bit, when set, indicates that we have had a fatal error somewhere,
1224 * either inside the journaling layer or indicated to us by the client
1225 * (eg. ext3), and that we and should not commit any further
1226 * transactions.
1227 */
1228
1229static inline int is_journal_aborted(journal_t *journal)
1230{
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001231 return journal->j_flags & JBD2_ABORT;
Dave Kleikamp470decc2006-10-11 01:20:57 -07001232}
1233
1234static inline int is_handle_aborted(handle_t *handle)
1235{
1236 if (handle->h_aborted)
1237 return 1;
1238 return is_journal_aborted(handle->h_transaction->t_journal);
1239}
1240
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001241static inline void jbd2_journal_abort_handle(handle_t *handle)
Dave Kleikamp470decc2006-10-11 01:20:57 -07001242{
1243 handle->h_aborted = 1;
1244}
1245
1246#endif /* __KERNEL__ */
1247
1248/* Comparison functions for transaction IDs: perform comparisons using
1249 * modulo arithmetic so that they work over sequence number wraps. */
1250
1251static inline int tid_gt(tid_t x, tid_t y)
1252{
1253 int difference = (x - y);
1254 return (difference > 0);
1255}
1256
1257static inline int tid_geq(tid_t x, tid_t y)
1258{
1259 int difference = (x - y);
1260 return (difference >= 0);
1261}
1262
Mingming Caof7f4bcc2006-10-11 01:20:59 -07001263extern int jbd2_journal_blocks_per_page(struct inode *inode);
Zach Brownb517bea2006-10-11 01:21:08 -07001264extern size_t journal_tag_bytes(journal_t *journal);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001265
1266/*
1267 * Return the minimum number of blocks which must be free in the journal
1268 * before a new transaction may be started. Must be called under j_state_lock.
1269 */
1270static inline int jbd_space_needed(journal_t *journal)
1271{
1272 int nblocks = journal->j_max_transaction_buffers;
1273 if (journal->j_committing_transaction)
Theodore Ts'oa51dca92010-08-02 08:43:25 -04001274 nblocks += atomic_read(&journal->j_committing_transaction->
1275 t_outstanding_credits);
Dave Kleikamp470decc2006-10-11 01:20:57 -07001276 return nblocks;
1277}
1278
1279/*
1280 * Definitions which augment the buffer_head layer
1281 */
1282
1283/* journaling buffer types */
1284#define BJ_None 0 /* Not journaled */
Jan Kara87c89c22008-07-11 19:27:31 -04001285#define BJ_Metadata 1 /* Normal journaled metadata */
1286#define BJ_Forget 2 /* Buffer superseded by this transaction */
1287#define BJ_IO 3 /* Buffer is for temporary IO use */
1288#define BJ_Shadow 4 /* Buffer contents being shadowed to the log */
1289#define BJ_LogCtl 5 /* Buffer contains log descriptors */
1290#define BJ_Reserved 6 /* Buffer is reserved for access by journal */
1291#define BJ_Types 7
Dave Kleikamp470decc2006-10-11 01:20:57 -07001292
1293extern int jbd_blocks_per_page(struct inode *inode);
1294
Behan Websterfd47d3e2012-11-08 11:24:46 -05001295/* JBD uses a CRC32 checksum */
1296#define JBD_MAX_CHECKSUM_SIZE 4
1297
Darrick J. Wong01b5adc2012-05-27 07:50:56 -04001298static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1299 const void *address, unsigned int length)
1300{
1301 struct {
1302 struct shash_desc shash;
Behan Websterfd47d3e2012-11-08 11:24:46 -05001303 char ctx[JBD_MAX_CHECKSUM_SIZE];
Darrick J. Wong01b5adc2012-05-27 07:50:56 -04001304 } desc;
1305 int err;
1306
Behan Websterfd47d3e2012-11-08 11:24:46 -05001307 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1308 JBD_MAX_CHECKSUM_SIZE);
1309
Darrick J. Wong01b5adc2012-05-27 07:50:56 -04001310 desc.shash.tfm = journal->j_chksum_driver;
1311 desc.shash.flags = 0;
1312 *(u32 *)desc.ctx = crc;
1313
1314 err = crypto_shash_update(&desc.shash, address, length);
1315 BUG_ON(err);
1316
1317 return *(u32 *)desc.ctx;
1318}
1319
Dave Kleikamp470decc2006-10-11 01:20:57 -07001320#ifdef __KERNEL__
1321
1322#define buffer_trace_init(bh) do {} while (0)
1323#define print_buffer_fields(bh) do {} while (0)
1324#define print_buffer_trace(bh) do {} while (0)
1325#define BUFFER_TRACE(bh, info) do {} while (0)
1326#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1327#define JBUFFER_TRACE(jh, info) do {} while (0)
1328
1329#endif /* __KERNEL__ */
1330
Mingming Caocd02ff02007-10-16 18:38:25 -04001331#endif /* _LINUX_JBD2_H */