]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/jbd2.h
mod_devicetable: fix PHY module format
[mirror_ubuntu-bionic-kernel.git] / include / linux / jbd2.h
CommitLineData
470decc6 1/*
f7f4bccb 2 * linux/include/linux/jbd2.h
470decc6
DK
3 *
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
cd02ff0b
MC
16#ifndef _LINUX_JBD2_H
17#define _LINUX_JBD2_H
470decc6
DK
18
19/* Allow this file to be included directly into e2fsprogs */
20#ifndef __KERNEL__
21#include "jfs_compat.h"
f7f4bccb 22#define JBD2_DEBUG
470decc6
DK
23#else
24
25#include <linux/types.h>
26#include <linux/buffer_head.h>
27#include <linux/journal-head.h>
28#include <linux/stddef.h>
470decc6
DK
29#include <linux/mutex.h>
30#include <linux/timer.h>
5a0e3ad6 31#include <linux/slab.h>
c290ea01 32#include <linux/bit_spinlock.h>
01b5adce 33#include <crypto/hash.h>
470decc6
DK
34#endif
35
36#define journal_oom_retry 1
37
38/*
cd02ff0b 39 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
470decc6 40 * certain classes of error which can occur due to failed IOs. Under
cd02ff0b 41 * normal use we want ext4 to continue after such errors, because
470decc6
DK
42 * hardware _can_ fail, but for debugging purposes when running tests on
43 * known-good hardware we may want to trap these errors.
44 */
cd02ff0b 45#undef JBD2_PARANOID_IOFAIL
470decc6
DK
46
47/*
48 * The default maximum commit age, in seconds.
49 */
cd02ff0b 50#define JBD2_DEFAULT_MAX_COMMIT_AGE 5
470decc6 51
e23291b9 52#ifdef CONFIG_JBD2_DEBUG
470decc6 53/*
cd02ff0b 54 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
470decc6 55 * consistency checks. By default we don't do this unless
e23291b9 56 * CONFIG_JBD2_DEBUG is on.
470decc6 57 */
cd02ff0b 58#define JBD2_EXPENSIVE_CHECKING
b6e96d00 59extern ushort jbd2_journal_enable_debug;
169f1a2a
PG
60void __jbd2_debug(int level, const char *file, const char *func,
61 unsigned int line, const char *fmt, ...);
470decc6 62
169f1a2a
PG
63#define jbd_debug(n, fmt, a...) \
64 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
470decc6 65#else
169f1a2a 66#define jbd_debug(n, fmt, a...) /**/
470decc6
DK
67#endif
68
d2eecb03
TT
69extern void *jbd2_alloc(size_t size, gfp_t flags);
70extern void jbd2_free(void *ptr, size_t size);
af1e76d6 71
f7f4bccb 72#define JBD2_MIN_JOURNAL_BLOCKS 1024
470decc6
DK
73
74#ifdef __KERNEL__
75
76/**
77 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
78 *
79 * All filesystem modifications made by the process go
80 * through this handle. Recursive operations (such as quota operations)
81 * are gathered into a single update.
82 *
83 * The buffer credits field is used to account for journaled buffers
84 * being modified by the running process. To ensure that there is
85 * enough log space for all outstanding operations, we need to limit the
86 * number of outstanding buffers possible at any time. When the
87 * operation completes, any buffer credits not used are credited back to
88 * the transaction, so that at all times we know how many buffers the
89 * outstanding updates on a transaction might possibly touch.
90 *
91 * This is an opaque datatype.
92 **/
8aefcd55 93typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
470decc6
DK
94
95
96/**
97 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
98 *
99 * journal_t is linked to from the fs superblock structure.
100 *
101 * We use the journal_t to keep track of all outstanding transaction
102 * activity on the filesystem, and to manage the state of the log
103 * writing process.
104 *
105 * This is an opaque datatype.
106 **/
107typedef struct journal_s journal_t; /* Journal control structure */
108#endif
109
110/*
111 * Internal structures used by the logging mechanism:
112 */
113
f7f4bccb 114#define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
470decc6
DK
115
116/*
117 * On-disk structures
118 */
119
120/*
121 * Descriptor block types:
122 */
123
f7f4bccb
MC
124#define JBD2_DESCRIPTOR_BLOCK 1
125#define JBD2_COMMIT_BLOCK 2
126#define JBD2_SUPERBLOCK_V1 3
127#define JBD2_SUPERBLOCK_V2 4
128#define JBD2_REVOKE_BLOCK 5
470decc6
DK
129
130/*
131 * Standard header for all descriptor blocks:
132 */
133typedef struct journal_header_s
134{
135 __be32 h_magic;
136 __be32 h_blocktype;
137 __be32 h_sequence;
138} journal_header_t;
139
818d276c
GS
140/*
141 * Checksum types.
142 */
143#define JBD2_CRC32_CHKSUM 1
144#define JBD2_MD5_CHKSUM 2
145#define JBD2_SHA1_CHKSUM 3
8f888ef8 146#define JBD2_CRC32C_CHKSUM 4
818d276c
GS
147
148#define JBD2_CRC32_CHKSUM_SIZE 4
149
150#define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
151/*
152 * Commit block header for storing transactional checksums:
8f888ef8
DW
153 *
154 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
155 * fields are used to store a checksum of the descriptor and data blocks.
156 *
157 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
158 * field is used to store crc32c(uuid+commit_block). Each journal metadata
159 * block gets its own checksum, and data block checksums are stored in
160 * journal_block_tag (in the descriptor). The other h_chksum* fields are
161 * not used.
162 *
db9ee220
DW
163 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
164 * journal_block_tag3_t to store a full 32-bit checksum. Everything else
165 * is the same as v2.
166 *
167 * Checksum v1, v2, and v3 are mutually exclusive features.
818d276c
GS
168 */
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];
736603ab
TT
177 __be64 h_commit_sec;
178 __be32 h_commit_nsec;
818d276c 179};
470decc6
DK
180
181/*
b517bea1
ZB
182 * 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.
470decc6 186 */
db9ee220
DW
187typedef struct journal_block_tag3_s
188{
189 __be32 t_blocknr; /* The on-disk block number */
190 __be32 t_flags; /* See below */
191 __be32 t_blocknr_high; /* most-significant high 32bits. */
192 __be32 t_checksum; /* crc32c(uuid+seq+block) */
193} journal_block_tag3_t;
194
470decc6
DK
195typedef struct journal_block_tag_s
196{
197 __be32 t_blocknr; /* The on-disk block number */
8f888ef8
DW
198 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
199 __be16 t_flags; /* See below */
b517bea1 200 __be32 t_blocknr_high; /* most-significant high 32bits. */
470decc6
DK
201} journal_block_tag_t;
202
1101cd4d 203/* Tail of descriptor or revoke block, for checksumming */
8f888ef8
DW
204struct jbd2_journal_block_tail {
205 __be32 t_checksum; /* crc32c(uuid+descr_block) */
206};
207
470decc6
DK
208/*
209 * The revoke descriptor: used on disk to describe a series of blocks to
210 * be revoked from the log
211 */
f7f4bccb 212typedef struct jbd2_journal_revoke_header_s
470decc6
DK
213{
214 journal_header_t r_header;
215 __be32 r_count; /* Count of bytes used in the block */
f7f4bccb 216} jbd2_journal_revoke_header_t;
470decc6 217
470decc6 218/* Definitions for the journal tag flags word: */
f7f4bccb
MC
219#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 */
470decc6
DK
223
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 */
f7f4bccb 245 /* Error value, as set by jbd2_journal_abort(). */
470decc6
DK
246 __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 */
8f888ef8
DW
266 __u8 s_checksum_type; /* checksum type */
267 __u8 s_padding2[3];
268 __u32 s_padding[42];
269 __be32 s_checksum; /* crc32c(superblock) */
470decc6
DK
270
271/* 0x0100 */
272 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
273/* 0x0400 */
274} journal_superblock_t;
275
56316a0d 276/* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
f7f4bccb 277#define JBD2_HAS_COMPAT_FEATURE(j,mask) \
470decc6
DK
278 ((j)->j_format_version >= 2 && \
279 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
f7f4bccb 280#define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
470decc6
DK
281 ((j)->j_format_version >= 2 && \
282 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
f7f4bccb 283#define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
470decc6
DK
284 ((j)->j_format_version >= 2 && \
285 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
286
56316a0d 287#define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
818d276c
GS
288
289#define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
290#define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
291#define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
8f888ef8 292#define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
db9ee220 293#define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010
470decc6 294
56316a0d
DW
295/* See "journal feature predicate functions" below */
296
470decc6 297/* Features known to this kernel version: */
818d276c 298#define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
f7f4bccb 299#define JBD2_KNOWN_ROCOMPAT_FEATURES 0
b517bea1 300#define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
818d276c 301 JBD2_FEATURE_INCOMPAT_64BIT | \
e93376c2 302 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
db9ee220
DW
303 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
304 JBD2_FEATURE_INCOMPAT_CSUM_V3)
470decc6
DK
305
306#ifdef __KERNEL__
307
308#include <linux/fs.h>
309#include <linux/sched.h>
b34090e5
JK
310
311enum jbd_state_bits {
312 BH_JBD /* Has an attached ext3 journal_head */
313 = BH_PrivateStart,
314 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
315 BH_Freed, /* Has been freed (truncated) */
316 BH_Revoked, /* Has been revoked from the log */
317 BH_RevokeValid, /* Revoked flag is valid */
318 BH_JBDDirty, /* Is dirty but journaled */
319 BH_State, /* Pins most journal_head state */
320 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
321 BH_Shadow, /* IO on shadow buffer is running */
322 BH_Verified, /* Metadata block has been verified ok */
323 BH_JBDPrivateStart, /* First bit available for private use by FS */
324};
325
326BUFFER_FNS(JBD, jbd)
327BUFFER_FNS(JWrite, jwrite)
328BUFFER_FNS(JBDDirty, jbddirty)
329TAS_BUFFER_FNS(JBDDirty, jbddirty)
330BUFFER_FNS(Revoked, revoked)
331TAS_BUFFER_FNS(Revoked, revoked)
332BUFFER_FNS(RevokeValid, revokevalid)
333TAS_BUFFER_FNS(RevokeValid, revokevalid)
334BUFFER_FNS(Freed, freed)
335BUFFER_FNS(Shadow, shadow)
336BUFFER_FNS(Verified, verified)
337
c290ea01
JK
338static inline struct buffer_head *jh2bh(struct journal_head *jh)
339{
340 return jh->b_bh;
341}
342
343static inline struct journal_head *bh2jh(struct buffer_head *bh)
344{
345 return bh->b_private;
346}
347
348static inline void jbd_lock_bh_state(struct buffer_head *bh)
349{
350 bit_spin_lock(BH_State, &bh->b_state);
351}
352
353static inline int jbd_trylock_bh_state(struct buffer_head *bh)
354{
355 return bit_spin_trylock(BH_State, &bh->b_state);
356}
357
358static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
359{
360 return bit_spin_is_locked(BH_State, &bh->b_state);
361}
362
363static inline void jbd_unlock_bh_state(struct buffer_head *bh)
364{
365 bit_spin_unlock(BH_State, &bh->b_state);
366}
367
368static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
369{
370 bit_spin_lock(BH_JournalHead, &bh->b_state);
371}
372
373static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
374{
375 bit_spin_unlock(BH_JournalHead, &bh->b_state);
376}
470decc6 377
36df53f4 378#define J_ASSERT(assert) BUG_ON(!(assert))
470decc6 379
470decc6
DK
380#define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
381#define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
470decc6 382
cd02ff0b 383#if defined(JBD2_PARANOID_IOFAIL)
470decc6
DK
384#define J_EXPECT(expr, why...) J_ASSERT(expr)
385#define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
386#define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
387#else
388#define __journal_expect(expr, why...) \
389 ({ \
390 int val = (expr); \
391 if (!val) { \
392 printk(KERN_ERR \
08ec8c38
TT
393 "JBD2 unexpected failure: %s: %s;\n", \
394 __func__, #expr); \
470decc6
DK
395 printk(KERN_ERR why "\n"); \
396 } \
397 val; \
398 })
399#define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
400#define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
401#define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
402#endif
403
c851ed54
JK
404/* Flags in jbd_inode->i_flags */
405#define __JI_COMMIT_RUNNING 0
41617e1a
JK
406#define __JI_WRITE_DATA 1
407#define __JI_WAIT_DATA 2
408
409/*
410 * Commit of the inode data in progress. We use this flag to protect us from
c851ed54
JK
411 * concurrent deletion of inode. We cannot use reference to inode for this
412 * since we cannot afford doing last iput() on behalf of kjournald
413 */
414#define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
41617e1a
JK
415/* Write allocated dirty buffers in this inode before commit */
416#define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
417/* Wait for outstanding data writes for this inode before commit */
418#define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
c851ed54
JK
419
420/**
1980a53c
TH
421 * struct jbd_inode - The jbd_inode type is the structure linking inodes in
422 * ordered mode present in a transaction so that we can sync them during commit.
c851ed54
JK
423 */
424struct jbd2_inode {
1980a53c
TH
425 /**
426 * @i_transaction:
427 *
428 * Which transaction does this inode belong to? Either the running
429 * transaction or the committing one. [j_list_lock]
430 */
c851ed54
JK
431 transaction_t *i_transaction;
432
1980a53c
TH
433 /**
434 * @i_next_transaction:
435 *
436 * Pointer to the running transaction modifying inode's data in case
437 * there is already a committing transaction touching it. [j_list_lock]
438 */
c851ed54
JK
439 transaction_t *i_next_transaction;
440
1980a53c
TH
441 /**
442 * @i_list: List of inodes in the i_transaction [j_list_lock]
443 */
c851ed54
JK
444 struct list_head i_list;
445
1980a53c
TH
446 /**
447 * @i_vfs_inode:
448 *
449 * VFS inode this inode belongs to [constant for lifetime of structure]
450 */
c851ed54
JK
451 struct inode *i_vfs_inode;
452
1980a53c
TH
453 /**
454 * @i_flags: Flags of inode [j_list_lock]
455 */
39e3ac25 456 unsigned long i_flags;
d55ca21d
RZ
457
458 /**
459 * @i_dirty_start:
460 *
461 * Offset in bytes where the dirty range for this inode starts.
462 * [j_list_lock]
463 */
464 loff_t i_dirty_start;
465
466 /**
467 * @i_dirty_end:
468 *
469 * Inclusive offset in bytes where the dirty range for this inode
470 * ends. [j_list_lock]
471 */
472 loff_t i_dirty_end;
c851ed54
JK
473};
474
f7f4bccb 475struct jbd2_revoke_table_s;
470decc6
DK
476
477/**
478 * struct handle_s - The handle_s type is the concrete type associated with
479 * handle_t.
480 * @h_transaction: Which compound transaction is this update a part of?
1980a53c
TH
481 * @h_journal: Which journal handle belongs to - used iff h_reserved set.
482 * @h_rsv_handle: Handle reserved for finishing the logical operation.
470decc6 483 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
1980a53c
TH
484 * @h_ref: Reference count on this handle.
485 * @h_err: Field for caller's use to track errors through large fs operations.
486 * @h_sync: Flag for sync-on-close.
487 * @h_jdata: Flag to force data journaling.
488 * @h_reserved: Flag for handle for reserved credits.
489 * @h_aborted: Flag indicating fatal error on handle.
490 * @h_type: For handle statistics.
491 * @h_line_no: For handle statistics.
492 * @h_start_jiffies: Handle Start time.
493 * @h_requested_credits: Holds @h_buffer_credits after handle is started.
494 * @saved_alloc_context: Saved context while transaction is open.
470decc6
DK
495 **/
496
497/* Docbook can't yet cope with the bit fields, but will leave the documentation
498 * in so it can be fixed later.
499 */
500
8aefcd55 501struct jbd2_journal_handle
470decc6 502{
8f7d89f3 503 union {
8f7d89f3
JK
504 transaction_t *h_transaction;
505 /* Which journal handle belongs to - used iff h_reserved set */
506 journal_t *h_journal;
507 };
508
8f7d89f3 509 handle_t *h_rsv_handle;
470decc6 510 int h_buffer_credits;
470decc6 511 int h_ref;
470decc6
DK
512 int h_err;
513
514 /* Flags [no locking] */
1980a53c
TH
515 unsigned int h_sync: 1;
516 unsigned int h_jdata: 1;
517 unsigned int h_reserved: 1;
518 unsigned int h_aborted: 1;
519 unsigned int h_type: 8;
520 unsigned int h_line_no: 16;
343d9c28
TT
521
522 unsigned long h_start_jiffies;
523 unsigned int h_requested_credits;
81378da6
MH
524
525 unsigned int saved_alloc_context;
470decc6
DK
526};
527
528
8e85fb3f
JL
529/*
530 * Some stats for checkpoint phase
531 */
532struct transaction_chp_stats_s {
533 unsigned long cs_chp_time;
bf699327
TT
534 __u32 cs_forced_to_close;
535 __u32 cs_written;
536 __u32 cs_dropped;
8e85fb3f
JL
537};
538
470decc6
DK
539/* The transaction_t type is the guts of the journaling mechanism. It
540 * tracks a compound transaction through its various states:
541 *
542 * RUNNING: accepting new updates
543 * LOCKED: Updates still running but we don't accept new ones
544 * RUNDOWN: Updates are tidying up but have finished requesting
545 * new buffers to modify (state not used for now)
546 * FLUSH: All updates complete, but we are still writing to disk
547 * COMMIT: All data on disk, writing commit record
548 * FINISHED: We still have to keep the transaction for checkpointing.
549 *
550 * The transaction keeps track of all of the buffers modified by a
551 * running transaction, and all of the buffers committed but not yet
552 * flushed to home for finished transactions.
553 */
554
555/*
556 * Lock ranking:
557 *
558 * j_list_lock
559 * ->jbd_lock_bh_journal_head() (This is "innermost")
560 *
561 * j_state_lock
562 * ->jbd_lock_bh_state()
563 *
564 * jbd_lock_bh_state()
565 * ->j_list_lock
566 *
567 * j_state_lock
568 * ->t_handle_lock
569 *
570 * j_state_lock
571 * ->j_list_lock (journal_unmap_buffer)
572 *
573 */
574
575struct transaction_s
576{
577 /* Pointer to the journal for this transaction. [no locking] */
578 journal_t *t_journal;
579
580 /* Sequence number for this transaction [no locking] */
581 tid_t t_tid;
582
583 /*
584 * Transaction's current state
f7f4bccb 585 * [no locking - only kjournald2 alters this]
f5a7a6b0
JK
586 * [j_list_lock] guards transition of a transaction into T_FINISHED
587 * state and subsequent call of __jbd2_journal_drop_transaction()
470decc6
DK
588 * FIXME: needs barriers
589 * KLUDGE: [use j_state_lock]
590 */
591 enum {
592 T_RUNNING,
593 T_LOCKED,
470decc6
DK
594 T_FLUSH,
595 T_COMMIT,
bbd2be36
JK
596 T_COMMIT_DFLUSH,
597 T_COMMIT_JFLUSH,
794446c6 598 T_COMMIT_CALLBACK,
470decc6
DK
599 T_FINISHED
600 } t_state;
601
602 /*
603 * Where in the log does this transaction's commit start? [no locking]
604 */
605 unsigned long t_log_start;
606
607 /* Number of buffers on the t_buffers list [j_list_lock] */
608 int t_nr_buffers;
609
610 /*
611 * Doubly-linked circular list of all buffers reserved but not yet
612 * modified by this transaction [j_list_lock]
613 */
614 struct journal_head *t_reserved_list;
615
470decc6
DK
616 /*
617 * Doubly-linked circular list of all metadata buffers owned by this
618 * transaction [j_list_lock]
619 */
620 struct journal_head *t_buffers;
621
470decc6
DK
622 /*
623 * Doubly-linked circular list of all forget buffers (superseded
624 * buffers which we can un-checkpoint once this transaction commits)
625 * [j_list_lock]
626 */
627 struct journal_head *t_forget;
628
629 /*
630 * Doubly-linked circular list of all buffers still to be flushed before
631 * this transaction can be checkpointed. [j_list_lock]
632 */
633 struct journal_head *t_checkpoint_list;
634
635 /*
636 * Doubly-linked circular list of all buffers submitted for IO while
637 * checkpointing. [j_list_lock]
638 */
639 struct journal_head *t_checkpoint_io_list;
640
470decc6
DK
641 /*
642 * Doubly-linked circular list of metadata buffers being shadowed by log
643 * IO. The IO buffers on the iobuf list and the shadow buffers on this
644 * list match each other one for one at all times. [j_list_lock]
645 */
646 struct journal_head *t_shadow_list;
647
c851ed54
JK
648 /*
649 * List of inodes whose data we've modified in data=ordered mode.
650 * [j_list_lock]
651 */
652 struct list_head t_inode_list;
653
470decc6
DK
654 /*
655 * Protects info related to handles
656 */
657 spinlock_t t_handle_lock;
658
8e85fb3f
JL
659 /*
660 * Longest time some handle had to wait for running transaction
661 */
662 unsigned long t_max_wait;
663
664 /*
665 * When transaction started
666 */
667 unsigned long t_start;
668
9fff24aa
TT
669 /*
670 * When commit was requested
671 */
672 unsigned long t_requested;
673
8e85fb3f
JL
674 /*
675 * Checkpointing stats [j_checkpoint_sem]
676 */
677 struct transaction_chp_stats_s t_chp_stats;
678
470decc6
DK
679 /*
680 * Number of outstanding updates running on this transaction
681 * [t_handle_lock]
682 */
a51dca9c 683 atomic_t t_updates;
470decc6
DK
684
685 /*
686 * Number of buffers reserved for use by all handles in this transaction
687 * handle but not yet modified. [t_handle_lock]
688 */
a51dca9c 689 atomic_t t_outstanding_credits;
470decc6
DK
690
691 /*
692 * Forward and backward links for the circular list of all transactions
693 * awaiting checkpoint. [j_list_lock]
694 */
695 transaction_t *t_cpnext, *t_cpprev;
696
697 /*
698 * When will the transaction expire (become due for commit), in jiffies?
699 * [no locking]
700 */
701 unsigned long t_expires;
702
e07f7183
JB
703 /*
704 * When this transaction started, in nanoseconds [no locking]
705 */
706 ktime_t t_start_time;
707
470decc6
DK
708 /*
709 * How many handles used this transaction? [t_handle_lock]
710 */
8dd42046 711 atomic_t t_handle_count;
470decc6 712
7058548c
TT
713 /*
714 * This transaction is being forced and some process is
715 * waiting for it to finish.
716 */
0ccff1a4 717 unsigned int t_synchronous_commit:1;
81be12c8
JK
718
719 /* Disk flush needs to be sent to fs partition [no locking] */
720 int t_need_data_flush;
7058548c 721
3e624fc7
TT
722 /*
723 * For use by the filesystem to store fs-specific data
724 * structures associated with the transaction
725 */
726 struct list_head t_private_list;
470decc6
DK
727};
728
8e85fb3f
JL
729struct transaction_run_stats_s {
730 unsigned long rs_wait;
9fff24aa 731 unsigned long rs_request_delay;
8e85fb3f
JL
732 unsigned long rs_running;
733 unsigned long rs_locked;
734 unsigned long rs_flushing;
735 unsigned long rs_logging;
736
bf699327
TT
737 __u32 rs_handle_count;
738 __u32 rs_blocks;
739 __u32 rs_blocks_logged;
8e85fb3f
JL
740};
741
742struct transaction_stats_s {
8e85fb3f 743 unsigned long ts_tid;
9fff24aa 744 unsigned long ts_requested;
bf699327 745 struct transaction_run_stats_s run;
8e85fb3f
JL
746};
747
8e85fb3f
JL
748static inline unsigned long
749jbd2_time_diff(unsigned long start, unsigned long end)
750{
751 if (end >= start)
752 return end - start;
753
754 return end + (MAX_JIFFY_OFFSET - start);
755}
756
1a0d3786
TT
757#define JBD2_NR_BATCH 64
758
470decc6
DK
759/**
760 * struct journal_s - The journal_s type is the concrete type associated with
761 * journal_t.
470decc6 762 */
470decc6
DK
763struct journal_s
764{
1980a53c
TH
765 /**
766 * @j_flags: General journaling state flags [j_state_lock]
767 */
470decc6
DK
768 unsigned long j_flags;
769
1980a53c
TH
770 /**
771 * @j_errno:
772 *
470decc6
DK
773 * Is there an outstanding uncleared error on the journal (from a prior
774 * abort)? [j_state_lock]
775 */
776 int j_errno;
777
1980a53c
TH
778 /**
779 * @j_sb_buffer: The first part of the superblock buffer.
780 */
470decc6 781 struct buffer_head *j_sb_buffer;
1980a53c
TH
782
783 /**
784 * @j_superblock: The second part of the superblock buffer.
785 */
470decc6
DK
786 journal_superblock_t *j_superblock;
787
1980a53c
TH
788 /**
789 * @j_format_version: Version of the superblock format.
790 */
470decc6
DK
791 int j_format_version;
792
1980a53c
TH
793 /**
794 * @j_state_lock: Protect the various scalars in the journal.
470decc6 795 */
a931da6a 796 rwlock_t j_state_lock;
470decc6 797
1980a53c
TH
798 /**
799 * @j_barrier_count:
800 *
470decc6
DK
801 * Number of processes waiting to create a barrier lock [j_state_lock]
802 */
803 int j_barrier_count;
804
1980a53c
TH
805 /**
806 * @j_barrier: The barrier lock itself.
807 */
470decc6
DK
808 struct mutex j_barrier;
809
1980a53c
TH
810 /**
811 * @j_running_transaction:
812 *
470decc6
DK
813 * Transactions: The current running transaction...
814 * [j_state_lock] [caller holding open handle]
815 */
816 transaction_t *j_running_transaction;
817
1980a53c
TH
818 /**
819 * @j_committing_transaction:
820 *
470decc6
DK
821 * the transaction we are pushing to disk
822 * [j_state_lock] [caller holding open handle]
823 */
824 transaction_t *j_committing_transaction;
825
1980a53c
TH
826 /**
827 * @j_checkpoint_transactions:
828 *
470decc6
DK
829 * ... and a linked circular list of all transactions waiting for
830 * checkpointing. [j_list_lock]
831 */
832 transaction_t *j_checkpoint_transactions;
833
1980a53c
TH
834 /**
835 * @j_wait_transaction_locked:
836 *
470decc6 837 * Wait queue for waiting for a locked transaction to start committing,
1980a53c 838 * or for a barrier lock to be released.
470decc6
DK
839 */
840 wait_queue_head_t j_wait_transaction_locked;
841
1980a53c
TH
842 /**
843 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
844 */
470decc6
DK
845 wait_queue_head_t j_wait_done_commit;
846
1980a53c
TH
847 /**
848 * @j_wait_commit: Wait queue to trigger commit.
849 */
470decc6
DK
850 wait_queue_head_t j_wait_commit;
851
1980a53c
TH
852 /**
853 * @j_wait_updates: Wait queue to wait for updates to complete.
854 */
470decc6
DK
855 wait_queue_head_t j_wait_updates;
856
1980a53c
TH
857 /**
858 * @j_wait_reserved:
859 *
860 * Wait queue to wait for reserved buffer credits to drop.
861 */
8f7d89f3
JK
862 wait_queue_head_t j_wait_reserved;
863
1980a53c
TH
864 /**
865 * @j_checkpoint_mutex:
866 *
867 * Semaphore for locking against concurrent checkpoints.
868 */
470decc6
DK
869 struct mutex j_checkpoint_mutex;
870
1980a53c
TH
871 /**
872 * @j_chkpt_bhs:
873 *
1a0d3786
TT
874 * List of buffer heads used by the checkpoint routine. This
875 * was moved from jbd2_log_do_checkpoint() to reduce stack
876 * usage. Access to this array is controlled by the
1980a53c 877 * @j_checkpoint_mutex. [j_checkpoint_mutex]
1a0d3786
TT
878 */
879 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
1980a53c
TH
880
881 /**
882 * @j_head:
883 *
470decc6
DK
884 * Journal head: identifies the first unused block in the journal.
885 * [j_state_lock]
886 */
887 unsigned long j_head;
888
1980a53c
TH
889 /**
890 * @j_tail:
891 *
470decc6
DK
892 * Journal tail: identifies the oldest still-used block in the journal.
893 * [j_state_lock]
894 */
895 unsigned long j_tail;
896
1980a53c
TH
897 /**
898 * @j_free:
899 *
470decc6
DK
900 * Journal free: how many free blocks are there in the journal?
901 * [j_state_lock]
902 */
903 unsigned long j_free;
904
1980a53c
TH
905 /**
906 * @j_first:
907 *
908 * The block number of the first usable block in the journal
909 * [j_state_lock].
470decc6
DK
910 */
911 unsigned long j_first;
1980a53c
TH
912
913 /**
914 * @j_last:
915 *
916 * The block number one beyond the last usable block in the journal
917 * [j_state_lock].
918 */
470decc6
DK
919 unsigned long j_last;
920
1980a53c
TH
921 /**
922 * @j_dev: Device where we store the journal.
470decc6
DK
923 */
924 struct block_device *j_dev;
1980a53c
TH
925
926 /**
927 * @j_blocksize: Block size for the location where we store the journal.
928 */
470decc6 929 int j_blocksize;
1980a53c
TH
930
931 /**
932 * @j_blk_offset:
933 *
934 * Starting block offset into the device where we store the journal.
935 */
05496769 936 unsigned long long j_blk_offset;
1980a53c
TH
937
938 /**
939 * @j_devname: Journal device name.
940 */
05496769 941 char j_devname[BDEVNAME_SIZE+24];
470decc6 942
1980a53c
TH
943 /**
944 * @j_fs_dev:
945 *
470decc6
DK
946 * Device which holds the client fs. For internal journal this will be
947 * equal to j_dev.
948 */
949 struct block_device *j_fs_dev;
950
1980a53c
TH
951 /**
952 * @j_maxlen: Total maximum capacity of the journal region on disk.
953 */
470decc6
DK
954 unsigned int j_maxlen;
955
1980a53c
TH
956 /**
957 * @j_reserved_credits:
958 *
959 * Number of buffers reserved from the running transaction.
960 */
8f7d89f3
JK
961 atomic_t j_reserved_credits;
962
1980a53c
TH
963 /**
964 * @j_list_lock: Protects the buffer lists and internal buffer state.
470decc6
DK
965 */
966 spinlock_t j_list_lock;
967
1980a53c
TH
968 /**
969 * @j_inode:
970 *
971 * Optional inode where we store the journal. If present, all
972 * journal block numbers are mapped into this inode via bmap().
973 */
470decc6
DK
974 struct inode *j_inode;
975
1980a53c
TH
976 /**
977 * @j_tail_sequence:
978 *
470decc6
DK
979 * Sequence number of the oldest transaction in the log [j_state_lock]
980 */
981 tid_t j_tail_sequence;
982
1980a53c
TH
983 /**
984 * @j_transaction_sequence:
985 *
470decc6
DK
986 * Sequence number of the next transaction to grant [j_state_lock]
987 */
988 tid_t j_transaction_sequence;
989
1980a53c
TH
990 /**
991 * @j_commit_sequence:
992 *
470decc6
DK
993 * Sequence number of the most recently committed transaction
994 * [j_state_lock].
995 */
996 tid_t j_commit_sequence;
997
1980a53c
TH
998 /**
999 * @j_commit_request:
1000 *
470decc6
DK
1001 * Sequence number of the most recent transaction wanting commit
1002 * [j_state_lock]
1003 */
1004 tid_t j_commit_request;
1005
1980a53c
TH
1006 /**
1007 * @j_uuid:
1008 *
470decc6
DK
1009 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1010 * backed by this journal. This will eventually be replaced by an array
1011 * of uuids, allowing us to index multiple devices within a single
1012 * journal and to perform atomic updates across them.
1013 */
1014 __u8 j_uuid[16];
1015
1980a53c
TH
1016 /**
1017 * @j_task: Pointer to the current commit thread for this journal.
1018 */
470decc6
DK
1019 struct task_struct *j_task;
1020
1980a53c
TH
1021 /**
1022 * @j_max_transaction_buffers:
1023 *
470decc6 1024 * Maximum number of metadata buffers to allow in a single compound
1980a53c 1025 * commit transaction.
470decc6
DK
1026 */
1027 int j_max_transaction_buffers;
1028
1980a53c
TH
1029 /**
1030 * @j_commit_interval:
1031 *
470decc6
DK
1032 * What is the maximum transaction lifetime before we begin a commit?
1033 */
1034 unsigned long j_commit_interval;
1035
1980a53c
TH
1036 /**
1037 * @j_commit_timer: The timer used to wakeup the commit thread.
1038 */
470decc6
DK
1039 struct timer_list j_commit_timer;
1040
1980a53c
TH
1041 /**
1042 * @j_revoke_lock: Protect the revoke table.
470decc6
DK
1043 */
1044 spinlock_t j_revoke_lock;
1980a53c
TH
1045
1046 /**
1047 * @j_revoke:
1048 *
1049 * The revoke table - maintains the list of revoked blocks in the
1050 * current transaction.
1051 */
f7f4bccb 1052 struct jbd2_revoke_table_s *j_revoke;
1980a53c
TH
1053
1054 /**
1055 * @j_revoke_table: Alternate revoke tables for j_revoke.
1056 */
f7f4bccb 1057 struct jbd2_revoke_table_s *j_revoke_table[2];
470decc6 1058
1980a53c
TH
1059 /**
1060 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
470decc6
DK
1061 */
1062 struct buffer_head **j_wbuf;
1980a53c
TH
1063
1064 /**
1065 * @j_wbufsize:
1066 *
1067 * Size of @j_wbuf array.
1068 */
470decc6
DK
1069 int j_wbufsize;
1070
1980a53c
TH
1071 /**
1072 * @j_last_sync_writer:
1073 *
1074 * The pid of the last person to run a synchronous operation
1075 * through the journal.
e07f7183 1076 */
470decc6
DK
1077 pid_t j_last_sync_writer;
1078
1980a53c
TH
1079 /**
1080 * @j_average_commit_time:
1081 *
1082 * The average amount of time in nanoseconds it takes to commit a
e07f7183
JB
1083 * transaction to disk. [j_state_lock]
1084 */
1085 u64 j_average_commit_time;
1086
1980a53c
TH
1087 /**
1088 * @j_min_batch_time:
1089 *
1090 * Minimum time that we should wait for additional filesystem operations
1091 * to get batched into a synchronous handle in microseconds.
30773840
TT
1092 */
1093 u32 j_min_batch_time;
1980a53c
TH
1094
1095 /**
1096 * @j_max_batch_time:
1097 *
1098 * Maximum time that we should wait for additional filesystem operations
1099 * to get batched into a synchronous handle in microseconds.
1100 */
30773840
TT
1101 u32 j_max_batch_time;
1102
1980a53c
TH
1103 /**
1104 * @j_commit_callback:
1105 *
1106 * This function is called when a transaction is closed.
1107 */
3e624fc7
TT
1108 void (*j_commit_callback)(journal_t *,
1109 transaction_t *);
1110
470decc6 1111 /*
8e85fb3f
JL
1112 * Journal statistics
1113 */
1980a53c
TH
1114
1115 /**
1116 * @j_history_lock: Protect the transactions statistics history.
1117 */
8e85fb3f 1118 spinlock_t j_history_lock;
1980a53c
TH
1119
1120 /**
1121 * @j_proc_entry: procfs entry for the jbd statistics directory.
1122 */
8e85fb3f 1123 struct proc_dir_entry *j_proc_entry;
1980a53c
TH
1124
1125 /**
1126 * @j_stats: Overall statistics.
1127 */
8e85fb3f
JL
1128 struct transaction_stats_s j_stats;
1129
1980a53c
TH
1130 /**
1131 * @j_failed_commit: Failed journal commit ID.
1132 */
624080ed
TT
1133 unsigned int j_failed_commit;
1134
1980a53c
TH
1135 /**
1136 * @j_private:
1137 *
470decc6 1138 * An opaque pointer to fs-private information. ext3 puts its
1980a53c 1139 * superblock pointer here.
470decc6
DK
1140 */
1141 void *j_private;
01b5adce 1142
1980a53c
TH
1143 /**
1144 * @j_chksum_driver:
1145 *
1146 * Reference to checksum algorithm driver via cryptoapi.
1147 */
01b5adce 1148 struct crypto_shash *j_chksum_driver;
4fd5ea43 1149
1980a53c
TH
1150 /**
1151 * @j_csum_seed:
1152 *
1153 * Precomputed journal UUID checksum for seeding other checksums.
1154 */
4fd5ea43 1155 __u32 j_csum_seed;
ab714aff
JK
1156
1157#ifdef CONFIG_DEBUG_LOCK_ALLOC
1980a53c
TH
1158 /**
1159 * @j_trans_commit_map:
1160 *
ab714aff
JK
1161 * Lockdep entity to track transaction commit dependencies. Handles
1162 * hold this "lock" for read, when we wait for commit, we acquire the
1163 * "lock" for writing. This matches the properties of jbd2 journalling
1164 * where the running transaction has to wait for all handles to be
1165 * dropped to commit that transaction and also acquiring a handle may
1166 * require transaction commit to finish.
1167 */
1168 struct lockdep_map j_trans_commit_map;
1169#endif
470decc6
DK
1170};
1171
1eaa566d
JK
1172#define jbd2_might_wait_for_commit(j) \
1173 do { \
1174 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1175 rwsem_release(&j->j_trans_commit_map, 1, _THIS_IP_); \
1176 } while (0)
1177
56316a0d
DW
1178/* journal feature predicate functions */
1179#define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1180static inline bool jbd2_has_feature_##name(journal_t *j) \
1181{ \
1182 return ((j)->j_format_version >= 2 && \
1183 ((j)->j_superblock->s_feature_compat & \
1184 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1185} \
1186static inline void jbd2_set_feature_##name(journal_t *j) \
1187{ \
1188 (j)->j_superblock->s_feature_compat |= \
1189 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1190} \
1191static inline void jbd2_clear_feature_##name(journal_t *j) \
1192{ \
1193 (j)->j_superblock->s_feature_compat &= \
1194 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1195}
1196
1197#define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1198static inline bool jbd2_has_feature_##name(journal_t *j) \
1199{ \
1200 return ((j)->j_format_version >= 2 && \
1201 ((j)->j_superblock->s_feature_ro_compat & \
1202 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1203} \
1204static inline void jbd2_set_feature_##name(journal_t *j) \
1205{ \
1206 (j)->j_superblock->s_feature_ro_compat |= \
1207 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1208} \
1209static inline void jbd2_clear_feature_##name(journal_t *j) \
1210{ \
1211 (j)->j_superblock->s_feature_ro_compat &= \
1212 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1213}
1214
1215#define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1216static inline bool jbd2_has_feature_##name(journal_t *j) \
1217{ \
1218 return ((j)->j_format_version >= 2 && \
1219 ((j)->j_superblock->s_feature_incompat & \
1220 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1221} \
1222static inline void jbd2_set_feature_##name(journal_t *j) \
1223{ \
1224 (j)->j_superblock->s_feature_incompat |= \
1225 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1226} \
1227static inline void jbd2_clear_feature_##name(journal_t *j) \
1228{ \
1229 (j)->j_superblock->s_feature_incompat &= \
1230 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1231}
1232
1233JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM)
1234
1235JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE)
1236JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
1237JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT)
1238JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2)
1239JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3)
1240
470decc6
DK
1241/*
1242 * Journal flag definitions
1243 */
f7f4bccb
MC
1244#define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1245#define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1246#define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1247#define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1248#define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1249#define JBD2_BARRIER 0x020 /* Use IDE barriers */
5bf5683a
HK
1250#define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1251 * data write error in ordered
1252 * mode */
4327ba52 1253#define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */
470decc6
DK
1254
1255/*
1256 * Function declarations for the journaling transaction and buffer
1257 * management
1258 */
1259
1260/* Filing buffers */
f7f4bccb 1261extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
f7f4bccb
MC
1262extern void __jbd2_journal_refile_buffer(struct journal_head *);
1263extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1264extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
470decc6 1265extern void __journal_free_buffer(struct journal_head *bh);
f7f4bccb 1266extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
470decc6 1267extern void __journal_clean_data_list(transaction_t *transaction);
f5113eff
JK
1268static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1269{
1270 list_add_tail(&bh->b_assoc_buffers, head);
1271}
1272static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1273{
1274 list_del_init(&bh->b_assoc_buffers);
1275}
470decc6
DK
1276
1277/* Log buffer allocation */
32ab6715 1278struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1101cd4d 1279void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
18eba7aa 1280int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
79feb521
JK
1281int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1282 unsigned long *block);
6f6a6fda 1283int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
3339578f 1284void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
470decc6
DK
1285
1286/* Commit management */
f7f4bccb 1287extern void jbd2_journal_commit_transaction(journal_t *);
470decc6
DK
1288
1289/* Checkpoint list management */
841df7df 1290void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
f7f4bccb 1291int __jbd2_journal_remove_checkpoint(struct journal_head *);
841df7df 1292void jbd2_journal_destroy_checkpoint(journal_t *journal);
f7f4bccb 1293void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
470decc6 1294
e06c8227
JB
1295
1296/*
1297 * Triggers
1298 */
1299
1300struct jbd2_buffer_trigger_type {
1301 /*
13ceef09
JK
1302 * Fired a the moment data to write to the journal are known to be
1303 * stable - so either at the moment b_frozen_data is created or just
1304 * before a buffer is written to the journal. mapped_data is a mapped
1305 * buffer that is the frozen data for commit.
e06c8227 1306 */
13ceef09 1307 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
e06c8227
JB
1308 struct buffer_head *bh, void *mapped_data,
1309 size_t size);
1310
1311 /*
1312 * Fired during journal abort for dirty buffers that will not be
1313 * committed.
1314 */
1315 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1316 struct buffer_head *bh);
1317};
1318
13ceef09 1319extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
e06c8227
JB
1320 void *mapped_data,
1321 struct jbd2_buffer_trigger_type *triggers);
1322extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1323 struct jbd2_buffer_trigger_type *triggers);
1324
470decc6 1325/* Buffer IO */
f5113eff
JK
1326extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1327 struct journal_head *jh_in,
1328 struct buffer_head **bh_out,
1329 sector_t blocknr);
470decc6
DK
1330
1331/* Transaction locking */
1332extern void __wait_on_journal (journal_t *);
1333
0c2022ec
YY
1334/* Transaction cache support */
1335extern void jbd2_journal_destroy_transaction_cache(void);
1336extern int jbd2_journal_init_transaction_cache(void);
1337extern void jbd2_journal_free_transaction(transaction_t *);
1338
470decc6
DK
1339/*
1340 * Journal locking.
1341 *
1342 * We need to lock the journal during transaction state changes so that nobody
1343 * ever tries to take a handle on the running transaction while we are in the
1344 * middle of moving it to the commit phase. j_state_lock does this.
1345 *
1346 * Note that the locking is completely interrupt unsafe. We never touch
1347 * journal structures from interrupts.
1348 */
1349
1350static inline handle_t *journal_current_handle(void)
1351{
1352 return current->journal_info;
1353}
1354
1355/* The journaling code user interface:
1356 *
1357 * Create and destroy handles
1358 * Register buffer modifications against the current transaction.
1359 */
1360
f7f4bccb 1361extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
8f7d89f3
JK
1362extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1363 gfp_t gfp_mask, unsigned int type,
1364 unsigned int line_no);
47def826 1365extern int jbd2_journal_restart(handle_t *, int nblocks);
d2159fb7 1366extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
8f7d89f3
JK
1367extern int jbd2_journal_start_reserved(handle_t *handle,
1368 unsigned int type, unsigned int line_no);
1369extern void jbd2_journal_free_reserved(handle_t *handle);
f7f4bccb
MC
1370extern int jbd2_journal_extend (handle_t *, int nblocks);
1371extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1372extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1373extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
e06c8227
JB
1374void jbd2_journal_set_triggers(struct buffer_head *,
1375 struct jbd2_buffer_trigger_type *type);
f7f4bccb 1376extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
f7f4bccb 1377extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
470decc6 1378extern void journal_sync_buffer (struct buffer_head *);
53e87268 1379extern int jbd2_journal_invalidatepage(journal_t *,
259709b0 1380 struct page *, unsigned int, unsigned int);
f7f4bccb
MC
1381extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1382extern int jbd2_journal_stop(handle_t *);
1383extern int jbd2_journal_flush (journal_t *);
1384extern void jbd2_journal_lock_updates (journal_t *);
1385extern void jbd2_journal_unlock_updates (journal_t *);
470decc6 1386
f7f4bccb 1387extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
470decc6 1388 struct block_device *fs_dev,
18eba7aa 1389 unsigned long long start, int len, int bsize);
f7f4bccb
MC
1390extern journal_t * jbd2_journal_init_inode (struct inode *);
1391extern int jbd2_journal_update_format (journal_t *);
1392extern int jbd2_journal_check_used_features
470decc6 1393 (journal_t *, unsigned long, unsigned long, unsigned long);
f7f4bccb 1394extern int jbd2_journal_check_available_features
470decc6 1395 (journal_t *, unsigned long, unsigned long, unsigned long);
f7f4bccb 1396extern int jbd2_journal_set_features
470decc6 1397 (journal_t *, unsigned long, unsigned long, unsigned long);
818d276c
GS
1398extern void jbd2_journal_clear_features
1399 (journal_t *, unsigned long, unsigned long, unsigned long);
f7f4bccb 1400extern int jbd2_journal_load (journal_t *journal);
44519faf 1401extern int jbd2_journal_destroy (journal_t *);
f7f4bccb
MC
1402extern int jbd2_journal_recover (journal_t *journal);
1403extern int jbd2_journal_wipe (journal_t *, int);
1404extern int jbd2_journal_skip_recovery (journal_t *);
d796c52e 1405extern void jbd2_journal_update_sb_errno(journal_t *);
6f6a6fda 1406extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
79feb521 1407 unsigned long, int);
f7f4bccb
MC
1408extern void __jbd2_journal_abort_hard (journal_t *);
1409extern void jbd2_journal_abort (journal_t *, int);
1410extern int jbd2_journal_errno (journal_t *);
1411extern void jbd2_journal_ack_err (journal_t *);
1412extern int jbd2_journal_clear_err (journal_t *);
18eba7aa 1413extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
f7f4bccb 1414extern int jbd2_journal_force_commit(journal_t *);
9ff86446 1415extern int jbd2_journal_force_commit_nested(journal_t *);
41617e1a
JK
1416extern int jbd2_journal_inode_add_write(handle_t *handle, struct jbd2_inode *inode);
1417extern int jbd2_journal_inode_add_wait(handle_t *handle, struct jbd2_inode *inode);
d55ca21d
RZ
1418extern int jbd2_journal_inode_ranged_write(handle_t *handle,
1419 struct jbd2_inode *inode, loff_t start_byte,
1420 loff_t length);
1421extern int jbd2_journal_inode_ranged_wait(handle_t *handle,
1422 struct jbd2_inode *inode, loff_t start_byte,
1423 loff_t length);
7f5aa215
JK
1424extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1425 struct jbd2_inode *inode, loff_t new_size);
c851ed54
JK
1426extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1427extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
470decc6
DK
1428
1429/*
1430 * journal_head management
1431 */
f7f4bccb
MC
1432struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1433struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
f7f4bccb 1434void jbd2_journal_put_journal_head(struct journal_head *jh);
470decc6
DK
1435
1436/*
1437 * handle management
1438 */
e18b890b 1439extern struct kmem_cache *jbd2_handle_cache;
470decc6 1440
af1e76d6 1441static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
470decc6 1442{
28daf4fa 1443 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
470decc6
DK
1444}
1445
af1e76d6 1446static inline void jbd2_free_handle(handle_t *handle)
470decc6 1447{
f7f4bccb 1448 kmem_cache_free(jbd2_handle_cache, handle);
470decc6
DK
1449}
1450
8aefcd55
TT
1451/*
1452 * jbd2_inode management (optional, for those file systems that want to use
1453 * dynamically allocated jbd2_inode structures)
1454 */
1455extern struct kmem_cache *jbd2_inode_cache;
1456
1457static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1458{
1459 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1460}
1461
1462static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1463{
1464 kmem_cache_free(jbd2_inode_cache, jinode);
1465}
1466
470decc6
DK
1467/* Primary revoke support */
1468#define JOURNAL_REVOKE_DEFAULT_HASH 256
f7f4bccb
MC
1469extern int jbd2_journal_init_revoke(journal_t *, int);
1470extern void jbd2_journal_destroy_revoke_caches(void);
1471extern int jbd2_journal_init_revoke_caches(void);
470decc6 1472
f7f4bccb 1473extern void jbd2_journal_destroy_revoke(journal_t *);
18eba7aa 1474extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
f7f4bccb 1475extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
9bcf976c
JK
1476extern void jbd2_journal_write_revoke_records(transaction_t *transaction,
1477 struct list_head *log_bufs);
470decc6
DK
1478
1479/* Recovery revoke support */
18eba7aa
MC
1480extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1481extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
f7f4bccb
MC
1482extern void jbd2_journal_clear_revoke(journal_t *);
1483extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1ba37268 1484extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
470decc6
DK
1485
1486/*
1487 * The log thread user interface:
1488 *
1489 * Request space in the current transaction, and force transaction commit
1490 * transitions on demand.
1491 */
1492
f7f4bccb
MC
1493int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1494int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1495int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
f7f4bccb 1496int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
b8a6176c 1497int jbd2_transaction_committed(journal_t *journal, tid_t tid);
d76a3a77 1498int jbd2_complete_transaction(journal_t *journal, tid_t tid);
f7f4bccb 1499int jbd2_log_do_checkpoint(journal_t *journal);
bbd2be36 1500int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
470decc6 1501
f7f4bccb 1502void __jbd2_log_wait_for_space(journal_t *journal);
fb68407b
AK
1503extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1504extern int jbd2_cleanup_journal_tail(journal_t *);
470decc6 1505
470decc6
DK
1506/*
1507 * is_journal_abort
1508 *
f7f4bccb 1509 * Simple test wrapper function to test the JBD2_ABORT state flag. This
470decc6
DK
1510 * bit, when set, indicates that we have had a fatal error somewhere,
1511 * either inside the journaling layer or indicated to us by the client
1512 * (eg. ext3), and that we and should not commit any further
1513 * transactions.
1514 */
1515
1516static inline int is_journal_aborted(journal_t *journal)
1517{
f7f4bccb 1518 return journal->j_flags & JBD2_ABORT;
470decc6
DK
1519}
1520
1521static inline int is_handle_aborted(handle_t *handle)
1522{
41a5b913 1523 if (handle->h_aborted || !handle->h_transaction)
470decc6
DK
1524 return 1;
1525 return is_journal_aborted(handle->h_transaction->t_journal);
1526}
1527
f7f4bccb 1528static inline void jbd2_journal_abort_handle(handle_t *handle)
470decc6
DK
1529{
1530 handle->h_aborted = 1;
1531}
1532
1533#endif /* __KERNEL__ */
1534
1535/* Comparison functions for transaction IDs: perform comparisons using
1536 * modulo arithmetic so that they work over sequence number wraps. */
1537
1538static inline int tid_gt(tid_t x, tid_t y)
1539{
1540 int difference = (x - y);
1541 return (difference > 0);
1542}
1543
1544static inline int tid_geq(tid_t x, tid_t y)
1545{
1546 int difference = (x - y);
1547 return (difference >= 0);
1548}
1549
f7f4bccb 1550extern int jbd2_journal_blocks_per_page(struct inode *inode);
b517bea1 1551extern size_t journal_tag_bytes(journal_t *journal);
470decc6 1552
8595798c
DW
1553static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1554{
56316a0d 1555 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
8595798c
DW
1556}
1557
db9ee220
DW
1558static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1559{
8595798c
DW
1560 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1561 journal->j_chksum_driver == NULL);
db9ee220 1562
8595798c 1563 return journal->j_chksum_driver != NULL;
db9ee220
DW
1564}
1565
76c39904
JK
1566/*
1567 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
1568 * transaction control blocks.
1569 */
1570#define JBD2_CONTROL_BLOCKS_SHIFT 5
1571
470decc6
DK
1572/*
1573 * Return the minimum number of blocks which must be free in the journal
1574 * before a new transaction may be started. Must be called under j_state_lock.
1575 */
76c39904 1576static inline int jbd2_space_needed(journal_t *journal)
470decc6
DK
1577{
1578 int nblocks = journal->j_max_transaction_buffers;
76c39904
JK
1579 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT);
1580}
1581
1582/*
1583 * Return number of free blocks in the log. Must be called under j_state_lock.
1584 */
1585static inline unsigned long jbd2_log_space_left(journal_t *journal)
1586{
1587 /* Allow for rounding errors */
32502f2d 1588 long free = journal->j_free - 32;
76c39904
JK
1589
1590 if (journal->j_committing_transaction) {
1591 unsigned long committing = atomic_read(&journal->
1592 j_committing_transaction->t_outstanding_credits);
1593
1594 /* Transaction + control blocks */
1595 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT);
1596 }
32502f2d 1597 return max_t(long, free, 0);
470decc6
DK
1598}
1599
1600/*
1601 * Definitions which augment the buffer_head layer
1602 */
1603
1604/* journaling buffer types */
1605#define BJ_None 0 /* Not journaled */
87c89c23
JK
1606#define BJ_Metadata 1 /* Normal journaled metadata */
1607#define BJ_Forget 2 /* Buffer superseded by this transaction */
f5113eff 1608#define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
e5a120ae
JK
1609#define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1610#define BJ_Types 5
470decc6
DK
1611
1612extern int jbd_blocks_per_page(struct inode *inode);
1613
fd47d3e1
BW
1614/* JBD uses a CRC32 checksum */
1615#define JBD_MAX_CHECKSUM_SIZE 4
1616
01b5adce
DW
1617static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1618 const void *address, unsigned int length)
1619{
1620 struct {
1621 struct shash_desc shash;
fd47d3e1 1622 char ctx[JBD_MAX_CHECKSUM_SIZE];
01b5adce
DW
1623 } desc;
1624 int err;
1625
fd47d3e1
BW
1626 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1627 JBD_MAX_CHECKSUM_SIZE);
1628
01b5adce
DW
1629 desc.shash.tfm = journal->j_chksum_driver;
1630 desc.shash.flags = 0;
1631 *(u32 *)desc.ctx = crc;
1632
1633 err = crypto_shash_update(&desc.shash, address, length);
1634 BUG_ON(err);
1635
1636 return *(u32 *)desc.ctx;
1637}
1638
06a407f1
DM
1639/* Return most recent uncommitted transaction */
1640static inline tid_t jbd2_get_latest_transaction(journal_t *journal)
1641{
1642 tid_t tid;
1643
1644 read_lock(&journal->j_state_lock);
1645 tid = journal->j_commit_request;
1646 if (journal->j_running_transaction)
1647 tid = journal->j_running_transaction->t_tid;
1648 read_unlock(&journal->j_state_lock);
1649 return tid;
1650}
1651
470decc6
DK
1652#ifdef __KERNEL__
1653
1654#define buffer_trace_init(bh) do {} while (0)
1655#define print_buffer_fields(bh) do {} while (0)
1656#define print_buffer_trace(bh) do {} while (0)
1657#define BUFFER_TRACE(bh, info) do {} while (0)
1658#define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1659#define JBUFFER_TRACE(jh, info) do {} while (0)
1660
1661#endif /* __KERNEL__ */
1662
6a797d27
DW
1663#define EFSBADCRC EBADMSG /* Bad CRC detected */
1664#define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
1665
cd02ff0b 1666#endif /* _LINUX_JBD2_H */