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