1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * linux/include/linux/jbd2.h
4 *
5 * Written by Stephen C. Tweedie <sct@redhat.com>
6 *
7 * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
8 *
9 * Definitions for transaction data structures for the buffer cache
10 * filesystem journaling support.
11 */
12
13 #ifndef _LINUX_JBD2_H
14 #define _LINUX_JBD2_H
15
16 /* Allow this file to be included directly into e2fsprogs */
17 #ifndef __KERNEL__
18 #include "jfs_compat.h"
19 #define JBD2_DEBUG
20 #else
21
22 #include <linux/types.h>
23 #include <linux/buffer_head.h>
24 #include <linux/journal-head.h>
25 #include <linux/stddef.h>
26 #include <linux/mutex.h>
27 #include <linux/timer.h>
28 #include <linux/slab.h>
29 #include <linux/bit_spinlock.h>
30 #include <linux/blkdev.h>
31 #include <crypto/hash.h>
32 #endif
33
34 #define journal_oom_retry 1
35
36 /*
37 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
38 * certain classes of error which can occur due to failed IOs. Under
39 * normal use we want ext4 to continue after such errors, because
40 * hardware _can_ fail, but for debugging purposes when running tests on
41 * known-good hardware we may want to trap these errors.
42 */
43 #undef JBD2_PARANOID_IOFAIL
44
45 /*
46 * The default maximum commit age, in seconds.
47 */
48 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
49
50 #ifdef CONFIG_JBD2_DEBUG
51 /*
52 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
53 * consistency checks. By default we don't do this unless
54 * CONFIG_JBD2_DEBUG is on.
55 */
56 #define JBD2_EXPENSIVE_CHECKING
57 void __jbd2_debug(int level, const char *file, const char *func,
58 unsigned int line, const char *fmt, ...);
59
60 #define jbd2_debug(n, fmt, a...) \
61 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
62 #else
63 #define jbd2_debug(n, fmt, a...) no_printk(fmt, ##a)
64 #endif
65
66 extern void *jbd2_alloc(size_t size, gfp_t flags);
67 extern void jbd2_free(void *ptr, size_t size);
68
69 #define JBD2_MIN_JOURNAL_BLOCKS 1024
70 #define JBD2_DEFAULT_FAST_COMMIT_BLOCKS 256
71
72 #ifdef __KERNEL__
73
74 /**
75 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
76 *
77 * All filesystem modifications made by the process go
78 * through this handle. Recursive operations (such as quota operations)
79 * are gathered into a single update.
80 *
81 * The buffer credits field is used to account for journaled buffers
82 * being modified by the running process. To ensure that there is
83 * enough log space for all outstanding operations, we need to limit the
84 * number of outstanding buffers possible at any time. When the
85 * operation completes, any buffer credits not used are credited back to
86 * the transaction, so that at all times we know how many buffers the
87 * outstanding updates on a transaction might possibly touch.
88 *
89 * This is an opaque datatype.
90 **/
91 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
92
93
94 /**
95 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
96 *
97 * journal_t is linked to from the fs superblock structure.
98 *
99 * We use the journal_t to keep track of all outstanding transaction
100 * activity on the filesystem, and to manage the state of the log
101 * writing process.
102 *
103 * This is an opaque datatype.
104 **/
105 typedef struct journal_s journal_t; /* Journal control structure */
106 #endif
107
108 /*
109 * Internal structures used by the logging mechanism:
110 */
111
112 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
113
114 /*
115 * On-disk structures
116 */
117
118 /*
119 * Descriptor block types:
120 */
121
122 #define JBD2_DESCRIPTOR_BLOCK 1
123 #define JBD2_COMMIT_BLOCK 2
124 #define JBD2_SUPERBLOCK_V1 3
125 #define JBD2_SUPERBLOCK_V2 4
126 #define JBD2_REVOKE_BLOCK 5
127
128 /*
129 * Standard header for all descriptor blocks:
130 */
131 typedef struct journal_header_s
132 {
133 __be32 h_magic;
134 __be32 h_blocktype;
135 __be32 h_sequence;
136 } journal_header_t;
137
138 /*
139 * Checksum types.
140 */
141 #define JBD2_CRC32_CHKSUM 1
142 #define JBD2_MD5_CHKSUM 2
143 #define JBD2_SHA1_CHKSUM 3
144 #define JBD2_CRC32C_CHKSUM 4
145
146 #define JBD2_CRC32_CHKSUM_SIZE 4
147
148 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
149 /*
150 * Commit block header for storing transactional checksums:
151 *
152 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
153 * fields are used to store a checksum of the descriptor and data blocks.
154 *
155 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
156 * field is used to store crc32c(uuid+commit_block). Each journal metadata
157 * block gets its own checksum, and data block checksums are stored in
158 * journal_block_tag (in the descriptor). The other h_chksum* fields are
159 * not used.
160 *
161 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
162 * journal_block_tag3_t to store a full 32-bit checksum. Everything else
163 * is the same as v2.
164 *
165 * Checksum v1, v2, and v3 are mutually exclusive features.
166 */
167 struct commit_header {
168 __be32 h_magic;
169 __be32 h_blocktype;
170 __be32 h_sequence;
171 unsigned char h_chksum_type;
172 unsigned char h_chksum_size;
173 unsigned char h_padding[2];
174 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
175 __be64 h_commit_sec;
176 __be32 h_commit_nsec;
177 };
178
179 /*
180 * The block tag: used to describe a single buffer in the journal.
181 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
182 * raw struct shouldn't be used for pointer math or sizeof() - use
183 * journal_tag_bytes(journal) instead to compute this.
184 */
185 typedef struct journal_block_tag3_s
186 {
187 __be32 t_blocknr; /* The on-disk block number */
188 __be32 t_flags; /* See below */
189 __be32 t_blocknr_high; /* most-significant high 32bits. */
190 __be32 t_checksum; /* crc32c(uuid+seq+block) */
191 } journal_block_tag3_t;
192
193 typedef struct journal_block_tag_s
194 {
195 __be32 t_blocknr; /* The on-disk block number */
196 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
197 __be16 t_flags; /* See below */
198 __be32 t_blocknr_high; /* most-significant high 32bits. */
199 } journal_block_tag_t;
200
201 /* Tail of descriptor or revoke block, for checksumming */
202 struct jbd2_journal_block_tail {
203 __be32 t_checksum; /* crc32c(uuid+descr_block) */
204 };
205
206 /*
207 * The revoke descriptor: used on disk to describe a series of blocks to
208 * be revoked from the log
209 */
210 typedef struct jbd2_journal_revoke_header_s
211 {
212 journal_header_t r_header;
213 __be32 r_count; /* Count of bytes used in the block */
214 } jbd2_journal_revoke_header_t;
215
216 /* Definitions for the journal tag flags word: */
217 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
218 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
219 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
220 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
221
222
223 /*
224 * The journal superblock. All fields are in big-endian byte order.
225 */
226 typedef struct journal_superblock_s
227 {
228 /* 0x0000 */
229 journal_header_t s_header;
230
231 /* 0x000C */
232 /* Static information describing the journal */
233 __be32 s_blocksize; /* journal device blocksize */
234 __be32 s_maxlen; /* total blocks in journal file */
235 __be32 s_first; /* first block of log information */
236
237 /* 0x0018 */
238 /* Dynamic information describing the current state of the log */
239 __be32 s_sequence; /* first commit ID expected in log */
240 __be32 s_start; /* blocknr of start of log */
241
242 /* 0x0020 */
243 /* Error value, as set by jbd2_journal_abort(). */
244 __be32 s_errno;
245
246 /* 0x0024 */
247 /* Remaining fields are only valid in a version-2 superblock */
248 __be32 s_feature_compat; /* compatible feature set */
249 __be32 s_feature_incompat; /* incompatible feature set */
250 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
251 /* 0x0030 */
252 __u8 s_uuid[16]; /* 128-bit uuid for journal */
253
254 /* 0x0040 */
255 __be32 s_nr_users; /* Nr of filesystems sharing log */
256
257 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
258
259 /* 0x0048 */
260 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
261 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
262
263 /* 0x0050 */
264 __u8 s_checksum_type; /* checksum type */
265 __u8 s_padding2[3];
266 /* 0x0054 */
267 __be32 s_num_fc_blks; /* Number of fast commit blocks */
268 __be32 s_head; /* blocknr of head of log, only uptodate
269 * while the filesystem is clean */
270 /* 0x005C */
271 __u32 s_padding[40];
272 __be32 s_checksum; /* crc32c(superblock) */
273
274 /* 0x0100 */
275 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
276 /* 0x0400 */
277 } journal_superblock_t;
278
279 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
280
281 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
282 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
283 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
284 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
285 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010
286 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT 0x00000020
287
288 /* See "journal feature predicate functions" below */
289
290 /* Features known to this kernel version: */
291 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
292 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
293 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
294 JBD2_FEATURE_INCOMPAT_64BIT | \
295 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
296 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
297 JBD2_FEATURE_INCOMPAT_CSUM_V3 | \
298 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)
299
300 #ifdef __KERNEL__
301
302 #include <linux/fs.h>
303 #include <linux/sched.h>
304
305 enum jbd_state_bits {
306 BH_JBD /* Has an attached ext3 journal_head */
307 = BH_PrivateStart,
308 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
309 BH_Freed, /* Has been freed (truncated) */
310 BH_Revoked, /* Has been revoked from the log */
311 BH_RevokeValid, /* Revoked flag is valid */
312 BH_JBDDirty, /* Is dirty but journaled */
313 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
314 BH_Shadow, /* IO on shadow buffer is running */
315 BH_Verified, /* Metadata block has been verified ok */
316 BH_JBDPrivateStart, /* First bit available for private use by FS */
317 };
318
BUFFER_FNS(JBD,jbd)319 BUFFER_FNS(JBD, jbd)
320 BUFFER_FNS(JWrite, jwrite)
321 BUFFER_FNS(JBDDirty, jbddirty)
322 TAS_BUFFER_FNS(JBDDirty, jbddirty)
323 BUFFER_FNS(Revoked, revoked)
324 TAS_BUFFER_FNS(Revoked, revoked)
325 BUFFER_FNS(RevokeValid, revokevalid)
326 TAS_BUFFER_FNS(RevokeValid, revokevalid)
327 BUFFER_FNS(Freed, freed)
328 BUFFER_FNS(Shadow, shadow)
329 BUFFER_FNS(Verified, verified)
330
331 static inline struct buffer_head *jh2bh(struct journal_head *jh)
332 {
333 return jh->b_bh;
334 }
335
bh2jh(struct buffer_head * bh)336 static inline struct journal_head *bh2jh(struct buffer_head *bh)
337 {
338 return bh->b_private;
339 }
340
jbd_lock_bh_journal_head(struct buffer_head * bh)341 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
342 {
343 bit_spin_lock(BH_JournalHead, &bh->b_state);
344 }
345
jbd_unlock_bh_journal_head(struct buffer_head * bh)346 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
347 {
348 bit_spin_unlock(BH_JournalHead, &bh->b_state);
349 }
350
351 #define J_ASSERT(assert) BUG_ON(!(assert))
352
353 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
354 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
355
356 #if defined(JBD2_PARANOID_IOFAIL)
357 #define J_EXPECT(expr, why...) J_ASSERT(expr)
358 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
359 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
360 #else
361 #define __journal_expect(expr, why...) \
362 ({ \
363 int val = (expr); \
364 if (!val) { \
365 printk(KERN_ERR \
366 "JBD2 unexpected failure: %s: %s;\n", \
367 __func__, #expr); \
368 printk(KERN_ERR why "\n"); \
369 } \
370 val; \
371 })
372 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
373 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
374 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
375 #endif
376
377 /* Flags in jbd_inode->i_flags */
378 #define __JI_COMMIT_RUNNING 0
379 #define __JI_WRITE_DATA 1
380 #define __JI_WAIT_DATA 2
381
382 /*
383 * Commit of the inode data in progress. We use this flag to protect us from
384 * concurrent deletion of inode. We cannot use reference to inode for this
385 * since we cannot afford doing last iput() on behalf of kjournald
386 */
387 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
388 /* Write allocated dirty buffers in this inode before commit */
389 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
390 /* Wait for outstanding data writes for this inode before commit */
391 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
392
393 /**
394 * struct jbd2_inode - The jbd_inode type is the structure linking inodes in
395 * ordered mode present in a transaction so that we can sync them during commit.
396 */
397 struct jbd2_inode {
398 /**
399 * @i_transaction:
400 *
401 * Which transaction does this inode belong to? Either the running
402 * transaction or the committing one. [j_list_lock]
403 */
404 transaction_t *i_transaction;
405
406 /**
407 * @i_next_transaction:
408 *
409 * Pointer to the running transaction modifying inode's data in case
410 * there is already a committing transaction touching it. [j_list_lock]
411 */
412 transaction_t *i_next_transaction;
413
414 /**
415 * @i_list: List of inodes in the i_transaction [j_list_lock]
416 */
417 struct list_head i_list;
418
419 /**
420 * @i_vfs_inode:
421 *
422 * VFS inode this inode belongs to [constant for lifetime of structure]
423 */
424 struct inode *i_vfs_inode;
425
426 /**
427 * @i_flags: Flags of inode [j_list_lock]
428 */
429 unsigned long i_flags;
430
431 /**
432 * @i_dirty_start:
433 *
434 * Offset in bytes where the dirty range for this inode starts.
435 * [j_list_lock]
436 */
437 loff_t i_dirty_start;
438
439 /**
440 * @i_dirty_end:
441 *
442 * Inclusive offset in bytes where the dirty range for this inode
443 * ends. [j_list_lock]
444 */
445 loff_t i_dirty_end;
446 };
447
448 struct jbd2_revoke_table_s;
449
450 /**
451 * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete
452 * type associated with handle_t.
453 * @h_transaction: Which compound transaction is this update a part of?
454 * @h_journal: Which journal handle belongs to - used iff h_reserved set.
455 * @h_rsv_handle: Handle reserved for finishing the logical operation.
456 * @h_total_credits: Number of remaining buffers we are allowed to add to
457 * journal. These are dirty buffers and revoke descriptor blocks.
458 * @h_revoke_credits: Number of remaining revoke records available for handle
459 * @h_ref: Reference count on this handle.
460 * @h_err: Field for caller's use to track errors through large fs operations.
461 * @h_sync: Flag for sync-on-close.
462 * @h_jdata: Flag to force data journaling.
463 * @h_reserved: Flag for handle for reserved credits.
464 * @h_aborted: Flag indicating fatal error on handle.
465 * @h_type: For handle statistics.
466 * @h_line_no: For handle statistics.
467 * @h_start_jiffies: Handle Start time.
468 * @h_requested_credits: Holds @h_total_credits after handle is started.
469 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
470 * @saved_alloc_context: Saved context while transaction is open.
471 **/
472
473 /* Docbook can't yet cope with the bit fields, but will leave the documentation
474 * in so it can be fixed later.
475 */
476
477 struct jbd2_journal_handle
478 {
479 union {
480 transaction_t *h_transaction;
481 /* Which journal handle belongs to - used iff h_reserved set */
482 journal_t *h_journal;
483 };
484
485 handle_t *h_rsv_handle;
486 int h_total_credits;
487 int h_revoke_credits;
488 int h_revoke_credits_requested;
489 int h_ref;
490 int h_err;
491
492 /* Flags [no locking] */
493 unsigned int h_sync: 1;
494 unsigned int h_jdata: 1;
495 unsigned int h_reserved: 1;
496 unsigned int h_aborted: 1;
497 unsigned int h_type: 8;
498 unsigned int h_line_no: 16;
499
500 unsigned long h_start_jiffies;
501 unsigned int h_requested_credits;
502
503 unsigned int saved_alloc_context;
504 };
505
506
507 /*
508 * Some stats for checkpoint phase
509 */
510 struct transaction_chp_stats_s {
511 unsigned long cs_chp_time;
512 __u32 cs_forced_to_close;
513 __u32 cs_written;
514 __u32 cs_dropped;
515 };
516
517 /* The transaction_t type is the guts of the journaling mechanism. It
518 * tracks a compound transaction through its various states:
519 *
520 * RUNNING: accepting new updates
521 * LOCKED: Updates still running but we don't accept new ones
522 * RUNDOWN: Updates are tidying up but have finished requesting
523 * new buffers to modify (state not used for now)
524 * FLUSH: All updates complete, but we are still writing to disk
525 * COMMIT: All data on disk, writing commit record
526 * FINISHED: We still have to keep the transaction for checkpointing.
527 *
528 * The transaction keeps track of all of the buffers modified by a
529 * running transaction, and all of the buffers committed but not yet
530 * flushed to home for finished transactions.
531 * (Locking Documentation improved by LockDoc)
532 */
533
534 /*
535 * Lock ranking:
536 *
537 * j_list_lock
538 * ->jbd_lock_bh_journal_head() (This is "innermost")
539 *
540 * j_state_lock
541 * ->b_state_lock
542 *
543 * b_state_lock
544 * ->j_list_lock
545 *
546 * j_state_lock
547 * ->j_list_lock (journal_unmap_buffer)
548 *
549 */
550
551 struct transaction_s
552 {
553 /* Pointer to the journal for this transaction. [no locking] */
554 journal_t *t_journal;
555
556 /* Sequence number for this transaction [no locking] */
557 tid_t t_tid;
558
559 /*
560 * Transaction's current state
561 * [no locking - only kjournald2 alters this]
562 * [j_list_lock] guards transition of a transaction into T_FINISHED
563 * state and subsequent call of __jbd2_journal_drop_transaction()
564 * FIXME: needs barriers
565 * KLUDGE: [use j_state_lock]
566 */
567 enum {
568 T_RUNNING,
569 T_LOCKED,
570 T_SWITCH,
571 T_FLUSH,
572 T_COMMIT,
573 T_COMMIT_DFLUSH,
574 T_COMMIT_JFLUSH,
575 T_COMMIT_CALLBACK,
576 T_FINISHED
577 } t_state;
578
579 /*
580 * Where in the log does this transaction's commit start? [no locking]
581 */
582 unsigned long t_log_start;
583
584 /*
585 * Number of buffers on the t_buffers list [j_list_lock, no locks
586 * needed for jbd2 thread]
587 */
588 int t_nr_buffers;
589
590 /*
591 * Doubly-linked circular list of all buffers reserved but not yet
592 * modified by this transaction [j_list_lock, no locks needed fo
593 * jbd2 thread]
594 */
595 struct journal_head *t_reserved_list;
596
597 /*
598 * Doubly-linked circular list of all metadata buffers owned by this
599 * transaction [j_list_lock, no locks needed for jbd2 thread]
600 */
601 struct journal_head *t_buffers;
602
603 /*
604 * Doubly-linked circular list of all forget buffers (superseded
605 * buffers which we can un-checkpoint once this transaction commits)
606 * [j_list_lock]
607 */
608 struct journal_head *t_forget;
609
610 /*
611 * Doubly-linked circular list of all buffers still to be flushed before
612 * this transaction can be checkpointed. [j_list_lock]
613 */
614 struct journal_head *t_checkpoint_list;
615
616 /*
617 * Doubly-linked circular list of metadata buffers being
618 * shadowed by log IO. The IO buffers on the iobuf list and
619 * the shadow buffers on this list match each other one for
620 * one at all times. [j_list_lock, no locks needed for jbd2
621 * thread]
622 */
623 struct journal_head *t_shadow_list;
624
625 /*
626 * List of inodes associated with the transaction; e.g., ext4 uses
627 * this to track inodes in data=ordered and data=journal mode that
628 * need special handling on transaction commit; also used by ocfs2.
629 * [j_list_lock]
630 */
631 struct list_head t_inode_list;
632
633 /*
634 * Longest time some handle had to wait for running transaction
635 */
636 unsigned long t_max_wait;
637
638 /*
639 * When transaction started
640 */
641 unsigned long t_start;
642
643 /*
644 * When commit was requested [j_state_lock]
645 */
646 unsigned long t_requested;
647
648 /*
649 * Checkpointing stats [j_list_lock]
650 */
651 struct transaction_chp_stats_s t_chp_stats;
652
653 /*
654 * Number of outstanding updates running on this transaction
655 * [none]
656 */
657 atomic_t t_updates;
658
659 /*
660 * Number of blocks reserved for this transaction in the journal.
661 * This is including all credits reserved when starting transaction
662 * handles as well as all journal descriptor blocks needed for this
663 * transaction. [none]
664 */
665 atomic_t t_outstanding_credits;
666
667 /*
668 * Number of revoke records for this transaction added by already
669 * stopped handles. [none]
670 */
671 atomic_t t_outstanding_revokes;
672
673 /*
674 * How many handles used this transaction? [none]
675 */
676 atomic_t t_handle_count;
677
678 /*
679 * Forward and backward links for the circular list of all transactions
680 * awaiting checkpoint. [j_list_lock]
681 */
682 transaction_t *t_cpnext, *t_cpprev;
683
684 /*
685 * When will the transaction expire (become due for commit), in jiffies?
686 * [no locking]
687 */
688 unsigned long t_expires;
689
690 /*
691 * When this transaction started, in nanoseconds [no locking]
692 */
693 ktime_t t_start_time;
694
695 /*
696 * This transaction is being forced and some process is
697 * waiting for it to finish.
698 */
699 unsigned int t_synchronous_commit:1;
700
701 /* Disk flush needs to be sent to fs partition [no locking] */
702 int t_need_data_flush;
703
704 /*
705 * For use by the filesystem to store fs-specific data
706 * structures associated with the transaction
707 */
708 struct list_head t_private_list;
709 };
710
711 struct transaction_run_stats_s {
712 unsigned long rs_wait;
713 unsigned long rs_request_delay;
714 unsigned long rs_running;
715 unsigned long rs_locked;
716 unsigned long rs_flushing;
717 unsigned long rs_logging;
718
719 __u32 rs_handle_count;
720 __u32 rs_blocks;
721 __u32 rs_blocks_logged;
722 };
723
724 struct transaction_stats_s {
725 unsigned long ts_tid;
726 unsigned long ts_requested;
727 struct transaction_run_stats_s run;
728 };
729
730 static inline unsigned long
jbd2_time_diff(unsigned long start,unsigned long end)731 jbd2_time_diff(unsigned long start, unsigned long end)
732 {
733 if (end >= start)
734 return end - start;
735
736 return end + (MAX_JIFFY_OFFSET - start);
737 }
738
739 #define JBD2_NR_BATCH 64
740
741 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
742
743 #define JBD2_FC_REPLAY_STOP 0
744 #define JBD2_FC_REPLAY_CONTINUE 1
745
746 /**
747 * struct journal_s - The journal_s type is the concrete type associated with
748 * journal_t.
749 */
750 struct journal_s
751 {
752 /**
753 * @j_flags: General journaling state flags [j_state_lock,
754 * no lock for quick racy checks]
755 */
756 unsigned long j_flags;
757
758 /**
759 * @j_errno:
760 *
761 * Is there an outstanding uncleared error on the journal (from a prior
762 * abort)? [j_state_lock]
763 */
764 int j_errno;
765
766 /**
767 * @j_abort_mutex: Lock the whole aborting procedure.
768 */
769 struct mutex j_abort_mutex;
770
771 /**
772 * @j_sb_buffer: The first part of the superblock buffer.
773 */
774 struct buffer_head *j_sb_buffer;
775
776 /**
777 * @j_superblock: The second part of the superblock buffer.
778 */
779 journal_superblock_t *j_superblock;
780
781 /**
782 * @j_state_lock: Protect the various scalars in the journal.
783 */
784 rwlock_t j_state_lock;
785
786 /**
787 * @j_barrier_count:
788 *
789 * Number of processes waiting to create a barrier lock [j_state_lock,
790 * no lock for quick racy checks]
791 */
792 int j_barrier_count;
793
794 /**
795 * @j_barrier: The barrier lock itself.
796 */
797 struct mutex j_barrier;
798
799 /**
800 * @j_running_transaction:
801 *
802 * Transactions: The current running transaction...
803 * [j_state_lock, no lock for quick racy checks] [caller holding
804 * open handle]
805 */
806 transaction_t *j_running_transaction;
807
808 /**
809 * @j_committing_transaction:
810 *
811 * the transaction we are pushing to disk
812 * [j_state_lock] [caller holding open handle]
813 */
814 transaction_t *j_committing_transaction;
815
816 /**
817 * @j_checkpoint_transactions:
818 *
819 * ... and a linked circular list of all transactions waiting for
820 * checkpointing. [j_list_lock]
821 */
822 transaction_t *j_checkpoint_transactions;
823
824 /**
825 * @j_wait_transaction_locked:
826 *
827 * Wait queue for waiting for a locked transaction to start committing,
828 * or for a barrier lock to be released.
829 */
830 wait_queue_head_t j_wait_transaction_locked;
831
832 /**
833 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
834 */
835 wait_queue_head_t j_wait_done_commit;
836
837 /**
838 * @j_wait_commit: Wait queue to trigger commit.
839 */
840 wait_queue_head_t j_wait_commit;
841
842 /**
843 * @j_wait_updates: Wait queue to wait for updates to complete.
844 */
845 wait_queue_head_t j_wait_updates;
846
847 /**
848 * @j_wait_reserved:
849 *
850 * Wait queue to wait for reserved buffer credits to drop.
851 */
852 wait_queue_head_t j_wait_reserved;
853
854 /**
855 * @j_fc_wait:
856 *
857 * Wait queue to wait for completion of async fast commits.
858 */
859 wait_queue_head_t j_fc_wait;
860
861 /**
862 * @j_checkpoint_mutex:
863 *
864 * Semaphore for locking against concurrent checkpoints.
865 */
866 struct mutex j_checkpoint_mutex;
867
868 /**
869 * @j_chkpt_bhs:
870 *
871 * List of buffer heads used by the checkpoint routine. This
872 * was moved from jbd2_log_do_checkpoint() to reduce stack
873 * usage. Access to this array is controlled by the
874 * @j_checkpoint_mutex. [j_checkpoint_mutex]
875 */
876 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
877
878 /**
879 * @j_shrinker:
880 *
881 * Journal head shrinker, reclaim buffer's journal head which
882 * has been written back.
883 */
884 struct shrinker *j_shrinker;
885
886 /**
887 * @j_checkpoint_jh_count:
888 *
889 * Number of journal buffers on the checkpoint list. [j_list_lock]
890 */
891 struct percpu_counter j_checkpoint_jh_count;
892
893 /**
894 * @j_shrink_transaction:
895 *
896 * Record next transaction will shrink on the checkpoint list.
897 * [j_list_lock]
898 */
899 transaction_t *j_shrink_transaction;
900
901 /**
902 * @j_head:
903 *
904 * Journal head: identifies the first unused block in the journal.
905 * [j_state_lock]
906 */
907 unsigned long j_head;
908
909 /**
910 * @j_tail:
911 *
912 * Journal tail: identifies the oldest still-used block in the journal.
913 * [j_state_lock]
914 */
915 unsigned long j_tail;
916
917 /**
918 * @j_free:
919 *
920 * Journal free: how many free blocks are there in the journal?
921 * [j_state_lock]
922 */
923 unsigned long j_free;
924
925 /**
926 * @j_first:
927 *
928 * The block number of the first usable block in the journal
929 * [j_state_lock].
930 */
931 unsigned long j_first;
932
933 /**
934 * @j_last:
935 *
936 * The block number one beyond the last usable block in the journal
937 * [j_state_lock].
938 */
939 unsigned long j_last;
940
941 /**
942 * @j_fc_first:
943 *
944 * The block number of the first fast commit block in the journal
945 * [j_state_lock].
946 */
947 unsigned long j_fc_first;
948
949 /**
950 * @j_fc_off:
951 *
952 * Number of fast commit blocks currently allocated. Accessed only
953 * during fast commit. Currently only process can do fast commit, so
954 * this field is not protected by any lock.
955 */
956 unsigned long j_fc_off;
957
958 /**
959 * @j_fc_last:
960 *
961 * The block number one beyond the last fast commit block in the journal
962 * [j_state_lock].
963 */
964 unsigned long j_fc_last;
965
966 /**
967 * @j_dev: Device where we store the journal.
968 */
969 struct block_device *j_dev;
970
971 /**
972 * @j_blocksize: Block size for the location where we store the journal.
973 */
974 int j_blocksize;
975
976 /**
977 * @j_blk_offset:
978 *
979 * Starting block offset into the device where we store the journal.
980 */
981 unsigned long long j_blk_offset;
982
983 /**
984 * @j_devname: Journal device name.
985 */
986 char j_devname[BDEVNAME_SIZE+24];
987
988 /**
989 * @j_fs_dev:
990 *
991 * Device which holds the client fs. For internal journal this will be
992 * equal to j_dev.
993 */
994 struct block_device *j_fs_dev;
995
996 /**
997 * @j_fs_dev_wb_err:
998 *
999 * Records the errseq of the client fs's backing block device.
1000 */
1001 errseq_t j_fs_dev_wb_err;
1002
1003 /**
1004 * @j_total_len: Total maximum capacity of the journal region on disk.
1005 */
1006 unsigned int j_total_len;
1007
1008 /**
1009 * @j_reserved_credits:
1010 *
1011 * Number of buffers reserved from the running transaction.
1012 */
1013 atomic_t j_reserved_credits;
1014
1015 /**
1016 * @j_list_lock: Protects the buffer lists and internal buffer state.
1017 */
1018 spinlock_t j_list_lock;
1019
1020 /**
1021 * @j_inode:
1022 *
1023 * Optional inode where we store the journal. If present, all
1024 * journal block numbers are mapped into this inode via bmap().
1025 */
1026 struct inode *j_inode;
1027
1028 /**
1029 * @j_tail_sequence:
1030 *
1031 * Sequence number of the oldest transaction in the log [j_state_lock]
1032 */
1033 tid_t j_tail_sequence;
1034
1035 /**
1036 * @j_transaction_sequence:
1037 *
1038 * Sequence number of the next transaction to grant [j_state_lock]
1039 */
1040 tid_t j_transaction_sequence;
1041
1042 /**
1043 * @j_commit_sequence:
1044 *
1045 * Sequence number of the most recently committed transaction
1046 * [j_state_lock, no lock for quick racy checks]
1047 */
1048 tid_t j_commit_sequence;
1049
1050 /**
1051 * @j_commit_request:
1052 *
1053 * Sequence number of the most recent transaction wanting commit
1054 * [j_state_lock, no lock for quick racy checks]
1055 */
1056 tid_t j_commit_request;
1057
1058 /**
1059 * @j_uuid:
1060 *
1061 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1062 * backed by this journal. This will eventually be replaced by an array
1063 * of uuids, allowing us to index multiple devices within a single
1064 * journal and to perform atomic updates across them.
1065 */
1066 __u8 j_uuid[16];
1067
1068 /**
1069 * @j_task: Pointer to the current commit thread for this journal.
1070 */
1071 struct task_struct *j_task;
1072
1073 /**
1074 * @j_max_transaction_buffers:
1075 *
1076 * Maximum number of metadata buffers to allow in a single compound
1077 * commit transaction.
1078 */
1079 int j_max_transaction_buffers;
1080
1081 /**
1082 * @j_revoke_records_per_block:
1083 *
1084 * Number of revoke records that fit in one descriptor block.
1085 */
1086 int j_revoke_records_per_block;
1087
1088 /**
1089 * @j_transaction_overhead_buffers:
1090 *
1091 * Number of blocks each transaction needs for its own bookkeeping
1092 */
1093 int j_transaction_overhead_buffers;
1094
1095 /**
1096 * @j_commit_interval:
1097 *
1098 * What is the maximum transaction lifetime before we begin a commit?
1099 */
1100 unsigned long j_commit_interval;
1101
1102 /**
1103 * @j_commit_timer: The timer used to wakeup the commit thread.
1104 */
1105 struct timer_list j_commit_timer;
1106
1107 /**
1108 * @j_revoke_lock: Protect the revoke table.
1109 */
1110 spinlock_t j_revoke_lock;
1111
1112 /**
1113 * @j_revoke:
1114 *
1115 * The revoke table - maintains the list of revoked blocks in the
1116 * current transaction.
1117 */
1118 struct jbd2_revoke_table_s *j_revoke;
1119
1120 /**
1121 * @j_revoke_table: Alternate revoke tables for j_revoke.
1122 */
1123 struct jbd2_revoke_table_s *j_revoke_table[2];
1124
1125 /**
1126 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1127 */
1128 struct buffer_head **j_wbuf;
1129
1130 /**
1131 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only
1132 * during a fast commit. Currently only process can do fast commit, so
1133 * this field is not protected by any lock.
1134 */
1135 struct buffer_head **j_fc_wbuf;
1136
1137 /**
1138 * @j_wbufsize:
1139 *
1140 * Size of @j_wbuf array.
1141 */
1142 int j_wbufsize;
1143
1144 /**
1145 * @j_fc_wbufsize:
1146 *
1147 * Size of @j_fc_wbuf array.
1148 */
1149 int j_fc_wbufsize;
1150
1151 /**
1152 * @j_last_sync_writer:
1153 *
1154 * The pid of the last person to run a synchronous operation
1155 * through the journal.
1156 */
1157 pid_t j_last_sync_writer;
1158
1159 /**
1160 * @j_average_commit_time:
1161 *
1162 * The average amount of time in nanoseconds it takes to commit a
1163 * transaction to disk. [j_state_lock]
1164 */
1165 u64 j_average_commit_time;
1166
1167 /**
1168 * @j_min_batch_time:
1169 *
1170 * Minimum time that we should wait for additional filesystem operations
1171 * to get batched into a synchronous handle in microseconds.
1172 */
1173 u32 j_min_batch_time;
1174
1175 /**
1176 * @j_max_batch_time:
1177 *
1178 * Maximum time that we should wait for additional filesystem operations
1179 * to get batched into a synchronous handle in microseconds.
1180 */
1181 u32 j_max_batch_time;
1182
1183 /**
1184 * @j_commit_callback:
1185 *
1186 * This function is called when a transaction is closed.
1187 */
1188 void (*j_commit_callback)(journal_t *,
1189 transaction_t *);
1190
1191 /**
1192 * @j_submit_inode_data_buffers:
1193 *
1194 * This function is called for all inodes associated with the
1195 * committing transaction marked with JI_WRITE_DATA flag
1196 * before we start to write out the transaction to the journal.
1197 */
1198 int (*j_submit_inode_data_buffers)
1199 (struct jbd2_inode *);
1200
1201 /**
1202 * @j_finish_inode_data_buffers:
1203 *
1204 * This function is called for all inodes associated with the
1205 * committing transaction marked with JI_WAIT_DATA flag
1206 * after we have written the transaction to the journal
1207 * but before we write out the commit block.
1208 */
1209 int (*j_finish_inode_data_buffers)
1210 (struct jbd2_inode *);
1211
1212 /*
1213 * Journal statistics
1214 */
1215
1216 /**
1217 * @j_history_lock: Protect the transactions statistics history.
1218 */
1219 spinlock_t j_history_lock;
1220
1221 /**
1222 * @j_proc_entry: procfs entry for the jbd statistics directory.
1223 */
1224 struct proc_dir_entry *j_proc_entry;
1225
1226 /**
1227 * @j_stats: Overall statistics.
1228 */
1229 struct transaction_stats_s j_stats;
1230
1231 /**
1232 * @j_failed_commit: Failed journal commit ID.
1233 */
1234 unsigned int j_failed_commit;
1235
1236 /**
1237 * @j_private:
1238 *
1239 * An opaque pointer to fs-private information. ext3 puts its
1240 * superblock pointer here.
1241 */
1242 void *j_private;
1243
1244 /**
1245 * @j_chksum_driver:
1246 *
1247 * Reference to checksum algorithm driver via cryptoapi.
1248 */
1249 struct crypto_shash *j_chksum_driver;
1250
1251 /**
1252 * @j_csum_seed:
1253 *
1254 * Precomputed journal UUID checksum for seeding other checksums.
1255 */
1256 __u32 j_csum_seed;
1257
1258 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1259 /**
1260 * @j_trans_commit_map:
1261 *
1262 * Lockdep entity to track transaction commit dependencies. Handles
1263 * hold this "lock" for read, when we wait for commit, we acquire the
1264 * "lock" for writing. This matches the properties of jbd2 journalling
1265 * where the running transaction has to wait for all handles to be
1266 * dropped to commit that transaction and also acquiring a handle may
1267 * require transaction commit to finish.
1268 */
1269 struct lockdep_map j_trans_commit_map;
1270 #endif
1271
1272 /**
1273 * @j_fc_cleanup_callback:
1274 *
1275 * Clean-up after fast commit or full commit. JBD2 calls this function
1276 * after every commit operation.
1277 */
1278 void (*j_fc_cleanup_callback)(struct journal_s *journal, int full, tid_t tid);
1279
1280 /**
1281 * @j_fc_replay_callback:
1282 *
1283 * File-system specific function that performs replay of a fast
1284 * commit. JBD2 calls this function for each fast commit block found in
1285 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1286 * to indicate that the block was processed correctly and more fast
1287 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1288 * indicates the end of replay (no more blocks remaining). A negative
1289 * return value indicates error.
1290 */
1291 int (*j_fc_replay_callback)(struct journal_s *journal,
1292 struct buffer_head *bh,
1293 enum passtype pass, int off,
1294 tid_t expected_commit_id);
1295
1296 /**
1297 * @j_bmap:
1298 *
1299 * Bmap function that should be used instead of the generic
1300 * VFS bmap function.
1301 */
1302 int (*j_bmap)(struct journal_s *journal, sector_t *block);
1303 };
1304
1305 #define jbd2_might_wait_for_commit(j) \
1306 do { \
1307 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1308 rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1309 } while (0)
1310
1311 /*
1312 * We can support any known requested features iff the
1313 * superblock is not in version 1. Otherwise we fail to support any
1314 * extended sb features.
1315 */
jbd2_format_support_feature(journal_t * j)1316 static inline bool jbd2_format_support_feature(journal_t *j)
1317 {
1318 return j->j_superblock->s_header.h_blocktype !=
1319 cpu_to_be32(JBD2_SUPERBLOCK_V1);
1320 }
1321
1322 /* journal feature predicate functions */
1323 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1324 static inline bool jbd2_has_feature_##name(journal_t *j) \
1325 { \
1326 return (jbd2_format_support_feature(j) && \
1327 ((j)->j_superblock->s_feature_compat & \
1328 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1329 } \
1330 static inline void jbd2_set_feature_##name(journal_t *j) \
1331 { \
1332 (j)->j_superblock->s_feature_compat |= \
1333 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1334 } \
1335 static inline void jbd2_clear_feature_##name(journal_t *j) \
1336 { \
1337 (j)->j_superblock->s_feature_compat &= \
1338 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1339 }
1340
1341 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1342 static inline bool jbd2_has_feature_##name(journal_t *j) \
1343 { \
1344 return (jbd2_format_support_feature(j) && \
1345 ((j)->j_superblock->s_feature_ro_compat & \
1346 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1347 } \
1348 static inline void jbd2_set_feature_##name(journal_t *j) \
1349 { \
1350 (j)->j_superblock->s_feature_ro_compat |= \
1351 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1352 } \
1353 static inline void jbd2_clear_feature_##name(journal_t *j) \
1354 { \
1355 (j)->j_superblock->s_feature_ro_compat &= \
1356 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1357 }
1358
1359 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1360 static inline bool jbd2_has_feature_##name(journal_t *j) \
1361 { \
1362 return (jbd2_format_support_feature(j) && \
1363 ((j)->j_superblock->s_feature_incompat & \
1364 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1365 } \
1366 static inline void jbd2_set_feature_##name(journal_t *j) \
1367 { \
1368 (j)->j_superblock->s_feature_incompat |= \
1369 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1370 } \
1371 static inline void jbd2_clear_feature_##name(journal_t *j) \
1372 { \
1373 (j)->j_superblock->s_feature_incompat &= \
1374 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1375 }
1376
1377 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM)
1378
1379 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE)
1380 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
1381 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT)
1382 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2)
1383 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3)
1384 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit, FAST_COMMIT)
1385
1386 /* Journal high priority write IO operation flags */
1387 #define JBD2_JOURNAL_REQ_FLAGS (REQ_META | REQ_SYNC | REQ_IDLE)
1388
1389 /*
1390 * Journal flag definitions
1391 */
1392 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1393 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1394 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1395 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1396 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1397 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
1398 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1399 * data write error in ordered
1400 * mode */
1401 #define JBD2_CYCLE_RECORD 0x080 /* Journal cycled record log on
1402 * clean and empty filesystem
1403 * logging area */
1404 #define JBD2_FAST_COMMIT_ONGOING 0x100 /* Fast commit is ongoing */
1405 #define JBD2_FULL_COMMIT_ONGOING 0x200 /* Full commit is ongoing */
1406 #define JBD2_JOURNAL_FLUSH_DISCARD 0x0001
1407 #define JBD2_JOURNAL_FLUSH_ZEROOUT 0x0002
1408 #define JBD2_JOURNAL_FLUSH_VALID (JBD2_JOURNAL_FLUSH_DISCARD | \
1409 JBD2_JOURNAL_FLUSH_ZEROOUT)
1410
1411 /*
1412 * Function declarations for the journaling transaction and buffer
1413 * management
1414 */
1415
1416 /* Filing buffers */
1417 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1418 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1419 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1420 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1421 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
jbd2_file_log_bh(struct list_head * head,struct buffer_head * bh)1422 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1423 {
1424 list_add_tail(&bh->b_assoc_buffers, head);
1425 }
jbd2_unfile_log_bh(struct buffer_head * bh)1426 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1427 {
1428 list_del_init(&bh->b_assoc_buffers);
1429 }
1430
1431 /* Log buffer allocation */
1432 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1433 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1434 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1435 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1436 unsigned long *block);
1437 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1438 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1439
1440 /* Commit management */
1441 extern void jbd2_journal_commit_transaction(journal_t *);
1442
1443 /* Checkpoint list management */
1444 enum jbd2_shrink_type {JBD2_SHRINK_DESTROY, JBD2_SHRINK_BUSY_STOP, JBD2_SHRINK_BUSY_SKIP};
1445
1446 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, enum jbd2_shrink_type type);
1447 unsigned long jbd2_journal_shrink_checkpoint_list(journal_t *journal, unsigned long *nr_to_scan);
1448 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1449 int jbd2_journal_try_remove_checkpoint(struct journal_head *jh);
1450 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1451 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1452
1453
1454 /*
1455 * Triggers
1456 */
1457
1458 struct jbd2_buffer_trigger_type {
1459 /*
1460 * Fired a the moment data to write to the journal are known to be
1461 * stable - so either at the moment b_frozen_data is created or just
1462 * before a buffer is written to the journal. mapped_data is a mapped
1463 * buffer that is the frozen data for commit.
1464 */
1465 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1466 struct buffer_head *bh, void *mapped_data,
1467 size_t size);
1468
1469 /*
1470 * Fired during journal abort for dirty buffers that will not be
1471 * committed.
1472 */
1473 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1474 struct buffer_head *bh);
1475 };
1476
1477 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1478 void *mapped_data,
1479 struct jbd2_buffer_trigger_type *triggers);
1480 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1481 struct jbd2_buffer_trigger_type *triggers);
1482
1483 /* Buffer IO */
1484 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1485 struct journal_head *jh_in,
1486 struct buffer_head **bh_out,
1487 sector_t blocknr);
1488
1489 /* Transaction cache support */
1490 extern void jbd2_journal_destroy_transaction_cache(void);
1491 extern int __init jbd2_journal_init_transaction_cache(void);
1492 extern void jbd2_journal_free_transaction(transaction_t *);
1493
1494 /*
1495 * Journal locking.
1496 *
1497 * We need to lock the journal during transaction state changes so that nobody
1498 * ever tries to take a handle on the running transaction while we are in the
1499 * middle of moving it to the commit phase. j_state_lock does this.
1500 *
1501 * Note that the locking is completely interrupt unsafe. We never touch
1502 * journal structures from interrupts.
1503 */
1504
journal_current_handle(void)1505 static inline handle_t *journal_current_handle(void)
1506 {
1507 return current->journal_info;
1508 }
1509
1510 /* The journaling code user interface:
1511 *
1512 * Create and destroy handles
1513 * Register buffer modifications against the current transaction.
1514 */
1515
1516 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1517 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1518 int revoke_records, gfp_t gfp_mask,
1519 unsigned int type, unsigned int line_no);
1520 extern int jbd2_journal_restart(handle_t *, int nblocks);
1521 extern int jbd2__journal_restart(handle_t *, int nblocks,
1522 int revoke_records, gfp_t gfp_mask);
1523 extern int jbd2_journal_start_reserved(handle_t *handle,
1524 unsigned int type, unsigned int line_no);
1525 extern void jbd2_journal_free_reserved(handle_t *handle);
1526 extern int jbd2_journal_extend(handle_t *handle, int nblocks,
1527 int revoke_records);
1528 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1529 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1530 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1531 void jbd2_journal_set_triggers(struct buffer_head *,
1532 struct jbd2_buffer_trigger_type *type);
1533 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1534 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1535 int jbd2_journal_invalidate_folio(journal_t *, struct folio *,
1536 size_t offset, size_t length);
1537 bool jbd2_journal_try_to_free_buffers(journal_t *journal, struct folio *folio);
1538 extern int jbd2_journal_stop(handle_t *);
1539 extern int jbd2_journal_flush(journal_t *journal, unsigned int flags);
1540 extern void jbd2_journal_lock_updates (journal_t *);
1541 extern void jbd2_journal_unlock_updates (journal_t *);
1542
1543 void jbd2_journal_wait_updates(journal_t *);
1544
1545 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1546 struct block_device *fs_dev,
1547 unsigned long long start, int len, int bsize);
1548 extern journal_t * jbd2_journal_init_inode (struct inode *);
1549 extern int jbd2_journal_update_format (journal_t *);
1550 extern int jbd2_journal_check_used_features
1551 (journal_t *, unsigned long, unsigned long, unsigned long);
1552 extern int jbd2_journal_check_available_features
1553 (journal_t *, unsigned long, unsigned long, unsigned long);
1554 extern int jbd2_journal_set_features
1555 (journal_t *, unsigned long, unsigned long, unsigned long);
1556 extern void jbd2_journal_clear_features
1557 (journal_t *, unsigned long, unsigned long, unsigned long);
1558 extern int jbd2_journal_load (journal_t *journal);
1559 extern int jbd2_journal_destroy (journal_t *);
1560 extern int jbd2_journal_recover (journal_t *journal);
1561 extern int jbd2_journal_wipe (journal_t *, int);
1562 extern int jbd2_journal_skip_recovery (journal_t *);
1563 extern void jbd2_journal_update_sb_errno(journal_t *);
1564 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
1565 unsigned long, blk_opf_t);
1566 extern void jbd2_journal_abort (journal_t *, int);
1567 extern int jbd2_journal_errno (journal_t *);
1568 extern void jbd2_journal_ack_err (journal_t *);
1569 extern int jbd2_journal_clear_err (journal_t *);
1570 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1571 extern int jbd2_journal_force_commit(journal_t *);
1572 extern int jbd2_journal_force_commit_nested(journal_t *);
1573 extern int jbd2_journal_inode_ranged_write(handle_t *handle,
1574 struct jbd2_inode *inode, loff_t start_byte,
1575 loff_t length);
1576 extern int jbd2_journal_inode_ranged_wait(handle_t *handle,
1577 struct jbd2_inode *inode, loff_t start_byte,
1578 loff_t length);
1579 extern int jbd2_journal_finish_inode_data_buffers(
1580 struct jbd2_inode *jinode);
1581 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1582 struct jbd2_inode *inode, loff_t new_size);
1583 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1584 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1585
1586 /*
1587 * journal_head management
1588 */
1589 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1590 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1591 void jbd2_journal_put_journal_head(struct journal_head *jh);
1592
1593 /*
1594 * handle management
1595 */
1596 extern struct kmem_cache *jbd2_handle_cache;
1597
1598 /*
1599 * This specialized allocator has to be a macro for its allocations to be
1600 * accounted separately (to have a separate alloc_tag). The typecast is
1601 * intentional to enforce typesafety.
1602 */
1603 #define jbd2_alloc_handle(_gfp_flags) \
1604 ((handle_t *)kmem_cache_zalloc(jbd2_handle_cache, _gfp_flags))
1605
jbd2_free_handle(handle_t * handle)1606 static inline void jbd2_free_handle(handle_t *handle)
1607 {
1608 kmem_cache_free(jbd2_handle_cache, handle);
1609 }
1610
1611 /*
1612 * jbd2_inode management (optional, for those file systems that want to use
1613 * dynamically allocated jbd2_inode structures)
1614 */
1615 extern struct kmem_cache *jbd2_inode_cache;
1616
1617 /*
1618 * This specialized allocator has to be a macro for its allocations to be
1619 * accounted separately (to have a separate alloc_tag). The typecast is
1620 * intentional to enforce typesafety.
1621 */
1622 #define jbd2_alloc_inode(_gfp_flags) \
1623 ((struct jbd2_inode *)kmem_cache_alloc(jbd2_inode_cache, _gfp_flags))
1624
jbd2_free_inode(struct jbd2_inode * jinode)1625 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1626 {
1627 kmem_cache_free(jbd2_inode_cache, jinode);
1628 }
1629
1630 /* Primary revoke support */
1631 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1632 extern int jbd2_journal_init_revoke(journal_t *, int);
1633 extern void jbd2_journal_destroy_revoke_record_cache(void);
1634 extern void jbd2_journal_destroy_revoke_table_cache(void);
1635 extern int __init jbd2_journal_init_revoke_record_cache(void);
1636 extern int __init jbd2_journal_init_revoke_table_cache(void);
1637
1638 extern void jbd2_journal_destroy_revoke(journal_t *);
1639 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1640 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1641 extern void jbd2_journal_write_revoke_records(transaction_t *transaction,
1642 struct list_head *log_bufs);
1643
1644 /* Recovery revoke support */
1645 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1646 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1647 extern void jbd2_journal_clear_revoke(journal_t *);
1648 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1649 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
1650
1651 /*
1652 * The log thread user interface:
1653 *
1654 * Request space in the current transaction, and force transaction commit
1655 * transitions on demand.
1656 */
1657
1658 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1659 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1660 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1661 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1662 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1663 int jbd2_log_do_checkpoint(journal_t *journal);
1664 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1665
1666 void __jbd2_log_wait_for_space(journal_t *journal);
1667 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1668 extern int jbd2_cleanup_journal_tail(journal_t *);
1669
1670 /* Fast commit related APIs */
1671 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1672 int jbd2_fc_end_commit(journal_t *journal);
1673 int jbd2_fc_end_commit_fallback(journal_t *journal);
1674 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1675 int jbd2_submit_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1676 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1677 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1678 void jbd2_fc_release_bufs(journal_t *journal);
1679
1680 /*
1681 * is_journal_abort
1682 *
1683 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1684 * bit, when set, indicates that we have had a fatal error somewhere,
1685 * either inside the journaling layer or indicated to us by the client
1686 * (eg. ext3), and that we and should not commit any further
1687 * transactions.
1688 */
1689
is_journal_aborted(journal_t * journal)1690 static inline int is_journal_aborted(journal_t *journal)
1691 {
1692 return journal->j_flags & JBD2_ABORT;
1693 }
1694
is_handle_aborted(handle_t * handle)1695 static inline int is_handle_aborted(handle_t *handle)
1696 {
1697 if (handle->h_aborted || !handle->h_transaction)
1698 return 1;
1699 return is_journal_aborted(handle->h_transaction->t_journal);
1700 }
1701
jbd2_journal_abort_handle(handle_t * handle)1702 static inline void jbd2_journal_abort_handle(handle_t *handle)
1703 {
1704 handle->h_aborted = 1;
1705 }
1706
jbd2_init_fs_dev_write_error(journal_t * journal)1707 static inline void jbd2_init_fs_dev_write_error(journal_t *journal)
1708 {
1709 struct address_space *mapping = journal->j_fs_dev->bd_mapping;
1710
1711 /*
1712 * Save the original wb_err value of client fs's bdev mapping which
1713 * could be used to detect the client fs's metadata async write error.
1714 */
1715 errseq_check_and_advance(&mapping->wb_err, &journal->j_fs_dev_wb_err);
1716 }
1717
jbd2_check_fs_dev_write_error(journal_t * journal)1718 static inline int jbd2_check_fs_dev_write_error(journal_t *journal)
1719 {
1720 struct address_space *mapping = journal->j_fs_dev->bd_mapping;
1721
1722 return errseq_check(&mapping->wb_err,
1723 READ_ONCE(journal->j_fs_dev_wb_err));
1724 }
1725
1726 #endif /* __KERNEL__ */
1727
1728 /* Comparison functions for transaction IDs: perform comparisons using
1729 * modulo arithmetic so that they work over sequence number wraps. */
1730
tid_gt(tid_t x,tid_t y)1731 static inline int tid_gt(tid_t x, tid_t y)
1732 {
1733 int difference = (x - y);
1734 return (difference > 0);
1735 }
1736
tid_geq(tid_t x,tid_t y)1737 static inline int tid_geq(tid_t x, tid_t y)
1738 {
1739 int difference = (x - y);
1740 return (difference >= 0);
1741 }
1742
1743 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1744 extern size_t journal_tag_bytes(journal_t *journal);
1745
jbd2_journal_has_csum_v2or3_feature(journal_t * j)1746 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1747 {
1748 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1749 }
1750
jbd2_journal_has_csum_v2or3(journal_t * journal)1751 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1752 {
1753 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1754 journal->j_chksum_driver == NULL);
1755
1756 return journal->j_chksum_driver != NULL;
1757 }
1758
jbd2_journal_get_num_fc_blks(journal_superblock_t * jsb)1759 static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb)
1760 {
1761 int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks);
1762
1763 return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS;
1764 }
1765
1766 /*
1767 * Return number of free blocks in the log. Must be called under j_state_lock.
1768 */
jbd2_log_space_left(journal_t * journal)1769 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1770 {
1771 /* Allow for rounding errors */
1772 long free = journal->j_free - 32;
1773
1774 if (journal->j_committing_transaction) {
1775 free -= atomic_read(&journal->
1776 j_committing_transaction->t_outstanding_credits);
1777 }
1778 return max_t(long, free, 0);
1779 }
1780
1781 /*
1782 * Definitions which augment the buffer_head layer
1783 */
1784
1785 /* journaling buffer types */
1786 #define BJ_None 0 /* Not journaled */
1787 #define BJ_Metadata 1 /* Normal journaled metadata */
1788 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1789 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
1790 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1791 #define BJ_Types 5
1792
1793 /* JBD uses a CRC32 checksum */
1794 #define JBD_MAX_CHECKSUM_SIZE 4
1795
jbd2_chksum(journal_t * journal,u32 crc,const void * address,unsigned int length)1796 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1797 const void *address, unsigned int length)
1798 {
1799 DEFINE_RAW_FLEX(struct shash_desc, desc, __ctx,
1800 DIV_ROUND_UP(JBD_MAX_CHECKSUM_SIZE,
1801 sizeof(*((struct shash_desc *)0)->__ctx)));
1802 int err;
1803
1804 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1805 JBD_MAX_CHECKSUM_SIZE);
1806
1807 desc->tfm = journal->j_chksum_driver;
1808 *(u32 *)desc->__ctx = crc;
1809
1810 err = crypto_shash_update(desc, address, length);
1811 BUG_ON(err);
1812
1813 return *(u32 *)desc->__ctx;
1814 }
1815
1816 /* Return most recent uncommitted transaction */
jbd2_get_latest_transaction(journal_t * journal)1817 static inline tid_t jbd2_get_latest_transaction(journal_t *journal)
1818 {
1819 tid_t tid;
1820
1821 read_lock(&journal->j_state_lock);
1822 tid = journal->j_commit_request;
1823 if (journal->j_running_transaction)
1824 tid = journal->j_running_transaction->t_tid;
1825 read_unlock(&journal->j_state_lock);
1826 return tid;
1827 }
1828
jbd2_handle_buffer_credits(handle_t * handle)1829 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1830 {
1831 journal_t *journal;
1832
1833 if (!handle->h_reserved)
1834 journal = handle->h_transaction->t_journal;
1835 else
1836 journal = handle->h_journal;
1837
1838 return handle->h_total_credits -
1839 DIV_ROUND_UP(handle->h_revoke_credits_requested,
1840 journal->j_revoke_records_per_block);
1841 }
1842
1843 #ifdef __KERNEL__
1844
1845 #define buffer_trace_init(bh) do {} while (0)
1846 #define print_buffer_fields(bh) do {} while (0)
1847 #define print_buffer_trace(bh) do {} while (0)
1848 #define BUFFER_TRACE(bh, info) do {} while (0)
1849 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1850 #define JBUFFER_TRACE(jh, info) do {} while (0)
1851
1852 #endif /* __KERNEL__ */
1853
1854 #define EFSBADCRC EBADMSG /* Bad CRC detected */
1855 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
1856
1857 #endif /* _LINUX_JBD2_H */
1858