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 <crypto/hash.h>
31 #endif
32
33 #define journal_oom_retry 1
34
35 /*
36 * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
37 * certain classes of error which can occur due to failed IOs. Under
38 * normal use we want ext4 to continue after such errors, because
39 * hardware _can_ fail, but for debugging purposes when running tests on
40 * known-good hardware we may want to trap these errors.
41 */
42 #undef JBD2_PARANOID_IOFAIL
43
44 /*
45 * The default maximum commit age, in seconds.
46 */
47 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
48
49 #ifdef CONFIG_JBD2_DEBUG
50 /*
51 * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
52 * consistency checks. By default we don't do this unless
53 * CONFIG_JBD2_DEBUG is on.
54 */
55 #define JBD2_EXPENSIVE_CHECKING
56 extern ushort jbd2_journal_enable_debug;
57 void __jbd2_debug(int level, const char *file, const char *func,
58 unsigned int line, const char *fmt, ...);
59
60 #define jbd_debug(n, fmt, a...) \
61 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
62 #else
63 #define jbd_debug(n, 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
71 #ifdef __KERNEL__
72
73 /**
74 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
75 *
76 * All filesystem modifications made by the process go
77 * through this handle. Recursive operations (such as quota operations)
78 * are gathered into a single update.
79 *
80 * The buffer credits field is used to account for journaled buffers
81 * being modified by the running process. To ensure that there is
82 * enough log space for all outstanding operations, we need to limit the
83 * number of outstanding buffers possible at any time. When the
84 * operation completes, any buffer credits not used are credited back to
85 * the transaction, so that at all times we know how many buffers the
86 * outstanding updates on a transaction might possibly touch.
87 *
88 * This is an opaque datatype.
89 **/
90 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
91
92
93 /**
94 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
95 *
96 * journal_t is linked to from the fs superblock structure.
97 *
98 * We use the journal_t to keep track of all outstanding transaction
99 * activity on the filesystem, and to manage the state of the log
100 * writing process.
101 *
102 * This is an opaque datatype.
103 **/
104 typedef struct journal_s journal_t; /* Journal control structure */
105 #endif
106
107 /*
108 * Internal structures used by the logging mechanism:
109 */
110
111 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
112
113 /*
114 * On-disk structures
115 */
116
117 /*
118 * Descriptor block types:
119 */
120
121 #define JBD2_DESCRIPTOR_BLOCK 1
122 #define JBD2_COMMIT_BLOCK 2
123 #define JBD2_SUPERBLOCK_V1 3
124 #define JBD2_SUPERBLOCK_V2 4
125 #define JBD2_REVOKE_BLOCK 5
126
127 /*
128 * Standard header for all descriptor blocks:
129 */
130 typedef struct journal_header_s
131 {
132 __be32 h_magic;
133 __be32 h_blocktype;
134 __be32 h_sequence;
135 } journal_header_t;
136
137 /*
138 * Checksum types.
139 */
140 #define JBD2_CRC32_CHKSUM 1
141 #define JBD2_MD5_CHKSUM 2
142 #define JBD2_SHA1_CHKSUM 3
143 #define JBD2_CRC32C_CHKSUM 4
144
145 #define JBD2_CRC32_CHKSUM_SIZE 4
146
147 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
148 /*
149 * Commit block header for storing transactional checksums:
150 *
151 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
152 * fields are used to store a checksum of the descriptor and data blocks.
153 *
154 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
155 * field is used to store crc32c(uuid+commit_block). Each journal metadata
156 * block gets its own checksum, and data block checksums are stored in
157 * journal_block_tag (in the descriptor). The other h_chksum* fields are
158 * not used.
159 *
160 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
161 * journal_block_tag3_t to store a full 32-bit checksum. Everything else
162 * is the same as v2.
163 *
164 * Checksum v1, v2, and v3 are mutually exclusive features.
165 */
166 struct commit_header {
167 __be32 h_magic;
168 __be32 h_blocktype;
169 __be32 h_sequence;
170 unsigned char h_chksum_type;
171 unsigned char h_chksum_size;
172 unsigned char h_padding[2];
173 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
174 __be64 h_commit_sec;
175 __be32 h_commit_nsec;
176 };
177
178 /*
179 * The block tag: used to describe a single buffer in the journal.
180 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
181 * raw struct shouldn't be used for pointer math or sizeof() - use
182 * journal_tag_bytes(journal) instead to compute this.
183 */
184 typedef struct journal_block_tag3_s
185 {
186 __be32 t_blocknr; /* The on-disk block number */
187 __be32 t_flags; /* See below */
188 __be32 t_blocknr_high; /* most-significant high 32bits. */
189 __be32 t_checksum; /* crc32c(uuid+seq+block) */
190 } journal_block_tag3_t;
191
192 typedef struct journal_block_tag_s
193 {
194 __be32 t_blocknr; /* The on-disk block number */
195 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
196 __be16 t_flags; /* See below */
197 __be32 t_blocknr_high; /* most-significant high 32bits. */
198 } journal_block_tag_t;
199
200 /* Tail of descriptor or revoke block, for checksumming */
201 struct jbd2_journal_block_tail {
202 __be32 t_checksum; /* crc32c(uuid+descr_block) */
203 };
204
205 /*
206 * The revoke descriptor: used on disk to describe a series of blocks to
207 * be revoked from the log
208 */
209 typedef struct jbd2_journal_revoke_header_s
210 {
211 journal_header_t r_header;
212 __be32 r_count; /* Count of bytes used in the block */
213 } jbd2_journal_revoke_header_t;
214
215 /* Definitions for the journal tag flags word: */
216 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
217 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
218 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
219 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
220
221
222 /*
223 * The journal superblock. All fields are in big-endian byte order.
224 */
225 typedef struct journal_superblock_s
226 {
227 /* 0x0000 */
228 journal_header_t s_header;
229
230 /* 0x000C */
231 /* Static information describing the journal */
232 __be32 s_blocksize; /* journal device blocksize */
233 __be32 s_maxlen; /* total blocks in journal file */
234 __be32 s_first; /* first block of log information */
235
236 /* 0x0018 */
237 /* Dynamic information describing the current state of the log */
238 __be32 s_sequence; /* first commit ID expected in log */
239 __be32 s_start; /* blocknr of start of log */
240
241 /* 0x0020 */
242 /* Error value, as set by jbd2_journal_abort(). */
243 __be32 s_errno;
244
245 /* 0x0024 */
246 /* Remaining fields are only valid in a version-2 superblock */
247 __be32 s_feature_compat; /* compatible feature set */
248 __be32 s_feature_incompat; /* incompatible feature set */
249 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
250 /* 0x0030 */
251 __u8 s_uuid[16]; /* 128-bit uuid for journal */
252
253 /* 0x0040 */
254 __be32 s_nr_users; /* Nr of filesystems sharing log */
255
256 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
257
258 /* 0x0048 */
259 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
260 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
261
262 /* 0x0050 */
263 __u8 s_checksum_type; /* checksum type */
264 __u8 s_padding2[3];
265 __u32 s_padding[42];
266 __be32 s_checksum; /* crc32c(superblock) */
267
268 /* 0x0100 */
269 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
270 /* 0x0400 */
271 } journal_superblock_t;
272
273 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
274 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
275 ((j)->j_format_version >= 2 && \
276 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
277 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
278 ((j)->j_format_version >= 2 && \
279 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
280 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
281 ((j)->j_format_version >= 2 && \
282 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
283
284 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
285
286 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
287 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
288 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
289 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
290 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010
291
292 /* See "journal feature predicate functions" below */
293
294 /* Features known to this kernel version: */
295 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
296 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
297 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
298 JBD2_FEATURE_INCOMPAT_64BIT | \
299 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
300 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
301 JBD2_FEATURE_INCOMPAT_CSUM_V3)
302
303 #ifdef __KERNEL__
304
305 #include <linux/fs.h>
306 #include <linux/sched.h>
307
308 enum jbd_state_bits {
309 BH_JBD /* Has an attached ext3 journal_head */
310 = BH_PrivateStart,
311 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
312 BH_Freed, /* Has been freed (truncated) */
313 BH_Revoked, /* Has been revoked from the log */
314 BH_RevokeValid, /* Revoked flag is valid */
315 BH_JBDDirty, /* Is dirty but journaled */
316 BH_State, /* Pins most journal_head state */
317 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
318 BH_Shadow, /* IO on shadow buffer is running */
319 BH_Verified, /* Metadata block has been verified ok */
320 BH_JBDPrivateStart, /* First bit available for private use by FS */
321 };
322
323 BUFFER_FNS(JBD, jbd)
324 BUFFER_FNS(JWrite, jwrite)
325 BUFFER_FNS(JBDDirty, jbddirty)
326 TAS_BUFFER_FNS(JBDDirty, jbddirty)
327 BUFFER_FNS(Revoked, revoked)
328 TAS_BUFFER_FNS(Revoked, revoked)
329 BUFFER_FNS(RevokeValid, revokevalid)
330 TAS_BUFFER_FNS(RevokeValid, revokevalid)
331 BUFFER_FNS(Freed, freed)
332 BUFFER_FNS(Shadow, shadow)
333 BUFFER_FNS(Verified, verified)
334
335 static inline struct buffer_head *jh2bh(struct journal_head *jh)
336 {
337 return jh->b_bh;
338 }
339
340 static inline struct journal_head *bh2jh(struct buffer_head *bh)
341 {
342 return bh->b_private;
343 }
344
345 static inline void jbd_lock_bh_state(struct buffer_head *bh)
346 {
347 bit_spin_lock(BH_State, &bh->b_state);
348 }
349
350 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
351 {
352 return bit_spin_trylock(BH_State, &bh->b_state);
353 }
354
355 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
356 {
357 return bit_spin_is_locked(BH_State, &bh->b_state);
358 }
359
360 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
361 {
362 bit_spin_unlock(BH_State, &bh->b_state);
363 }
364
365 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
366 {
367 bit_spin_lock(BH_JournalHead, &bh->b_state);
368 }
369
370 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
371 {
372 bit_spin_unlock(BH_JournalHead, &bh->b_state);
373 }
374
375 #define J_ASSERT(assert) BUG_ON(!(assert))
376
377 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
378 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
379
380 #if defined(JBD2_PARANOID_IOFAIL)
381 #define J_EXPECT(expr, why...) J_ASSERT(expr)
382 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
383 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
384 #else
385 #define __journal_expect(expr, why...) \
386 ({ \
387 int val = (expr); \
388 if (!val) { \
389 printk(KERN_ERR \
390 "JBD2 unexpected failure: %s: %s;\n", \
391 __func__, #expr); \
392 printk(KERN_ERR why "\n"); \
393 } \
394 val; \
395 })
396 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
397 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
398 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
399 #endif
400
401 /* Flags in jbd_inode->i_flags */
402 #define __JI_COMMIT_RUNNING 0
403 #define __JI_WRITE_DATA 1
404 #define __JI_WAIT_DATA 2
405
406 /*
407 * Commit of the inode data in progress. We use this flag to protect us from
408 * concurrent deletion of inode. We cannot use reference to inode for this
409 * since we cannot afford doing last iput() on behalf of kjournald
410 */
411 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
412 /* Write allocated dirty buffers in this inode before commit */
413 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
414 /* Wait for outstanding data writes for this inode before commit */
415 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
416
417 /**
418 * struct jbd_inode - The jbd_inode type is the structure linking inodes in
419 * ordered mode present in a transaction so that we can sync them during commit.
420 */
421 struct jbd2_inode {
422 /**
423 * @i_transaction:
424 *
425 * Which transaction does this inode belong to? Either the running
426 * transaction or the committing one. [j_list_lock]
427 */
428 transaction_t *i_transaction;
429
430 /**
431 * @i_next_transaction:
432 *
433 * Pointer to the running transaction modifying inode's data in case
434 * there is already a committing transaction touching it. [j_list_lock]
435 */
436 transaction_t *i_next_transaction;
437
438 /**
439 * @i_list: List of inodes in the i_transaction [j_list_lock]
440 */
441 struct list_head i_list;
442
443 /**
444 * @i_vfs_inode:
445 *
446 * VFS inode this inode belongs to [constant for lifetime of structure]
447 */
448 struct inode *i_vfs_inode;
449
450 /**
451 * @i_flags: Flags of inode [j_list_lock]
452 */
453 unsigned long i_flags;
454
455 /**
456 * @i_dirty_start:
457 *
458 * Offset in bytes where the dirty range for this inode starts.
459 * [j_list_lock]
460 */
461 loff_t i_dirty_start;
462
463 /**
464 * @i_dirty_end:
465 *
466 * Inclusive offset in bytes where the dirty range for this inode
467 * ends. [j_list_lock]
468 */
469 loff_t i_dirty_end;
470 };
471
472 struct jbd2_revoke_table_s;
473
474 /**
475 * struct handle_s - The handle_s type is the concrete type associated with
476 * handle_t.
477 * @h_transaction: Which compound transaction is this update a part of?
478 * @h_journal: Which journal handle belongs to - used iff h_reserved set.
479 * @h_rsv_handle: Handle reserved for finishing the logical operation.
480 * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
481 * @h_ref: Reference count on this handle.
482 * @h_err: Field for caller's use to track errors through large fs operations.
483 * @h_sync: Flag for sync-on-close.
484 * @h_jdata: Flag to force data journaling.
485 * @h_reserved: Flag for handle for reserved credits.
486 * @h_aborted: Flag indicating fatal error on handle.
487 * @h_type: For handle statistics.
488 * @h_line_no: For handle statistics.
489 * @h_start_jiffies: Handle Start time.
490 * @h_requested_credits: Holds @h_buffer_credits after handle is started.
491 * @saved_alloc_context: Saved context while transaction is open.
492 **/
493
494 /* Docbook can't yet cope with the bit fields, but will leave the documentation
495 * in so it can be fixed later.
496 */
497
498 struct jbd2_journal_handle
499 {
500 union {
501 transaction_t *h_transaction;
502 /* Which journal handle belongs to - used iff h_reserved set */
503 journal_t *h_journal;
504 };
505
506 handle_t *h_rsv_handle;
507 int h_buffer_credits;
508 int h_ref;
509 int h_err;
510
511 /* Flags [no locking] */
512 unsigned int h_sync: 1;
513 unsigned int h_jdata: 1;
514 unsigned int h_reserved: 1;
515 unsigned int h_aborted: 1;
516 unsigned int h_type: 8;
517 unsigned int h_line_no: 16;
518
519 unsigned long h_start_jiffies;
520 unsigned int h_requested_credits;
521
522 unsigned int saved_alloc_context;
523 };
524
525
526 /*
527 * Some stats for checkpoint phase
528 */
529 struct transaction_chp_stats_s {
530 unsigned long cs_chp_time;
531 __u32 cs_forced_to_close;
532 __u32 cs_written;
533 __u32 cs_dropped;
534 };
535
536 /* The transaction_t type is the guts of the journaling mechanism. It
537 * tracks a compound transaction through its various states:
538 *
539 * RUNNING: accepting new updates
540 * LOCKED: Updates still running but we don't accept new ones
541 * RUNDOWN: Updates are tidying up but have finished requesting
542 * new buffers to modify (state not used for now)
543 * FLUSH: All updates complete, but we are still writing to disk
544 * COMMIT: All data on disk, writing commit record
545 * FINISHED: We still have to keep the transaction for checkpointing.
546 *
547 * The transaction keeps track of all of the buffers modified by a
548 * running transaction, and all of the buffers committed but not yet
549 * flushed to home for finished transactions.
550 */
551
552 /*
553 * Lock ranking:
554 *
555 * j_list_lock
556 * ->jbd_lock_bh_journal_head() (This is "innermost")
557 *
558 * j_state_lock
559 * ->jbd_lock_bh_state()
560 *
561 * jbd_lock_bh_state()
562 * ->j_list_lock
563 *
564 * j_state_lock
565 * ->t_handle_lock
566 *
567 * j_state_lock
568 * ->j_list_lock (journal_unmap_buffer)
569 *
570 */
571
572 struct transaction_s
573 {
574 /* Pointer to the journal for this transaction. [no locking] */
575 journal_t *t_journal;
576
577 /* Sequence number for this transaction [no locking] */
578 tid_t t_tid;
579
580 /*
581 * Transaction's current state
582 * [no locking - only kjournald2 alters this]
583 * [j_list_lock] guards transition of a transaction into T_FINISHED
584 * state and subsequent call of __jbd2_journal_drop_transaction()
585 * FIXME: needs barriers
586 * KLUDGE: [use j_state_lock]
587 */
588 enum {
589 T_RUNNING,
590 T_LOCKED,
591 T_SWITCH,
592 T_FLUSH,
593 T_COMMIT,
594 T_COMMIT_DFLUSH,
595 T_COMMIT_JFLUSH,
596 T_COMMIT_CALLBACK,
597 T_FINISHED
598 } t_state;
599
600 /*
601 * Where in the log does this transaction's commit start? [no locking]
602 */
603 unsigned long t_log_start;
604
605 /* Number of buffers on the t_buffers list [j_list_lock] */
606 int t_nr_buffers;
607
608 /*
609 * Doubly-linked circular list of all buffers reserved but not yet
610 * modified by this transaction [j_list_lock]
611 */
612 struct journal_head *t_reserved_list;
613
614 /*
615 * Doubly-linked circular list of all metadata buffers owned by this
616 * transaction [j_list_lock]
617 */
618 struct journal_head *t_buffers;
619
620 /*
621 * Doubly-linked circular list of all forget buffers (superseded
622 * buffers which we can un-checkpoint once this transaction commits)
623 * [j_list_lock]
624 */
625 struct journal_head *t_forget;
626
627 /*
628 * Doubly-linked circular list of all buffers still to be flushed before
629 * this transaction can be checkpointed. [j_list_lock]
630 */
631 struct journal_head *t_checkpoint_list;
632
633 /*
634 * Doubly-linked circular list of all buffers submitted for IO while
635 * checkpointing. [j_list_lock]
636 */
637 struct journal_head *t_checkpoint_io_list;
638
639 /*
640 * Doubly-linked circular list of metadata buffers being shadowed by log
641 * IO. The IO buffers on the iobuf list and the shadow buffers on this
642 * list match each other one for one at all times. [j_list_lock]
643 */
644 struct journal_head *t_shadow_list;
645
646 /*
647 * List of inodes whose data we've modified in data=ordered mode.
648 * [j_list_lock]
649 */
650 struct list_head t_inode_list;
651
652 /*
653 * Protects info related to handles
654 */
655 spinlock_t t_handle_lock;
656
657 /*
658 * Longest time some handle had to wait for running transaction
659 */
660 unsigned long t_max_wait;
661
662 /*
663 * When transaction started
664 */
665 unsigned long t_start;
666
667 /*
668 * When commit was requested
669 */
670 unsigned long t_requested;
671
672 /*
673 * Checkpointing stats [j_checkpoint_sem]
674 */
675 struct transaction_chp_stats_s t_chp_stats;
676
677 /*
678 * Number of outstanding updates running on this transaction
679 * [none]
680 */
681 atomic_t t_updates;
682
683 /*
684 * Number of buffers reserved for use by all handles in this transaction
685 * handle but not yet modified. [none]
686 */
687 atomic_t t_outstanding_credits;
688
689 /*
690 * Forward and backward links for the circular list of all transactions
691 * awaiting checkpoint. [j_list_lock]
692 */
693 transaction_t *t_cpnext, *t_cpprev;
694
695 /*
696 * When will the transaction expire (become due for commit), in jiffies?
697 * [no locking]
698 */
699 unsigned long t_expires;
700
701 /*
702 * When this transaction started, in nanoseconds [no locking]
703 */
704 ktime_t t_start_time;
705
706 /*
707 * How many handles used this transaction? [none]
708 */
709 atomic_t t_handle_count;
710
711 /*
712 * This transaction is being forced and some process is
713 * waiting for it to finish.
714 */
715 unsigned int t_synchronous_commit:1;
716
717 /* Disk flush needs to be sent to fs partition [no locking] */
718 int t_need_data_flush;
719
720 /*
721 * For use by the filesystem to store fs-specific data
722 * structures associated with the transaction
723 */
724 struct list_head t_private_list;
725 };
726
727 struct transaction_run_stats_s {
728 unsigned long rs_wait;
729 unsigned long rs_request_delay;
730 unsigned long rs_running;
731 unsigned long rs_locked;
732 unsigned long rs_flushing;
733 unsigned long rs_logging;
734
735 __u32 rs_handle_count;
736 __u32 rs_blocks;
737 __u32 rs_blocks_logged;
738 };
739
740 struct transaction_stats_s {
741 unsigned long ts_tid;
742 unsigned long ts_requested;
743 struct transaction_run_stats_s run;
744 };
745
746 static inline unsigned long
747 jbd2_time_diff(unsigned long start, unsigned long end)
748 {
749 if (end >= start)
750 return end - start;
751
752 return end + (MAX_JIFFY_OFFSET - start);
753 }
754
755 #define JBD2_NR_BATCH 64
756
757 /**
758 * struct journal_s - The journal_s type is the concrete type associated with
759 * journal_t.
760 */
761 struct journal_s
762 {
763 /**
764 * @j_flags: General journaling state flags [j_state_lock]
765 */
766 unsigned long j_flags;
767
768 /**
769 * @j_errno:
770 *
771 * Is there an outstanding uncleared error on the journal (from a prior
772 * abort)? [j_state_lock]
773 */
774 int j_errno;
775
776 /**
777 * @j_sb_buffer: The first part of the superblock buffer.
778 */
779 struct buffer_head *j_sb_buffer;
780
781 /**
782 * @j_superblock: The second part of the superblock buffer.
783 */
784 journal_superblock_t *j_superblock;
785
786 /**
787 * @j_format_version: Version of the superblock format.
788 */
789 int j_format_version;
790
791 /**
792 * @j_state_lock: Protect the various scalars in the journal.
793 */
794 rwlock_t j_state_lock;
795
796 /**
797 * @j_barrier_count:
798 *
799 * Number of processes waiting to create a barrier lock [j_state_lock]
800 */
801 int j_barrier_count;
802
803 /**
804 * @j_barrier: The barrier lock itself.
805 */
806 struct mutex j_barrier;
807
808 /**
809 * @j_running_transaction:
810 *
811 * Transactions: The current running transaction...
812 * [j_state_lock] [caller holding open handle]
813 */
814 transaction_t *j_running_transaction;
815
816 /**
817 * @j_committing_transaction:
818 *
819 * the transaction we are pushing to disk
820 * [j_state_lock] [caller holding open handle]
821 */
822 transaction_t *j_committing_transaction;
823
824 /**
825 * @j_checkpoint_transactions:
826 *
827 * ... and a linked circular list of all transactions waiting for
828 * checkpointing. [j_list_lock]
829 */
830 transaction_t *j_checkpoint_transactions;
831
832 /**
833 * @j_wait_transaction_locked:
834 *
835 * Wait queue for waiting for a locked transaction to start committing,
836 * or for a barrier lock to be released.
837 */
838 wait_queue_head_t j_wait_transaction_locked;
839
840 /**
841 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
842 */
843 wait_queue_head_t j_wait_done_commit;
844
845 /**
846 * @j_wait_commit: Wait queue to trigger commit.
847 */
848 wait_queue_head_t j_wait_commit;
849
850 /**
851 * @j_wait_updates: Wait queue to wait for updates to complete.
852 */
853 wait_queue_head_t j_wait_updates;
854
855 /**
856 * @j_wait_reserved:
857 *
858 * Wait queue to wait for reserved buffer credits to drop.
859 */
860 wait_queue_head_t j_wait_reserved;
861
862 /**
863 * @j_checkpoint_mutex:
864 *
865 * Semaphore for locking against concurrent checkpoints.
866 */
867 struct mutex j_checkpoint_mutex;
868
869 /**
870 * @j_chkpt_bhs:
871 *
872 * List of buffer heads used by the checkpoint routine. This
873 * was moved from jbd2_log_do_checkpoint() to reduce stack
874 * usage. Access to this array is controlled by the
875 * @j_checkpoint_mutex. [j_checkpoint_mutex]
876 */
877 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
878
879 /**
880 * @j_head:
881 *
882 * Journal head: identifies the first unused block in the journal.
883 * [j_state_lock]
884 */
885 unsigned long j_head;
886
887 /**
888 * @j_tail:
889 *
890 * Journal tail: identifies the oldest still-used block in the journal.
891 * [j_state_lock]
892 */
893 unsigned long j_tail;
894
895 /**
896 * @j_free:
897 *
898 * Journal free: how many free blocks are there in the journal?
899 * [j_state_lock]
900 */
901 unsigned long j_free;
902
903 /**
904 * @j_first:
905 *
906 * The block number of the first usable block in the journal
907 * [j_state_lock].
908 */
909 unsigned long j_first;
910
911 /**
912 * @j_last:
913 *
914 * The block number one beyond the last usable block in the journal
915 * [j_state_lock].
916 */
917 unsigned long j_last;
918
919 /**
920 * @j_dev: Device where we store the journal.
921 */
922 struct block_device *j_dev;
923
924 /**
925 * @j_blocksize: Block size for the location where we store the journal.
926 */
927 int j_blocksize;
928
929 /**
930 * @j_blk_offset:
931 *
932 * Starting block offset into the device where we store the journal.
933 */
934 unsigned long long j_blk_offset;
935
936 /**
937 * @j_devname: Journal device name.
938 */
939 char j_devname[BDEVNAME_SIZE+24];
940
941 /**
942 * @j_fs_dev:
943 *
944 * Device which holds the client fs. For internal journal this will be
945 * equal to j_dev.
946 */
947 struct block_device *j_fs_dev;
948
949 /**
950 * @j_maxlen: Total maximum capacity of the journal region on disk.
951 */
952 unsigned int j_maxlen;
953
954 /**
955 * @j_reserved_credits:
956 *
957 * Number of buffers reserved from the running transaction.
958 */
959 atomic_t j_reserved_credits;
960
961 /**
962 * @j_list_lock: Protects the buffer lists and internal buffer state.
963 */
964 spinlock_t j_list_lock;
965
966 /**
967 * @j_inode:
968 *
969 * Optional inode where we store the journal. If present, all
970 * journal block numbers are mapped into this inode via bmap().
971 */
972 struct inode *j_inode;
973
974 /**
975 * @j_tail_sequence:
976 *
977 * Sequence number of the oldest transaction in the log [j_state_lock]
978 */
979 tid_t j_tail_sequence;
980
981 /**
982 * @j_transaction_sequence:
983 *
984 * Sequence number of the next transaction to grant [j_state_lock]
985 */
986 tid_t j_transaction_sequence;
987
988 /**
989 * @j_commit_sequence:
990 *
991 * Sequence number of the most recently committed transaction
992 * [j_state_lock].
993 */
994 tid_t j_commit_sequence;
995
996 /**
997 * @j_commit_request:
998 *
999 * Sequence number of the most recent transaction wanting commit
1000 * [j_state_lock]
1001 */
1002 tid_t j_commit_request;
1003
1004 /**
1005 * @j_uuid:
1006 *
1007 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1008 * backed by this journal. This will eventually be replaced by an array
1009 * of uuids, allowing us to index multiple devices within a single
1010 * journal and to perform atomic updates across them.
1011 */
1012 __u8 j_uuid[16];
1013
1014 /**
1015 * @j_task: Pointer to the current commit thread for this journal.
1016 */
1017 struct task_struct *j_task;
1018
1019 /**
1020 * @j_max_transaction_buffers:
1021 *
1022 * Maximum number of metadata buffers to allow in a single compound
1023 * commit transaction.
1024 */
1025 int j_max_transaction_buffers;
1026
1027 /**
1028 * @j_commit_interval:
1029 *
1030 * What is the maximum transaction lifetime before we begin a commit?
1031 */
1032 unsigned long j_commit_interval;
1033
1034 /**
1035 * @j_commit_timer: The timer used to wakeup the commit thread.
1036 */
1037 struct timer_list j_commit_timer;
1038
1039 /**
1040 * @j_revoke_lock: Protect the revoke table.
1041 */
1042 spinlock_t j_revoke_lock;
1043
1044 /**
1045 * @j_revoke:
1046 *
1047 * The revoke table - maintains the list of revoked blocks in the
1048 * current transaction.
1049 */
1050 struct jbd2_revoke_table_s *j_revoke;
1051
1052 /**
1053 * @j_revoke_table: Alternate revoke tables for j_revoke.
1054 */
1055 struct jbd2_revoke_table_s *j_revoke_table[2];
1056
1057 /**
1058 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1059 */
1060 struct buffer_head **j_wbuf;
1061
1062 /**
1063 * @j_wbufsize:
1064 *
1065 * Size of @j_wbuf array.
1066 */
1067 int j_wbufsize;
1068
1069 /**
1070 * @j_last_sync_writer:
1071 *
1072 * The pid of the last person to run a synchronous operation
1073 * through the journal.
1074 */
1075 pid_t j_last_sync_writer;
1076
1077 /**
1078 * @j_average_commit_time:
1079 *
1080 * The average amount of time in nanoseconds it takes to commit a
1081 * transaction to disk. [j_state_lock]
1082 */
1083 u64 j_average_commit_time;
1084
1085 /**
1086 * @j_min_batch_time:
1087 *
1088 * Minimum time that we should wait for additional filesystem operations
1089 * to get batched into a synchronous handle in microseconds.
1090 */
1091 u32 j_min_batch_time;
1092
1093 /**
1094 * @j_max_batch_time:
1095 *
1096 * Maximum time that we should wait for additional filesystem operations
1097 * to get batched into a synchronous handle in microseconds.
1098 */
1099 u32 j_max_batch_time;
1100
1101 /**
1102 * @j_commit_callback:
1103 *
1104 * This function is called when a transaction is closed.
1105 */
1106 void (*j_commit_callback)(journal_t *,
1107 transaction_t *);
1108
1109 /*
1110 * Journal statistics
1111 */
1112
1113 /**
1114 * @j_history_lock: Protect the transactions statistics history.
1115 */
1116 spinlock_t j_history_lock;
1117
1118 /**
1119 * @j_proc_entry: procfs entry for the jbd statistics directory.
1120 */
1121 struct proc_dir_entry *j_proc_entry;
1122
1123 /**
1124 * @j_stats: Overall statistics.
1125 */
1126 struct transaction_stats_s j_stats;
1127
1128 /**
1129 * @j_failed_commit: Failed journal commit ID.
1130 */
1131 unsigned int j_failed_commit;
1132
1133 /**
1134 * @j_private:
1135 *
1136 * An opaque pointer to fs-private information. ext3 puts its
1137 * superblock pointer here.
1138 */
1139 void *j_private;
1140
1141 /**
1142 * @j_chksum_driver:
1143 *
1144 * Reference to checksum algorithm driver via cryptoapi.
1145 */
1146 struct crypto_shash *j_chksum_driver;
1147
1148 /**
1149 * @j_csum_seed:
1150 *
1151 * Precomputed journal UUID checksum for seeding other checksums.
1152 */
1153 __u32 j_csum_seed;
1154
1155 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1156 /**
1157 * @j_trans_commit_map:
1158 *
1159 * Lockdep entity to track transaction commit dependencies. Handles
1160 * hold this "lock" for read, when we wait for commit, we acquire the
1161 * "lock" for writing. This matches the properties of jbd2 journalling
1162 * where the running transaction has to wait for all handles to be
1163 * dropped to commit that transaction and also acquiring a handle may
1164 * require transaction commit to finish.
1165 */
1166 struct lockdep_map j_trans_commit_map;
1167 #endif
1168 };
1169
1170 #define jbd2_might_wait_for_commit(j) \
1171 do { \
1172 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1173 rwsem_release(&j->j_trans_commit_map, 1, _THIS_IP_); \
1174 } while (0)
1175
1176 /* journal feature predicate functions */
1177 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1178 static inline bool jbd2_has_feature_##name(journal_t *j) \
1179 { \
1180 return ((j)->j_format_version >= 2 && \
1181 ((j)->j_superblock->s_feature_compat & \
1182 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1183 } \
1184 static inline void jbd2_set_feature_##name(journal_t *j) \
1185 { \
1186 (j)->j_superblock->s_feature_compat |= \
1187 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1188 } \
1189 static inline void jbd2_clear_feature_##name(journal_t *j) \
1190 { \
1191 (j)->j_superblock->s_feature_compat &= \
1192 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1193 }
1194
1195 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1196 static inline bool jbd2_has_feature_##name(journal_t *j) \
1197 { \
1198 return ((j)->j_format_version >= 2 && \
1199 ((j)->j_superblock->s_feature_ro_compat & \
1200 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1201 } \
1202 static inline void jbd2_set_feature_##name(journal_t *j) \
1203 { \
1204 (j)->j_superblock->s_feature_ro_compat |= \
1205 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1206 } \
1207 static inline void jbd2_clear_feature_##name(journal_t *j) \
1208 { \
1209 (j)->j_superblock->s_feature_ro_compat &= \
1210 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1211 }
1212
1213 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1214 static inline bool jbd2_has_feature_##name(journal_t *j) \
1215 { \
1216 return ((j)->j_format_version >= 2 && \
1217 ((j)->j_superblock->s_feature_incompat & \
1218 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1219 } \
1220 static inline void jbd2_set_feature_##name(journal_t *j) \
1221 { \
1222 (j)->j_superblock->s_feature_incompat |= \
1223 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1224 } \
1225 static inline void jbd2_clear_feature_##name(journal_t *j) \
1226 { \
1227 (j)->j_superblock->s_feature_incompat &= \
1228 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1229 }
1230
1231 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM)
1232
1233 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE)
1234 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
1235 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT)
1236 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2)
1237 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3)
1238
1239 /*
1240 * Journal flag definitions
1241 */
1242 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1243 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1244 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1245 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1246 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1247 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
1248 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1249 * data write error in ordered
1250 * mode */
1251 #define JBD2_REC_ERR 0x080 /* The errno in the sb has been recorded */
1252
1253 /*
1254 * Function declarations for the journaling transaction and buffer
1255 * management
1256 */
1257
1258 /* Filing buffers */
1259 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1260 extern void __jbd2_journal_refile_buffer(struct journal_head *);
1261 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1262 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1263 extern void __journal_free_buffer(struct journal_head *bh);
1264 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1265 extern void __journal_clean_data_list(transaction_t *transaction);
1266 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1267 {
1268 list_add_tail(&bh->b_assoc_buffers, head);
1269 }
1270 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1271 {
1272 list_del_init(&bh->b_assoc_buffers);
1273 }
1274
1275 /* Log buffer allocation */
1276 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1277 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1278 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1279 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1280 unsigned long *block);
1281 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1282 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1283
1284 /* Commit management */
1285 extern void jbd2_journal_commit_transaction(journal_t *);
1286
1287 /* Checkpoint list management */
1288 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1289 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1290 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1291 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1292
1293
1294 /*
1295 * Triggers
1296 */
1297
1298 struct jbd2_buffer_trigger_type {
1299 /*
1300 * Fired a the moment data to write to the journal are known to be
1301 * stable - so either at the moment b_frozen_data is created or just
1302 * before a buffer is written to the journal. mapped_data is a mapped
1303 * buffer that is the frozen data for commit.
1304 */
1305 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1306 struct buffer_head *bh, void *mapped_data,
1307 size_t size);
1308
1309 /*
1310 * Fired during journal abort for dirty buffers that will not be
1311 * committed.
1312 */
1313 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1314 struct buffer_head *bh);
1315 };
1316
1317 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1318 void *mapped_data,
1319 struct jbd2_buffer_trigger_type *triggers);
1320 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1321 struct jbd2_buffer_trigger_type *triggers);
1322
1323 /* Buffer IO */
1324 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1325 struct journal_head *jh_in,
1326 struct buffer_head **bh_out,
1327 sector_t blocknr);
1328
1329 /* Transaction locking */
1330 extern void __wait_on_journal (journal_t *);
1331
1332 /* Transaction cache support */
1333 extern void jbd2_journal_destroy_transaction_cache(void);
1334 extern int __init jbd2_journal_init_transaction_cache(void);
1335 extern void jbd2_journal_free_transaction(transaction_t *);
1336
1337 /*
1338 * Journal locking.
1339 *
1340 * We need to lock the journal during transaction state changes so that nobody
1341 * ever tries to take a handle on the running transaction while we are in the
1342 * middle of moving it to the commit phase. j_state_lock does this.
1343 *
1344 * Note that the locking is completely interrupt unsafe. We never touch
1345 * journal structures from interrupts.
1346 */
1347
1348 static inline handle_t *journal_current_handle(void)
1349 {
1350 return current->journal_info;
1351 }
1352
1353 /* The journaling code user interface:
1354 *
1355 * Create and destroy handles
1356 * Register buffer modifications against the current transaction.
1357 */
1358
1359 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1360 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1361 gfp_t gfp_mask, unsigned int type,
1362 unsigned int line_no);
1363 extern int jbd2_journal_restart(handle_t *, int nblocks);
1364 extern int jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
1365 extern int jbd2_journal_start_reserved(handle_t *handle,
1366 unsigned int type, unsigned int line_no);
1367 extern void jbd2_journal_free_reserved(handle_t *handle);
1368 extern int jbd2_journal_extend (handle_t *, int nblocks);
1369 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1370 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1371 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1372 void jbd2_journal_set_triggers(struct buffer_head *,
1373 struct jbd2_buffer_trigger_type *type);
1374 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1375 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1376 extern int jbd2_journal_invalidatepage(journal_t *,
1377 struct page *, unsigned int, unsigned int);
1378 extern int jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1379 extern int jbd2_journal_stop(handle_t *);
1380 extern int jbd2_journal_flush (journal_t *);
1381 extern void jbd2_journal_lock_updates (journal_t *);
1382 extern void jbd2_journal_unlock_updates (journal_t *);
1383
1384 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1385 struct block_device *fs_dev,
1386 unsigned long long start, int len, int bsize);
1387 extern journal_t * jbd2_journal_init_inode (struct inode *);
1388 extern int jbd2_journal_update_format (journal_t *);
1389 extern int jbd2_journal_check_used_features
1390 (journal_t *, unsigned long, unsigned long, unsigned long);
1391 extern int jbd2_journal_check_available_features
1392 (journal_t *, unsigned long, unsigned long, unsigned long);
1393 extern int jbd2_journal_set_features
1394 (journal_t *, unsigned long, unsigned long, unsigned long);
1395 extern void jbd2_journal_clear_features
1396 (journal_t *, unsigned long, unsigned long, unsigned long);
1397 extern int jbd2_journal_load (journal_t *journal);
1398 extern int jbd2_journal_destroy (journal_t *);
1399 extern int jbd2_journal_recover (journal_t *journal);
1400 extern int jbd2_journal_wipe (journal_t *, int);
1401 extern int jbd2_journal_skip_recovery (journal_t *);
1402 extern void jbd2_journal_update_sb_errno(journal_t *);
1403 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
1404 unsigned long, int);
1405 extern void __jbd2_journal_abort_hard (journal_t *);
1406 extern void jbd2_journal_abort (journal_t *, int);
1407 extern int jbd2_journal_errno (journal_t *);
1408 extern void jbd2_journal_ack_err (journal_t *);
1409 extern int jbd2_journal_clear_err (journal_t *);
1410 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1411 extern int jbd2_journal_force_commit(journal_t *);
1412 extern int jbd2_journal_force_commit_nested(journal_t *);
1413 extern int jbd2_journal_inode_ranged_write(handle_t *handle,
1414 struct jbd2_inode *inode, loff_t start_byte,
1415 loff_t length);
1416 extern int jbd2_journal_inode_ranged_wait(handle_t *handle,
1417 struct jbd2_inode *inode, loff_t start_byte,
1418 loff_t length);
1419 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1420 struct jbd2_inode *inode, loff_t new_size);
1421 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1422 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1423
1424 /*
1425 * journal_head management
1426 */
1427 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1428 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1429 void jbd2_journal_put_journal_head(struct journal_head *jh);
1430
1431 /*
1432 * handle management
1433 */
1434 extern struct kmem_cache *jbd2_handle_cache;
1435
1436 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1437 {
1438 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1439 }
1440
1441 static inline void jbd2_free_handle(handle_t *handle)
1442 {
1443 kmem_cache_free(jbd2_handle_cache, handle);
1444 }
1445
1446 /*
1447 * jbd2_inode management (optional, for those file systems that want to use
1448 * dynamically allocated jbd2_inode structures)
1449 */
1450 extern struct kmem_cache *jbd2_inode_cache;
1451
1452 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1453 {
1454 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1455 }
1456
1457 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1458 {
1459 kmem_cache_free(jbd2_inode_cache, jinode);
1460 }
1461
1462 /* Primary revoke support */
1463 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1464 extern int jbd2_journal_init_revoke(journal_t *, int);
1465 extern void jbd2_journal_destroy_revoke_record_cache(void);
1466 extern void jbd2_journal_destroy_revoke_table_cache(void);
1467 extern int __init jbd2_journal_init_revoke_record_cache(void);
1468 extern int __init jbd2_journal_init_revoke_table_cache(void);
1469
1470 extern void jbd2_journal_destroy_revoke(journal_t *);
1471 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1472 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1473 extern void jbd2_journal_write_revoke_records(transaction_t *transaction,
1474 struct list_head *log_bufs);
1475
1476 /* Recovery revoke support */
1477 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1478 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1479 extern void jbd2_journal_clear_revoke(journal_t *);
1480 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1481 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
1482
1483 /*
1484 * The log thread user interface:
1485 *
1486 * Request space in the current transaction, and force transaction commit
1487 * transitions on demand.
1488 */
1489
1490 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1491 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1492 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1493 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1494 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1495 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1496 int jbd2_log_do_checkpoint(journal_t *journal);
1497 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1498
1499 void __jbd2_log_wait_for_space(journal_t *journal);
1500 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1501 extern int jbd2_cleanup_journal_tail(journal_t *);
1502
1503 /*
1504 * is_journal_abort
1505 *
1506 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1507 * bit, when set, indicates that we have had a fatal error somewhere,
1508 * either inside the journaling layer or indicated to us by the client
1509 * (eg. ext3), and that we and should not commit any further
1510 * transactions.
1511 */
1512
1513 static inline int is_journal_aborted(journal_t *journal)
1514 {
1515 return journal->j_flags & JBD2_ABORT;
1516 }
1517
1518 static inline int is_handle_aborted(handle_t *handle)
1519 {
1520 if (handle->h_aborted || !handle->h_transaction)
1521 return 1;
1522 return is_journal_aborted(handle->h_transaction->t_journal);
1523 }
1524
1525 static inline void jbd2_journal_abort_handle(handle_t *handle)
1526 {
1527 handle->h_aborted = 1;
1528 }
1529
1530 #endif /* __KERNEL__ */
1531
1532 /* Comparison functions for transaction IDs: perform comparisons using
1533 * modulo arithmetic so that they work over sequence number wraps. */
1534
1535 static inline int tid_gt(tid_t x, tid_t y)
1536 {
1537 int difference = (x - y);
1538 return (difference > 0);
1539 }
1540
1541 static inline int tid_geq(tid_t x, tid_t y)
1542 {
1543 int difference = (x - y);
1544 return (difference >= 0);
1545 }
1546
1547 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1548 extern size_t journal_tag_bytes(journal_t *journal);
1549
1550 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1551 {
1552 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1553 }
1554
1555 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1556 {
1557 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1558 journal->j_chksum_driver == NULL);
1559
1560 return journal->j_chksum_driver != NULL;
1561 }
1562
1563 /*
1564 * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
1565 * transaction control blocks.
1566 */
1567 #define JBD2_CONTROL_BLOCKS_SHIFT 5
1568
1569 /*
1570 * Return the minimum number of blocks which must be free in the journal
1571 * before a new transaction may be started. Must be called under j_state_lock.
1572 */
1573 static inline int jbd2_space_needed(journal_t *journal)
1574 {
1575 int nblocks = journal->j_max_transaction_buffers;
1576 return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT);
1577 }
1578
1579 /*
1580 * Return number of free blocks in the log. Must be called under j_state_lock.
1581 */
1582 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1583 {
1584 /* Allow for rounding errors */
1585 long free = journal->j_free - 32;
1586
1587 if (journal->j_committing_transaction) {
1588 unsigned long committing = atomic_read(&journal->
1589 j_committing_transaction->t_outstanding_credits);
1590
1591 /* Transaction + control blocks */
1592 free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT);
1593 }
1594 return max_t(long, free, 0);
1595 }
1596
1597 /*
1598 * Definitions which augment the buffer_head layer
1599 */
1600
1601 /* journaling buffer types */
1602 #define BJ_None 0 /* Not journaled */
1603 #define BJ_Metadata 1 /* Normal journaled metadata */
1604 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1605 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
1606 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1607 #define BJ_Types 5
1608
1609 extern int jbd_blocks_per_page(struct inode *inode);
1610
1611 /* JBD uses a CRC32 checksum */
1612 #define JBD_MAX_CHECKSUM_SIZE 4
1613
1614 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1615 const void *address, unsigned int length)
1616 {
1617 struct {
1618 struct shash_desc shash;
1619 char ctx[JBD_MAX_CHECKSUM_SIZE];
1620 } desc;
1621 int err;
1622
1623 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1624 JBD_MAX_CHECKSUM_SIZE);
1625
1626 desc.shash.tfm = journal->j_chksum_driver;
1627 *(u32 *)desc.ctx = crc;
1628
1629 err = crypto_shash_update(&desc.shash, address, length);
1630 BUG_ON(err);
1631
1632 return *(u32 *)desc.ctx;
1633 }
1634
1635 /* Return most recent uncommitted transaction */
1636 static inline tid_t jbd2_get_latest_transaction(journal_t *journal)
1637 {
1638 tid_t tid;
1639
1640 read_lock(&journal->j_state_lock);
1641 tid = journal->j_commit_request;
1642 if (journal->j_running_transaction)
1643 tid = journal->j_running_transaction->t_tid;
1644 read_unlock(&journal->j_state_lock);
1645 return tid;
1646 }
1647
1648 #ifdef __KERNEL__
1649
1650 #define buffer_trace_init(bh) do {} while (0)
1651 #define print_buffer_fields(bh) do {} while (0)
1652 #define print_buffer_trace(bh) do {} while (0)
1653 #define BUFFER_TRACE(bh, info) do {} while (0)
1654 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1655 #define JBUFFER_TRACE(jh, info) do {} while (0)
1656
1657 #endif /* __KERNEL__ */
1658
1659 #define EFSBADCRC EBADMSG /* Bad CRC detected */
1660 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
1661
1662 #endif /* _LINUX_JBD2_H */