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 extern ushort jbd2_journal_enable_debug;
58 void __jbd2_debug(int level, const char *file, const char *func,
59 unsigned int line, const char *fmt, ...);
60
61 #define jbd_debug(n, fmt, a...) \
62 __jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
63 #else
64 #define jbd_debug(n, fmt, a...) /**/
65 #endif
66
67 extern void *jbd2_alloc(size_t size, gfp_t flags);
68 extern void jbd2_free(void *ptr, size_t size);
69
70 #define JBD2_MIN_JOURNAL_BLOCKS 1024
71 #define JBD2_MIN_FC_BLOCKS 256
72
73 #ifdef __KERNEL__
74
75 /**
76 * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
77 *
78 * All filesystem modifications made by the process go
79 * through this handle. Recursive operations (such as quota operations)
80 * are gathered into a single update.
81 *
82 * The buffer credits field is used to account for journaled buffers
83 * being modified by the running process. To ensure that there is
84 * enough log space for all outstanding operations, we need to limit the
85 * number of outstanding buffers possible at any time. When the
86 * operation completes, any buffer credits not used are credited back to
87 * the transaction, so that at all times we know how many buffers the
88 * outstanding updates on a transaction might possibly touch.
89 *
90 * This is an opaque datatype.
91 **/
92 typedef struct jbd2_journal_handle handle_t; /* Atomic operation type */
93
94
95 /**
96 * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
97 *
98 * journal_t is linked to from the fs superblock structure.
99 *
100 * We use the journal_t to keep track of all outstanding transaction
101 * activity on the filesystem, and to manage the state of the log
102 * writing process.
103 *
104 * This is an opaque datatype.
105 **/
106 typedef struct journal_s journal_t; /* Journal control structure */
107 #endif
108
109 /*
110 * Internal structures used by the logging mechanism:
111 */
112
113 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
114
115 /*
116 * On-disk structures
117 */
118
119 /*
120 * Descriptor block types:
121 */
122
123 #define JBD2_DESCRIPTOR_BLOCK 1
124 #define JBD2_COMMIT_BLOCK 2
125 #define JBD2_SUPERBLOCK_V1 3
126 #define JBD2_SUPERBLOCK_V2 4
127 #define JBD2_REVOKE_BLOCK 5
128
129 /*
130 * Standard header for all descriptor blocks:
131 */
132 typedef struct journal_header_s
133 {
134 __be32 h_magic;
135 __be32 h_blocktype;
136 __be32 h_sequence;
137 } journal_header_t;
138
139 /*
140 * Checksum types.
141 */
142 #define JBD2_CRC32_CHKSUM 1
143 #define JBD2_MD5_CHKSUM 2
144 #define JBD2_SHA1_CHKSUM 3
145 #define JBD2_CRC32C_CHKSUM 4
146
147 #define JBD2_CRC32_CHKSUM_SIZE 4
148
149 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
150 /*
151 * Commit block header for storing transactional checksums:
152 *
153 * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
154 * fields are used to store a checksum of the descriptor and data blocks.
155 *
156 * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
157 * field is used to store crc32c(uuid+commit_block). Each journal metadata
158 * block gets its own checksum, and data block checksums are stored in
159 * journal_block_tag (in the descriptor). The other h_chksum* fields are
160 * not used.
161 *
162 * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
163 * journal_block_tag3_t to store a full 32-bit checksum. Everything else
164 * is the same as v2.
165 *
166 * Checksum v1, v2, and v3 are mutually exclusive features.
167 */
168 struct commit_header {
169 __be32 h_magic;
170 __be32 h_blocktype;
171 __be32 h_sequence;
172 unsigned char h_chksum_type;
173 unsigned char h_chksum_size;
174 unsigned char h_padding[2];
175 __be32 h_chksum[JBD2_CHECKSUM_BYTES];
176 __be64 h_commit_sec;
177 __be32 h_commit_nsec;
178 };
179
180 /*
181 * The block tag: used to describe a single buffer in the journal.
182 * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
183 * raw struct shouldn't be used for pointer math or sizeof() - use
184 * journal_tag_bytes(journal) instead to compute this.
185 */
186 typedef struct journal_block_tag3_s
187 {
188 __be32 t_blocknr; /* The on-disk block number */
189 __be32 t_flags; /* See below */
190 __be32 t_blocknr_high; /* most-significant high 32bits. */
191 __be32 t_checksum; /* crc32c(uuid+seq+block) */
192 } journal_block_tag3_t;
193
194 typedef struct journal_block_tag_s
195 {
196 __be32 t_blocknr; /* The on-disk block number */
197 __be16 t_checksum; /* truncated crc32c(uuid+seq+block) */
198 __be16 t_flags; /* See below */
199 __be32 t_blocknr_high; /* most-significant high 32bits. */
200 } journal_block_tag_t;
201
202 /* Tail of descriptor or revoke block, for checksumming */
203 struct jbd2_journal_block_tail {
204 __be32 t_checksum; /* crc32c(uuid+descr_block) */
205 };
206
207 /*
208 * The revoke descriptor: used on disk to describe a series of blocks to
209 * be revoked from the log
210 */
211 typedef struct jbd2_journal_revoke_header_s
212 {
213 journal_header_t r_header;
214 __be32 r_count; /* Count of bytes used in the block */
215 } jbd2_journal_revoke_header_t;
216
217 /* Definitions for the journal tag flags word: */
218 #define JBD2_FLAG_ESCAPE 1 /* on-disk block is escaped */
219 #define JBD2_FLAG_SAME_UUID 2 /* block has same uuid as previous */
220 #define JBD2_FLAG_DELETED 4 /* block deleted by this transaction */
221 #define JBD2_FLAG_LAST_TAG 8 /* last tag in this descriptor block */
222
223
224 /*
225 * The journal superblock. All fields are in big-endian byte order.
226 */
227 typedef struct journal_superblock_s
228 {
229 /* 0x0000 */
230 journal_header_t s_header;
231
232 /* 0x000C */
233 /* Static information describing the journal */
234 __be32 s_blocksize; /* journal device blocksize */
235 __be32 s_maxlen; /* total blocks in journal file */
236 __be32 s_first; /* first block of log information */
237
238 /* 0x0018 */
239 /* Dynamic information describing the current state of the log */
240 __be32 s_sequence; /* first commit ID expected in log */
241 __be32 s_start; /* blocknr of start of log */
242
243 /* 0x0020 */
244 /* Error value, as set by jbd2_journal_abort(). */
245 __be32 s_errno;
246
247 /* 0x0024 */
248 /* Remaining fields are only valid in a version-2 superblock */
249 __be32 s_feature_compat; /* compatible feature set */
250 __be32 s_feature_incompat; /* incompatible feature set */
251 __be32 s_feature_ro_compat; /* readonly-compatible feature set */
252 /* 0x0030 */
253 __u8 s_uuid[16]; /* 128-bit uuid for journal */
254
255 /* 0x0040 */
256 __be32 s_nr_users; /* Nr of filesystems sharing log */
257
258 __be32 s_dynsuper; /* Blocknr of dynamic superblock copy*/
259
260 /* 0x0048 */
261 __be32 s_max_transaction; /* Limit of journal blocks per trans.*/
262 __be32 s_max_trans_data; /* Limit of data blocks per trans. */
263
264 /* 0x0050 */
265 __u8 s_checksum_type; /* checksum type */
266 __u8 s_padding2[3];
267 /* 0x0054 */
268 __be32 s_num_fc_blks; /* Number of fast commit blocks */
269 /* 0x0058 */
270 __u32 s_padding[41];
271 __be32 s_checksum; /* crc32c(superblock) */
272
273 /* 0x0100 */
274 __u8 s_users[16*48]; /* ids of all fs'es sharing the log */
275 /* 0x0400 */
276 } journal_superblock_t;
277
278 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
279 #define JBD2_HAS_COMPAT_FEATURE(j,mask) \
280 ((j)->j_format_version >= 2 && \
281 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
282 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask) \
283 ((j)->j_format_version >= 2 && \
284 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
285 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask) \
286 ((j)->j_format_version >= 2 && \
287 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
288
289 #define JBD2_FEATURE_COMPAT_CHECKSUM 0x00000001
290
291 #define JBD2_FEATURE_INCOMPAT_REVOKE 0x00000001
292 #define JBD2_FEATURE_INCOMPAT_64BIT 0x00000002
293 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT 0x00000004
294 #define JBD2_FEATURE_INCOMPAT_CSUM_V2 0x00000008
295 #define JBD2_FEATURE_INCOMPAT_CSUM_V3 0x00000010
296 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT 0x00000020
297
298 /* See "journal feature predicate functions" below */
299
300 /* Features known to this kernel version: */
301 #define JBD2_KNOWN_COMPAT_FEATURES JBD2_FEATURE_COMPAT_CHECKSUM
302 #define JBD2_KNOWN_ROCOMPAT_FEATURES 0
303 #define JBD2_KNOWN_INCOMPAT_FEATURES (JBD2_FEATURE_INCOMPAT_REVOKE | \
304 JBD2_FEATURE_INCOMPAT_64BIT | \
305 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
306 JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
307 JBD2_FEATURE_INCOMPAT_CSUM_V3 | \
308 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)
309
310 #ifdef __KERNEL__
311
312 #include <linux/fs.h>
313 #include <linux/sched.h>
314
315 enum jbd_state_bits {
316 BH_JBD /* Has an attached ext3 journal_head */
317 = BH_PrivateStart,
318 BH_JWrite, /* Being written to log (@@@ DEBUGGING) */
319 BH_Freed, /* Has been freed (truncated) */
320 BH_Revoked, /* Has been revoked from the log */
321 BH_RevokeValid, /* Revoked flag is valid */
322 BH_JBDDirty, /* Is dirty but journaled */
323 BH_JournalHead, /* Pins bh->b_private and jh->b_bh */
324 BH_Shadow, /* IO on shadow buffer is running */
325 BH_Verified, /* Metadata block has been verified ok */
326 BH_JBDPrivateStart, /* First bit available for private use by FS */
327 };
328
BUFFER_FNS(JBD,jbd)329 BUFFER_FNS(JBD, jbd)
330 BUFFER_FNS(JWrite, jwrite)
331 BUFFER_FNS(JBDDirty, jbddirty)
332 TAS_BUFFER_FNS(JBDDirty, jbddirty)
333 BUFFER_FNS(Revoked, revoked)
334 TAS_BUFFER_FNS(Revoked, revoked)
335 BUFFER_FNS(RevokeValid, revokevalid)
336 TAS_BUFFER_FNS(RevokeValid, revokevalid)
337 BUFFER_FNS(Freed, freed)
338 BUFFER_FNS(Shadow, shadow)
339 BUFFER_FNS(Verified, verified)
340
341 static inline struct buffer_head *jh2bh(struct journal_head *jh)
342 {
343 return jh->b_bh;
344 }
345
bh2jh(struct buffer_head * bh)346 static inline struct journal_head *bh2jh(struct buffer_head *bh)
347 {
348 return bh->b_private;
349 }
350
jbd_lock_bh_journal_head(struct buffer_head * bh)351 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
352 {
353 bit_spin_lock(BH_JournalHead, &bh->b_state);
354 }
355
jbd_unlock_bh_journal_head(struct buffer_head * bh)356 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
357 {
358 bit_spin_unlock(BH_JournalHead, &bh->b_state);
359 }
360
361 #define J_ASSERT(assert) BUG_ON(!(assert))
362
363 #define J_ASSERT_BH(bh, expr) J_ASSERT(expr)
364 #define J_ASSERT_JH(jh, expr) J_ASSERT(expr)
365
366 #if defined(JBD2_PARANOID_IOFAIL)
367 #define J_EXPECT(expr, why...) J_ASSERT(expr)
368 #define J_EXPECT_BH(bh, expr, why...) J_ASSERT_BH(bh, expr)
369 #define J_EXPECT_JH(jh, expr, why...) J_ASSERT_JH(jh, expr)
370 #else
371 #define __journal_expect(expr, why...) \
372 ({ \
373 int val = (expr); \
374 if (!val) { \
375 printk(KERN_ERR \
376 "JBD2 unexpected failure: %s: %s;\n", \
377 __func__, #expr); \
378 printk(KERN_ERR why "\n"); \
379 } \
380 val; \
381 })
382 #define J_EXPECT(expr, why...) __journal_expect(expr, ## why)
383 #define J_EXPECT_BH(bh, expr, why...) __journal_expect(expr, ## why)
384 #define J_EXPECT_JH(jh, expr, why...) __journal_expect(expr, ## why)
385 #endif
386
387 /* Flags in jbd_inode->i_flags */
388 #define __JI_COMMIT_RUNNING 0
389 #define __JI_WRITE_DATA 1
390 #define __JI_WAIT_DATA 2
391
392 /*
393 * Commit of the inode data in progress. We use this flag to protect us from
394 * concurrent deletion of inode. We cannot use reference to inode for this
395 * since we cannot afford doing last iput() on behalf of kjournald
396 */
397 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
398 /* Write allocated dirty buffers in this inode before commit */
399 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
400 /* Wait for outstanding data writes for this inode before commit */
401 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
402
403 /**
404 * struct jbd2_inode - The jbd_inode type is the structure linking inodes in
405 * ordered mode present in a transaction so that we can sync them during commit.
406 */
407 struct jbd2_inode {
408 /**
409 * @i_transaction:
410 *
411 * Which transaction does this inode belong to? Either the running
412 * transaction or the committing one. [j_list_lock]
413 */
414 transaction_t *i_transaction;
415
416 /**
417 * @i_next_transaction:
418 *
419 * Pointer to the running transaction modifying inode's data in case
420 * there is already a committing transaction touching it. [j_list_lock]
421 */
422 transaction_t *i_next_transaction;
423
424 /**
425 * @i_list: List of inodes in the i_transaction [j_list_lock]
426 */
427 struct list_head i_list;
428
429 /**
430 * @i_vfs_inode:
431 *
432 * VFS inode this inode belongs to [constant for lifetime of structure]
433 */
434 struct inode *i_vfs_inode;
435
436 /**
437 * @i_flags: Flags of inode [j_list_lock]
438 */
439 unsigned long i_flags;
440
441 /**
442 * @i_dirty_start:
443 *
444 * Offset in bytes where the dirty range for this inode starts.
445 * [j_list_lock]
446 */
447 loff_t i_dirty_start;
448
449 /**
450 * @i_dirty_end:
451 *
452 * Inclusive offset in bytes where the dirty range for this inode
453 * ends. [j_list_lock]
454 */
455 loff_t i_dirty_end;
456 };
457
458 struct jbd2_revoke_table_s;
459
460 /**
461 * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete
462 * type associated with handle_t.
463 * @h_transaction: Which compound transaction is this update a part of?
464 * @h_journal: Which journal handle belongs to - used iff h_reserved set.
465 * @h_rsv_handle: Handle reserved for finishing the logical operation.
466 * @h_total_credits: Number of remaining buffers we are allowed to add to
467 * journal. These are dirty buffers and revoke descriptor blocks.
468 * @h_revoke_credits: Number of remaining revoke records available for handle
469 * @h_ref: Reference count on this handle.
470 * @h_err: Field for caller's use to track errors through large fs operations.
471 * @h_sync: Flag for sync-on-close.
472 * @h_jdata: Flag to force data journaling.
473 * @h_reserved: Flag for handle for reserved credits.
474 * @h_aborted: Flag indicating fatal error on handle.
475 * @h_type: For handle statistics.
476 * @h_line_no: For handle statistics.
477 * @h_start_jiffies: Handle Start time.
478 * @h_requested_credits: Holds @h_total_credits after handle is started.
479 * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
480 * @saved_alloc_context: Saved context while transaction is open.
481 **/
482
483 /* Docbook can't yet cope with the bit fields, but will leave the documentation
484 * in so it can be fixed later.
485 */
486
487 struct jbd2_journal_handle
488 {
489 union {
490 transaction_t *h_transaction;
491 /* Which journal handle belongs to - used iff h_reserved set */
492 journal_t *h_journal;
493 };
494
495 handle_t *h_rsv_handle;
496 int h_total_credits;
497 int h_revoke_credits;
498 int h_revoke_credits_requested;
499 int h_ref;
500 int h_err;
501
502 /* Flags [no locking] */
503 unsigned int h_sync: 1;
504 unsigned int h_jdata: 1;
505 unsigned int h_reserved: 1;
506 unsigned int h_aborted: 1;
507 unsigned int h_type: 8;
508 unsigned int h_line_no: 16;
509
510 unsigned long h_start_jiffies;
511 unsigned int h_requested_credits;
512
513 unsigned int saved_alloc_context;
514 };
515
516
517 /*
518 * Some stats for checkpoint phase
519 */
520 struct transaction_chp_stats_s {
521 unsigned long cs_chp_time;
522 __u32 cs_forced_to_close;
523 __u32 cs_written;
524 __u32 cs_dropped;
525 };
526
527 /* The transaction_t type is the guts of the journaling mechanism. It
528 * tracks a compound transaction through its various states:
529 *
530 * RUNNING: accepting new updates
531 * LOCKED: Updates still running but we don't accept new ones
532 * RUNDOWN: Updates are tidying up but have finished requesting
533 * new buffers to modify (state not used for now)
534 * FLUSH: All updates complete, but we are still writing to disk
535 * COMMIT: All data on disk, writing commit record
536 * FINISHED: We still have to keep the transaction for checkpointing.
537 *
538 * The transaction keeps track of all of the buffers modified by a
539 * running transaction, and all of the buffers committed but not yet
540 * flushed to home for finished transactions.
541 */
542
543 /*
544 * Lock ranking:
545 *
546 * j_list_lock
547 * ->jbd_lock_bh_journal_head() (This is "innermost")
548 *
549 * j_state_lock
550 * ->b_state_lock
551 *
552 * b_state_lock
553 * ->j_list_lock
554 *
555 * j_state_lock
556 * ->t_handle_lock
557 *
558 * j_state_lock
559 * ->j_list_lock (journal_unmap_buffer)
560 *
561 */
562
563 struct transaction_s
564 {
565 /* Pointer to the journal for this transaction. [no locking] */
566 journal_t *t_journal;
567
568 /* Sequence number for this transaction [no locking] */
569 tid_t t_tid;
570
571 /*
572 * Transaction's current state
573 * [no locking - only kjournald2 alters this]
574 * [j_list_lock] guards transition of a transaction into T_FINISHED
575 * state and subsequent call of __jbd2_journal_drop_transaction()
576 * FIXME: needs barriers
577 * KLUDGE: [use j_state_lock]
578 */
579 enum {
580 T_RUNNING,
581 T_LOCKED,
582 T_SWITCH,
583 T_FLUSH,
584 T_COMMIT,
585 T_COMMIT_DFLUSH,
586 T_COMMIT_JFLUSH,
587 T_COMMIT_CALLBACK,
588 T_FINISHED
589 } t_state;
590
591 /*
592 * Where in the log does this transaction's commit start? [no locking]
593 */
594 unsigned long t_log_start;
595
596 /* Number of buffers on the t_buffers list [j_list_lock] */
597 int t_nr_buffers;
598
599 /*
600 * Doubly-linked circular list of all buffers reserved but not yet
601 * modified by this transaction [j_list_lock]
602 */
603 struct journal_head *t_reserved_list;
604
605 /*
606 * Doubly-linked circular list of all metadata buffers owned by this
607 * transaction [j_list_lock]
608 */
609 struct journal_head *t_buffers;
610
611 /*
612 * Doubly-linked circular list of all forget buffers (superseded
613 * buffers which we can un-checkpoint once this transaction commits)
614 * [j_list_lock]
615 */
616 struct journal_head *t_forget;
617
618 /*
619 * Doubly-linked circular list of all buffers still to be flushed before
620 * this transaction can be checkpointed. [j_list_lock]
621 */
622 struct journal_head *t_checkpoint_list;
623
624 /*
625 * Doubly-linked circular list of all buffers submitted for IO while
626 * checkpointing. [j_list_lock]
627 */
628 struct journal_head *t_checkpoint_io_list;
629
630 /*
631 * Doubly-linked circular list of metadata buffers being shadowed by log
632 * IO. The IO buffers on the iobuf list and the shadow buffers on this
633 * list match each other one for one at all times. [j_list_lock]
634 */
635 struct journal_head *t_shadow_list;
636
637 /*
638 * List of inodes associated with the transaction; e.g., ext4 uses
639 * this to track inodes in data=ordered and data=journal mode that
640 * need special handling on transaction commit; also used by ocfs2.
641 * [j_list_lock]
642 */
643 struct list_head t_inode_list;
644
645 /*
646 * Protects info related to handles
647 */
648 spinlock_t t_handle_lock;
649
650 /*
651 * Longest time some handle had to wait for running transaction
652 */
653 unsigned long t_max_wait;
654
655 /*
656 * When transaction started
657 */
658 unsigned long t_start;
659
660 /*
661 * When commit was requested
662 */
663 unsigned long t_requested;
664
665 /*
666 * Checkpointing stats [j_checkpoint_sem]
667 */
668 struct transaction_chp_stats_s t_chp_stats;
669
670 /*
671 * Number of outstanding updates running on this transaction
672 * [none]
673 */
674 atomic_t t_updates;
675
676 /*
677 * Number of blocks reserved for this transaction in the journal.
678 * This is including all credits reserved when starting transaction
679 * handles as well as all journal descriptor blocks needed for this
680 * transaction. [none]
681 */
682 atomic_t t_outstanding_credits;
683
684 /*
685 * Number of revoke records for this transaction added by already
686 * stopped handles. [none]
687 */
688 atomic_t t_outstanding_revokes;
689
690 /*
691 * How many handles used this transaction? [none]
692 */
693 atomic_t t_handle_count;
694
695 /*
696 * Forward and backward links for the circular list of all transactions
697 * awaiting checkpoint. [j_list_lock]
698 */
699 transaction_t *t_cpnext, *t_cpprev;
700
701 /*
702 * When will the transaction expire (become due for commit), in jiffies?
703 * [no locking]
704 */
705 unsigned long t_expires;
706
707 /*
708 * When this transaction started, in nanoseconds [no locking]
709 */
710 ktime_t t_start_time;
711
712 /*
713 * This transaction is being forced and some process is
714 * waiting for it to finish.
715 */
716 unsigned int t_synchronous_commit:1;
717
718 /* Disk flush needs to be sent to fs partition [no locking] */
719 int t_need_data_flush;
720
721 /*
722 * For use by the filesystem to store fs-specific data
723 * structures associated with the transaction
724 */
725 struct list_head t_private_list;
726 };
727
728 struct transaction_run_stats_s {
729 unsigned long rs_wait;
730 unsigned long rs_request_delay;
731 unsigned long rs_running;
732 unsigned long rs_locked;
733 unsigned long rs_flushing;
734 unsigned long rs_logging;
735
736 __u32 rs_handle_count;
737 __u32 rs_blocks;
738 __u32 rs_blocks_logged;
739 };
740
741 struct transaction_stats_s {
742 unsigned long ts_tid;
743 unsigned long ts_requested;
744 struct transaction_run_stats_s run;
745 };
746
747 static inline unsigned long
jbd2_time_diff(unsigned long start,unsigned long end)748 jbd2_time_diff(unsigned long start, unsigned long end)
749 {
750 if (end >= start)
751 return end - start;
752
753 return end + (MAX_JIFFY_OFFSET - start);
754 }
755
756 #define JBD2_NR_BATCH 64
757
758 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
759
760 #define JBD2_FC_REPLAY_STOP 0
761 #define JBD2_FC_REPLAY_CONTINUE 1
762
763 /**
764 * struct journal_s - The journal_s type is the concrete type associated with
765 * journal_t.
766 */
767 struct journal_s
768 {
769 /**
770 * @j_flags: General journaling state flags [j_state_lock]
771 */
772 unsigned long j_flags;
773
774 /**
775 * @j_atomic_flags: Atomic journaling state flags.
776 */
777 unsigned long j_atomic_flags;
778
779 /**
780 * @j_errno:
781 *
782 * Is there an outstanding uncleared error on the journal (from a prior
783 * abort)? [j_state_lock]
784 */
785 int j_errno;
786
787 /**
788 * @j_abort_mutex: Lock the whole aborting procedure.
789 */
790 struct mutex j_abort_mutex;
791
792 /**
793 * @j_sb_buffer: The first part of the superblock buffer.
794 */
795 struct buffer_head *j_sb_buffer;
796
797 /**
798 * @j_superblock: The second part of the superblock buffer.
799 */
800 journal_superblock_t *j_superblock;
801
802 /**
803 * @j_format_version: Version of the superblock format.
804 */
805 int j_format_version;
806
807 /**
808 * @j_state_lock: Protect the various scalars in the journal.
809 */
810 rwlock_t j_state_lock;
811
812 /**
813 * @j_barrier_count:
814 *
815 * Number of processes waiting to create a barrier lock [j_state_lock]
816 */
817 int j_barrier_count;
818
819 /**
820 * @j_barrier: The barrier lock itself.
821 */
822 struct mutex j_barrier;
823
824 /**
825 * @j_running_transaction:
826 *
827 * Transactions: The current running transaction...
828 * [j_state_lock] [caller holding open handle]
829 */
830 transaction_t *j_running_transaction;
831
832 /**
833 * @j_committing_transaction:
834 *
835 * the transaction we are pushing to disk
836 * [j_state_lock] [caller holding open handle]
837 */
838 transaction_t *j_committing_transaction;
839
840 /**
841 * @j_checkpoint_transactions:
842 *
843 * ... and a linked circular list of all transactions waiting for
844 * checkpointing. [j_list_lock]
845 */
846 transaction_t *j_checkpoint_transactions;
847
848 /**
849 * @j_wait_transaction_locked:
850 *
851 * Wait queue for waiting for a locked transaction to start committing,
852 * or for a barrier lock to be released.
853 */
854 wait_queue_head_t j_wait_transaction_locked;
855
856 /**
857 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
858 */
859 wait_queue_head_t j_wait_done_commit;
860
861 /**
862 * @j_wait_commit: Wait queue to trigger commit.
863 */
864 wait_queue_head_t j_wait_commit;
865
866 /**
867 * @j_wait_updates: Wait queue to wait for updates to complete.
868 */
869 wait_queue_head_t j_wait_updates;
870
871 /**
872 * @j_wait_reserved:
873 *
874 * Wait queue to wait for reserved buffer credits to drop.
875 */
876 wait_queue_head_t j_wait_reserved;
877
878 /**
879 * @j_fc_wait:
880 *
881 * Wait queue to wait for completion of async fast commits.
882 */
883 wait_queue_head_t j_fc_wait;
884
885 /**
886 * @j_checkpoint_mutex:
887 *
888 * Semaphore for locking against concurrent checkpoints.
889 */
890 struct mutex j_checkpoint_mutex;
891
892 /**
893 * @j_chkpt_bhs:
894 *
895 * List of buffer heads used by the checkpoint routine. This
896 * was moved from jbd2_log_do_checkpoint() to reduce stack
897 * usage. Access to this array is controlled by the
898 * @j_checkpoint_mutex. [j_checkpoint_mutex]
899 */
900 struct buffer_head *j_chkpt_bhs[JBD2_NR_BATCH];
901
902 /**
903 * @j_shrinker:
904 *
905 * Journal head shrinker, reclaim buffer's journal head which
906 * has been written back.
907 */
908 struct shrinker j_shrinker;
909
910 /**
911 * @j_checkpoint_jh_count:
912 *
913 * Number of journal buffers on the checkpoint list. [j_list_lock]
914 */
915 struct percpu_counter j_checkpoint_jh_count;
916
917 /**
918 * @j_shrink_transaction:
919 *
920 * Record next transaction will shrink on the checkpoint list.
921 * [j_list_lock]
922 */
923 transaction_t *j_shrink_transaction;
924
925 /**
926 * @j_head:
927 *
928 * Journal head: identifies the first unused block in the journal.
929 * [j_state_lock]
930 */
931 unsigned long j_head;
932
933 /**
934 * @j_tail:
935 *
936 * Journal tail: identifies the oldest still-used block in the journal.
937 * [j_state_lock]
938 */
939 unsigned long j_tail;
940
941 /**
942 * @j_free:
943 *
944 * Journal free: how many free blocks are there in the journal?
945 * [j_state_lock]
946 */
947 unsigned long j_free;
948
949 /**
950 * @j_first:
951 *
952 * The block number of the first usable block in the journal
953 * [j_state_lock].
954 */
955 unsigned long j_first;
956
957 /**
958 * @j_last:
959 *
960 * The block number one beyond the last usable block in the journal
961 * [j_state_lock].
962 */
963 unsigned long j_last;
964
965 /**
966 * @j_fc_first:
967 *
968 * The block number of the first fast commit block in the journal
969 * [j_state_lock].
970 */
971 unsigned long j_fc_first;
972
973 /**
974 * @j_fc_off:
975 *
976 * Number of fast commit blocks currently allocated. Accessed only
977 * during fast commit. Currently only process can do fast commit, so
978 * this field is not protected by any lock.
979 */
980 unsigned long j_fc_off;
981
982 /**
983 * @j_fc_last:
984 *
985 * The block number one beyond the last fast commit block in the journal
986 * [j_state_lock].
987 */
988 unsigned long j_fc_last;
989
990 /**
991 * @j_dev: Device where we store the journal.
992 */
993 struct block_device *j_dev;
994
995 /**
996 * @j_blocksize: Block size for the location where we store the journal.
997 */
998 int j_blocksize;
999
1000 /**
1001 * @j_blk_offset:
1002 *
1003 * Starting block offset into the device where we store the journal.
1004 */
1005 unsigned long long j_blk_offset;
1006
1007 /**
1008 * @j_devname: Journal device name.
1009 */
1010 char j_devname[BDEVNAME_SIZE+24];
1011
1012 /**
1013 * @j_fs_dev:
1014 *
1015 * Device which holds the client fs. For internal journal this will be
1016 * equal to j_dev.
1017 */
1018 struct block_device *j_fs_dev;
1019
1020 /**
1021 * @j_total_len: Total maximum capacity of the journal region on disk.
1022 */
1023 unsigned int j_total_len;
1024
1025 /**
1026 * @j_reserved_credits:
1027 *
1028 * Number of buffers reserved from the running transaction.
1029 */
1030 atomic_t j_reserved_credits;
1031
1032 /**
1033 * @j_list_lock: Protects the buffer lists and internal buffer state.
1034 */
1035 spinlock_t j_list_lock;
1036
1037 /**
1038 * @j_inode:
1039 *
1040 * Optional inode where we store the journal. If present, all
1041 * journal block numbers are mapped into this inode via bmap().
1042 */
1043 struct inode *j_inode;
1044
1045 /**
1046 * @j_tail_sequence:
1047 *
1048 * Sequence number of the oldest transaction in the log [j_state_lock]
1049 */
1050 tid_t j_tail_sequence;
1051
1052 /**
1053 * @j_transaction_sequence:
1054 *
1055 * Sequence number of the next transaction to grant [j_state_lock]
1056 */
1057 tid_t j_transaction_sequence;
1058
1059 /**
1060 * @j_commit_sequence:
1061 *
1062 * Sequence number of the most recently committed transaction
1063 * [j_state_lock].
1064 */
1065 tid_t j_commit_sequence;
1066
1067 /**
1068 * @j_commit_request:
1069 *
1070 * Sequence number of the most recent transaction wanting commit
1071 * [j_state_lock]
1072 */
1073 tid_t j_commit_request;
1074
1075 /**
1076 * @j_uuid:
1077 *
1078 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1079 * backed by this journal. This will eventually be replaced by an array
1080 * of uuids, allowing us to index multiple devices within a single
1081 * journal and to perform atomic updates across them.
1082 */
1083 __u8 j_uuid[16];
1084
1085 /**
1086 * @j_task: Pointer to the current commit thread for this journal.
1087 */
1088 struct task_struct *j_task;
1089
1090 /**
1091 * @j_max_transaction_buffers:
1092 *
1093 * Maximum number of metadata buffers to allow in a single compound
1094 * commit transaction.
1095 */
1096 int j_max_transaction_buffers;
1097
1098 /**
1099 * @j_revoke_records_per_block:
1100 *
1101 * Number of revoke records that fit in one descriptor block.
1102 */
1103 int j_revoke_records_per_block;
1104
1105 /**
1106 * @j_commit_interval:
1107 *
1108 * What is the maximum transaction lifetime before we begin a commit?
1109 */
1110 unsigned long j_commit_interval;
1111
1112 /**
1113 * @j_commit_timer: The timer used to wakeup the commit thread.
1114 */
1115 struct timer_list j_commit_timer;
1116
1117 /**
1118 * @j_revoke_lock: Protect the revoke table.
1119 */
1120 spinlock_t j_revoke_lock;
1121
1122 /**
1123 * @j_revoke:
1124 *
1125 * The revoke table - maintains the list of revoked blocks in the
1126 * current transaction.
1127 */
1128 struct jbd2_revoke_table_s *j_revoke;
1129
1130 /**
1131 * @j_revoke_table: Alternate revoke tables for j_revoke.
1132 */
1133 struct jbd2_revoke_table_s *j_revoke_table[2];
1134
1135 /**
1136 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1137 */
1138 struct buffer_head **j_wbuf;
1139
1140 /**
1141 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only
1142 * during a fast commit. Currently only process can do fast commit, so
1143 * this field is not protected by any lock.
1144 */
1145 struct buffer_head **j_fc_wbuf;
1146
1147 /**
1148 * @j_wbufsize:
1149 *
1150 * Size of @j_wbuf array.
1151 */
1152 int j_wbufsize;
1153
1154 /**
1155 * @j_fc_wbufsize:
1156 *
1157 * Size of @j_fc_wbuf array.
1158 */
1159 int j_fc_wbufsize;
1160
1161 /**
1162 * @j_last_sync_writer:
1163 *
1164 * The pid of the last person to run a synchronous operation
1165 * through the journal.
1166 */
1167 pid_t j_last_sync_writer;
1168
1169 /**
1170 * @j_average_commit_time:
1171 *
1172 * The average amount of time in nanoseconds it takes to commit a
1173 * transaction to disk. [j_state_lock]
1174 */
1175 u64 j_average_commit_time;
1176
1177 /**
1178 * @j_min_batch_time:
1179 *
1180 * Minimum time that we should wait for additional filesystem operations
1181 * to get batched into a synchronous handle in microseconds.
1182 */
1183 u32 j_min_batch_time;
1184
1185 /**
1186 * @j_max_batch_time:
1187 *
1188 * Maximum time that we should wait for additional filesystem operations
1189 * to get batched into a synchronous handle in microseconds.
1190 */
1191 u32 j_max_batch_time;
1192
1193 /**
1194 * @j_commit_callback:
1195 *
1196 * This function is called when a transaction is closed.
1197 */
1198 void (*j_commit_callback)(journal_t *,
1199 transaction_t *);
1200
1201 /**
1202 * @j_submit_inode_data_buffers:
1203 *
1204 * This function is called for all inodes associated with the
1205 * committing transaction marked with JI_WRITE_DATA flag
1206 * before we start to write out the transaction to the journal.
1207 */
1208 int (*j_submit_inode_data_buffers)
1209 (struct jbd2_inode *);
1210
1211 /**
1212 * @j_finish_inode_data_buffers:
1213 *
1214 * This function is called for all inodes associated with the
1215 * committing transaction marked with JI_WAIT_DATA flag
1216 * after we have written the transaction to the journal
1217 * but before we write out the commit block.
1218 */
1219 int (*j_finish_inode_data_buffers)
1220 (struct jbd2_inode *);
1221
1222 /*
1223 * Journal statistics
1224 */
1225
1226 /**
1227 * @j_history_lock: Protect the transactions statistics history.
1228 */
1229 spinlock_t j_history_lock;
1230
1231 /**
1232 * @j_proc_entry: procfs entry for the jbd statistics directory.
1233 */
1234 struct proc_dir_entry *j_proc_entry;
1235
1236 /**
1237 * @j_stats: Overall statistics.
1238 */
1239 struct transaction_stats_s j_stats;
1240
1241 /**
1242 * @j_failed_commit: Failed journal commit ID.
1243 */
1244 unsigned int j_failed_commit;
1245
1246 /**
1247 * @j_private:
1248 *
1249 * An opaque pointer to fs-private information. ext3 puts its
1250 * superblock pointer here.
1251 */
1252 void *j_private;
1253
1254 /**
1255 * @j_chksum_driver:
1256 *
1257 * Reference to checksum algorithm driver via cryptoapi.
1258 */
1259 struct crypto_shash *j_chksum_driver;
1260
1261 /**
1262 * @j_csum_seed:
1263 *
1264 * Precomputed journal UUID checksum for seeding other checksums.
1265 */
1266 __u32 j_csum_seed;
1267
1268 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1269 /**
1270 * @j_trans_commit_map:
1271 *
1272 * Lockdep entity to track transaction commit dependencies. Handles
1273 * hold this "lock" for read, when we wait for commit, we acquire the
1274 * "lock" for writing. This matches the properties of jbd2 journalling
1275 * where the running transaction has to wait for all handles to be
1276 * dropped to commit that transaction and also acquiring a handle may
1277 * require transaction commit to finish.
1278 */
1279 struct lockdep_map j_trans_commit_map;
1280 #endif
1281
1282 /**
1283 * @j_fc_cleanup_callback:
1284 *
1285 * Clean-up after fast commit or full commit. JBD2 calls this function
1286 * after every commit operation.
1287 */
1288 void (*j_fc_cleanup_callback)(struct journal_s *journal, int);
1289
1290 /**
1291 * @j_fc_replay_callback:
1292 *
1293 * File-system specific function that performs replay of a fast
1294 * commit. JBD2 calls this function for each fast commit block found in
1295 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1296 * to indicate that the block was processed correctly and more fast
1297 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1298 * indicates the end of replay (no more blocks remaining). A negative
1299 * return value indicates error.
1300 */
1301 int (*j_fc_replay_callback)(struct journal_s *journal,
1302 struct buffer_head *bh,
1303 enum passtype pass, int off,
1304 tid_t expected_commit_id);
1305 };
1306
1307 #define jbd2_might_wait_for_commit(j) \
1308 do { \
1309 rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1310 rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1311 } while (0)
1312
1313 /* journal feature predicate functions */
1314 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1315 static inline bool jbd2_has_feature_##name(journal_t *j) \
1316 { \
1317 return ((j)->j_format_version >= 2 && \
1318 ((j)->j_superblock->s_feature_compat & \
1319 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1320 } \
1321 static inline void jbd2_set_feature_##name(journal_t *j) \
1322 { \
1323 (j)->j_superblock->s_feature_compat |= \
1324 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1325 } \
1326 static inline void jbd2_clear_feature_##name(journal_t *j) \
1327 { \
1328 (j)->j_superblock->s_feature_compat &= \
1329 ~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1330 }
1331
1332 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1333 static inline bool jbd2_has_feature_##name(journal_t *j) \
1334 { \
1335 return ((j)->j_format_version >= 2 && \
1336 ((j)->j_superblock->s_feature_ro_compat & \
1337 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1338 } \
1339 static inline void jbd2_set_feature_##name(journal_t *j) \
1340 { \
1341 (j)->j_superblock->s_feature_ro_compat |= \
1342 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1343 } \
1344 static inline void jbd2_clear_feature_##name(journal_t *j) \
1345 { \
1346 (j)->j_superblock->s_feature_ro_compat &= \
1347 ~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1348 }
1349
1350 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1351 static inline bool jbd2_has_feature_##name(journal_t *j) \
1352 { \
1353 return ((j)->j_format_version >= 2 && \
1354 ((j)->j_superblock->s_feature_incompat & \
1355 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1356 } \
1357 static inline void jbd2_set_feature_##name(journal_t *j) \
1358 { \
1359 (j)->j_superblock->s_feature_incompat |= \
1360 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1361 } \
1362 static inline void jbd2_clear_feature_##name(journal_t *j) \
1363 { \
1364 (j)->j_superblock->s_feature_incompat &= \
1365 ~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1366 }
1367
1368 JBD2_FEATURE_COMPAT_FUNCS(checksum, CHECKSUM)
1369
1370 JBD2_FEATURE_INCOMPAT_FUNCS(revoke, REVOKE)
1371 JBD2_FEATURE_INCOMPAT_FUNCS(64bit, 64BIT)
1372 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit, ASYNC_COMMIT)
1373 JBD2_FEATURE_INCOMPAT_FUNCS(csum2, CSUM_V2)
1374 JBD2_FEATURE_INCOMPAT_FUNCS(csum3, CSUM_V3)
1375 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit, FAST_COMMIT)
1376
1377 /*
1378 * Journal flag definitions
1379 */
1380 #define JBD2_UNMOUNT 0x001 /* Journal thread is being destroyed */
1381 #define JBD2_ABORT 0x002 /* Journaling has been aborted for errors. */
1382 #define JBD2_ACK_ERR 0x004 /* The errno in the sb has been acked */
1383 #define JBD2_FLUSHED 0x008 /* The journal superblock has been flushed */
1384 #define JBD2_LOADED 0x010 /* The journal superblock has been loaded */
1385 #define JBD2_BARRIER 0x020 /* Use IDE barriers */
1386 #define JBD2_ABORT_ON_SYNCDATA_ERR 0x040 /* Abort the journal on file
1387 * data write error in ordered
1388 * mode */
1389 #define JBD2_FAST_COMMIT_ONGOING 0x100 /* Fast commit is ongoing */
1390 #define JBD2_FULL_COMMIT_ONGOING 0x200 /* Full commit is ongoing */
1391
1392 /*
1393 * Journal atomic flag definitions
1394 */
1395 #define JBD2_CHECKPOINT_IO_ERROR 0x001 /* Detect io error while writing
1396 * buffer back to disk */
1397
1398 /*
1399 * Function declarations for the journaling transaction and buffer
1400 * management
1401 */
1402
1403 /* Filing buffers */
1404 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1405 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1406 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1407 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1408 extern void __journal_free_buffer(struct journal_head *bh);
1409 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1410 extern void __journal_clean_data_list(transaction_t *transaction);
jbd2_file_log_bh(struct list_head * head,struct buffer_head * bh)1411 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1412 {
1413 list_add_tail(&bh->b_assoc_buffers, head);
1414 }
jbd2_unfile_log_bh(struct buffer_head * bh)1415 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1416 {
1417 list_del_init(&bh->b_assoc_buffers);
1418 }
1419
1420 /* Log buffer allocation */
1421 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1422 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1423 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1424 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1425 unsigned long *block);
1426 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1427 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1428
1429 /* Commit management */
1430 extern void jbd2_journal_commit_transaction(journal_t *);
1431
1432 /* Checkpoint list management */
1433 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1434 unsigned long jbd2_journal_shrink_checkpoint_list(journal_t *journal, unsigned long *nr_to_scan);
1435 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1436 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1437 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1438
1439
1440 /*
1441 * Triggers
1442 */
1443
1444 struct jbd2_buffer_trigger_type {
1445 /*
1446 * Fired a the moment data to write to the journal are known to be
1447 * stable - so either at the moment b_frozen_data is created or just
1448 * before a buffer is written to the journal. mapped_data is a mapped
1449 * buffer that is the frozen data for commit.
1450 */
1451 void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1452 struct buffer_head *bh, void *mapped_data,
1453 size_t size);
1454
1455 /*
1456 * Fired during journal abort for dirty buffers that will not be
1457 * committed.
1458 */
1459 void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1460 struct buffer_head *bh);
1461 };
1462
1463 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1464 void *mapped_data,
1465 struct jbd2_buffer_trigger_type *triggers);
1466 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1467 struct jbd2_buffer_trigger_type *triggers);
1468
1469 /* Buffer IO */
1470 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1471 struct journal_head *jh_in,
1472 struct buffer_head **bh_out,
1473 sector_t blocknr);
1474
1475 /* Transaction locking */
1476 extern void __wait_on_journal (journal_t *);
1477
1478 /* Transaction cache support */
1479 extern void jbd2_journal_destroy_transaction_cache(void);
1480 extern int __init jbd2_journal_init_transaction_cache(void);
1481 extern void jbd2_journal_free_transaction(transaction_t *);
1482
1483 /*
1484 * Journal locking.
1485 *
1486 * We need to lock the journal during transaction state changes so that nobody
1487 * ever tries to take a handle on the running transaction while we are in the
1488 * middle of moving it to the commit phase. j_state_lock does this.
1489 *
1490 * Note that the locking is completely interrupt unsafe. We never touch
1491 * journal structures from interrupts.
1492 */
1493
journal_current_handle(void)1494 static inline handle_t *journal_current_handle(void)
1495 {
1496 return current->journal_info;
1497 }
1498
1499 /* The journaling code user interface:
1500 *
1501 * Create and destroy handles
1502 * Register buffer modifications against the current transaction.
1503 */
1504
1505 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1506 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1507 int revoke_records, gfp_t gfp_mask,
1508 unsigned int type, unsigned int line_no);
1509 extern int jbd2_journal_restart(handle_t *, int nblocks);
1510 extern int jbd2__journal_restart(handle_t *, int nblocks,
1511 int revoke_records, gfp_t gfp_mask);
1512 extern int jbd2_journal_start_reserved(handle_t *handle,
1513 unsigned int type, unsigned int line_no);
1514 extern void jbd2_journal_free_reserved(handle_t *handle);
1515 extern int jbd2_journal_extend(handle_t *handle, int nblocks,
1516 int revoke_records);
1517 extern int jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1518 extern int jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1519 extern int jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1520 void jbd2_journal_set_triggers(struct buffer_head *,
1521 struct jbd2_buffer_trigger_type *type);
1522 extern int jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1523 extern int jbd2_journal_forget (handle_t *, struct buffer_head *);
1524 extern int jbd2_journal_invalidatepage(journal_t *,
1525 struct page *, unsigned int, unsigned int);
1526 extern int jbd2_journal_try_to_free_buffers(journal_t *journal, struct page *page);
1527 extern int jbd2_journal_stop(handle_t *);
1528 extern int jbd2_journal_flush (journal_t *);
1529 extern void jbd2_journal_lock_updates (journal_t *);
1530 extern void jbd2_journal_unlock_updates (journal_t *);
1531
1532 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1533 struct block_device *fs_dev,
1534 unsigned long long start, int len, int bsize);
1535 extern journal_t * jbd2_journal_init_inode (struct inode *);
1536 extern int jbd2_journal_update_format (journal_t *);
1537 extern int jbd2_journal_check_used_features
1538 (journal_t *, unsigned long, unsigned long, unsigned long);
1539 extern int jbd2_journal_check_available_features
1540 (journal_t *, unsigned long, unsigned long, unsigned long);
1541 extern int jbd2_journal_set_features
1542 (journal_t *, unsigned long, unsigned long, unsigned long);
1543 extern void jbd2_journal_clear_features
1544 (journal_t *, unsigned long, unsigned long, unsigned long);
1545 extern int jbd2_journal_load (journal_t *journal);
1546 extern int jbd2_journal_destroy (journal_t *);
1547 extern int jbd2_journal_recover (journal_t *journal);
1548 extern int jbd2_journal_wipe (journal_t *, int);
1549 extern int jbd2_journal_skip_recovery (journal_t *);
1550 extern void jbd2_journal_update_sb_errno(journal_t *);
1551 extern int jbd2_journal_update_sb_log_tail (journal_t *, tid_t,
1552 unsigned long, int);
1553 extern void jbd2_journal_abort (journal_t *, int);
1554 extern int jbd2_journal_errno (journal_t *);
1555 extern void jbd2_journal_ack_err (journal_t *);
1556 extern int jbd2_journal_clear_err (journal_t *);
1557 extern int jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1558 extern int jbd2_journal_force_commit(journal_t *);
1559 extern int jbd2_journal_force_commit_nested(journal_t *);
1560 extern int jbd2_journal_inode_ranged_write(handle_t *handle,
1561 struct jbd2_inode *inode, loff_t start_byte,
1562 loff_t length);
1563 extern int jbd2_journal_inode_ranged_wait(handle_t *handle,
1564 struct jbd2_inode *inode, loff_t start_byte,
1565 loff_t length);
1566 extern int jbd2_journal_submit_inode_data_buffers(
1567 struct jbd2_inode *jinode);
1568 extern int jbd2_journal_finish_inode_data_buffers(
1569 struct jbd2_inode *jinode);
1570 extern int jbd2_journal_begin_ordered_truncate(journal_t *journal,
1571 struct jbd2_inode *inode, loff_t new_size);
1572 extern void jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1573 extern void jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1574
1575 /*
1576 * journal_head management
1577 */
1578 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1579 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1580 void jbd2_journal_put_journal_head(struct journal_head *jh);
1581
1582 /*
1583 * handle management
1584 */
1585 extern struct kmem_cache *jbd2_handle_cache;
1586
jbd2_alloc_handle(gfp_t gfp_flags)1587 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1588 {
1589 return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1590 }
1591
jbd2_free_handle(handle_t * handle)1592 static inline void jbd2_free_handle(handle_t *handle)
1593 {
1594 kmem_cache_free(jbd2_handle_cache, handle);
1595 }
1596
1597 /*
1598 * jbd2_inode management (optional, for those file systems that want to use
1599 * dynamically allocated jbd2_inode structures)
1600 */
1601 extern struct kmem_cache *jbd2_inode_cache;
1602
jbd2_alloc_inode(gfp_t gfp_flags)1603 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1604 {
1605 return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1606 }
1607
jbd2_free_inode(struct jbd2_inode * jinode)1608 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1609 {
1610 kmem_cache_free(jbd2_inode_cache, jinode);
1611 }
1612
1613 /* Primary revoke support */
1614 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1615 extern int jbd2_journal_init_revoke(journal_t *, int);
1616 extern void jbd2_journal_destroy_revoke_record_cache(void);
1617 extern void jbd2_journal_destroy_revoke_table_cache(void);
1618 extern int __init jbd2_journal_init_revoke_record_cache(void);
1619 extern int __init jbd2_journal_init_revoke_table_cache(void);
1620
1621 extern void jbd2_journal_destroy_revoke(journal_t *);
1622 extern int jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1623 extern int jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1624 extern void jbd2_journal_write_revoke_records(transaction_t *transaction,
1625 struct list_head *log_bufs);
1626
1627 /* Recovery revoke support */
1628 extern int jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1629 extern int jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1630 extern void jbd2_journal_clear_revoke(journal_t *);
1631 extern void jbd2_journal_switch_revoke_table(journal_t *journal);
1632 extern void jbd2_clear_buffer_revoked_flags(journal_t *journal);
1633
1634 /*
1635 * The log thread user interface:
1636 *
1637 * Request space in the current transaction, and force transaction commit
1638 * transitions on demand.
1639 */
1640
1641 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1642 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1643 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1644 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1645 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1646 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1647 int jbd2_log_do_checkpoint(journal_t *journal);
1648 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1649
1650 void __jbd2_log_wait_for_space(journal_t *journal);
1651 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1652 extern int jbd2_cleanup_journal_tail(journal_t *);
1653
1654 /* Fast commit related APIs */
1655 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1656 int jbd2_fc_end_commit(journal_t *journal);
1657 int jbd2_fc_end_commit_fallback(journal_t *journal);
1658 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1659 int jbd2_submit_inode_data(struct jbd2_inode *jinode);
1660 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1661 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1662 int jbd2_fc_release_bufs(journal_t *journal);
1663
jbd2_journal_get_max_txn_bufs(journal_t * journal)1664 static inline int jbd2_journal_get_max_txn_bufs(journal_t *journal)
1665 {
1666 return (journal->j_total_len - journal->j_fc_wbufsize) / 4;
1667 }
1668
1669 /*
1670 * is_journal_abort
1671 *
1672 * Simple test wrapper function to test the JBD2_ABORT state flag. This
1673 * bit, when set, indicates that we have had a fatal error somewhere,
1674 * either inside the journaling layer or indicated to us by the client
1675 * (eg. ext3), and that we and should not commit any further
1676 * transactions.
1677 */
1678
is_journal_aborted(journal_t * journal)1679 static inline int is_journal_aborted(journal_t *journal)
1680 {
1681 return journal->j_flags & JBD2_ABORT;
1682 }
1683
is_handle_aborted(handle_t * handle)1684 static inline int is_handle_aborted(handle_t *handle)
1685 {
1686 if (handle->h_aborted || !handle->h_transaction)
1687 return 1;
1688 return is_journal_aborted(handle->h_transaction->t_journal);
1689 }
1690
jbd2_journal_abort_handle(handle_t * handle)1691 static inline void jbd2_journal_abort_handle(handle_t *handle)
1692 {
1693 handle->h_aborted = 1;
1694 }
1695
1696 #endif /* __KERNEL__ */
1697
1698 /* Comparison functions for transaction IDs: perform comparisons using
1699 * modulo arithmetic so that they work over sequence number wraps. */
1700
tid_gt(tid_t x,tid_t y)1701 static inline int tid_gt(tid_t x, tid_t y)
1702 {
1703 int difference = (x - y);
1704 return (difference > 0);
1705 }
1706
tid_geq(tid_t x,tid_t y)1707 static inline int tid_geq(tid_t x, tid_t y)
1708 {
1709 int difference = (x - y);
1710 return (difference >= 0);
1711 }
1712
1713 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1714 extern size_t journal_tag_bytes(journal_t *journal);
1715
jbd2_journal_has_csum_v2or3_feature(journal_t * j)1716 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1717 {
1718 return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1719 }
1720
jbd2_journal_has_csum_v2or3(journal_t * journal)1721 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1722 {
1723 WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1724 journal->j_chksum_driver == NULL);
1725
1726 return journal->j_chksum_driver != NULL;
1727 }
1728
1729 /*
1730 * Return number of free blocks in the log. Must be called under j_state_lock.
1731 */
jbd2_log_space_left(journal_t * journal)1732 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1733 {
1734 /* Allow for rounding errors */
1735 long free = journal->j_free - 32;
1736
1737 if (journal->j_committing_transaction) {
1738 free -= atomic_read(&journal->
1739 j_committing_transaction->t_outstanding_credits);
1740 }
1741 return max_t(long, free, 0);
1742 }
1743
1744 /*
1745 * Definitions which augment the buffer_head layer
1746 */
1747
1748 /* journaling buffer types */
1749 #define BJ_None 0 /* Not journaled */
1750 #define BJ_Metadata 1 /* Normal journaled metadata */
1751 #define BJ_Forget 2 /* Buffer superseded by this transaction */
1752 #define BJ_Shadow 3 /* Buffer contents being shadowed to the log */
1753 #define BJ_Reserved 4 /* Buffer is reserved for access by journal */
1754 #define BJ_Types 5
1755
1756 extern int jbd_blocks_per_page(struct inode *inode);
1757
1758 /* JBD uses a CRC32 checksum */
1759 #define JBD_MAX_CHECKSUM_SIZE 4
1760
jbd2_chksum(journal_t * journal,u32 crc,const void * address,unsigned int length)1761 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1762 const void *address, unsigned int length)
1763 {
1764 struct {
1765 struct shash_desc shash;
1766 char ctx[JBD_MAX_CHECKSUM_SIZE];
1767 } desc;
1768 int err;
1769
1770 BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1771 JBD_MAX_CHECKSUM_SIZE);
1772
1773 desc.shash.tfm = journal->j_chksum_driver;
1774 *(u32 *)desc.ctx = crc;
1775
1776 err = crypto_shash_update(&desc.shash, address, length);
1777 BUG_ON(err);
1778
1779 return *(u32 *)desc.ctx;
1780 }
1781
1782 /* Return most recent uncommitted transaction */
jbd2_get_latest_transaction(journal_t * journal)1783 static inline tid_t jbd2_get_latest_transaction(journal_t *journal)
1784 {
1785 tid_t tid;
1786
1787 read_lock(&journal->j_state_lock);
1788 tid = journal->j_commit_request;
1789 if (journal->j_running_transaction)
1790 tid = journal->j_running_transaction->t_tid;
1791 read_unlock(&journal->j_state_lock);
1792 return tid;
1793 }
1794
jbd2_handle_buffer_credits(handle_t * handle)1795 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1796 {
1797 journal_t *journal;
1798
1799 if (!handle->h_reserved)
1800 journal = handle->h_transaction->t_journal;
1801 else
1802 journal = handle->h_journal;
1803
1804 return handle->h_total_credits -
1805 DIV_ROUND_UP(handle->h_revoke_credits_requested,
1806 journal->j_revoke_records_per_block);
1807 }
1808
1809 #ifdef __KERNEL__
1810
1811 #define buffer_trace_init(bh) do {} while (0)
1812 #define print_buffer_fields(bh) do {} while (0)
1813 #define print_buffer_trace(bh) do {} while (0)
1814 #define BUFFER_TRACE(bh, info) do {} while (0)
1815 #define BUFFER_TRACE2(bh, bh2, info) do {} while (0)
1816 #define JBUFFER_TRACE(jh, info) do {} while (0)
1817
1818 #endif /* __KERNEL__ */
1819
1820 #define EFSBADCRC EBADMSG /* Bad CRC detected */
1821 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
1822
1823 #endif /* _LINUX_JBD2_H */
1824