xref: /linux/include/linux/jbd2.h (revision 4359a011e259a4608afc7fb3635370c9d4ba5943)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * linux/include/linux/jbd2.h
4  *
5  * Written by Stephen C. Tweedie <sct@redhat.com>
6  *
7  * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
8  *
9  * Definitions for transaction data structures for the buffer cache
10  * filesystem journaling support.
11  */
12 
13 #ifndef _LINUX_JBD2_H
14 #define _LINUX_JBD2_H
15 
16 /* Allow this file to be included directly into e2fsprogs */
17 #ifndef __KERNEL__
18 #include "jfs_compat.h"
19 #define JBD2_DEBUG
20 #else
21 
22 #include <linux/types.h>
23 #include <linux/buffer_head.h>
24 #include <linux/journal-head.h>
25 #include <linux/stddef.h>
26 #include <linux/mutex.h>
27 #include <linux/timer.h>
28 #include <linux/slab.h>
29 #include <linux/bit_spinlock.h>
30 #include <linux/blkdev.h>
31 #include <crypto/hash.h>
32 #endif
33 
34 #define journal_oom_retry 1
35 
36 /*
37  * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
38  * certain classes of error which can occur due to failed IOs.  Under
39  * normal use we want ext4 to continue after such errors, because
40  * hardware _can_ fail, but for debugging purposes when running tests on
41  * known-good hardware we may want to trap these errors.
42  */
43 #undef JBD2_PARANOID_IOFAIL
44 
45 /*
46  * The default maximum commit age, in seconds.
47  */
48 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
49 
50 #ifdef CONFIG_JBD2_DEBUG
51 /*
52  * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
53  * consistency checks.  By default we don't do this unless
54  * CONFIG_JBD2_DEBUG is on.
55  */
56 #define JBD2_EXPENSIVE_CHECKING
57 void __jbd2_debug(int level, const char *file, const char *func,
58 		  unsigned int line, const char *fmt, ...);
59 
60 #define jbd2_debug(n, fmt, a...) \
61 	__jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
62 #else
63 #define jbd2_debug(n, fmt, a...)  no_printk(fmt, ##a)
64 #endif
65 
66 extern void *jbd2_alloc(size_t size, gfp_t flags);
67 extern void jbd2_free(void *ptr, size_t size);
68 
69 #define JBD2_MIN_JOURNAL_BLOCKS 1024
70 #define JBD2_DEFAULT_FAST_COMMIT_BLOCKS 256
71 
72 #ifdef __KERNEL__
73 
74 /**
75  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
76  *
77  * All filesystem modifications made by the process go
78  * through this handle.  Recursive operations (such as quota operations)
79  * are gathered into a single update.
80  *
81  * The buffer credits field is used to account for journaled buffers
82  * being modified by the running process.  To ensure that there is
83  * enough log space for all outstanding operations, we need to limit the
84  * number of outstanding buffers possible at any time.  When the
85  * operation completes, any buffer credits not used are credited back to
86  * the transaction, so that at all times we know how many buffers the
87  * outstanding updates on a transaction might possibly touch.
88  *
89  * This is an opaque datatype.
90  **/
91 typedef struct jbd2_journal_handle handle_t;	/* Atomic operation type */
92 
93 
94 /**
95  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
96  *
97  * journal_t is linked to from the fs superblock structure.
98  *
99  * We use the journal_t to keep track of all outstanding transaction
100  * activity on the filesystem, and to manage the state of the log
101  * writing process.
102  *
103  * This is an opaque datatype.
104  **/
105 typedef struct journal_s	journal_t;	/* Journal control structure */
106 #endif
107 
108 /*
109  * Internal structures used by the logging mechanism:
110  */
111 
112 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
113 
114 /*
115  * On-disk structures
116  */
117 
118 /*
119  * Descriptor block types:
120  */
121 
122 #define JBD2_DESCRIPTOR_BLOCK	1
123 #define JBD2_COMMIT_BLOCK	2
124 #define JBD2_SUPERBLOCK_V1	3
125 #define JBD2_SUPERBLOCK_V2	4
126 #define JBD2_REVOKE_BLOCK	5
127 
128 /*
129  * Standard header for all descriptor blocks:
130  */
131 typedef struct journal_header_s
132 {
133 	__be32		h_magic;
134 	__be32		h_blocktype;
135 	__be32		h_sequence;
136 } journal_header_t;
137 
138 /*
139  * Checksum types.
140  */
141 #define JBD2_CRC32_CHKSUM   1
142 #define JBD2_MD5_CHKSUM     2
143 #define JBD2_SHA1_CHKSUM    3
144 #define JBD2_CRC32C_CHKSUM  4
145 
146 #define JBD2_CRC32_CHKSUM_SIZE 4
147 
148 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
149 /*
150  * Commit block header for storing transactional checksums:
151  *
152  * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
153  * fields are used to store a checksum of the descriptor and data blocks.
154  *
155  * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
156  * field is used to store crc32c(uuid+commit_block).  Each journal metadata
157  * block gets its own checksum, and data block checksums are stored in
158  * journal_block_tag (in the descriptor).  The other h_chksum* fields are
159  * not used.
160  *
161  * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
162  * journal_block_tag3_t to store a full 32-bit checksum.  Everything else
163  * is the same as v2.
164  *
165  * Checksum v1, v2, and v3 are mutually exclusive features.
166  */
167 struct commit_header {
168 	__be32		h_magic;
169 	__be32          h_blocktype;
170 	__be32          h_sequence;
171 	unsigned char   h_chksum_type;
172 	unsigned char   h_chksum_size;
173 	unsigned char 	h_padding[2];
174 	__be32 		h_chksum[JBD2_CHECKSUM_BYTES];
175 	__be64		h_commit_sec;
176 	__be32		h_commit_nsec;
177 };
178 
179 /*
180  * The block tag: used to describe a single buffer in the journal.
181  * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
182  * raw struct shouldn't be used for pointer math or sizeof() - use
183  * journal_tag_bytes(journal) instead to compute this.
184  */
185 typedef struct journal_block_tag3_s
186 {
187 	__be32		t_blocknr;	/* The on-disk block number */
188 	__be32		t_flags;	/* See below */
189 	__be32		t_blocknr_high; /* most-significant high 32bits. */
190 	__be32		t_checksum;	/* crc32c(uuid+seq+block) */
191 } journal_block_tag3_t;
192 
193 typedef struct journal_block_tag_s
194 {
195 	__be32		t_blocknr;	/* The on-disk block number */
196 	__be16		t_checksum;	/* truncated crc32c(uuid+seq+block) */
197 	__be16		t_flags;	/* See below */
198 	__be32		t_blocknr_high; /* most-significant high 32bits. */
199 } journal_block_tag_t;
200 
201 /* Tail of descriptor or revoke block, for checksumming */
202 struct jbd2_journal_block_tail {
203 	__be32		t_checksum;	/* crc32c(uuid+descr_block) */
204 };
205 
206 /*
207  * The revoke descriptor: used on disk to describe a series of blocks to
208  * be revoked from the log
209  */
210 typedef struct jbd2_journal_revoke_header_s
211 {
212 	journal_header_t r_header;
213 	__be32		 r_count;	/* Count of bytes used in the block */
214 } jbd2_journal_revoke_header_t;
215 
216 /* Definitions for the journal tag flags word: */
217 #define JBD2_FLAG_ESCAPE		1	/* on-disk block is escaped */
218 #define JBD2_FLAG_SAME_UUID	2	/* block has same uuid as previous */
219 #define JBD2_FLAG_DELETED	4	/* block deleted by this transaction */
220 #define JBD2_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
221 
222 
223 /*
224  * The journal superblock.  All fields are in big-endian byte order.
225  */
226 typedef struct journal_superblock_s
227 {
228 /* 0x0000 */
229 	journal_header_t s_header;
230 
231 /* 0x000C */
232 	/* Static information describing the journal */
233 	__be32	s_blocksize;		/* journal device blocksize */
234 	__be32	s_maxlen;		/* total blocks in journal file */
235 	__be32	s_first;		/* first block of log information */
236 
237 /* 0x0018 */
238 	/* Dynamic information describing the current state of the log */
239 	__be32	s_sequence;		/* first commit ID expected in log */
240 	__be32	s_start;		/* blocknr of start of log */
241 
242 /* 0x0020 */
243 	/* Error value, as set by jbd2_journal_abort(). */
244 	__be32	s_errno;
245 
246 /* 0x0024 */
247 	/* Remaining fields are only valid in a version-2 superblock */
248 	__be32	s_feature_compat;	/* compatible feature set */
249 	__be32	s_feature_incompat;	/* incompatible feature set */
250 	__be32	s_feature_ro_compat;	/* readonly-compatible feature set */
251 /* 0x0030 */
252 	__u8	s_uuid[16];		/* 128-bit uuid for journal */
253 
254 /* 0x0040 */
255 	__be32	s_nr_users;		/* Nr of filesystems sharing log */
256 
257 	__be32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
258 
259 /* 0x0048 */
260 	__be32	s_max_transaction;	/* Limit of journal blocks per trans.*/
261 	__be32	s_max_trans_data;	/* Limit of data blocks per trans. */
262 
263 /* 0x0050 */
264 	__u8	s_checksum_type;	/* checksum type */
265 	__u8	s_padding2[3];
266 /* 0x0054 */
267 	__be32	s_num_fc_blks;		/* Number of fast commit blocks */
268 /* 0x0058 */
269 	__u32	s_padding[41];
270 	__be32	s_checksum;		/* crc32c(superblock) */
271 
272 /* 0x0100 */
273 	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
274 /* 0x0400 */
275 } journal_superblock_t;
276 
277 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
278 #define JBD2_HAS_COMPAT_FEATURE(j,mask)					\
279 	((j)->j_format_version >= 2 &&					\
280 	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
281 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask)				\
282 	((j)->j_format_version >= 2 &&					\
283 	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
284 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask)				\
285 	((j)->j_format_version >= 2 &&					\
286 	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
287 
288 #define JBD2_FEATURE_COMPAT_CHECKSUM		0x00000001
289 
290 #define JBD2_FEATURE_INCOMPAT_REVOKE		0x00000001
291 #define JBD2_FEATURE_INCOMPAT_64BIT		0x00000002
292 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
293 #define JBD2_FEATURE_INCOMPAT_CSUM_V2		0x00000008
294 #define JBD2_FEATURE_INCOMPAT_CSUM_V3		0x00000010
295 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT	0x00000020
296 
297 /* See "journal feature predicate functions" below */
298 
299 /* Features known to this kernel version: */
300 #define JBD2_KNOWN_COMPAT_FEATURES	JBD2_FEATURE_COMPAT_CHECKSUM
301 #define JBD2_KNOWN_ROCOMPAT_FEATURES	0
302 #define JBD2_KNOWN_INCOMPAT_FEATURES	(JBD2_FEATURE_INCOMPAT_REVOKE | \
303 					JBD2_FEATURE_INCOMPAT_64BIT | \
304 					JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
305 					JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
306 					JBD2_FEATURE_INCOMPAT_CSUM_V3 | \
307 					JBD2_FEATURE_INCOMPAT_FAST_COMMIT)
308 
309 #ifdef __KERNEL__
310 
311 #include <linux/fs.h>
312 #include <linux/sched.h>
313 
314 enum jbd_state_bits {
315 	BH_JBD			/* Has an attached ext3 journal_head */
316 	  = BH_PrivateStart,
317 	BH_JWrite,		/* Being written to log (@@@ DEBUGGING) */
318 	BH_Freed,		/* Has been freed (truncated) */
319 	BH_Revoked,		/* Has been revoked from the log */
320 	BH_RevokeValid,		/* Revoked flag is valid */
321 	BH_JBDDirty,		/* Is dirty but journaled */
322 	BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
323 	BH_Shadow,		/* IO on shadow buffer is running */
324 	BH_Verified,		/* Metadata block has been verified ok */
325 	BH_JBDPrivateStart,	/* First bit available for private use by FS */
326 };
327 
328 BUFFER_FNS(JBD, jbd)
329 BUFFER_FNS(JWrite, jwrite)
330 BUFFER_FNS(JBDDirty, jbddirty)
331 TAS_BUFFER_FNS(JBDDirty, jbddirty)
332 BUFFER_FNS(Revoked, revoked)
333 TAS_BUFFER_FNS(Revoked, revoked)
334 BUFFER_FNS(RevokeValid, revokevalid)
335 TAS_BUFFER_FNS(RevokeValid, revokevalid)
336 BUFFER_FNS(Freed, freed)
337 BUFFER_FNS(Shadow, shadow)
338 BUFFER_FNS(Verified, verified)
339 
340 static inline struct buffer_head *jh2bh(struct journal_head *jh)
341 {
342 	return jh->b_bh;
343 }
344 
345 static inline struct journal_head *bh2jh(struct buffer_head *bh)
346 {
347 	return bh->b_private;
348 }
349 
350 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
351 {
352 	bit_spin_lock(BH_JournalHead, &bh->b_state);
353 }
354 
355 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
356 {
357 	bit_spin_unlock(BH_JournalHead, &bh->b_state);
358 }
359 
360 #define J_ASSERT(assert)	BUG_ON(!(assert))
361 
362 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
363 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
364 
365 #if defined(JBD2_PARANOID_IOFAIL)
366 #define J_EXPECT(expr, why...)		J_ASSERT(expr)
367 #define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
368 #define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
369 #else
370 #define __journal_expect(expr, why...)					     \
371 	({								     \
372 		int val = (expr);					     \
373 		if (!val) {						     \
374 			printk(KERN_ERR					     \
375 			       "JBD2 unexpected failure: %s: %s;\n",	     \
376 			       __func__, #expr);			     \
377 			printk(KERN_ERR why "\n");			     \
378 		}							     \
379 		val;							     \
380 	})
381 #define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
382 #define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
383 #define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
384 #endif
385 
386 /* Flags in jbd_inode->i_flags */
387 #define __JI_COMMIT_RUNNING 0
388 #define __JI_WRITE_DATA 1
389 #define __JI_WAIT_DATA 2
390 
391 /*
392  * Commit of the inode data in progress. We use this flag to protect us from
393  * concurrent deletion of inode. We cannot use reference to inode for this
394  * since we cannot afford doing last iput() on behalf of kjournald
395  */
396 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
397 /* Write allocated dirty buffers in this inode before commit */
398 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
399 /* Wait for outstanding data writes for this inode before commit */
400 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
401 
402 /**
403  * struct jbd2_inode - The jbd_inode type is the structure linking inodes in
404  * ordered mode present in a transaction so that we can sync them during commit.
405  */
406 struct jbd2_inode {
407 	/**
408 	 * @i_transaction:
409 	 *
410 	 * Which transaction does this inode belong to? Either the running
411 	 * transaction or the committing one. [j_list_lock]
412 	 */
413 	transaction_t *i_transaction;
414 
415 	/**
416 	 * @i_next_transaction:
417 	 *
418 	 * Pointer to the running transaction modifying inode's data in case
419 	 * there is already a committing transaction touching it. [j_list_lock]
420 	 */
421 	transaction_t *i_next_transaction;
422 
423 	/**
424 	 * @i_list: List of inodes in the i_transaction [j_list_lock]
425 	 */
426 	struct list_head i_list;
427 
428 	/**
429 	 * @i_vfs_inode:
430 	 *
431 	 * VFS inode this inode belongs to [constant for lifetime of structure]
432 	 */
433 	struct inode *i_vfs_inode;
434 
435 	/**
436 	 * @i_flags: Flags of inode [j_list_lock]
437 	 */
438 	unsigned long i_flags;
439 
440 	/**
441 	 * @i_dirty_start:
442 	 *
443 	 * Offset in bytes where the dirty range for this inode starts.
444 	 * [j_list_lock]
445 	 */
446 	loff_t i_dirty_start;
447 
448 	/**
449 	 * @i_dirty_end:
450 	 *
451 	 * Inclusive offset in bytes where the dirty range for this inode
452 	 * ends. [j_list_lock]
453 	 */
454 	loff_t i_dirty_end;
455 };
456 
457 struct jbd2_revoke_table_s;
458 
459 /**
460  * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete
461  *     type associated with handle_t.
462  * @h_transaction: Which compound transaction is this update a part of?
463  * @h_journal: Which journal handle belongs to - used iff h_reserved set.
464  * @h_rsv_handle: Handle reserved for finishing the logical operation.
465  * @h_total_credits: Number of remaining buffers we are allowed to add to
466  *	journal. These are dirty buffers and revoke descriptor blocks.
467  * @h_revoke_credits: Number of remaining revoke records available for handle
468  * @h_ref: Reference count on this handle.
469  * @h_err: Field for caller's use to track errors through large fs operations.
470  * @h_sync: Flag for sync-on-close.
471  * @h_jdata: Flag to force data journaling.
472  * @h_reserved: Flag for handle for reserved credits.
473  * @h_aborted: Flag indicating fatal error on handle.
474  * @h_type: For handle statistics.
475  * @h_line_no: For handle statistics.
476  * @h_start_jiffies: Handle Start time.
477  * @h_requested_credits: Holds @h_total_credits after handle is started.
478  * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
479  * @saved_alloc_context: Saved context while transaction is open.
480  **/
481 
482 /* Docbook can't yet cope with the bit fields, but will leave the documentation
483  * in so it can be fixed later.
484  */
485 
486 struct jbd2_journal_handle
487 {
488 	union {
489 		transaction_t	*h_transaction;
490 		/* Which journal handle belongs to - used iff h_reserved set */
491 		journal_t	*h_journal;
492 	};
493 
494 	handle_t		*h_rsv_handle;
495 	int			h_total_credits;
496 	int			h_revoke_credits;
497 	int			h_revoke_credits_requested;
498 	int			h_ref;
499 	int			h_err;
500 
501 	/* Flags [no locking] */
502 	unsigned int	h_sync:		1;
503 	unsigned int	h_jdata:	1;
504 	unsigned int	h_reserved:	1;
505 	unsigned int	h_aborted:	1;
506 	unsigned int	h_type:		8;
507 	unsigned int	h_line_no:	16;
508 
509 	unsigned long		h_start_jiffies;
510 	unsigned int		h_requested_credits;
511 
512 	unsigned int		saved_alloc_context;
513 };
514 
515 
516 /*
517  * Some stats for checkpoint phase
518  */
519 struct transaction_chp_stats_s {
520 	unsigned long		cs_chp_time;
521 	__u32			cs_forced_to_close;
522 	__u32			cs_written;
523 	__u32			cs_dropped;
524 };
525 
526 /* The transaction_t type is the guts of the journaling mechanism.  It
527  * tracks a compound transaction through its various states:
528  *
529  * RUNNING:	accepting new updates
530  * LOCKED:	Updates still running but we don't accept new ones
531  * RUNDOWN:	Updates are tidying up but have finished requesting
532  *		new buffers to modify (state not used for now)
533  * FLUSH:       All updates complete, but we are still writing to disk
534  * COMMIT:      All data on disk, writing commit record
535  * FINISHED:	We still have to keep the transaction for checkpointing.
536  *
537  * The transaction keeps track of all of the buffers modified by a
538  * running transaction, and all of the buffers committed but not yet
539  * flushed to home for finished transactions.
540  * (Locking Documentation improved by LockDoc)
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  *    ->j_list_lock			(journal_unmap_buffer)
557  *
558  */
559 
560 struct transaction_s
561 {
562 	/* Pointer to the journal for this transaction. [no locking] */
563 	journal_t		*t_journal;
564 
565 	/* Sequence number for this transaction [no locking] */
566 	tid_t			t_tid;
567 
568 	/*
569 	 * Transaction's current state
570 	 * [no locking - only kjournald2 alters this]
571 	 * [j_list_lock] guards transition of a transaction into T_FINISHED
572 	 * state and subsequent call of __jbd2_journal_drop_transaction()
573 	 * FIXME: needs barriers
574 	 * KLUDGE: [use j_state_lock]
575 	 */
576 	enum {
577 		T_RUNNING,
578 		T_LOCKED,
579 		T_SWITCH,
580 		T_FLUSH,
581 		T_COMMIT,
582 		T_COMMIT_DFLUSH,
583 		T_COMMIT_JFLUSH,
584 		T_COMMIT_CALLBACK,
585 		T_FINISHED
586 	}			t_state;
587 
588 	/*
589 	 * Where in the log does this transaction's commit start? [no locking]
590 	 */
591 	unsigned long		t_log_start;
592 
593 	/*
594 	 * Number of buffers on the t_buffers list [j_list_lock, no locks
595 	 * needed for jbd2 thread]
596 	 */
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, no locks needed fo
602 	 * jbd2 thread]
603 	 */
604 	struct journal_head	*t_reserved_list;
605 
606 	/*
607 	 * Doubly-linked circular list of all metadata buffers owned by this
608 	 * transaction [j_list_lock, no locks needed for jbd2 thread]
609 	 */
610 	struct journal_head	*t_buffers;
611 
612 	/*
613 	 * Doubly-linked circular list of all forget buffers (superseded
614 	 * buffers which we can un-checkpoint once this transaction commits)
615 	 * [j_list_lock]
616 	 */
617 	struct journal_head	*t_forget;
618 
619 	/*
620 	 * Doubly-linked circular list of all buffers still to be flushed before
621 	 * this transaction can be checkpointed. [j_list_lock]
622 	 */
623 	struct journal_head	*t_checkpoint_list;
624 
625 	/*
626 	 * Doubly-linked circular list of all buffers submitted for IO while
627 	 * checkpointing. [j_list_lock]
628 	 */
629 	struct journal_head	*t_checkpoint_io_list;
630 
631 	/*
632 	 * Doubly-linked circular list of metadata buffers being
633 	 * shadowed by log IO.  The IO buffers on the iobuf list and
634 	 * the shadow buffers on this list match each other one for
635 	 * one at all times. [j_list_lock, no locks needed for jbd2
636 	 * thread]
637 	 */
638 	struct journal_head	*t_shadow_list;
639 
640 	/*
641 	 * List of inodes associated with the transaction; e.g., ext4 uses
642 	 * this to track inodes in data=ordered and data=journal mode that
643 	 * need special handling on transaction commit; also used by ocfs2.
644 	 * [j_list_lock]
645 	 */
646 	struct list_head	t_inode_list;
647 
648 	/*
649 	 * Protects info related to handles
650 	 */
651 	spinlock_t		t_handle_lock;
652 
653 	/*
654 	 * Longest time some handle had to wait for running transaction
655 	 */
656 	unsigned long		t_max_wait;
657 
658 	/*
659 	 * When transaction started
660 	 */
661 	unsigned long		t_start;
662 
663 	/*
664 	 * When commit was requested [j_state_lock]
665 	 */
666 	unsigned long		t_requested;
667 
668 	/*
669 	 * Checkpointing stats [j_list_lock]
670 	 */
671 	struct transaction_chp_stats_s t_chp_stats;
672 
673 	/*
674 	 * Number of outstanding updates running on this transaction
675 	 * [none]
676 	 */
677 	atomic_t		t_updates;
678 
679 	/*
680 	 * Number of blocks reserved for this transaction in the journal.
681 	 * This is including all credits reserved when starting transaction
682 	 * handles as well as all journal descriptor blocks needed for this
683 	 * transaction. [none]
684 	 */
685 	atomic_t		t_outstanding_credits;
686 
687 	/*
688 	 * Number of revoke records for this transaction added by already
689 	 * stopped handles. [none]
690 	 */
691 	atomic_t		t_outstanding_revokes;
692 
693 	/*
694 	 * How many handles used this transaction? [none]
695 	 */
696 	atomic_t		t_handle_count;
697 
698 	/*
699 	 * Forward and backward links for the circular list of all transactions
700 	 * awaiting checkpoint. [j_list_lock]
701 	 */
702 	transaction_t		*t_cpnext, *t_cpprev;
703 
704 	/*
705 	 * When will the transaction expire (become due for commit), in jiffies?
706 	 * [no locking]
707 	 */
708 	unsigned long		t_expires;
709 
710 	/*
711 	 * When this transaction started, in nanoseconds [no locking]
712 	 */
713 	ktime_t			t_start_time;
714 
715 	/*
716 	 * This transaction is being forced and some process is
717 	 * waiting for it to finish.
718 	 */
719 	unsigned int t_synchronous_commit:1;
720 
721 	/* Disk flush needs to be sent to fs partition [no locking] */
722 	int			t_need_data_flush;
723 
724 	/*
725 	 * For use by the filesystem to store fs-specific data
726 	 * structures associated with the transaction
727 	 */
728 	struct list_head	t_private_list;
729 };
730 
731 struct transaction_run_stats_s {
732 	unsigned long		rs_wait;
733 	unsigned long		rs_request_delay;
734 	unsigned long		rs_running;
735 	unsigned long		rs_locked;
736 	unsigned long		rs_flushing;
737 	unsigned long		rs_logging;
738 
739 	__u32			rs_handle_count;
740 	__u32			rs_blocks;
741 	__u32			rs_blocks_logged;
742 };
743 
744 struct transaction_stats_s {
745 	unsigned long		ts_tid;
746 	unsigned long		ts_requested;
747 	struct transaction_run_stats_s run;
748 };
749 
750 static inline unsigned long
751 jbd2_time_diff(unsigned long start, unsigned long end)
752 {
753 	if (end >= start)
754 		return end - start;
755 
756 	return end + (MAX_JIFFY_OFFSET - start);
757 }
758 
759 #define JBD2_NR_BATCH	64
760 
761 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
762 
763 #define JBD2_FC_REPLAY_STOP	0
764 #define JBD2_FC_REPLAY_CONTINUE	1
765 
766 /**
767  * struct journal_s - The journal_s type is the concrete type associated with
768  *     journal_t.
769  */
770 struct journal_s
771 {
772 	/**
773 	 * @j_flags: General journaling state flags [j_state_lock,
774 	 * no lock for quick racy checks]
775 	 */
776 	unsigned long		j_flags;
777 
778 	/**
779 	 * @j_atomic_flags: Atomic journaling state flags.
780 	 */
781 	unsigned long		j_atomic_flags;
782 
783 	/**
784 	 * @j_errno:
785 	 *
786 	 * Is there an outstanding uncleared error on the journal (from a prior
787 	 * abort)? [j_state_lock]
788 	 */
789 	int			j_errno;
790 
791 	/**
792 	 * @j_abort_mutex: Lock the whole aborting procedure.
793 	 */
794 	struct mutex		j_abort_mutex;
795 
796 	/**
797 	 * @j_sb_buffer: The first part of the superblock buffer.
798 	 */
799 	struct buffer_head	*j_sb_buffer;
800 
801 	/**
802 	 * @j_superblock: The second part of the superblock buffer.
803 	 */
804 	journal_superblock_t	*j_superblock;
805 
806 	/**
807 	 * @j_format_version: Version of the superblock format.
808 	 */
809 	int			j_format_version;
810 
811 	/**
812 	 * @j_state_lock: Protect the various scalars in the journal.
813 	 */
814 	rwlock_t		j_state_lock;
815 
816 	/**
817 	 * @j_barrier_count:
818 	 *
819 	 * Number of processes waiting to create a barrier lock [j_state_lock,
820 	 * no lock for quick racy checks]
821 	 */
822 	int			j_barrier_count;
823 
824 	/**
825 	 * @j_barrier: The barrier lock itself.
826 	 */
827 	struct mutex		j_barrier;
828 
829 	/**
830 	 * @j_running_transaction:
831 	 *
832 	 * Transactions: The current running transaction...
833 	 * [j_state_lock, no lock for quick racy checks] [caller holding
834 	 * open handle]
835 	 */
836 	transaction_t		*j_running_transaction;
837 
838 	/**
839 	 * @j_committing_transaction:
840 	 *
841 	 * the transaction we are pushing to disk
842 	 * [j_state_lock] [caller holding open handle]
843 	 */
844 	transaction_t		*j_committing_transaction;
845 
846 	/**
847 	 * @j_checkpoint_transactions:
848 	 *
849 	 * ... and a linked circular list of all transactions waiting for
850 	 * checkpointing. [j_list_lock]
851 	 */
852 	transaction_t		*j_checkpoint_transactions;
853 
854 	/**
855 	 * @j_wait_transaction_locked:
856 	 *
857 	 * Wait queue for waiting for a locked transaction to start committing,
858 	 * or for a barrier lock to be released.
859 	 */
860 	wait_queue_head_t	j_wait_transaction_locked;
861 
862 	/**
863 	 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
864 	 */
865 	wait_queue_head_t	j_wait_done_commit;
866 
867 	/**
868 	 * @j_wait_commit: Wait queue to trigger commit.
869 	 */
870 	wait_queue_head_t	j_wait_commit;
871 
872 	/**
873 	 * @j_wait_updates: Wait queue to wait for updates to complete.
874 	 */
875 	wait_queue_head_t	j_wait_updates;
876 
877 	/**
878 	 * @j_wait_reserved:
879 	 *
880 	 * Wait queue to wait for reserved buffer credits to drop.
881 	 */
882 	wait_queue_head_t	j_wait_reserved;
883 
884 	/**
885 	 * @j_fc_wait:
886 	 *
887 	 * Wait queue to wait for completion of async fast commits.
888 	 */
889 	wait_queue_head_t	j_fc_wait;
890 
891 	/**
892 	 * @j_checkpoint_mutex:
893 	 *
894 	 * Semaphore for locking against concurrent checkpoints.
895 	 */
896 	struct mutex		j_checkpoint_mutex;
897 
898 	/**
899 	 * @j_chkpt_bhs:
900 	 *
901 	 * List of buffer heads used by the checkpoint routine.  This
902 	 * was moved from jbd2_log_do_checkpoint() to reduce stack
903 	 * usage.  Access to this array is controlled by the
904 	 * @j_checkpoint_mutex.  [j_checkpoint_mutex]
905 	 */
906 	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
907 
908 	/**
909 	 * @j_shrinker:
910 	 *
911 	 * Journal head shrinker, reclaim buffer's journal head which
912 	 * has been written back.
913 	 */
914 	struct shrinker		j_shrinker;
915 
916 	/**
917 	 * @j_checkpoint_jh_count:
918 	 *
919 	 * Number of journal buffers on the checkpoint list. [j_list_lock]
920 	 */
921 	struct percpu_counter	j_checkpoint_jh_count;
922 
923 	/**
924 	 * @j_shrink_transaction:
925 	 *
926 	 * Record next transaction will shrink on the checkpoint list.
927 	 * [j_list_lock]
928 	 */
929 	transaction_t		*j_shrink_transaction;
930 
931 	/**
932 	 * @j_head:
933 	 *
934 	 * Journal head: identifies the first unused block in the journal.
935 	 * [j_state_lock]
936 	 */
937 	unsigned long		j_head;
938 
939 	/**
940 	 * @j_tail:
941 	 *
942 	 * Journal tail: identifies the oldest still-used block in the journal.
943 	 * [j_state_lock]
944 	 */
945 	unsigned long		j_tail;
946 
947 	/**
948 	 * @j_free:
949 	 *
950 	 * Journal free: how many free blocks are there in the journal?
951 	 * [j_state_lock]
952 	 */
953 	unsigned long		j_free;
954 
955 	/**
956 	 * @j_first:
957 	 *
958 	 * The block number of the first usable block in the journal
959 	 * [j_state_lock].
960 	 */
961 	unsigned long		j_first;
962 
963 	/**
964 	 * @j_last:
965 	 *
966 	 * The block number one beyond the last usable block in the journal
967 	 * [j_state_lock].
968 	 */
969 	unsigned long		j_last;
970 
971 	/**
972 	 * @j_fc_first:
973 	 *
974 	 * The block number of the first fast commit block in the journal
975 	 * [j_state_lock].
976 	 */
977 	unsigned long		j_fc_first;
978 
979 	/**
980 	 * @j_fc_off:
981 	 *
982 	 * Number of fast commit blocks currently allocated. Accessed only
983 	 * during fast commit. Currently only process can do fast commit, so
984 	 * this field is not protected by any lock.
985 	 */
986 	unsigned long		j_fc_off;
987 
988 	/**
989 	 * @j_fc_last:
990 	 *
991 	 * The block number one beyond the last fast commit block in the journal
992 	 * [j_state_lock].
993 	 */
994 	unsigned long		j_fc_last;
995 
996 	/**
997 	 * @j_dev: Device where we store the journal.
998 	 */
999 	struct block_device	*j_dev;
1000 
1001 	/**
1002 	 * @j_blocksize: Block size for the location where we store the journal.
1003 	 */
1004 	int			j_blocksize;
1005 
1006 	/**
1007 	 * @j_blk_offset:
1008 	 *
1009 	 * Starting block offset into the device where we store the journal.
1010 	 */
1011 	unsigned long long	j_blk_offset;
1012 
1013 	/**
1014 	 * @j_devname: Journal device name.
1015 	 */
1016 	char			j_devname[BDEVNAME_SIZE+24];
1017 
1018 	/**
1019 	 * @j_fs_dev:
1020 	 *
1021 	 * Device which holds the client fs.  For internal journal this will be
1022 	 * equal to j_dev.
1023 	 */
1024 	struct block_device	*j_fs_dev;
1025 
1026 	/**
1027 	 * @j_total_len: Total maximum capacity of the journal region on disk.
1028 	 */
1029 	unsigned int		j_total_len;
1030 
1031 	/**
1032 	 * @j_reserved_credits:
1033 	 *
1034 	 * Number of buffers reserved from the running transaction.
1035 	 */
1036 	atomic_t		j_reserved_credits;
1037 
1038 	/**
1039 	 * @j_list_lock: Protects the buffer lists and internal buffer state.
1040 	 */
1041 	spinlock_t		j_list_lock;
1042 
1043 	/**
1044 	 * @j_inode:
1045 	 *
1046 	 * Optional inode where we store the journal.  If present, all
1047 	 * journal block numbers are mapped into this inode via bmap().
1048 	 */
1049 	struct inode		*j_inode;
1050 
1051 	/**
1052 	 * @j_tail_sequence:
1053 	 *
1054 	 * Sequence number of the oldest transaction in the log [j_state_lock]
1055 	 */
1056 	tid_t			j_tail_sequence;
1057 
1058 	/**
1059 	 * @j_transaction_sequence:
1060 	 *
1061 	 * Sequence number of the next transaction to grant [j_state_lock]
1062 	 */
1063 	tid_t			j_transaction_sequence;
1064 
1065 	/**
1066 	 * @j_commit_sequence:
1067 	 *
1068 	 * Sequence number of the most recently committed transaction
1069 	 * [j_state_lock, no lock for quick racy checks]
1070 	 */
1071 	tid_t			j_commit_sequence;
1072 
1073 	/**
1074 	 * @j_commit_request:
1075 	 *
1076 	 * Sequence number of the most recent transaction wanting commit
1077 	 * [j_state_lock, no lock for quick racy checks]
1078 	 */
1079 	tid_t			j_commit_request;
1080 
1081 	/**
1082 	 * @j_uuid:
1083 	 *
1084 	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1085 	 * backed by this journal.  This will eventually be replaced by an array
1086 	 * of uuids, allowing us to index multiple devices within a single
1087 	 * journal and to perform atomic updates across them.
1088 	 */
1089 	__u8			j_uuid[16];
1090 
1091 	/**
1092 	 * @j_task: Pointer to the current commit thread for this journal.
1093 	 */
1094 	struct task_struct	*j_task;
1095 
1096 	/**
1097 	 * @j_max_transaction_buffers:
1098 	 *
1099 	 * Maximum number of metadata buffers to allow in a single compound
1100 	 * commit transaction.
1101 	 */
1102 	int			j_max_transaction_buffers;
1103 
1104 	/**
1105 	 * @j_revoke_records_per_block:
1106 	 *
1107 	 * Number of revoke records that fit in one descriptor block.
1108 	 */
1109 	int			j_revoke_records_per_block;
1110 
1111 	/**
1112 	 * @j_commit_interval:
1113 	 *
1114 	 * What is the maximum transaction lifetime before we begin a commit?
1115 	 */
1116 	unsigned long		j_commit_interval;
1117 
1118 	/**
1119 	 * @j_commit_timer: The timer used to wakeup the commit thread.
1120 	 */
1121 	struct timer_list	j_commit_timer;
1122 
1123 	/**
1124 	 * @j_revoke_lock: Protect the revoke table.
1125 	 */
1126 	spinlock_t		j_revoke_lock;
1127 
1128 	/**
1129 	 * @j_revoke:
1130 	 *
1131 	 * The revoke table - maintains the list of revoked blocks in the
1132 	 * current transaction.
1133 	 */
1134 	struct jbd2_revoke_table_s *j_revoke;
1135 
1136 	/**
1137 	 * @j_revoke_table: Alternate revoke tables for j_revoke.
1138 	 */
1139 	struct jbd2_revoke_table_s *j_revoke_table[2];
1140 
1141 	/**
1142 	 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1143 	 */
1144 	struct buffer_head	**j_wbuf;
1145 
1146 	/**
1147 	 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only
1148 	 * during a fast commit. Currently only process can do fast commit, so
1149 	 * this field is not protected by any lock.
1150 	 */
1151 	struct buffer_head	**j_fc_wbuf;
1152 
1153 	/**
1154 	 * @j_wbufsize:
1155 	 *
1156 	 * Size of @j_wbuf array.
1157 	 */
1158 	int			j_wbufsize;
1159 
1160 	/**
1161 	 * @j_fc_wbufsize:
1162 	 *
1163 	 * Size of @j_fc_wbuf array.
1164 	 */
1165 	int			j_fc_wbufsize;
1166 
1167 	/**
1168 	 * @j_last_sync_writer:
1169 	 *
1170 	 * The pid of the last person to run a synchronous operation
1171 	 * through the journal.
1172 	 */
1173 	pid_t			j_last_sync_writer;
1174 
1175 	/**
1176 	 * @j_average_commit_time:
1177 	 *
1178 	 * The average amount of time in nanoseconds it takes to commit a
1179 	 * transaction to disk. [j_state_lock]
1180 	 */
1181 	u64			j_average_commit_time;
1182 
1183 	/**
1184 	 * @j_min_batch_time:
1185 	 *
1186 	 * Minimum time that we should wait for additional filesystem operations
1187 	 * to get batched into a synchronous handle in microseconds.
1188 	 */
1189 	u32			j_min_batch_time;
1190 
1191 	/**
1192 	 * @j_max_batch_time:
1193 	 *
1194 	 * Maximum time that we should wait for additional filesystem operations
1195 	 * to get batched into a synchronous handle in microseconds.
1196 	 */
1197 	u32			j_max_batch_time;
1198 
1199 	/**
1200 	 * @j_commit_callback:
1201 	 *
1202 	 * This function is called when a transaction is closed.
1203 	 */
1204 	void			(*j_commit_callback)(journal_t *,
1205 						     transaction_t *);
1206 
1207 	/**
1208 	 * @j_submit_inode_data_buffers:
1209 	 *
1210 	 * This function is called for all inodes associated with the
1211 	 * committing transaction marked with JI_WRITE_DATA flag
1212 	 * before we start to write out the transaction to the journal.
1213 	 */
1214 	int			(*j_submit_inode_data_buffers)
1215 					(struct jbd2_inode *);
1216 
1217 	/**
1218 	 * @j_finish_inode_data_buffers:
1219 	 *
1220 	 * This function is called for all inodes associated with the
1221 	 * committing transaction marked with JI_WAIT_DATA flag
1222 	 * after we have written the transaction to the journal
1223 	 * but before we write out the commit block.
1224 	 */
1225 	int			(*j_finish_inode_data_buffers)
1226 					(struct jbd2_inode *);
1227 
1228 	/*
1229 	 * Journal statistics
1230 	 */
1231 
1232 	/**
1233 	 * @j_history_lock: Protect the transactions statistics history.
1234 	 */
1235 	spinlock_t		j_history_lock;
1236 
1237 	/**
1238 	 * @j_proc_entry: procfs entry for the jbd statistics directory.
1239 	 */
1240 	struct proc_dir_entry	*j_proc_entry;
1241 
1242 	/**
1243 	 * @j_stats: Overall statistics.
1244 	 */
1245 	struct transaction_stats_s j_stats;
1246 
1247 	/**
1248 	 * @j_failed_commit: Failed journal commit ID.
1249 	 */
1250 	unsigned int		j_failed_commit;
1251 
1252 	/**
1253 	 * @j_private:
1254 	 *
1255 	 * An opaque pointer to fs-private information.  ext3 puts its
1256 	 * superblock pointer here.
1257 	 */
1258 	void *j_private;
1259 
1260 	/**
1261 	 * @j_chksum_driver:
1262 	 *
1263 	 * Reference to checksum algorithm driver via cryptoapi.
1264 	 */
1265 	struct crypto_shash *j_chksum_driver;
1266 
1267 	/**
1268 	 * @j_csum_seed:
1269 	 *
1270 	 * Precomputed journal UUID checksum for seeding other checksums.
1271 	 */
1272 	__u32 j_csum_seed;
1273 
1274 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1275 	/**
1276 	 * @j_trans_commit_map:
1277 	 *
1278 	 * Lockdep entity to track transaction commit dependencies. Handles
1279 	 * hold this "lock" for read, when we wait for commit, we acquire the
1280 	 * "lock" for writing. This matches the properties of jbd2 journalling
1281 	 * where the running transaction has to wait for all handles to be
1282 	 * dropped to commit that transaction and also acquiring a handle may
1283 	 * require transaction commit to finish.
1284 	 */
1285 	struct lockdep_map	j_trans_commit_map;
1286 #endif
1287 
1288 	/**
1289 	 * @j_fc_cleanup_callback:
1290 	 *
1291 	 * Clean-up after fast commit or full commit. JBD2 calls this function
1292 	 * after every commit operation.
1293 	 */
1294 	void (*j_fc_cleanup_callback)(struct journal_s *journal, int full, tid_t tid);
1295 
1296 	/**
1297 	 * @j_fc_replay_callback:
1298 	 *
1299 	 * File-system specific function that performs replay of a fast
1300 	 * commit. JBD2 calls this function for each fast commit block found in
1301 	 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1302 	 * to indicate that the block was processed correctly and more fast
1303 	 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1304 	 * indicates the end of replay (no more blocks remaining). A negative
1305 	 * return value indicates error.
1306 	 */
1307 	int (*j_fc_replay_callback)(struct journal_s *journal,
1308 				    struct buffer_head *bh,
1309 				    enum passtype pass, int off,
1310 				    tid_t expected_commit_id);
1311 };
1312 
1313 #define jbd2_might_wait_for_commit(j) \
1314 	do { \
1315 		rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1316 		rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1317 	} while (0)
1318 
1319 /* journal feature predicate functions */
1320 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1321 static inline bool jbd2_has_feature_##name(journal_t *j) \
1322 { \
1323 	return ((j)->j_format_version >= 2 && \
1324 		((j)->j_superblock->s_feature_compat & \
1325 		 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1326 } \
1327 static inline void jbd2_set_feature_##name(journal_t *j) \
1328 { \
1329 	(j)->j_superblock->s_feature_compat |= \
1330 		cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1331 } \
1332 static inline void jbd2_clear_feature_##name(journal_t *j) \
1333 { \
1334 	(j)->j_superblock->s_feature_compat &= \
1335 		~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1336 }
1337 
1338 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1339 static inline bool jbd2_has_feature_##name(journal_t *j) \
1340 { \
1341 	return ((j)->j_format_version >= 2 && \
1342 		((j)->j_superblock->s_feature_ro_compat & \
1343 		 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1344 } \
1345 static inline void jbd2_set_feature_##name(journal_t *j) \
1346 { \
1347 	(j)->j_superblock->s_feature_ro_compat |= \
1348 		cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1349 } \
1350 static inline void jbd2_clear_feature_##name(journal_t *j) \
1351 { \
1352 	(j)->j_superblock->s_feature_ro_compat &= \
1353 		~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1354 }
1355 
1356 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1357 static inline bool jbd2_has_feature_##name(journal_t *j) \
1358 { \
1359 	return ((j)->j_format_version >= 2 && \
1360 		((j)->j_superblock->s_feature_incompat & \
1361 		 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1362 } \
1363 static inline void jbd2_set_feature_##name(journal_t *j) \
1364 { \
1365 	(j)->j_superblock->s_feature_incompat |= \
1366 		cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1367 } \
1368 static inline void jbd2_clear_feature_##name(journal_t *j) \
1369 { \
1370 	(j)->j_superblock->s_feature_incompat &= \
1371 		~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1372 }
1373 
1374 JBD2_FEATURE_COMPAT_FUNCS(checksum,		CHECKSUM)
1375 
1376 JBD2_FEATURE_INCOMPAT_FUNCS(revoke,		REVOKE)
1377 JBD2_FEATURE_INCOMPAT_FUNCS(64bit,		64BIT)
1378 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit,	ASYNC_COMMIT)
1379 JBD2_FEATURE_INCOMPAT_FUNCS(csum2,		CSUM_V2)
1380 JBD2_FEATURE_INCOMPAT_FUNCS(csum3,		CSUM_V3)
1381 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit,	FAST_COMMIT)
1382 
1383 /*
1384  * Journal flag definitions
1385  */
1386 #define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1387 #define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1388 #define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1389 #define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1390 #define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1391 #define JBD2_BARRIER	0x020	/* Use IDE barriers */
1392 #define JBD2_ABORT_ON_SYNCDATA_ERR	0x040	/* Abort the journal on file
1393 						 * data write error in ordered
1394 						 * mode */
1395 #define JBD2_FAST_COMMIT_ONGOING	0x100	/* Fast commit is ongoing */
1396 #define JBD2_FULL_COMMIT_ONGOING	0x200	/* Full commit is ongoing */
1397 #define JBD2_JOURNAL_FLUSH_DISCARD	0x0001
1398 #define JBD2_JOURNAL_FLUSH_ZEROOUT	0x0002
1399 #define JBD2_JOURNAL_FLUSH_VALID	(JBD2_JOURNAL_FLUSH_DISCARD | \
1400 					JBD2_JOURNAL_FLUSH_ZEROOUT)
1401 
1402 /*
1403  * Journal atomic flag definitions
1404  */
1405 #define JBD2_CHECKPOINT_IO_ERROR	0x001	/* Detect io error while writing
1406 						 * buffer back to disk */
1407 
1408 /*
1409  * Function declarations for the journaling transaction and buffer
1410  * management
1411  */
1412 
1413 /* Filing buffers */
1414 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1415 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1416 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1417 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1418 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1419 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1420 {
1421 	list_add_tail(&bh->b_assoc_buffers, head);
1422 }
1423 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1424 {
1425 	list_del_init(&bh->b_assoc_buffers);
1426 }
1427 
1428 /* Log buffer allocation */
1429 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1430 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1431 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1432 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1433 			      unsigned long *block);
1434 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1435 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1436 
1437 /* Commit management */
1438 extern void jbd2_journal_commit_transaction(journal_t *);
1439 
1440 /* Checkpoint list management */
1441 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1442 unsigned long jbd2_journal_shrink_checkpoint_list(journal_t *journal, unsigned long *nr_to_scan);
1443 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1444 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1445 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1446 
1447 
1448 /*
1449  * Triggers
1450  */
1451 
1452 struct jbd2_buffer_trigger_type {
1453 	/*
1454 	 * Fired a the moment data to write to the journal are known to be
1455 	 * stable - so either at the moment b_frozen_data is created or just
1456 	 * before a buffer is written to the journal.  mapped_data is a mapped
1457 	 * buffer that is the frozen data for commit.
1458 	 */
1459 	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1460 			 struct buffer_head *bh, void *mapped_data,
1461 			 size_t size);
1462 
1463 	/*
1464 	 * Fired during journal abort for dirty buffers that will not be
1465 	 * committed.
1466 	 */
1467 	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1468 			struct buffer_head *bh);
1469 };
1470 
1471 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1472 				       void *mapped_data,
1473 				       struct jbd2_buffer_trigger_type *triggers);
1474 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1475 				      struct jbd2_buffer_trigger_type *triggers);
1476 
1477 /* Buffer IO */
1478 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1479 					      struct journal_head *jh_in,
1480 					      struct buffer_head **bh_out,
1481 					      sector_t blocknr);
1482 
1483 /* Transaction cache support */
1484 extern void jbd2_journal_destroy_transaction_cache(void);
1485 extern int __init jbd2_journal_init_transaction_cache(void);
1486 extern void jbd2_journal_free_transaction(transaction_t *);
1487 
1488 /*
1489  * Journal locking.
1490  *
1491  * We need to lock the journal during transaction state changes so that nobody
1492  * ever tries to take a handle on the running transaction while we are in the
1493  * middle of moving it to the commit phase.  j_state_lock does this.
1494  *
1495  * Note that the locking is completely interrupt unsafe.  We never touch
1496  * journal structures from interrupts.
1497  */
1498 
1499 static inline handle_t *journal_current_handle(void)
1500 {
1501 	return current->journal_info;
1502 }
1503 
1504 /* The journaling code user interface:
1505  *
1506  * Create and destroy handles
1507  * Register buffer modifications against the current transaction.
1508  */
1509 
1510 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1511 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1512 				     int revoke_records, gfp_t gfp_mask,
1513 				     unsigned int type, unsigned int line_no);
1514 extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1515 extern int	 jbd2__journal_restart(handle_t *, int nblocks,
1516 				       int revoke_records, gfp_t gfp_mask);
1517 extern int	 jbd2_journal_start_reserved(handle_t *handle,
1518 				unsigned int type, unsigned int line_no);
1519 extern void	 jbd2_journal_free_reserved(handle_t *handle);
1520 extern int	 jbd2_journal_extend(handle_t *handle, int nblocks,
1521 				     int revoke_records);
1522 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1523 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1524 extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1525 void		 jbd2_journal_set_triggers(struct buffer_head *,
1526 					   struct jbd2_buffer_trigger_type *type);
1527 extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1528 extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1529 int jbd2_journal_invalidate_folio(journal_t *, struct folio *,
1530 					size_t offset, size_t length);
1531 bool jbd2_journal_try_to_free_buffers(journal_t *journal, struct folio *folio);
1532 extern int	 jbd2_journal_stop(handle_t *);
1533 extern int	 jbd2_journal_flush(journal_t *journal, unsigned int flags);
1534 extern void	 jbd2_journal_lock_updates (journal_t *);
1535 extern void	 jbd2_journal_unlock_updates (journal_t *);
1536 
1537 void jbd2_journal_wait_updates(journal_t *);
1538 
1539 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1540 				struct block_device *fs_dev,
1541 				unsigned long long start, int len, int bsize);
1542 extern journal_t * jbd2_journal_init_inode (struct inode *);
1543 extern int	   jbd2_journal_update_format (journal_t *);
1544 extern int	   jbd2_journal_check_used_features
1545 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1546 extern int	   jbd2_journal_check_available_features
1547 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1548 extern int	   jbd2_journal_set_features
1549 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1550 extern void	   jbd2_journal_clear_features
1551 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1552 extern int	   jbd2_journal_load       (journal_t *journal);
1553 extern int	   jbd2_journal_destroy    (journal_t *);
1554 extern int	   jbd2_journal_recover    (journal_t *journal);
1555 extern int	   jbd2_journal_wipe       (journal_t *, int);
1556 extern int	   jbd2_journal_skip_recovery	(journal_t *);
1557 extern void	   jbd2_journal_update_sb_errno(journal_t *);
1558 extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1559 				unsigned long, blk_opf_t);
1560 extern void	   jbd2_journal_abort      (journal_t *, int);
1561 extern int	   jbd2_journal_errno      (journal_t *);
1562 extern void	   jbd2_journal_ack_err    (journal_t *);
1563 extern int	   jbd2_journal_clear_err  (journal_t *);
1564 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1565 extern int	   jbd2_journal_force_commit(journal_t *);
1566 extern int	   jbd2_journal_force_commit_nested(journal_t *);
1567 extern int	   jbd2_journal_inode_ranged_write(handle_t *handle,
1568 			struct jbd2_inode *inode, loff_t start_byte,
1569 			loff_t length);
1570 extern int	   jbd2_journal_inode_ranged_wait(handle_t *handle,
1571 			struct jbd2_inode *inode, loff_t start_byte,
1572 			loff_t length);
1573 extern int	   jbd2_journal_submit_inode_data_buffers(
1574 			struct jbd2_inode *jinode);
1575 extern int	   jbd2_journal_finish_inode_data_buffers(
1576 			struct jbd2_inode *jinode);
1577 extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1578 				struct jbd2_inode *inode, loff_t new_size);
1579 extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1580 extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1581 
1582 /*
1583  * journal_head management
1584  */
1585 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1586 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1587 void jbd2_journal_put_journal_head(struct journal_head *jh);
1588 
1589 /*
1590  * handle management
1591  */
1592 extern struct kmem_cache *jbd2_handle_cache;
1593 
1594 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1595 {
1596 	return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1597 }
1598 
1599 static inline void jbd2_free_handle(handle_t *handle)
1600 {
1601 	kmem_cache_free(jbd2_handle_cache, handle);
1602 }
1603 
1604 /*
1605  * jbd2_inode management (optional, for those file systems that want to use
1606  * dynamically allocated jbd2_inode structures)
1607  */
1608 extern struct kmem_cache *jbd2_inode_cache;
1609 
1610 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1611 {
1612 	return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1613 }
1614 
1615 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1616 {
1617 	kmem_cache_free(jbd2_inode_cache, jinode);
1618 }
1619 
1620 /* Primary revoke support */
1621 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1622 extern int	   jbd2_journal_init_revoke(journal_t *, int);
1623 extern void	   jbd2_journal_destroy_revoke_record_cache(void);
1624 extern void	   jbd2_journal_destroy_revoke_table_cache(void);
1625 extern int __init jbd2_journal_init_revoke_record_cache(void);
1626 extern int __init jbd2_journal_init_revoke_table_cache(void);
1627 
1628 extern void	   jbd2_journal_destroy_revoke(journal_t *);
1629 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1630 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1631 extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
1632 						     struct list_head *log_bufs);
1633 
1634 /* Recovery revoke support */
1635 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1636 extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1637 extern void	jbd2_journal_clear_revoke(journal_t *);
1638 extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1639 extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1640 
1641 /*
1642  * The log thread user interface:
1643  *
1644  * Request space in the current transaction, and force transaction commit
1645  * transitions on demand.
1646  */
1647 
1648 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1649 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1650 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1651 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1652 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1653 int jbd2_log_do_checkpoint(journal_t *journal);
1654 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1655 
1656 void __jbd2_log_wait_for_space(journal_t *journal);
1657 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1658 extern int jbd2_cleanup_journal_tail(journal_t *);
1659 
1660 /* Fast commit related APIs */
1661 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1662 int jbd2_fc_end_commit(journal_t *journal);
1663 int jbd2_fc_end_commit_fallback(journal_t *journal);
1664 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1665 int jbd2_submit_inode_data(struct jbd2_inode *jinode);
1666 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1667 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1668 int jbd2_fc_release_bufs(journal_t *journal);
1669 
1670 static inline int jbd2_journal_get_max_txn_bufs(journal_t *journal)
1671 {
1672 	return (journal->j_total_len - journal->j_fc_wbufsize) / 4;
1673 }
1674 
1675 /*
1676  * is_journal_abort
1677  *
1678  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1679  * bit, when set, indicates that we have had a fatal error somewhere,
1680  * either inside the journaling layer or indicated to us by the client
1681  * (eg. ext3), and that we and should not commit any further
1682  * transactions.
1683  */
1684 
1685 static inline int is_journal_aborted(journal_t *journal)
1686 {
1687 	return journal->j_flags & JBD2_ABORT;
1688 }
1689 
1690 static inline int is_handle_aborted(handle_t *handle)
1691 {
1692 	if (handle->h_aborted || !handle->h_transaction)
1693 		return 1;
1694 	return is_journal_aborted(handle->h_transaction->t_journal);
1695 }
1696 
1697 static inline void jbd2_journal_abort_handle(handle_t *handle)
1698 {
1699 	handle->h_aborted = 1;
1700 }
1701 
1702 #endif /* __KERNEL__   */
1703 
1704 /* Comparison functions for transaction IDs: perform comparisons using
1705  * modulo arithmetic so that they work over sequence number wraps. */
1706 
1707 static inline int tid_gt(tid_t x, tid_t y)
1708 {
1709 	int difference = (x - y);
1710 	return (difference > 0);
1711 }
1712 
1713 static inline int tid_geq(tid_t x, tid_t y)
1714 {
1715 	int difference = (x - y);
1716 	return (difference >= 0);
1717 }
1718 
1719 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1720 extern size_t journal_tag_bytes(journal_t *journal);
1721 
1722 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1723 {
1724 	return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1725 }
1726 
1727 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1728 {
1729 	WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1730 		     journal->j_chksum_driver == NULL);
1731 
1732 	return journal->j_chksum_driver != NULL;
1733 }
1734 
1735 static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb)
1736 {
1737 	int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks);
1738 
1739 	return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS;
1740 }
1741 
1742 /*
1743  * Return number of free blocks in the log. Must be called under j_state_lock.
1744  */
1745 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1746 {
1747 	/* Allow for rounding errors */
1748 	long free = journal->j_free - 32;
1749 
1750 	if (journal->j_committing_transaction) {
1751 		free -= atomic_read(&journal->
1752                         j_committing_transaction->t_outstanding_credits);
1753 	}
1754 	return max_t(long, free, 0);
1755 }
1756 
1757 /*
1758  * Definitions which augment the buffer_head layer
1759  */
1760 
1761 /* journaling buffer types */
1762 #define BJ_None		0	/* Not journaled */
1763 #define BJ_Metadata	1	/* Normal journaled metadata */
1764 #define BJ_Forget	2	/* Buffer superseded by this transaction */
1765 #define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1766 #define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1767 #define BJ_Types	5
1768 
1769 /* JBD uses a CRC32 checksum */
1770 #define JBD_MAX_CHECKSUM_SIZE 4
1771 
1772 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1773 			      const void *address, unsigned int length)
1774 {
1775 	struct {
1776 		struct shash_desc shash;
1777 		char ctx[JBD_MAX_CHECKSUM_SIZE];
1778 	} desc;
1779 	int err;
1780 
1781 	BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1782 		JBD_MAX_CHECKSUM_SIZE);
1783 
1784 	desc.shash.tfm = journal->j_chksum_driver;
1785 	*(u32 *)desc.ctx = crc;
1786 
1787 	err = crypto_shash_update(&desc.shash, address, length);
1788 	BUG_ON(err);
1789 
1790 	return *(u32 *)desc.ctx;
1791 }
1792 
1793 /* Return most recent uncommitted transaction */
1794 static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1795 {
1796 	tid_t tid;
1797 
1798 	read_lock(&journal->j_state_lock);
1799 	tid = journal->j_commit_request;
1800 	if (journal->j_running_transaction)
1801 		tid = journal->j_running_transaction->t_tid;
1802 	read_unlock(&journal->j_state_lock);
1803 	return tid;
1804 }
1805 
1806 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1807 {
1808 	journal_t *journal;
1809 
1810 	if (!handle->h_reserved)
1811 		journal = handle->h_transaction->t_journal;
1812 	else
1813 		journal = handle->h_journal;
1814 
1815 	return handle->h_total_credits -
1816 		DIV_ROUND_UP(handle->h_revoke_credits_requested,
1817 			     journal->j_revoke_records_per_block);
1818 }
1819 
1820 #ifdef __KERNEL__
1821 
1822 #define buffer_trace_init(bh)	do {} while (0)
1823 #define print_buffer_fields(bh)	do {} while (0)
1824 #define print_buffer_trace(bh)	do {} while (0)
1825 #define BUFFER_TRACE(bh, info)	do {} while (0)
1826 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1827 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1828 
1829 #endif	/* __KERNEL__ */
1830 
1831 #define EFSBADCRC	EBADMSG		/* Bad CRC detected */
1832 #define EFSCORRUPTED	EUCLEAN		/* Filesystem is corrupted */
1833 
1834 #endif	/* _LINUX_JBD2_H */
1835