xref: /linux/include/linux/jbd2.h (revision 9009b455811b0fa1f6b0adfa94db136984db5a38)
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...)  no_printk(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_DEFAULT_FAST_COMMIT_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 
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 
346 static inline struct journal_head *bh2jh(struct buffer_head *bh)
347 {
348 	return bh->b_private;
349 }
350 
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 
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  * (Locking Documentation improved by LockDoc)
542  */
543 
544 /*
545  * Lock ranking:
546  *
547  *    j_list_lock
548  *      ->jbd_lock_bh_journal_head()	(This is "innermost")
549  *
550  *    j_state_lock
551  *    ->b_state_lock
552  *
553  *    b_state_lock
554  *    ->j_list_lock
555  *
556  *    j_state_lock
557  *    ->t_handle_lock
558  *
559  *    j_state_lock
560  *    ->j_list_lock			(journal_unmap_buffer)
561  *
562  */
563 
564 struct transaction_s
565 {
566 	/* Pointer to the journal for this transaction. [no locking] */
567 	journal_t		*t_journal;
568 
569 	/* Sequence number for this transaction [no locking] */
570 	tid_t			t_tid;
571 
572 	/*
573 	 * Transaction's current state
574 	 * [no locking - only kjournald2 alters this]
575 	 * [j_list_lock] guards transition of a transaction into T_FINISHED
576 	 * state and subsequent call of __jbd2_journal_drop_transaction()
577 	 * FIXME: needs barriers
578 	 * KLUDGE: [use j_state_lock]
579 	 */
580 	enum {
581 		T_RUNNING,
582 		T_LOCKED,
583 		T_SWITCH,
584 		T_FLUSH,
585 		T_COMMIT,
586 		T_COMMIT_DFLUSH,
587 		T_COMMIT_JFLUSH,
588 		T_COMMIT_CALLBACK,
589 		T_FINISHED
590 	}			t_state;
591 
592 	/*
593 	 * Where in the log does this transaction's commit start? [no locking]
594 	 */
595 	unsigned long		t_log_start;
596 
597 	/*
598 	 * Number of buffers on the t_buffers list [j_list_lock, no locks
599 	 * needed for jbd2 thread]
600 	 */
601 	int			t_nr_buffers;
602 
603 	/*
604 	 * Doubly-linked circular list of all buffers reserved but not yet
605 	 * modified by this transaction [j_list_lock, no locks needed fo
606 	 * jbd2 thread]
607 	 */
608 	struct journal_head	*t_reserved_list;
609 
610 	/*
611 	 * Doubly-linked circular list of all metadata buffers owned by this
612 	 * transaction [j_list_lock, no locks needed for jbd2 thread]
613 	 */
614 	struct journal_head	*t_buffers;
615 
616 	/*
617 	 * Doubly-linked circular list of all forget buffers (superseded
618 	 * buffers which we can un-checkpoint once this transaction commits)
619 	 * [j_list_lock]
620 	 */
621 	struct journal_head	*t_forget;
622 
623 	/*
624 	 * Doubly-linked circular list of all buffers still to be flushed before
625 	 * this transaction can be checkpointed. [j_list_lock]
626 	 */
627 	struct journal_head	*t_checkpoint_list;
628 
629 	/*
630 	 * Doubly-linked circular list of all buffers submitted for IO while
631 	 * checkpointing. [j_list_lock]
632 	 */
633 	struct journal_head	*t_checkpoint_io_list;
634 
635 	/*
636 	 * Doubly-linked circular list of metadata buffers being
637 	 * shadowed by log IO.  The IO buffers on the iobuf list and
638 	 * the shadow buffers on this list match each other one for
639 	 * one at all times. [j_list_lock, no locks needed for jbd2
640 	 * thread]
641 	 */
642 	struct journal_head	*t_shadow_list;
643 
644 	/*
645 	 * List of inodes associated with the transaction; e.g., ext4 uses
646 	 * this to track inodes in data=ordered and data=journal mode that
647 	 * need special handling on transaction commit; also used by ocfs2.
648 	 * [j_list_lock]
649 	 */
650 	struct list_head	t_inode_list;
651 
652 	/*
653 	 * Protects info related to handles
654 	 */
655 	spinlock_t		t_handle_lock;
656 
657 	/*
658 	 * Longest time some handle had to wait for running transaction
659 	 */
660 	unsigned long		t_max_wait;
661 
662 	/*
663 	 * When transaction started
664 	 */
665 	unsigned long		t_start;
666 
667 	/*
668 	 * When commit was requested [j_state_lock]
669 	 */
670 	unsigned long		t_requested;
671 
672 	/*
673 	 * Checkpointing stats [j_list_lock]
674 	 */
675 	struct transaction_chp_stats_s t_chp_stats;
676 
677 	/*
678 	 * Number of outstanding updates running on this transaction
679 	 * [none]
680 	 */
681 	atomic_t		t_updates;
682 
683 	/*
684 	 * Number of blocks reserved for this transaction in the journal.
685 	 * This is including all credits reserved when starting transaction
686 	 * handles as well as all journal descriptor blocks needed for this
687 	 * transaction. [none]
688 	 */
689 	atomic_t		t_outstanding_credits;
690 
691 	/*
692 	 * Number of revoke records for this transaction added by already
693 	 * stopped handles. [none]
694 	 */
695 	atomic_t		t_outstanding_revokes;
696 
697 	/*
698 	 * How many handles used this transaction? [none]
699 	 */
700 	atomic_t		t_handle_count;
701 
702 	/*
703 	 * Forward and backward links for the circular list of all transactions
704 	 * awaiting checkpoint. [j_list_lock]
705 	 */
706 	transaction_t		*t_cpnext, *t_cpprev;
707 
708 	/*
709 	 * When will the transaction expire (become due for commit), in jiffies?
710 	 * [no locking]
711 	 */
712 	unsigned long		t_expires;
713 
714 	/*
715 	 * When this transaction started, in nanoseconds [no locking]
716 	 */
717 	ktime_t			t_start_time;
718 
719 	/*
720 	 * This transaction is being forced and some process is
721 	 * waiting for it to finish.
722 	 */
723 	unsigned int t_synchronous_commit:1;
724 
725 	/* Disk flush needs to be sent to fs partition [no locking] */
726 	int			t_need_data_flush;
727 
728 	/*
729 	 * For use by the filesystem to store fs-specific data
730 	 * structures associated with the transaction
731 	 */
732 	struct list_head	t_private_list;
733 };
734 
735 struct transaction_run_stats_s {
736 	unsigned long		rs_wait;
737 	unsigned long		rs_request_delay;
738 	unsigned long		rs_running;
739 	unsigned long		rs_locked;
740 	unsigned long		rs_flushing;
741 	unsigned long		rs_logging;
742 
743 	__u32			rs_handle_count;
744 	__u32			rs_blocks;
745 	__u32			rs_blocks_logged;
746 };
747 
748 struct transaction_stats_s {
749 	unsigned long		ts_tid;
750 	unsigned long		ts_requested;
751 	struct transaction_run_stats_s run;
752 };
753 
754 static inline unsigned long
755 jbd2_time_diff(unsigned long start, unsigned long end)
756 {
757 	if (end >= start)
758 		return end - start;
759 
760 	return end + (MAX_JIFFY_OFFSET - start);
761 }
762 
763 #define JBD2_NR_BATCH	64
764 
765 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
766 
767 #define JBD2_FC_REPLAY_STOP	0
768 #define JBD2_FC_REPLAY_CONTINUE	1
769 
770 /**
771  * struct journal_s - The journal_s type is the concrete type associated with
772  *     journal_t.
773  */
774 struct journal_s
775 {
776 	/**
777 	 * @j_flags: General journaling state flags [j_state_lock,
778 	 * no lock for quick racy checks]
779 	 */
780 	unsigned long		j_flags;
781 
782 	/**
783 	 * @j_errno:
784 	 *
785 	 * Is there an outstanding uncleared error on the journal (from a prior
786 	 * abort)? [j_state_lock]
787 	 */
788 	int			j_errno;
789 
790 	/**
791 	 * @j_abort_mutex: Lock the whole aborting procedure.
792 	 */
793 	struct mutex		j_abort_mutex;
794 
795 	/**
796 	 * @j_sb_buffer: The first part of the superblock buffer.
797 	 */
798 	struct buffer_head	*j_sb_buffer;
799 
800 	/**
801 	 * @j_superblock: The second part of the superblock buffer.
802 	 */
803 	journal_superblock_t	*j_superblock;
804 
805 	/**
806 	 * @j_format_version: Version of the superblock format.
807 	 */
808 	int			j_format_version;
809 
810 	/**
811 	 * @j_state_lock: Protect the various scalars in the journal.
812 	 */
813 	rwlock_t		j_state_lock;
814 
815 	/**
816 	 * @j_barrier_count:
817 	 *
818 	 * Number of processes waiting to create a barrier lock [j_state_lock,
819 	 * no lock for quick racy checks]
820 	 */
821 	int			j_barrier_count;
822 
823 	/**
824 	 * @j_barrier: The barrier lock itself.
825 	 */
826 	struct mutex		j_barrier;
827 
828 	/**
829 	 * @j_running_transaction:
830 	 *
831 	 * Transactions: The current running transaction...
832 	 * [j_state_lock, no lock for quick racy checks] [caller holding
833 	 * open handle]
834 	 */
835 	transaction_t		*j_running_transaction;
836 
837 	/**
838 	 * @j_committing_transaction:
839 	 *
840 	 * the transaction we are pushing to disk
841 	 * [j_state_lock] [caller holding open handle]
842 	 */
843 	transaction_t		*j_committing_transaction;
844 
845 	/**
846 	 * @j_checkpoint_transactions:
847 	 *
848 	 * ... and a linked circular list of all transactions waiting for
849 	 * checkpointing. [j_list_lock]
850 	 */
851 	transaction_t		*j_checkpoint_transactions;
852 
853 	/**
854 	 * @j_wait_transaction_locked:
855 	 *
856 	 * Wait queue for waiting for a locked transaction to start committing,
857 	 * or for a barrier lock to be released.
858 	 */
859 	wait_queue_head_t	j_wait_transaction_locked;
860 
861 	/**
862 	 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
863 	 */
864 	wait_queue_head_t	j_wait_done_commit;
865 
866 	/**
867 	 * @j_wait_commit: Wait queue to trigger commit.
868 	 */
869 	wait_queue_head_t	j_wait_commit;
870 
871 	/**
872 	 * @j_wait_updates: Wait queue to wait for updates to complete.
873 	 */
874 	wait_queue_head_t	j_wait_updates;
875 
876 	/**
877 	 * @j_wait_reserved:
878 	 *
879 	 * Wait queue to wait for reserved buffer credits to drop.
880 	 */
881 	wait_queue_head_t	j_wait_reserved;
882 
883 	/**
884 	 * @j_fc_wait:
885 	 *
886 	 * Wait queue to wait for completion of async fast commits.
887 	 */
888 	wait_queue_head_t	j_fc_wait;
889 
890 	/**
891 	 * @j_checkpoint_mutex:
892 	 *
893 	 * Semaphore for locking against concurrent checkpoints.
894 	 */
895 	struct mutex		j_checkpoint_mutex;
896 
897 	/**
898 	 * @j_chkpt_bhs:
899 	 *
900 	 * List of buffer heads used by the checkpoint routine.  This
901 	 * was moved from jbd2_log_do_checkpoint() to reduce stack
902 	 * usage.  Access to this array is controlled by the
903 	 * @j_checkpoint_mutex.  [j_checkpoint_mutex]
904 	 */
905 	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
906 
907 	/**
908 	 * @j_head:
909 	 *
910 	 * Journal head: identifies the first unused block in the journal.
911 	 * [j_state_lock]
912 	 */
913 	unsigned long		j_head;
914 
915 	/**
916 	 * @j_tail:
917 	 *
918 	 * Journal tail: identifies the oldest still-used block in the journal.
919 	 * [j_state_lock]
920 	 */
921 	unsigned long		j_tail;
922 
923 	/**
924 	 * @j_free:
925 	 *
926 	 * Journal free: how many free blocks are there in the journal?
927 	 * [j_state_lock]
928 	 */
929 	unsigned long		j_free;
930 
931 	/**
932 	 * @j_first:
933 	 *
934 	 * The block number of the first usable block in the journal
935 	 * [j_state_lock].
936 	 */
937 	unsigned long		j_first;
938 
939 	/**
940 	 * @j_last:
941 	 *
942 	 * The block number one beyond the last usable block in the journal
943 	 * [j_state_lock].
944 	 */
945 	unsigned long		j_last;
946 
947 	/**
948 	 * @j_fc_first:
949 	 *
950 	 * The block number of the first fast commit block in the journal
951 	 * [j_state_lock].
952 	 */
953 	unsigned long		j_fc_first;
954 
955 	/**
956 	 * @j_fc_off:
957 	 *
958 	 * Number of fast commit blocks currently allocated. Accessed only
959 	 * during fast commit. Currently only process can do fast commit, so
960 	 * this field is not protected by any lock.
961 	 */
962 	unsigned long		j_fc_off;
963 
964 	/**
965 	 * @j_fc_last:
966 	 *
967 	 * The block number one beyond the last fast commit block in the journal
968 	 * [j_state_lock].
969 	 */
970 	unsigned long		j_fc_last;
971 
972 	/**
973 	 * @j_dev: Device where we store the journal.
974 	 */
975 	struct block_device	*j_dev;
976 
977 	/**
978 	 * @j_blocksize: Block size for the location where we store the journal.
979 	 */
980 	int			j_blocksize;
981 
982 	/**
983 	 * @j_blk_offset:
984 	 *
985 	 * Starting block offset into the device where we store the journal.
986 	 */
987 	unsigned long long	j_blk_offset;
988 
989 	/**
990 	 * @j_devname: Journal device name.
991 	 */
992 	char			j_devname[BDEVNAME_SIZE+24];
993 
994 	/**
995 	 * @j_fs_dev:
996 	 *
997 	 * Device which holds the client fs.  For internal journal this will be
998 	 * equal to j_dev.
999 	 */
1000 	struct block_device	*j_fs_dev;
1001 
1002 	/**
1003 	 * @j_total_len: Total maximum capacity of the journal region on disk.
1004 	 */
1005 	unsigned int		j_total_len;
1006 
1007 	/**
1008 	 * @j_reserved_credits:
1009 	 *
1010 	 * Number of buffers reserved from the running transaction.
1011 	 */
1012 	atomic_t		j_reserved_credits;
1013 
1014 	/**
1015 	 * @j_list_lock: Protects the buffer lists and internal buffer state.
1016 	 */
1017 	spinlock_t		j_list_lock;
1018 
1019 	/**
1020 	 * @j_inode:
1021 	 *
1022 	 * Optional inode where we store the journal.  If present, all
1023 	 * journal block numbers are mapped into this inode via bmap().
1024 	 */
1025 	struct inode		*j_inode;
1026 
1027 	/**
1028 	 * @j_tail_sequence:
1029 	 *
1030 	 * Sequence number of the oldest transaction in the log [j_state_lock]
1031 	 */
1032 	tid_t			j_tail_sequence;
1033 
1034 	/**
1035 	 * @j_transaction_sequence:
1036 	 *
1037 	 * Sequence number of the next transaction to grant [j_state_lock]
1038 	 */
1039 	tid_t			j_transaction_sequence;
1040 
1041 	/**
1042 	 * @j_commit_sequence:
1043 	 *
1044 	 * Sequence number of the most recently committed transaction
1045 	 * [j_state_lock, no lock for quick racy checks]
1046 	 */
1047 	tid_t			j_commit_sequence;
1048 
1049 	/**
1050 	 * @j_commit_request:
1051 	 *
1052 	 * Sequence number of the most recent transaction wanting commit
1053 	 * [j_state_lock, no lock for quick racy checks]
1054 	 */
1055 	tid_t			j_commit_request;
1056 
1057 	/**
1058 	 * @j_uuid:
1059 	 *
1060 	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1061 	 * backed by this journal.  This will eventually be replaced by an array
1062 	 * of uuids, allowing us to index multiple devices within a single
1063 	 * journal and to perform atomic updates across them.
1064 	 */
1065 	__u8			j_uuid[16];
1066 
1067 	/**
1068 	 * @j_task: Pointer to the current commit thread for this journal.
1069 	 */
1070 	struct task_struct	*j_task;
1071 
1072 	/**
1073 	 * @j_max_transaction_buffers:
1074 	 *
1075 	 * Maximum number of metadata buffers to allow in a single compound
1076 	 * commit transaction.
1077 	 */
1078 	int			j_max_transaction_buffers;
1079 
1080 	/**
1081 	 * @j_revoke_records_per_block:
1082 	 *
1083 	 * Number of revoke records that fit in one descriptor block.
1084 	 */
1085 	int			j_revoke_records_per_block;
1086 
1087 	/**
1088 	 * @j_commit_interval:
1089 	 *
1090 	 * What is the maximum transaction lifetime before we begin a commit?
1091 	 */
1092 	unsigned long		j_commit_interval;
1093 
1094 	/**
1095 	 * @j_commit_timer: The timer used to wakeup the commit thread.
1096 	 */
1097 	struct timer_list	j_commit_timer;
1098 
1099 	/**
1100 	 * @j_revoke_lock: Protect the revoke table.
1101 	 */
1102 	spinlock_t		j_revoke_lock;
1103 
1104 	/**
1105 	 * @j_revoke:
1106 	 *
1107 	 * The revoke table - maintains the list of revoked blocks in the
1108 	 * current transaction.
1109 	 */
1110 	struct jbd2_revoke_table_s *j_revoke;
1111 
1112 	/**
1113 	 * @j_revoke_table: Alternate revoke tables for j_revoke.
1114 	 */
1115 	struct jbd2_revoke_table_s *j_revoke_table[2];
1116 
1117 	/**
1118 	 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1119 	 */
1120 	struct buffer_head	**j_wbuf;
1121 
1122 	/**
1123 	 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only
1124 	 * during a fast commit. Currently only process can do fast commit, so
1125 	 * this field is not protected by any lock.
1126 	 */
1127 	struct buffer_head	**j_fc_wbuf;
1128 
1129 	/**
1130 	 * @j_wbufsize:
1131 	 *
1132 	 * Size of @j_wbuf array.
1133 	 */
1134 	int			j_wbufsize;
1135 
1136 	/**
1137 	 * @j_fc_wbufsize:
1138 	 *
1139 	 * Size of @j_fc_wbuf array.
1140 	 */
1141 	int			j_fc_wbufsize;
1142 
1143 	/**
1144 	 * @j_last_sync_writer:
1145 	 *
1146 	 * The pid of the last person to run a synchronous operation
1147 	 * through the journal.
1148 	 */
1149 	pid_t			j_last_sync_writer;
1150 
1151 	/**
1152 	 * @j_average_commit_time:
1153 	 *
1154 	 * The average amount of time in nanoseconds it takes to commit a
1155 	 * transaction to disk. [j_state_lock]
1156 	 */
1157 	u64			j_average_commit_time;
1158 
1159 	/**
1160 	 * @j_min_batch_time:
1161 	 *
1162 	 * Minimum time that we should wait for additional filesystem operations
1163 	 * to get batched into a synchronous handle in microseconds.
1164 	 */
1165 	u32			j_min_batch_time;
1166 
1167 	/**
1168 	 * @j_max_batch_time:
1169 	 *
1170 	 * Maximum time that we should wait for additional filesystem operations
1171 	 * to get batched into a synchronous handle in microseconds.
1172 	 */
1173 	u32			j_max_batch_time;
1174 
1175 	/**
1176 	 * @j_commit_callback:
1177 	 *
1178 	 * This function is called when a transaction is closed.
1179 	 */
1180 	void			(*j_commit_callback)(journal_t *,
1181 						     transaction_t *);
1182 
1183 	/**
1184 	 * @j_submit_inode_data_buffers:
1185 	 *
1186 	 * This function is called for all inodes associated with the
1187 	 * committing transaction marked with JI_WRITE_DATA flag
1188 	 * before we start to write out the transaction to the journal.
1189 	 */
1190 	int			(*j_submit_inode_data_buffers)
1191 					(struct jbd2_inode *);
1192 
1193 	/**
1194 	 * @j_finish_inode_data_buffers:
1195 	 *
1196 	 * This function is called for all inodes associated with the
1197 	 * committing transaction marked with JI_WAIT_DATA flag
1198 	 * after we have written the transaction to the journal
1199 	 * but before we write out the commit block.
1200 	 */
1201 	int			(*j_finish_inode_data_buffers)
1202 					(struct jbd2_inode *);
1203 
1204 	/*
1205 	 * Journal statistics
1206 	 */
1207 
1208 	/**
1209 	 * @j_history_lock: Protect the transactions statistics history.
1210 	 */
1211 	spinlock_t		j_history_lock;
1212 
1213 	/**
1214 	 * @j_proc_entry: procfs entry for the jbd statistics directory.
1215 	 */
1216 	struct proc_dir_entry	*j_proc_entry;
1217 
1218 	/**
1219 	 * @j_stats: Overall statistics.
1220 	 */
1221 	struct transaction_stats_s j_stats;
1222 
1223 	/**
1224 	 * @j_failed_commit: Failed journal commit ID.
1225 	 */
1226 	unsigned int		j_failed_commit;
1227 
1228 	/**
1229 	 * @j_private:
1230 	 *
1231 	 * An opaque pointer to fs-private information.  ext3 puts its
1232 	 * superblock pointer here.
1233 	 */
1234 	void *j_private;
1235 
1236 	/**
1237 	 * @j_chksum_driver:
1238 	 *
1239 	 * Reference to checksum algorithm driver via cryptoapi.
1240 	 */
1241 	struct crypto_shash *j_chksum_driver;
1242 
1243 	/**
1244 	 * @j_csum_seed:
1245 	 *
1246 	 * Precomputed journal UUID checksum for seeding other checksums.
1247 	 */
1248 	__u32 j_csum_seed;
1249 
1250 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1251 	/**
1252 	 * @j_trans_commit_map:
1253 	 *
1254 	 * Lockdep entity to track transaction commit dependencies. Handles
1255 	 * hold this "lock" for read, when we wait for commit, we acquire the
1256 	 * "lock" for writing. This matches the properties of jbd2 journalling
1257 	 * where the running transaction has to wait for all handles to be
1258 	 * dropped to commit that transaction and also acquiring a handle may
1259 	 * require transaction commit to finish.
1260 	 */
1261 	struct lockdep_map	j_trans_commit_map;
1262 #endif
1263 
1264 	/**
1265 	 * @j_fc_cleanup_callback:
1266 	 *
1267 	 * Clean-up after fast commit or full commit. JBD2 calls this function
1268 	 * after every commit operation.
1269 	 */
1270 	void (*j_fc_cleanup_callback)(struct journal_s *journal, int);
1271 
1272 	/**
1273 	 * @j_fc_replay_callback:
1274 	 *
1275 	 * File-system specific function that performs replay of a fast
1276 	 * commit. JBD2 calls this function for each fast commit block found in
1277 	 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1278 	 * to indicate that the block was processed correctly and more fast
1279 	 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1280 	 * indicates the end of replay (no more blocks remaining). A negative
1281 	 * return value indicates error.
1282 	 */
1283 	int (*j_fc_replay_callback)(struct journal_s *journal,
1284 				    struct buffer_head *bh,
1285 				    enum passtype pass, int off,
1286 				    tid_t expected_commit_id);
1287 };
1288 
1289 #define jbd2_might_wait_for_commit(j) \
1290 	do { \
1291 		rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1292 		rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1293 	} while (0)
1294 
1295 /* journal feature predicate functions */
1296 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1297 static inline bool jbd2_has_feature_##name(journal_t *j) \
1298 { \
1299 	return ((j)->j_format_version >= 2 && \
1300 		((j)->j_superblock->s_feature_compat & \
1301 		 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1302 } \
1303 static inline void jbd2_set_feature_##name(journal_t *j) \
1304 { \
1305 	(j)->j_superblock->s_feature_compat |= \
1306 		cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1307 } \
1308 static inline void jbd2_clear_feature_##name(journal_t *j) \
1309 { \
1310 	(j)->j_superblock->s_feature_compat &= \
1311 		~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1312 }
1313 
1314 #define JBD2_FEATURE_RO_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_ro_compat & \
1319 		 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1320 } \
1321 static inline void jbd2_set_feature_##name(journal_t *j) \
1322 { \
1323 	(j)->j_superblock->s_feature_ro_compat |= \
1324 		cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1325 } \
1326 static inline void jbd2_clear_feature_##name(journal_t *j) \
1327 { \
1328 	(j)->j_superblock->s_feature_ro_compat &= \
1329 		~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1330 }
1331 
1332 #define JBD2_FEATURE_INCOMPAT_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_incompat & \
1337 		 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1338 } \
1339 static inline void jbd2_set_feature_##name(journal_t *j) \
1340 { \
1341 	(j)->j_superblock->s_feature_incompat |= \
1342 		cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1343 } \
1344 static inline void jbd2_clear_feature_##name(journal_t *j) \
1345 { \
1346 	(j)->j_superblock->s_feature_incompat &= \
1347 		~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1348 }
1349 
1350 JBD2_FEATURE_COMPAT_FUNCS(checksum,		CHECKSUM)
1351 
1352 JBD2_FEATURE_INCOMPAT_FUNCS(revoke,		REVOKE)
1353 JBD2_FEATURE_INCOMPAT_FUNCS(64bit,		64BIT)
1354 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit,	ASYNC_COMMIT)
1355 JBD2_FEATURE_INCOMPAT_FUNCS(csum2,		CSUM_V2)
1356 JBD2_FEATURE_INCOMPAT_FUNCS(csum3,		CSUM_V3)
1357 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit,	FAST_COMMIT)
1358 
1359 /*
1360  * Journal flag definitions
1361  */
1362 #define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1363 #define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1364 #define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1365 #define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1366 #define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1367 #define JBD2_BARRIER	0x020	/* Use IDE barriers */
1368 #define JBD2_ABORT_ON_SYNCDATA_ERR	0x040	/* Abort the journal on file
1369 						 * data write error in ordered
1370 						 * mode */
1371 #define JBD2_FAST_COMMIT_ONGOING	0x100	/* Fast commit is ongoing */
1372 #define JBD2_FULL_COMMIT_ONGOING	0x200	/* Full commit is ongoing */
1373 
1374 /*
1375  * Function declarations for the journaling transaction and buffer
1376  * management
1377  */
1378 
1379 /* Filing buffers */
1380 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1381 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1382 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1383 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1384 extern void __journal_free_buffer(struct journal_head *bh);
1385 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1386 extern void __journal_clean_data_list(transaction_t *transaction);
1387 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1388 {
1389 	list_add_tail(&bh->b_assoc_buffers, head);
1390 }
1391 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1392 {
1393 	list_del_init(&bh->b_assoc_buffers);
1394 }
1395 
1396 /* Log buffer allocation */
1397 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1398 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1399 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1400 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1401 			      unsigned long *block);
1402 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1403 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1404 
1405 /* Commit management */
1406 extern void jbd2_journal_commit_transaction(journal_t *);
1407 
1408 /* Checkpoint list management */
1409 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1410 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1411 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1412 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1413 
1414 
1415 /*
1416  * Triggers
1417  */
1418 
1419 struct jbd2_buffer_trigger_type {
1420 	/*
1421 	 * Fired a the moment data to write to the journal are known to be
1422 	 * stable - so either at the moment b_frozen_data is created or just
1423 	 * before a buffer is written to the journal.  mapped_data is a mapped
1424 	 * buffer that is the frozen data for commit.
1425 	 */
1426 	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1427 			 struct buffer_head *bh, void *mapped_data,
1428 			 size_t size);
1429 
1430 	/*
1431 	 * Fired during journal abort for dirty buffers that will not be
1432 	 * committed.
1433 	 */
1434 	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1435 			struct buffer_head *bh);
1436 };
1437 
1438 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1439 				       void *mapped_data,
1440 				       struct jbd2_buffer_trigger_type *triggers);
1441 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1442 				      struct jbd2_buffer_trigger_type *triggers);
1443 
1444 /* Buffer IO */
1445 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1446 					      struct journal_head *jh_in,
1447 					      struct buffer_head **bh_out,
1448 					      sector_t blocknr);
1449 
1450 /* Transaction locking */
1451 extern void		__wait_on_journal (journal_t *);
1452 
1453 /* Transaction cache support */
1454 extern void jbd2_journal_destroy_transaction_cache(void);
1455 extern int __init jbd2_journal_init_transaction_cache(void);
1456 extern void jbd2_journal_free_transaction(transaction_t *);
1457 
1458 /*
1459  * Journal locking.
1460  *
1461  * We need to lock the journal during transaction state changes so that nobody
1462  * ever tries to take a handle on the running transaction while we are in the
1463  * middle of moving it to the commit phase.  j_state_lock does this.
1464  *
1465  * Note that the locking is completely interrupt unsafe.  We never touch
1466  * journal structures from interrupts.
1467  */
1468 
1469 static inline handle_t *journal_current_handle(void)
1470 {
1471 	return current->journal_info;
1472 }
1473 
1474 /* The journaling code user interface:
1475  *
1476  * Create and destroy handles
1477  * Register buffer modifications against the current transaction.
1478  */
1479 
1480 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1481 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1482 				     int revoke_records, gfp_t gfp_mask,
1483 				     unsigned int type, unsigned int line_no);
1484 extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1485 extern int	 jbd2__journal_restart(handle_t *, int nblocks,
1486 				       int revoke_records, gfp_t gfp_mask);
1487 extern int	 jbd2_journal_start_reserved(handle_t *handle,
1488 				unsigned int type, unsigned int line_no);
1489 extern void	 jbd2_journal_free_reserved(handle_t *handle);
1490 extern int	 jbd2_journal_extend(handle_t *handle, int nblocks,
1491 				     int revoke_records);
1492 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1493 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1494 extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1495 void		 jbd2_journal_set_triggers(struct buffer_head *,
1496 					   struct jbd2_buffer_trigger_type *type);
1497 extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1498 extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1499 extern int	 jbd2_journal_invalidatepage(journal_t *,
1500 				struct page *, unsigned int, unsigned int);
1501 extern int	 jbd2_journal_try_to_free_buffers(journal_t *journal, struct page *page);
1502 extern int	 jbd2_journal_stop(handle_t *);
1503 extern int	 jbd2_journal_flush (journal_t *);
1504 extern void	 jbd2_journal_lock_updates (journal_t *);
1505 extern void	 jbd2_journal_unlock_updates (journal_t *);
1506 
1507 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1508 				struct block_device *fs_dev,
1509 				unsigned long long start, int len, int bsize);
1510 extern journal_t * jbd2_journal_init_inode (struct inode *);
1511 extern int	   jbd2_journal_update_format (journal_t *);
1512 extern int	   jbd2_journal_check_used_features
1513 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1514 extern int	   jbd2_journal_check_available_features
1515 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1516 extern int	   jbd2_journal_set_features
1517 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1518 extern void	   jbd2_journal_clear_features
1519 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1520 extern int	   jbd2_journal_load       (journal_t *journal);
1521 extern int	   jbd2_journal_destroy    (journal_t *);
1522 extern int	   jbd2_journal_recover    (journal_t *journal);
1523 extern int	   jbd2_journal_wipe       (journal_t *, int);
1524 extern int	   jbd2_journal_skip_recovery	(journal_t *);
1525 extern void	   jbd2_journal_update_sb_errno(journal_t *);
1526 extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1527 				unsigned long, int);
1528 extern void	   jbd2_journal_abort      (journal_t *, int);
1529 extern int	   jbd2_journal_errno      (journal_t *);
1530 extern void	   jbd2_journal_ack_err    (journal_t *);
1531 extern int	   jbd2_journal_clear_err  (journal_t *);
1532 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1533 extern int	   jbd2_journal_force_commit(journal_t *);
1534 extern int	   jbd2_journal_force_commit_nested(journal_t *);
1535 extern int	   jbd2_journal_inode_ranged_write(handle_t *handle,
1536 			struct jbd2_inode *inode, loff_t start_byte,
1537 			loff_t length);
1538 extern int	   jbd2_journal_inode_ranged_wait(handle_t *handle,
1539 			struct jbd2_inode *inode, loff_t start_byte,
1540 			loff_t length);
1541 extern int	   jbd2_journal_submit_inode_data_buffers(
1542 			struct jbd2_inode *jinode);
1543 extern int	   jbd2_journal_finish_inode_data_buffers(
1544 			struct jbd2_inode *jinode);
1545 extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1546 				struct jbd2_inode *inode, loff_t new_size);
1547 extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1548 extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1549 
1550 /*
1551  * journal_head management
1552  */
1553 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1554 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1555 void jbd2_journal_put_journal_head(struct journal_head *jh);
1556 
1557 /*
1558  * handle management
1559  */
1560 extern struct kmem_cache *jbd2_handle_cache;
1561 
1562 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1563 {
1564 	return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1565 }
1566 
1567 static inline void jbd2_free_handle(handle_t *handle)
1568 {
1569 	kmem_cache_free(jbd2_handle_cache, handle);
1570 }
1571 
1572 /*
1573  * jbd2_inode management (optional, for those file systems that want to use
1574  * dynamically allocated jbd2_inode structures)
1575  */
1576 extern struct kmem_cache *jbd2_inode_cache;
1577 
1578 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1579 {
1580 	return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1581 }
1582 
1583 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1584 {
1585 	kmem_cache_free(jbd2_inode_cache, jinode);
1586 }
1587 
1588 /* Primary revoke support */
1589 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1590 extern int	   jbd2_journal_init_revoke(journal_t *, int);
1591 extern void	   jbd2_journal_destroy_revoke_record_cache(void);
1592 extern void	   jbd2_journal_destroy_revoke_table_cache(void);
1593 extern int __init jbd2_journal_init_revoke_record_cache(void);
1594 extern int __init jbd2_journal_init_revoke_table_cache(void);
1595 
1596 extern void	   jbd2_journal_destroy_revoke(journal_t *);
1597 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1598 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1599 extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
1600 						     struct list_head *log_bufs);
1601 
1602 /* Recovery revoke support */
1603 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1604 extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1605 extern void	jbd2_journal_clear_revoke(journal_t *);
1606 extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1607 extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1608 
1609 /*
1610  * The log thread user interface:
1611  *
1612  * Request space in the current transaction, and force transaction commit
1613  * transitions on demand.
1614  */
1615 
1616 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1617 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1618 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1619 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1620 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1621 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1622 int jbd2_log_do_checkpoint(journal_t *journal);
1623 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1624 
1625 void __jbd2_log_wait_for_space(journal_t *journal);
1626 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1627 extern int jbd2_cleanup_journal_tail(journal_t *);
1628 
1629 /* Fast commit related APIs */
1630 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1631 int jbd2_fc_end_commit(journal_t *journal);
1632 int jbd2_fc_end_commit_fallback(journal_t *journal);
1633 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1634 int jbd2_submit_inode_data(struct jbd2_inode *jinode);
1635 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1636 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1637 int jbd2_fc_release_bufs(journal_t *journal);
1638 
1639 static inline int jbd2_journal_get_max_txn_bufs(journal_t *journal)
1640 {
1641 	return (journal->j_total_len - journal->j_fc_wbufsize) / 4;
1642 }
1643 
1644 /*
1645  * is_journal_abort
1646  *
1647  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1648  * bit, when set, indicates that we have had a fatal error somewhere,
1649  * either inside the journaling layer or indicated to us by the client
1650  * (eg. ext3), and that we and should not commit any further
1651  * transactions.
1652  */
1653 
1654 static inline int is_journal_aborted(journal_t *journal)
1655 {
1656 	return journal->j_flags & JBD2_ABORT;
1657 }
1658 
1659 static inline int is_handle_aborted(handle_t *handle)
1660 {
1661 	if (handle->h_aborted || !handle->h_transaction)
1662 		return 1;
1663 	return is_journal_aborted(handle->h_transaction->t_journal);
1664 }
1665 
1666 static inline void jbd2_journal_abort_handle(handle_t *handle)
1667 {
1668 	handle->h_aborted = 1;
1669 }
1670 
1671 #endif /* __KERNEL__   */
1672 
1673 /* Comparison functions for transaction IDs: perform comparisons using
1674  * modulo arithmetic so that they work over sequence number wraps. */
1675 
1676 static inline int tid_gt(tid_t x, tid_t y)
1677 {
1678 	int difference = (x - y);
1679 	return (difference > 0);
1680 }
1681 
1682 static inline int tid_geq(tid_t x, tid_t y)
1683 {
1684 	int difference = (x - y);
1685 	return (difference >= 0);
1686 }
1687 
1688 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1689 extern size_t journal_tag_bytes(journal_t *journal);
1690 
1691 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1692 {
1693 	return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1694 }
1695 
1696 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1697 {
1698 	WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1699 		     journal->j_chksum_driver == NULL);
1700 
1701 	return journal->j_chksum_driver != NULL;
1702 }
1703 
1704 static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb)
1705 {
1706 	int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks);
1707 
1708 	return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS;
1709 }
1710 
1711 /*
1712  * Return number of free blocks in the log. Must be called under j_state_lock.
1713  */
1714 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1715 {
1716 	/* Allow for rounding errors */
1717 	long free = journal->j_free - 32;
1718 
1719 	if (journal->j_committing_transaction) {
1720 		free -= atomic_read(&journal->
1721                         j_committing_transaction->t_outstanding_credits);
1722 	}
1723 	return max_t(long, free, 0);
1724 }
1725 
1726 /*
1727  * Definitions which augment the buffer_head layer
1728  */
1729 
1730 /* journaling buffer types */
1731 #define BJ_None		0	/* Not journaled */
1732 #define BJ_Metadata	1	/* Normal journaled metadata */
1733 #define BJ_Forget	2	/* Buffer superseded by this transaction */
1734 #define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1735 #define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1736 #define BJ_Types	5
1737 
1738 extern int jbd_blocks_per_page(struct inode *inode);
1739 
1740 /* JBD uses a CRC32 checksum */
1741 #define JBD_MAX_CHECKSUM_SIZE 4
1742 
1743 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1744 			      const void *address, unsigned int length)
1745 {
1746 	struct {
1747 		struct shash_desc shash;
1748 		char ctx[JBD_MAX_CHECKSUM_SIZE];
1749 	} desc;
1750 	int err;
1751 
1752 	BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1753 		JBD_MAX_CHECKSUM_SIZE);
1754 
1755 	desc.shash.tfm = journal->j_chksum_driver;
1756 	*(u32 *)desc.ctx = crc;
1757 
1758 	err = crypto_shash_update(&desc.shash, address, length);
1759 	BUG_ON(err);
1760 
1761 	return *(u32 *)desc.ctx;
1762 }
1763 
1764 /* Return most recent uncommitted transaction */
1765 static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1766 {
1767 	tid_t tid;
1768 
1769 	read_lock(&journal->j_state_lock);
1770 	tid = journal->j_commit_request;
1771 	if (journal->j_running_transaction)
1772 		tid = journal->j_running_transaction->t_tid;
1773 	read_unlock(&journal->j_state_lock);
1774 	return tid;
1775 }
1776 
1777 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1778 {
1779 	journal_t *journal;
1780 
1781 	if (!handle->h_reserved)
1782 		journal = handle->h_transaction->t_journal;
1783 	else
1784 		journal = handle->h_journal;
1785 
1786 	return handle->h_total_credits -
1787 		DIV_ROUND_UP(handle->h_revoke_credits_requested,
1788 			     journal->j_revoke_records_per_block);
1789 }
1790 
1791 #ifdef __KERNEL__
1792 
1793 #define buffer_trace_init(bh)	do {} while (0)
1794 #define print_buffer_fields(bh)	do {} while (0)
1795 #define print_buffer_trace(bh)	do {} while (0)
1796 #define BUFFER_TRACE(bh, info)	do {} while (0)
1797 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1798 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1799 
1800 #endif	/* __KERNEL__ */
1801 
1802 #define EFSBADCRC	EBADMSG		/* Bad CRC detected */
1803 #define EFSCORRUPTED	EUCLEAN		/* Filesystem is corrupted */
1804 
1805 #endif	/* _LINUX_JBD2_H */
1806