xref: /linux/include/linux/jbd2.h (revision 43bb48c38e817b5f89fce340f49436a605e47e66)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * linux/include/linux/jbd2.h
4  *
5  * Written by Stephen C. Tweedie <sct@redhat.com>
6  *
7  * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
8  *
9  * Definitions for transaction data structures for the buffer cache
10  * filesystem journaling support.
11  */
12 
13 #ifndef _LINUX_JBD2_H
14 #define _LINUX_JBD2_H
15 
16 /* Allow this file to be included directly into e2fsprogs */
17 #ifndef __KERNEL__
18 #include "jfs_compat.h"
19 #define JBD2_DEBUG
20 #else
21 
22 #include <linux/types.h>
23 #include <linux/buffer_head.h>
24 #include <linux/journal-head.h>
25 #include <linux/stddef.h>
26 #include <linux/mutex.h>
27 #include <linux/timer.h>
28 #include <linux/slab.h>
29 #include <linux/bit_spinlock.h>
30 #include <linux/blkdev.h>
31 #include <crypto/hash.h>
32 #endif
33 
34 #define journal_oom_retry 1
35 
36 /*
37  * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
38  * certain classes of error which can occur due to failed IOs.  Under
39  * normal use we want ext4 to continue after such errors, because
40  * hardware _can_ fail, but for debugging purposes when running tests on
41  * known-good hardware we may want to trap these errors.
42  */
43 #undef JBD2_PARANOID_IOFAIL
44 
45 /*
46  * The default maximum commit age, in seconds.
47  */
48 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
49 
50 #ifdef CONFIG_JBD2_DEBUG
51 /*
52  * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
53  * consistency checks.  By default we don't do this unless
54  * CONFIG_JBD2_DEBUG is on.
55  */
56 #define JBD2_EXPENSIVE_CHECKING
57 extern ushort jbd2_journal_enable_debug;
58 void __jbd2_debug(int level, const char *file, const char *func,
59 		  unsigned int line, const char *fmt, ...);
60 
61 #define jbd_debug(n, fmt, a...) \
62 	__jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
63 #else
64 #define jbd_debug(n, fmt, a...)    /**/
65 #endif
66 
67 extern void *jbd2_alloc(size_t size, gfp_t flags);
68 extern void jbd2_free(void *ptr, size_t size);
69 
70 #define JBD2_MIN_JOURNAL_BLOCKS 1024
71 
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 	__u32	s_padding[42];
267 	__be32	s_checksum;		/* crc32c(superblock) */
268 
269 /* 0x0100 */
270 	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
271 /* 0x0400 */
272 } journal_superblock_t;
273 
274 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
275 #define JBD2_HAS_COMPAT_FEATURE(j,mask)					\
276 	((j)->j_format_version >= 2 &&					\
277 	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
278 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask)				\
279 	((j)->j_format_version >= 2 &&					\
280 	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
281 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask)				\
282 	((j)->j_format_version >= 2 &&					\
283 	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
284 
285 #define JBD2_FEATURE_COMPAT_CHECKSUM		0x00000001
286 
287 #define JBD2_FEATURE_INCOMPAT_REVOKE		0x00000001
288 #define JBD2_FEATURE_INCOMPAT_64BIT		0x00000002
289 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
290 #define JBD2_FEATURE_INCOMPAT_CSUM_V2		0x00000008
291 #define JBD2_FEATURE_INCOMPAT_CSUM_V3		0x00000010
292 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT	0x00000020
293 
294 /* See "journal feature predicate functions" below */
295 
296 /* Features known to this kernel version: */
297 #define JBD2_KNOWN_COMPAT_FEATURES	JBD2_FEATURE_COMPAT_CHECKSUM
298 #define JBD2_KNOWN_ROCOMPAT_FEATURES	0
299 #define JBD2_KNOWN_INCOMPAT_FEATURES	(JBD2_FEATURE_INCOMPAT_REVOKE | \
300 					JBD2_FEATURE_INCOMPAT_64BIT | \
301 					JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
302 					JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
303 					JBD2_FEATURE_INCOMPAT_CSUM_V3 | \
304 					JBD2_FEATURE_INCOMPAT_FAST_COMMIT)
305 
306 #ifdef __KERNEL__
307 
308 #include <linux/fs.h>
309 #include <linux/sched.h>
310 
311 enum jbd_state_bits {
312 	BH_JBD			/* Has an attached ext3 journal_head */
313 	  = BH_PrivateStart,
314 	BH_JWrite,		/* Being written to log (@@@ DEBUGGING) */
315 	BH_Freed,		/* Has been freed (truncated) */
316 	BH_Revoked,		/* Has been revoked from the log */
317 	BH_RevokeValid,		/* Revoked flag is valid */
318 	BH_JBDDirty,		/* Is dirty but journaled */
319 	BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
320 	BH_Shadow,		/* IO on shadow buffer is running */
321 	BH_Verified,		/* Metadata block has been verified ok */
322 	BH_JBDPrivateStart,	/* First bit available for private use by FS */
323 };
324 
325 BUFFER_FNS(JBD, jbd)
326 BUFFER_FNS(JWrite, jwrite)
327 BUFFER_FNS(JBDDirty, jbddirty)
328 TAS_BUFFER_FNS(JBDDirty, jbddirty)
329 BUFFER_FNS(Revoked, revoked)
330 TAS_BUFFER_FNS(Revoked, revoked)
331 BUFFER_FNS(RevokeValid, revokevalid)
332 TAS_BUFFER_FNS(RevokeValid, revokevalid)
333 BUFFER_FNS(Freed, freed)
334 BUFFER_FNS(Shadow, shadow)
335 BUFFER_FNS(Verified, verified)
336 
337 static inline struct buffer_head *jh2bh(struct journal_head *jh)
338 {
339 	return jh->b_bh;
340 }
341 
342 static inline struct journal_head *bh2jh(struct buffer_head *bh)
343 {
344 	return bh->b_private;
345 }
346 
347 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
348 {
349 	bit_spin_lock(BH_JournalHead, &bh->b_state);
350 }
351 
352 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
353 {
354 	bit_spin_unlock(BH_JournalHead, &bh->b_state);
355 }
356 
357 #define J_ASSERT(assert)	BUG_ON(!(assert))
358 
359 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
360 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
361 
362 #if defined(JBD2_PARANOID_IOFAIL)
363 #define J_EXPECT(expr, why...)		J_ASSERT(expr)
364 #define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
365 #define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
366 #else
367 #define __journal_expect(expr, why...)					     \
368 	({								     \
369 		int val = (expr);					     \
370 		if (!val) {						     \
371 			printk(KERN_ERR					     \
372 			       "JBD2 unexpected failure: %s: %s;\n",	     \
373 			       __func__, #expr);			     \
374 			printk(KERN_ERR why "\n");			     \
375 		}							     \
376 		val;							     \
377 	})
378 #define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
379 #define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
380 #define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
381 #endif
382 
383 /* Flags in jbd_inode->i_flags */
384 #define __JI_COMMIT_RUNNING 0
385 #define __JI_WRITE_DATA 1
386 #define __JI_WAIT_DATA 2
387 
388 /*
389  * Commit of the inode data in progress. We use this flag to protect us from
390  * concurrent deletion of inode. We cannot use reference to inode for this
391  * since we cannot afford doing last iput() on behalf of kjournald
392  */
393 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
394 /* Write allocated dirty buffers in this inode before commit */
395 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
396 /* Wait for outstanding data writes for this inode before commit */
397 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
398 
399 /**
400  * struct jbd_inode - The jbd_inode type is the structure linking inodes in
401  * ordered mode present in a transaction so that we can sync them during commit.
402  */
403 struct jbd2_inode {
404 	/**
405 	 * @i_transaction:
406 	 *
407 	 * Which transaction does this inode belong to? Either the running
408 	 * transaction or the committing one. [j_list_lock]
409 	 */
410 	transaction_t *i_transaction;
411 
412 	/**
413 	 * @i_next_transaction:
414 	 *
415 	 * Pointer to the running transaction modifying inode's data in case
416 	 * there is already a committing transaction touching it. [j_list_lock]
417 	 */
418 	transaction_t *i_next_transaction;
419 
420 	/**
421 	 * @i_list: List of inodes in the i_transaction [j_list_lock]
422 	 */
423 	struct list_head i_list;
424 
425 	/**
426 	 * @i_vfs_inode:
427 	 *
428 	 * VFS inode this inode belongs to [constant for lifetime of structure]
429 	 */
430 	struct inode *i_vfs_inode;
431 
432 	/**
433 	 * @i_flags: Flags of inode [j_list_lock]
434 	 */
435 	unsigned long i_flags;
436 
437 	/**
438 	 * @i_dirty_start:
439 	 *
440 	 * Offset in bytes where the dirty range for this inode starts.
441 	 * [j_list_lock]
442 	 */
443 	loff_t i_dirty_start;
444 
445 	/**
446 	 * @i_dirty_end:
447 	 *
448 	 * Inclusive offset in bytes where the dirty range for this inode
449 	 * ends. [j_list_lock]
450 	 */
451 	loff_t i_dirty_end;
452 };
453 
454 struct jbd2_revoke_table_s;
455 
456 /**
457  * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete
458  *     type associated with handle_t.
459  * @h_transaction: Which compound transaction is this update a part of?
460  * @h_journal: Which journal handle belongs to - used iff h_reserved set.
461  * @h_rsv_handle: Handle reserved for finishing the logical operation.
462  * @h_total_credits: Number of remaining buffers we are allowed to add to
463  *	journal. These are dirty buffers and revoke descriptor blocks.
464  * @h_revoke_credits: Number of remaining revoke records available for handle
465  * @h_ref: Reference count on this handle.
466  * @h_err: Field for caller's use to track errors through large fs operations.
467  * @h_sync: Flag for sync-on-close.
468  * @h_jdata: Flag to force data journaling.
469  * @h_reserved: Flag for handle for reserved credits.
470  * @h_aborted: Flag indicating fatal error on handle.
471  * @h_type: For handle statistics.
472  * @h_line_no: For handle statistics.
473  * @h_start_jiffies: Handle Start time.
474  * @h_requested_credits: Holds @h_total_credits after handle is started.
475  * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
476  * @saved_alloc_context: Saved context while transaction is open.
477  **/
478 
479 /* Docbook can't yet cope with the bit fields, but will leave the documentation
480  * in so it can be fixed later.
481  */
482 
483 struct jbd2_journal_handle
484 {
485 	union {
486 		transaction_t	*h_transaction;
487 		/* Which journal handle belongs to - used iff h_reserved set */
488 		journal_t	*h_journal;
489 	};
490 
491 	handle_t		*h_rsv_handle;
492 	int			h_total_credits;
493 	int			h_revoke_credits;
494 	int			h_revoke_credits_requested;
495 	int			h_ref;
496 	int			h_err;
497 
498 	/* Flags [no locking] */
499 	unsigned int	h_sync:		1;
500 	unsigned int	h_jdata:	1;
501 	unsigned int	h_reserved:	1;
502 	unsigned int	h_aborted:	1;
503 	unsigned int	h_type:		8;
504 	unsigned int	h_line_no:	16;
505 
506 	unsigned long		h_start_jiffies;
507 	unsigned int		h_requested_credits;
508 
509 	unsigned int		saved_alloc_context;
510 };
511 
512 
513 /*
514  * Some stats for checkpoint phase
515  */
516 struct transaction_chp_stats_s {
517 	unsigned long		cs_chp_time;
518 	__u32			cs_forced_to_close;
519 	__u32			cs_written;
520 	__u32			cs_dropped;
521 };
522 
523 /* The transaction_t type is the guts of the journaling mechanism.  It
524  * tracks a compound transaction through its various states:
525  *
526  * RUNNING:	accepting new updates
527  * LOCKED:	Updates still running but we don't accept new ones
528  * RUNDOWN:	Updates are tidying up but have finished requesting
529  *		new buffers to modify (state not used for now)
530  * FLUSH:       All updates complete, but we are still writing to disk
531  * COMMIT:      All data on disk, writing commit record
532  * FINISHED:	We still have to keep the transaction for checkpointing.
533  *
534  * The transaction keeps track of all of the buffers modified by a
535  * running transaction, and all of the buffers committed but not yet
536  * flushed to home for finished transactions.
537  */
538 
539 /*
540  * Lock ranking:
541  *
542  *    j_list_lock
543  *      ->jbd_lock_bh_journal_head()	(This is "innermost")
544  *
545  *    j_state_lock
546  *    ->b_state_lock
547  *
548  *    b_state_lock
549  *    ->j_list_lock
550  *
551  *    j_state_lock
552  *    ->t_handle_lock
553  *
554  *    j_state_lock
555  *    ->j_list_lock			(journal_unmap_buffer)
556  *
557  */
558 
559 struct transaction_s
560 {
561 	/* Pointer to the journal for this transaction. [no locking] */
562 	journal_t		*t_journal;
563 
564 	/* Sequence number for this transaction [no locking] */
565 	tid_t			t_tid;
566 
567 	/*
568 	 * Transaction's current state
569 	 * [no locking - only kjournald2 alters this]
570 	 * [j_list_lock] guards transition of a transaction into T_FINISHED
571 	 * state and subsequent call of __jbd2_journal_drop_transaction()
572 	 * FIXME: needs barriers
573 	 * KLUDGE: [use j_state_lock]
574 	 */
575 	enum {
576 		T_RUNNING,
577 		T_LOCKED,
578 		T_SWITCH,
579 		T_FLUSH,
580 		T_COMMIT,
581 		T_COMMIT_DFLUSH,
582 		T_COMMIT_JFLUSH,
583 		T_COMMIT_CALLBACK,
584 		T_FINISHED
585 	}			t_state;
586 
587 	/*
588 	 * Where in the log does this transaction's commit start? [no locking]
589 	 */
590 	unsigned long		t_log_start;
591 
592 	/* Number of buffers on the t_buffers list [j_list_lock] */
593 	int			t_nr_buffers;
594 
595 	/*
596 	 * Doubly-linked circular list of all buffers reserved but not yet
597 	 * modified by this transaction [j_list_lock]
598 	 */
599 	struct journal_head	*t_reserved_list;
600 
601 	/*
602 	 * Doubly-linked circular list of all metadata buffers owned by this
603 	 * transaction [j_list_lock]
604 	 */
605 	struct journal_head	*t_buffers;
606 
607 	/*
608 	 * Doubly-linked circular list of all forget buffers (superseded
609 	 * buffers which we can un-checkpoint once this transaction commits)
610 	 * [j_list_lock]
611 	 */
612 	struct journal_head	*t_forget;
613 
614 	/*
615 	 * Doubly-linked circular list of all buffers still to be flushed before
616 	 * this transaction can be checkpointed. [j_list_lock]
617 	 */
618 	struct journal_head	*t_checkpoint_list;
619 
620 	/*
621 	 * Doubly-linked circular list of all buffers submitted for IO while
622 	 * checkpointing. [j_list_lock]
623 	 */
624 	struct journal_head	*t_checkpoint_io_list;
625 
626 	/*
627 	 * Doubly-linked circular list of metadata buffers being shadowed by log
628 	 * IO.  The IO buffers on the iobuf list and the shadow buffers on this
629 	 * list match each other one for one at all times. [j_list_lock]
630 	 */
631 	struct journal_head	*t_shadow_list;
632 
633 	/*
634 	 * List of inodes associated with the transaction; e.g., ext4 uses
635 	 * this to track inodes in data=ordered and data=journal mode that
636 	 * need special handling on transaction commit; also used by ocfs2.
637 	 * [j_list_lock]
638 	 */
639 	struct list_head	t_inode_list;
640 
641 	/*
642 	 * Protects info related to handles
643 	 */
644 	spinlock_t		t_handle_lock;
645 
646 	/*
647 	 * Longest time some handle had to wait for running transaction
648 	 */
649 	unsigned long		t_max_wait;
650 
651 	/*
652 	 * When transaction started
653 	 */
654 	unsigned long		t_start;
655 
656 	/*
657 	 * When commit was requested
658 	 */
659 	unsigned long		t_requested;
660 
661 	/*
662 	 * Checkpointing stats [j_checkpoint_sem]
663 	 */
664 	struct transaction_chp_stats_s t_chp_stats;
665 
666 	/*
667 	 * Number of outstanding updates running on this transaction
668 	 * [none]
669 	 */
670 	atomic_t		t_updates;
671 
672 	/*
673 	 * Number of blocks reserved for this transaction in the journal.
674 	 * This is including all credits reserved when starting transaction
675 	 * handles as well as all journal descriptor blocks needed for this
676 	 * transaction. [none]
677 	 */
678 	atomic_t		t_outstanding_credits;
679 
680 	/*
681 	 * Number of revoke records for this transaction added by already
682 	 * stopped handles. [none]
683 	 */
684 	atomic_t		t_outstanding_revokes;
685 
686 	/*
687 	 * How many handles used this transaction? [none]
688 	 */
689 	atomic_t		t_handle_count;
690 
691 	/*
692 	 * Forward and backward links for the circular list of all transactions
693 	 * awaiting checkpoint. [j_list_lock]
694 	 */
695 	transaction_t		*t_cpnext, *t_cpprev;
696 
697 	/*
698 	 * When will the transaction expire (become due for commit), in jiffies?
699 	 * [no locking]
700 	 */
701 	unsigned long		t_expires;
702 
703 	/*
704 	 * When this transaction started, in nanoseconds [no locking]
705 	 */
706 	ktime_t			t_start_time;
707 
708 	/*
709 	 * This transaction is being forced and some process is
710 	 * waiting for it to finish.
711 	 */
712 	unsigned int t_synchronous_commit:1;
713 
714 	/* Disk flush needs to be sent to fs partition [no locking] */
715 	int			t_need_data_flush;
716 
717 	/*
718 	 * For use by the filesystem to store fs-specific data
719 	 * structures associated with the transaction
720 	 */
721 	struct list_head	t_private_list;
722 };
723 
724 struct transaction_run_stats_s {
725 	unsigned long		rs_wait;
726 	unsigned long		rs_request_delay;
727 	unsigned long		rs_running;
728 	unsigned long		rs_locked;
729 	unsigned long		rs_flushing;
730 	unsigned long		rs_logging;
731 
732 	__u32			rs_handle_count;
733 	__u32			rs_blocks;
734 	__u32			rs_blocks_logged;
735 };
736 
737 struct transaction_stats_s {
738 	unsigned long		ts_tid;
739 	unsigned long		ts_requested;
740 	struct transaction_run_stats_s run;
741 };
742 
743 static inline unsigned long
744 jbd2_time_diff(unsigned long start, unsigned long end)
745 {
746 	if (end >= start)
747 		return end - start;
748 
749 	return end + (MAX_JIFFY_OFFSET - start);
750 }
751 
752 #define JBD2_NR_BATCH	64
753 
754 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
755 
756 #define JBD2_FC_REPLAY_STOP	0
757 #define JBD2_FC_REPLAY_CONTINUE	1
758 
759 /**
760  * struct journal_s - The journal_s type is the concrete type associated with
761  *     journal_t.
762  */
763 struct journal_s
764 {
765 	/**
766 	 * @j_flags: General journaling state flags [j_state_lock]
767 	 */
768 	unsigned long		j_flags;
769 
770 	/**
771 	 * @j_errno:
772 	 *
773 	 * Is there an outstanding uncleared error on the journal (from a prior
774 	 * abort)? [j_state_lock]
775 	 */
776 	int			j_errno;
777 
778 	/**
779 	 * @j_abort_mutex: Lock the whole aborting procedure.
780 	 */
781 	struct mutex		j_abort_mutex;
782 
783 	/**
784 	 * @j_sb_buffer: The first part of the superblock buffer.
785 	 */
786 	struct buffer_head	*j_sb_buffer;
787 
788 	/**
789 	 * @j_superblock: The second part of the superblock buffer.
790 	 */
791 	journal_superblock_t	*j_superblock;
792 
793 	/**
794 	 * @j_format_version: Version of the superblock format.
795 	 */
796 	int			j_format_version;
797 
798 	/**
799 	 * @j_state_lock: Protect the various scalars in the journal.
800 	 */
801 	rwlock_t		j_state_lock;
802 
803 	/**
804 	 * @j_barrier_count:
805 	 *
806 	 * Number of processes waiting to create a barrier lock [j_state_lock]
807 	 */
808 	int			j_barrier_count;
809 
810 	/**
811 	 * @j_barrier: The barrier lock itself.
812 	 */
813 	struct mutex		j_barrier;
814 
815 	/**
816 	 * @j_running_transaction:
817 	 *
818 	 * Transactions: The current running transaction...
819 	 * [j_state_lock] [caller holding open handle]
820 	 */
821 	transaction_t		*j_running_transaction;
822 
823 	/**
824 	 * @j_committing_transaction:
825 	 *
826 	 * the transaction we are pushing to disk
827 	 * [j_state_lock] [caller holding open handle]
828 	 */
829 	transaction_t		*j_committing_transaction;
830 
831 	/**
832 	 * @j_checkpoint_transactions:
833 	 *
834 	 * ... and a linked circular list of all transactions waiting for
835 	 * checkpointing. [j_list_lock]
836 	 */
837 	transaction_t		*j_checkpoint_transactions;
838 
839 	/**
840 	 * @j_wait_transaction_locked:
841 	 *
842 	 * Wait queue for waiting for a locked transaction to start committing,
843 	 * or for a barrier lock to be released.
844 	 */
845 	wait_queue_head_t	j_wait_transaction_locked;
846 
847 	/**
848 	 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
849 	 */
850 	wait_queue_head_t	j_wait_done_commit;
851 
852 	/**
853 	 * @j_wait_commit: Wait queue to trigger commit.
854 	 */
855 	wait_queue_head_t	j_wait_commit;
856 
857 	/**
858 	 * @j_wait_updates: Wait queue to wait for updates to complete.
859 	 */
860 	wait_queue_head_t	j_wait_updates;
861 
862 	/**
863 	 * @j_wait_reserved:
864 	 *
865 	 * Wait queue to wait for reserved buffer credits to drop.
866 	 */
867 	wait_queue_head_t	j_wait_reserved;
868 
869 	/**
870 	 * @j_fc_wait:
871 	 *
872 	 * Wait queue to wait for completion of async fast commits.
873 	 */
874 	wait_queue_head_t	j_fc_wait;
875 
876 	/**
877 	 * @j_checkpoint_mutex:
878 	 *
879 	 * Semaphore for locking against concurrent checkpoints.
880 	 */
881 	struct mutex		j_checkpoint_mutex;
882 
883 	/**
884 	 * @j_chkpt_bhs:
885 	 *
886 	 * List of buffer heads used by the checkpoint routine.  This
887 	 * was moved from jbd2_log_do_checkpoint() to reduce stack
888 	 * usage.  Access to this array is controlled by the
889 	 * @j_checkpoint_mutex.  [j_checkpoint_mutex]
890 	 */
891 	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
892 
893 	/**
894 	 * @j_head:
895 	 *
896 	 * Journal head: identifies the first unused block in the journal.
897 	 * [j_state_lock]
898 	 */
899 	unsigned long		j_head;
900 
901 	/**
902 	 * @j_tail:
903 	 *
904 	 * Journal tail: identifies the oldest still-used block in the journal.
905 	 * [j_state_lock]
906 	 */
907 	unsigned long		j_tail;
908 
909 	/**
910 	 * @j_free:
911 	 *
912 	 * Journal free: how many free blocks are there in the journal?
913 	 * [j_state_lock]
914 	 */
915 	unsigned long		j_free;
916 
917 	/**
918 	 * @j_first:
919 	 *
920 	 * The block number of the first usable block in the journal
921 	 * [j_state_lock].
922 	 */
923 	unsigned long		j_first;
924 
925 	/**
926 	 * @j_last:
927 	 *
928 	 * The block number one beyond the last usable block in the journal
929 	 * [j_state_lock].
930 	 */
931 	unsigned long		j_last;
932 
933 	/**
934 	 * @j_fc_first:
935 	 *
936 	 * The block number of the first fast commit block in the journal
937 	 * [j_state_lock].
938 	 */
939 	unsigned long		j_fc_first;
940 
941 	/**
942 	 * @j_fc_off:
943 	 *
944 	 * Number of fast commit blocks currently allocated.
945 	 * [j_state_lock].
946 	 */
947 	unsigned long		j_fc_off;
948 
949 	/**
950 	 * @j_fc_last:
951 	 *
952 	 * The block number one beyond the last fast commit block in the journal
953 	 * [j_state_lock].
954 	 */
955 	unsigned long		j_fc_last;
956 
957 	/**
958 	 * @j_dev: Device where we store the journal.
959 	 */
960 	struct block_device	*j_dev;
961 
962 	/**
963 	 * @j_blocksize: Block size for the location where we store the journal.
964 	 */
965 	int			j_blocksize;
966 
967 	/**
968 	 * @j_blk_offset:
969 	 *
970 	 * Starting block offset into the device where we store the journal.
971 	 */
972 	unsigned long long	j_blk_offset;
973 
974 	/**
975 	 * @j_devname: Journal device name.
976 	 */
977 	char			j_devname[BDEVNAME_SIZE+24];
978 
979 	/**
980 	 * @j_fs_dev:
981 	 *
982 	 * Device which holds the client fs.  For internal journal this will be
983 	 * equal to j_dev.
984 	 */
985 	struct block_device	*j_fs_dev;
986 
987 	/**
988 	 * @j_maxlen: Total maximum capacity of the journal region on disk.
989 	 */
990 	unsigned int		j_maxlen;
991 
992 	/**
993 	 * @j_reserved_credits:
994 	 *
995 	 * Number of buffers reserved from the running transaction.
996 	 */
997 	atomic_t		j_reserved_credits;
998 
999 	/**
1000 	 * @j_list_lock: Protects the buffer lists and internal buffer state.
1001 	 */
1002 	spinlock_t		j_list_lock;
1003 
1004 	/**
1005 	 * @j_inode:
1006 	 *
1007 	 * Optional inode where we store the journal.  If present, all
1008 	 * journal block numbers are mapped into this inode via bmap().
1009 	 */
1010 	struct inode		*j_inode;
1011 
1012 	/**
1013 	 * @j_tail_sequence:
1014 	 *
1015 	 * Sequence number of the oldest transaction in the log [j_state_lock]
1016 	 */
1017 	tid_t			j_tail_sequence;
1018 
1019 	/**
1020 	 * @j_transaction_sequence:
1021 	 *
1022 	 * Sequence number of the next transaction to grant [j_state_lock]
1023 	 */
1024 	tid_t			j_transaction_sequence;
1025 
1026 	/**
1027 	 * @j_commit_sequence:
1028 	 *
1029 	 * Sequence number of the most recently committed transaction
1030 	 * [j_state_lock].
1031 	 */
1032 	tid_t			j_commit_sequence;
1033 
1034 	/**
1035 	 * @j_commit_request:
1036 	 *
1037 	 * Sequence number of the most recent transaction wanting commit
1038 	 * [j_state_lock]
1039 	 */
1040 	tid_t			j_commit_request;
1041 
1042 	/**
1043 	 * @j_uuid:
1044 	 *
1045 	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1046 	 * backed by this journal.  This will eventually be replaced by an array
1047 	 * of uuids, allowing us to index multiple devices within a single
1048 	 * journal and to perform atomic updates across them.
1049 	 */
1050 	__u8			j_uuid[16];
1051 
1052 	/**
1053 	 * @j_task: Pointer to the current commit thread for this journal.
1054 	 */
1055 	struct task_struct	*j_task;
1056 
1057 	/**
1058 	 * @j_max_transaction_buffers:
1059 	 *
1060 	 * Maximum number of metadata buffers to allow in a single compound
1061 	 * commit transaction.
1062 	 */
1063 	int			j_max_transaction_buffers;
1064 
1065 	/**
1066 	 * @j_revoke_records_per_block:
1067 	 *
1068 	 * Number of revoke records that fit in one descriptor block.
1069 	 */
1070 	int			j_revoke_records_per_block;
1071 
1072 	/**
1073 	 * @j_commit_interval:
1074 	 *
1075 	 * What is the maximum transaction lifetime before we begin a commit?
1076 	 */
1077 	unsigned long		j_commit_interval;
1078 
1079 	/**
1080 	 * @j_commit_timer: The timer used to wakeup the commit thread.
1081 	 */
1082 	struct timer_list	j_commit_timer;
1083 
1084 	/**
1085 	 * @j_revoke_lock: Protect the revoke table.
1086 	 */
1087 	spinlock_t		j_revoke_lock;
1088 
1089 	/**
1090 	 * @j_revoke:
1091 	 *
1092 	 * The revoke table - maintains the list of revoked blocks in the
1093 	 * current transaction.
1094 	 */
1095 	struct jbd2_revoke_table_s *j_revoke;
1096 
1097 	/**
1098 	 * @j_revoke_table: Alternate revoke tables for j_revoke.
1099 	 */
1100 	struct jbd2_revoke_table_s *j_revoke_table[2];
1101 
1102 	/**
1103 	 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1104 	 */
1105 	struct buffer_head	**j_wbuf;
1106 
1107 	/**
1108 	 * @j_fc_wbuf: Array of fast commit bhs for
1109 	 * jbd2_journal_commit_transaction.
1110 	 */
1111 	struct buffer_head	**j_fc_wbuf;
1112 
1113 	/**
1114 	 * @j_wbufsize:
1115 	 *
1116 	 * Size of @j_wbuf array.
1117 	 */
1118 	int			j_wbufsize;
1119 
1120 	/**
1121 	 * @j_fc_wbufsize:
1122 	 *
1123 	 * Size of @j_fc_wbuf array.
1124 	 */
1125 	int			j_fc_wbufsize;
1126 
1127 	/**
1128 	 * @j_last_sync_writer:
1129 	 *
1130 	 * The pid of the last person to run a synchronous operation
1131 	 * through the journal.
1132 	 */
1133 	pid_t			j_last_sync_writer;
1134 
1135 	/**
1136 	 * @j_average_commit_time:
1137 	 *
1138 	 * The average amount of time in nanoseconds it takes to commit a
1139 	 * transaction to disk. [j_state_lock]
1140 	 */
1141 	u64			j_average_commit_time;
1142 
1143 	/**
1144 	 * @j_min_batch_time:
1145 	 *
1146 	 * Minimum time that we should wait for additional filesystem operations
1147 	 * to get batched into a synchronous handle in microseconds.
1148 	 */
1149 	u32			j_min_batch_time;
1150 
1151 	/**
1152 	 * @j_max_batch_time:
1153 	 *
1154 	 * Maximum time that we should wait for additional filesystem operations
1155 	 * to get batched into a synchronous handle in microseconds.
1156 	 */
1157 	u32			j_max_batch_time;
1158 
1159 	/**
1160 	 * @j_commit_callback:
1161 	 *
1162 	 * This function is called when a transaction is closed.
1163 	 */
1164 	void			(*j_commit_callback)(journal_t *,
1165 						     transaction_t *);
1166 
1167 	/**
1168 	 * @j_submit_inode_data_buffers:
1169 	 *
1170 	 * This function is called for all inodes associated with the
1171 	 * committing transaction marked with JI_WRITE_DATA flag
1172 	 * before we start to write out the transaction to the journal.
1173 	 */
1174 	int			(*j_submit_inode_data_buffers)
1175 					(struct jbd2_inode *);
1176 
1177 	/**
1178 	 * @j_finish_inode_data_buffers:
1179 	 *
1180 	 * This function is called for all inodes associated with the
1181 	 * committing transaction marked with JI_WAIT_DATA flag
1182 	 * after we have written the transaction to the journal
1183 	 * but before we write out the commit block.
1184 	 */
1185 	int			(*j_finish_inode_data_buffers)
1186 					(struct jbd2_inode *);
1187 
1188 	/*
1189 	 * Journal statistics
1190 	 */
1191 
1192 	/**
1193 	 * @j_history_lock: Protect the transactions statistics history.
1194 	 */
1195 	spinlock_t		j_history_lock;
1196 
1197 	/**
1198 	 * @j_proc_entry: procfs entry for the jbd statistics directory.
1199 	 */
1200 	struct proc_dir_entry	*j_proc_entry;
1201 
1202 	/**
1203 	 * @j_stats: Overall statistics.
1204 	 */
1205 	struct transaction_stats_s j_stats;
1206 
1207 	/**
1208 	 * @j_failed_commit: Failed journal commit ID.
1209 	 */
1210 	unsigned int		j_failed_commit;
1211 
1212 	/**
1213 	 * @j_private:
1214 	 *
1215 	 * An opaque pointer to fs-private information.  ext3 puts its
1216 	 * superblock pointer here.
1217 	 */
1218 	void *j_private;
1219 
1220 	/**
1221 	 * @j_chksum_driver:
1222 	 *
1223 	 * Reference to checksum algorithm driver via cryptoapi.
1224 	 */
1225 	struct crypto_shash *j_chksum_driver;
1226 
1227 	/**
1228 	 * @j_csum_seed:
1229 	 *
1230 	 * Precomputed journal UUID checksum for seeding other checksums.
1231 	 */
1232 	__u32 j_csum_seed;
1233 
1234 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1235 	/**
1236 	 * @j_trans_commit_map:
1237 	 *
1238 	 * Lockdep entity to track transaction commit dependencies. Handles
1239 	 * hold this "lock" for read, when we wait for commit, we acquire the
1240 	 * "lock" for writing. This matches the properties of jbd2 journalling
1241 	 * where the running transaction has to wait for all handles to be
1242 	 * dropped to commit that transaction and also acquiring a handle may
1243 	 * require transaction commit to finish.
1244 	 */
1245 	struct lockdep_map	j_trans_commit_map;
1246 #endif
1247 
1248 	/**
1249 	 * @j_fc_cleanup_callback:
1250 	 *
1251 	 * Clean-up after fast commit or full commit. JBD2 calls this function
1252 	 * after every commit operation.
1253 	 */
1254 	void (*j_fc_cleanup_callback)(struct journal_s *journal, int);
1255 
1256 	/*
1257 	 * @j_fc_replay_callback:
1258 	 *
1259 	 * File-system specific function that performs replay of a fast
1260 	 * commit. JBD2 calls this function for each fast commit block found in
1261 	 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1262 	 * to indicate that the block was processed correctly and more fast
1263 	 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1264 	 * indicates the end of replay (no more blocks remaining). A negative
1265 	 * return value indicates error.
1266 	 */
1267 	int (*j_fc_replay_callback)(struct journal_s *journal,
1268 				    struct buffer_head *bh,
1269 				    enum passtype pass, int off,
1270 				    tid_t expected_commit_id);
1271 };
1272 
1273 #define jbd2_might_wait_for_commit(j) \
1274 	do { \
1275 		rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1276 		rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1277 	} while (0)
1278 
1279 /* journal feature predicate functions */
1280 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1281 static inline bool jbd2_has_feature_##name(journal_t *j) \
1282 { \
1283 	return ((j)->j_format_version >= 2 && \
1284 		((j)->j_superblock->s_feature_compat & \
1285 		 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1286 } \
1287 static inline void jbd2_set_feature_##name(journal_t *j) \
1288 { \
1289 	(j)->j_superblock->s_feature_compat |= \
1290 		cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1291 } \
1292 static inline void jbd2_clear_feature_##name(journal_t *j) \
1293 { \
1294 	(j)->j_superblock->s_feature_compat &= \
1295 		~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1296 }
1297 
1298 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1299 static inline bool jbd2_has_feature_##name(journal_t *j) \
1300 { \
1301 	return ((j)->j_format_version >= 2 && \
1302 		((j)->j_superblock->s_feature_ro_compat & \
1303 		 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1304 } \
1305 static inline void jbd2_set_feature_##name(journal_t *j) \
1306 { \
1307 	(j)->j_superblock->s_feature_ro_compat |= \
1308 		cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1309 } \
1310 static inline void jbd2_clear_feature_##name(journal_t *j) \
1311 { \
1312 	(j)->j_superblock->s_feature_ro_compat &= \
1313 		~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1314 }
1315 
1316 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1317 static inline bool jbd2_has_feature_##name(journal_t *j) \
1318 { \
1319 	return ((j)->j_format_version >= 2 && \
1320 		((j)->j_superblock->s_feature_incompat & \
1321 		 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1322 } \
1323 static inline void jbd2_set_feature_##name(journal_t *j) \
1324 { \
1325 	(j)->j_superblock->s_feature_incompat |= \
1326 		cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1327 } \
1328 static inline void jbd2_clear_feature_##name(journal_t *j) \
1329 { \
1330 	(j)->j_superblock->s_feature_incompat &= \
1331 		~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1332 }
1333 
1334 JBD2_FEATURE_COMPAT_FUNCS(checksum,		CHECKSUM)
1335 
1336 JBD2_FEATURE_INCOMPAT_FUNCS(revoke,		REVOKE)
1337 JBD2_FEATURE_INCOMPAT_FUNCS(64bit,		64BIT)
1338 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit,	ASYNC_COMMIT)
1339 JBD2_FEATURE_INCOMPAT_FUNCS(csum2,		CSUM_V2)
1340 JBD2_FEATURE_INCOMPAT_FUNCS(csum3,		CSUM_V3)
1341 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit,	FAST_COMMIT)
1342 
1343 /*
1344  * Journal flag definitions
1345  */
1346 #define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1347 #define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1348 #define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1349 #define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1350 #define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1351 #define JBD2_BARRIER	0x020	/* Use IDE barriers */
1352 #define JBD2_ABORT_ON_SYNCDATA_ERR	0x040	/* Abort the journal on file
1353 						 * data write error in ordered
1354 						 * mode */
1355 #define JBD2_FAST_COMMIT_ONGOING	0x100	/* Fast commit is ongoing */
1356 #define JBD2_FULL_COMMIT_ONGOING	0x200	/* Full commit is ongoing */
1357 
1358 /*
1359  * Function declarations for the journaling transaction and buffer
1360  * management
1361  */
1362 
1363 /* Filing buffers */
1364 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1365 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1366 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1367 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1368 extern void __journal_free_buffer(struct journal_head *bh);
1369 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1370 extern void __journal_clean_data_list(transaction_t *transaction);
1371 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1372 {
1373 	list_add_tail(&bh->b_assoc_buffers, head);
1374 }
1375 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1376 {
1377 	list_del_init(&bh->b_assoc_buffers);
1378 }
1379 
1380 /* Log buffer allocation */
1381 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1382 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1383 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1384 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1385 			      unsigned long *block);
1386 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1387 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1388 
1389 /* Commit management */
1390 extern void jbd2_journal_commit_transaction(journal_t *);
1391 
1392 /* Checkpoint list management */
1393 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1394 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1395 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1396 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1397 
1398 
1399 /*
1400  * Triggers
1401  */
1402 
1403 struct jbd2_buffer_trigger_type {
1404 	/*
1405 	 * Fired a the moment data to write to the journal are known to be
1406 	 * stable - so either at the moment b_frozen_data is created or just
1407 	 * before a buffer is written to the journal.  mapped_data is a mapped
1408 	 * buffer that is the frozen data for commit.
1409 	 */
1410 	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1411 			 struct buffer_head *bh, void *mapped_data,
1412 			 size_t size);
1413 
1414 	/*
1415 	 * Fired during journal abort for dirty buffers that will not be
1416 	 * committed.
1417 	 */
1418 	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1419 			struct buffer_head *bh);
1420 };
1421 
1422 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1423 				       void *mapped_data,
1424 				       struct jbd2_buffer_trigger_type *triggers);
1425 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1426 				      struct jbd2_buffer_trigger_type *triggers);
1427 
1428 /* Buffer IO */
1429 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1430 					      struct journal_head *jh_in,
1431 					      struct buffer_head **bh_out,
1432 					      sector_t blocknr);
1433 
1434 /* Transaction locking */
1435 extern void		__wait_on_journal (journal_t *);
1436 
1437 /* Transaction cache support */
1438 extern void jbd2_journal_destroy_transaction_cache(void);
1439 extern int __init jbd2_journal_init_transaction_cache(void);
1440 extern void jbd2_journal_free_transaction(transaction_t *);
1441 
1442 /*
1443  * Journal locking.
1444  *
1445  * We need to lock the journal during transaction state changes so that nobody
1446  * ever tries to take a handle on the running transaction while we are in the
1447  * middle of moving it to the commit phase.  j_state_lock does this.
1448  *
1449  * Note that the locking is completely interrupt unsafe.  We never touch
1450  * journal structures from interrupts.
1451  */
1452 
1453 static inline handle_t *journal_current_handle(void)
1454 {
1455 	return current->journal_info;
1456 }
1457 
1458 /* The journaling code user interface:
1459  *
1460  * Create and destroy handles
1461  * Register buffer modifications against the current transaction.
1462  */
1463 
1464 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1465 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1466 				     int revoke_records, gfp_t gfp_mask,
1467 				     unsigned int type, unsigned int line_no);
1468 extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1469 extern int	 jbd2__journal_restart(handle_t *, int nblocks,
1470 				       int revoke_records, gfp_t gfp_mask);
1471 extern int	 jbd2_journal_start_reserved(handle_t *handle,
1472 				unsigned int type, unsigned int line_no);
1473 extern void	 jbd2_journal_free_reserved(handle_t *handle);
1474 extern int	 jbd2_journal_extend(handle_t *handle, int nblocks,
1475 				     int revoke_records);
1476 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1477 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1478 extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1479 void		 jbd2_journal_set_triggers(struct buffer_head *,
1480 					   struct jbd2_buffer_trigger_type *type);
1481 extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1482 extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1483 extern int	 jbd2_journal_invalidatepage(journal_t *,
1484 				struct page *, unsigned int, unsigned int);
1485 extern int	 jbd2_journal_try_to_free_buffers(journal_t *journal, struct page *page);
1486 extern int	 jbd2_journal_stop(handle_t *);
1487 extern int	 jbd2_journal_flush (journal_t *);
1488 extern void	 jbd2_journal_lock_updates (journal_t *);
1489 extern void	 jbd2_journal_unlock_updates (journal_t *);
1490 
1491 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1492 				struct block_device *fs_dev,
1493 				unsigned long long start, int len, int bsize);
1494 extern journal_t * jbd2_journal_init_inode (struct inode *);
1495 extern int	   jbd2_journal_update_format (journal_t *);
1496 extern int	   jbd2_journal_check_used_features
1497 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1498 extern int	   jbd2_journal_check_available_features
1499 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1500 extern int	   jbd2_journal_set_features
1501 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1502 extern void	   jbd2_journal_clear_features
1503 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1504 extern int	   jbd2_journal_load       (journal_t *journal);
1505 extern int	   jbd2_journal_destroy    (journal_t *);
1506 extern int	   jbd2_journal_recover    (journal_t *journal);
1507 extern int	   jbd2_journal_wipe       (journal_t *, int);
1508 extern int	   jbd2_journal_skip_recovery	(journal_t *);
1509 extern void	   jbd2_journal_update_sb_errno(journal_t *);
1510 extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1511 				unsigned long, int);
1512 extern void	   jbd2_journal_abort      (journal_t *, int);
1513 extern int	   jbd2_journal_errno      (journal_t *);
1514 extern void	   jbd2_journal_ack_err    (journal_t *);
1515 extern int	   jbd2_journal_clear_err  (journal_t *);
1516 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1517 extern int	   jbd2_journal_force_commit(journal_t *);
1518 extern int	   jbd2_journal_force_commit_nested(journal_t *);
1519 extern int	   jbd2_journal_inode_ranged_write(handle_t *handle,
1520 			struct jbd2_inode *inode, loff_t start_byte,
1521 			loff_t length);
1522 extern int	   jbd2_journal_inode_ranged_wait(handle_t *handle,
1523 			struct jbd2_inode *inode, loff_t start_byte,
1524 			loff_t length);
1525 extern int	   jbd2_journal_submit_inode_data_buffers(
1526 			struct jbd2_inode *jinode);
1527 extern int	   jbd2_journal_finish_inode_data_buffers(
1528 			struct jbd2_inode *jinode);
1529 extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1530 				struct jbd2_inode *inode, loff_t new_size);
1531 extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1532 extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1533 
1534 /*
1535  * journal_head management
1536  */
1537 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1538 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1539 void jbd2_journal_put_journal_head(struct journal_head *jh);
1540 
1541 /*
1542  * handle management
1543  */
1544 extern struct kmem_cache *jbd2_handle_cache;
1545 
1546 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1547 {
1548 	return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1549 }
1550 
1551 static inline void jbd2_free_handle(handle_t *handle)
1552 {
1553 	kmem_cache_free(jbd2_handle_cache, handle);
1554 }
1555 
1556 /*
1557  * jbd2_inode management (optional, for those file systems that want to use
1558  * dynamically allocated jbd2_inode structures)
1559  */
1560 extern struct kmem_cache *jbd2_inode_cache;
1561 
1562 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1563 {
1564 	return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1565 }
1566 
1567 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1568 {
1569 	kmem_cache_free(jbd2_inode_cache, jinode);
1570 }
1571 
1572 /* Primary revoke support */
1573 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1574 extern int	   jbd2_journal_init_revoke(journal_t *, int);
1575 extern void	   jbd2_journal_destroy_revoke_record_cache(void);
1576 extern void	   jbd2_journal_destroy_revoke_table_cache(void);
1577 extern int __init jbd2_journal_init_revoke_record_cache(void);
1578 extern int __init jbd2_journal_init_revoke_table_cache(void);
1579 
1580 extern void	   jbd2_journal_destroy_revoke(journal_t *);
1581 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1582 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1583 extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
1584 						     struct list_head *log_bufs);
1585 
1586 /* Recovery revoke support */
1587 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1588 extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1589 extern void	jbd2_journal_clear_revoke(journal_t *);
1590 extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1591 extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1592 
1593 /*
1594  * The log thread user interface:
1595  *
1596  * Request space in the current transaction, and force transaction commit
1597  * transitions on demand.
1598  */
1599 
1600 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1601 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1602 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1603 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1604 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1605 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1606 int jbd2_log_do_checkpoint(journal_t *journal);
1607 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1608 
1609 void __jbd2_log_wait_for_space(journal_t *journal);
1610 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1611 extern int jbd2_cleanup_journal_tail(journal_t *);
1612 
1613 /* Fast commit related APIs */
1614 int jbd2_fc_init(journal_t *journal, int num_fc_blks);
1615 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1616 int jbd2_fc_end_commit(journal_t *journal);
1617 int jbd2_fc_end_commit_fallback(journal_t *journal, tid_t tid);
1618 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1619 int jbd2_submit_inode_data(struct jbd2_inode *jinode);
1620 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1621 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1622 int jbd2_fc_release_bufs(journal_t *journal);
1623 
1624 /*
1625  * is_journal_abort
1626  *
1627  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1628  * bit, when set, indicates that we have had a fatal error somewhere,
1629  * either inside the journaling layer or indicated to us by the client
1630  * (eg. ext3), and that we and should not commit any further
1631  * transactions.
1632  */
1633 
1634 static inline int is_journal_aborted(journal_t *journal)
1635 {
1636 	return journal->j_flags & JBD2_ABORT;
1637 }
1638 
1639 static inline int is_handle_aborted(handle_t *handle)
1640 {
1641 	if (handle->h_aborted || !handle->h_transaction)
1642 		return 1;
1643 	return is_journal_aborted(handle->h_transaction->t_journal);
1644 }
1645 
1646 static inline void jbd2_journal_abort_handle(handle_t *handle)
1647 {
1648 	handle->h_aborted = 1;
1649 }
1650 
1651 #endif /* __KERNEL__   */
1652 
1653 /* Comparison functions for transaction IDs: perform comparisons using
1654  * modulo arithmetic so that they work over sequence number wraps. */
1655 
1656 static inline int tid_gt(tid_t x, tid_t y)
1657 {
1658 	int difference = (x - y);
1659 	return (difference > 0);
1660 }
1661 
1662 static inline int tid_geq(tid_t x, tid_t y)
1663 {
1664 	int difference = (x - y);
1665 	return (difference >= 0);
1666 }
1667 
1668 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1669 extern size_t journal_tag_bytes(journal_t *journal);
1670 
1671 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1672 {
1673 	return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1674 }
1675 
1676 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1677 {
1678 	WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1679 		     journal->j_chksum_driver == NULL);
1680 
1681 	return journal->j_chksum_driver != NULL;
1682 }
1683 
1684 /*
1685  * Return number of free blocks in the log. Must be called under j_state_lock.
1686  */
1687 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1688 {
1689 	/* Allow for rounding errors */
1690 	long free = journal->j_free - 32;
1691 
1692 	if (journal->j_committing_transaction) {
1693 		free -= atomic_read(&journal->
1694                         j_committing_transaction->t_outstanding_credits);
1695 	}
1696 	return max_t(long, free, 0);
1697 }
1698 
1699 /*
1700  * Definitions which augment the buffer_head layer
1701  */
1702 
1703 /* journaling buffer types */
1704 #define BJ_None		0	/* Not journaled */
1705 #define BJ_Metadata	1	/* Normal journaled metadata */
1706 #define BJ_Forget	2	/* Buffer superseded by this transaction */
1707 #define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1708 #define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1709 #define BJ_Types	5
1710 
1711 extern int jbd_blocks_per_page(struct inode *inode);
1712 
1713 /* JBD uses a CRC32 checksum */
1714 #define JBD_MAX_CHECKSUM_SIZE 4
1715 
1716 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1717 			      const void *address, unsigned int length)
1718 {
1719 	struct {
1720 		struct shash_desc shash;
1721 		char ctx[JBD_MAX_CHECKSUM_SIZE];
1722 	} desc;
1723 	int err;
1724 
1725 	BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1726 		JBD_MAX_CHECKSUM_SIZE);
1727 
1728 	desc.shash.tfm = journal->j_chksum_driver;
1729 	*(u32 *)desc.ctx = crc;
1730 
1731 	err = crypto_shash_update(&desc.shash, address, length);
1732 	BUG_ON(err);
1733 
1734 	return *(u32 *)desc.ctx;
1735 }
1736 
1737 /* Return most recent uncommitted transaction */
1738 static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1739 {
1740 	tid_t tid;
1741 
1742 	read_lock(&journal->j_state_lock);
1743 	tid = journal->j_commit_request;
1744 	if (journal->j_running_transaction)
1745 		tid = journal->j_running_transaction->t_tid;
1746 	read_unlock(&journal->j_state_lock);
1747 	return tid;
1748 }
1749 
1750 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1751 {
1752 	journal_t *journal;
1753 
1754 	if (!handle->h_reserved)
1755 		journal = handle->h_transaction->t_journal;
1756 	else
1757 		journal = handle->h_journal;
1758 
1759 	return handle->h_total_credits -
1760 		DIV_ROUND_UP(handle->h_revoke_credits_requested,
1761 			     journal->j_revoke_records_per_block);
1762 }
1763 
1764 #ifdef __KERNEL__
1765 
1766 #define buffer_trace_init(bh)	do {} while (0)
1767 #define print_buffer_fields(bh)	do {} while (0)
1768 #define print_buffer_trace(bh)	do {} while (0)
1769 #define BUFFER_TRACE(bh, info)	do {} while (0)
1770 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1771 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1772 
1773 #endif	/* __KERNEL__ */
1774 
1775 #define EFSBADCRC	EBADMSG		/* Bad CRC detected */
1776 #define EFSCORRUPTED	EUCLEAN		/* Filesystem is corrupted */
1777 
1778 #endif	/* _LINUX_JBD2_H */
1779