xref: /linux/include/linux/jbd2.h (revision e58e871becec2d3b04ed91c0c16fe8deac9c9dfa)
1 /*
2  * linux/include/linux/jbd2.h
3  *
4  * Written by Stephen C. Tweedie <sct@redhat.com>
5  *
6  * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
7  *
8  * This file is part of the Linux kernel and is made available under
9  * the terms of the GNU General Public License, version 2, or at your
10  * option, any later version, incorporated herein by reference.
11  *
12  * Definitions for transaction data structures for the buffer cache
13  * filesystem journaling support.
14  */
15 
16 #ifndef _LINUX_JBD2_H
17 #define _LINUX_JBD2_H
18 
19 /* Allow this file to be included directly into e2fsprogs */
20 #ifndef __KERNEL__
21 #include "jfs_compat.h"
22 #define JBD2_DEBUG
23 #else
24 
25 #include <linux/types.h>
26 #include <linux/buffer_head.h>
27 #include <linux/journal-head.h>
28 #include <linux/stddef.h>
29 #include <linux/mutex.h>
30 #include <linux/timer.h>
31 #include <linux/slab.h>
32 #include <linux/bit_spinlock.h>
33 #include <crypto/hash.h>
34 #endif
35 
36 #define journal_oom_retry 1
37 
38 /*
39  * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
40  * certain classes of error which can occur due to failed IOs.  Under
41  * normal use we want ext4 to continue after such errors, because
42  * hardware _can_ fail, but for debugging purposes when running tests on
43  * known-good hardware we may want to trap these errors.
44  */
45 #undef JBD2_PARANOID_IOFAIL
46 
47 /*
48  * The default maximum commit age, in seconds.
49  */
50 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
51 
52 #ifdef CONFIG_JBD2_DEBUG
53 /*
54  * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
55  * consistency checks.  By default we don't do this unless
56  * CONFIG_JBD2_DEBUG is on.
57  */
58 #define JBD2_EXPENSIVE_CHECKING
59 extern ushort jbd2_journal_enable_debug;
60 void __jbd2_debug(int level, const char *file, const char *func,
61 		  unsigned int line, const char *fmt, ...);
62 
63 #define jbd_debug(n, fmt, a...) \
64 	__jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
65 #else
66 #define jbd_debug(n, fmt, a...)    /**/
67 #endif
68 
69 extern void *jbd2_alloc(size_t size, gfp_t flags);
70 extern void jbd2_free(void *ptr, size_t size);
71 
72 #define JBD2_MIN_JOURNAL_BLOCKS 1024
73 
74 #ifdef __KERNEL__
75 
76 /**
77  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
78  *
79  * All filesystem modifications made by the process go
80  * through this handle.  Recursive operations (such as quota operations)
81  * are gathered into a single update.
82  *
83  * The buffer credits field is used to account for journaled buffers
84  * being modified by the running process.  To ensure that there is
85  * enough log space for all outstanding operations, we need to limit the
86  * number of outstanding buffers possible at any time.  When the
87  * operation completes, any buffer credits not used are credited back to
88  * the transaction, so that at all times we know how many buffers the
89  * outstanding updates on a transaction might possibly touch.
90  *
91  * This is an opaque datatype.
92  **/
93 typedef struct jbd2_journal_handle handle_t;	/* Atomic operation type */
94 
95 
96 /**
97  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
98  *
99  * journal_t is linked to from the fs superblock structure.
100  *
101  * We use the journal_t to keep track of all outstanding transaction
102  * activity on the filesystem, and to manage the state of the log
103  * writing process.
104  *
105  * This is an opaque datatype.
106  **/
107 typedef struct journal_s	journal_t;	/* Journal control structure */
108 #endif
109 
110 /*
111  * Internal structures used by the logging mechanism:
112  */
113 
114 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
115 
116 /*
117  * On-disk structures
118  */
119 
120 /*
121  * Descriptor block types:
122  */
123 
124 #define JBD2_DESCRIPTOR_BLOCK	1
125 #define JBD2_COMMIT_BLOCK	2
126 #define JBD2_SUPERBLOCK_V1	3
127 #define JBD2_SUPERBLOCK_V2	4
128 #define JBD2_REVOKE_BLOCK	5
129 
130 /*
131  * Standard header for all descriptor blocks:
132  */
133 typedef struct journal_header_s
134 {
135 	__be32		h_magic;
136 	__be32		h_blocktype;
137 	__be32		h_sequence;
138 } journal_header_t;
139 
140 /*
141  * Checksum types.
142  */
143 #define JBD2_CRC32_CHKSUM   1
144 #define JBD2_MD5_CHKSUM     2
145 #define JBD2_SHA1_CHKSUM    3
146 #define JBD2_CRC32C_CHKSUM  4
147 
148 #define JBD2_CRC32_CHKSUM_SIZE 4
149 
150 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
151 /*
152  * Commit block header for storing transactional checksums:
153  *
154  * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
155  * fields are used to store a checksum of the descriptor and data blocks.
156  *
157  * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
158  * field is used to store crc32c(uuid+commit_block).  Each journal metadata
159  * block gets its own checksum, and data block checksums are stored in
160  * journal_block_tag (in the descriptor).  The other h_chksum* fields are
161  * not used.
162  *
163  * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
164  * journal_block_tag3_t to store a full 32-bit checksum.  Everything else
165  * is the same as v2.
166  *
167  * Checksum v1, v2, and v3 are mutually exclusive features.
168  */
169 struct commit_header {
170 	__be32		h_magic;
171 	__be32          h_blocktype;
172 	__be32          h_sequence;
173 	unsigned char   h_chksum_type;
174 	unsigned char   h_chksum_size;
175 	unsigned char 	h_padding[2];
176 	__be32 		h_chksum[JBD2_CHECKSUM_BYTES];
177 	__be64		h_commit_sec;
178 	__be32		h_commit_nsec;
179 };
180 
181 /*
182  * The block tag: used to describe a single buffer in the journal.
183  * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
184  * raw struct shouldn't be used for pointer math or sizeof() - use
185  * journal_tag_bytes(journal) instead to compute this.
186  */
187 typedef struct journal_block_tag3_s
188 {
189 	__be32		t_blocknr;	/* The on-disk block number */
190 	__be32		t_flags;	/* See below */
191 	__be32		t_blocknr_high; /* most-significant high 32bits. */
192 	__be32		t_checksum;	/* crc32c(uuid+seq+block) */
193 } journal_block_tag3_t;
194 
195 typedef struct journal_block_tag_s
196 {
197 	__be32		t_blocknr;	/* The on-disk block number */
198 	__be16		t_checksum;	/* truncated crc32c(uuid+seq+block) */
199 	__be16		t_flags;	/* See below */
200 	__be32		t_blocknr_high; /* most-significant high 32bits. */
201 } journal_block_tag_t;
202 
203 /* Tail of descriptor or revoke block, for checksumming */
204 struct jbd2_journal_block_tail {
205 	__be32		t_checksum;	/* crc32c(uuid+descr_block) */
206 };
207 
208 /*
209  * The revoke descriptor: used on disk to describe a series of blocks to
210  * be revoked from the log
211  */
212 typedef struct jbd2_journal_revoke_header_s
213 {
214 	journal_header_t r_header;
215 	__be32		 r_count;	/* Count of bytes used in the block */
216 } jbd2_journal_revoke_header_t;
217 
218 /* Definitions for the journal tag flags word: */
219 #define JBD2_FLAG_ESCAPE		1	/* on-disk block is escaped */
220 #define JBD2_FLAG_SAME_UUID	2	/* block has same uuid as previous */
221 #define JBD2_FLAG_DELETED	4	/* block deleted by this transaction */
222 #define JBD2_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
223 
224 
225 /*
226  * The journal superblock.  All fields are in big-endian byte order.
227  */
228 typedef struct journal_superblock_s
229 {
230 /* 0x0000 */
231 	journal_header_t s_header;
232 
233 /* 0x000C */
234 	/* Static information describing the journal */
235 	__be32	s_blocksize;		/* journal device blocksize */
236 	__be32	s_maxlen;		/* total blocks in journal file */
237 	__be32	s_first;		/* first block of log information */
238 
239 /* 0x0018 */
240 	/* Dynamic information describing the current state of the log */
241 	__be32	s_sequence;		/* first commit ID expected in log */
242 	__be32	s_start;		/* blocknr of start of log */
243 
244 /* 0x0020 */
245 	/* Error value, as set by jbd2_journal_abort(). */
246 	__be32	s_errno;
247 
248 /* 0x0024 */
249 	/* Remaining fields are only valid in a version-2 superblock */
250 	__be32	s_feature_compat;	/* compatible feature set */
251 	__be32	s_feature_incompat;	/* incompatible feature set */
252 	__be32	s_feature_ro_compat;	/* readonly-compatible feature set */
253 /* 0x0030 */
254 	__u8	s_uuid[16];		/* 128-bit uuid for journal */
255 
256 /* 0x0040 */
257 	__be32	s_nr_users;		/* Nr of filesystems sharing log */
258 
259 	__be32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
260 
261 /* 0x0048 */
262 	__be32	s_max_transaction;	/* Limit of journal blocks per trans.*/
263 	__be32	s_max_trans_data;	/* Limit of data blocks per trans. */
264 
265 /* 0x0050 */
266 	__u8	s_checksum_type;	/* checksum type */
267 	__u8	s_padding2[3];
268 	__u32	s_padding[42];
269 	__be32	s_checksum;		/* crc32c(superblock) */
270 
271 /* 0x0100 */
272 	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
273 /* 0x0400 */
274 } journal_superblock_t;
275 
276 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
277 #define JBD2_HAS_COMPAT_FEATURE(j,mask)					\
278 	((j)->j_format_version >= 2 &&					\
279 	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
280 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask)				\
281 	((j)->j_format_version >= 2 &&					\
282 	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
283 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask)				\
284 	((j)->j_format_version >= 2 &&					\
285 	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
286 
287 #define JBD2_FEATURE_COMPAT_CHECKSUM		0x00000001
288 
289 #define JBD2_FEATURE_INCOMPAT_REVOKE		0x00000001
290 #define JBD2_FEATURE_INCOMPAT_64BIT		0x00000002
291 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
292 #define JBD2_FEATURE_INCOMPAT_CSUM_V2		0x00000008
293 #define JBD2_FEATURE_INCOMPAT_CSUM_V3		0x00000010
294 
295 /* See "journal feature predicate functions" below */
296 
297 /* Features known to this kernel version: */
298 #define JBD2_KNOWN_COMPAT_FEATURES	JBD2_FEATURE_COMPAT_CHECKSUM
299 #define JBD2_KNOWN_ROCOMPAT_FEATURES	0
300 #define JBD2_KNOWN_INCOMPAT_FEATURES	(JBD2_FEATURE_INCOMPAT_REVOKE | \
301 					JBD2_FEATURE_INCOMPAT_64BIT | \
302 					JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
303 					JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
304 					JBD2_FEATURE_INCOMPAT_CSUM_V3)
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_State,		/* Pins most journal_head state */
320 	BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
321 	BH_Shadow,		/* IO on shadow buffer is running */
322 	BH_Verified,		/* Metadata block has been verified ok */
323 	BH_JBDPrivateStart,	/* First bit available for private use by FS */
324 };
325 
326 BUFFER_FNS(JBD, jbd)
327 BUFFER_FNS(JWrite, jwrite)
328 BUFFER_FNS(JBDDirty, jbddirty)
329 TAS_BUFFER_FNS(JBDDirty, jbddirty)
330 BUFFER_FNS(Revoked, revoked)
331 TAS_BUFFER_FNS(Revoked, revoked)
332 BUFFER_FNS(RevokeValid, revokevalid)
333 TAS_BUFFER_FNS(RevokeValid, revokevalid)
334 BUFFER_FNS(Freed, freed)
335 BUFFER_FNS(Shadow, shadow)
336 BUFFER_FNS(Verified, verified)
337 
338 static inline struct buffer_head *jh2bh(struct journal_head *jh)
339 {
340 	return jh->b_bh;
341 }
342 
343 static inline struct journal_head *bh2jh(struct buffer_head *bh)
344 {
345 	return bh->b_private;
346 }
347 
348 static inline void jbd_lock_bh_state(struct buffer_head *bh)
349 {
350 	bit_spin_lock(BH_State, &bh->b_state);
351 }
352 
353 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
354 {
355 	return bit_spin_trylock(BH_State, &bh->b_state);
356 }
357 
358 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
359 {
360 	return bit_spin_is_locked(BH_State, &bh->b_state);
361 }
362 
363 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
364 {
365 	bit_spin_unlock(BH_State, &bh->b_state);
366 }
367 
368 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
369 {
370 	bit_spin_lock(BH_JournalHead, &bh->b_state);
371 }
372 
373 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
374 {
375 	bit_spin_unlock(BH_JournalHead, &bh->b_state);
376 }
377 
378 #define J_ASSERT(assert)	BUG_ON(!(assert))
379 
380 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
381 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
382 
383 #if defined(JBD2_PARANOID_IOFAIL)
384 #define J_EXPECT(expr, why...)		J_ASSERT(expr)
385 #define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
386 #define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
387 #else
388 #define __journal_expect(expr, why...)					     \
389 	({								     \
390 		int val = (expr);					     \
391 		if (!val) {						     \
392 			printk(KERN_ERR					     \
393 			       "JBD2 unexpected failure: %s: %s;\n",	     \
394 			       __func__, #expr);			     \
395 			printk(KERN_ERR why "\n");			     \
396 		}							     \
397 		val;							     \
398 	})
399 #define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
400 #define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
401 #define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
402 #endif
403 
404 /* Flags in jbd_inode->i_flags */
405 #define __JI_COMMIT_RUNNING 0
406 #define __JI_WRITE_DATA 1
407 #define __JI_WAIT_DATA 2
408 
409 /*
410  * Commit of the inode data in progress. We use this flag to protect us from
411  * concurrent deletion of inode. We cannot use reference to inode for this
412  * since we cannot afford doing last iput() on behalf of kjournald
413  */
414 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
415 /* Write allocated dirty buffers in this inode before commit */
416 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
417 /* Wait for outstanding data writes for this inode before commit */
418 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
419 
420 /**
421  * struct jbd_inode is the structure linking inodes in ordered mode
422  *   present in a transaction so that we can sync them during commit.
423  */
424 struct jbd2_inode {
425 	/* Which transaction does this inode belong to? Either the running
426 	 * transaction or the committing one. [j_list_lock] */
427 	transaction_t *i_transaction;
428 
429 	/* Pointer to the running transaction modifying inode's data in case
430 	 * there is already a committing transaction touching it. [j_list_lock] */
431 	transaction_t *i_next_transaction;
432 
433 	/* List of inodes in the i_transaction [j_list_lock] */
434 	struct list_head i_list;
435 
436 	/* VFS inode this inode belongs to [constant during the lifetime
437 	 * of the structure] */
438 	struct inode *i_vfs_inode;
439 
440 	/* Flags of inode [j_list_lock] */
441 	unsigned long i_flags;
442 };
443 
444 struct jbd2_revoke_table_s;
445 
446 /**
447  * struct handle_s - The handle_s type is the concrete type associated with
448  *     handle_t.
449  * @h_transaction: Which compound transaction is this update a part of?
450  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
451  * @h_ref: Reference count on this handle
452  * @h_err: Field for caller's use to track errors through large fs operations
453  * @h_sync: flag for sync-on-close
454  * @h_jdata: flag to force data journaling
455  * @h_aborted: flag indicating fatal error on handle
456  **/
457 
458 /* Docbook can't yet cope with the bit fields, but will leave the documentation
459  * in so it can be fixed later.
460  */
461 
462 struct jbd2_journal_handle
463 {
464 	union {
465 		/* Which compound transaction is this update a part of? */
466 		transaction_t	*h_transaction;
467 		/* Which journal handle belongs to - used iff h_reserved set */
468 		journal_t	*h_journal;
469 	};
470 
471 	/* Handle reserved for finishing the logical operation */
472 	handle_t		*h_rsv_handle;
473 
474 	/* Number of remaining buffers we are allowed to dirty: */
475 	int			h_buffer_credits;
476 
477 	/* Reference count on this handle */
478 	int			h_ref;
479 
480 	/* Field for caller's use to track errors through large fs */
481 	/* operations */
482 	int			h_err;
483 
484 	/* Flags [no locking] */
485 	unsigned int	h_sync:		1;	/* sync-on-close */
486 	unsigned int	h_jdata:	1;	/* force data journaling */
487 	unsigned int	h_reserved:	1;	/* handle with reserved credits */
488 	unsigned int	h_aborted:	1;	/* fatal error on handle */
489 	unsigned int	h_type:		8;	/* for handle statistics */
490 	unsigned int	h_line_no:	16;	/* for handle statistics */
491 
492 	unsigned long		h_start_jiffies;
493 	unsigned int		h_requested_credits;
494 
495 	unsigned int		saved_alloc_context;
496 };
497 
498 
499 /*
500  * Some stats for checkpoint phase
501  */
502 struct transaction_chp_stats_s {
503 	unsigned long		cs_chp_time;
504 	__u32			cs_forced_to_close;
505 	__u32			cs_written;
506 	__u32			cs_dropped;
507 };
508 
509 /* The transaction_t type is the guts of the journaling mechanism.  It
510  * tracks a compound transaction through its various states:
511  *
512  * RUNNING:	accepting new updates
513  * LOCKED:	Updates still running but we don't accept new ones
514  * RUNDOWN:	Updates are tidying up but have finished requesting
515  *		new buffers to modify (state not used for now)
516  * FLUSH:       All updates complete, but we are still writing to disk
517  * COMMIT:      All data on disk, writing commit record
518  * FINISHED:	We still have to keep the transaction for checkpointing.
519  *
520  * The transaction keeps track of all of the buffers modified by a
521  * running transaction, and all of the buffers committed but not yet
522  * flushed to home for finished transactions.
523  */
524 
525 /*
526  * Lock ranking:
527  *
528  *    j_list_lock
529  *      ->jbd_lock_bh_journal_head()	(This is "innermost")
530  *
531  *    j_state_lock
532  *    ->jbd_lock_bh_state()
533  *
534  *    jbd_lock_bh_state()
535  *    ->j_list_lock
536  *
537  *    j_state_lock
538  *    ->t_handle_lock
539  *
540  *    j_state_lock
541  *    ->j_list_lock			(journal_unmap_buffer)
542  *
543  */
544 
545 struct transaction_s
546 {
547 	/* Pointer to the journal for this transaction. [no locking] */
548 	journal_t		*t_journal;
549 
550 	/* Sequence number for this transaction [no locking] */
551 	tid_t			t_tid;
552 
553 	/*
554 	 * Transaction's current state
555 	 * [no locking - only kjournald2 alters this]
556 	 * [j_list_lock] guards transition of a transaction into T_FINISHED
557 	 * state and subsequent call of __jbd2_journal_drop_transaction()
558 	 * FIXME: needs barriers
559 	 * KLUDGE: [use j_state_lock]
560 	 */
561 	enum {
562 		T_RUNNING,
563 		T_LOCKED,
564 		T_FLUSH,
565 		T_COMMIT,
566 		T_COMMIT_DFLUSH,
567 		T_COMMIT_JFLUSH,
568 		T_COMMIT_CALLBACK,
569 		T_FINISHED
570 	}			t_state;
571 
572 	/*
573 	 * Where in the log does this transaction's commit start? [no locking]
574 	 */
575 	unsigned long		t_log_start;
576 
577 	/* Number of buffers on the t_buffers list [j_list_lock] */
578 	int			t_nr_buffers;
579 
580 	/*
581 	 * Doubly-linked circular list of all buffers reserved but not yet
582 	 * modified by this transaction [j_list_lock]
583 	 */
584 	struct journal_head	*t_reserved_list;
585 
586 	/*
587 	 * Doubly-linked circular list of all metadata buffers owned by this
588 	 * transaction [j_list_lock]
589 	 */
590 	struct journal_head	*t_buffers;
591 
592 	/*
593 	 * Doubly-linked circular list of all forget buffers (superseded
594 	 * buffers which we can un-checkpoint once this transaction commits)
595 	 * [j_list_lock]
596 	 */
597 	struct journal_head	*t_forget;
598 
599 	/*
600 	 * Doubly-linked circular list of all buffers still to be flushed before
601 	 * this transaction can be checkpointed. [j_list_lock]
602 	 */
603 	struct journal_head	*t_checkpoint_list;
604 
605 	/*
606 	 * Doubly-linked circular list of all buffers submitted for IO while
607 	 * checkpointing. [j_list_lock]
608 	 */
609 	struct journal_head	*t_checkpoint_io_list;
610 
611 	/*
612 	 * Doubly-linked circular list of metadata buffers being shadowed by log
613 	 * IO.  The IO buffers on the iobuf list and the shadow buffers on this
614 	 * list match each other one for one at all times. [j_list_lock]
615 	 */
616 	struct journal_head	*t_shadow_list;
617 
618 	/*
619 	 * List of inodes whose data we've modified in data=ordered mode.
620 	 * [j_list_lock]
621 	 */
622 	struct list_head	t_inode_list;
623 
624 	/*
625 	 * Protects info related to handles
626 	 */
627 	spinlock_t		t_handle_lock;
628 
629 	/*
630 	 * Longest time some handle had to wait for running transaction
631 	 */
632 	unsigned long		t_max_wait;
633 
634 	/*
635 	 * When transaction started
636 	 */
637 	unsigned long		t_start;
638 
639 	/*
640 	 * When commit was requested
641 	 */
642 	unsigned long		t_requested;
643 
644 	/*
645 	 * Checkpointing stats [j_checkpoint_sem]
646 	 */
647 	struct transaction_chp_stats_s t_chp_stats;
648 
649 	/*
650 	 * Number of outstanding updates running on this transaction
651 	 * [t_handle_lock]
652 	 */
653 	atomic_t		t_updates;
654 
655 	/*
656 	 * Number of buffers reserved for use by all handles in this transaction
657 	 * handle but not yet modified. [t_handle_lock]
658 	 */
659 	atomic_t		t_outstanding_credits;
660 
661 	/*
662 	 * Forward and backward links for the circular list of all transactions
663 	 * awaiting checkpoint. [j_list_lock]
664 	 */
665 	transaction_t		*t_cpnext, *t_cpprev;
666 
667 	/*
668 	 * When will the transaction expire (become due for commit), in jiffies?
669 	 * [no locking]
670 	 */
671 	unsigned long		t_expires;
672 
673 	/*
674 	 * When this transaction started, in nanoseconds [no locking]
675 	 */
676 	ktime_t			t_start_time;
677 
678 	/*
679 	 * How many handles used this transaction? [t_handle_lock]
680 	 */
681 	atomic_t		t_handle_count;
682 
683 	/*
684 	 * This transaction is being forced and some process is
685 	 * waiting for it to finish.
686 	 */
687 	unsigned int t_synchronous_commit:1;
688 
689 	/* Disk flush needs to be sent to fs partition [no locking] */
690 	int			t_need_data_flush;
691 
692 	/*
693 	 * For use by the filesystem to store fs-specific data
694 	 * structures associated with the transaction
695 	 */
696 	struct list_head	t_private_list;
697 };
698 
699 struct transaction_run_stats_s {
700 	unsigned long		rs_wait;
701 	unsigned long		rs_request_delay;
702 	unsigned long		rs_running;
703 	unsigned long		rs_locked;
704 	unsigned long		rs_flushing;
705 	unsigned long		rs_logging;
706 
707 	__u32			rs_handle_count;
708 	__u32			rs_blocks;
709 	__u32			rs_blocks_logged;
710 };
711 
712 struct transaction_stats_s {
713 	unsigned long		ts_tid;
714 	unsigned long		ts_requested;
715 	struct transaction_run_stats_s run;
716 };
717 
718 static inline unsigned long
719 jbd2_time_diff(unsigned long start, unsigned long end)
720 {
721 	if (end >= start)
722 		return end - start;
723 
724 	return end + (MAX_JIFFY_OFFSET - start);
725 }
726 
727 #define JBD2_NR_BATCH	64
728 
729 /**
730  * struct journal_s - The journal_s type is the concrete type associated with
731  *     journal_t.
732  * @j_flags:  General journaling state flags
733  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
734  *     prior abort)?
735  * @j_sb_buffer: First part of superblock buffer
736  * @j_superblock: Second part of superblock buffer
737  * @j_format_version: Version of the superblock format
738  * @j_state_lock: Protect the various scalars in the journal
739  * @j_barrier_count:  Number of processes waiting to create a barrier lock
740  * @j_barrier: The barrier lock itself
741  * @j_running_transaction: The current running transaction..
742  * @j_committing_transaction: the transaction we are pushing to disk
743  * @j_checkpoint_transactions: a linked circular list of all transactions
744  *  waiting for checkpointing
745  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
746  *  to start committing, or for a barrier lock to be released
747  * @j_wait_done_commit: Wait queue for waiting for commit to complete
748  * @j_wait_commit: Wait queue to trigger commit
749  * @j_wait_updates: Wait queue to wait for updates to complete
750  * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop
751  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
752  * @j_head: Journal head - identifies the first unused block in the journal
753  * @j_tail: Journal tail - identifies the oldest still-used block in the
754  *  journal.
755  * @j_free: Journal free - how many free blocks are there in the journal?
756  * @j_first: The block number of the first usable block
757  * @j_last: The block number one beyond the last usable block
758  * @j_dev: Device where we store the journal
759  * @j_blocksize: blocksize for the location where we store the journal.
760  * @j_blk_offset: starting block offset for into the device where we store the
761  *     journal
762  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
763  *     be equal to j_dev
764  * @j_reserved_credits: Number of buffers reserved from the running transaction
765  * @j_maxlen: Total maximum capacity of the journal region on disk.
766  * @j_list_lock: Protects the buffer lists and internal buffer state.
767  * @j_inode: Optional inode where we store the journal.  If present, all journal
768  *     block numbers are mapped into this inode via bmap().
769  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
770  * @j_transaction_sequence: Sequence number of the next transaction to grant
771  * @j_commit_sequence: Sequence number of the most recently committed
772  *  transaction
773  * @j_commit_request: Sequence number of the most recent transaction wanting
774  *     commit
775  * @j_uuid: Uuid of client object.
776  * @j_task: Pointer to the current commit thread for this journal
777  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
778  *     single compound commit transaction
779  * @j_commit_interval: What is the maximum transaction lifetime before we begin
780  *  a commit?
781  * @j_commit_timer:  The timer used to wakeup the commit thread
782  * @j_revoke_lock: Protect the revoke table
783  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
784  *     current transaction.
785  * @j_revoke_table: alternate revoke tables for j_revoke
786  * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
787  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
788  *	number that will fit in j_blocksize
789  * @j_last_sync_writer: most recent pid which did a synchronous write
790  * @j_history_lock: Protect the transactions statistics history
791  * @j_proc_entry: procfs entry for the jbd statistics directory
792  * @j_stats: Overall statistics
793  * @j_private: An opaque pointer to fs-private information.
794  * @j_trans_commit_map: Lockdep entity to track transaction commit dependencies
795  */
796 
797 struct journal_s
798 {
799 	/* General journaling state flags [j_state_lock] */
800 	unsigned long		j_flags;
801 
802 	/*
803 	 * Is there an outstanding uncleared error on the journal (from a prior
804 	 * abort)? [j_state_lock]
805 	 */
806 	int			j_errno;
807 
808 	/* The superblock buffer */
809 	struct buffer_head	*j_sb_buffer;
810 	journal_superblock_t	*j_superblock;
811 
812 	/* Version of the superblock format */
813 	int			j_format_version;
814 
815 	/*
816 	 * Protect the various scalars in the journal
817 	 */
818 	rwlock_t		j_state_lock;
819 
820 	/*
821 	 * Number of processes waiting to create a barrier lock [j_state_lock]
822 	 */
823 	int			j_barrier_count;
824 
825 	/* The barrier lock itself */
826 	struct mutex		j_barrier;
827 
828 	/*
829 	 * Transactions: The current running transaction...
830 	 * [j_state_lock] [caller holding open handle]
831 	 */
832 	transaction_t		*j_running_transaction;
833 
834 	/*
835 	 * the transaction we are pushing to disk
836 	 * [j_state_lock] [caller holding open handle]
837 	 */
838 	transaction_t		*j_committing_transaction;
839 
840 	/*
841 	 * ... and a linked circular list of all transactions waiting for
842 	 * checkpointing. [j_list_lock]
843 	 */
844 	transaction_t		*j_checkpoint_transactions;
845 
846 	/*
847 	 * Wait queue for waiting for a locked transaction to start committing,
848 	 * or for a barrier lock to be released
849 	 */
850 	wait_queue_head_t	j_wait_transaction_locked;
851 
852 	/* Wait queue for waiting for commit to complete */
853 	wait_queue_head_t	j_wait_done_commit;
854 
855 	/* Wait queue to trigger commit */
856 	wait_queue_head_t	j_wait_commit;
857 
858 	/* Wait queue to wait for updates to complete */
859 	wait_queue_head_t	j_wait_updates;
860 
861 	/* Wait queue to wait for reserved buffer credits to drop */
862 	wait_queue_head_t	j_wait_reserved;
863 
864 	/* Semaphore for locking against concurrent checkpoints */
865 	struct mutex		j_checkpoint_mutex;
866 
867 	/*
868 	 * List of buffer heads used by the checkpoint routine.  This
869 	 * was moved from jbd2_log_do_checkpoint() to reduce stack
870 	 * usage.  Access to this array is controlled by the
871 	 * j_checkpoint_mutex.  [j_checkpoint_mutex]
872 	 */
873 	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
874 
875 	/*
876 	 * Journal head: identifies the first unused block in the journal.
877 	 * [j_state_lock]
878 	 */
879 	unsigned long		j_head;
880 
881 	/*
882 	 * Journal tail: identifies the oldest still-used block in the journal.
883 	 * [j_state_lock]
884 	 */
885 	unsigned long		j_tail;
886 
887 	/*
888 	 * Journal free: how many free blocks are there in the journal?
889 	 * [j_state_lock]
890 	 */
891 	unsigned long		j_free;
892 
893 	/*
894 	 * Journal start and end: the block numbers of the first usable block
895 	 * and one beyond the last usable block in the journal. [j_state_lock]
896 	 */
897 	unsigned long		j_first;
898 	unsigned long		j_last;
899 
900 	/*
901 	 * Device, blocksize and starting block offset for the location where we
902 	 * store the journal.
903 	 */
904 	struct block_device	*j_dev;
905 	int			j_blocksize;
906 	unsigned long long	j_blk_offset;
907 	char			j_devname[BDEVNAME_SIZE+24];
908 
909 	/*
910 	 * Device which holds the client fs.  For internal journal this will be
911 	 * equal to j_dev.
912 	 */
913 	struct block_device	*j_fs_dev;
914 
915 	/* Total maximum capacity of the journal region on disk. */
916 	unsigned int		j_maxlen;
917 
918 	/* Number of buffers reserved from the running transaction */
919 	atomic_t		j_reserved_credits;
920 
921 	/*
922 	 * Protects the buffer lists and internal buffer state.
923 	 */
924 	spinlock_t		j_list_lock;
925 
926 	/* Optional inode where we store the journal.  If present, all */
927 	/* journal block numbers are mapped into this inode via */
928 	/* bmap(). */
929 	struct inode		*j_inode;
930 
931 	/*
932 	 * Sequence number of the oldest transaction in the log [j_state_lock]
933 	 */
934 	tid_t			j_tail_sequence;
935 
936 	/*
937 	 * Sequence number of the next transaction to grant [j_state_lock]
938 	 */
939 	tid_t			j_transaction_sequence;
940 
941 	/*
942 	 * Sequence number of the most recently committed transaction
943 	 * [j_state_lock].
944 	 */
945 	tid_t			j_commit_sequence;
946 
947 	/*
948 	 * Sequence number of the most recent transaction wanting commit
949 	 * [j_state_lock]
950 	 */
951 	tid_t			j_commit_request;
952 
953 	/*
954 	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
955 	 * backed by this journal.  This will eventually be replaced by an array
956 	 * of uuids, allowing us to index multiple devices within a single
957 	 * journal and to perform atomic updates across them.
958 	 */
959 	__u8			j_uuid[16];
960 
961 	/* Pointer to the current commit thread for this journal */
962 	struct task_struct	*j_task;
963 
964 	/*
965 	 * Maximum number of metadata buffers to allow in a single compound
966 	 * commit transaction
967 	 */
968 	int			j_max_transaction_buffers;
969 
970 	/*
971 	 * What is the maximum transaction lifetime before we begin a commit?
972 	 */
973 	unsigned long		j_commit_interval;
974 
975 	/* The timer used to wakeup the commit thread: */
976 	struct timer_list	j_commit_timer;
977 
978 	/*
979 	 * The revoke table: maintains the list of revoked blocks in the
980 	 * current transaction.  [j_revoke_lock]
981 	 */
982 	spinlock_t		j_revoke_lock;
983 	struct jbd2_revoke_table_s *j_revoke;
984 	struct jbd2_revoke_table_s *j_revoke_table[2];
985 
986 	/*
987 	 * array of bhs for jbd2_journal_commit_transaction
988 	 */
989 	struct buffer_head	**j_wbuf;
990 	int			j_wbufsize;
991 
992 	/*
993 	 * this is the pid of hte last person to run a synchronous operation
994 	 * through the journal
995 	 */
996 	pid_t			j_last_sync_writer;
997 
998 	/*
999 	 * the average amount of time in nanoseconds it takes to commit a
1000 	 * transaction to disk. [j_state_lock]
1001 	 */
1002 	u64			j_average_commit_time;
1003 
1004 	/*
1005 	 * minimum and maximum times that we should wait for
1006 	 * additional filesystem operations to get batched into a
1007 	 * synchronous handle in microseconds
1008 	 */
1009 	u32			j_min_batch_time;
1010 	u32			j_max_batch_time;
1011 
1012 	/* This function is called when a transaction is closed */
1013 	void			(*j_commit_callback)(journal_t *,
1014 						     transaction_t *);
1015 
1016 	/*
1017 	 * Journal statistics
1018 	 */
1019 	spinlock_t		j_history_lock;
1020 	struct proc_dir_entry	*j_proc_entry;
1021 	struct transaction_stats_s j_stats;
1022 
1023 	/* Failed journal commit ID */
1024 	unsigned int		j_failed_commit;
1025 
1026 	/*
1027 	 * An opaque pointer to fs-private information.  ext3 puts its
1028 	 * superblock pointer here
1029 	 */
1030 	void *j_private;
1031 
1032 	/* Reference to checksum algorithm driver via cryptoapi */
1033 	struct crypto_shash *j_chksum_driver;
1034 
1035 	/* Precomputed journal UUID checksum for seeding other checksums */
1036 	__u32 j_csum_seed;
1037 
1038 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1039 	/*
1040 	 * Lockdep entity to track transaction commit dependencies. Handles
1041 	 * hold this "lock" for read, when we wait for commit, we acquire the
1042 	 * "lock" for writing. This matches the properties of jbd2 journalling
1043 	 * where the running transaction has to wait for all handles to be
1044 	 * dropped to commit that transaction and also acquiring a handle may
1045 	 * require transaction commit to finish.
1046 	 */
1047 	struct lockdep_map	j_trans_commit_map;
1048 #endif
1049 };
1050 
1051 #define jbd2_might_wait_for_commit(j) \
1052 	do { \
1053 		rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1054 		rwsem_release(&j->j_trans_commit_map, 1, _THIS_IP_); \
1055 	} while (0)
1056 
1057 /* journal feature predicate functions */
1058 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1059 static inline bool jbd2_has_feature_##name(journal_t *j) \
1060 { \
1061 	return ((j)->j_format_version >= 2 && \
1062 		((j)->j_superblock->s_feature_compat & \
1063 		 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1064 } \
1065 static inline void jbd2_set_feature_##name(journal_t *j) \
1066 { \
1067 	(j)->j_superblock->s_feature_compat |= \
1068 		cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1069 } \
1070 static inline void jbd2_clear_feature_##name(journal_t *j) \
1071 { \
1072 	(j)->j_superblock->s_feature_compat &= \
1073 		~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1074 }
1075 
1076 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1077 static inline bool jbd2_has_feature_##name(journal_t *j) \
1078 { \
1079 	return ((j)->j_format_version >= 2 && \
1080 		((j)->j_superblock->s_feature_ro_compat & \
1081 		 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1082 } \
1083 static inline void jbd2_set_feature_##name(journal_t *j) \
1084 { \
1085 	(j)->j_superblock->s_feature_ro_compat |= \
1086 		cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1087 } \
1088 static inline void jbd2_clear_feature_##name(journal_t *j) \
1089 { \
1090 	(j)->j_superblock->s_feature_ro_compat &= \
1091 		~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1092 }
1093 
1094 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1095 static inline bool jbd2_has_feature_##name(journal_t *j) \
1096 { \
1097 	return ((j)->j_format_version >= 2 && \
1098 		((j)->j_superblock->s_feature_incompat & \
1099 		 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1100 } \
1101 static inline void jbd2_set_feature_##name(journal_t *j) \
1102 { \
1103 	(j)->j_superblock->s_feature_incompat |= \
1104 		cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1105 } \
1106 static inline void jbd2_clear_feature_##name(journal_t *j) \
1107 { \
1108 	(j)->j_superblock->s_feature_incompat &= \
1109 		~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1110 }
1111 
1112 JBD2_FEATURE_COMPAT_FUNCS(checksum,		CHECKSUM)
1113 
1114 JBD2_FEATURE_INCOMPAT_FUNCS(revoke,		REVOKE)
1115 JBD2_FEATURE_INCOMPAT_FUNCS(64bit,		64BIT)
1116 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit,	ASYNC_COMMIT)
1117 JBD2_FEATURE_INCOMPAT_FUNCS(csum2,		CSUM_V2)
1118 JBD2_FEATURE_INCOMPAT_FUNCS(csum3,		CSUM_V3)
1119 
1120 /*
1121  * Journal flag definitions
1122  */
1123 #define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1124 #define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1125 #define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1126 #define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1127 #define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1128 #define JBD2_BARRIER	0x020	/* Use IDE barriers */
1129 #define JBD2_ABORT_ON_SYNCDATA_ERR	0x040	/* Abort the journal on file
1130 						 * data write error in ordered
1131 						 * mode */
1132 #define JBD2_REC_ERR	0x080	/* The errno in the sb has been recorded */
1133 
1134 /*
1135  * Function declarations for the journaling transaction and buffer
1136  * management
1137  */
1138 
1139 /* Filing buffers */
1140 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1141 extern void __jbd2_journal_refile_buffer(struct journal_head *);
1142 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1143 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1144 extern void __journal_free_buffer(struct journal_head *bh);
1145 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1146 extern void __journal_clean_data_list(transaction_t *transaction);
1147 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1148 {
1149 	list_add_tail(&bh->b_assoc_buffers, head);
1150 }
1151 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1152 {
1153 	list_del_init(&bh->b_assoc_buffers);
1154 }
1155 
1156 /* Log buffer allocation */
1157 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1158 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1159 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1160 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1161 			      unsigned long *block);
1162 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1163 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1164 
1165 /* Commit management */
1166 extern void jbd2_journal_commit_transaction(journal_t *);
1167 
1168 /* Checkpoint list management */
1169 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1170 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1171 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1172 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1173 
1174 
1175 /*
1176  * Triggers
1177  */
1178 
1179 struct jbd2_buffer_trigger_type {
1180 	/*
1181 	 * Fired a the moment data to write to the journal are known to be
1182 	 * stable - so either at the moment b_frozen_data is created or just
1183 	 * before a buffer is written to the journal.  mapped_data is a mapped
1184 	 * buffer that is the frozen data for commit.
1185 	 */
1186 	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1187 			 struct buffer_head *bh, void *mapped_data,
1188 			 size_t size);
1189 
1190 	/*
1191 	 * Fired during journal abort for dirty buffers that will not be
1192 	 * committed.
1193 	 */
1194 	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1195 			struct buffer_head *bh);
1196 };
1197 
1198 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1199 				       void *mapped_data,
1200 				       struct jbd2_buffer_trigger_type *triggers);
1201 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1202 				      struct jbd2_buffer_trigger_type *triggers);
1203 
1204 /* Buffer IO */
1205 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1206 					      struct journal_head *jh_in,
1207 					      struct buffer_head **bh_out,
1208 					      sector_t blocknr);
1209 
1210 /* Transaction locking */
1211 extern void		__wait_on_journal (journal_t *);
1212 
1213 /* Transaction cache support */
1214 extern void jbd2_journal_destroy_transaction_cache(void);
1215 extern int  jbd2_journal_init_transaction_cache(void);
1216 extern void jbd2_journal_free_transaction(transaction_t *);
1217 
1218 /*
1219  * Journal locking.
1220  *
1221  * We need to lock the journal during transaction state changes so that nobody
1222  * ever tries to take a handle on the running transaction while we are in the
1223  * middle of moving it to the commit phase.  j_state_lock does this.
1224  *
1225  * Note that the locking is completely interrupt unsafe.  We never touch
1226  * journal structures from interrupts.
1227  */
1228 
1229 static inline handle_t *journal_current_handle(void)
1230 {
1231 	return current->journal_info;
1232 }
1233 
1234 /* The journaling code user interface:
1235  *
1236  * Create and destroy handles
1237  * Register buffer modifications against the current transaction.
1238  */
1239 
1240 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1241 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1242 				     gfp_t gfp_mask, unsigned int type,
1243 				     unsigned int line_no);
1244 extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1245 extern int	 jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
1246 extern int	 jbd2_journal_start_reserved(handle_t *handle,
1247 				unsigned int type, unsigned int line_no);
1248 extern void	 jbd2_journal_free_reserved(handle_t *handle);
1249 extern int	 jbd2_journal_extend (handle_t *, int nblocks);
1250 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1251 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1252 extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1253 void		 jbd2_journal_set_triggers(struct buffer_head *,
1254 					   struct jbd2_buffer_trigger_type *type);
1255 extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1256 extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1257 extern void	 journal_sync_buffer (struct buffer_head *);
1258 extern int	 jbd2_journal_invalidatepage(journal_t *,
1259 				struct page *, unsigned int, unsigned int);
1260 extern int	 jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1261 extern int	 jbd2_journal_stop(handle_t *);
1262 extern int	 jbd2_journal_flush (journal_t *);
1263 extern void	 jbd2_journal_lock_updates (journal_t *);
1264 extern void	 jbd2_journal_unlock_updates (journal_t *);
1265 
1266 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1267 				struct block_device *fs_dev,
1268 				unsigned long long start, int len, int bsize);
1269 extern journal_t * jbd2_journal_init_inode (struct inode *);
1270 extern int	   jbd2_journal_update_format (journal_t *);
1271 extern int	   jbd2_journal_check_used_features
1272 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1273 extern int	   jbd2_journal_check_available_features
1274 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1275 extern int	   jbd2_journal_set_features
1276 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1277 extern void	   jbd2_journal_clear_features
1278 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1279 extern int	   jbd2_journal_load       (journal_t *journal);
1280 extern int	   jbd2_journal_destroy    (journal_t *);
1281 extern int	   jbd2_journal_recover    (journal_t *journal);
1282 extern int	   jbd2_journal_wipe       (journal_t *, int);
1283 extern int	   jbd2_journal_skip_recovery	(journal_t *);
1284 extern void	   jbd2_journal_update_sb_errno(journal_t *);
1285 extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1286 				unsigned long, int);
1287 extern void	   __jbd2_journal_abort_hard	(journal_t *);
1288 extern void	   jbd2_journal_abort      (journal_t *, int);
1289 extern int	   jbd2_journal_errno      (journal_t *);
1290 extern void	   jbd2_journal_ack_err    (journal_t *);
1291 extern int	   jbd2_journal_clear_err  (journal_t *);
1292 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1293 extern int	   jbd2_journal_force_commit(journal_t *);
1294 extern int	   jbd2_journal_force_commit_nested(journal_t *);
1295 extern int	   jbd2_journal_inode_add_write(handle_t *handle, struct jbd2_inode *inode);
1296 extern int	   jbd2_journal_inode_add_wait(handle_t *handle, struct jbd2_inode *inode);
1297 extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1298 				struct jbd2_inode *inode, loff_t new_size);
1299 extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1300 extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1301 
1302 /*
1303  * journal_head management
1304  */
1305 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1306 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1307 void jbd2_journal_put_journal_head(struct journal_head *jh);
1308 
1309 /*
1310  * handle management
1311  */
1312 extern struct kmem_cache *jbd2_handle_cache;
1313 
1314 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1315 {
1316 	return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1317 }
1318 
1319 static inline void jbd2_free_handle(handle_t *handle)
1320 {
1321 	kmem_cache_free(jbd2_handle_cache, handle);
1322 }
1323 
1324 /*
1325  * jbd2_inode management (optional, for those file systems that want to use
1326  * dynamically allocated jbd2_inode structures)
1327  */
1328 extern struct kmem_cache *jbd2_inode_cache;
1329 
1330 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1331 {
1332 	return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1333 }
1334 
1335 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1336 {
1337 	kmem_cache_free(jbd2_inode_cache, jinode);
1338 }
1339 
1340 /* Primary revoke support */
1341 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1342 extern int	   jbd2_journal_init_revoke(journal_t *, int);
1343 extern void	   jbd2_journal_destroy_revoke_caches(void);
1344 extern int	   jbd2_journal_init_revoke_caches(void);
1345 
1346 extern void	   jbd2_journal_destroy_revoke(journal_t *);
1347 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1348 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1349 extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
1350 						     struct list_head *log_bufs);
1351 
1352 /* Recovery revoke support */
1353 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1354 extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1355 extern void	jbd2_journal_clear_revoke(journal_t *);
1356 extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1357 extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1358 
1359 /*
1360  * The log thread user interface:
1361  *
1362  * Request space in the current transaction, and force transaction commit
1363  * transitions on demand.
1364  */
1365 
1366 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1367 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1368 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1369 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1370 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1371 int jbd2_log_do_checkpoint(journal_t *journal);
1372 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1373 
1374 void __jbd2_log_wait_for_space(journal_t *journal);
1375 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1376 extern int jbd2_cleanup_journal_tail(journal_t *);
1377 
1378 /*
1379  * is_journal_abort
1380  *
1381  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1382  * bit, when set, indicates that we have had a fatal error somewhere,
1383  * either inside the journaling layer or indicated to us by the client
1384  * (eg. ext3), and that we and should not commit any further
1385  * transactions.
1386  */
1387 
1388 static inline int is_journal_aborted(journal_t *journal)
1389 {
1390 	return journal->j_flags & JBD2_ABORT;
1391 }
1392 
1393 static inline int is_handle_aborted(handle_t *handle)
1394 {
1395 	if (handle->h_aborted || !handle->h_transaction)
1396 		return 1;
1397 	return is_journal_aborted(handle->h_transaction->t_journal);
1398 }
1399 
1400 static inline void jbd2_journal_abort_handle(handle_t *handle)
1401 {
1402 	handle->h_aborted = 1;
1403 }
1404 
1405 #endif /* __KERNEL__   */
1406 
1407 /* Comparison functions for transaction IDs: perform comparisons using
1408  * modulo arithmetic so that they work over sequence number wraps. */
1409 
1410 static inline int tid_gt(tid_t x, tid_t y)
1411 {
1412 	int difference = (x - y);
1413 	return (difference > 0);
1414 }
1415 
1416 static inline int tid_geq(tid_t x, tid_t y)
1417 {
1418 	int difference = (x - y);
1419 	return (difference >= 0);
1420 }
1421 
1422 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1423 extern size_t journal_tag_bytes(journal_t *journal);
1424 
1425 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1426 {
1427 	return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1428 }
1429 
1430 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1431 {
1432 	WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1433 		     journal->j_chksum_driver == NULL);
1434 
1435 	return journal->j_chksum_driver != NULL;
1436 }
1437 
1438 /*
1439  * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
1440  * transaction control blocks.
1441  */
1442 #define JBD2_CONTROL_BLOCKS_SHIFT 5
1443 
1444 /*
1445  * Return the minimum number of blocks which must be free in the journal
1446  * before a new transaction may be started.  Must be called under j_state_lock.
1447  */
1448 static inline int jbd2_space_needed(journal_t *journal)
1449 {
1450 	int nblocks = journal->j_max_transaction_buffers;
1451 	return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT);
1452 }
1453 
1454 /*
1455  * Return number of free blocks in the log. Must be called under j_state_lock.
1456  */
1457 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1458 {
1459 	/* Allow for rounding errors */
1460 	unsigned long free = journal->j_free - 32;
1461 
1462 	if (journal->j_committing_transaction) {
1463 		unsigned long committing = atomic_read(&journal->
1464 			j_committing_transaction->t_outstanding_credits);
1465 
1466 		/* Transaction + control blocks */
1467 		free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT);
1468 	}
1469 	return free;
1470 }
1471 
1472 /*
1473  * Definitions which augment the buffer_head layer
1474  */
1475 
1476 /* journaling buffer types */
1477 #define BJ_None		0	/* Not journaled */
1478 #define BJ_Metadata	1	/* Normal journaled metadata */
1479 #define BJ_Forget	2	/* Buffer superseded by this transaction */
1480 #define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1481 #define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1482 #define BJ_Types	5
1483 
1484 extern int jbd_blocks_per_page(struct inode *inode);
1485 
1486 /* JBD uses a CRC32 checksum */
1487 #define JBD_MAX_CHECKSUM_SIZE 4
1488 
1489 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1490 			      const void *address, unsigned int length)
1491 {
1492 	struct {
1493 		struct shash_desc shash;
1494 		char ctx[JBD_MAX_CHECKSUM_SIZE];
1495 	} desc;
1496 	int err;
1497 
1498 	BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1499 		JBD_MAX_CHECKSUM_SIZE);
1500 
1501 	desc.shash.tfm = journal->j_chksum_driver;
1502 	desc.shash.flags = 0;
1503 	*(u32 *)desc.ctx = crc;
1504 
1505 	err = crypto_shash_update(&desc.shash, address, length);
1506 	BUG_ON(err);
1507 
1508 	return *(u32 *)desc.ctx;
1509 }
1510 
1511 /* Return most recent uncommitted transaction */
1512 static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1513 {
1514 	tid_t tid;
1515 
1516 	read_lock(&journal->j_state_lock);
1517 	tid = journal->j_commit_request;
1518 	if (journal->j_running_transaction)
1519 		tid = journal->j_running_transaction->t_tid;
1520 	read_unlock(&journal->j_state_lock);
1521 	return tid;
1522 }
1523 
1524 #ifdef __KERNEL__
1525 
1526 #define buffer_trace_init(bh)	do {} while (0)
1527 #define print_buffer_fields(bh)	do {} while (0)
1528 #define print_buffer_trace(bh)	do {} while (0)
1529 #define BUFFER_TRACE(bh, info)	do {} while (0)
1530 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1531 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1532 
1533 #endif	/* __KERNEL__ */
1534 
1535 #define EFSBADCRC	EBADMSG		/* Bad CRC detected */
1536 #define EFSCORRUPTED	EUCLEAN		/* Filesystem is corrupted */
1537 
1538 #endif	/* _LINUX_JBD2_H */
1539