xref: /linux/include/linux/jbd2.h (revision b85d45947951d23cb22d90caecf4c1eb81342c96)
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 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 /* Tail of revoke block, for checksumming */
219 struct jbd2_journal_revoke_tail {
220 	__be32		r_checksum;	/* crc32c(uuid+revoke_block) */
221 };
222 
223 /* Definitions for the journal tag flags word: */
224 #define JBD2_FLAG_ESCAPE		1	/* on-disk block is escaped */
225 #define JBD2_FLAG_SAME_UUID	2	/* block has same uuid as previous */
226 #define JBD2_FLAG_DELETED	4	/* block deleted by this transaction */
227 #define JBD2_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
228 
229 
230 /*
231  * The journal superblock.  All fields are in big-endian byte order.
232  */
233 typedef struct journal_superblock_s
234 {
235 /* 0x0000 */
236 	journal_header_t s_header;
237 
238 /* 0x000C */
239 	/* Static information describing the journal */
240 	__be32	s_blocksize;		/* journal device blocksize */
241 	__be32	s_maxlen;		/* total blocks in journal file */
242 	__be32	s_first;		/* first block of log information */
243 
244 /* 0x0018 */
245 	/* Dynamic information describing the current state of the log */
246 	__be32	s_sequence;		/* first commit ID expected in log */
247 	__be32	s_start;		/* blocknr of start of log */
248 
249 /* 0x0020 */
250 	/* Error value, as set by jbd2_journal_abort(). */
251 	__be32	s_errno;
252 
253 /* 0x0024 */
254 	/* Remaining fields are only valid in a version-2 superblock */
255 	__be32	s_feature_compat;	/* compatible feature set */
256 	__be32	s_feature_incompat;	/* incompatible feature set */
257 	__be32	s_feature_ro_compat;	/* readonly-compatible feature set */
258 /* 0x0030 */
259 	__u8	s_uuid[16];		/* 128-bit uuid for journal */
260 
261 /* 0x0040 */
262 	__be32	s_nr_users;		/* Nr of filesystems sharing log */
263 
264 	__be32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
265 
266 /* 0x0048 */
267 	__be32	s_max_transaction;	/* Limit of journal blocks per trans.*/
268 	__be32	s_max_trans_data;	/* Limit of data blocks per trans. */
269 
270 /* 0x0050 */
271 	__u8	s_checksum_type;	/* checksum type */
272 	__u8	s_padding2[3];
273 	__u32	s_padding[42];
274 	__be32	s_checksum;		/* crc32c(superblock) */
275 
276 /* 0x0100 */
277 	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
278 /* 0x0400 */
279 } journal_superblock_t;
280 
281 #define JBD2_HAS_COMPAT_FEATURE(j,mask)					\
282 	((j)->j_format_version >= 2 &&					\
283 	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
284 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask)				\
285 	((j)->j_format_version >= 2 &&					\
286 	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
287 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask)				\
288 	((j)->j_format_version >= 2 &&					\
289 	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
290 
291 #define JBD2_FEATURE_COMPAT_CHECKSUM	0x00000001
292 
293 #define JBD2_FEATURE_INCOMPAT_REVOKE		0x00000001
294 #define JBD2_FEATURE_INCOMPAT_64BIT		0x00000002
295 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
296 #define JBD2_FEATURE_INCOMPAT_CSUM_V2		0x00000008
297 #define JBD2_FEATURE_INCOMPAT_CSUM_V3		0x00000010
298 
299 /* Features known to this kernel version: */
300 #define JBD2_KNOWN_COMPAT_FEATURES	JBD2_FEATURE_COMPAT_CHECKSUM
301 #define JBD2_KNOWN_ROCOMPAT_FEATURES	0
302 #define JBD2_KNOWN_INCOMPAT_FEATURES	(JBD2_FEATURE_INCOMPAT_REVOKE | \
303 					JBD2_FEATURE_INCOMPAT_64BIT | \
304 					JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
305 					JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
306 					JBD2_FEATURE_INCOMPAT_CSUM_V3)
307 
308 #ifdef __KERNEL__
309 
310 #include <linux/fs.h>
311 #include <linux/sched.h>
312 
313 enum jbd_state_bits {
314 	BH_JBD			/* Has an attached ext3 journal_head */
315 	  = BH_PrivateStart,
316 	BH_JWrite,		/* Being written to log (@@@ DEBUGGING) */
317 	BH_Freed,		/* Has been freed (truncated) */
318 	BH_Revoked,		/* Has been revoked from the log */
319 	BH_RevokeValid,		/* Revoked flag is valid */
320 	BH_JBDDirty,		/* Is dirty but journaled */
321 	BH_State,		/* Pins most journal_head state */
322 	BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
323 	BH_Shadow,		/* IO on shadow buffer is running */
324 	BH_Verified,		/* Metadata block has been verified ok */
325 	BH_JBDPrivateStart,	/* First bit available for private use by FS */
326 };
327 
328 BUFFER_FNS(JBD, jbd)
329 BUFFER_FNS(JWrite, jwrite)
330 BUFFER_FNS(JBDDirty, jbddirty)
331 TAS_BUFFER_FNS(JBDDirty, jbddirty)
332 BUFFER_FNS(Revoked, revoked)
333 TAS_BUFFER_FNS(Revoked, revoked)
334 BUFFER_FNS(RevokeValid, revokevalid)
335 TAS_BUFFER_FNS(RevokeValid, revokevalid)
336 BUFFER_FNS(Freed, freed)
337 BUFFER_FNS(Shadow, shadow)
338 BUFFER_FNS(Verified, verified)
339 
340 static inline struct buffer_head *jh2bh(struct journal_head *jh)
341 {
342 	return jh->b_bh;
343 }
344 
345 static inline struct journal_head *bh2jh(struct buffer_head *bh)
346 {
347 	return bh->b_private;
348 }
349 
350 static inline void jbd_lock_bh_state(struct buffer_head *bh)
351 {
352 	bit_spin_lock(BH_State, &bh->b_state);
353 }
354 
355 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
356 {
357 	return bit_spin_trylock(BH_State, &bh->b_state);
358 }
359 
360 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
361 {
362 	return bit_spin_is_locked(BH_State, &bh->b_state);
363 }
364 
365 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
366 {
367 	bit_spin_unlock(BH_State, &bh->b_state);
368 }
369 
370 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
371 {
372 	bit_spin_lock(BH_JournalHead, &bh->b_state);
373 }
374 
375 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
376 {
377 	bit_spin_unlock(BH_JournalHead, &bh->b_state);
378 }
379 
380 #define J_ASSERT(assert)	BUG_ON(!(assert))
381 
382 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
383 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
384 
385 #if defined(JBD2_PARANOID_IOFAIL)
386 #define J_EXPECT(expr, why...)		J_ASSERT(expr)
387 #define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
388 #define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
389 #else
390 #define __journal_expect(expr, why...)					     \
391 	({								     \
392 		int val = (expr);					     \
393 		if (!val) {						     \
394 			printk(KERN_ERR					     \
395 			       "JBD2 unexpected failure: %s: %s;\n",	     \
396 			       __func__, #expr);			     \
397 			printk(KERN_ERR why "\n");			     \
398 		}							     \
399 		val;							     \
400 	})
401 #define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
402 #define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
403 #define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
404 #endif
405 
406 /* Flags in jbd_inode->i_flags */
407 #define __JI_COMMIT_RUNNING 0
408 /* Commit of the inode data in progress. We use this flag to protect us from
409  * concurrent deletion of inode. We cannot use reference to inode for this
410  * since we cannot afford doing last iput() on behalf of kjournald
411  */
412 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
413 
414 /**
415  * struct jbd_inode is the structure linking inodes in ordered mode
416  *   present in a transaction so that we can sync them during commit.
417  */
418 struct jbd2_inode {
419 	/* Which transaction does this inode belong to? Either the running
420 	 * transaction or the committing one. [j_list_lock] */
421 	transaction_t *i_transaction;
422 
423 	/* Pointer to the running transaction modifying inode's data in case
424 	 * there is already a committing transaction touching it. [j_list_lock] */
425 	transaction_t *i_next_transaction;
426 
427 	/* List of inodes in the i_transaction [j_list_lock] */
428 	struct list_head i_list;
429 
430 	/* VFS inode this inode belongs to [constant during the lifetime
431 	 * of the structure] */
432 	struct inode *i_vfs_inode;
433 
434 	/* Flags of inode [j_list_lock] */
435 	unsigned long i_flags;
436 };
437 
438 struct jbd2_revoke_table_s;
439 
440 /**
441  * struct handle_s - The handle_s type is the concrete type associated with
442  *     handle_t.
443  * @h_transaction: Which compound transaction is this update a part of?
444  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
445  * @h_ref: Reference count on this handle
446  * @h_err: Field for caller's use to track errors through large fs operations
447  * @h_sync: flag for sync-on-close
448  * @h_jdata: flag to force data journaling
449  * @h_aborted: flag indicating fatal error on handle
450  **/
451 
452 /* Docbook can't yet cope with the bit fields, but will leave the documentation
453  * in so it can be fixed later.
454  */
455 
456 struct jbd2_journal_handle
457 {
458 	union {
459 		/* Which compound transaction is this update a part of? */
460 		transaction_t	*h_transaction;
461 		/* Which journal handle belongs to - used iff h_reserved set */
462 		journal_t	*h_journal;
463 	};
464 
465 	/* Handle reserved for finishing the logical operation */
466 	handle_t		*h_rsv_handle;
467 
468 	/* Number of remaining buffers we are allowed to dirty: */
469 	int			h_buffer_credits;
470 
471 	/* Reference count on this handle */
472 	int			h_ref;
473 
474 	/* Field for caller's use to track errors through large fs */
475 	/* operations */
476 	int			h_err;
477 
478 	/* Flags [no locking] */
479 	unsigned int	h_sync:		1;	/* sync-on-close */
480 	unsigned int	h_jdata:	1;	/* force data journaling */
481 	unsigned int	h_reserved:	1;	/* handle with reserved credits */
482 	unsigned int	h_aborted:	1;	/* fatal error on handle */
483 	unsigned int	h_type:		8;	/* for handle statistics */
484 	unsigned int	h_line_no:	16;	/* for handle statistics */
485 
486 	unsigned long		h_start_jiffies;
487 	unsigned int		h_requested_credits;
488 
489 #ifdef CONFIG_DEBUG_LOCK_ALLOC
490 	struct lockdep_map	h_lockdep_map;
491 #endif
492 };
493 
494 
495 /*
496  * Some stats for checkpoint phase
497  */
498 struct transaction_chp_stats_s {
499 	unsigned long		cs_chp_time;
500 	__u32			cs_forced_to_close;
501 	__u32			cs_written;
502 	__u32			cs_dropped;
503 };
504 
505 /* The transaction_t type is the guts of the journaling mechanism.  It
506  * tracks a compound transaction through its various states:
507  *
508  * RUNNING:	accepting new updates
509  * LOCKED:	Updates still running but we don't accept new ones
510  * RUNDOWN:	Updates are tidying up but have finished requesting
511  *		new buffers to modify (state not used for now)
512  * FLUSH:       All updates complete, but we are still writing to disk
513  * COMMIT:      All data on disk, writing commit record
514  * FINISHED:	We still have to keep the transaction for checkpointing.
515  *
516  * The transaction keeps track of all of the buffers modified by a
517  * running transaction, and all of the buffers committed but not yet
518  * flushed to home for finished transactions.
519  */
520 
521 /*
522  * Lock ranking:
523  *
524  *    j_list_lock
525  *      ->jbd_lock_bh_journal_head()	(This is "innermost")
526  *
527  *    j_state_lock
528  *    ->jbd_lock_bh_state()
529  *
530  *    jbd_lock_bh_state()
531  *    ->j_list_lock
532  *
533  *    j_state_lock
534  *    ->t_handle_lock
535  *
536  *    j_state_lock
537  *    ->j_list_lock			(journal_unmap_buffer)
538  *
539  */
540 
541 struct transaction_s
542 {
543 	/* Pointer to the journal for this transaction. [no locking] */
544 	journal_t		*t_journal;
545 
546 	/* Sequence number for this transaction [no locking] */
547 	tid_t			t_tid;
548 
549 	/*
550 	 * Transaction's current state
551 	 * [no locking - only kjournald2 alters this]
552 	 * [j_list_lock] guards transition of a transaction into T_FINISHED
553 	 * state and subsequent call of __jbd2_journal_drop_transaction()
554 	 * FIXME: needs barriers
555 	 * KLUDGE: [use j_state_lock]
556 	 */
557 	enum {
558 		T_RUNNING,
559 		T_LOCKED,
560 		T_FLUSH,
561 		T_COMMIT,
562 		T_COMMIT_DFLUSH,
563 		T_COMMIT_JFLUSH,
564 		T_COMMIT_CALLBACK,
565 		T_FINISHED
566 	}			t_state;
567 
568 	/*
569 	 * Where in the log does this transaction's commit start? [no locking]
570 	 */
571 	unsigned long		t_log_start;
572 
573 	/* Number of buffers on the t_buffers list [j_list_lock] */
574 	int			t_nr_buffers;
575 
576 	/*
577 	 * Doubly-linked circular list of all buffers reserved but not yet
578 	 * modified by this transaction [j_list_lock]
579 	 */
580 	struct journal_head	*t_reserved_list;
581 
582 	/*
583 	 * Doubly-linked circular list of all metadata buffers owned by this
584 	 * transaction [j_list_lock]
585 	 */
586 	struct journal_head	*t_buffers;
587 
588 	/*
589 	 * Doubly-linked circular list of all forget buffers (superseded
590 	 * buffers which we can un-checkpoint once this transaction commits)
591 	 * [j_list_lock]
592 	 */
593 	struct journal_head	*t_forget;
594 
595 	/*
596 	 * Doubly-linked circular list of all buffers still to be flushed before
597 	 * this transaction can be checkpointed. [j_list_lock]
598 	 */
599 	struct journal_head	*t_checkpoint_list;
600 
601 	/*
602 	 * Doubly-linked circular list of all buffers submitted for IO while
603 	 * checkpointing. [j_list_lock]
604 	 */
605 	struct journal_head	*t_checkpoint_io_list;
606 
607 	/*
608 	 * Doubly-linked circular list of metadata buffers being shadowed by log
609 	 * IO.  The IO buffers on the iobuf list and the shadow buffers on this
610 	 * list match each other one for one at all times. [j_list_lock]
611 	 */
612 	struct journal_head	*t_shadow_list;
613 
614 	/*
615 	 * List of inodes whose data we've modified in data=ordered mode.
616 	 * [j_list_lock]
617 	 */
618 	struct list_head	t_inode_list;
619 
620 	/*
621 	 * Protects info related to handles
622 	 */
623 	spinlock_t		t_handle_lock;
624 
625 	/*
626 	 * Longest time some handle had to wait for running transaction
627 	 */
628 	unsigned long		t_max_wait;
629 
630 	/*
631 	 * When transaction started
632 	 */
633 	unsigned long		t_start;
634 
635 	/*
636 	 * When commit was requested
637 	 */
638 	unsigned long		t_requested;
639 
640 	/*
641 	 * Checkpointing stats [j_checkpoint_sem]
642 	 */
643 	struct transaction_chp_stats_s t_chp_stats;
644 
645 	/*
646 	 * Number of outstanding updates running on this transaction
647 	 * [t_handle_lock]
648 	 */
649 	atomic_t		t_updates;
650 
651 	/*
652 	 * Number of buffers reserved for use by all handles in this transaction
653 	 * handle but not yet modified. [t_handle_lock]
654 	 */
655 	atomic_t		t_outstanding_credits;
656 
657 	/*
658 	 * Forward and backward links for the circular list of all transactions
659 	 * awaiting checkpoint. [j_list_lock]
660 	 */
661 	transaction_t		*t_cpnext, *t_cpprev;
662 
663 	/*
664 	 * When will the transaction expire (become due for commit), in jiffies?
665 	 * [no locking]
666 	 */
667 	unsigned long		t_expires;
668 
669 	/*
670 	 * When this transaction started, in nanoseconds [no locking]
671 	 */
672 	ktime_t			t_start_time;
673 
674 	/*
675 	 * How many handles used this transaction? [t_handle_lock]
676 	 */
677 	atomic_t		t_handle_count;
678 
679 	/*
680 	 * This transaction is being forced and some process is
681 	 * waiting for it to finish.
682 	 */
683 	unsigned int t_synchronous_commit:1;
684 
685 	/* Disk flush needs to be sent to fs partition [no locking] */
686 	int			t_need_data_flush;
687 
688 	/*
689 	 * For use by the filesystem to store fs-specific data
690 	 * structures associated with the transaction
691 	 */
692 	struct list_head	t_private_list;
693 };
694 
695 struct transaction_run_stats_s {
696 	unsigned long		rs_wait;
697 	unsigned long		rs_request_delay;
698 	unsigned long		rs_running;
699 	unsigned long		rs_locked;
700 	unsigned long		rs_flushing;
701 	unsigned long		rs_logging;
702 
703 	__u32			rs_handle_count;
704 	__u32			rs_blocks;
705 	__u32			rs_blocks_logged;
706 };
707 
708 struct transaction_stats_s {
709 	unsigned long		ts_tid;
710 	unsigned long		ts_requested;
711 	struct transaction_run_stats_s run;
712 };
713 
714 static inline unsigned long
715 jbd2_time_diff(unsigned long start, unsigned long end)
716 {
717 	if (end >= start)
718 		return end - start;
719 
720 	return end + (MAX_JIFFY_OFFSET - start);
721 }
722 
723 #define JBD2_NR_BATCH	64
724 
725 /**
726  * struct journal_s - The journal_s type is the concrete type associated with
727  *     journal_t.
728  * @j_flags:  General journaling state flags
729  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
730  *     prior abort)?
731  * @j_sb_buffer: First part of superblock buffer
732  * @j_superblock: Second part of superblock buffer
733  * @j_format_version: Version of the superblock format
734  * @j_state_lock: Protect the various scalars in the journal
735  * @j_barrier_count:  Number of processes waiting to create a barrier lock
736  * @j_barrier: The barrier lock itself
737  * @j_running_transaction: The current running transaction..
738  * @j_committing_transaction: the transaction we are pushing to disk
739  * @j_checkpoint_transactions: a linked circular list of all transactions
740  *  waiting for checkpointing
741  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
742  *  to start committing, or for a barrier lock to be released
743  * @j_wait_done_commit: Wait queue for waiting for commit to complete
744  * @j_wait_commit: Wait queue to trigger commit
745  * @j_wait_updates: Wait queue to wait for updates to complete
746  * @j_wait_reserved: Wait queue to wait for reserved buffer credits to drop
747  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
748  * @j_head: Journal head - identifies the first unused block in the journal
749  * @j_tail: Journal tail - identifies the oldest still-used block in the
750  *  journal.
751  * @j_free: Journal free - how many free blocks are there in the journal?
752  * @j_first: The block number of the first usable block
753  * @j_last: The block number one beyond the last usable block
754  * @j_dev: Device where we store the journal
755  * @j_blocksize: blocksize for the location where we store the journal.
756  * @j_blk_offset: starting block offset for into the device where we store the
757  *     journal
758  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
759  *     be equal to j_dev
760  * @j_reserved_credits: Number of buffers reserved from the running transaction
761  * @j_maxlen: Total maximum capacity of the journal region on disk.
762  * @j_list_lock: Protects the buffer lists and internal buffer state.
763  * @j_inode: Optional inode where we store the journal.  If present, all journal
764  *     block numbers are mapped into this inode via bmap().
765  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
766  * @j_transaction_sequence: Sequence number of the next transaction to grant
767  * @j_commit_sequence: Sequence number of the most recently committed
768  *  transaction
769  * @j_commit_request: Sequence number of the most recent transaction wanting
770  *     commit
771  * @j_uuid: Uuid of client object.
772  * @j_task: Pointer to the current commit thread for this journal
773  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
774  *     single compound commit transaction
775  * @j_commit_interval: What is the maximum transaction lifetime before we begin
776  *  a commit?
777  * @j_commit_timer:  The timer used to wakeup the commit thread
778  * @j_revoke_lock: Protect the revoke table
779  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
780  *     current transaction.
781  * @j_revoke_table: alternate revoke tables for j_revoke
782  * @j_wbuf: array of buffer_heads for jbd2_journal_commit_transaction
783  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
784  *	number that will fit in j_blocksize
785  * @j_last_sync_writer: most recent pid which did a synchronous write
786  * @j_history: Buffer storing the transactions statistics history
787  * @j_history_max: Maximum number of transactions in the statistics history
788  * @j_history_cur: Current number of transactions in the statistics history
789  * @j_history_lock: Protect the transactions statistics history
790  * @j_proc_entry: procfs entry for the jbd statistics directory
791  * @j_stats: Overall statistics
792  * @j_private: An opaque pointer to fs-private information.
793  */
794 
795 struct journal_s
796 {
797 	/* General journaling state flags [j_state_lock] */
798 	unsigned long		j_flags;
799 
800 	/*
801 	 * Is there an outstanding uncleared error on the journal (from a prior
802 	 * abort)? [j_state_lock]
803 	 */
804 	int			j_errno;
805 
806 	/* The superblock buffer */
807 	struct buffer_head	*j_sb_buffer;
808 	journal_superblock_t	*j_superblock;
809 
810 	/* Version of the superblock format */
811 	int			j_format_version;
812 
813 	/*
814 	 * Protect the various scalars in the journal
815 	 */
816 	rwlock_t		j_state_lock;
817 
818 	/*
819 	 * Number of processes waiting to create a barrier lock [j_state_lock]
820 	 */
821 	int			j_barrier_count;
822 
823 	/* The barrier lock itself */
824 	struct mutex		j_barrier;
825 
826 	/*
827 	 * Transactions: The current running transaction...
828 	 * [j_state_lock] [caller holding open handle]
829 	 */
830 	transaction_t		*j_running_transaction;
831 
832 	/*
833 	 * the transaction we are pushing to disk
834 	 * [j_state_lock] [caller holding open handle]
835 	 */
836 	transaction_t		*j_committing_transaction;
837 
838 	/*
839 	 * ... and a linked circular list of all transactions waiting for
840 	 * checkpointing. [j_list_lock]
841 	 */
842 	transaction_t		*j_checkpoint_transactions;
843 
844 	/*
845 	 * Wait queue for waiting for a locked transaction to start committing,
846 	 * or for a barrier lock to be released
847 	 */
848 	wait_queue_head_t	j_wait_transaction_locked;
849 
850 	/* Wait queue for waiting for commit to complete */
851 	wait_queue_head_t	j_wait_done_commit;
852 
853 	/* Wait queue to trigger commit */
854 	wait_queue_head_t	j_wait_commit;
855 
856 	/* Wait queue to wait for updates to complete */
857 	wait_queue_head_t	j_wait_updates;
858 
859 	/* Wait queue to wait for reserved buffer credits to drop */
860 	wait_queue_head_t	j_wait_reserved;
861 
862 	/* Semaphore for locking against concurrent checkpoints */
863 	struct mutex		j_checkpoint_mutex;
864 
865 	/*
866 	 * List of buffer heads used by the checkpoint routine.  This
867 	 * was moved from jbd2_log_do_checkpoint() to reduce stack
868 	 * usage.  Access to this array is controlled by the
869 	 * j_checkpoint_mutex.  [j_checkpoint_mutex]
870 	 */
871 	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
872 
873 	/*
874 	 * Journal head: identifies the first unused block in the journal.
875 	 * [j_state_lock]
876 	 */
877 	unsigned long		j_head;
878 
879 	/*
880 	 * Journal tail: identifies the oldest still-used block in the journal.
881 	 * [j_state_lock]
882 	 */
883 	unsigned long		j_tail;
884 
885 	/*
886 	 * Journal free: how many free blocks are there in the journal?
887 	 * [j_state_lock]
888 	 */
889 	unsigned long		j_free;
890 
891 	/*
892 	 * Journal start and end: the block numbers of the first usable block
893 	 * and one beyond the last usable block in the journal. [j_state_lock]
894 	 */
895 	unsigned long		j_first;
896 	unsigned long		j_last;
897 
898 	/*
899 	 * Device, blocksize and starting block offset for the location where we
900 	 * store the journal.
901 	 */
902 	struct block_device	*j_dev;
903 	int			j_blocksize;
904 	unsigned long long	j_blk_offset;
905 	char			j_devname[BDEVNAME_SIZE+24];
906 
907 	/*
908 	 * Device which holds the client fs.  For internal journal this will be
909 	 * equal to j_dev.
910 	 */
911 	struct block_device	*j_fs_dev;
912 
913 	/* Total maximum capacity of the journal region on disk. */
914 	unsigned int		j_maxlen;
915 
916 	/* Number of buffers reserved from the running transaction */
917 	atomic_t		j_reserved_credits;
918 
919 	/*
920 	 * Protects the buffer lists and internal buffer state.
921 	 */
922 	spinlock_t		j_list_lock;
923 
924 	/* Optional inode where we store the journal.  If present, all */
925 	/* journal block numbers are mapped into this inode via */
926 	/* bmap(). */
927 	struct inode		*j_inode;
928 
929 	/*
930 	 * Sequence number of the oldest transaction in the log [j_state_lock]
931 	 */
932 	tid_t			j_tail_sequence;
933 
934 	/*
935 	 * Sequence number of the next transaction to grant [j_state_lock]
936 	 */
937 	tid_t			j_transaction_sequence;
938 
939 	/*
940 	 * Sequence number of the most recently committed transaction
941 	 * [j_state_lock].
942 	 */
943 	tid_t			j_commit_sequence;
944 
945 	/*
946 	 * Sequence number of the most recent transaction wanting commit
947 	 * [j_state_lock]
948 	 */
949 	tid_t			j_commit_request;
950 
951 	/*
952 	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
953 	 * backed by this journal.  This will eventually be replaced by an array
954 	 * of uuids, allowing us to index multiple devices within a single
955 	 * journal and to perform atomic updates across them.
956 	 */
957 	__u8			j_uuid[16];
958 
959 	/* Pointer to the current commit thread for this journal */
960 	struct task_struct	*j_task;
961 
962 	/*
963 	 * Maximum number of metadata buffers to allow in a single compound
964 	 * commit transaction
965 	 */
966 	int			j_max_transaction_buffers;
967 
968 	/*
969 	 * What is the maximum transaction lifetime before we begin a commit?
970 	 */
971 	unsigned long		j_commit_interval;
972 
973 	/* The timer used to wakeup the commit thread: */
974 	struct timer_list	j_commit_timer;
975 
976 	/*
977 	 * The revoke table: maintains the list of revoked blocks in the
978 	 * current transaction.  [j_revoke_lock]
979 	 */
980 	spinlock_t		j_revoke_lock;
981 	struct jbd2_revoke_table_s *j_revoke;
982 	struct jbd2_revoke_table_s *j_revoke_table[2];
983 
984 	/*
985 	 * array of bhs for jbd2_journal_commit_transaction
986 	 */
987 	struct buffer_head	**j_wbuf;
988 	int			j_wbufsize;
989 
990 	/*
991 	 * this is the pid of hte last person to run a synchronous operation
992 	 * through the journal
993 	 */
994 	pid_t			j_last_sync_writer;
995 
996 	/*
997 	 * the average amount of time in nanoseconds it takes to commit a
998 	 * transaction to disk. [j_state_lock]
999 	 */
1000 	u64			j_average_commit_time;
1001 
1002 	/*
1003 	 * minimum and maximum times that we should wait for
1004 	 * additional filesystem operations to get batched into a
1005 	 * synchronous handle in microseconds
1006 	 */
1007 	u32			j_min_batch_time;
1008 	u32			j_max_batch_time;
1009 
1010 	/* This function is called when a transaction is closed */
1011 	void			(*j_commit_callback)(journal_t *,
1012 						     transaction_t *);
1013 
1014 	/*
1015 	 * Journal statistics
1016 	 */
1017 	spinlock_t		j_history_lock;
1018 	struct proc_dir_entry	*j_proc_entry;
1019 	struct transaction_stats_s j_stats;
1020 
1021 	/* Failed journal commit ID */
1022 	unsigned int		j_failed_commit;
1023 
1024 	/*
1025 	 * An opaque pointer to fs-private information.  ext3 puts its
1026 	 * superblock pointer here
1027 	 */
1028 	void *j_private;
1029 
1030 	/* Reference to checksum algorithm driver via cryptoapi */
1031 	struct crypto_shash *j_chksum_driver;
1032 
1033 	/* Precomputed journal UUID checksum for seeding other checksums */
1034 	__u32 j_csum_seed;
1035 };
1036 
1037 /*
1038  * Journal flag definitions
1039  */
1040 #define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1041 #define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1042 #define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1043 #define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1044 #define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1045 #define JBD2_BARRIER	0x020	/* Use IDE barriers */
1046 #define JBD2_ABORT_ON_SYNCDATA_ERR	0x040	/* Abort the journal on file
1047 						 * data write error in ordered
1048 						 * mode */
1049 
1050 /*
1051  * Function declarations for the journaling transaction and buffer
1052  * management
1053  */
1054 
1055 /* Filing buffers */
1056 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1057 extern void __jbd2_journal_refile_buffer(struct journal_head *);
1058 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1059 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1060 extern void __journal_free_buffer(struct journal_head *bh);
1061 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1062 extern void __journal_clean_data_list(transaction_t *transaction);
1063 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1064 {
1065 	list_add_tail(&bh->b_assoc_buffers, head);
1066 }
1067 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1068 {
1069 	list_del_init(&bh->b_assoc_buffers);
1070 }
1071 
1072 /* Log buffer allocation */
1073 struct buffer_head *jbd2_journal_get_descriptor_buffer(journal_t *journal);
1074 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1075 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1076 			      unsigned long *block);
1077 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1078 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1079 
1080 /* Commit management */
1081 extern void jbd2_journal_commit_transaction(journal_t *);
1082 
1083 /* Checkpoint list management */
1084 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1085 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1086 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1087 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1088 
1089 
1090 /*
1091  * Triggers
1092  */
1093 
1094 struct jbd2_buffer_trigger_type {
1095 	/*
1096 	 * Fired a the moment data to write to the journal are known to be
1097 	 * stable - so either at the moment b_frozen_data is created or just
1098 	 * before a buffer is written to the journal.  mapped_data is a mapped
1099 	 * buffer that is the frozen data for commit.
1100 	 */
1101 	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1102 			 struct buffer_head *bh, void *mapped_data,
1103 			 size_t size);
1104 
1105 	/*
1106 	 * Fired during journal abort for dirty buffers that will not be
1107 	 * committed.
1108 	 */
1109 	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1110 			struct buffer_head *bh);
1111 };
1112 
1113 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1114 				       void *mapped_data,
1115 				       struct jbd2_buffer_trigger_type *triggers);
1116 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1117 				      struct jbd2_buffer_trigger_type *triggers);
1118 
1119 /* Buffer IO */
1120 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1121 					      struct journal_head *jh_in,
1122 					      struct buffer_head **bh_out,
1123 					      sector_t blocknr);
1124 
1125 /* Transaction locking */
1126 extern void		__wait_on_journal (journal_t *);
1127 
1128 /* Transaction cache support */
1129 extern void jbd2_journal_destroy_transaction_cache(void);
1130 extern int  jbd2_journal_init_transaction_cache(void);
1131 extern void jbd2_journal_free_transaction(transaction_t *);
1132 
1133 /*
1134  * Journal locking.
1135  *
1136  * We need to lock the journal during transaction state changes so that nobody
1137  * ever tries to take a handle on the running transaction while we are in the
1138  * middle of moving it to the commit phase.  j_state_lock does this.
1139  *
1140  * Note that the locking is completely interrupt unsafe.  We never touch
1141  * journal structures from interrupts.
1142  */
1143 
1144 static inline handle_t *journal_current_handle(void)
1145 {
1146 	return current->journal_info;
1147 }
1148 
1149 /* The journaling code user interface:
1150  *
1151  * Create and destroy handles
1152  * Register buffer modifications against the current transaction.
1153  */
1154 
1155 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1156 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1157 				     gfp_t gfp_mask, unsigned int type,
1158 				     unsigned int line_no);
1159 extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1160 extern int	 jbd2__journal_restart(handle_t *, int nblocks, gfp_t gfp_mask);
1161 extern int	 jbd2_journal_start_reserved(handle_t *handle,
1162 				unsigned int type, unsigned int line_no);
1163 extern void	 jbd2_journal_free_reserved(handle_t *handle);
1164 extern int	 jbd2_journal_extend (handle_t *, int nblocks);
1165 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1166 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1167 extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1168 void		 jbd2_journal_set_triggers(struct buffer_head *,
1169 					   struct jbd2_buffer_trigger_type *type);
1170 extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1171 extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1172 extern void	 journal_sync_buffer (struct buffer_head *);
1173 extern int	 jbd2_journal_invalidatepage(journal_t *,
1174 				struct page *, unsigned int, unsigned int);
1175 extern int	 jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1176 extern int	 jbd2_journal_stop(handle_t *);
1177 extern int	 jbd2_journal_flush (journal_t *);
1178 extern void	 jbd2_journal_lock_updates (journal_t *);
1179 extern void	 jbd2_journal_unlock_updates (journal_t *);
1180 
1181 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1182 				struct block_device *fs_dev,
1183 				unsigned long long start, int len, int bsize);
1184 extern journal_t * jbd2_journal_init_inode (struct inode *);
1185 extern int	   jbd2_journal_update_format (journal_t *);
1186 extern int	   jbd2_journal_check_used_features
1187 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1188 extern int	   jbd2_journal_check_available_features
1189 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1190 extern int	   jbd2_journal_set_features
1191 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1192 extern void	   jbd2_journal_clear_features
1193 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1194 extern int	   jbd2_journal_load       (journal_t *journal);
1195 extern int	   jbd2_journal_destroy    (journal_t *);
1196 extern int	   jbd2_journal_recover    (journal_t *journal);
1197 extern int	   jbd2_journal_wipe       (journal_t *, int);
1198 extern int	   jbd2_journal_skip_recovery	(journal_t *);
1199 extern void	   jbd2_journal_update_sb_errno(journal_t *);
1200 extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1201 				unsigned long, int);
1202 extern void	   __jbd2_journal_abort_hard	(journal_t *);
1203 extern void	   jbd2_journal_abort      (journal_t *, int);
1204 extern int	   jbd2_journal_errno      (journal_t *);
1205 extern void	   jbd2_journal_ack_err    (journal_t *);
1206 extern int	   jbd2_journal_clear_err  (journal_t *);
1207 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1208 extern int	   jbd2_journal_force_commit(journal_t *);
1209 extern int	   jbd2_journal_force_commit_nested(journal_t *);
1210 extern int	   jbd2_journal_file_inode(handle_t *handle, struct jbd2_inode *inode);
1211 extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1212 				struct jbd2_inode *inode, loff_t new_size);
1213 extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1214 extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1215 
1216 /*
1217  * journal_head management
1218  */
1219 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1220 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1221 void jbd2_journal_put_journal_head(struct journal_head *jh);
1222 
1223 /*
1224  * handle management
1225  */
1226 extern struct kmem_cache *jbd2_handle_cache;
1227 
1228 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1229 {
1230 	return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1231 }
1232 
1233 static inline void jbd2_free_handle(handle_t *handle)
1234 {
1235 	kmem_cache_free(jbd2_handle_cache, handle);
1236 }
1237 
1238 /*
1239  * jbd2_inode management (optional, for those file systems that want to use
1240  * dynamically allocated jbd2_inode structures)
1241  */
1242 extern struct kmem_cache *jbd2_inode_cache;
1243 
1244 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1245 {
1246 	return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1247 }
1248 
1249 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1250 {
1251 	kmem_cache_free(jbd2_inode_cache, jinode);
1252 }
1253 
1254 /* Primary revoke support */
1255 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1256 extern int	   jbd2_journal_init_revoke(journal_t *, int);
1257 extern void	   jbd2_journal_destroy_revoke_caches(void);
1258 extern int	   jbd2_journal_init_revoke_caches(void);
1259 
1260 extern void	   jbd2_journal_destroy_revoke(journal_t *);
1261 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1262 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1263 extern void	   jbd2_journal_write_revoke_records(journal_t *journal,
1264 						     transaction_t *transaction,
1265 						     struct list_head *log_bufs,
1266 						     int write_op);
1267 
1268 /* Recovery revoke support */
1269 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1270 extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1271 extern void	jbd2_journal_clear_revoke(journal_t *);
1272 extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1273 extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1274 
1275 /*
1276  * The log thread user interface:
1277  *
1278  * Request space in the current transaction, and force transaction commit
1279  * transitions on demand.
1280  */
1281 
1282 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1283 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1284 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1285 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1286 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1287 int jbd2_log_do_checkpoint(journal_t *journal);
1288 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1289 
1290 void __jbd2_log_wait_for_space(journal_t *journal);
1291 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1292 extern int jbd2_cleanup_journal_tail(journal_t *);
1293 
1294 /*
1295  * is_journal_abort
1296  *
1297  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1298  * bit, when set, indicates that we have had a fatal error somewhere,
1299  * either inside the journaling layer or indicated to us by the client
1300  * (eg. ext3), and that we and should not commit any further
1301  * transactions.
1302  */
1303 
1304 static inline int is_journal_aborted(journal_t *journal)
1305 {
1306 	return journal->j_flags & JBD2_ABORT;
1307 }
1308 
1309 static inline int is_handle_aborted(handle_t *handle)
1310 {
1311 	if (handle->h_aborted || !handle->h_transaction)
1312 		return 1;
1313 	return is_journal_aborted(handle->h_transaction->t_journal);
1314 }
1315 
1316 static inline void jbd2_journal_abort_handle(handle_t *handle)
1317 {
1318 	handle->h_aborted = 1;
1319 }
1320 
1321 #endif /* __KERNEL__   */
1322 
1323 /* Comparison functions for transaction IDs: perform comparisons using
1324  * modulo arithmetic so that they work over sequence number wraps. */
1325 
1326 static inline int tid_gt(tid_t x, tid_t y)
1327 {
1328 	int difference = (x - y);
1329 	return (difference > 0);
1330 }
1331 
1332 static inline int tid_geq(tid_t x, tid_t y)
1333 {
1334 	int difference = (x - y);
1335 	return (difference >= 0);
1336 }
1337 
1338 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1339 extern size_t journal_tag_bytes(journal_t *journal);
1340 
1341 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1342 {
1343 	if (JBD2_HAS_INCOMPAT_FEATURE(journal, JBD2_FEATURE_INCOMPAT_CSUM_V2) ||
1344 	    JBD2_HAS_INCOMPAT_FEATURE(journal, JBD2_FEATURE_INCOMPAT_CSUM_V3))
1345 		return 1;
1346 
1347 	return 0;
1348 }
1349 
1350 /*
1351  * We reserve t_outstanding_credits >> JBD2_CONTROL_BLOCKS_SHIFT for
1352  * transaction control blocks.
1353  */
1354 #define JBD2_CONTROL_BLOCKS_SHIFT 5
1355 
1356 /*
1357  * Return the minimum number of blocks which must be free in the journal
1358  * before a new transaction may be started.  Must be called under j_state_lock.
1359  */
1360 static inline int jbd2_space_needed(journal_t *journal)
1361 {
1362 	int nblocks = journal->j_max_transaction_buffers;
1363 	return nblocks + (nblocks >> JBD2_CONTROL_BLOCKS_SHIFT);
1364 }
1365 
1366 /*
1367  * Return number of free blocks in the log. Must be called under j_state_lock.
1368  */
1369 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1370 {
1371 	/* Allow for rounding errors */
1372 	unsigned long free = journal->j_free - 32;
1373 
1374 	if (journal->j_committing_transaction) {
1375 		unsigned long committing = atomic_read(&journal->
1376 			j_committing_transaction->t_outstanding_credits);
1377 
1378 		/* Transaction + control blocks */
1379 		free -= committing + (committing >> JBD2_CONTROL_BLOCKS_SHIFT);
1380 	}
1381 	return free;
1382 }
1383 
1384 /*
1385  * Definitions which augment the buffer_head layer
1386  */
1387 
1388 /* journaling buffer types */
1389 #define BJ_None		0	/* Not journaled */
1390 #define BJ_Metadata	1	/* Normal journaled metadata */
1391 #define BJ_Forget	2	/* Buffer superseded by this transaction */
1392 #define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1393 #define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1394 #define BJ_Types	5
1395 
1396 extern int jbd_blocks_per_page(struct inode *inode);
1397 
1398 /* JBD uses a CRC32 checksum */
1399 #define JBD_MAX_CHECKSUM_SIZE 4
1400 
1401 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1402 			      const void *address, unsigned int length)
1403 {
1404 	struct {
1405 		struct shash_desc shash;
1406 		char ctx[JBD_MAX_CHECKSUM_SIZE];
1407 	} desc;
1408 	int err;
1409 
1410 	BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1411 		JBD_MAX_CHECKSUM_SIZE);
1412 
1413 	desc.shash.tfm = journal->j_chksum_driver;
1414 	desc.shash.flags = 0;
1415 	*(u32 *)desc.ctx = crc;
1416 
1417 	err = crypto_shash_update(&desc.shash, address, length);
1418 	BUG_ON(err);
1419 
1420 	return *(u32 *)desc.ctx;
1421 }
1422 
1423 /* Return most recent uncommitted transaction */
1424 static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1425 {
1426 	tid_t tid;
1427 
1428 	read_lock(&journal->j_state_lock);
1429 	tid = journal->j_commit_request;
1430 	if (journal->j_running_transaction)
1431 		tid = journal->j_running_transaction->t_tid;
1432 	read_unlock(&journal->j_state_lock);
1433 	return tid;
1434 }
1435 
1436 #ifdef __KERNEL__
1437 
1438 #define buffer_trace_init(bh)	do {} while (0)
1439 #define print_buffer_fields(bh)	do {} while (0)
1440 #define print_buffer_trace(bh)	do {} while (0)
1441 #define BUFFER_TRACE(bh, info)	do {} while (0)
1442 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1443 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1444 
1445 #endif	/* __KERNEL__ */
1446 
1447 #endif	/* _LINUX_JBD2_H */
1448