xref: /linux/include/linux/jbd2.h (revision 37744feebc086908fd89760650f458ab19071750)
1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * linux/include/linux/jbd2.h
4  *
5  * Written by Stephen C. Tweedie <sct@redhat.com>
6  *
7  * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
8  *
9  * Definitions for transaction data structures for the buffer cache
10  * filesystem journaling support.
11  */
12 
13 #ifndef _LINUX_JBD2_H
14 #define _LINUX_JBD2_H
15 
16 /* Allow this file to be included directly into e2fsprogs */
17 #ifndef __KERNEL__
18 #include "jfs_compat.h"
19 #define JBD2_DEBUG
20 #else
21 
22 #include <linux/types.h>
23 #include <linux/buffer_head.h>
24 #include <linux/journal-head.h>
25 #include <linux/stddef.h>
26 #include <linux/mutex.h>
27 #include <linux/timer.h>
28 #include <linux/slab.h>
29 #include <linux/bit_spinlock.h>
30 #include <crypto/hash.h>
31 #endif
32 
33 #define journal_oom_retry 1
34 
35 /*
36  * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
37  * certain classes of error which can occur due to failed IOs.  Under
38  * normal use we want ext4 to continue after such errors, because
39  * hardware _can_ fail, but for debugging purposes when running tests on
40  * known-good hardware we may want to trap these errors.
41  */
42 #undef JBD2_PARANOID_IOFAIL
43 
44 /*
45  * The default maximum commit age, in seconds.
46  */
47 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
48 
49 #ifdef CONFIG_JBD2_DEBUG
50 /*
51  * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
52  * consistency checks.  By default we don't do this unless
53  * CONFIG_JBD2_DEBUG is on.
54  */
55 #define JBD2_EXPENSIVE_CHECKING
56 extern ushort jbd2_journal_enable_debug;
57 void __jbd2_debug(int level, const char *file, const char *func,
58 		  unsigned int line, const char *fmt, ...);
59 
60 #define jbd_debug(n, fmt, a...) \
61 	__jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
62 #else
63 #define jbd_debug(n, fmt, a...)    /**/
64 #endif
65 
66 extern void *jbd2_alloc(size_t size, gfp_t flags);
67 extern void jbd2_free(void *ptr, size_t size);
68 
69 #define JBD2_MIN_JOURNAL_BLOCKS 1024
70 
71 #ifdef __KERNEL__
72 
73 /**
74  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
75  *
76  * All filesystem modifications made by the process go
77  * through this handle.  Recursive operations (such as quota operations)
78  * are gathered into a single update.
79  *
80  * The buffer credits field is used to account for journaled buffers
81  * being modified by the running process.  To ensure that there is
82  * enough log space for all outstanding operations, we need to limit the
83  * number of outstanding buffers possible at any time.  When the
84  * operation completes, any buffer credits not used are credited back to
85  * the transaction, so that at all times we know how many buffers the
86  * outstanding updates on a transaction might possibly touch.
87  *
88  * This is an opaque datatype.
89  **/
90 typedef struct jbd2_journal_handle handle_t;	/* Atomic operation type */
91 
92 
93 /**
94  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
95  *
96  * journal_t is linked to from the fs superblock structure.
97  *
98  * We use the journal_t to keep track of all outstanding transaction
99  * activity on the filesystem, and to manage the state of the log
100  * writing process.
101  *
102  * This is an opaque datatype.
103  **/
104 typedef struct journal_s	journal_t;	/* Journal control structure */
105 #endif
106 
107 /*
108  * Internal structures used by the logging mechanism:
109  */
110 
111 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
112 
113 /*
114  * On-disk structures
115  */
116 
117 /*
118  * Descriptor block types:
119  */
120 
121 #define JBD2_DESCRIPTOR_BLOCK	1
122 #define JBD2_COMMIT_BLOCK	2
123 #define JBD2_SUPERBLOCK_V1	3
124 #define JBD2_SUPERBLOCK_V2	4
125 #define JBD2_REVOKE_BLOCK	5
126 
127 /*
128  * Standard header for all descriptor blocks:
129  */
130 typedef struct journal_header_s
131 {
132 	__be32		h_magic;
133 	__be32		h_blocktype;
134 	__be32		h_sequence;
135 } journal_header_t;
136 
137 /*
138  * Checksum types.
139  */
140 #define JBD2_CRC32_CHKSUM   1
141 #define JBD2_MD5_CHKSUM     2
142 #define JBD2_SHA1_CHKSUM    3
143 #define JBD2_CRC32C_CHKSUM  4
144 
145 #define JBD2_CRC32_CHKSUM_SIZE 4
146 
147 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
148 /*
149  * Commit block header for storing transactional checksums:
150  *
151  * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
152  * fields are used to store a checksum of the descriptor and data blocks.
153  *
154  * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
155  * field is used to store crc32c(uuid+commit_block).  Each journal metadata
156  * block gets its own checksum, and data block checksums are stored in
157  * journal_block_tag (in the descriptor).  The other h_chksum* fields are
158  * not used.
159  *
160  * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
161  * journal_block_tag3_t to store a full 32-bit checksum.  Everything else
162  * is the same as v2.
163  *
164  * Checksum v1, v2, and v3 are mutually exclusive features.
165  */
166 struct commit_header {
167 	__be32		h_magic;
168 	__be32          h_blocktype;
169 	__be32          h_sequence;
170 	unsigned char   h_chksum_type;
171 	unsigned char   h_chksum_size;
172 	unsigned char 	h_padding[2];
173 	__be32 		h_chksum[JBD2_CHECKSUM_BYTES];
174 	__be64		h_commit_sec;
175 	__be32		h_commit_nsec;
176 };
177 
178 /*
179  * The block tag: used to describe a single buffer in the journal.
180  * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
181  * raw struct shouldn't be used for pointer math or sizeof() - use
182  * journal_tag_bytes(journal) instead to compute this.
183  */
184 typedef struct journal_block_tag3_s
185 {
186 	__be32		t_blocknr;	/* The on-disk block number */
187 	__be32		t_flags;	/* See below */
188 	__be32		t_blocknr_high; /* most-significant high 32bits. */
189 	__be32		t_checksum;	/* crc32c(uuid+seq+block) */
190 } journal_block_tag3_t;
191 
192 typedef struct journal_block_tag_s
193 {
194 	__be32		t_blocknr;	/* The on-disk block number */
195 	__be16		t_checksum;	/* truncated crc32c(uuid+seq+block) */
196 	__be16		t_flags;	/* See below */
197 	__be32		t_blocknr_high; /* most-significant high 32bits. */
198 } journal_block_tag_t;
199 
200 /* Tail of descriptor or revoke block, for checksumming */
201 struct jbd2_journal_block_tail {
202 	__be32		t_checksum;	/* crc32c(uuid+descr_block) */
203 };
204 
205 /*
206  * The revoke descriptor: used on disk to describe a series of blocks to
207  * be revoked from the log
208  */
209 typedef struct jbd2_journal_revoke_header_s
210 {
211 	journal_header_t r_header;
212 	__be32		 r_count;	/* Count of bytes used in the block */
213 } jbd2_journal_revoke_header_t;
214 
215 /* Definitions for the journal tag flags word: */
216 #define JBD2_FLAG_ESCAPE		1	/* on-disk block is escaped */
217 #define JBD2_FLAG_SAME_UUID	2	/* block has same uuid as previous */
218 #define JBD2_FLAG_DELETED	4	/* block deleted by this transaction */
219 #define JBD2_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
220 
221 
222 /*
223  * The journal superblock.  All fields are in big-endian byte order.
224  */
225 typedef struct journal_superblock_s
226 {
227 /* 0x0000 */
228 	journal_header_t s_header;
229 
230 /* 0x000C */
231 	/* Static information describing the journal */
232 	__be32	s_blocksize;		/* journal device blocksize */
233 	__be32	s_maxlen;		/* total blocks in journal file */
234 	__be32	s_first;		/* first block of log information */
235 
236 /* 0x0018 */
237 	/* Dynamic information describing the current state of the log */
238 	__be32	s_sequence;		/* first commit ID expected in log */
239 	__be32	s_start;		/* blocknr of start of log */
240 
241 /* 0x0020 */
242 	/* Error value, as set by jbd2_journal_abort(). */
243 	__be32	s_errno;
244 
245 /* 0x0024 */
246 	/* Remaining fields are only valid in a version-2 superblock */
247 	__be32	s_feature_compat;	/* compatible feature set */
248 	__be32	s_feature_incompat;	/* incompatible feature set */
249 	__be32	s_feature_ro_compat;	/* readonly-compatible feature set */
250 /* 0x0030 */
251 	__u8	s_uuid[16];		/* 128-bit uuid for journal */
252 
253 /* 0x0040 */
254 	__be32	s_nr_users;		/* Nr of filesystems sharing log */
255 
256 	__be32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
257 
258 /* 0x0048 */
259 	__be32	s_max_transaction;	/* Limit of journal blocks per trans.*/
260 	__be32	s_max_trans_data;	/* Limit of data blocks per trans. */
261 
262 /* 0x0050 */
263 	__u8	s_checksum_type;	/* checksum type */
264 	__u8	s_padding2[3];
265 	__u32	s_padding[42];
266 	__be32	s_checksum;		/* crc32c(superblock) */
267 
268 /* 0x0100 */
269 	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
270 /* 0x0400 */
271 } journal_superblock_t;
272 
273 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
274 #define JBD2_HAS_COMPAT_FEATURE(j,mask)					\
275 	((j)->j_format_version >= 2 &&					\
276 	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
277 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask)				\
278 	((j)->j_format_version >= 2 &&					\
279 	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
280 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask)				\
281 	((j)->j_format_version >= 2 &&					\
282 	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
283 
284 #define JBD2_FEATURE_COMPAT_CHECKSUM		0x00000001
285 
286 #define JBD2_FEATURE_INCOMPAT_REVOKE		0x00000001
287 #define JBD2_FEATURE_INCOMPAT_64BIT		0x00000002
288 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
289 #define JBD2_FEATURE_INCOMPAT_CSUM_V2		0x00000008
290 #define JBD2_FEATURE_INCOMPAT_CSUM_V3		0x00000010
291 
292 /* See "journal feature predicate functions" below */
293 
294 /* Features known to this kernel version: */
295 #define JBD2_KNOWN_COMPAT_FEATURES	JBD2_FEATURE_COMPAT_CHECKSUM
296 #define JBD2_KNOWN_ROCOMPAT_FEATURES	0
297 #define JBD2_KNOWN_INCOMPAT_FEATURES	(JBD2_FEATURE_INCOMPAT_REVOKE | \
298 					JBD2_FEATURE_INCOMPAT_64BIT | \
299 					JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
300 					JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
301 					JBD2_FEATURE_INCOMPAT_CSUM_V3)
302 
303 #ifdef __KERNEL__
304 
305 #include <linux/fs.h>
306 #include <linux/sched.h>
307 
308 enum jbd_state_bits {
309 	BH_JBD			/* Has an attached ext3 journal_head */
310 	  = BH_PrivateStart,
311 	BH_JWrite,		/* Being written to log (@@@ DEBUGGING) */
312 	BH_Freed,		/* Has been freed (truncated) */
313 	BH_Revoked,		/* Has been revoked from the log */
314 	BH_RevokeValid,		/* Revoked flag is valid */
315 	BH_JBDDirty,		/* Is dirty but journaled */
316 	BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
317 	BH_Shadow,		/* IO on shadow buffer is running */
318 	BH_Verified,		/* Metadata block has been verified ok */
319 	BH_JBDPrivateStart,	/* First bit available for private use by FS */
320 };
321 
322 BUFFER_FNS(JBD, jbd)
323 BUFFER_FNS(JWrite, jwrite)
324 BUFFER_FNS(JBDDirty, jbddirty)
325 TAS_BUFFER_FNS(JBDDirty, jbddirty)
326 BUFFER_FNS(Revoked, revoked)
327 TAS_BUFFER_FNS(Revoked, revoked)
328 BUFFER_FNS(RevokeValid, revokevalid)
329 TAS_BUFFER_FNS(RevokeValid, revokevalid)
330 BUFFER_FNS(Freed, freed)
331 BUFFER_FNS(Shadow, shadow)
332 BUFFER_FNS(Verified, verified)
333 
334 static inline struct buffer_head *jh2bh(struct journal_head *jh)
335 {
336 	return jh->b_bh;
337 }
338 
339 static inline struct journal_head *bh2jh(struct buffer_head *bh)
340 {
341 	return bh->b_private;
342 }
343 
344 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
345 {
346 	bit_spin_lock(BH_JournalHead, &bh->b_state);
347 }
348 
349 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
350 {
351 	bit_spin_unlock(BH_JournalHead, &bh->b_state);
352 }
353 
354 #define J_ASSERT(assert)	BUG_ON(!(assert))
355 
356 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
357 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
358 
359 #if defined(JBD2_PARANOID_IOFAIL)
360 #define J_EXPECT(expr, why...)		J_ASSERT(expr)
361 #define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
362 #define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
363 #else
364 #define __journal_expect(expr, why...)					     \
365 	({								     \
366 		int val = (expr);					     \
367 		if (!val) {						     \
368 			printk(KERN_ERR					     \
369 			       "JBD2 unexpected failure: %s: %s;\n",	     \
370 			       __func__, #expr);			     \
371 			printk(KERN_ERR why "\n");			     \
372 		}							     \
373 		val;							     \
374 	})
375 #define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
376 #define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
377 #define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
378 #endif
379 
380 /* Flags in jbd_inode->i_flags */
381 #define __JI_COMMIT_RUNNING 0
382 #define __JI_WRITE_DATA 1
383 #define __JI_WAIT_DATA 2
384 
385 /*
386  * Commit of the inode data in progress. We use this flag to protect us from
387  * concurrent deletion of inode. We cannot use reference to inode for this
388  * since we cannot afford doing last iput() on behalf of kjournald
389  */
390 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
391 /* Write allocated dirty buffers in this inode before commit */
392 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
393 /* Wait for outstanding data writes for this inode before commit */
394 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
395 
396 /**
397  * struct jbd_inode - The jbd_inode type is the structure linking inodes in
398  * ordered mode present in a transaction so that we can sync them during commit.
399  */
400 struct jbd2_inode {
401 	/**
402 	 * @i_transaction:
403 	 *
404 	 * Which transaction does this inode belong to? Either the running
405 	 * transaction or the committing one. [j_list_lock]
406 	 */
407 	transaction_t *i_transaction;
408 
409 	/**
410 	 * @i_next_transaction:
411 	 *
412 	 * Pointer to the running transaction modifying inode's data in case
413 	 * there is already a committing transaction touching it. [j_list_lock]
414 	 */
415 	transaction_t *i_next_transaction;
416 
417 	/**
418 	 * @i_list: List of inodes in the i_transaction [j_list_lock]
419 	 */
420 	struct list_head i_list;
421 
422 	/**
423 	 * @i_vfs_inode:
424 	 *
425 	 * VFS inode this inode belongs to [constant for lifetime of structure]
426 	 */
427 	struct inode *i_vfs_inode;
428 
429 	/**
430 	 * @i_flags: Flags of inode [j_list_lock]
431 	 */
432 	unsigned long i_flags;
433 
434 	/**
435 	 * @i_dirty_start:
436 	 *
437 	 * Offset in bytes where the dirty range for this inode starts.
438 	 * [j_list_lock]
439 	 */
440 	loff_t i_dirty_start;
441 
442 	/**
443 	 * @i_dirty_end:
444 	 *
445 	 * Inclusive offset in bytes where the dirty range for this inode
446 	 * ends. [j_list_lock]
447 	 */
448 	loff_t i_dirty_end;
449 };
450 
451 struct jbd2_revoke_table_s;
452 
453 /**
454  * struct handle_s - The handle_s type is the concrete type associated with
455  *     handle_t.
456  * @h_transaction: Which compound transaction is this update a part of?
457  * @h_journal: Which journal handle belongs to - used iff h_reserved set.
458  * @h_rsv_handle: Handle reserved for finishing the logical operation.
459  * @h_total_credits: Number of remaining buffers we are allowed to add to
460  *	journal. These are dirty buffers and revoke descriptor blocks.
461  * @h_revoke_credits: Number of remaining revoke records available for handle
462  * @h_ref: Reference count on this handle.
463  * @h_err: Field for caller's use to track errors through large fs operations.
464  * @h_sync: Flag for sync-on-close.
465  * @h_jdata: Flag to force data journaling.
466  * @h_reserved: Flag for handle for reserved credits.
467  * @h_aborted: Flag indicating fatal error on handle.
468  * @h_type: For handle statistics.
469  * @h_line_no: For handle statistics.
470  * @h_start_jiffies: Handle Start time.
471  * @h_requested_credits: Holds @h_total_credits after handle is started.
472  * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
473  * @saved_alloc_context: Saved context while transaction is open.
474  **/
475 
476 /* Docbook can't yet cope with the bit fields, but will leave the documentation
477  * in so it can be fixed later.
478  */
479 
480 struct jbd2_journal_handle
481 {
482 	union {
483 		transaction_t	*h_transaction;
484 		/* Which journal handle belongs to - used iff h_reserved set */
485 		journal_t	*h_journal;
486 	};
487 
488 	handle_t		*h_rsv_handle;
489 	int			h_total_credits;
490 	int			h_revoke_credits;
491 	int			h_revoke_credits_requested;
492 	int			h_ref;
493 	int			h_err;
494 
495 	/* Flags [no locking] */
496 	unsigned int	h_sync:		1;
497 	unsigned int	h_jdata:	1;
498 	unsigned int	h_reserved:	1;
499 	unsigned int	h_aborted:	1;
500 	unsigned int	h_type:		8;
501 	unsigned int	h_line_no:	16;
502 
503 	unsigned long		h_start_jiffies;
504 	unsigned int		h_requested_credits;
505 
506 	unsigned int		saved_alloc_context;
507 };
508 
509 
510 /*
511  * Some stats for checkpoint phase
512  */
513 struct transaction_chp_stats_s {
514 	unsigned long		cs_chp_time;
515 	__u32			cs_forced_to_close;
516 	__u32			cs_written;
517 	__u32			cs_dropped;
518 };
519 
520 /* The transaction_t type is the guts of the journaling mechanism.  It
521  * tracks a compound transaction through its various states:
522  *
523  * RUNNING:	accepting new updates
524  * LOCKED:	Updates still running but we don't accept new ones
525  * RUNDOWN:	Updates are tidying up but have finished requesting
526  *		new buffers to modify (state not used for now)
527  * FLUSH:       All updates complete, but we are still writing to disk
528  * COMMIT:      All data on disk, writing commit record
529  * FINISHED:	We still have to keep the transaction for checkpointing.
530  *
531  * The transaction keeps track of all of the buffers modified by a
532  * running transaction, and all of the buffers committed but not yet
533  * flushed to home for finished transactions.
534  */
535 
536 /*
537  * Lock ranking:
538  *
539  *    j_list_lock
540  *      ->jbd_lock_bh_journal_head()	(This is "innermost")
541  *
542  *    j_state_lock
543  *    ->b_state_lock
544  *
545  *    b_state_lock
546  *    ->j_list_lock
547  *
548  *    j_state_lock
549  *    ->t_handle_lock
550  *
551  *    j_state_lock
552  *    ->j_list_lock			(journal_unmap_buffer)
553  *
554  */
555 
556 struct transaction_s
557 {
558 	/* Pointer to the journal for this transaction. [no locking] */
559 	journal_t		*t_journal;
560 
561 	/* Sequence number for this transaction [no locking] */
562 	tid_t			t_tid;
563 
564 	/*
565 	 * Transaction's current state
566 	 * [no locking - only kjournald2 alters this]
567 	 * [j_list_lock] guards transition of a transaction into T_FINISHED
568 	 * state and subsequent call of __jbd2_journal_drop_transaction()
569 	 * FIXME: needs barriers
570 	 * KLUDGE: [use j_state_lock]
571 	 */
572 	enum {
573 		T_RUNNING,
574 		T_LOCKED,
575 		T_SWITCH,
576 		T_FLUSH,
577 		T_COMMIT,
578 		T_COMMIT_DFLUSH,
579 		T_COMMIT_JFLUSH,
580 		T_COMMIT_CALLBACK,
581 		T_FINISHED
582 	}			t_state;
583 
584 	/*
585 	 * Where in the log does this transaction's commit start? [no locking]
586 	 */
587 	unsigned long		t_log_start;
588 
589 	/* Number of buffers on the t_buffers list [j_list_lock] */
590 	int			t_nr_buffers;
591 
592 	/*
593 	 * Doubly-linked circular list of all buffers reserved but not yet
594 	 * modified by this transaction [j_list_lock]
595 	 */
596 	struct journal_head	*t_reserved_list;
597 
598 	/*
599 	 * Doubly-linked circular list of all metadata buffers owned by this
600 	 * transaction [j_list_lock]
601 	 */
602 	struct journal_head	*t_buffers;
603 
604 	/*
605 	 * Doubly-linked circular list of all forget buffers (superseded
606 	 * buffers which we can un-checkpoint once this transaction commits)
607 	 * [j_list_lock]
608 	 */
609 	struct journal_head	*t_forget;
610 
611 	/*
612 	 * Doubly-linked circular list of all buffers still to be flushed before
613 	 * this transaction can be checkpointed. [j_list_lock]
614 	 */
615 	struct journal_head	*t_checkpoint_list;
616 
617 	/*
618 	 * Doubly-linked circular list of all buffers submitted for IO while
619 	 * checkpointing. [j_list_lock]
620 	 */
621 	struct journal_head	*t_checkpoint_io_list;
622 
623 	/*
624 	 * Doubly-linked circular list of metadata buffers being shadowed by log
625 	 * IO.  The IO buffers on the iobuf list and the shadow buffers on this
626 	 * list match each other one for one at all times. [j_list_lock]
627 	 */
628 	struct journal_head	*t_shadow_list;
629 
630 	/*
631 	 * List of inodes whose data we've modified in data=ordered mode.
632 	 * [j_list_lock]
633 	 */
634 	struct list_head	t_inode_list;
635 
636 	/*
637 	 * Protects info related to handles
638 	 */
639 	spinlock_t		t_handle_lock;
640 
641 	/*
642 	 * Longest time some handle had to wait for running transaction
643 	 */
644 	unsigned long		t_max_wait;
645 
646 	/*
647 	 * When transaction started
648 	 */
649 	unsigned long		t_start;
650 
651 	/*
652 	 * When commit was requested
653 	 */
654 	unsigned long		t_requested;
655 
656 	/*
657 	 * Checkpointing stats [j_checkpoint_sem]
658 	 */
659 	struct transaction_chp_stats_s t_chp_stats;
660 
661 	/*
662 	 * Number of outstanding updates running on this transaction
663 	 * [none]
664 	 */
665 	atomic_t		t_updates;
666 
667 	/*
668 	 * Number of blocks reserved for this transaction in the journal.
669 	 * This is including all credits reserved when starting transaction
670 	 * handles as well as all journal descriptor blocks needed for this
671 	 * transaction. [none]
672 	 */
673 	atomic_t		t_outstanding_credits;
674 
675 	/*
676 	 * Number of revoke records for this transaction added by already
677 	 * stopped handles. [none]
678 	 */
679 	atomic_t		t_outstanding_revokes;
680 
681 	/*
682 	 * How many handles used this transaction? [none]
683 	 */
684 	atomic_t		t_handle_count;
685 
686 	/*
687 	 * Forward and backward links for the circular list of all transactions
688 	 * awaiting checkpoint. [j_list_lock]
689 	 */
690 	transaction_t		*t_cpnext, *t_cpprev;
691 
692 	/*
693 	 * When will the transaction expire (become due for commit), in jiffies?
694 	 * [no locking]
695 	 */
696 	unsigned long		t_expires;
697 
698 	/*
699 	 * When this transaction started, in nanoseconds [no locking]
700 	 */
701 	ktime_t			t_start_time;
702 
703 	/*
704 	 * This transaction is being forced and some process is
705 	 * waiting for it to finish.
706 	 */
707 	unsigned int t_synchronous_commit:1;
708 
709 	/* Disk flush needs to be sent to fs partition [no locking] */
710 	int			t_need_data_flush;
711 
712 	/*
713 	 * For use by the filesystem to store fs-specific data
714 	 * structures associated with the transaction
715 	 */
716 	struct list_head	t_private_list;
717 };
718 
719 struct transaction_run_stats_s {
720 	unsigned long		rs_wait;
721 	unsigned long		rs_request_delay;
722 	unsigned long		rs_running;
723 	unsigned long		rs_locked;
724 	unsigned long		rs_flushing;
725 	unsigned long		rs_logging;
726 
727 	__u32			rs_handle_count;
728 	__u32			rs_blocks;
729 	__u32			rs_blocks_logged;
730 };
731 
732 struct transaction_stats_s {
733 	unsigned long		ts_tid;
734 	unsigned long		ts_requested;
735 	struct transaction_run_stats_s run;
736 };
737 
738 static inline unsigned long
739 jbd2_time_diff(unsigned long start, unsigned long end)
740 {
741 	if (end >= start)
742 		return end - start;
743 
744 	return end + (MAX_JIFFY_OFFSET - start);
745 }
746 
747 #define JBD2_NR_BATCH	64
748 
749 /**
750  * struct journal_s - The journal_s type is the concrete type associated with
751  *     journal_t.
752  */
753 struct journal_s
754 {
755 	/**
756 	 * @j_flags: General journaling state flags [j_state_lock]
757 	 */
758 	unsigned long		j_flags;
759 
760 	/**
761 	 * @j_errno:
762 	 *
763 	 * Is there an outstanding uncleared error on the journal (from a prior
764 	 * abort)? [j_state_lock]
765 	 */
766 	int			j_errno;
767 
768 	/**
769 	 * @j_sb_buffer: The first part of the superblock buffer.
770 	 */
771 	struct buffer_head	*j_sb_buffer;
772 
773 	/**
774 	 * @j_superblock: The second part of the superblock buffer.
775 	 */
776 	journal_superblock_t	*j_superblock;
777 
778 	/**
779 	 * @j_format_version: Version of the superblock format.
780 	 */
781 	int			j_format_version;
782 
783 	/**
784 	 * @j_state_lock: Protect the various scalars in the journal.
785 	 */
786 	rwlock_t		j_state_lock;
787 
788 	/**
789 	 * @j_barrier_count:
790 	 *
791 	 * Number of processes waiting to create a barrier lock [j_state_lock]
792 	 */
793 	int			j_barrier_count;
794 
795 	/**
796 	 * @j_barrier: The barrier lock itself.
797 	 */
798 	struct mutex		j_barrier;
799 
800 	/**
801 	 * @j_running_transaction:
802 	 *
803 	 * Transactions: The current running transaction...
804 	 * [j_state_lock] [caller holding open handle]
805 	 */
806 	transaction_t		*j_running_transaction;
807 
808 	/**
809 	 * @j_committing_transaction:
810 	 *
811 	 * the transaction we are pushing to disk
812 	 * [j_state_lock] [caller holding open handle]
813 	 */
814 	transaction_t		*j_committing_transaction;
815 
816 	/**
817 	 * @j_checkpoint_transactions:
818 	 *
819 	 * ... and a linked circular list of all transactions waiting for
820 	 * checkpointing. [j_list_lock]
821 	 */
822 	transaction_t		*j_checkpoint_transactions;
823 
824 	/**
825 	 * @j_wait_transaction_locked:
826 	 *
827 	 * Wait queue for waiting for a locked transaction to start committing,
828 	 * or for a barrier lock to be released.
829 	 */
830 	wait_queue_head_t	j_wait_transaction_locked;
831 
832 	/**
833 	 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
834 	 */
835 	wait_queue_head_t	j_wait_done_commit;
836 
837 	/**
838 	 * @j_wait_commit: Wait queue to trigger commit.
839 	 */
840 	wait_queue_head_t	j_wait_commit;
841 
842 	/**
843 	 * @j_wait_updates: Wait queue to wait for updates to complete.
844 	 */
845 	wait_queue_head_t	j_wait_updates;
846 
847 	/**
848 	 * @j_wait_reserved:
849 	 *
850 	 * Wait queue to wait for reserved buffer credits to drop.
851 	 */
852 	wait_queue_head_t	j_wait_reserved;
853 
854 	/**
855 	 * @j_checkpoint_mutex:
856 	 *
857 	 * Semaphore for locking against concurrent checkpoints.
858 	 */
859 	struct mutex		j_checkpoint_mutex;
860 
861 	/**
862 	 * @j_chkpt_bhs:
863 	 *
864 	 * List of buffer heads used by the checkpoint routine.  This
865 	 * was moved from jbd2_log_do_checkpoint() to reduce stack
866 	 * usage.  Access to this array is controlled by the
867 	 * @j_checkpoint_mutex.  [j_checkpoint_mutex]
868 	 */
869 	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
870 
871 	/**
872 	 * @j_head:
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 	 * @j_tail:
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 	 * @j_free:
889 	 *
890 	 * Journal free: how many free blocks are there in the journal?
891 	 * [j_state_lock]
892 	 */
893 	unsigned long		j_free;
894 
895 	/**
896 	 * @j_first:
897 	 *
898 	 * The block number of the first usable block in the journal
899 	 * [j_state_lock].
900 	 */
901 	unsigned long		j_first;
902 
903 	/**
904 	 * @j_last:
905 	 *
906 	 * The block number one beyond the last usable block in the journal
907 	 * [j_state_lock].
908 	 */
909 	unsigned long		j_last;
910 
911 	/**
912 	 * @j_dev: Device where we store the journal.
913 	 */
914 	struct block_device	*j_dev;
915 
916 	/**
917 	 * @j_blocksize: Block size for the location where we store the journal.
918 	 */
919 	int			j_blocksize;
920 
921 	/**
922 	 * @j_blk_offset:
923 	 *
924 	 * Starting block offset into the device where we store the journal.
925 	 */
926 	unsigned long long	j_blk_offset;
927 
928 	/**
929 	 * @j_devname: Journal device name.
930 	 */
931 	char			j_devname[BDEVNAME_SIZE+24];
932 
933 	/**
934 	 * @j_fs_dev:
935 	 *
936 	 * Device which holds the client fs.  For internal journal this will be
937 	 * equal to j_dev.
938 	 */
939 	struct block_device	*j_fs_dev;
940 
941 	/**
942 	 * @j_maxlen: Total maximum capacity of the journal region on disk.
943 	 */
944 	unsigned int		j_maxlen;
945 
946 	/**
947 	 * @j_reserved_credits:
948 	 *
949 	 * Number of buffers reserved from the running transaction.
950 	 */
951 	atomic_t		j_reserved_credits;
952 
953 	/**
954 	 * @j_list_lock: Protects the buffer lists and internal buffer state.
955 	 */
956 	spinlock_t		j_list_lock;
957 
958 	/**
959 	 * @j_inode:
960 	 *
961 	 * Optional inode where we store the journal.  If present, all
962 	 * journal block numbers are mapped into this inode via bmap().
963 	 */
964 	struct inode		*j_inode;
965 
966 	/**
967 	 * @j_tail_sequence:
968 	 *
969 	 * Sequence number of the oldest transaction in the log [j_state_lock]
970 	 */
971 	tid_t			j_tail_sequence;
972 
973 	/**
974 	 * @j_transaction_sequence:
975 	 *
976 	 * Sequence number of the next transaction to grant [j_state_lock]
977 	 */
978 	tid_t			j_transaction_sequence;
979 
980 	/**
981 	 * @j_commit_sequence:
982 	 *
983 	 * Sequence number of the most recently committed transaction
984 	 * [j_state_lock].
985 	 */
986 	tid_t			j_commit_sequence;
987 
988 	/**
989 	 * @j_commit_request:
990 	 *
991 	 * Sequence number of the most recent transaction wanting commit
992 	 * [j_state_lock]
993 	 */
994 	tid_t			j_commit_request;
995 
996 	/**
997 	 * @j_uuid:
998 	 *
999 	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1000 	 * backed by this journal.  This will eventually be replaced by an array
1001 	 * of uuids, allowing us to index multiple devices within a single
1002 	 * journal and to perform atomic updates across them.
1003 	 */
1004 	__u8			j_uuid[16];
1005 
1006 	/**
1007 	 * @j_task: Pointer to the current commit thread for this journal.
1008 	 */
1009 	struct task_struct	*j_task;
1010 
1011 	/**
1012 	 * @j_max_transaction_buffers:
1013 	 *
1014 	 * Maximum number of metadata buffers to allow in a single compound
1015 	 * commit transaction.
1016 	 */
1017 	int			j_max_transaction_buffers;
1018 
1019 	/**
1020 	 * @j_revoke_records_per_block:
1021 	 *
1022 	 * Number of revoke records that fit in one descriptor block.
1023 	 */
1024 	int			j_revoke_records_per_block;
1025 
1026 	/**
1027 	 * @j_commit_interval:
1028 	 *
1029 	 * What is the maximum transaction lifetime before we begin a commit?
1030 	 */
1031 	unsigned long		j_commit_interval;
1032 
1033 	/**
1034 	 * @j_commit_timer: The timer used to wakeup the commit thread.
1035 	 */
1036 	struct timer_list	j_commit_timer;
1037 
1038 	/**
1039 	 * @j_revoke_lock: Protect the revoke table.
1040 	 */
1041 	spinlock_t		j_revoke_lock;
1042 
1043 	/**
1044 	 * @j_revoke:
1045 	 *
1046 	 * The revoke table - maintains the list of revoked blocks in the
1047 	 * current transaction.
1048 	 */
1049 	struct jbd2_revoke_table_s *j_revoke;
1050 
1051 	/**
1052 	 * @j_revoke_table: Alternate revoke tables for j_revoke.
1053 	 */
1054 	struct jbd2_revoke_table_s *j_revoke_table[2];
1055 
1056 	/**
1057 	 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1058 	 */
1059 	struct buffer_head	**j_wbuf;
1060 
1061 	/**
1062 	 * @j_wbufsize:
1063 	 *
1064 	 * Size of @j_wbuf array.
1065 	 */
1066 	int			j_wbufsize;
1067 
1068 	/**
1069 	 * @j_last_sync_writer:
1070 	 *
1071 	 * The pid of the last person to run a synchronous operation
1072 	 * through the journal.
1073 	 */
1074 	pid_t			j_last_sync_writer;
1075 
1076 	/**
1077 	 * @j_average_commit_time:
1078 	 *
1079 	 * The average amount of time in nanoseconds it takes to commit a
1080 	 * transaction to disk. [j_state_lock]
1081 	 */
1082 	u64			j_average_commit_time;
1083 
1084 	/**
1085 	 * @j_min_batch_time:
1086 	 *
1087 	 * Minimum time that we should wait for additional filesystem operations
1088 	 * to get batched into a synchronous handle in microseconds.
1089 	 */
1090 	u32			j_min_batch_time;
1091 
1092 	/**
1093 	 * @j_max_batch_time:
1094 	 *
1095 	 * Maximum time that we should wait for additional filesystem operations
1096 	 * to get batched into a synchronous handle in microseconds.
1097 	 */
1098 	u32			j_max_batch_time;
1099 
1100 	/**
1101 	 * @j_commit_callback:
1102 	 *
1103 	 * This function is called when a transaction is closed.
1104 	 */
1105 	void			(*j_commit_callback)(journal_t *,
1106 						     transaction_t *);
1107 
1108 	/*
1109 	 * Journal statistics
1110 	 */
1111 
1112 	/**
1113 	 * @j_history_lock: Protect the transactions statistics history.
1114 	 */
1115 	spinlock_t		j_history_lock;
1116 
1117 	/**
1118 	 * @j_proc_entry: procfs entry for the jbd statistics directory.
1119 	 */
1120 	struct proc_dir_entry	*j_proc_entry;
1121 
1122 	/**
1123 	 * @j_stats: Overall statistics.
1124 	 */
1125 	struct transaction_stats_s j_stats;
1126 
1127 	/**
1128 	 * @j_failed_commit: Failed journal commit ID.
1129 	 */
1130 	unsigned int		j_failed_commit;
1131 
1132 	/**
1133 	 * @j_private:
1134 	 *
1135 	 * An opaque pointer to fs-private information.  ext3 puts its
1136 	 * superblock pointer here.
1137 	 */
1138 	void *j_private;
1139 
1140 	/**
1141 	 * @j_chksum_driver:
1142 	 *
1143 	 * Reference to checksum algorithm driver via cryptoapi.
1144 	 */
1145 	struct crypto_shash *j_chksum_driver;
1146 
1147 	/**
1148 	 * @j_csum_seed:
1149 	 *
1150 	 * Precomputed journal UUID checksum for seeding other checksums.
1151 	 */
1152 	__u32 j_csum_seed;
1153 
1154 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1155 	/**
1156 	 * @j_trans_commit_map:
1157 	 *
1158 	 * Lockdep entity to track transaction commit dependencies. Handles
1159 	 * hold this "lock" for read, when we wait for commit, we acquire the
1160 	 * "lock" for writing. This matches the properties of jbd2 journalling
1161 	 * where the running transaction has to wait for all handles to be
1162 	 * dropped to commit that transaction and also acquiring a handle may
1163 	 * require transaction commit to finish.
1164 	 */
1165 	struct lockdep_map	j_trans_commit_map;
1166 #endif
1167 };
1168 
1169 #define jbd2_might_wait_for_commit(j) \
1170 	do { \
1171 		rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1172 		rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1173 	} while (0)
1174 
1175 /* journal feature predicate functions */
1176 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1177 static inline bool jbd2_has_feature_##name(journal_t *j) \
1178 { \
1179 	return ((j)->j_format_version >= 2 && \
1180 		((j)->j_superblock->s_feature_compat & \
1181 		 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1182 } \
1183 static inline void jbd2_set_feature_##name(journal_t *j) \
1184 { \
1185 	(j)->j_superblock->s_feature_compat |= \
1186 		cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1187 } \
1188 static inline void jbd2_clear_feature_##name(journal_t *j) \
1189 { \
1190 	(j)->j_superblock->s_feature_compat &= \
1191 		~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1192 }
1193 
1194 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1195 static inline bool jbd2_has_feature_##name(journal_t *j) \
1196 { \
1197 	return ((j)->j_format_version >= 2 && \
1198 		((j)->j_superblock->s_feature_ro_compat & \
1199 		 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1200 } \
1201 static inline void jbd2_set_feature_##name(journal_t *j) \
1202 { \
1203 	(j)->j_superblock->s_feature_ro_compat |= \
1204 		cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1205 } \
1206 static inline void jbd2_clear_feature_##name(journal_t *j) \
1207 { \
1208 	(j)->j_superblock->s_feature_ro_compat &= \
1209 		~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1210 }
1211 
1212 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1213 static inline bool jbd2_has_feature_##name(journal_t *j) \
1214 { \
1215 	return ((j)->j_format_version >= 2 && \
1216 		((j)->j_superblock->s_feature_incompat & \
1217 		 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1218 } \
1219 static inline void jbd2_set_feature_##name(journal_t *j) \
1220 { \
1221 	(j)->j_superblock->s_feature_incompat |= \
1222 		cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1223 } \
1224 static inline void jbd2_clear_feature_##name(journal_t *j) \
1225 { \
1226 	(j)->j_superblock->s_feature_incompat &= \
1227 		~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1228 }
1229 
1230 JBD2_FEATURE_COMPAT_FUNCS(checksum,		CHECKSUM)
1231 
1232 JBD2_FEATURE_INCOMPAT_FUNCS(revoke,		REVOKE)
1233 JBD2_FEATURE_INCOMPAT_FUNCS(64bit,		64BIT)
1234 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit,	ASYNC_COMMIT)
1235 JBD2_FEATURE_INCOMPAT_FUNCS(csum2,		CSUM_V2)
1236 JBD2_FEATURE_INCOMPAT_FUNCS(csum3,		CSUM_V3)
1237 
1238 /*
1239  * Journal flag definitions
1240  */
1241 #define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1242 #define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1243 #define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1244 #define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1245 #define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1246 #define JBD2_BARRIER	0x020	/* Use IDE barriers */
1247 #define JBD2_ABORT_ON_SYNCDATA_ERR	0x040	/* Abort the journal on file
1248 						 * data write error in ordered
1249 						 * mode */
1250 #define JBD2_REC_ERR	0x080	/* The errno in the sb has been recorded */
1251 
1252 /*
1253  * Function declarations for the journaling transaction and buffer
1254  * management
1255  */
1256 
1257 /* Filing buffers */
1258 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1259 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1260 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1261 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1262 extern void __journal_free_buffer(struct journal_head *bh);
1263 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1264 extern void __journal_clean_data_list(transaction_t *transaction);
1265 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1266 {
1267 	list_add_tail(&bh->b_assoc_buffers, head);
1268 }
1269 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1270 {
1271 	list_del_init(&bh->b_assoc_buffers);
1272 }
1273 
1274 /* Log buffer allocation */
1275 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1276 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1277 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1278 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1279 			      unsigned long *block);
1280 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1281 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1282 
1283 /* Commit management */
1284 extern void jbd2_journal_commit_transaction(journal_t *);
1285 
1286 /* Checkpoint list management */
1287 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1288 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1289 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1290 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1291 
1292 
1293 /*
1294  * Triggers
1295  */
1296 
1297 struct jbd2_buffer_trigger_type {
1298 	/*
1299 	 * Fired a the moment data to write to the journal are known to be
1300 	 * stable - so either at the moment b_frozen_data is created or just
1301 	 * before a buffer is written to the journal.  mapped_data is a mapped
1302 	 * buffer that is the frozen data for commit.
1303 	 */
1304 	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1305 			 struct buffer_head *bh, void *mapped_data,
1306 			 size_t size);
1307 
1308 	/*
1309 	 * Fired during journal abort for dirty buffers that will not be
1310 	 * committed.
1311 	 */
1312 	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1313 			struct buffer_head *bh);
1314 };
1315 
1316 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1317 				       void *mapped_data,
1318 				       struct jbd2_buffer_trigger_type *triggers);
1319 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1320 				      struct jbd2_buffer_trigger_type *triggers);
1321 
1322 /* Buffer IO */
1323 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1324 					      struct journal_head *jh_in,
1325 					      struct buffer_head **bh_out,
1326 					      sector_t blocknr);
1327 
1328 /* Transaction locking */
1329 extern void		__wait_on_journal (journal_t *);
1330 
1331 /* Transaction cache support */
1332 extern void jbd2_journal_destroy_transaction_cache(void);
1333 extern int __init jbd2_journal_init_transaction_cache(void);
1334 extern void jbd2_journal_free_transaction(transaction_t *);
1335 
1336 /*
1337  * Journal locking.
1338  *
1339  * We need to lock the journal during transaction state changes so that nobody
1340  * ever tries to take a handle on the running transaction while we are in the
1341  * middle of moving it to the commit phase.  j_state_lock does this.
1342  *
1343  * Note that the locking is completely interrupt unsafe.  We never touch
1344  * journal structures from interrupts.
1345  */
1346 
1347 static inline handle_t *journal_current_handle(void)
1348 {
1349 	return current->journal_info;
1350 }
1351 
1352 /* The journaling code user interface:
1353  *
1354  * Create and destroy handles
1355  * Register buffer modifications against the current transaction.
1356  */
1357 
1358 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1359 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1360 				     int revoke_records, gfp_t gfp_mask,
1361 				     unsigned int type, unsigned int line_no);
1362 extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1363 extern int	 jbd2__journal_restart(handle_t *, int nblocks,
1364 				       int revoke_records, gfp_t gfp_mask);
1365 extern int	 jbd2_journal_start_reserved(handle_t *handle,
1366 				unsigned int type, unsigned int line_no);
1367 extern void	 jbd2_journal_free_reserved(handle_t *handle);
1368 extern int	 jbd2_journal_extend(handle_t *handle, int nblocks,
1369 				     int revoke_records);
1370 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1371 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1372 extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1373 void		 jbd2_journal_set_triggers(struct buffer_head *,
1374 					   struct jbd2_buffer_trigger_type *type);
1375 extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1376 extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1377 extern int	 jbd2_journal_invalidatepage(journal_t *,
1378 				struct page *, unsigned int, unsigned int);
1379 extern int	 jbd2_journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
1380 extern int	 jbd2_journal_stop(handle_t *);
1381 extern int	 jbd2_journal_flush (journal_t *);
1382 extern void	 jbd2_journal_lock_updates (journal_t *);
1383 extern void	 jbd2_journal_unlock_updates (journal_t *);
1384 
1385 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1386 				struct block_device *fs_dev,
1387 				unsigned long long start, int len, int bsize);
1388 extern journal_t * jbd2_journal_init_inode (struct inode *);
1389 extern int	   jbd2_journal_update_format (journal_t *);
1390 extern int	   jbd2_journal_check_used_features
1391 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1392 extern int	   jbd2_journal_check_available_features
1393 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1394 extern int	   jbd2_journal_set_features
1395 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1396 extern void	   jbd2_journal_clear_features
1397 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1398 extern int	   jbd2_journal_load       (journal_t *journal);
1399 extern int	   jbd2_journal_destroy    (journal_t *);
1400 extern int	   jbd2_journal_recover    (journal_t *journal);
1401 extern int	   jbd2_journal_wipe       (journal_t *, int);
1402 extern int	   jbd2_journal_skip_recovery	(journal_t *);
1403 extern void	   jbd2_journal_update_sb_errno(journal_t *);
1404 extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1405 				unsigned long, int);
1406 extern void	   jbd2_journal_abort      (journal_t *, int);
1407 extern int	   jbd2_journal_errno      (journal_t *);
1408 extern void	   jbd2_journal_ack_err    (journal_t *);
1409 extern int	   jbd2_journal_clear_err  (journal_t *);
1410 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1411 extern int	   jbd2_journal_force_commit(journal_t *);
1412 extern int	   jbd2_journal_force_commit_nested(journal_t *);
1413 extern int	   jbd2_journal_inode_ranged_write(handle_t *handle,
1414 			struct jbd2_inode *inode, loff_t start_byte,
1415 			loff_t length);
1416 extern int	   jbd2_journal_inode_ranged_wait(handle_t *handle,
1417 			struct jbd2_inode *inode, loff_t start_byte,
1418 			loff_t length);
1419 extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1420 				struct jbd2_inode *inode, loff_t new_size);
1421 extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1422 extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1423 
1424 /*
1425  * journal_head management
1426  */
1427 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1428 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1429 void jbd2_journal_put_journal_head(struct journal_head *jh);
1430 
1431 /*
1432  * handle management
1433  */
1434 extern struct kmem_cache *jbd2_handle_cache;
1435 
1436 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1437 {
1438 	return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1439 }
1440 
1441 static inline void jbd2_free_handle(handle_t *handle)
1442 {
1443 	kmem_cache_free(jbd2_handle_cache, handle);
1444 }
1445 
1446 /*
1447  * jbd2_inode management (optional, for those file systems that want to use
1448  * dynamically allocated jbd2_inode structures)
1449  */
1450 extern struct kmem_cache *jbd2_inode_cache;
1451 
1452 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1453 {
1454 	return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1455 }
1456 
1457 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1458 {
1459 	kmem_cache_free(jbd2_inode_cache, jinode);
1460 }
1461 
1462 /* Primary revoke support */
1463 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1464 extern int	   jbd2_journal_init_revoke(journal_t *, int);
1465 extern void	   jbd2_journal_destroy_revoke_record_cache(void);
1466 extern void	   jbd2_journal_destroy_revoke_table_cache(void);
1467 extern int __init jbd2_journal_init_revoke_record_cache(void);
1468 extern int __init jbd2_journal_init_revoke_table_cache(void);
1469 
1470 extern void	   jbd2_journal_destroy_revoke(journal_t *);
1471 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1472 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1473 extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
1474 						     struct list_head *log_bufs);
1475 
1476 /* Recovery revoke support */
1477 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1478 extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1479 extern void	jbd2_journal_clear_revoke(journal_t *);
1480 extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1481 extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1482 
1483 /*
1484  * The log thread user interface:
1485  *
1486  * Request space in the current transaction, and force transaction commit
1487  * transitions on demand.
1488  */
1489 
1490 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1491 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1492 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1493 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1494 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1495 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1496 int jbd2_log_do_checkpoint(journal_t *journal);
1497 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1498 
1499 void __jbd2_log_wait_for_space(journal_t *journal);
1500 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1501 extern int jbd2_cleanup_journal_tail(journal_t *);
1502 
1503 /*
1504  * is_journal_abort
1505  *
1506  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1507  * bit, when set, indicates that we have had a fatal error somewhere,
1508  * either inside the journaling layer or indicated to us by the client
1509  * (eg. ext3), and that we and should not commit any further
1510  * transactions.
1511  */
1512 
1513 static inline int is_journal_aborted(journal_t *journal)
1514 {
1515 	return journal->j_flags & JBD2_ABORT;
1516 }
1517 
1518 static inline int is_handle_aborted(handle_t *handle)
1519 {
1520 	if (handle->h_aborted || !handle->h_transaction)
1521 		return 1;
1522 	return is_journal_aborted(handle->h_transaction->t_journal);
1523 }
1524 
1525 static inline void jbd2_journal_abort_handle(handle_t *handle)
1526 {
1527 	handle->h_aborted = 1;
1528 }
1529 
1530 #endif /* __KERNEL__   */
1531 
1532 /* Comparison functions for transaction IDs: perform comparisons using
1533  * modulo arithmetic so that they work over sequence number wraps. */
1534 
1535 static inline int tid_gt(tid_t x, tid_t y)
1536 {
1537 	int difference = (x - y);
1538 	return (difference > 0);
1539 }
1540 
1541 static inline int tid_geq(tid_t x, tid_t y)
1542 {
1543 	int difference = (x - y);
1544 	return (difference >= 0);
1545 }
1546 
1547 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1548 extern size_t journal_tag_bytes(journal_t *journal);
1549 
1550 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1551 {
1552 	return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1553 }
1554 
1555 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1556 {
1557 	WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1558 		     journal->j_chksum_driver == NULL);
1559 
1560 	return journal->j_chksum_driver != NULL;
1561 }
1562 
1563 /*
1564  * Return number of free blocks in the log. Must be called under j_state_lock.
1565  */
1566 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1567 {
1568 	/* Allow for rounding errors */
1569 	long free = journal->j_free - 32;
1570 
1571 	if (journal->j_committing_transaction) {
1572 		free -= atomic_read(&journal->
1573                         j_committing_transaction->t_outstanding_credits);
1574 	}
1575 	return max_t(long, free, 0);
1576 }
1577 
1578 /*
1579  * Definitions which augment the buffer_head layer
1580  */
1581 
1582 /* journaling buffer types */
1583 #define BJ_None		0	/* Not journaled */
1584 #define BJ_Metadata	1	/* Normal journaled metadata */
1585 #define BJ_Forget	2	/* Buffer superseded by this transaction */
1586 #define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1587 #define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1588 #define BJ_Types	5
1589 
1590 extern int jbd_blocks_per_page(struct inode *inode);
1591 
1592 /* JBD uses a CRC32 checksum */
1593 #define JBD_MAX_CHECKSUM_SIZE 4
1594 
1595 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1596 			      const void *address, unsigned int length)
1597 {
1598 	struct {
1599 		struct shash_desc shash;
1600 		char ctx[JBD_MAX_CHECKSUM_SIZE];
1601 	} desc;
1602 	int err;
1603 
1604 	BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1605 		JBD_MAX_CHECKSUM_SIZE);
1606 
1607 	desc.shash.tfm = journal->j_chksum_driver;
1608 	*(u32 *)desc.ctx = crc;
1609 
1610 	err = crypto_shash_update(&desc.shash, address, length);
1611 	BUG_ON(err);
1612 
1613 	return *(u32 *)desc.ctx;
1614 }
1615 
1616 /* Return most recent uncommitted transaction */
1617 static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1618 {
1619 	tid_t tid;
1620 
1621 	read_lock(&journal->j_state_lock);
1622 	tid = journal->j_commit_request;
1623 	if (journal->j_running_transaction)
1624 		tid = journal->j_running_transaction->t_tid;
1625 	read_unlock(&journal->j_state_lock);
1626 	return tid;
1627 }
1628 
1629 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1630 {
1631 	journal_t *journal;
1632 
1633 	if (!handle->h_reserved)
1634 		journal = handle->h_transaction->t_journal;
1635 	else
1636 		journal = handle->h_journal;
1637 
1638 	return handle->h_total_credits -
1639 		DIV_ROUND_UP(handle->h_revoke_credits_requested,
1640 			     journal->j_revoke_records_per_block);
1641 }
1642 
1643 #ifdef __KERNEL__
1644 
1645 #define buffer_trace_init(bh)	do {} while (0)
1646 #define print_buffer_fields(bh)	do {} while (0)
1647 #define print_buffer_trace(bh)	do {} while (0)
1648 #define BUFFER_TRACE(bh, info)	do {} while (0)
1649 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1650 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1651 
1652 #endif	/* __KERNEL__ */
1653 
1654 #define EFSBADCRC	EBADMSG		/* Bad CRC detected */
1655 #define EFSCORRUPTED	EUCLEAN		/* Filesystem is corrupted */
1656 
1657 #endif	/* _LINUX_JBD2_H */
1658