1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3  * linux/include/linux/jbd2.h
4  *
5  * Written by Stephen C. Tweedie <sct@redhat.com>
6  *
7  * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
8  *
9  * Definitions for transaction data structures for the buffer cache
10  * filesystem journaling support.
11  */
12 
13 #ifndef _LINUX_JBD2_H
14 #define _LINUX_JBD2_H
15 
16 /* Allow this file to be included directly into e2fsprogs */
17 #ifndef __KERNEL__
18 #include "jfs_compat.h"
19 #define JBD2_DEBUG
20 #else
21 
22 #include <linux/types.h>
23 #include <linux/buffer_head.h>
24 #include <linux/journal-head.h>
25 #include <linux/stddef.h>
26 #include <linux/mutex.h>
27 #include <linux/timer.h>
28 #include <linux/slab.h>
29 #include <linux/bit_spinlock.h>
30 #include <linux/blkdev.h>
31 #include <crypto/hash.h>
32 #endif
33 
34 #define journal_oom_retry 1
35 
36 /*
37  * Define JBD2_PARANIOD_IOFAIL to cause a kernel BUG() if ext4 finds
38  * certain classes of error which can occur due to failed IOs.  Under
39  * normal use we want ext4 to continue after such errors, because
40  * hardware _can_ fail, but for debugging purposes when running tests on
41  * known-good hardware we may want to trap these errors.
42  */
43 #undef JBD2_PARANOID_IOFAIL
44 
45 /*
46  * The default maximum commit age, in seconds.
47  */
48 #define JBD2_DEFAULT_MAX_COMMIT_AGE 5
49 
50 #ifdef CONFIG_JBD2_DEBUG
51 /*
52  * Define JBD2_EXPENSIVE_CHECKING to enable more expensive internal
53  * consistency checks.  By default we don't do this unless
54  * CONFIG_JBD2_DEBUG is on.
55  */
56 #define JBD2_EXPENSIVE_CHECKING
57 extern ushort jbd2_journal_enable_debug;
58 void __jbd2_debug(int level, const char *file, const char *func,
59 		  unsigned int line, const char *fmt, ...);
60 
61 #define jbd_debug(n, fmt, a...) \
62 	__jbd2_debug((n), __FILE__, __func__, __LINE__, (fmt), ##a)
63 #else
64 #define jbd_debug(n, fmt, a...)  no_printk(fmt, ##a)
65 #endif
66 
67 extern void *jbd2_alloc(size_t size, gfp_t flags);
68 extern void jbd2_free(void *ptr, size_t size);
69 
70 #define JBD2_MIN_JOURNAL_BLOCKS 1024
71 #define JBD2_DEFAULT_FAST_COMMIT_BLOCKS 256
72 
73 #ifdef __KERNEL__
74 
75 /**
76  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
77  *
78  * All filesystem modifications made by the process go
79  * through this handle.  Recursive operations (such as quota operations)
80  * are gathered into a single update.
81  *
82  * The buffer credits field is used to account for journaled buffers
83  * being modified by the running process.  To ensure that there is
84  * enough log space for all outstanding operations, we need to limit the
85  * number of outstanding buffers possible at any time.  When the
86  * operation completes, any buffer credits not used are credited back to
87  * the transaction, so that at all times we know how many buffers the
88  * outstanding updates on a transaction might possibly touch.
89  *
90  * This is an opaque datatype.
91  **/
92 typedef struct jbd2_journal_handle handle_t;	/* Atomic operation type */
93 
94 
95 /**
96  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
97  *
98  * journal_t is linked to from the fs superblock structure.
99  *
100  * We use the journal_t to keep track of all outstanding transaction
101  * activity on the filesystem, and to manage the state of the log
102  * writing process.
103  *
104  * This is an opaque datatype.
105  **/
106 typedef struct journal_s	journal_t;	/* Journal control structure */
107 #endif
108 
109 /*
110  * Internal structures used by the logging mechanism:
111  */
112 
113 #define JBD2_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
114 
115 /*
116  * On-disk structures
117  */
118 
119 /*
120  * Descriptor block types:
121  */
122 
123 #define JBD2_DESCRIPTOR_BLOCK	1
124 #define JBD2_COMMIT_BLOCK	2
125 #define JBD2_SUPERBLOCK_V1	3
126 #define JBD2_SUPERBLOCK_V2	4
127 #define JBD2_REVOKE_BLOCK	5
128 
129 /*
130  * Standard header for all descriptor blocks:
131  */
132 typedef struct journal_header_s
133 {
134 	__be32		h_magic;
135 	__be32		h_blocktype;
136 	__be32		h_sequence;
137 } journal_header_t;
138 
139 /*
140  * Checksum types.
141  */
142 #define JBD2_CRC32_CHKSUM   1
143 #define JBD2_MD5_CHKSUM     2
144 #define JBD2_SHA1_CHKSUM    3
145 #define JBD2_CRC32C_CHKSUM  4
146 
147 #define JBD2_CRC32_CHKSUM_SIZE 4
148 
149 #define JBD2_CHECKSUM_BYTES (32 / sizeof(u32))
150 /*
151  * Commit block header for storing transactional checksums:
152  *
153  * NOTE: If FEATURE_COMPAT_CHECKSUM (checksum v1) is set, the h_chksum*
154  * fields are used to store a checksum of the descriptor and data blocks.
155  *
156  * If FEATURE_INCOMPAT_CSUM_V2 (checksum v2) is set, then the h_chksum
157  * field is used to store crc32c(uuid+commit_block).  Each journal metadata
158  * block gets its own checksum, and data block checksums are stored in
159  * journal_block_tag (in the descriptor).  The other h_chksum* fields are
160  * not used.
161  *
162  * If FEATURE_INCOMPAT_CSUM_V3 is set, the descriptor block uses
163  * journal_block_tag3_t to store a full 32-bit checksum.  Everything else
164  * is the same as v2.
165  *
166  * Checksum v1, v2, and v3 are mutually exclusive features.
167  */
168 struct commit_header {
169 	__be32		h_magic;
170 	__be32          h_blocktype;
171 	__be32          h_sequence;
172 	unsigned char   h_chksum_type;
173 	unsigned char   h_chksum_size;
174 	unsigned char 	h_padding[2];
175 	__be32 		h_chksum[JBD2_CHECKSUM_BYTES];
176 	__be64		h_commit_sec;
177 	__be32		h_commit_nsec;
178 };
179 
180 /*
181  * The block tag: used to describe a single buffer in the journal.
182  * t_blocknr_high is only used if INCOMPAT_64BIT is set, so this
183  * raw struct shouldn't be used for pointer math or sizeof() - use
184  * journal_tag_bytes(journal) instead to compute this.
185  */
186 typedef struct journal_block_tag3_s
187 {
188 	__be32		t_blocknr;	/* The on-disk block number */
189 	__be32		t_flags;	/* See below */
190 	__be32		t_blocknr_high; /* most-significant high 32bits. */
191 	__be32		t_checksum;	/* crc32c(uuid+seq+block) */
192 } journal_block_tag3_t;
193 
194 typedef struct journal_block_tag_s
195 {
196 	__be32		t_blocknr;	/* The on-disk block number */
197 	__be16		t_checksum;	/* truncated crc32c(uuid+seq+block) */
198 	__be16		t_flags;	/* See below */
199 	__be32		t_blocknr_high; /* most-significant high 32bits. */
200 } journal_block_tag_t;
201 
202 /* Tail of descriptor or revoke block, for checksumming */
203 struct jbd2_journal_block_tail {
204 	__be32		t_checksum;	/* crc32c(uuid+descr_block) */
205 };
206 
207 /*
208  * The revoke descriptor: used on disk to describe a series of blocks to
209  * be revoked from the log
210  */
211 typedef struct jbd2_journal_revoke_header_s
212 {
213 	journal_header_t r_header;
214 	__be32		 r_count;	/* Count of bytes used in the block */
215 } jbd2_journal_revoke_header_t;
216 
217 /* Definitions for the journal tag flags word: */
218 #define JBD2_FLAG_ESCAPE		1	/* on-disk block is escaped */
219 #define JBD2_FLAG_SAME_UUID	2	/* block has same uuid as previous */
220 #define JBD2_FLAG_DELETED	4	/* block deleted by this transaction */
221 #define JBD2_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
222 
223 
224 /*
225  * The journal superblock.  All fields are in big-endian byte order.
226  */
227 typedef struct journal_superblock_s
228 {
229 /* 0x0000 */
230 	journal_header_t s_header;
231 
232 /* 0x000C */
233 	/* Static information describing the journal */
234 	__be32	s_blocksize;		/* journal device blocksize */
235 	__be32	s_maxlen;		/* total blocks in journal file */
236 	__be32	s_first;		/* first block of log information */
237 
238 /* 0x0018 */
239 	/* Dynamic information describing the current state of the log */
240 	__be32	s_sequence;		/* first commit ID expected in log */
241 	__be32	s_start;		/* blocknr of start of log */
242 
243 /* 0x0020 */
244 	/* Error value, as set by jbd2_journal_abort(). */
245 	__be32	s_errno;
246 
247 /* 0x0024 */
248 	/* Remaining fields are only valid in a version-2 superblock */
249 	__be32	s_feature_compat;	/* compatible feature set */
250 	__be32	s_feature_incompat;	/* incompatible feature set */
251 	__be32	s_feature_ro_compat;	/* readonly-compatible feature set */
252 /* 0x0030 */
253 	__u8	s_uuid[16];		/* 128-bit uuid for journal */
254 
255 /* 0x0040 */
256 	__be32	s_nr_users;		/* Nr of filesystems sharing log */
257 
258 	__be32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
259 
260 /* 0x0048 */
261 	__be32	s_max_transaction;	/* Limit of journal blocks per trans.*/
262 	__be32	s_max_trans_data;	/* Limit of data blocks per trans. */
263 
264 /* 0x0050 */
265 	__u8	s_checksum_type;	/* checksum type */
266 	__u8	s_padding2[3];
267 /* 0x0054 */
268 	__be32	s_num_fc_blks;		/* Number of fast commit blocks */
269 /* 0x0058 */
270 	__u32	s_padding[41];
271 	__be32	s_checksum;		/* crc32c(superblock) */
272 
273 /* 0x0100 */
274 	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
275 /* 0x0400 */
276 } journal_superblock_t;
277 
278 /* Use the jbd2_{has,set,clear}_feature_* helpers; these will be removed */
279 #define JBD2_HAS_COMPAT_FEATURE(j,mask)					\
280 	((j)->j_format_version >= 2 &&					\
281 	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
282 #define JBD2_HAS_RO_COMPAT_FEATURE(j,mask)				\
283 	((j)->j_format_version >= 2 &&					\
284 	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
285 #define JBD2_HAS_INCOMPAT_FEATURE(j,mask)				\
286 	((j)->j_format_version >= 2 &&					\
287 	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
288 
289 #define JBD2_FEATURE_COMPAT_CHECKSUM		0x00000001
290 
291 #define JBD2_FEATURE_INCOMPAT_REVOKE		0x00000001
292 #define JBD2_FEATURE_INCOMPAT_64BIT		0x00000002
293 #define JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
294 #define JBD2_FEATURE_INCOMPAT_CSUM_V2		0x00000008
295 #define JBD2_FEATURE_INCOMPAT_CSUM_V3		0x00000010
296 #define JBD2_FEATURE_INCOMPAT_FAST_COMMIT	0x00000020
297 
298 /* See "journal feature predicate functions" below */
299 
300 /* Features known to this kernel version: */
301 #define JBD2_KNOWN_COMPAT_FEATURES	JBD2_FEATURE_COMPAT_CHECKSUM
302 #define JBD2_KNOWN_ROCOMPAT_FEATURES	0
303 #define JBD2_KNOWN_INCOMPAT_FEATURES	(JBD2_FEATURE_INCOMPAT_REVOKE | \
304 					JBD2_FEATURE_INCOMPAT_64BIT | \
305 					JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | \
306 					JBD2_FEATURE_INCOMPAT_CSUM_V2 | \
307 					JBD2_FEATURE_INCOMPAT_CSUM_V3 | \
308 					JBD2_FEATURE_INCOMPAT_FAST_COMMIT)
309 
310 #ifdef __KERNEL__
311 
312 #include <linux/fs.h>
313 #include <linux/sched.h>
314 
315 enum jbd_state_bits {
316 	BH_JBD			/* Has an attached ext3 journal_head */
317 	  = BH_PrivateStart,
318 	BH_JWrite,		/* Being written to log (@@@ DEBUGGING) */
319 	BH_Freed,		/* Has been freed (truncated) */
320 	BH_Revoked,		/* Has been revoked from the log */
321 	BH_RevokeValid,		/* Revoked flag is valid */
322 	BH_JBDDirty,		/* Is dirty but journaled */
323 	BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
324 	BH_Shadow,		/* IO on shadow buffer is running */
325 	BH_Verified,		/* Metadata block has been verified ok */
326 	BH_JBDPrivateStart,	/* First bit available for private use by FS */
327 };
328 
BUFFER_FNS(JBD,jbd)329 BUFFER_FNS(JBD, jbd)
330 BUFFER_FNS(JWrite, jwrite)
331 BUFFER_FNS(JBDDirty, jbddirty)
332 TAS_BUFFER_FNS(JBDDirty, jbddirty)
333 BUFFER_FNS(Revoked, revoked)
334 TAS_BUFFER_FNS(Revoked, revoked)
335 BUFFER_FNS(RevokeValid, revokevalid)
336 TAS_BUFFER_FNS(RevokeValid, revokevalid)
337 BUFFER_FNS(Freed, freed)
338 BUFFER_FNS(Shadow, shadow)
339 BUFFER_FNS(Verified, verified)
340 
341 static inline struct buffer_head *jh2bh(struct journal_head *jh)
342 {
343 	return jh->b_bh;
344 }
345 
bh2jh(struct buffer_head * bh)346 static inline struct journal_head *bh2jh(struct buffer_head *bh)
347 {
348 	return bh->b_private;
349 }
350 
jbd_lock_bh_journal_head(struct buffer_head * bh)351 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
352 {
353 	bit_spin_lock(BH_JournalHead, &bh->b_state);
354 }
355 
jbd_unlock_bh_journal_head(struct buffer_head * bh)356 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
357 {
358 	bit_spin_unlock(BH_JournalHead, &bh->b_state);
359 }
360 
361 #define J_ASSERT(assert)	BUG_ON(!(assert))
362 
363 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
364 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
365 
366 #if defined(JBD2_PARANOID_IOFAIL)
367 #define J_EXPECT(expr, why...)		J_ASSERT(expr)
368 #define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
369 #define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
370 #else
371 #define __journal_expect(expr, why...)					     \
372 	({								     \
373 		int val = (expr);					     \
374 		if (!val) {						     \
375 			printk(KERN_ERR					     \
376 			       "JBD2 unexpected failure: %s: %s;\n",	     \
377 			       __func__, #expr);			     \
378 			printk(KERN_ERR why "\n");			     \
379 		}							     \
380 		val;							     \
381 	})
382 #define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
383 #define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
384 #define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
385 #endif
386 
387 /* Flags in jbd_inode->i_flags */
388 #define __JI_COMMIT_RUNNING 0
389 #define __JI_WRITE_DATA 1
390 #define __JI_WAIT_DATA 2
391 
392 /*
393  * Commit of the inode data in progress. We use this flag to protect us from
394  * concurrent deletion of inode. We cannot use reference to inode for this
395  * since we cannot afford doing last iput() on behalf of kjournald
396  */
397 #define JI_COMMIT_RUNNING (1 << __JI_COMMIT_RUNNING)
398 /* Write allocated dirty buffers in this inode before commit */
399 #define JI_WRITE_DATA (1 << __JI_WRITE_DATA)
400 /* Wait for outstanding data writes for this inode before commit */
401 #define JI_WAIT_DATA (1 << __JI_WAIT_DATA)
402 
403 /**
404  * struct jbd2_inode - The jbd_inode type is the structure linking inodes in
405  * ordered mode present in a transaction so that we can sync them during commit.
406  */
407 struct jbd2_inode {
408 	/**
409 	 * @i_transaction:
410 	 *
411 	 * Which transaction does this inode belong to? Either the running
412 	 * transaction or the committing one. [j_list_lock]
413 	 */
414 	transaction_t *i_transaction;
415 
416 	/**
417 	 * @i_next_transaction:
418 	 *
419 	 * Pointer to the running transaction modifying inode's data in case
420 	 * there is already a committing transaction touching it. [j_list_lock]
421 	 */
422 	transaction_t *i_next_transaction;
423 
424 	/**
425 	 * @i_list: List of inodes in the i_transaction [j_list_lock]
426 	 */
427 	struct list_head i_list;
428 
429 	/**
430 	 * @i_vfs_inode:
431 	 *
432 	 * VFS inode this inode belongs to [constant for lifetime of structure]
433 	 */
434 	struct inode *i_vfs_inode;
435 
436 	/**
437 	 * @i_flags: Flags of inode [j_list_lock]
438 	 */
439 	unsigned long i_flags;
440 
441 	/**
442 	 * @i_dirty_start:
443 	 *
444 	 * Offset in bytes where the dirty range for this inode starts.
445 	 * [j_list_lock]
446 	 */
447 	loff_t i_dirty_start;
448 
449 	/**
450 	 * @i_dirty_end:
451 	 *
452 	 * Inclusive offset in bytes where the dirty range for this inode
453 	 * ends. [j_list_lock]
454 	 */
455 	loff_t i_dirty_end;
456 };
457 
458 struct jbd2_revoke_table_s;
459 
460 /**
461  * struct jbd2_journal_handle - The jbd2_journal_handle type is the concrete
462  *     type associated with handle_t.
463  * @h_transaction: Which compound transaction is this update a part of?
464  * @h_journal: Which journal handle belongs to - used iff h_reserved set.
465  * @h_rsv_handle: Handle reserved for finishing the logical operation.
466  * @h_total_credits: Number of remaining buffers we are allowed to add to
467  *	journal. These are dirty buffers and revoke descriptor blocks.
468  * @h_revoke_credits: Number of remaining revoke records available for handle
469  * @h_ref: Reference count on this handle.
470  * @h_err: Field for caller's use to track errors through large fs operations.
471  * @h_sync: Flag for sync-on-close.
472  * @h_jdata: Flag to force data journaling.
473  * @h_reserved: Flag for handle for reserved credits.
474  * @h_aborted: Flag indicating fatal error on handle.
475  * @h_type: For handle statistics.
476  * @h_line_no: For handle statistics.
477  * @h_start_jiffies: Handle Start time.
478  * @h_requested_credits: Holds @h_total_credits after handle is started.
479  * @h_revoke_credits_requested: Holds @h_revoke_credits after handle is started.
480  * @saved_alloc_context: Saved context while transaction is open.
481  **/
482 
483 /* Docbook can't yet cope with the bit fields, but will leave the documentation
484  * in so it can be fixed later.
485  */
486 
487 struct jbd2_journal_handle
488 {
489 	union {
490 		transaction_t	*h_transaction;
491 		/* Which journal handle belongs to - used iff h_reserved set */
492 		journal_t	*h_journal;
493 	};
494 
495 	handle_t		*h_rsv_handle;
496 	int			h_total_credits;
497 	int			h_revoke_credits;
498 	int			h_revoke_credits_requested;
499 	int			h_ref;
500 	int			h_err;
501 
502 	/* Flags [no locking] */
503 	unsigned int	h_sync:		1;
504 	unsigned int	h_jdata:	1;
505 	unsigned int	h_reserved:	1;
506 	unsigned int	h_aborted:	1;
507 	unsigned int	h_type:		8;
508 	unsigned int	h_line_no:	16;
509 
510 	unsigned long		h_start_jiffies;
511 	unsigned int		h_requested_credits;
512 
513 	unsigned int		saved_alloc_context;
514 };
515 
516 
517 /*
518  * Some stats for checkpoint phase
519  */
520 struct transaction_chp_stats_s {
521 	unsigned long		cs_chp_time;
522 	__u32			cs_forced_to_close;
523 	__u32			cs_written;
524 	__u32			cs_dropped;
525 };
526 
527 /* The transaction_t type is the guts of the journaling mechanism.  It
528  * tracks a compound transaction through its various states:
529  *
530  * RUNNING:	accepting new updates
531  * LOCKED:	Updates still running but we don't accept new ones
532  * RUNDOWN:	Updates are tidying up but have finished requesting
533  *		new buffers to modify (state not used for now)
534  * FLUSH:       All updates complete, but we are still writing to disk
535  * COMMIT:      All data on disk, writing commit record
536  * FINISHED:	We still have to keep the transaction for checkpointing.
537  *
538  * The transaction keeps track of all of the buffers modified by a
539  * running transaction, and all of the buffers committed but not yet
540  * flushed to home for finished transactions.
541  * (Locking Documentation improved by LockDoc)
542  */
543 
544 /*
545  * Lock ranking:
546  *
547  *    j_list_lock
548  *      ->jbd_lock_bh_journal_head()	(This is "innermost")
549  *
550  *    j_state_lock
551  *    ->b_state_lock
552  *
553  *    b_state_lock
554  *    ->j_list_lock
555  *
556  *    j_state_lock
557  *    ->j_list_lock			(journal_unmap_buffer)
558  *
559  */
560 
561 struct transaction_s
562 {
563 	/* Pointer to the journal for this transaction. [no locking] */
564 	journal_t		*t_journal;
565 
566 	/* Sequence number for this transaction [no locking] */
567 	tid_t			t_tid;
568 
569 	/*
570 	 * Transaction's current state
571 	 * [no locking - only kjournald2 alters this]
572 	 * [j_list_lock] guards transition of a transaction into T_FINISHED
573 	 * state and subsequent call of __jbd2_journal_drop_transaction()
574 	 * FIXME: needs barriers
575 	 * KLUDGE: [use j_state_lock]
576 	 */
577 	enum {
578 		T_RUNNING,
579 		T_LOCKED,
580 		T_SWITCH,
581 		T_FLUSH,
582 		T_COMMIT,
583 		T_COMMIT_DFLUSH,
584 		T_COMMIT_JFLUSH,
585 		T_COMMIT_CALLBACK,
586 		T_FINISHED
587 	}			t_state;
588 
589 	/*
590 	 * Where in the log does this transaction's commit start? [no locking]
591 	 */
592 	unsigned long		t_log_start;
593 
594 	/*
595 	 * Number of buffers on the t_buffers list [j_list_lock, no locks
596 	 * needed for jbd2 thread]
597 	 */
598 	int			t_nr_buffers;
599 
600 	/*
601 	 * Doubly-linked circular list of all buffers reserved but not yet
602 	 * modified by this transaction [j_list_lock, no locks needed fo
603 	 * jbd2 thread]
604 	 */
605 	struct journal_head	*t_reserved_list;
606 
607 	/*
608 	 * Doubly-linked circular list of all metadata buffers owned by this
609 	 * transaction [j_list_lock, no locks needed for jbd2 thread]
610 	 */
611 	struct journal_head	*t_buffers;
612 
613 	/*
614 	 * Doubly-linked circular list of all forget buffers (superseded
615 	 * buffers which we can un-checkpoint once this transaction commits)
616 	 * [j_list_lock]
617 	 */
618 	struct journal_head	*t_forget;
619 
620 	/*
621 	 * Doubly-linked circular list of all buffers still to be flushed before
622 	 * this transaction can be checkpointed. [j_list_lock]
623 	 */
624 	struct journal_head	*t_checkpoint_list;
625 
626 	/*
627 	 * Doubly-linked circular list of all buffers submitted for IO while
628 	 * checkpointing. [j_list_lock]
629 	 */
630 	struct journal_head	*t_checkpoint_io_list;
631 
632 	/*
633 	 * Doubly-linked circular list of metadata buffers being
634 	 * shadowed by log IO.  The IO buffers on the iobuf list and
635 	 * the shadow buffers on this list match each other one for
636 	 * one at all times. [j_list_lock, no locks needed for jbd2
637 	 * thread]
638 	 */
639 	struct journal_head	*t_shadow_list;
640 
641 	/*
642 	 * List of inodes associated with the transaction; e.g., ext4 uses
643 	 * this to track inodes in data=ordered and data=journal mode that
644 	 * need special handling on transaction commit; also used by ocfs2.
645 	 * [j_list_lock]
646 	 */
647 	struct list_head	t_inode_list;
648 
649 	/*
650 	 * Protects info related to handles
651 	 */
652 	spinlock_t		t_handle_lock;
653 
654 	/*
655 	 * Longest time some handle had to wait for running transaction
656 	 */
657 	unsigned long		t_max_wait;
658 
659 	/*
660 	 * When transaction started
661 	 */
662 	unsigned long		t_start;
663 
664 	/*
665 	 * When commit was requested [j_state_lock]
666 	 */
667 	unsigned long		t_requested;
668 
669 	/*
670 	 * Checkpointing stats [j_list_lock]
671 	 */
672 	struct transaction_chp_stats_s t_chp_stats;
673 
674 	/*
675 	 * Number of outstanding updates running on this transaction
676 	 * [none]
677 	 */
678 	atomic_t		t_updates;
679 
680 	/*
681 	 * Number of blocks reserved for this transaction in the journal.
682 	 * This is including all credits reserved when starting transaction
683 	 * handles as well as all journal descriptor blocks needed for this
684 	 * transaction. [none]
685 	 */
686 	atomic_t		t_outstanding_credits;
687 
688 	/*
689 	 * Number of revoke records for this transaction added by already
690 	 * stopped handles. [none]
691 	 */
692 	atomic_t		t_outstanding_revokes;
693 
694 	/*
695 	 * How many handles used this transaction? [none]
696 	 */
697 	atomic_t		t_handle_count;
698 
699 	/*
700 	 * Forward and backward links for the circular list of all transactions
701 	 * awaiting checkpoint. [j_list_lock]
702 	 */
703 	transaction_t		*t_cpnext, *t_cpprev;
704 
705 	/*
706 	 * When will the transaction expire (become due for commit), in jiffies?
707 	 * [no locking]
708 	 */
709 	unsigned long		t_expires;
710 
711 	/*
712 	 * When this transaction started, in nanoseconds [no locking]
713 	 */
714 	ktime_t			t_start_time;
715 
716 	/*
717 	 * This transaction is being forced and some process is
718 	 * waiting for it to finish.
719 	 */
720 	unsigned int t_synchronous_commit:1;
721 
722 	/* Disk flush needs to be sent to fs partition [no locking] */
723 	int			t_need_data_flush;
724 
725 	/*
726 	 * For use by the filesystem to store fs-specific data
727 	 * structures associated with the transaction
728 	 */
729 	struct list_head	t_private_list;
730 };
731 
732 struct transaction_run_stats_s {
733 	unsigned long		rs_wait;
734 	unsigned long		rs_request_delay;
735 	unsigned long		rs_running;
736 	unsigned long		rs_locked;
737 	unsigned long		rs_flushing;
738 	unsigned long		rs_logging;
739 
740 	__u32			rs_handle_count;
741 	__u32			rs_blocks;
742 	__u32			rs_blocks_logged;
743 };
744 
745 struct transaction_stats_s {
746 	unsigned long		ts_tid;
747 	unsigned long		ts_requested;
748 	struct transaction_run_stats_s run;
749 };
750 
751 static inline unsigned long
jbd2_time_diff(unsigned long start,unsigned long end)752 jbd2_time_diff(unsigned long start, unsigned long end)
753 {
754 	if (end >= start)
755 		return end - start;
756 
757 	return end + (MAX_JIFFY_OFFSET - start);
758 }
759 
760 #define JBD2_NR_BATCH	64
761 
762 enum passtype {PASS_SCAN, PASS_REVOKE, PASS_REPLAY};
763 
764 #define JBD2_FC_REPLAY_STOP	0
765 #define JBD2_FC_REPLAY_CONTINUE	1
766 
767 /**
768  * struct journal_s - The journal_s type is the concrete type associated with
769  *     journal_t.
770  */
771 struct journal_s
772 {
773 	/**
774 	 * @j_flags: General journaling state flags [j_state_lock,
775 	 * no lock for quick racy checks]
776 	 */
777 	unsigned long		j_flags;
778 
779 	/**
780 	 * @j_atomic_flags: Atomic journaling state flags.
781 	 */
782 	unsigned long		j_atomic_flags;
783 
784 	/**
785 	 * @j_errno:
786 	 *
787 	 * Is there an outstanding uncleared error on the journal (from a prior
788 	 * abort)? [j_state_lock]
789 	 */
790 	int			j_errno;
791 
792 	/**
793 	 * @j_abort_mutex: Lock the whole aborting procedure.
794 	 */
795 	struct mutex		j_abort_mutex;
796 
797 	/**
798 	 * @j_sb_buffer: The first part of the superblock buffer.
799 	 */
800 	struct buffer_head	*j_sb_buffer;
801 
802 	/**
803 	 * @j_superblock: The second part of the superblock buffer.
804 	 */
805 	journal_superblock_t	*j_superblock;
806 
807 	/**
808 	 * @j_format_version: Version of the superblock format.
809 	 */
810 	int			j_format_version;
811 
812 	/**
813 	 * @j_state_lock: Protect the various scalars in the journal.
814 	 */
815 	rwlock_t		j_state_lock;
816 
817 	/**
818 	 * @j_barrier_count:
819 	 *
820 	 * Number of processes waiting to create a barrier lock [j_state_lock,
821 	 * no lock for quick racy checks]
822 	 */
823 	int			j_barrier_count;
824 
825 	/**
826 	 * @j_barrier: The barrier lock itself.
827 	 */
828 	struct mutex		j_barrier;
829 
830 	/**
831 	 * @j_running_transaction:
832 	 *
833 	 * Transactions: The current running transaction...
834 	 * [j_state_lock, no lock for quick racy checks] [caller holding
835 	 * open handle]
836 	 */
837 	transaction_t		*j_running_transaction;
838 
839 	/**
840 	 * @j_committing_transaction:
841 	 *
842 	 * the transaction we are pushing to disk
843 	 * [j_state_lock] [caller holding open handle]
844 	 */
845 	transaction_t		*j_committing_transaction;
846 
847 	/**
848 	 * @j_checkpoint_transactions:
849 	 *
850 	 * ... and a linked circular list of all transactions waiting for
851 	 * checkpointing. [j_list_lock]
852 	 */
853 	transaction_t		*j_checkpoint_transactions;
854 
855 	/**
856 	 * @j_wait_transaction_locked:
857 	 *
858 	 * Wait queue for waiting for a locked transaction to start committing,
859 	 * or for a barrier lock to be released.
860 	 */
861 	wait_queue_head_t	j_wait_transaction_locked;
862 
863 	/**
864 	 * @j_wait_done_commit: Wait queue for waiting for commit to complete.
865 	 */
866 	wait_queue_head_t	j_wait_done_commit;
867 
868 	/**
869 	 * @j_wait_commit: Wait queue to trigger commit.
870 	 */
871 	wait_queue_head_t	j_wait_commit;
872 
873 	/**
874 	 * @j_wait_updates: Wait queue to wait for updates to complete.
875 	 */
876 	wait_queue_head_t	j_wait_updates;
877 
878 	/**
879 	 * @j_wait_reserved:
880 	 *
881 	 * Wait queue to wait for reserved buffer credits to drop.
882 	 */
883 	wait_queue_head_t	j_wait_reserved;
884 
885 	/**
886 	 * @j_fc_wait:
887 	 *
888 	 * Wait queue to wait for completion of async fast commits.
889 	 */
890 	wait_queue_head_t	j_fc_wait;
891 
892 	/**
893 	 * @j_checkpoint_mutex:
894 	 *
895 	 * Semaphore for locking against concurrent checkpoints.
896 	 */
897 	struct mutex		j_checkpoint_mutex;
898 
899 	/**
900 	 * @j_chkpt_bhs:
901 	 *
902 	 * List of buffer heads used by the checkpoint routine.  This
903 	 * was moved from jbd2_log_do_checkpoint() to reduce stack
904 	 * usage.  Access to this array is controlled by the
905 	 * @j_checkpoint_mutex.  [j_checkpoint_mutex]
906 	 */
907 	struct buffer_head	*j_chkpt_bhs[JBD2_NR_BATCH];
908 
909 	/**
910 	 * @j_shrinker:
911 	 *
912 	 * Journal head shrinker, reclaim buffer's journal head which
913 	 * has been written back.
914 	 */
915 	struct shrinker		j_shrinker;
916 
917 	/**
918 	 * @j_checkpoint_jh_count:
919 	 *
920 	 * Number of journal buffers on the checkpoint list. [j_list_lock]
921 	 */
922 	struct percpu_counter	j_checkpoint_jh_count;
923 
924 	/**
925 	 * @j_shrink_transaction:
926 	 *
927 	 * Record next transaction will shrink on the checkpoint list.
928 	 * [j_list_lock]
929 	 */
930 	transaction_t		*j_shrink_transaction;
931 
932 	/**
933 	 * @j_head:
934 	 *
935 	 * Journal head: identifies the first unused block in the journal.
936 	 * [j_state_lock]
937 	 */
938 	unsigned long		j_head;
939 
940 	/**
941 	 * @j_tail:
942 	 *
943 	 * Journal tail: identifies the oldest still-used block in the journal.
944 	 * [j_state_lock]
945 	 */
946 	unsigned long		j_tail;
947 
948 	/**
949 	 * @j_free:
950 	 *
951 	 * Journal free: how many free blocks are there in the journal?
952 	 * [j_state_lock]
953 	 */
954 	unsigned long		j_free;
955 
956 	/**
957 	 * @j_first:
958 	 *
959 	 * The block number of the first usable block in the journal
960 	 * [j_state_lock].
961 	 */
962 	unsigned long		j_first;
963 
964 	/**
965 	 * @j_last:
966 	 *
967 	 * The block number one beyond the last usable block in the journal
968 	 * [j_state_lock].
969 	 */
970 	unsigned long		j_last;
971 
972 	/**
973 	 * @j_fc_first:
974 	 *
975 	 * The block number of the first fast commit block in the journal
976 	 * [j_state_lock].
977 	 */
978 	unsigned long		j_fc_first;
979 
980 	/**
981 	 * @j_fc_off:
982 	 *
983 	 * Number of fast commit blocks currently allocated. Accessed only
984 	 * during fast commit. Currently only process can do fast commit, so
985 	 * this field is not protected by any lock.
986 	 */
987 	unsigned long		j_fc_off;
988 
989 	/**
990 	 * @j_fc_last:
991 	 *
992 	 * The block number one beyond the last fast commit block in the journal
993 	 * [j_state_lock].
994 	 */
995 	unsigned long		j_fc_last;
996 
997 	/**
998 	 * @j_dev: Device where we store the journal.
999 	 */
1000 	struct block_device	*j_dev;
1001 
1002 	/**
1003 	 * @j_blocksize: Block size for the location where we store the journal.
1004 	 */
1005 	int			j_blocksize;
1006 
1007 	/**
1008 	 * @j_blk_offset:
1009 	 *
1010 	 * Starting block offset into the device where we store the journal.
1011 	 */
1012 	unsigned long long	j_blk_offset;
1013 
1014 	/**
1015 	 * @j_devname: Journal device name.
1016 	 */
1017 	char			j_devname[BDEVNAME_SIZE+24];
1018 
1019 	/**
1020 	 * @j_fs_dev:
1021 	 *
1022 	 * Device which holds the client fs.  For internal journal this will be
1023 	 * equal to j_dev.
1024 	 */
1025 	struct block_device	*j_fs_dev;
1026 
1027 	/**
1028 	 * @j_total_len: Total maximum capacity of the journal region on disk.
1029 	 */
1030 	unsigned int		j_total_len;
1031 
1032 	/**
1033 	 * @j_reserved_credits:
1034 	 *
1035 	 * Number of buffers reserved from the running transaction.
1036 	 */
1037 	atomic_t		j_reserved_credits;
1038 
1039 	/**
1040 	 * @j_list_lock: Protects the buffer lists and internal buffer state.
1041 	 */
1042 	spinlock_t		j_list_lock;
1043 
1044 	/**
1045 	 * @j_inode:
1046 	 *
1047 	 * Optional inode where we store the journal.  If present, all
1048 	 * journal block numbers are mapped into this inode via bmap().
1049 	 */
1050 	struct inode		*j_inode;
1051 
1052 	/**
1053 	 * @j_tail_sequence:
1054 	 *
1055 	 * Sequence number of the oldest transaction in the log [j_state_lock]
1056 	 */
1057 	tid_t			j_tail_sequence;
1058 
1059 	/**
1060 	 * @j_transaction_sequence:
1061 	 *
1062 	 * Sequence number of the next transaction to grant [j_state_lock]
1063 	 */
1064 	tid_t			j_transaction_sequence;
1065 
1066 	/**
1067 	 * @j_commit_sequence:
1068 	 *
1069 	 * Sequence number of the most recently committed transaction
1070 	 * [j_state_lock, no lock for quick racy checks]
1071 	 */
1072 	tid_t			j_commit_sequence;
1073 
1074 	/**
1075 	 * @j_commit_request:
1076 	 *
1077 	 * Sequence number of the most recent transaction wanting commit
1078 	 * [j_state_lock, no lock for quick racy checks]
1079 	 */
1080 	tid_t			j_commit_request;
1081 
1082 	/**
1083 	 * @j_uuid:
1084 	 *
1085 	 * Journal uuid: identifies the object (filesystem, LVM volume etc)
1086 	 * backed by this journal.  This will eventually be replaced by an array
1087 	 * of uuids, allowing us to index multiple devices within a single
1088 	 * journal and to perform atomic updates across them.
1089 	 */
1090 	__u8			j_uuid[16];
1091 
1092 	/**
1093 	 * @j_task: Pointer to the current commit thread for this journal.
1094 	 */
1095 	struct task_struct	*j_task;
1096 
1097 	/**
1098 	 * @j_max_transaction_buffers:
1099 	 *
1100 	 * Maximum number of metadata buffers to allow in a single compound
1101 	 * commit transaction.
1102 	 */
1103 	int			j_max_transaction_buffers;
1104 
1105 	/**
1106 	 * @j_revoke_records_per_block:
1107 	 *
1108 	 * Number of revoke records that fit in one descriptor block.
1109 	 */
1110 	int			j_revoke_records_per_block;
1111 
1112 	/**
1113 	 * @j_commit_interval:
1114 	 *
1115 	 * What is the maximum transaction lifetime before we begin a commit?
1116 	 */
1117 	unsigned long		j_commit_interval;
1118 
1119 	/**
1120 	 * @j_commit_timer: The timer used to wakeup the commit thread.
1121 	 */
1122 	struct timer_list	j_commit_timer;
1123 
1124 	/**
1125 	 * @j_revoke_lock: Protect the revoke table.
1126 	 */
1127 	spinlock_t		j_revoke_lock;
1128 
1129 	/**
1130 	 * @j_revoke:
1131 	 *
1132 	 * The revoke table - maintains the list of revoked blocks in the
1133 	 * current transaction.
1134 	 */
1135 	struct jbd2_revoke_table_s *j_revoke;
1136 
1137 	/**
1138 	 * @j_revoke_table: Alternate revoke tables for j_revoke.
1139 	 */
1140 	struct jbd2_revoke_table_s *j_revoke_table[2];
1141 
1142 	/**
1143 	 * @j_wbuf: Array of bhs for jbd2_journal_commit_transaction.
1144 	 */
1145 	struct buffer_head	**j_wbuf;
1146 
1147 	/**
1148 	 * @j_fc_wbuf: Array of fast commit bhs for fast commit. Accessed only
1149 	 * during a fast commit. Currently only process can do fast commit, so
1150 	 * this field is not protected by any lock.
1151 	 */
1152 	struct buffer_head	**j_fc_wbuf;
1153 
1154 	/**
1155 	 * @j_wbufsize:
1156 	 *
1157 	 * Size of @j_wbuf array.
1158 	 */
1159 	int			j_wbufsize;
1160 
1161 	/**
1162 	 * @j_fc_wbufsize:
1163 	 *
1164 	 * Size of @j_fc_wbuf array.
1165 	 */
1166 	int			j_fc_wbufsize;
1167 
1168 	/**
1169 	 * @j_last_sync_writer:
1170 	 *
1171 	 * The pid of the last person to run a synchronous operation
1172 	 * through the journal.
1173 	 */
1174 	pid_t			j_last_sync_writer;
1175 
1176 	/**
1177 	 * @j_average_commit_time:
1178 	 *
1179 	 * The average amount of time in nanoseconds it takes to commit a
1180 	 * transaction to disk. [j_state_lock]
1181 	 */
1182 	u64			j_average_commit_time;
1183 
1184 	/**
1185 	 * @j_min_batch_time:
1186 	 *
1187 	 * Minimum time that we should wait for additional filesystem operations
1188 	 * to get batched into a synchronous handle in microseconds.
1189 	 */
1190 	u32			j_min_batch_time;
1191 
1192 	/**
1193 	 * @j_max_batch_time:
1194 	 *
1195 	 * Maximum time that we should wait for additional filesystem operations
1196 	 * to get batched into a synchronous handle in microseconds.
1197 	 */
1198 	u32			j_max_batch_time;
1199 
1200 	/**
1201 	 * @j_commit_callback:
1202 	 *
1203 	 * This function is called when a transaction is closed.
1204 	 */
1205 	void			(*j_commit_callback)(journal_t *,
1206 						     transaction_t *);
1207 
1208 	/**
1209 	 * @j_submit_inode_data_buffers:
1210 	 *
1211 	 * This function is called for all inodes associated with the
1212 	 * committing transaction marked with JI_WRITE_DATA flag
1213 	 * before we start to write out the transaction to the journal.
1214 	 */
1215 	int			(*j_submit_inode_data_buffers)
1216 					(struct jbd2_inode *);
1217 
1218 	/**
1219 	 * @j_finish_inode_data_buffers:
1220 	 *
1221 	 * This function is called for all inodes associated with the
1222 	 * committing transaction marked with JI_WAIT_DATA flag
1223 	 * after we have written the transaction to the journal
1224 	 * but before we write out the commit block.
1225 	 */
1226 	int			(*j_finish_inode_data_buffers)
1227 					(struct jbd2_inode *);
1228 
1229 	/*
1230 	 * Journal statistics
1231 	 */
1232 
1233 	/**
1234 	 * @j_history_lock: Protect the transactions statistics history.
1235 	 */
1236 	spinlock_t		j_history_lock;
1237 
1238 	/**
1239 	 * @j_proc_entry: procfs entry for the jbd statistics directory.
1240 	 */
1241 	struct proc_dir_entry	*j_proc_entry;
1242 
1243 	/**
1244 	 * @j_stats: Overall statistics.
1245 	 */
1246 	struct transaction_stats_s j_stats;
1247 
1248 	/**
1249 	 * @j_failed_commit: Failed journal commit ID.
1250 	 */
1251 	unsigned int		j_failed_commit;
1252 
1253 	/**
1254 	 * @j_private:
1255 	 *
1256 	 * An opaque pointer to fs-private information.  ext3 puts its
1257 	 * superblock pointer here.
1258 	 */
1259 	void *j_private;
1260 
1261 	/**
1262 	 * @j_chksum_driver:
1263 	 *
1264 	 * Reference to checksum algorithm driver via cryptoapi.
1265 	 */
1266 	struct crypto_shash *j_chksum_driver;
1267 
1268 	/**
1269 	 * @j_csum_seed:
1270 	 *
1271 	 * Precomputed journal UUID checksum for seeding other checksums.
1272 	 */
1273 	__u32 j_csum_seed;
1274 
1275 #ifdef CONFIG_DEBUG_LOCK_ALLOC
1276 	/**
1277 	 * @j_trans_commit_map:
1278 	 *
1279 	 * Lockdep entity to track transaction commit dependencies. Handles
1280 	 * hold this "lock" for read, when we wait for commit, we acquire the
1281 	 * "lock" for writing. This matches the properties of jbd2 journalling
1282 	 * where the running transaction has to wait for all handles to be
1283 	 * dropped to commit that transaction and also acquiring a handle may
1284 	 * require transaction commit to finish.
1285 	 */
1286 	struct lockdep_map	j_trans_commit_map;
1287 #endif
1288 
1289 	/**
1290 	 * @j_fc_cleanup_callback:
1291 	 *
1292 	 * Clean-up after fast commit or full commit. JBD2 calls this function
1293 	 * after every commit operation.
1294 	 */
1295 	void (*j_fc_cleanup_callback)(struct journal_s *journal, int full, tid_t tid);
1296 
1297 	/**
1298 	 * @j_fc_replay_callback:
1299 	 *
1300 	 * File-system specific function that performs replay of a fast
1301 	 * commit. JBD2 calls this function for each fast commit block found in
1302 	 * the journal. This function should return JBD2_FC_REPLAY_CONTINUE
1303 	 * to indicate that the block was processed correctly and more fast
1304 	 * commit replay should continue. Return value of JBD2_FC_REPLAY_STOP
1305 	 * indicates the end of replay (no more blocks remaining). A negative
1306 	 * return value indicates error.
1307 	 */
1308 	int (*j_fc_replay_callback)(struct journal_s *journal,
1309 				    struct buffer_head *bh,
1310 				    enum passtype pass, int off,
1311 				    tid_t expected_commit_id);
1312 };
1313 
1314 #define jbd2_might_wait_for_commit(j) \
1315 	do { \
1316 		rwsem_acquire(&j->j_trans_commit_map, 0, 0, _THIS_IP_); \
1317 		rwsem_release(&j->j_trans_commit_map, _THIS_IP_); \
1318 	} while (0)
1319 
1320 /* journal feature predicate functions */
1321 #define JBD2_FEATURE_COMPAT_FUNCS(name, flagname) \
1322 static inline bool jbd2_has_feature_##name(journal_t *j) \
1323 { \
1324 	return ((j)->j_format_version >= 2 && \
1325 		((j)->j_superblock->s_feature_compat & \
1326 		 cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname)) != 0); \
1327 } \
1328 static inline void jbd2_set_feature_##name(journal_t *j) \
1329 { \
1330 	(j)->j_superblock->s_feature_compat |= \
1331 		cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1332 } \
1333 static inline void jbd2_clear_feature_##name(journal_t *j) \
1334 { \
1335 	(j)->j_superblock->s_feature_compat &= \
1336 		~cpu_to_be32(JBD2_FEATURE_COMPAT_##flagname); \
1337 }
1338 
1339 #define JBD2_FEATURE_RO_COMPAT_FUNCS(name, flagname) \
1340 static inline bool jbd2_has_feature_##name(journal_t *j) \
1341 { \
1342 	return ((j)->j_format_version >= 2 && \
1343 		((j)->j_superblock->s_feature_ro_compat & \
1344 		 cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname)) != 0); \
1345 } \
1346 static inline void jbd2_set_feature_##name(journal_t *j) \
1347 { \
1348 	(j)->j_superblock->s_feature_ro_compat |= \
1349 		cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1350 } \
1351 static inline void jbd2_clear_feature_##name(journal_t *j) \
1352 { \
1353 	(j)->j_superblock->s_feature_ro_compat &= \
1354 		~cpu_to_be32(JBD2_FEATURE_RO_COMPAT_##flagname); \
1355 }
1356 
1357 #define JBD2_FEATURE_INCOMPAT_FUNCS(name, flagname) \
1358 static inline bool jbd2_has_feature_##name(journal_t *j) \
1359 { \
1360 	return ((j)->j_format_version >= 2 && \
1361 		((j)->j_superblock->s_feature_incompat & \
1362 		 cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname)) != 0); \
1363 } \
1364 static inline void jbd2_set_feature_##name(journal_t *j) \
1365 { \
1366 	(j)->j_superblock->s_feature_incompat |= \
1367 		cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1368 } \
1369 static inline void jbd2_clear_feature_##name(journal_t *j) \
1370 { \
1371 	(j)->j_superblock->s_feature_incompat &= \
1372 		~cpu_to_be32(JBD2_FEATURE_INCOMPAT_##flagname); \
1373 }
1374 
1375 JBD2_FEATURE_COMPAT_FUNCS(checksum,		CHECKSUM)
1376 
1377 JBD2_FEATURE_INCOMPAT_FUNCS(revoke,		REVOKE)
1378 JBD2_FEATURE_INCOMPAT_FUNCS(64bit,		64BIT)
1379 JBD2_FEATURE_INCOMPAT_FUNCS(async_commit,	ASYNC_COMMIT)
1380 JBD2_FEATURE_INCOMPAT_FUNCS(csum2,		CSUM_V2)
1381 JBD2_FEATURE_INCOMPAT_FUNCS(csum3,		CSUM_V3)
1382 JBD2_FEATURE_INCOMPAT_FUNCS(fast_commit,	FAST_COMMIT)
1383 
1384 /*
1385  * Journal flag definitions
1386  */
1387 #define JBD2_UNMOUNT	0x001	/* Journal thread is being destroyed */
1388 #define JBD2_ABORT	0x002	/* Journaling has been aborted for errors. */
1389 #define JBD2_ACK_ERR	0x004	/* The errno in the sb has been acked */
1390 #define JBD2_FLUSHED	0x008	/* The journal superblock has been flushed */
1391 #define JBD2_LOADED	0x010	/* The journal superblock has been loaded */
1392 #define JBD2_BARRIER	0x020	/* Use IDE barriers */
1393 #define JBD2_ABORT_ON_SYNCDATA_ERR	0x040	/* Abort the journal on file
1394 						 * data write error in ordered
1395 						 * mode */
1396 #define JBD2_FAST_COMMIT_ONGOING	0x100	/* Fast commit is ongoing */
1397 #define JBD2_FULL_COMMIT_ONGOING	0x200	/* Full commit is ongoing */
1398 #define JBD2_JOURNAL_FLUSH_DISCARD	0x0001
1399 #define JBD2_JOURNAL_FLUSH_ZEROOUT	0x0002
1400 #define JBD2_JOURNAL_FLUSH_VALID	(JBD2_JOURNAL_FLUSH_DISCARD | \
1401 					JBD2_JOURNAL_FLUSH_ZEROOUT)
1402 
1403 /*
1404  * Journal atomic flag definitions
1405  */
1406 #define JBD2_CHECKPOINT_IO_ERROR	0x001	/* Detect io error while writing
1407 						 * buffer back to disk */
1408 
1409 /*
1410  * Function declarations for the journaling transaction and buffer
1411  * management
1412  */
1413 
1414 /* Filing buffers */
1415 extern void jbd2_journal_unfile_buffer(journal_t *, struct journal_head *);
1416 extern bool __jbd2_journal_refile_buffer(struct journal_head *);
1417 extern void jbd2_journal_refile_buffer(journal_t *, struct journal_head *);
1418 extern void __jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
1419 extern void jbd2_journal_file_buffer(struct journal_head *, transaction_t *, int);
jbd2_file_log_bh(struct list_head * head,struct buffer_head * bh)1420 static inline void jbd2_file_log_bh(struct list_head *head, struct buffer_head *bh)
1421 {
1422 	list_add_tail(&bh->b_assoc_buffers, head);
1423 }
jbd2_unfile_log_bh(struct buffer_head * bh)1424 static inline void jbd2_unfile_log_bh(struct buffer_head *bh)
1425 {
1426 	list_del_init(&bh->b_assoc_buffers);
1427 }
1428 
1429 /* Log buffer allocation */
1430 struct buffer_head *jbd2_journal_get_descriptor_buffer(transaction_t *, int);
1431 void jbd2_descriptor_block_csum_set(journal_t *, struct buffer_head *);
1432 int jbd2_journal_next_log_block(journal_t *, unsigned long long *);
1433 int jbd2_journal_get_log_tail(journal_t *journal, tid_t *tid,
1434 			      unsigned long *block);
1435 int __jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1436 void jbd2_update_log_tail(journal_t *journal, tid_t tid, unsigned long block);
1437 
1438 /* Commit management */
1439 extern void jbd2_journal_commit_transaction(journal_t *);
1440 
1441 /* Checkpoint list management */
1442 void __jbd2_journal_clean_checkpoint_list(journal_t *journal, bool destroy);
1443 unsigned long jbd2_journal_shrink_checkpoint_list(journal_t *journal, unsigned long *nr_to_scan);
1444 int __jbd2_journal_remove_checkpoint(struct journal_head *);
1445 void jbd2_journal_destroy_checkpoint(journal_t *journal);
1446 void __jbd2_journal_insert_checkpoint(struct journal_head *, transaction_t *);
1447 
1448 
1449 /*
1450  * Triggers
1451  */
1452 
1453 struct jbd2_buffer_trigger_type {
1454 	/*
1455 	 * Fired a the moment data to write to the journal are known to be
1456 	 * stable - so either at the moment b_frozen_data is created or just
1457 	 * before a buffer is written to the journal.  mapped_data is a mapped
1458 	 * buffer that is the frozen data for commit.
1459 	 */
1460 	void (*t_frozen)(struct jbd2_buffer_trigger_type *type,
1461 			 struct buffer_head *bh, void *mapped_data,
1462 			 size_t size);
1463 
1464 	/*
1465 	 * Fired during journal abort for dirty buffers that will not be
1466 	 * committed.
1467 	 */
1468 	void (*t_abort)(struct jbd2_buffer_trigger_type *type,
1469 			struct buffer_head *bh);
1470 };
1471 
1472 extern void jbd2_buffer_frozen_trigger(struct journal_head *jh,
1473 				       void *mapped_data,
1474 				       struct jbd2_buffer_trigger_type *triggers);
1475 extern void jbd2_buffer_abort_trigger(struct journal_head *jh,
1476 				      struct jbd2_buffer_trigger_type *triggers);
1477 
1478 /* Buffer IO */
1479 extern int jbd2_journal_write_metadata_buffer(transaction_t *transaction,
1480 					      struct journal_head *jh_in,
1481 					      struct buffer_head **bh_out,
1482 					      sector_t blocknr);
1483 
1484 /* Transaction cache support */
1485 extern void jbd2_journal_destroy_transaction_cache(void);
1486 extern int __init jbd2_journal_init_transaction_cache(void);
1487 extern void jbd2_journal_free_transaction(transaction_t *);
1488 
1489 /*
1490  * Journal locking.
1491  *
1492  * We need to lock the journal during transaction state changes so that nobody
1493  * ever tries to take a handle on the running transaction while we are in the
1494  * middle of moving it to the commit phase.  j_state_lock does this.
1495  *
1496  * Note that the locking is completely interrupt unsafe.  We never touch
1497  * journal structures from interrupts.
1498  */
1499 
journal_current_handle(void)1500 static inline handle_t *journal_current_handle(void)
1501 {
1502 	return current->journal_info;
1503 }
1504 
1505 /* The journaling code user interface:
1506  *
1507  * Create and destroy handles
1508  * Register buffer modifications against the current transaction.
1509  */
1510 
1511 extern handle_t *jbd2_journal_start(journal_t *, int nblocks);
1512 extern handle_t *jbd2__journal_start(journal_t *, int blocks, int rsv_blocks,
1513 				     int revoke_records, gfp_t gfp_mask,
1514 				     unsigned int type, unsigned int line_no);
1515 extern int	 jbd2_journal_restart(handle_t *, int nblocks);
1516 extern int	 jbd2__journal_restart(handle_t *, int nblocks,
1517 				       int revoke_records, gfp_t gfp_mask);
1518 extern int	 jbd2_journal_start_reserved(handle_t *handle,
1519 				unsigned int type, unsigned int line_no);
1520 extern void	 jbd2_journal_free_reserved(handle_t *handle);
1521 extern int	 jbd2_journal_extend(handle_t *handle, int nblocks,
1522 				     int revoke_records);
1523 extern int	 jbd2_journal_get_write_access(handle_t *, struct buffer_head *);
1524 extern int	 jbd2_journal_get_create_access (handle_t *, struct buffer_head *);
1525 extern int	 jbd2_journal_get_undo_access(handle_t *, struct buffer_head *);
1526 void		 jbd2_journal_set_triggers(struct buffer_head *,
1527 					   struct jbd2_buffer_trigger_type *type);
1528 extern int	 jbd2_journal_dirty_metadata (handle_t *, struct buffer_head *);
1529 extern int	 jbd2_journal_forget (handle_t *, struct buffer_head *);
1530 int jbd2_journal_invalidate_folio(journal_t *, struct folio *,
1531 					size_t offset, size_t length);
1532 bool jbd2_journal_try_to_free_buffers(journal_t *journal, struct folio *folio);
1533 extern int	 jbd2_journal_stop(handle_t *);
1534 extern int	 jbd2_journal_flush(journal_t *journal, unsigned int flags);
1535 extern void	 jbd2_journal_lock_updates (journal_t *);
1536 extern void	 jbd2_journal_unlock_updates (journal_t *);
1537 
1538 void jbd2_journal_wait_updates(journal_t *);
1539 
1540 extern journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1541 				struct block_device *fs_dev,
1542 				unsigned long long start, int len, int bsize);
1543 extern journal_t * jbd2_journal_init_inode (struct inode *);
1544 extern int	   jbd2_journal_update_format (journal_t *);
1545 extern int	   jbd2_journal_check_used_features
1546 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1547 extern int	   jbd2_journal_check_available_features
1548 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1549 extern int	   jbd2_journal_set_features
1550 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1551 extern void	   jbd2_journal_clear_features
1552 		   (journal_t *, unsigned long, unsigned long, unsigned long);
1553 extern int	   jbd2_journal_load       (journal_t *journal);
1554 extern int	   jbd2_journal_destroy    (journal_t *);
1555 extern int	   jbd2_journal_recover    (journal_t *journal);
1556 extern int	   jbd2_journal_wipe       (journal_t *, int);
1557 extern int	   jbd2_journal_skip_recovery	(journal_t *);
1558 extern void	   jbd2_journal_update_sb_errno(journal_t *);
1559 extern int	   jbd2_journal_update_sb_log_tail	(journal_t *, tid_t,
1560 				unsigned long, int);
1561 extern void	   jbd2_journal_abort      (journal_t *, int);
1562 extern int	   jbd2_journal_errno      (journal_t *);
1563 extern void	   jbd2_journal_ack_err    (journal_t *);
1564 extern int	   jbd2_journal_clear_err  (journal_t *);
1565 extern int	   jbd2_journal_bmap(journal_t *, unsigned long, unsigned long long *);
1566 extern int	   jbd2_journal_force_commit(journal_t *);
1567 extern int	   jbd2_journal_force_commit_nested(journal_t *);
1568 extern int	   jbd2_journal_inode_ranged_write(handle_t *handle,
1569 			struct jbd2_inode *inode, loff_t start_byte,
1570 			loff_t length);
1571 extern int	   jbd2_journal_inode_ranged_wait(handle_t *handle,
1572 			struct jbd2_inode *inode, loff_t start_byte,
1573 			loff_t length);
1574 extern int	   jbd2_journal_submit_inode_data_buffers(
1575 			struct jbd2_inode *jinode);
1576 extern int	   jbd2_journal_finish_inode_data_buffers(
1577 			struct jbd2_inode *jinode);
1578 extern int	   jbd2_journal_begin_ordered_truncate(journal_t *journal,
1579 				struct jbd2_inode *inode, loff_t new_size);
1580 extern void	   jbd2_journal_init_jbd_inode(struct jbd2_inode *jinode, struct inode *inode);
1581 extern void	   jbd2_journal_release_jbd_inode(journal_t *journal, struct jbd2_inode *jinode);
1582 
1583 /*
1584  * journal_head management
1585  */
1586 struct journal_head *jbd2_journal_add_journal_head(struct buffer_head *bh);
1587 struct journal_head *jbd2_journal_grab_journal_head(struct buffer_head *bh);
1588 void jbd2_journal_put_journal_head(struct journal_head *jh);
1589 
1590 /*
1591  * handle management
1592  */
1593 extern struct kmem_cache *jbd2_handle_cache;
1594 
jbd2_alloc_handle(gfp_t gfp_flags)1595 static inline handle_t *jbd2_alloc_handle(gfp_t gfp_flags)
1596 {
1597 	return kmem_cache_zalloc(jbd2_handle_cache, gfp_flags);
1598 }
1599 
jbd2_free_handle(handle_t * handle)1600 static inline void jbd2_free_handle(handle_t *handle)
1601 {
1602 	kmem_cache_free(jbd2_handle_cache, handle);
1603 }
1604 
1605 /*
1606  * jbd2_inode management (optional, for those file systems that want to use
1607  * dynamically allocated jbd2_inode structures)
1608  */
1609 extern struct kmem_cache *jbd2_inode_cache;
1610 
jbd2_alloc_inode(gfp_t gfp_flags)1611 static inline struct jbd2_inode *jbd2_alloc_inode(gfp_t gfp_flags)
1612 {
1613 	return kmem_cache_alloc(jbd2_inode_cache, gfp_flags);
1614 }
1615 
jbd2_free_inode(struct jbd2_inode * jinode)1616 static inline void jbd2_free_inode(struct jbd2_inode *jinode)
1617 {
1618 	kmem_cache_free(jbd2_inode_cache, jinode);
1619 }
1620 
1621 /* Primary revoke support */
1622 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1623 extern int	   jbd2_journal_init_revoke(journal_t *, int);
1624 extern void	   jbd2_journal_destroy_revoke_record_cache(void);
1625 extern void	   jbd2_journal_destroy_revoke_table_cache(void);
1626 extern int __init jbd2_journal_init_revoke_record_cache(void);
1627 extern int __init jbd2_journal_init_revoke_table_cache(void);
1628 
1629 extern void	   jbd2_journal_destroy_revoke(journal_t *);
1630 extern int	   jbd2_journal_revoke (handle_t *, unsigned long long, struct buffer_head *);
1631 extern int	   jbd2_journal_cancel_revoke(handle_t *, struct journal_head *);
1632 extern void	   jbd2_journal_write_revoke_records(transaction_t *transaction,
1633 						     struct list_head *log_bufs);
1634 
1635 /* Recovery revoke support */
1636 extern int	jbd2_journal_set_revoke(journal_t *, unsigned long long, tid_t);
1637 extern int	jbd2_journal_test_revoke(journal_t *, unsigned long long, tid_t);
1638 extern void	jbd2_journal_clear_revoke(journal_t *);
1639 extern void	jbd2_journal_switch_revoke_table(journal_t *journal);
1640 extern void	jbd2_clear_buffer_revoked_flags(journal_t *journal);
1641 
1642 /*
1643  * The log thread user interface:
1644  *
1645  * Request space in the current transaction, and force transaction commit
1646  * transitions on demand.
1647  */
1648 
1649 int jbd2_log_start_commit(journal_t *journal, tid_t tid);
1650 int __jbd2_log_start_commit(journal_t *journal, tid_t tid);
1651 int jbd2_journal_start_commit(journal_t *journal, tid_t *tid);
1652 int jbd2_log_wait_commit(journal_t *journal, tid_t tid);
1653 int jbd2_transaction_committed(journal_t *journal, tid_t tid);
1654 int jbd2_complete_transaction(journal_t *journal, tid_t tid);
1655 int jbd2_log_do_checkpoint(journal_t *journal);
1656 int jbd2_trans_will_send_data_barrier(journal_t *journal, tid_t tid);
1657 
1658 void __jbd2_log_wait_for_space(journal_t *journal);
1659 extern void __jbd2_journal_drop_transaction(journal_t *, transaction_t *);
1660 extern int jbd2_cleanup_journal_tail(journal_t *);
1661 
1662 /* Fast commit related APIs */
1663 int jbd2_fc_begin_commit(journal_t *journal, tid_t tid);
1664 int jbd2_fc_end_commit(journal_t *journal);
1665 int jbd2_fc_end_commit_fallback(journal_t *journal);
1666 int jbd2_fc_get_buf(journal_t *journal, struct buffer_head **bh_out);
1667 int jbd2_submit_inode_data(struct jbd2_inode *jinode);
1668 int jbd2_wait_inode_data(journal_t *journal, struct jbd2_inode *jinode);
1669 int jbd2_fc_wait_bufs(journal_t *journal, int num_blks);
1670 int jbd2_fc_release_bufs(journal_t *journal);
1671 
jbd2_journal_get_max_txn_bufs(journal_t * journal)1672 static inline int jbd2_journal_get_max_txn_bufs(journal_t *journal)
1673 {
1674 	return (journal->j_total_len - journal->j_fc_wbufsize) / 4;
1675 }
1676 
1677 /*
1678  * is_journal_abort
1679  *
1680  * Simple test wrapper function to test the JBD2_ABORT state flag.  This
1681  * bit, when set, indicates that we have had a fatal error somewhere,
1682  * either inside the journaling layer or indicated to us by the client
1683  * (eg. ext3), and that we and should not commit any further
1684  * transactions.
1685  */
1686 
is_journal_aborted(journal_t * journal)1687 static inline int is_journal_aborted(journal_t *journal)
1688 {
1689 	return journal->j_flags & JBD2_ABORT;
1690 }
1691 
is_handle_aborted(handle_t * handle)1692 static inline int is_handle_aborted(handle_t *handle)
1693 {
1694 	if (handle->h_aborted || !handle->h_transaction)
1695 		return 1;
1696 	return is_journal_aborted(handle->h_transaction->t_journal);
1697 }
1698 
jbd2_journal_abort_handle(handle_t * handle)1699 static inline void jbd2_journal_abort_handle(handle_t *handle)
1700 {
1701 	handle->h_aborted = 1;
1702 }
1703 
1704 #endif /* __KERNEL__   */
1705 
1706 /* Comparison functions for transaction IDs: perform comparisons using
1707  * modulo arithmetic so that they work over sequence number wraps. */
1708 
tid_gt(tid_t x,tid_t y)1709 static inline int tid_gt(tid_t x, tid_t y)
1710 {
1711 	int difference = (x - y);
1712 	return (difference > 0);
1713 }
1714 
tid_geq(tid_t x,tid_t y)1715 static inline int tid_geq(tid_t x, tid_t y)
1716 {
1717 	int difference = (x - y);
1718 	return (difference >= 0);
1719 }
1720 
1721 extern int jbd2_journal_blocks_per_page(struct inode *inode);
1722 extern size_t journal_tag_bytes(journal_t *journal);
1723 
jbd2_journal_has_csum_v2or3_feature(journal_t * j)1724 static inline bool jbd2_journal_has_csum_v2or3_feature(journal_t *j)
1725 {
1726 	return jbd2_has_feature_csum2(j) || jbd2_has_feature_csum3(j);
1727 }
1728 
jbd2_journal_has_csum_v2or3(journal_t * journal)1729 static inline int jbd2_journal_has_csum_v2or3(journal_t *journal)
1730 {
1731 	WARN_ON_ONCE(jbd2_journal_has_csum_v2or3_feature(journal) &&
1732 		     journal->j_chksum_driver == NULL);
1733 
1734 	return journal->j_chksum_driver != NULL;
1735 }
1736 
jbd2_journal_get_num_fc_blks(journal_superblock_t * jsb)1737 static inline int jbd2_journal_get_num_fc_blks(journal_superblock_t *jsb)
1738 {
1739 	int num_fc_blocks = be32_to_cpu(jsb->s_num_fc_blks);
1740 
1741 	return num_fc_blocks ? num_fc_blocks : JBD2_DEFAULT_FAST_COMMIT_BLOCKS;
1742 }
1743 
1744 /*
1745  * Return number of free blocks in the log. Must be called under j_state_lock.
1746  */
jbd2_log_space_left(journal_t * journal)1747 static inline unsigned long jbd2_log_space_left(journal_t *journal)
1748 {
1749 	/* Allow for rounding errors */
1750 	long free = journal->j_free - 32;
1751 
1752 	if (journal->j_committing_transaction) {
1753 		free -= atomic_read(&journal->
1754                         j_committing_transaction->t_outstanding_credits);
1755 	}
1756 	return max_t(long, free, 0);
1757 }
1758 
1759 /*
1760  * Definitions which augment the buffer_head layer
1761  */
1762 
1763 /* journaling buffer types */
1764 #define BJ_None		0	/* Not journaled */
1765 #define BJ_Metadata	1	/* Normal journaled metadata */
1766 #define BJ_Forget	2	/* Buffer superseded by this transaction */
1767 #define BJ_Shadow	3	/* Buffer contents being shadowed to the log */
1768 #define BJ_Reserved	4	/* Buffer is reserved for access by journal */
1769 #define BJ_Types	5
1770 
1771 /* JBD uses a CRC32 checksum */
1772 #define JBD_MAX_CHECKSUM_SIZE 4
1773 
jbd2_chksum(journal_t * journal,u32 crc,const void * address,unsigned int length)1774 static inline u32 jbd2_chksum(journal_t *journal, u32 crc,
1775 			      const void *address, unsigned int length)
1776 {
1777 	struct {
1778 		struct shash_desc shash;
1779 		char ctx[JBD_MAX_CHECKSUM_SIZE];
1780 	} desc;
1781 	int err;
1782 
1783 	BUG_ON(crypto_shash_descsize(journal->j_chksum_driver) >
1784 		JBD_MAX_CHECKSUM_SIZE);
1785 
1786 	desc.shash.tfm = journal->j_chksum_driver;
1787 	*(u32 *)desc.ctx = crc;
1788 
1789 	err = crypto_shash_update(&desc.shash, address, length);
1790 	BUG_ON(err);
1791 
1792 	return *(u32 *)desc.ctx;
1793 }
1794 
1795 /* Return most recent uncommitted transaction */
jbd2_get_latest_transaction(journal_t * journal)1796 static inline tid_t  jbd2_get_latest_transaction(journal_t *journal)
1797 {
1798 	tid_t tid;
1799 
1800 	read_lock(&journal->j_state_lock);
1801 	tid = journal->j_commit_request;
1802 	if (journal->j_running_transaction)
1803 		tid = journal->j_running_transaction->t_tid;
1804 	read_unlock(&journal->j_state_lock);
1805 	return tid;
1806 }
1807 
jbd2_handle_buffer_credits(handle_t * handle)1808 static inline int jbd2_handle_buffer_credits(handle_t *handle)
1809 {
1810 	journal_t *journal;
1811 
1812 	if (!handle->h_reserved)
1813 		journal = handle->h_transaction->t_journal;
1814 	else
1815 		journal = handle->h_journal;
1816 
1817 	return handle->h_total_credits -
1818 		DIV_ROUND_UP(handle->h_revoke_credits_requested,
1819 			     journal->j_revoke_records_per_block);
1820 }
1821 
1822 #ifdef __KERNEL__
1823 
1824 #define buffer_trace_init(bh)	do {} while (0)
1825 #define print_buffer_fields(bh)	do {} while (0)
1826 #define print_buffer_trace(bh)	do {} while (0)
1827 #define BUFFER_TRACE(bh, info)	do {} while (0)
1828 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1829 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1830 
1831 #endif	/* __KERNEL__ */
1832 
1833 #define EFSBADCRC	EBADMSG		/* Bad CRC detected */
1834 #define EFSCORRUPTED	EUCLEAN		/* Filesystem is corrupted */
1835 
1836 #endif	/* _LINUX_JBD2_H */
1837