xref: /reactos/drivers/filesystems/ext2/inc/linux/jbd.h (revision c2c66aff)
1 /*
2  * linux/include/linux/jbd.h
3  *
4  * Written by Stephen C. Tweedie <sct@redhat.com>
5  *
6  * Copyright 1998-2000 Red Hat, Inc --- All Rights Reserved
7  *
8  * This file is part of the Linux kernel and is made available under
9  * the terms of the GNU General Public License, version 2, or at your
10  * option, any later version, incorporated herein by reference.
11  *
12  * Definitions for transaction data structures for the buffer cache
13  * filesystem journaling support.
14  */
15 
16 #ifndef _LINUX_JBD_H
17 #define _LINUX_JBD_H
18 
19 /* Allow this file to be included directly into e2fsprogs */
20 #ifndef __KERNEL__
21 #include "jfs_compat.h"
22 #define JFS_DEBUG
23 #define jfs_debug jbd_debug
24 #else
25 
26 #include <linux/module.h>
27 #include <linux/buffer_head.h>
28 #include <linux/journal-head.h>
29 #include <linux/stddef.h>
30 #include <linux/bit_spinlock.h>
31 #include <linux/mutex.h>
32 #include <linux/timer.h>
33 #include <linux/lockdep.h>
34 
35 #include <asm/semaphore.h>
36 
37 #define journal_oom_retry 1
38 
39 #define jbdlock_t FAST_MUTEX
40 
jbd_lock_init(jbdlock_t * lock)41 static inline void jbd_lock_init(jbdlock_t *lock)
42 {
43     ExInitializeFastMutex(lock);
44 }
45 
jbd_lock(jbdlock_t * lock)46 static inline void jbd_lock(jbdlock_t *lock)
47 {
48     ExAcquireFastMutex(lock);
49 }
50 
jbd_unlock(jbdlock_t * lock)51 static inline void jbd_unlock(jbdlock_t *lock)
52 {
53     ExReleaseFastMutex(lock);
54 }
55 
assert_jbd_locked(jbdlock_t * lock)56 static inline int assert_jbd_locked(jbdlock_t *lock)
57 {
58     int rc = 1;
59 
60     if (ExTryToAcquireFastMutex(lock)) {
61         ExReleaseFastMutex(lock);
62         rc = 0;
63     }
64 
65     return rc;
66 }
67 
68 
69 /*
70  * Define JBD_PARANIOD_IOFAIL to cause a kernel BUG() if ext3 finds
71  * certain classes of error which can occur due to failed IOs.  Under
72  * normal use we want ext3 to continue after such errors, because
73  * hardware _can_ fail, but for debugging purposes when running tests on
74  * known-good hardware we may want to trap these errors.
75  */
76 #undef JBD_PARANOID_IOFAIL
77 
78 /*
79  * The default maximum commit age, in seconds.
80  */
81 #define JBD_DEFAULT_MAX_COMMIT_AGE 5
82 
83 #ifdef CONFIG_JBD_DEBUG
84 /*
85  * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
86  * consistency checks.  By default we don't do this unless
87  * CONFIG_JBD_DEBUG is on.
88  */
89 #define JBD_EXPENSIVE_CHECKING
90 extern u8 journal_enable_debug;
91 
92 #define jbd_debug(n, f, a...)						\
93 	do {								\
94 		if ((n) <= journal_enable_debug) {			\
95 			printk (KERN_DEBUG "(%s, %d): %s: ",		\
96 				__FILE__, __LINE__, __FUNCTION__);	\
97 			printk (f, ## a);				\
98 		}							\
99 	} while (0)
100 #else
101 #define jbd_debug
102 #endif
103 
jbd_alloc(size_t size,gfp_t flags)104 static inline void *jbd_alloc(size_t size, gfp_t flags)
105 {
106     return kmalloc(size, flags);
107     /*__get_free_pages(flags, get_order(size));*/
108 }
109 
jbd_free(void * ptr,size_t size)110 static inline void jbd_free(void *ptr, size_t size)
111 {
112     if (ptr) {
113         kfree(ptr);
114     }
115     /* free_pages((unsigned long)ptr, get_order(size)); */
116 };
117 
118 #define JFS_MIN_JOURNAL_BLOCKS 1024
119 
120 
121 /**
122  * typedef handle_t - The handle_t type represents a single atomic update being performed by some process.
123  *
124  * All filesystem modifications made by the process go
125  * through this handle.  Recursive operations (such as quota operations)
126  * are gathered into a single update.
127  *
128  * The buffer credits field is used to account for journaled buffers
129  * being modified by the running process.  To ensure that there is
130  * enough log space for all outstanding operations, we need to limit the
131  * number of outstanding buffers possible at any time.  When the
132  * operation completes, any buffer credits not used are credited back to
133  * the transaction, so that at all times we know how many buffers the
134  * outstanding updates on a transaction might possibly touch.
135  *
136  * This is an opaque datatype.
137  **/
138 typedef struct handle_s		handle_t;	/* Atomic operation type */
139 
140 
141 /**
142  * typedef journal_t - The journal_t maintains all of the journaling state information for a single filesystem.
143  *
144  * journal_t is linked to from the fs superblock structure.
145  *
146  * We use the journal_t to keep track of all outstanding transaction
147  * activity on the filesystem, and to manage the state of the log
148  * writing process.
149  *
150  * This is an opaque datatype.
151  **/
152 typedef struct journal_s	journal_t;	/* Journal control structure */
153 #endif
154 
155 /*
156  * Internal structures used by the logging mechanism:
157  */
158 
159 #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
160 
161 /*
162  * On-disk structures
163  */
164 
165 /*
166  * Descriptor block types:
167  */
168 
169 #define JFS_DESCRIPTOR_BLOCK	1
170 #define JFS_COMMIT_BLOCK	2
171 #define JFS_SUPERBLOCK_V1	3
172 #define JFS_SUPERBLOCK_V2	4
173 #define JFS_REVOKE_BLOCK	5
174 
175 /*
176  * Standard header for all descriptor blocks:
177  */
178 typedef struct journal_header_s
179 {
180     __be32		h_magic;
181     __be32		h_blocktype;
182     __be32		h_sequence;
183 } journal_header_t;
184 
185 
186 /*
187  * The block tag: used to describe a single buffer in the journal
188  */
189 typedef struct journal_block_tag_s
190 {
191     __be32		t_blocknr;	/* The on-disk block number */
192     __be32		t_flags;	/* See below */
193 } journal_block_tag_t;
194 
195 /*
196  * The revoke descriptor: used on disk to describe a series of blocks to
197  * be revoked from the log
198  */
199 typedef struct journal_revoke_header_s
200 {
201     journal_header_t r_header;
202     __be32		 r_count;	/* Count of bytes used in the block */
203 } journal_revoke_header_t;
204 
205 
206 /* Definitions for the journal tag flags word: */
207 #define JFS_FLAG_ESCAPE		1	/* on-disk block is escaped */
208 #define JFS_FLAG_SAME_UUID	2	/* block has same uuid as previous */
209 #define JFS_FLAG_DELETED	4	/* block deleted by this transaction */
210 #define JFS_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
211 
212 
213 /*
214  * The journal superblock.  All fields are in big-endian byte order.
215  */
216 typedef struct journal_superblock_s
217 {
218     /* 0x0000 */
219     journal_header_t s_header;
220 
221     /* 0x000C */
222     /* Static information describing the journal */
223     __be32	s_blocksize;		/* journal device blocksize */
224     __be32	s_maxlen;		/* total blocks in journal file */
225     __be32	s_first;		/* first block of log information */
226 
227     /* 0x0018 */
228     /* Dynamic information describing the current state of the log */
229     __be32	s_sequence;		/* first commit ID expected in log */
230     __be32	s_start;		/* blocknr of start of log */
231 
232     /* 0x0020 */
233     /* Error value, as set by journal_abort(). */
234     __be32	s_errno;
235 
236     /* 0x0024 */
237     /* Remaining fields are only valid in a version-2 superblock */
238     __be32	s_feature_compat;	/* compatible feature set */
239     __be32	s_feature_incompat;	/* incompatible feature set */
240     __be32	s_feature_ro_compat;	/* readonly-compatible feature set */
241     /* 0x0030 */
242     __u8	s_uuid[16];		/* 128-bit uuid for journal */
243 
244     /* 0x0040 */
245     __be32	s_nr_users;		/* Nr of filesystems sharing log */
246 
247     __be32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
248 
249     /* 0x0048 */
250     __be32	s_max_transaction;	/* Limit of journal blocks per trans.*/
251     __be32	s_max_trans_data;	/* Limit of data blocks per trans. */
252 
253     /* 0x0050 */
254     __u32	s_padding[44];
255 
256     /* 0x0100 */
257     __u8	s_users[16*48];		/* ids of all fs'es sharing the log */
258     /* 0x0400 */
259 } journal_superblock_t;
260 
261 #define JFS_HAS_COMPAT_FEATURE(j,mask)					\
262 	((j)->j_format_version >= 2 &&					\
263 	 ((j)->j_superblock->s_feature_compat & cpu_to_be32((mask))))
264 #define JFS_HAS_RO_COMPAT_FEATURE(j,mask)				\
265 	((j)->j_format_version >= 2 &&					\
266 	 ((j)->j_superblock->s_feature_ro_compat & cpu_to_be32((mask))))
267 #define JFS_HAS_INCOMPAT_FEATURE(j,mask)				\
268 	((j)->j_format_version >= 2 &&					\
269 	 ((j)->j_superblock->s_feature_incompat & cpu_to_be32((mask))))
270 
271 #define JFS_FEATURE_INCOMPAT_REVOKE	0x00000001
272 
273 /* Features known to this kernel version: */
274 #define JFS_KNOWN_COMPAT_FEATURES	0
275 #define JFS_KNOWN_ROCOMPAT_FEATURES	0
276 #define JFS_KNOWN_INCOMPAT_FEATURES	JFS_FEATURE_INCOMPAT_REVOKE
277 
278 #ifdef __KERNEL__
279 
280 #include <linux/fs.h>
281 #include <linux/sched.h>
282 
283 #define J_ASSERT  ASSERT
284 
285 #if defined(CONFIG_BUFFER_DEBUG)
286 void buffer_assertion_failure(struct buffer_head *bh);
287 #define J_ASSERT_BH(bh, expr)						\
288 	do {								\
289 		if (!(expr))						\
290 			buffer_assertion_failure(bh);			\
291 		J_ASSERT(expr);						\
292 	} while (0)
293 #define J_ASSERT_JH(jh, expr)	J_ASSERT_BH(jh2bh(jh), expr)
294 #else
295 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
296 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
297 #endif
298 
299 #if defined(JBD_PARANOID_IOFAIL)
300 #define J_EXPECT(expr, why...)		J_ASSERT(expr)
301 #define J_EXPECT_BH(bh, expr, why...)	J_ASSERT_BH(bh, expr)
302 #define J_EXPECT_JH(jh, expr, why...)	J_ASSERT_JH(jh, expr)
303 #else
304 #if 0
305 #define __journal_expect(expr, why...)					     \
306 	({								     \
307 		int val = (expr);					     \
308 		if (!val) {						     \
309 			printk(KERN_ERR					     \
310 				"EXT3-fs unexpected failure: %s;\n",# expr); \
311 			printk(KERN_ERR why "\n");			     \
312 		}							     \
313 		val;							     \
314 	})
315 #define J_EXPECT(expr, why...)		__journal_expect(expr, ## why)
316 #define J_EXPECT_BH(bh, expr, why...)	__journal_expect(expr, ## why)
317 #define J_EXPECT_JH(jh, expr, why...)	__journal_expect(expr, ## why)
318 #endif
319 #define __journal_expect
320 
321 #define J_EXPECT
322 #define J_EXPECT_BH
323 #define J_EXPECT_JH
324 
325 #endif
326 
327 enum jbd_state_bits {
328     BH_JBD			/* Has an attached ext3 journal_head */
329     = BH_PrivateStart,
330     BH_JWrite,		/* Being written to log (@@@ DEBUGGING) */
331     BH_Freed,		/* Has been freed (truncated) */
332     BH_Revoked,		/* Has been revoked from the log */
333     BH_RevokeValid,		/* Revoked flag is valid */
334     BH_JBDDirty,		/* Is dirty but journaled */
335     BH_State,		/* Pins most journal_head state */
336     BH_JournalHead,		/* Pins bh->b_private and jh->b_bh */
337     BH_Unshadow,		/* Dummy bit, for BJ_Shadow wakeup filtering */
338 };
339 
BUFFER_FNS(JBD,jbd)340 BUFFER_FNS(JBD, jbd)
341 BUFFER_FNS(JWrite, jwrite)
342 BUFFER_FNS(JBDDirty, jbddirty)
343 TAS_BUFFER_FNS(JBDDirty, jbddirty)
344 BUFFER_FNS(Revoked, revoked)
345 TAS_BUFFER_FNS(Revoked, revoked)
346 BUFFER_FNS(RevokeValid, revokevalid)
347 TAS_BUFFER_FNS(RevokeValid, revokevalid)
348 BUFFER_FNS(Freed, freed)
349 
350 static inline struct buffer_head *jh2bh(struct journal_head *jh)
351 {
352     return jh->b_bh;
353 }
354 
bh2jh(struct buffer_head * bh)355 static inline struct journal_head *bh2jh(struct buffer_head *bh)
356 {
357     return bh->b_private;
358 }
359 
jbd_lock_bh_state(struct buffer_head * bh)360 static inline void jbd_lock_bh_state(struct buffer_head *bh)
361 {
362     bit_spin_lock(BH_State, &bh->b_state);
363 }
364 
jbd_trylock_bh_state(struct buffer_head * bh)365 static inline int jbd_trylock_bh_state(struct buffer_head *bh)
366 {
367     return bit_spin_trylock(BH_State, &bh->b_state);
368 }
369 
jbd_is_locked_bh_state(struct buffer_head * bh)370 static inline int jbd_is_locked_bh_state(struct buffer_head *bh)
371 {
372     return bit_spin_is_locked(BH_State, &bh->b_state);
373 }
374 
jbd_unlock_bh_state(struct buffer_head * bh)375 static inline void jbd_unlock_bh_state(struct buffer_head *bh)
376 {
377     bit_spin_unlock(BH_State, &bh->b_state);
378 }
379 
jbd_lock_bh_journal_head(struct buffer_head * bh)380 static inline void jbd_lock_bh_journal_head(struct buffer_head *bh)
381 {
382     bit_spin_lock(BH_JournalHead, &bh->b_state);
383 }
384 
jbd_unlock_bh_journal_head(struct buffer_head * bh)385 static inline void jbd_unlock_bh_journal_head(struct buffer_head *bh)
386 {
387     bit_spin_unlock(BH_JournalHead, &bh->b_state);
388 }
389 
390 struct jbd_revoke_table_s;
391 
392 /**
393  * struct handle_s - The handle_s type is the concrete type associated with
394  *     handle_t.
395  * @h_transaction: Which compound transaction is this update a part of?
396  * @h_buffer_credits: Number of remaining buffers we are allowed to dirty.
397  * @h_ref: Reference count on this handle
398  * @h_err: Field for caller's use to track errors through large fs operations
399  * @h_sync: flag for sync-on-close
400  * @h_jdata: flag to force data journaling
401  * @h_aborted: flag indicating fatal error on handle
402  * @h_lockdep_map: lockdep info for debugging lock problems
403  **/
404 
405 /* Docbook can't yet cope with the bit fields, but will leave the documentation
406  * in so it can be fixed later.
407  */
408 
409 struct handle_s
410 {
411     /* Which compound transaction is this update a part of? */
412     transaction_t		*h_transaction;
413 
414     /* Number of remaining buffers we are allowed to dirty: */
415     int			h_buffer_credits;
416 
417     /* Reference count on this handle */
418     int			h_ref;
419 
420     /* Field for caller's use to track errors through large fs */
421     /* operations */
422     int			h_err;
423 
424     /* Flags [no locking] */
425 unsigned int	h_sync:
426     1;	/* sync-on-close */
427 unsigned int	h_jdata:
428     1;	/* force data journaling */
429 unsigned int	h_aborted:
430     1;	/* fatal error on handle */
431 
432 #ifdef CONFIG_DEBUG_LOCK_ALLOC
433     struct lockdep_map	h_lockdep_map;
434 #endif
435 };
436 
437 
438 /* The transaction_t type is the guts of the journaling mechanism.  It
439  * tracks a compound transaction through its various states:
440  *
441  * RUNNING:	accepting new updates
442  * LOCKED:	Updates still running but we don't accept new ones
443  * RUNDOWN:	Updates are tidying up but have finished requesting
444  *		new buffers to modify (state not used for now)
445  * FLUSH:       All updates complete, but we are still writing to disk
446  * COMMIT:      All data on disk, writing commit record
447  * FINISHED:	We still have to keep the transaction for checkpointing.
448  *
449  * The transaction keeps track of all of the buffers modified by a
450  * running transaction, and all of the buffers committed but not yet
451  * flushed to home for finished transactions.
452  */
453 
454 /*
455  * Lock ranking:
456  *
457  *    j_list_lock
458  *      ->jbd_lock_bh_journal_head()	(This is "innermost")
459  *
460  *    j_state_lock
461  *    ->jbd_lock_bh_state()
462  *
463  *    jbd_lock_bh_state()
464  *    ->j_list_lock
465  *
466  *    j_state_lock
467  *    ->t_handle_lock
468  *
469  *    j_state_lock
470  *    ->j_list_lock			(journal_unmap_buffer)
471  *
472  */
473 
474 struct transaction_s
475 {
476     /* Pointer to the journal for this transaction. [no locking] */
477     journal_t		*t_journal;
478 
479     /* Sequence number for this transaction [no locking] */
480     tid_t			t_tid;
481 
482     /*
483      * Transaction's current state
484      * [no locking - only kjournald alters this]
485      * [j_list_lock] guards transition of a transaction into T_FINISHED
486      * state and subsequent call of __journal_drop_transaction()
487      * FIXME: needs barriers
488      * KLUDGE: [use j_state_lock]
489      */
490     enum {
491         T_RUNNING,
492         T_LOCKED,
493         T_RUNDOWN,
494         T_FLUSH,
495         T_COMMIT,
496         T_FINISHED
497     }			t_state;
498 
499     /*
500      * Where in the log does this transaction's commit start? [no locking]
501      */
502     unsigned long		t_log_start;
503 
504     /* Number of buffers on the t_buffers list [j_list_lock] */
505     int			t_nr_buffers;
506 
507     /*
508      * Doubly-linked circular list of all buffers reserved but not yet
509      * modified by this transaction [j_list_lock]
510      */
511     struct journal_head	*t_reserved_list;
512 
513     /*
514      * Doubly-linked circular list of all buffers under writeout during
515      * commit [j_list_lock]
516      */
517     struct journal_head	*t_locked_list;
518 
519     /*
520      * Doubly-linked circular list of all metadata buffers owned by this
521      * transaction [j_list_lock]
522      */
523     struct journal_head	*t_buffers;
524 
525     /*
526      * Doubly-linked circular list of all data buffers still to be
527      * flushed before this transaction can be committed [j_list_lock]
528      */
529     struct journal_head	*t_sync_datalist;
530 
531     /*
532      * Doubly-linked circular list of all forget buffers (superseded
533      * buffers which we can un-checkpoint once this transaction commits)
534      * [j_list_lock]
535      */
536     struct journal_head	*t_forget;
537 
538     /*
539      * Doubly-linked circular list of all buffers still to be flushed before
540      * this transaction can be checkpointed. [j_list_lock]
541      */
542     struct journal_head	*t_checkpoint_list;
543 
544     /*
545      * Doubly-linked circular list of all buffers submitted for IO while
546      * checkpointing. [j_list_lock]
547      */
548     struct journal_head	*t_checkpoint_io_list;
549 
550     /*
551      * Doubly-linked circular list of temporary buffers currently undergoing
552      * IO in the log [j_list_lock]
553      */
554     struct journal_head	*t_iobuf_list;
555 
556     /*
557      * Doubly-linked circular list of metadata buffers being shadowed by log
558      * IO.  The IO buffers on the iobuf list and the shadow buffers on this
559      * list match each other one for one at all times. [j_list_lock]
560      */
561     struct journal_head	*t_shadow_list;
562 
563     /*
564      * Doubly-linked circular list of control buffers being written to the
565      * log. [j_list_lock]
566      */
567     struct journal_head	*t_log_list;
568 
569     /*
570      * Protects info related to handles
571      */
572     jbdlock_t		t_handle_lock;
573 
574     /*
575      * Number of outstanding updates running on this transaction
576      * [t_handle_lock]
577      */
578     int			t_updates;
579 
580     /*
581      * Number of buffers reserved for use by all handles in this transaction
582      * handle but not yet modified. [t_handle_lock]
583      */
584     int			t_outstanding_credits;
585 
586     /*
587      * Forward and backward links for the circular list of all transactions
588      * awaiting checkpoint. [j_list_lock]
589      */
590     transaction_t		*t_cpnext, *t_cpprev;
591 
592     /*
593      * When will the transaction expire (become due for commit), in jiffies?
594      * [no locking]
595      */
596     unsigned long		t_expires;
597 
598     /*
599      * How many handles used this transaction? [t_handle_lock]
600      */
601     int t_handle_count;
602 
603 };
604 
605 /**
606  * struct journal_s - The journal_s type is the concrete type associated with
607  *     journal_t.
608  * @j_flags:  General journaling state flags
609  * @j_errno:  Is there an outstanding uncleared error on the journal (from a
610  *     prior abort)?
611  * @j_sb_buffer: First part of superblock buffer
612  * @j_superblock: Second part of superblock buffer
613  * @j_format_version: Version of the superblock format
614  * @j_state_lock: Protect the various scalars in the journal
615  * @j_barrier_count:  Number of processes waiting to create a barrier lock
616  * @j_barrier: The barrier lock itself
617  * @j_running_transaction: The current running transaction..
618  * @j_committing_transaction: the transaction we are pushing to disk
619  * @j_checkpoint_transactions: a linked circular list of all transactions
620  *  waiting for checkpointing
621  * @j_wait_transaction_locked: Wait queue for waiting for a locked transaction
622  *  to start committing, or for a barrier lock to be released
623  * @j_wait_logspace: Wait queue for waiting for checkpointing to complete
624  * @j_wait_done_commit: Wait queue for waiting for commit to complete
625  * @j_wait_checkpoint:  Wait queue to trigger checkpointing
626  * @j_wait_commit: Wait queue to trigger commit
627  * @j_wait_updates: Wait queue to wait for updates to complete
628  * @j_checkpoint_mutex: Mutex for locking against concurrent checkpoints
629  * @j_head: Journal head - identifies the first unused block in the journal
630  * @j_tail: Journal tail - identifies the oldest still-used block in the
631  *  journal.
632  * @j_free: Journal free - how many free blocks are there in the journal?
633  * @j_first: The block number of the first usable block
634  * @j_last: The block number one beyond the last usable block
635  * @j_dev: Device where we store the journal
636  * @j_blocksize: blocksize for the location where we store the journal.
637  * @j_blk_offset: starting block offset for into the device where we store the
638  *     journal
639  * @j_fs_dev: Device which holds the client fs.  For internal journal this will
640  *     be equal to j_dev
641  * @j_maxlen: Total maximum capacity of the journal region on disk.
642  * @j_list_lock: Protects the buffer lists and internal buffer state.
643  * @j_inode: Optional inode where we store the journal.  If present, all journal
644  *     block numbers are mapped into this inode via bmap().
645  * @j_tail_sequence:  Sequence number of the oldest transaction in the log
646  * @j_transaction_sequence: Sequence number of the next transaction to grant
647  * @j_commit_sequence: Sequence number of the most recently committed
648  *  transaction
649  * @j_commit_request: Sequence number of the most recent transaction wanting
650  *     commit
651  * @j_uuid: Uuid of client object.
652  * @j_task: Pointer to the current commit thread for this journal
653  * @j_max_transaction_buffers:  Maximum number of metadata buffers to allow in a
654  *     single compound commit transaction
655  * @j_commit_interval: What is the maximum transaction lifetime before we begin
656  *  a commit?
657  * @j_commit_timer:  The timer used to wakeup the commit thread
658  * @j_revoke_lock: Protect the revoke table
659  * @j_revoke: The revoke table - maintains the list of revoked blocks in the
660  *     current transaction.
661  * @j_revoke_table: alternate revoke tables for j_revoke
662  * @j_wbuf: array of buffer_heads for journal_commit_transaction
663  * @j_wbufsize: maximum number of buffer_heads allowed in j_wbuf, the
664  *	number that will fit in j_blocksize
665  * @j_last_sync_writer: most recent pid which did a synchronous write
666  * @j_private: An opaque pointer to fs-private information.
667  */
668 
669 struct journal_s
670 {
671     /* General journaling state flags [j_state_lock] */
672     unsigned long		j_flags;
673 
674     /*
675      * Is there an outstanding uncleared error on the journal (from a prior
676      * abort)? [j_state_lock]
677      */
678     int			j_errno;
679 
680     /* The superblock buffer */
681     struct buffer_head	*j_sb_buffer;
682     journal_superblock_t	*j_superblock;
683 
684     /* Version of the superblock format */
685     int			j_format_version;
686 
687     /*
688      * Protect the various scalars in the journal
689      */
690     jbdlock_t		j_state_lock;
691 
692     /*
693      * Number of processes waiting to create a barrier lock [j_state_lock]
694      */
695     int			j_barrier_count;
696 
697     /* The barrier lock itself */
698     struct mutex		j_barrier;
699 
700     /*
701      * Transactions: The current running transaction...
702      * [j_state_lock] [caller holding open handle]
703      */
704     transaction_t		*j_running_transaction;
705 
706     /*
707      * the transaction we are pushing to disk
708      * [j_state_lock] [caller holding open handle]
709      */
710     transaction_t		*j_committing_transaction;
711 
712     /*
713      * ... and a linked circular list of all transactions waiting for
714      * checkpointing. [j_list_lock]
715      */
716     transaction_t		*j_checkpoint_transactions;
717 
718     /*
719      * Wait queue for waiting for a locked transaction to start committing,
720      * or for a barrier lock to be released
721      */
722     wait_queue_head_t	j_wait_transaction_locked;
723 
724     /* Wait queue for waiting for checkpointing to complete */
725     wait_queue_head_t	j_wait_logspace;
726 
727     /* Wait queue for waiting for commit to complete */
728     wait_queue_head_t	j_wait_done_commit;
729 
730     /* Wait queue to trigger checkpointing */
731     wait_queue_head_t	j_wait_checkpoint;
732 
733     /* Wait queue to trigger commit */
734     wait_queue_head_t	j_wait_commit;
735 
736     /* Wait queue to wait for updates to complete */
737     wait_queue_head_t	j_wait_updates;
738 
739     /* Semaphore for locking against concurrent checkpoints */
740     struct mutex		j_checkpoint_mutex;
741 
742     /*
743      * Journal head: identifies the first unused block in the journal.
744      * [j_state_lock]
745      */
746     unsigned long		j_head;
747 
748     /*
749      * Journal tail: identifies the oldest still-used block in the journal.
750      * [j_state_lock]
751      */
752     unsigned long		j_tail;
753 
754     /*
755      * Journal free: how many free blocks are there in the journal?
756      * [j_state_lock]
757      */
758     unsigned long		j_free;
759 
760     /*
761      * Journal start and end: the block numbers of the first usable block
762      * and one beyond the last usable block in the journal. [j_state_lock]
763      */
764     unsigned long		j_first;
765     unsigned long		j_last;
766 
767     /*
768      * Device, blocksize and starting block offset for the location where we
769      * store the journal.
770      */
771     struct block_device	*j_dev;
772     int			j_blocksize;
773     unsigned long		j_blk_offset;
774 
775     /*
776      * Device which holds the client fs.  For internal journal this will be
777      * equal to j_dev.
778      */
779     struct block_device	*j_fs_dev;
780 
781     /* Total maximum capacity of the journal region on disk. */
782     unsigned int		j_maxlen;
783 
784     /*
785      * Protects the buffer lists and internal buffer state.
786      */
787     jbdlock_t		j_list_lock;
788 
789     /* Optional inode where we store the journal.  If present, all */
790     /* journal block numbers are mapped into this inode via */
791     /* bmap(). */
792     struct inode		*j_inode;
793 
794     /*
795      * Sequence number of the oldest transaction in the log [j_state_lock]
796      */
797     tid_t			j_tail_sequence;
798 
799     /*
800      * Sequence number of the next transaction to grant [j_state_lock]
801      */
802     tid_t			j_transaction_sequence;
803 
804     /*
805      * Sequence number of the most recently committed transaction
806      * [j_state_lock].
807      */
808     tid_t			j_commit_sequence;
809 
810     /*
811      * Sequence number of the most recent transaction wanting commit
812      * [j_state_lock]
813      */
814     tid_t			j_commit_request;
815 
816     /*
817      * Journal uuid: identifies the object (filesystem, LVM volume etc)
818      * backed by this journal.  This will eventually be replaced by an array
819      * of uuids, allowing us to index multiple devices within a single
820      * journal and to perform atomic updates across them.
821      */
822     __u8			j_uuid[16];
823 
824     /* Pointer to the current commit thread for this journal */
825     struct task_struct	*j_task;
826 
827     /*
828      * Maximum number of metadata buffers to allow in a single compound
829      * commit transaction
830      */
831     int			j_max_transaction_buffers;
832 
833     /*
834      * What is the maximum transaction lifetime before we begin a commit?
835      */
836     unsigned long		j_commit_interval;
837 
838     /* The timer used to wakeup the commit thread: */
839     struct timer_list	j_commit_timer;
840 
841     /*
842      * The revoke table: maintains the list of revoked blocks in the
843      * current transaction.  [j_revoke_lock]
844      */
845     jbdlock_t		j_revoke_lock;
846     struct jbd_revoke_table_s *j_revoke;
847     struct jbd_revoke_table_s *j_revoke_table[2];
848 
849     /*
850      * array of bhs for journal_commit_transaction
851      */
852     struct buffer_head	**j_wbuf;
853     int			j_wbufsize;
854 
855     pid_t			j_last_sync_writer;
856 
857     /*
858      * An opaque pointer to fs-private information.  ext3 puts its
859      * superblock pointer here
860      */
861     void *j_private;
862 };
863 
864 /*
865  * Journal flag definitions
866  */
867 #define JFS_UNMOUNT	0x001	/* Journal thread is being destroyed */
868 #define JFS_ABORT	0x002	/* Journaling has been aborted for errors. */
869 #define JFS_ACK_ERR	0x004	/* The errno in the sb has been acked */
870 #define JFS_FLUSHED	0x008	/* The journal superblock has been flushed */
871 #define JFS_LOADED	0x010	/* The journal superblock has been loaded */
872 #define JFS_BARRIER	0x020	/* Use IDE barriers */
873 
874 /*
875  * Function declarations for the journaling transaction and buffer
876  * management
877  */
878 
879 /* Filing buffers */
880 extern void journal_unfile_buffer(journal_t *, struct journal_head *);
881 extern void __journal_unfile_buffer(struct journal_head *);
882 extern void __journal_refile_buffer(struct journal_head *);
883 extern void journal_refile_buffer(journal_t *, struct journal_head *);
884 extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
885 extern void __journal_free_buffer(struct journal_head *bh);
886 extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
887 extern void __journal_clean_data_list(transaction_t *transaction);
888 
889 /* Log buffer allocation */
890 extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
891 int journal_next_log_block(journal_t *, unsigned long *);
892 
893 /* Commit management */
894 extern void journal_commit_transaction(journal_t *);
895 
896 /* Checkpoint list management */
897 int __journal_clean_checkpoint_list(journal_t *journal);
898 int __journal_remove_checkpoint(struct journal_head *);
899 void __journal_insert_checkpoint(struct journal_head *, transaction_t *);
900 
901 /* Buffer IO */
902 extern int
903     journal_write_metadata_buffer(transaction_t	  *transaction,
904                                   struct journal_head  *jh_in,
905                                   struct journal_head **jh_out,
906                                   unsigned long	   blocknr);
907 
908 /* Transaction locking */
909 extern void		__wait_on_journal (journal_t *);
910 
911 /*
912  * Journal locking.
913  *
914  * We need to lock the journal during transaction state changes so that nobody
915  * ever tries to take a handle on the running transaction while we are in the
916  * middle of moving it to the commit phase.  j_state_lock does this.
917  *
918  * Note that the locking is completely interrupt unsafe.  We never touch
919  * journal structures from interrupts.
920  */
921 
journal_current_handle(void)922 static inline handle_t *journal_current_handle(void)
923 {
924     return NULL;
925     /* return current->journal_info; */
926 #ifdef _MSC_VER
927 #pragma message("WARNINGS: journal_current_handle NOT implemented")
928 #endif
929 }
930 
931 /* The journaling code user interface:
932  *
933  * Create and destroy handles
934  * Register buffer modifications against the current transaction.
935  */
936 
937 extern handle_t *journal_start(journal_t *, int nblocks);
938 extern int	 journal_restart (handle_t *, int nblocks);
939 extern int	 journal_extend (handle_t *, int nblocks);
940 extern int	 journal_get_write_access(handle_t *, struct buffer_head *);
941 extern int	 journal_get_create_access (handle_t *, struct buffer_head *);
942 extern int	 journal_get_undo_access(handle_t *, struct buffer_head *);
943 extern int	 journal_dirty_data (handle_t *, struct buffer_head *);
944 extern int	 journal_dirty_metadata (handle_t *, struct buffer_head *);
945 extern void	 journal_release_buffer (handle_t *, struct buffer_head *);
946 extern int	 journal_forget (handle_t *, struct buffer_head *);
947 extern void	 journal_sync_buffer (struct buffer_head *);
948 extern void	 journal_invalidatepage(journal_t *,
949                                         struct page *, unsigned long);
950 extern int	 journal_try_to_free_buffers(journal_t *, struct page *, gfp_t);
951 extern int	 journal_stop(handle_t *);
952 extern int	 journal_flush (journal_t *);
953 extern void	 journal_lock_updates (journal_t *);
954 extern void	 journal_unlock_updates (journal_t *);
955 
956 extern journal_t * journal_init_dev(struct block_device *bdev,
957                                         struct block_device *fs_dev,
958                                         int start, int len, int bsize);
959 extern journal_t * journal_init_inode (struct inode *);
960 extern int	   journal_update_format (journal_t *);
961 extern int	   journal_check_used_features
962     (journal_t *, unsigned long, unsigned long, unsigned long);
963 extern int	   journal_check_available_features
964     (journal_t *, unsigned long, unsigned long, unsigned long);
965 extern int	   journal_set_features
966     (journal_t *, unsigned long, unsigned long, unsigned long);
967 extern int	   journal_create     (journal_t *);
968 extern int	   journal_load       (journal_t *journal);
969 extern void	   journal_destroy    (journal_t *);
970 extern int	   journal_recover    (journal_t *journal);
971 extern int	   journal_wipe       (journal_t *, int);
972 extern int	   journal_skip_recovery	(journal_t *);
973 extern void	   journal_update_superblock	(journal_t *, int);
974 void	   journal_abort      (journal_t *, int);
975 extern int	   journal_errno      (journal_t *);
976 extern void	   journal_ack_err    (journal_t *);
977 extern int	   journal_clear_err  (journal_t *);
978 extern int	   journal_bmap(journal_t *, unsigned long, unsigned long *);
979 extern int	   journal_force_commit(journal_t *);
980 
981 /* added by matt */
982 void journal_wipe_recovery(journal_t *journal);
983 
984 /*
985  * journal_head management
986  */
987 struct journal_head *journal_add_journal_head(struct buffer_head *bh);
988 struct journal_head *journal_grab_journal_head(struct buffer_head *bh);
989 void journal_remove_journal_head(struct buffer_head *bh);
990 void journal_put_journal_head(struct journal_head *jh);
991 
992 /*
993  * handle management
994  */
995 extern struct kmem_cache *jbd_handle_cache;
996 
jbd_alloc_handle(gfp_t gfp_flags)997 static inline handle_t *jbd_alloc_handle(gfp_t gfp_flags)
998 {
999     return (handle_t *)kmem_cache_alloc(jbd_handle_cache, gfp_flags);
1000 }
1001 
jbd_free_handle(handle_t * handle)1002 static inline void jbd_free_handle(handle_t *handle)
1003 {
1004     kmem_cache_free(jbd_handle_cache, handle);
1005 }
1006 
1007 /* Primary revoke support */
1008 #define JOURNAL_REVOKE_DEFAULT_HASH 256
1009 extern int	   journal_init_revoke(journal_t *, int);
1010 extern void	   journal_destroy_revoke_caches(void);
1011 extern int	   journal_init_revoke_caches(void);
1012 
1013 extern void	   journal_destroy_revoke(journal_t *);
1014 extern int	   journal_revoke (handle_t *,
1015                                   unsigned long, struct buffer_head *);
1016 extern int	   journal_cancel_revoke(handle_t *, struct journal_head *);
1017 extern void	   journal_write_revoke_records(journal_t *, transaction_t *);
1018 
1019 /* Recovery revoke support */
1020 extern int	journal_set_revoke(journal_t *, unsigned long, tid_t);
1021 extern int	journal_test_revoke(journal_t *, unsigned long, tid_t);
1022 extern void	journal_clear_revoke(journal_t *);
1023 extern void	journal_switch_revoke_table(journal_t *journal);
1024 
1025 /*
1026  * The log thread user interface:
1027  *
1028  * Request space in the current transaction, and force transaction commit
1029  * transitions on demand.
1030  */
1031 
1032 int __log_space_left(journal_t *); /* Called with journal locked */
1033 int log_start_commit(journal_t *journal, tid_t tid);
1034 int __log_start_commit(journal_t *journal, tid_t tid);
1035 int journal_start_commit(journal_t *journal, tid_t *tid);
1036 int journal_force_commit_nested(journal_t *journal);
1037 int log_wait_commit(journal_t *journal, tid_t tid);
1038 int log_do_checkpoint(journal_t *journal);
1039 
1040 void __log_wait_for_space(journal_t *journal);
1041 extern void	__journal_drop_transaction(journal_t *, transaction_t *);
1042 extern int	cleanup_journal_tail(journal_t *);
1043 
1044 /* Debugging code only: */
1045 
1046 #define jbd_ENOSYS() \
1047 do {								           \
1048 	printk (KERN_ERR "JBD unimplemented function %s\n", __FUNCTION__); \
1049 	current->state = TASK_UNINTERRUPTIBLE;			           \
1050 	schedule();						           \
1051 } while (1)
1052 
1053 /*
1054  * is_journal_abort
1055  *
1056  * Simple test wrapper function to test the JFS_ABORT state flag.  This
1057  * bit, when set, indicates that we have had a fatal error somewhere,
1058  * either inside the journaling layer or indicated to us by the client
1059  * (eg. ext3), and that we and should not commit any further
1060  * transactions.
1061  */
1062 
is_journal_aborted(journal_t * journal)1063 static inline int is_journal_aborted(journal_t *journal)
1064 {
1065     return journal->j_flags & JFS_ABORT;
1066 }
1067 
is_handle_aborted(handle_t * handle)1068 static inline int is_handle_aborted(handle_t *handle)
1069 {
1070     if (handle->h_aborted)
1071         return 1;
1072     return is_journal_aborted(handle->h_transaction->t_journal);
1073 }
1074 
journal_abort_handle(handle_t * handle)1075 static inline void journal_abort_handle(handle_t *handle)
1076 {
1077     handle->h_aborted = 1;
1078 }
1079 
1080 #endif /* __KERNEL__   */
1081 
1082 /* Comparison functions for transaction IDs: perform comparisons using
1083  * modulo arithmetic so that they work over sequence number wraps. */
1084 
tid_gt(tid_t x,tid_t y)1085 static inline int tid_gt(tid_t x, tid_t y)
1086 {
1087     int difference = (x - y);
1088     return (difference > 0);
1089 }
1090 
tid_geq(tid_t x,tid_t y)1091 static inline int tid_geq(tid_t x, tid_t y)
1092 {
1093     int difference = (x - y);
1094     return (difference >= 0);
1095 }
1096 
1097 extern int journal_blocks_per_page(struct inode *inode);
1098 
1099 /*
1100  * Return the minimum number of blocks which must be free in the journal
1101  * before a new transaction may be started.  Must be called under j_state_lock.
1102  */
jbd_space_needed(journal_t * journal)1103 static inline int jbd_space_needed(journal_t *journal)
1104 {
1105     int nblocks = journal->j_max_transaction_buffers;
1106     if (journal->j_committing_transaction)
1107         nblocks += journal->j_committing_transaction->
1108                    t_outstanding_credits;
1109     return nblocks;
1110 }
1111 
1112 /*
1113  * Definitions which augment the buffer_head layer
1114  */
1115 
1116 /* journaling buffer types */
1117 #define BJ_None		0	/* Not journaled */
1118 #define BJ_SyncData	1	/* Normal data: flush before commit */
1119 #define BJ_Metadata	2	/* Normal journaled metadata */
1120 #define BJ_Forget	3	/* Buffer superseded by this transaction */
1121 #define BJ_IO		4	/* Buffer is for temporary IO use */
1122 #define BJ_Shadow	5	/* Buffer contents being shadowed to the log */
1123 #define BJ_LogCtl	6	/* Buffer contains log descriptors */
1124 #define BJ_Reserved	7	/* Buffer is reserved for access by journal */
1125 #define BJ_Locked	8	/* Locked for I/O during commit */
1126 #define BJ_Types	9
1127 
1128 extern int jbd_blocks_per_page(struct inode *inode);
1129 
1130 #ifdef __KERNEL__
1131 
1132 #define buffer_trace_init(bh)	do {} while (0)
1133 #define print_buffer_fields(bh)	do {} while (0)
1134 #define print_buffer_trace(bh)	do {} while (0)
1135 #define BUFFER_TRACE(bh, info)	do {} while (0)
1136 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
1137 #define JBUFFER_TRACE(jh, info)	do {} while (0)
1138 
1139 #endif	/* __KERNEL__ */
1140 
1141 #endif	/* _LINUX_JBD_H */
1142