Home | History | Annotate | Download | only in ext2fs
      1 /*
      2  * linux/include/linux/jbd.h
      3  *
      4  * Written by Stephen C. Tweedie <sct (at) 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 #if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE) || !defined(__KERNEL__)
     20 
     21 /* Allow this file to be included directly into e2fsprogs */
     22 #ifndef __KERNEL__
     23 #include "jfs_compat.h"
     24 #define JFS_DEBUG
     25 #define jfs_debug jbd_debug
     26 #else
     27 
     28 #include <linux/journal-head.h>
     29 #include <linux/stddef.h>
     30 #include <asm/semaphore.h>
     31 #endif
     32 
     33 #ifndef __GNUC__
     34 #define __FUNCTION__ ""
     35 #endif
     36 
     37 #define journal_oom_retry 1
     38 
     39 #ifdef __STDC__
     40 #ifdef CONFIG_JBD_DEBUG
     41 /*
     42  * Define JBD_EXPENSIVE_CHECKING to enable more expensive internal
     43  * consistency checks.  By default we don't do this unless
     44  * CONFIG_JBD_DEBUG is on.
     45  */
     46 #define JBD_EXPENSIVE_CHECKING
     47 extern int journal_enable_debug;
     48 
     49 #define jbd_debug(n, f, a...)						\
     50 	do {								\
     51 		if ((n) <= journal_enable_debug) {			\
     52 			printk (KERN_DEBUG "(%s, %d): %s: ",		\
     53 				__FILE__, __LINE__, __FUNCTION__);	\
     54 		  	printk (f, ## a);				\
     55 		}							\
     56 	} while (0)
     57 #else
     58 #ifdef __GNUC__
     59 #define jbd_debug(f, a...)	/**/
     60 #else
     61 #define jbd_debug(f, ...)	/**/
     62 #endif
     63 #endif
     64 #else
     65 #define jbd_debug(x)		/* AIX doesn't do STDC */
     66 #endif
     67 
     68 extern void * __jbd_kmalloc (char *where, size_t size, int flags, int retry);
     69 #define jbd_kmalloc(size, flags) \
     70 	__jbd_kmalloc(__FUNCTION__, (size), (flags), journal_oom_retry)
     71 #define jbd_rep_kmalloc(size, flags) \
     72 	__jbd_kmalloc(__FUNCTION__, (size), (flags), 1)
     73 
     74 #define JFS_MIN_JOURNAL_BLOCKS 1024
     75 
     76 #ifdef __KERNEL__
     77 typedef struct handle_s		handle_t;	/* Atomic operation type */
     78 typedef struct journal_s	journal_t;	/* Journal control structure */
     79 #endif
     80 
     81 /*
     82  * Internal structures used by the logging mechanism:
     83  */
     84 
     85 #define JFS_MAGIC_NUMBER 0xc03b3998U /* The first 4 bytes of /dev/random! */
     86 
     87 /*
     88  * On-disk structures
     89  */
     90 
     91 /*
     92  * Descriptor block types:
     93  */
     94 
     95 #define JFS_DESCRIPTOR_BLOCK	1
     96 #define JFS_COMMIT_BLOCK	2
     97 #define JFS_SUPERBLOCK_V1	3
     98 #define JFS_SUPERBLOCK_V2	4
     99 #define JFS_REVOKE_BLOCK	5
    100 
    101 /*
    102  * Standard header for all descriptor blocks:
    103  */
    104 typedef struct journal_header_s
    105 {
    106 	__u32		h_magic;
    107 	__u32		h_blocktype;
    108 	__u32		h_sequence;
    109 } journal_header_t;
    110 
    111 /*
    112  * Checksum types.
    113  */
    114 #define JBD2_CRC32_CHKSUM   1
    115 #define JBD2_MD5_CHKSUM     2
    116 #define JBD2_SHA1_CHKSUM    3
    117 
    118 #define JBD2_CRC32_CHKSUM_SIZE 4
    119 
    120 #define JBD2_CHECKSUM_BYTES (32 / sizeof(__u32))
    121 /*
    122  * Commit block header for storing transactional checksums:
    123  */
    124 struct commit_header {
    125 	__u32		h_magic;
    126 	__u32		h_blocktype;
    127 	__u32		h_sequence;
    128 	unsigned char	h_chksum_type;
    129 	unsigned char	h_chksum_size;
    130 	unsigned char	h_padding[2];
    131 	__u32		h_chksum[JBD2_CHECKSUM_BYTES];
    132 	__u64		h_commit_sec;
    133 	__u32		h_commit_nsec;
    134 };
    135 
    136 /*
    137  * The block tag: used to describe a single buffer in the journal
    138  */
    139 typedef struct journal_block_tag_s
    140 {
    141 	__u32		t_blocknr;	/* The on-disk block number */
    142 	__u32		t_flags;	/* See below */
    143 	__u32		t_blocknr_high; /* most-significant high 32bits. */
    144 } journal_block_tag_t;
    145 
    146 #define JBD_TAG_SIZE64 (sizeof(journal_block_tag_t))
    147 #define JBD_TAG_SIZE32 (8)
    148 
    149 /*
    150  * The revoke descriptor: used on disk to describe a series of blocks to
    151  * be revoked from the log
    152  */
    153 typedef struct journal_revoke_header_s
    154 {
    155 	journal_header_t r_header;
    156 	int		 r_count;	/* Count of bytes used in the block */
    157 } journal_revoke_header_t;
    158 
    159 
    160 /* Definitions for the journal tag flags word: */
    161 #define JFS_FLAG_ESCAPE		1	/* on-disk block is escaped */
    162 #define JFS_FLAG_SAME_UUID	2	/* block has same uuid as previous */
    163 #define JFS_FLAG_DELETED	4	/* block deleted by this transaction */
    164 #define JFS_FLAG_LAST_TAG	8	/* last tag in this descriptor block */
    165 
    166 
    167 /*
    168  * The journal superblock.  All fields are in big-endian byte order.
    169  */
    170 typedef struct journal_superblock_s
    171 {
    172 /* 0x0000 */
    173 	journal_header_t s_header;
    174 
    175 /* 0x000C */
    176 	/* Static information describing the journal */
    177 	__u32	s_blocksize;		/* journal device blocksize */
    178 	__u32	s_maxlen;		/* total blocks in journal file */
    179 	__u32	s_first;		/* first block of log information */
    180 
    181 /* 0x0018 */
    182 	/* Dynamic information describing the current state of the log */
    183 	__u32	s_sequence;		/* first commit ID expected in log */
    184 	__u32	s_start;		/* blocknr of start of log */
    185 
    186 /* 0x0020 */
    187 	/* Error value, as set by journal_abort(). */
    188 	__s32	s_errno;
    189 
    190 /* 0x0024 */
    191 	/* Remaining fields are only valid in a version-2 superblock */
    192 	__u32	s_feature_compat; 	/* compatible feature set */
    193 	__u32	s_feature_incompat; 	/* incompatible feature set */
    194 	__u32	s_feature_ro_compat; 	/* readonly-compatible feature set */
    195 /* 0x0030 */
    196 	__u8	s_uuid[16];		/* 128-bit uuid for journal */
    197 
    198 /* 0x0040 */
    199 	__u32	s_nr_users;		/* Nr of filesystems sharing log */
    200 
    201 	__u32	s_dynsuper;		/* Blocknr of dynamic superblock copy*/
    202 
    203 /* 0x0048 */
    204 	__u32	s_max_transaction;	/* Limit of journal blocks per trans.*/
    205 	__u32	s_max_trans_data;	/* Limit of data blocks per trans. */
    206 
    207 /* 0x0050 */
    208 	__u32	s_padding[44];
    209 
    210 /* 0x0100 */
    211 	__u8	s_users[16*48];		/* ids of all fs'es sharing the log */
    212 /* 0x0400 */
    213 } journal_superblock_t;
    214 
    215 #define JFS_HAS_COMPAT_FEATURE(j,mask)					\
    216 	((j)->j_format_version >= 2 &&					\
    217 	 ((j)->j_superblock->s_feature_compat & ext2fs_cpu_to_be32((mask))))
    218 #define JFS_HAS_RO_COMPAT_FEATURE(j,mask)				\
    219 	((j)->j_format_version >= 2 &&					\
    220 	 ((j)->j_superblock->s_feature_ro_compat & ext2fs_cpu_to_be32((mask))))
    221 #define JFS_HAS_INCOMPAT_FEATURE(j,mask)				\
    222 	((j)->j_format_version >= 2 &&					\
    223 	 ((j)->j_superblock->s_feature_incompat & ext2fs_cpu_to_be32((mask))))
    224 
    225 #define JFS_FEATURE_COMPAT_CHECKSUM	0x00000001
    226 
    227 #define JFS_FEATURE_INCOMPAT_REVOKE	0x00000001
    228 
    229 #define JFS_FEATURE_INCOMPAT_REVOKE		0x00000001
    230 #define JFS_FEATURE_INCOMPAT_64BIT		0x00000002
    231 #define JFS_FEATURE_INCOMPAT_ASYNC_COMMIT	0x00000004
    232 
    233 /* Features known to this kernel version: */
    234 #define JFS_KNOWN_COMPAT_FEATURES	0
    235 #define JFS_KNOWN_ROCOMPAT_FEATURES	0
    236 #define JFS_KNOWN_INCOMPAT_FEATURES	(JFS_FEATURE_INCOMPAT_REVOKE|\
    237 					 JFS_FEATURE_INCOMPAT_ASYNC_COMMIT|\
    238 					 JFS_FEATURE_INCOMPAT_64BIT)
    239 
    240 #ifdef __KERNEL__
    241 
    242 #include <linux/fs.h>
    243 #include <linux/sched.h>
    244 
    245 #define JBD_ASSERTIONS
    246 #ifdef JBD_ASSERTIONS
    247 #define J_ASSERT(assert)						\
    248 do {									\
    249 	if (!(assert)) {						\
    250 		printk (KERN_EMERG					\
    251 			"Assertion failure in %s() at %s:%d: \"%s\"\n",	\
    252 			__FUNCTION__, __FILE__, __LINE__, # assert);	\
    253 		BUG();							\
    254 	}								\
    255 } while (0)
    256 
    257 #if defined(CONFIG_BUFFER_DEBUG)
    258 void buffer_assertion_failure(struct buffer_head *bh);
    259 #define J_ASSERT_BH(bh, expr)						\
    260 	do {								\
    261 		if (!(expr))						\
    262 			buffer_assertion_failure(bh);			\
    263 		J_ASSERT(expr);						\
    264 	} while (0)
    265 #define J_ASSERT_JH(jh, expr)	J_ASSERT_BH(jh2bh(jh), expr)
    266 #else
    267 #define J_ASSERT_BH(bh, expr)	J_ASSERT(expr)
    268 #define J_ASSERT_JH(jh, expr)	J_ASSERT(expr)
    269 #endif
    270 
    271 #else
    272 #define J_ASSERT(assert)
    273 #endif		/* JBD_ASSERTIONS */
    274 
    275 enum jbd_state_bits {
    276 	BH_JWrite
    277 	  = BH_PrivateStart,	/* 1 if being written to log (@@@ DEBUGGING) */
    278 	BH_Freed,		/* 1 if buffer has been freed (truncated) */
    279 	BH_Revoked,		/* 1 if buffer has been revoked from the log */
    280 	BH_RevokeValid,		/* 1 if buffer revoked flag is valid */
    281 	BH_JBDDirty,		/* 1 if buffer is dirty but journaled */
    282 };
    283 
    284 /* Return true if the buffer is one which JBD is managing */
    285 static inline int buffer_jbd(struct buffer_head *bh)
    286 {
    287 	return __buffer_state(bh, JBD);
    288 }
    289 
    290 static inline struct buffer_head *jh2bh(struct journal_head *jh)
    291 {
    292 	return jh->b_bh;
    293 }
    294 
    295 static inline struct journal_head *bh2jh(struct buffer_head *bh)
    296 {
    297 	return bh->b_private;
    298 }
    299 
    300 struct jbd_revoke_table_s;
    301 
    302 /* The handle_t type represents a single atomic update being performed
    303  * by some process.  All filesystem modifications made by the process go
    304  * through this handle.  Recursive operations (such as quota operations)
    305  * are gathered into a single update.
    306  *
    307  * The buffer credits field is used to account for journaled buffers
    308  * being modified by the running process.  To ensure that there is
    309  * enough log space for all outstanding operations, we need to limit the
    310  * number of outstanding buffers possible at any time.  When the
    311  * operation completes, any buffer credits not used are credited back to
    312  * the transaction, so that at all times we know how many buffers the
    313  * outstanding updates on a transaction might possibly touch. */
    314 
    315 struct handle_s
    316 {
    317 	/* Which compound transaction is this update a part of? */
    318 	transaction_t	      * h_transaction;
    319 
    320 	/* Number of remaining buffers we are allowed to dirty: */
    321 	int			h_buffer_credits;
    322 
    323 	/* Reference count on this handle */
    324 	int			h_ref;
    325 
    326 	/* Field for caller's use to track errors through large fs
    327 	   operations */
    328 	int			h_err;
    329 
    330 	/* Flags */
    331 	unsigned int	h_sync:		1;	/* sync-on-close */
    332 	unsigned int	h_jdata:	1;	/* force data journaling */
    333 	unsigned int	h_aborted:	1;	/* fatal error on handle */
    334 };
    335 
    336 
    337 /* The transaction_t type is the guts of the journaling mechanism.  It
    338  * tracks a compound transaction through its various states:
    339  *
    340  * RUNNING:	accepting new updates
    341  * LOCKED:	Updates still running but we don't accept new ones
    342  * RUNDOWN:	Updates are tidying up but have finished requesting
    343  *		new buffers to modify (state not used for now)
    344  * FLUSH:       All updates complete, but we are still writing to disk
    345  * COMMIT:      All data on disk, writing commit record
    346  * FINISHED:	We still have to keep the transaction for checkpointing.
    347  *
    348  * The transaction keeps track of all of the buffers modified by a
    349  * running transaction, and all of the buffers committed but not yet
    350  * flushed to home for finished transactions.
    351  */
    352 
    353 struct transaction_s
    354 {
    355 	/* Pointer to the journal for this transaction. */
    356 	journal_t *		t_journal;
    357 
    358 	/* Sequence number for this transaction */
    359 	tid_t			t_tid;
    360 
    361 	/* Transaction's current state */
    362 	enum {
    363 		T_RUNNING,
    364 		T_LOCKED,
    365 		T_RUNDOWN,
    366 		T_FLUSH,
    367 		T_COMMIT,
    368 		T_FINISHED
    369 	}			t_state;
    370 
    371 	/* Where in the log does this transaction's commit start? */
    372 	unsigned long		t_log_start;
    373 
    374 	/* Doubly-linked circular list of all inodes owned by this
    375            transaction */	/* AKPM: unused */
    376 	struct inode *		t_ilist;
    377 
    378 	/* Number of buffers on the t_buffers list */
    379 	int			t_nr_buffers;
    380 
    381 	/* Doubly-linked circular list of all buffers reserved but not
    382            yet modified by this transaction */
    383 	struct journal_head *	t_reserved_list;
    384 
    385 	/* Doubly-linked circular list of all metadata buffers owned by this
    386            transaction */
    387 	struct journal_head *	t_buffers;
    388 
    389 	/*
    390 	 * Doubly-linked circular list of all data buffers still to be
    391 	 * flushed before this transaction can be committed.
    392 	 * Protected by journal_datalist_lock.
    393 	 */
    394 	struct journal_head *	t_sync_datalist;
    395 
    396 	/*
    397 	 * Doubly-linked circular list of all writepage data buffers
    398 	 * still to be written before this transaction can be committed.
    399 	 * Protected by journal_datalist_lock.
    400 	 */
    401 	struct journal_head *	t_async_datalist;
    402 
    403 	/* Doubly-linked circular list of all forget buffers (superceded
    404            buffers which we can un-checkpoint once this transaction
    405            commits) */
    406 	struct journal_head *	t_forget;
    407 
    408 	/*
    409 	 * Doubly-linked circular list of all buffers still to be
    410 	 * flushed before this transaction can be checkpointed.
    411 	 */
    412 	/* Protected by journal_datalist_lock */
    413 	struct journal_head *	t_checkpoint_list;
    414 
    415 	/* Doubly-linked circular list of temporary buffers currently
    416            undergoing IO in the log */
    417 	struct journal_head *	t_iobuf_list;
    418 
    419 	/* Doubly-linked circular list of metadata buffers being
    420            shadowed by log IO.  The IO buffers on the iobuf list and the
    421            shadow buffers on this list match each other one for one at
    422            all times. */
    423 	struct journal_head *	t_shadow_list;
    424 
    425 	/* Doubly-linked circular list of control buffers being written
    426            to the log. */
    427 	struct journal_head *	t_log_list;
    428 
    429 	/* Number of outstanding updates running on this transaction */
    430 	int			t_updates;
    431 
    432 	/* Number of buffers reserved for use by all handles in this
    433 	 * transaction handle but not yet modified. */
    434 	int			t_outstanding_credits;
    435 
    436 	/*
    437 	 * Forward and backward links for the circular list of all
    438 	 * transactions awaiting checkpoint.
    439 	 */
    440 	/* Protected by journal_datalist_lock */
    441 	transaction_t		*t_cpnext, *t_cpprev;
    442 
    443 	/* When will the transaction expire (become due for commit), in
    444 	 * jiffies ? */
    445 	unsigned long		t_expires;
    446 
    447 	/* How many handles used this transaction? */
    448 	int t_handle_count;
    449 };
    450 
    451 
    452 /* The journal_t maintains all of the journaling state information for a
    453  * single filesystem.  It is linked to from the fs superblock structure.
    454  *
    455  * We use the journal_t to keep track of all outstanding transaction
    456  * activity on the filesystem, and to manage the state of the log
    457  * writing process. */
    458 
    459 struct journal_s
    460 {
    461 	/* General journaling state flags */
    462 	unsigned long		j_flags;
    463 
    464 	/* Is there an outstanding uncleared error on the journal (from
    465 	 * a prior abort)? */
    466 	int			j_errno;
    467 
    468 	/* The superblock buffer */
    469 	struct buffer_head *	j_sb_buffer;
    470 	journal_superblock_t *	j_superblock;
    471 
    472 	/* Version of the superblock format */
    473 	int			j_format_version;
    474 
    475 	/* Number of processes waiting to create a barrier lock */
    476 	int			j_barrier_count;
    477 
    478 	/* The barrier lock itself */
    479 	struct semaphore	j_barrier;
    480 
    481 	/* Transactions: The current running transaction... */
    482 	transaction_t *		j_running_transaction;
    483 
    484 	/* ... the transaction we are pushing to disk ... */
    485 	transaction_t *		j_committing_transaction;
    486 
    487 	/* ... and a linked circular list of all transactions waiting
    488 	 * for checkpointing. */
    489 	/* Protected by journal_datalist_lock */
    490 	transaction_t *		j_checkpoint_transactions;
    491 
    492 	/* Wait queue for waiting for a locked transaction to start
    493            committing, or for a barrier lock to be released */
    494 	wait_queue_head_t	j_wait_transaction_locked;
    495 
    496 	/* Wait queue for waiting for checkpointing to complete */
    497 	wait_queue_head_t	j_wait_logspace;
    498 
    499 	/* Wait queue for waiting for commit to complete */
    500 	wait_queue_head_t	j_wait_done_commit;
    501 
    502 	/* Wait queue to trigger checkpointing */
    503 	wait_queue_head_t	j_wait_checkpoint;
    504 
    505 	/* Wait queue to trigger commit */
    506 	wait_queue_head_t	j_wait_commit;
    507 
    508 	/* Wait queue to wait for updates to complete */
    509 	wait_queue_head_t	j_wait_updates;
    510 
    511 	/* Semaphore for locking against concurrent checkpoints */
    512 	struct semaphore 	j_checkpoint_sem;
    513 
    514 	/* The main journal lock, used by lock_journal() */
    515 	struct semaphore	j_sem;
    516 
    517 	/* Journal head: identifies the first unused block in the journal. */
    518 	unsigned long		j_head;
    519 
    520 	/* Journal tail: identifies the oldest still-used block in the
    521 	 * journal. */
    522 	unsigned long		j_tail;
    523 
    524 	/* Journal free: how many free blocks are there in the journal? */
    525 	unsigned long		j_free;
    526 
    527 	/* Journal start and end: the block numbers of the first usable
    528 	 * block and one beyond the last usable block in the journal. */
    529 	unsigned long		j_first, j_last;
    530 
    531 	/* Device, blocksize and starting block offset for the location
    532 	 * where we store the journal. */
    533 	kdev_t			j_dev;
    534 	int			j_blocksize;
    535 	unsigned int		j_blk_offset;
    536 
    537 	/* Device which holds the client fs.  For internal journal this
    538 	 * will be equal to j_dev. */
    539 	kdev_t			j_fs_dev;
    540 
    541 	/* Total maximum capacity of the journal region on disk. */
    542 	unsigned int		j_maxlen;
    543 
    544 	/* Optional inode where we store the journal.  If present, all
    545 	 * journal block numbers are mapped into this inode via
    546 	 * bmap(). */
    547 	struct inode *		j_inode;
    548 
    549 	/* Sequence number of the oldest transaction in the log */
    550 	tid_t			j_tail_sequence;
    551 	/* Sequence number of the next transaction to grant */
    552 	tid_t			j_transaction_sequence;
    553 	/* Sequence number of the most recently committed transaction */
    554 	tid_t			j_commit_sequence;
    555 	/* Sequence number of the most recent transaction wanting commit */
    556 	tid_t			j_commit_request;
    557 
    558 	/* Journal uuid: identifies the object (filesystem, LVM volume
    559 	 * etc) backed by this journal.  This will eventually be
    560 	 * replaced by an array of uuids, allowing us to index multiple
    561 	 * devices within a single journal and to perform atomic updates
    562 	 * across them.  */
    563 
    564 	__u8			j_uuid[16];
    565 
    566 	/* Pointer to the current commit thread for this journal */
    567 	struct task_struct *	j_task;
    568 
    569 	/* Maximum number of metadata buffers to allow in a single
    570 	 * compound commit transaction */
    571 	int			j_max_transaction_buffers;
    572 
    573 	/* What is the maximum transaction lifetime before we begin a
    574 	 * commit? */
    575 	unsigned long		j_commit_interval;
    576 
    577 	/* The timer used to wakeup the commit thread: */
    578 	struct timer_list *	j_commit_timer;
    579 	int			j_commit_timer_active;
    580 
    581 	/* Link all journals together - system-wide */
    582 	struct list_head	j_all_journals;
    583 
    584 	/* The revoke table: maintains the list of revoked blocks in the
    585            current transaction. */
    586 	struct jbd_revoke_table_s *j_revoke;
    587 
    588 	/* Failed journal commit ID */
    589 	unsigned int		j_failed_commit;
    590 };
    591 
    592 /*
    593  * Journal flag definitions
    594  */
    595 #define JFS_UNMOUNT	0x001	/* Journal thread is being destroyed */
    596 #define JFS_ABORT	0x002	/* Journaling has been aborted for errors. */
    597 #define JFS_ACK_ERR	0x004	/* The errno in the sb has been acked */
    598 #define JFS_FLUSHED	0x008	/* The journal superblock has been flushed */
    599 #define JFS_LOADED	0x010	/* The journal superblock has been loaded */
    600 
    601 /*
    602  * Function declarations for the journaling transaction and buffer
    603  * management
    604  */
    605 
    606 /* Filing buffers */
    607 extern void __journal_unfile_buffer(struct journal_head *);
    608 extern void journal_unfile_buffer(struct journal_head *);
    609 extern void __journal_refile_buffer(struct journal_head *);
    610 extern void journal_refile_buffer(struct journal_head *);
    611 extern void __journal_file_buffer(struct journal_head *, transaction_t *, int);
    612 extern void __journal_free_buffer(struct journal_head *bh);
    613 extern void journal_file_buffer(struct journal_head *, transaction_t *, int);
    614 extern void __journal_clean_data_list(transaction_t *transaction);
    615 
    616 /* Log buffer allocation */
    617 extern struct journal_head * journal_get_descriptor_buffer(journal_t *);
    618 extern unsigned long journal_next_log_block(journal_t *);
    619 
    620 /* Commit management */
    621 extern void journal_commit_transaction(journal_t *);
    622 
    623 /* Checkpoint list management */
    624 int __journal_clean_checkpoint_list(journal_t *journal);
    625 extern void journal_remove_checkpoint(struct journal_head *);
    626 extern void __journal_remove_checkpoint(struct journal_head *);
    627 extern void journal_insert_checkpoint(struct journal_head *, transaction_t *);
    628 extern void __journal_insert_checkpoint(struct journal_head *,transaction_t *);
    629 
    630 /* Buffer IO */
    631 extern int
    632 journal_write_metadata_buffer(transaction_t	  *transaction,
    633 			      struct journal_head  *jh_in,
    634 			      struct journal_head **jh_out,
    635 			      int		   blocknr);
    636 
    637 /* Transaction locking */
    638 extern void		__wait_on_journal (journal_t *);
    639 
    640 /*
    641  * Journal locking.
    642  *
    643  * We need to lock the journal during transaction state changes so that
    644  * nobody ever tries to take a handle on the running transaction while
    645  * we are in the middle of moving it to the commit phase.
    646  *
    647  * Note that the locking is completely interrupt unsafe.  We never touch
    648  * journal structures from interrupts.
    649  *
    650  * In 2.2, the BKL was required for lock_journal.  This is no longer
    651  * the case.
    652  */
    653 
    654 static inline void lock_journal(journal_t *journal)
    655 {
    656 	down(&journal->j_sem);
    657 }
    658 
    659 /* This returns zero if we acquired the semaphore */
    660 static inline int try_lock_journal(journal_t * journal)
    661 {
    662 	return down_trylock(&journal->j_sem);
    663 }
    664 
    665 static inline void unlock_journal(journal_t * journal)
    666 {
    667 	up(&journal->j_sem);
    668 }
    669 
    670 
    671 static inline handle_t *journal_current_handle(void)
    672 {
    673 	return current->journal_info;
    674 }
    675 
    676 /* The journaling code user interface:
    677  *
    678  * Create and destroy handles
    679  * Register buffer modifications against the current transaction.
    680  */
    681 
    682 extern handle_t *journal_start(journal_t *, int nblocks);
    683 extern handle_t *journal_try_start(journal_t *, int nblocks);
    684 extern int	 journal_restart (handle_t *, int nblocks);
    685 extern int	 journal_extend (handle_t *, int nblocks);
    686 extern int	 journal_get_write_access (handle_t *, struct buffer_head *);
    687 extern int	 journal_get_create_access (handle_t *, struct buffer_head *);
    688 extern int	 journal_get_undo_access (handle_t *, struct buffer_head *);
    689 extern int	 journal_dirty_data (handle_t *,
    690 				struct buffer_head *, int async);
    691 extern int	 journal_dirty_metadata (handle_t *, struct buffer_head *);
    692 extern void	 journal_release_buffer (handle_t *, struct buffer_head *);
    693 extern void	 journal_forget (handle_t *, struct buffer_head *);
    694 extern void	 journal_sync_buffer (struct buffer_head *);
    695 extern int	 journal_flushpage(journal_t *, struct page *, unsigned long);
    696 extern int	 journal_try_to_free_buffers(journal_t *, struct page *, int);
    697 extern int	 journal_stop(handle_t *);
    698 extern int	 journal_flush (journal_t *);
    699 
    700 extern void	 journal_lock_updates (journal_t *);
    701 extern void	 journal_unlock_updates (journal_t *);
    702 
    703 extern journal_t * journal_init_dev(kdev_t dev, kdev_t fs_dev,
    704 				int start, int len, int bsize);
    705 extern journal_t * journal_init_inode (struct inode *);
    706 extern int	   journal_update_format (journal_t *);
    707 extern int	   journal_check_used_features
    708 		   (journal_t *, unsigned long, unsigned long, unsigned long);
    709 extern int	   journal_check_available_features
    710 		   (journal_t *, unsigned long, unsigned long, unsigned long);
    711 extern int	   journal_set_features
    712 		   (journal_t *, unsigned long, unsigned long, unsigned long);
    713 extern int	   journal_create     (journal_t *);
    714 extern int	   journal_load       (journal_t *journal);
    715 extern void	   journal_destroy    (journal_t *);
    716 extern int	   journal_recover    (journal_t *journal);
    717 extern int	   journal_wipe       (journal_t *, int);
    718 extern int	   journal_skip_recovery (journal_t *);
    719 extern void	   journal_update_superblock (journal_t *, int);
    720 extern void	   __journal_abort      (journal_t *);
    721 extern void	   journal_abort      (journal_t *, int);
    722 extern int	   journal_errno      (journal_t *);
    723 extern void	   journal_ack_err    (journal_t *);
    724 extern int	   journal_clear_err  (journal_t *);
    725 extern unsigned long journal_bmap(journal_t *journal, unsigned long blocknr);
    726 extern int	    journal_force_commit(journal_t *journal);
    727 
    728 /*
    729  * journal_head management
    730  */
    731 extern struct journal_head
    732 		*journal_add_journal_head(struct buffer_head *bh);
    733 extern void	journal_remove_journal_head(struct buffer_head *bh);
    734 extern void	__journal_remove_journal_head(struct buffer_head *bh);
    735 extern void	journal_unlock_journal_head(struct journal_head *jh);
    736 
    737 /* Primary revoke support */
    738 #define JOURNAL_REVOKE_DEFAULT_HASH 256
    739 extern int	   journal_init_revoke(journal_t *, int);
    740 extern void	   journal_destroy_revoke_caches(void);
    741 extern int	   journal_init_revoke_caches(void);
    742 
    743 extern void	   journal_destroy_revoke(journal_t *);
    744 extern int	   journal_revoke (handle_t *,
    745 				unsigned long, struct buffer_head *);
    746 extern int	   journal_cancel_revoke(handle_t *, struct journal_head *);
    747 extern void	   journal_write_revoke_records(journal_t *, transaction_t *);
    748 
    749 /* Recovery revoke support */
    750 extern int	   journal_set_revoke(journal_t *, unsigned long, tid_t);
    751 extern int	   journal_test_revoke(journal_t *, unsigned long, tid_t);
    752 extern void	   journal_clear_revoke(journal_t *);
    753 extern void	   journal_brelse_array(struct buffer_head *b[], int n);
    754 
    755 /* The log thread user interface:
    756  *
    757  * Request space in the current transaction, and force transaction commit
    758  * transitions on demand.
    759  */
    760 
    761 extern int	log_space_left (journal_t *); /* Called with journal locked */
    762 extern tid_t	log_start_commit (journal_t *, transaction_t *);
    763 extern void	log_wait_commit (journal_t *, tid_t);
    764 extern int	log_do_checkpoint (journal_t *, int);
    765 
    766 extern void	log_wait_for_space(journal_t *, int nblocks);
    767 extern void	__journal_drop_transaction(journal_t *, transaction_t *);
    768 extern int	cleanup_journal_tail(journal_t *);
    769 
    770 /* Reduce journal memory usage by flushing */
    771 extern void shrink_journal_memory(void);
    772 
    773 /* Debugging code only: */
    774 
    775 #define jbd_ENOSYS() \
    776 do {								      \
    777 	printk (KERN_ERR "JBD unimplemented function " __FUNCTION__); \
    778 	current->state = TASK_UNINTERRUPTIBLE;			      \
    779 	schedule();						      \
    780 } while (1)
    781 
    782 /*
    783  * is_journal_abort
    784  *
    785  * Simple test wrapper function to test the JFS_ABORT state flag.  This
    786  * bit, when set, indicates that we have had a fatal error somewhere,
    787  * either inside the journaling layer or indicated to us by the client
    788  * (eg. ext3), and that we and should not commit any further
    789  * transactions.
    790  */
    791 
    792 static inline int is_journal_aborted(journal_t *journal)
    793 {
    794 	return journal->j_flags & JFS_ABORT;
    795 }
    796 
    797 static inline int is_handle_aborted(handle_t *handle)
    798 {
    799 	if (handle->h_aborted)
    800 		return 1;
    801 	return is_journal_aborted(handle->h_transaction->t_journal);
    802 }
    803 
    804 static inline void journal_abort_handle(handle_t *handle)
    805 {
    806 	handle->h_aborted = 1;
    807 }
    808 
    809 /* Not all architectures define BUG() */
    810 #ifndef BUG
    811 #define BUG() do { \
    812         printk("kernel BUG at %s:%d!\n", __FILE__, __LINE__); \
    813 	* ((char *) 0) = 0; \
    814  } while (0)
    815 #endif /* BUG */
    816 
    817 #else
    818 
    819 extern int	   journal_recover    (journal_t *journal);
    820 extern int	   journal_skip_recovery (journal_t *);
    821 
    822 /* Primary revoke support */
    823 extern int	   journal_init_revoke(journal_t *, int);
    824 extern void	   journal_destroy_revoke_caches(void);
    825 extern int	   journal_init_revoke_caches(void);
    826 
    827 /* Recovery revoke support */
    828 extern int	   journal_set_revoke(journal_t *, unsigned long, tid_t);
    829 extern int	   journal_test_revoke(journal_t *, unsigned long, tid_t);
    830 extern void	   journal_clear_revoke(journal_t *);
    831 extern void	   journal_brelse_array(struct buffer_head *b[], int n);
    832 
    833 extern void	   journal_destroy_revoke(journal_t *);
    834 #endif /* __KERNEL__   */
    835 
    836 static inline int tid_gt(tid_t x, tid_t y) EXT2FS_ATTR((unused));
    837 static inline int tid_geq(tid_t x, tid_t y) EXT2FS_ATTR((unused));
    838 
    839 /* Comparison functions for transaction IDs: perform comparisons using
    840  * modulo arithmetic so that they work over sequence number wraps. */
    841 
    842 static inline int tid_gt(tid_t x, tid_t y)
    843 {
    844 	int difference = (x - y);
    845 	return (difference > 0);
    846 }
    847 
    848 static inline int tid_geq(tid_t x, tid_t y)
    849 {
    850 	int difference = (x - y);
    851 	return (difference >= 0);
    852 }
    853 
    854 extern int journal_blocks_per_page(struct inode *inode);
    855 
    856 /*
    857  * Definitions which augment the buffer_head layer
    858  */
    859 
    860 /* journaling buffer types */
    861 #define BJ_None		0	/* Not journaled */
    862 #define BJ_SyncData	1	/* Normal data: flush before commit */
    863 #define BJ_AsyncData	2	/* writepage data: wait on it before commit */
    864 #define BJ_Metadata	3	/* Normal journaled metadata */
    865 #define BJ_Forget	4	/* Buffer superceded by this transaction */
    866 #define BJ_IO		5	/* Buffer is for temporary IO use */
    867 #define BJ_Shadow	6	/* Buffer contents being shadowed to the log */
    868 #define BJ_LogCtl	7	/* Buffer contains log descriptors */
    869 #define BJ_Reserved	8	/* Buffer is reserved for access by journal */
    870 #define BJ_Types	9
    871 
    872 extern int jbd_blocks_per_page(struct inode *inode);
    873 
    874 #ifdef __KERNEL__
    875 
    876 extern spinlock_t jh_splice_lock;
    877 /*
    878  * Once `expr1' has been found true, take jh_splice_lock
    879  * and then reevaluate everything.
    880  */
    881 #define SPLICE_LOCK(expr1, expr2)				\
    882 	({							\
    883 		int ret = (expr1);				\
    884 		if (ret) {					\
    885 			spin_lock(&jh_splice_lock);		\
    886 			ret = (expr1) && (expr2);		\
    887 			spin_unlock(&jh_splice_lock);		\
    888 		}						\
    889 		ret;						\
    890 	})
    891 
    892 /*
    893  * A number of buffer state predicates.  They test for
    894  * buffer_jbd() because they are used in core kernel code.
    895  *
    896  * These will be racy on SMP unless we're *sure* that the
    897  * buffer won't be detached from the journalling system
    898  * in parallel.
    899  */
    900 
    901 /* Return true if the buffer is on journal list `list' */
    902 static inline int buffer_jlist_eq(struct buffer_head *bh, int list)
    903 {
    904 	return SPLICE_LOCK(buffer_jbd(bh), bh2jh(bh)->b_jlist == list);
    905 }
    906 
    907 /* Return true if this bufer is dirty wrt the journal */
    908 static inline int buffer_jdirty(struct buffer_head *bh)
    909 {
    910 	return buffer_jbd(bh) && __buffer_state(bh, JBDDirty);
    911 }
    912 
    913 /* Return true if it's a data buffer which journalling is managing */
    914 static inline int buffer_jbd_data(struct buffer_head *bh)
    915 {
    916 	return SPLICE_LOCK(buffer_jbd(bh),
    917 			bh2jh(bh)->b_jlist == BJ_SyncData ||
    918 			bh2jh(bh)->b_jlist == BJ_AsyncData);
    919 }
    920 
    921 #ifdef CONFIG_SMP
    922 #define assert_spin_locked(lock)	J_ASSERT(spin_is_locked(lock))
    923 #else
    924 #define assert_spin_locked(lock)	do {} while(0)
    925 #endif
    926 
    927 #define buffer_trace_init(bh)	do {} while (0)
    928 #define print_buffer_fields(bh)	do {} while (0)
    929 #define print_buffer_trace(bh)	do {} while (0)
    930 #define BUFFER_TRACE(bh, info)	do {} while (0)
    931 #define BUFFER_TRACE2(bh, bh2, info)	do {} while (0)
    932 #define JBUFFER_TRACE(jh, info)	do {} while (0)
    933 
    934 #endif	/* __KERNEL__ */
    935 
    936 #endif	/* CONFIG_JBD || CONFIG_JBD_MODULE || !__KERNEL__ */
    937 
    938 /*
    939  * Compatibility no-ops which allow the kernel to compile without CONFIG_JBD
    940  * go here.
    941  */
    942 
    943 #if defined(__KERNEL__) && !(defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE))
    944 
    945 #define J_ASSERT(expr)			do {} while (0)
    946 #define J_ASSERT_BH(bh, expr)		do {} while (0)
    947 #define buffer_jbd(bh)			0
    948 #define buffer_jlist_eq(bh, val)	0
    949 #define journal_buffer_journal_lru(bh)	0
    950 
    951 #endif	/* defined(__KERNEL__) && !defined(CONFIG_JBD) */
    952 #endif	/* _LINUX_JBD_H */
    953