4 * Written by Stephen C. Tweedie <sct@redhat.com>, 1999
6 * Copyright 1998--1999 Red Hat corp --- All Rights Reserved
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.
12 * Ext4-specific journaling extensions.
19 #include <linux/jbd2.h>
22 #define EXT4_JOURNAL(inode) (EXT4_SB((inode)->i_sb)->s_journal)
24 /* Define the number of blocks we need to account to a transaction to
25 * modify one block of data.
27 * We may have to touch one inode, one bitmap buffer, up to three
28 * indirection blocks, the group and superblock summaries, and the data
29 * block to complete the transaction.
31 * For extents-enabled fs we may have to allocate and modify up to
32 * 5 levels of tree + root which are stored in the inode. */
34 #define EXT4_SINGLEDATA_TRANS_BLOCKS(sb) \
35 (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS) \
38 /* Extended attribute operations touch at most two data buffers,
39 * two bitmap buffers, and two group summaries, in addition to the inode
40 * and the superblock, which are already accounted for. */
42 #define EXT4_XATTR_TRANS_BLOCKS 6U
44 /* Define the minimum size for a transaction which modifies data. This
45 * needs to take into account the fact that we may end up modifying two
46 * quota files too (one for the group, one for the user quota). The
47 * superblock only gets updated once, of course, so don't bother
48 * counting that again for the quota updates. */
50 #define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
51 EXT4_XATTR_TRANS_BLOCKS - 2 + \
52 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
55 * Define the number of metadata blocks we need to account to modify data.
57 * This include super block, inode block, quota blocks and xattr blocks
59 #define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \
60 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
62 /* Delete operations potentially hit one directory's namespace plus an
63 * entire inode, plus arbitrary amounts of bitmap/indirection data. Be
64 * generous. We can grow the delete transaction later if necessary. */
66 #define EXT4_DELETE_TRANS_BLOCKS(sb) (2 * EXT4_DATA_TRANS_BLOCKS(sb) + 64)
68 /* Define an arbitrary limit for the amount of data we will anticipate
69 * writing to any given transaction. For unbounded transactions such as
70 * write(2) and truncate(2) we can write more than this, but we always
71 * start off at the maximum transaction size and grow the transaction
72 * optimistically as we go. */
74 #define EXT4_MAX_TRANS_DATA 64U
76 /* We break up a large truncate or write transaction once the handle's
77 * buffer credits gets this low, we need either to extend the
78 * transaction or to start a new one. Reserve enough space here for
79 * inode, bitmap, superblock, group and indirection updates for at least
80 * one block, plus two quota updates. Quota allocations are not
83 #define EXT4_RESERVE_TRANS_BLOCKS 12U
85 #define EXT4_INDEX_EXTRA_TRANS_BLOCKS 8
88 /* Amount of blocks needed for quota update - we know that the structure was
89 * allocated so we need to update only data block */
90 #define EXT4_QUOTA_TRANS_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
91 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
93 /* Amount of blocks needed for quota insert/delete - we do some block writes
94 * but inode, sb and group updates are done only once */
95 #define EXT4_QUOTA_INIT_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
96 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
97 (DQUOT_INIT_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
98 +3+DQUOT_INIT_REWRITE) : 0)
100 #define EXT4_QUOTA_DEL_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
101 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
102 (DQUOT_DEL_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
103 +3+DQUOT_DEL_REWRITE) : 0)
105 #define EXT4_QUOTA_TRANS_BLOCKS(sb) 0
106 #define EXT4_QUOTA_INIT_BLOCKS(sb) 0
107 #define EXT4_QUOTA_DEL_BLOCKS(sb) 0
109 #define EXT4_MAXQUOTAS_TRANS_BLOCKS(sb) (MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
110 #define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
111 #define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (MAXQUOTAS*EXT4_QUOTA_DEL_BLOCKS(sb))
114 * struct ext4_journal_cb_entry - Base structure for callback information.
116 * This struct is a 'seed' structure for a using with your own callback
117 * structs. If you are using callbacks you must allocate one of these
118 * or another struct of your own definition which has this struct
119 * as it's first element and pass it to ext4_journal_callback_add().
121 struct ext4_journal_cb_entry
{
122 /* list information for other callbacks attached to the same handle */
123 struct list_head jce_list
;
125 /* Function to call with this callback structure */
126 void (*jce_func
)(struct super_block
*sb
,
127 struct ext4_journal_cb_entry
*jce
, int error
);
129 /* user data goes here */
133 * ext4_journal_callback_add: add a function to call after transaction commit
134 * @handle: active journal transaction handle to register callback on
135 * @func: callback function to call after the transaction has committed:
136 * @sb: superblock of current filesystem for transaction
137 * @jce: returned journal callback data
138 * @rc: journal state at commit (0 = transaction committed properly)
139 * @jce: journal callback data (internal and function private data struct)
141 * The registered function will be called in the context of the journal thread
142 * after the transaction for which the handle was created has completed.
144 * No locks are held when the callback function is called, so it is safe to
145 * call blocking functions from within the callback, but the callback should
146 * not block or run for too long, or the filesystem will be blocked waiting for
147 * the next transaction to commit. No journaling functions can be used, or
148 * there is a risk of deadlock.
150 * There is no guaranteed calling order of multiple registered callbacks on
151 * the same transaction.
153 static inline void ext4_journal_callback_add(handle_t
*handle
,
154 void (*func
)(struct super_block
*sb
,
155 struct ext4_journal_cb_entry
*jce
,
157 struct ext4_journal_cb_entry
*jce
)
159 struct ext4_sb_info
*sbi
=
160 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
162 /* Add the jce to transaction's private list */
163 jce
->jce_func
= func
;
164 spin_lock(&sbi
->s_md_lock
);
165 list_add_tail(&jce
->jce_list
, &handle
->h_transaction
->t_private_list
);
166 spin_unlock(&sbi
->s_md_lock
);
170 * ext4_journal_callback_del: delete a registered callback
171 * @handle: active journal transaction handle on which callback was registered
172 * @jce: registered journal callback entry to unregister
174 static inline void ext4_journal_callback_del(handle_t
*handle
,
175 struct ext4_journal_cb_entry
*jce
)
177 struct ext4_sb_info
*sbi
=
178 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
180 spin_lock(&sbi
->s_md_lock
);
181 list_del_init(&jce
->jce_list
);
182 spin_unlock(&sbi
->s_md_lock
);
186 ext4_mark_iloc_dirty(handle_t
*handle
,
188 struct ext4_iloc
*iloc
);
191 * On success, We end up with an outstanding reference count against
192 * iloc->bh. This _must_ be cleaned up later.
195 int ext4_reserve_inode_write(handle_t
*handle
, struct inode
*inode
,
196 struct ext4_iloc
*iloc
);
198 int ext4_mark_inode_dirty(handle_t
*handle
, struct inode
*inode
);
201 * Wrapper functions with which ext4 calls into JBD.
203 void ext4_journal_abort_handle(const char *caller
, unsigned int line
,
205 struct buffer_head
*bh
, handle_t
*handle
, int err
);
207 int __ext4_journal_get_write_access(const char *where
, unsigned int line
,
208 handle_t
*handle
, struct buffer_head
*bh
);
210 int __ext4_forget(const char *where
, unsigned int line
, handle_t
*handle
,
211 int is_metadata
, struct inode
*inode
,
212 struct buffer_head
*bh
, ext4_fsblk_t blocknr
);
214 int __ext4_journal_get_create_access(const char *where
, unsigned int line
,
215 handle_t
*handle
, struct buffer_head
*bh
);
217 int __ext4_handle_dirty_metadata(const char *where
, unsigned int line
,
218 handle_t
*handle
, struct inode
*inode
,
219 struct buffer_head
*bh
);
221 int __ext4_handle_dirty_super(const char *where
, unsigned int line
,
222 handle_t
*handle
, struct super_block
*sb
);
224 #define ext4_journal_get_write_access(handle, bh) \
225 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
226 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
227 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
229 #define ext4_journal_get_create_access(handle, bh) \
230 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
231 #define ext4_handle_dirty_metadata(handle, inode, bh) \
232 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
234 #define ext4_handle_dirty_super(handle, sb) \
235 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
237 handle_t
*ext4_journal_start_sb(struct super_block
*sb
, int nblocks
);
238 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
);
240 #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
242 /* Note: Do not use this for NULL handles. This is only to determine if
243 * a properly allocated handle is using a journal or not. */
244 static inline int ext4_handle_valid(handle_t
*handle
)
246 if ((unsigned long)handle
< EXT4_NOJOURNAL_MAX_REF_COUNT
)
251 static inline void ext4_handle_sync(handle_t
*handle
)
253 if (ext4_handle_valid(handle
))
257 static inline void ext4_handle_release_buffer(handle_t
*handle
,
258 struct buffer_head
*bh
)
260 if (ext4_handle_valid(handle
))
261 jbd2_journal_release_buffer(handle
, bh
);
264 static inline int ext4_handle_is_aborted(handle_t
*handle
)
266 if (ext4_handle_valid(handle
))
267 return is_handle_aborted(handle
);
271 static inline int ext4_handle_has_enough_credits(handle_t
*handle
, int needed
)
273 if (ext4_handle_valid(handle
) && handle
->h_buffer_credits
< needed
)
278 static inline handle_t
*ext4_journal_start(struct inode
*inode
, int nblocks
)
280 return ext4_journal_start_sb(inode
->i_sb
, nblocks
);
283 #define ext4_journal_stop(handle) \
284 __ext4_journal_stop(__func__, __LINE__, (handle))
286 static inline handle_t
*ext4_journal_current_handle(void)
288 return journal_current_handle();
291 static inline int ext4_journal_extend(handle_t
*handle
, int nblocks
)
293 if (ext4_handle_valid(handle
))
294 return jbd2_journal_extend(handle
, nblocks
);
298 static inline int ext4_journal_restart(handle_t
*handle
, int nblocks
)
300 if (ext4_handle_valid(handle
))
301 return jbd2_journal_restart(handle
, nblocks
);
305 static inline int ext4_journal_blocks_per_page(struct inode
*inode
)
307 if (EXT4_JOURNAL(inode
) != NULL
)
308 return jbd2_journal_blocks_per_page(inode
);
312 static inline int ext4_journal_force_commit(journal_t
*journal
)
315 return jbd2_journal_force_commit(journal
);
319 static inline int ext4_jbd2_file_inode(handle_t
*handle
, struct inode
*inode
)
321 if (ext4_handle_valid(handle
))
322 return jbd2_journal_file_inode(handle
, EXT4_I(inode
)->jinode
);
326 static inline void ext4_update_inode_fsync_trans(handle_t
*handle
,
330 struct ext4_inode_info
*ei
= EXT4_I(inode
);
332 if (ext4_handle_valid(handle
)) {
333 ei
->i_sync_tid
= handle
->h_transaction
->t_tid
;
335 ei
->i_datasync_tid
= handle
->h_transaction
->t_tid
;
340 int ext4_force_commit(struct super_block
*sb
);
343 * Ext4 inode journal modes
345 #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
346 #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
347 #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
349 static inline int ext4_inode_journal_mode(struct inode
*inode
)
351 if (EXT4_JOURNAL(inode
) == NULL
)
352 return EXT4_INODE_WRITEBACK_DATA_MODE
; /* writeback */
353 /* We do not support data journalling with delayed allocation */
354 if (!S_ISREG(inode
->i_mode
) ||
355 test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
356 return EXT4_INODE_JOURNAL_DATA_MODE
; /* journal data */
357 if (ext4_test_inode_flag(inode
, EXT4_INODE_JOURNAL_DATA
) &&
358 !test_opt(inode
->i_sb
, DELALLOC
))
359 return EXT4_INODE_JOURNAL_DATA_MODE
; /* journal data */
360 if (test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
361 return EXT4_INODE_ORDERED_DATA_MODE
; /* ordered */
362 if (test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
363 return EXT4_INODE_WRITEBACK_DATA_MODE
; /* writeback */
368 static inline int ext4_should_journal_data(struct inode
*inode
)
370 return ext4_inode_journal_mode(inode
) & EXT4_INODE_JOURNAL_DATA_MODE
;
373 static inline int ext4_should_order_data(struct inode
*inode
)
375 return ext4_inode_journal_mode(inode
) & EXT4_INODE_ORDERED_DATA_MODE
;
378 static inline int ext4_should_writeback_data(struct inode
*inode
)
380 return ext4_inode_journal_mode(inode
) & EXT4_INODE_WRITEBACK_DATA_MODE
;
384 * This function controls whether or not we should try to go down the
385 * dioread_nolock code paths, which makes it safe to avoid taking
386 * i_mutex for direct I/O reads. This only works for extent-based
387 * files, and it doesn't work if data journaling is enabled, since the
388 * dioread_nolock code uses b_private to pass information back to the
389 * I/O completion handler, and this conflicts with the jbd's use of
392 static inline int ext4_should_dioread_nolock(struct inode
*inode
)
394 if (!test_opt(inode
->i_sb
, DIOREAD_NOLOCK
))
396 if (!S_ISREG(inode
->i_mode
))
398 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
400 if (ext4_should_journal_data(inode
))
405 #endif /* _EXT4_JBD2_H */