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, data block (for each of these we need bitmap + group
33 * summaries), root which is stored in the inode, sb
36 #define EXT4_SINGLEDATA_TRANS_BLOCKS(sb) \
37 (ext4_has_feature_extents(sb) ? 20U : 8U)
39 /* Extended attribute operations touch at most two data buffers,
40 * two bitmap buffers, and two group summaries, in addition to the inode
41 * and the superblock, which are already accounted for. */
43 #define EXT4_XATTR_TRANS_BLOCKS 6U
45 /* Define the minimum size for a transaction which modifies data. This
46 * needs to take into account the fact that we may end up modifying two
47 * quota files too (one for the group, one for the user quota). The
48 * superblock only gets updated once, of course, so don't bother
49 * counting that again for the quota updates. */
51 #define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
52 EXT4_XATTR_TRANS_BLOCKS - 2 + \
53 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
56 * Define the number of metadata blocks we need to account to modify data.
58 * This include super block, inode block, quota blocks and xattr blocks
60 #define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \
61 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
63 /* Define an arbitrary limit for the amount of data we will anticipate
64 * writing to any given transaction. For unbounded transactions such as
65 * write(2) and truncate(2) we can write more than this, but we always
66 * start off at the maximum transaction size and grow the transaction
67 * optimistically as we go. */
69 #define EXT4_MAX_TRANS_DATA 64U
71 /* We break up a large truncate or write transaction once the handle's
72 * buffer credits gets this low, we need either to extend the
73 * transaction or to start a new one. Reserve enough space here for
74 * inode, bitmap, superblock, group and indirection updates for at least
75 * one block, plus two quota updates. Quota allocations are not
78 #define EXT4_RESERVE_TRANS_BLOCKS 12U
81 * Number of credits needed if we need to insert an entry into a
82 * directory. For each new index block, we need 4 blocks (old index
83 * block, new index block, bitmap block, bg summary). For normal
84 * htree directories there are 2 levels; if the largedir feature
85 * enabled it's 3 levels.
87 #define EXT4_INDEX_EXTRA_TRANS_BLOCKS 12U
90 /* Amount of blocks needed for quota update - we know that the structure was
91 * allocated so we need to update only data block */
92 #define EXT4_QUOTA_TRANS_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
93 ext4_has_feature_quota(sb)) ? 1 : 0)
94 /* Amount of blocks needed for quota insert/delete - we do some block writes
95 * but inode, sb and group updates are done only once */
96 #define EXT4_QUOTA_INIT_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
97 ext4_has_feature_quota(sb)) ?\
98 (DQUOT_INIT_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
99 +3+DQUOT_INIT_REWRITE) : 0)
101 #define EXT4_QUOTA_DEL_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
102 ext4_has_feature_quota(sb)) ?\
103 (DQUOT_DEL_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
104 +3+DQUOT_DEL_REWRITE) : 0)
106 #define EXT4_QUOTA_TRANS_BLOCKS(sb) 0
107 #define EXT4_QUOTA_INIT_BLOCKS(sb) 0
108 #define EXT4_QUOTA_DEL_BLOCKS(sb) 0
110 #define EXT4_MAXQUOTAS_TRANS_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
111 #define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
112 #define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_DEL_BLOCKS(sb))
115 * Ext4 handle operation types -- for logging purposes
117 #define EXT4_HT_MISC 0
118 #define EXT4_HT_INODE 1
119 #define EXT4_HT_WRITE_PAGE 2
120 #define EXT4_HT_MAP_BLOCKS 3
121 #define EXT4_HT_DIR 4
122 #define EXT4_HT_TRUNCATE 5
123 #define EXT4_HT_QUOTA 6
124 #define EXT4_HT_RESIZE 7
125 #define EXT4_HT_MIGRATE 8
126 #define EXT4_HT_MOVE_EXTENTS 9
127 #define EXT4_HT_XATTR 10
128 #define EXT4_HT_EXT_CONVERT 11
129 #define EXT4_HT_MAX 12
132 * struct ext4_journal_cb_entry - Base structure for callback information.
134 * This struct is a 'seed' structure for a using with your own callback
135 * structs. If you are using callbacks you must allocate one of these
136 * or another struct of your own definition which has this struct
137 * as it's first element and pass it to ext4_journal_callback_add().
139 struct ext4_journal_cb_entry
{
140 /* list information for other callbacks attached to the same handle */
141 struct list_head jce_list
;
143 /* Function to call with this callback structure */
144 void (*jce_func
)(struct super_block
*sb
,
145 struct ext4_journal_cb_entry
*jce
, int error
);
147 /* user data goes here */
151 * ext4_journal_callback_add: add a function to call after transaction commit
152 * @handle: active journal transaction handle to register callback on
153 * @func: callback function to call after the transaction has committed:
154 * @sb: superblock of current filesystem for transaction
155 * @jce: returned journal callback data
156 * @rc: journal state at commit (0 = transaction committed properly)
157 * @jce: journal callback data (internal and function private data struct)
159 * The registered function will be called in the context of the journal thread
160 * after the transaction for which the handle was created has completed.
162 * No locks are held when the callback function is called, so it is safe to
163 * call blocking functions from within the callback, but the callback should
164 * not block or run for too long, or the filesystem will be blocked waiting for
165 * the next transaction to commit. No journaling functions can be used, or
166 * there is a risk of deadlock.
168 * There is no guaranteed calling order of multiple registered callbacks on
169 * the same transaction.
171 static inline void _ext4_journal_callback_add(handle_t
*handle
,
172 struct ext4_journal_cb_entry
*jce
)
174 /* Add the jce to transaction's private list */
175 list_add_tail(&jce
->jce_list
, &handle
->h_transaction
->t_private_list
);
178 static inline void ext4_journal_callback_add(handle_t
*handle
,
179 void (*func
)(struct super_block
*sb
,
180 struct ext4_journal_cb_entry
*jce
,
182 struct ext4_journal_cb_entry
*jce
)
184 struct ext4_sb_info
*sbi
=
185 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
187 /* Add the jce to transaction's private list */
188 jce
->jce_func
= func
;
189 spin_lock(&sbi
->s_md_lock
);
190 _ext4_journal_callback_add(handle
, jce
);
191 spin_unlock(&sbi
->s_md_lock
);
196 * ext4_journal_callback_del: delete a registered callback
197 * @handle: active journal transaction handle on which callback was registered
198 * @jce: registered journal callback entry to unregister
199 * Return true if object was successfully removed
201 static inline bool ext4_journal_callback_try_del(handle_t
*handle
,
202 struct ext4_journal_cb_entry
*jce
)
205 struct ext4_sb_info
*sbi
=
206 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
208 spin_lock(&sbi
->s_md_lock
);
209 deleted
= !list_empty(&jce
->jce_list
);
210 list_del_init(&jce
->jce_list
);
211 spin_unlock(&sbi
->s_md_lock
);
216 ext4_mark_iloc_dirty(handle_t
*handle
,
218 struct ext4_iloc
*iloc
);
221 * On success, We end up with an outstanding reference count against
222 * iloc->bh. This _must_ be cleaned up later.
225 int ext4_reserve_inode_write(handle_t
*handle
, struct inode
*inode
,
226 struct ext4_iloc
*iloc
);
228 int ext4_mark_inode_dirty(handle_t
*handle
, struct inode
*inode
);
230 int ext4_expand_extra_isize(struct inode
*inode
,
231 unsigned int new_extra_isize
,
232 struct ext4_iloc
*iloc
);
234 * Wrapper functions with which ext4 calls into JBD.
236 int __ext4_journal_get_write_access(const char *where
, unsigned int line
,
237 handle_t
*handle
, struct buffer_head
*bh
);
239 int __ext4_forget(const char *where
, unsigned int line
, handle_t
*handle
,
240 int is_metadata
, struct inode
*inode
,
241 struct buffer_head
*bh
, ext4_fsblk_t blocknr
);
243 int __ext4_journal_get_create_access(const char *where
, unsigned int line
,
244 handle_t
*handle
, struct buffer_head
*bh
);
246 int __ext4_handle_dirty_metadata(const char *where
, unsigned int line
,
247 handle_t
*handle
, struct inode
*inode
,
248 struct buffer_head
*bh
);
250 int __ext4_handle_dirty_super(const char *where
, unsigned int line
,
251 handle_t
*handle
, struct super_block
*sb
);
253 #define ext4_journal_get_write_access(handle, bh) \
254 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
255 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
256 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
258 #define ext4_journal_get_create_access(handle, bh) \
259 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
260 #define ext4_handle_dirty_metadata(handle, inode, bh) \
261 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
263 #define ext4_handle_dirty_super(handle, sb) \
264 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
266 handle_t
*__ext4_journal_start_sb(struct super_block
*sb
, unsigned int line
,
267 int type
, int blocks
, int rsv_blocks
);
268 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
);
270 #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
272 /* Note: Do not use this for NULL handles. This is only to determine if
273 * a properly allocated handle is using a journal or not. */
274 static inline int ext4_handle_valid(handle_t
*handle
)
276 if ((unsigned long)handle
< EXT4_NOJOURNAL_MAX_REF_COUNT
)
281 static inline void ext4_handle_sync(handle_t
*handle
)
283 if (ext4_handle_valid(handle
))
287 static inline int ext4_handle_is_aborted(handle_t
*handle
)
289 if (ext4_handle_valid(handle
))
290 return is_handle_aborted(handle
);
294 static inline int ext4_handle_has_enough_credits(handle_t
*handle
, int needed
)
296 if (ext4_handle_valid(handle
) && handle
->h_buffer_credits
< needed
)
301 #define ext4_journal_start_sb(sb, type, nblocks) \
302 __ext4_journal_start_sb((sb), __LINE__, (type), (nblocks), 0)
304 #define ext4_journal_start(inode, type, nblocks) \
305 __ext4_journal_start((inode), __LINE__, (type), (nblocks), 0)
307 #define ext4_journal_start_with_reserve(inode, type, blocks, rsv_blocks) \
308 __ext4_journal_start((inode), __LINE__, (type), (blocks), (rsv_blocks))
310 static inline handle_t
*__ext4_journal_start(struct inode
*inode
,
311 unsigned int line
, int type
,
312 int blocks
, int rsv_blocks
)
314 return __ext4_journal_start_sb(inode
->i_sb
, line
, type
, blocks
,
318 #define ext4_journal_stop(handle) \
319 __ext4_journal_stop(__func__, __LINE__, (handle))
321 #define ext4_journal_start_reserved(handle, type) \
322 __ext4_journal_start_reserved((handle), __LINE__, (type))
324 handle_t
*__ext4_journal_start_reserved(handle_t
*handle
, unsigned int line
,
327 static inline void ext4_journal_free_reserved(handle_t
*handle
)
329 if (ext4_handle_valid(handle
))
330 jbd2_journal_free_reserved(handle
);
333 static inline handle_t
*ext4_journal_current_handle(void)
335 return journal_current_handle();
338 static inline int ext4_journal_extend(handle_t
*handle
, int nblocks
)
340 if (ext4_handle_valid(handle
))
341 return jbd2_journal_extend(handle
, nblocks
);
345 static inline int ext4_journal_restart(handle_t
*handle
, int nblocks
)
347 if (ext4_handle_valid(handle
))
348 return jbd2_journal_restart(handle
, nblocks
);
352 static inline int ext4_journal_blocks_per_page(struct inode
*inode
)
354 if (EXT4_JOURNAL(inode
) != NULL
)
355 return jbd2_journal_blocks_per_page(inode
);
359 static inline int ext4_journal_force_commit(journal_t
*journal
)
362 return jbd2_journal_force_commit(journal
);
366 static inline int ext4_jbd2_inode_add_write(handle_t
*handle
,
369 if (ext4_handle_valid(handle
))
370 return jbd2_journal_inode_add_write(handle
,
371 EXT4_I(inode
)->jinode
);
375 static inline int ext4_jbd2_inode_add_wait(handle_t
*handle
,
378 if (ext4_handle_valid(handle
))
379 return jbd2_journal_inode_add_wait(handle
,
380 EXT4_I(inode
)->jinode
);
384 static inline void ext4_update_inode_fsync_trans(handle_t
*handle
,
388 struct ext4_inode_info
*ei
= EXT4_I(inode
);
390 if (ext4_handle_valid(handle
)) {
391 ei
->i_sync_tid
= handle
->h_transaction
->t_tid
;
393 ei
->i_datasync_tid
= handle
->h_transaction
->t_tid
;
398 int ext4_force_commit(struct super_block
*sb
);
401 * Ext4 inode journal modes
403 #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
404 #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
405 #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
407 static inline int ext4_inode_journal_mode(struct inode
*inode
)
409 if (EXT4_JOURNAL(inode
) == NULL
)
410 return EXT4_INODE_WRITEBACK_DATA_MODE
; /* writeback */
411 /* We do not support data journalling with delayed allocation */
412 if (!S_ISREG(inode
->i_mode
) ||
413 test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
||
414 (ext4_test_inode_flag(inode
, EXT4_INODE_JOURNAL_DATA
) &&
415 !test_opt(inode
->i_sb
, DELALLOC
))) {
416 /* We do not support data journalling for encrypted data */
417 if (S_ISREG(inode
->i_mode
) && ext4_encrypted_inode(inode
))
418 return EXT4_INODE_ORDERED_DATA_MODE
; /* ordered */
419 return EXT4_INODE_JOURNAL_DATA_MODE
; /* journal data */
421 if (test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
422 return EXT4_INODE_ORDERED_DATA_MODE
; /* ordered */
423 if (test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
424 return EXT4_INODE_WRITEBACK_DATA_MODE
; /* writeback */
428 static inline int ext4_should_journal_data(struct inode
*inode
)
430 return ext4_inode_journal_mode(inode
) & EXT4_INODE_JOURNAL_DATA_MODE
;
433 static inline int ext4_should_order_data(struct inode
*inode
)
435 return ext4_inode_journal_mode(inode
) & EXT4_INODE_ORDERED_DATA_MODE
;
438 static inline int ext4_should_writeback_data(struct inode
*inode
)
440 return ext4_inode_journal_mode(inode
) & EXT4_INODE_WRITEBACK_DATA_MODE
;
444 * This function controls whether or not we should try to go down the
445 * dioread_nolock code paths, which makes it safe to avoid taking
446 * i_mutex for direct I/O reads. This only works for extent-based
447 * files, and it doesn't work if data journaling is enabled, since the
448 * dioread_nolock code uses b_private to pass information back to the
449 * I/O completion handler, and this conflicts with the jbd's use of
452 static inline int ext4_should_dioread_nolock(struct inode
*inode
)
454 if (!test_opt(inode
->i_sb
, DIOREAD_NOLOCK
))
456 if (!S_ISREG(inode
->i_mode
))
458 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
460 if (ext4_should_journal_data(inode
))
465 #endif /* _EXT4_JBD2_H */