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_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS) \
40 /* Extended attribute operations touch at most two data buffers,
41 * two bitmap buffers, and two group summaries, in addition to the inode
42 * and the superblock, which are already accounted for. */
44 #define EXT4_XATTR_TRANS_BLOCKS 6U
46 /* Define the minimum size for a transaction which modifies data. This
47 * needs to take into account the fact that we may end up modifying two
48 * quota files too (one for the group, one for the user quota). The
49 * superblock only gets updated once, of course, so don't bother
50 * counting that again for the quota updates. */
52 #define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
53 EXT4_XATTR_TRANS_BLOCKS - 2 + \
54 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
57 * Define the number of metadata blocks we need to account to modify data.
59 * This include super block, inode block, quota blocks and xattr blocks
61 #define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \
62 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
64 /* Define an arbitrary limit for the amount of data we will anticipate
65 * writing to any given transaction. For unbounded transactions such as
66 * write(2) and truncate(2) we can write more than this, but we always
67 * start off at the maximum transaction size and grow the transaction
68 * optimistically as we go. */
70 #define EXT4_MAX_TRANS_DATA 64U
72 /* We break up a large truncate or write transaction once the handle's
73 * buffer credits gets this low, we need either to extend the
74 * transaction or to start a new one. Reserve enough space here for
75 * inode, bitmap, superblock, group and indirection updates for at least
76 * one block, plus two quota updates. Quota allocations are not
79 #define EXT4_RESERVE_TRANS_BLOCKS 12U
81 #define EXT4_INDEX_EXTRA_TRANS_BLOCKS 8
84 /* Amount of blocks needed for quota update - we know that the structure was
85 * allocated so we need to update only data block */
86 #define EXT4_QUOTA_TRANS_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
87 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
89 /* Amount of blocks needed for quota insert/delete - we do some block writes
90 * but inode, sb and group updates are done only once */
91 #define EXT4_QUOTA_INIT_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
92 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
93 (DQUOT_INIT_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
94 +3+DQUOT_INIT_REWRITE) : 0)
96 #define EXT4_QUOTA_DEL_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
97 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\
98 (DQUOT_DEL_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
99 +3+DQUOT_DEL_REWRITE) : 0)
101 #define EXT4_QUOTA_TRANS_BLOCKS(sb) 0
102 #define EXT4_QUOTA_INIT_BLOCKS(sb) 0
103 #define EXT4_QUOTA_DEL_BLOCKS(sb) 0
105 #define EXT4_MAXQUOTAS_TRANS_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
106 #define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
107 #define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_DEL_BLOCKS(sb))
109 static inline int ext4_jbd2_credits_xattr(struct inode
*inode
)
111 int credits
= EXT4_DATA_TRANS_BLOCKS(inode
->i_sb
);
114 * In case of inline data, we may push out the data to a block,
115 * so we need to reserve credits for this eventuality
117 if (ext4_has_inline_data(inode
))
118 credits
+= ext4_writepage_trans_blocks(inode
) + 1;
124 * Ext4 handle operation types -- for logging purposes
126 #define EXT4_HT_MISC 0
127 #define EXT4_HT_INODE 1
128 #define EXT4_HT_WRITE_PAGE 2
129 #define EXT4_HT_MAP_BLOCKS 3
130 #define EXT4_HT_DIR 4
131 #define EXT4_HT_TRUNCATE 5
132 #define EXT4_HT_QUOTA 6
133 #define EXT4_HT_RESIZE 7
134 #define EXT4_HT_MIGRATE 8
135 #define EXT4_HT_MOVE_EXTENTS 9
136 #define EXT4_HT_XATTR 10
137 #define EXT4_HT_EXT_CONVERT 11
138 #define EXT4_HT_MAX 12
141 * struct ext4_journal_cb_entry - Base structure for callback information.
143 * This struct is a 'seed' structure for a using with your own callback
144 * structs. If you are using callbacks you must allocate one of these
145 * or another struct of your own definition which has this struct
146 * as it's first element and pass it to ext4_journal_callback_add().
148 struct ext4_journal_cb_entry
{
149 /* list information for other callbacks attached to the same handle */
150 struct list_head jce_list
;
152 /* Function to call with this callback structure */
153 void (*jce_func
)(struct super_block
*sb
,
154 struct ext4_journal_cb_entry
*jce
, int error
);
156 /* user data goes here */
160 * ext4_journal_callback_add: add a function to call after transaction commit
161 * @handle: active journal transaction handle to register callback on
162 * @func: callback function to call after the transaction has committed:
163 * @sb: superblock of current filesystem for transaction
164 * @jce: returned journal callback data
165 * @rc: journal state at commit (0 = transaction committed properly)
166 * @jce: journal callback data (internal and function private data struct)
168 * The registered function will be called in the context of the journal thread
169 * after the transaction for which the handle was created has completed.
171 * No locks are held when the callback function is called, so it is safe to
172 * call blocking functions from within the callback, but the callback should
173 * not block or run for too long, or the filesystem will be blocked waiting for
174 * the next transaction to commit. No journaling functions can be used, or
175 * there is a risk of deadlock.
177 * There is no guaranteed calling order of multiple registered callbacks on
178 * the same transaction.
180 static inline void ext4_journal_callback_add(handle_t
*handle
,
181 void (*func
)(struct super_block
*sb
,
182 struct ext4_journal_cb_entry
*jce
,
184 struct ext4_journal_cb_entry
*jce
)
186 struct ext4_sb_info
*sbi
=
187 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
189 /* Add the jce to transaction's private list */
190 jce
->jce_func
= func
;
191 spin_lock(&sbi
->s_md_lock
);
192 list_add_tail(&jce
->jce_list
, &handle
->h_transaction
->t_private_list
);
193 spin_unlock(&sbi
->s_md_lock
);
197 * ext4_journal_callback_del: delete a registered callback
198 * @handle: active journal transaction handle on which callback was registered
199 * @jce: registered journal callback entry to unregister
200 * Return true if object was successfully removed
202 static inline bool ext4_journal_callback_try_del(handle_t
*handle
,
203 struct ext4_journal_cb_entry
*jce
)
206 struct ext4_sb_info
*sbi
=
207 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
209 spin_lock(&sbi
->s_md_lock
);
210 deleted
= !list_empty(&jce
->jce_list
);
211 list_del_init(&jce
->jce_list
);
212 spin_unlock(&sbi
->s_md_lock
);
217 ext4_mark_iloc_dirty(handle_t
*handle
,
219 struct ext4_iloc
*iloc
);
222 * On success, We end up with an outstanding reference count against
223 * iloc->bh. This _must_ be cleaned up later.
226 int ext4_reserve_inode_write(handle_t
*handle
, struct inode
*inode
,
227 struct ext4_iloc
*iloc
);
229 int ext4_mark_inode_dirty(handle_t
*handle
, struct inode
*inode
);
232 * Wrapper functions with which ext4 calls into JBD.
234 int __ext4_journal_get_write_access(const char *where
, unsigned int line
,
235 handle_t
*handle
, struct buffer_head
*bh
);
237 int __ext4_forget(const char *where
, unsigned int line
, handle_t
*handle
,
238 int is_metadata
, struct inode
*inode
,
239 struct buffer_head
*bh
, ext4_fsblk_t blocknr
);
241 int __ext4_journal_get_create_access(const char *where
, unsigned int line
,
242 handle_t
*handle
, struct buffer_head
*bh
);
244 int __ext4_handle_dirty_metadata(const char *where
, unsigned int line
,
245 handle_t
*handle
, struct inode
*inode
,
246 struct buffer_head
*bh
);
248 int __ext4_handle_dirty_super(const char *where
, unsigned int line
,
249 handle_t
*handle
, struct super_block
*sb
);
251 #define ext4_journal_get_write_access(handle, bh) \
252 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
253 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
254 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
256 #define ext4_journal_get_create_access(handle, bh) \
257 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
258 #define ext4_handle_dirty_metadata(handle, inode, bh) \
259 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
261 #define ext4_handle_dirty_super(handle, sb) \
262 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
264 handle_t
*__ext4_journal_start_sb(struct super_block
*sb
, unsigned int line
,
265 int type
, int blocks
, int rsv_blocks
);
266 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
);
268 #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
270 /* Note: Do not use this for NULL handles. This is only to determine if
271 * a properly allocated handle is using a journal or not. */
272 static inline int ext4_handle_valid(handle_t
*handle
)
274 if ((unsigned long)handle
< EXT4_NOJOURNAL_MAX_REF_COUNT
)
279 static inline void ext4_handle_sync(handle_t
*handle
)
281 if (ext4_handle_valid(handle
))
285 static inline int ext4_handle_is_aborted(handle_t
*handle
)
287 if (ext4_handle_valid(handle
))
288 return is_handle_aborted(handle
);
292 static inline int ext4_handle_has_enough_credits(handle_t
*handle
, int needed
)
294 if (ext4_handle_valid(handle
) && handle
->h_buffer_credits
< needed
)
299 #define ext4_journal_start_sb(sb, type, nblocks) \
300 __ext4_journal_start_sb((sb), __LINE__, (type), (nblocks), 0)
302 #define ext4_journal_start(inode, type, nblocks) \
303 __ext4_journal_start((inode), __LINE__, (type), (nblocks), 0)
305 #define ext4_journal_start_with_reserve(inode, type, blocks, rsv_blocks) \
306 __ext4_journal_start((inode), __LINE__, (type), (blocks), (rsv_blocks))
308 static inline handle_t
*__ext4_journal_start(struct inode
*inode
,
309 unsigned int line
, int type
,
310 int blocks
, int rsv_blocks
)
312 return __ext4_journal_start_sb(inode
->i_sb
, line
, type
, blocks
,
316 #define ext4_journal_stop(handle) \
317 __ext4_journal_stop(__func__, __LINE__, (handle))
319 #define ext4_journal_start_reserved(handle, type) \
320 __ext4_journal_start_reserved((handle), __LINE__, (type))
322 handle_t
*__ext4_journal_start_reserved(handle_t
*handle
, unsigned int line
,
325 static inline void ext4_journal_free_reserved(handle_t
*handle
)
327 if (ext4_handle_valid(handle
))
328 jbd2_journal_free_reserved(handle
);
331 static inline handle_t
*ext4_journal_current_handle(void)
333 return journal_current_handle();
336 static inline int ext4_journal_extend(handle_t
*handle
, int nblocks
)
338 if (ext4_handle_valid(handle
))
339 return jbd2_journal_extend(handle
, nblocks
);
343 static inline int ext4_journal_restart(handle_t
*handle
, int nblocks
)
345 if (ext4_handle_valid(handle
))
346 return jbd2_journal_restart(handle
, nblocks
);
350 static inline int ext4_journal_blocks_per_page(struct inode
*inode
)
352 if (EXT4_JOURNAL(inode
) != NULL
)
353 return jbd2_journal_blocks_per_page(inode
);
357 static inline int ext4_journal_force_commit(journal_t
*journal
)
360 return jbd2_journal_force_commit(journal
);
364 static inline int ext4_jbd2_file_inode(handle_t
*handle
, struct inode
*inode
)
366 if (ext4_handle_valid(handle
))
367 return jbd2_journal_file_inode(handle
, EXT4_I(inode
)->jinode
);
371 static inline void ext4_update_inode_fsync_trans(handle_t
*handle
,
375 struct ext4_inode_info
*ei
= EXT4_I(inode
);
377 if (ext4_handle_valid(handle
)) {
378 ei
->i_sync_tid
= handle
->h_transaction
->t_tid
;
380 ei
->i_datasync_tid
= handle
->h_transaction
->t_tid
;
385 int ext4_force_commit(struct super_block
*sb
);
388 * Ext4 inode journal modes
390 #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
391 #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
392 #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
394 static inline int ext4_inode_journal_mode(struct inode
*inode
)
396 if (EXT4_JOURNAL(inode
) == NULL
)
397 return EXT4_INODE_WRITEBACK_DATA_MODE
; /* writeback */
398 /* We do not support data journalling with delayed allocation */
399 if (!S_ISREG(inode
->i_mode
) ||
400 test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
401 return EXT4_INODE_JOURNAL_DATA_MODE
; /* journal data */
402 if (ext4_test_inode_flag(inode
, EXT4_INODE_JOURNAL_DATA
) &&
403 !test_opt(inode
->i_sb
, DELALLOC
))
404 return EXT4_INODE_JOURNAL_DATA_MODE
; /* journal data */
405 if (test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_ORDERED_DATA
)
406 return EXT4_INODE_ORDERED_DATA_MODE
; /* ordered */
407 if (test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_WRITEBACK_DATA
)
408 return EXT4_INODE_WRITEBACK_DATA_MODE
; /* writeback */
413 static inline int ext4_should_journal_data(struct inode
*inode
)
415 return ext4_inode_journal_mode(inode
) & EXT4_INODE_JOURNAL_DATA_MODE
;
418 static inline int ext4_should_order_data(struct inode
*inode
)
420 return ext4_inode_journal_mode(inode
) & EXT4_INODE_ORDERED_DATA_MODE
;
423 static inline int ext4_should_writeback_data(struct inode
*inode
)
425 return ext4_inode_journal_mode(inode
) & EXT4_INODE_WRITEBACK_DATA_MODE
;
429 * This function controls whether or not we should try to go down the
430 * dioread_nolock code paths, which makes it safe to avoid taking
431 * i_mutex for direct I/O reads. This only works for extent-based
432 * files, and it doesn't work if data journaling is enabled, since the
433 * dioread_nolock code uses b_private to pass information back to the
434 * I/O completion handler, and this conflicts with the jbd's use of
437 static inline int ext4_should_dioread_nolock(struct inode
*inode
)
439 if (!test_opt(inode
->i_sb
, DIOREAD_NOLOCK
))
441 if (!S_ISREG(inode
->i_mode
))
443 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
445 if (ext4_should_journal_data(inode
))
450 #endif /* _EXT4_JBD2_H */