1 // SPDX-License-Identifier: GPL-2.0+
5 * Written by Stephen C. Tweedie <sct@redhat.com>, 1999
7 * Copyright 1998--1999 Red Hat corp --- All Rights Reserved
9 * Ext4-specific journaling extensions.
16 #include <linux/jbd2.h>
19 #define EXT4_JOURNAL(inode) (EXT4_SB((inode)->i_sb)->s_journal)
21 /* Define the number of blocks we need to account to a transaction to
22 * modify one block of data.
24 * We may have to touch one inode, one bitmap buffer, up to three
25 * indirection blocks, the group and superblock summaries, and the data
26 * block to complete the transaction.
28 * For extents-enabled fs we may have to allocate and modify up to
29 * 5 levels of tree, data block (for each of these we need bitmap + group
30 * summaries), root which is stored in the inode, sb
33 #define EXT4_SINGLEDATA_TRANS_BLOCKS(sb) \
34 (ext4_has_feature_extents(sb) ? 20U : 8U)
36 /* Extended attribute operations touch at most two data buffers,
37 * two bitmap buffers, and two group summaries, in addition to the inode
38 * and the superblock, which are already accounted for. */
40 #define EXT4_XATTR_TRANS_BLOCKS 6U
42 /* Define the minimum size for a transaction which modifies data. This
43 * needs to take into account the fact that we may end up modifying two
44 * quota files too (one for the group, one for the user quota). The
45 * superblock only gets updated once, of course, so don't bother
46 * counting that again for the quota updates. */
48 #define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
49 EXT4_XATTR_TRANS_BLOCKS - 2 + \
50 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
53 * Define the number of metadata blocks we need to account to modify data.
55 * This include super block, inode block, quota blocks and xattr blocks
57 #define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \
58 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
60 /* Define an arbitrary limit for the amount of data we will anticipate
61 * writing to any given transaction. For unbounded transactions such as
62 * write(2) and truncate(2) we can write more than this, but we always
63 * start off at the maximum transaction size and grow the transaction
64 * optimistically as we go. */
66 #define EXT4_MAX_TRANS_DATA 64U
68 /* We break up a large truncate or write transaction once the handle's
69 * buffer credits gets this low, we need either to extend the
70 * transaction or to start a new one. Reserve enough space here for
71 * inode, bitmap, superblock, group and indirection updates for at least
72 * one block, plus two quota updates. Quota allocations are not
75 #define EXT4_RESERVE_TRANS_BLOCKS 12U
78 * Number of credits needed if we need to insert an entry into a
79 * directory. For each new index block, we need 4 blocks (old index
80 * block, new index block, bitmap block, bg summary). For normal
81 * htree directories there are 2 levels; if the largedir feature
82 * enabled it's 3 levels.
84 #define EXT4_INDEX_EXTRA_TRANS_BLOCKS 12U
87 /* Amount of blocks needed for quota update - we know that the structure was
88 * allocated so we need to update only data block */
89 #define EXT4_QUOTA_TRANS_BLOCKS(sb) ((ext4_quota_capable(sb)) ? 1 : 0)
90 /* Amount of blocks needed for quota insert/delete - we do some block writes
91 * but inode, sb and group updates are done only once */
92 #define EXT4_QUOTA_INIT_BLOCKS(sb) ((ext4_quota_capable(sb)) ?\
93 (DQUOT_INIT_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
94 +3+DQUOT_INIT_REWRITE) : 0)
96 #define EXT4_QUOTA_DEL_BLOCKS(sb) ((ext4_quota_capable(sb)) ?\
97 (DQUOT_DEL_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
98 +3+DQUOT_DEL_REWRITE) : 0)
100 #define EXT4_QUOTA_TRANS_BLOCKS(sb) 0
101 #define EXT4_QUOTA_INIT_BLOCKS(sb) 0
102 #define EXT4_QUOTA_DEL_BLOCKS(sb) 0
104 #define EXT4_MAXQUOTAS_TRANS_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
105 #define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
106 #define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_DEL_BLOCKS(sb))
109 * Ext4 handle operation types -- for logging purposes
111 #define EXT4_HT_MISC 0
112 #define EXT4_HT_INODE 1
113 #define EXT4_HT_WRITE_PAGE 2
114 #define EXT4_HT_MAP_BLOCKS 3
115 #define EXT4_HT_DIR 4
116 #define EXT4_HT_TRUNCATE 5
117 #define EXT4_HT_QUOTA 6
118 #define EXT4_HT_RESIZE 7
119 #define EXT4_HT_MIGRATE 8
120 #define EXT4_HT_MOVE_EXTENTS 9
121 #define EXT4_HT_XATTR 10
122 #define EXT4_HT_EXT_CONVERT 11
123 #define EXT4_HT_MAX 12
126 * struct ext4_journal_cb_entry - Base structure for callback information.
128 * This struct is a 'seed' structure for a using with your own callback
129 * structs. If you are using callbacks you must allocate one of these
130 * or another struct of your own definition which has this struct
131 * as it's first element and pass it to ext4_journal_callback_add().
133 struct ext4_journal_cb_entry
{
134 /* list information for other callbacks attached to the same handle */
135 struct list_head jce_list
;
137 /* Function to call with this callback structure */
138 void (*jce_func
)(struct super_block
*sb
,
139 struct ext4_journal_cb_entry
*jce
, int error
);
141 /* user data goes here */
145 * ext4_journal_callback_add: add a function to call after transaction commit
146 * @handle: active journal transaction handle to register callback on
147 * @func: callback function to call after the transaction has committed:
148 * @sb: superblock of current filesystem for transaction
149 * @jce: returned journal callback data
150 * @rc: journal state at commit (0 = transaction committed properly)
151 * @jce: journal callback data (internal and function private data struct)
153 * The registered function will be called in the context of the journal thread
154 * after the transaction for which the handle was created has completed.
156 * No locks are held when the callback function is called, so it is safe to
157 * call blocking functions from within the callback, but the callback should
158 * not block or run for too long, or the filesystem will be blocked waiting for
159 * the next transaction to commit. No journaling functions can be used, or
160 * there is a risk of deadlock.
162 * There is no guaranteed calling order of multiple registered callbacks on
163 * the same transaction.
165 static inline void _ext4_journal_callback_add(handle_t
*handle
,
166 struct ext4_journal_cb_entry
*jce
)
168 /* Add the jce to transaction's private list */
169 list_add_tail(&jce
->jce_list
, &handle
->h_transaction
->t_private_list
);
172 static inline void ext4_journal_callback_add(handle_t
*handle
,
173 void (*func
)(struct super_block
*sb
,
174 struct ext4_journal_cb_entry
*jce
,
176 struct ext4_journal_cb_entry
*jce
)
178 struct ext4_sb_info
*sbi
=
179 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
181 /* Add the jce to transaction's private list */
182 jce
->jce_func
= func
;
183 spin_lock(&sbi
->s_md_lock
);
184 _ext4_journal_callback_add(handle
, jce
);
185 spin_unlock(&sbi
->s_md_lock
);
190 * ext4_journal_callback_del: delete a registered callback
191 * @handle: active journal transaction handle on which callback was registered
192 * @jce: registered journal callback entry to unregister
193 * Return true if object was successfully removed
195 static inline bool ext4_journal_callback_try_del(handle_t
*handle
,
196 struct ext4_journal_cb_entry
*jce
)
199 struct ext4_sb_info
*sbi
=
200 EXT4_SB(handle
->h_transaction
->t_journal
->j_private
);
202 spin_lock(&sbi
->s_md_lock
);
203 deleted
= !list_empty(&jce
->jce_list
);
204 list_del_init(&jce
->jce_list
);
205 spin_unlock(&sbi
->s_md_lock
);
210 ext4_mark_iloc_dirty(handle_t
*handle
,
212 struct ext4_iloc
*iloc
);
215 * On success, We end up with an outstanding reference count against
216 * iloc->bh. This _must_ be cleaned up later.
219 int ext4_reserve_inode_write(handle_t
*handle
, struct inode
*inode
,
220 struct ext4_iloc
*iloc
);
222 #define ext4_mark_inode_dirty(__h, __i) \
223 __ext4_mark_inode_dirty((__h), (__i), __func__, __LINE__)
224 int __ext4_mark_inode_dirty(handle_t
*handle
, struct inode
*inode
,
225 const char *func
, unsigned int line
);
227 int ext4_expand_extra_isize(struct inode
*inode
,
228 unsigned int new_extra_isize
,
229 struct ext4_iloc
*iloc
);
231 * Wrapper functions with which ext4 calls into JBD.
233 int __ext4_journal_get_write_access(const char *where
, unsigned int line
,
234 handle_t
*handle
, struct super_block
*sb
,
235 struct buffer_head
*bh
,
236 enum ext4_journal_trigger_type trigger_type
);
238 int __ext4_forget(const char *where
, unsigned int line
, handle_t
*handle
,
239 int is_metadata
, struct inode
*inode
,
240 struct buffer_head
*bh
, ext4_fsblk_t blocknr
);
242 int __ext4_journal_get_create_access(const char *where
, unsigned int line
,
243 handle_t
*handle
, struct super_block
*sb
,
244 struct buffer_head
*bh
,
245 enum ext4_journal_trigger_type trigger_type
);
247 int __ext4_handle_dirty_metadata(const char *where
, unsigned int line
,
248 handle_t
*handle
, struct inode
*inode
,
249 struct buffer_head
*bh
);
251 #define ext4_journal_get_write_access(handle, sb, bh, trigger_type) \
252 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (sb), \
253 (bh), (trigger_type))
254 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
255 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
257 #define ext4_journal_get_create_access(handle, sb, bh, trigger_type) \
258 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (sb), \
259 (bh), (trigger_type))
260 #define ext4_handle_dirty_metadata(handle, inode, bh) \
261 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
264 handle_t
*__ext4_journal_start_sb(struct inode
*inode
, struct super_block
*sb
,
265 unsigned int line
, int type
, int blocks
,
266 int rsv_blocks
, int revoke_creds
);
267 int __ext4_journal_stop(const char *where
, unsigned int line
, handle_t
*handle
);
269 #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
271 /* Note: Do not use this for NULL handles. This is only to determine if
272 * a properly allocated handle is using a journal or not. */
273 static inline int ext4_handle_valid(handle_t
*handle
)
275 if ((unsigned long)handle
< EXT4_NOJOURNAL_MAX_REF_COUNT
)
280 static inline void ext4_handle_sync(handle_t
*handle
)
282 if (ext4_handle_valid(handle
))
286 static inline int ext4_handle_is_aborted(handle_t
*handle
)
288 if (ext4_handle_valid(handle
))
289 return is_handle_aborted(handle
);
293 static inline int ext4_free_metadata_revoke_credits(struct super_block
*sb
,
296 /* Freeing each metadata block can result in freeing one cluster */
297 return blocks
* EXT4_SB(sb
)->s_cluster_ratio
;
300 static inline int ext4_trans_default_revoke_credits(struct super_block
*sb
)
302 return ext4_free_metadata_revoke_credits(sb
, 8);
305 #define ext4_journal_start_sb(sb, type, nblocks) \
306 __ext4_journal_start_sb(NULL, (sb), __LINE__, (type), (nblocks), 0,\
307 ext4_trans_default_revoke_credits(sb))
309 #define ext4_journal_start(inode, type, nblocks) \
310 __ext4_journal_start((inode), __LINE__, (type), (nblocks), 0, \
311 ext4_trans_default_revoke_credits((inode)->i_sb))
313 #define ext4_journal_start_with_reserve(inode, type, blocks, rsv_blocks)\
314 __ext4_journal_start((inode), __LINE__, (type), (blocks), (rsv_blocks),\
315 ext4_trans_default_revoke_credits((inode)->i_sb))
317 #define ext4_journal_start_with_revoke(inode, type, blocks, revoke_creds) \
318 __ext4_journal_start((inode), __LINE__, (type), (blocks), 0, \
321 static inline handle_t
*__ext4_journal_start(struct inode
*inode
,
322 unsigned int line
, int type
,
323 int blocks
, int rsv_blocks
,
326 return __ext4_journal_start_sb(inode
, inode
->i_sb
, line
, type
, blocks
,
327 rsv_blocks
, revoke_creds
);
330 #define ext4_journal_stop(handle) \
331 __ext4_journal_stop(__func__, __LINE__, (handle))
333 #define ext4_journal_start_reserved(handle, type) \
334 __ext4_journal_start_reserved((handle), __LINE__, (type))
336 handle_t
*__ext4_journal_start_reserved(handle_t
*handle
, unsigned int line
,
339 static inline handle_t
*ext4_journal_current_handle(void)
341 return journal_current_handle();
344 static inline int ext4_journal_extend(handle_t
*handle
, int nblocks
, int revoke
)
346 if (ext4_handle_valid(handle
))
347 return jbd2_journal_extend(handle
, nblocks
, revoke
);
351 static inline int ext4_journal_restart(handle_t
*handle
, int nblocks
,
354 if (ext4_handle_valid(handle
))
355 return jbd2__journal_restart(handle
, nblocks
, revoke
, GFP_NOFS
);
359 int __ext4_journal_ensure_credits(handle_t
*handle
, int check_cred
,
360 int extend_cred
, int revoke_cred
);
364 * Ensure @handle has at least @check_creds credits available. If not,
365 * transaction will be extended or restarted to contain at least @extend_cred
366 * credits. Before restarting transaction @fn is executed to allow for cleanup
367 * before the transaction is restarted.
369 * The return value is < 0 in case of error, 0 in case the handle has enough
370 * credits or transaction extension succeeded, 1 in case transaction had to be
373 #define ext4_journal_ensure_credits_fn(handle, check_cred, extend_cred, \
376 __label__ __ensure_end; \
377 int err = __ext4_journal_ensure_credits((handle), (check_cred), \
378 (extend_cred), (revoke_cred)); \
385 err = ext4_journal_restart((handle), (extend_cred), (revoke_cred)); \
393 * Ensure given handle has at least requested amount of credits available,
394 * possibly restarting transaction if needed. We also make sure the transaction
395 * has space for at least ext4_trans_default_revoke_credits(sb) revoke records
396 * as freeing one or two blocks is very common pattern and requesting this is
399 static inline int ext4_journal_ensure_credits(handle_t
*handle
, int credits
,
402 return ext4_journal_ensure_credits_fn(handle
, credits
, credits
,
406 static inline int ext4_journal_blocks_per_page(struct inode
*inode
)
408 if (EXT4_JOURNAL(inode
) != NULL
)
409 return jbd2_journal_blocks_per_page(inode
);
413 static inline int ext4_journal_force_commit(journal_t
*journal
)
416 return jbd2_journal_force_commit(journal
);
420 static inline int ext4_jbd2_inode_add_write(handle_t
*handle
,
421 struct inode
*inode
, loff_t start_byte
, loff_t length
)
423 if (ext4_handle_valid(handle
))
424 return jbd2_journal_inode_ranged_write(handle
,
425 EXT4_I(inode
)->jinode
, start_byte
, length
);
429 static inline int ext4_jbd2_inode_add_wait(handle_t
*handle
,
430 struct inode
*inode
, loff_t start_byte
, loff_t length
)
432 if (ext4_handle_valid(handle
))
433 return jbd2_journal_inode_ranged_wait(handle
,
434 EXT4_I(inode
)->jinode
, start_byte
, length
);
438 static inline void ext4_update_inode_fsync_trans(handle_t
*handle
,
442 struct ext4_inode_info
*ei
= EXT4_I(inode
);
444 if (ext4_handle_valid(handle
) && !is_handle_aborted(handle
)) {
445 ei
->i_sync_tid
= handle
->h_transaction
->t_tid
;
447 ei
->i_datasync_tid
= handle
->h_transaction
->t_tid
;
452 int ext4_force_commit(struct super_block
*sb
);
455 * Ext4 inode journal modes
457 #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
458 #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
459 #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
461 int ext4_inode_journal_mode(struct inode
*inode
);
463 static inline int ext4_should_journal_data(struct inode
*inode
)
465 return ext4_inode_journal_mode(inode
) & EXT4_INODE_JOURNAL_DATA_MODE
;
468 static inline int ext4_should_order_data(struct inode
*inode
)
470 return ext4_inode_journal_mode(inode
) & EXT4_INODE_ORDERED_DATA_MODE
;
473 static inline int ext4_should_writeback_data(struct inode
*inode
)
475 return ext4_inode_journal_mode(inode
) & EXT4_INODE_WRITEBACK_DATA_MODE
;
478 static inline int ext4_free_data_revoke_credits(struct inode
*inode
, int blocks
)
480 if (test_opt(inode
->i_sb
, DATA_FLAGS
) == EXT4_MOUNT_JOURNAL_DATA
)
482 if (!ext4_should_journal_data(inode
))
485 * Data blocks in one extent are contiguous, just account for partial
486 * clusters at extent boundaries
488 return blocks
+ 2*(EXT4_SB(inode
->i_sb
)->s_cluster_ratio
- 1);
492 * This function controls whether or not we should try to go down the
493 * dioread_nolock code paths, which makes it safe to avoid taking
494 * i_rwsem for direct I/O reads. This only works for extent-based
495 * files, and it doesn't work if data journaling is enabled, since the
496 * dioread_nolock code uses b_private to pass information back to the
497 * I/O completion handler, and this conflicts with the jbd's use of
500 static inline int ext4_should_dioread_nolock(struct inode
*inode
)
502 if (!test_opt(inode
->i_sb
, DIOREAD_NOLOCK
))
504 if (!S_ISREG(inode
->i_mode
))
506 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
508 if (ext4_should_journal_data(inode
))
510 /* temporary fix to prevent generic/422 test failures */
511 if (!test_opt(inode
->i_sb
, DELALLOC
))
516 #endif /* _EXT4_JBD2_H */