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
80 #define EXT4_INDEX_EXTRA_TRANS_BLOCKS 8
83 /* Amount of blocks needed for quota update - we know that the structure was
84 * allocated so we need to update only data block */
85 #define EXT4_QUOTA_TRANS_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
86 ext4_has_feature_quota(sb)) ? 1 : 0)
87 /* Amount of blocks needed for quota insert/delete - we do some block writes
88 * but inode, sb and group updates are done only once */
89 #define EXT4_QUOTA_INIT_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
90 ext4_has_feature_quota(sb)) ?\
91 (DQUOT_INIT_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
92 +3+DQUOT_INIT_REWRITE) : 0)
94 #define EXT4_QUOTA_DEL_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\
95 ext4_has_feature_quota(sb)) ?\
96 (DQUOT_DEL_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\
97 +3+DQUOT_DEL_REWRITE) : 0)
99 #define EXT4_QUOTA_TRANS_BLOCKS(sb) 0
100 #define EXT4_QUOTA_INIT_BLOCKS(sb) 0
101 #define EXT4_QUOTA_DEL_BLOCKS(sb) 0
103 #define EXT4_MAXQUOTAS_TRANS_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
104 #define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
105 #define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (EXT4_MAXQUOTAS*EXT4_QUOTA_DEL_BLOCKS(sb))
107 static inline int ext4_jbd2_credits_xattr(struct inode *inode)
109 int credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb);
112 * In case of inline data, we may push out the data to a block,
113 * so we need to reserve credits for this eventuality
115 if (ext4_has_inline_data(inode))
116 credits += ext4_writepage_trans_blocks(inode) + 1;
122 * Ext4 handle operation types -- for logging purposes
124 #define EXT4_HT_MISC 0
125 #define EXT4_HT_INODE 1
126 #define EXT4_HT_WRITE_PAGE 2
127 #define EXT4_HT_MAP_BLOCKS 3
128 #define EXT4_HT_DIR 4
129 #define EXT4_HT_TRUNCATE 5
130 #define EXT4_HT_QUOTA 6
131 #define EXT4_HT_RESIZE 7
132 #define EXT4_HT_MIGRATE 8
133 #define EXT4_HT_MOVE_EXTENTS 9
134 #define EXT4_HT_XATTR 10
135 #define EXT4_HT_EXT_CONVERT 11
136 #define EXT4_HT_MAX 12
139 * struct ext4_journal_cb_entry - Base structure for callback information.
141 * This struct is a 'seed' structure for a using with your own callback
142 * structs. If you are using callbacks you must allocate one of these
143 * or another struct of your own definition which has this struct
144 * as it's first element and pass it to ext4_journal_callback_add().
146 struct ext4_journal_cb_entry {
147 /* list information for other callbacks attached to the same handle */
148 struct list_head jce_list;
150 /* Function to call with this callback structure */
151 void (*jce_func)(struct super_block *sb,
152 struct ext4_journal_cb_entry *jce, int error);
154 /* user data goes here */
158 * ext4_journal_callback_add: add a function to call after transaction commit
159 * @handle: active journal transaction handle to register callback on
160 * @func: callback function to call after the transaction has committed:
161 * @sb: superblock of current filesystem for transaction
162 * @jce: returned journal callback data
163 * @rc: journal state at commit (0 = transaction committed properly)
164 * @jce: journal callback data (internal and function private data struct)
166 * The registered function will be called in the context of the journal thread
167 * after the transaction for which the handle was created has completed.
169 * No locks are held when the callback function is called, so it is safe to
170 * call blocking functions from within the callback, but the callback should
171 * not block or run for too long, or the filesystem will be blocked waiting for
172 * the next transaction to commit. No journaling functions can be used, or
173 * there is a risk of deadlock.
175 * There is no guaranteed calling order of multiple registered callbacks on
176 * the same transaction.
178 static inline void _ext4_journal_callback_add(handle_t *handle,
179 struct ext4_journal_cb_entry *jce)
181 /* Add the jce to transaction's private list */
182 list_add_tail(&jce->jce_list, &handle->h_transaction->t_private_list);
185 static inline void ext4_journal_callback_add(handle_t *handle,
186 void (*func)(struct super_block *sb,
187 struct ext4_journal_cb_entry *jce,
189 struct ext4_journal_cb_entry *jce)
191 struct ext4_sb_info *sbi =
192 EXT4_SB(handle->h_transaction->t_journal->j_private);
194 /* Add the jce to transaction's private list */
195 jce->jce_func = func;
196 spin_lock(&sbi->s_md_lock);
197 _ext4_journal_callback_add(handle, jce);
198 spin_unlock(&sbi->s_md_lock);
203 * ext4_journal_callback_del: delete a registered callback
204 * @handle: active journal transaction handle on which callback was registered
205 * @jce: registered journal callback entry to unregister
206 * Return true if object was successfully removed
208 static inline bool ext4_journal_callback_try_del(handle_t *handle,
209 struct ext4_journal_cb_entry *jce)
212 struct ext4_sb_info *sbi =
213 EXT4_SB(handle->h_transaction->t_journal->j_private);
215 spin_lock(&sbi->s_md_lock);
216 deleted = !list_empty(&jce->jce_list);
217 list_del_init(&jce->jce_list);
218 spin_unlock(&sbi->s_md_lock);
223 ext4_mark_iloc_dirty(handle_t *handle,
225 struct ext4_iloc *iloc);
228 * On success, We end up with an outstanding reference count against
229 * iloc->bh. This _must_ be cleaned up later.
232 int ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
233 struct ext4_iloc *iloc);
235 int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode);
238 * Wrapper functions with which ext4 calls into JBD.
240 int __ext4_journal_get_write_access(const char *where, unsigned int line,
241 handle_t *handle, struct buffer_head *bh);
243 int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
244 int is_metadata, struct inode *inode,
245 struct buffer_head *bh, ext4_fsblk_t blocknr);
247 int __ext4_journal_get_create_access(const char *where, unsigned int line,
248 handle_t *handle, struct buffer_head *bh);
250 int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
251 handle_t *handle, struct inode *inode,
252 struct buffer_head *bh);
254 int __ext4_handle_dirty_super(const char *where, unsigned int line,
255 handle_t *handle, struct super_block *sb);
257 #define ext4_journal_get_write_access(handle, bh) \
258 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
259 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
260 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
262 #define ext4_journal_get_create_access(handle, bh) \
263 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
264 #define ext4_handle_dirty_metadata(handle, inode, bh) \
265 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
267 #define ext4_handle_dirty_super(handle, sb) \
268 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
270 handle_t *__ext4_journal_start_sb(struct super_block *sb, unsigned int line,
271 int type, int blocks, int rsv_blocks);
272 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle);
274 #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
276 /* Note: Do not use this for NULL handles. This is only to determine if
277 * a properly allocated handle is using a journal or not. */
278 static inline int ext4_handle_valid(handle_t *handle)
280 if ((unsigned long)handle < EXT4_NOJOURNAL_MAX_REF_COUNT)
285 static inline void ext4_handle_sync(handle_t *handle)
287 if (ext4_handle_valid(handle))
291 static inline int ext4_handle_is_aborted(handle_t *handle)
293 if (ext4_handle_valid(handle))
294 return is_handle_aborted(handle);
298 static inline int ext4_handle_has_enough_credits(handle_t *handle, int needed)
300 if (ext4_handle_valid(handle) && handle->h_buffer_credits < needed)
305 #define ext4_journal_start_sb(sb, type, nblocks) \
306 __ext4_journal_start_sb((sb), __LINE__, (type), (nblocks), 0)
308 #define ext4_journal_start(inode, type, nblocks) \
309 __ext4_journal_start((inode), __LINE__, (type), (nblocks), 0)
311 #define ext4_journal_start_with_reserve(inode, type, blocks, rsv_blocks) \
312 __ext4_journal_start((inode), __LINE__, (type), (blocks), (rsv_blocks))
314 static inline handle_t *__ext4_journal_start(struct inode *inode,
315 unsigned int line, int type,
316 int blocks, int rsv_blocks)
318 return __ext4_journal_start_sb(inode->i_sb, line, type, blocks,
322 #define ext4_journal_stop(handle) \
323 __ext4_journal_stop(__func__, __LINE__, (handle))
325 #define ext4_journal_start_reserved(handle, type) \
326 __ext4_journal_start_reserved((handle), __LINE__, (type))
328 handle_t *__ext4_journal_start_reserved(handle_t *handle, unsigned int line,
331 static inline void ext4_journal_free_reserved(handle_t *handle)
333 if (ext4_handle_valid(handle))
334 jbd2_journal_free_reserved(handle);
337 static inline handle_t *ext4_journal_current_handle(void)
339 return journal_current_handle();
342 static inline int ext4_journal_extend(handle_t *handle, int nblocks)
344 if (ext4_handle_valid(handle))
345 return jbd2_journal_extend(handle, nblocks);
349 static inline int ext4_journal_restart(handle_t *handle, int nblocks)
351 if (ext4_handle_valid(handle))
352 return jbd2_journal_restart(handle, nblocks);
356 static inline int ext4_journal_blocks_per_page(struct inode *inode)
358 if (EXT4_JOURNAL(inode) != NULL)
359 return jbd2_journal_blocks_per_page(inode);
363 static inline int ext4_journal_force_commit(journal_t *journal)
366 return jbd2_journal_force_commit(journal);
370 static inline int ext4_jbd2_inode_add_write(handle_t *handle,
373 if (ext4_handle_valid(handle))
374 return jbd2_journal_inode_add_write(handle,
375 EXT4_I(inode)->jinode);
379 static inline int ext4_jbd2_inode_add_wait(handle_t *handle,
382 if (ext4_handle_valid(handle))
383 return jbd2_journal_inode_add_wait(handle,
384 EXT4_I(inode)->jinode);
388 static inline void ext4_update_inode_fsync_trans(handle_t *handle,
392 struct ext4_inode_info *ei = EXT4_I(inode);
394 if (ext4_handle_valid(handle) && !is_handle_aborted(handle)) {
395 ei->i_sync_tid = handle->h_transaction->t_tid;
397 ei->i_datasync_tid = handle->h_transaction->t_tid;
402 int ext4_force_commit(struct super_block *sb);
405 * Ext4 inode journal modes
407 #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
408 #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
409 #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
411 static inline int ext4_inode_journal_mode(struct inode *inode)
413 if (EXT4_JOURNAL(inode) == NULL)
414 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
415 /* We do not support data journalling with delayed allocation */
416 if (!S_ISREG(inode->i_mode) ||
417 test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ||
418 (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) &&
419 !test_opt(inode->i_sb, DELALLOC))) {
420 /* We do not support data journalling for encrypted data */
421 if (S_ISREG(inode->i_mode) && ext4_encrypted_inode(inode))
422 return EXT4_INODE_ORDERED_DATA_MODE; /* ordered */
423 return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */
425 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
426 return EXT4_INODE_ORDERED_DATA_MODE; /* ordered */
427 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
428 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
432 static inline int ext4_should_journal_data(struct inode *inode)
434 return ext4_inode_journal_mode(inode) & EXT4_INODE_JOURNAL_DATA_MODE;
437 static inline int ext4_should_order_data(struct inode *inode)
439 return ext4_inode_journal_mode(inode) & EXT4_INODE_ORDERED_DATA_MODE;
442 static inline int ext4_should_writeback_data(struct inode *inode)
444 return ext4_inode_journal_mode(inode) & EXT4_INODE_WRITEBACK_DATA_MODE;
448 * This function controls whether or not we should try to go down the
449 * dioread_nolock code paths, which makes it safe to avoid taking
450 * i_mutex for direct I/O reads. This only works for extent-based
451 * files, and it doesn't work if data journaling is enabled, since the
452 * dioread_nolock code uses b_private to pass information back to the
453 * I/O completion handler, and this conflicts with the jbd's use of
456 static inline int ext4_should_dioread_nolock(struct inode *inode)
458 if (!test_opt(inode->i_sb, DIOREAD_NOLOCK))
460 if (!S_ISREG(inode->i_mode))
462 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
464 if (ext4_should_journal_data(inode))
469 #endif /* _EXT4_JBD2_H */