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) (MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
106 #define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
107 #define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (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 sucessfully 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 void ext4_journal_abort_handle(const char *caller, unsigned int line,
236 struct buffer_head *bh, handle_t *handle, int err);
238 int __ext4_journal_get_write_access(const char *where, unsigned int line,
239 handle_t *handle, struct buffer_head *bh);
241 int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
242 int is_metadata, struct inode *inode,
243 struct buffer_head *bh, ext4_fsblk_t blocknr);
245 int __ext4_journal_get_create_access(const char *where, unsigned int line,
246 handle_t *handle, struct buffer_head *bh);
248 int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
249 handle_t *handle, struct inode *inode,
250 struct buffer_head *bh);
252 int __ext4_handle_dirty_super(const char *where, unsigned int line,
253 handle_t *handle, struct super_block *sb);
255 #define ext4_journal_get_write_access(handle, bh) \
256 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
257 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
258 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
260 #define ext4_journal_get_create_access(handle, bh) \
261 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
262 #define ext4_handle_dirty_metadata(handle, inode, bh) \
263 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
265 #define ext4_handle_dirty_super(handle, sb) \
266 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
268 handle_t *__ext4_journal_start_sb(struct super_block *sb, unsigned int line,
269 int type, int blocks, int rsv_blocks);
270 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle);
272 #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
274 /* Note: Do not use this for NULL handles. This is only to determine if
275 * a properly allocated handle is using a journal or not. */
276 static inline int ext4_handle_valid(handle_t *handle)
278 if ((unsigned long)handle < EXT4_NOJOURNAL_MAX_REF_COUNT)
283 static inline void ext4_handle_sync(handle_t *handle)
285 if (ext4_handle_valid(handle))
289 static inline int ext4_handle_is_aborted(handle_t *handle)
291 if (ext4_handle_valid(handle))
292 return is_handle_aborted(handle);
296 static inline int ext4_handle_has_enough_credits(handle_t *handle, int needed)
298 if (ext4_handle_valid(handle) && handle->h_buffer_credits < needed)
303 #define ext4_journal_start_sb(sb, type, nblocks) \
304 __ext4_journal_start_sb((sb), __LINE__, (type), (nblocks), 0)
306 #define ext4_journal_start(inode, type, nblocks) \
307 __ext4_journal_start((inode), __LINE__, (type), (nblocks), 0)
309 #define ext4_journal_start_with_reserve(inode, type, blocks, rsv_blocks) \
310 __ext4_journal_start((inode), __LINE__, (type), (blocks), (rsv_blocks))
312 static inline handle_t *__ext4_journal_start(struct inode *inode,
313 unsigned int line, int type,
314 int blocks, int rsv_blocks)
316 return __ext4_journal_start_sb(inode->i_sb, line, type, blocks,
320 #define ext4_journal_stop(handle) \
321 __ext4_journal_stop(__func__, __LINE__, (handle))
323 #define ext4_journal_start_reserved(handle, type) \
324 __ext4_journal_start_reserved((handle), __LINE__, (type))
326 handle_t *__ext4_journal_start_reserved(handle_t *handle, unsigned int line,
329 static inline void ext4_journal_free_reserved(handle_t *handle)
331 if (ext4_handle_valid(handle))
332 jbd2_journal_free_reserved(handle);
335 static inline handle_t *ext4_journal_current_handle(void)
337 return journal_current_handle();
340 static inline int ext4_journal_extend(handle_t *handle, int nblocks)
342 if (ext4_handle_valid(handle))
343 return jbd2_journal_extend(handle, nblocks);
347 static inline int ext4_journal_restart(handle_t *handle, int nblocks)
349 if (ext4_handle_valid(handle))
350 return jbd2_journal_restart(handle, nblocks);
354 static inline int ext4_journal_blocks_per_page(struct inode *inode)
356 if (EXT4_JOURNAL(inode) != NULL)
357 return jbd2_journal_blocks_per_page(inode);
361 static inline int ext4_journal_force_commit(journal_t *journal)
364 return jbd2_journal_force_commit(journal);
368 static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode)
370 if (ext4_handle_valid(handle))
371 return jbd2_journal_file_inode(handle, EXT4_I(inode)->jinode);
375 static inline void ext4_update_inode_fsync_trans(handle_t *handle,
379 struct ext4_inode_info *ei = EXT4_I(inode);
381 if (ext4_handle_valid(handle)) {
382 ei->i_sync_tid = handle->h_transaction->t_tid;
384 ei->i_datasync_tid = handle->h_transaction->t_tid;
389 int ext4_force_commit(struct super_block *sb);
392 * Ext4 inode journal modes
394 #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
395 #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
396 #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
398 static inline int ext4_inode_journal_mode(struct inode *inode)
400 if (EXT4_JOURNAL(inode) == NULL)
401 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
402 /* We do not support data journalling with delayed allocation */
403 if (!S_ISREG(inode->i_mode) ||
404 test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
405 return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */
406 if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) &&
407 !test_opt(inode->i_sb, DELALLOC))
408 return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */
409 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
410 return EXT4_INODE_ORDERED_DATA_MODE; /* ordered */
411 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
412 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
417 static inline int ext4_should_journal_data(struct inode *inode)
419 return ext4_inode_journal_mode(inode) & EXT4_INODE_JOURNAL_DATA_MODE;
422 static inline int ext4_should_order_data(struct inode *inode)
424 return ext4_inode_journal_mode(inode) & EXT4_INODE_ORDERED_DATA_MODE;
427 static inline int ext4_should_writeback_data(struct inode *inode)
429 return ext4_inode_journal_mode(inode) & EXT4_INODE_WRITEBACK_DATA_MODE;
433 * This function controls whether or not we should try to go down the
434 * dioread_nolock code paths, which makes it safe to avoid taking
435 * i_mutex for direct I/O reads. This only works for extent-based
436 * files, and it doesn't work if data journaling is enabled, since the
437 * dioread_nolock code uses b_private to pass information back to the
438 * I/O completion handler, and this conflicts with the jbd's use of
441 static inline int ext4_should_dioread_nolock(struct inode *inode)
443 if (!test_opt(inode->i_sb, DIOREAD_NOLOCK))
445 if (!S_ISREG(inode->i_mode))
447 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
449 if (ext4_should_journal_data(inode))
454 #endif /* _EXT4_JBD2_H */