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 + root which are stored in the inode. */
34 #define EXT4_SINGLEDATA_TRANS_BLOCKS(sb) \
35 (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS) \
38 /* Extended attribute operations touch at most two data buffers,
39 * two bitmap buffers, and two group summaries, in addition to the inode
40 * and the superblock, which are already accounted for. */
42 #define EXT4_XATTR_TRANS_BLOCKS 6U
44 /* Define the minimum size for a transaction which modifies data. This
45 * needs to take into account the fact that we may end up modifying two
46 * quota files too (one for the group, one for the user quota). The
47 * superblock only gets updated once, of course, so don't bother
48 * counting that again for the quota updates. */
50 #define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
51 EXT4_XATTR_TRANS_BLOCKS - 2 + \
52 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
55 * Define the number of metadata blocks we need to account to modify data.
57 * This include super block, inode block, quota blocks and xattr blocks
59 #define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \
60 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
62 /* Delete operations potentially hit one directory's namespace plus an
63 * entire inode, plus arbitrary amounts of bitmap/indirection data. Be
64 * generous. We can grow the delete transaction later if necessary. */
66 #define EXT4_DELETE_TRANS_BLOCKS(sb) (2 * EXT4_DATA_TRANS_BLOCKS(sb) + 64)
68 /* Define an arbitrary limit for the amount of data we will anticipate
69 * writing to any given transaction. For unbounded transactions such as
70 * write(2) and truncate(2) we can write more than this, but we always
71 * start off at the maximum transaction size and grow the transaction
72 * optimistically as we go. */
74 #define EXT4_MAX_TRANS_DATA 64U
76 /* We break up a large truncate or write transaction once the handle's
77 * buffer credits gets this low, we need either to extend the
78 * transaction or to start a new one. Reserve enough space here for
79 * inode, bitmap, superblock, group and indirection updates for at least
80 * one block, plus two quota updates. Quota allocations are not
83 #define EXT4_RESERVE_TRANS_BLOCKS 12U
85 #define EXT4_INDEX_EXTRA_TRANS_BLOCKS 8
88 /* Amount of blocks needed for quota update - we know that the structure was
89 * allocated so we need to update only data block */
90 #define EXT4_QUOTA_TRANS_BLOCKS(sb) (test_opt(sb, QUOTA) ? 1 : 0)
91 /* Amount of blocks needed for quota insert/delete - we do some block writes
92 * but inode, sb and group updates are done only once */
93 #define EXT4_QUOTA_INIT_BLOCKS(sb) (test_opt(sb, QUOTA) ? (DQUOT_INIT_ALLOC*\
94 (EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)+3+DQUOT_INIT_REWRITE) : 0)
96 #define EXT4_QUOTA_DEL_BLOCKS(sb) (test_opt(sb, QUOTA) ? (DQUOT_DEL_ALLOC*\
97 (EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)+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) (MAXQUOTAS*EXT4_QUOTA_TRANS_BLOCKS(sb))
104 #define EXT4_MAXQUOTAS_INIT_BLOCKS(sb) (MAXQUOTAS*EXT4_QUOTA_INIT_BLOCKS(sb))
105 #define EXT4_MAXQUOTAS_DEL_BLOCKS(sb) (MAXQUOTAS*EXT4_QUOTA_DEL_BLOCKS(sb))
108 * struct ext4_journal_cb_entry - Base structure for callback information.
110 * This struct is a 'seed' structure for a using with your own callback
111 * structs. If you are using callbacks you must allocate one of these
112 * or another struct of your own definition which has this struct
113 * as it's first element and pass it to ext4_journal_callback_add().
115 struct ext4_journal_cb_entry {
116 /* list information for other callbacks attached to the same handle */
117 struct list_head jce_list;
119 /* Function to call with this callback structure */
120 void (*jce_func)(struct super_block *sb,
121 struct ext4_journal_cb_entry *jce, int error);
123 /* user data goes here */
127 * ext4_journal_callback_add: add a function to call after transaction commit
128 * @handle: active journal transaction handle to register callback on
129 * @func: callback function to call after the transaction has committed:
130 * @sb: superblock of current filesystem for transaction
131 * @jce: returned journal callback data
132 * @rc: journal state at commit (0 = transaction committed properly)
133 * @jce: journal callback data (internal and function private data struct)
135 * The registered function will be called in the context of the journal thread
136 * after the transaction for which the handle was created has completed.
138 * No locks are held when the callback function is called, so it is safe to
139 * call blocking functions from within the callback, but the callback should
140 * not block or run for too long, or the filesystem will be blocked waiting for
141 * the next transaction to commit. No journaling functions can be used, or
142 * there is a risk of deadlock.
144 * There is no guaranteed calling order of multiple registered callbacks on
145 * the same transaction.
147 static inline void ext4_journal_callback_add(handle_t *handle,
148 void (*func)(struct super_block *sb,
149 struct ext4_journal_cb_entry *jce,
151 struct ext4_journal_cb_entry *jce)
153 struct ext4_sb_info *sbi =
154 EXT4_SB(handle->h_transaction->t_journal->j_private);
156 /* Add the jce to transaction's private list */
157 jce->jce_func = func;
158 spin_lock(&sbi->s_md_lock);
159 list_add_tail(&jce->jce_list, &handle->h_transaction->t_private_list);
160 spin_unlock(&sbi->s_md_lock);
164 * ext4_journal_callback_del: delete a registered callback
165 * @handle: active journal transaction handle on which callback was registered
166 * @jce: registered journal callback entry to unregister
168 static inline void ext4_journal_callback_del(handle_t *handle,
169 struct ext4_journal_cb_entry *jce)
171 struct ext4_sb_info *sbi =
172 EXT4_SB(handle->h_transaction->t_journal->j_private);
174 spin_lock(&sbi->s_md_lock);
175 list_del_init(&jce->jce_list);
176 spin_unlock(&sbi->s_md_lock);
180 ext4_mark_iloc_dirty(handle_t *handle,
182 struct ext4_iloc *iloc);
185 * On success, We end up with an outstanding reference count against
186 * iloc->bh. This _must_ be cleaned up later.
189 int ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
190 struct ext4_iloc *iloc);
192 int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode);
195 * Wrapper functions with which ext4 calls into JBD.
197 void ext4_journal_abort_handle(const char *caller, unsigned int line,
199 struct buffer_head *bh, handle_t *handle, int err);
201 int __ext4_journal_get_write_access(const char *where, unsigned int line,
202 handle_t *handle, struct buffer_head *bh);
204 int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
205 int is_metadata, struct inode *inode,
206 struct buffer_head *bh, ext4_fsblk_t blocknr);
208 int __ext4_journal_get_create_access(const char *where, unsigned int line,
209 handle_t *handle, struct buffer_head *bh);
211 int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
212 handle_t *handle, struct inode *inode,
213 struct buffer_head *bh);
215 int __ext4_handle_dirty_super(const char *where, unsigned int line,
216 handle_t *handle, struct super_block *sb,
219 #define ext4_journal_get_write_access(handle, bh) \
220 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
221 #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
222 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
224 #define ext4_journal_get_create_access(handle, bh) \
225 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
226 #define ext4_handle_dirty_metadata(handle, inode, bh) \
227 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
229 #define ext4_handle_dirty_super_now(handle, sb) \
230 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb), 1)
231 #define ext4_handle_dirty_super(handle, sb) \
232 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb), 0)
234 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks);
235 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle);
237 #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
239 /* Note: Do not use this for NULL handles. This is only to determine if
240 * a properly allocated handle is using a journal or not. */
241 static inline int ext4_handle_valid(handle_t *handle)
243 if ((unsigned long)handle < EXT4_NOJOURNAL_MAX_REF_COUNT)
248 static inline void ext4_handle_sync(handle_t *handle)
250 if (ext4_handle_valid(handle))
254 static inline void ext4_handle_release_buffer(handle_t *handle,
255 struct buffer_head *bh)
257 if (ext4_handle_valid(handle))
258 jbd2_journal_release_buffer(handle, bh);
261 static inline int ext4_handle_is_aborted(handle_t *handle)
263 if (ext4_handle_valid(handle))
264 return is_handle_aborted(handle);
268 static inline int ext4_handle_has_enough_credits(handle_t *handle, int needed)
270 if (ext4_handle_valid(handle) && handle->h_buffer_credits < needed)
275 static inline handle_t *ext4_journal_start(struct inode *inode, int nblocks)
277 return ext4_journal_start_sb(inode->i_sb, nblocks);
280 #define ext4_journal_stop(handle) \
281 __ext4_journal_stop(__func__, __LINE__, (handle))
283 static inline handle_t *ext4_journal_current_handle(void)
285 return journal_current_handle();
288 static inline int ext4_journal_extend(handle_t *handle, int nblocks)
290 if (ext4_handle_valid(handle))
291 return jbd2_journal_extend(handle, nblocks);
295 static inline int ext4_journal_restart(handle_t *handle, int nblocks)
297 if (ext4_handle_valid(handle))
298 return jbd2_journal_restart(handle, nblocks);
302 static inline int ext4_journal_blocks_per_page(struct inode *inode)
304 if (EXT4_JOURNAL(inode) != NULL)
305 return jbd2_journal_blocks_per_page(inode);
309 static inline int ext4_journal_force_commit(journal_t *journal)
312 return jbd2_journal_force_commit(journal);
316 static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode)
318 if (ext4_handle_valid(handle))
319 return jbd2_journal_file_inode(handle, EXT4_I(inode)->jinode);
323 static inline void ext4_update_inode_fsync_trans(handle_t *handle,
327 struct ext4_inode_info *ei = EXT4_I(inode);
329 if (ext4_handle_valid(handle)) {
330 ei->i_sync_tid = handle->h_transaction->t_tid;
332 ei->i_datasync_tid = handle->h_transaction->t_tid;
337 int ext4_force_commit(struct super_block *sb);
340 * Ext4 inode journal modes
342 #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
343 #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
344 #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
346 static inline int ext4_inode_journal_mode(struct inode *inode)
348 if (EXT4_JOURNAL(inode) == NULL)
349 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
350 /* We do not support data journalling with delayed allocation */
351 if (!S_ISREG(inode->i_mode) ||
352 test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
353 return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */
354 if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) &&
355 !test_opt(inode->i_sb, DELALLOC))
356 return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */
357 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
358 return EXT4_INODE_ORDERED_DATA_MODE; /* ordered */
359 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
360 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
365 static inline int ext4_should_journal_data(struct inode *inode)
367 return ext4_inode_journal_mode(inode) & EXT4_INODE_JOURNAL_DATA_MODE;
370 static inline int ext4_should_order_data(struct inode *inode)
372 return ext4_inode_journal_mode(inode) & EXT4_INODE_ORDERED_DATA_MODE;
375 static inline int ext4_should_writeback_data(struct inode *inode)
377 return ext4_inode_journal_mode(inode) & EXT4_INODE_WRITEBACK_DATA_MODE;
381 * This function controls whether or not we should try to go down the
382 * dioread_nolock code paths, which makes it safe to avoid taking
383 * i_mutex for direct I/O reads. This only works for extent-based
384 * files, and it doesn't work if data journaling is enabled, since the
385 * dioread_nolock code uses b_private to pass information back to the
386 * I/O completion handler, and this conflicts with the jbd's use of
389 static inline int ext4_should_dioread_nolock(struct inode *inode)
391 if (!test_opt(inode->i_sb, DIOREAD_NOLOCK))
393 if (!S_ISREG(inode->i_mode))
395 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
397 if (ext4_should_journal_data(inode))
402 #endif /* _EXT4_JBD2_H */