Commit | Line | Data |
---|---|---|
470decc6 | 1 | /* |
3dcf5451 | 2 | * ext4_jbd2.h |
470decc6 DK |
3 | * |
4 | * Written by Stephen C. Tweedie <sct@redhat.com>, 1999 | |
5 | * | |
6 | * Copyright 1998--1999 Red Hat corp --- All Rights Reserved | |
7 | * | |
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. | |
11 | * | |
12 | * Ext4-specific journaling extensions. | |
13 | */ | |
14 | ||
3dcf5451 CH |
15 | #ifndef _EXT4_JBD2_H |
16 | #define _EXT4_JBD2_H | |
470decc6 DK |
17 | |
18 | #include <linux/fs.h> | |
f7f4bccb | 19 | #include <linux/jbd2.h> |
3dcf5451 | 20 | #include "ext4.h" |
470decc6 DK |
21 | |
22 | #define EXT4_JOURNAL(inode) (EXT4_SB((inode)->i_sb)->s_journal) | |
23 | ||
24 | /* Define the number of blocks we need to account to a transaction to | |
25 | * modify one block of data. | |
26 | * | |
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 | |
a86c6181 AT |
29 | * block to complete the transaction. |
30 | * | |
d0d856e8 RD |
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. */ | |
470decc6 | 33 | |
a86c6181 | 34 | #define EXT4_SINGLEDATA_TRANS_BLOCKS(sb) \ |
83982b6f TT |
35 | (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS) \ |
36 | ? 27U : 8U) | |
470decc6 DK |
37 | |
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. */ | |
41 | ||
42 | #define EXT4_XATTR_TRANS_BLOCKS 6U | |
43 | ||
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. */ | |
49 | ||
a86c6181 | 50 | #define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \ |
470decc6 | 51 | EXT4_XATTR_TRANS_BLOCKS - 2 + \ |
5aca07eb | 52 | EXT4_MAXQUOTAS_TRANS_BLOCKS(sb)) |
470decc6 | 53 | |
a02908f1 MC |
54 | /* |
55 | * Define the number of metadata blocks we need to account to modify data. | |
56 | * | |
57 | * This include super block, inode block, quota blocks and xattr blocks | |
58 | */ | |
59 | #define EXT4_META_TRANS_BLOCKS(sb) (EXT4_XATTR_TRANS_BLOCKS + \ | |
5aca07eb | 60 | EXT4_MAXQUOTAS_TRANS_BLOCKS(sb)) |
a02908f1 | 61 | |
470decc6 DK |
62 | /* Define an arbitrary limit for the amount of data we will anticipate |
63 | * writing to any given transaction. For unbounded transactions such as | |
64 | * write(2) and truncate(2) we can write more than this, but we always | |
65 | * start off at the maximum transaction size and grow the transaction | |
66 | * optimistically as we go. */ | |
67 | ||
68 | #define EXT4_MAX_TRANS_DATA 64U | |
69 | ||
70 | /* We break up a large truncate or write transaction once the handle's | |
71 | * buffer credits gets this low, we need either to extend the | |
72 | * transaction or to start a new one. Reserve enough space here for | |
73 | * inode, bitmap, superblock, group and indirection updates for at least | |
74 | * one block, plus two quota updates. Quota allocations are not | |
75 | * needed. */ | |
76 | ||
77 | #define EXT4_RESERVE_TRANS_BLOCKS 12U | |
78 | ||
79 | #define EXT4_INDEX_EXTRA_TRANS_BLOCKS 8 | |
80 | ||
81 | #ifdef CONFIG_QUOTA | |
82 | /* Amount of blocks needed for quota update - we know that the structure was | |
21f97697 | 83 | * allocated so we need to update only data block */ |
7c319d32 AK |
84 | #define EXT4_QUOTA_TRANS_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\ |
85 | EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\ | |
86 | 1 : 0) | |
470decc6 DK |
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 */ | |
7c319d32 AK |
89 | #define EXT4_QUOTA_INIT_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\ |
90 | EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\ | |
91 | (DQUOT_INIT_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\ | |
92 | +3+DQUOT_INIT_REWRITE) : 0) | |
93 | ||
94 | #define EXT4_QUOTA_DEL_BLOCKS(sb) ((test_opt(sb, QUOTA) ||\ | |
95 | EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) ?\ | |
96 | (DQUOT_DEL_ALLOC*(EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)\ | |
97 | +3+DQUOT_DEL_REWRITE) : 0) | |
470decc6 DK |
98 | #else |
99 | #define EXT4_QUOTA_TRANS_BLOCKS(sb) 0 | |
100 | #define EXT4_QUOTA_INIT_BLOCKS(sb) 0 | |
101 | #define EXT4_QUOTA_DEL_BLOCKS(sb) 0 | |
102 | #endif | |
5aca07eb DM |
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)) | |
470decc6 | 106 | |
95eaefbd TT |
107 | static inline int ext4_jbd2_credits_xattr(struct inode *inode) |
108 | { | |
109 | int credits = EXT4_DATA_TRANS_BLOCKS(inode->i_sb); | |
110 | ||
111 | /* | |
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 | |
114 | */ | |
115 | if (ext4_has_inline_data(inode)) | |
116 | credits += ext4_writepage_trans_blocks(inode) + 1; | |
117 | return credits; | |
118 | } | |
119 | ||
120 | ||
9924a92a TT |
121 | /* |
122 | * Ext4 handle operation types -- for logging purposes | |
123 | */ | |
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_MAX 11 | |
136 | ||
18aadd47 BJ |
137 | /** |
138 | * struct ext4_journal_cb_entry - Base structure for callback information. | |
139 | * | |
140 | * This struct is a 'seed' structure for a using with your own callback | |
141 | * structs. If you are using callbacks you must allocate one of these | |
142 | * or another struct of your own definition which has this struct | |
143 | * as it's first element and pass it to ext4_journal_callback_add(). | |
144 | */ | |
145 | struct ext4_journal_cb_entry { | |
146 | /* list information for other callbacks attached to the same handle */ | |
147 | struct list_head jce_list; | |
148 | ||
149 | /* Function to call with this callback structure */ | |
150 | void (*jce_func)(struct super_block *sb, | |
151 | struct ext4_journal_cb_entry *jce, int error); | |
152 | ||
153 | /* user data goes here */ | |
154 | }; | |
155 | ||
156 | /** | |
157 | * ext4_journal_callback_add: add a function to call after transaction commit | |
158 | * @handle: active journal transaction handle to register callback on | |
159 | * @func: callback function to call after the transaction has committed: | |
160 | * @sb: superblock of current filesystem for transaction | |
161 | * @jce: returned journal callback data | |
162 | * @rc: journal state at commit (0 = transaction committed properly) | |
163 | * @jce: journal callback data (internal and function private data struct) | |
164 | * | |
165 | * The registered function will be called in the context of the journal thread | |
166 | * after the transaction for which the handle was created has completed. | |
167 | * | |
168 | * No locks are held when the callback function is called, so it is safe to | |
169 | * call blocking functions from within the callback, but the callback should | |
170 | * not block or run for too long, or the filesystem will be blocked waiting for | |
171 | * the next transaction to commit. No journaling functions can be used, or | |
172 | * there is a risk of deadlock. | |
173 | * | |
174 | * There is no guaranteed calling order of multiple registered callbacks on | |
175 | * the same transaction. | |
176 | */ | |
177 | static inline void ext4_journal_callback_add(handle_t *handle, | |
178 | void (*func)(struct super_block *sb, | |
179 | struct ext4_journal_cb_entry *jce, | |
180 | int rc), | |
181 | struct ext4_journal_cb_entry *jce) | |
182 | { | |
183 | struct ext4_sb_info *sbi = | |
184 | EXT4_SB(handle->h_transaction->t_journal->j_private); | |
185 | ||
186 | /* Add the jce to transaction's private list */ | |
187 | jce->jce_func = func; | |
188 | spin_lock(&sbi->s_md_lock); | |
189 | list_add_tail(&jce->jce_list, &handle->h_transaction->t_private_list); | |
190 | spin_unlock(&sbi->s_md_lock); | |
191 | } | |
192 | ||
193 | /** | |
194 | * ext4_journal_callback_del: delete a registered callback | |
195 | * @handle: active journal transaction handle on which callback was registered | |
196 | * @jce: registered journal callback entry to unregister | |
197 | */ | |
198 | static inline void ext4_journal_callback_del(handle_t *handle, | |
199 | struct ext4_journal_cb_entry *jce) | |
200 | { | |
201 | struct ext4_sb_info *sbi = | |
202 | EXT4_SB(handle->h_transaction->t_journal->j_private); | |
203 | ||
204 | spin_lock(&sbi->s_md_lock); | |
205 | list_del_init(&jce->jce_list); | |
206 | spin_unlock(&sbi->s_md_lock); | |
207 | } | |
208 | ||
470decc6 DK |
209 | int |
210 | ext4_mark_iloc_dirty(handle_t *handle, | |
211 | struct inode *inode, | |
212 | struct ext4_iloc *iloc); | |
213 | ||
214 | /* | |
215 | * On success, We end up with an outstanding reference count against | |
216 | * iloc->bh. This _must_ be cleaned up later. | |
217 | */ | |
218 | ||
219 | int ext4_reserve_inode_write(handle_t *handle, struct inode *inode, | |
220 | struct ext4_iloc *iloc); | |
221 | ||
222 | int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode); | |
223 | ||
224 | /* | |
e4684b3f | 225 | * Wrapper functions with which ext4 calls into JBD. |
470decc6 | 226 | */ |
90c7201b TT |
227 | void ext4_journal_abort_handle(const char *caller, unsigned int line, |
228 | const char *err_fn, | |
8984d137 | 229 | struct buffer_head *bh, handle_t *handle, int err); |
470decc6 | 230 | |
90c7201b TT |
231 | int __ext4_journal_get_write_access(const char *where, unsigned int line, |
232 | handle_t *handle, struct buffer_head *bh); | |
470decc6 | 233 | |
90c7201b TT |
234 | int __ext4_forget(const char *where, unsigned int line, handle_t *handle, |
235 | int is_metadata, struct inode *inode, | |
236 | struct buffer_head *bh, ext4_fsblk_t blocknr); | |
d6797d14 | 237 | |
90c7201b | 238 | int __ext4_journal_get_create_access(const char *where, unsigned int line, |
8984d137 | 239 | handle_t *handle, struct buffer_head *bh); |
470decc6 | 240 | |
90c7201b TT |
241 | int __ext4_handle_dirty_metadata(const char *where, unsigned int line, |
242 | handle_t *handle, struct inode *inode, | |
243 | struct buffer_head *bh); | |
470decc6 | 244 | |
90c7201b | 245 | int __ext4_handle_dirty_super(const char *where, unsigned int line, |
b50924c2 | 246 | handle_t *handle, struct super_block *sb); |
a0375156 | 247 | |
470decc6 | 248 | #define ext4_journal_get_write_access(handle, bh) \ |
90c7201b | 249 | __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh)) |
d6797d14 | 250 | #define ext4_forget(handle, is_metadata, inode, bh, block_nr) \ |
90c7201b TT |
251 | __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \ |
252 | (bh), (block_nr)) | |
470decc6 | 253 | #define ext4_journal_get_create_access(handle, bh) \ |
90c7201b | 254 | __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh)) |
0390131b | 255 | #define ext4_handle_dirty_metadata(handle, inode, bh) \ |
90c7201b TT |
256 | __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \ |
257 | (bh)) | |
a0375156 | 258 | #define ext4_handle_dirty_super(handle, sb) \ |
b50924c2 | 259 | __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb)) |
470decc6 | 260 | |
9924a92a TT |
261 | handle_t *__ext4_journal_start_sb(struct super_block *sb, unsigned int line, |
262 | int type, int nblocks); | |
c398eda0 | 263 | int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle); |
470decc6 | 264 | |
d3d1faf6 | 265 | #define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096) |
0390131b | 266 | |
d3d1faf6 CW |
267 | /* Note: Do not use this for NULL handles. This is only to determine if |
268 | * a properly allocated handle is using a journal or not. */ | |
0390131b FM |
269 | static inline int ext4_handle_valid(handle_t *handle) |
270 | { | |
d3d1faf6 | 271 | if ((unsigned long)handle < EXT4_NOJOURNAL_MAX_REF_COUNT) |
0390131b FM |
272 | return 0; |
273 | return 1; | |
274 | } | |
275 | ||
276 | static inline void ext4_handle_sync(handle_t *handle) | |
277 | { | |
278 | if (ext4_handle_valid(handle)) | |
279 | handle->h_sync = 1; | |
280 | } | |
281 | ||
0390131b FM |
282 | static inline int ext4_handle_is_aborted(handle_t *handle) |
283 | { | |
284 | if (ext4_handle_valid(handle)) | |
285 | return is_handle_aborted(handle); | |
286 | return 0; | |
287 | } | |
288 | ||
289 | static inline int ext4_handle_has_enough_credits(handle_t *handle, int needed) | |
290 | { | |
291 | if (ext4_handle_valid(handle) && handle->h_buffer_credits < needed) | |
292 | return 0; | |
293 | return 1; | |
294 | } | |
295 | ||
9924a92a TT |
296 | #define ext4_journal_start_sb(sb, type, nblocks) \ |
297 | __ext4_journal_start_sb((sb), __LINE__, (type), (nblocks)) | |
298 | ||
299 | #define ext4_journal_start(inode, type, nblocks) \ | |
300 | __ext4_journal_start((inode), __LINE__, (type), (nblocks)) | |
301 | ||
302 | static inline handle_t *__ext4_journal_start(struct inode *inode, | |
303 | unsigned int line, int type, | |
304 | int nblocks) | |
470decc6 | 305 | { |
9924a92a | 306 | return __ext4_journal_start_sb(inode->i_sb, line, type, nblocks); |
470decc6 DK |
307 | } |
308 | ||
309 | #define ext4_journal_stop(handle) \ | |
c398eda0 | 310 | __ext4_journal_stop(__func__, __LINE__, (handle)) |
470decc6 DK |
311 | |
312 | static inline handle_t *ext4_journal_current_handle(void) | |
313 | { | |
314 | return journal_current_handle(); | |
315 | } | |
316 | ||
317 | static inline int ext4_journal_extend(handle_t *handle, int nblocks) | |
318 | { | |
0390131b FM |
319 | if (ext4_handle_valid(handle)) |
320 | return jbd2_journal_extend(handle, nblocks); | |
321 | return 0; | |
470decc6 DK |
322 | } |
323 | ||
324 | static inline int ext4_journal_restart(handle_t *handle, int nblocks) | |
325 | { | |
0390131b FM |
326 | if (ext4_handle_valid(handle)) |
327 | return jbd2_journal_restart(handle, nblocks); | |
328 | return 0; | |
470decc6 DK |
329 | } |
330 | ||
331 | static inline int ext4_journal_blocks_per_page(struct inode *inode) | |
332 | { | |
0390131b FM |
333 | if (EXT4_JOURNAL(inode) != NULL) |
334 | return jbd2_journal_blocks_per_page(inode); | |
335 | return 0; | |
470decc6 DK |
336 | } |
337 | ||
338 | static inline int ext4_journal_force_commit(journal_t *journal) | |
339 | { | |
0390131b FM |
340 | if (journal) |
341 | return jbd2_journal_force_commit(journal); | |
342 | return 0; | |
470decc6 DK |
343 | } |
344 | ||
678aaf48 JK |
345 | static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode) |
346 | { | |
0390131b | 347 | if (ext4_handle_valid(handle)) |
8aefcd55 | 348 | return jbd2_journal_file_inode(handle, EXT4_I(inode)->jinode); |
0390131b | 349 | return 0; |
678aaf48 JK |
350 | } |
351 | ||
b436b9be JK |
352 | static inline void ext4_update_inode_fsync_trans(handle_t *handle, |
353 | struct inode *inode, | |
354 | int datasync) | |
355 | { | |
356 | struct ext4_inode_info *ei = EXT4_I(inode); | |
357 | ||
358 | if (ext4_handle_valid(handle)) { | |
359 | ei->i_sync_tid = handle->h_transaction->t_tid; | |
360 | if (datasync) | |
361 | ei->i_datasync_tid = handle->h_transaction->t_tid; | |
362 | } | |
363 | } | |
364 | ||
470decc6 DK |
365 | /* super.c */ |
366 | int ext4_force_commit(struct super_block *sb); | |
367 | ||
3d2b1582 LC |
368 | /* |
369 | * Ext4 inode journal modes | |
370 | */ | |
371 | #define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */ | |
372 | #define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */ | |
373 | #define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */ | |
374 | ||
375 | static inline int ext4_inode_journal_mode(struct inode *inode) | |
470decc6 | 376 | { |
0390131b | 377 | if (EXT4_JOURNAL(inode) == NULL) |
3d2b1582 LC |
378 | return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */ |
379 | /* We do not support data journalling with delayed allocation */ | |
380 | if (!S_ISREG(inode->i_mode) || | |
381 | test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) | |
382 | return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */ | |
383 | if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) && | |
384 | !test_opt(inode->i_sb, DELALLOC)) | |
385 | return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */ | |
386 | if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) | |
387 | return EXT4_INODE_ORDERED_DATA_MODE; /* ordered */ | |
388 | if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) | |
389 | return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */ | |
390 | else | |
391 | BUG(); | |
392 | } | |
393 | ||
394 | static inline int ext4_should_journal_data(struct inode *inode) | |
395 | { | |
396 | return ext4_inode_journal_mode(inode) & EXT4_INODE_JOURNAL_DATA_MODE; | |
470decc6 DK |
397 | } |
398 | ||
399 | static inline int ext4_should_order_data(struct inode *inode) | |
400 | { | |
3d2b1582 | 401 | return ext4_inode_journal_mode(inode) & EXT4_INODE_ORDERED_DATA_MODE; |
470decc6 DK |
402 | } |
403 | ||
404 | static inline int ext4_should_writeback_data(struct inode *inode) | |
405 | { | |
3d2b1582 | 406 | return ext4_inode_journal_mode(inode) & EXT4_INODE_WRITEBACK_DATA_MODE; |
470decc6 DK |
407 | } |
408 | ||
744692dc JZ |
409 | /* |
410 | * This function controls whether or not we should try to go down the | |
411 | * dioread_nolock code paths, which makes it safe to avoid taking | |
412 | * i_mutex for direct I/O reads. This only works for extent-based | |
206f7ab4 CH |
413 | * files, and it doesn't work if data journaling is enabled, since the |
414 | * dioread_nolock code uses b_private to pass information back to the | |
415 | * I/O completion handler, and this conflicts with the jbd's use of | |
416 | * b_private. | |
744692dc JZ |
417 | */ |
418 | static inline int ext4_should_dioread_nolock(struct inode *inode) | |
419 | { | |
420 | if (!test_opt(inode->i_sb, DIOREAD_NOLOCK)) | |
421 | return 0; | |
744692dc JZ |
422 | if (!S_ISREG(inode->i_mode)) |
423 | return 0; | |
12e9b892 | 424 | if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) |
744692dc JZ |
425 | return 0; |
426 | if (ext4_should_journal_data(inode)) | |
427 | return 0; | |
428 | return 1; | |
429 | } | |
430 | ||
3dcf5451 | 431 | #endif /* _EXT4_JBD2_H */ |