blob: 83b20fcf9400b11b28185470f8feef309faa8252 [file] [log] [blame]
Dave Kleikamp470decc2006-10-11 01:20:57 -07001/*
Christoph Hellwig3dcf5452008-04-29 18:13:32 -04002 * ext4_jbd2.h
Dave Kleikamp470decc2006-10-11 01:20:57 -07003 *
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
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040015#ifndef _EXT4_JBD2_H
16#define _EXT4_JBD2_H
Dave Kleikamp470decc2006-10-11 01:20:57 -070017
18#include <linux/fs.h>
Mingming Caof7f4bcc2006-10-11 01:20:59 -070019#include <linux/jbd2.h>
Christoph Hellwig3dcf5452008-04-29 18:13:32 -040020#include "ext4.h"
Dave Kleikamp470decc2006-10-11 01:20:57 -070021
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
Alex Tomasa86c6182006-10-11 01:21:03 -070029 * block to complete the transaction.
30 *
Randy Dunlapd0d856e2006-10-11 01:21:07 -070031 * 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. */
Dave Kleikamp470decc2006-10-11 01:20:57 -070033
Alex Tomasa86c6182006-10-11 01:21:03 -070034#define EXT4_SINGLEDATA_TRANS_BLOCKS(sb) \
Theodore Ts'o83982b62009-01-06 14:53:16 -050035 (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS) \
36 ? 27U : 8U)
Dave Kleikamp470decc2006-10-11 01:20:57 -070037
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
Alex Tomasa86c6182006-10-11 01:21:03 -070050#define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
Dave Kleikamp470decc2006-10-11 01:20:57 -070051 EXT4_XATTR_TRANS_BLOCKS - 2 + \
Dmitry Monakhov5aca07e2009-12-08 22:42:15 -050052 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
Dave Kleikamp470decc2006-10-11 01:20:57 -070053
Mingming Caoa02908f2008-08-19 22:16:07 -040054/*
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 + \
Dmitry Monakhov5aca07e2009-12-08 22:42:15 -050060 EXT4_MAXQUOTAS_TRANS_BLOCKS(sb))
Mingming Caoa02908f2008-08-19 22:16:07 -040061
Dave Kleikamp470decc2006-10-11 01:20:57 -070062/* 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. */
65
66#define EXT4_DELETE_TRANS_BLOCKS(sb) (2 * EXT4_DATA_TRANS_BLOCKS(sb) + 64)
67
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. */
73
74#define EXT4_MAX_TRANS_DATA 64U
75
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
81 * needed. */
82
83#define EXT4_RESERVE_TRANS_BLOCKS 12U
84
85#define EXT4_INDEX_EXTRA_TRANS_BLOCKS 8
86
87#ifdef CONFIG_QUOTA
88/* Amount of blocks needed for quota update - we know that the structure was
Jan Kara21f97692011-04-04 15:33:39 -040089 * allocated so we need to update only data block */
90#define EXT4_QUOTA_TRANS_BLOCKS(sb) (test_opt(sb, QUOTA) ? 1 : 0)
Dave Kleikamp470decc2006-10-11 01:20:57 -070091/* 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*\
Alex Tomasa86c6182006-10-11 01:21:03 -070094 (EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)+3+DQUOT_INIT_REWRITE) : 0)
Dmitry Monakhov5aca07e2009-12-08 22:42:15 -050095
Dave Kleikamp470decc2006-10-11 01:20:57 -070096#define EXT4_QUOTA_DEL_BLOCKS(sb) (test_opt(sb, QUOTA) ? (DQUOT_DEL_ALLOC*\
Alex Tomasa86c6182006-10-11 01:21:03 -070097 (EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)+3+DQUOT_DEL_REWRITE) : 0)
Dave Kleikamp470decc2006-10-11 01:20:57 -070098#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
Dmitry Monakhov5aca07e2009-12-08 22:42:15 -0500103#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))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700106
Bobi Jam18aadd42012-02-20 17:53:02 -0500107/**
108 * struct ext4_journal_cb_entry - Base structure for callback information.
109 *
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().
114 */
115struct ext4_journal_cb_entry {
116 /* list information for other callbacks attached to the same handle */
117 struct list_head jce_list;
118
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);
122
123 /* user data goes here */
124};
125
126/**
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)
134 *
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.
137 *
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.
143 *
144 * There is no guaranteed calling order of multiple registered callbacks on
145 * the same transaction.
146 */
147static inline void ext4_journal_callback_add(handle_t *handle,
148 void (*func)(struct super_block *sb,
149 struct ext4_journal_cb_entry *jce,
150 int rc),
151 struct ext4_journal_cb_entry *jce)
152{
153 struct ext4_sb_info *sbi =
154 EXT4_SB(handle->h_transaction->t_journal->j_private);
155
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);
161}
162
163/**
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
167 */
168static inline void ext4_journal_callback_del(handle_t *handle,
169 struct ext4_journal_cb_entry *jce)
170{
171 struct ext4_sb_info *sbi =
172 EXT4_SB(handle->h_transaction->t_journal->j_private);
173
174 spin_lock(&sbi->s_md_lock);
175 list_del_init(&jce->jce_list);
176 spin_unlock(&sbi->s_md_lock);
177}
178
Dave Kleikamp470decc2006-10-11 01:20:57 -0700179int
180ext4_mark_iloc_dirty(handle_t *handle,
181 struct inode *inode,
182 struct ext4_iloc *iloc);
183
184/*
185 * On success, We end up with an outstanding reference count against
186 * iloc->bh. This _must_ be cleaned up later.
187 */
188
189int ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
190 struct ext4_iloc *iloc);
191
192int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode);
193
194/*
Theodore Ts'oe4684b32009-11-24 11:05:59 -0500195 * Wrapper functions with which ext4 calls into JBD.
Dave Kleikamp470decc2006-10-11 01:20:57 -0700196 */
Theodore Ts'o90c72012010-06-29 14:53:24 -0400197void ext4_journal_abort_handle(const char *caller, unsigned int line,
198 const char *err_fn,
Andrew Morton8984d132006-12-06 20:37:15 -0800199 struct buffer_head *bh, handle_t *handle, int err);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700200
Theodore Ts'o90c72012010-06-29 14:53:24 -0400201int __ext4_journal_get_write_access(const char *where, unsigned int line,
202 handle_t *handle, struct buffer_head *bh);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700203
Theodore Ts'o90c72012010-06-29 14:53:24 -0400204int __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);
Theodore Ts'od6797d12009-11-22 20:52:12 -0500207
Theodore Ts'o90c72012010-06-29 14:53:24 -0400208int __ext4_journal_get_create_access(const char *where, unsigned int line,
Andrew Morton8984d132006-12-06 20:37:15 -0800209 handle_t *handle, struct buffer_head *bh);
210
Theodore Ts'o90c72012010-06-29 14:53:24 -0400211int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
212 handle_t *handle, struct inode *inode,
213 struct buffer_head *bh);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700214
Theodore Ts'o90c72012010-06-29 14:53:24 -0400215int __ext4_handle_dirty_super(const char *where, unsigned int line,
216 handle_t *handle, struct super_block *sb);
Theodore Ts'oa0375152010-06-11 23:14:04 -0400217
Dave Kleikamp470decc2006-10-11 01:20:57 -0700218#define ext4_journal_get_write_access(handle, bh) \
Theodore Ts'o90c72012010-06-29 14:53:24 -0400219 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
Theodore Ts'od6797d12009-11-22 20:52:12 -0500220#define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
Theodore Ts'o90c72012010-06-29 14:53:24 -0400221 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
222 (bh), (block_nr))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700223#define ext4_journal_get_create_access(handle, bh) \
Theodore Ts'o90c72012010-06-29 14:53:24 -0400224 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
Frank Mayhar03901312009-01-07 00:06:22 -0500225#define ext4_handle_dirty_metadata(handle, inode, bh) \
Theodore Ts'o90c72012010-06-29 14:53:24 -0400226 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
227 (bh))
Theodore Ts'oa0375152010-06-11 23:14:04 -0400228#define ext4_handle_dirty_super(handle, sb) \
Theodore Ts'o90c72012010-06-29 14:53:24 -0400229 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700230
Dave Kleikamp470decc2006-10-11 01:20:57 -0700231handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks);
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400232int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle);
Dave Kleikamp470decc2006-10-11 01:20:57 -0700233
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400234#define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
Frank Mayhar03901312009-01-07 00:06:22 -0500235
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400236/* Note: Do not use this for NULL handles. This is only to determine if
237 * a properly allocated handle is using a journal or not. */
Frank Mayhar03901312009-01-07 00:06:22 -0500238static inline int ext4_handle_valid(handle_t *handle)
239{
Curt Wohlgemuthd3d1faf2009-09-29 11:01:03 -0400240 if ((unsigned long)handle < EXT4_NOJOURNAL_MAX_REF_COUNT)
Frank Mayhar03901312009-01-07 00:06:22 -0500241 return 0;
242 return 1;
243}
244
245static inline void ext4_handle_sync(handle_t *handle)
246{
247 if (ext4_handle_valid(handle))
248 handle->h_sync = 1;
249}
250
251static inline void ext4_handle_release_buffer(handle_t *handle,
252 struct buffer_head *bh)
253{
254 if (ext4_handle_valid(handle))
255 jbd2_journal_release_buffer(handle, bh);
256}
257
258static inline int ext4_handle_is_aborted(handle_t *handle)
259{
260 if (ext4_handle_valid(handle))
261 return is_handle_aborted(handle);
262 return 0;
263}
264
265static inline int ext4_handle_has_enough_credits(handle_t *handle, int needed)
266{
267 if (ext4_handle_valid(handle) && handle->h_buffer_credits < needed)
268 return 0;
269 return 1;
270}
271
Dave Kleikamp470decc2006-10-11 01:20:57 -0700272static inline handle_t *ext4_journal_start(struct inode *inode, int nblocks)
273{
274 return ext4_journal_start_sb(inode->i_sb, nblocks);
275}
276
277#define ext4_journal_stop(handle) \
Theodore Ts'oc398eda2010-07-27 11:56:40 -0400278 __ext4_journal_stop(__func__, __LINE__, (handle))
Dave Kleikamp470decc2006-10-11 01:20:57 -0700279
280static inline handle_t *ext4_journal_current_handle(void)
281{
282 return journal_current_handle();
283}
284
285static inline int ext4_journal_extend(handle_t *handle, int nblocks)
286{
Frank Mayhar03901312009-01-07 00:06:22 -0500287 if (ext4_handle_valid(handle))
288 return jbd2_journal_extend(handle, nblocks);
289 return 0;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700290}
291
292static inline int ext4_journal_restart(handle_t *handle, int nblocks)
293{
Frank Mayhar03901312009-01-07 00:06:22 -0500294 if (ext4_handle_valid(handle))
295 return jbd2_journal_restart(handle, nblocks);
296 return 0;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700297}
298
299static inline int ext4_journal_blocks_per_page(struct inode *inode)
300{
Frank Mayhar03901312009-01-07 00:06:22 -0500301 if (EXT4_JOURNAL(inode) != NULL)
302 return jbd2_journal_blocks_per_page(inode);
303 return 0;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700304}
305
306static inline int ext4_journal_force_commit(journal_t *journal)
307{
Frank Mayhar03901312009-01-07 00:06:22 -0500308 if (journal)
309 return jbd2_journal_force_commit(journal);
310 return 0;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700311}
312
Jan Kara678aaf42008-07-11 19:27:31 -0400313static inline int ext4_jbd2_file_inode(handle_t *handle, struct inode *inode)
314{
Frank Mayhar03901312009-01-07 00:06:22 -0500315 if (ext4_handle_valid(handle))
Theodore Ts'o8aefcd52011-01-10 12:29:43 -0500316 return jbd2_journal_file_inode(handle, EXT4_I(inode)->jinode);
Frank Mayhar03901312009-01-07 00:06:22 -0500317 return 0;
Jan Kara678aaf42008-07-11 19:27:31 -0400318}
319
Jan Karab436b9b2009-12-08 23:51:10 -0500320static inline void ext4_update_inode_fsync_trans(handle_t *handle,
321 struct inode *inode,
322 int datasync)
323{
324 struct ext4_inode_info *ei = EXT4_I(inode);
325
326 if (ext4_handle_valid(handle)) {
327 ei->i_sync_tid = handle->h_transaction->t_tid;
328 if (datasync)
329 ei->i_datasync_tid = handle->h_transaction->t_tid;
330 }
331}
332
Dave Kleikamp470decc2006-10-11 01:20:57 -0700333/* super.c */
334int ext4_force_commit(struct super_block *sb);
335
Lukas Czerner3d2b1582012-02-20 17:53:00 -0500336/*
337 * Ext4 inode journal modes
338 */
339#define EXT4_INODE_JOURNAL_DATA_MODE 0x01 /* journal data mode */
340#define EXT4_INODE_ORDERED_DATA_MODE 0x02 /* ordered data mode */
341#define EXT4_INODE_WRITEBACK_DATA_MODE 0x04 /* writeback data mode */
342
343static inline int ext4_inode_journal_mode(struct inode *inode)
Dave Kleikamp470decc2006-10-11 01:20:57 -0700344{
Frank Mayhar03901312009-01-07 00:06:22 -0500345 if (EXT4_JOURNAL(inode) == NULL)
Lukas Czerner3d2b1582012-02-20 17:53:00 -0500346 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
347 /* We do not support data journalling with delayed allocation */
348 if (!S_ISREG(inode->i_mode) ||
349 test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
350 return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */
351 if (ext4_test_inode_flag(inode, EXT4_INODE_JOURNAL_DATA) &&
352 !test_opt(inode->i_sb, DELALLOC))
353 return EXT4_INODE_JOURNAL_DATA_MODE; /* journal data */
354 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
355 return EXT4_INODE_ORDERED_DATA_MODE; /* ordered */
356 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
357 return EXT4_INODE_WRITEBACK_DATA_MODE; /* writeback */
358 else
359 BUG();
360}
361
362static inline int ext4_should_journal_data(struct inode *inode)
363{
364 return ext4_inode_journal_mode(inode) & EXT4_INODE_JOURNAL_DATA_MODE;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700365}
366
367static inline int ext4_should_order_data(struct inode *inode)
368{
Lukas Czerner3d2b1582012-02-20 17:53:00 -0500369 return ext4_inode_journal_mode(inode) & EXT4_INODE_ORDERED_DATA_MODE;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700370}
371
372static inline int ext4_should_writeback_data(struct inode *inode)
373{
Lukas Czerner3d2b1582012-02-20 17:53:00 -0500374 return ext4_inode_journal_mode(inode) & EXT4_INODE_WRITEBACK_DATA_MODE;
Dave Kleikamp470decc2006-10-11 01:20:57 -0700375}
376
Jiaying Zhang744692d2010-03-04 16:14:02 -0500377/*
378 * This function controls whether or not we should try to go down the
379 * dioread_nolock code paths, which makes it safe to avoid taking
380 * i_mutex for direct I/O reads. This only works for extent-based
Christoph Hellwig206f7ab2010-06-14 14:42:49 -0400381 * files, and it doesn't work if data journaling is enabled, since the
382 * dioread_nolock code uses b_private to pass information back to the
383 * I/O completion handler, and this conflicts with the jbd's use of
384 * b_private.
Jiaying Zhang744692d2010-03-04 16:14:02 -0500385 */
386static inline int ext4_should_dioread_nolock(struct inode *inode)
387{
388 if (!test_opt(inode->i_sb, DIOREAD_NOLOCK))
389 return 0;
Jiaying Zhang744692d2010-03-04 16:14:02 -0500390 if (!S_ISREG(inode->i_mode))
391 return 0;
Dmitry Monakhov12e9b892010-05-16 22:00:00 -0400392 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
Jiaying Zhang744692d2010-03-04 16:14:02 -0500393 return 0;
394 if (ext4_should_journal_data(inode))
395 return 0;
396 return 1;
397}
398
Christoph Hellwig3dcf5452008-04-29 18:13:32 -0400399#endif /* _EXT4_JBD2_H */