treewide: Replace GPLv2 boilerplate/reference with SPDX - rule 145
[linux-block.git] / fs / ocfs2 / journal.h
CommitLineData
328970de 1/* SPDX-License-Identifier: GPL-2.0-or-later */
ccd979bd
MF
2/* -*- mode: c; c-basic-offset: 8; -*-
3 * vim: noexpandtab sw=8 ts=8 sts=0:
4 *
5 * journal.h
6 *
7 * Defines journalling api and structures.
8 *
9 * Copyright (C) 2003, 2005 Oracle. All rights reserved.
ccd979bd
MF
10 */
11
12#ifndef OCFS2_JOURNAL_H
13#define OCFS2_JOURNAL_H
14
15#include <linux/fs.h>
53ef99ca 16#include <linux/jbd2.h>
ccd979bd 17
ccd979bd
MF
18enum ocfs2_journal_state {
19 OCFS2_JOURNAL_FREE = 0,
20 OCFS2_JOURNAL_LOADED,
21 OCFS2_JOURNAL_IN_SHUTDOWN,
22};
23
24struct ocfs2_super;
25struct ocfs2_dinode;
ccd979bd 26
96a6c64b
SM
27/*
28 * The recovery_list is a simple linked list of node numbers to recover.
29 * It is protected by the recovery_lock.
30 */
31
32struct ocfs2_recovery_map {
33 unsigned int rm_used;
34 unsigned int *rm_entries;
35};
36
37
ccd979bd
MF
38struct ocfs2_journal {
39 enum ocfs2_journal_state j_state; /* Journals current state */
40
41 journal_t *j_journal; /* The kernels journal type */
42 struct inode *j_inode; /* Kernel inode pointing to
43 * this journal */
44 struct ocfs2_super *j_osb; /* pointer to the super
45 * block for the node
46 * we're currently
47 * running on -- not
48 * necessarily the super
49 * block from the node
50 * which we usually run
51 * from (recovery,
52 * etc) */
53 struct buffer_head *j_bh; /* Journal disk inode block */
54 atomic_t j_num_trans; /* Number of transactions
55 * currently in the system. */
83fd9c7f 56 spinlock_t j_lock;
ccd979bd
MF
57 unsigned long j_trans_id;
58 struct rw_semaphore j_trans_barrier;
59 wait_queue_head_t j_checkpointed;
60
83fd9c7f 61 /* both fields protected by j_lock*/
ccd979bd
MF
62 struct list_head j_la_cleanups;
63 struct work_struct j_recovery_work;
64};
65
66extern spinlock_t trans_inc_lock;
67
68/* wrap j_trans_id so we never have it equal to zero. */
69static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j)
70{
71 unsigned long old_id;
72 spin_lock(&trans_inc_lock);
73 old_id = j->j_trans_id++;
74 if (unlikely(!j->j_trans_id))
75 j->j_trans_id = 1;
76 spin_unlock(&trans_inc_lock);
77 return old_id;
78}
79
0cf2f763
JB
80static inline void ocfs2_set_ci_lock_trans(struct ocfs2_journal *journal,
81 struct ocfs2_caching_info *ci)
ccd979bd
MF
82{
83 spin_lock(&trans_inc_lock);
0cf2f763 84 ci->ci_last_trans = journal->j_trans_id;
ccd979bd
MF
85 spin_unlock(&trans_inc_lock);
86}
87
88/* Used to figure out whether it's safe to drop a metadata lock on an
0cf2f763 89 * cached object. Returns true if all the object's changes have been
ccd979bd
MF
90 * checkpointed to disk. You should be holding the spinlock on the
91 * metadata lock while calling this to be sure that nobody can take
92 * the lock and put it on another transaction. */
0cf2f763 93static inline int ocfs2_ci_fully_checkpointed(struct ocfs2_caching_info *ci)
ccd979bd
MF
94{
95 int ret;
0cf2f763
JB
96 struct ocfs2_journal *journal =
97 OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
ccd979bd
MF
98
99 spin_lock(&trans_inc_lock);
0cf2f763 100 ret = time_after(journal->j_trans_id, ci->ci_last_trans);
ccd979bd
MF
101 spin_unlock(&trans_inc_lock);
102 return ret;
103}
104
0cf2f763
JB
105/* convenience function to check if an object backed by struct
106 * ocfs2_caching_info is still new (has never hit disk) Will do you a
107 * favor and set created_trans = 0 when you've
108 * been checkpointed. returns '1' if the ci is still new. */
109static inline int ocfs2_ci_is_new(struct ocfs2_caching_info *ci)
ccd979bd
MF
110{
111 int ret;
0cf2f763
JB
112 struct ocfs2_journal *journal =
113 OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
114
115 spin_lock(&trans_inc_lock);
116 ret = !(time_after(journal->j_trans_id, ci->ci_created_trans));
117 if (!ret)
118 ci->ci_created_trans = 0;
119 spin_unlock(&trans_inc_lock);
120 return ret;
121}
ccd979bd 122
0cf2f763
JB
123/* Wrapper for inodes so we can check system files */
124static inline int ocfs2_inode_is_new(struct inode *inode)
125{
ccd979bd
MF
126 /* System files are never "new" as they're written out by
127 * mkfs. This helps us early during mount, before we have the
128 * journal open and j_trans_id could be junk. */
129 if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
130 return 0;
0cf2f763
JB
131
132 return ocfs2_ci_is_new(INODE_CACHE(inode));
ccd979bd
MF
133}
134
0cf2f763
JB
135static inline void ocfs2_ci_set_new(struct ocfs2_super *osb,
136 struct ocfs2_caching_info *ci)
ccd979bd
MF
137{
138 spin_lock(&trans_inc_lock);
0cf2f763 139 ci->ci_created_trans = osb->journal->j_trans_id;
ccd979bd
MF
140 spin_unlock(&trans_inc_lock);
141}
142
ccd979bd 143/* Exported only for the journal struct init code in super.c. Do not call. */
df152c24 144void ocfs2_orphan_scan_init(struct ocfs2_super *osb);
8b712cd5 145void ocfs2_orphan_scan_start(struct ocfs2_super *osb);
83273932
SE
146void ocfs2_orphan_scan_stop(struct ocfs2_super *osb);
147void ocfs2_orphan_scan_exit(struct ocfs2_super *osb);
148
c4028958 149void ocfs2_complete_recovery(struct work_struct *work);
553abd04
JB
150void ocfs2_wait_for_recovery(struct ocfs2_super *osb);
151
152int ocfs2_recovery_init(struct ocfs2_super *osb);
153void ocfs2_recovery_exit(struct ocfs2_super *osb);
ccd979bd 154
9140db04 155int ocfs2_compute_replay_slots(struct ocfs2_super *osb);
ccd979bd
MF
156/*
157 * Journal Control:
158 * Initialize, Load, Shutdown, Wipe a journal.
159 *
160 * ocfs2_journal_init - Initialize journal structures in the OSB.
161 * ocfs2_journal_load - Load the given journal off disk. Replay it if
162 * there's transactions still in there.
163 * ocfs2_journal_shutdown - Shutdown a journal, this will flush all
164 * uncommitted, uncheckpointed transactions.
165 * ocfs2_journal_wipe - Wipe transactions from a journal. Optionally
166 * zero out each block.
167 * ocfs2_recovery_thread - Perform recovery on a node. osb is our own osb.
168 * ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
169 * event on.
170 * ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
171 */
172void ocfs2_set_journal_params(struct ocfs2_super *osb);
173int ocfs2_journal_init(struct ocfs2_journal *journal,
174 int *dirty);
175void ocfs2_journal_shutdown(struct ocfs2_super *osb);
176int ocfs2_journal_wipe(struct ocfs2_journal *journal,
177 int full);
539d8264
SM
178int ocfs2_journal_load(struct ocfs2_journal *journal, int local,
179 int replayed);
ccd979bd
MF
180int ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
181void ocfs2_recovery_thread(struct ocfs2_super *osb,
182 int node_num);
183int ocfs2_mark_dead_nodes(struct ocfs2_super *osb);
184void ocfs2_complete_mount_recovery(struct ocfs2_super *osb);
2205363d 185void ocfs2_complete_quota_recovery(struct ocfs2_super *osb);
ccd979bd
MF
186
187static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb)
188{
ccd979bd
MF
189 wake_up(&osb->checkpoint_event);
190}
191
192static inline void ocfs2_checkpoint_inode(struct inode *inode)
193{
194 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
195
c271c5c2
SM
196 if (ocfs2_mount_local(osb))
197 return;
198
0cf2f763 199 if (!ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))) {
ccd979bd
MF
200 /* WARNING: This only kicks off a single
201 * checkpoint. If someone races you and adds more
202 * metadata to the journal, you won't know, and will
25985edc 203 * wind up waiting *a lot* longer than necessary. Right
ccd979bd
MF
204 * now we only use this in clear_inode so that's
205 * OK. */
206 ocfs2_start_checkpoint(osb);
207
208 wait_event(osb->journal->j_checkpointed,
0cf2f763 209 ocfs2_ci_fully_checkpointed(INODE_CACHE(inode)));
ccd979bd
MF
210 }
211}
212
213/*
214 * Transaction Handling:
215 * Manage the lifetime of a transaction handle.
216 *
ccd979bd
MF
217 * ocfs2_start_trans - Begin a transaction. Give it an upper estimate of
218 * the number of blocks that will be changed during
219 * this handle.
1fabe148
MF
220 * ocfs2_commit_trans - Complete a handle. It might return -EIO if
221 * the journal was aborted. The majority of paths don't
222 * check the return value as an error there comes too
223 * late to do anything (and will be picked up in a
224 * later transaction).
ccd979bd
MF
225 * ocfs2_extend_trans - Extend a handle by nblocks credits. This may
226 * commit the handle to disk in the process, but will
227 * not release any locks taken during the transaction.
50655ae9 228 * ocfs2_journal_access* - Notify the handle that we want to journal this
ccd979bd
MF
229 * buffer. Will have to call ocfs2_journal_dirty once
230 * we've actually dirtied it. Type is one of . or .
50655ae9
JB
231 * Always call the specific flavor of
232 * ocfs2_journal_access_*() unless you intend to
233 * manage the checksum by hand.
ccd979bd 234 * ocfs2_journal_dirty - Mark a journalled buffer as having dirty data.
2b4e30fb
JB
235 * ocfs2_jbd2_file_inode - Mark an inode so that its data goes out before
236 * the current handle commits.
ccd979bd
MF
237 */
238
239/* You must always start_trans with a number of buffs > 0, but it's
240 * perfectly legal to go through an entire transaction without having
241 * dirtied any buffers. */
1fabe148 242handle_t *ocfs2_start_trans(struct ocfs2_super *osb,
ccd979bd 243 int max_buffs);
1fabe148
MF
244int ocfs2_commit_trans(struct ocfs2_super *osb,
245 handle_t *handle);
1fc58146 246int ocfs2_extend_trans(handle_t *handle, int nblocks);
2b1e55c3
YL
247int ocfs2_allocate_extend_trans(handle_t *handle,
248 int thresh);
249
250/*
251 * Define an arbitrary limit for the amount of data we will anticipate
252 * writing to any given transaction. For unbounded transactions such as
253 * fallocate(2) we can write more than this, but we always
254 * start off at the maximum transaction size and grow the transaction
255 * optimistically as we go.
256 */
257#define OCFS2_MAX_TRANS_DATA 64U
ccd979bd
MF
258
259/*
260 * Create access is for when we get a newly created buffer and we're
261 * not gonna read it off disk, but rather fill it ourselves. Right
262 * now, we don't do anything special with this (it turns into a write
263 * request), but this is a good placeholder in case we do...
264 *
265 * Write access is for when we read a block off disk and are going to
266 * modify it. This way the journalling layer knows it may need to make
267 * a copy of that block (if it's part of another, uncommitted
268 * transaction) before we do so.
269 */
270#define OCFS2_JOURNAL_ACCESS_CREATE 0
271#define OCFS2_JOURNAL_ACCESS_WRITE 1
272#define OCFS2_JOURNAL_ACCESS_UNDO 2
273
13723d00 274
50655ae9 275/* ocfs2_inode */
0cf2f763 276int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci,
13723d00 277 struct buffer_head *bh, int type);
50655ae9 278/* ocfs2_extent_block */
0cf2f763 279int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci,
50655ae9 280 struct buffer_head *bh, int type);
93c97087
TM
281/* ocfs2_refcount_block */
282int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci,
283 struct buffer_head *bh, int type);
50655ae9 284/* ocfs2_group_desc */
0cf2f763 285int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci,
50655ae9
JB
286 struct buffer_head *bh, int type);
287/* ocfs2_xattr_block */
0cf2f763 288int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci,
50655ae9
JB
289 struct buffer_head *bh, int type);
290/* quota blocks */
0cf2f763 291int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci,
50655ae9
JB
292 struct buffer_head *bh, int type);
293/* dirblock */
0cf2f763 294int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci,
50655ae9 295 struct buffer_head *bh, int type);
9b7895ef 296/* ocfs2_dx_root_block */
0cf2f763 297int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci,
9b7895ef
MF
298 struct buffer_head *bh, int type);
299/* ocfs2_dx_leaf */
0cf2f763 300int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci,
9b7895ef 301 struct buffer_head *bh, int type);
50655ae9 302/* Anything that has no ecc */
0cf2f763 303int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
50655ae9
JB
304 struct buffer_head *bh, int type);
305
ccd979bd
MF
306/*
307 * A word about the journal_access/journal_dirty "dance". It is
308 * entirely legal to journal_access a buffer more than once (as long
309 * as the access type is the same -- I'm not sure what will happen if
310 * access type is different but this should never happen anyway) It is
311 * also legal to journal_dirty a buffer more than once. In fact, you
312 * can even journal_access a buffer after you've done a
313 * journal_access/journal_dirty pair. The only thing you cannot do
314 * however, is journal_dirty a buffer which you haven't yet passed to
315 * journal_access at least once.
316 *
317 * That said, 99% of the time this doesn't matter and this is what the
318 * path looks like:
319 *
320 * <read a bh>
321 * ocfs2_journal_access(handle, bh, OCFS2_JOURNAL_ACCESS_WRITE);
322 * <modify the bh>
323 * ocfs2_journal_dirty(handle, bh);
324 */
ec20cec7 325void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh);
ccd979bd
MF
326
327/*
328 * Credit Macros:
329 * Convenience macros to calculate number of credits needed.
330 *
331 * For convenience sake, I have a set of macros here which calculate
332 * the *maximum* number of sectors which will be changed for various
333 * metadata updates.
334 */
335
336/* simple file updates like chmod, etc. */
337#define OCFS2_INODE_UPDATE_CREDITS 1
338
cf1d6c76
TY
339/* extended attribute block update */
340#define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1
341
0584974a
JK
342/* Update of a single quota block */
343#define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1
344
a90714c1 345/* global quotafile inode update, data block */
0584974a
JK
346#define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
347 OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
a90714c1 348
0584974a 349#define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS
a90714c1
JK
350/*
351 * The two writes below can accidentally see global info dirty due
352 * to set_info() quotactl so make them prepared for the writes.
353 */
354/* quota data block, global info */
355/* Write to local quota file */
0584974a
JK
356#define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
357 OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
a90714c1
JK
358
359/* global quota data block, local quota data block, global quota inode,
360 * global quota info */
0584974a
JK
361#define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
362 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
a90714c1
JK
363
364static inline int ocfs2_quota_trans_credits(struct super_block *sb)
365{
366 int credits = 0;
367
368 if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA))
369 credits += OCFS2_QWRITE_CREDITS;
370 if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA))
371 credits += OCFS2_QWRITE_CREDITS;
372 return credits;
373}
374
d659072f
TM
375/* group extend. inode update and last group update. */
376#define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
377
7909f2bf
TM
378/* group add. inode update and the new group update. */
379#define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
380
ccd979bd
MF
381/* get one bit out of a suballocator: dinode + group descriptor +
382 * prev. group desc. if we relink. */
383#define OCFS2_SUBALLOC_ALLOC (3)
384
a90714c1
JK
385static inline int ocfs2_inline_to_extents_credits(struct super_block *sb)
386{
387 return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS +
388 ocfs2_quota_trans_credits(sb);
389}
1afc32b9 390
ccd979bd
MF
391/* dinode + group descriptor update. We don't relink on free yet. */
392#define OCFS2_SUBALLOC_FREE (2)
393
394#define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS
395#define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE \
396 + OCFS2_TRUNCATE_LOG_UPDATE)
397
a90714c1
JK
398static inline int ocfs2_remove_extent_credits(struct super_block *sb)
399{
400 return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS +
401 ocfs2_quota_trans_credits(sb);
402}
063c4561 403
705773a6
JK
404/* data block for new dir/symlink, allocation of directory block, dx_root
405 * update for free list */
406#define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + OCFS2_SUBALLOC_ALLOC + 1)
ccd979bd 407
9b7895ef
MF
408static inline int ocfs2_add_dir_index_credits(struct super_block *sb)
409{
410 /* 1 block for index, 2 allocs (data, metadata), 1 clusters
411 * worth of blocks for initial extent. */
412 return 1 + 2 * OCFS2_SUBALLOC_ALLOC +
413 ocfs2_clusters_to_blocks(sb, 1);
414}
415
416/* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode
417 * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr
418 * blocks + quota update */
419static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir,
420 int xattr_credits)
a90714c1 421{
9b7895ef
MF
422 int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS;
423
424 if (is_dir)
425 dir_credits += ocfs2_add_dir_index_credits(sb);
426
427 return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits +
a90714c1
JK
428 ocfs2_quota_trans_credits(sb);
429}
ccd979bd
MF
430
431/* local alloc metadata change + main bitmap updates */
432#define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS \
433 + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE)
434
435/* used when we don't need an allocation change for a dir extend. One
436 * for the dinode, one for the new block. */
437#define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2)
438
a90714c1 439/* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota
0393afea
XH
440 * update on dir + index leaf + dx root update for free list +
441 * previous dirblock update in the free list */
a90714c1
JK
442static inline int ocfs2_link_credits(struct super_block *sb)
443{
0393afea 444 return 2*OCFS2_INODE_UPDATE_CREDITS + 4 +
a90714c1
JK
445 ocfs2_quota_trans_credits(sb);
446}
ccd979bd
MF
447
448/* inode + dir inode (if we unlink a dir), + dir entry block + orphan
e7c17e43 449 * dir inode link + dir inode index leaf + dir index root */
a90714c1
JK
450static inline int ocfs2_unlink_credits(struct super_block *sb)
451{
452 /* The quota update from ocfs2_link_credits is unused here... */
e7c17e43 453 return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb);
a90714c1 454}
ccd979bd
MF
455
456/* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
dfa13f39
JB
457 * inode alloc group descriptor + orphan dir index root +
458 * orphan dir index leaf */
459#define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 4)
ccd979bd 460
06ee5c75
JQ
461/* dinode + orphan dir dinode + extent tree leaf block + orphan dir entry +
462 * orphan dir index root + orphan dir index leaf */
463#define OCFS2_INODE_ADD_TO_ORPHAN_CREDITS (2 * OCFS2_INODE_UPDATE_CREDITS + 4)
464#define OCFS2_INODE_DEL_FROM_ORPHAN_CREDITS OCFS2_INODE_ADD_TO_ORPHAN_CREDITS
465
ccd979bd
MF
466/* dinode update, old dir dinode update, new dir dinode update, old
467 * dir dir entry, new dir dir entry, dir entry update for renaming
9b7895ef 468 * directory + target unlink + 3 x dir index leaves */
a90714c1
JK
469static inline int ocfs2_rename_credits(struct super_block *sb)
470{
9b7895ef 471 return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb);
a90714c1 472}
ccd979bd 473
cf1d6c76
TY
474/* global bitmap dinode, group desc., relinked group,
475 * suballocator dinode, group desc., relinked group,
476 * dinode, xattr block */
477#define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \
478 + OCFS2_INODE_UPDATE_CREDITS \
479 + OCFS2_XATTR_BLOCK_UPDATE_CREDITS)
480
9b7895ef
MF
481/* inode update, removal of dx root block from allocator */
482#define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
483 OCFS2_SUBALLOC_FREE)
484
4ed8a6bb
MF
485static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb)
486{
487 int credits = 1 + OCFS2_SUBALLOC_ALLOC;
488
489 credits += ocfs2_clusters_to_blocks(sb, 1);
490 credits += ocfs2_quota_trans_credits(sb);
491
492 return credits;
493}
494
8bf396de
TM
495/* inode update, new refcount block and its allocation credits. */
496#define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \
497 + OCFS2_SUBALLOC_ALLOC)
498
499/* inode and the refcount block update. */
500#define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
501
502/*
503 * inode and the refcount block update.
504 * It doesn't include the credits for sub alloc change.
505 * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added.
506 */
507#define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
508
bcbbb24a
TM
509/* 2 metadata alloc, 2 new blocks and root refcount block */
510#define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3)
511
811f933d
TM
512/*
513 * Please note that the caller must make sure that root_el is the root
514 * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
515 * the result may be wrong.
516 */
ccd979bd 517static inline int ocfs2_calc_extend_credits(struct super_block *sb,
06f9da6e 518 struct ocfs2_extent_list *root_el)
ccd979bd 519{
811f933d 520 int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks;
ccd979bd
MF
521
522 /* bitmap dinode, group desc. + relinked group. */
523 bitmap_blocks = OCFS2_SUBALLOC_ALLOC;
524
525 /* we might need to shift tree depth so lets assume an
526 * absolute worst case of complete fragmentation. Even with
527 * that, we only need one update for the dinode, and then
528 * however many metadata chunks needed * a remaining suballoc
529 * alloc. */
530 sysfile_bitmap_blocks = 1 +
811f933d 531 (OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(root_el);
ccd979bd
MF
532
533 /* this does not include *new* metadata blocks, which are
811f933d 534 * accounted for in sysfile_bitmap_blocks. root_el +
ccd979bd
MF
535 * prev. last_eb_blk + blocks along edge of tree.
536 * calc_symlink_credits passes because we just need 1
537 * credit for the dinode there. */
811f933d 538 extent_blocks = 1 + 1 + le16_to_cpu(root_el->l_tree_depth);
ccd979bd 539
a90714c1 540 return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks +
6115ea28 541 ocfs2_quota_trans_credits(sb);
ccd979bd
MF
542}
543
544static inline int ocfs2_calc_symlink_credits(struct super_block *sb)
545{
9b7895ef 546 int blocks = ocfs2_mknod_credits(sb, 0, 0);
ccd979bd
MF
547
548 /* links can be longer than one block so we may update many
549 * within our single allocated extent. */
550 blocks += ocfs2_clusters_to_blocks(sb, 1);
551
a90714c1 552 return blocks + ocfs2_quota_trans_credits(sb);
ccd979bd
MF
553}
554
555static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
556 unsigned int cpg)
557{
558 int blocks;
559 int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1;
560 /* parent inode update + new block group header + bitmap inode update
561 + bitmap blocks affected */
562 blocks = 1 + 1 + 1 + bitmap_blocks;
563 return blocks;
564}
565
8b06bc59
JB
566/*
567 * Allocating a discontiguous block group requires the credits from
568 * ocfs2_calc_group_alloc_credits() as well as enough credits to fill
569 * the group descriptor's extent list. The caller already has started
570 * the transaction with ocfs2_calc_group_alloc_credits(). They extend
571 * it with these credits.
572 */
573static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb)
574{
575 return ocfs2_extent_recs_per_gd(sb);
576}
577
ccd979bd
MF
578static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
579 unsigned int clusters_to_del,
580 struct ocfs2_dinode *fe,
581 struct ocfs2_extent_list *last_el)
582{
583 /* for dinode + all headers in this pass + update to next leaf */
584 u16 next_free = le16_to_cpu(last_el->l_next_free_rec);
585 u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth);
586 int credits = 1 + tree_depth + 1;
587 int i;
588
589 i = next_free - 1;
590 BUG_ON(i < 0);
591
592 /* We may be deleting metadata blocks, so metadata alloc dinode +
593 one desc. block for each possible delete. */
594 if (tree_depth && next_free == 1 &&
e48edee2 595 ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del)
ccd979bd
MF
596 credits += 1 + tree_depth;
597
598 /* update to the truncate log. */
599 credits += OCFS2_TRUNCATE_LOG_UPDATE;
600
a90714c1
JK
601 credits += ocfs2_quota_trans_credits(sb);
602
ccd979bd
MF
603 return credits;
604}
605
2b4e30fb
JB
606static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode)
607{
41617e1a 608 return jbd2_journal_inode_add_write(handle, &OCFS2_I(inode)->ip_jinode);
2b4e30fb
JB
609}
610
611static inline int ocfs2_begin_ordered_truncate(struct inode *inode,
612 loff_t new_size)
613{
7f5aa215
JK
614 return jbd2_journal_begin_ordered_truncate(
615 OCFS2_SB(inode->i_sb)->journal->j_journal,
616 &OCFS2_I(inode)->ip_jinode,
617 new_size);
2b4e30fb
JB
618}
619
2931cdcb
DW
620static inline void ocfs2_update_inode_fsync_trans(handle_t *handle,
621 struct inode *inode,
622 int datasync)
623{
624 struct ocfs2_inode_info *oi = OCFS2_I(inode);
625
626 oi->i_sync_tid = handle->h_transaction->t_tid;
627 if (datasync)
628 oi->i_datasync_tid = handle->h_transaction->t_tid;
629}
630
ccd979bd 631#endif /* OCFS2_JOURNAL_H */