blob: 619dd7f6c053c1562863707bebc7c3c6e1753b52 [file] [log] [blame]
Mark Fashehccd979b2005-12-15 14:31:24 -08001/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * journal.h
5 *
6 * Defines journalling api and structures.
7 *
8 * Copyright (C) 2003, 2005 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#ifndef OCFS2_JOURNAL_H
27#define OCFS2_JOURNAL_H
28
29#include <linux/fs.h>
Mark Fasheh53ef99c2008-11-18 16:53:43 -080030#include <linux/jbd2.h>
Mark Fashehccd979b2005-12-15 14:31:24 -080031
Mark Fashehccd979b2005-12-15 14:31:24 -080032enum ocfs2_journal_state {
33 OCFS2_JOURNAL_FREE = 0,
34 OCFS2_JOURNAL_LOADED,
35 OCFS2_JOURNAL_IN_SHUTDOWN,
36};
37
38struct ocfs2_super;
39struct ocfs2_dinode;
Mark Fashehccd979b2005-12-15 14:31:24 -080040
Sunil Mushran96a6c642008-12-16 15:49:16 -080041/*
42 * The recovery_list is a simple linked list of node numbers to recover.
43 * It is protected by the recovery_lock.
44 */
45
46struct ocfs2_recovery_map {
47 unsigned int rm_used;
48 unsigned int *rm_entries;
49};
50
51
Mark Fashehccd979b2005-12-15 14:31:24 -080052struct ocfs2_journal {
53 enum ocfs2_journal_state j_state; /* Journals current state */
54
55 journal_t *j_journal; /* The kernels journal type */
56 struct inode *j_inode; /* Kernel inode pointing to
57 * this journal */
58 struct ocfs2_super *j_osb; /* pointer to the super
59 * block for the node
60 * we're currently
61 * running on -- not
62 * necessarily the super
63 * block from the node
64 * which we usually run
65 * from (recovery,
66 * etc) */
67 struct buffer_head *j_bh; /* Journal disk inode block */
68 atomic_t j_num_trans; /* Number of transactions
69 * currently in the system. */
70 unsigned long j_trans_id;
71 struct rw_semaphore j_trans_barrier;
72 wait_queue_head_t j_checkpointed;
73
74 spinlock_t j_lock;
75 struct list_head j_la_cleanups;
76 struct work_struct j_recovery_work;
77};
78
79extern spinlock_t trans_inc_lock;
80
81/* wrap j_trans_id so we never have it equal to zero. */
82static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j)
83{
84 unsigned long old_id;
85 spin_lock(&trans_inc_lock);
86 old_id = j->j_trans_id++;
87 if (unlikely(!j->j_trans_id))
88 j->j_trans_id = 1;
89 spin_unlock(&trans_inc_lock);
90 return old_id;
91}
92
93static inline void ocfs2_set_inode_lock_trans(struct ocfs2_journal *journal,
94 struct inode *inode)
95{
96 spin_lock(&trans_inc_lock);
97 OCFS2_I(inode)->ip_last_trans = journal->j_trans_id;
98 spin_unlock(&trans_inc_lock);
99}
100
101/* Used to figure out whether it's safe to drop a metadata lock on an
102 * inode. Returns true if all the inodes changes have been
103 * checkpointed to disk. You should be holding the spinlock on the
104 * metadata lock while calling this to be sure that nobody can take
105 * the lock and put it on another transaction. */
106static inline int ocfs2_inode_fully_checkpointed(struct inode *inode)
107{
108 int ret;
109 struct ocfs2_journal *journal = OCFS2_SB(inode->i_sb)->journal;
110
111 spin_lock(&trans_inc_lock);
112 ret = time_after(journal->j_trans_id, OCFS2_I(inode)->ip_last_trans);
113 spin_unlock(&trans_inc_lock);
114 return ret;
115}
116
117/* convenience function to check if an inode is still new (has never
118 * hit disk) Will do you a favor and set created_trans = 0 when you've
119 * been checkpointed. returns '1' if the inode is still new. */
120static inline int ocfs2_inode_is_new(struct inode *inode)
121{
122 int ret;
123
124 /* System files are never "new" as they're written out by
125 * mkfs. This helps us early during mount, before we have the
126 * journal open and j_trans_id could be junk. */
127 if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
128 return 0;
129 spin_lock(&trans_inc_lock);
130 ret = !(time_after(OCFS2_SB(inode->i_sb)->journal->j_trans_id,
131 OCFS2_I(inode)->ip_created_trans));
132 if (!ret)
133 OCFS2_I(inode)->ip_created_trans = 0;
134 spin_unlock(&trans_inc_lock);
135 return ret;
136}
137
138static inline void ocfs2_inode_set_new(struct ocfs2_super *osb,
139 struct inode *inode)
140{
141 spin_lock(&trans_inc_lock);
142 OCFS2_I(inode)->ip_created_trans = osb->journal->j_trans_id;
143 spin_unlock(&trans_inc_lock);
144}
145
Mark Fashehccd979b2005-12-15 14:31:24 -0800146/* Exported only for the journal struct init code in super.c. Do not call. */
David Howellsc4028952006-11-22 14:57:56 +0000147void ocfs2_complete_recovery(struct work_struct *work);
Joel Becker553abd02008-02-01 12:03:57 -0800148void ocfs2_wait_for_recovery(struct ocfs2_super *osb);
149
150int ocfs2_recovery_init(struct ocfs2_super *osb);
151void ocfs2_recovery_exit(struct ocfs2_super *osb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800152
Srinivas Eeda9140db02009-03-06 14:21:46 -0800153int ocfs2_compute_replay_slots(struct ocfs2_super *osb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800154/*
155 * Journal Control:
156 * Initialize, Load, Shutdown, Wipe a journal.
157 *
158 * ocfs2_journal_init - Initialize journal structures in the OSB.
159 * ocfs2_journal_load - Load the given journal off disk. Replay it if
160 * there's transactions still in there.
161 * ocfs2_journal_shutdown - Shutdown a journal, this will flush all
162 * uncommitted, uncheckpointed transactions.
163 * ocfs2_journal_wipe - Wipe transactions from a journal. Optionally
164 * zero out each block.
165 * ocfs2_recovery_thread - Perform recovery on a node. osb is our own osb.
166 * ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
167 * event on.
168 * ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
169 */
170void ocfs2_set_journal_params(struct ocfs2_super *osb);
171int ocfs2_journal_init(struct ocfs2_journal *journal,
172 int *dirty);
173void ocfs2_journal_shutdown(struct ocfs2_super *osb);
174int ocfs2_journal_wipe(struct ocfs2_journal *journal,
175 int full);
Sunil Mushran539d8262008-07-14 17:31:10 -0700176int ocfs2_journal_load(struct ocfs2_journal *journal, int local,
177 int replayed);
Mark Fashehccd979b2005-12-15 14:31:24 -0800178int ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
179void ocfs2_recovery_thread(struct ocfs2_super *osb,
180 int node_num);
181int ocfs2_mark_dead_nodes(struct ocfs2_super *osb);
182void ocfs2_complete_mount_recovery(struct ocfs2_super *osb);
Jan Kara22053632008-10-20 23:50:38 +0200183void ocfs2_complete_quota_recovery(struct ocfs2_super *osb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800184
185static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb)
186{
187 atomic_set(&osb->needs_checkpoint, 1);
188 wake_up(&osb->checkpoint_event);
189}
190
191static inline void ocfs2_checkpoint_inode(struct inode *inode)
192{
193 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
194
Sunil Mushranc271c5c2006-12-05 17:56:35 -0800195 if (ocfs2_mount_local(osb))
196 return;
197
Mark Fashehccd979b2005-12-15 14:31:24 -0800198 if (!ocfs2_inode_fully_checkpointed(inode)) {
199 /* WARNING: This only kicks off a single
200 * checkpoint. If someone races you and adds more
201 * metadata to the journal, you won't know, and will
202 * wind up waiting *alot* longer than necessary. Right
203 * now we only use this in clear_inode so that's
204 * OK. */
205 ocfs2_start_checkpoint(osb);
206
207 wait_event(osb->journal->j_checkpointed,
208 ocfs2_inode_fully_checkpointed(inode));
209 }
210}
211
212/*
213 * Transaction Handling:
214 * Manage the lifetime of a transaction handle.
215 *
Mark Fashehccd979b2005-12-15 14:31:24 -0800216 * ocfs2_start_trans - Begin a transaction. Give it an upper estimate of
217 * the number of blocks that will be changed during
218 * this handle.
Mark Fasheh1fabe142006-10-09 18:11:45 -0700219 * ocfs2_commit_trans - Complete a handle. It might return -EIO if
220 * the journal was aborted. The majority of paths don't
221 * check the return value as an error there comes too
222 * late to do anything (and will be picked up in a
223 * later transaction).
Mark Fashehccd979b2005-12-15 14:31:24 -0800224 * ocfs2_extend_trans - Extend a handle by nblocks credits. This may
225 * commit the handle to disk in the process, but will
226 * not release any locks taken during the transaction.
Joel Becker50655ae2008-09-11 15:53:07 -0700227 * ocfs2_journal_access* - Notify the handle that we want to journal this
Mark Fashehccd979b2005-12-15 14:31:24 -0800228 * buffer. Will have to call ocfs2_journal_dirty once
229 * we've actually dirtied it. Type is one of . or .
Joel Becker50655ae2008-09-11 15:53:07 -0700230 * Always call the specific flavor of
231 * ocfs2_journal_access_*() unless you intend to
232 * manage the checksum by hand.
Mark Fashehccd979b2005-12-15 14:31:24 -0800233 * ocfs2_journal_dirty - Mark a journalled buffer as having dirty data.
Joel Becker2b4e30f2008-09-03 20:03:41 -0700234 * ocfs2_jbd2_file_inode - Mark an inode so that its data goes out before
235 * the current handle commits.
Mark Fashehccd979b2005-12-15 14:31:24 -0800236 */
237
238/* You must always start_trans with a number of buffs > 0, but it's
239 * perfectly legal to go through an entire transaction without having
240 * dirtied any buffers. */
Mark Fasheh1fabe142006-10-09 18:11:45 -0700241handle_t *ocfs2_start_trans(struct ocfs2_super *osb,
Mark Fashehccd979b2005-12-15 14:31:24 -0800242 int max_buffs);
Mark Fasheh1fabe142006-10-09 18:11:45 -0700243int ocfs2_commit_trans(struct ocfs2_super *osb,
244 handle_t *handle);
Mark Fasheh1fc58142006-10-05 14:15:36 -0700245int ocfs2_extend_trans(handle_t *handle, int nblocks);
Mark Fashehccd979b2005-12-15 14:31:24 -0800246
247/*
248 * Create access is for when we get a newly created buffer and we're
249 * not gonna read it off disk, but rather fill it ourselves. Right
250 * now, we don't do anything special with this (it turns into a write
251 * request), but this is a good placeholder in case we do...
252 *
253 * Write access is for when we read a block off disk and are going to
254 * modify it. This way the journalling layer knows it may need to make
255 * a copy of that block (if it's part of another, uncommitted
256 * transaction) before we do so.
257 */
258#define OCFS2_JOURNAL_ACCESS_CREATE 0
259#define OCFS2_JOURNAL_ACCESS_WRITE 1
260#define OCFS2_JOURNAL_ACCESS_UNDO 2
261
Joel Becker13723d02008-10-17 19:25:01 -0700262
Joel Becker50655ae2008-09-11 15:53:07 -0700263/* ocfs2_inode */
264int ocfs2_journal_access_di(handle_t *handle, struct inode *inode,
Joel Becker13723d02008-10-17 19:25:01 -0700265 struct buffer_head *bh, int type);
Joel Becker50655ae2008-09-11 15:53:07 -0700266/* ocfs2_extent_block */
267int ocfs2_journal_access_eb(handle_t *handle, struct inode *inode,
268 struct buffer_head *bh, int type);
269/* ocfs2_group_desc */
270int ocfs2_journal_access_gd(handle_t *handle, struct inode *inode,
271 struct buffer_head *bh, int type);
272/* ocfs2_xattr_block */
273int ocfs2_journal_access_xb(handle_t *handle, struct inode *inode,
274 struct buffer_head *bh, int type);
275/* quota blocks */
276int ocfs2_journal_access_dq(handle_t *handle, struct inode *inode,
277 struct buffer_head *bh, int type);
278/* dirblock */
279int ocfs2_journal_access_db(handle_t *handle, struct inode *inode,
280 struct buffer_head *bh, int type);
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800281/* ocfs2_dx_root_block */
282int ocfs2_journal_access_dr(handle_t *handle, struct inode *inode,
283 struct buffer_head *bh, int type);
284/* ocfs2_dx_leaf */
285int ocfs2_journal_access_dl(handle_t *handle, struct inode *inode,
286 struct buffer_head *bh, int type);
Joel Becker50655ae2008-09-11 15:53:07 -0700287/* Anything that has no ecc */
288int ocfs2_journal_access(handle_t *handle, struct inode *inode,
289 struct buffer_head *bh, int type);
290
Mark Fashehccd979b2005-12-15 14:31:24 -0800291/*
292 * A word about the journal_access/journal_dirty "dance". It is
293 * entirely legal to journal_access a buffer more than once (as long
294 * as the access type is the same -- I'm not sure what will happen if
295 * access type is different but this should never happen anyway) It is
296 * also legal to journal_dirty a buffer more than once. In fact, you
297 * can even journal_access a buffer after you've done a
298 * journal_access/journal_dirty pair. The only thing you cannot do
299 * however, is journal_dirty a buffer which you haven't yet passed to
300 * journal_access at least once.
301 *
302 * That said, 99% of the time this doesn't matter and this is what the
303 * path looks like:
304 *
305 * <read a bh>
306 * ocfs2_journal_access(handle, bh, OCFS2_JOURNAL_ACCESS_WRITE);
307 * <modify the bh>
308 * ocfs2_journal_dirty(handle, bh);
309 */
Mark Fasheh1fabe142006-10-09 18:11:45 -0700310int ocfs2_journal_dirty(handle_t *handle,
Mark Fashehccd979b2005-12-15 14:31:24 -0800311 struct buffer_head *bh);
Mark Fashehccd979b2005-12-15 14:31:24 -0800312
313/*
314 * Credit Macros:
315 * Convenience macros to calculate number of credits needed.
316 *
317 * For convenience sake, I have a set of macros here which calculate
318 * the *maximum* number of sectors which will be changed for various
319 * metadata updates.
320 */
321
322/* simple file updates like chmod, etc. */
323#define OCFS2_INODE_UPDATE_CREDITS 1
324
Tiger Yangcf1d6c72008-08-18 17:11:00 +0800325/* extended attribute block update */
326#define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1
327
Jan Karaa90714c2008-10-09 19:38:40 +0200328/* global quotafile inode update, data block */
329#define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
330
331/*
332 * The two writes below can accidentally see global info dirty due
333 * to set_info() quotactl so make them prepared for the writes.
334 */
335/* quota data block, global info */
336/* Write to local quota file */
337#define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + 1)
338
339/* global quota data block, local quota data block, global quota inode,
340 * global quota info */
341#define OCFS2_QSYNC_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 3)
342
343static inline int ocfs2_quota_trans_credits(struct super_block *sb)
344{
345 int credits = 0;
346
347 if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA))
348 credits += OCFS2_QWRITE_CREDITS;
349 if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA))
350 credits += OCFS2_QWRITE_CREDITS;
351 return credits;
352}
353
354/* Number of credits needed for removing quota structure from file */
355int ocfs2_calc_qdel_credits(struct super_block *sb, int type);
356/* Number of credits needed for initialization of new quota structure */
357int ocfs2_calc_qinit_credits(struct super_block *sb, int type);
358
Tao Mad6590722007-12-18 15:47:03 +0800359/* group extend. inode update and last group update. */
360#define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
361
Tao Ma7909f2b2007-12-18 15:47:25 +0800362/* group add. inode update and the new group update. */
363#define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
364
Mark Fashehccd979b2005-12-15 14:31:24 -0800365/* get one bit out of a suballocator: dinode + group descriptor +
366 * prev. group desc. if we relink. */
367#define OCFS2_SUBALLOC_ALLOC (3)
368
Jan Karaa90714c2008-10-09 19:38:40 +0200369static inline int ocfs2_inline_to_extents_credits(struct super_block *sb)
370{
371 return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS +
372 ocfs2_quota_trans_credits(sb);
373}
Mark Fasheh1afc32b2007-09-07 14:46:51 -0700374
Mark Fashehccd979b2005-12-15 14:31:24 -0800375/* dinode + group descriptor update. We don't relink on free yet. */
376#define OCFS2_SUBALLOC_FREE (2)
377
378#define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS
379#define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE \
380 + OCFS2_TRUNCATE_LOG_UPDATE)
381
Jan Karaa90714c2008-10-09 19:38:40 +0200382static inline int ocfs2_remove_extent_credits(struct super_block *sb)
383{
384 return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS +
385 ocfs2_quota_trans_credits(sb);
386}
Mark Fasheh063c4562007-07-03 13:34:11 -0700387
Mark Fashehccd979b2005-12-15 14:31:24 -0800388/* data block for new dir/symlink, 2 for bitmap updates (bitmap fe +
Mark Fashehe7c17e42009-01-29 18:17:46 -0800389 * bitmap block for the new bit) dx_root update for free list */
390#define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + 2 + 1)
Mark Fashehccd979b2005-12-15 14:31:24 -0800391
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800392static inline int ocfs2_add_dir_index_credits(struct super_block *sb)
Jan Karaa90714c2008-10-09 19:38:40 +0200393{
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800394 /* 1 block for index, 2 allocs (data, metadata), 1 clusters
395 * worth of blocks for initial extent. */
396 return 1 + 2 * OCFS2_SUBALLOC_ALLOC +
397 ocfs2_clusters_to_blocks(sb, 1);
398}
399
400/* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode
401 * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr
402 * blocks + quota update */
403static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir,
404 int xattr_credits)
405{
406 int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS;
407
408 if (is_dir)
409 dir_credits += ocfs2_add_dir_index_credits(sb);
410
411 return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits +
Jan Karaa90714c2008-10-09 19:38:40 +0200412 ocfs2_quota_trans_credits(sb);
413}
Mark Fashehccd979b2005-12-15 14:31:24 -0800414
415/* local alloc metadata change + main bitmap updates */
416#define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS \
417 + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE)
418
419/* used when we don't need an allocation change for a dir extend. One
420 * for the dinode, one for the new block. */
421#define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2)
422
Jan Karaa90714c2008-10-09 19:38:40 +0200423/* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota
Mark Fashehe7c17e42009-01-29 18:17:46 -0800424 * update on dir + index leaf + dx root update for free list */
Jan Karaa90714c2008-10-09 19:38:40 +0200425static inline int ocfs2_link_credits(struct super_block *sb)
426{
Mark Fashehe7c17e42009-01-29 18:17:46 -0800427 return 2*OCFS2_INODE_UPDATE_CREDITS + 3 +
Jan Karaa90714c2008-10-09 19:38:40 +0200428 ocfs2_quota_trans_credits(sb);
429}
Mark Fashehccd979b2005-12-15 14:31:24 -0800430
431/* inode + dir inode (if we unlink a dir), + dir entry block + orphan
Mark Fashehe7c17e42009-01-29 18:17:46 -0800432 * dir inode link + dir inode index leaf + dir index root */
Jan Karaa90714c2008-10-09 19:38:40 +0200433static inline int ocfs2_unlink_credits(struct super_block *sb)
434{
435 /* The quota update from ocfs2_link_credits is unused here... */
Mark Fashehe7c17e42009-01-29 18:17:46 -0800436 return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb);
Jan Karaa90714c2008-10-09 19:38:40 +0200437}
Mark Fashehccd979b2005-12-15 14:31:24 -0800438
439/* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800440 * inode alloc group descriptor + orphan dir index leaf */
441#define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 3)
Mark Fashehccd979b2005-12-15 14:31:24 -0800442
443/* dinode update, old dir dinode update, new dir dinode update, old
444 * dir dir entry, new dir dir entry, dir entry update for renaming
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800445 * directory + target unlink + 3 x dir index leaves */
Jan Karaa90714c2008-10-09 19:38:40 +0200446static inline int ocfs2_rename_credits(struct super_block *sb)
447{
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800448 return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb);
Jan Karaa90714c2008-10-09 19:38:40 +0200449}
Mark Fashehccd979b2005-12-15 14:31:24 -0800450
Tiger Yangcf1d6c72008-08-18 17:11:00 +0800451/* global bitmap dinode, group desc., relinked group,
452 * suballocator dinode, group desc., relinked group,
453 * dinode, xattr block */
454#define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \
455 + OCFS2_INODE_UPDATE_CREDITS \
456 + OCFS2_XATTR_BLOCK_UPDATE_CREDITS)
457
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800458/* inode update, removal of dx root block from allocator */
459#define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
460 OCFS2_SUBALLOC_FREE)
461
Mark Fasheh4ed8a6b2008-11-24 17:02:08 -0800462static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb)
463{
464 int credits = 1 + OCFS2_SUBALLOC_ALLOC;
465
466 credits += ocfs2_clusters_to_blocks(sb, 1);
467 credits += ocfs2_quota_trans_credits(sb);
468
469 return credits;
470}
471
Tao Ma811f9332008-08-18 17:38:43 +0800472/*
473 * Please note that the caller must make sure that root_el is the root
474 * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
475 * the result may be wrong.
476 */
Mark Fashehccd979b2005-12-15 14:31:24 -0800477static inline int ocfs2_calc_extend_credits(struct super_block *sb,
Tao Ma811f9332008-08-18 17:38:43 +0800478 struct ocfs2_extent_list *root_el,
Mark Fashehccd979b2005-12-15 14:31:24 -0800479 u32 bits_wanted)
480{
Tao Ma811f9332008-08-18 17:38:43 +0800481 int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks;
Mark Fashehccd979b2005-12-15 14:31:24 -0800482
483 /* bitmap dinode, group desc. + relinked group. */
484 bitmap_blocks = OCFS2_SUBALLOC_ALLOC;
485
486 /* we might need to shift tree depth so lets assume an
487 * absolute worst case of complete fragmentation. Even with
488 * that, we only need one update for the dinode, and then
489 * however many metadata chunks needed * a remaining suballoc
490 * alloc. */
491 sysfile_bitmap_blocks = 1 +
Tao Ma811f9332008-08-18 17:38:43 +0800492 (OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(root_el);
Mark Fashehccd979b2005-12-15 14:31:24 -0800493
494 /* this does not include *new* metadata blocks, which are
Tao Ma811f9332008-08-18 17:38:43 +0800495 * accounted for in sysfile_bitmap_blocks. root_el +
Mark Fashehccd979b2005-12-15 14:31:24 -0800496 * prev. last_eb_blk + blocks along edge of tree.
497 * calc_symlink_credits passes because we just need 1
498 * credit for the dinode there. */
Tao Ma811f9332008-08-18 17:38:43 +0800499 extent_blocks = 1 + 1 + le16_to_cpu(root_el->l_tree_depth);
Mark Fashehccd979b2005-12-15 14:31:24 -0800500
Jan Karaa90714c2008-10-09 19:38:40 +0200501 return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks +
502 ocfs2_quota_trans_credits(sb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800503}
504
505static inline int ocfs2_calc_symlink_credits(struct super_block *sb)
506{
Mark Fasheh9b7895e2008-11-12 16:27:44 -0800507 int blocks = ocfs2_mknod_credits(sb, 0, 0);
Mark Fashehccd979b2005-12-15 14:31:24 -0800508
509 /* links can be longer than one block so we may update many
510 * within our single allocated extent. */
511 blocks += ocfs2_clusters_to_blocks(sb, 1);
512
Jan Karaa90714c2008-10-09 19:38:40 +0200513 return blocks + ocfs2_quota_trans_credits(sb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800514}
515
516static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
517 unsigned int cpg)
518{
519 int blocks;
520 int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1;
521 /* parent inode update + new block group header + bitmap inode update
522 + bitmap blocks affected */
523 blocks = 1 + 1 + 1 + bitmap_blocks;
524 return blocks;
525}
526
527static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
528 unsigned int clusters_to_del,
529 struct ocfs2_dinode *fe,
530 struct ocfs2_extent_list *last_el)
531{
532 /* for dinode + all headers in this pass + update to next leaf */
533 u16 next_free = le16_to_cpu(last_el->l_next_free_rec);
534 u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth);
535 int credits = 1 + tree_depth + 1;
536 int i;
537
538 i = next_free - 1;
539 BUG_ON(i < 0);
540
541 /* We may be deleting metadata blocks, so metadata alloc dinode +
542 one desc. block for each possible delete. */
543 if (tree_depth && next_free == 1 &&
Mark Fashehe48edee2007-03-07 16:46:57 -0800544 ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del)
Mark Fashehccd979b2005-12-15 14:31:24 -0800545 credits += 1 + tree_depth;
546
547 /* update to the truncate log. */
548 credits += OCFS2_TRUNCATE_LOG_UPDATE;
549
Jan Karaa90714c2008-10-09 19:38:40 +0200550 credits += ocfs2_quota_trans_credits(sb);
551
Mark Fashehccd979b2005-12-15 14:31:24 -0800552 return credits;
553}
554
Joel Becker2b4e30f2008-09-03 20:03:41 -0700555static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode)
556{
557 return jbd2_journal_file_inode(handle, &OCFS2_I(inode)->ip_jinode);
558}
559
560static inline int ocfs2_begin_ordered_truncate(struct inode *inode,
561 loff_t new_size)
562{
Jan Kara7f5aa212009-02-10 11:15:34 -0500563 return jbd2_journal_begin_ordered_truncate(
564 OCFS2_SB(inode->i_sb)->journal->j_journal,
565 &OCFS2_I(inode)->ip_jinode,
566 new_size);
Joel Becker2b4e30f2008-09-03 20:03:41 -0700567}
568
Mark Fashehccd979b2005-12-15 14:31:24 -0800569#endif /* OCFS2_JOURNAL_H */