]>
Commit | Line | Data |
---|---|---|
ccd979bd MF |
1 | /* -*- 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> | |
53ef99ca | 30 | #include <linux/jbd2.h> |
ccd979bd | 31 | |
ccd979bd MF |
32 | enum ocfs2_journal_state { |
33 | OCFS2_JOURNAL_FREE = 0, | |
34 | OCFS2_JOURNAL_LOADED, | |
35 | OCFS2_JOURNAL_IN_SHUTDOWN, | |
36 | }; | |
37 | ||
38 | struct ocfs2_super; | |
39 | struct ocfs2_dinode; | |
ccd979bd | 40 | |
96a6c64b SM |
41 | /* |
42 | * The recovery_list is a simple linked list of node numbers to recover. | |
43 | * It is protected by the recovery_lock. | |
44 | */ | |
45 | ||
46 | struct ocfs2_recovery_map { | |
47 | unsigned int rm_used; | |
48 | unsigned int *rm_entries; | |
49 | }; | |
50 | ||
51 | ||
ccd979bd MF |
52 | struct 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 | ||
79 | extern spinlock_t trans_inc_lock; | |
80 | ||
81 | /* wrap j_trans_id so we never have it equal to zero. */ | |
82 | static 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 | ||
0cf2f763 JB |
93 | static inline void ocfs2_set_ci_lock_trans(struct ocfs2_journal *journal, |
94 | struct ocfs2_caching_info *ci) | |
ccd979bd MF |
95 | { |
96 | spin_lock(&trans_inc_lock); | |
0cf2f763 | 97 | ci->ci_last_trans = journal->j_trans_id; |
ccd979bd MF |
98 | spin_unlock(&trans_inc_lock); |
99 | } | |
100 | ||
101 | /* Used to figure out whether it's safe to drop a metadata lock on an | |
0cf2f763 | 102 | * cached object. Returns true if all the object's changes have been |
ccd979bd MF |
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. */ | |
0cf2f763 | 106 | static inline int ocfs2_ci_fully_checkpointed(struct ocfs2_caching_info *ci) |
ccd979bd MF |
107 | { |
108 | int ret; | |
0cf2f763 JB |
109 | struct ocfs2_journal *journal = |
110 | OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal; | |
ccd979bd MF |
111 | |
112 | spin_lock(&trans_inc_lock); | |
0cf2f763 | 113 | ret = time_after(journal->j_trans_id, ci->ci_last_trans); |
ccd979bd MF |
114 | spin_unlock(&trans_inc_lock); |
115 | return ret; | |
116 | } | |
117 | ||
0cf2f763 JB |
118 | /* convenience function to check if an object backed by struct |
119 | * ocfs2_caching_info is still new (has never hit disk) Will do you a | |
120 | * favor and set created_trans = 0 when you've | |
121 | * been checkpointed. returns '1' if the ci is still new. */ | |
122 | static inline int ocfs2_ci_is_new(struct ocfs2_caching_info *ci) | |
ccd979bd MF |
123 | { |
124 | int ret; | |
0cf2f763 JB |
125 | struct ocfs2_journal *journal = |
126 | OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal; | |
127 | ||
128 | spin_lock(&trans_inc_lock); | |
129 | ret = !(time_after(journal->j_trans_id, ci->ci_created_trans)); | |
130 | if (!ret) | |
131 | ci->ci_created_trans = 0; | |
132 | spin_unlock(&trans_inc_lock); | |
133 | return ret; | |
134 | } | |
ccd979bd | 135 | |
0cf2f763 JB |
136 | /* Wrapper for inodes so we can check system files */ |
137 | static inline int ocfs2_inode_is_new(struct inode *inode) | |
138 | { | |
ccd979bd MF |
139 | /* System files are never "new" as they're written out by |
140 | * mkfs. This helps us early during mount, before we have the | |
141 | * journal open and j_trans_id could be junk. */ | |
142 | if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE) | |
143 | return 0; | |
0cf2f763 JB |
144 | |
145 | return ocfs2_ci_is_new(INODE_CACHE(inode)); | |
ccd979bd MF |
146 | } |
147 | ||
0cf2f763 JB |
148 | static inline void ocfs2_ci_set_new(struct ocfs2_super *osb, |
149 | struct ocfs2_caching_info *ci) | |
ccd979bd MF |
150 | { |
151 | spin_lock(&trans_inc_lock); | |
0cf2f763 | 152 | ci->ci_created_trans = osb->journal->j_trans_id; |
ccd979bd MF |
153 | spin_unlock(&trans_inc_lock); |
154 | } | |
155 | ||
ccd979bd | 156 | /* Exported only for the journal struct init code in super.c. Do not call. */ |
df152c24 | 157 | void ocfs2_orphan_scan_init(struct ocfs2_super *osb); |
8b712cd5 | 158 | void ocfs2_orphan_scan_start(struct ocfs2_super *osb); |
83273932 SE |
159 | void ocfs2_orphan_scan_stop(struct ocfs2_super *osb); |
160 | void ocfs2_orphan_scan_exit(struct ocfs2_super *osb); | |
161 | ||
c4028958 | 162 | void ocfs2_complete_recovery(struct work_struct *work); |
553abd04 JB |
163 | void ocfs2_wait_for_recovery(struct ocfs2_super *osb); |
164 | ||
165 | int ocfs2_recovery_init(struct ocfs2_super *osb); | |
166 | void ocfs2_recovery_exit(struct ocfs2_super *osb); | |
ccd979bd | 167 | |
9140db04 | 168 | int ocfs2_compute_replay_slots(struct ocfs2_super *osb); |
ccd979bd MF |
169 | /* |
170 | * Journal Control: | |
171 | * Initialize, Load, Shutdown, Wipe a journal. | |
172 | * | |
173 | * ocfs2_journal_init - Initialize journal structures in the OSB. | |
174 | * ocfs2_journal_load - Load the given journal off disk. Replay it if | |
175 | * there's transactions still in there. | |
176 | * ocfs2_journal_shutdown - Shutdown a journal, this will flush all | |
177 | * uncommitted, uncheckpointed transactions. | |
178 | * ocfs2_journal_wipe - Wipe transactions from a journal. Optionally | |
179 | * zero out each block. | |
180 | * ocfs2_recovery_thread - Perform recovery on a node. osb is our own osb. | |
181 | * ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat | |
182 | * event on. | |
183 | * ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint. | |
184 | */ | |
185 | void ocfs2_set_journal_params(struct ocfs2_super *osb); | |
186 | int ocfs2_journal_init(struct ocfs2_journal *journal, | |
187 | int *dirty); | |
188 | void ocfs2_journal_shutdown(struct ocfs2_super *osb); | |
189 | int ocfs2_journal_wipe(struct ocfs2_journal *journal, | |
190 | int full); | |
539d8264 SM |
191 | int ocfs2_journal_load(struct ocfs2_journal *journal, int local, |
192 | int replayed); | |
ccd979bd MF |
193 | int ocfs2_check_journals_nolocks(struct ocfs2_super *osb); |
194 | void ocfs2_recovery_thread(struct ocfs2_super *osb, | |
195 | int node_num); | |
196 | int ocfs2_mark_dead_nodes(struct ocfs2_super *osb); | |
197 | void ocfs2_complete_mount_recovery(struct ocfs2_super *osb); | |
2205363d | 198 | void ocfs2_complete_quota_recovery(struct ocfs2_super *osb); |
ccd979bd MF |
199 | |
200 | static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb) | |
201 | { | |
202 | atomic_set(&osb->needs_checkpoint, 1); | |
203 | wake_up(&osb->checkpoint_event); | |
204 | } | |
205 | ||
206 | static inline void ocfs2_checkpoint_inode(struct inode *inode) | |
207 | { | |
208 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
209 | ||
c271c5c2 SM |
210 | if (ocfs2_mount_local(osb)) |
211 | return; | |
212 | ||
0cf2f763 | 213 | if (!ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))) { |
ccd979bd MF |
214 | /* WARNING: This only kicks off a single |
215 | * checkpoint. If someone races you and adds more | |
216 | * metadata to the journal, you won't know, and will | |
217 | * wind up waiting *alot* longer than necessary. Right | |
218 | * now we only use this in clear_inode so that's | |
219 | * OK. */ | |
220 | ocfs2_start_checkpoint(osb); | |
221 | ||
222 | wait_event(osb->journal->j_checkpointed, | |
0cf2f763 | 223 | ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))); |
ccd979bd MF |
224 | } |
225 | } | |
226 | ||
227 | /* | |
228 | * Transaction Handling: | |
229 | * Manage the lifetime of a transaction handle. | |
230 | * | |
ccd979bd MF |
231 | * ocfs2_start_trans - Begin a transaction. Give it an upper estimate of |
232 | * the number of blocks that will be changed during | |
233 | * this handle. | |
1fabe148 MF |
234 | * ocfs2_commit_trans - Complete a handle. It might return -EIO if |
235 | * the journal was aborted. The majority of paths don't | |
236 | * check the return value as an error there comes too | |
237 | * late to do anything (and will be picked up in a | |
238 | * later transaction). | |
ccd979bd MF |
239 | * ocfs2_extend_trans - Extend a handle by nblocks credits. This may |
240 | * commit the handle to disk in the process, but will | |
241 | * not release any locks taken during the transaction. | |
50655ae9 | 242 | * ocfs2_journal_access* - Notify the handle that we want to journal this |
ccd979bd MF |
243 | * buffer. Will have to call ocfs2_journal_dirty once |
244 | * we've actually dirtied it. Type is one of . or . | |
50655ae9 JB |
245 | * Always call the specific flavor of |
246 | * ocfs2_journal_access_*() unless you intend to | |
247 | * manage the checksum by hand. | |
ccd979bd | 248 | * ocfs2_journal_dirty - Mark a journalled buffer as having dirty data. |
2b4e30fb JB |
249 | * ocfs2_jbd2_file_inode - Mark an inode so that its data goes out before |
250 | * the current handle commits. | |
ccd979bd MF |
251 | */ |
252 | ||
253 | /* You must always start_trans with a number of buffs > 0, but it's | |
254 | * perfectly legal to go through an entire transaction without having | |
255 | * dirtied any buffers. */ | |
1fabe148 | 256 | handle_t *ocfs2_start_trans(struct ocfs2_super *osb, |
ccd979bd | 257 | int max_buffs); |
1fabe148 MF |
258 | int ocfs2_commit_trans(struct ocfs2_super *osb, |
259 | handle_t *handle); | |
1fc58146 | 260 | int ocfs2_extend_trans(handle_t *handle, int nblocks); |
ccd979bd MF |
261 | |
262 | /* | |
263 | * Create access is for when we get a newly created buffer and we're | |
264 | * not gonna read it off disk, but rather fill it ourselves. Right | |
265 | * now, we don't do anything special with this (it turns into a write | |
266 | * request), but this is a good placeholder in case we do... | |
267 | * | |
268 | * Write access is for when we read a block off disk and are going to | |
269 | * modify it. This way the journalling layer knows it may need to make | |
270 | * a copy of that block (if it's part of another, uncommitted | |
271 | * transaction) before we do so. | |
272 | */ | |
273 | #define OCFS2_JOURNAL_ACCESS_CREATE 0 | |
274 | #define OCFS2_JOURNAL_ACCESS_WRITE 1 | |
275 | #define OCFS2_JOURNAL_ACCESS_UNDO 2 | |
276 | ||
13723d00 | 277 | |
50655ae9 | 278 | /* ocfs2_inode */ |
0cf2f763 | 279 | int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci, |
13723d00 | 280 | struct buffer_head *bh, int type); |
50655ae9 | 281 | /* ocfs2_extent_block */ |
0cf2f763 | 282 | int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 | 283 | struct buffer_head *bh, int type); |
93c97087 TM |
284 | /* ocfs2_refcount_block */ |
285 | int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci, | |
286 | struct buffer_head *bh, int type); | |
50655ae9 | 287 | /* ocfs2_group_desc */ |
0cf2f763 | 288 | int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
289 | struct buffer_head *bh, int type); |
290 | /* ocfs2_xattr_block */ | |
0cf2f763 | 291 | int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
292 | struct buffer_head *bh, int type); |
293 | /* quota blocks */ | |
0cf2f763 | 294 | int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
295 | struct buffer_head *bh, int type); |
296 | /* dirblock */ | |
0cf2f763 | 297 | int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 | 298 | struct buffer_head *bh, int type); |
9b7895ef | 299 | /* ocfs2_dx_root_block */ |
0cf2f763 | 300 | int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci, |
9b7895ef MF |
301 | struct buffer_head *bh, int type); |
302 | /* ocfs2_dx_leaf */ | |
0cf2f763 | 303 | int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci, |
9b7895ef | 304 | struct buffer_head *bh, int type); |
50655ae9 | 305 | /* Anything that has no ecc */ |
0cf2f763 | 306 | int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
307 | struct buffer_head *bh, int type); |
308 | ||
ccd979bd MF |
309 | /* |
310 | * A word about the journal_access/journal_dirty "dance". It is | |
311 | * entirely legal to journal_access a buffer more than once (as long | |
312 | * as the access type is the same -- I'm not sure what will happen if | |
313 | * access type is different but this should never happen anyway) It is | |
314 | * also legal to journal_dirty a buffer more than once. In fact, you | |
315 | * can even journal_access a buffer after you've done a | |
316 | * journal_access/journal_dirty pair. The only thing you cannot do | |
317 | * however, is journal_dirty a buffer which you haven't yet passed to | |
318 | * journal_access at least once. | |
319 | * | |
320 | * That said, 99% of the time this doesn't matter and this is what the | |
321 | * path looks like: | |
322 | * | |
323 | * <read a bh> | |
324 | * ocfs2_journal_access(handle, bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
325 | * <modify the bh> | |
326 | * ocfs2_journal_dirty(handle, bh); | |
327 | */ | |
1fabe148 | 328 | int ocfs2_journal_dirty(handle_t *handle, |
ccd979bd | 329 | struct buffer_head *bh); |
ccd979bd MF |
330 | |
331 | /* | |
332 | * Credit Macros: | |
333 | * Convenience macros to calculate number of credits needed. | |
334 | * | |
335 | * For convenience sake, I have a set of macros here which calculate | |
336 | * the *maximum* number of sectors which will be changed for various | |
337 | * metadata updates. | |
338 | */ | |
339 | ||
340 | /* simple file updates like chmod, etc. */ | |
341 | #define OCFS2_INODE_UPDATE_CREDITS 1 | |
342 | ||
cf1d6c76 TY |
343 | /* extended attribute block update */ |
344 | #define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1 | |
345 | ||
0584974a JK |
346 | /* Update of a single quota block */ |
347 | #define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1 | |
348 | ||
a90714c1 | 349 | /* global quotafile inode update, data block */ |
0584974a JK |
350 | #define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \ |
351 | OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 | 352 | |
0584974a | 353 | #define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS |
a90714c1 JK |
354 | /* |
355 | * The two writes below can accidentally see global info dirty due | |
356 | * to set_info() quotactl so make them prepared for the writes. | |
357 | */ | |
358 | /* quota data block, global info */ | |
359 | /* Write to local quota file */ | |
0584974a JK |
360 | #define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ |
361 | OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 JK |
362 | |
363 | /* global quota data block, local quota data block, global quota inode, | |
364 | * global quota info */ | |
0584974a JK |
365 | #define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ |
366 | 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 JK |
367 | |
368 | static inline int ocfs2_quota_trans_credits(struct super_block *sb) | |
369 | { | |
370 | int credits = 0; | |
371 | ||
372 | if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) | |
373 | credits += OCFS2_QWRITE_CREDITS; | |
374 | if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) | |
375 | credits += OCFS2_QWRITE_CREDITS; | |
376 | return credits; | |
377 | } | |
378 | ||
d659072f TM |
379 | /* group extend. inode update and last group update. */ |
380 | #define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
381 | ||
7909f2bf TM |
382 | /* group add. inode update and the new group update. */ |
383 | #define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
384 | ||
ccd979bd MF |
385 | /* get one bit out of a suballocator: dinode + group descriptor + |
386 | * prev. group desc. if we relink. */ | |
387 | #define OCFS2_SUBALLOC_ALLOC (3) | |
388 | ||
a90714c1 JK |
389 | static inline int ocfs2_inline_to_extents_credits(struct super_block *sb) |
390 | { | |
391 | return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS + | |
392 | ocfs2_quota_trans_credits(sb); | |
393 | } | |
1afc32b9 | 394 | |
ccd979bd MF |
395 | /* dinode + group descriptor update. We don't relink on free yet. */ |
396 | #define OCFS2_SUBALLOC_FREE (2) | |
397 | ||
398 | #define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS | |
399 | #define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE \ | |
400 | + OCFS2_TRUNCATE_LOG_UPDATE) | |
401 | ||
a90714c1 JK |
402 | static inline int ocfs2_remove_extent_credits(struct super_block *sb) |
403 | { | |
404 | return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS + | |
405 | ocfs2_quota_trans_credits(sb); | |
406 | } | |
063c4561 | 407 | |
ccd979bd | 408 | /* data block for new dir/symlink, 2 for bitmap updates (bitmap fe + |
e7c17e43 MF |
409 | * bitmap block for the new bit) dx_root update for free list */ |
410 | #define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + 2 + 1) | |
ccd979bd | 411 | |
9b7895ef MF |
412 | static inline int ocfs2_add_dir_index_credits(struct super_block *sb) |
413 | { | |
414 | /* 1 block for index, 2 allocs (data, metadata), 1 clusters | |
415 | * worth of blocks for initial extent. */ | |
416 | return 1 + 2 * OCFS2_SUBALLOC_ALLOC + | |
417 | ocfs2_clusters_to_blocks(sb, 1); | |
418 | } | |
419 | ||
420 | /* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode | |
421 | * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr | |
422 | * blocks + quota update */ | |
423 | static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir, | |
424 | int xattr_credits) | |
a90714c1 | 425 | { |
9b7895ef MF |
426 | int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS; |
427 | ||
428 | if (is_dir) | |
429 | dir_credits += ocfs2_add_dir_index_credits(sb); | |
430 | ||
431 | return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits + | |
a90714c1 JK |
432 | ocfs2_quota_trans_credits(sb); |
433 | } | |
ccd979bd MF |
434 | |
435 | /* local alloc metadata change + main bitmap updates */ | |
436 | #define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS \ | |
437 | + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE) | |
438 | ||
439 | /* used when we don't need an allocation change for a dir extend. One | |
440 | * for the dinode, one for the new block. */ | |
441 | #define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2) | |
442 | ||
a90714c1 | 443 | /* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota |
e7c17e43 | 444 | * update on dir + index leaf + dx root update for free list */ |
a90714c1 JK |
445 | static inline int ocfs2_link_credits(struct super_block *sb) |
446 | { | |
e7c17e43 | 447 | return 2*OCFS2_INODE_UPDATE_CREDITS + 3 + |
a90714c1 JK |
448 | ocfs2_quota_trans_credits(sb); |
449 | } | |
ccd979bd MF |
450 | |
451 | /* inode + dir inode (if we unlink a dir), + dir entry block + orphan | |
e7c17e43 | 452 | * dir inode link + dir inode index leaf + dir index root */ |
a90714c1 JK |
453 | static inline int ocfs2_unlink_credits(struct super_block *sb) |
454 | { | |
455 | /* The quota update from ocfs2_link_credits is unused here... */ | |
e7c17e43 | 456 | return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb); |
a90714c1 | 457 | } |
ccd979bd MF |
458 | |
459 | /* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry + | |
dfa13f39 JB |
460 | * inode alloc group descriptor + orphan dir index root + |
461 | * orphan dir index leaf */ | |
462 | #define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 4) | |
ccd979bd MF |
463 | |
464 | /* dinode update, old dir dinode update, new dir dinode update, old | |
465 | * dir dir entry, new dir dir entry, dir entry update for renaming | |
9b7895ef | 466 | * directory + target unlink + 3 x dir index leaves */ |
a90714c1 JK |
467 | static inline int ocfs2_rename_credits(struct super_block *sb) |
468 | { | |
9b7895ef | 469 | return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb); |
a90714c1 | 470 | } |
ccd979bd | 471 | |
cf1d6c76 TY |
472 | /* global bitmap dinode, group desc., relinked group, |
473 | * suballocator dinode, group desc., relinked group, | |
474 | * dinode, xattr block */ | |
475 | #define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \ | |
476 | + OCFS2_INODE_UPDATE_CREDITS \ | |
477 | + OCFS2_XATTR_BLOCK_UPDATE_CREDITS) | |
478 | ||
9b7895ef MF |
479 | /* inode update, removal of dx root block from allocator */ |
480 | #define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \ | |
481 | OCFS2_SUBALLOC_FREE) | |
482 | ||
4ed8a6bb MF |
483 | static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb) |
484 | { | |
485 | int credits = 1 + OCFS2_SUBALLOC_ALLOC; | |
486 | ||
487 | credits += ocfs2_clusters_to_blocks(sb, 1); | |
488 | credits += ocfs2_quota_trans_credits(sb); | |
489 | ||
490 | return credits; | |
491 | } | |
492 | ||
8bf396de TM |
493 | /* inode update, new refcount block and its allocation credits. */ |
494 | #define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \ | |
495 | + OCFS2_SUBALLOC_ALLOC) | |
496 | ||
497 | /* inode and the refcount block update. */ | |
498 | #define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
499 | ||
500 | /* | |
501 | * inode and the refcount block update. | |
502 | * It doesn't include the credits for sub alloc change. | |
503 | * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added. | |
504 | */ | |
505 | #define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
506 | ||
bcbbb24a TM |
507 | /* 2 metadata alloc, 2 new blocks and root refcount block */ |
508 | #define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3) | |
509 | ||
811f933d TM |
510 | /* |
511 | * Please note that the caller must make sure that root_el is the root | |
512 | * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise | |
513 | * the result may be wrong. | |
514 | */ | |
ccd979bd | 515 | static inline int ocfs2_calc_extend_credits(struct super_block *sb, |
811f933d | 516 | struct ocfs2_extent_list *root_el, |
ccd979bd MF |
517 | u32 bits_wanted) |
518 | { | |
811f933d | 519 | int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks; |
ccd979bd MF |
520 | |
521 | /* bitmap dinode, group desc. + relinked group. */ | |
522 | bitmap_blocks = OCFS2_SUBALLOC_ALLOC; | |
523 | ||
524 | /* we might need to shift tree depth so lets assume an | |
525 | * absolute worst case of complete fragmentation. Even with | |
526 | * that, we only need one update for the dinode, and then | |
527 | * however many metadata chunks needed * a remaining suballoc | |
528 | * alloc. */ | |
529 | sysfile_bitmap_blocks = 1 + | |
811f933d | 530 | (OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(root_el); |
ccd979bd MF |
531 | |
532 | /* this does not include *new* metadata blocks, which are | |
811f933d | 533 | * accounted for in sysfile_bitmap_blocks. root_el + |
ccd979bd MF |
534 | * prev. last_eb_blk + blocks along edge of tree. |
535 | * calc_symlink_credits passes because we just need 1 | |
536 | * credit for the dinode there. */ | |
811f933d | 537 | extent_blocks = 1 + 1 + le16_to_cpu(root_el->l_tree_depth); |
ccd979bd | 538 | |
a90714c1 JK |
539 | return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks + |
540 | ocfs2_quota_trans_credits(sb); | |
ccd979bd MF |
541 | } |
542 | ||
543 | static inline int ocfs2_calc_symlink_credits(struct super_block *sb) | |
544 | { | |
9b7895ef | 545 | int blocks = ocfs2_mknod_credits(sb, 0, 0); |
ccd979bd MF |
546 | |
547 | /* links can be longer than one block so we may update many | |
548 | * within our single allocated extent. */ | |
549 | blocks += ocfs2_clusters_to_blocks(sb, 1); | |
550 | ||
a90714c1 | 551 | return blocks + ocfs2_quota_trans_credits(sb); |
ccd979bd MF |
552 | } |
553 | ||
554 | static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb, | |
555 | unsigned int cpg) | |
556 | { | |
557 | int blocks; | |
558 | int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1; | |
559 | /* parent inode update + new block group header + bitmap inode update | |
560 | + bitmap blocks affected */ | |
561 | blocks = 1 + 1 + 1 + bitmap_blocks; | |
562 | return blocks; | |
563 | } | |
564 | ||
565 | static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb, | |
566 | unsigned int clusters_to_del, | |
567 | struct ocfs2_dinode *fe, | |
568 | struct ocfs2_extent_list *last_el) | |
569 | { | |
570 | /* for dinode + all headers in this pass + update to next leaf */ | |
571 | u16 next_free = le16_to_cpu(last_el->l_next_free_rec); | |
572 | u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth); | |
573 | int credits = 1 + tree_depth + 1; | |
574 | int i; | |
575 | ||
576 | i = next_free - 1; | |
577 | BUG_ON(i < 0); | |
578 | ||
579 | /* We may be deleting metadata blocks, so metadata alloc dinode + | |
580 | one desc. block for each possible delete. */ | |
581 | if (tree_depth && next_free == 1 && | |
e48edee2 | 582 | ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del) |
ccd979bd MF |
583 | credits += 1 + tree_depth; |
584 | ||
585 | /* update to the truncate log. */ | |
586 | credits += OCFS2_TRUNCATE_LOG_UPDATE; | |
587 | ||
a90714c1 JK |
588 | credits += ocfs2_quota_trans_credits(sb); |
589 | ||
ccd979bd MF |
590 | return credits; |
591 | } | |
592 | ||
2b4e30fb JB |
593 | static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode) |
594 | { | |
595 | return jbd2_journal_file_inode(handle, &OCFS2_I(inode)->ip_jinode); | |
596 | } | |
597 | ||
598 | static inline int ocfs2_begin_ordered_truncate(struct inode *inode, | |
599 | loff_t new_size) | |
600 | { | |
7f5aa215 JK |
601 | return jbd2_journal_begin_ordered_truncate( |
602 | OCFS2_SB(inode->i_sb)->journal->j_journal, | |
603 | &OCFS2_I(inode)->ip_jinode, | |
604 | new_size); | |
2b4e30fb JB |
605 | } |
606 | ||
ccd979bd | 607 | #endif /* OCFS2_JOURNAL_H */ |