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. */ | |
83fd9c7f | 70 | spinlock_t j_lock; |
ccd979bd MF |
71 | unsigned long j_trans_id; |
72 | struct rw_semaphore j_trans_barrier; | |
73 | wait_queue_head_t j_checkpointed; | |
74 | ||
83fd9c7f | 75 | /* both fields protected by j_lock*/ |
ccd979bd MF |
76 | struct list_head j_la_cleanups; |
77 | struct work_struct j_recovery_work; | |
78 | }; | |
79 | ||
80 | extern spinlock_t trans_inc_lock; | |
81 | ||
82 | /* wrap j_trans_id so we never have it equal to zero. */ | |
83 | static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j) | |
84 | { | |
85 | unsigned long old_id; | |
86 | spin_lock(&trans_inc_lock); | |
87 | old_id = j->j_trans_id++; | |
88 | if (unlikely(!j->j_trans_id)) | |
89 | j->j_trans_id = 1; | |
90 | spin_unlock(&trans_inc_lock); | |
91 | return old_id; | |
92 | } | |
93 | ||
0cf2f763 JB |
94 | static inline void ocfs2_set_ci_lock_trans(struct ocfs2_journal *journal, |
95 | struct ocfs2_caching_info *ci) | |
ccd979bd MF |
96 | { |
97 | spin_lock(&trans_inc_lock); | |
0cf2f763 | 98 | ci->ci_last_trans = journal->j_trans_id; |
ccd979bd MF |
99 | spin_unlock(&trans_inc_lock); |
100 | } | |
101 | ||
102 | /* Used to figure out whether it's safe to drop a metadata lock on an | |
0cf2f763 | 103 | * cached object. Returns true if all the object's changes have been |
ccd979bd MF |
104 | * checkpointed to disk. You should be holding the spinlock on the |
105 | * metadata lock while calling this to be sure that nobody can take | |
106 | * the lock and put it on another transaction. */ | |
0cf2f763 | 107 | static inline int ocfs2_ci_fully_checkpointed(struct ocfs2_caching_info *ci) |
ccd979bd MF |
108 | { |
109 | int ret; | |
0cf2f763 JB |
110 | struct ocfs2_journal *journal = |
111 | OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal; | |
ccd979bd MF |
112 | |
113 | spin_lock(&trans_inc_lock); | |
0cf2f763 | 114 | ret = time_after(journal->j_trans_id, ci->ci_last_trans); |
ccd979bd MF |
115 | spin_unlock(&trans_inc_lock); |
116 | return ret; | |
117 | } | |
118 | ||
0cf2f763 JB |
119 | /* convenience function to check if an object backed by struct |
120 | * ocfs2_caching_info is still new (has never hit disk) Will do you a | |
121 | * favor and set created_trans = 0 when you've | |
122 | * been checkpointed. returns '1' if the ci is still new. */ | |
123 | static inline int ocfs2_ci_is_new(struct ocfs2_caching_info *ci) | |
ccd979bd MF |
124 | { |
125 | int ret; | |
0cf2f763 JB |
126 | struct ocfs2_journal *journal = |
127 | OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal; | |
128 | ||
129 | spin_lock(&trans_inc_lock); | |
130 | ret = !(time_after(journal->j_trans_id, ci->ci_created_trans)); | |
131 | if (!ret) | |
132 | ci->ci_created_trans = 0; | |
133 | spin_unlock(&trans_inc_lock); | |
134 | return ret; | |
135 | } | |
ccd979bd | 136 | |
0cf2f763 JB |
137 | /* Wrapper for inodes so we can check system files */ |
138 | static inline int ocfs2_inode_is_new(struct inode *inode) | |
139 | { | |
ccd979bd MF |
140 | /* System files are never "new" as they're written out by |
141 | * mkfs. This helps us early during mount, before we have the | |
142 | * journal open and j_trans_id could be junk. */ | |
143 | if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE) | |
144 | return 0; | |
0cf2f763 JB |
145 | |
146 | return ocfs2_ci_is_new(INODE_CACHE(inode)); | |
ccd979bd MF |
147 | } |
148 | ||
0cf2f763 JB |
149 | static inline void ocfs2_ci_set_new(struct ocfs2_super *osb, |
150 | struct ocfs2_caching_info *ci) | |
ccd979bd MF |
151 | { |
152 | spin_lock(&trans_inc_lock); | |
0cf2f763 | 153 | ci->ci_created_trans = osb->journal->j_trans_id; |
ccd979bd MF |
154 | spin_unlock(&trans_inc_lock); |
155 | } | |
156 | ||
ccd979bd | 157 | /* Exported only for the journal struct init code in super.c. Do not call. */ |
df152c24 | 158 | void ocfs2_orphan_scan_init(struct ocfs2_super *osb); |
8b712cd5 | 159 | void ocfs2_orphan_scan_start(struct ocfs2_super *osb); |
83273932 SE |
160 | void ocfs2_orphan_scan_stop(struct ocfs2_super *osb); |
161 | void ocfs2_orphan_scan_exit(struct ocfs2_super *osb); | |
162 | ||
c4028958 | 163 | void ocfs2_complete_recovery(struct work_struct *work); |
553abd04 JB |
164 | void ocfs2_wait_for_recovery(struct ocfs2_super *osb); |
165 | ||
166 | int ocfs2_recovery_init(struct ocfs2_super *osb); | |
167 | void ocfs2_recovery_exit(struct ocfs2_super *osb); | |
ccd979bd | 168 | |
9140db04 | 169 | int ocfs2_compute_replay_slots(struct ocfs2_super *osb); |
ccd979bd MF |
170 | /* |
171 | * Journal Control: | |
172 | * Initialize, Load, Shutdown, Wipe a journal. | |
173 | * | |
174 | * ocfs2_journal_init - Initialize journal structures in the OSB. | |
175 | * ocfs2_journal_load - Load the given journal off disk. Replay it if | |
176 | * there's transactions still in there. | |
177 | * ocfs2_journal_shutdown - Shutdown a journal, this will flush all | |
178 | * uncommitted, uncheckpointed transactions. | |
179 | * ocfs2_journal_wipe - Wipe transactions from a journal. Optionally | |
180 | * zero out each block. | |
181 | * ocfs2_recovery_thread - Perform recovery on a node. osb is our own osb. | |
182 | * ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat | |
183 | * event on. | |
184 | * ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint. | |
185 | */ | |
186 | void ocfs2_set_journal_params(struct ocfs2_super *osb); | |
187 | int ocfs2_journal_init(struct ocfs2_journal *journal, | |
188 | int *dirty); | |
189 | void ocfs2_journal_shutdown(struct ocfs2_super *osb); | |
190 | int ocfs2_journal_wipe(struct ocfs2_journal *journal, | |
191 | int full); | |
539d8264 SM |
192 | int ocfs2_journal_load(struct ocfs2_journal *journal, int local, |
193 | int replayed); | |
ccd979bd MF |
194 | int ocfs2_check_journals_nolocks(struct ocfs2_super *osb); |
195 | void ocfs2_recovery_thread(struct ocfs2_super *osb, | |
196 | int node_num); | |
197 | int ocfs2_mark_dead_nodes(struct ocfs2_super *osb); | |
198 | void ocfs2_complete_mount_recovery(struct ocfs2_super *osb); | |
2205363d | 199 | void ocfs2_complete_quota_recovery(struct ocfs2_super *osb); |
ccd979bd MF |
200 | |
201 | static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb) | |
202 | { | |
ccd979bd MF |
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 | |
25985edc | 217 | * wind up waiting *a lot* longer than necessary. Right |
ccd979bd MF |
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); |
2b1e55c3 YL |
261 | int ocfs2_allocate_extend_trans(handle_t *handle, |
262 | int thresh); | |
263 | ||
264 | /* | |
265 | * Define an arbitrary limit for the amount of data we will anticipate | |
266 | * writing to any given transaction. For unbounded transactions such as | |
267 | * fallocate(2) we can write more than this, but we always | |
268 | * start off at the maximum transaction size and grow the transaction | |
269 | * optimistically as we go. | |
270 | */ | |
271 | #define OCFS2_MAX_TRANS_DATA 64U | |
ccd979bd MF |
272 | |
273 | /* | |
274 | * Create access is for when we get a newly created buffer and we're | |
275 | * not gonna read it off disk, but rather fill it ourselves. Right | |
276 | * now, we don't do anything special with this (it turns into a write | |
277 | * request), but this is a good placeholder in case we do... | |
278 | * | |
279 | * Write access is for when we read a block off disk and are going to | |
280 | * modify it. This way the journalling layer knows it may need to make | |
281 | * a copy of that block (if it's part of another, uncommitted | |
282 | * transaction) before we do so. | |
283 | */ | |
284 | #define OCFS2_JOURNAL_ACCESS_CREATE 0 | |
285 | #define OCFS2_JOURNAL_ACCESS_WRITE 1 | |
286 | #define OCFS2_JOURNAL_ACCESS_UNDO 2 | |
287 | ||
13723d00 | 288 | |
50655ae9 | 289 | /* ocfs2_inode */ |
0cf2f763 | 290 | int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci, |
13723d00 | 291 | struct buffer_head *bh, int type); |
50655ae9 | 292 | /* ocfs2_extent_block */ |
0cf2f763 | 293 | int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 | 294 | struct buffer_head *bh, int type); |
93c97087 TM |
295 | /* ocfs2_refcount_block */ |
296 | int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci, | |
297 | struct buffer_head *bh, int type); | |
50655ae9 | 298 | /* ocfs2_group_desc */ |
0cf2f763 | 299 | int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
300 | struct buffer_head *bh, int type); |
301 | /* ocfs2_xattr_block */ | |
0cf2f763 | 302 | int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
303 | struct buffer_head *bh, int type); |
304 | /* quota blocks */ | |
0cf2f763 | 305 | int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
306 | struct buffer_head *bh, int type); |
307 | /* dirblock */ | |
0cf2f763 | 308 | int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 | 309 | struct buffer_head *bh, int type); |
9b7895ef | 310 | /* ocfs2_dx_root_block */ |
0cf2f763 | 311 | int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci, |
9b7895ef MF |
312 | struct buffer_head *bh, int type); |
313 | /* ocfs2_dx_leaf */ | |
0cf2f763 | 314 | int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci, |
9b7895ef | 315 | struct buffer_head *bh, int type); |
50655ae9 | 316 | /* Anything that has no ecc */ |
0cf2f763 | 317 | int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
318 | struct buffer_head *bh, int type); |
319 | ||
ccd979bd MF |
320 | /* |
321 | * A word about the journal_access/journal_dirty "dance". It is | |
322 | * entirely legal to journal_access a buffer more than once (as long | |
323 | * as the access type is the same -- I'm not sure what will happen if | |
324 | * access type is different but this should never happen anyway) It is | |
325 | * also legal to journal_dirty a buffer more than once. In fact, you | |
326 | * can even journal_access a buffer after you've done a | |
327 | * journal_access/journal_dirty pair. The only thing you cannot do | |
328 | * however, is journal_dirty a buffer which you haven't yet passed to | |
329 | * journal_access at least once. | |
330 | * | |
331 | * That said, 99% of the time this doesn't matter and this is what the | |
332 | * path looks like: | |
333 | * | |
334 | * <read a bh> | |
335 | * ocfs2_journal_access(handle, bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
336 | * <modify the bh> | |
337 | * ocfs2_journal_dirty(handle, bh); | |
338 | */ | |
ec20cec7 | 339 | void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh); |
ccd979bd MF |
340 | |
341 | /* | |
342 | * Credit Macros: | |
343 | * Convenience macros to calculate number of credits needed. | |
344 | * | |
345 | * For convenience sake, I have a set of macros here which calculate | |
346 | * the *maximum* number of sectors which will be changed for various | |
347 | * metadata updates. | |
348 | */ | |
349 | ||
350 | /* simple file updates like chmod, etc. */ | |
351 | #define OCFS2_INODE_UPDATE_CREDITS 1 | |
352 | ||
cf1d6c76 TY |
353 | /* extended attribute block update */ |
354 | #define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1 | |
355 | ||
0584974a JK |
356 | /* Update of a single quota block */ |
357 | #define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1 | |
358 | ||
a90714c1 | 359 | /* global quotafile inode update, data block */ |
0584974a JK |
360 | #define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \ |
361 | OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 | 362 | |
0584974a | 363 | #define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS |
a90714c1 JK |
364 | /* |
365 | * The two writes below can accidentally see global info dirty due | |
366 | * to set_info() quotactl so make them prepared for the writes. | |
367 | */ | |
368 | /* quota data block, global info */ | |
369 | /* Write to local quota file */ | |
0584974a JK |
370 | #define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ |
371 | OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 JK |
372 | |
373 | /* global quota data block, local quota data block, global quota inode, | |
374 | * global quota info */ | |
0584974a JK |
375 | #define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ |
376 | 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 JK |
377 | |
378 | static inline int ocfs2_quota_trans_credits(struct super_block *sb) | |
379 | { | |
380 | int credits = 0; | |
381 | ||
382 | if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) | |
383 | credits += OCFS2_QWRITE_CREDITS; | |
384 | if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) | |
385 | credits += OCFS2_QWRITE_CREDITS; | |
386 | return credits; | |
387 | } | |
388 | ||
d659072f TM |
389 | /* group extend. inode update and last group update. */ |
390 | #define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
391 | ||
7909f2bf TM |
392 | /* group add. inode update and the new group update. */ |
393 | #define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
394 | ||
ccd979bd MF |
395 | /* get one bit out of a suballocator: dinode + group descriptor + |
396 | * prev. group desc. if we relink. */ | |
397 | #define OCFS2_SUBALLOC_ALLOC (3) | |
398 | ||
a90714c1 JK |
399 | static inline int ocfs2_inline_to_extents_credits(struct super_block *sb) |
400 | { | |
401 | return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS + | |
402 | ocfs2_quota_trans_credits(sb); | |
403 | } | |
1afc32b9 | 404 | |
ccd979bd MF |
405 | /* dinode + group descriptor update. We don't relink on free yet. */ |
406 | #define OCFS2_SUBALLOC_FREE (2) | |
407 | ||
408 | #define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS | |
409 | #define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE \ | |
410 | + OCFS2_TRUNCATE_LOG_UPDATE) | |
411 | ||
a90714c1 JK |
412 | static inline int ocfs2_remove_extent_credits(struct super_block *sb) |
413 | { | |
414 | return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS + | |
415 | ocfs2_quota_trans_credits(sb); | |
416 | } | |
063c4561 | 417 | |
705773a6 JK |
418 | /* data block for new dir/symlink, allocation of directory block, dx_root |
419 | * update for free list */ | |
420 | #define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + OCFS2_SUBALLOC_ALLOC + 1) | |
ccd979bd | 421 | |
9b7895ef MF |
422 | static inline int ocfs2_add_dir_index_credits(struct super_block *sb) |
423 | { | |
424 | /* 1 block for index, 2 allocs (data, metadata), 1 clusters | |
425 | * worth of blocks for initial extent. */ | |
426 | return 1 + 2 * OCFS2_SUBALLOC_ALLOC + | |
427 | ocfs2_clusters_to_blocks(sb, 1); | |
428 | } | |
429 | ||
430 | /* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode | |
431 | * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr | |
432 | * blocks + quota update */ | |
433 | static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir, | |
434 | int xattr_credits) | |
a90714c1 | 435 | { |
9b7895ef MF |
436 | int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS; |
437 | ||
438 | if (is_dir) | |
439 | dir_credits += ocfs2_add_dir_index_credits(sb); | |
440 | ||
441 | return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits + | |
a90714c1 JK |
442 | ocfs2_quota_trans_credits(sb); |
443 | } | |
ccd979bd MF |
444 | |
445 | /* local alloc metadata change + main bitmap updates */ | |
446 | #define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS \ | |
447 | + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE) | |
448 | ||
449 | /* used when we don't need an allocation change for a dir extend. One | |
450 | * for the dinode, one for the new block. */ | |
451 | #define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2) | |
452 | ||
a90714c1 | 453 | /* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota |
0393afea XH |
454 | * update on dir + index leaf + dx root update for free list + |
455 | * previous dirblock update in the free list */ | |
a90714c1 JK |
456 | static inline int ocfs2_link_credits(struct super_block *sb) |
457 | { | |
0393afea | 458 | return 2*OCFS2_INODE_UPDATE_CREDITS + 4 + |
a90714c1 JK |
459 | ocfs2_quota_trans_credits(sb); |
460 | } | |
ccd979bd MF |
461 | |
462 | /* inode + dir inode (if we unlink a dir), + dir entry block + orphan | |
e7c17e43 | 463 | * dir inode link + dir inode index leaf + dir index root */ |
a90714c1 JK |
464 | static inline int ocfs2_unlink_credits(struct super_block *sb) |
465 | { | |
466 | /* The quota update from ocfs2_link_credits is unused here... */ | |
e7c17e43 | 467 | return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb); |
a90714c1 | 468 | } |
ccd979bd MF |
469 | |
470 | /* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry + | |
dfa13f39 JB |
471 | * inode alloc group descriptor + orphan dir index root + |
472 | * orphan dir index leaf */ | |
473 | #define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 4) | |
ccd979bd | 474 | |
06ee5c75 JQ |
475 | /* dinode + orphan dir dinode + extent tree leaf block + orphan dir entry + |
476 | * orphan dir index root + orphan dir index leaf */ | |
477 | #define OCFS2_INODE_ADD_TO_ORPHAN_CREDITS (2 * OCFS2_INODE_UPDATE_CREDITS + 4) | |
478 | #define OCFS2_INODE_DEL_FROM_ORPHAN_CREDITS OCFS2_INODE_ADD_TO_ORPHAN_CREDITS | |
479 | ||
ccd979bd MF |
480 | /* dinode update, old dir dinode update, new dir dinode update, old |
481 | * dir dir entry, new dir dir entry, dir entry update for renaming | |
9b7895ef | 482 | * directory + target unlink + 3 x dir index leaves */ |
a90714c1 JK |
483 | static inline int ocfs2_rename_credits(struct super_block *sb) |
484 | { | |
9b7895ef | 485 | return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb); |
a90714c1 | 486 | } |
ccd979bd | 487 | |
cf1d6c76 TY |
488 | /* global bitmap dinode, group desc., relinked group, |
489 | * suballocator dinode, group desc., relinked group, | |
490 | * dinode, xattr block */ | |
491 | #define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \ | |
492 | + OCFS2_INODE_UPDATE_CREDITS \ | |
493 | + OCFS2_XATTR_BLOCK_UPDATE_CREDITS) | |
494 | ||
9b7895ef MF |
495 | /* inode update, removal of dx root block from allocator */ |
496 | #define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \ | |
497 | OCFS2_SUBALLOC_FREE) | |
498 | ||
4ed8a6bb MF |
499 | static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb) |
500 | { | |
501 | int credits = 1 + OCFS2_SUBALLOC_ALLOC; | |
502 | ||
503 | credits += ocfs2_clusters_to_blocks(sb, 1); | |
504 | credits += ocfs2_quota_trans_credits(sb); | |
505 | ||
506 | return credits; | |
507 | } | |
508 | ||
8bf396de TM |
509 | /* inode update, new refcount block and its allocation credits. */ |
510 | #define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \ | |
511 | + OCFS2_SUBALLOC_ALLOC) | |
512 | ||
513 | /* inode and the refcount block update. */ | |
514 | #define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
515 | ||
516 | /* | |
517 | * inode and the refcount block update. | |
518 | * It doesn't include the credits for sub alloc change. | |
519 | * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added. | |
520 | */ | |
521 | #define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
522 | ||
bcbbb24a TM |
523 | /* 2 metadata alloc, 2 new blocks and root refcount block */ |
524 | #define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3) | |
525 | ||
811f933d TM |
526 | /* |
527 | * Please note that the caller must make sure that root_el is the root | |
528 | * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise | |
529 | * the result may be wrong. | |
530 | */ | |
ccd979bd | 531 | static inline int ocfs2_calc_extend_credits(struct super_block *sb, |
06f9da6e | 532 | struct ocfs2_extent_list *root_el) |
ccd979bd | 533 | { |
811f933d | 534 | int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks; |
ccd979bd MF |
535 | |
536 | /* bitmap dinode, group desc. + relinked group. */ | |
537 | bitmap_blocks = OCFS2_SUBALLOC_ALLOC; | |
538 | ||
539 | /* we might need to shift tree depth so lets assume an | |
540 | * absolute worst case of complete fragmentation. Even with | |
541 | * that, we only need one update for the dinode, and then | |
542 | * however many metadata chunks needed * a remaining suballoc | |
543 | * alloc. */ | |
544 | sysfile_bitmap_blocks = 1 + | |
811f933d | 545 | (OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(root_el); |
ccd979bd MF |
546 | |
547 | /* this does not include *new* metadata blocks, which are | |
811f933d | 548 | * accounted for in sysfile_bitmap_blocks. root_el + |
ccd979bd MF |
549 | * prev. last_eb_blk + blocks along edge of tree. |
550 | * calc_symlink_credits passes because we just need 1 | |
551 | * credit for the dinode there. */ | |
811f933d | 552 | extent_blocks = 1 + 1 + le16_to_cpu(root_el->l_tree_depth); |
ccd979bd | 553 | |
a90714c1 | 554 | return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks + |
6115ea28 | 555 | ocfs2_quota_trans_credits(sb); |
ccd979bd MF |
556 | } |
557 | ||
558 | static inline int ocfs2_calc_symlink_credits(struct super_block *sb) | |
559 | { | |
9b7895ef | 560 | int blocks = ocfs2_mknod_credits(sb, 0, 0); |
ccd979bd MF |
561 | |
562 | /* links can be longer than one block so we may update many | |
563 | * within our single allocated extent. */ | |
564 | blocks += ocfs2_clusters_to_blocks(sb, 1); | |
565 | ||
a90714c1 | 566 | return blocks + ocfs2_quota_trans_credits(sb); |
ccd979bd MF |
567 | } |
568 | ||
569 | static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb, | |
570 | unsigned int cpg) | |
571 | { | |
572 | int blocks; | |
573 | int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1; | |
574 | /* parent inode update + new block group header + bitmap inode update | |
575 | + bitmap blocks affected */ | |
576 | blocks = 1 + 1 + 1 + bitmap_blocks; | |
577 | return blocks; | |
578 | } | |
579 | ||
8b06bc59 JB |
580 | /* |
581 | * Allocating a discontiguous block group requires the credits from | |
582 | * ocfs2_calc_group_alloc_credits() as well as enough credits to fill | |
583 | * the group descriptor's extent list. The caller already has started | |
584 | * the transaction with ocfs2_calc_group_alloc_credits(). They extend | |
585 | * it with these credits. | |
586 | */ | |
587 | static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb) | |
588 | { | |
589 | return ocfs2_extent_recs_per_gd(sb); | |
590 | } | |
591 | ||
ccd979bd MF |
592 | static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb, |
593 | unsigned int clusters_to_del, | |
594 | struct ocfs2_dinode *fe, | |
595 | struct ocfs2_extent_list *last_el) | |
596 | { | |
597 | /* for dinode + all headers in this pass + update to next leaf */ | |
598 | u16 next_free = le16_to_cpu(last_el->l_next_free_rec); | |
599 | u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth); | |
600 | int credits = 1 + tree_depth + 1; | |
601 | int i; | |
602 | ||
603 | i = next_free - 1; | |
604 | BUG_ON(i < 0); | |
605 | ||
606 | /* We may be deleting metadata blocks, so metadata alloc dinode + | |
607 | one desc. block for each possible delete. */ | |
608 | if (tree_depth && next_free == 1 && | |
e48edee2 | 609 | ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del) |
ccd979bd MF |
610 | credits += 1 + tree_depth; |
611 | ||
612 | /* update to the truncate log. */ | |
613 | credits += OCFS2_TRUNCATE_LOG_UPDATE; | |
614 | ||
a90714c1 JK |
615 | credits += ocfs2_quota_trans_credits(sb); |
616 | ||
ccd979bd MF |
617 | return credits; |
618 | } | |
619 | ||
2b4e30fb JB |
620 | static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode) |
621 | { | |
622 | return jbd2_journal_file_inode(handle, &OCFS2_I(inode)->ip_jinode); | |
623 | } | |
624 | ||
625 | static inline int ocfs2_begin_ordered_truncate(struct inode *inode, | |
626 | loff_t new_size) | |
627 | { | |
7f5aa215 JK |
628 | return jbd2_journal_begin_ordered_truncate( |
629 | OCFS2_SB(inode->i_sb)->journal->j_journal, | |
630 | &OCFS2_I(inode)->ip_jinode, | |
631 | new_size); | |
2b4e30fb JB |
632 | } |
633 | ||
2931cdcb DW |
634 | static inline void ocfs2_update_inode_fsync_trans(handle_t *handle, |
635 | struct inode *inode, | |
636 | int datasync) | |
637 | { | |
638 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
639 | ||
640 | oi->i_sync_tid = handle->h_transaction->t_tid; | |
641 | if (datasync) | |
642 | oi->i_datasync_tid = handle->h_transaction->t_tid; | |
643 | } | |
644 | ||
ccd979bd | 645 | #endif /* OCFS2_JOURNAL_H */ |