]>
Commit | Line | Data |
---|---|---|
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 |
18 | enum ocfs2_journal_state { |
19 | OCFS2_JOURNAL_FREE = 0, | |
20 | OCFS2_JOURNAL_LOADED, | |
21 | OCFS2_JOURNAL_IN_SHUTDOWN, | |
22 | }; | |
23 | ||
24 | struct ocfs2_super; | |
25 | struct 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 | ||
32 | struct ocfs2_recovery_map { | |
33 | unsigned int rm_used; | |
34 | unsigned int *rm_entries; | |
35 | }; | |
36 | ||
37 | ||
ccd979bd MF |
38 | struct 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 | ||
66 | extern spinlock_t trans_inc_lock; | |
67 | ||
68 | /* wrap j_trans_id so we never have it equal to zero. */ | |
69 | static 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 |
80 | static 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 | 93 | static 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. */ | |
109 | static 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 */ |
124 | static 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 |
135 | static 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 | 144 | void ocfs2_orphan_scan_init(struct ocfs2_super *osb); |
8b712cd5 | 145 | void ocfs2_orphan_scan_start(struct ocfs2_super *osb); |
83273932 | 146 | void ocfs2_orphan_scan_stop(struct ocfs2_super *osb); |
83273932 | 147 | |
c4028958 | 148 | void ocfs2_complete_recovery(struct work_struct *work); |
553abd04 JB |
149 | void ocfs2_wait_for_recovery(struct ocfs2_super *osb); |
150 | ||
151 | int ocfs2_recovery_init(struct ocfs2_super *osb); | |
152 | void ocfs2_recovery_exit(struct ocfs2_super *osb); | |
ccd979bd | 153 | |
9140db04 | 154 | int ocfs2_compute_replay_slots(struct ocfs2_super *osb); |
ccd979bd MF |
155 | /* |
156 | * Journal Control: | |
157 | * Initialize, Load, Shutdown, Wipe a journal. | |
158 | * | |
159 | * ocfs2_journal_init - Initialize journal structures in the OSB. | |
160 | * ocfs2_journal_load - Load the given journal off disk. Replay it if | |
161 | * there's transactions still in there. | |
162 | * ocfs2_journal_shutdown - Shutdown a journal, this will flush all | |
163 | * uncommitted, uncheckpointed transactions. | |
164 | * ocfs2_journal_wipe - Wipe transactions from a journal. Optionally | |
165 | * zero out each block. | |
166 | * ocfs2_recovery_thread - Perform recovery on a node. osb is our own osb. | |
167 | * ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat | |
168 | * event on. | |
169 | * ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint. | |
170 | */ | |
171 | void ocfs2_set_journal_params(struct ocfs2_super *osb); | |
172 | int ocfs2_journal_init(struct ocfs2_journal *journal, | |
173 | int *dirty); | |
174 | void ocfs2_journal_shutdown(struct ocfs2_super *osb); | |
175 | int ocfs2_journal_wipe(struct ocfs2_journal *journal, | |
176 | int full); | |
539d8264 SM |
177 | int ocfs2_journal_load(struct ocfs2_journal *journal, int local, |
178 | int replayed); | |
ccd979bd MF |
179 | int ocfs2_check_journals_nolocks(struct ocfs2_super *osb); |
180 | void ocfs2_recovery_thread(struct ocfs2_super *osb, | |
181 | int node_num); | |
182 | int ocfs2_mark_dead_nodes(struct ocfs2_super *osb); | |
183 | void ocfs2_complete_mount_recovery(struct ocfs2_super *osb); | |
2205363d | 184 | void ocfs2_complete_quota_recovery(struct ocfs2_super *osb); |
ccd979bd MF |
185 | |
186 | static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb) | |
187 | { | |
ccd979bd MF |
188 | wake_up(&osb->checkpoint_event); |
189 | } | |
190 | ||
191 | static inline void ocfs2_checkpoint_inode(struct inode *inode) | |
192 | { | |
193 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
194 | ||
c271c5c2 SM |
195 | if (ocfs2_mount_local(osb)) |
196 | return; | |
197 | ||
0cf2f763 | 198 | if (!ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))) { |
ccd979bd MF |
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 | |
25985edc | 202 | * wind up waiting *a lot* longer than necessary. Right |
ccd979bd MF |
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, | |
0cf2f763 | 208 | ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))); |
ccd979bd MF |
209 | } |
210 | } | |
211 | ||
212 | /* | |
213 | * Transaction Handling: | |
214 | * Manage the lifetime of a transaction handle. | |
215 | * | |
ccd979bd MF |
216 | * 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. | |
1fabe148 MF |
219 | * 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). | |
ccd979bd MF |
224 | * 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. | |
50655ae9 | 227 | * ocfs2_journal_access* - Notify the handle that we want to journal this |
ccd979bd MF |
228 | * buffer. Will have to call ocfs2_journal_dirty once |
229 | * we've actually dirtied it. Type is one of . or . | |
50655ae9 JB |
230 | * Always call the specific flavor of |
231 | * ocfs2_journal_access_*() unless you intend to | |
232 | * manage the checksum by hand. | |
ccd979bd | 233 | * ocfs2_journal_dirty - Mark a journalled buffer as having dirty data. |
bbd0f327 JQ |
234 | * ocfs2_jbd2_inode_add_write - Mark an inode with range so that its data goes |
235 | * out before the current handle commits. | |
ccd979bd MF |
236 | */ |
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. */ | |
1fabe148 | 241 | handle_t *ocfs2_start_trans(struct ocfs2_super *osb, |
ccd979bd | 242 | int max_buffs); |
1fabe148 MF |
243 | int ocfs2_commit_trans(struct ocfs2_super *osb, |
244 | handle_t *handle); | |
1fc58146 | 245 | int ocfs2_extend_trans(handle_t *handle, int nblocks); |
2b1e55c3 YL |
246 | int ocfs2_allocate_extend_trans(handle_t *handle, |
247 | int thresh); | |
248 | ||
249 | /* | |
250 | * Define an arbitrary limit for the amount of data we will anticipate | |
251 | * writing to any given transaction. For unbounded transactions such as | |
252 | * fallocate(2) we can write more than this, but we always | |
253 | * start off at the maximum transaction size and grow the transaction | |
254 | * optimistically as we go. | |
255 | */ | |
256 | #define OCFS2_MAX_TRANS_DATA 64U | |
ccd979bd MF |
257 | |
258 | /* | |
259 | * Create access is for when we get a newly created buffer and we're | |
260 | * not gonna read it off disk, but rather fill it ourselves. Right | |
261 | * now, we don't do anything special with this (it turns into a write | |
262 | * request), but this is a good placeholder in case we do... | |
263 | * | |
264 | * Write access is for when we read a block off disk and are going to | |
265 | * modify it. This way the journalling layer knows it may need to make | |
266 | * a copy of that block (if it's part of another, uncommitted | |
267 | * transaction) before we do so. | |
268 | */ | |
269 | #define OCFS2_JOURNAL_ACCESS_CREATE 0 | |
270 | #define OCFS2_JOURNAL_ACCESS_WRITE 1 | |
271 | #define OCFS2_JOURNAL_ACCESS_UNDO 2 | |
272 | ||
13723d00 | 273 | |
50655ae9 | 274 | /* ocfs2_inode */ |
0cf2f763 | 275 | int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci, |
13723d00 | 276 | struct buffer_head *bh, int type); |
50655ae9 | 277 | /* ocfs2_extent_block */ |
0cf2f763 | 278 | int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 | 279 | struct buffer_head *bh, int type); |
93c97087 TM |
280 | /* ocfs2_refcount_block */ |
281 | int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci, | |
282 | struct buffer_head *bh, int type); | |
50655ae9 | 283 | /* ocfs2_group_desc */ |
0cf2f763 | 284 | int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
285 | struct buffer_head *bh, int type); |
286 | /* ocfs2_xattr_block */ | |
0cf2f763 | 287 | int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
288 | struct buffer_head *bh, int type); |
289 | /* quota blocks */ | |
0cf2f763 | 290 | int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
291 | struct buffer_head *bh, int type); |
292 | /* dirblock */ | |
0cf2f763 | 293 | int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 | 294 | struct buffer_head *bh, int type); |
9b7895ef | 295 | /* ocfs2_dx_root_block */ |
0cf2f763 | 296 | int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci, |
9b7895ef MF |
297 | struct buffer_head *bh, int type); |
298 | /* ocfs2_dx_leaf */ | |
0cf2f763 | 299 | int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci, |
9b7895ef | 300 | struct buffer_head *bh, int type); |
50655ae9 | 301 | /* Anything that has no ecc */ |
0cf2f763 | 302 | int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci, |
50655ae9 JB |
303 | struct buffer_head *bh, int type); |
304 | ||
ccd979bd MF |
305 | /* |
306 | * A word about the journal_access/journal_dirty "dance". It is | |
307 | * entirely legal to journal_access a buffer more than once (as long | |
308 | * as the access type is the same -- I'm not sure what will happen if | |
309 | * access type is different but this should never happen anyway) It is | |
310 | * also legal to journal_dirty a buffer more than once. In fact, you | |
311 | * can even journal_access a buffer after you've done a | |
312 | * journal_access/journal_dirty pair. The only thing you cannot do | |
313 | * however, is journal_dirty a buffer which you haven't yet passed to | |
314 | * journal_access at least once. | |
315 | * | |
316 | * That said, 99% of the time this doesn't matter and this is what the | |
317 | * path looks like: | |
318 | * | |
319 | * <read a bh> | |
320 | * ocfs2_journal_access(handle, bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
321 | * <modify the bh> | |
322 | * ocfs2_journal_dirty(handle, bh); | |
323 | */ | |
ec20cec7 | 324 | void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh); |
ccd979bd MF |
325 | |
326 | /* | |
327 | * Credit Macros: | |
328 | * Convenience macros to calculate number of credits needed. | |
329 | * | |
330 | * For convenience sake, I have a set of macros here which calculate | |
331 | * the *maximum* number of sectors which will be changed for various | |
332 | * metadata updates. | |
333 | */ | |
334 | ||
335 | /* simple file updates like chmod, etc. */ | |
336 | #define OCFS2_INODE_UPDATE_CREDITS 1 | |
337 | ||
cf1d6c76 TY |
338 | /* extended attribute block update */ |
339 | #define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1 | |
340 | ||
0584974a JK |
341 | /* Update of a single quota block */ |
342 | #define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1 | |
343 | ||
a90714c1 | 344 | /* global quotafile inode update, data block */ |
0584974a JK |
345 | #define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \ |
346 | OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 | 347 | |
0584974a | 348 | #define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS |
a90714c1 JK |
349 | /* |
350 | * The two writes below can accidentally see global info dirty due | |
351 | * to set_info() quotactl so make them prepared for the writes. | |
352 | */ | |
353 | /* quota data block, global info */ | |
354 | /* Write to local quota file */ | |
0584974a JK |
355 | #define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ |
356 | OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 JK |
357 | |
358 | /* global quota data block, local quota data block, global quota inode, | |
359 | * global quota info */ | |
0584974a JK |
360 | #define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \ |
361 | 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS) | |
a90714c1 JK |
362 | |
363 | static inline int ocfs2_quota_trans_credits(struct super_block *sb) | |
364 | { | |
365 | int credits = 0; | |
366 | ||
367 | if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) | |
368 | credits += OCFS2_QWRITE_CREDITS; | |
369 | if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) | |
370 | credits += OCFS2_QWRITE_CREDITS; | |
371 | return credits; | |
372 | } | |
373 | ||
d659072f TM |
374 | /* group extend. inode update and last group update. */ |
375 | #define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
376 | ||
7909f2bf TM |
377 | /* group add. inode update and the new group update. */ |
378 | #define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
379 | ||
ccd979bd MF |
380 | /* get one bit out of a suballocator: dinode + group descriptor + |
381 | * prev. group desc. if we relink. */ | |
382 | #define OCFS2_SUBALLOC_ALLOC (3) | |
383 | ||
a90714c1 JK |
384 | static inline int ocfs2_inline_to_extents_credits(struct super_block *sb) |
385 | { | |
386 | return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS + | |
387 | ocfs2_quota_trans_credits(sb); | |
388 | } | |
1afc32b9 | 389 | |
ccd979bd MF |
390 | /* dinode + group descriptor update. We don't relink on free yet. */ |
391 | #define OCFS2_SUBALLOC_FREE (2) | |
392 | ||
393 | #define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS | |
394 | #define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE \ | |
395 | + OCFS2_TRUNCATE_LOG_UPDATE) | |
396 | ||
a90714c1 JK |
397 | static inline int ocfs2_remove_extent_credits(struct super_block *sb) |
398 | { | |
399 | return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS + | |
400 | ocfs2_quota_trans_credits(sb); | |
401 | } | |
063c4561 | 402 | |
705773a6 JK |
403 | /* data block for new dir/symlink, allocation of directory block, dx_root |
404 | * update for free list */ | |
405 | #define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + OCFS2_SUBALLOC_ALLOC + 1) | |
ccd979bd | 406 | |
9b7895ef MF |
407 | static inline int ocfs2_add_dir_index_credits(struct super_block *sb) |
408 | { | |
409 | /* 1 block for index, 2 allocs (data, metadata), 1 clusters | |
410 | * worth of blocks for initial extent. */ | |
411 | return 1 + 2 * OCFS2_SUBALLOC_ALLOC + | |
412 | ocfs2_clusters_to_blocks(sb, 1); | |
413 | } | |
414 | ||
415 | /* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode | |
416 | * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr | |
417 | * blocks + quota update */ | |
418 | static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir, | |
419 | int xattr_credits) | |
a90714c1 | 420 | { |
9b7895ef MF |
421 | int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS; |
422 | ||
423 | if (is_dir) | |
424 | dir_credits += ocfs2_add_dir_index_credits(sb); | |
425 | ||
426 | return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits + | |
a90714c1 JK |
427 | ocfs2_quota_trans_credits(sb); |
428 | } | |
ccd979bd MF |
429 | |
430 | /* local alloc metadata change + main bitmap updates */ | |
431 | #define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS \ | |
432 | + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE) | |
433 | ||
434 | /* used when we don't need an allocation change for a dir extend. One | |
435 | * for the dinode, one for the new block. */ | |
436 | #define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2) | |
437 | ||
a90714c1 | 438 | /* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota |
0393afea XH |
439 | * update on dir + index leaf + dx root update for free list + |
440 | * previous dirblock update in the free list */ | |
a90714c1 JK |
441 | static inline int ocfs2_link_credits(struct super_block *sb) |
442 | { | |
bf5a5264 | 443 | return 2 * OCFS2_INODE_UPDATE_CREDITS + 4 + |
a90714c1 JK |
444 | ocfs2_quota_trans_credits(sb); |
445 | } | |
ccd979bd MF |
446 | |
447 | /* inode + dir inode (if we unlink a dir), + dir entry block + orphan | |
e7c17e43 | 448 | * dir inode link + dir inode index leaf + dir index root */ |
a90714c1 JK |
449 | static inline int ocfs2_unlink_credits(struct super_block *sb) |
450 | { | |
451 | /* The quota update from ocfs2_link_credits is unused here... */ | |
e7c17e43 | 452 | return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb); |
a90714c1 | 453 | } |
ccd979bd MF |
454 | |
455 | /* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry + | |
dfa13f39 JB |
456 | * inode alloc group descriptor + orphan dir index root + |
457 | * orphan dir index leaf */ | |
458 | #define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 4) | |
ccd979bd | 459 | |
06ee5c75 JQ |
460 | /* dinode + orphan dir dinode + extent tree leaf block + orphan dir entry + |
461 | * orphan dir index root + orphan dir index leaf */ | |
462 | #define OCFS2_INODE_ADD_TO_ORPHAN_CREDITS (2 * OCFS2_INODE_UPDATE_CREDITS + 4) | |
463 | #define OCFS2_INODE_DEL_FROM_ORPHAN_CREDITS OCFS2_INODE_ADD_TO_ORPHAN_CREDITS | |
464 | ||
ccd979bd MF |
465 | /* dinode update, old dir dinode update, new dir dinode update, old |
466 | * dir dir entry, new dir dir entry, dir entry update for renaming | |
9b7895ef | 467 | * directory + target unlink + 3 x dir index leaves */ |
a90714c1 JK |
468 | static inline int ocfs2_rename_credits(struct super_block *sb) |
469 | { | |
9b7895ef | 470 | return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb); |
a90714c1 | 471 | } |
ccd979bd | 472 | |
cf1d6c76 TY |
473 | /* global bitmap dinode, group desc., relinked group, |
474 | * suballocator dinode, group desc., relinked group, | |
475 | * dinode, xattr block */ | |
476 | #define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \ | |
477 | + OCFS2_INODE_UPDATE_CREDITS \ | |
478 | + OCFS2_XATTR_BLOCK_UPDATE_CREDITS) | |
479 | ||
9b7895ef MF |
480 | /* inode update, removal of dx root block from allocator */ |
481 | #define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \ | |
482 | OCFS2_SUBALLOC_FREE) | |
483 | ||
4ed8a6bb MF |
484 | static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb) |
485 | { | |
486 | int credits = 1 + OCFS2_SUBALLOC_ALLOC; | |
487 | ||
488 | credits += ocfs2_clusters_to_blocks(sb, 1); | |
489 | credits += ocfs2_quota_trans_credits(sb); | |
490 | ||
491 | return credits; | |
492 | } | |
493 | ||
8bf396de TM |
494 | /* inode update, new refcount block and its allocation credits. */ |
495 | #define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \ | |
496 | + OCFS2_SUBALLOC_ALLOC) | |
497 | ||
498 | /* inode and the refcount block update. */ | |
499 | #define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
500 | ||
501 | /* | |
502 | * inode and the refcount block update. | |
503 | * It doesn't include the credits for sub alloc change. | |
504 | * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added. | |
505 | */ | |
506 | #define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1) | |
507 | ||
bcbbb24a TM |
508 | /* 2 metadata alloc, 2 new blocks and root refcount block */ |
509 | #define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3) | |
510 | ||
811f933d TM |
511 | /* |
512 | * Please note that the caller must make sure that root_el is the root | |
513 | * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise | |
514 | * the result may be wrong. | |
515 | */ | |
ccd979bd | 516 | static inline int ocfs2_calc_extend_credits(struct super_block *sb, |
06f9da6e | 517 | struct ocfs2_extent_list *root_el) |
ccd979bd | 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 | 539 | return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks + |
6115ea28 | 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 | ||
8b06bc59 JB |
565 | /* |
566 | * Allocating a discontiguous block group requires the credits from | |
567 | * ocfs2_calc_group_alloc_credits() as well as enough credits to fill | |
568 | * the group descriptor's extent list. The caller already has started | |
569 | * the transaction with ocfs2_calc_group_alloc_credits(). They extend | |
570 | * it with these credits. | |
571 | */ | |
572 | static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb) | |
573 | { | |
574 | return ocfs2_extent_recs_per_gd(sb); | |
575 | } | |
576 | ||
bbd0f327 JQ |
577 | static inline int ocfs2_jbd2_inode_add_write(handle_t *handle, struct inode *inode, |
578 | loff_t start_byte, loff_t length) | |
2b4e30fb | 579 | { |
bbd0f327 JQ |
580 | return jbd2_journal_inode_ranged_write(handle, |
581 | &OCFS2_I(inode)->ip_jinode, | |
582 | start_byte, length); | |
2b4e30fb JB |
583 | } |
584 | ||
585 | static inline int ocfs2_begin_ordered_truncate(struct inode *inode, | |
586 | loff_t new_size) | |
587 | { | |
7f5aa215 JK |
588 | return jbd2_journal_begin_ordered_truncate( |
589 | OCFS2_SB(inode->i_sb)->journal->j_journal, | |
590 | &OCFS2_I(inode)->ip_jinode, | |
591 | new_size); | |
2b4e30fb JB |
592 | } |
593 | ||
2931cdcb DW |
594 | static inline void ocfs2_update_inode_fsync_trans(handle_t *handle, |
595 | struct inode *inode, | |
596 | int datasync) | |
597 | { | |
598 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
599 | ||
9f16ca48 | 600 | if (!is_handle_aborted(handle)) { |
601 | oi->i_sync_tid = handle->h_transaction->t_tid; | |
602 | if (datasync) | |
603 | oi->i_datasync_tid = handle->h_transaction->t_tid; | |
604 | } | |
2931cdcb DW |
605 | } |
606 | ||
ccd979bd | 607 | #endif /* OCFS2_JOURNAL_H */ |