journal.h 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648
  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. #ifndef OCFS2_JOURNAL_H
  26. #define OCFS2_JOURNAL_H
  27. #include <linux/fs.h>
  28. #include <linux/jbd2.h>
  29. enum ocfs2_journal_state {
  30. OCFS2_JOURNAL_FREE = 0,
  31. OCFS2_JOURNAL_LOADED,
  32. OCFS2_JOURNAL_IN_SHUTDOWN,
  33. };
  34. struct ocfs2_super;
  35. struct ocfs2_dinode;
  36. /*
  37. * The recovery_list is a simple linked list of node numbers to recover.
  38. * It is protected by the recovery_lock.
  39. */
  40. struct ocfs2_recovery_map {
  41. unsigned int rm_used;
  42. unsigned int *rm_entries;
  43. };
  44. struct ocfs2_journal {
  45. enum ocfs2_journal_state j_state; /* Journals current state */
  46. journal_t *j_journal; /* The kernels journal type */
  47. struct inode *j_inode; /* Kernel inode pointing to
  48. * this journal */
  49. struct ocfs2_super *j_osb; /* pointer to the super
  50. * block for the node
  51. * we're currently
  52. * running on -- not
  53. * necessarily the super
  54. * block from the node
  55. * which we usually run
  56. * from (recovery,
  57. * etc) */
  58. struct buffer_head *j_bh; /* Journal disk inode block */
  59. atomic_t j_num_trans; /* Number of transactions
  60. * currently in the system. */
  61. spinlock_t j_lock;
  62. unsigned long j_trans_id;
  63. struct rw_semaphore j_trans_barrier;
  64. wait_queue_head_t j_checkpointed;
  65. /* both fields protected by j_lock*/
  66. struct list_head j_la_cleanups;
  67. struct work_struct j_recovery_work;
  68. };
  69. extern spinlock_t trans_inc_lock;
  70. /* wrap j_trans_id so we never have it equal to zero. */
  71. static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j)
  72. {
  73. unsigned long old_id;
  74. spin_lock(&trans_inc_lock);
  75. old_id = j->j_trans_id++;
  76. if (unlikely(!j->j_trans_id))
  77. j->j_trans_id = 1;
  78. spin_unlock(&trans_inc_lock);
  79. return old_id;
  80. }
  81. static inline void ocfs2_set_ci_lock_trans(struct ocfs2_journal *journal,
  82. struct ocfs2_caching_info *ci)
  83. {
  84. spin_lock(&trans_inc_lock);
  85. ci->ci_last_trans = journal->j_trans_id;
  86. spin_unlock(&trans_inc_lock);
  87. }
  88. /* Used to figure out whether it's safe to drop a metadata lock on an
  89. * cached object. Returns true if all the object's changes have been
  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. */
  93. static inline int ocfs2_ci_fully_checkpointed(struct ocfs2_caching_info *ci)
  94. {
  95. int ret;
  96. struct ocfs2_journal *journal =
  97. OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
  98. spin_lock(&trans_inc_lock);
  99. ret = time_after(journal->j_trans_id, ci->ci_last_trans);
  100. spin_unlock(&trans_inc_lock);
  101. return ret;
  102. }
  103. /* convenience function to check if an object backed by struct
  104. * ocfs2_caching_info is still new (has never hit disk) Will do you a
  105. * favor and set created_trans = 0 when you've
  106. * been checkpointed. returns '1' if the ci is still new. */
  107. static inline int ocfs2_ci_is_new(struct ocfs2_caching_info *ci)
  108. {
  109. int ret;
  110. struct ocfs2_journal *journal =
  111. OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
  112. spin_lock(&trans_inc_lock);
  113. ret = !(time_after(journal->j_trans_id, ci->ci_created_trans));
  114. if (!ret)
  115. ci->ci_created_trans = 0;
  116. spin_unlock(&trans_inc_lock);
  117. return ret;
  118. }
  119. /* Wrapper for inodes so we can check system files */
  120. static inline int ocfs2_inode_is_new(struct inode *inode)
  121. {
  122. /* System files are never "new" as they're written out by
  123. * mkfs. This helps us early during mount, before we have the
  124. * journal open and j_trans_id could be junk. */
  125. if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
  126. return 0;
  127. return ocfs2_ci_is_new(INODE_CACHE(inode));
  128. }
  129. static inline void ocfs2_ci_set_new(struct ocfs2_super *osb,
  130. struct ocfs2_caching_info *ci)
  131. {
  132. spin_lock(&trans_inc_lock);
  133. ci->ci_created_trans = osb->journal->j_trans_id;
  134. spin_unlock(&trans_inc_lock);
  135. }
  136. /* Exported only for the journal struct init code in super.c. Do not call. */
  137. void ocfs2_orphan_scan_init(struct ocfs2_super *osb);
  138. void ocfs2_orphan_scan_start(struct ocfs2_super *osb);
  139. void ocfs2_orphan_scan_stop(struct ocfs2_super *osb);
  140. void ocfs2_orphan_scan_exit(struct ocfs2_super *osb);
  141. void ocfs2_complete_recovery(struct work_struct *work);
  142. void ocfs2_wait_for_recovery(struct ocfs2_super *osb);
  143. int ocfs2_recovery_init(struct ocfs2_super *osb);
  144. void ocfs2_recovery_exit(struct ocfs2_super *osb);
  145. int ocfs2_compute_replay_slots(struct ocfs2_super *osb);
  146. /*
  147. * Journal Control:
  148. * Initialize, Load, Shutdown, Wipe a journal.
  149. *
  150. * ocfs2_journal_init - Initialize journal structures in the OSB.
  151. * ocfs2_journal_load - Load the given journal off disk. Replay it if
  152. * there's transactions still in there.
  153. * ocfs2_journal_shutdown - Shutdown a journal, this will flush all
  154. * uncommitted, uncheckpointed transactions.
  155. * ocfs2_journal_wipe - Wipe transactions from a journal. Optionally
  156. * zero out each block.
  157. * ocfs2_recovery_thread - Perform recovery on a node. osb is our own osb.
  158. * ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
  159. * event on.
  160. * ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
  161. */
  162. void ocfs2_set_journal_params(struct ocfs2_super *osb);
  163. int ocfs2_journal_init(struct ocfs2_journal *journal,
  164. int *dirty);
  165. void ocfs2_journal_shutdown(struct ocfs2_super *osb);
  166. int ocfs2_journal_wipe(struct ocfs2_journal *journal,
  167. int full);
  168. int ocfs2_journal_load(struct ocfs2_journal *journal, int local,
  169. int replayed);
  170. int ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
  171. void ocfs2_recovery_thread(struct ocfs2_super *osb,
  172. int node_num);
  173. int ocfs2_mark_dead_nodes(struct ocfs2_super *osb);
  174. void ocfs2_complete_mount_recovery(struct ocfs2_super *osb);
  175. void ocfs2_complete_quota_recovery(struct ocfs2_super *osb);
  176. static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb)
  177. {
  178. wake_up(&osb->checkpoint_event);
  179. }
  180. static inline void ocfs2_checkpoint_inode(struct inode *inode)
  181. {
  182. struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
  183. if (ocfs2_mount_local(osb))
  184. return;
  185. if (!ocfs2_ci_fully_checkpointed(INODE_CACHE(inode))) {
  186. /* WARNING: This only kicks off a single
  187. * checkpoint. If someone races you and adds more
  188. * metadata to the journal, you won't know, and will
  189. * wind up waiting *a lot* longer than necessary. Right
  190. * now we only use this in clear_inode so that's
  191. * OK. */
  192. ocfs2_start_checkpoint(osb);
  193. wait_event(osb->journal->j_checkpointed,
  194. ocfs2_ci_fully_checkpointed(INODE_CACHE(inode)));
  195. }
  196. }
  197. /*
  198. * Transaction Handling:
  199. * Manage the lifetime of a transaction handle.
  200. *
  201. * ocfs2_start_trans - Begin a transaction. Give it an upper estimate of
  202. * the number of blocks that will be changed during
  203. * this handle.
  204. * ocfs2_commit_trans - Complete a handle. It might return -EIO if
  205. * the journal was aborted. The majority of paths don't
  206. * check the return value as an error there comes too
  207. * late to do anything (and will be picked up in a
  208. * later transaction).
  209. * ocfs2_extend_trans - Extend a handle by nblocks credits. This may
  210. * commit the handle to disk in the process, but will
  211. * not release any locks taken during the transaction.
  212. * ocfs2_journal_access* - Notify the handle that we want to journal this
  213. * buffer. Will have to call ocfs2_journal_dirty once
  214. * we've actually dirtied it. Type is one of . or .
  215. * Always call the specific flavor of
  216. * ocfs2_journal_access_*() unless you intend to
  217. * manage the checksum by hand.
  218. * ocfs2_journal_dirty - Mark a journalled buffer as having dirty data.
  219. * ocfs2_jbd2_file_inode - Mark an inode so that its data goes out before
  220. * the current handle commits.
  221. */
  222. /* You must always start_trans with a number of buffs > 0, but it's
  223. * perfectly legal to go through an entire transaction without having
  224. * dirtied any buffers. */
  225. handle_t *ocfs2_start_trans(struct ocfs2_super *osb,
  226. int max_buffs);
  227. int ocfs2_commit_trans(struct ocfs2_super *osb,
  228. handle_t *handle);
  229. int ocfs2_extend_trans(handle_t *handle, int nblocks);
  230. int ocfs2_allocate_extend_trans(handle_t *handle,
  231. int thresh);
  232. /*
  233. * Define an arbitrary limit for the amount of data we will anticipate
  234. * writing to any given transaction. For unbounded transactions such as
  235. * fallocate(2) we can write more than this, but we always
  236. * start off at the maximum transaction size and grow the transaction
  237. * optimistically as we go.
  238. */
  239. #define OCFS2_MAX_TRANS_DATA 64U
  240. /*
  241. * Create access is for when we get a newly created buffer and we're
  242. * not gonna read it off disk, but rather fill it ourselves. Right
  243. * now, we don't do anything special with this (it turns into a write
  244. * request), but this is a good placeholder in case we do...
  245. *
  246. * Write access is for when we read a block off disk and are going to
  247. * modify it. This way the journalling layer knows it may need to make
  248. * a copy of that block (if it's part of another, uncommitted
  249. * transaction) before we do so.
  250. */
  251. #define OCFS2_JOURNAL_ACCESS_CREATE 0
  252. #define OCFS2_JOURNAL_ACCESS_WRITE 1
  253. #define OCFS2_JOURNAL_ACCESS_UNDO 2
  254. /* ocfs2_inode */
  255. int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci,
  256. struct buffer_head *bh, int type);
  257. /* ocfs2_extent_block */
  258. int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci,
  259. struct buffer_head *bh, int type);
  260. /* ocfs2_refcount_block */
  261. int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci,
  262. struct buffer_head *bh, int type);
  263. /* ocfs2_group_desc */
  264. int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci,
  265. struct buffer_head *bh, int type);
  266. /* ocfs2_xattr_block */
  267. int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci,
  268. struct buffer_head *bh, int type);
  269. /* quota blocks */
  270. int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci,
  271. struct buffer_head *bh, int type);
  272. /* dirblock */
  273. int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci,
  274. struct buffer_head *bh, int type);
  275. /* ocfs2_dx_root_block */
  276. int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci,
  277. struct buffer_head *bh, int type);
  278. /* ocfs2_dx_leaf */
  279. int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci,
  280. struct buffer_head *bh, int type);
  281. /* Anything that has no ecc */
  282. int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
  283. struct buffer_head *bh, int type);
  284. /*
  285. * A word about the journal_access/journal_dirty "dance". It is
  286. * entirely legal to journal_access a buffer more than once (as long
  287. * as the access type is the same -- I'm not sure what will happen if
  288. * access type is different but this should never happen anyway) It is
  289. * also legal to journal_dirty a buffer more than once. In fact, you
  290. * can even journal_access a buffer after you've done a
  291. * journal_access/journal_dirty pair. The only thing you cannot do
  292. * however, is journal_dirty a buffer which you haven't yet passed to
  293. * journal_access at least once.
  294. *
  295. * That said, 99% of the time this doesn't matter and this is what the
  296. * path looks like:
  297. *
  298. * <read a bh>
  299. * ocfs2_journal_access(handle, bh, OCFS2_JOURNAL_ACCESS_WRITE);
  300. * <modify the bh>
  301. * ocfs2_journal_dirty(handle, bh);
  302. */
  303. void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh);
  304. /*
  305. * Credit Macros:
  306. * Convenience macros to calculate number of credits needed.
  307. *
  308. * For convenience sake, I have a set of macros here which calculate
  309. * the *maximum* number of sectors which will be changed for various
  310. * metadata updates.
  311. */
  312. /* simple file updates like chmod, etc. */
  313. #define OCFS2_INODE_UPDATE_CREDITS 1
  314. /* extended attribute block update */
  315. #define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1
  316. /* Update of a single quota block */
  317. #define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1
  318. /* global quotafile inode update, data block */
  319. #define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
  320. OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
  321. #define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS
  322. /*
  323. * The two writes below can accidentally see global info dirty due
  324. * to set_info() quotactl so make them prepared for the writes.
  325. */
  326. /* quota data block, global info */
  327. /* Write to local quota file */
  328. #define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
  329. OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
  330. /* global quota data block, local quota data block, global quota inode,
  331. * global quota info */
  332. #define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
  333. 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
  334. static inline int ocfs2_quota_trans_credits(struct super_block *sb)
  335. {
  336. int credits = 0;
  337. if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA))
  338. credits += OCFS2_QWRITE_CREDITS;
  339. if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA))
  340. credits += OCFS2_QWRITE_CREDITS;
  341. return credits;
  342. }
  343. /* group extend. inode update and last group update. */
  344. #define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  345. /* group add. inode update and the new group update. */
  346. #define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  347. /* get one bit out of a suballocator: dinode + group descriptor +
  348. * prev. group desc. if we relink. */
  349. #define OCFS2_SUBALLOC_ALLOC (3)
  350. static inline int ocfs2_inline_to_extents_credits(struct super_block *sb)
  351. {
  352. return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS +
  353. ocfs2_quota_trans_credits(sb);
  354. }
  355. /* dinode + group descriptor update. We don't relink on free yet. */
  356. #define OCFS2_SUBALLOC_FREE (2)
  357. #define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS
  358. #define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE \
  359. + OCFS2_TRUNCATE_LOG_UPDATE)
  360. static inline int ocfs2_remove_extent_credits(struct super_block *sb)
  361. {
  362. return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS +
  363. ocfs2_quota_trans_credits(sb);
  364. }
  365. /* data block for new dir/symlink, allocation of directory block, dx_root
  366. * update for free list */
  367. #define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + OCFS2_SUBALLOC_ALLOC + 1)
  368. static inline int ocfs2_add_dir_index_credits(struct super_block *sb)
  369. {
  370. /* 1 block for index, 2 allocs (data, metadata), 1 clusters
  371. * worth of blocks for initial extent. */
  372. return 1 + 2 * OCFS2_SUBALLOC_ALLOC +
  373. ocfs2_clusters_to_blocks(sb, 1);
  374. }
  375. /* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode
  376. * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr
  377. * blocks + quota update */
  378. static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir,
  379. int xattr_credits)
  380. {
  381. int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS;
  382. if (is_dir)
  383. dir_credits += ocfs2_add_dir_index_credits(sb);
  384. return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits +
  385. ocfs2_quota_trans_credits(sb);
  386. }
  387. /* local alloc metadata change + main bitmap updates */
  388. #define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS \
  389. + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE)
  390. /* used when we don't need an allocation change for a dir extend. One
  391. * for the dinode, one for the new block. */
  392. #define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2)
  393. /* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota
  394. * update on dir + index leaf + dx root update for free list +
  395. * previous dirblock update in the free list */
  396. static inline int ocfs2_link_credits(struct super_block *sb)
  397. {
  398. return 2*OCFS2_INODE_UPDATE_CREDITS + 4 +
  399. ocfs2_quota_trans_credits(sb);
  400. }
  401. /* inode + dir inode (if we unlink a dir), + dir entry block + orphan
  402. * dir inode link + dir inode index leaf + dir index root */
  403. static inline int ocfs2_unlink_credits(struct super_block *sb)
  404. {
  405. /* The quota update from ocfs2_link_credits is unused here... */
  406. return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb);
  407. }
  408. /* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
  409. * inode alloc group descriptor + orphan dir index root +
  410. * orphan dir index leaf */
  411. #define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 4)
  412. /* dinode + orphan dir dinode + extent tree leaf block + orphan dir entry +
  413. * orphan dir index root + orphan dir index leaf */
  414. #define OCFS2_INODE_ADD_TO_ORPHAN_CREDITS (2 * OCFS2_INODE_UPDATE_CREDITS + 4)
  415. #define OCFS2_INODE_DEL_FROM_ORPHAN_CREDITS OCFS2_INODE_ADD_TO_ORPHAN_CREDITS
  416. /* dinode update, old dir dinode update, new dir dinode update, old
  417. * dir dir entry, new dir dir entry, dir entry update for renaming
  418. * directory + target unlink + 3 x dir index leaves */
  419. static inline int ocfs2_rename_credits(struct super_block *sb)
  420. {
  421. return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb);
  422. }
  423. /* global bitmap dinode, group desc., relinked group,
  424. * suballocator dinode, group desc., relinked group,
  425. * dinode, xattr block */
  426. #define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \
  427. + OCFS2_INODE_UPDATE_CREDITS \
  428. + OCFS2_XATTR_BLOCK_UPDATE_CREDITS)
  429. /* inode update, removal of dx root block from allocator */
  430. #define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
  431. OCFS2_SUBALLOC_FREE)
  432. static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb)
  433. {
  434. int credits = 1 + OCFS2_SUBALLOC_ALLOC;
  435. credits += ocfs2_clusters_to_blocks(sb, 1);
  436. credits += ocfs2_quota_trans_credits(sb);
  437. return credits;
  438. }
  439. /* inode update, new refcount block and its allocation credits. */
  440. #define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \
  441. + OCFS2_SUBALLOC_ALLOC)
  442. /* inode and the refcount block update. */
  443. #define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  444. /*
  445. * inode and the refcount block update.
  446. * It doesn't include the credits for sub alloc change.
  447. * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added.
  448. */
  449. #define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  450. /* 2 metadata alloc, 2 new blocks and root refcount block */
  451. #define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3)
  452. /*
  453. * Please note that the caller must make sure that root_el is the root
  454. * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
  455. * the result may be wrong.
  456. */
  457. static inline int ocfs2_calc_extend_credits(struct super_block *sb,
  458. struct ocfs2_extent_list *root_el)
  459. {
  460. int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks;
  461. /* bitmap dinode, group desc. + relinked group. */
  462. bitmap_blocks = OCFS2_SUBALLOC_ALLOC;
  463. /* we might need to shift tree depth so lets assume an
  464. * absolute worst case of complete fragmentation. Even with
  465. * that, we only need one update for the dinode, and then
  466. * however many metadata chunks needed * a remaining suballoc
  467. * alloc. */
  468. sysfile_bitmap_blocks = 1 +
  469. (OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(root_el);
  470. /* this does not include *new* metadata blocks, which are
  471. * accounted for in sysfile_bitmap_blocks. root_el +
  472. * prev. last_eb_blk + blocks along edge of tree.
  473. * calc_symlink_credits passes because we just need 1
  474. * credit for the dinode there. */
  475. extent_blocks = 1 + 1 + le16_to_cpu(root_el->l_tree_depth);
  476. return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks +
  477. ocfs2_quota_trans_credits(sb);
  478. }
  479. static inline int ocfs2_calc_symlink_credits(struct super_block *sb)
  480. {
  481. int blocks = ocfs2_mknod_credits(sb, 0, 0);
  482. /* links can be longer than one block so we may update many
  483. * within our single allocated extent. */
  484. blocks += ocfs2_clusters_to_blocks(sb, 1);
  485. return blocks + ocfs2_quota_trans_credits(sb);
  486. }
  487. static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
  488. unsigned int cpg)
  489. {
  490. int blocks;
  491. int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1;
  492. /* parent inode update + new block group header + bitmap inode update
  493. + bitmap blocks affected */
  494. blocks = 1 + 1 + 1 + bitmap_blocks;
  495. return blocks;
  496. }
  497. /*
  498. * Allocating a discontiguous block group requires the credits from
  499. * ocfs2_calc_group_alloc_credits() as well as enough credits to fill
  500. * the group descriptor's extent list. The caller already has started
  501. * the transaction with ocfs2_calc_group_alloc_credits(). They extend
  502. * it with these credits.
  503. */
  504. static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb)
  505. {
  506. return ocfs2_extent_recs_per_gd(sb);
  507. }
  508. static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
  509. unsigned int clusters_to_del,
  510. struct ocfs2_dinode *fe,
  511. struct ocfs2_extent_list *last_el)
  512. {
  513. /* for dinode + all headers in this pass + update to next leaf */
  514. u16 next_free = le16_to_cpu(last_el->l_next_free_rec);
  515. u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth);
  516. int credits = 1 + tree_depth + 1;
  517. int i;
  518. i = next_free - 1;
  519. BUG_ON(i < 0);
  520. /* We may be deleting metadata blocks, so metadata alloc dinode +
  521. one desc. block for each possible delete. */
  522. if (tree_depth && next_free == 1 &&
  523. ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del)
  524. credits += 1 + tree_depth;
  525. /* update to the truncate log. */
  526. credits += OCFS2_TRUNCATE_LOG_UPDATE;
  527. credits += ocfs2_quota_trans_credits(sb);
  528. return credits;
  529. }
  530. static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode)
  531. {
  532. return jbd2_journal_inode_add_write(handle, &OCFS2_I(inode)->ip_jinode);
  533. }
  534. static inline int ocfs2_begin_ordered_truncate(struct inode *inode,
  535. loff_t new_size)
  536. {
  537. return jbd2_journal_begin_ordered_truncate(
  538. OCFS2_SB(inode->i_sb)->journal->j_journal,
  539. &OCFS2_I(inode)->ip_jinode,
  540. new_size);
  541. }
  542. static inline void ocfs2_update_inode_fsync_trans(handle_t *handle,
  543. struct inode *inode,
  544. int datasync)
  545. {
  546. struct ocfs2_inode_info *oi = OCFS2_I(inode);
  547. if (!is_handle_aborted(handle)) {
  548. oi->i_sync_tid = handle->h_transaction->t_tid;
  549. if (datasync)
  550. oi->i_datasync_tid = handle->h_transaction->t_tid;
  551. }
  552. }
  553. #endif /* OCFS2_JOURNAL_H */