journal.h 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618
  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. unsigned long j_trans_id;
  62. struct rw_semaphore j_trans_barrier;
  63. wait_queue_head_t j_checkpointed;
  64. spinlock_t j_lock;
  65. struct list_head j_la_cleanups;
  66. struct work_struct j_recovery_work;
  67. };
  68. extern spinlock_t trans_inc_lock;
  69. /* wrap j_trans_id so we never have it equal to zero. */
  70. static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j)
  71. {
  72. unsigned long old_id;
  73. spin_lock(&trans_inc_lock);
  74. old_id = j->j_trans_id++;
  75. if (unlikely(!j->j_trans_id))
  76. j->j_trans_id = 1;
  77. spin_unlock(&trans_inc_lock);
  78. return old_id;
  79. }
  80. static inline void ocfs2_set_ci_lock_trans(struct ocfs2_journal *journal,
  81. struct ocfs2_caching_info *ci)
  82. {
  83. spin_lock(&trans_inc_lock);
  84. ci->ci_last_trans = journal->j_trans_id;
  85. spin_unlock(&trans_inc_lock);
  86. }
  87. /* Used to figure out whether it's safe to drop a metadata lock on an
  88. * cached object. Returns true if all the object's changes have been
  89. * checkpointed to disk. You should be holding the spinlock on the
  90. * metadata lock while calling this to be sure that nobody can take
  91. * the lock and put it on another transaction. */
  92. static inline int ocfs2_ci_fully_checkpointed(struct ocfs2_caching_info *ci)
  93. {
  94. int ret;
  95. struct ocfs2_journal *journal =
  96. OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
  97. spin_lock(&trans_inc_lock);
  98. ret = time_after(journal->j_trans_id, ci->ci_last_trans);
  99. spin_unlock(&trans_inc_lock);
  100. return ret;
  101. }
  102. /* convenience function to check if an object backed by struct
  103. * ocfs2_caching_info is still new (has never hit disk) Will do you a
  104. * favor and set created_trans = 0 when you've
  105. * been checkpointed. returns '1' if the ci is still new. */
  106. static inline int ocfs2_ci_is_new(struct ocfs2_caching_info *ci)
  107. {
  108. int ret;
  109. struct ocfs2_journal *journal =
  110. OCFS2_SB(ocfs2_metadata_cache_get_super(ci))->journal;
  111. spin_lock(&trans_inc_lock);
  112. ret = !(time_after(journal->j_trans_id, ci->ci_created_trans));
  113. if (!ret)
  114. ci->ci_created_trans = 0;
  115. spin_unlock(&trans_inc_lock);
  116. return ret;
  117. }
  118. /* Wrapper for inodes so we can check system files */
  119. static inline int ocfs2_inode_is_new(struct inode *inode)
  120. {
  121. /* System files are never "new" as they're written out by
  122. * mkfs. This helps us early during mount, before we have the
  123. * journal open and j_trans_id could be junk. */
  124. if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
  125. return 0;
  126. return ocfs2_ci_is_new(INODE_CACHE(inode));
  127. }
  128. static inline void ocfs2_ci_set_new(struct ocfs2_super *osb,
  129. struct ocfs2_caching_info *ci)
  130. {
  131. spin_lock(&trans_inc_lock);
  132. ci->ci_created_trans = osb->journal->j_trans_id;
  133. spin_unlock(&trans_inc_lock);
  134. }
  135. /* Exported only for the journal struct init code in super.c. Do not call. */
  136. void ocfs2_orphan_scan_init(struct ocfs2_super *osb);
  137. void ocfs2_orphan_scan_start(struct ocfs2_super *osb);
  138. void ocfs2_orphan_scan_stop(struct ocfs2_super *osb);
  139. void ocfs2_orphan_scan_exit(struct ocfs2_super *osb);
  140. void ocfs2_complete_recovery(struct work_struct *work);
  141. void ocfs2_wait_for_recovery(struct ocfs2_super *osb);
  142. int ocfs2_recovery_init(struct ocfs2_super *osb);
  143. void ocfs2_recovery_exit(struct ocfs2_super *osb);
  144. int ocfs2_compute_replay_slots(struct ocfs2_super *osb);
  145. /*
  146. * Journal Control:
  147. * Initialize, Load, Shutdown, Wipe a journal.
  148. *
  149. * ocfs2_journal_init - Initialize journal structures in the OSB.
  150. * ocfs2_journal_load - Load the given journal off disk. Replay it if
  151. * there's transactions still in there.
  152. * ocfs2_journal_shutdown - Shutdown a journal, this will flush all
  153. * uncommitted, uncheckpointed transactions.
  154. * ocfs2_journal_wipe - Wipe transactions from a journal. Optionally
  155. * zero out each block.
  156. * ocfs2_recovery_thread - Perform recovery on a node. osb is our own osb.
  157. * ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
  158. * event on.
  159. * ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
  160. */
  161. void ocfs2_set_journal_params(struct ocfs2_super *osb);
  162. int ocfs2_journal_init(struct ocfs2_journal *journal,
  163. int *dirty);
  164. void ocfs2_journal_shutdown(struct ocfs2_super *osb);
  165. int ocfs2_journal_wipe(struct ocfs2_journal *journal,
  166. int full);
  167. int ocfs2_journal_load(struct ocfs2_journal *journal, int local,
  168. int replayed);
  169. int ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
  170. void ocfs2_recovery_thread(struct ocfs2_super *osb,
  171. int node_num);
  172. int ocfs2_mark_dead_nodes(struct ocfs2_super *osb);
  173. void ocfs2_complete_mount_recovery(struct ocfs2_super *osb);
  174. void ocfs2_complete_quota_recovery(struct ocfs2_super *osb);
  175. static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb)
  176. {
  177. atomic_set(&osb->needs_checkpoint, 1);
  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 *alot* 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. /*
  231. * Create access is for when we get a newly created buffer and we're
  232. * not gonna read it off disk, but rather fill it ourselves. Right
  233. * now, we don't do anything special with this (it turns into a write
  234. * request), but this is a good placeholder in case we do...
  235. *
  236. * Write access is for when we read a block off disk and are going to
  237. * modify it. This way the journalling layer knows it may need to make
  238. * a copy of that block (if it's part of another, uncommitted
  239. * transaction) before we do so.
  240. */
  241. #define OCFS2_JOURNAL_ACCESS_CREATE 0
  242. #define OCFS2_JOURNAL_ACCESS_WRITE 1
  243. #define OCFS2_JOURNAL_ACCESS_UNDO 2
  244. /* ocfs2_inode */
  245. int ocfs2_journal_access_di(handle_t *handle, struct ocfs2_caching_info *ci,
  246. struct buffer_head *bh, int type);
  247. /* ocfs2_extent_block */
  248. int ocfs2_journal_access_eb(handle_t *handle, struct ocfs2_caching_info *ci,
  249. struct buffer_head *bh, int type);
  250. /* ocfs2_refcount_block */
  251. int ocfs2_journal_access_rb(handle_t *handle, struct ocfs2_caching_info *ci,
  252. struct buffer_head *bh, int type);
  253. /* ocfs2_group_desc */
  254. int ocfs2_journal_access_gd(handle_t *handle, struct ocfs2_caching_info *ci,
  255. struct buffer_head *bh, int type);
  256. /* ocfs2_xattr_block */
  257. int ocfs2_journal_access_xb(handle_t *handle, struct ocfs2_caching_info *ci,
  258. struct buffer_head *bh, int type);
  259. /* quota blocks */
  260. int ocfs2_journal_access_dq(handle_t *handle, struct ocfs2_caching_info *ci,
  261. struct buffer_head *bh, int type);
  262. /* dirblock */
  263. int ocfs2_journal_access_db(handle_t *handle, struct ocfs2_caching_info *ci,
  264. struct buffer_head *bh, int type);
  265. /* ocfs2_dx_root_block */
  266. int ocfs2_journal_access_dr(handle_t *handle, struct ocfs2_caching_info *ci,
  267. struct buffer_head *bh, int type);
  268. /* ocfs2_dx_leaf */
  269. int ocfs2_journal_access_dl(handle_t *handle, struct ocfs2_caching_info *ci,
  270. struct buffer_head *bh, int type);
  271. /* Anything that has no ecc */
  272. int ocfs2_journal_access(handle_t *handle, struct ocfs2_caching_info *ci,
  273. struct buffer_head *bh, int type);
  274. /*
  275. * A word about the journal_access/journal_dirty "dance". It is
  276. * entirely legal to journal_access a buffer more than once (as long
  277. * as the access type is the same -- I'm not sure what will happen if
  278. * access type is different but this should never happen anyway) It is
  279. * also legal to journal_dirty a buffer more than once. In fact, you
  280. * can even journal_access a buffer after you've done a
  281. * journal_access/journal_dirty pair. The only thing you cannot do
  282. * however, is journal_dirty a buffer which you haven't yet passed to
  283. * journal_access at least once.
  284. *
  285. * That said, 99% of the time this doesn't matter and this is what the
  286. * path looks like:
  287. *
  288. * <read a bh>
  289. * ocfs2_journal_access(handle, bh, OCFS2_JOURNAL_ACCESS_WRITE);
  290. * <modify the bh>
  291. * ocfs2_journal_dirty(handle, bh);
  292. */
  293. void ocfs2_journal_dirty(handle_t *handle, struct buffer_head *bh);
  294. /*
  295. * Credit Macros:
  296. * Convenience macros to calculate number of credits needed.
  297. *
  298. * For convenience sake, I have a set of macros here which calculate
  299. * the *maximum* number of sectors which will be changed for various
  300. * metadata updates.
  301. */
  302. /* simple file updates like chmod, etc. */
  303. #define OCFS2_INODE_UPDATE_CREDITS 1
  304. /* extended attribute block update */
  305. #define OCFS2_XATTR_BLOCK_UPDATE_CREDITS 1
  306. /* Update of a single quota block */
  307. #define OCFS2_QUOTA_BLOCK_UPDATE_CREDITS 1
  308. /* global quotafile inode update, data block */
  309. #define OCFS2_QINFO_WRITE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
  310. OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
  311. #define OCFS2_LOCAL_QINFO_WRITE_CREDITS OCFS2_QUOTA_BLOCK_UPDATE_CREDITS
  312. /*
  313. * The two writes below can accidentally see global info dirty due
  314. * to set_info() quotactl so make them prepared for the writes.
  315. */
  316. /* quota data block, global info */
  317. /* Write to local quota file */
  318. #define OCFS2_QWRITE_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
  319. OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
  320. /* global quota data block, local quota data block, global quota inode,
  321. * global quota info */
  322. #define OCFS2_QSYNC_CREDITS (OCFS2_QINFO_WRITE_CREDITS + \
  323. 2 * OCFS2_QUOTA_BLOCK_UPDATE_CREDITS)
  324. static inline int ocfs2_quota_trans_credits(struct super_block *sb)
  325. {
  326. int credits = 0;
  327. if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA))
  328. credits += OCFS2_QWRITE_CREDITS;
  329. if (OCFS2_HAS_RO_COMPAT_FEATURE(sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA))
  330. credits += OCFS2_QWRITE_CREDITS;
  331. return credits;
  332. }
  333. /* group extend. inode update and last group update. */
  334. #define OCFS2_GROUP_EXTEND_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  335. /* group add. inode update and the new group update. */
  336. #define OCFS2_GROUP_ADD_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  337. /* get one bit out of a suballocator: dinode + group descriptor +
  338. * prev. group desc. if we relink. */
  339. #define OCFS2_SUBALLOC_ALLOC (3)
  340. static inline int ocfs2_inline_to_extents_credits(struct super_block *sb)
  341. {
  342. return OCFS2_SUBALLOC_ALLOC + OCFS2_INODE_UPDATE_CREDITS +
  343. ocfs2_quota_trans_credits(sb);
  344. }
  345. /* dinode + group descriptor update. We don't relink on free yet. */
  346. #define OCFS2_SUBALLOC_FREE (2)
  347. #define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS
  348. #define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE \
  349. + OCFS2_TRUNCATE_LOG_UPDATE)
  350. static inline int ocfs2_remove_extent_credits(struct super_block *sb)
  351. {
  352. return OCFS2_TRUNCATE_LOG_UPDATE + OCFS2_INODE_UPDATE_CREDITS +
  353. ocfs2_quota_trans_credits(sb);
  354. }
  355. /* data block for new dir/symlink, 2 for bitmap updates (bitmap fe +
  356. * bitmap block for the new bit) dx_root update for free list */
  357. #define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + 2 + 1)
  358. static inline int ocfs2_add_dir_index_credits(struct super_block *sb)
  359. {
  360. /* 1 block for index, 2 allocs (data, metadata), 1 clusters
  361. * worth of blocks for initial extent. */
  362. return 1 + 2 * OCFS2_SUBALLOC_ALLOC +
  363. ocfs2_clusters_to_blocks(sb, 1);
  364. }
  365. /* parent fe, parent block, new file entry, index leaf, inode alloc fe, inode
  366. * alloc group descriptor + mkdir/symlink blocks + dir blocks + xattr
  367. * blocks + quota update */
  368. static inline int ocfs2_mknod_credits(struct super_block *sb, int is_dir,
  369. int xattr_credits)
  370. {
  371. int dir_credits = OCFS2_DIR_LINK_ADDITIONAL_CREDITS;
  372. if (is_dir)
  373. dir_credits += ocfs2_add_dir_index_credits(sb);
  374. return 4 + OCFS2_SUBALLOC_ALLOC + dir_credits + xattr_credits +
  375. ocfs2_quota_trans_credits(sb);
  376. }
  377. /* local alloc metadata change + main bitmap updates */
  378. #define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS \
  379. + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE)
  380. /* used when we don't need an allocation change for a dir extend. One
  381. * for the dinode, one for the new block. */
  382. #define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2)
  383. /* file update (nlink, etc) + directory mtime/ctime + dir entry block + quota
  384. * update on dir + index leaf + dx root update for free list */
  385. static inline int ocfs2_link_credits(struct super_block *sb)
  386. {
  387. return 2*OCFS2_INODE_UPDATE_CREDITS + 3 +
  388. ocfs2_quota_trans_credits(sb);
  389. }
  390. /* inode + dir inode (if we unlink a dir), + dir entry block + orphan
  391. * dir inode link + dir inode index leaf + dir index root */
  392. static inline int ocfs2_unlink_credits(struct super_block *sb)
  393. {
  394. /* The quota update from ocfs2_link_credits is unused here... */
  395. return 2 * OCFS2_INODE_UPDATE_CREDITS + 3 + ocfs2_link_credits(sb);
  396. }
  397. /* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
  398. * inode alloc group descriptor + orphan dir index root +
  399. * orphan dir index leaf */
  400. #define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 4)
  401. /* dinode update, old dir dinode update, new dir dinode update, old
  402. * dir dir entry, new dir dir entry, dir entry update for renaming
  403. * directory + target unlink + 3 x dir index leaves */
  404. static inline int ocfs2_rename_credits(struct super_block *sb)
  405. {
  406. return 3 * OCFS2_INODE_UPDATE_CREDITS + 6 + ocfs2_unlink_credits(sb);
  407. }
  408. /* global bitmap dinode, group desc., relinked group,
  409. * suballocator dinode, group desc., relinked group,
  410. * dinode, xattr block */
  411. #define OCFS2_XATTR_BLOCK_CREATE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + \
  412. + OCFS2_INODE_UPDATE_CREDITS \
  413. + OCFS2_XATTR_BLOCK_UPDATE_CREDITS)
  414. /* inode update, removal of dx root block from allocator */
  415. #define OCFS2_DX_ROOT_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + \
  416. OCFS2_SUBALLOC_FREE)
  417. static inline int ocfs2_calc_dxi_expand_credits(struct super_block *sb)
  418. {
  419. int credits = 1 + OCFS2_SUBALLOC_ALLOC;
  420. credits += ocfs2_clusters_to_blocks(sb, 1);
  421. credits += ocfs2_quota_trans_credits(sb);
  422. return credits;
  423. }
  424. /* inode update, new refcount block and its allocation credits. */
  425. #define OCFS2_REFCOUNT_TREE_CREATE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1 \
  426. + OCFS2_SUBALLOC_ALLOC)
  427. /* inode and the refcount block update. */
  428. #define OCFS2_REFCOUNT_TREE_SET_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  429. /*
  430. * inode and the refcount block update.
  431. * It doesn't include the credits for sub alloc change.
  432. * So if we need to free the bit, OCFS2_SUBALLOC_FREE needs to be added.
  433. */
  434. #define OCFS2_REFCOUNT_TREE_REMOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS + 1)
  435. /* 2 metadata alloc, 2 new blocks and root refcount block */
  436. #define OCFS2_EXPAND_REFCOUNT_TREE_CREDITS (OCFS2_SUBALLOC_ALLOC * 2 + 3)
  437. /*
  438. * Please note that the caller must make sure that root_el is the root
  439. * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
  440. * the result may be wrong.
  441. */
  442. static inline int ocfs2_calc_extend_credits(struct super_block *sb,
  443. struct ocfs2_extent_list *root_el,
  444. u32 bits_wanted)
  445. {
  446. int bitmap_blocks, sysfile_bitmap_blocks, extent_blocks;
  447. /* bitmap dinode, group desc. + relinked group. */
  448. bitmap_blocks = OCFS2_SUBALLOC_ALLOC;
  449. /* we might need to shift tree depth so lets assume an
  450. * absolute worst case of complete fragmentation. Even with
  451. * that, we only need one update for the dinode, and then
  452. * however many metadata chunks needed * a remaining suballoc
  453. * alloc. */
  454. sysfile_bitmap_blocks = 1 +
  455. (OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(root_el);
  456. /* this does not include *new* metadata blocks, which are
  457. * accounted for in sysfile_bitmap_blocks. root_el +
  458. * prev. last_eb_blk + blocks along edge of tree.
  459. * calc_symlink_credits passes because we just need 1
  460. * credit for the dinode there. */
  461. extent_blocks = 1 + 1 + le16_to_cpu(root_el->l_tree_depth);
  462. return bitmap_blocks + sysfile_bitmap_blocks + extent_blocks +
  463. ocfs2_quota_trans_credits(sb);
  464. }
  465. static inline int ocfs2_calc_symlink_credits(struct super_block *sb)
  466. {
  467. int blocks = ocfs2_mknod_credits(sb, 0, 0);
  468. /* links can be longer than one block so we may update many
  469. * within our single allocated extent. */
  470. blocks += ocfs2_clusters_to_blocks(sb, 1);
  471. return blocks + ocfs2_quota_trans_credits(sb);
  472. }
  473. static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
  474. unsigned int cpg)
  475. {
  476. int blocks;
  477. int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1;
  478. /* parent inode update + new block group header + bitmap inode update
  479. + bitmap blocks affected */
  480. blocks = 1 + 1 + 1 + bitmap_blocks;
  481. return blocks;
  482. }
  483. /*
  484. * Allocating a discontiguous block group requires the credits from
  485. * ocfs2_calc_group_alloc_credits() as well as enough credits to fill
  486. * the group descriptor's extent list. The caller already has started
  487. * the transaction with ocfs2_calc_group_alloc_credits(). They extend
  488. * it with these credits.
  489. */
  490. static inline int ocfs2_calc_bg_discontig_credits(struct super_block *sb)
  491. {
  492. return ocfs2_extent_recs_per_gd(sb);
  493. }
  494. static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
  495. unsigned int clusters_to_del,
  496. struct ocfs2_dinode *fe,
  497. struct ocfs2_extent_list *last_el)
  498. {
  499. /* for dinode + all headers in this pass + update to next leaf */
  500. u16 next_free = le16_to_cpu(last_el->l_next_free_rec);
  501. u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth);
  502. int credits = 1 + tree_depth + 1;
  503. int i;
  504. i = next_free - 1;
  505. BUG_ON(i < 0);
  506. /* We may be deleting metadata blocks, so metadata alloc dinode +
  507. one desc. block for each possible delete. */
  508. if (tree_depth && next_free == 1 &&
  509. ocfs2_rec_clusters(last_el, &last_el->l_recs[i]) == clusters_to_del)
  510. credits += 1 + tree_depth;
  511. /* update to the truncate log. */
  512. credits += OCFS2_TRUNCATE_LOG_UPDATE;
  513. credits += ocfs2_quota_trans_credits(sb);
  514. return credits;
  515. }
  516. static inline int ocfs2_jbd2_file_inode(handle_t *handle, struct inode *inode)
  517. {
  518. return jbd2_journal_file_inode(handle, &OCFS2_I(inode)->ip_jinode);
  519. }
  520. static inline int ocfs2_begin_ordered_truncate(struct inode *inode,
  521. loff_t new_size)
  522. {
  523. return jbd2_journal_begin_ordered_truncate(
  524. OCFS2_SB(inode->i_sb)->journal->j_journal,
  525. &OCFS2_I(inode)->ip_jinode,
  526. new_size);
  527. }
  528. #endif /* OCFS2_JOURNAL_H */