alloc.h 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325
  1. /* -*- mode: c; c-basic-offset: 8; -*-
  2. * vim: noexpandtab sw=8 ts=8 sts=0:
  3. *
  4. * alloc.h
  5. *
  6. * Function prototypes
  7. *
  8. * Copyright (C) 2002, 2004 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_ALLOC_H
  26. #define OCFS2_ALLOC_H
  27. /*
  28. * For xattr tree leaf, we limit the leaf byte size to be 64K.
  29. */
  30. #define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536
  31. /*
  32. * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract
  33. * the b-tree operations in ocfs2. Now all the b-tree operations are not
  34. * limited to ocfs2_dinode only. Any data which need to allocate clusters
  35. * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree
  36. * and operation.
  37. *
  38. * ocfs2_extent_tree becomes the first-class object for extent tree
  39. * manipulation. Callers of the alloc.c code need to fill it via one of
  40. * the ocfs2_init_*_extent_tree() operations below.
  41. *
  42. * ocfs2_extent_tree contains info for the root of the b-tree, it must have a
  43. * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree
  44. * functions. It needs the ocfs2_caching_info structure associated with
  45. * I/O on the tree. With metadata ecc, we now call different journal_access
  46. * functions for each type of metadata, so it must have the
  47. * root_journal_access function.
  48. * ocfs2_extent_tree_operations abstract the normal operations we do for
  49. * the root of extent b-tree.
  50. */
  51. struct ocfs2_extent_tree_operations;
  52. struct ocfs2_extent_tree {
  53. struct ocfs2_extent_tree_operations *et_ops;
  54. struct buffer_head *et_root_bh;
  55. struct ocfs2_extent_list *et_root_el;
  56. struct ocfs2_caching_info *et_ci;
  57. ocfs2_journal_access_func et_root_journal_access;
  58. void *et_object;
  59. unsigned int et_max_leaf_clusters;
  60. };
  61. /*
  62. * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the
  63. * specified object buffer.
  64. */
  65. void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et,
  66. struct ocfs2_caching_info *ci,
  67. struct buffer_head *bh);
  68. void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et,
  69. struct ocfs2_caching_info *ci,
  70. struct buffer_head *bh);
  71. struct ocfs2_xattr_value_buf;
  72. void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et,
  73. struct ocfs2_caching_info *ci,
  74. struct ocfs2_xattr_value_buf *vb);
  75. void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et,
  76. struct ocfs2_caching_info *ci,
  77. struct buffer_head *bh);
  78. void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et,
  79. struct ocfs2_caching_info *ci,
  80. struct buffer_head *bh);
  81. /*
  82. * Read an extent block into *bh. If *bh is NULL, a bh will be
  83. * allocated. This is a cached read. The extent block will be validated
  84. * with ocfs2_validate_extent_block().
  85. */
  86. int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno,
  87. struct buffer_head **bh);
  88. struct ocfs2_alloc_context;
  89. int ocfs2_insert_extent(handle_t *handle,
  90. struct ocfs2_extent_tree *et,
  91. u32 cpos,
  92. u64 start_blk,
  93. u32 new_clusters,
  94. u8 flags,
  95. struct ocfs2_alloc_context *meta_ac);
  96. enum ocfs2_alloc_restarted {
  97. RESTART_NONE = 0,
  98. RESTART_TRANS,
  99. RESTART_META
  100. };
  101. int ocfs2_add_clusters_in_btree(handle_t *handle,
  102. struct ocfs2_extent_tree *et,
  103. u32 *logical_offset,
  104. u32 clusters_to_add,
  105. int mark_unwritten,
  106. struct ocfs2_alloc_context *data_ac,
  107. struct ocfs2_alloc_context *meta_ac,
  108. enum ocfs2_alloc_restarted *reason_ret);
  109. struct ocfs2_cached_dealloc_ctxt;
  110. struct ocfs2_path;
  111. int ocfs2_split_extent(handle_t *handle,
  112. struct ocfs2_extent_tree *et,
  113. struct ocfs2_path *path,
  114. int split_index,
  115. struct ocfs2_extent_rec *split_rec,
  116. struct ocfs2_alloc_context *meta_ac,
  117. struct ocfs2_cached_dealloc_ctxt *dealloc);
  118. int ocfs2_mark_extent_written(struct inode *inode,
  119. struct ocfs2_extent_tree *et,
  120. handle_t *handle, u32 cpos, u32 len, u32 phys,
  121. struct ocfs2_alloc_context *meta_ac,
  122. struct ocfs2_cached_dealloc_ctxt *dealloc);
  123. int ocfs2_change_extent_flag(handle_t *handle,
  124. struct ocfs2_extent_tree *et,
  125. u32 cpos, u32 len, u32 phys,
  126. struct ocfs2_alloc_context *meta_ac,
  127. struct ocfs2_cached_dealloc_ctxt *dealloc,
  128. int new_flags, int clear_flags);
  129. int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et,
  130. u32 cpos, u32 len,
  131. struct ocfs2_alloc_context *meta_ac,
  132. struct ocfs2_cached_dealloc_ctxt *dealloc);
  133. int ocfs2_remove_btree_range(struct inode *inode,
  134. struct ocfs2_extent_tree *et,
  135. u32 cpos, u32 phys_cpos, u32 len, int flags,
  136. struct ocfs2_cached_dealloc_ctxt *dealloc,
  137. u64 refcount_loc, bool refcount_tree_locked);
  138. int ocfs2_num_free_extents(struct ocfs2_super *osb,
  139. struct ocfs2_extent_tree *et);
  140. /*
  141. * how many new metadata chunks would an allocation need at maximum?
  142. *
  143. * Please note that the caller must make sure that root_el is the root
  144. * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
  145. * the result may be wrong.
  146. */
  147. static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el)
  148. {
  149. /*
  150. * Rather than do all the work of determining how much we need
  151. * (involves a ton of reads and locks), just ask for the
  152. * maximal limit. That's a tree depth shift. So, one block for
  153. * level of the tree (current l_tree_depth), one block for the
  154. * new tree_depth==0 extent_block, and one block at the new
  155. * top-of-the tree.
  156. */
  157. return le16_to_cpu(root_el->l_tree_depth) + 2;
  158. }
  159. void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di);
  160. void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di);
  161. int ocfs2_convert_inline_data_to_extents(struct inode *inode,
  162. struct buffer_head *di_bh);
  163. int ocfs2_truncate_log_init(struct ocfs2_super *osb);
  164. void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
  165. void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
  166. int cancel);
  167. int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
  168. int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
  169. int slot_num,
  170. struct ocfs2_dinode **tl_copy);
  171. int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
  172. struct ocfs2_dinode *tl_copy);
  173. int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb);
  174. int ocfs2_truncate_log_append(struct ocfs2_super *osb,
  175. handle_t *handle,
  176. u64 start_blk,
  177. unsigned int num_clusters);
  178. int __ocfs2_flush_truncate_log(struct ocfs2_super *osb);
  179. /*
  180. * Process local structure which describes the block unlinks done
  181. * during an operation. This is populated via
  182. * ocfs2_cache_block_dealloc().
  183. *
  184. * ocfs2_run_deallocs() should be called after the potentially
  185. * de-allocating routines. No journal handles should be open, and most
  186. * locks should have been dropped.
  187. */
  188. struct ocfs2_cached_dealloc_ctxt {
  189. struct ocfs2_per_slot_free_list *c_first_suballocator;
  190. struct ocfs2_cached_block_free *c_global_allocator;
  191. };
  192. static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
  193. {
  194. c->c_first_suballocator = NULL;
  195. c->c_global_allocator = NULL;
  196. }
  197. int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
  198. u64 blkno, unsigned int bit);
  199. int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
  200. int type, int slot, u64 suballoc, u64 blkno,
  201. unsigned int bit);
  202. static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
  203. {
  204. return c->c_global_allocator != NULL;
  205. }
  206. int ocfs2_run_deallocs(struct ocfs2_super *osb,
  207. struct ocfs2_cached_dealloc_ctxt *ctxt);
  208. struct ocfs2_truncate_context {
  209. struct ocfs2_cached_dealloc_ctxt tc_dealloc;
  210. int tc_ext_alloc_locked; /* is it cluster locked? */
  211. /* these get destroyed once it's passed to ocfs2_commit_truncate. */
  212. struct buffer_head *tc_last_eb_bh;
  213. };
  214. int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
  215. u64 range_start, u64 range_end);
  216. int ocfs2_commit_truncate(struct ocfs2_super *osb,
  217. struct inode *inode,
  218. struct buffer_head *di_bh);
  219. int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
  220. unsigned int start, unsigned int end, int trunc);
  221. int ocfs2_find_leaf(struct ocfs2_caching_info *ci,
  222. struct ocfs2_extent_list *root_el, u32 cpos,
  223. struct buffer_head **leaf_bh);
  224. int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster);
  225. int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range);
  226. /*
  227. * Helper function to look at the # of clusters in an extent record.
  228. */
  229. static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el,
  230. struct ocfs2_extent_rec *rec)
  231. {
  232. /*
  233. * Cluster count in extent records is slightly different
  234. * between interior nodes and leaf nodes. This is to support
  235. * unwritten extents which need a flags field in leaf node
  236. * records, thus shrinking the available space for a clusters
  237. * field.
  238. */
  239. if (el->l_tree_depth)
  240. return le32_to_cpu(rec->e_int_clusters);
  241. else
  242. return le16_to_cpu(rec->e_leaf_clusters);
  243. }
  244. /*
  245. * This is only valid for leaf nodes, which are the only ones that can
  246. * have empty extents anyway.
  247. */
  248. static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
  249. {
  250. return !rec->e_leaf_clusters;
  251. }
  252. int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end,
  253. struct page **pages, int *num);
  254. void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
  255. unsigned int from, unsigned int to,
  256. struct page *page, int zero, u64 *phys);
  257. /*
  258. * Structures which describe a path through a btree, and functions to
  259. * manipulate them.
  260. *
  261. * The idea here is to be as generic as possible with the tree
  262. * manipulation code.
  263. */
  264. struct ocfs2_path_item {
  265. struct buffer_head *bh;
  266. struct ocfs2_extent_list *el;
  267. };
  268. #define OCFS2_MAX_PATH_DEPTH 5
  269. struct ocfs2_path {
  270. int p_tree_depth;
  271. ocfs2_journal_access_func p_root_access;
  272. struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH];
  273. };
  274. #define path_root_bh(_path) ((_path)->p_node[0].bh)
  275. #define path_root_el(_path) ((_path)->p_node[0].el)
  276. #define path_root_access(_path)((_path)->p_root_access)
  277. #define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
  278. #define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
  279. #define path_num_items(_path) ((_path)->p_tree_depth + 1)
  280. void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root);
  281. void ocfs2_free_path(struct ocfs2_path *path);
  282. int ocfs2_find_path(struct ocfs2_caching_info *ci,
  283. struct ocfs2_path *path,
  284. u32 cpos);
  285. struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path);
  286. struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et);
  287. int ocfs2_path_bh_journal_access(handle_t *handle,
  288. struct ocfs2_caching_info *ci,
  289. struct ocfs2_path *path,
  290. int idx);
  291. int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
  292. handle_t *handle,
  293. struct ocfs2_path *path);
  294. int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
  295. struct ocfs2_path *path, u32 *cpos);
  296. int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
  297. struct ocfs2_path *path, u32 *cpos);
  298. int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
  299. struct ocfs2_path *left,
  300. struct ocfs2_path *right);
  301. #endif /* OCFS2_ALLOC_H */