1ccd979bdSMark Fasheh /* -*- mode: c; c-basic-offset: 8; -*- 2ccd979bdSMark Fasheh * vim: noexpandtab sw=8 ts=8 sts=0: 3ccd979bdSMark Fasheh * 4ccd979bdSMark Fasheh * alloc.h 5ccd979bdSMark Fasheh * 6ccd979bdSMark Fasheh * Function prototypes 7ccd979bdSMark Fasheh * 8ccd979bdSMark Fasheh * Copyright (C) 2002, 2004 Oracle. All rights reserved. 9ccd979bdSMark Fasheh * 10ccd979bdSMark Fasheh * This program is free software; you can redistribute it and/or 11ccd979bdSMark Fasheh * modify it under the terms of the GNU General Public 12ccd979bdSMark Fasheh * License as published by the Free Software Foundation; either 13ccd979bdSMark Fasheh * version 2 of the License, or (at your option) any later version. 14ccd979bdSMark Fasheh * 15ccd979bdSMark Fasheh * This program is distributed in the hope that it will be useful, 16ccd979bdSMark Fasheh * but WITHOUT ANY WARRANTY; without even the implied warranty of 17ccd979bdSMark Fasheh * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 18ccd979bdSMark Fasheh * General Public License for more details. 19ccd979bdSMark Fasheh * 20ccd979bdSMark Fasheh * You should have received a copy of the GNU General Public 21ccd979bdSMark Fasheh * License along with this program; if not, write to the 22ccd979bdSMark Fasheh * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 23ccd979bdSMark Fasheh * Boston, MA 021110-1307, USA. 24ccd979bdSMark Fasheh */ 25ccd979bdSMark Fasheh 26ccd979bdSMark Fasheh #ifndef OCFS2_ALLOC_H 27ccd979bdSMark Fasheh #define OCFS2_ALLOC_H 28ccd979bdSMark Fasheh 29e7d4cb6bSTao Ma 30ca12b7c4STao Ma /* 31ca12b7c4STao Ma * For xattr tree leaf, we limit the leaf byte size to be 64K. 32ca12b7c4STao Ma */ 33ca12b7c4STao Ma #define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536 34ca12b7c4STao Ma 35f99b9b7cSJoel Becker /* 36f99b9b7cSJoel Becker * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract 37f99b9b7cSJoel Becker * the b-tree operations in ocfs2. Now all the b-tree operations are not 38f99b9b7cSJoel Becker * limited to ocfs2_dinode only. Any data which need to allocate clusters 39f99b9b7cSJoel Becker * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree 40f99b9b7cSJoel Becker * and operation. 41f99b9b7cSJoel Becker * 42f99b9b7cSJoel Becker * ocfs2_extent_tree becomes the first-class object for extent tree 43f99b9b7cSJoel Becker * manipulation. Callers of the alloc.c code need to fill it via one of 448d6220d6SJoel Becker * the ocfs2_init_*_extent_tree() operations below. 45f99b9b7cSJoel Becker * 46f99b9b7cSJoel Becker * ocfs2_extent_tree contains info for the root of the b-tree, it must have a 47f99b9b7cSJoel Becker * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree 48d9a0a1f8SJoel Becker * functions. It needs the ocfs2_caching_info structure associated with 49d9a0a1f8SJoel Becker * I/O on the tree. With metadata ecc, we now call different journal_access 5013723d00SJoel Becker * functions for each type of metadata, so it must have the 5113723d00SJoel Becker * root_journal_access function. 52f99b9b7cSJoel Becker * ocfs2_extent_tree_operations abstract the normal operations we do for 53f99b9b7cSJoel Becker * the root of extent b-tree. 54f99b9b7cSJoel Becker */ 55f99b9b7cSJoel Becker struct ocfs2_extent_tree_operations; 56f99b9b7cSJoel Becker struct ocfs2_extent_tree { 57f99b9b7cSJoel Becker struct ocfs2_extent_tree_operations *et_ops; 58f99b9b7cSJoel Becker struct buffer_head *et_root_bh; 59f99b9b7cSJoel Becker struct ocfs2_extent_list *et_root_el; 60d9a0a1f8SJoel Becker struct ocfs2_caching_info *et_ci; 6113723d00SJoel Becker ocfs2_journal_access_func et_root_journal_access; 62f99b9b7cSJoel Becker void *et_object; 63f99b9b7cSJoel Becker unsigned int et_max_leaf_clusters; 64f99b9b7cSJoel Becker }; 65f99b9b7cSJoel Becker 66f99b9b7cSJoel Becker /* 678d6220d6SJoel Becker * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the 688d6220d6SJoel Becker * specified object buffer. 69f99b9b7cSJoel Becker */ 708d6220d6SJoel Becker void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, 715e404e9eSJoel Becker struct ocfs2_caching_info *ci, 72f99b9b7cSJoel Becker struct buffer_head *bh); 738d6220d6SJoel Becker void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, 745e404e9eSJoel Becker struct ocfs2_caching_info *ci, 75f99b9b7cSJoel Becker struct buffer_head *bh); 762a50a743SJoel Becker struct ocfs2_xattr_value_buf; 778d6220d6SJoel Becker void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, 785e404e9eSJoel Becker struct ocfs2_caching_info *ci, 792a50a743SJoel Becker struct ocfs2_xattr_value_buf *vb); 809b7895efSMark Fasheh void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et, 815e404e9eSJoel Becker struct ocfs2_caching_info *ci, 829b7895efSMark Fasheh struct buffer_head *bh); 83fe924415STao Ma void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et, 84fe924415STao Ma struct ocfs2_caching_info *ci, 85fe924415STao Ma struct buffer_head *bh); 86f99b9b7cSJoel Becker 875e96581aSJoel Becker /* 885e96581aSJoel Becker * Read an extent block into *bh. If *bh is NULL, a bh will be 895e96581aSJoel Becker * allocated. This is a cached read. The extent block will be validated 905e96581aSJoel Becker * with ocfs2_validate_extent_block(). 915e96581aSJoel Becker */ 923d03a305SJoel Becker int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno, 935e96581aSJoel Becker struct buffer_head **bh); 945e96581aSJoel Becker 95f99b9b7cSJoel Becker struct ocfs2_alloc_context; 96cc79d8c1SJoel Becker int ocfs2_insert_extent(handle_t *handle, 97f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 98ba492615STao Ma u32 cpos, 99ba492615STao Ma u64 start_blk, 100ba492615STao Ma u32 new_clusters, 101ba492615STao Ma u8 flags, 102ba492615STao Ma struct ocfs2_alloc_context *meta_ac); 103f99b9b7cSJoel Becker 1040eb8d47eSTao Ma enum ocfs2_alloc_restarted { 1050eb8d47eSTao Ma RESTART_NONE = 0, 1060eb8d47eSTao Ma RESTART_TRANS, 1070eb8d47eSTao Ma RESTART_META 1080eb8d47eSTao Ma }; 109cbee7e1aSJoel Becker int ocfs2_add_clusters_in_btree(handle_t *handle, 110cbee7e1aSJoel Becker struct ocfs2_extent_tree *et, 1110eb8d47eSTao Ma u32 *logical_offset, 1120eb8d47eSTao Ma u32 clusters_to_add, 1130eb8d47eSTao Ma int mark_unwritten, 1140eb8d47eSTao Ma struct ocfs2_alloc_context *data_ac, 1150eb8d47eSTao Ma struct ocfs2_alloc_context *meta_ac, 116f99b9b7cSJoel Becker enum ocfs2_alloc_restarted *reason_ret); 117328d5752SMark Fasheh struct ocfs2_cached_dealloc_ctxt; 118e2e9f608STao Ma struct ocfs2_path; 119e2e9f608STao Ma int ocfs2_split_extent(handle_t *handle, 120e2e9f608STao Ma struct ocfs2_extent_tree *et, 121e2e9f608STao Ma struct ocfs2_path *path, 122e2e9f608STao Ma int split_index, 123e2e9f608STao Ma struct ocfs2_extent_rec *split_rec, 124e2e9f608STao Ma struct ocfs2_alloc_context *meta_ac, 125e2e9f608STao Ma struct ocfs2_cached_dealloc_ctxt *dealloc); 126f99b9b7cSJoel Becker int ocfs2_mark_extent_written(struct inode *inode, 127f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 128328d5752SMark Fasheh handle_t *handle, u32 cpos, u32 len, u32 phys, 129328d5752SMark Fasheh struct ocfs2_alloc_context *meta_ac, 130f99b9b7cSJoel Becker struct ocfs2_cached_dealloc_ctxt *dealloc); 1311aa75feaSTao Ma int ocfs2_change_extent_flag(handle_t *handle, 1321aa75feaSTao Ma struct ocfs2_extent_tree *et, 1331aa75feaSTao Ma u32 cpos, u32 len, u32 phys, 1341aa75feaSTao Ma struct ocfs2_alloc_context *meta_ac, 1351aa75feaSTao Ma struct ocfs2_cached_dealloc_ctxt *dealloc, 1361aa75feaSTao Ma int new_flags, int clear_flags); 137dbdcf6a4SJoel Becker int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et, 138dbdcf6a4SJoel Becker u32 cpos, u32 len, 139063c4561SMark Fasheh struct ocfs2_alloc_context *meta_ac, 140f99b9b7cSJoel Becker struct ocfs2_cached_dealloc_ctxt *dealloc); 141fecc0112SMark Fasheh int ocfs2_remove_btree_range(struct inode *inode, 142fecc0112SMark Fasheh struct ocfs2_extent_tree *et, 143fecc0112SMark Fasheh u32 cpos, u32 phys_cpos, u32 len, 144fecc0112SMark Fasheh struct ocfs2_cached_dealloc_ctxt *dealloc); 145fecc0112SMark Fasheh 146ccd979bdSMark Fasheh int ocfs2_num_free_extents(struct ocfs2_super *osb, 147f99b9b7cSJoel Becker struct ocfs2_extent_tree *et); 148e7d4cb6bSTao Ma 149811f933dSTao Ma /* 150811f933dSTao Ma * how many new metadata chunks would an allocation need at maximum? 151811f933dSTao Ma * 152811f933dSTao Ma * Please note that the caller must make sure that root_el is the root 153811f933dSTao Ma * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise 154811f933dSTao Ma * the result may be wrong. 155811f933dSTao Ma */ 156811f933dSTao Ma static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el) 157ccd979bdSMark Fasheh { 158ccd979bdSMark Fasheh /* 159ccd979bdSMark Fasheh * Rather than do all the work of determining how much we need 160ccd979bdSMark Fasheh * (involves a ton of reads and locks), just ask for the 161ccd979bdSMark Fasheh * maximal limit. That's a tree depth shift. So, one block for 162ccd979bdSMark Fasheh * level of the tree (current l_tree_depth), one block for the 163ccd979bdSMark Fasheh * new tree_depth==0 extent_block, and one block at the new 164ccd979bdSMark Fasheh * top-of-the tree. 165ccd979bdSMark Fasheh */ 166811f933dSTao Ma return le16_to_cpu(root_el->l_tree_depth) + 2; 167ccd979bdSMark Fasheh } 168ccd979bdSMark Fasheh 1695b6a3a2bSMark Fasheh void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di); 1701afc32b9SMark Fasheh void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di); 1711afc32b9SMark Fasheh int ocfs2_convert_inline_data_to_extents(struct inode *inode, 1721afc32b9SMark Fasheh struct buffer_head *di_bh); 1731afc32b9SMark Fasheh 174ccd979bdSMark Fasheh int ocfs2_truncate_log_init(struct ocfs2_super *osb); 175ccd979bdSMark Fasheh void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb); 176ccd979bdSMark Fasheh void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, 177ccd979bdSMark Fasheh int cancel); 178ccd979bdSMark Fasheh int ocfs2_flush_truncate_log(struct ocfs2_super *osb); 179ccd979bdSMark Fasheh int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, 180ccd979bdSMark Fasheh int slot_num, 181ccd979bdSMark Fasheh struct ocfs2_dinode **tl_copy); 182ccd979bdSMark Fasheh int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, 183ccd979bdSMark Fasheh struct ocfs2_dinode *tl_copy); 184063c4561SMark Fasheh int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb); 185063c4561SMark Fasheh int ocfs2_truncate_log_append(struct ocfs2_super *osb, 186063c4561SMark Fasheh handle_t *handle, 187063c4561SMark Fasheh u64 start_blk, 188063c4561SMark Fasheh unsigned int num_clusters); 189063c4561SMark Fasheh int __ocfs2_flush_truncate_log(struct ocfs2_super *osb); 190ccd979bdSMark Fasheh 1912b604351SMark Fasheh /* 1922b604351SMark Fasheh * Process local structure which describes the block unlinks done 1932b604351SMark Fasheh * during an operation. This is populated via 1942b604351SMark Fasheh * ocfs2_cache_block_dealloc(). 1952b604351SMark Fasheh * 1962b604351SMark Fasheh * ocfs2_run_deallocs() should be called after the potentially 1972b604351SMark Fasheh * de-allocating routines. No journal handles should be open, and most 1982b604351SMark Fasheh * locks should have been dropped. 1992b604351SMark Fasheh */ 2002b604351SMark Fasheh struct ocfs2_cached_dealloc_ctxt { 2012b604351SMark Fasheh struct ocfs2_per_slot_free_list *c_first_suballocator; 2022891d290STao Ma struct ocfs2_cached_block_free *c_global_allocator; 2032b604351SMark Fasheh }; 2042b604351SMark Fasheh static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c) 2052b604351SMark Fasheh { 2062b604351SMark Fasheh c->c_first_suballocator = NULL; 2072891d290STao Ma c->c_global_allocator = NULL; 2082b604351SMark Fasheh } 2092891d290STao Ma int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, 2102891d290STao Ma u64 blkno, unsigned int bit); 2111823cb0bSTao Ma int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, 2121823cb0bSTao Ma int type, int slot, u64 blkno, 2131823cb0bSTao Ma unsigned int bit); 21478f30c31STao Ma static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c) 21578f30c31STao Ma { 21678f30c31STao Ma return c->c_global_allocator != NULL; 21778f30c31STao Ma } 2182b604351SMark Fasheh int ocfs2_run_deallocs(struct ocfs2_super *osb, 2192b604351SMark Fasheh struct ocfs2_cached_dealloc_ctxt *ctxt); 2202b604351SMark Fasheh 221ccd979bdSMark Fasheh struct ocfs2_truncate_context { 22259a5e416SMark Fasheh struct ocfs2_cached_dealloc_ctxt tc_dealloc; 223ccd979bdSMark Fasheh int tc_ext_alloc_locked; /* is it cluster locked? */ 224ccd979bdSMark Fasheh /* these get destroyed once it's passed to ocfs2_commit_truncate. */ 225ccd979bdSMark Fasheh struct buffer_head *tc_last_eb_bh; 226ccd979bdSMark Fasheh }; 227ccd979bdSMark Fasheh 22835edec1dSMark Fasheh int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, 22935edec1dSMark Fasheh u64 range_start, u64 range_end); 230ccd979bdSMark Fasheh int ocfs2_prepare_truncate(struct ocfs2_super *osb, 231ccd979bdSMark Fasheh struct inode *inode, 232ccd979bdSMark Fasheh struct buffer_head *fe_bh, 233ccd979bdSMark Fasheh struct ocfs2_truncate_context **tc); 234ccd979bdSMark Fasheh int ocfs2_commit_truncate(struct ocfs2_super *osb, 235ccd979bdSMark Fasheh struct inode *inode, 236ccd979bdSMark Fasheh struct buffer_head *fe_bh, 237ccd979bdSMark Fasheh struct ocfs2_truncate_context *tc); 2381afc32b9SMark Fasheh int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, 2391afc32b9SMark Fasheh unsigned int start, unsigned int end, int trunc); 240ccd979bdSMark Fasheh 241facdb77fSJoel Becker int ocfs2_find_leaf(struct ocfs2_caching_info *ci, 242facdb77fSJoel Becker struct ocfs2_extent_list *root_el, u32 cpos, 243facdb77fSJoel Becker struct buffer_head **leaf_bh); 244328d5752SMark Fasheh int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster); 245363041a5SMark Fasheh 246e48edee2SMark Fasheh /* 247e48edee2SMark Fasheh * Helper function to look at the # of clusters in an extent record. 248e48edee2SMark Fasheh */ 249e48edee2SMark Fasheh static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el, 250e48edee2SMark Fasheh struct ocfs2_extent_rec *rec) 251e48edee2SMark Fasheh { 252e48edee2SMark Fasheh /* 253e48edee2SMark Fasheh * Cluster count in extent records is slightly different 254e48edee2SMark Fasheh * between interior nodes and leaf nodes. This is to support 255e48edee2SMark Fasheh * unwritten extents which need a flags field in leaf node 256e48edee2SMark Fasheh * records, thus shrinking the available space for a clusters 257e48edee2SMark Fasheh * field. 258e48edee2SMark Fasheh */ 259e48edee2SMark Fasheh if (el->l_tree_depth) 260e48edee2SMark Fasheh return le32_to_cpu(rec->e_int_clusters); 261e48edee2SMark Fasheh else 262e48edee2SMark Fasheh return le16_to_cpu(rec->e_leaf_clusters); 263e48edee2SMark Fasheh } 264e48edee2SMark Fasheh 26500dc417fSMark Fasheh /* 26600dc417fSMark Fasheh * This is only valid for leaf nodes, which are the only ones that can 26700dc417fSMark Fasheh * have empty extents anyway. 26800dc417fSMark Fasheh */ 26900dc417fSMark Fasheh static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec) 27000dc417fSMark Fasheh { 27100dc417fSMark Fasheh return !rec->e_leaf_clusters; 27200dc417fSMark Fasheh } 27300dc417fSMark Fasheh 274e2e9f608STao Ma /* 275e2e9f608STao Ma * Structures which describe a path through a btree, and functions to 276e2e9f608STao Ma * manipulate them. 277e2e9f608STao Ma * 278e2e9f608STao Ma * The idea here is to be as generic as possible with the tree 279e2e9f608STao Ma * manipulation code. 280e2e9f608STao Ma */ 281e2e9f608STao Ma struct ocfs2_path_item { 282e2e9f608STao Ma struct buffer_head *bh; 283e2e9f608STao Ma struct ocfs2_extent_list *el; 284e2e9f608STao Ma }; 285e2e9f608STao Ma 286e2e9f608STao Ma #define OCFS2_MAX_PATH_DEPTH 5 287e2e9f608STao Ma 288e2e9f608STao Ma struct ocfs2_path { 289e2e9f608STao Ma int p_tree_depth; 290e2e9f608STao Ma ocfs2_journal_access_func p_root_access; 291e2e9f608STao Ma struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH]; 292e2e9f608STao Ma }; 293e2e9f608STao Ma 294e2e9f608STao Ma #define path_root_bh(_path) ((_path)->p_node[0].bh) 295e2e9f608STao Ma #define path_root_el(_path) ((_path)->p_node[0].el) 296e2e9f608STao Ma #define path_root_access(_path)((_path)->p_root_access) 297e2e9f608STao Ma #define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh) 298e2e9f608STao Ma #define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el) 299e2e9f608STao Ma #define path_num_items(_path) ((_path)->p_tree_depth + 1) 300e2e9f608STao Ma 301e2e9f608STao Ma void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root); 302e2e9f608STao Ma void ocfs2_free_path(struct ocfs2_path *path); 303e2e9f608STao Ma int ocfs2_find_path(struct ocfs2_caching_info *ci, 304e2e9f608STao Ma struct ocfs2_path *path, 305e2e9f608STao Ma u32 cpos); 306e2e9f608STao Ma struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path); 307e2e9f608STao Ma struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et); 308e2e9f608STao Ma int ocfs2_path_bh_journal_access(handle_t *handle, 309e2e9f608STao Ma struct ocfs2_caching_info *ci, 310e2e9f608STao Ma struct ocfs2_path *path, 311e2e9f608STao Ma int idx); 312e2e9f608STao Ma int ocfs2_journal_access_path(struct ocfs2_caching_info *ci, 313e2e9f608STao Ma handle_t *handle, 314e2e9f608STao Ma struct ocfs2_path *path); 315ccd979bdSMark Fasheh #endif /* OCFS2_ALLOC_H */ 316