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 26 #ifndef OCFS2_ALLOC_H 27 #define OCFS2_ALLOC_H 28 29 30 /* 31 * For xattr tree leaf, we limit the leaf byte size to be 64K. 32 */ 33 #define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536 34 35 /* 36 * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract 37 * the b-tree operations in ocfs2. Now all the b-tree operations are not 38 * limited to ocfs2_dinode only. Any data which need to allocate clusters 39 * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree 40 * and operation. 41 * 42 * ocfs2_extent_tree becomes the first-class object for extent tree 43 * manipulation. Callers of the alloc.c code need to fill it via one of 44 * the ocfs2_init_*_extent_tree() operations below. 45 * 46 * ocfs2_extent_tree contains info for the root of the b-tree, it must have a 47 * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree 48 * functions. It needs the ocfs2_caching_info structure associated with 49 * I/O on the tree. With metadata ecc, we now call different journal_access 50 * functions for each type of metadata, so it must have the 51 * root_journal_access function. 52 * ocfs2_extent_tree_operations abstract the normal operations we do for 53 * the root of extent b-tree. 54 */ 55 struct ocfs2_extent_tree_operations; 56 struct ocfs2_extent_tree { 57 struct ocfs2_extent_tree_operations *et_ops; 58 struct buffer_head *et_root_bh; 59 struct ocfs2_extent_list *et_root_el; 60 struct ocfs2_caching_info *et_ci; 61 ocfs2_journal_access_func et_root_journal_access; 62 void *et_object; 63 unsigned int et_max_leaf_clusters; 64 }; 65 66 /* 67 * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the 68 * specified object buffer. 69 */ 70 void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, 71 struct inode *inode, 72 struct buffer_head *bh); 73 void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, 74 struct inode *inode, 75 struct buffer_head *bh); 76 struct ocfs2_xattr_value_buf; 77 void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, 78 struct inode *inode, 79 struct ocfs2_xattr_value_buf *vb); 80 void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et, 81 struct inode *inode, 82 struct buffer_head *bh); 83 84 /* 85 * Read an extent block into *bh. If *bh is NULL, a bh will be 86 * allocated. This is a cached read. The extent block will be validated 87 * with ocfs2_validate_extent_block(). 88 */ 89 int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno, 90 struct buffer_head **bh); 91 92 struct ocfs2_alloc_context; 93 int ocfs2_insert_extent(handle_t *handle, 94 struct ocfs2_extent_tree *et, 95 u32 cpos, 96 u64 start_blk, 97 u32 new_clusters, 98 u8 flags, 99 struct ocfs2_alloc_context *meta_ac); 100 101 enum ocfs2_alloc_restarted { 102 RESTART_NONE = 0, 103 RESTART_TRANS, 104 RESTART_META 105 }; 106 int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, 107 struct inode *inode, 108 u32 *logical_offset, 109 u32 clusters_to_add, 110 int mark_unwritten, 111 struct ocfs2_extent_tree *et, 112 handle_t *handle, 113 struct ocfs2_alloc_context *data_ac, 114 struct ocfs2_alloc_context *meta_ac, 115 enum ocfs2_alloc_restarted *reason_ret); 116 struct ocfs2_cached_dealloc_ctxt; 117 int ocfs2_mark_extent_written(struct inode *inode, 118 struct ocfs2_extent_tree *et, 119 handle_t *handle, u32 cpos, u32 len, u32 phys, 120 struct ocfs2_alloc_context *meta_ac, 121 struct ocfs2_cached_dealloc_ctxt *dealloc); 122 int ocfs2_remove_extent(struct inode *inode, 123 struct ocfs2_extent_tree *et, 124 u32 cpos, u32 len, handle_t *handle, 125 struct ocfs2_alloc_context *meta_ac, 126 struct ocfs2_cached_dealloc_ctxt *dealloc); 127 int ocfs2_remove_btree_range(struct inode *inode, 128 struct ocfs2_extent_tree *et, 129 u32 cpos, u32 phys_cpos, u32 len, 130 struct ocfs2_cached_dealloc_ctxt *dealloc); 131 132 int ocfs2_num_free_extents(struct ocfs2_super *osb, 133 struct ocfs2_extent_tree *et); 134 135 /* 136 * how many new metadata chunks would an allocation need at maximum? 137 * 138 * Please note that the caller must make sure that root_el is the root 139 * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise 140 * the result may be wrong. 141 */ 142 static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el) 143 { 144 /* 145 * Rather than do all the work of determining how much we need 146 * (involves a ton of reads and locks), just ask for the 147 * maximal limit. That's a tree depth shift. So, one block for 148 * level of the tree (current l_tree_depth), one block for the 149 * new tree_depth==0 extent_block, and one block at the new 150 * top-of-the tree. 151 */ 152 return le16_to_cpu(root_el->l_tree_depth) + 2; 153 } 154 155 void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di); 156 void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di); 157 int ocfs2_convert_inline_data_to_extents(struct inode *inode, 158 struct buffer_head *di_bh); 159 160 int ocfs2_truncate_log_init(struct ocfs2_super *osb); 161 void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb); 162 void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, 163 int cancel); 164 int ocfs2_flush_truncate_log(struct ocfs2_super *osb); 165 int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, 166 int slot_num, 167 struct ocfs2_dinode **tl_copy); 168 int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, 169 struct ocfs2_dinode *tl_copy); 170 int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb); 171 int ocfs2_truncate_log_append(struct ocfs2_super *osb, 172 handle_t *handle, 173 u64 start_blk, 174 unsigned int num_clusters); 175 int __ocfs2_flush_truncate_log(struct ocfs2_super *osb); 176 177 /* 178 * Process local structure which describes the block unlinks done 179 * during an operation. This is populated via 180 * ocfs2_cache_block_dealloc(). 181 * 182 * ocfs2_run_deallocs() should be called after the potentially 183 * de-allocating routines. No journal handles should be open, and most 184 * locks should have been dropped. 185 */ 186 struct ocfs2_cached_dealloc_ctxt { 187 struct ocfs2_per_slot_free_list *c_first_suballocator; 188 struct ocfs2_cached_block_free *c_global_allocator; 189 }; 190 static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c) 191 { 192 c->c_first_suballocator = NULL; 193 c->c_global_allocator = NULL; 194 } 195 int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, 196 u64 blkno, unsigned int bit); 197 static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c) 198 { 199 return c->c_global_allocator != NULL; 200 } 201 int ocfs2_run_deallocs(struct ocfs2_super *osb, 202 struct ocfs2_cached_dealloc_ctxt *ctxt); 203 204 struct ocfs2_truncate_context { 205 struct ocfs2_cached_dealloc_ctxt tc_dealloc; 206 int tc_ext_alloc_locked; /* is it cluster locked? */ 207 /* these get destroyed once it's passed to ocfs2_commit_truncate. */ 208 struct buffer_head *tc_last_eb_bh; 209 }; 210 211 int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, 212 u64 range_start, u64 range_end); 213 int ocfs2_prepare_truncate(struct ocfs2_super *osb, 214 struct inode *inode, 215 struct buffer_head *fe_bh, 216 struct ocfs2_truncate_context **tc); 217 int ocfs2_commit_truncate(struct ocfs2_super *osb, 218 struct inode *inode, 219 struct buffer_head *fe_bh, 220 struct ocfs2_truncate_context *tc); 221 int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, 222 unsigned int start, unsigned int end, int trunc); 223 224 int ocfs2_find_leaf(struct ocfs2_caching_info *ci, 225 struct ocfs2_extent_list *root_el, u32 cpos, 226 struct buffer_head **leaf_bh); 227 int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster); 228 229 /* 230 * Helper function to look at the # of clusters in an extent record. 231 */ 232 static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el, 233 struct ocfs2_extent_rec *rec) 234 { 235 /* 236 * Cluster count in extent records is slightly different 237 * between interior nodes and leaf nodes. This is to support 238 * unwritten extents which need a flags field in leaf node 239 * records, thus shrinking the available space for a clusters 240 * field. 241 */ 242 if (el->l_tree_depth) 243 return le32_to_cpu(rec->e_int_clusters); 244 else 245 return le16_to_cpu(rec->e_leaf_clusters); 246 } 247 248 /* 249 * This is only valid for leaf nodes, which are the only ones that can 250 * have empty extents anyway. 251 */ 252 static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec) 253 { 254 return !rec->e_leaf_clusters; 255 } 256 257 #endif /* OCFS2_ALLOC_H */ 258