1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
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.
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.
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.
31 * For xattr tree leaf, we limit the leaf byte size to be 64K.
33 #define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536
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
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.
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.
55 struct ocfs2_extent_tree_operations;
56 struct ocfs2_extent_tree {
57 const 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;
63 unsigned int et_max_leaf_clusters;
64 struct ocfs2_cached_dealloc_ctxt *et_dealloc;
68 * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the
69 * specified object buffer.
71 void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et,
72 struct ocfs2_caching_info *ci,
73 struct buffer_head *bh);
74 void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et,
75 struct ocfs2_caching_info *ci,
76 struct buffer_head *bh);
77 struct ocfs2_xattr_value_buf;
78 void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et,
79 struct ocfs2_caching_info *ci,
80 struct ocfs2_xattr_value_buf *vb);
81 void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et,
82 struct ocfs2_caching_info *ci,
83 struct buffer_head *bh);
84 void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et,
85 struct ocfs2_caching_info *ci,
86 struct buffer_head *bh);
89 * Read an extent block into *bh. If *bh is NULL, a bh will be
90 * allocated. This is a cached read. The extent block will be validated
91 * with ocfs2_validate_extent_block().
93 int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno,
94 struct buffer_head **bh);
96 struct ocfs2_alloc_context;
97 int ocfs2_insert_extent(handle_t *handle,
98 struct ocfs2_extent_tree *et,
103 struct ocfs2_alloc_context *meta_ac);
105 enum ocfs2_alloc_restarted {
110 int ocfs2_add_clusters_in_btree(handle_t *handle,
111 struct ocfs2_extent_tree *et,
115 struct ocfs2_alloc_context *data_ac,
116 struct ocfs2_alloc_context *meta_ac,
117 enum ocfs2_alloc_restarted *reason_ret);
118 struct ocfs2_cached_dealloc_ctxt;
120 int ocfs2_split_extent(handle_t *handle,
121 struct ocfs2_extent_tree *et,
122 struct ocfs2_path *path,
124 struct ocfs2_extent_rec *split_rec,
125 struct ocfs2_alloc_context *meta_ac,
126 struct ocfs2_cached_dealloc_ctxt *dealloc);
127 int ocfs2_mark_extent_written(struct inode *inode,
128 struct ocfs2_extent_tree *et,
129 handle_t *handle, u32 cpos, u32 len, u32 phys,
130 struct ocfs2_alloc_context *meta_ac,
131 struct ocfs2_cached_dealloc_ctxt *dealloc);
132 int ocfs2_change_extent_flag(handle_t *handle,
133 struct ocfs2_extent_tree *et,
134 u32 cpos, u32 len, u32 phys,
135 struct ocfs2_alloc_context *meta_ac,
136 struct ocfs2_cached_dealloc_ctxt *dealloc,
137 int new_flags, int clear_flags);
138 int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et,
140 struct ocfs2_alloc_context *meta_ac,
141 struct ocfs2_cached_dealloc_ctxt *dealloc);
142 int ocfs2_remove_btree_range(struct inode *inode,
143 struct ocfs2_extent_tree *et,
144 u32 cpos, u32 phys_cpos, u32 len, int flags,
145 struct ocfs2_cached_dealloc_ctxt *dealloc,
146 u64 refcount_loc, bool refcount_tree_locked);
148 int ocfs2_num_free_extents(struct ocfs2_extent_tree *et);
151 * how many new metadata chunks would an allocation need at maximum?
153 * Please note that the caller must make sure that root_el is the root
154 * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
155 * the result may be wrong.
157 static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el)
160 * Rather than do all the work of determining how much we need
161 * (involves a ton of reads and locks), just ask for the
162 * maximal limit. That's a tree depth shift. So, one block for
163 * level of the tree (current l_tree_depth), one block for the
164 * new tree_depth==0 extent_block, and one block at the new
167 return le16_to_cpu(root_el->l_tree_depth) + 2;
170 void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di);
171 void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di);
172 int ocfs2_convert_inline_data_to_extents(struct inode *inode,
173 struct buffer_head *di_bh);
175 int ocfs2_truncate_log_init(struct ocfs2_super *osb);
176 void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
177 void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
179 int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
180 int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
182 struct ocfs2_dinode **tl_copy);
183 int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
184 struct ocfs2_dinode *tl_copy);
185 int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb);
186 int ocfs2_truncate_log_append(struct ocfs2_super *osb,
189 unsigned int num_clusters);
190 int __ocfs2_flush_truncate_log(struct ocfs2_super *osb);
191 int ocfs2_try_to_free_truncate_log(struct ocfs2_super *osb,
192 unsigned int needed);
195 * Process local structure which describes the block unlinks done
196 * during an operation. This is populated via
197 * ocfs2_cache_block_dealloc().
199 * ocfs2_run_deallocs() should be called after the potentially
200 * de-allocating routines. No journal handles should be open, and most
201 * locks should have been dropped.
203 struct ocfs2_cached_dealloc_ctxt {
204 struct ocfs2_per_slot_free_list *c_first_suballocator;
205 struct ocfs2_cached_block_free *c_global_allocator;
207 static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
209 c->c_first_suballocator = NULL;
210 c->c_global_allocator = NULL;
212 int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
213 u64 blkno, unsigned int bit);
214 int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
215 int type, int slot, u64 suballoc, u64 blkno,
217 static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
219 return c->c_global_allocator != NULL;
221 int ocfs2_run_deallocs(struct ocfs2_super *osb,
222 struct ocfs2_cached_dealloc_ctxt *ctxt);
224 struct ocfs2_truncate_context {
225 struct ocfs2_cached_dealloc_ctxt tc_dealloc;
226 int tc_ext_alloc_locked; /* is it cluster locked? */
227 /* these get destroyed once it's passed to ocfs2_commit_truncate. */
228 struct buffer_head *tc_last_eb_bh;
231 int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
232 u64 range_start, u64 range_end);
233 int ocfs2_commit_truncate(struct ocfs2_super *osb,
235 struct buffer_head *di_bh);
236 int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
237 unsigned int start, unsigned int end, int trunc);
239 int ocfs2_find_leaf(struct ocfs2_caching_info *ci,
240 struct ocfs2_extent_list *root_el, u32 cpos,
241 struct buffer_head **leaf_bh);
242 int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster);
244 int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range);
246 * Helper function to look at the # of clusters in an extent record.
248 static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el,
249 struct ocfs2_extent_rec *rec)
252 * Cluster count in extent records is slightly different
253 * between interior nodes and leaf nodes. This is to support
254 * unwritten extents which need a flags field in leaf node
255 * records, thus shrinking the available space for a clusters
258 if (el->l_tree_depth)
259 return le32_to_cpu(rec->e_int_clusters);
261 return le16_to_cpu(rec->e_leaf_clusters);
265 * This is only valid for leaf nodes, which are the only ones that can
266 * have empty extents anyway.
268 static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
270 return !rec->e_leaf_clusters;
273 int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end,
274 struct page **pages, int *num);
275 void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
276 unsigned int from, unsigned int to,
277 struct page *page, int zero, u64 *phys);
279 * Structures which describe a path through a btree, and functions to
282 * The idea here is to be as generic as possible with the tree
285 struct ocfs2_path_item {
286 struct buffer_head *bh;
287 struct ocfs2_extent_list *el;
290 #define OCFS2_MAX_PATH_DEPTH 5
294 ocfs2_journal_access_func p_root_access;
295 struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH];
298 #define path_root_bh(_path) ((_path)->p_node[0].bh)
299 #define path_root_el(_path) ((_path)->p_node[0].el)
300 #define path_root_access(_path)((_path)->p_root_access)
301 #define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
302 #define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
303 #define path_num_items(_path) ((_path)->p_tree_depth + 1)
305 void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root);
306 void ocfs2_free_path(struct ocfs2_path *path);
307 int ocfs2_find_path(struct ocfs2_caching_info *ci,
308 struct ocfs2_path *path,
310 struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path);
311 struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et);
312 int ocfs2_path_bh_journal_access(handle_t *handle,
313 struct ocfs2_caching_info *ci,
314 struct ocfs2_path *path,
316 int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
318 struct ocfs2_path *path);
319 int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
320 struct ocfs2_path *path, u32 *cpos);
321 int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
322 struct ocfs2_path *path, u32 *cpos);
323 int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
324 struct ocfs2_path *left,
325 struct ocfs2_path *right);
326 #endif /* OCFS2_ALLOC_H */