1 /* SPDX-License-Identifier: GPL-2.0-or-later */
2 /*
3 * alloc.h
4 *
5 * Function prototypes
6 *
7 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
8 */
9
10 #ifndef OCFS2_ALLOC_H
11 #define OCFS2_ALLOC_H
12
13
14 /*
15 * For xattr tree leaf, we limit the leaf byte size to be 64K.
16 */
17 #define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536
18
19 /*
20 * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract
21 * the b-tree operations in ocfs2. Now all the b-tree operations are not
22 * limited to ocfs2_dinode only. Any data which need to allocate clusters
23 * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree
24 * and operation.
25 *
26 * ocfs2_extent_tree becomes the first-class object for extent tree
27 * manipulation. Callers of the alloc.c code need to fill it via one of
28 * the ocfs2_init_*_extent_tree() operations below.
29 *
30 * ocfs2_extent_tree contains info for the root of the b-tree, it must have a
31 * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree
32 * functions. It needs the ocfs2_caching_info structure associated with
33 * I/O on the tree. With metadata ecc, we now call different journal_access
34 * functions for each type of metadata, so it must have the
35 * root_journal_access function.
36 * ocfs2_extent_tree_operations abstract the normal operations we do for
37 * the root of extent b-tree.
38 */
39 struct ocfs2_extent_tree_operations;
40 struct ocfs2_extent_tree {
41 const struct ocfs2_extent_tree_operations *et_ops;
42 struct buffer_head *et_root_bh;
43 struct ocfs2_extent_list *et_root_el;
44 struct ocfs2_caching_info *et_ci;
45 ocfs2_journal_access_func et_root_journal_access;
46 void *et_object;
47 unsigned int et_max_leaf_clusters;
48 struct ocfs2_cached_dealloc_ctxt *et_dealloc;
49 };
50
51 /*
52 * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the
53 * specified object buffer.
54 */
55 void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et,
56 struct ocfs2_caching_info *ci,
57 struct buffer_head *bh);
58 void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et,
59 struct ocfs2_caching_info *ci,
60 struct buffer_head *bh);
61 struct ocfs2_xattr_value_buf;
62 void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et,
63 struct ocfs2_caching_info *ci,
64 struct ocfs2_xattr_value_buf *vb);
65 void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et,
66 struct ocfs2_caching_info *ci,
67 struct buffer_head *bh);
68 void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et,
69 struct ocfs2_caching_info *ci,
70 struct buffer_head *bh);
71
72 /*
73 * Read an extent block into *bh. If *bh is NULL, a bh will be
74 * allocated. This is a cached read. The extent block will be validated
75 * with ocfs2_validate_extent_block().
76 */
77 int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno,
78 struct buffer_head **bh);
79
80 struct ocfs2_alloc_context;
81 int ocfs2_insert_extent(handle_t *handle,
82 struct ocfs2_extent_tree *et,
83 u32 cpos,
84 u64 start_blk,
85 u32 new_clusters,
86 u8 flags,
87 struct ocfs2_alloc_context *meta_ac);
88
89 enum ocfs2_alloc_restarted {
90 RESTART_NONE = 0,
91 RESTART_TRANS,
92 RESTART_META
93 };
94 int ocfs2_add_clusters_in_btree(handle_t *handle,
95 struct ocfs2_extent_tree *et,
96 u32 *logical_offset,
97 u32 clusters_to_add,
98 int mark_unwritten,
99 struct ocfs2_alloc_context *data_ac,
100 struct ocfs2_alloc_context *meta_ac,
101 enum ocfs2_alloc_restarted *reason_ret);
102 struct ocfs2_cached_dealloc_ctxt;
103 struct ocfs2_path;
104 int ocfs2_split_extent(handle_t *handle,
105 struct ocfs2_extent_tree *et,
106 struct ocfs2_path *path,
107 int split_index,
108 struct ocfs2_extent_rec *split_rec,
109 struct ocfs2_alloc_context *meta_ac,
110 struct ocfs2_cached_dealloc_ctxt *dealloc);
111 int ocfs2_mark_extent_written(struct inode *inode,
112 struct ocfs2_extent_tree *et,
113 handle_t *handle, u32 cpos, u32 len, u32 phys,
114 struct ocfs2_alloc_context *meta_ac,
115 struct ocfs2_cached_dealloc_ctxt *dealloc);
116 int ocfs2_change_extent_flag(handle_t *handle,
117 struct ocfs2_extent_tree *et,
118 u32 cpos, u32 len, u32 phys,
119 struct ocfs2_alloc_context *meta_ac,
120 struct ocfs2_cached_dealloc_ctxt *dealloc,
121 int new_flags, int clear_flags);
122 int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et,
123 u32 cpos, u32 len,
124 struct ocfs2_alloc_context *meta_ac,
125 struct ocfs2_cached_dealloc_ctxt *dealloc);
126 int ocfs2_remove_btree_range(struct inode *inode,
127 struct ocfs2_extent_tree *et,
128 u32 cpos, u32 phys_cpos, u32 len, int flags,
129 struct ocfs2_cached_dealloc_ctxt *dealloc,
130 u64 refcount_loc, bool refcount_tree_locked);
131
132 int ocfs2_num_free_extents(struct ocfs2_extent_tree *et);
133
134 /*
135 * how many new metadata chunks would an allocation need at maximum?
136 *
137 * Please note that the caller must make sure that root_el is the root
138 * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
139 * the result may be wrong.
140 */
ocfs2_extend_meta_needed(struct ocfs2_extent_list * root_el)141 static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el)
142 {
143 /*
144 * Rather than do all the work of determining how much we need
145 * (involves a ton of reads and locks), just ask for the
146 * maximal limit. That's a tree depth shift. So, one block for
147 * level of the tree (current l_tree_depth), one block for the
148 * new tree_depth==0 extent_block, and one block at the new
149 * top-of-the tree.
150 */
151 return le16_to_cpu(root_el->l_tree_depth) + 2;
152 }
153
154 void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di);
155 void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di);
156 int ocfs2_convert_inline_data_to_extents(struct inode *inode,
157 struct buffer_head *di_bh);
158
159 int ocfs2_truncate_log_init(struct ocfs2_super *osb);
160 void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
161 void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
162 int cancel);
163 int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
164 int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
165 int slot_num,
166 struct ocfs2_dinode **tl_copy);
167 int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
168 struct ocfs2_dinode *tl_copy);
169 int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb);
170 int ocfs2_truncate_log_append(struct ocfs2_super *osb,
171 handle_t *handle,
172 u64 start_blk,
173 unsigned int num_clusters);
174 int __ocfs2_flush_truncate_log(struct ocfs2_super *osb);
175 int ocfs2_try_to_free_truncate_log(struct ocfs2_super *osb,
176 unsigned int needed);
177
178 /*
179 * Process local structure which describes the block unlinks done
180 * during an operation. This is populated via
181 * ocfs2_cache_block_dealloc().
182 *
183 * ocfs2_run_deallocs() should be called after the potentially
184 * de-allocating routines. No journal handles should be open, and most
185 * locks should have been dropped.
186 */
187 struct ocfs2_cached_dealloc_ctxt {
188 struct ocfs2_per_slot_free_list *c_first_suballocator;
189 struct ocfs2_cached_block_free *c_global_allocator;
190 };
ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt * c)191 static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
192 {
193 c->c_first_suballocator = NULL;
194 c->c_global_allocator = NULL;
195 }
196 int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
197 u64 blkno, unsigned int bit);
198 int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
199 int type, int slot, u64 suballoc, u64 blkno,
200 unsigned int bit);
ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt * c)201 static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
202 {
203 return c->c_global_allocator != NULL;
204 }
205 int ocfs2_run_deallocs(struct ocfs2_super *osb,
206 struct ocfs2_cached_dealloc_ctxt *ctxt);
207
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
215 int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
216 u64 range_start, u64 range_end);
217 int ocfs2_commit_truncate(struct ocfs2_super *osb,
218 struct inode *inode,
219 struct buffer_head *di_bh);
220 int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
221 unsigned int start, unsigned int end, int trunc);
222
223 int ocfs2_find_leaf(struct ocfs2_caching_info *ci,
224 struct ocfs2_extent_list *root_el, u32 cpos,
225 struct buffer_head **leaf_bh);
226 int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster);
227
228 int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range);
229 /*
230 * Helper function to look at the # of clusters in an extent record.
231 */
ocfs2_rec_clusters(struct ocfs2_extent_list * el,struct ocfs2_extent_rec * rec)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 */
ocfs2_is_empty_extent(struct ocfs2_extent_rec * rec)252 static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
253 {
254 return !rec->e_leaf_clusters;
255 }
256
257 int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end,
258 struct page **pages, int *num);
259 void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
260 unsigned int from, unsigned int to,
261 struct page *page, int zero, u64 *phys);
262 /*
263 * Structures which describe a path through a btree, and functions to
264 * manipulate them.
265 *
266 * The idea here is to be as generic as possible with the tree
267 * manipulation code.
268 */
269 struct ocfs2_path_item {
270 struct buffer_head *bh;
271 struct ocfs2_extent_list *el;
272 };
273
274 #define OCFS2_MAX_PATH_DEPTH 5
275
276 struct ocfs2_path {
277 int p_tree_depth;
278 ocfs2_journal_access_func p_root_access;
279 struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH];
280 };
281
282 #define path_root_bh(_path) ((_path)->p_node[0].bh)
283 #define path_root_el(_path) ((_path)->p_node[0].el)
284 #define path_root_access(_path)((_path)->p_root_access)
285 #define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
286 #define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
287 #define path_num_items(_path) ((_path)->p_tree_depth + 1)
288
289 void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root);
290 void ocfs2_free_path(struct ocfs2_path *path);
291 int ocfs2_find_path(struct ocfs2_caching_info *ci,
292 struct ocfs2_path *path,
293 u32 cpos);
294 struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path);
295 struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et);
296 int ocfs2_path_bh_journal_access(handle_t *handle,
297 struct ocfs2_caching_info *ci,
298 struct ocfs2_path *path,
299 int idx);
300 int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
301 handle_t *handle,
302 struct ocfs2_path *path);
303 int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
304 struct ocfs2_path *path, u32 *cpos);
305 int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
306 struct ocfs2_path *path, u32 *cpos);
307 int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
308 struct ocfs2_path *left,
309 struct ocfs2_path *right);
310 #endif /* OCFS2_ALLOC_H */
311