blob: 3bd08a03251c32fe904af81cb49305f72dba1e41 [file] [log] [blame]
Mark Fashehccd979b2005-12-15 14:31:24 -08001/* -*- 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
Tao Mae7d4cb62008-08-18 17:38:44 +080029
Tao Maca12b7c2008-08-18 17:38:52 +080030/*
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
Joel Beckerf99b9b72008-08-20 19:36:33 -070035/*
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
Joel Becker8d6220d2008-08-22 12:46:09 -070044 * the ocfs2_init_*_extent_tree() operations below.
Joel Beckerf99b9b72008-08-20 19:36:33 -070045 *
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
Joel Beckerd9a0a1f2009-02-12 17:32:34 -080048 * 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
Joel Becker13723d02008-10-17 19:25:01 -070050 * functions for each type of metadata, so it must have the
51 * root_journal_access function.
Joel Beckerf99b9b72008-08-20 19:36:33 -070052 * ocfs2_extent_tree_operations abstract the normal operations we do for
53 * the root of extent b-tree.
54 */
55struct ocfs2_extent_tree_operations;
56struct 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;
Joel Beckerd9a0a1f2009-02-12 17:32:34 -080060 struct ocfs2_caching_info *et_ci;
Joel Becker13723d02008-10-17 19:25:01 -070061 ocfs2_journal_access_func et_root_journal_access;
Joel Beckerf99b9b72008-08-20 19:36:33 -070062 void *et_object;
63 unsigned int et_max_leaf_clusters;
64};
65
66/*
Joel Becker8d6220d2008-08-22 12:46:09 -070067 * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the
68 * specified object buffer.
Joel Beckerf99b9b72008-08-20 19:36:33 -070069 */
Joel Becker8d6220d2008-08-22 12:46:09 -070070void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et,
Joel Becker5e404e92009-02-13 03:54:22 -080071 struct ocfs2_caching_info *ci,
Joel Becker8d6220d2008-08-22 12:46:09 -070072 struct buffer_head *bh);
73void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et,
Joel Becker5e404e92009-02-13 03:54:22 -080074 struct ocfs2_caching_info *ci,
Joel Becker8d6220d2008-08-22 12:46:09 -070075 struct buffer_head *bh);
Joel Becker2a50a742008-12-09 14:24:33 -080076struct ocfs2_xattr_value_buf;
Joel Becker8d6220d2008-08-22 12:46:09 -070077void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et,
Joel Becker5e404e92009-02-13 03:54:22 -080078 struct ocfs2_caching_info *ci,
Joel Becker2a50a742008-12-09 14:24:33 -080079 struct ocfs2_xattr_value_buf *vb);
Mark Fasheh9b7895e2008-11-12 16:27:44 -080080void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et,
Joel Becker5e404e92009-02-13 03:54:22 -080081 struct ocfs2_caching_info *ci,
Mark Fasheh9b7895e2008-11-12 16:27:44 -080082 struct buffer_head *bh);
Tao Mafe924412009-08-18 11:22:25 +080083void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et,
84 struct ocfs2_caching_info *ci,
85 struct buffer_head *bh);
Joel Beckerf99b9b72008-08-20 19:36:33 -070086
Joel Becker5e965812008-11-13 14:49:16 -080087/*
88 * Read an extent block into *bh. If *bh is NULL, a bh will be
89 * allocated. This is a cached read. The extent block will be validated
90 * with ocfs2_validate_extent_block().
91 */
Joel Becker3d03a302009-02-12 17:49:26 -080092int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno,
Joel Becker5e965812008-11-13 14:49:16 -080093 struct buffer_head **bh);
94
Mark Fashehccd979b2005-12-15 14:31:24 -080095struct ocfs2_alloc_context;
Joel Beckercc79d8c2009-02-13 03:24:43 -080096int ocfs2_insert_extent(handle_t *handle,
Joel Beckerf99b9b72008-08-20 19:36:33 -070097 struct ocfs2_extent_tree *et,
98 u32 cpos,
99 u64 start_blk,
100 u32 new_clusters,
101 u8 flags,
102 struct ocfs2_alloc_context *meta_ac);
103
Tao Ma0eb8d472008-08-18 17:38:45 +0800104enum ocfs2_alloc_restarted {
105 RESTART_NONE = 0,
106 RESTART_TRANS,
107 RESTART_META
108};
Joel Beckercbee7e12009-02-13 03:34:15 -0800109int ocfs2_add_clusters_in_btree(handle_t *handle,
110 struct ocfs2_extent_tree *et,
Tao Ma0eb8d472008-08-18 17:38:45 +0800111 u32 *logical_offset,
112 u32 clusters_to_add,
113 int mark_unwritten,
Tao Ma0eb8d472008-08-18 17:38:45 +0800114 struct ocfs2_alloc_context *data_ac,
115 struct ocfs2_alloc_context *meta_ac,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700116 enum ocfs2_alloc_restarted *reason_ret);
Mark Fasheh328d5752007-06-18 10:48:04 -0700117struct ocfs2_cached_dealloc_ctxt;
Tao Mae2e9f602009-08-18 11:22:34 +0800118struct ocfs2_path;
119int ocfs2_split_extent(handle_t *handle,
120 struct ocfs2_extent_tree *et,
121 struct ocfs2_path *path,
122 int split_index,
123 struct ocfs2_extent_rec *split_rec,
124 struct ocfs2_alloc_context *meta_ac,
125 struct ocfs2_cached_dealloc_ctxt *dealloc);
Joel Beckerf99b9b72008-08-20 19:36:33 -0700126int ocfs2_mark_extent_written(struct inode *inode,
127 struct ocfs2_extent_tree *et,
Mark Fasheh328d5752007-06-18 10:48:04 -0700128 handle_t *handle, u32 cpos, u32 len, u32 phys,
129 struct ocfs2_alloc_context *meta_ac,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700130 struct ocfs2_cached_dealloc_ctxt *dealloc);
Tao Ma1aa75fe2009-08-18 11:28:39 +0800131int ocfs2_change_extent_flag(handle_t *handle,
132 struct ocfs2_extent_tree *et,
133 u32 cpos, u32 len, u32 phys,
134 struct ocfs2_alloc_context *meta_ac,
135 struct ocfs2_cached_dealloc_ctxt *dealloc,
136 int new_flags, int clear_flags);
Joel Beckerdbdcf6a2009-02-13 03:41:26 -0800137int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et,
138 u32 cpos, u32 len,
Mark Fasheh063c4562007-07-03 13:34:11 -0700139 struct ocfs2_alloc_context *meta_ac,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700140 struct ocfs2_cached_dealloc_ctxt *dealloc);
Mark Fashehfecc0112008-11-12 15:16:38 -0800141int ocfs2_remove_btree_range(struct inode *inode,
142 struct ocfs2_extent_tree *et,
Tristan Ye78f94672010-05-11 17:54:42 +0800143 u32 cpos, u32 phys_cpos, u32 len, int flags,
144 struct ocfs2_cached_dealloc_ctxt *dealloc,
145 u64 refcount_loc);
Mark Fashehfecc0112008-11-12 15:16:38 -0800146
Mark Fashehccd979b2005-12-15 14:31:24 -0800147int ocfs2_num_free_extents(struct ocfs2_super *osb,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700148 struct ocfs2_extent_tree *et);
Tao Mae7d4cb62008-08-18 17:38:44 +0800149
Tao Ma811f9332008-08-18 17:38:43 +0800150/*
151 * how many new metadata chunks would an allocation need at maximum?
152 *
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.
156 */
157static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el)
Mark Fashehccd979b2005-12-15 14:31:24 -0800158{
159 /*
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
165 * top-of-the tree.
166 */
Tao Ma811f9332008-08-18 17:38:43 +0800167 return le16_to_cpu(root_el->l_tree_depth) + 2;
Mark Fashehccd979b2005-12-15 14:31:24 -0800168}
169
Mark Fasheh5b6a3a22007-09-13 16:33:54 -0700170void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di);
Mark Fasheh1afc32b2007-09-07 14:46:51 -0700171void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di);
172int ocfs2_convert_inline_data_to_extents(struct inode *inode,
173 struct buffer_head *di_bh);
174
Mark Fashehccd979b2005-12-15 14:31:24 -0800175int ocfs2_truncate_log_init(struct ocfs2_super *osb);
176void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
177void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
178 int cancel);
179int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
180int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
181 int slot_num,
182 struct ocfs2_dinode **tl_copy);
183int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
184 struct ocfs2_dinode *tl_copy);
Mark Fasheh063c4562007-07-03 13:34:11 -0700185int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb);
186int ocfs2_truncate_log_append(struct ocfs2_super *osb,
187 handle_t *handle,
188 u64 start_blk,
189 unsigned int num_clusters);
190int __ocfs2_flush_truncate_log(struct ocfs2_super *osb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800191
Mark Fasheh2b604352007-06-22 15:45:27 -0700192/*
193 * Process local structure which describes the block unlinks done
194 * during an operation. This is populated via
195 * ocfs2_cache_block_dealloc().
196 *
197 * ocfs2_run_deallocs() should be called after the potentially
198 * de-allocating routines. No journal handles should be open, and most
199 * locks should have been dropped.
200 */
201struct ocfs2_cached_dealloc_ctxt {
202 struct ocfs2_per_slot_free_list *c_first_suballocator;
Tao Ma2891d292008-11-12 08:26:58 +0800203 struct ocfs2_cached_block_free *c_global_allocator;
Mark Fasheh2b604352007-06-22 15:45:27 -0700204};
205static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
206{
207 c->c_first_suballocator = NULL;
Tao Ma2891d292008-11-12 08:26:58 +0800208 c->c_global_allocator = NULL;
Mark Fasheh2b604352007-06-22 15:45:27 -0700209}
Tao Ma2891d292008-11-12 08:26:58 +0800210int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
211 u64 blkno, unsigned int bit);
Tao Ma1823cb02009-08-18 11:24:49 +0800212int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
Tao Ma74380c42010-03-22 14:20:18 +0800213 int type, int slot, u64 suballoc, u64 blkno,
Tao Ma1823cb02009-08-18 11:24:49 +0800214 unsigned int bit);
Tao Ma78f30c32008-11-12 08:27:00 +0800215static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
216{
217 return c->c_global_allocator != NULL;
218}
Mark Fasheh2b604352007-06-22 15:45:27 -0700219int ocfs2_run_deallocs(struct ocfs2_super *osb,
220 struct ocfs2_cached_dealloc_ctxt *ctxt);
221
Mark Fashehccd979b2005-12-15 14:31:24 -0800222struct ocfs2_truncate_context {
Mark Fasheh59a5e412007-06-22 15:52:36 -0700223 struct ocfs2_cached_dealloc_ctxt tc_dealloc;
Mark Fashehccd979b2005-12-15 14:31:24 -0800224 int tc_ext_alloc_locked; /* is it cluster locked? */
225 /* these get destroyed once it's passed to ocfs2_commit_truncate. */
226 struct buffer_head *tc_last_eb_bh;
227};
228
Mark Fasheh35edec12007-07-06 14:41:18 -0700229int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
230 u64 range_start, u64 range_end);
Mark Fashehccd979b2005-12-15 14:31:24 -0800231int ocfs2_commit_truncate(struct ocfs2_super *osb,
232 struct inode *inode,
Tristan Ye78f94672010-05-11 17:54:42 +0800233 struct buffer_head *di_bh);
Mark Fasheh1afc32b2007-09-07 14:46:51 -0700234int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
235 unsigned int start, unsigned int end, int trunc);
Mark Fashehccd979b2005-12-15 14:31:24 -0800236
Joel Beckerfacdb772009-02-12 18:08:48 -0800237int ocfs2_find_leaf(struct ocfs2_caching_info *ci,
238 struct ocfs2_extent_list *root_el, u32 cpos,
239 struct buffer_head **leaf_bh);
Mark Fasheh328d5752007-06-18 10:48:04 -0700240int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster);
Mark Fasheh363041a2007-01-17 12:31:35 -0800241
Mark Fashehe48edee2007-03-07 16:46:57 -0800242/*
243 * Helper function to look at the # of clusters in an extent record.
244 */
245static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el,
246 struct ocfs2_extent_rec *rec)
247{
248 /*
249 * Cluster count in extent records is slightly different
250 * between interior nodes and leaf nodes. This is to support
251 * unwritten extents which need a flags field in leaf node
252 * records, thus shrinking the available space for a clusters
253 * field.
254 */
255 if (el->l_tree_depth)
256 return le32_to_cpu(rec->e_int_clusters);
257 else
258 return le16_to_cpu(rec->e_leaf_clusters);
259}
260
Mark Fasheh00dc4172008-10-03 17:32:11 -0400261/*
262 * This is only valid for leaf nodes, which are the only ones that can
263 * have empty extents anyway.
264 */
265static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
266{
267 return !rec->e_leaf_clusters;
268}
269
Tao Ma6f70fa52009-08-25 08:05:12 +0800270int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end,
271 struct page **pages, int *num);
272void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
273 unsigned int from, unsigned int to,
274 struct page *page, int zero, u64 *phys);
Tao Mae2e9f602009-08-18 11:22:34 +0800275/*
276 * Structures which describe a path through a btree, and functions to
277 * manipulate them.
278 *
279 * The idea here is to be as generic as possible with the tree
280 * manipulation code.
281 */
282struct ocfs2_path_item {
283 struct buffer_head *bh;
284 struct ocfs2_extent_list *el;
285};
286
287#define OCFS2_MAX_PATH_DEPTH 5
288
289struct ocfs2_path {
290 int p_tree_depth;
291 ocfs2_journal_access_func p_root_access;
292 struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH];
293};
294
295#define path_root_bh(_path) ((_path)->p_node[0].bh)
296#define path_root_el(_path) ((_path)->p_node[0].el)
297#define path_root_access(_path)((_path)->p_root_access)
298#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
299#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
300#define path_num_items(_path) ((_path)->p_tree_depth + 1)
301
302void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root);
303void ocfs2_free_path(struct ocfs2_path *path);
304int ocfs2_find_path(struct ocfs2_caching_info *ci,
305 struct ocfs2_path *path,
306 u32 cpos);
307struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path);
308struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et);
309int ocfs2_path_bh_journal_access(handle_t *handle,
310 struct ocfs2_caching_info *ci,
311 struct ocfs2_path *path,
312 int idx);
313int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
314 handle_t *handle,
315 struct ocfs2_path *path);
Tao Ma38a04e42009-11-30 14:32:19 +0800316int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
317 struct ocfs2_path *path, u32 *cpos);
Tristan Yeee149a72010-05-11 17:54:44 +0800318int ocfs2_find_cpos_for_left_leaf(struct super_block *sb,
319 struct ocfs2_path *path, u32 *cpos);
Tao Ma38a04e42009-11-30 14:32:19 +0800320int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
321 struct ocfs2_path *left,
322 struct ocfs2_path *right);
Mark Fashehccd979b2005-12-15 14:31:24 -0800323#endif /* OCFS2_ALLOC_H */