blob: cceff5c37f47c5872284d5ab43fafab9f9e27db1 [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 Becker13723d02008-10-17 19:25:01 -070048 * functions. With metadata ecc, we now call different journal_access
49 * functions for each type of metadata, so it must have the
50 * root_journal_access function.
Joel Beckerf99b9b72008-08-20 19:36:33 -070051 * ocfs2_extent_tree_operations abstract the normal operations we do for
52 * the root of extent b-tree.
53 */
54struct ocfs2_extent_tree_operations;
55struct ocfs2_extent_tree {
56 struct ocfs2_extent_tree_operations *et_ops;
57 struct buffer_head *et_root_bh;
58 struct ocfs2_extent_list *et_root_el;
Joel Becker13723d02008-10-17 19:25:01 -070059 ocfs2_journal_access_func et_root_journal_access;
Joel Beckerf99b9b72008-08-20 19:36:33 -070060 void *et_object;
61 unsigned int et_max_leaf_clusters;
62};
63
64/*
Joel Becker8d6220d2008-08-22 12:46:09 -070065 * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the
66 * specified object buffer.
Joel Beckerf99b9b72008-08-20 19:36:33 -070067 */
Joel Becker8d6220d2008-08-22 12:46:09 -070068void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et,
69 struct inode *inode,
70 struct buffer_head *bh);
71void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et,
Joel Beckerf99b9b72008-08-20 19:36:33 -070072 struct inode *inode,
Joel Becker8d6220d2008-08-22 12:46:09 -070073 struct buffer_head *bh);
Joel Becker2a50a742008-12-09 14:24:33 -080074struct ocfs2_xattr_value_buf;
Joel Becker8d6220d2008-08-22 12:46:09 -070075void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et,
76 struct inode *inode,
Joel Becker2a50a742008-12-09 14:24:33 -080077 struct ocfs2_xattr_value_buf *vb);
Joel Beckerf99b9b72008-08-20 19:36:33 -070078
Joel Becker5e965812008-11-13 14:49:16 -080079/*
80 * Read an extent block into *bh. If *bh is NULL, a bh will be
81 * allocated. This is a cached read. The extent block will be validated
82 * with ocfs2_validate_extent_block().
83 */
84int ocfs2_read_extent_block(struct inode *inode, u64 eb_blkno,
85 struct buffer_head **bh);
86
Mark Fashehccd979b2005-12-15 14:31:24 -080087struct ocfs2_alloc_context;
Joel Beckerf99b9b72008-08-20 19:36:33 -070088int ocfs2_insert_extent(struct ocfs2_super *osb,
89 handle_t *handle,
90 struct inode *inode,
91 struct ocfs2_extent_tree *et,
92 u32 cpos,
93 u64 start_blk,
94 u32 new_clusters,
95 u8 flags,
96 struct ocfs2_alloc_context *meta_ac);
97
Tao Ma0eb8d472008-08-18 17:38:45 +080098enum ocfs2_alloc_restarted {
99 RESTART_NONE = 0,
100 RESTART_TRANS,
101 RESTART_META
102};
103int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb,
104 struct inode *inode,
105 u32 *logical_offset,
106 u32 clusters_to_add,
107 int mark_unwritten,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700108 struct ocfs2_extent_tree *et,
Tao Ma0eb8d472008-08-18 17:38:45 +0800109 handle_t *handle,
110 struct ocfs2_alloc_context *data_ac,
111 struct ocfs2_alloc_context *meta_ac,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700112 enum ocfs2_alloc_restarted *reason_ret);
Mark Fasheh328d5752007-06-18 10:48:04 -0700113struct ocfs2_cached_dealloc_ctxt;
Joel Beckerf99b9b72008-08-20 19:36:33 -0700114int ocfs2_mark_extent_written(struct inode *inode,
115 struct ocfs2_extent_tree *et,
Mark Fasheh328d5752007-06-18 10:48:04 -0700116 handle_t *handle, u32 cpos, u32 len, u32 phys,
117 struct ocfs2_alloc_context *meta_ac,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700118 struct ocfs2_cached_dealloc_ctxt *dealloc);
119int ocfs2_remove_extent(struct inode *inode,
120 struct ocfs2_extent_tree *et,
Mark Fasheh063c4562007-07-03 13:34:11 -0700121 u32 cpos, u32 len, handle_t *handle,
122 struct ocfs2_alloc_context *meta_ac,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700123 struct ocfs2_cached_dealloc_ctxt *dealloc);
Mark Fashehfecc0112008-11-12 15:16:38 -0800124int ocfs2_remove_btree_range(struct inode *inode,
125 struct ocfs2_extent_tree *et,
126 u32 cpos, u32 phys_cpos, u32 len,
127 struct ocfs2_cached_dealloc_ctxt *dealloc);
128
Mark Fashehccd979b2005-12-15 14:31:24 -0800129int ocfs2_num_free_extents(struct ocfs2_super *osb,
130 struct inode *inode,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700131 struct ocfs2_extent_tree *et);
Tao Mae7d4cb62008-08-18 17:38:44 +0800132
Tao Ma811f9332008-08-18 17:38:43 +0800133/*
134 * how many new metadata chunks would an allocation need at maximum?
135 *
136 * Please note that the caller must make sure that root_el is the root
137 * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
138 * the result may be wrong.
139 */
140static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el)
Mark Fashehccd979b2005-12-15 14:31:24 -0800141{
142 /*
143 * Rather than do all the work of determining how much we need
144 * (involves a ton of reads and locks), just ask for the
145 * maximal limit. That's a tree depth shift. So, one block for
146 * level of the tree (current l_tree_depth), one block for the
147 * new tree_depth==0 extent_block, and one block at the new
148 * top-of-the tree.
149 */
Tao Ma811f9332008-08-18 17:38:43 +0800150 return le16_to_cpu(root_el->l_tree_depth) + 2;
Mark Fashehccd979b2005-12-15 14:31:24 -0800151}
152
Mark Fasheh5b6a3a22007-09-13 16:33:54 -0700153void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di);
Mark Fasheh1afc32b2007-09-07 14:46:51 -0700154void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di);
155int ocfs2_convert_inline_data_to_extents(struct inode *inode,
156 struct buffer_head *di_bh);
157
Mark Fashehccd979b2005-12-15 14:31:24 -0800158int ocfs2_truncate_log_init(struct ocfs2_super *osb);
159void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
160void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
161 int cancel);
162int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
163int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
164 int slot_num,
165 struct ocfs2_dinode **tl_copy);
166int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
167 struct ocfs2_dinode *tl_copy);
Mark Fasheh063c4562007-07-03 13:34:11 -0700168int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb);
169int ocfs2_truncate_log_append(struct ocfs2_super *osb,
170 handle_t *handle,
171 u64 start_blk,
172 unsigned int num_clusters);
173int __ocfs2_flush_truncate_log(struct ocfs2_super *osb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800174
Mark Fasheh2b604352007-06-22 15:45:27 -0700175/*
176 * Process local structure which describes the block unlinks done
177 * during an operation. This is populated via
178 * ocfs2_cache_block_dealloc().
179 *
180 * ocfs2_run_deallocs() should be called after the potentially
181 * de-allocating routines. No journal handles should be open, and most
182 * locks should have been dropped.
183 */
184struct ocfs2_cached_dealloc_ctxt {
185 struct ocfs2_per_slot_free_list *c_first_suballocator;
Tao Ma2891d292008-11-12 08:26:58 +0800186 struct ocfs2_cached_block_free *c_global_allocator;
Mark Fasheh2b604352007-06-22 15:45:27 -0700187};
188static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
189{
190 c->c_first_suballocator = NULL;
Tao Ma2891d292008-11-12 08:26:58 +0800191 c->c_global_allocator = NULL;
Mark Fasheh2b604352007-06-22 15:45:27 -0700192}
Tao Ma2891d292008-11-12 08:26:58 +0800193int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
194 u64 blkno, unsigned int bit);
Tao Ma78f30c32008-11-12 08:27:00 +0800195static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
196{
197 return c->c_global_allocator != NULL;
198}
Mark Fasheh2b604352007-06-22 15:45:27 -0700199int ocfs2_run_deallocs(struct ocfs2_super *osb,
200 struct ocfs2_cached_dealloc_ctxt *ctxt);
201
Mark Fashehccd979b2005-12-15 14:31:24 -0800202struct ocfs2_truncate_context {
Mark Fasheh59a5e412007-06-22 15:52:36 -0700203 struct ocfs2_cached_dealloc_ctxt tc_dealloc;
Mark Fashehccd979b2005-12-15 14:31:24 -0800204 int tc_ext_alloc_locked; /* is it cluster locked? */
205 /* these get destroyed once it's passed to ocfs2_commit_truncate. */
206 struct buffer_head *tc_last_eb_bh;
207};
208
Mark Fasheh35edec12007-07-06 14:41:18 -0700209int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
210 u64 range_start, u64 range_end);
Mark Fashehccd979b2005-12-15 14:31:24 -0800211int ocfs2_prepare_truncate(struct ocfs2_super *osb,
212 struct inode *inode,
213 struct buffer_head *fe_bh,
214 struct ocfs2_truncate_context **tc);
215int ocfs2_commit_truncate(struct ocfs2_super *osb,
216 struct inode *inode,
217 struct buffer_head *fe_bh,
218 struct ocfs2_truncate_context *tc);
Mark Fasheh1afc32b2007-09-07 14:46:51 -0700219int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
220 unsigned int start, unsigned int end, int trunc);
Mark Fashehccd979b2005-12-15 14:31:24 -0800221
Mark Fasheh363041a2007-01-17 12:31:35 -0800222int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el,
223 u32 cpos, struct buffer_head **leaf_bh);
Mark Fasheh328d5752007-06-18 10:48:04 -0700224int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster);
Mark Fasheh363041a2007-01-17 12:31:35 -0800225
Mark Fashehe48edee2007-03-07 16:46:57 -0800226/*
227 * Helper function to look at the # of clusters in an extent record.
228 */
229static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el,
230 struct ocfs2_extent_rec *rec)
231{
232 /*
233 * Cluster count in extent records is slightly different
234 * between interior nodes and leaf nodes. This is to support
235 * unwritten extents which need a flags field in leaf node
236 * records, thus shrinking the available space for a clusters
237 * field.
238 */
239 if (el->l_tree_depth)
240 return le32_to_cpu(rec->e_int_clusters);
241 else
242 return le16_to_cpu(rec->e_leaf_clusters);
243}
244
Mark Fasheh00dc4172008-10-03 17:32:11 -0400245/*
246 * This is only valid for leaf nodes, which are the only ones that can
247 * have empty extents anyway.
248 */
249static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
250{
251 return !rec->e_leaf_clusters;
252}
253
Mark Fashehccd979b2005-12-15 14:31:24 -0800254#endif /* OCFS2_ALLOC_H */