blob: 1db4359ccb90b61521a0167be7581eb91e823d83 [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,
143 u32 cpos, u32 phys_cpos, u32 len,
144 struct ocfs2_cached_dealloc_ctxt *dealloc);
145
Mark Fashehccd979b2005-12-15 14:31:24 -0800146int ocfs2_num_free_extents(struct ocfs2_super *osb,
Joel Beckerf99b9b72008-08-20 19:36:33 -0700147 struct ocfs2_extent_tree *et);
Tao Mae7d4cb62008-08-18 17:38:44 +0800148
Tao Ma811f9332008-08-18 17:38:43 +0800149/*
150 * how many new metadata chunks would an allocation need at maximum?
151 *
152 * Please note that the caller must make sure that root_el is the root
153 * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise
154 * the result may be wrong.
155 */
156static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el)
Mark Fashehccd979b2005-12-15 14:31:24 -0800157{
158 /*
159 * Rather than do all the work of determining how much we need
160 * (involves a ton of reads and locks), just ask for the
161 * maximal limit. That's a tree depth shift. So, one block for
162 * level of the tree (current l_tree_depth), one block for the
163 * new tree_depth==0 extent_block, and one block at the new
164 * top-of-the tree.
165 */
Tao Ma811f9332008-08-18 17:38:43 +0800166 return le16_to_cpu(root_el->l_tree_depth) + 2;
Mark Fashehccd979b2005-12-15 14:31:24 -0800167}
168
Mark Fasheh5b6a3a22007-09-13 16:33:54 -0700169void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di);
Mark Fasheh1afc32b2007-09-07 14:46:51 -0700170void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di);
171int ocfs2_convert_inline_data_to_extents(struct inode *inode,
172 struct buffer_head *di_bh);
173
Mark Fashehccd979b2005-12-15 14:31:24 -0800174int ocfs2_truncate_log_init(struct ocfs2_super *osb);
175void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
176void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
177 int cancel);
178int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
179int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
180 int slot_num,
181 struct ocfs2_dinode **tl_copy);
182int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
183 struct ocfs2_dinode *tl_copy);
Mark Fasheh063c4562007-07-03 13:34:11 -0700184int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb);
185int ocfs2_truncate_log_append(struct ocfs2_super *osb,
186 handle_t *handle,
187 u64 start_blk,
188 unsigned int num_clusters);
189int __ocfs2_flush_truncate_log(struct ocfs2_super *osb);
Mark Fashehccd979b2005-12-15 14:31:24 -0800190
Mark Fasheh2b604352007-06-22 15:45:27 -0700191/*
192 * Process local structure which describes the block unlinks done
193 * during an operation. This is populated via
194 * ocfs2_cache_block_dealloc().
195 *
196 * ocfs2_run_deallocs() should be called after the potentially
197 * de-allocating routines. No journal handles should be open, and most
198 * locks should have been dropped.
199 */
200struct ocfs2_cached_dealloc_ctxt {
201 struct ocfs2_per_slot_free_list *c_first_suballocator;
Tao Ma2891d292008-11-12 08:26:58 +0800202 struct ocfs2_cached_block_free *c_global_allocator;
Mark Fasheh2b604352007-06-22 15:45:27 -0700203};
204static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c)
205{
206 c->c_first_suballocator = NULL;
Tao Ma2891d292008-11-12 08:26:58 +0800207 c->c_global_allocator = NULL;
Mark Fasheh2b604352007-06-22 15:45:27 -0700208}
Tao Ma2891d292008-11-12 08:26:58 +0800209int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
210 u64 blkno, unsigned int bit);
Tao Ma1823cb02009-08-18 11:24:49 +0800211int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt,
212 int type, int slot, u64 blkno,
213 unsigned int bit);
Tao Ma78f30c32008-11-12 08:27:00 +0800214static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c)
215{
216 return c->c_global_allocator != NULL;
217}
Mark Fasheh2b604352007-06-22 15:45:27 -0700218int ocfs2_run_deallocs(struct ocfs2_super *osb,
219 struct ocfs2_cached_dealloc_ctxt *ctxt);
220
Mark Fashehccd979b2005-12-15 14:31:24 -0800221struct ocfs2_truncate_context {
Mark Fasheh59a5e412007-06-22 15:52:36 -0700222 struct ocfs2_cached_dealloc_ctxt tc_dealloc;
Mark Fashehccd979b2005-12-15 14:31:24 -0800223 int tc_ext_alloc_locked; /* is it cluster locked? */
224 /* these get destroyed once it's passed to ocfs2_commit_truncate. */
225 struct buffer_head *tc_last_eb_bh;
226};
227
Mark Fasheh35edec12007-07-06 14:41:18 -0700228int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle,
229 u64 range_start, u64 range_end);
Mark Fashehccd979b2005-12-15 14:31:24 -0800230int ocfs2_prepare_truncate(struct ocfs2_super *osb,
231 struct inode *inode,
232 struct buffer_head *fe_bh,
233 struct ocfs2_truncate_context **tc);
234int ocfs2_commit_truncate(struct ocfs2_super *osb,
235 struct inode *inode,
236 struct buffer_head *fe_bh,
237 struct ocfs2_truncate_context *tc);
Mark Fasheh1afc32b2007-09-07 14:46:51 -0700238int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh,
239 unsigned int start, unsigned int end, int trunc);
Mark Fashehccd979b2005-12-15 14:31:24 -0800240
Joel Beckerfacdb772009-02-12 18:08:48 -0800241int ocfs2_find_leaf(struct ocfs2_caching_info *ci,
242 struct ocfs2_extent_list *root_el, u32 cpos,
243 struct buffer_head **leaf_bh);
Mark Fasheh328d5752007-06-18 10:48:04 -0700244int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster);
Mark Fasheh363041a2007-01-17 12:31:35 -0800245
Mark Fashehe48edee2007-03-07 16:46:57 -0800246/*
247 * Helper function to look at the # of clusters in an extent record.
248 */
249static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el,
250 struct ocfs2_extent_rec *rec)
251{
252 /*
253 * Cluster count in extent records is slightly different
254 * between interior nodes and leaf nodes. This is to support
255 * unwritten extents which need a flags field in leaf node
256 * records, thus shrinking the available space for a clusters
257 * field.
258 */
259 if (el->l_tree_depth)
260 return le32_to_cpu(rec->e_int_clusters);
261 else
262 return le16_to_cpu(rec->e_leaf_clusters);
263}
264
Mark Fasheh00dc4172008-10-03 17:32:11 -0400265/*
266 * This is only valid for leaf nodes, which are the only ones that can
267 * have empty extents anyway.
268 */
269static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec)
270{
271 return !rec->e_leaf_clusters;
272}
273
Tao Ma6f70fa52009-08-25 08:05:12 +0800274int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end,
275 struct page **pages, int *num);
276void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle,
277 unsigned int from, unsigned int to,
278 struct page *page, int zero, u64 *phys);
Tao Mae2e9f602009-08-18 11:22:34 +0800279/*
280 * Structures which describe a path through a btree, and functions to
281 * manipulate them.
282 *
283 * The idea here is to be as generic as possible with the tree
284 * manipulation code.
285 */
286struct ocfs2_path_item {
287 struct buffer_head *bh;
288 struct ocfs2_extent_list *el;
289};
290
291#define OCFS2_MAX_PATH_DEPTH 5
292
293struct ocfs2_path {
294 int p_tree_depth;
295 ocfs2_journal_access_func p_root_access;
296 struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH];
297};
298
299#define path_root_bh(_path) ((_path)->p_node[0].bh)
300#define path_root_el(_path) ((_path)->p_node[0].el)
301#define path_root_access(_path)((_path)->p_root_access)
302#define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh)
303#define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el)
304#define path_num_items(_path) ((_path)->p_tree_depth + 1)
305
306void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root);
307void ocfs2_free_path(struct ocfs2_path *path);
308int ocfs2_find_path(struct ocfs2_caching_info *ci,
309 struct ocfs2_path *path,
310 u32 cpos);
311struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path);
312struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et);
313int ocfs2_path_bh_journal_access(handle_t *handle,
314 struct ocfs2_caching_info *ci,
315 struct ocfs2_path *path,
316 int idx);
317int ocfs2_journal_access_path(struct ocfs2_caching_info *ci,
318 handle_t *handle,
319 struct ocfs2_path *path);
Tao Ma38a04e42009-11-30 14:32:19 +0800320int ocfs2_find_cpos_for_right_leaf(struct super_block *sb,
321 struct ocfs2_path *path, u32 *cpos);
322int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et,
323 struct ocfs2_path *left,
324 struct ocfs2_path *right);
Mark Fashehccd979b2005-12-15 14:31:24 -0800325#endif /* OCFS2_ALLOC_H */