Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 1 | /* -*- 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 Ma | e7d4cb6 | 2008-08-18 17:38:44 +0800 | [diff] [blame] | 29 | |
Tao Ma | ca12b7c | 2008-08-18 17:38:52 +0800 | [diff] [blame] | 30 | /* |
| 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 Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 35 | /* |
| 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 Becker | 8d6220d | 2008-08-22 12:46:09 -0700 | [diff] [blame] | 44 | * the ocfs2_init_*_extent_tree() operations below. |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 45 | * |
| 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. |
| 49 | * ocfs2_extent_tree_operations abstract the normal operations we do for |
| 50 | * the root of extent b-tree. |
| 51 | */ |
| 52 | struct ocfs2_extent_tree_operations; |
| 53 | struct ocfs2_extent_tree { |
| 54 | struct ocfs2_extent_tree_operations *et_ops; |
| 55 | struct buffer_head *et_root_bh; |
| 56 | struct ocfs2_extent_list *et_root_el; |
| 57 | void *et_object; |
| 58 | unsigned int et_max_leaf_clusters; |
| 59 | }; |
| 60 | |
| 61 | /* |
Joel Becker | 8d6220d | 2008-08-22 12:46:09 -0700 | [diff] [blame] | 62 | * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the |
| 63 | * specified object buffer. |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 64 | */ |
Joel Becker | 8d6220d | 2008-08-22 12:46:09 -0700 | [diff] [blame] | 65 | void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, |
| 66 | struct inode *inode, |
| 67 | struct buffer_head *bh); |
| 68 | void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 69 | struct inode *inode, |
Joel Becker | 8d6220d | 2008-08-22 12:46:09 -0700 | [diff] [blame] | 70 | struct buffer_head *bh); |
| 71 | void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, |
| 72 | struct inode *inode, |
| 73 | struct buffer_head *bh, |
| 74 | struct ocfs2_xattr_value_root *xv); |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 75 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 76 | struct ocfs2_alloc_context; |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 77 | int ocfs2_insert_extent(struct ocfs2_super *osb, |
| 78 | handle_t *handle, |
| 79 | struct inode *inode, |
| 80 | struct ocfs2_extent_tree *et, |
| 81 | u32 cpos, |
| 82 | u64 start_blk, |
| 83 | u32 new_clusters, |
| 84 | u8 flags, |
| 85 | struct ocfs2_alloc_context *meta_ac); |
| 86 | |
Tao Ma | 0eb8d47 | 2008-08-18 17:38:45 +0800 | [diff] [blame] | 87 | enum ocfs2_alloc_restarted { |
| 88 | RESTART_NONE = 0, |
| 89 | RESTART_TRANS, |
| 90 | RESTART_META |
| 91 | }; |
| 92 | int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, |
| 93 | struct inode *inode, |
| 94 | u32 *logical_offset, |
| 95 | u32 clusters_to_add, |
| 96 | int mark_unwritten, |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 97 | struct ocfs2_extent_tree *et, |
Tao Ma | 0eb8d47 | 2008-08-18 17:38:45 +0800 | [diff] [blame] | 98 | handle_t *handle, |
| 99 | struct ocfs2_alloc_context *data_ac, |
| 100 | struct ocfs2_alloc_context *meta_ac, |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 101 | enum ocfs2_alloc_restarted *reason_ret); |
Mark Fasheh | 328d575 | 2007-06-18 10:48:04 -0700 | [diff] [blame] | 102 | struct ocfs2_cached_dealloc_ctxt; |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 103 | int ocfs2_mark_extent_written(struct inode *inode, |
| 104 | struct ocfs2_extent_tree *et, |
Mark Fasheh | 328d575 | 2007-06-18 10:48:04 -0700 | [diff] [blame] | 105 | handle_t *handle, u32 cpos, u32 len, u32 phys, |
| 106 | struct ocfs2_alloc_context *meta_ac, |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 107 | struct ocfs2_cached_dealloc_ctxt *dealloc); |
| 108 | int ocfs2_remove_extent(struct inode *inode, |
| 109 | struct ocfs2_extent_tree *et, |
Mark Fasheh | 063c456 | 2007-07-03 13:34:11 -0700 | [diff] [blame] | 110 | u32 cpos, u32 len, handle_t *handle, |
| 111 | struct ocfs2_alloc_context *meta_ac, |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 112 | struct ocfs2_cached_dealloc_ctxt *dealloc); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 113 | int ocfs2_num_free_extents(struct ocfs2_super *osb, |
| 114 | struct inode *inode, |
Joel Becker | f99b9b7 | 2008-08-20 19:36:33 -0700 | [diff] [blame] | 115 | struct ocfs2_extent_tree *et); |
Tao Ma | e7d4cb6 | 2008-08-18 17:38:44 +0800 | [diff] [blame] | 116 | |
Tao Ma | 811f933 | 2008-08-18 17:38:43 +0800 | [diff] [blame] | 117 | /* |
| 118 | * how many new metadata chunks would an allocation need at maximum? |
| 119 | * |
| 120 | * Please note that the caller must make sure that root_el is the root |
| 121 | * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise |
| 122 | * the result may be wrong. |
| 123 | */ |
| 124 | static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el) |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 125 | { |
| 126 | /* |
| 127 | * Rather than do all the work of determining how much we need |
| 128 | * (involves a ton of reads and locks), just ask for the |
| 129 | * maximal limit. That's a tree depth shift. So, one block for |
| 130 | * level of the tree (current l_tree_depth), one block for the |
| 131 | * new tree_depth==0 extent_block, and one block at the new |
| 132 | * top-of-the tree. |
| 133 | */ |
Tao Ma | 811f933 | 2008-08-18 17:38:43 +0800 | [diff] [blame] | 134 | return le16_to_cpu(root_el->l_tree_depth) + 2; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 135 | } |
| 136 | |
Mark Fasheh | 5b6a3a2 | 2007-09-13 16:33:54 -0700 | [diff] [blame] | 137 | void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di); |
Mark Fasheh | 1afc32b | 2007-09-07 14:46:51 -0700 | [diff] [blame] | 138 | void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di); |
| 139 | int ocfs2_convert_inline_data_to_extents(struct inode *inode, |
| 140 | struct buffer_head *di_bh); |
| 141 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 142 | int ocfs2_truncate_log_init(struct ocfs2_super *osb); |
| 143 | void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb); |
| 144 | void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, |
| 145 | int cancel); |
| 146 | int ocfs2_flush_truncate_log(struct ocfs2_super *osb); |
| 147 | int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, |
| 148 | int slot_num, |
| 149 | struct ocfs2_dinode **tl_copy); |
| 150 | int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, |
| 151 | struct ocfs2_dinode *tl_copy); |
Mark Fasheh | 063c456 | 2007-07-03 13:34:11 -0700 | [diff] [blame] | 152 | int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb); |
| 153 | int ocfs2_truncate_log_append(struct ocfs2_super *osb, |
| 154 | handle_t *handle, |
| 155 | u64 start_blk, |
| 156 | unsigned int num_clusters); |
| 157 | int __ocfs2_flush_truncate_log(struct ocfs2_super *osb); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 158 | |
Mark Fasheh | 2b60435 | 2007-06-22 15:45:27 -0700 | [diff] [blame] | 159 | /* |
| 160 | * Process local structure which describes the block unlinks done |
| 161 | * during an operation. This is populated via |
| 162 | * ocfs2_cache_block_dealloc(). |
| 163 | * |
| 164 | * ocfs2_run_deallocs() should be called after the potentially |
| 165 | * de-allocating routines. No journal handles should be open, and most |
| 166 | * locks should have been dropped. |
| 167 | */ |
| 168 | struct ocfs2_cached_dealloc_ctxt { |
| 169 | struct ocfs2_per_slot_free_list *c_first_suballocator; |
Tao Ma | 2891d29 | 2008-11-12 08:26:58 +0800 | [diff] [blame] | 170 | struct ocfs2_cached_block_free *c_global_allocator; |
Mark Fasheh | 2b60435 | 2007-06-22 15:45:27 -0700 | [diff] [blame] | 171 | }; |
| 172 | static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c) |
| 173 | { |
| 174 | c->c_first_suballocator = NULL; |
Tao Ma | 2891d29 | 2008-11-12 08:26:58 +0800 | [diff] [blame] | 175 | c->c_global_allocator = NULL; |
Mark Fasheh | 2b60435 | 2007-06-22 15:45:27 -0700 | [diff] [blame] | 176 | } |
Tao Ma | 2891d29 | 2008-11-12 08:26:58 +0800 | [diff] [blame] | 177 | int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, |
| 178 | u64 blkno, unsigned int bit); |
Tao Ma | 78f30c3 | 2008-11-12 08:27:00 +0800 | [diff] [blame^] | 179 | static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c) |
| 180 | { |
| 181 | return c->c_global_allocator != NULL; |
| 182 | } |
Mark Fasheh | 2b60435 | 2007-06-22 15:45:27 -0700 | [diff] [blame] | 183 | int ocfs2_run_deallocs(struct ocfs2_super *osb, |
| 184 | struct ocfs2_cached_dealloc_ctxt *ctxt); |
| 185 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 186 | struct ocfs2_truncate_context { |
Mark Fasheh | 59a5e41 | 2007-06-22 15:52:36 -0700 | [diff] [blame] | 187 | struct ocfs2_cached_dealloc_ctxt tc_dealloc; |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 188 | int tc_ext_alloc_locked; /* is it cluster locked? */ |
| 189 | /* these get destroyed once it's passed to ocfs2_commit_truncate. */ |
| 190 | struct buffer_head *tc_last_eb_bh; |
| 191 | }; |
| 192 | |
Mark Fasheh | 35edec1 | 2007-07-06 14:41:18 -0700 | [diff] [blame] | 193 | int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, |
| 194 | u64 range_start, u64 range_end); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 195 | int ocfs2_prepare_truncate(struct ocfs2_super *osb, |
| 196 | struct inode *inode, |
| 197 | struct buffer_head *fe_bh, |
| 198 | struct ocfs2_truncate_context **tc); |
| 199 | int ocfs2_commit_truncate(struct ocfs2_super *osb, |
| 200 | struct inode *inode, |
| 201 | struct buffer_head *fe_bh, |
| 202 | struct ocfs2_truncate_context *tc); |
Mark Fasheh | 1afc32b | 2007-09-07 14:46:51 -0700 | [diff] [blame] | 203 | int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, |
| 204 | unsigned int start, unsigned int end, int trunc); |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 205 | |
Mark Fasheh | 363041a | 2007-01-17 12:31:35 -0800 | [diff] [blame] | 206 | int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el, |
| 207 | u32 cpos, struct buffer_head **leaf_bh); |
Mark Fasheh | 328d575 | 2007-06-18 10:48:04 -0700 | [diff] [blame] | 208 | int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster); |
Mark Fasheh | 363041a | 2007-01-17 12:31:35 -0800 | [diff] [blame] | 209 | |
Mark Fasheh | e48edee | 2007-03-07 16:46:57 -0800 | [diff] [blame] | 210 | /* |
| 211 | * Helper function to look at the # of clusters in an extent record. |
| 212 | */ |
| 213 | static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el, |
| 214 | struct ocfs2_extent_rec *rec) |
| 215 | { |
| 216 | /* |
| 217 | * Cluster count in extent records is slightly different |
| 218 | * between interior nodes and leaf nodes. This is to support |
| 219 | * unwritten extents which need a flags field in leaf node |
| 220 | * records, thus shrinking the available space for a clusters |
| 221 | * field. |
| 222 | */ |
| 223 | if (el->l_tree_depth) |
| 224 | return le32_to_cpu(rec->e_int_clusters); |
| 225 | else |
| 226 | return le16_to_cpu(rec->e_leaf_clusters); |
| 227 | } |
| 228 | |
Mark Fasheh | 00dc417 | 2008-10-03 17:32:11 -0400 | [diff] [blame] | 229 | /* |
| 230 | * This is only valid for leaf nodes, which are the only ones that can |
| 231 | * have empty extents anyway. |
| 232 | */ |
| 233 | static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec) |
| 234 | { |
| 235 | return !rec->e_leaf_clusters; |
| 236 | } |
| 237 | |
Mark Fasheh | ccd979b | 2005-12-15 14:31:24 -0800 | [diff] [blame] | 238 | #endif /* OCFS2_ALLOC_H */ |