1ccd979bdSMark Fasheh /* -*- mode: c; c-basic-offset: 8; -*- 2ccd979bdSMark Fasheh * vim: noexpandtab sw=8 ts=8 sts=0: 3ccd979bdSMark Fasheh * 4ccd979bdSMark Fasheh * alloc.h 5ccd979bdSMark Fasheh * 6ccd979bdSMark Fasheh * Function prototypes 7ccd979bdSMark Fasheh * 8ccd979bdSMark Fasheh * Copyright (C) 2002, 2004 Oracle. All rights reserved. 9ccd979bdSMark Fasheh * 10ccd979bdSMark Fasheh * This program is free software; you can redistribute it and/or 11ccd979bdSMark Fasheh * modify it under the terms of the GNU General Public 12ccd979bdSMark Fasheh * License as published by the Free Software Foundation; either 13ccd979bdSMark Fasheh * version 2 of the License, or (at your option) any later version. 14ccd979bdSMark Fasheh * 15ccd979bdSMark Fasheh * This program is distributed in the hope that it will be useful, 16ccd979bdSMark Fasheh * but WITHOUT ANY WARRANTY; without even the implied warranty of 17ccd979bdSMark Fasheh * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 18ccd979bdSMark Fasheh * General Public License for more details. 19ccd979bdSMark Fasheh * 20ccd979bdSMark Fasheh * You should have received a copy of the GNU General Public 21ccd979bdSMark Fasheh * License along with this program; if not, write to the 22ccd979bdSMark Fasheh * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 23ccd979bdSMark Fasheh * Boston, MA 021110-1307, USA. 24ccd979bdSMark Fasheh */ 25ccd979bdSMark Fasheh 26ccd979bdSMark Fasheh #ifndef OCFS2_ALLOC_H 27ccd979bdSMark Fasheh #define OCFS2_ALLOC_H 28ccd979bdSMark Fasheh 29e7d4cb6bSTao Ma 30ca12b7c4STao Ma /* 31ca12b7c4STao Ma * For xattr tree leaf, we limit the leaf byte size to be 64K. 32ca12b7c4STao Ma */ 33ca12b7c4STao Ma #define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536 34ca12b7c4STao Ma 35f99b9b7cSJoel Becker /* 36f99b9b7cSJoel Becker * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract 37f99b9b7cSJoel Becker * the b-tree operations in ocfs2. Now all the b-tree operations are not 38f99b9b7cSJoel Becker * limited to ocfs2_dinode only. Any data which need to allocate clusters 39f99b9b7cSJoel Becker * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree 40f99b9b7cSJoel Becker * and operation. 41f99b9b7cSJoel Becker * 42f99b9b7cSJoel Becker * ocfs2_extent_tree becomes the first-class object for extent tree 43f99b9b7cSJoel Becker * manipulation. Callers of the alloc.c code need to fill it via one of 448d6220d6SJoel Becker * the ocfs2_init_*_extent_tree() operations below. 45f99b9b7cSJoel Becker * 46f99b9b7cSJoel Becker * ocfs2_extent_tree contains info for the root of the b-tree, it must have a 47f99b9b7cSJoel Becker * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree 48f99b9b7cSJoel Becker * functions. 49f99b9b7cSJoel Becker * ocfs2_extent_tree_operations abstract the normal operations we do for 50f99b9b7cSJoel Becker * the root of extent b-tree. 51f99b9b7cSJoel Becker */ 52f99b9b7cSJoel Becker struct ocfs2_extent_tree_operations; 53f99b9b7cSJoel Becker struct ocfs2_extent_tree { 54f99b9b7cSJoel Becker struct ocfs2_extent_tree_operations *et_ops; 55f99b9b7cSJoel Becker struct buffer_head *et_root_bh; 56f99b9b7cSJoel Becker struct ocfs2_extent_list *et_root_el; 57f99b9b7cSJoel Becker void *et_object; 58f99b9b7cSJoel Becker unsigned int et_max_leaf_clusters; 59f99b9b7cSJoel Becker }; 60f99b9b7cSJoel Becker 61f99b9b7cSJoel Becker /* 628d6220d6SJoel Becker * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the 638d6220d6SJoel Becker * specified object buffer. 64f99b9b7cSJoel Becker */ 658d6220d6SJoel Becker void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, 66f56654c4STao Ma struct inode *inode, 67f99b9b7cSJoel Becker struct buffer_head *bh); 688d6220d6SJoel Becker void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, 69ccd979bdSMark Fasheh struct inode *inode, 70f99b9b7cSJoel Becker struct buffer_head *bh); 718d6220d6SJoel Becker void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, 72f99b9b7cSJoel Becker struct inode *inode, 73f99b9b7cSJoel Becker struct buffer_head *bh, 741a09f556SJoel Becker struct ocfs2_xattr_value_root *xv); 75f99b9b7cSJoel Becker 765e96581aSJoel Becker /* 775e96581aSJoel Becker * Read an extent block into *bh. If *bh is NULL, a bh will be 785e96581aSJoel Becker * allocated. This is a cached read. The extent block will be validated 795e96581aSJoel Becker * with ocfs2_validate_extent_block(). 805e96581aSJoel Becker */ 815e96581aSJoel Becker int ocfs2_read_extent_block(struct inode *inode, u64 eb_blkno, 825e96581aSJoel Becker struct buffer_head **bh); 835e96581aSJoel Becker 84f99b9b7cSJoel Becker struct ocfs2_alloc_context; 85f99b9b7cSJoel Becker int ocfs2_insert_extent(struct ocfs2_super *osb, 86ba492615STao Ma handle_t *handle, 87ba492615STao Ma struct inode *inode, 88f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 89ba492615STao Ma u32 cpos, 90ba492615STao Ma u64 start_blk, 91ba492615STao Ma u32 new_clusters, 92ba492615STao Ma u8 flags, 93ba492615STao Ma struct ocfs2_alloc_context *meta_ac); 94f99b9b7cSJoel Becker 950eb8d47eSTao Ma enum ocfs2_alloc_restarted { 960eb8d47eSTao Ma RESTART_NONE = 0, 970eb8d47eSTao Ma RESTART_TRANS, 980eb8d47eSTao Ma RESTART_META 990eb8d47eSTao Ma }; 1000eb8d47eSTao Ma int ocfs2_add_clusters_in_btree(struct ocfs2_super *osb, 1010eb8d47eSTao Ma struct inode *inode, 1020eb8d47eSTao Ma u32 *logical_offset, 1030eb8d47eSTao Ma u32 clusters_to_add, 1040eb8d47eSTao Ma int mark_unwritten, 105f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 1060eb8d47eSTao Ma handle_t *handle, 1070eb8d47eSTao Ma struct ocfs2_alloc_context *data_ac, 1080eb8d47eSTao Ma struct ocfs2_alloc_context *meta_ac, 109f99b9b7cSJoel Becker enum ocfs2_alloc_restarted *reason_ret); 110328d5752SMark Fasheh struct ocfs2_cached_dealloc_ctxt; 111f99b9b7cSJoel Becker int ocfs2_mark_extent_written(struct inode *inode, 112f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 113328d5752SMark Fasheh handle_t *handle, u32 cpos, u32 len, u32 phys, 114328d5752SMark Fasheh struct ocfs2_alloc_context *meta_ac, 115f99b9b7cSJoel Becker struct ocfs2_cached_dealloc_ctxt *dealloc); 116f99b9b7cSJoel Becker int ocfs2_remove_extent(struct inode *inode, 117f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 118063c4561SMark Fasheh u32 cpos, u32 len, handle_t *handle, 119063c4561SMark Fasheh struct ocfs2_alloc_context *meta_ac, 120f99b9b7cSJoel Becker struct ocfs2_cached_dealloc_ctxt *dealloc); 121fecc0112SMark Fasheh int ocfs2_remove_btree_range(struct inode *inode, 122fecc0112SMark Fasheh struct ocfs2_extent_tree *et, 123fecc0112SMark Fasheh u32 cpos, u32 phys_cpos, u32 len, 124fecc0112SMark Fasheh struct ocfs2_cached_dealloc_ctxt *dealloc); 125fecc0112SMark Fasheh 126ccd979bdSMark Fasheh int ocfs2_num_free_extents(struct ocfs2_super *osb, 127ccd979bdSMark Fasheh struct inode *inode, 128f99b9b7cSJoel Becker struct ocfs2_extent_tree *et); 129e7d4cb6bSTao Ma 130811f933dSTao Ma /* 131811f933dSTao Ma * how many new metadata chunks would an allocation need at maximum? 132811f933dSTao Ma * 133811f933dSTao Ma * Please note that the caller must make sure that root_el is the root 134811f933dSTao Ma * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise 135811f933dSTao Ma * the result may be wrong. 136811f933dSTao Ma */ 137811f933dSTao Ma static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el) 138ccd979bdSMark Fasheh { 139ccd979bdSMark Fasheh /* 140ccd979bdSMark Fasheh * Rather than do all the work of determining how much we need 141ccd979bdSMark Fasheh * (involves a ton of reads and locks), just ask for the 142ccd979bdSMark Fasheh * maximal limit. That's a tree depth shift. So, one block for 143ccd979bdSMark Fasheh * level of the tree (current l_tree_depth), one block for the 144ccd979bdSMark Fasheh * new tree_depth==0 extent_block, and one block at the new 145ccd979bdSMark Fasheh * top-of-the tree. 146ccd979bdSMark Fasheh */ 147811f933dSTao Ma return le16_to_cpu(root_el->l_tree_depth) + 2; 148ccd979bdSMark Fasheh } 149ccd979bdSMark Fasheh 1505b6a3a2bSMark Fasheh void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di); 1511afc32b9SMark Fasheh void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di); 1521afc32b9SMark Fasheh int ocfs2_convert_inline_data_to_extents(struct inode *inode, 1531afc32b9SMark Fasheh struct buffer_head *di_bh); 1541afc32b9SMark Fasheh 155ccd979bdSMark Fasheh int ocfs2_truncate_log_init(struct ocfs2_super *osb); 156ccd979bdSMark Fasheh void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb); 157ccd979bdSMark Fasheh void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, 158ccd979bdSMark Fasheh int cancel); 159ccd979bdSMark Fasheh int ocfs2_flush_truncate_log(struct ocfs2_super *osb); 160ccd979bdSMark Fasheh int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, 161ccd979bdSMark Fasheh int slot_num, 162ccd979bdSMark Fasheh struct ocfs2_dinode **tl_copy); 163ccd979bdSMark Fasheh int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, 164ccd979bdSMark Fasheh struct ocfs2_dinode *tl_copy); 165063c4561SMark Fasheh int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb); 166063c4561SMark Fasheh int ocfs2_truncate_log_append(struct ocfs2_super *osb, 167063c4561SMark Fasheh handle_t *handle, 168063c4561SMark Fasheh u64 start_blk, 169063c4561SMark Fasheh unsigned int num_clusters); 170063c4561SMark Fasheh int __ocfs2_flush_truncate_log(struct ocfs2_super *osb); 171ccd979bdSMark Fasheh 1722b604351SMark Fasheh /* 1732b604351SMark Fasheh * Process local structure which describes the block unlinks done 1742b604351SMark Fasheh * during an operation. This is populated via 1752b604351SMark Fasheh * ocfs2_cache_block_dealloc(). 1762b604351SMark Fasheh * 1772b604351SMark Fasheh * ocfs2_run_deallocs() should be called after the potentially 1782b604351SMark Fasheh * de-allocating routines. No journal handles should be open, and most 1792b604351SMark Fasheh * locks should have been dropped. 1802b604351SMark Fasheh */ 1812b604351SMark Fasheh struct ocfs2_cached_dealloc_ctxt { 1822b604351SMark Fasheh struct ocfs2_per_slot_free_list *c_first_suballocator; 1832891d290STao Ma struct ocfs2_cached_block_free *c_global_allocator; 1842b604351SMark Fasheh }; 1852b604351SMark Fasheh static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c) 1862b604351SMark Fasheh { 1872b604351SMark Fasheh c->c_first_suballocator = NULL; 1882891d290STao Ma c->c_global_allocator = NULL; 1892b604351SMark Fasheh } 1902891d290STao Ma int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, 1912891d290STao Ma u64 blkno, unsigned int bit); 19278f30c31STao Ma static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c) 19378f30c31STao Ma { 19478f30c31STao Ma return c->c_global_allocator != NULL; 19578f30c31STao Ma } 1962b604351SMark Fasheh int ocfs2_run_deallocs(struct ocfs2_super *osb, 1972b604351SMark Fasheh struct ocfs2_cached_dealloc_ctxt *ctxt); 1982b604351SMark Fasheh 199ccd979bdSMark Fasheh struct ocfs2_truncate_context { 20059a5e416SMark Fasheh struct ocfs2_cached_dealloc_ctxt tc_dealloc; 201ccd979bdSMark Fasheh int tc_ext_alloc_locked; /* is it cluster locked? */ 202ccd979bdSMark Fasheh /* these get destroyed once it's passed to ocfs2_commit_truncate. */ 203ccd979bdSMark Fasheh struct buffer_head *tc_last_eb_bh; 204ccd979bdSMark Fasheh }; 205ccd979bdSMark Fasheh 20635edec1dSMark Fasheh int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, 20735edec1dSMark Fasheh u64 range_start, u64 range_end); 208ccd979bdSMark Fasheh int ocfs2_prepare_truncate(struct ocfs2_super *osb, 209ccd979bdSMark Fasheh struct inode *inode, 210ccd979bdSMark Fasheh struct buffer_head *fe_bh, 211ccd979bdSMark Fasheh struct ocfs2_truncate_context **tc); 212ccd979bdSMark Fasheh int ocfs2_commit_truncate(struct ocfs2_super *osb, 213ccd979bdSMark Fasheh struct inode *inode, 214ccd979bdSMark Fasheh struct buffer_head *fe_bh, 215ccd979bdSMark Fasheh struct ocfs2_truncate_context *tc); 2161afc32b9SMark Fasheh int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, 2171afc32b9SMark Fasheh unsigned int start, unsigned int end, int trunc); 218ccd979bdSMark Fasheh 219363041a5SMark Fasheh int ocfs2_find_leaf(struct inode *inode, struct ocfs2_extent_list *root_el, 220363041a5SMark Fasheh u32 cpos, struct buffer_head **leaf_bh); 221328d5752SMark Fasheh int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster); 222363041a5SMark Fasheh 223e48edee2SMark Fasheh /* 224e48edee2SMark Fasheh * Helper function to look at the # of clusters in an extent record. 225e48edee2SMark Fasheh */ 226e48edee2SMark Fasheh static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el, 227e48edee2SMark Fasheh struct ocfs2_extent_rec *rec) 228e48edee2SMark Fasheh { 229e48edee2SMark Fasheh /* 230e48edee2SMark Fasheh * Cluster count in extent records is slightly different 231e48edee2SMark Fasheh * between interior nodes and leaf nodes. This is to support 232e48edee2SMark Fasheh * unwritten extents which need a flags field in leaf node 233e48edee2SMark Fasheh * records, thus shrinking the available space for a clusters 234e48edee2SMark Fasheh * field. 235e48edee2SMark Fasheh */ 236e48edee2SMark Fasheh if (el->l_tree_depth) 237e48edee2SMark Fasheh return le32_to_cpu(rec->e_int_clusters); 238e48edee2SMark Fasheh else 239e48edee2SMark Fasheh return le16_to_cpu(rec->e_leaf_clusters); 240e48edee2SMark Fasheh } 241e48edee2SMark Fasheh 24200dc417fSMark Fasheh /* 24300dc417fSMark Fasheh * This is only valid for leaf nodes, which are the only ones that can 24400dc417fSMark Fasheh * have empty extents anyway. 24500dc417fSMark Fasheh */ 24600dc417fSMark Fasheh static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec) 24700dc417fSMark Fasheh { 24800dc417fSMark Fasheh return !rec->e_leaf_clusters; 24900dc417fSMark Fasheh } 25000dc417fSMark Fasheh 251ccd979bdSMark Fasheh #endif /* OCFS2_ALLOC_H */ 252