1328970deSThomas Gleixner /* SPDX-License-Identifier: GPL-2.0-or-later */ 2ccd979bdSMark Fasheh /* -*- mode: c; c-basic-offset: 8; -*- 3ccd979bdSMark Fasheh * vim: noexpandtab sw=8 ts=8 sts=0: 4ccd979bdSMark Fasheh * 5ccd979bdSMark Fasheh * alloc.h 6ccd979bdSMark Fasheh * 7ccd979bdSMark Fasheh * Function prototypes 8ccd979bdSMark Fasheh * 9ccd979bdSMark Fasheh * Copyright (C) 2002, 2004 Oracle. All rights reserved. 10ccd979bdSMark Fasheh */ 11ccd979bdSMark Fasheh 12ccd979bdSMark Fasheh #ifndef OCFS2_ALLOC_H 13ccd979bdSMark Fasheh #define OCFS2_ALLOC_H 14ccd979bdSMark Fasheh 15e7d4cb6bSTao Ma 16ca12b7c4STao Ma /* 17ca12b7c4STao Ma * For xattr tree leaf, we limit the leaf byte size to be 64K. 18ca12b7c4STao Ma */ 19ca12b7c4STao Ma #define OCFS2_MAX_XATTR_TREE_LEAF_SIZE 65536 20ca12b7c4STao Ma 21f99b9b7cSJoel Becker /* 22f99b9b7cSJoel Becker * ocfs2_extent_tree and ocfs2_extent_tree_operations are used to abstract 23f99b9b7cSJoel Becker * the b-tree operations in ocfs2. Now all the b-tree operations are not 24f99b9b7cSJoel Becker * limited to ocfs2_dinode only. Any data which need to allocate clusters 25f99b9b7cSJoel Becker * to store can use b-tree. And it only needs to implement its ocfs2_extent_tree 26f99b9b7cSJoel Becker * and operation. 27f99b9b7cSJoel Becker * 28f99b9b7cSJoel Becker * ocfs2_extent_tree becomes the first-class object for extent tree 29f99b9b7cSJoel Becker * manipulation. Callers of the alloc.c code need to fill it via one of 308d6220d6SJoel Becker * the ocfs2_init_*_extent_tree() operations below. 31f99b9b7cSJoel Becker * 32f99b9b7cSJoel Becker * ocfs2_extent_tree contains info for the root of the b-tree, it must have a 33f99b9b7cSJoel Becker * root ocfs2_extent_list and a root_bh so that they can be used in the b-tree 34d9a0a1f8SJoel Becker * functions. It needs the ocfs2_caching_info structure associated with 35d9a0a1f8SJoel Becker * I/O on the tree. With metadata ecc, we now call different journal_access 3613723d00SJoel Becker * functions for each type of metadata, so it must have the 3713723d00SJoel Becker * root_journal_access function. 38f99b9b7cSJoel Becker * ocfs2_extent_tree_operations abstract the normal operations we do for 39f99b9b7cSJoel Becker * the root of extent b-tree. 40f99b9b7cSJoel Becker */ 41f99b9b7cSJoel Becker struct ocfs2_extent_tree_operations; 42f99b9b7cSJoel Becker struct ocfs2_extent_tree { 439e62dc09SJulia Lawall const struct ocfs2_extent_tree_operations *et_ops; 44f99b9b7cSJoel Becker struct buffer_head *et_root_bh; 45f99b9b7cSJoel Becker struct ocfs2_extent_list *et_root_el; 46d9a0a1f8SJoel Becker struct ocfs2_caching_info *et_ci; 4713723d00SJoel Becker ocfs2_journal_access_func et_root_journal_access; 48f99b9b7cSJoel Becker void *et_object; 49f99b9b7cSJoel Becker unsigned int et_max_leaf_clusters; 5071a36944SChangwei Ge struct ocfs2_cached_dealloc_ctxt *et_dealloc; 51f99b9b7cSJoel Becker }; 52f99b9b7cSJoel Becker 53f99b9b7cSJoel Becker /* 548d6220d6SJoel Becker * ocfs2_init_*_extent_tree() will fill an ocfs2_extent_tree from the 558d6220d6SJoel Becker * specified object buffer. 56f99b9b7cSJoel Becker */ 578d6220d6SJoel Becker void ocfs2_init_dinode_extent_tree(struct ocfs2_extent_tree *et, 585e404e9eSJoel Becker struct ocfs2_caching_info *ci, 59f99b9b7cSJoel Becker struct buffer_head *bh); 608d6220d6SJoel Becker void ocfs2_init_xattr_tree_extent_tree(struct ocfs2_extent_tree *et, 615e404e9eSJoel Becker struct ocfs2_caching_info *ci, 62f99b9b7cSJoel Becker struct buffer_head *bh); 632a50a743SJoel Becker struct ocfs2_xattr_value_buf; 648d6220d6SJoel Becker void ocfs2_init_xattr_value_extent_tree(struct ocfs2_extent_tree *et, 655e404e9eSJoel Becker struct ocfs2_caching_info *ci, 662a50a743SJoel Becker struct ocfs2_xattr_value_buf *vb); 679b7895efSMark Fasheh void ocfs2_init_dx_root_extent_tree(struct ocfs2_extent_tree *et, 685e404e9eSJoel Becker struct ocfs2_caching_info *ci, 699b7895efSMark Fasheh struct buffer_head *bh); 70fe924415STao Ma void ocfs2_init_refcount_extent_tree(struct ocfs2_extent_tree *et, 71fe924415STao Ma struct ocfs2_caching_info *ci, 72fe924415STao Ma struct buffer_head *bh); 73f99b9b7cSJoel Becker 745e96581aSJoel Becker /* 755e96581aSJoel Becker * Read an extent block into *bh. If *bh is NULL, a bh will be 765e96581aSJoel Becker * allocated. This is a cached read. The extent block will be validated 775e96581aSJoel Becker * with ocfs2_validate_extent_block(). 785e96581aSJoel Becker */ 793d03a305SJoel Becker int ocfs2_read_extent_block(struct ocfs2_caching_info *ci, u64 eb_blkno, 805e96581aSJoel Becker struct buffer_head **bh); 815e96581aSJoel Becker 82f99b9b7cSJoel Becker struct ocfs2_alloc_context; 83cc79d8c1SJoel Becker int ocfs2_insert_extent(handle_t *handle, 84f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 85ba492615STao Ma u32 cpos, 86ba492615STao Ma u64 start_blk, 87ba492615STao Ma u32 new_clusters, 88ba492615STao Ma u8 flags, 89ba492615STao Ma struct ocfs2_alloc_context *meta_ac); 90f99b9b7cSJoel Becker 910eb8d47eSTao Ma enum ocfs2_alloc_restarted { 920eb8d47eSTao Ma RESTART_NONE = 0, 930eb8d47eSTao Ma RESTART_TRANS, 940eb8d47eSTao Ma RESTART_META 950eb8d47eSTao Ma }; 96cbee7e1aSJoel Becker int ocfs2_add_clusters_in_btree(handle_t *handle, 97cbee7e1aSJoel Becker struct ocfs2_extent_tree *et, 980eb8d47eSTao Ma u32 *logical_offset, 990eb8d47eSTao Ma u32 clusters_to_add, 1000eb8d47eSTao Ma int mark_unwritten, 1010eb8d47eSTao Ma struct ocfs2_alloc_context *data_ac, 1020eb8d47eSTao Ma struct ocfs2_alloc_context *meta_ac, 103f99b9b7cSJoel Becker enum ocfs2_alloc_restarted *reason_ret); 104328d5752SMark Fasheh struct ocfs2_cached_dealloc_ctxt; 105e2e9f608STao Ma struct ocfs2_path; 106e2e9f608STao Ma int ocfs2_split_extent(handle_t *handle, 107e2e9f608STao Ma struct ocfs2_extent_tree *et, 108e2e9f608STao Ma struct ocfs2_path *path, 109e2e9f608STao Ma int split_index, 110e2e9f608STao Ma struct ocfs2_extent_rec *split_rec, 111e2e9f608STao Ma struct ocfs2_alloc_context *meta_ac, 112e2e9f608STao Ma struct ocfs2_cached_dealloc_ctxt *dealloc); 113f99b9b7cSJoel Becker int ocfs2_mark_extent_written(struct inode *inode, 114f99b9b7cSJoel Becker struct ocfs2_extent_tree *et, 115328d5752SMark Fasheh handle_t *handle, u32 cpos, u32 len, u32 phys, 116328d5752SMark Fasheh struct ocfs2_alloc_context *meta_ac, 117f99b9b7cSJoel Becker struct ocfs2_cached_dealloc_ctxt *dealloc); 1181aa75feaSTao Ma int ocfs2_change_extent_flag(handle_t *handle, 1191aa75feaSTao Ma struct ocfs2_extent_tree *et, 1201aa75feaSTao Ma u32 cpos, u32 len, u32 phys, 1211aa75feaSTao Ma struct ocfs2_alloc_context *meta_ac, 1221aa75feaSTao Ma struct ocfs2_cached_dealloc_ctxt *dealloc, 1231aa75feaSTao Ma int new_flags, int clear_flags); 124dbdcf6a4SJoel Becker int ocfs2_remove_extent(handle_t *handle, struct ocfs2_extent_tree *et, 125dbdcf6a4SJoel Becker u32 cpos, u32 len, 126063c4561SMark Fasheh struct ocfs2_alloc_context *meta_ac, 127f99b9b7cSJoel Becker struct ocfs2_cached_dealloc_ctxt *dealloc); 128fecc0112SMark Fasheh int ocfs2_remove_btree_range(struct inode *inode, 129fecc0112SMark Fasheh struct ocfs2_extent_tree *et, 13078f94673STristan Ye u32 cpos, u32 phys_cpos, u32 len, int flags, 13178f94673STristan Ye struct ocfs2_cached_dealloc_ctxt *dealloc, 132f62f12b3SJunxiao Bi u64 refcount_loc, bool refcount_tree_locked); 133fecc0112SMark Fasheh 134964f14a0SJun Piao int ocfs2_num_free_extents(struct ocfs2_extent_tree *et); 135e7d4cb6bSTao Ma 136811f933dSTao Ma /* 137811f933dSTao Ma * how many new metadata chunks would an allocation need at maximum? 138811f933dSTao Ma * 139811f933dSTao Ma * Please note that the caller must make sure that root_el is the root 140811f933dSTao Ma * of extent tree. So for an inode, it should be &fe->id2.i_list. Otherwise 141811f933dSTao Ma * the result may be wrong. 142811f933dSTao Ma */ 143811f933dSTao Ma static inline int ocfs2_extend_meta_needed(struct ocfs2_extent_list *root_el) 144ccd979bdSMark Fasheh { 145ccd979bdSMark Fasheh /* 146ccd979bdSMark Fasheh * Rather than do all the work of determining how much we need 147ccd979bdSMark Fasheh * (involves a ton of reads and locks), just ask for the 148ccd979bdSMark Fasheh * maximal limit. That's a tree depth shift. So, one block for 149ccd979bdSMark Fasheh * level of the tree (current l_tree_depth), one block for the 150ccd979bdSMark Fasheh * new tree_depth==0 extent_block, and one block at the new 151ccd979bdSMark Fasheh * top-of-the tree. 152ccd979bdSMark Fasheh */ 153811f933dSTao Ma return le16_to_cpu(root_el->l_tree_depth) + 2; 154ccd979bdSMark Fasheh } 155ccd979bdSMark Fasheh 1565b6a3a2bSMark Fasheh void ocfs2_dinode_new_extent_list(struct inode *inode, struct ocfs2_dinode *di); 1571afc32b9SMark Fasheh void ocfs2_set_inode_data_inline(struct inode *inode, struct ocfs2_dinode *di); 1581afc32b9SMark Fasheh int ocfs2_convert_inline_data_to_extents(struct inode *inode, 1591afc32b9SMark Fasheh struct buffer_head *di_bh); 1601afc32b9SMark Fasheh 161ccd979bdSMark Fasheh int ocfs2_truncate_log_init(struct ocfs2_super *osb); 162ccd979bdSMark Fasheh void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb); 163ccd979bdSMark Fasheh void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb, 164ccd979bdSMark Fasheh int cancel); 165ccd979bdSMark Fasheh int ocfs2_flush_truncate_log(struct ocfs2_super *osb); 166ccd979bdSMark Fasheh int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb, 167ccd979bdSMark Fasheh int slot_num, 168ccd979bdSMark Fasheh struct ocfs2_dinode **tl_copy); 169ccd979bdSMark Fasheh int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb, 170ccd979bdSMark Fasheh struct ocfs2_dinode *tl_copy); 171063c4561SMark Fasheh int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb); 172063c4561SMark Fasheh int ocfs2_truncate_log_append(struct ocfs2_super *osb, 173063c4561SMark Fasheh handle_t *handle, 174063c4561SMark Fasheh u64 start_blk, 175063c4561SMark Fasheh unsigned int num_clusters); 176063c4561SMark Fasheh int __ocfs2_flush_truncate_log(struct ocfs2_super *osb); 1772070ad1aSEric Ren int ocfs2_try_to_free_truncate_log(struct ocfs2_super *osb, 1782070ad1aSEric Ren unsigned int needed); 179ccd979bdSMark Fasheh 1802b604351SMark Fasheh /* 1812b604351SMark Fasheh * Process local structure which describes the block unlinks done 1822b604351SMark Fasheh * during an operation. This is populated via 1832b604351SMark Fasheh * ocfs2_cache_block_dealloc(). 1842b604351SMark Fasheh * 1852b604351SMark Fasheh * ocfs2_run_deallocs() should be called after the potentially 1862b604351SMark Fasheh * de-allocating routines. No journal handles should be open, and most 1872b604351SMark Fasheh * locks should have been dropped. 1882b604351SMark Fasheh */ 1892b604351SMark Fasheh struct ocfs2_cached_dealloc_ctxt { 1902b604351SMark Fasheh struct ocfs2_per_slot_free_list *c_first_suballocator; 1912891d290STao Ma struct ocfs2_cached_block_free *c_global_allocator; 1922b604351SMark Fasheh }; 1932b604351SMark Fasheh static inline void ocfs2_init_dealloc_ctxt(struct ocfs2_cached_dealloc_ctxt *c) 1942b604351SMark Fasheh { 1952b604351SMark Fasheh c->c_first_suballocator = NULL; 1962891d290STao Ma c->c_global_allocator = NULL; 1972b604351SMark Fasheh } 1982891d290STao Ma int ocfs2_cache_cluster_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, 1992891d290STao Ma u64 blkno, unsigned int bit); 2001823cb0bSTao Ma int ocfs2_cache_block_dealloc(struct ocfs2_cached_dealloc_ctxt *ctxt, 20174380c47STao Ma int type, int slot, u64 suballoc, u64 blkno, 2021823cb0bSTao Ma unsigned int bit); 20378f30c31STao Ma static inline int ocfs2_dealloc_has_cluster(struct ocfs2_cached_dealloc_ctxt *c) 20478f30c31STao Ma { 20578f30c31STao Ma return c->c_global_allocator != NULL; 20678f30c31STao Ma } 2072b604351SMark Fasheh int ocfs2_run_deallocs(struct ocfs2_super *osb, 2082b604351SMark Fasheh struct ocfs2_cached_dealloc_ctxt *ctxt); 2092b604351SMark Fasheh 210ccd979bdSMark Fasheh struct ocfs2_truncate_context { 21159a5e416SMark Fasheh struct ocfs2_cached_dealloc_ctxt tc_dealloc; 212ccd979bdSMark Fasheh int tc_ext_alloc_locked; /* is it cluster locked? */ 213ccd979bdSMark Fasheh /* these get destroyed once it's passed to ocfs2_commit_truncate. */ 214ccd979bdSMark Fasheh struct buffer_head *tc_last_eb_bh; 215ccd979bdSMark Fasheh }; 216ccd979bdSMark Fasheh 21735edec1dSMark Fasheh int ocfs2_zero_range_for_truncate(struct inode *inode, handle_t *handle, 21835edec1dSMark Fasheh u64 range_start, u64 range_end); 219ccd979bdSMark Fasheh int ocfs2_commit_truncate(struct ocfs2_super *osb, 220ccd979bdSMark Fasheh struct inode *inode, 22178f94673STristan Ye struct buffer_head *di_bh); 2221afc32b9SMark Fasheh int ocfs2_truncate_inline(struct inode *inode, struct buffer_head *di_bh, 2231afc32b9SMark Fasheh unsigned int start, unsigned int end, int trunc); 224ccd979bdSMark Fasheh 225facdb77fSJoel Becker int ocfs2_find_leaf(struct ocfs2_caching_info *ci, 226facdb77fSJoel Becker struct ocfs2_extent_list *root_el, u32 cpos, 227facdb77fSJoel Becker struct buffer_head **leaf_bh); 228328d5752SMark Fasheh int ocfs2_search_extent_list(struct ocfs2_extent_list *el, u32 v_cluster); 229363041a5SMark Fasheh 230e80de36dSTao Ma int ocfs2_trim_fs(struct super_block *sb, struct fstrim_range *range); 231e48edee2SMark Fasheh /* 232e48edee2SMark Fasheh * Helper function to look at the # of clusters in an extent record. 233e48edee2SMark Fasheh */ 234e48edee2SMark Fasheh static inline unsigned int ocfs2_rec_clusters(struct ocfs2_extent_list *el, 235e48edee2SMark Fasheh struct ocfs2_extent_rec *rec) 236e48edee2SMark Fasheh { 237e48edee2SMark Fasheh /* 238e48edee2SMark Fasheh * Cluster count in extent records is slightly different 239e48edee2SMark Fasheh * between interior nodes and leaf nodes. This is to support 240e48edee2SMark Fasheh * unwritten extents which need a flags field in leaf node 241e48edee2SMark Fasheh * records, thus shrinking the available space for a clusters 242e48edee2SMark Fasheh * field. 243e48edee2SMark Fasheh */ 244e48edee2SMark Fasheh if (el->l_tree_depth) 245e48edee2SMark Fasheh return le32_to_cpu(rec->e_int_clusters); 246e48edee2SMark Fasheh else 247e48edee2SMark Fasheh return le16_to_cpu(rec->e_leaf_clusters); 248e48edee2SMark Fasheh } 249e48edee2SMark Fasheh 25000dc417fSMark Fasheh /* 25100dc417fSMark Fasheh * This is only valid for leaf nodes, which are the only ones that can 25200dc417fSMark Fasheh * have empty extents anyway. 25300dc417fSMark Fasheh */ 25400dc417fSMark Fasheh static inline int ocfs2_is_empty_extent(struct ocfs2_extent_rec *rec) 25500dc417fSMark Fasheh { 25600dc417fSMark Fasheh return !rec->e_leaf_clusters; 25700dc417fSMark Fasheh } 25800dc417fSMark Fasheh 2596f70fa51STao Ma int ocfs2_grab_pages(struct inode *inode, loff_t start, loff_t end, 2606f70fa51STao Ma struct page **pages, int *num); 2616f70fa51STao Ma void ocfs2_map_and_dirty_page(struct inode *inode, handle_t *handle, 2626f70fa51STao Ma unsigned int from, unsigned int to, 2636f70fa51STao Ma struct page *page, int zero, u64 *phys); 264e2e9f608STao Ma /* 265e2e9f608STao Ma * Structures which describe a path through a btree, and functions to 266e2e9f608STao Ma * manipulate them. 267e2e9f608STao Ma * 268e2e9f608STao Ma * The idea here is to be as generic as possible with the tree 269e2e9f608STao Ma * manipulation code. 270e2e9f608STao Ma */ 271e2e9f608STao Ma struct ocfs2_path_item { 272e2e9f608STao Ma struct buffer_head *bh; 273e2e9f608STao Ma struct ocfs2_extent_list *el; 274e2e9f608STao Ma }; 275e2e9f608STao Ma 276e2e9f608STao Ma #define OCFS2_MAX_PATH_DEPTH 5 277e2e9f608STao Ma 278e2e9f608STao Ma struct ocfs2_path { 279e2e9f608STao Ma int p_tree_depth; 280e2e9f608STao Ma ocfs2_journal_access_func p_root_access; 281e2e9f608STao Ma struct ocfs2_path_item p_node[OCFS2_MAX_PATH_DEPTH]; 282e2e9f608STao Ma }; 283e2e9f608STao Ma 284e2e9f608STao Ma #define path_root_bh(_path) ((_path)->p_node[0].bh) 285e2e9f608STao Ma #define path_root_el(_path) ((_path)->p_node[0].el) 286e2e9f608STao Ma #define path_root_access(_path)((_path)->p_root_access) 287e2e9f608STao Ma #define path_leaf_bh(_path) ((_path)->p_node[(_path)->p_tree_depth].bh) 288e2e9f608STao Ma #define path_leaf_el(_path) ((_path)->p_node[(_path)->p_tree_depth].el) 289e2e9f608STao Ma #define path_num_items(_path) ((_path)->p_tree_depth + 1) 290e2e9f608STao Ma 291e2e9f608STao Ma void ocfs2_reinit_path(struct ocfs2_path *path, int keep_root); 292e2e9f608STao Ma void ocfs2_free_path(struct ocfs2_path *path); 293e2e9f608STao Ma int ocfs2_find_path(struct ocfs2_caching_info *ci, 294e2e9f608STao Ma struct ocfs2_path *path, 295e2e9f608STao Ma u32 cpos); 296e2e9f608STao Ma struct ocfs2_path *ocfs2_new_path_from_path(struct ocfs2_path *path); 297e2e9f608STao Ma struct ocfs2_path *ocfs2_new_path_from_et(struct ocfs2_extent_tree *et); 298e2e9f608STao Ma int ocfs2_path_bh_journal_access(handle_t *handle, 299e2e9f608STao Ma struct ocfs2_caching_info *ci, 300e2e9f608STao Ma struct ocfs2_path *path, 301e2e9f608STao Ma int idx); 302e2e9f608STao Ma int ocfs2_journal_access_path(struct ocfs2_caching_info *ci, 303e2e9f608STao Ma handle_t *handle, 304e2e9f608STao Ma struct ocfs2_path *path); 30538a04e43STao Ma int ocfs2_find_cpos_for_right_leaf(struct super_block *sb, 30638a04e43STao Ma struct ocfs2_path *path, u32 *cpos); 307ee149a7cSTristan Ye int ocfs2_find_cpos_for_left_leaf(struct super_block *sb, 308ee149a7cSTristan Ye struct ocfs2_path *path, u32 *cpos); 30938a04e43STao Ma int ocfs2_find_subtree_root(struct ocfs2_extent_tree *et, 31038a04e43STao Ma struct ocfs2_path *left, 31138a04e43STao Ma struct ocfs2_path *right); 312ccd979bdSMark Fasheh #endif /* OCFS2_ALLOC_H */ 313