// SPDX-License-Identifier: GPL-2.0-only /* * refcounttree.c * * Copyright (C) 2009 Oracle. All rights reserved. */ #include <linux/sort.h> #include <cluster/masklog.h> #include "ocfs2.h" #include "inode.h" #include "alloc.h" #include "suballoc.h" #include "journal.h" #include "uptodate.h" #include "super.h" #include "buffer_head_io.h" #include "blockcheck.h" #include "refcounttree.h" #include "sysfile.h" #include "dlmglue.h" #include "extent_map.h" #include "aops.h" #include "xattr.h" #include "namei.h" #include "ocfs2_trace.h" #include "file.h" #include "symlink.h" #include <linux/bio.h> #include <linux/blkdev.h> #include <linux/slab.h> #include <linux/writeback.h> #include <linux/pagevec.h> #include <linux/swap.h> #include <linux/security.h> #include <linux/fsnotify.h> #include <linux/quotaops.h> #include <linux/namei.h> #include <linux/mount.h> #include <linux/posix_acl.h> struct ocfs2_cow_context { … }; static inline struct ocfs2_refcount_tree * cache_info_to_refcount(struct ocfs2_caching_info *ci) { … } static int ocfs2_validate_refcount_block(struct super_block *sb, struct buffer_head *bh) { … } static int ocfs2_read_refcount_block(struct ocfs2_caching_info *ci, u64 rb_blkno, struct buffer_head **bh) { … } static u64 ocfs2_refcount_cache_owner(struct ocfs2_caching_info *ci) { … } static struct super_block * ocfs2_refcount_cache_get_super(struct ocfs2_caching_info *ci) { … } static void ocfs2_refcount_cache_lock(struct ocfs2_caching_info *ci) __acquires(&rf->rf_lock) { … } static void ocfs2_refcount_cache_unlock(struct ocfs2_caching_info *ci) __releases(&rf->rf_lock) { … } static void ocfs2_refcount_cache_io_lock(struct ocfs2_caching_info *ci) { … } static void ocfs2_refcount_cache_io_unlock(struct ocfs2_caching_info *ci) { … } static const struct ocfs2_caching_operations ocfs2_refcount_caching_ops = …; static struct ocfs2_refcount_tree * ocfs2_find_refcount_tree(struct ocfs2_super *osb, u64 blkno) { … } /* osb_lock is already locked. */ static void ocfs2_insert_refcount_tree(struct ocfs2_super *osb, struct ocfs2_refcount_tree *new) { … } static void ocfs2_free_refcount_tree(struct ocfs2_refcount_tree *tree) { … } static inline void ocfs2_erase_refcount_tree_from_list_no_lock(struct ocfs2_super *osb, struct ocfs2_refcount_tree *tree) { … } static void ocfs2_erase_refcount_tree_from_list(struct ocfs2_super *osb, struct ocfs2_refcount_tree *tree) { … } static void ocfs2_kref_remove_refcount_tree(struct kref *kref) { … } static inline void ocfs2_refcount_tree_get(struct ocfs2_refcount_tree *tree) { … } static inline void ocfs2_refcount_tree_put(struct ocfs2_refcount_tree *tree) { … } static inline void ocfs2_init_refcount_tree_ci(struct ocfs2_refcount_tree *new, struct super_block *sb) { … } static inline void ocfs2_init_refcount_tree_lock(struct ocfs2_super *osb, struct ocfs2_refcount_tree *new, u64 rf_blkno, u32 generation) { … } static struct ocfs2_refcount_tree* ocfs2_allocate_refcount_tree(struct ocfs2_super *osb, u64 rf_blkno) { … } static int ocfs2_get_refcount_tree(struct ocfs2_super *osb, u64 rf_blkno, struct ocfs2_refcount_tree **ret_tree) { … } static int ocfs2_get_refcount_block(struct inode *inode, u64 *ref_blkno) { … } static int __ocfs2_lock_refcount_tree(struct ocfs2_super *osb, struct ocfs2_refcount_tree *tree, int rw) { … } /* * Lock the refcount tree pointed by ref_blkno and return the tree. * In most case, we lock the tree and read the refcount block. * So read it here if the caller really needs it. * * If the tree has been re-created by other node, it will free the * old one and re-create it. */ int ocfs2_lock_refcount_tree(struct ocfs2_super *osb, u64 ref_blkno, int rw, struct ocfs2_refcount_tree **ret_tree, struct buffer_head **ref_bh) { … } void ocfs2_unlock_refcount_tree(struct ocfs2_super *osb, struct ocfs2_refcount_tree *tree, int rw) { … } void ocfs2_purge_refcount_trees(struct ocfs2_super *osb) { … } /* * Create a refcount tree for an inode. * We take for granted that the inode is already locked. */ static int ocfs2_create_refcount_tree(struct inode *inode, struct buffer_head *di_bh) { … } static int ocfs2_set_refcount_tree(struct inode *inode, struct buffer_head *di_bh, u64 refcount_loc) { … } int ocfs2_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh) { … } static void ocfs2_find_refcount_rec_in_rl(struct ocfs2_caching_info *ci, struct buffer_head *ref_leaf_bh, u64 cpos, unsigned int len, struct ocfs2_refcount_rec *ret_rec, int *index) { … } /* * Try to remove refcount tree. The mechanism is: * 1) Check whether i_clusters == 0, if no, exit. * 2) check whether we have i_xattr_loc in dinode. if yes, exit. * 3) Check whether we have inline xattr stored outside, if yes, exit. * 4) Remove the tree. */ int ocfs2_try_remove_refcount_tree(struct inode *inode, struct buffer_head *di_bh) { … } /* * Find the end range for a leaf refcount block indicated by * el->l_recs[index].e_blkno. */ static int ocfs2_get_refcount_cpos_end(struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct ocfs2_extent_block *eb, struct ocfs2_extent_list *el, int index, u32 *cpos_end) { … } /* * Given a cpos and len, try to find the refcount record which contains cpos. * 1. If cpos can be found in one refcount record, return the record. * 2. If cpos can't be found, return a fake record which start from cpos * and end at a small value between cpos+len and start of the next record. * This fake record has r_refcount = 0. */ static int ocfs2_get_refcount_rec(struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, u64 cpos, unsigned int len, struct ocfs2_refcount_rec *ret_rec, int *index, struct buffer_head **ret_bh) { … } enum ocfs2_ref_rec_contig { … }; static enum ocfs2_ref_rec_contig ocfs2_refcount_rec_adjacent(struct ocfs2_refcount_block *rb, int index) { … } static enum ocfs2_ref_rec_contig ocfs2_refcount_rec_contig(struct ocfs2_refcount_block *rb, int index) { … } static void ocfs2_rotate_refcount_rec_left(struct ocfs2_refcount_block *rb, int index) { … } /* * Merge the refcount rec if we are contiguous with the adjacent recs. */ static void ocfs2_refcount_rec_merge(struct ocfs2_refcount_block *rb, int index) { … } /* * Change the refcount indexed by "index" in ref_bh. * If refcount reaches 0, remove it. */ static int ocfs2_change_refcount_rec(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_leaf_bh, int index, int merge, int change) { … } static int ocfs2_expand_inline_ref_root(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head **ref_leaf_bh, struct ocfs2_alloc_context *meta_ac) { … } static int ocfs2_refcount_rec_no_intersect(struct ocfs2_refcount_rec *prev, struct ocfs2_refcount_rec *next) { … } static int cmp_refcount_rec_by_low_cpos(const void *a, const void *b) { … } static int cmp_refcount_rec_by_cpos(const void *a, const void *b) { … } /* * The refcount cpos are ordered by their 64bit cpos, * But we will use the low 32 bit to be the e_cpos in the b-tree. * So we need to make sure that this pos isn't intersected with others. * * Note: The refcount block is already sorted by their low 32 bit cpos, * So just try the middle pos first, and we will exit when we find * the good position. */ static int ocfs2_find_refcount_split_pos(struct ocfs2_refcount_list *rl, u32 *split_pos, int *split_index) { … } static int ocfs2_divide_leaf_refcount_block(struct buffer_head *ref_leaf_bh, struct buffer_head *new_bh, u32 *split_cpos) { … } static int ocfs2_new_leaf_refcount_block(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head *ref_leaf_bh, struct ocfs2_alloc_context *meta_ac) { … } static int ocfs2_expand_refcount_tree(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head *ref_leaf_bh, struct ocfs2_alloc_context *meta_ac) { … } /* * Adjust the extent rec in b-tree representing ref_leaf_bh. * * Only called when we have inserted a new refcount rec at index 0 * which means ocfs2_extent_rec.e_cpos may need some change. */ static int ocfs2_adjust_refcount_rec(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head *ref_leaf_bh, struct ocfs2_refcount_rec *rec) { … } static int ocfs2_insert_refcount_rec(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head *ref_leaf_bh, struct ocfs2_refcount_rec *rec, int index, int merge, struct ocfs2_alloc_context *meta_ac) { … } /* * Split the refcount_rec indexed by "index" in ref_leaf_bh. * This is much simple than our b-tree code. * split_rec is the new refcount rec we want to insert. * If split_rec->r_refcount > 0, we are changing the refcount(in case we * increase refcount or decrease a refcount to non-zero). * If split_rec->r_refcount == 0, we are punching a hole in current refcount * rec( in case we decrease a refcount to zero). */ static int ocfs2_split_refcount_rec(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head *ref_leaf_bh, struct ocfs2_refcount_rec *split_rec, int index, int merge, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } static int __ocfs2_increase_refcount(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, u64 cpos, u32 len, int merge, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } static int ocfs2_remove_refcount_extent(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head *ref_leaf_bh, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } int ocfs2_increase_refcount(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, u64 cpos, u32 len, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } static int ocfs2_decrease_refcount_rec(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, struct buffer_head *ref_leaf_bh, int index, u64 cpos, unsigned int len, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } static int __ocfs2_decrease_refcount(handle_t *handle, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, u64 cpos, u32 len, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc, int delete) { … } /* Caller must hold refcount tree lock. */ int ocfs2_decrease_refcount(struct inode *inode, handle_t *handle, u32 cpos, u32 len, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc, int delete) { … } /* * Mark the already-existing extent at cpos as refcounted for len clusters. * This adds the refcount extent flag. * * If the existing extent is larger than the request, initiate a * split. An attempt will be made at merging with adjacent extents. * * The caller is responsible for passing down meta_ac if we'll need it. */ static int ocfs2_mark_extent_refcounted(struct inode *inode, struct ocfs2_extent_tree *et, handle_t *handle, u32 cpos, u32 len, u32 phys, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } /* * Given some contiguous physical clusters, calculate what we need * for modifying their refcount. */ static int ocfs2_calc_refcount_meta_credits(struct super_block *sb, struct ocfs2_caching_info *ci, struct buffer_head *ref_root_bh, u64 start_cpos, u32 clusters, int *meta_add, int *credits) { … } /* * For refcount tree, we will decrease some contiguous clusters * refcount count, so just go through it to see how many blocks * we gonna touch and whether we need to create new blocks. * * Normally the refcount blocks store these refcount should be * contiguous also, so that we can get the number easily. * We will at most add split 2 refcount records and 2 more * refcount blocks, so just check it in a rough way. * * Caller must hold refcount tree lock. */ int ocfs2_prepare_refcount_change_for_del(struct inode *inode, u64 refcount_loc, u64 phys_blkno, u32 clusters, int *credits, int *ref_blocks) { … } #define MAX_CONTIG_BYTES … static inline unsigned int ocfs2_cow_contig_clusters(struct super_block *sb) { … } static inline unsigned int ocfs2_cow_contig_mask(struct super_block *sb) { … } /* * Given an extent that starts at 'start' and an I/O that starts at 'cpos', * find an offset (start + (n * contig_clusters)) that is closest to cpos * while still being less than or equal to it. * * The goal is to break the extent at a multiple of contig_clusters. */ static inline unsigned int ocfs2_cow_align_start(struct super_block *sb, unsigned int start, unsigned int cpos) { … } /* * Given a cluster count of len, pad it out so that it is a multiple * of contig_clusters. */ static inline unsigned int ocfs2_cow_align_length(struct super_block *sb, unsigned int len) { … } /* * Calculate out the start and number of virtual clusters we need to CoW. * * cpos is vitual start cluster position we want to do CoW in a * file and write_len is the cluster length. * max_cpos is the place where we want to stop CoW intentionally. * * Normal we will start CoW from the beginning of extent record cotaining cpos. * We try to break up extents on boundaries of MAX_CONTIG_BYTES so that we * get good I/O from the resulting extent tree. */ static int ocfs2_refcount_cal_cow_clusters(struct inode *inode, struct ocfs2_extent_list *el, u32 cpos, u32 write_len, u32 max_cpos, u32 *cow_start, u32 *cow_len) { … } /* * Prepare meta_ac, data_ac and calculate credits when we want to add some * num_clusters in data_tree "et" and change the refcount for the old * clusters(starting form p_cluster) in the refcount tree. * * Note: * 1. since we may split the old tree, so we at most will need num_clusters + 2 * more new leaf records. * 2. In some case, we may not need to reserve new clusters(e.g, reflink), so * just give data_ac = NULL. */ static int ocfs2_lock_refcount_allocators(struct super_block *sb, u32 p_cluster, u32 num_clusters, struct ocfs2_extent_tree *et, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, struct ocfs2_alloc_context **meta_ac, struct ocfs2_alloc_context **data_ac, int *credits) { … } static int ocfs2_clear_cow_buffer(handle_t *handle, struct buffer_head *bh) { … } int ocfs2_duplicate_clusters_by_page(handle_t *handle, struct inode *inode, u32 cpos, u32 old_cluster, u32 new_cluster, u32 new_len) { … } int ocfs2_duplicate_clusters_by_jbd(handle_t *handle, struct inode *inode, u32 cpos, u32 old_cluster, u32 new_cluster, u32 new_len) { … } static int ocfs2_clear_ext_refcount(handle_t *handle, struct ocfs2_extent_tree *et, u32 cpos, u32 p_cluster, u32 len, unsigned int ext_flags, struct ocfs2_alloc_context *meta_ac, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } static int ocfs2_replace_clusters(handle_t *handle, struct ocfs2_cow_context *context, u32 cpos, u32 old, u32 new, u32 len, unsigned int ext_flags) { … } int ocfs2_cow_sync_writeback(struct super_block *sb, struct inode *inode, u32 cpos, u32 num_clusters) { … } static int ocfs2_di_get_clusters(struct ocfs2_cow_context *context, u32 v_cluster, u32 *p_cluster, u32 *num_clusters, unsigned int *extent_flags) { … } static int ocfs2_make_clusters_writable(struct super_block *sb, struct ocfs2_cow_context *context, u32 cpos, u32 p_cluster, u32 num_clusters, unsigned int e_flags) { … } static int ocfs2_replace_cow(struct ocfs2_cow_context *context) { … } /* * Starting at cpos, try to CoW write_len clusters. Don't CoW * past max_cpos. This will stop when it runs into a hole or an * unrefcounted extent. */ static int ocfs2_refcount_cow_hunk(struct inode *inode, struct buffer_head *di_bh, u32 cpos, u32 write_len, u32 max_cpos) { … } /* * CoW any and all clusters between cpos and cpos+write_len. * Don't CoW past max_cpos. If this returns successfully, all * clusters between cpos and cpos+write_len are safe to modify. */ int ocfs2_refcount_cow(struct inode *inode, struct buffer_head *di_bh, u32 cpos, u32 write_len, u32 max_cpos) { … } static int ocfs2_xattr_value_get_clusters(struct ocfs2_cow_context *context, u32 v_cluster, u32 *p_cluster, u32 *num_clusters, unsigned int *extent_flags) { … } /* * Given a xattr value root, calculate the most meta/credits we need for * refcount tree change if we truncate it to 0. */ int ocfs2_refcounted_xattr_delete_need(struct inode *inode, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, struct ocfs2_xattr_value_root *xv, int *meta_add, int *credits) { … } /* * Do CoW for xattr. */ int ocfs2_refcount_cow_xattr(struct inode *inode, struct ocfs2_dinode *di, struct ocfs2_xattr_value_buf *vb, struct ocfs2_refcount_tree *ref_tree, struct buffer_head *ref_root_bh, u32 cpos, u32 write_len, struct ocfs2_post_refcount *post) { … } /* * Insert a new extent into refcount tree and mark a extent rec * as refcounted in the dinode tree. */ int ocfs2_add_refcount_flag(struct inode *inode, struct ocfs2_extent_tree *data_et, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, u32 cpos, u32 p_cluster, u32 num_clusters, struct ocfs2_cached_dealloc_ctxt *dealloc, struct ocfs2_post_refcount *post) { … } static int ocfs2_change_ctime(struct inode *inode, struct buffer_head *di_bh) { … } static int ocfs2_attach_refcount_tree(struct inode *inode, struct buffer_head *di_bh) { … } static int ocfs2_add_refcounted_extent(struct inode *inode, struct ocfs2_extent_tree *et, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, u32 cpos, u32 p_cluster, u32 num_clusters, unsigned int ext_flags, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } static int ocfs2_duplicate_inline_data(struct inode *s_inode, struct buffer_head *s_bh, struct inode *t_inode, struct buffer_head *t_bh) { … } static int ocfs2_duplicate_extent_list(struct inode *s_inode, struct inode *t_inode, struct buffer_head *t_bh, struct ocfs2_caching_info *ref_ci, struct buffer_head *ref_root_bh, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } /* * change the new file's attributes to the src. * * reflink creates a snapshot of a file, that means the attributes * must be identical except for three exceptions - nlink, ino, and ctime. */ static int ocfs2_complete_reflink(struct inode *s_inode, struct buffer_head *s_bh, struct inode *t_inode, struct buffer_head *t_bh, bool preserve) { … } static int ocfs2_create_reflink_node(struct inode *s_inode, struct buffer_head *s_bh, struct inode *t_inode, struct buffer_head *t_bh, bool preserve) { … } static int __ocfs2_reflink(struct dentry *old_dentry, struct buffer_head *old_bh, struct inode *new_inode, bool preserve) { … } static int ocfs2_reflink(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, bool preserve) { … } /* * Below here are the bits used by OCFS2_IOC_REFLINK() to fake * sys_reflink(). This will go away when vfs_reflink() exists in * fs/namei.c. */ /* copied from may_create in VFS. */ static inline int ocfs2_may_create(struct inode *dir, struct dentry *child) { … } /** * ocfs2_vfs_reflink - Create a reference-counted link * * @old_dentry: source dentry + inode * @dir: directory to create the target * @new_dentry: target dentry * @preserve: if true, preserve all file attributes */ static int ocfs2_vfs_reflink(struct dentry *old_dentry, struct inode *dir, struct dentry *new_dentry, bool preserve) { … } /* * Most codes are copied from sys_linkat. */ int ocfs2_reflink_ioctl(struct inode *inode, const char __user *oldname, const char __user *newname, bool preserve) { … } /* Update destination inode size, if necessary. */ int ocfs2_reflink_update_dest(struct inode *dest, struct buffer_head *d_bh, loff_t newlen) { … } /* Remap the range pos_in:len in s_inode to pos_out:len in t_inode. */ static loff_t ocfs2_reflink_remap_extent(struct inode *s_inode, struct buffer_head *s_bh, loff_t pos_in, struct inode *t_inode, struct buffer_head *t_bh, loff_t pos_out, loff_t len, struct ocfs2_cached_dealloc_ctxt *dealloc) { … } /* Set up refcount tree and remap s_inode to t_inode. */ loff_t ocfs2_reflink_remap_blocks(struct inode *s_inode, struct buffer_head *s_bh, loff_t pos_in, struct inode *t_inode, struct buffer_head *t_bh, loff_t pos_out, loff_t len) { … } /* Lock an inode and grab a bh pointing to the inode. */ int ocfs2_reflink_inodes_lock(struct inode *s_inode, struct buffer_head **bh_s, struct inode *t_inode, struct buffer_head **bh_t) { … } /* Unlock both inodes and release buffers. */ void ocfs2_reflink_inodes_unlock(struct inode *s_inode, struct buffer_head *s_bh, struct inode *t_inode, struct buffer_head *t_bh) { … }
Codebase Browser
linux