/* SPDX-License-Identifier: GPL-2.0 */1/*2* Copyright (C) 2007 Oracle. All rights reserved.3*/45#ifndef BTRFS_INODE_H6#define BTRFS_INODE_H78#include <linux/hash.h>9#include <linux/refcount.h>10#include <linux/spinlock.h>11#include <linux/mutex.h>12#include <linux/rwsem.h>13#include <linux/fs.h>14#include <linux/mm.h>15#include <linux/compiler.h>16#include <linux/fscrypt.h>17#include <linux/lockdep.h>18#include <uapi/linux/btrfs_tree.h>19#include <trace/events/btrfs.h>20#include "block-rsv.h"21#include "extent_map.h"22#include "extent_io.h"23#include "extent-io-tree.h"24#include "ordered-data.h"25#include "delayed-inode.h"2627struct extent_state;28struct posix_acl;29struct iov_iter;30struct writeback_control;31struct btrfs_root;32struct btrfs_fs_info;33struct btrfs_trans_handle;3435/*36* Since we search a directory based on f_pos (struct dir_context::pos) we have37* to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so38* everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).39*/40#define BTRFS_DIR_START_INDEX 24142/*43* ordered_data_close is set by truncate when a file that used44* to have good data has been truncated to zero. When it is set45* the btrfs file release call will add this inode to the46* ordered operations list so that we make sure to flush out any47* new data the application may have written before commit.48*/49enum {50BTRFS_INODE_FLUSH_ON_CLOSE,51BTRFS_INODE_DUMMY,52BTRFS_INODE_IN_DEFRAG,53BTRFS_INODE_HAS_ASYNC_EXTENT,54/*55* Always set under the VFS' inode lock, otherwise it can cause races56* during fsync (we start as a fast fsync and then end up in a full57* fsync racing with ordered extent completion).58*/59BTRFS_INODE_NEEDS_FULL_SYNC,60BTRFS_INODE_COPY_EVERYTHING,61BTRFS_INODE_HAS_PROPS,62BTRFS_INODE_SNAPSHOT_FLUSH,63/*64* Set and used when logging an inode and it serves to signal that an65* inode does not have xattrs, so subsequent fsyncs can avoid searching66* for xattrs to log. This bit must be cleared whenever a xattr is added67* to an inode.68*/69BTRFS_INODE_NO_XATTRS,70/*71* Set when we are in a context where we need to start a transaction and72* have dirty pages with the respective file range locked. This is to73* ensure that when reserving space for the transaction, if we are low74* on available space and need to flush delalloc, we will not flush75* delalloc for this inode, because that could result in a deadlock (on76* the file range, inode's io_tree).77*/78BTRFS_INODE_NO_DELALLOC_FLUSH,79/*80* Set when we are working on enabling verity for a file. Computing and81* writing the whole Merkle tree can take a while so we want to prevent82* races where two separate tasks attempt to simultaneously start verity83* on the same file.84*/85BTRFS_INODE_VERITY_IN_PROGRESS,86/* Set when this inode is a free space inode. */87BTRFS_INODE_FREE_SPACE_INODE,88/* Set when there are no capabilities in XATTs for the inode. */89BTRFS_INODE_NO_CAP_XATTR,90/*91* Set if an error happened when doing a COW write before submitting a92* bio or during writeback. Used for both buffered writes and direct IO93* writes. This is to signal a fast fsync that it has to wait for94* ordered extents to complete and therefore not log extent maps that95* point to unwritten extents (when an ordered extent completes and it96* has the BTRFS_ORDERED_IOERR flag set, it drops extent maps in its97* range).98*/99BTRFS_INODE_COW_WRITE_ERROR,100/*101* Indicate this is a directory that points to a subvolume for which102* there is no root reference item. That's a case like the following:103*104* $ btrfs subvolume create /mnt/parent105* $ btrfs subvolume create /mnt/parent/child106* $ btrfs subvolume snapshot /mnt/parent /mnt/snap107*108* If subvolume "parent" is root 256, subvolume "child" is root 257 and109* snapshot "snap" is root 258, then there's no root reference item (key110* BTRFS_ROOT_REF_KEY in the root tree) for the subvolume "child"111* associated to root 258 (the snapshot) - there's only for the root112* of the "parent" subvolume (root 256). In the chunk root we have a113* (256 BTRFS_ROOT_REF_KEY 257) key but we don't have a114* (258 BTRFS_ROOT_REF_KEY 257) key - the sames goes for backrefs, we115* have a (257 BTRFS_ROOT_BACKREF_KEY 256) but we don't have a116* (257 BTRFS_ROOT_BACKREF_KEY 258) key.117*118* So when opening the "child" dentry from the snapshot's directory,119* we don't find a root ref item and we create a stub inode. This is120* done at new_simple_dir(), called from btrfs_lookup_dentry().121*/122BTRFS_INODE_ROOT_STUB,123};124125/* in memory btrfs inode */126struct btrfs_inode {127/* which subvolume this inode belongs to */128struct btrfs_root *root;129130#if BITS_PER_LONG == 32131/*132* The objectid of the corresponding BTRFS_INODE_ITEM_KEY.133* On 64 bits platforms we can get it from vfs_inode.i_ino, which is an134* unsigned long and therefore 64 bits on such platforms.135*/136u64 objectid;137#endif138139/* Cached value of inode property 'compression'. */140u8 prop_compress;141142/*143* Force compression on the file using the defrag ioctl, could be144* different from prop_compress and takes precedence if set.145*/146u8 defrag_compress;147s8 defrag_compress_level;148149/*150* Lock for counters and all fields used to determine if the inode is in151* the log or not (last_trans, last_sub_trans, last_log_commit,152* logged_trans), to access/update delalloc_bytes, new_delalloc_bytes,153* defrag_bytes, disk_i_size, outstanding_extents, csum_bytes and to154* update the VFS' inode number of bytes used.155* Also protects setting struct file::private_data.156*/157spinlock_t lock;158159/* the extent_tree has caches of all the extent mappings to disk */160struct extent_map_tree extent_tree;161162/* the io_tree does range state (DIRTY, LOCKED etc) */163struct extent_io_tree io_tree;164165/*166* Keep track of where the inode has extent items mapped in order to167* make sure the i_size adjustments are accurate. Not required when the168* filesystem is NO_HOLES, the status can't be set while mounted as169* it's a mkfs-time feature.170*/171struct extent_io_tree *file_extent_tree;172173/* held while logging the inode in tree-log.c */174struct mutex log_mutex;175176/*177* Counters to keep track of the number of extent item's we may use due178* to delalloc and such. outstanding_extents is the number of extent179* items we think we'll end up using, and reserved_extents is the number180* of extent items we've reserved metadata for. Protected by 'lock'.181*/182unsigned outstanding_extents;183184/* used to order data wrt metadata */185spinlock_t ordered_tree_lock;186struct rb_root ordered_tree;187struct rb_node *ordered_tree_last;188189/* list of all the delalloc inodes in the FS. There are times we need190* to write all the delalloc pages to disk, and this list is used191* to walk them all.192*/193struct list_head delalloc_inodes;194195unsigned long runtime_flags;196197/* full 64 bit generation number, struct vfs_inode doesn't have a big198* enough field for this.199*/200u64 generation;201202/*203* ID of the transaction handle that last modified this inode.204* Protected by 'lock'.205*/206u64 last_trans;207208/*209* ID of the transaction that last logged this inode.210* Protected by 'lock'.211*/212u64 logged_trans;213214/*215* Log transaction ID when this inode was last modified.216* Protected by 'lock'.217*/218int last_sub_trans;219220/* A local copy of root's last_log_commit. Protected by 'lock'. */221int last_log_commit;222223union {224/*225* Total number of bytes pending delalloc, used by stat to226* calculate the real block usage of the file. This is used227* only for files. Protected by 'lock'.228*/229u64 delalloc_bytes;230/*231* The lowest possible index of the next dir index key which232* points to an inode that needs to be logged.233* This is used only for directories.234* Use the helpers btrfs_get_first_dir_index_to_log() and235* btrfs_set_first_dir_index_to_log() to access this field.236*/237u64 first_dir_index_to_log;238};239240union {241/*242* Total number of bytes pending delalloc that fall within a file243* range that is either a hole or beyond EOF (and no prealloc extent244* exists in the range). This is always <= delalloc_bytes and this245* is used only for files. Protected by 'lock'.246*/247u64 new_delalloc_bytes;248/*249* The offset of the last dir index key that was logged.250* This is used only for directories. Protected by 'log_mutex'.251*/252u64 last_dir_index_offset;253};254255union {256/*257* Total number of bytes pending defrag, used by stat to check whether258* it needs COW. Protected by 'lock'.259* Used by inodes other than the data relocation inode.260*/261u64 defrag_bytes;262263/*264* Logical address of the block group being relocated.265* Used only by the data relocation inode.266*/267u64 reloc_block_group_start;268};269270/*271* The size of the file stored in the metadata on disk. data=ordered272* means the in-memory i_size might be larger than the size on disk273* because not all the blocks are written yet. Protected by 'lock'.274*/275u64 disk_i_size;276277union {278/*279* If this is a directory then index_cnt is the counter for the280* index number for new files that are created. For an empty281* directory, this must be initialized to BTRFS_DIR_START_INDEX.282*/283u64 index_cnt;284285/*286* If this is not a directory, this is the number of bytes287* outstanding that are going to need csums. This is used in288* ENOSPC accounting. Protected by 'lock'.289*/290u64 csum_bytes;291};292293/* Cache the directory index number to speed the dir/file remove */294u64 dir_index;295296/* the fsync log has some corner cases that mean we have to check297* directories to see if any unlinks have been done before298* the directory was logged. See tree-log.c for all the299* details300*/301u64 last_unlink_trans;302303union {304/*305* The id/generation of the last transaction where this inode306* was either the source or the destination of a clone/dedupe307* operation. Used when logging an inode to know if there are308* shared extents that need special care when logging checksum309* items, to avoid duplicate checksum items in a log (which can310* lead to a corruption where we end up with missing checksum311* ranges after log replay). Protected by the VFS inode lock.312* Used for regular files only.313*/314u64 last_reflink_trans;315316/*317* In case this a root stub inode (BTRFS_INODE_ROOT_STUB flag set),318* the ID of that root.319*/320u64 ref_root_id;321};322323/* Backwards incompatible flags, lower half of inode_item::flags */324u32 flags;325/* Read-only compatibility flags, upper half of inode_item::flags */326u32 ro_flags;327328struct btrfs_block_rsv block_rsv;329330struct btrfs_delayed_node *delayed_node;331332/* File creation time. */333u64 i_otime_sec;334u32 i_otime_nsec;335336/* Hook into fs_info->delayed_iputs */337struct list_head delayed_iput;338339struct rw_semaphore i_mmap_lock;340struct inode vfs_inode;341};342343static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)344{345return READ_ONCE(inode->first_dir_index_to_log);346}347348static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,349u64 index)350{351WRITE_ONCE(inode->first_dir_index_to_log, index);352}353354/* Type checked and const-preserving VFS inode -> btrfs inode. */355#define BTRFS_I(_inode) \356_Generic(_inode, \357struct inode *: container_of(_inode, struct btrfs_inode, vfs_inode), \358const struct inode *: (const struct btrfs_inode *)container_of( \359_inode, const struct btrfs_inode, vfs_inode))360361static inline unsigned long btrfs_inode_hash(u64 objectid,362const struct btrfs_root *root)363{364u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);365366#if BITS_PER_LONG == 32367h = (h >> 32) ^ (h & 0xffffffff);368#endif369370return (unsigned long)h;371}372373#if BITS_PER_LONG == 32374375/*376* On 32 bit systems the i_ino of struct inode is 32 bits (unsigned long), so377* we use the inode's location objectid which is a u64 to avoid truncation.378*/379static inline u64 btrfs_ino(const struct btrfs_inode *inode)380{381u64 ino = inode->objectid;382383if (test_bit(BTRFS_INODE_ROOT_STUB, &inode->runtime_flags))384ino = inode->vfs_inode.i_ino;385return ino;386}387388#else389390static inline u64 btrfs_ino(const struct btrfs_inode *inode)391{392return inode->vfs_inode.i_ino;393}394395#endif396397static inline void btrfs_get_inode_key(const struct btrfs_inode *inode,398struct btrfs_key *key)399{400key->objectid = btrfs_ino(inode);401key->type = BTRFS_INODE_ITEM_KEY;402key->offset = 0;403}404405static inline void btrfs_set_inode_number(struct btrfs_inode *inode, u64 ino)406{407#if BITS_PER_LONG == 32408inode->objectid = ino;409#endif410inode->vfs_inode.i_ino = ino;411}412413static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)414{415i_size_write(&inode->vfs_inode, size);416inode->disk_i_size = size;417}418419static inline bool btrfs_is_free_space_inode(const struct btrfs_inode *inode)420{421return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);422}423424static inline bool is_data_inode(const struct btrfs_inode *inode)425{426return btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID;427}428429static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,430int mod)431{432lockdep_assert_held(&inode->lock);433inode->outstanding_extents += mod;434if (btrfs_is_free_space_inode(inode))435return;436trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),437mod, inode->outstanding_extents);438}439440/*441* Called every time after doing a buffered, direct IO or memory mapped write.442*443* This is to ensure that if we write to a file that was previously fsynced in444* the current transaction, then try to fsync it again in the same transaction,445* we will know that there were changes in the file and that it needs to be446* logged.447*/448static inline void btrfs_set_inode_last_sub_trans(struct btrfs_inode *inode)449{450spin_lock(&inode->lock);451inode->last_sub_trans = inode->root->log_transid;452spin_unlock(&inode->lock);453}454455/*456* Should be called while holding the inode's VFS lock in exclusive mode, or457* while holding the inode's mmap lock (struct btrfs_inode::i_mmap_lock) in458* either shared or exclusive mode, or in a context where no one else can access459* the inode concurrently (during inode creation or when loading an inode from460* disk).461*/462static inline void btrfs_set_inode_full_sync(struct btrfs_inode *inode)463{464set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &inode->runtime_flags);465/*466* The inode may have been part of a reflink operation in the last467* transaction that modified it, and then a fsync has reset the468* last_reflink_trans to avoid subsequent fsyncs in the same469* transaction to do unnecessary work. So update last_reflink_trans470* to the last_trans value (we have to be pessimistic and assume a471* reflink happened).472*473* The ->last_trans is protected by the inode's spinlock and we can474* have a concurrent ordered extent completion update it. Also set475* last_reflink_trans to ->last_trans only if the former is less than476* the later, because we can be called in a context where477* last_reflink_trans was set to the current transaction generation478* while ->last_trans was not yet updated in the current transaction,479* and therefore has a lower value.480*/481spin_lock(&inode->lock);482if (inode->last_reflink_trans < inode->last_trans)483inode->last_reflink_trans = inode->last_trans;484spin_unlock(&inode->lock);485}486487static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)488{489bool ret = false;490491spin_lock(&inode->lock);492if (inode->logged_trans == generation &&493inode->last_sub_trans <= inode->last_log_commit &&494inode->last_sub_trans <= btrfs_get_root_last_log_commit(inode->root))495ret = true;496spin_unlock(&inode->lock);497return ret;498}499500/*501* Check if the inode has flags compatible with compression502*/503static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)504{505if (inode->flags & BTRFS_INODE_NODATACOW ||506inode->flags & BTRFS_INODE_NODATASUM)507return false;508return true;509}510511static inline void btrfs_assert_inode_locked(struct btrfs_inode *inode)512{513/* Immediately trigger a crash if the inode is not locked. */514ASSERT(inode_is_locked(&inode->vfs_inode));515/* Trigger a splat in dmesg if this task is not holding the lock. */516lockdep_assert_held(&inode->vfs_inode.i_rwsem);517}518519static inline void btrfs_update_inode_mapping_flags(struct btrfs_inode *inode)520{521if (inode->flags & BTRFS_INODE_NODATASUM)522mapping_clear_stable_writes(inode->vfs_inode.i_mapping);523else524mapping_set_stable_writes(inode->vfs_inode.i_mapping);525}526527static inline void btrfs_set_inode_mapping_order(struct btrfs_inode *inode)528{529/* Metadata inode should not reach here. */530ASSERT(is_data_inode(inode));531532/* We only allow BITS_PER_LONGS blocks for each bitmap. */533#ifdef CONFIG_BTRFS_EXPERIMENTAL534mapping_set_folio_order_range(inode->vfs_inode.i_mapping, 0,535ilog2(((BITS_PER_LONG << inode->root->fs_info->sectorsize_bits)536>> PAGE_SHIFT)));537#endif538}539540/* Array of bytes with variable length, hexadecimal format 0x1234 */541#define CSUM_FMT "0x%*phN"542#define CSUM_FMT_VALUE(size, bytes) size, bytes543544int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, void *kaddr, u8 *csum,545const u8 * const csum_expected);546bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,547u32 bio_offset, struct bio_vec *bv);548noinline int can_nocow_extent(struct btrfs_inode *inode, u64 offset, u64 *len,549struct btrfs_file_extent *file_extent,550bool nowait);551552void btrfs_del_delalloc_inode(struct btrfs_inode *inode);553struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);554int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);555int btrfs_unlink_inode(struct btrfs_trans_handle *trans,556struct btrfs_inode *dir, struct btrfs_inode *inode,557const struct fscrypt_str *name);558int btrfs_add_link(struct btrfs_trans_handle *trans,559struct btrfs_inode *parent_inode, struct btrfs_inode *inode,560const struct fscrypt_str *name, int add_backref, u64 index);561int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);562int btrfs_truncate_block(struct btrfs_inode *inode, u64 offset, u64 start, u64 end);563564int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);565int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,566bool in_reclaim_context);567int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,568unsigned int extra_bits,569struct extent_state **cached_state);570571struct btrfs_new_inode_args {572/* Input */573struct inode *dir;574struct dentry *dentry;575struct inode *inode;576bool orphan;577bool subvol;578579/* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */580struct posix_acl *default_acl;581struct posix_acl *acl;582struct fscrypt_name fname;583};584585int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,586unsigned int *trans_num_items);587int btrfs_create_new_inode(struct btrfs_trans_handle *trans,588struct btrfs_new_inode_args *args);589void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);590struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,591struct inode *dir);592void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,593u32 bits);594void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,595struct extent_state *state, u32 bits);596void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,597struct extent_state *other);598void btrfs_split_delalloc_extent(struct btrfs_inode *inode,599struct extent_state *orig, u64 split);600void btrfs_evict_inode(struct inode *inode);601struct inode *btrfs_alloc_inode(struct super_block *sb);602void btrfs_destroy_inode(struct inode *inode);603void btrfs_free_inode(struct inode *inode);604int btrfs_drop_inode(struct inode *inode);605int __init btrfs_init_cachep(void);606void __cold btrfs_destroy_cachep(void);607struct btrfs_inode *btrfs_iget_path(u64 ino, struct btrfs_root *root,608struct btrfs_path *path);609struct btrfs_inode *btrfs_iget(u64 ino, struct btrfs_root *root);610struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,611struct folio *folio, u64 start, u64 len);612int btrfs_update_inode(struct btrfs_trans_handle *trans,613struct btrfs_inode *inode);614int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,615struct btrfs_inode *inode);616int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);617int btrfs_orphan_cleanup(struct btrfs_root *root);618int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);619void btrfs_add_delayed_iput(struct btrfs_inode *inode);620void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);621int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);622int btrfs_prealloc_file_range(struct inode *inode, int mode,623u64 start, u64 num_bytes, u64 min_size,624loff_t actual_len, u64 *alloc_hint);625int btrfs_prealloc_file_range_trans(struct inode *inode,626struct btrfs_trans_handle *trans, int mode,627u64 start, u64 num_bytes, u64 min_size,628loff_t actual_len, u64 *alloc_hint);629int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct folio *locked_folio,630u64 start, u64 end, struct writeback_control *wbc);631int btrfs_writepage_cow_fixup(struct folio *folio);632int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,633int compress_type);634int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,635u64 disk_bytenr, u64 disk_io_size,636struct page **pages, void *uring_ctx);637ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,638struct btrfs_ioctl_encoded_io_args *encoded,639struct extent_state **cached_state,640u64 *disk_bytenr, u64 *disk_io_size);641ssize_t btrfs_encoded_read_regular(struct kiocb *iocb, struct iov_iter *iter,642u64 start, u64 lockend,643struct extent_state **cached_state,644u64 disk_bytenr, u64 disk_io_size,645size_t count, bool compressed, bool *unlocked);646ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,647const struct btrfs_ioctl_encoded_io_args *encoded);648649struct btrfs_inode *btrfs_find_first_inode(struct btrfs_root *root, u64 min_ino);650651extern const struct dentry_operations btrfs_dentry_operations;652653/* Inode locking type flags, by default the exclusive lock is taken. */654enum btrfs_ilock_type {655ENUM_BIT(BTRFS_ILOCK_SHARED),656ENUM_BIT(BTRFS_ILOCK_TRY),657ENUM_BIT(BTRFS_ILOCK_MMAP),658};659660int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);661void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);662void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,663const u64 del_bytes);664void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);665u64 btrfs_get_extent_allocation_hint(struct btrfs_inode *inode, u64 start,666u64 num_bytes);667struct extent_map *btrfs_create_io_em(struct btrfs_inode *inode, u64 start,668const struct btrfs_file_extent *file_extent,669int type);670671#endif672673674