647 lines · c
1/* SPDX-License-Identifier: GPL-2.0 */2/*3 * Copyright (C) 2007 Oracle. All rights reserved.4 */5 6#ifndef BTRFS_INODE_H7#define BTRFS_INODE_H8 9#include <linux/hash.h>10#include <linux/refcount.h>11#include <linux/spinlock.h>12#include <linux/mutex.h>13#include <linux/rwsem.h>14#include <linux/fs.h>15#include <linux/mm.h>16#include <linux/compiler.h>17#include <linux/fscrypt.h>18#include <linux/lockdep.h>19#include <uapi/linux/btrfs_tree.h>20#include <trace/events/btrfs.h>21#include "block-rsv.h"22#include "extent_map.h"23#include "extent_io.h"24#include "extent-io-tree.h"25#include "ordered-data.h"26#include "delayed-inode.h"27 28struct extent_state;29struct posix_acl;30struct iov_iter;31struct writeback_control;32struct btrfs_root;33struct btrfs_fs_info;34struct btrfs_trans_handle;35 36/*37 * Since we search a directory based on f_pos (struct dir_context::pos) we have38 * to start at 2 since '.' and '..' have f_pos of 0 and 1 respectively, so39 * everybody else has to start at 2 (see btrfs_real_readdir() and dir_emit_dots()).40 */41#define BTRFS_DIR_START_INDEX 242 43/*44 * ordered_data_close is set by truncate when a file that used45 * to have good data has been truncated to zero. When it is set46 * the btrfs file release call will add this inode to the47 * ordered operations list so that we make sure to flush out any48 * new data the application may have written before commit.49 */50enum {51 BTRFS_INODE_FLUSH_ON_CLOSE,52 BTRFS_INODE_DUMMY,53 BTRFS_INODE_IN_DEFRAG,54 BTRFS_INODE_HAS_ASYNC_EXTENT,55 /*56 * Always set under the VFS' inode lock, otherwise it can cause races57 * during fsync (we start as a fast fsync and then end up in a full58 * fsync racing with ordered extent completion).59 */60 BTRFS_INODE_NEEDS_FULL_SYNC,61 BTRFS_INODE_COPY_EVERYTHING,62 BTRFS_INODE_HAS_PROPS,63 BTRFS_INODE_SNAPSHOT_FLUSH,64 /*65 * Set and used when logging an inode and it serves to signal that an66 * inode does not have xattrs, so subsequent fsyncs can avoid searching67 * for xattrs to log. This bit must be cleared whenever a xattr is added68 * to an inode.69 */70 BTRFS_INODE_NO_XATTRS,71 /*72 * Set when we are in a context where we need to start a transaction and73 * have dirty pages with the respective file range locked. This is to74 * ensure that when reserving space for the transaction, if we are low75 * on available space and need to flush delalloc, we will not flush76 * delalloc for this inode, because that could result in a deadlock (on77 * the file range, inode's io_tree).78 */79 BTRFS_INODE_NO_DELALLOC_FLUSH,80 /*81 * Set when we are working on enabling verity for a file. Computing and82 * writing the whole Merkle tree can take a while so we want to prevent83 * races where two separate tasks attempt to simultaneously start verity84 * on the same file.85 */86 BTRFS_INODE_VERITY_IN_PROGRESS,87 /* Set when this inode is a free space inode. */88 BTRFS_INODE_FREE_SPACE_INODE,89 /* Set when there are no capabilities in XATTs for the inode. */90 BTRFS_INODE_NO_CAP_XATTR,91 /*92 * Set if an error happened when doing a COW write before submitting a93 * bio or during writeback. Used for both buffered writes and direct IO94 * writes. This is to signal a fast fsync that it has to wait for95 * ordered extents to complete and therefore not log extent maps that96 * point to unwritten extents (when an ordered extent completes and it97 * has the BTRFS_ORDERED_IOERR flag set, it drops extent maps in its98 * range).99 */100 BTRFS_INODE_COW_WRITE_ERROR,101 /*102 * Indicate this is a directory that points to a subvolume for which103 * there is no root reference item. That's a case like the following:104 *105 * $ btrfs subvolume create /mnt/parent106 * $ btrfs subvolume create /mnt/parent/child107 * $ btrfs subvolume snapshot /mnt/parent /mnt/snap108 *109 * If subvolume "parent" is root 256, subvolume "child" is root 257 and110 * snapshot "snap" is root 258, then there's no root reference item (key111 * BTRFS_ROOT_REF_KEY in the root tree) for the subvolume "child"112 * associated to root 258 (the snapshot) - there's only for the root113 * of the "parent" subvolume (root 256). In the chunk root we have a114 * (256 BTRFS_ROOT_REF_KEY 257) key but we don't have a115 * (258 BTRFS_ROOT_REF_KEY 257) key - the sames goes for backrefs, we116 * have a (257 BTRFS_ROOT_BACKREF_KEY 256) but we don't have a117 * (257 BTRFS_ROOT_BACKREF_KEY 258) key.118 *119 * So when opening the "child" dentry from the snapshot's directory,120 * we don't find a root ref item and we create a stub inode. This is121 * done at new_simple_dir(), called from btrfs_lookup_dentry().122 */123 BTRFS_INODE_ROOT_STUB,124};125 126/* in memory btrfs inode */127struct btrfs_inode {128 /* which subvolume this inode belongs to */129 struct btrfs_root *root;130 131#if BITS_PER_LONG == 32132 /*133 * The objectid of the corresponding BTRFS_INODE_ITEM_KEY.134 * On 64 bits platforms we can get it from vfs_inode.i_ino, which is an135 * unsigned long and therefore 64 bits on such platforms.136 */137 u64 objectid;138#endif139 140 /* Cached value of inode property 'compression'. */141 u8 prop_compress;142 143 /*144 * Force compression on the file using the defrag ioctl, could be145 * different from prop_compress and takes precedence if set.146 */147 u8 defrag_compress;148 149 /*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 */157 spinlock_t lock;158 159 /* the extent_tree has caches of all the extent mappings to disk */160 struct extent_map_tree extent_tree;161 162 /* the io_tree does range state (DIRTY, LOCKED etc) */163 struct extent_io_tree io_tree;164 165 /*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 */171 struct extent_io_tree *file_extent_tree;172 173 /* held while logging the inode in tree-log.c */174 struct mutex log_mutex;175 176 /*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 */182 unsigned outstanding_extents;183 184 /* used to order data wrt metadata */185 spinlock_t ordered_tree_lock;186 struct rb_root ordered_tree;187 struct rb_node *ordered_tree_last;188 189 /* 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 */193 struct list_head delalloc_inodes;194 195 unsigned long runtime_flags;196 197 /* full 64 bit generation number, struct vfs_inode doesn't have a big198 * enough field for this.199 */200 u64 generation;201 202 /*203 * ID of the transaction handle that last modified this inode.204 * Protected by 'lock'.205 */206 u64 last_trans;207 208 /*209 * ID of the transaction that last logged this inode.210 * Protected by 'lock'.211 */212 u64 logged_trans;213 214 /*215 * Log transaction ID when this inode was last modified.216 * Protected by 'lock'.217 */218 int last_sub_trans;219 220 /* A local copy of root's last_log_commit. Protected by 'lock'. */221 int last_log_commit;222 223 union {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 */229 u64 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 */237 u64 first_dir_index_to_log;238 };239 240 union {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 */247 u64 new_delalloc_bytes;248 /*249 * The offset of the last dir index key that was logged.250 * This is used only for directories.251 */252 u64 last_dir_index_offset;253 };254 255 union {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 */261 u64 defrag_bytes;262 263 /*264 * Logical address of the block group being relocated.265 * Used only by the data relocation inode.266 */267 u64 reloc_block_group_start;268 };269 270 /*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 */275 u64 disk_i_size;276 277 union {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 */283 u64 index_cnt;284 285 /*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 */290 u64 csum_bytes;291 };292 293 /* Cache the directory index number to speed the dir/file remove */294 u64 dir_index;295 296 /* 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 */301 u64 last_unlink_trans;302 303 union {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 */314 u64 last_reflink_trans;315 316 /*317 * In case this a root stub inode (BTRFS_INODE_ROOT_STUB flag set),318 * the ID of that root.319 */320 u64 ref_root_id;321 };322 323 /* Backwards incompatible flags, lower half of inode_item::flags */324 u32 flags;325 /* Read-only compatibility flags, upper half of inode_item::flags */326 u32 ro_flags;327 328 struct btrfs_block_rsv block_rsv;329 330 struct btrfs_delayed_node *delayed_node;331 332 /* File creation time. */333 u64 i_otime_sec;334 u32 i_otime_nsec;335 336 /* Hook into fs_info->delayed_iputs */337 struct list_head delayed_iput;338 339 struct rw_semaphore i_mmap_lock;340 struct inode vfs_inode;341};342 343static inline u64 btrfs_get_first_dir_index_to_log(const struct btrfs_inode *inode)344{345 return READ_ONCE(inode->first_dir_index_to_log);346}347 348static inline void btrfs_set_first_dir_index_to_log(struct btrfs_inode *inode,349 u64 index)350{351 WRITE_ONCE(inode->first_dir_index_to_log, index);352}353 354/* Type checked and const-preserving VFS inode -> btrfs inode. */355#define BTRFS_I(_inode) \356 _Generic(_inode, \357 struct inode *: container_of(_inode, struct btrfs_inode, vfs_inode), \358 const struct inode *: (const struct btrfs_inode *)container_of( \359 _inode, const struct btrfs_inode, vfs_inode))360 361static inline unsigned long btrfs_inode_hash(u64 objectid,362 const struct btrfs_root *root)363{364 u64 h = objectid ^ (root->root_key.objectid * GOLDEN_RATIO_PRIME);365 366#if BITS_PER_LONG == 32367 h = (h >> 32) ^ (h & 0xffffffff);368#endif369 370 return (unsigned long)h;371}372 373#if BITS_PER_LONG == 32374 375/*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{381 u64 ino = inode->objectid;382 383 if (test_bit(BTRFS_INODE_ROOT_STUB, &inode->runtime_flags))384 ino = inode->vfs_inode.i_ino;385 return ino;386}387 388#else389 390static inline u64 btrfs_ino(const struct btrfs_inode *inode)391{392 return inode->vfs_inode.i_ino;393}394 395#endif396 397static inline void btrfs_get_inode_key(const struct btrfs_inode *inode,398 struct btrfs_key *key)399{400 key->objectid = btrfs_ino(inode);401 key->type = BTRFS_INODE_ITEM_KEY;402 key->offset = 0;403}404 405static inline void btrfs_set_inode_number(struct btrfs_inode *inode, u64 ino)406{407#if BITS_PER_LONG == 32408 inode->objectid = ino;409#endif410 inode->vfs_inode.i_ino = ino;411}412 413static inline void btrfs_i_size_write(struct btrfs_inode *inode, u64 size)414{415 i_size_write(&inode->vfs_inode, size);416 inode->disk_i_size = size;417}418 419static inline bool btrfs_is_free_space_inode(const struct btrfs_inode *inode)420{421 return test_bit(BTRFS_INODE_FREE_SPACE_INODE, &inode->runtime_flags);422}423 424static inline bool is_data_inode(const struct btrfs_inode *inode)425{426 return btrfs_ino(inode) != BTRFS_BTREE_INODE_OBJECTID;427}428 429static inline void btrfs_mod_outstanding_extents(struct btrfs_inode *inode,430 int mod)431{432 lockdep_assert_held(&inode->lock);433 inode->outstanding_extents += mod;434 if (btrfs_is_free_space_inode(inode))435 return;436 trace_btrfs_inode_mod_outstanding_extents(inode->root, btrfs_ino(inode),437 mod, inode->outstanding_extents);438}439 440/*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{450 spin_lock(&inode->lock);451 inode->last_sub_trans = inode->root->log_transid;452 spin_unlock(&inode->lock);453}454 455/*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{464 set_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 */481 spin_lock(&inode->lock);482 if (inode->last_reflink_trans < inode->last_trans)483 inode->last_reflink_trans = inode->last_trans;484 spin_unlock(&inode->lock);485}486 487static inline bool btrfs_inode_in_log(struct btrfs_inode *inode, u64 generation)488{489 bool ret = false;490 491 spin_lock(&inode->lock);492 if (inode->logged_trans == generation &&493 inode->last_sub_trans <= inode->last_log_commit &&494 inode->last_sub_trans <= btrfs_get_root_last_log_commit(inode->root))495 ret = true;496 spin_unlock(&inode->lock);497 return ret;498}499 500/*501 * Check if the inode has flags compatible with compression502 */503static inline bool btrfs_inode_can_compress(const struct btrfs_inode *inode)504{505 if (inode->flags & BTRFS_INODE_NODATACOW ||506 inode->flags & BTRFS_INODE_NODATASUM)507 return false;508 return true;509}510 511static inline void btrfs_assert_inode_locked(struct btrfs_inode *inode)512{513 /* Immediately trigger a crash if the inode is not locked. */514 ASSERT(inode_is_locked(&inode->vfs_inode));515 /* Trigger a splat in dmesg if this task is not holding the lock. */516 lockdep_assert_held(&inode->vfs_inode.i_rwsem);517}518 519/* Array of bytes with variable length, hexadecimal format 0x1234 */520#define CSUM_FMT "0x%*phN"521#define CSUM_FMT_VALUE(size, bytes) size, bytes522 523int btrfs_check_sector_csum(struct btrfs_fs_info *fs_info, struct page *page,524 u32 pgoff, u8 *csum, const u8 * const csum_expected);525bool btrfs_data_csum_ok(struct btrfs_bio *bbio, struct btrfs_device *dev,526 u32 bio_offset, struct bio_vec *bv);527noinline int can_nocow_extent(struct inode *inode, u64 offset, u64 *len,528 struct btrfs_file_extent *file_extent,529 bool nowait, bool strict);530 531void btrfs_del_delalloc_inode(struct btrfs_inode *inode);532struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry);533int btrfs_set_inode_index(struct btrfs_inode *dir, u64 *index);534int btrfs_unlink_inode(struct btrfs_trans_handle *trans,535 struct btrfs_inode *dir, struct btrfs_inode *inode,536 const struct fscrypt_str *name);537int btrfs_add_link(struct btrfs_trans_handle *trans,538 struct btrfs_inode *parent_inode, struct btrfs_inode *inode,539 const struct fscrypt_str *name, int add_backref, u64 index);540int btrfs_delete_subvolume(struct btrfs_inode *dir, struct dentry *dentry);541int btrfs_truncate_block(struct btrfs_inode *inode, loff_t from, loff_t len,542 int front);543 544int btrfs_start_delalloc_snapshot(struct btrfs_root *root, bool in_reclaim_context);545int btrfs_start_delalloc_roots(struct btrfs_fs_info *fs_info, long nr,546 bool in_reclaim_context);547int btrfs_set_extent_delalloc(struct btrfs_inode *inode, u64 start, u64 end,548 unsigned int extra_bits,549 struct extent_state **cached_state);550 551struct btrfs_new_inode_args {552 /* Input */553 struct inode *dir;554 struct dentry *dentry;555 struct inode *inode;556 bool orphan;557 bool subvol;558 559 /* Output from btrfs_new_inode_prepare(), input to btrfs_create_new_inode(). */560 struct posix_acl *default_acl;561 struct posix_acl *acl;562 struct fscrypt_name fname;563};564 565int btrfs_new_inode_prepare(struct btrfs_new_inode_args *args,566 unsigned int *trans_num_items);567int btrfs_create_new_inode(struct btrfs_trans_handle *trans,568 struct btrfs_new_inode_args *args);569void btrfs_new_inode_args_destroy(struct btrfs_new_inode_args *args);570struct inode *btrfs_new_subvol_inode(struct mnt_idmap *idmap,571 struct inode *dir);572 void btrfs_set_delalloc_extent(struct btrfs_inode *inode, struct extent_state *state,573 u32 bits);574void btrfs_clear_delalloc_extent(struct btrfs_inode *inode,575 struct extent_state *state, u32 bits);576void btrfs_merge_delalloc_extent(struct btrfs_inode *inode, struct extent_state *new,577 struct extent_state *other);578void btrfs_split_delalloc_extent(struct btrfs_inode *inode,579 struct extent_state *orig, u64 split);580void btrfs_set_range_writeback(struct btrfs_inode *inode, u64 start, u64 end);581void btrfs_evict_inode(struct inode *inode);582struct inode *btrfs_alloc_inode(struct super_block *sb);583void btrfs_destroy_inode(struct inode *inode);584void btrfs_free_inode(struct inode *inode);585int btrfs_drop_inode(struct inode *inode);586int __init btrfs_init_cachep(void);587void __cold btrfs_destroy_cachep(void);588struct inode *btrfs_iget_path(u64 ino, struct btrfs_root *root,589 struct btrfs_path *path);590struct inode *btrfs_iget(u64 ino, struct btrfs_root *root);591struct extent_map *btrfs_get_extent(struct btrfs_inode *inode,592 struct folio *folio, u64 start, u64 len);593int btrfs_update_inode(struct btrfs_trans_handle *trans,594 struct btrfs_inode *inode);595int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,596 struct btrfs_inode *inode);597int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct btrfs_inode *inode);598int btrfs_orphan_cleanup(struct btrfs_root *root);599int btrfs_cont_expand(struct btrfs_inode *inode, loff_t oldsize, loff_t size);600void btrfs_add_delayed_iput(struct btrfs_inode *inode);601void btrfs_run_delayed_iputs(struct btrfs_fs_info *fs_info);602int btrfs_wait_on_delayed_iputs(struct btrfs_fs_info *fs_info);603int btrfs_prealloc_file_range(struct inode *inode, int mode,604 u64 start, u64 num_bytes, u64 min_size,605 loff_t actual_len, u64 *alloc_hint);606int btrfs_prealloc_file_range_trans(struct inode *inode,607 struct btrfs_trans_handle *trans, int mode,608 u64 start, u64 num_bytes, u64 min_size,609 loff_t actual_len, u64 *alloc_hint);610int btrfs_run_delalloc_range(struct btrfs_inode *inode, struct folio *locked_folio,611 u64 start, u64 end, struct writeback_control *wbc);612int btrfs_writepage_cow_fixup(struct folio *folio);613int btrfs_encoded_io_compression_from_extent(struct btrfs_fs_info *fs_info,614 int compress_type);615int btrfs_encoded_read_regular_fill_pages(struct btrfs_inode *inode,616 u64 file_offset, u64 disk_bytenr,617 u64 disk_io_size,618 struct page **pages);619ssize_t btrfs_encoded_read(struct kiocb *iocb, struct iov_iter *iter,620 struct btrfs_ioctl_encoded_io_args *encoded);621ssize_t btrfs_do_encoded_write(struct kiocb *iocb, struct iov_iter *from,622 const struct btrfs_ioctl_encoded_io_args *encoded);623 624struct btrfs_inode *btrfs_find_first_inode(struct btrfs_root *root, u64 min_ino);625 626extern const struct dentry_operations btrfs_dentry_operations;627 628/* Inode locking type flags, by default the exclusive lock is taken. */629enum btrfs_ilock_type {630 ENUM_BIT(BTRFS_ILOCK_SHARED),631 ENUM_BIT(BTRFS_ILOCK_TRY),632 ENUM_BIT(BTRFS_ILOCK_MMAP),633};634 635int btrfs_inode_lock(struct btrfs_inode *inode, unsigned int ilock_flags);636void btrfs_inode_unlock(struct btrfs_inode *inode, unsigned int ilock_flags);637void btrfs_update_inode_bytes(struct btrfs_inode *inode, const u64 add_bytes,638 const u64 del_bytes);639void btrfs_assert_inode_range_clean(struct btrfs_inode *inode, u64 start, u64 end);640u64 btrfs_get_extent_allocation_hint(struct btrfs_inode *inode, u64 start,641 u64 num_bytes);642struct extent_map *btrfs_create_io_em(struct btrfs_inode *inode, u64 start,643 const struct btrfs_file_extent *file_extent,644 int type);645 646#endif647