1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
6 #ifndef __XFS_INODE_H__
7 #define __XFS_INODE_H__
9 #include "xfs_inode_buf.h"
10 #include "xfs_inode_fork.h"
11 #include "xfs_inode_util.h"
14 * Kernel only inode definitions
20 struct xfs_inode_log_item;
25 typedef struct xfs_inode {
26 /* Inode linking and identification information. */
27 struct xfs_mount *i_mount; /* fs mount struct ptr */
28 struct xfs_dquot *i_udquot; /* user dquot */
29 struct xfs_dquot *i_gdquot; /* group dquot */
30 struct xfs_dquot *i_pdquot; /* project dquot */
32 /* Inode location stuff */
33 xfs_ino_t i_ino; /* inode number (agno/agino)*/
34 struct xfs_imap i_imap; /* location for xfs_imap() */
36 /* Extent information. */
37 struct xfs_ifork *i_cowfp; /* copy on write extents */
38 struct xfs_ifork i_df; /* data fork */
39 struct xfs_ifork i_af; /* attribute fork */
41 /* Transaction and locking information. */
42 struct xfs_inode_log_item *i_itemp; /* logging information */
43 struct rw_semaphore i_lock; /* inode lock */
44 atomic_t i_pincount; /* inode pin count */
45 struct llist_node i_gclist; /* deferred inactivation list */
48 * Bitsets of inode metadata that have been checked and/or are sick.
49 * Callers must hold i_flags_lock before accessing this field.
54 spinlock_t i_flags_lock; /* inode i_flags lock */
55 /* Miscellaneous state. */
56 unsigned long i_flags; /* see defined flags below */
57 uint64_t i_delayed_blks; /* count of delay alloc blks */
58 xfs_fsize_t i_disk_size; /* number of bytes in file */
59 xfs_rfsblock_t i_nblocks; /* # of direct & btree blocks */
60 prid_t i_projid; /* owner's project id */
61 xfs_extlen_t i_extsize; /* basic/minimum extent size */
63 * i_used_blocks is used for zoned rtrmap inodes,
64 * i_cowextsize is used for other v3 inodes,
65 * i_flushiter for v1/2 inodes
68 uint32_t i_used_blocks; /* used blocks in RTG */
69 xfs_extlen_t i_cowextsize; /* basic cow extent size */
70 uint16_t i_flushiter; /* incremented on flush */
72 uint8_t i_forkoff; /* attr fork offset >> 3 */
73 enum xfs_metafile_type i_metatype; /* XFS_METAFILE_* */
74 uint16_t i_diflags; /* XFS_DIFLAG_... */
75 uint64_t i_diflags2; /* XFS_DIFLAG2_... */
76 struct timespec64 i_crtime; /* time created */
79 * Unlinked list pointers. These point to the next and previous inodes
80 * in the AGI unlinked bucket list, respectively. These fields can
81 * only be updated with the AGI locked.
83 * i_next_unlinked caches di_next_unlinked.
85 xfs_agino_t i_next_unlinked;
88 * If the inode is not on an unlinked list, this field is zero. If the
89 * inode is the first element in an unlinked list, this field is
90 * NULLAGINO. Otherwise, i_prev_unlinked points to the previous inode
91 * in the unlinked list.
93 xfs_agino_t i_prev_unlinked;
96 struct inode i_vnode; /* embedded VFS inode */
98 /* pending io completions */
99 spinlock_t i_ioend_lock;
100 struct work_struct i_ioend_work;
101 struct list_head i_ioend_list;
104 static inline bool xfs_inode_on_unlinked_list(const struct xfs_inode *ip)
106 return ip->i_prev_unlinked != 0;
109 static inline bool xfs_inode_has_attr_fork(const struct xfs_inode *ip)
111 return ip->i_forkoff > 0;
114 static inline struct xfs_ifork *
116 struct xfs_inode *ip,
123 if (!xfs_inode_has_attr_fork(ip))
134 static inline unsigned int xfs_inode_fork_boff(struct xfs_inode *ip)
136 return ip->i_forkoff << 3;
139 static inline unsigned int xfs_inode_data_fork_size(struct xfs_inode *ip)
141 if (xfs_inode_has_attr_fork(ip))
142 return xfs_inode_fork_boff(ip);
144 return XFS_LITINO(ip->i_mount);
147 static inline unsigned int xfs_inode_attr_fork_size(struct xfs_inode *ip)
149 if (xfs_inode_has_attr_fork(ip))
150 return XFS_LITINO(ip->i_mount) - xfs_inode_fork_boff(ip);
154 static inline unsigned int
156 struct xfs_inode *ip,
161 return xfs_inode_data_fork_size(ip);
163 return xfs_inode_attr_fork_size(ip);
169 /* Convert from vfs inode to xfs inode */
170 static inline struct xfs_inode *XFS_I(struct inode *inode)
172 return container_of(inode, struct xfs_inode, i_vnode);
175 /* convert from xfs inode to vfs inode */
176 static inline struct inode *VFS_I(struct xfs_inode *ip)
181 /* convert from const xfs inode to const vfs inode */
182 static inline const struct inode *VFS_IC(const struct xfs_inode *ip)
188 * For regular files we only update the on-disk filesize when actually
189 * writing data back to disk. Until then only the copy in the VFS inode
192 static inline xfs_fsize_t XFS_ISIZE(struct xfs_inode *ip)
194 if (S_ISREG(VFS_I(ip)->i_mode))
195 return i_size_read(VFS_I(ip));
196 return ip->i_disk_size;
200 * If this I/O goes past the on-disk inode size update it unless it would
201 * be past the current in-core inode size.
203 static inline xfs_fsize_t
204 xfs_new_eof(struct xfs_inode *ip, xfs_fsize_t new_size)
206 xfs_fsize_t i_size = i_size_read(VFS_I(ip));
208 if (new_size > i_size || new_size < 0)
210 return new_size > ip->i_disk_size ? new_size : 0;
214 * i_flags helper functions
217 __xfs_iflags_set(xfs_inode_t *ip, unsigned long flags)
219 ip->i_flags |= flags;
223 xfs_iflags_set(xfs_inode_t *ip, unsigned long flags)
225 spin_lock(&ip->i_flags_lock);
226 __xfs_iflags_set(ip, flags);
227 spin_unlock(&ip->i_flags_lock);
231 xfs_iflags_clear(xfs_inode_t *ip, unsigned long flags)
233 spin_lock(&ip->i_flags_lock);
234 ip->i_flags &= ~flags;
235 spin_unlock(&ip->i_flags_lock);
239 __xfs_iflags_test(const struct xfs_inode *ip, unsigned long flags)
241 return (ip->i_flags & flags);
245 xfs_iflags_test(xfs_inode_t *ip, unsigned long flags)
248 spin_lock(&ip->i_flags_lock);
249 ret = __xfs_iflags_test(ip, flags);
250 spin_unlock(&ip->i_flags_lock);
255 xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned long flags)
259 spin_lock(&ip->i_flags_lock);
260 ret = ip->i_flags & flags;
262 ip->i_flags &= ~flags;
263 spin_unlock(&ip->i_flags_lock);
268 xfs_iflags_test_and_set(xfs_inode_t *ip, unsigned long flags)
272 spin_lock(&ip->i_flags_lock);
273 ret = ip->i_flags & flags;
275 ip->i_flags |= flags;
276 spin_unlock(&ip->i_flags_lock);
280 static inline bool xfs_is_reflink_inode(const struct xfs_inode *ip)
282 return ip->i_diflags2 & XFS_DIFLAG2_REFLINK;
285 static inline bool xfs_is_metadir_inode(const struct xfs_inode *ip)
287 return ip->i_diflags2 & XFS_DIFLAG2_METADATA;
290 static inline bool xfs_is_internal_inode(const struct xfs_inode *ip)
292 struct xfs_mount *mp = ip->i_mount;
294 /* Any file in the metadata directory tree is a metadata inode. */
295 if (xfs_has_metadir(mp))
296 return xfs_is_metadir_inode(ip);
299 * Before metadata directories, the only metadata inodes were the
300 * three quota files, the realtime bitmap, and the realtime summary.
302 return ip->i_ino == mp->m_sb.sb_rbmino ||
303 ip->i_ino == mp->m_sb.sb_rsumino ||
304 xfs_is_quota_inode(&mp->m_sb, ip->i_ino);
307 static inline bool xfs_is_zoned_inode(const struct xfs_inode *ip)
309 return xfs_has_zoned(ip->i_mount) && XFS_IS_REALTIME_INODE(ip);
312 bool xfs_is_always_cow_inode(const struct xfs_inode *ip);
314 static inline bool xfs_is_cow_inode(const struct xfs_inode *ip)
316 return xfs_is_reflink_inode(ip) || xfs_is_always_cow_inode(ip);
319 static inline bool xfs_inode_has_filedata(const struct xfs_inode *ip)
321 return ip->i_df.if_nextents > 0 || ip->i_delayed_blks > 0;
325 * Check if an inode has any data in the COW fork. This might be often false
326 * even for inodes with the reflink flag when there is no pending COW operation.
328 static inline bool xfs_inode_has_cow_data(const struct xfs_inode *ip)
330 return ip->i_cowfp && ip->i_cowfp->if_bytes;
333 static inline bool xfs_inode_has_bigtime(const struct xfs_inode *ip)
335 return ip->i_diflags2 & XFS_DIFLAG2_BIGTIME;
338 static inline bool xfs_inode_has_large_extent_counts(const struct xfs_inode *ip)
340 return ip->i_diflags2 & XFS_DIFLAG2_NREXT64;
344 * Decide if this file is a realtime file whose data allocation unit is larger
345 * than a single filesystem block.
347 static inline bool xfs_inode_has_bigrtalloc(const struct xfs_inode *ip)
349 return XFS_IS_REALTIME_INODE(ip) && ip->i_mount->m_sb.sb_rextsize > 1;
353 * Return the buftarg used for data allocations on a given inode.
355 #define xfs_inode_buftarg(ip) \
356 (XFS_IS_REALTIME_INODE(ip) ? \
357 (ip)->i_mount->m_rtdev_targp : (ip)->i_mount->m_ddev_targp)
359 static inline bool xfs_inode_can_hw_atomic_write(const struct xfs_inode *ip)
361 return xfs_inode_buftarg(ip)->bt_bdev_awu_max > 0;
365 * In-core inode flags.
367 #define XFS_IRECLAIM (1 << 0) /* started reclaiming this inode */
368 #define XFS_ISTALE (1 << 1) /* inode has been staled */
369 #define XFS_IRECLAIMABLE (1 << 2) /* inode can be reclaimed */
370 #define XFS_INEW (1 << 3) /* inode has just been allocated */
371 #define XFS_IPRESERVE_DM_FIELDS (1 << 4) /* has legacy DMAPI fields set */
372 #define XFS_ITRUNCATED (1 << 5) /* truncated down so flush-on-close */
373 #define XFS_EOFBLOCKS_RELEASED (1 << 6) /* eofblocks were freed in ->release */
374 #define XFS_IFLUSHING (1 << 7) /* inode is being flushed */
375 #define __XFS_IPINNED_BIT 8 /* wakeup key for zero pin count */
376 #define XFS_IPINNED (1 << __XFS_IPINNED_BIT)
377 #define XFS_IEOFBLOCKS (1 << 9) /* has the preallocblocks tag set */
378 #define XFS_NEED_INACTIVE (1 << 10) /* see XFS_INACTIVATING below */
380 * If this unlinked inode is in the middle of recovery, don't let drop_inode
381 * truncate and free the inode. This can happen if we iget the inode during
382 * log recovery to replay a bmap operation on the inode.
384 #define XFS_IRECOVERY (1 << 11)
385 #define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */
388 * If we need to update on-disk metadata before this IRECLAIMABLE inode can be
389 * freed, then NEED_INACTIVE will be set. Once we start the updates, the
390 * INACTIVATING bit will be set to keep iget away from this inode. After the
391 * inactivation completes, both flags will be cleared and the inode is a
392 * plain old IRECLAIMABLE inode.
394 #define XFS_INACTIVATING (1 << 13)
396 /* Quotacheck is running but inode has not been added to quota counts. */
397 #define XFS_IQUOTAUNCHECKED (1 << 14)
400 * Remap in progress. Callers that wish to update file data while
401 * holding a shared IOLOCK or MMAPLOCK must drop the lock and retake
402 * the lock in exclusive mode. Relocking the file will block until
403 * IREMAPPING is cleared.
405 #define XFS_IREMAPPING (1U << 15)
407 /* All inode state flags related to inode reclaim. */
408 #define XFS_ALL_IRECLAIM_FLAGS (XFS_IRECLAIMABLE | \
410 XFS_NEED_INACTIVE | \
414 * Per-lifetime flags need to be reset when re-using a reclaimable inode during
415 * inode lookup. This prevents unintended behaviour on the new inode from
418 #define XFS_IRECLAIM_RESET_FLAGS \
419 (XFS_IRECLAIMABLE | XFS_IRECLAIM | \
420 XFS_EOFBLOCKS_RELEASED | XFS_ITRUNCATED | XFS_NEED_INACTIVE | \
421 XFS_INACTIVATING | XFS_IQUOTAUNCHECKED)
424 * Flags for inode locking.
425 * Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
426 * 1<<16 - 1<<32-1 -- lockdep annotation (integers)
428 #define XFS_IOLOCK_EXCL (1u << 0)
429 #define XFS_IOLOCK_SHARED (1u << 1)
430 #define XFS_ILOCK_EXCL (1u << 2)
431 #define XFS_ILOCK_SHARED (1u << 3)
432 #define XFS_MMAPLOCK_EXCL (1u << 4)
433 #define XFS_MMAPLOCK_SHARED (1u << 5)
435 #define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
436 | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
437 | XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
439 #define XFS_LOCK_FLAGS \
440 { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
441 { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
442 { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
443 { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
444 { XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \
445 { XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" }
449 * Flags for lockdep annotations.
451 * XFS_LOCK_PARENT - for directory operations that require locking a
452 * parent directory inode and a child entry inode. IOLOCK requires nesting,
453 * MMAPLOCK does not support this class, ILOCK requires a single subclass
454 * to differentiate parent from child.
456 * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
457 * inodes do not participate in the normal lock order, and thus have their
460 * XFS_LOCK_INUMORDER - for locking several inodes at the some time
461 * with xfs_lock_inodes(). This flag is used as the starting subclass
462 * and each subsequent lock acquired will increment the subclass by one.
463 * However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
464 * limited to the subclasses we can represent via nesting. We need at least
465 * 5 inodes nest depth for the ILOCK through rename, and we also have to support
466 * XFS_ILOCK_PARENT, which gives 6 subclasses. That's 6 of the 8 subclasses
467 * supported by lockdep.
469 * This also means we have to number the sub-classes in the lowest bits of
470 * the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
471 * mask and we can't use bit-masking to build the subclasses. What a mess.
476 * 16-19 XFS_IOLOCK_SHIFT dependencies
477 * 20-23 XFS_MMAPLOCK_SHIFT dependencies
478 * 24-31 XFS_ILOCK_SHIFT dependencies
491 * 0-4 subclass values
492 * 5 PARENT subclass (not nestable)
497 #define XFS_IOLOCK_SHIFT 16
498 #define XFS_IOLOCK_MAX_SUBCLASS 3
499 #define XFS_IOLOCK_DEP_MASK 0x000f0000u
501 #define XFS_MMAPLOCK_SHIFT 20
502 #define XFS_MMAPLOCK_NUMORDER 0
503 #define XFS_MMAPLOCK_MAX_SUBCLASS 3
504 #define XFS_MMAPLOCK_DEP_MASK 0x00f00000u
506 #define XFS_ILOCK_SHIFT 24
507 #define XFS_ILOCK_PARENT_VAL 5u
508 #define XFS_ILOCK_MAX_SUBCLASS (XFS_ILOCK_PARENT_VAL - 1)
509 #define XFS_ILOCK_DEP_MASK 0xff000000u
510 #define XFS_ILOCK_PARENT (XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
512 #define XFS_LOCK_SUBCLASS_MASK (XFS_IOLOCK_DEP_MASK | \
513 XFS_MMAPLOCK_DEP_MASK | \
516 #define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \
518 #define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \
519 >> XFS_MMAPLOCK_SHIFT)
520 #define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \
524 * Layouts are broken in the BREAK_WRITE case to ensure that
525 * layout-holders do not collide with local writes. Additionally,
526 * layouts are broken in the BREAK_UNMAP case to make sure the
527 * layout-holder has a consistent view of the file's extent map. While
528 * BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
529 * BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
532 enum layout_break_reason {
538 * For multiple groups support: if S_ISGID bit is set in the parent
539 * directory, group of new file is set to that of the parent, and
540 * new subdirectory gets S_ISGID bit from parent.
542 #define XFS_INHERIT_GID(pip) \
543 (xfs_has_grpid((pip)->i_mount) || (VFS_I(pip)->i_mode & S_ISGID))
545 int xfs_inactive(struct xfs_inode *ip);
546 int xfs_lookup(struct xfs_inode *dp, const struct xfs_name *name,
547 struct xfs_inode **ipp, struct xfs_name *ci_name);
548 int xfs_create(const struct xfs_icreate_args *iargs,
549 struct xfs_name *name, struct xfs_inode **ipp);
550 int xfs_create_tmpfile(const struct xfs_icreate_args *iargs,
551 struct xfs_inode **ipp);
552 int xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
553 struct xfs_inode *ip);
554 int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
555 struct xfs_name *target_name);
556 int xfs_rename(struct mnt_idmap *idmap,
557 struct xfs_inode *src_dp, struct xfs_name *src_name,
558 struct xfs_inode *src_ip, struct xfs_inode *target_dp,
559 struct xfs_name *target_name,
560 struct xfs_inode *target_ip, unsigned int flags);
562 void xfs_ilock(xfs_inode_t *, uint);
563 int xfs_ilock_nowait(xfs_inode_t *, uint);
564 void xfs_iunlock(xfs_inode_t *, uint);
565 void xfs_ilock_demote(xfs_inode_t *, uint);
566 void xfs_assert_ilocked(struct xfs_inode *, uint);
567 uint xfs_ilock_data_map_shared(struct xfs_inode *);
568 uint xfs_ilock_attr_map_shared(struct xfs_inode *);
570 int xfs_ifree(struct xfs_trans *, struct xfs_inode *);
571 int xfs_itruncate_extents_flags(struct xfs_trans **,
572 struct xfs_inode *, int, xfs_fsize_t, int);
573 void xfs_iext_realloc(xfs_inode_t *, int, int);
575 int xfs_log_force_inode(struct xfs_inode *ip);
576 void xfs_iunpin_wait(xfs_inode_t *);
577 #define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
579 int xfs_iflush_cluster(struct xfs_buf *);
580 void xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
581 struct xfs_inode *ip1, uint ip1_mode);
583 int xfs_icreate(struct xfs_trans *tp, xfs_ino_t ino,
584 const struct xfs_icreate_args *args, struct xfs_inode **ipp);
587 xfs_itruncate_extents(
588 struct xfs_trans **tpp,
589 struct xfs_inode *ip,
591 xfs_fsize_t new_size)
593 return xfs_itruncate_extents_flags(tpp, ip, whichfork, new_size, 0);
596 int xfs_break_dax_layouts(struct inode *inode);
597 int xfs_break_layouts(struct inode *inode, uint *iolock,
598 enum layout_break_reason reason);
600 static inline void xfs_update_stable_writes(struct xfs_inode *ip)
602 if (bdev_stable_writes(xfs_inode_buftarg(ip)->bt_bdev))
603 mapping_set_stable_writes(VFS_I(ip)->i_mapping);
605 mapping_clear_stable_writes(VFS_I(ip)->i_mapping);
609 * When setting up a newly allocated inode, we need to call
610 * xfs_finish_inode_setup() once the inode is fully instantiated at
611 * the VFS level to prevent the rest of the world seeing the inode
612 * before we've completed instantiation. Otherwise we can do it
613 * the moment the inode lookup is complete.
615 static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
617 xfs_iflags_clear(ip, XFS_INEW);
619 unlock_new_inode(VFS_I(ip));
622 static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
626 xfs_finish_inode_setup(ip);
629 void xfs_irele(struct xfs_inode *ip);
631 extern struct kmem_cache *xfs_inode_cache;
633 /* The default CoW extent size hint. */
634 #define XFS_DEFAULT_COWEXTSZ_HINT 32
636 bool xfs_inode_needs_inactive(struct xfs_inode *ip);
638 struct xfs_inode *xfs_iunlink_lookup(struct xfs_perag *pag, xfs_agino_t agino);
639 int xfs_iunlink_reload_next(struct xfs_trans *tp, struct xfs_buf *agibp,
640 xfs_agino_t prev_agino, xfs_agino_t next_agino);
642 void xfs_end_io(struct work_struct *work);
644 int xfs_ilock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
645 void xfs_iunlock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
646 void xfs_iunlock2_remapping(struct xfs_inode *ip1, struct xfs_inode *ip2);
647 void xfs_lock_inodes(struct xfs_inode **ips, int inodes, uint lock_mode);
648 void xfs_sort_inodes(struct xfs_inode **i_tab, unsigned int num_inodes);
651 xfs_inode_unlinked_incomplete(
652 const struct xfs_inode *ip)
654 return VFS_IC(ip)->i_nlink == 0 && !xfs_inode_on_unlinked_list(ip);
656 int xfs_inode_reload_unlinked_bucket(struct xfs_trans *tp, struct xfs_inode *ip);
657 int xfs_inode_reload_unlinked(struct xfs_inode *ip);
659 bool xfs_ifork_zapped(const struct xfs_inode *ip, int whichfork);
660 void xfs_inode_count_blocks(struct xfs_trans *tp, struct xfs_inode *ip,
661 xfs_filblks_t *dblocks, xfs_filblks_t *rblocks);
662 unsigned int xfs_inode_alloc_unitsize(struct xfs_inode *ip);
664 int xfs_icreate_dqalloc(const struct xfs_icreate_args *args,
665 struct xfs_dquot **udqpp, struct xfs_dquot **gdqpp,
666 struct xfs_dquot **pdqpp);
668 #endif /* __XFS_INODE_H__ */