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"
13 * Kernel only inode definitions
19 struct xfs_inode_log_item
;
24 typedef struct xfs_inode
{
25 /* Inode linking and identification information. */
26 struct xfs_mount
*i_mount
; /* fs mount struct ptr */
27 struct xfs_dquot
*i_udquot
; /* user dquot */
28 struct xfs_dquot
*i_gdquot
; /* group dquot */
29 struct xfs_dquot
*i_pdquot
; /* project dquot */
31 /* Inode location stuff */
32 xfs_ino_t i_ino
; /* inode number (agno/agino)*/
33 struct xfs_imap i_imap
; /* location for xfs_imap() */
35 /* Extent information. */
36 struct xfs_ifork
*i_cowfp
; /* copy on write extents */
37 struct xfs_ifork i_df
; /* data fork */
38 struct xfs_ifork i_af
; /* attribute fork */
40 /* Transaction and locking information. */
41 struct xfs_inode_log_item
*i_itemp
; /* logging information */
42 mrlock_t i_lock
; /* inode lock */
43 atomic_t i_pincount
; /* inode pin count */
44 struct llist_node i_gclist
; /* deferred inactivation list */
47 * Bitsets of inode metadata that have been checked and/or are sick.
48 * Callers must hold i_flags_lock before accessing this field.
53 spinlock_t i_flags_lock
; /* inode i_flags lock */
54 /* Miscellaneous state. */
55 unsigned long i_flags
; /* see defined flags below */
56 uint64_t i_delayed_blks
; /* count of delay alloc blks */
57 xfs_fsize_t i_disk_size
; /* number of bytes in file */
58 xfs_rfsblock_t i_nblocks
; /* # of direct & btree blocks */
59 prid_t i_projid
; /* owner's project id */
60 xfs_extlen_t i_extsize
; /* basic/minimum extent size */
61 /* cowextsize is only used for v3 inodes, flushiter for v1/2 */
63 xfs_extlen_t i_cowextsize
; /* basic cow extent size */
64 uint16_t i_flushiter
; /* incremented on flush */
66 uint8_t i_forkoff
; /* attr fork offset >> 3 */
67 uint16_t i_diflags
; /* XFS_DIFLAG_... */
68 uint64_t i_diflags2
; /* XFS_DIFLAG2_... */
69 struct timespec64 i_crtime
; /* time created */
72 * Unlinked list pointers. These point to the next and previous inodes
73 * in the AGI unlinked bucket list, respectively. These fields can
74 * only be updated with the AGI locked.
76 * i_next_unlinked caches di_next_unlinked.
78 xfs_agino_t i_next_unlinked
;
81 * If the inode is not on an unlinked list, this field is zero. If the
82 * inode is the first element in an unlinked list, this field is
83 * NULLAGINO. Otherwise, i_prev_unlinked points to the previous inode
84 * in the unlinked list.
86 xfs_agino_t i_prev_unlinked
;
89 struct inode i_vnode
; /* embedded VFS inode */
91 /* pending io completions */
92 spinlock_t i_ioend_lock
;
93 struct work_struct i_ioend_work
;
94 struct list_head i_ioend_list
;
97 static inline bool xfs_inode_on_unlinked_list(const struct xfs_inode
*ip
)
99 return ip
->i_prev_unlinked
!= 0;
102 static inline bool xfs_inode_has_attr_fork(struct xfs_inode
*ip
)
104 return ip
->i_forkoff
> 0;
107 static inline struct xfs_ifork
*
109 struct xfs_inode
*ip
,
116 if (!xfs_inode_has_attr_fork(ip
))
127 static inline unsigned int xfs_inode_fork_boff(struct xfs_inode
*ip
)
129 return ip
->i_forkoff
<< 3;
132 static inline unsigned int xfs_inode_data_fork_size(struct xfs_inode
*ip
)
134 if (xfs_inode_has_attr_fork(ip
))
135 return xfs_inode_fork_boff(ip
);
137 return XFS_LITINO(ip
->i_mount
);
140 static inline unsigned int xfs_inode_attr_fork_size(struct xfs_inode
*ip
)
142 if (xfs_inode_has_attr_fork(ip
))
143 return XFS_LITINO(ip
->i_mount
) - xfs_inode_fork_boff(ip
);
147 static inline unsigned int
149 struct xfs_inode
*ip
,
154 return xfs_inode_data_fork_size(ip
);
156 return xfs_inode_attr_fork_size(ip
);
162 /* Convert from vfs inode to xfs inode */
163 static inline struct xfs_inode
*XFS_I(struct inode
*inode
)
165 return container_of(inode
, struct xfs_inode
, i_vnode
);
168 /* convert from xfs inode to vfs inode */
169 static inline struct inode
*VFS_I(struct xfs_inode
*ip
)
175 * For regular files we only update the on-disk filesize when actually
176 * writing data back to disk. Until then only the copy in the VFS inode
179 static inline xfs_fsize_t
XFS_ISIZE(struct xfs_inode
*ip
)
181 if (S_ISREG(VFS_I(ip
)->i_mode
))
182 return i_size_read(VFS_I(ip
));
183 return ip
->i_disk_size
;
187 * If this I/O goes past the on-disk inode size update it unless it would
188 * be past the current in-core inode size.
190 static inline xfs_fsize_t
191 xfs_new_eof(struct xfs_inode
*ip
, xfs_fsize_t new_size
)
193 xfs_fsize_t i_size
= i_size_read(VFS_I(ip
));
195 if (new_size
> i_size
|| new_size
< 0)
197 return new_size
> ip
->i_disk_size
? new_size
: 0;
201 * i_flags helper functions
204 __xfs_iflags_set(xfs_inode_t
*ip
, unsigned short flags
)
206 ip
->i_flags
|= flags
;
210 xfs_iflags_set(xfs_inode_t
*ip
, unsigned short flags
)
212 spin_lock(&ip
->i_flags_lock
);
213 __xfs_iflags_set(ip
, flags
);
214 spin_unlock(&ip
->i_flags_lock
);
218 xfs_iflags_clear(xfs_inode_t
*ip
, unsigned short flags
)
220 spin_lock(&ip
->i_flags_lock
);
221 ip
->i_flags
&= ~flags
;
222 spin_unlock(&ip
->i_flags_lock
);
226 __xfs_iflags_test(xfs_inode_t
*ip
, unsigned short flags
)
228 return (ip
->i_flags
& flags
);
232 xfs_iflags_test(xfs_inode_t
*ip
, unsigned short flags
)
235 spin_lock(&ip
->i_flags_lock
);
236 ret
= __xfs_iflags_test(ip
, flags
);
237 spin_unlock(&ip
->i_flags_lock
);
242 xfs_iflags_test_and_clear(xfs_inode_t
*ip
, unsigned short flags
)
246 spin_lock(&ip
->i_flags_lock
);
247 ret
= ip
->i_flags
& flags
;
249 ip
->i_flags
&= ~flags
;
250 spin_unlock(&ip
->i_flags_lock
);
255 xfs_iflags_test_and_set(xfs_inode_t
*ip
, unsigned short 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_get_initial_prid(struct xfs_inode
*dp
)
270 if (dp
->i_diflags
& XFS_DIFLAG_PROJINHERIT
)
273 return XFS_PROJID_DEFAULT
;
276 static inline bool xfs_is_reflink_inode(struct xfs_inode
*ip
)
278 return ip
->i_diflags2
& XFS_DIFLAG2_REFLINK
;
281 static inline bool xfs_is_metadata_inode(struct xfs_inode
*ip
)
283 struct xfs_mount
*mp
= ip
->i_mount
;
285 return ip
== mp
->m_rbmip
|| ip
== mp
->m_rsumip
||
286 xfs_is_quota_inode(&mp
->m_sb
, ip
->i_ino
);
290 * Check if an inode has any data in the COW fork. This might be often false
291 * even for inodes with the reflink flag when there is no pending COW operation.
293 static inline bool xfs_inode_has_cow_data(struct xfs_inode
*ip
)
295 return ip
->i_cowfp
&& ip
->i_cowfp
->if_bytes
;
298 static inline bool xfs_inode_has_bigtime(struct xfs_inode
*ip
)
300 return ip
->i_diflags2
& XFS_DIFLAG2_BIGTIME
;
303 static inline bool xfs_inode_has_large_extent_counts(struct xfs_inode
*ip
)
305 return ip
->i_diflags2
& XFS_DIFLAG2_NREXT64
;
309 * Return the buftarg used for data allocations on a given inode.
311 #define xfs_inode_buftarg(ip) \
312 (XFS_IS_REALTIME_INODE(ip) ? \
313 (ip)->i_mount->m_rtdev_targp : (ip)->i_mount->m_ddev_targp)
316 * In-core inode flags.
318 #define XFS_IRECLAIM (1 << 0) /* started reclaiming this inode */
319 #define XFS_ISTALE (1 << 1) /* inode has been staled */
320 #define XFS_IRECLAIMABLE (1 << 2) /* inode can be reclaimed */
321 #define XFS_INEW (1 << 3) /* inode has just been allocated */
322 #define XFS_IPRESERVE_DM_FIELDS (1 << 4) /* has legacy DMAPI fields set */
323 #define XFS_ITRUNCATED (1 << 5) /* truncated down so flush-on-close */
324 #define XFS_IDIRTY_RELEASE (1 << 6) /* dirty release already seen */
325 #define XFS_IFLUSHING (1 << 7) /* inode is being flushed */
326 #define __XFS_IPINNED_BIT 8 /* wakeup key for zero pin count */
327 #define XFS_IPINNED (1 << __XFS_IPINNED_BIT)
328 #define XFS_IEOFBLOCKS (1 << 9) /* has the preallocblocks tag set */
329 #define XFS_NEED_INACTIVE (1 << 10) /* see XFS_INACTIVATING below */
331 * If this unlinked inode is in the middle of recovery, don't let drop_inode
332 * truncate and free the inode. This can happen if we iget the inode during
333 * log recovery to replay a bmap operation on the inode.
335 #define XFS_IRECOVERY (1 << 11)
336 #define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */
339 * If we need to update on-disk metadata before this IRECLAIMABLE inode can be
340 * freed, then NEED_INACTIVE will be set. Once we start the updates, the
341 * INACTIVATING bit will be set to keep iget away from this inode. After the
342 * inactivation completes, both flags will be cleared and the inode is a
343 * plain old IRECLAIMABLE inode.
345 #define XFS_INACTIVATING (1 << 13)
347 /* Quotacheck is running but inode has not been added to quota counts. */
348 #define XFS_IQUOTAUNCHECKED (1 << 14)
350 /* All inode state flags related to inode reclaim. */
351 #define XFS_ALL_IRECLAIM_FLAGS (XFS_IRECLAIMABLE | \
353 XFS_NEED_INACTIVE | \
357 * Per-lifetime flags need to be reset when re-using a reclaimable inode during
358 * inode lookup. This prevents unintended behaviour on the new inode from
361 #define XFS_IRECLAIM_RESET_FLAGS \
362 (XFS_IRECLAIMABLE | XFS_IRECLAIM | \
363 XFS_IDIRTY_RELEASE | XFS_ITRUNCATED | XFS_NEED_INACTIVE | \
364 XFS_INACTIVATING | XFS_IQUOTAUNCHECKED)
367 * Flags for inode locking.
368 * Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
369 * 1<<16 - 1<<32-1 -- lockdep annotation (integers)
371 #define XFS_IOLOCK_EXCL (1u << 0)
372 #define XFS_IOLOCK_SHARED (1u << 1)
373 #define XFS_ILOCK_EXCL (1u << 2)
374 #define XFS_ILOCK_SHARED (1u << 3)
375 #define XFS_MMAPLOCK_EXCL (1u << 4)
376 #define XFS_MMAPLOCK_SHARED (1u << 5)
378 #define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
379 | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
380 | XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
382 #define XFS_LOCK_FLAGS \
383 { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
384 { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
385 { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
386 { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
387 { XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \
388 { XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" }
392 * Flags for lockdep annotations.
394 * XFS_LOCK_PARENT - for directory operations that require locking a
395 * parent directory inode and a child entry inode. IOLOCK requires nesting,
396 * MMAPLOCK does not support this class, ILOCK requires a single subclass
397 * to differentiate parent from child.
399 * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
400 * inodes do not participate in the normal lock order, and thus have their
403 * XFS_LOCK_INUMORDER - for locking several inodes at the some time
404 * with xfs_lock_inodes(). This flag is used as the starting subclass
405 * and each subsequent lock acquired will increment the subclass by one.
406 * However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
407 * limited to the subclasses we can represent via nesting. We need at least
408 * 5 inodes nest depth for the ILOCK through rename, and we also have to support
409 * XFS_ILOCK_PARENT, which gives 6 subclasses. Then we have XFS_ILOCK_RTBITMAP
410 * and XFS_ILOCK_RTSUM, which are another 2 unique subclasses, so that's all
411 * 8 subclasses supported by lockdep.
413 * This also means we have to number the sub-classes in the lowest bits of
414 * the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
415 * mask and we can't use bit-masking to build the subclasses. What a mess.
420 * 16-19 XFS_IOLOCK_SHIFT dependencies
421 * 20-23 XFS_MMAPLOCK_SHIFT dependencies
422 * 24-31 XFS_ILOCK_SHIFT dependencies
435 * 0-4 subclass values
436 * 5 PARENT subclass (not nestable)
437 * 6 RTBITMAP subclass (not nestable)
438 * 7 RTSUM subclass (not nestable)
441 #define XFS_IOLOCK_SHIFT 16
442 #define XFS_IOLOCK_MAX_SUBCLASS 3
443 #define XFS_IOLOCK_DEP_MASK 0x000f0000u
445 #define XFS_MMAPLOCK_SHIFT 20
446 #define XFS_MMAPLOCK_NUMORDER 0
447 #define XFS_MMAPLOCK_MAX_SUBCLASS 3
448 #define XFS_MMAPLOCK_DEP_MASK 0x00f00000u
450 #define XFS_ILOCK_SHIFT 24
451 #define XFS_ILOCK_PARENT_VAL 5u
452 #define XFS_ILOCK_MAX_SUBCLASS (XFS_ILOCK_PARENT_VAL - 1)
453 #define XFS_ILOCK_RTBITMAP_VAL 6u
454 #define XFS_ILOCK_RTSUM_VAL 7u
455 #define XFS_ILOCK_DEP_MASK 0xff000000u
456 #define XFS_ILOCK_PARENT (XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
457 #define XFS_ILOCK_RTBITMAP (XFS_ILOCK_RTBITMAP_VAL << XFS_ILOCK_SHIFT)
458 #define XFS_ILOCK_RTSUM (XFS_ILOCK_RTSUM_VAL << XFS_ILOCK_SHIFT)
460 #define XFS_LOCK_SUBCLASS_MASK (XFS_IOLOCK_DEP_MASK | \
461 XFS_MMAPLOCK_DEP_MASK | \
464 #define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \
466 #define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \
467 >> XFS_MMAPLOCK_SHIFT)
468 #define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \
472 * Layouts are broken in the BREAK_WRITE case to ensure that
473 * layout-holders do not collide with local writes. Additionally,
474 * layouts are broken in the BREAK_UNMAP case to make sure the
475 * layout-holder has a consistent view of the file's extent map. While
476 * BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
477 * BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
480 enum layout_break_reason
{
486 * For multiple groups support: if S_ISGID bit is set in the parent
487 * directory, group of new file is set to that of the parent, and
488 * new subdirectory gets S_ISGID bit from parent.
490 #define XFS_INHERIT_GID(pip) \
491 (xfs_has_grpid((pip)->i_mount) || (VFS_I(pip)->i_mode & S_ISGID))
493 int xfs_release(struct xfs_inode
*ip
);
494 int xfs_inactive(struct xfs_inode
*ip
);
495 int xfs_lookup(struct xfs_inode
*dp
, const struct xfs_name
*name
,
496 struct xfs_inode
**ipp
, struct xfs_name
*ci_name
);
497 int xfs_create(struct mnt_idmap
*idmap
,
498 struct xfs_inode
*dp
, struct xfs_name
*name
,
499 umode_t mode
, dev_t rdev
, bool need_xattr
,
500 struct xfs_inode
**ipp
);
501 int xfs_create_tmpfile(struct mnt_idmap
*idmap
,
502 struct xfs_inode
*dp
, umode_t mode
,
503 struct xfs_inode
**ipp
);
504 int xfs_remove(struct xfs_inode
*dp
, struct xfs_name
*name
,
505 struct xfs_inode
*ip
);
506 int xfs_link(struct xfs_inode
*tdp
, struct xfs_inode
*sip
,
507 struct xfs_name
*target_name
);
508 int xfs_rename(struct mnt_idmap
*idmap
,
509 struct xfs_inode
*src_dp
, struct xfs_name
*src_name
,
510 struct xfs_inode
*src_ip
, struct xfs_inode
*target_dp
,
511 struct xfs_name
*target_name
,
512 struct xfs_inode
*target_ip
, unsigned int flags
);
514 void xfs_ilock(xfs_inode_t
*, uint
);
515 int xfs_ilock_nowait(xfs_inode_t
*, uint
);
516 void xfs_iunlock(xfs_inode_t
*, uint
);
517 void xfs_ilock_demote(xfs_inode_t
*, uint
);
518 bool xfs_isilocked(struct xfs_inode
*, uint
);
519 uint
xfs_ilock_data_map_shared(struct xfs_inode
*);
520 uint
xfs_ilock_attr_map_shared(struct xfs_inode
*);
522 uint
xfs_ip2xflags(struct xfs_inode
*);
523 int xfs_ifree(struct xfs_trans
*, struct xfs_inode
*);
524 int xfs_itruncate_extents_flags(struct xfs_trans
**,
525 struct xfs_inode
*, int, xfs_fsize_t
, int);
526 void xfs_iext_realloc(xfs_inode_t
*, int, int);
528 int xfs_log_force_inode(struct xfs_inode
*ip
);
529 void xfs_iunpin_wait(xfs_inode_t
*);
530 #define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
532 int xfs_iflush_cluster(struct xfs_buf
*);
533 void xfs_lock_two_inodes(struct xfs_inode
*ip0
, uint ip0_mode
,
534 struct xfs_inode
*ip1
, uint ip1_mode
);
536 xfs_extlen_t
xfs_get_extsz_hint(struct xfs_inode
*ip
);
537 xfs_extlen_t
xfs_get_cowextsz_hint(struct xfs_inode
*ip
);
539 int xfs_init_new_inode(struct mnt_idmap
*idmap
, struct xfs_trans
*tp
,
540 struct xfs_inode
*pip
, xfs_ino_t ino
, umode_t mode
,
541 xfs_nlink_t nlink
, dev_t rdev
, prid_t prid
, bool init_xattrs
,
542 struct xfs_inode
**ipp
);
545 xfs_itruncate_extents(
546 struct xfs_trans
**tpp
,
547 struct xfs_inode
*ip
,
549 xfs_fsize_t new_size
)
551 return xfs_itruncate_extents_flags(tpp
, ip
, whichfork
, new_size
, 0);
554 /* from xfs_file.c */
555 int xfs_break_dax_layouts(struct inode
*inode
, bool *retry
);
556 int xfs_break_layouts(struct inode
*inode
, uint
*iolock
,
557 enum layout_break_reason reason
);
559 /* from xfs_iops.c */
560 extern void xfs_setup_inode(struct xfs_inode
*ip
);
561 extern void xfs_setup_iops(struct xfs_inode
*ip
);
562 extern void xfs_diflags_to_iflags(struct xfs_inode
*ip
, bool init
);
565 * When setting up a newly allocated inode, we need to call
566 * xfs_finish_inode_setup() once the inode is fully instantiated at
567 * the VFS level to prevent the rest of the world seeing the inode
568 * before we've completed instantiation. Otherwise we can do it
569 * the moment the inode lookup is complete.
571 static inline void xfs_finish_inode_setup(struct xfs_inode
*ip
)
573 xfs_iflags_clear(ip
, XFS_INEW
);
575 unlock_new_inode(VFS_I(ip
));
578 static inline void xfs_setup_existing_inode(struct xfs_inode
*ip
)
582 xfs_finish_inode_setup(ip
);
585 void xfs_irele(struct xfs_inode
*ip
);
587 extern struct kmem_cache
*xfs_inode_cache
;
589 /* The default CoW extent size hint. */
590 #define XFS_DEFAULT_COWEXTSZ_HINT 32
592 bool xfs_inode_needs_inactive(struct xfs_inode
*ip
);
594 void xfs_end_io(struct work_struct
*work
);
596 int xfs_ilock2_io_mmap(struct xfs_inode
*ip1
, struct xfs_inode
*ip2
);
597 void xfs_iunlock2_io_mmap(struct xfs_inode
*ip1
, struct xfs_inode
*ip2
);
600 xfs_inode_unlinked_incomplete(
601 struct xfs_inode
*ip
)
603 return VFS_I(ip
)->i_nlink
== 0 && !xfs_inode_on_unlinked_list(ip
);
605 int xfs_inode_reload_unlinked_bucket(struct xfs_trans
*tp
, struct xfs_inode
*ip
);
606 int xfs_inode_reload_unlinked(struct xfs_inode
*ip
);
608 #endif /* __XFS_INODE_H__ */