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_afp; /* attribute fork pointer */
37 struct xfs_ifork *i_cowfp; /* copy on write extents */
38 struct xfs_ifork i_df; /* data 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 struct inode i_vnode; /* embedded VFS inode */
74 /* pending io completions */
75 spinlock_t i_ioend_lock;
76 struct work_struct i_ioend_work;
77 struct list_head i_ioend_list;
80 /* Convert from vfs inode to xfs inode */
81 static inline struct xfs_inode *XFS_I(struct inode *inode)
83 return container_of(inode, struct xfs_inode, i_vnode);
86 /* convert from xfs inode to vfs inode */
87 static inline struct inode *VFS_I(struct xfs_inode *ip)
93 * For regular files we only update the on-disk filesize when actually
94 * writing data back to disk. Until then only the copy in the VFS inode
97 static inline xfs_fsize_t XFS_ISIZE(struct xfs_inode *ip)
99 if (S_ISREG(VFS_I(ip)->i_mode))
100 return i_size_read(VFS_I(ip));
101 return ip->i_disk_size;
105 * If this I/O goes past the on-disk inode size update it unless it would
106 * be past the current in-core inode size.
108 static inline xfs_fsize_t
109 xfs_new_eof(struct xfs_inode *ip, xfs_fsize_t new_size)
111 xfs_fsize_t i_size = i_size_read(VFS_I(ip));
113 if (new_size > i_size || new_size < 0)
115 return new_size > ip->i_disk_size ? new_size : 0;
119 * i_flags helper functions
122 __xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
124 ip->i_flags |= flags;
128 xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
130 spin_lock(&ip->i_flags_lock);
131 __xfs_iflags_set(ip, flags);
132 spin_unlock(&ip->i_flags_lock);
136 xfs_iflags_clear(xfs_inode_t *ip, unsigned short flags)
138 spin_lock(&ip->i_flags_lock);
139 ip->i_flags &= ~flags;
140 spin_unlock(&ip->i_flags_lock);
144 __xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
146 return (ip->i_flags & flags);
150 xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
153 spin_lock(&ip->i_flags_lock);
154 ret = __xfs_iflags_test(ip, flags);
155 spin_unlock(&ip->i_flags_lock);
160 xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
164 spin_lock(&ip->i_flags_lock);
165 ret = ip->i_flags & flags;
167 ip->i_flags &= ~flags;
168 spin_unlock(&ip->i_flags_lock);
173 xfs_iflags_test_and_set(xfs_inode_t *ip, unsigned short flags)
177 spin_lock(&ip->i_flags_lock);
178 ret = ip->i_flags & flags;
180 ip->i_flags |= flags;
181 spin_unlock(&ip->i_flags_lock);
186 xfs_get_initial_prid(struct xfs_inode *dp)
188 if (dp->i_diflags & XFS_DIFLAG_PROJINHERIT)
191 return XFS_PROJID_DEFAULT;
194 static inline bool xfs_is_reflink_inode(struct xfs_inode *ip)
196 return ip->i_diflags2 & XFS_DIFLAG2_REFLINK;
199 static inline bool xfs_is_metadata_inode(struct xfs_inode *ip)
201 struct xfs_mount *mp = ip->i_mount;
203 return ip == mp->m_rbmip || ip == mp->m_rsumip ||
204 xfs_is_quota_inode(&mp->m_sb, ip->i_ino);
208 * Check if an inode has any data in the COW fork. This might be often false
209 * even for inodes with the reflink flag when there is no pending COW operation.
211 static inline bool xfs_inode_has_cow_data(struct xfs_inode *ip)
213 return ip->i_cowfp && ip->i_cowfp->if_bytes;
216 static inline bool xfs_inode_has_bigtime(struct xfs_inode *ip)
218 return ip->i_diflags2 & XFS_DIFLAG2_BIGTIME;
222 * Return the buftarg used for data allocations on a given inode.
224 #define xfs_inode_buftarg(ip) \
225 (XFS_IS_REALTIME_INODE(ip) ? \
226 (ip)->i_mount->m_rtdev_targp : (ip)->i_mount->m_ddev_targp)
229 * In-core inode flags.
231 #define XFS_IRECLAIM (1 << 0) /* started reclaiming this inode */
232 #define XFS_ISTALE (1 << 1) /* inode has been staled */
233 #define XFS_IRECLAIMABLE (1 << 2) /* inode can be reclaimed */
234 #define XFS_INEW (1 << 3) /* inode has just been allocated */
235 #define XFS_IPRESERVE_DM_FIELDS (1 << 4) /* has legacy DMAPI fields set */
236 #define XFS_ITRUNCATED (1 << 5) /* truncated down so flush-on-close */
237 #define XFS_IDIRTY_RELEASE (1 << 6) /* dirty release already seen */
238 #define XFS_IFLUSHING (1 << 7) /* inode is being flushed */
239 #define __XFS_IPINNED_BIT 8 /* wakeup key for zero pin count */
240 #define XFS_IPINNED (1 << __XFS_IPINNED_BIT)
241 #define XFS_IEOFBLOCKS (1 << 9) /* has the preallocblocks tag set */
242 #define XFS_NEED_INACTIVE (1 << 10) /* see XFS_INACTIVATING below */
244 * If this unlinked inode is in the middle of recovery, don't let drop_inode
245 * truncate and free the inode. This can happen if we iget the inode during
246 * log recovery to replay a bmap operation on the inode.
248 #define XFS_IRECOVERY (1 << 11)
249 #define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */
252 * If we need to update on-disk metadata before this IRECLAIMABLE inode can be
253 * freed, then NEED_INACTIVE will be set. Once we start the updates, the
254 * INACTIVATING bit will be set to keep iget away from this inode. After the
255 * inactivation completes, both flags will be cleared and the inode is a
256 * plain old IRECLAIMABLE inode.
258 #define XFS_INACTIVATING (1 << 13)
260 /* All inode state flags related to inode reclaim. */
261 #define XFS_ALL_IRECLAIM_FLAGS (XFS_IRECLAIMABLE | \
263 XFS_NEED_INACTIVE | \
267 * Per-lifetime flags need to be reset when re-using a reclaimable inode during
268 * inode lookup. This prevents unintended behaviour on the new inode from
271 #define XFS_IRECLAIM_RESET_FLAGS \
272 (XFS_IRECLAIMABLE | XFS_IRECLAIM | \
273 XFS_IDIRTY_RELEASE | XFS_ITRUNCATED | XFS_NEED_INACTIVE | \
277 * Flags for inode locking.
278 * Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
279 * 1<<16 - 1<<32-1 -- lockdep annotation (integers)
281 #define XFS_IOLOCK_EXCL (1<<0)
282 #define XFS_IOLOCK_SHARED (1<<1)
283 #define XFS_ILOCK_EXCL (1<<2)
284 #define XFS_ILOCK_SHARED (1<<3)
285 #define XFS_MMAPLOCK_EXCL (1<<4)
286 #define XFS_MMAPLOCK_SHARED (1<<5)
288 #define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
289 | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
290 | XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
292 #define XFS_LOCK_FLAGS \
293 { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
294 { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
295 { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
296 { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
297 { XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \
298 { XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" }
302 * Flags for lockdep annotations.
304 * XFS_LOCK_PARENT - for directory operations that require locking a
305 * parent directory inode and a child entry inode. IOLOCK requires nesting,
306 * MMAPLOCK does not support this class, ILOCK requires a single subclass
307 * to differentiate parent from child.
309 * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
310 * inodes do not participate in the normal lock order, and thus have their
313 * XFS_LOCK_INUMORDER - for locking several inodes at the some time
314 * with xfs_lock_inodes(). This flag is used as the starting subclass
315 * and each subsequent lock acquired will increment the subclass by one.
316 * However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
317 * limited to the subclasses we can represent via nesting. We need at least
318 * 5 inodes nest depth for the ILOCK through rename, and we also have to support
319 * XFS_ILOCK_PARENT, which gives 6 subclasses. Then we have XFS_ILOCK_RTBITMAP
320 * and XFS_ILOCK_RTSUM, which are another 2 unique subclasses, so that's all
321 * 8 subclasses supported by lockdep.
323 * This also means we have to number the sub-classes in the lowest bits of
324 * the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
325 * mask and we can't use bit-masking to build the subclasses. What a mess.
330 * 16-19 XFS_IOLOCK_SHIFT dependencies
331 * 20-23 XFS_MMAPLOCK_SHIFT dependencies
332 * 24-31 XFS_ILOCK_SHIFT dependencies
345 * 0-4 subclass values
346 * 5 PARENT subclass (not nestable)
347 * 6 RTBITMAP subclass (not nestable)
348 * 7 RTSUM subclass (not nestable)
351 #define XFS_IOLOCK_SHIFT 16
352 #define XFS_IOLOCK_MAX_SUBCLASS 3
353 #define XFS_IOLOCK_DEP_MASK 0x000f0000
355 #define XFS_MMAPLOCK_SHIFT 20
356 #define XFS_MMAPLOCK_NUMORDER 0
357 #define XFS_MMAPLOCK_MAX_SUBCLASS 3
358 #define XFS_MMAPLOCK_DEP_MASK 0x00f00000
360 #define XFS_ILOCK_SHIFT 24
361 #define XFS_ILOCK_PARENT_VAL 5
362 #define XFS_ILOCK_MAX_SUBCLASS (XFS_ILOCK_PARENT_VAL - 1)
363 #define XFS_ILOCK_RTBITMAP_VAL 6
364 #define XFS_ILOCK_RTSUM_VAL 7
365 #define XFS_ILOCK_DEP_MASK 0xff000000
366 #define XFS_ILOCK_PARENT (XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
367 #define XFS_ILOCK_RTBITMAP (XFS_ILOCK_RTBITMAP_VAL << XFS_ILOCK_SHIFT)
368 #define XFS_ILOCK_RTSUM (XFS_ILOCK_RTSUM_VAL << XFS_ILOCK_SHIFT)
370 #define XFS_LOCK_SUBCLASS_MASK (XFS_IOLOCK_DEP_MASK | \
371 XFS_MMAPLOCK_DEP_MASK | \
374 #define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \
376 #define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \
377 >> XFS_MMAPLOCK_SHIFT)
378 #define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \
382 * Layouts are broken in the BREAK_WRITE case to ensure that
383 * layout-holders do not collide with local writes. Additionally,
384 * layouts are broken in the BREAK_UNMAP case to make sure the
385 * layout-holder has a consistent view of the file's extent map. While
386 * BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
387 * BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
390 enum layout_break_reason {
396 * For multiple groups support: if S_ISGID bit is set in the parent
397 * directory, group of new file is set to that of the parent, and
398 * new subdirectory gets S_ISGID bit from parent.
400 #define XFS_INHERIT_GID(pip) \
401 (xfs_has_grpid((pip)->i_mount) || (VFS_I(pip)->i_mode & S_ISGID))
403 int xfs_release(struct xfs_inode *ip);
404 void xfs_inactive(struct xfs_inode *ip);
405 int xfs_lookup(struct xfs_inode *dp, const struct xfs_name *name,
406 struct xfs_inode **ipp, struct xfs_name *ci_name);
407 int xfs_create(struct user_namespace *mnt_userns,
408 struct xfs_inode *dp, struct xfs_name *name,
409 umode_t mode, dev_t rdev, bool need_xattr,
410 struct xfs_inode **ipp);
411 int xfs_create_tmpfile(struct user_namespace *mnt_userns,
412 struct xfs_inode *dp, umode_t mode,
413 struct xfs_inode **ipp);
414 int xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
415 struct xfs_inode *ip);
416 int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
417 struct xfs_name *target_name);
418 int xfs_rename(struct user_namespace *mnt_userns,
419 struct xfs_inode *src_dp, struct xfs_name *src_name,
420 struct xfs_inode *src_ip, struct xfs_inode *target_dp,
421 struct xfs_name *target_name,
422 struct xfs_inode *target_ip, unsigned int flags);
424 void xfs_ilock(xfs_inode_t *, uint);
425 int xfs_ilock_nowait(xfs_inode_t *, uint);
426 void xfs_iunlock(xfs_inode_t *, uint);
427 void xfs_ilock_demote(xfs_inode_t *, uint);
428 bool xfs_isilocked(struct xfs_inode *, uint);
429 uint xfs_ilock_data_map_shared(struct xfs_inode *);
430 uint xfs_ilock_attr_map_shared(struct xfs_inode *);
432 uint xfs_ip2xflags(struct xfs_inode *);
433 int xfs_ifree(struct xfs_trans *, struct xfs_inode *);
434 int xfs_itruncate_extents_flags(struct xfs_trans **,
435 struct xfs_inode *, int, xfs_fsize_t, int);
436 void xfs_iext_realloc(xfs_inode_t *, int, int);
438 int xfs_log_force_inode(struct xfs_inode *ip);
439 void xfs_iunpin_wait(xfs_inode_t *);
440 #define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
442 int xfs_iflush_cluster(struct xfs_buf *);
443 void xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
444 struct xfs_inode *ip1, uint ip1_mode);
446 xfs_extlen_t xfs_get_extsz_hint(struct xfs_inode *ip);
447 xfs_extlen_t xfs_get_cowextsz_hint(struct xfs_inode *ip);
449 int xfs_init_new_inode(struct user_namespace *mnt_userns, struct xfs_trans *tp,
450 struct xfs_inode *pip, xfs_ino_t ino, umode_t mode,
451 xfs_nlink_t nlink, dev_t rdev, prid_t prid, bool init_xattrs,
452 struct xfs_inode **ipp);
455 xfs_itruncate_extents(
456 struct xfs_trans **tpp,
457 struct xfs_inode *ip,
459 xfs_fsize_t new_size)
461 return xfs_itruncate_extents_flags(tpp, ip, whichfork, new_size, 0);
464 /* from xfs_file.c */
465 int xfs_break_layouts(struct inode *inode, uint *iolock,
466 enum layout_break_reason reason);
468 /* from xfs_iops.c */
469 extern void xfs_setup_inode(struct xfs_inode *ip);
470 extern void xfs_setup_iops(struct xfs_inode *ip);
471 extern void xfs_diflags_to_iflags(struct xfs_inode *ip, bool init);
474 * When setting up a newly allocated inode, we need to call
475 * xfs_finish_inode_setup() once the inode is fully instantiated at
476 * the VFS level to prevent the rest of the world seeing the inode
477 * before we've completed instantiation. Otherwise we can do it
478 * the moment the inode lookup is complete.
480 static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
482 xfs_iflags_clear(ip, XFS_INEW);
484 unlock_new_inode(VFS_I(ip));
487 static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
491 xfs_finish_inode_setup(ip);
494 void xfs_irele(struct xfs_inode *ip);
496 extern struct kmem_cache *xfs_inode_cache;
498 /* The default CoW extent size hint. */
499 #define XFS_DEFAULT_COWEXTSZ_HINT 32
501 bool xfs_inode_needs_inactive(struct xfs_inode *ip);
503 int xfs_iunlink_init(struct xfs_perag *pag);
504 void xfs_iunlink_destroy(struct xfs_perag *pag);
506 void xfs_end_io(struct work_struct *work);
508 int xfs_ilock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
509 void xfs_iunlock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
511 #endif /* __XFS_INODE_H__ */