xref: /linux/fs/xfs/xfs_inode.h (revision 987b741c52c7c6c68d46fbaeb95b8d1087f10b7f)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2003,2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #ifndef	__XFS_INODE_H__
7 #define	__XFS_INODE_H__
8 
9 #include "xfs_inode_buf.h"
10 #include "xfs_inode_fork.h"
11 
12 /*
13  * Kernel only inode definitions
14  */
15 struct xfs_dinode;
16 struct xfs_inode;
17 struct xfs_buf;
18 struct xfs_bmbt_irec;
19 struct xfs_inode_log_item;
20 struct xfs_mount;
21 struct xfs_trans;
22 struct xfs_dquot;
23 
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 */
30 
31 	/* Inode location stuff */
32 	xfs_ino_t		i_ino;		/* inode number (agno/agino)*/
33 	struct xfs_imap		i_imap;		/* location for xfs_imap() */
34 
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 */
39 
40 	/* Transaction and locking information. */
41 	struct xfs_inode_log_item *i_itemp;	/* logging information */
42 	mrlock_t		i_lock;		/* inode lock */
43 	mrlock_t		i_mmaplock;	/* inode mmap IO lock */
44 	atomic_t		i_pincount;	/* inode pin count */
45 
46 	/*
47 	 * Bitsets of inode metadata that have been checked and/or are sick.
48 	 * Callers must hold i_flags_lock before accessing this field.
49 	 */
50 	uint16_t		i_checked;
51 	uint16_t		i_sick;
52 
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 */
62 	union {
63 		xfs_extlen_t	i_cowextsize;	/* basic cow extent size */
64 		uint16_t	i_flushiter;	/* incremented on flush */
65 	};
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 */
70 
71 	/* VFS inode */
72 	struct inode		i_vnode;	/* embedded VFS inode */
73 
74 	/* pending io completions */
75 	spinlock_t		i_ioend_lock;
76 	struct work_struct	i_ioend_work;
77 	struct list_head	i_ioend_list;
78 } xfs_inode_t;
79 
80 /* Convert from vfs inode to xfs inode */
81 static inline struct xfs_inode *XFS_I(struct inode *inode)
82 {
83 	return container_of(inode, struct xfs_inode, i_vnode);
84 }
85 
86 /* convert from xfs inode to vfs inode */
87 static inline struct inode *VFS_I(struct xfs_inode *ip)
88 {
89 	return &ip->i_vnode;
90 }
91 
92 /*
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
95  * is uptodate.
96  */
97 static inline xfs_fsize_t XFS_ISIZE(struct xfs_inode *ip)
98 {
99 	if (S_ISREG(VFS_I(ip)->i_mode))
100 		return i_size_read(VFS_I(ip));
101 	return ip->i_disk_size;
102 }
103 
104 /*
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.
107  */
108 static inline xfs_fsize_t
109 xfs_new_eof(struct xfs_inode *ip, xfs_fsize_t new_size)
110 {
111 	xfs_fsize_t i_size = i_size_read(VFS_I(ip));
112 
113 	if (new_size > i_size || new_size < 0)
114 		new_size = i_size;
115 	return new_size > ip->i_disk_size ? new_size : 0;
116 }
117 
118 /*
119  * i_flags helper functions
120  */
121 static inline void
122 __xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
123 {
124 	ip->i_flags |= flags;
125 }
126 
127 static inline void
128 xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
129 {
130 	spin_lock(&ip->i_flags_lock);
131 	__xfs_iflags_set(ip, flags);
132 	spin_unlock(&ip->i_flags_lock);
133 }
134 
135 static inline void
136 xfs_iflags_clear(xfs_inode_t *ip, unsigned short flags)
137 {
138 	spin_lock(&ip->i_flags_lock);
139 	ip->i_flags &= ~flags;
140 	spin_unlock(&ip->i_flags_lock);
141 }
142 
143 static inline int
144 __xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
145 {
146 	return (ip->i_flags & flags);
147 }
148 
149 static inline int
150 xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
151 {
152 	int ret;
153 	spin_lock(&ip->i_flags_lock);
154 	ret = __xfs_iflags_test(ip, flags);
155 	spin_unlock(&ip->i_flags_lock);
156 	return ret;
157 }
158 
159 static inline int
160 xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
161 {
162 	int ret;
163 
164 	spin_lock(&ip->i_flags_lock);
165 	ret = ip->i_flags & flags;
166 	if (ret)
167 		ip->i_flags &= ~flags;
168 	spin_unlock(&ip->i_flags_lock);
169 	return ret;
170 }
171 
172 static inline int
173 xfs_iflags_test_and_set(xfs_inode_t *ip, unsigned short flags)
174 {
175 	int ret;
176 
177 	spin_lock(&ip->i_flags_lock);
178 	ret = ip->i_flags & flags;
179 	if (!ret)
180 		ip->i_flags |= flags;
181 	spin_unlock(&ip->i_flags_lock);
182 	return ret;
183 }
184 
185 static inline prid_t
186 xfs_get_initial_prid(struct xfs_inode *dp)
187 {
188 	if (dp->i_diflags & XFS_DIFLAG_PROJINHERIT)
189 		return dp->i_projid;
190 
191 	return XFS_PROJID_DEFAULT;
192 }
193 
194 static inline bool xfs_is_reflink_inode(struct xfs_inode *ip)
195 {
196 	return ip->i_diflags2 & XFS_DIFLAG2_REFLINK;
197 }
198 
199 static inline bool xfs_is_metadata_inode(struct xfs_inode *ip)
200 {
201 	struct xfs_mount	*mp = ip->i_mount;
202 
203 	return ip == mp->m_rbmip || ip == mp->m_rsumip ||
204 		xfs_is_quota_inode(&mp->m_sb, ip->i_ino);
205 }
206 
207 /*
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.
210  */
211 static inline bool xfs_inode_has_cow_data(struct xfs_inode *ip)
212 {
213 	return ip->i_cowfp && ip->i_cowfp->if_bytes;
214 }
215 
216 static inline bool xfs_inode_has_bigtime(struct xfs_inode *ip)
217 {
218 	return ip->i_diflags2 & XFS_DIFLAG2_BIGTIME;
219 }
220 
221 /*
222  * Return the buftarg used for data allocations on a given inode.
223  */
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)
227 
228 /*
229  * In-core inode flags.
230  */
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_BIT		3	 /* inode has just been allocated */
235 #define XFS_INEW		(1 << __XFS_INEW_BIT)
236 #define XFS_IPRESERVE_DM_FIELDS	(1 << 4) /* has legacy DMAPI fields set */
237 #define XFS_ITRUNCATED		(1 << 5) /* truncated down so flush-on-close */
238 #define XFS_IDIRTY_RELEASE	(1 << 6) /* dirty release already seen */
239 #define XFS_IFLUSHING		(1 << 7) /* inode is being flushed */
240 #define __XFS_IPINNED_BIT	8	 /* wakeup key for zero pin count */
241 #define XFS_IPINNED		(1 << __XFS_IPINNED_BIT)
242 #define XFS_IEOFBLOCKS		(1 << 9) /* has the preallocblocks tag set */
243 /*
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.
247  */
248 #define XFS_IRECOVERY		(1 << 11)
249 #define XFS_ICOWBLOCKS		(1 << 12)/* has the cowblocks tag set */
250 
251 /*
252  * Per-lifetime flags need to be reset when re-using a reclaimable inode during
253  * inode lookup. This prevents unintended behaviour on the new inode from
254  * ocurring.
255  */
256 #define XFS_IRECLAIM_RESET_FLAGS	\
257 	(XFS_IRECLAIMABLE | XFS_IRECLAIM | \
258 	 XFS_IDIRTY_RELEASE | XFS_ITRUNCATED)
259 
260 /*
261  * Flags for inode locking.
262  * Bit ranges:	1<<1  - 1<<16-1 -- iolock/ilock modes (bitfield)
263  *		1<<16 - 1<<32-1 -- lockdep annotation (integers)
264  */
265 #define	XFS_IOLOCK_EXCL		(1<<0)
266 #define	XFS_IOLOCK_SHARED	(1<<1)
267 #define	XFS_ILOCK_EXCL		(1<<2)
268 #define	XFS_ILOCK_SHARED	(1<<3)
269 #define	XFS_MMAPLOCK_EXCL	(1<<4)
270 #define	XFS_MMAPLOCK_SHARED	(1<<5)
271 
272 #define XFS_LOCK_MASK		(XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
273 				| XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
274 				| XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
275 
276 #define XFS_LOCK_FLAGS \
277 	{ XFS_IOLOCK_EXCL,	"IOLOCK_EXCL" }, \
278 	{ XFS_IOLOCK_SHARED,	"IOLOCK_SHARED" }, \
279 	{ XFS_ILOCK_EXCL,	"ILOCK_EXCL" }, \
280 	{ XFS_ILOCK_SHARED,	"ILOCK_SHARED" }, \
281 	{ XFS_MMAPLOCK_EXCL,	"MMAPLOCK_EXCL" }, \
282 	{ XFS_MMAPLOCK_SHARED,	"MMAPLOCK_SHARED" }
283 
284 
285 /*
286  * Flags for lockdep annotations.
287  *
288  * XFS_LOCK_PARENT - for directory operations that require locking a
289  * parent directory inode and a child entry inode. IOLOCK requires nesting,
290  * MMAPLOCK does not support this class, ILOCK requires a single subclass
291  * to differentiate parent from child.
292  *
293  * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
294  * inodes do not participate in the normal lock order, and thus have their
295  * own subclasses.
296  *
297  * XFS_LOCK_INUMORDER - for locking several inodes at the some time
298  * with xfs_lock_inodes().  This flag is used as the starting subclass
299  * and each subsequent lock acquired will increment the subclass by one.
300  * However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
301  * limited to the subclasses we can represent via nesting. We need at least
302  * 5 inodes nest depth for the ILOCK through rename, and we also have to support
303  * XFS_ILOCK_PARENT, which gives 6 subclasses. Then we have XFS_ILOCK_RTBITMAP
304  * and XFS_ILOCK_RTSUM, which are another 2 unique subclasses, so that's all
305  * 8 subclasses supported by lockdep.
306  *
307  * This also means we have to number the sub-classes in the lowest bits of
308  * the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
309  * mask and we can't use bit-masking to build the subclasses. What a mess.
310  *
311  * Bit layout:
312  *
313  * Bit		Lock Region
314  * 16-19	XFS_IOLOCK_SHIFT dependencies
315  * 20-23	XFS_MMAPLOCK_SHIFT dependencies
316  * 24-31	XFS_ILOCK_SHIFT dependencies
317  *
318  * IOLOCK values
319  *
320  * 0-3		subclass value
321  * 4-7		unused
322  *
323  * MMAPLOCK values
324  *
325  * 0-3		subclass value
326  * 4-7		unused
327  *
328  * ILOCK values
329  * 0-4		subclass values
330  * 5		PARENT subclass (not nestable)
331  * 6		RTBITMAP subclass (not nestable)
332  * 7		RTSUM subclass (not nestable)
333  *
334  */
335 #define XFS_IOLOCK_SHIFT		16
336 #define XFS_IOLOCK_MAX_SUBCLASS		3
337 #define XFS_IOLOCK_DEP_MASK		0x000f0000
338 
339 #define XFS_MMAPLOCK_SHIFT		20
340 #define XFS_MMAPLOCK_NUMORDER		0
341 #define XFS_MMAPLOCK_MAX_SUBCLASS	3
342 #define XFS_MMAPLOCK_DEP_MASK		0x00f00000
343 
344 #define XFS_ILOCK_SHIFT			24
345 #define XFS_ILOCK_PARENT_VAL		5
346 #define XFS_ILOCK_MAX_SUBCLASS		(XFS_ILOCK_PARENT_VAL - 1)
347 #define XFS_ILOCK_RTBITMAP_VAL		6
348 #define XFS_ILOCK_RTSUM_VAL		7
349 #define XFS_ILOCK_DEP_MASK		0xff000000
350 #define	XFS_ILOCK_PARENT		(XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
351 #define	XFS_ILOCK_RTBITMAP		(XFS_ILOCK_RTBITMAP_VAL << XFS_ILOCK_SHIFT)
352 #define	XFS_ILOCK_RTSUM			(XFS_ILOCK_RTSUM_VAL << XFS_ILOCK_SHIFT)
353 
354 #define XFS_LOCK_SUBCLASS_MASK	(XFS_IOLOCK_DEP_MASK | \
355 				 XFS_MMAPLOCK_DEP_MASK | \
356 				 XFS_ILOCK_DEP_MASK)
357 
358 #define XFS_IOLOCK_DEP(flags)	(((flags) & XFS_IOLOCK_DEP_MASK) \
359 					>> XFS_IOLOCK_SHIFT)
360 #define XFS_MMAPLOCK_DEP(flags)	(((flags) & XFS_MMAPLOCK_DEP_MASK) \
361 					>> XFS_MMAPLOCK_SHIFT)
362 #define XFS_ILOCK_DEP(flags)	(((flags) & XFS_ILOCK_DEP_MASK) \
363 					>> XFS_ILOCK_SHIFT)
364 
365 /*
366  * Layouts are broken in the BREAK_WRITE case to ensure that
367  * layout-holders do not collide with local writes. Additionally,
368  * layouts are broken in the BREAK_UNMAP case to make sure the
369  * layout-holder has a consistent view of the file's extent map. While
370  * BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
371  * BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
372  * go idle.
373  */
374 enum layout_break_reason {
375         BREAK_WRITE,
376         BREAK_UNMAP,
377 };
378 
379 /*
380  * For multiple groups support: if S_ISGID bit is set in the parent
381  * directory, group of new file is set to that of the parent, and
382  * new subdirectory gets S_ISGID bit from parent.
383  */
384 #define XFS_INHERIT_GID(pip)	\
385 	(((pip)->i_mount->m_flags & XFS_MOUNT_GRPID) || \
386 	 (VFS_I(pip)->i_mode & S_ISGID))
387 
388 int		xfs_release(struct xfs_inode *ip);
389 void		xfs_inactive(struct xfs_inode *ip);
390 int		xfs_lookup(struct xfs_inode *dp, struct xfs_name *name,
391 			   struct xfs_inode **ipp, struct xfs_name *ci_name);
392 int		xfs_create(struct user_namespace *mnt_userns,
393 			   struct xfs_inode *dp, struct xfs_name *name,
394 			   umode_t mode, dev_t rdev, bool need_xattr,
395 			   struct xfs_inode **ipp);
396 int		xfs_create_tmpfile(struct user_namespace *mnt_userns,
397 			   struct xfs_inode *dp, umode_t mode,
398 			   struct xfs_inode **ipp);
399 int		xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
400 			   struct xfs_inode *ip);
401 int		xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
402 			 struct xfs_name *target_name);
403 int		xfs_rename(struct user_namespace *mnt_userns,
404 			   struct xfs_inode *src_dp, struct xfs_name *src_name,
405 			   struct xfs_inode *src_ip, struct xfs_inode *target_dp,
406 			   struct xfs_name *target_name,
407 			   struct xfs_inode *target_ip, unsigned int flags);
408 
409 void		xfs_ilock(xfs_inode_t *, uint);
410 int		xfs_ilock_nowait(xfs_inode_t *, uint);
411 void		xfs_iunlock(xfs_inode_t *, uint);
412 void		xfs_ilock_demote(xfs_inode_t *, uint);
413 int		xfs_isilocked(xfs_inode_t *, uint);
414 uint		xfs_ilock_data_map_shared(struct xfs_inode *);
415 uint		xfs_ilock_attr_map_shared(struct xfs_inode *);
416 
417 uint		xfs_ip2xflags(struct xfs_inode *);
418 int		xfs_ifree(struct xfs_trans *, struct xfs_inode *);
419 int		xfs_itruncate_extents_flags(struct xfs_trans **,
420 				struct xfs_inode *, int, xfs_fsize_t, int);
421 void		xfs_iext_realloc(xfs_inode_t *, int, int);
422 
423 int		xfs_log_force_inode(struct xfs_inode *ip);
424 void		xfs_iunpin_wait(xfs_inode_t *);
425 #define xfs_ipincount(ip)	((unsigned int) atomic_read(&ip->i_pincount))
426 
427 int		xfs_iflush_cluster(struct xfs_buf *);
428 void		xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
429 				struct xfs_inode *ip1, uint ip1_mode);
430 
431 xfs_extlen_t	xfs_get_extsz_hint(struct xfs_inode *ip);
432 xfs_extlen_t	xfs_get_cowextsz_hint(struct xfs_inode *ip);
433 
434 int		xfs_dir_ialloc(struct user_namespace *mnt_userns,
435 			       struct xfs_trans **tpp, struct xfs_inode *dp,
436 			       umode_t mode, xfs_nlink_t nlink, dev_t dev,
437 			       prid_t prid, bool need_xattr,
438 			       struct xfs_inode **ipp);
439 
440 static inline int
441 xfs_itruncate_extents(
442 	struct xfs_trans	**tpp,
443 	struct xfs_inode	*ip,
444 	int			whichfork,
445 	xfs_fsize_t		new_size)
446 {
447 	return xfs_itruncate_extents_flags(tpp, ip, whichfork, new_size, 0);
448 }
449 
450 /* from xfs_file.c */
451 enum xfs_prealloc_flags {
452 	XFS_PREALLOC_SET	= (1 << 1),
453 	XFS_PREALLOC_CLEAR	= (1 << 2),
454 	XFS_PREALLOC_SYNC	= (1 << 3),
455 	XFS_PREALLOC_INVISIBLE	= (1 << 4),
456 };
457 
458 int	xfs_update_prealloc_flags(struct xfs_inode *ip,
459 				  enum xfs_prealloc_flags flags);
460 int	xfs_break_layouts(struct inode *inode, uint *iolock,
461 		enum layout_break_reason reason);
462 
463 /* from xfs_iops.c */
464 extern void xfs_setup_inode(struct xfs_inode *ip);
465 extern void xfs_setup_iops(struct xfs_inode *ip);
466 extern void xfs_diflags_to_iflags(struct xfs_inode *ip, bool init);
467 
468 /*
469  * When setting up a newly allocated inode, we need to call
470  * xfs_finish_inode_setup() once the inode is fully instantiated at
471  * the VFS level to prevent the rest of the world seeing the inode
472  * before we've completed instantiation. Otherwise we can do it
473  * the moment the inode lookup is complete.
474  */
475 static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
476 {
477 	xfs_iflags_clear(ip, XFS_INEW);
478 	barrier();
479 	unlock_new_inode(VFS_I(ip));
480 	wake_up_bit(&ip->i_flags, __XFS_INEW_BIT);
481 }
482 
483 static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
484 {
485 	xfs_setup_inode(ip);
486 	xfs_setup_iops(ip);
487 	xfs_finish_inode_setup(ip);
488 }
489 
490 void xfs_irele(struct xfs_inode *ip);
491 
492 extern struct kmem_zone	*xfs_inode_zone;
493 
494 /* The default CoW extent size hint. */
495 #define XFS_DEFAULT_COWEXTSZ_HINT 32
496 
497 int xfs_iunlink_init(struct xfs_perag *pag);
498 void xfs_iunlink_destroy(struct xfs_perag *pag);
499 
500 void xfs_end_io(struct work_struct *work);
501 
502 int xfs_ilock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
503 void xfs_iunlock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
504 
505 #endif	/* __XFS_INODE_H__ */
506