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 atomic_t i_pincount; /* inode pin count */
44 struct llist_node i_gclist; /* deferred inactivation list */
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 */
XFS_I(struct inode * 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 */
VFS_I(struct xfs_inode * ip)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 */
XFS_ISIZE(struct xfs_inode * ip)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
xfs_new_eof(struct xfs_inode * ip,xfs_fsize_t new_size)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
__xfs_iflags_set(xfs_inode_t * ip,unsigned short flags)122 __xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
123 {
124 ip->i_flags |= flags;
125 }
126
127 static inline void
xfs_iflags_set(xfs_inode_t * ip,unsigned short flags)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
xfs_iflags_clear(xfs_inode_t * ip,unsigned short flags)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
__xfs_iflags_test(xfs_inode_t * ip,unsigned short flags)144 __xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
145 {
146 return (ip->i_flags & flags);
147 }
148
149 static inline int
xfs_iflags_test(xfs_inode_t * ip,unsigned short flags)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
xfs_iflags_test_and_clear(xfs_inode_t * ip,unsigned short flags)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
xfs_iflags_test_and_set(xfs_inode_t * ip,unsigned short flags)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
xfs_get_initial_prid(struct xfs_inode * dp)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
xfs_is_reflink_inode(struct xfs_inode * ip)194 static inline bool xfs_is_reflink_inode(struct xfs_inode *ip)
195 {
196 return ip->i_diflags2 & XFS_DIFLAG2_REFLINK;
197 }
198
xfs_is_metadata_inode(struct xfs_inode * ip)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 */
xfs_inode_has_cow_data(struct xfs_inode * ip)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
xfs_inode_has_bigtime(struct xfs_inode * ip)216 static inline bool xfs_inode_has_bigtime(struct xfs_inode *ip)
217 {
218 return ip->i_diflags2 & XFS_DIFLAG2_BIGTIME;
219 }
220
xfs_inode_has_large_extent_counts(struct xfs_inode * ip)221 static inline bool xfs_inode_has_large_extent_counts(struct xfs_inode *ip)
222 {
223 return ip->i_diflags2 & XFS_DIFLAG2_NREXT64;
224 }
225
226 /*
227 * Return the buftarg used for data allocations on a given inode.
228 */
229 #define xfs_inode_buftarg(ip) \
230 (XFS_IS_REALTIME_INODE(ip) ? \
231 (ip)->i_mount->m_rtdev_targp : (ip)->i_mount->m_ddev_targp)
232
233 /*
234 * In-core inode flags.
235 */
236 #define XFS_IRECLAIM (1 << 0) /* started reclaiming this inode */
237 #define XFS_ISTALE (1 << 1) /* inode has been staled */
238 #define XFS_IRECLAIMABLE (1 << 2) /* inode can be reclaimed */
239 #define XFS_INEW (1 << 3) /* inode has just been allocated */
240 #define XFS_IPRESERVE_DM_FIELDS (1 << 4) /* has legacy DMAPI fields set */
241 #define XFS_ITRUNCATED (1 << 5) /* truncated down so flush-on-close */
242 #define XFS_IDIRTY_RELEASE (1 << 6) /* dirty release already seen */
243 #define XFS_IFLUSHING (1 << 7) /* inode is being flushed */
244 #define __XFS_IPINNED_BIT 8 /* wakeup key for zero pin count */
245 #define XFS_IPINNED (1 << __XFS_IPINNED_BIT)
246 #define XFS_IEOFBLOCKS (1 << 9) /* has the preallocblocks tag set */
247 #define XFS_NEED_INACTIVE (1 << 10) /* see XFS_INACTIVATING below */
248 /*
249 * If this unlinked inode is in the middle of recovery, don't let drop_inode
250 * truncate and free the inode. This can happen if we iget the inode during
251 * log recovery to replay a bmap operation on the inode.
252 */
253 #define XFS_IRECOVERY (1 << 11)
254 #define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */
255
256 /*
257 * If we need to update on-disk metadata before this IRECLAIMABLE inode can be
258 * freed, then NEED_INACTIVE will be set. Once we start the updates, the
259 * INACTIVATING bit will be set to keep iget away from this inode. After the
260 * inactivation completes, both flags will be cleared and the inode is a
261 * plain old IRECLAIMABLE inode.
262 */
263 #define XFS_INACTIVATING (1 << 13)
264
265 /* All inode state flags related to inode reclaim. */
266 #define XFS_ALL_IRECLAIM_FLAGS (XFS_IRECLAIMABLE | \
267 XFS_IRECLAIM | \
268 XFS_NEED_INACTIVE | \
269 XFS_INACTIVATING)
270
271 /*
272 * Per-lifetime flags need to be reset when re-using a reclaimable inode during
273 * inode lookup. This prevents unintended behaviour on the new inode from
274 * ocurring.
275 */
276 #define XFS_IRECLAIM_RESET_FLAGS \
277 (XFS_IRECLAIMABLE | XFS_IRECLAIM | \
278 XFS_IDIRTY_RELEASE | XFS_ITRUNCATED | XFS_NEED_INACTIVE | \
279 XFS_INACTIVATING)
280
281 /*
282 * Flags for inode locking.
283 * Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
284 * 1<<16 - 1<<32-1 -- lockdep annotation (integers)
285 */
286 #define XFS_IOLOCK_EXCL (1u << 0)
287 #define XFS_IOLOCK_SHARED (1u << 1)
288 #define XFS_ILOCK_EXCL (1u << 2)
289 #define XFS_ILOCK_SHARED (1u << 3)
290 #define XFS_MMAPLOCK_EXCL (1u << 4)
291 #define XFS_MMAPLOCK_SHARED (1u << 5)
292
293 #define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
294 | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
295 | XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
296
297 #define XFS_LOCK_FLAGS \
298 { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
299 { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
300 { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
301 { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
302 { XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \
303 { XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" }
304
305
306 /*
307 * Flags for lockdep annotations.
308 *
309 * XFS_LOCK_PARENT - for directory operations that require locking a
310 * parent directory inode and a child entry inode. IOLOCK requires nesting,
311 * MMAPLOCK does not support this class, ILOCK requires a single subclass
312 * to differentiate parent from child.
313 *
314 * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
315 * inodes do not participate in the normal lock order, and thus have their
316 * own subclasses.
317 *
318 * XFS_LOCK_INUMORDER - for locking several inodes at the some time
319 * with xfs_lock_inodes(). This flag is used as the starting subclass
320 * and each subsequent lock acquired will increment the subclass by one.
321 * However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
322 * limited to the subclasses we can represent via nesting. We need at least
323 * 5 inodes nest depth for the ILOCK through rename, and we also have to support
324 * XFS_ILOCK_PARENT, which gives 6 subclasses. Then we have XFS_ILOCK_RTBITMAP
325 * and XFS_ILOCK_RTSUM, which are another 2 unique subclasses, so that's all
326 * 8 subclasses supported by lockdep.
327 *
328 * This also means we have to number the sub-classes in the lowest bits of
329 * the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
330 * mask and we can't use bit-masking to build the subclasses. What a mess.
331 *
332 * Bit layout:
333 *
334 * Bit Lock Region
335 * 16-19 XFS_IOLOCK_SHIFT dependencies
336 * 20-23 XFS_MMAPLOCK_SHIFT dependencies
337 * 24-31 XFS_ILOCK_SHIFT dependencies
338 *
339 * IOLOCK values
340 *
341 * 0-3 subclass value
342 * 4-7 unused
343 *
344 * MMAPLOCK values
345 *
346 * 0-3 subclass value
347 * 4-7 unused
348 *
349 * ILOCK values
350 * 0-4 subclass values
351 * 5 PARENT subclass (not nestable)
352 * 6 RTBITMAP subclass (not nestable)
353 * 7 RTSUM subclass (not nestable)
354 *
355 */
356 #define XFS_IOLOCK_SHIFT 16
357 #define XFS_IOLOCK_MAX_SUBCLASS 3
358 #define XFS_IOLOCK_DEP_MASK 0x000f0000u
359
360 #define XFS_MMAPLOCK_SHIFT 20
361 #define XFS_MMAPLOCK_NUMORDER 0
362 #define XFS_MMAPLOCK_MAX_SUBCLASS 3
363 #define XFS_MMAPLOCK_DEP_MASK 0x00f00000u
364
365 #define XFS_ILOCK_SHIFT 24
366 #define XFS_ILOCK_PARENT_VAL 5u
367 #define XFS_ILOCK_MAX_SUBCLASS (XFS_ILOCK_PARENT_VAL - 1)
368 #define XFS_ILOCK_RTBITMAP_VAL 6u
369 #define XFS_ILOCK_RTSUM_VAL 7u
370 #define XFS_ILOCK_DEP_MASK 0xff000000u
371 #define XFS_ILOCK_PARENT (XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
372 #define XFS_ILOCK_RTBITMAP (XFS_ILOCK_RTBITMAP_VAL << XFS_ILOCK_SHIFT)
373 #define XFS_ILOCK_RTSUM (XFS_ILOCK_RTSUM_VAL << XFS_ILOCK_SHIFT)
374
375 #define XFS_LOCK_SUBCLASS_MASK (XFS_IOLOCK_DEP_MASK | \
376 XFS_MMAPLOCK_DEP_MASK | \
377 XFS_ILOCK_DEP_MASK)
378
379 #define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \
380 >> XFS_IOLOCK_SHIFT)
381 #define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \
382 >> XFS_MMAPLOCK_SHIFT)
383 #define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \
384 >> XFS_ILOCK_SHIFT)
385
386 /*
387 * Layouts are broken in the BREAK_WRITE case to ensure that
388 * layout-holders do not collide with local writes. Additionally,
389 * layouts are broken in the BREAK_UNMAP case to make sure the
390 * layout-holder has a consistent view of the file's extent map. While
391 * BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
392 * BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
393 * go idle.
394 */
395 enum layout_break_reason {
396 BREAK_WRITE,
397 BREAK_UNMAP,
398 };
399
400 /*
401 * For multiple groups support: if S_ISGID bit is set in the parent
402 * directory, group of new file is set to that of the parent, and
403 * new subdirectory gets S_ISGID bit from parent.
404 */
405 #define XFS_INHERIT_GID(pip) \
406 (xfs_has_grpid((pip)->i_mount) || (VFS_I(pip)->i_mode & S_ISGID))
407
408 int xfs_release(struct xfs_inode *ip);
409 void xfs_inactive(struct xfs_inode *ip);
410 int xfs_lookup(struct xfs_inode *dp, const struct xfs_name *name,
411 struct xfs_inode **ipp, struct xfs_name *ci_name);
412 int xfs_create(struct user_namespace *mnt_userns,
413 struct xfs_inode *dp, struct xfs_name *name,
414 umode_t mode, dev_t rdev, bool need_xattr,
415 struct xfs_inode **ipp);
416 int xfs_create_tmpfile(struct user_namespace *mnt_userns,
417 struct xfs_inode *dp, umode_t mode,
418 struct xfs_inode **ipp);
419 int xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
420 struct xfs_inode *ip);
421 int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
422 struct xfs_name *target_name);
423 int xfs_rename(struct user_namespace *mnt_userns,
424 struct xfs_inode *src_dp, struct xfs_name *src_name,
425 struct xfs_inode *src_ip, struct xfs_inode *target_dp,
426 struct xfs_name *target_name,
427 struct xfs_inode *target_ip, unsigned int flags);
428
429 void xfs_ilock(xfs_inode_t *, uint);
430 int xfs_ilock_nowait(xfs_inode_t *, uint);
431 void xfs_iunlock(xfs_inode_t *, uint);
432 void xfs_ilock_demote(xfs_inode_t *, uint);
433 bool xfs_isilocked(struct xfs_inode *, uint);
434 uint xfs_ilock_data_map_shared(struct xfs_inode *);
435 uint xfs_ilock_attr_map_shared(struct xfs_inode *);
436
437 uint xfs_ip2xflags(struct xfs_inode *);
438 int xfs_ifree(struct xfs_trans *, struct xfs_inode *);
439 int xfs_itruncate_extents_flags(struct xfs_trans **,
440 struct xfs_inode *, int, xfs_fsize_t, int);
441 void xfs_iext_realloc(xfs_inode_t *, int, int);
442
443 int xfs_log_force_inode(struct xfs_inode *ip);
444 void xfs_iunpin_wait(xfs_inode_t *);
445 #define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
446
447 int xfs_iflush_cluster(struct xfs_buf *);
448 void xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
449 struct xfs_inode *ip1, uint ip1_mode);
450
451 xfs_extlen_t xfs_get_extsz_hint(struct xfs_inode *ip);
452 xfs_extlen_t xfs_get_cowextsz_hint(struct xfs_inode *ip);
453
454 int xfs_init_new_inode(struct user_namespace *mnt_userns, struct xfs_trans *tp,
455 struct xfs_inode *pip, xfs_ino_t ino, umode_t mode,
456 xfs_nlink_t nlink, dev_t rdev, prid_t prid, bool init_xattrs,
457 struct xfs_inode **ipp);
458
459 static inline int
xfs_itruncate_extents(struct xfs_trans ** tpp,struct xfs_inode * ip,int whichfork,xfs_fsize_t new_size)460 xfs_itruncate_extents(
461 struct xfs_trans **tpp,
462 struct xfs_inode *ip,
463 int whichfork,
464 xfs_fsize_t new_size)
465 {
466 return xfs_itruncate_extents_flags(tpp, ip, whichfork, new_size, 0);
467 }
468
469 /* from xfs_file.c */
470 int xfs_break_layouts(struct inode *inode, uint *iolock,
471 enum layout_break_reason reason);
472
473 /* from xfs_iops.c */
474 extern void xfs_setup_inode(struct xfs_inode *ip);
475 extern void xfs_setup_iops(struct xfs_inode *ip);
476 extern void xfs_diflags_to_iflags(struct xfs_inode *ip, bool init);
477
478 /*
479 * When setting up a newly allocated inode, we need to call
480 * xfs_finish_inode_setup() once the inode is fully instantiated at
481 * the VFS level to prevent the rest of the world seeing the inode
482 * before we've completed instantiation. Otherwise we can do it
483 * the moment the inode lookup is complete.
484 */
xfs_finish_inode_setup(struct xfs_inode * ip)485 static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
486 {
487 xfs_iflags_clear(ip, XFS_INEW);
488 barrier();
489 unlock_new_inode(VFS_I(ip));
490 }
491
xfs_setup_existing_inode(struct xfs_inode * ip)492 static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
493 {
494 xfs_setup_inode(ip);
495 xfs_setup_iops(ip);
496 xfs_finish_inode_setup(ip);
497 }
498
499 void xfs_irele(struct xfs_inode *ip);
500
501 extern struct kmem_cache *xfs_inode_cache;
502
503 /* The default CoW extent size hint. */
504 #define XFS_DEFAULT_COWEXTSZ_HINT 32
505
506 bool xfs_inode_needs_inactive(struct xfs_inode *ip);
507
508 int xfs_iunlink_init(struct xfs_perag *pag);
509 void xfs_iunlink_destroy(struct xfs_perag *pag);
510
511 void xfs_end_io(struct work_struct *work);
512
513 int xfs_ilock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
514 void xfs_iunlock2_io_mmap(struct xfs_inode *ip1, struct xfs_inode *ip2);
515
516 #endif /* __XFS_INODE_H__ */
517