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 /* operations vectors */
41 const struct xfs_dir_ops *d_ops; /* directory ops vector */
42
43 /* Transaction and locking information. */
44 struct xfs_inode_log_item *i_itemp; /* logging information */
45 mrlock_t i_lock; /* inode lock */
46 mrlock_t i_mmaplock; /* inode mmap IO lock */
47 atomic_t i_pincount; /* inode pin count */
48
49 /*
50 * Bitsets of inode metadata that have been checked and/or are sick.
51 * Callers must hold i_flags_lock before accessing this field.
52 */
53 uint16_t i_checked;
54 uint16_t i_sick;
55
56 spinlock_t i_flags_lock; /* inode i_flags lock */
57 /* Miscellaneous state. */
58 unsigned long i_flags; /* see defined flags below */
59 uint64_t i_delayed_blks; /* count of delay alloc blks */
60
61 struct xfs_icdinode i_d; /* most of ondisk inode */
62
63 xfs_extnum_t i_cnextents; /* # of extents in cow fork */
64 unsigned int i_cformat; /* format of cow fork */
65
66 /* VFS inode */
67 struct inode i_vnode; /* embedded VFS inode */
68
69 /* pending io completions */
70 spinlock_t i_ioend_lock;
71 struct work_struct i_ioend_work;
72 struct list_head i_ioend_list;
73 } xfs_inode_t;
74
75 /* Convert from vfs inode to xfs inode */
XFS_I(struct inode * inode)76 static inline struct xfs_inode *XFS_I(struct inode *inode)
77 {
78 return container_of(inode, struct xfs_inode, i_vnode);
79 }
80
81 /* convert from xfs inode to vfs inode */
VFS_I(struct xfs_inode * ip)82 static inline struct inode *VFS_I(struct xfs_inode *ip)
83 {
84 return &ip->i_vnode;
85 }
86
87 /*
88 * For regular files we only update the on-disk filesize when actually
89 * writing data back to disk. Until then only the copy in the VFS inode
90 * is uptodate.
91 */
XFS_ISIZE(struct xfs_inode * ip)92 static inline xfs_fsize_t XFS_ISIZE(struct xfs_inode *ip)
93 {
94 if (S_ISREG(VFS_I(ip)->i_mode))
95 return i_size_read(VFS_I(ip));
96 return ip->i_d.di_size;
97 }
98
99 /*
100 * If this I/O goes past the on-disk inode size update it unless it would
101 * be past the current in-core inode size.
102 */
103 static inline xfs_fsize_t
xfs_new_eof(struct xfs_inode * ip,xfs_fsize_t new_size)104 xfs_new_eof(struct xfs_inode *ip, xfs_fsize_t new_size)
105 {
106 xfs_fsize_t i_size = i_size_read(VFS_I(ip));
107
108 if (new_size > i_size || new_size < 0)
109 new_size = i_size;
110 return new_size > ip->i_d.di_size ? new_size : 0;
111 }
112
113 /*
114 * i_flags helper functions
115 */
116 static inline void
__xfs_iflags_set(xfs_inode_t * ip,unsigned short flags)117 __xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
118 {
119 ip->i_flags |= flags;
120 }
121
122 static inline void
xfs_iflags_set(xfs_inode_t * ip,unsigned short flags)123 xfs_iflags_set(xfs_inode_t *ip, unsigned short flags)
124 {
125 spin_lock(&ip->i_flags_lock);
126 __xfs_iflags_set(ip, flags);
127 spin_unlock(&ip->i_flags_lock);
128 }
129
130 static inline void
xfs_iflags_clear(xfs_inode_t * ip,unsigned short flags)131 xfs_iflags_clear(xfs_inode_t *ip, unsigned short flags)
132 {
133 spin_lock(&ip->i_flags_lock);
134 ip->i_flags &= ~flags;
135 spin_unlock(&ip->i_flags_lock);
136 }
137
138 static inline int
__xfs_iflags_test(xfs_inode_t * ip,unsigned short flags)139 __xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
140 {
141 return (ip->i_flags & flags);
142 }
143
144 static inline int
xfs_iflags_test(xfs_inode_t * ip,unsigned short flags)145 xfs_iflags_test(xfs_inode_t *ip, unsigned short flags)
146 {
147 int ret;
148 spin_lock(&ip->i_flags_lock);
149 ret = __xfs_iflags_test(ip, flags);
150 spin_unlock(&ip->i_flags_lock);
151 return ret;
152 }
153
154 static inline int
xfs_iflags_test_and_clear(xfs_inode_t * ip,unsigned short flags)155 xfs_iflags_test_and_clear(xfs_inode_t *ip, unsigned short flags)
156 {
157 int ret;
158
159 spin_lock(&ip->i_flags_lock);
160 ret = ip->i_flags & flags;
161 if (ret)
162 ip->i_flags &= ~flags;
163 spin_unlock(&ip->i_flags_lock);
164 return ret;
165 }
166
167 static inline int
xfs_iflags_test_and_set(xfs_inode_t * ip,unsigned short flags)168 xfs_iflags_test_and_set(xfs_inode_t *ip, unsigned short flags)
169 {
170 int ret;
171
172 spin_lock(&ip->i_flags_lock);
173 ret = ip->i_flags & flags;
174 if (!ret)
175 ip->i_flags |= flags;
176 spin_unlock(&ip->i_flags_lock);
177 return ret;
178 }
179
180 static inline prid_t
xfs_get_initial_prid(struct xfs_inode * dp)181 xfs_get_initial_prid(struct xfs_inode *dp)
182 {
183 if (dp->i_d.di_flags & XFS_DIFLAG_PROJINHERIT)
184 return dp->i_d.di_projid;
185
186 return XFS_PROJID_DEFAULT;
187 }
188
xfs_is_reflink_inode(struct xfs_inode * ip)189 static inline bool xfs_is_reflink_inode(struct xfs_inode *ip)
190 {
191 return ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK;
192 }
193
194 /*
195 * Check if an inode has any data in the COW fork. This might be often false
196 * even for inodes with the reflink flag when there is no pending COW operation.
197 */
xfs_inode_has_cow_data(struct xfs_inode * ip)198 static inline bool xfs_inode_has_cow_data(struct xfs_inode *ip)
199 {
200 return ip->i_cowfp && ip->i_cowfp->if_bytes;
201 }
202
203 /*
204 * In-core inode flags.
205 */
206 #define XFS_IRECLAIM (1 << 0) /* started reclaiming this inode */
207 #define XFS_ISTALE (1 << 1) /* inode has been staled */
208 #define XFS_IRECLAIMABLE (1 << 2) /* inode can be reclaimed */
209 #define __XFS_INEW_BIT 3 /* inode has just been allocated */
210 #define XFS_INEW (1 << __XFS_INEW_BIT)
211 #define XFS_ITRUNCATED (1 << 5) /* truncated down so flush-on-close */
212 #define XFS_IDIRTY_RELEASE (1 << 6) /* dirty release already seen */
213 #define __XFS_IFLOCK_BIT 7 /* inode is being flushed right now */
214 #define XFS_IFLOCK (1 << __XFS_IFLOCK_BIT)
215 #define __XFS_IPINNED_BIT 8 /* wakeup key for zero pin count */
216 #define XFS_IPINNED (1 << __XFS_IPINNED_BIT)
217 #define XFS_IDONTCACHE (1 << 9) /* don't cache the inode long term */
218 #define XFS_IEOFBLOCKS (1 << 10)/* has the preallocblocks tag set */
219 /*
220 * If this unlinked inode is in the middle of recovery, don't let drop_inode
221 * truncate and free the inode. This can happen if we iget the inode during
222 * log recovery to replay a bmap operation on the inode.
223 */
224 #define XFS_IRECOVERY (1 << 11)
225 #define XFS_ICOWBLOCKS (1 << 12)/* has the cowblocks tag set */
226
227 /*
228 * Per-lifetime flags need to be reset when re-using a reclaimable inode during
229 * inode lookup. This prevents unintended behaviour on the new inode from
230 * ocurring.
231 */
232 #define XFS_IRECLAIM_RESET_FLAGS \
233 (XFS_IRECLAIMABLE | XFS_IRECLAIM | \
234 XFS_IDIRTY_RELEASE | XFS_ITRUNCATED)
235
236 /*
237 * Synchronize processes attempting to flush the in-core inode back to disk.
238 */
239
xfs_isiflocked(struct xfs_inode * ip)240 static inline int xfs_isiflocked(struct xfs_inode *ip)
241 {
242 return xfs_iflags_test(ip, XFS_IFLOCK);
243 }
244
245 extern void __xfs_iflock(struct xfs_inode *ip);
246
xfs_iflock_nowait(struct xfs_inode * ip)247 static inline int xfs_iflock_nowait(struct xfs_inode *ip)
248 {
249 return !xfs_iflags_test_and_set(ip, XFS_IFLOCK);
250 }
251
xfs_iflock(struct xfs_inode * ip)252 static inline void xfs_iflock(struct xfs_inode *ip)
253 {
254 if (!xfs_iflock_nowait(ip))
255 __xfs_iflock(ip);
256 }
257
xfs_ifunlock(struct xfs_inode * ip)258 static inline void xfs_ifunlock(struct xfs_inode *ip)
259 {
260 ASSERT(xfs_isiflocked(ip));
261 xfs_iflags_clear(ip, XFS_IFLOCK);
262 smp_mb();
263 wake_up_bit(&ip->i_flags, __XFS_IFLOCK_BIT);
264 }
265
266 /*
267 * Flags for inode locking.
268 * Bit ranges: 1<<1 - 1<<16-1 -- iolock/ilock modes (bitfield)
269 * 1<<16 - 1<<32-1 -- lockdep annotation (integers)
270 */
271 #define XFS_IOLOCK_EXCL (1<<0)
272 #define XFS_IOLOCK_SHARED (1<<1)
273 #define XFS_ILOCK_EXCL (1<<2)
274 #define XFS_ILOCK_SHARED (1<<3)
275 #define XFS_MMAPLOCK_EXCL (1<<4)
276 #define XFS_MMAPLOCK_SHARED (1<<5)
277
278 #define XFS_LOCK_MASK (XFS_IOLOCK_EXCL | XFS_IOLOCK_SHARED \
279 | XFS_ILOCK_EXCL | XFS_ILOCK_SHARED \
280 | XFS_MMAPLOCK_EXCL | XFS_MMAPLOCK_SHARED)
281
282 #define XFS_LOCK_FLAGS \
283 { XFS_IOLOCK_EXCL, "IOLOCK_EXCL" }, \
284 { XFS_IOLOCK_SHARED, "IOLOCK_SHARED" }, \
285 { XFS_ILOCK_EXCL, "ILOCK_EXCL" }, \
286 { XFS_ILOCK_SHARED, "ILOCK_SHARED" }, \
287 { XFS_MMAPLOCK_EXCL, "MMAPLOCK_EXCL" }, \
288 { XFS_MMAPLOCK_SHARED, "MMAPLOCK_SHARED" }
289
290
291 /*
292 * Flags for lockdep annotations.
293 *
294 * XFS_LOCK_PARENT - for directory operations that require locking a
295 * parent directory inode and a child entry inode. IOLOCK requires nesting,
296 * MMAPLOCK does not support this class, ILOCK requires a single subclass
297 * to differentiate parent from child.
298 *
299 * XFS_LOCK_RTBITMAP/XFS_LOCK_RTSUM - the realtime device bitmap and summary
300 * inodes do not participate in the normal lock order, and thus have their
301 * own subclasses.
302 *
303 * XFS_LOCK_INUMORDER - for locking several inodes at the some time
304 * with xfs_lock_inodes(). This flag is used as the starting subclass
305 * and each subsequent lock acquired will increment the subclass by one.
306 * However, MAX_LOCKDEP_SUBCLASSES == 8, which means we are greatly
307 * limited to the subclasses we can represent via nesting. We need at least
308 * 5 inodes nest depth for the ILOCK through rename, and we also have to support
309 * XFS_ILOCK_PARENT, which gives 6 subclasses. Then we have XFS_ILOCK_RTBITMAP
310 * and XFS_ILOCK_RTSUM, which are another 2 unique subclasses, so that's all
311 * 8 subclasses supported by lockdep.
312 *
313 * This also means we have to number the sub-classes in the lowest bits of
314 * the mask we keep, and we have to ensure we never exceed 3 bits of lockdep
315 * mask and we can't use bit-masking to build the subclasses. What a mess.
316 *
317 * Bit layout:
318 *
319 * Bit Lock Region
320 * 16-19 XFS_IOLOCK_SHIFT dependencies
321 * 20-23 XFS_MMAPLOCK_SHIFT dependencies
322 * 24-31 XFS_ILOCK_SHIFT dependencies
323 *
324 * IOLOCK values
325 *
326 * 0-3 subclass value
327 * 4-7 unused
328 *
329 * MMAPLOCK values
330 *
331 * 0-3 subclass value
332 * 4-7 unused
333 *
334 * ILOCK values
335 * 0-4 subclass values
336 * 5 PARENT subclass (not nestable)
337 * 6 RTBITMAP subclass (not nestable)
338 * 7 RTSUM subclass (not nestable)
339 *
340 */
341 #define XFS_IOLOCK_SHIFT 16
342 #define XFS_IOLOCK_MAX_SUBCLASS 3
343 #define XFS_IOLOCK_DEP_MASK 0x000f0000
344
345 #define XFS_MMAPLOCK_SHIFT 20
346 #define XFS_MMAPLOCK_NUMORDER 0
347 #define XFS_MMAPLOCK_MAX_SUBCLASS 3
348 #define XFS_MMAPLOCK_DEP_MASK 0x00f00000
349
350 #define XFS_ILOCK_SHIFT 24
351 #define XFS_ILOCK_PARENT_VAL 5
352 #define XFS_ILOCK_MAX_SUBCLASS (XFS_ILOCK_PARENT_VAL - 1)
353 #define XFS_ILOCK_RTBITMAP_VAL 6
354 #define XFS_ILOCK_RTSUM_VAL 7
355 #define XFS_ILOCK_DEP_MASK 0xff000000
356 #define XFS_ILOCK_PARENT (XFS_ILOCK_PARENT_VAL << XFS_ILOCK_SHIFT)
357 #define XFS_ILOCK_RTBITMAP (XFS_ILOCK_RTBITMAP_VAL << XFS_ILOCK_SHIFT)
358 #define XFS_ILOCK_RTSUM (XFS_ILOCK_RTSUM_VAL << XFS_ILOCK_SHIFT)
359
360 #define XFS_LOCK_SUBCLASS_MASK (XFS_IOLOCK_DEP_MASK | \
361 XFS_MMAPLOCK_DEP_MASK | \
362 XFS_ILOCK_DEP_MASK)
363
364 #define XFS_IOLOCK_DEP(flags) (((flags) & XFS_IOLOCK_DEP_MASK) \
365 >> XFS_IOLOCK_SHIFT)
366 #define XFS_MMAPLOCK_DEP(flags) (((flags) & XFS_MMAPLOCK_DEP_MASK) \
367 >> XFS_MMAPLOCK_SHIFT)
368 #define XFS_ILOCK_DEP(flags) (((flags) & XFS_ILOCK_DEP_MASK) \
369 >> XFS_ILOCK_SHIFT)
370
371 /*
372 * Layouts are broken in the BREAK_WRITE case to ensure that
373 * layout-holders do not collide with local writes. Additionally,
374 * layouts are broken in the BREAK_UNMAP case to make sure the
375 * layout-holder has a consistent view of the file's extent map. While
376 * BREAK_WRITE breaks can be satisfied by recalling FL_LAYOUT leases,
377 * BREAK_UNMAP breaks additionally require waiting for busy dax-pages to
378 * go idle.
379 */
380 enum layout_break_reason {
381 BREAK_WRITE,
382 BREAK_UNMAP,
383 };
384
385 /*
386 * For multiple groups support: if S_ISGID bit is set in the parent
387 * directory, group of new file is set to that of the parent, and
388 * new subdirectory gets S_ISGID bit from parent.
389 */
390 #define XFS_INHERIT_GID(pip) \
391 (((pip)->i_mount->m_flags & XFS_MOUNT_GRPID) || \
392 (VFS_I(pip)->i_mode & S_ISGID))
393
394 int xfs_release(struct xfs_inode *ip);
395 void xfs_inactive(struct xfs_inode *ip);
396 int xfs_lookup(struct xfs_inode *dp, struct xfs_name *name,
397 struct xfs_inode **ipp, struct xfs_name *ci_name);
398 int xfs_create(struct xfs_inode *dp, struct xfs_name *name,
399 umode_t mode, dev_t rdev, struct xfs_inode **ipp);
400 int xfs_create_tmpfile(struct xfs_inode *dp, umode_t mode,
401 struct xfs_inode **ipp);
402 int xfs_remove(struct xfs_inode *dp, struct xfs_name *name,
403 struct xfs_inode *ip);
404 int xfs_link(struct xfs_inode *tdp, struct xfs_inode *sip,
405 struct xfs_name *target_name);
406 int xfs_rename(struct xfs_inode *src_dp, struct xfs_name *src_name,
407 struct xfs_inode *src_ip, struct xfs_inode *target_dp,
408 struct xfs_name *target_name,
409 struct xfs_inode *target_ip, unsigned int flags);
410
411 void xfs_ilock(xfs_inode_t *, uint);
412 int xfs_ilock_nowait(xfs_inode_t *, uint);
413 void xfs_iunlock(xfs_inode_t *, uint);
414 void xfs_ilock_demote(xfs_inode_t *, uint);
415 int xfs_isilocked(xfs_inode_t *, uint);
416 uint xfs_ilock_data_map_shared(struct xfs_inode *);
417 uint xfs_ilock_attr_map_shared(struct xfs_inode *);
418
419 uint xfs_ip2xflags(struct xfs_inode *);
420 int xfs_ifree(struct xfs_trans *, struct xfs_inode *);
421 int xfs_itruncate_extents_flags(struct xfs_trans **,
422 struct xfs_inode *, int, xfs_fsize_t, int);
423 void xfs_iext_realloc(xfs_inode_t *, int, int);
424
425 int xfs_log_force_inode(struct xfs_inode *ip);
426 void xfs_iunpin_wait(xfs_inode_t *);
427 #define xfs_ipincount(ip) ((unsigned int) atomic_read(&ip->i_pincount))
428
429 int xfs_iflush(struct xfs_inode *, struct xfs_buf **);
430 void xfs_lock_two_inodes(struct xfs_inode *ip0, uint ip0_mode,
431 struct xfs_inode *ip1, uint ip1_mode);
432
433 xfs_extlen_t xfs_get_extsz_hint(struct xfs_inode *ip);
434 xfs_extlen_t xfs_get_cowextsz_hint(struct xfs_inode *ip);
435
436 int xfs_dir_ialloc(struct xfs_trans **, struct xfs_inode *, umode_t,
437 xfs_nlink_t, dev_t, prid_t,
438 struct xfs_inode **);
439
440 static inline int
xfs_itruncate_extents(struct xfs_trans ** tpp,struct xfs_inode * ip,int whichfork,xfs_fsize_t new_size)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
467 /*
468 * When setting up a newly allocated inode, we need to call
469 * xfs_finish_inode_setup() once the inode is fully instantiated at
470 * the VFS level to prevent the rest of the world seeing the inode
471 * before we've completed instantiation. Otherwise we can do it
472 * the moment the inode lookup is complete.
473 */
xfs_finish_inode_setup(struct xfs_inode * ip)474 static inline void xfs_finish_inode_setup(struct xfs_inode *ip)
475 {
476 xfs_iflags_clear(ip, XFS_INEW);
477 barrier();
478 unlock_new_inode(VFS_I(ip));
479 wake_up_bit(&ip->i_flags, __XFS_INEW_BIT);
480 }
481
xfs_setup_existing_inode(struct xfs_inode * ip)482 static inline void xfs_setup_existing_inode(struct xfs_inode *ip)
483 {
484 xfs_setup_inode(ip);
485 xfs_setup_iops(ip);
486 xfs_finish_inode_setup(ip);
487 }
488
489 void xfs_irele(struct xfs_inode *ip);
490
491 extern struct kmem_zone *xfs_inode_zone;
492
493 /* The default CoW extent size hint. */
494 #define XFS_DEFAULT_COWEXTSZ_HINT 32
495
496 bool xfs_inode_verify_forks(struct xfs_inode *ip);
497
498 int xfs_iunlink_init(struct xfs_perag *pag);
499 void xfs_iunlink_destroy(struct xfs_perag *pag);
500
501 void xfs_end_io(struct work_struct *work);
502
503 #endif /* __XFS_INODE_H__ */
504