xref: /linux/include/linux/fs.h (revision 11068e0b)
1 /* SPDX-License-Identifier: GPL-2.0 */
2 #ifndef _LINUX_FS_H
3 #define _LINUX_FS_H
4 
5 #include <linux/linkage.h>
6 #include <linux/wait_bit.h>
7 #include <linux/kdev_t.h>
8 #include <linux/dcache.h>
9 #include <linux/path.h>
10 #include <linux/stat.h>
11 #include <linux/cache.h>
12 #include <linux/list.h>
13 #include <linux/list_lru.h>
14 #include <linux/llist.h>
15 #include <linux/radix-tree.h>
16 #include <linux/xarray.h>
17 #include <linux/rbtree.h>
18 #include <linux/init.h>
19 #include <linux/pid.h>
20 #include <linux/bug.h>
21 #include <linux/mutex.h>
22 #include <linux/rwsem.h>
23 #include <linux/mm_types.h>
24 #include <linux/capability.h>
25 #include <linux/semaphore.h>
26 #include <linux/fcntl.h>
27 #include <linux/rculist_bl.h>
28 #include <linux/atomic.h>
29 #include <linux/shrinker.h>
30 #include <linux/migrate_mode.h>
31 #include <linux/uidgid.h>
32 #include <linux/lockdep.h>
33 #include <linux/percpu-rwsem.h>
34 #include <linux/workqueue.h>
35 #include <linux/delayed_call.h>
36 #include <linux/uuid.h>
37 #include <linux/errseq.h>
38 #include <linux/ioprio.h>
39 #include <linux/fs_types.h>
40 #include <linux/build_bug.h>
41 #include <linux/stddef.h>
42 #include <linux/mount.h>
43 #include <linux/cred.h>
44 #include <linux/mnt_idmapping.h>
45 #include <linux/slab.h>
46 #include <linux/maple_tree.h>
47 #include <linux/rw_hint.h>
48 
49 #include <asm/byteorder.h>
50 #include <uapi/linux/fs.h>
51 
52 struct backing_dev_info;
53 struct bdi_writeback;
54 struct bio;
55 struct io_comp_batch;
56 struct export_operations;
57 struct fiemap_extent_info;
58 struct hd_geometry;
59 struct iovec;
60 struct kiocb;
61 struct kobject;
62 struct pipe_inode_info;
63 struct poll_table_struct;
64 struct kstatfs;
65 struct vm_area_struct;
66 struct vfsmount;
67 struct cred;
68 struct swap_info_struct;
69 struct seq_file;
70 struct workqueue_struct;
71 struct iov_iter;
72 struct fscrypt_inode_info;
73 struct fscrypt_operations;
74 struct fsverity_info;
75 struct fsverity_operations;
76 struct fsnotify_mark_connector;
77 struct fsnotify_sb_info;
78 struct fs_context;
79 struct fs_parameter_spec;
80 struct fileattr;
81 struct iomap_ops;
82 
83 extern void __init inode_init(void);
84 extern void __init inode_init_early(void);
85 extern void __init files_init(void);
86 extern void __init files_maxfiles_init(void);
87 
88 extern unsigned long get_max_files(void);
89 extern unsigned int sysctl_nr_open;
90 
91 typedef __kernel_rwf_t rwf_t;
92 
93 struct buffer_head;
94 typedef int (get_block_t)(struct inode *inode, sector_t iblock,
95 			struct buffer_head *bh_result, int create);
96 typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
97 			ssize_t bytes, void *private);
98 
99 #define MAY_EXEC		0x00000001
100 #define MAY_WRITE		0x00000002
101 #define MAY_READ		0x00000004
102 #define MAY_APPEND		0x00000008
103 #define MAY_ACCESS		0x00000010
104 #define MAY_OPEN		0x00000020
105 #define MAY_CHDIR		0x00000040
106 /* called from RCU mode, don't block */
107 #define MAY_NOT_BLOCK		0x00000080
108 
109 /*
110  * flags in file.f_mode.  Note that FMODE_READ and FMODE_WRITE must correspond
111  * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
112  */
113 
114 /* file is open for reading */
115 #define FMODE_READ		((__force fmode_t)(1 << 0))
116 /* file is open for writing */
117 #define FMODE_WRITE		((__force fmode_t)(1 << 1))
118 /* file is seekable */
119 #define FMODE_LSEEK		((__force fmode_t)(1 << 2))
120 /* file can be accessed using pread */
121 #define FMODE_PREAD		((__force fmode_t)(1 << 3))
122 /* file can be accessed using pwrite */
123 #define FMODE_PWRITE		((__force fmode_t)(1 << 4))
124 /* File is opened for execution with sys_execve / sys_uselib */
125 #define FMODE_EXEC		((__force fmode_t)(1 << 5))
126 /* File writes are restricted (block device specific) */
127 #define FMODE_WRITE_RESTRICTED	((__force fmode_t)(1 << 6))
128 /* File supports atomic writes */
129 #define FMODE_CAN_ATOMIC_WRITE	((__force fmode_t)(1 << 7))
130 
131 /* FMODE_* bit 8 */
132 
133 /* 32bit hashes as llseek() offset (for directories) */
134 #define FMODE_32BITHASH         ((__force fmode_t)(1 << 9))
135 /* 64bit hashes as llseek() offset (for directories) */
136 #define FMODE_64BITHASH         ((__force fmode_t)(1 << 10))
137 
138 /*
139  * Don't update ctime and mtime.
140  *
141  * Currently a special hack for the XFS open_by_handle ioctl, but we'll
142  * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
143  */
144 #define FMODE_NOCMTIME		((__force fmode_t)(1 << 11))
145 
146 /* Expect random access pattern */
147 #define FMODE_RANDOM		((__force fmode_t)(1 << 12))
148 
149 /* FMODE_* bit 13 */
150 
151 /* File is opened with O_PATH; almost nothing can be done with it */
152 #define FMODE_PATH		((__force fmode_t)(1 << 14))
153 
154 /* File needs atomic accesses to f_pos */
155 #define FMODE_ATOMIC_POS	((__force fmode_t)(1 << 15))
156 /* Write access to underlying fs */
157 #define FMODE_WRITER		((__force fmode_t)(1 << 16))
158 /* Has read method(s) */
159 #define FMODE_CAN_READ          ((__force fmode_t)(1 << 17))
160 /* Has write method(s) */
161 #define FMODE_CAN_WRITE         ((__force fmode_t)(1 << 18))
162 
163 #define FMODE_OPENED		((__force fmode_t)(1 << 19))
164 #define FMODE_CREATED		((__force fmode_t)(1 << 20))
165 
166 /* File is stream-like */
167 #define FMODE_STREAM		((__force fmode_t)(1 << 21))
168 
169 /* File supports DIRECT IO */
170 #define	FMODE_CAN_ODIRECT	((__force fmode_t)(1 << 22))
171 
172 #define	FMODE_NOREUSE		((__force fmode_t)(1 << 23))
173 
174 /* FMODE_* bit 24 */
175 
176 /* File is embedded in backing_file object */
177 #define FMODE_BACKING		((__force fmode_t)(1 << 25))
178 
179 /* File was opened by fanotify and shouldn't generate fanotify events */
180 #define FMODE_NONOTIFY		((__force fmode_t)(1 << 26))
181 
182 /* File is capable of returning -EAGAIN if I/O will block */
183 #define FMODE_NOWAIT		((__force fmode_t)(1 << 27))
184 
185 /* File represents mount that needs unmounting */
186 #define FMODE_NEED_UNMOUNT	((__force fmode_t)(1 << 28))
187 
188 /* File does not contribute to nr_files count */
189 #define FMODE_NOACCOUNT		((__force fmode_t)(1 << 29))
190 
191 /*
192  * Attribute flags.  These should be or-ed together to figure out what
193  * has been changed!
194  */
195 #define ATTR_MODE	(1 << 0)
196 #define ATTR_UID	(1 << 1)
197 #define ATTR_GID	(1 << 2)
198 #define ATTR_SIZE	(1 << 3)
199 #define ATTR_ATIME	(1 << 4)
200 #define ATTR_MTIME	(1 << 5)
201 #define ATTR_CTIME	(1 << 6)
202 #define ATTR_ATIME_SET	(1 << 7)
203 #define ATTR_MTIME_SET	(1 << 8)
204 #define ATTR_FORCE	(1 << 9) /* Not a change, but a change it */
205 #define ATTR_KILL_SUID	(1 << 11)
206 #define ATTR_KILL_SGID	(1 << 12)
207 #define ATTR_FILE	(1 << 13)
208 #define ATTR_KILL_PRIV	(1 << 14)
209 #define ATTR_OPEN	(1 << 15) /* Truncating from open(O_TRUNC) */
210 #define ATTR_TIMES_SET	(1 << 16)
211 #define ATTR_TOUCH	(1 << 17)
212 #define ATTR_DELEG	(1 << 18) /* Delegated attrs. Don't break write delegations */
213 
214 /*
215  * Whiteout is represented by a char device.  The following constants define the
216  * mode and device number to use.
217  */
218 #define WHITEOUT_MODE 0
219 #define WHITEOUT_DEV 0
220 
221 /*
222  * This is the Inode Attributes structure, used for notify_change().  It
223  * uses the above definitions as flags, to know which values have changed.
224  * Also, in this manner, a Filesystem can look at only the values it cares
225  * about.  Basically, these are the attributes that the VFS layer can
226  * request to change from the FS layer.
227  *
228  * Derek Atkins <warlord@MIT.EDU> 94-10-20
229  */
230 struct iattr {
231 	unsigned int	ia_valid;
232 	umode_t		ia_mode;
233 	/*
234 	 * The two anonymous unions wrap structures with the same member.
235 	 *
236 	 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which
237 	 * are a dedicated type requiring the filesystem to use the dedicated
238 	 * helpers. Other filesystem can continue to use ia_{g,u}id until they
239 	 * have been ported.
240 	 *
241 	 * They always contain the same value. In other words FS_ALLOW_IDMAP
242 	 * pass down the same value on idmapped mounts as they would on regular
243 	 * mounts.
244 	 */
245 	union {
246 		kuid_t		ia_uid;
247 		vfsuid_t	ia_vfsuid;
248 	};
249 	union {
250 		kgid_t		ia_gid;
251 		vfsgid_t	ia_vfsgid;
252 	};
253 	loff_t		ia_size;
254 	struct timespec64 ia_atime;
255 	struct timespec64 ia_mtime;
256 	struct timespec64 ia_ctime;
257 
258 	/*
259 	 * Not an attribute, but an auxiliary info for filesystems wanting to
260 	 * implement an ftruncate() like method.  NOTE: filesystem should
261 	 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
262 	 */
263 	struct file	*ia_file;
264 };
265 
266 /*
267  * Includes for diskquotas.
268  */
269 #include <linux/quota.h>
270 
271 /*
272  * Maximum number of layers of fs stack.  Needs to be limited to
273  * prevent kernel stack overflow
274  */
275 #define FILESYSTEM_MAX_STACK_DEPTH 2
276 
277 /**
278  * enum positive_aop_returns - aop return codes with specific semantics
279  *
280  * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
281  * 			    completed, that the page is still locked, and
282  * 			    should be considered active.  The VM uses this hint
283  * 			    to return the page to the active list -- it won't
284  * 			    be a candidate for writeback again in the near
285  * 			    future.  Other callers must be careful to unlock
286  * 			    the page if they get this return.  Returned by
287  * 			    writepage();
288  *
289  * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
290  *  			unlocked it and the page might have been truncated.
291  *  			The caller should back up to acquiring a new page and
292  *  			trying again.  The aop will be taking reasonable
293  *  			precautions not to livelock.  If the caller held a page
294  *  			reference, it should drop it before retrying.  Returned
295  *  			by read_folio().
296  *
297  * address_space_operation functions return these large constants to indicate
298  * special semantics to the caller.  These are much larger than the bytes in a
299  * page to allow for functions that return the number of bytes operated on in a
300  * given page.
301  */
302 
303 enum positive_aop_returns {
304 	AOP_WRITEPAGE_ACTIVATE	= 0x80000,
305 	AOP_TRUNCATED_PAGE	= 0x80001,
306 };
307 
308 /*
309  * oh the beauties of C type declarations.
310  */
311 struct page;
312 struct address_space;
313 struct writeback_control;
314 struct readahead_control;
315 
316 /* Match RWF_* bits to IOCB bits */
317 #define IOCB_HIPRI		(__force int) RWF_HIPRI
318 #define IOCB_DSYNC		(__force int) RWF_DSYNC
319 #define IOCB_SYNC		(__force int) RWF_SYNC
320 #define IOCB_NOWAIT		(__force int) RWF_NOWAIT
321 #define IOCB_APPEND		(__force int) RWF_APPEND
322 #define IOCB_ATOMIC		(__force int) RWF_ATOMIC
323 
324 /* non-RWF related bits - start at 16 */
325 #define IOCB_EVENTFD		(1 << 16)
326 #define IOCB_DIRECT		(1 << 17)
327 #define IOCB_WRITE		(1 << 18)
328 /* iocb->ki_waitq is valid */
329 #define IOCB_WAITQ		(1 << 19)
330 #define IOCB_NOIO		(1 << 20)
331 /* can use bio alloc cache */
332 #define IOCB_ALLOC_CACHE	(1 << 21)
333 /*
334  * IOCB_DIO_CALLER_COMP can be set by the iocb owner, to indicate that the
335  * iocb completion can be passed back to the owner for execution from a safe
336  * context rather than needing to be punted through a workqueue. If this
337  * flag is set, the bio completion handling may set iocb->dio_complete to a
338  * handler function and iocb->private to context information for that handler.
339  * The issuer should call the handler with that context information from task
340  * context to complete the processing of the iocb. Note that while this
341  * provides a task context for the dio_complete() callback, it should only be
342  * used on the completion side for non-IO generating completions. It's fine to
343  * call blocking functions from this callback, but they should not wait for
344  * unrelated IO (like cache flushing, new IO generation, etc).
345  */
346 #define IOCB_DIO_CALLER_COMP	(1 << 22)
347 /* kiocb is a read or write operation submitted by fs/aio.c. */
348 #define IOCB_AIO_RW		(1 << 23)
349 
350 /* for use in trace events */
351 #define TRACE_IOCB_STRINGS \
352 	{ IOCB_HIPRI,		"HIPRI" }, \
353 	{ IOCB_DSYNC,		"DSYNC" }, \
354 	{ IOCB_SYNC,		"SYNC" }, \
355 	{ IOCB_NOWAIT,		"NOWAIT" }, \
356 	{ IOCB_APPEND,		"APPEND" }, \
357 	{ IOCB_ATOMIC,		"ATOMIC"}, \
358 	{ IOCB_EVENTFD,		"EVENTFD"}, \
359 	{ IOCB_DIRECT,		"DIRECT" }, \
360 	{ IOCB_WRITE,		"WRITE" }, \
361 	{ IOCB_WAITQ,		"WAITQ" }, \
362 	{ IOCB_NOIO,		"NOIO" }, \
363 	{ IOCB_ALLOC_CACHE,	"ALLOC_CACHE" }, \
364 	{ IOCB_DIO_CALLER_COMP,	"CALLER_COMP" }
365 
366 struct kiocb {
367 	struct file		*ki_filp;
368 	loff_t			ki_pos;
369 	void (*ki_complete)(struct kiocb *iocb, long ret);
370 	void			*private;
371 	int			ki_flags;
372 	u16			ki_ioprio; /* See linux/ioprio.h */
373 	union {
374 		/*
375 		 * Only used for async buffered reads, where it denotes the
376 		 * page waitqueue associated with completing the read. Valid
377 		 * IFF IOCB_WAITQ is set.
378 		 */
379 		struct wait_page_queue	*ki_waitq;
380 		/*
381 		 * Can be used for O_DIRECT IO, where the completion handling
382 		 * is punted back to the issuer of the IO. May only be set
383 		 * if IOCB_DIO_CALLER_COMP is set by the issuer, and the issuer
384 		 * must then check for presence of this handler when ki_complete
385 		 * is invoked. The data passed in to this handler must be
386 		 * assigned to ->private when dio_complete is assigned.
387 		 */
388 		ssize_t (*dio_complete)(void *data);
389 	};
390 };
391 
is_sync_kiocb(struct kiocb * kiocb)392 static inline bool is_sync_kiocb(struct kiocb *kiocb)
393 {
394 	return kiocb->ki_complete == NULL;
395 }
396 
397 struct address_space_operations {
398 	int (*writepage)(struct page *page, struct writeback_control *wbc);
399 	int (*read_folio)(struct file *, struct folio *);
400 
401 	/* Write back some dirty pages from this mapping. */
402 	int (*writepages)(struct address_space *, struct writeback_control *);
403 
404 	/* Mark a folio dirty.  Return true if this dirtied it */
405 	bool (*dirty_folio)(struct address_space *, struct folio *);
406 
407 	void (*readahead)(struct readahead_control *);
408 
409 	int (*write_begin)(struct file *, struct address_space *mapping,
410 				loff_t pos, unsigned len,
411 				struct folio **foliop, void **fsdata);
412 	int (*write_end)(struct file *, struct address_space *mapping,
413 				loff_t pos, unsigned len, unsigned copied,
414 				struct folio *folio, void *fsdata);
415 
416 	/* Unfortunately this kludge is needed for FIBMAP. Don't use it */
417 	sector_t (*bmap)(struct address_space *, sector_t);
418 	void (*invalidate_folio) (struct folio *, size_t offset, size_t len);
419 	bool (*release_folio)(struct folio *, gfp_t);
420 	void (*free_folio)(struct folio *folio);
421 	ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
422 	/*
423 	 * migrate the contents of a folio to the specified target. If
424 	 * migrate_mode is MIGRATE_ASYNC, it must not block.
425 	 */
426 	int (*migrate_folio)(struct address_space *, struct folio *dst,
427 			struct folio *src, enum migrate_mode);
428 	int (*launder_folio)(struct folio *);
429 	bool (*is_partially_uptodate) (struct folio *, size_t from,
430 			size_t count);
431 	void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb);
432 	int (*error_remove_folio)(struct address_space *, struct folio *);
433 
434 	/* swapfile support */
435 	int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
436 				sector_t *span);
437 	void (*swap_deactivate)(struct file *file);
438 	int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
439 };
440 
441 extern const struct address_space_operations empty_aops;
442 
443 /**
444  * struct address_space - Contents of a cacheable, mappable object.
445  * @host: Owner, either the inode or the block_device.
446  * @i_pages: Cached pages.
447  * @invalidate_lock: Guards coherency between page cache contents and
448  *   file offset->disk block mappings in the filesystem during invalidates.
449  *   It is also used to block modification of page cache contents through
450  *   memory mappings.
451  * @gfp_mask: Memory allocation flags to use for allocating pages.
452  * @i_mmap_writable: Number of VM_SHARED, VM_MAYWRITE mappings.
453  * @nr_thps: Number of THPs in the pagecache (non-shmem only).
454  * @i_mmap: Tree of private and shared mappings.
455  * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
456  * @nrpages: Number of page entries, protected by the i_pages lock.
457  * @writeback_index: Writeback starts here.
458  * @a_ops: Methods.
459  * @flags: Error bits and flags (AS_*).
460  * @wb_err: The most recent error which has occurred.
461  * @i_private_lock: For use by the owner of the address_space.
462  * @i_private_list: For use by the owner of the address_space.
463  * @i_private_data: For use by the owner of the address_space.
464  */
465 struct address_space {
466 	struct inode		*host;
467 	struct xarray		i_pages;
468 	struct rw_semaphore	invalidate_lock;
469 	gfp_t			gfp_mask;
470 	atomic_t		i_mmap_writable;
471 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
472 	/* number of thp, only for non-shmem files */
473 	atomic_t		nr_thps;
474 #endif
475 	struct rb_root_cached	i_mmap;
476 	unsigned long		nrpages;
477 	pgoff_t			writeback_index;
478 	const struct address_space_operations *a_ops;
479 	unsigned long		flags;
480 	errseq_t		wb_err;
481 	spinlock_t		i_private_lock;
482 	struct list_head	i_private_list;
483 	struct rw_semaphore	i_mmap_rwsem;
484 	void *			i_private_data;
485 } __attribute__((aligned(sizeof(long)))) __randomize_layout;
486 	/*
487 	 * On most architectures that alignment is already the case; but
488 	 * must be enforced here for CRIS, to let the least significant bit
489 	 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
490 	 */
491 
492 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */
493 #define PAGECACHE_TAG_DIRTY	XA_MARK_0
494 #define PAGECACHE_TAG_WRITEBACK	XA_MARK_1
495 #define PAGECACHE_TAG_TOWRITE	XA_MARK_2
496 
497 /*
498  * Returns true if any of the pages in the mapping are marked with the tag.
499  */
mapping_tagged(struct address_space * mapping,xa_mark_t tag)500 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
501 {
502 	return xa_marked(&mapping->i_pages, tag);
503 }
504 
i_mmap_lock_write(struct address_space * mapping)505 static inline void i_mmap_lock_write(struct address_space *mapping)
506 {
507 	down_write(&mapping->i_mmap_rwsem);
508 }
509 
i_mmap_trylock_write(struct address_space * mapping)510 static inline int i_mmap_trylock_write(struct address_space *mapping)
511 {
512 	return down_write_trylock(&mapping->i_mmap_rwsem);
513 }
514 
i_mmap_unlock_write(struct address_space * mapping)515 static inline void i_mmap_unlock_write(struct address_space *mapping)
516 {
517 	up_write(&mapping->i_mmap_rwsem);
518 }
519 
i_mmap_trylock_read(struct address_space * mapping)520 static inline int i_mmap_trylock_read(struct address_space *mapping)
521 {
522 	return down_read_trylock(&mapping->i_mmap_rwsem);
523 }
524 
i_mmap_lock_read(struct address_space * mapping)525 static inline void i_mmap_lock_read(struct address_space *mapping)
526 {
527 	down_read(&mapping->i_mmap_rwsem);
528 }
529 
i_mmap_unlock_read(struct address_space * mapping)530 static inline void i_mmap_unlock_read(struct address_space *mapping)
531 {
532 	up_read(&mapping->i_mmap_rwsem);
533 }
534 
i_mmap_assert_locked(struct address_space * mapping)535 static inline void i_mmap_assert_locked(struct address_space *mapping)
536 {
537 	lockdep_assert_held(&mapping->i_mmap_rwsem);
538 }
539 
i_mmap_assert_write_locked(struct address_space * mapping)540 static inline void i_mmap_assert_write_locked(struct address_space *mapping)
541 {
542 	lockdep_assert_held_write(&mapping->i_mmap_rwsem);
543 }
544 
545 /*
546  * Might pages of this file be mapped into userspace?
547  */
mapping_mapped(struct address_space * mapping)548 static inline int mapping_mapped(struct address_space *mapping)
549 {
550 	return	!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
551 }
552 
553 /*
554  * Might pages of this file have been modified in userspace?
555  * Note that i_mmap_writable counts all VM_SHARED, VM_MAYWRITE vmas: do_mmap
556  * marks vma as VM_SHARED if it is shared, and the file was opened for
557  * writing i.e. vma may be mprotected writable even if now readonly.
558  *
559  * If i_mmap_writable is negative, no new writable mappings are allowed. You
560  * can only deny writable mappings, if none exists right now.
561  */
mapping_writably_mapped(struct address_space * mapping)562 static inline int mapping_writably_mapped(struct address_space *mapping)
563 {
564 	return atomic_read(&mapping->i_mmap_writable) > 0;
565 }
566 
mapping_map_writable(struct address_space * mapping)567 static inline int mapping_map_writable(struct address_space *mapping)
568 {
569 	return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
570 		0 : -EPERM;
571 }
572 
mapping_unmap_writable(struct address_space * mapping)573 static inline void mapping_unmap_writable(struct address_space *mapping)
574 {
575 	atomic_dec(&mapping->i_mmap_writable);
576 }
577 
mapping_deny_writable(struct address_space * mapping)578 static inline int mapping_deny_writable(struct address_space *mapping)
579 {
580 	return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
581 		0 : -EBUSY;
582 }
583 
mapping_allow_writable(struct address_space * mapping)584 static inline void mapping_allow_writable(struct address_space *mapping)
585 {
586 	atomic_inc(&mapping->i_mmap_writable);
587 }
588 
589 /*
590  * Use sequence counter to get consistent i_size on 32-bit processors.
591  */
592 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
593 #include <linux/seqlock.h>
594 #define __NEED_I_SIZE_ORDERED
595 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
596 #else
597 #define i_size_ordered_init(inode) do { } while (0)
598 #endif
599 
600 struct posix_acl;
601 #define ACL_NOT_CACHED ((void *)(-1))
602 /*
603  * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
604  * cache the ACL.  This also means that ->get_inode_acl() can be called in RCU
605  * mode with the LOOKUP_RCU flag.
606  */
607 #define ACL_DONT_CACHE ((void *)(-3))
608 
609 static inline struct posix_acl *
uncached_acl_sentinel(struct task_struct * task)610 uncached_acl_sentinel(struct task_struct *task)
611 {
612 	return (void *)task + 1;
613 }
614 
615 static inline bool
is_uncached_acl(struct posix_acl * acl)616 is_uncached_acl(struct posix_acl *acl)
617 {
618 	return (long)acl & 1;
619 }
620 
621 #define IOP_FASTPERM	0x0001
622 #define IOP_LOOKUP	0x0002
623 #define IOP_NOFOLLOW	0x0004
624 #define IOP_XATTR	0x0008
625 #define IOP_DEFAULT_READLINK	0x0010
626 
627 /*
628  * Keep mostly read-only and often accessed (especially for
629  * the RCU path lookup and 'stat' data) fields at the beginning
630  * of the 'struct inode'
631  */
632 struct inode {
633 	umode_t			i_mode;
634 	unsigned short		i_opflags;
635 	kuid_t			i_uid;
636 	kgid_t			i_gid;
637 	unsigned int		i_flags;
638 
639 #ifdef CONFIG_FS_POSIX_ACL
640 	struct posix_acl	*i_acl;
641 	struct posix_acl	*i_default_acl;
642 #endif
643 
644 	const struct inode_operations	*i_op;
645 	struct super_block	*i_sb;
646 	struct address_space	*i_mapping;
647 
648 #ifdef CONFIG_SECURITY
649 	void			*i_security;
650 #endif
651 
652 	/* Stat data, not accessed from path walking */
653 	unsigned long		i_ino;
654 	/*
655 	 * Filesystems may only read i_nlink directly.  They shall use the
656 	 * following functions for modification:
657 	 *
658 	 *    (set|clear|inc|drop)_nlink
659 	 *    inode_(inc|dec)_link_count
660 	 */
661 	union {
662 		const unsigned int i_nlink;
663 		unsigned int __i_nlink;
664 	};
665 	dev_t			i_rdev;
666 	loff_t			i_size;
667 	time64_t		i_atime_sec;
668 	time64_t		i_mtime_sec;
669 	time64_t		i_ctime_sec;
670 	u32			i_atime_nsec;
671 	u32			i_mtime_nsec;
672 	u32			i_ctime_nsec;
673 	u32			i_generation;
674 	spinlock_t		i_lock;	/* i_blocks, i_bytes, maybe i_size */
675 	unsigned short          i_bytes;
676 	u8			i_blkbits;
677 	enum rw_hint		i_write_hint;
678 	blkcnt_t		i_blocks;
679 
680 #ifdef __NEED_I_SIZE_ORDERED
681 	seqcount_t		i_size_seqcount;
682 #endif
683 
684 	/* Misc */
685 	u32			i_state;
686 	/* 32-bit hole */
687 	struct rw_semaphore	i_rwsem;
688 
689 	unsigned long		dirtied_when;	/* jiffies of first dirtying */
690 	unsigned long		dirtied_time_when;
691 
692 	struct hlist_node	i_hash;
693 	struct list_head	i_io_list;	/* backing dev IO list */
694 #ifdef CONFIG_CGROUP_WRITEBACK
695 	struct bdi_writeback	*i_wb;		/* the associated cgroup wb */
696 
697 	/* foreign inode detection, see wbc_detach_inode() */
698 	int			i_wb_frn_winner;
699 	u16			i_wb_frn_avg_time;
700 	u16			i_wb_frn_history;
701 #endif
702 	struct list_head	i_lru;		/* inode LRU list */
703 	struct list_head	i_sb_list;
704 	struct list_head	i_wb_list;	/* backing dev writeback list */
705 	union {
706 		struct hlist_head	i_dentry;
707 		struct rcu_head		i_rcu;
708 	};
709 	atomic64_t		i_version;
710 	atomic64_t		i_sequence; /* see futex */
711 	atomic_t		i_count;
712 	atomic_t		i_dio_count;
713 	atomic_t		i_writecount;
714 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
715 	atomic_t		i_readcount; /* struct files open RO */
716 #endif
717 	union {
718 		const struct file_operations	*i_fop;	/* former ->i_op->default_file_ops */
719 		void (*free_inode)(struct inode *);
720 	};
721 	struct file_lock_context	*i_flctx;
722 	struct address_space	i_data;
723 	struct list_head	i_devices;
724 	union {
725 		struct pipe_inode_info	*i_pipe;
726 		struct cdev		*i_cdev;
727 		char			*i_link;
728 		unsigned		i_dir_seq;
729 	};
730 
731 
732 #ifdef CONFIG_FSNOTIFY
733 	__u32			i_fsnotify_mask; /* all events this inode cares about */
734 	/* 32-bit hole reserved for expanding i_fsnotify_mask */
735 	struct fsnotify_mark_connector __rcu	*i_fsnotify_marks;
736 #endif
737 
738 #ifdef CONFIG_FS_ENCRYPTION
739 	struct fscrypt_inode_info	*i_crypt_info;
740 #endif
741 
742 #ifdef CONFIG_FS_VERITY
743 	struct fsverity_info	*i_verity_info;
744 #endif
745 
746 	void			*i_private; /* fs or device private pointer */
747 } __randomize_layout;
748 
749 /*
750  * Get bit address from inode->i_state to use with wait_var_event()
751  * infrastructre.
752  */
753 #define inode_state_wait_address(inode, bit) ((char *)&(inode)->i_state + (bit))
754 
755 struct wait_queue_head *inode_bit_waitqueue(struct wait_bit_queue_entry *wqe,
756 					    struct inode *inode, u32 bit);
757 
inode_wake_up_bit(struct inode * inode,u32 bit)758 static inline void inode_wake_up_bit(struct inode *inode, u32 bit)
759 {
760 	/* Caller is responsible for correct memory barriers. */
761 	wake_up_var(inode_state_wait_address(inode, bit));
762 }
763 
764 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
765 
i_blocksize(const struct inode * node)766 static inline unsigned int i_blocksize(const struct inode *node)
767 {
768 	return (1 << node->i_blkbits);
769 }
770 
inode_unhashed(struct inode * inode)771 static inline int inode_unhashed(struct inode *inode)
772 {
773 	return hlist_unhashed(&inode->i_hash);
774 }
775 
776 /*
777  * __mark_inode_dirty expects inodes to be hashed.  Since we don't
778  * want special inodes in the fileset inode space, we make them
779  * appear hashed, but do not put on any lists.  hlist_del()
780  * will work fine and require no locking.
781  */
inode_fake_hash(struct inode * inode)782 static inline void inode_fake_hash(struct inode *inode)
783 {
784 	hlist_add_fake(&inode->i_hash);
785 }
786 
787 /*
788  * inode->i_mutex nesting subclasses for the lock validator:
789  *
790  * 0: the object of the current VFS operation
791  * 1: parent
792  * 2: child/target
793  * 3: xattr
794  * 4: second non-directory
795  * 5: second parent (when locking independent directories in rename)
796  *
797  * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
798  * non-directories at once.
799  *
800  * The locking order between these classes is
801  * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
802  */
803 enum inode_i_mutex_lock_class
804 {
805 	I_MUTEX_NORMAL,
806 	I_MUTEX_PARENT,
807 	I_MUTEX_CHILD,
808 	I_MUTEX_XATTR,
809 	I_MUTEX_NONDIR2,
810 	I_MUTEX_PARENT2,
811 };
812 
inode_lock(struct inode * inode)813 static inline void inode_lock(struct inode *inode)
814 {
815 	down_write(&inode->i_rwsem);
816 }
817 
inode_unlock(struct inode * inode)818 static inline void inode_unlock(struct inode *inode)
819 {
820 	up_write(&inode->i_rwsem);
821 }
822 
inode_lock_shared(struct inode * inode)823 static inline void inode_lock_shared(struct inode *inode)
824 {
825 	down_read(&inode->i_rwsem);
826 }
827 
inode_unlock_shared(struct inode * inode)828 static inline void inode_unlock_shared(struct inode *inode)
829 {
830 	up_read(&inode->i_rwsem);
831 }
832 
inode_trylock(struct inode * inode)833 static inline int inode_trylock(struct inode *inode)
834 {
835 	return down_write_trylock(&inode->i_rwsem);
836 }
837 
inode_trylock_shared(struct inode * inode)838 static inline int inode_trylock_shared(struct inode *inode)
839 {
840 	return down_read_trylock(&inode->i_rwsem);
841 }
842 
inode_is_locked(struct inode * inode)843 static inline int inode_is_locked(struct inode *inode)
844 {
845 	return rwsem_is_locked(&inode->i_rwsem);
846 }
847 
inode_lock_nested(struct inode * inode,unsigned subclass)848 static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
849 {
850 	down_write_nested(&inode->i_rwsem, subclass);
851 }
852 
inode_lock_shared_nested(struct inode * inode,unsigned subclass)853 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
854 {
855 	down_read_nested(&inode->i_rwsem, subclass);
856 }
857 
filemap_invalidate_lock(struct address_space * mapping)858 static inline void filemap_invalidate_lock(struct address_space *mapping)
859 {
860 	down_write(&mapping->invalidate_lock);
861 }
862 
filemap_invalidate_unlock(struct address_space * mapping)863 static inline void filemap_invalidate_unlock(struct address_space *mapping)
864 {
865 	up_write(&mapping->invalidate_lock);
866 }
867 
filemap_invalidate_lock_shared(struct address_space * mapping)868 static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
869 {
870 	down_read(&mapping->invalidate_lock);
871 }
872 
filemap_invalidate_trylock_shared(struct address_space * mapping)873 static inline int filemap_invalidate_trylock_shared(
874 					struct address_space *mapping)
875 {
876 	return down_read_trylock(&mapping->invalidate_lock);
877 }
878 
filemap_invalidate_unlock_shared(struct address_space * mapping)879 static inline void filemap_invalidate_unlock_shared(
880 					struct address_space *mapping)
881 {
882 	up_read(&mapping->invalidate_lock);
883 }
884 
885 void lock_two_nondirectories(struct inode *, struct inode*);
886 void unlock_two_nondirectories(struct inode *, struct inode*);
887 
888 void filemap_invalidate_lock_two(struct address_space *mapping1,
889 				 struct address_space *mapping2);
890 void filemap_invalidate_unlock_two(struct address_space *mapping1,
891 				   struct address_space *mapping2);
892 
893 
894 /*
895  * NOTE: in a 32bit arch with a preemptable kernel and
896  * an UP compile the i_size_read/write must be atomic
897  * with respect to the local cpu (unlike with preempt disabled),
898  * but they don't need to be atomic with respect to other cpus like in
899  * true SMP (so they need either to either locally disable irq around
900  * the read or for example on x86 they can be still implemented as a
901  * cmpxchg8b without the need of the lock prefix). For SMP compiles
902  * and 64bit archs it makes no difference if preempt is enabled or not.
903  */
i_size_read(const struct inode * inode)904 static inline loff_t i_size_read(const struct inode *inode)
905 {
906 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
907 	loff_t i_size;
908 	unsigned int seq;
909 
910 	do {
911 		seq = read_seqcount_begin(&inode->i_size_seqcount);
912 		i_size = inode->i_size;
913 	} while (read_seqcount_retry(&inode->i_size_seqcount, seq));
914 	return i_size;
915 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
916 	loff_t i_size;
917 
918 	preempt_disable();
919 	i_size = inode->i_size;
920 	preempt_enable();
921 	return i_size;
922 #else
923 	/* Pairs with smp_store_release() in i_size_write() */
924 	return smp_load_acquire(&inode->i_size);
925 #endif
926 }
927 
928 /*
929  * NOTE: unlike i_size_read(), i_size_write() does need locking around it
930  * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
931  * can be lost, resulting in subsequent i_size_read() calls spinning forever.
932  */
i_size_write(struct inode * inode,loff_t i_size)933 static inline void i_size_write(struct inode *inode, loff_t i_size)
934 {
935 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
936 	preempt_disable();
937 	write_seqcount_begin(&inode->i_size_seqcount);
938 	inode->i_size = i_size;
939 	write_seqcount_end(&inode->i_size_seqcount);
940 	preempt_enable();
941 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
942 	preempt_disable();
943 	inode->i_size = i_size;
944 	preempt_enable();
945 #else
946 	/*
947 	 * Pairs with smp_load_acquire() in i_size_read() to ensure
948 	 * changes related to inode size (such as page contents) are
949 	 * visible before we see the changed inode size.
950 	 */
951 	smp_store_release(&inode->i_size, i_size);
952 #endif
953 }
954 
iminor(const struct inode * inode)955 static inline unsigned iminor(const struct inode *inode)
956 {
957 	return MINOR(inode->i_rdev);
958 }
959 
imajor(const struct inode * inode)960 static inline unsigned imajor(const struct inode *inode)
961 {
962 	return MAJOR(inode->i_rdev);
963 }
964 
965 struct fown_struct {
966 	struct file *file;	/* backpointer for security modules */
967 	rwlock_t lock;          /* protects pid, uid, euid fields */
968 	struct pid *pid;	/* pid or -pgrp where SIGIO should be sent */
969 	enum pid_type pid_type;	/* Kind of process group SIGIO should be sent to */
970 	kuid_t uid, euid;	/* uid/euid of process setting the owner */
971 	int signum;		/* posix.1b rt signal to be delivered on IO */
972 };
973 
974 /**
975  * struct file_ra_state - Track a file's readahead state.
976  * @start: Where the most recent readahead started.
977  * @size: Number of pages read in the most recent readahead.
978  * @async_size: Numer of pages that were/are not needed immediately
979  *      and so were/are genuinely "ahead".  Start next readahead when
980  *      the first of these pages is accessed.
981  * @ra_pages: Maximum size of a readahead request, copied from the bdi.
982  * @mmap_miss: How many mmap accesses missed in the page cache.
983  * @prev_pos: The last byte in the most recent read request.
984  *
985  * When this structure is passed to ->readahead(), the "most recent"
986  * readahead means the current readahead.
987  */
988 struct file_ra_state {
989 	pgoff_t start;
990 	unsigned int size;
991 	unsigned int async_size;
992 	unsigned int ra_pages;
993 	unsigned int mmap_miss;
994 	loff_t prev_pos;
995 };
996 
997 /*
998  * Check if @index falls in the readahead windows.
999  */
ra_has_index(struct file_ra_state * ra,pgoff_t index)1000 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
1001 {
1002 	return (index >= ra->start &&
1003 		index <  ra->start + ra->size);
1004 }
1005 
1006 /**
1007  * struct file - Represents a file
1008  * @f_count: reference count
1009  * @f_lock: Protects f_ep, f_flags. Must not be taken from IRQ context.
1010  * @f_mode: FMODE_* flags often used in hotpaths
1011  * @f_op: file operations
1012  * @f_mapping: Contents of a cacheable, mappable object.
1013  * @private_data: filesystem or driver specific data
1014  * @f_inode: cached inode
1015  * @f_flags: file flags
1016  * @f_iocb_flags: iocb flags
1017  * @f_cred: stashed credentials of creator/opener
1018  * @f_path: path of the file
1019  * @f_pos_lock: lock protecting file position
1020  * @f_pipe: specific to pipes
1021  * @f_pos: file position
1022  * @f_security: LSM security context of this file
1023  * @f_owner: file owner
1024  * @f_wb_err: writeback error
1025  * @f_sb_err: per sb writeback errors
1026  * @f_ep: link of all epoll hooks for this file
1027  * @f_task_work: task work entry point
1028  * @f_llist: work queue entrypoint
1029  * @f_ra: file's readahead state
1030  * @f_freeptr: Pointer used by SLAB_TYPESAFE_BY_RCU file cache (don't touch.)
1031  */
1032 struct file {
1033 	atomic_long_t			f_count;
1034 	spinlock_t			f_lock;
1035 	fmode_t				f_mode;
1036 	const struct file_operations	*f_op;
1037 	struct address_space		*f_mapping;
1038 	void				*private_data;
1039 	struct inode			*f_inode;
1040 	unsigned int			f_flags;
1041 	unsigned int			f_iocb_flags;
1042 	const struct cred		*f_cred;
1043 	/* --- cacheline 1 boundary (64 bytes) --- */
1044 	struct path			f_path;
1045 	union {
1046 		/* regular files (with FMODE_ATOMIC_POS) and directories */
1047 		struct mutex		f_pos_lock;
1048 		/* pipes */
1049 		u64			f_pipe;
1050 	};
1051 	loff_t				f_pos;
1052 #ifdef CONFIG_SECURITY
1053 	void				*f_security;
1054 #endif
1055 	/* --- cacheline 2 boundary (128 bytes) --- */
1056 	struct fown_struct		*f_owner;
1057 	errseq_t			f_wb_err;
1058 	errseq_t			f_sb_err;
1059 #ifdef CONFIG_EPOLL
1060 	struct hlist_head		*f_ep;
1061 #endif
1062 	union {
1063 		struct callback_head	f_task_work;
1064 		struct llist_node	f_llist;
1065 		struct file_ra_state	f_ra;
1066 		freeptr_t		f_freeptr;
1067 	};
1068 	/* --- cacheline 3 boundary (192 bytes) --- */
1069 } __randomize_layout
1070   __attribute__((aligned(4)));	/* lest something weird decides that 2 is OK */
1071 
1072 struct file_handle {
1073 	__u32 handle_bytes;
1074 	int handle_type;
1075 	/* file identifier */
1076 	unsigned char f_handle[] __counted_by(handle_bytes);
1077 };
1078 
get_file(struct file * f)1079 static inline struct file *get_file(struct file *f)
1080 {
1081 	long prior = atomic_long_fetch_inc_relaxed(&f->f_count);
1082 	WARN_ONCE(!prior, "struct file::f_count incremented from zero; use-after-free condition present!\n");
1083 	return f;
1084 }
1085 
1086 struct file *get_file_rcu(struct file __rcu **f);
1087 struct file *get_file_active(struct file **f);
1088 
1089 #define file_count(x)	atomic_long_read(&(x)->f_count)
1090 
1091 #define	MAX_NON_LFS	((1UL<<31) - 1)
1092 
1093 /* Page cache limit. The filesystems should put that into their s_maxbytes
1094    limits, otherwise bad things can happen in VM. */
1095 #if BITS_PER_LONG==32
1096 #define MAX_LFS_FILESIZE	((loff_t)ULONG_MAX << PAGE_SHIFT)
1097 #elif BITS_PER_LONG==64
1098 #define MAX_LFS_FILESIZE 	((loff_t)LLONG_MAX)
1099 #endif
1100 
1101 /* legacy typedef, should eventually be removed */
1102 typedef void *fl_owner_t;
1103 
1104 struct file_lock;
1105 struct file_lease;
1106 
1107 /* The following constant reflects the upper bound of the file/locking space */
1108 #ifndef OFFSET_MAX
1109 #define OFFSET_MAX	type_max(loff_t)
1110 #define OFFT_OFFSET_MAX	type_max(off_t)
1111 #endif
1112 
1113 int file_f_owner_allocate(struct file *file);
file_f_owner(const struct file * file)1114 static inline struct fown_struct *file_f_owner(const struct file *file)
1115 {
1116 	return READ_ONCE(file->f_owner);
1117 }
1118 
1119 extern void send_sigio(struct fown_struct *fown, int fd, int band);
1120 
file_inode(const struct file * f)1121 static inline struct inode *file_inode(const struct file *f)
1122 {
1123 	return f->f_inode;
1124 }
1125 
1126 /*
1127  * file_dentry() is a relic from the days that overlayfs was using files with a
1128  * "fake" path, meaning, f_path on overlayfs and f_inode on underlying fs.
1129  * In those days, file_dentry() was needed to get the underlying fs dentry that
1130  * matches f_inode.
1131  * Files with "fake" path should not exist nowadays, so use an assertion to make
1132  * sure that file_dentry() was not papering over filesystem bugs.
1133  */
file_dentry(const struct file * file)1134 static inline struct dentry *file_dentry(const struct file *file)
1135 {
1136 	struct dentry *dentry = file->f_path.dentry;
1137 
1138 	WARN_ON_ONCE(d_inode(dentry) != file_inode(file));
1139 	return dentry;
1140 }
1141 
1142 struct fasync_struct {
1143 	rwlock_t		fa_lock;
1144 	int			magic;
1145 	int			fa_fd;
1146 	struct fasync_struct	*fa_next; /* singly linked list */
1147 	struct file		*fa_file;
1148 	struct rcu_head		fa_rcu;
1149 };
1150 
1151 #define FASYNC_MAGIC 0x4601
1152 
1153 /* SMP safe fasync helpers: */
1154 extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1155 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1156 extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1157 extern struct fasync_struct *fasync_alloc(void);
1158 extern void fasync_free(struct fasync_struct *);
1159 
1160 /* can be called from interrupts */
1161 extern void kill_fasync(struct fasync_struct **, int, int);
1162 
1163 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1164 extern int f_setown(struct file *filp, int who, int force);
1165 extern void f_delown(struct file *filp);
1166 extern pid_t f_getown(struct file *filp);
1167 extern int send_sigurg(struct file *file);
1168 
1169 /*
1170  * sb->s_flags.  Note that these mirror the equivalent MS_* flags where
1171  * represented in both.
1172  */
1173 #define SB_RDONLY       BIT(0)	/* Mount read-only */
1174 #define SB_NOSUID       BIT(1)	/* Ignore suid and sgid bits */
1175 #define SB_NODEV        BIT(2)	/* Disallow access to device special files */
1176 #define SB_NOEXEC       BIT(3)	/* Disallow program execution */
1177 #define SB_SYNCHRONOUS  BIT(4)	/* Writes are synced at once */
1178 #define SB_MANDLOCK     BIT(6)	/* Allow mandatory locks on an FS */
1179 #define SB_DIRSYNC      BIT(7)	/* Directory modifications are synchronous */
1180 #define SB_NOATIME      BIT(10)	/* Do not update access times. */
1181 #define SB_NODIRATIME   BIT(11)	/* Do not update directory access times */
1182 #define SB_SILENT       BIT(15)
1183 #define SB_POSIXACL     BIT(16)	/* Supports POSIX ACLs */
1184 #define SB_INLINECRYPT  BIT(17)	/* Use blk-crypto for encrypted files */
1185 #define SB_KERNMOUNT    BIT(22)	/* this is a kern_mount call */
1186 #define SB_I_VERSION    BIT(23)	/* Update inode I_version field */
1187 #define SB_LAZYTIME     BIT(25)	/* Update the on-disk [acm]times lazily */
1188 
1189 /* These sb flags are internal to the kernel */
1190 #define SB_DEAD         BIT(21)
1191 #define SB_DYING        BIT(24)
1192 #define SB_SUBMOUNT     BIT(26)
1193 #define SB_FORCE        BIT(27)
1194 #define SB_NOSEC        BIT(28)
1195 #define SB_BORN         BIT(29)
1196 #define SB_ACTIVE       BIT(30)
1197 #define SB_NOUSER       BIT(31)
1198 
1199 /* These flags relate to encoding and casefolding */
1200 #define SB_ENC_STRICT_MODE_FL	(1 << 0)
1201 
1202 #define sb_has_strict_encoding(sb) \
1203 	(sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1204 
1205 /*
1206  *	Umount options
1207  */
1208 
1209 #define MNT_FORCE	0x00000001	/* Attempt to forcibily umount */
1210 #define MNT_DETACH	0x00000002	/* Just detach from the tree */
1211 #define MNT_EXPIRE	0x00000004	/* Mark for expiry */
1212 #define UMOUNT_NOFOLLOW	0x00000008	/* Don't follow symlink on umount */
1213 #define UMOUNT_UNUSED	0x80000000	/* Flag guaranteed to be unused */
1214 
1215 /* sb->s_iflags */
1216 #define SB_I_CGROUPWB	0x00000001	/* cgroup-aware writeback enabled */
1217 #define SB_I_NOEXEC	0x00000002	/* Ignore executables on this fs */
1218 #define SB_I_NODEV	0x00000004	/* Ignore devices on this fs */
1219 #define SB_I_STABLE_WRITES 0x00000008	/* don't modify blks until WB is done */
1220 
1221 /* sb->s_iflags to limit user namespace mounts */
1222 #define SB_I_USERNS_VISIBLE		0x00000010 /* fstype already mounted */
1223 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE	0x00000020
1224 #define SB_I_UNTRUSTED_MOUNTER		0x00000040
1225 #define SB_I_EVM_HMAC_UNSUPPORTED	0x00000080
1226 
1227 #define SB_I_SKIP_SYNC	0x00000100	/* Skip superblock at global sync */
1228 #define SB_I_PERSB_BDI	0x00000200	/* has a per-sb bdi */
1229 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1230 #define SB_I_RETIRED	0x00000800	/* superblock shouldn't be reused */
1231 #define SB_I_NOUMASK	0x00001000	/* VFS does not apply umask */
1232 #define SB_I_NOIDMAP	0x00002000	/* No idmapped mounts on this superblock */
1233 
1234 /* Possible states of 'frozen' field */
1235 enum {
1236 	SB_UNFROZEN = 0,		/* FS is unfrozen */
1237 	SB_FREEZE_WRITE	= 1,		/* Writes, dir ops, ioctls frozen */
1238 	SB_FREEZE_PAGEFAULT = 2,	/* Page faults stopped as well */
1239 	SB_FREEZE_FS = 3,		/* For internal FS use (e.g. to stop
1240 					 * internal threads if needed) */
1241 	SB_FREEZE_COMPLETE = 4,		/* ->freeze_fs finished successfully */
1242 };
1243 
1244 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1245 
1246 struct sb_writers {
1247 	unsigned short			frozen;		/* Is sb frozen? */
1248 	int				freeze_kcount;	/* How many kernel freeze requests? */
1249 	int				freeze_ucount;	/* How many userspace freeze requests? */
1250 	struct percpu_rw_semaphore	rw_sem[SB_FREEZE_LEVELS];
1251 };
1252 
1253 struct super_block {
1254 	struct list_head	s_list;		/* Keep this first */
1255 	dev_t			s_dev;		/* search index; _not_ kdev_t */
1256 	unsigned char		s_blocksize_bits;
1257 	unsigned long		s_blocksize;
1258 	loff_t			s_maxbytes;	/* Max file size */
1259 	struct file_system_type	*s_type;
1260 	const struct super_operations	*s_op;
1261 	const struct dquot_operations	*dq_op;
1262 	const struct quotactl_ops	*s_qcop;
1263 	const struct export_operations *s_export_op;
1264 	unsigned long		s_flags;
1265 	unsigned long		s_iflags;	/* internal SB_I_* flags */
1266 	unsigned long		s_magic;
1267 	struct dentry		*s_root;
1268 	struct rw_semaphore	s_umount;
1269 	int			s_count;
1270 	atomic_t		s_active;
1271 #ifdef CONFIG_SECURITY
1272 	void                    *s_security;
1273 #endif
1274 	const struct xattr_handler * const *s_xattr;
1275 #ifdef CONFIG_FS_ENCRYPTION
1276 	const struct fscrypt_operations	*s_cop;
1277 	struct fscrypt_keyring	*s_master_keys; /* master crypto keys in use */
1278 #endif
1279 #ifdef CONFIG_FS_VERITY
1280 	const struct fsverity_operations *s_vop;
1281 #endif
1282 #if IS_ENABLED(CONFIG_UNICODE)
1283 	struct unicode_map *s_encoding;
1284 	__u16 s_encoding_flags;
1285 #endif
1286 	struct hlist_bl_head	s_roots;	/* alternate root dentries for NFS */
1287 	struct list_head	s_mounts;	/* list of mounts; _not_ for fs use */
1288 	struct block_device	*s_bdev;	/* can go away once we use an accessor for @s_bdev_file */
1289 	struct file		*s_bdev_file;
1290 	struct backing_dev_info *s_bdi;
1291 	struct mtd_info		*s_mtd;
1292 	struct hlist_node	s_instances;
1293 	unsigned int		s_quota_types;	/* Bitmask of supported quota types */
1294 	struct quota_info	s_dquot;	/* Diskquota specific options */
1295 
1296 	struct sb_writers	s_writers;
1297 
1298 	/*
1299 	 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1300 	 * s_fsnotify_info together for cache efficiency. They are frequently
1301 	 * accessed and rarely modified.
1302 	 */
1303 	void			*s_fs_info;	/* Filesystem private info */
1304 
1305 	/* Granularity of c/m/atime in ns (cannot be worse than a second) */
1306 	u32			s_time_gran;
1307 	/* Time limits for c/m/atime in seconds */
1308 	time64_t		   s_time_min;
1309 	time64_t		   s_time_max;
1310 #ifdef CONFIG_FSNOTIFY
1311 	u32			s_fsnotify_mask;
1312 	struct fsnotify_sb_info	*s_fsnotify_info;
1313 #endif
1314 
1315 	/*
1316 	 * q: why are s_id and s_sysfs_name not the same? both are human
1317 	 * readable strings that identify the filesystem
1318 	 * a: s_id is allowed to change at runtime; it's used in log messages,
1319 	 * and we want to when a device starts out as single device (s_id is dev
1320 	 * name) but then a device is hot added and we have to switch to
1321 	 * identifying it by UUID
1322 	 * but s_sysfs_name is a handle for programmatic access, and can't
1323 	 * change at runtime
1324 	 */
1325 	char			s_id[32];	/* Informational name */
1326 	uuid_t			s_uuid;		/* UUID */
1327 	u8			s_uuid_len;	/* Default 16, possibly smaller for weird filesystems */
1328 
1329 	/* if set, fs shows up under sysfs at /sys/fs/$FSTYP/s_sysfs_name */
1330 	char			s_sysfs_name[UUID_STRING_LEN + 1];
1331 
1332 	unsigned int		s_max_links;
1333 
1334 	/*
1335 	 * The next field is for VFS *only*. No filesystems have any business
1336 	 * even looking at it. You had been warned.
1337 	 */
1338 	struct mutex s_vfs_rename_mutex;	/* Kludge */
1339 
1340 	/*
1341 	 * Filesystem subtype.  If non-empty the filesystem type field
1342 	 * in /proc/mounts will be "type.subtype"
1343 	 */
1344 	const char *s_subtype;
1345 
1346 	const struct dentry_operations *s_d_op; /* default d_op for dentries */
1347 
1348 	struct shrinker *s_shrink;	/* per-sb shrinker handle */
1349 
1350 	/* Number of inodes with nlink == 0 but still referenced */
1351 	atomic_long_t s_remove_count;
1352 
1353 	/* Read-only state of the superblock is being changed */
1354 	int s_readonly_remount;
1355 
1356 	/* per-sb errseq_t for reporting writeback errors via syncfs */
1357 	errseq_t s_wb_err;
1358 
1359 	/* AIO completions deferred from interrupt context */
1360 	struct workqueue_struct *s_dio_done_wq;
1361 	struct hlist_head s_pins;
1362 
1363 	/*
1364 	 * Owning user namespace and default context in which to
1365 	 * interpret filesystem uids, gids, quotas, device nodes,
1366 	 * xattrs and security labels.
1367 	 */
1368 	struct user_namespace *s_user_ns;
1369 
1370 	/*
1371 	 * The list_lru structure is essentially just a pointer to a table
1372 	 * of per-node lru lists, each of which has its own spinlock.
1373 	 * There is no need to put them into separate cachelines.
1374 	 */
1375 	struct list_lru		s_dentry_lru;
1376 	struct list_lru		s_inode_lru;
1377 	struct rcu_head		rcu;
1378 	struct work_struct	destroy_work;
1379 
1380 	struct mutex		s_sync_lock;	/* sync serialisation lock */
1381 
1382 	/*
1383 	 * Indicates how deep in a filesystem stack this SB is
1384 	 */
1385 	int s_stack_depth;
1386 
1387 	/* s_inode_list_lock protects s_inodes */
1388 	spinlock_t		s_inode_list_lock ____cacheline_aligned_in_smp;
1389 	struct list_head	s_inodes;	/* all inodes */
1390 
1391 	spinlock_t		s_inode_wblist_lock;
1392 	struct list_head	s_inodes_wb;	/* writeback inodes */
1393 } __randomize_layout;
1394 
i_user_ns(const struct inode * inode)1395 static inline struct user_namespace *i_user_ns(const struct inode *inode)
1396 {
1397 	return inode->i_sb->s_user_ns;
1398 }
1399 
1400 /* Helper functions so that in most cases filesystems will
1401  * not need to deal directly with kuid_t and kgid_t and can
1402  * instead deal with the raw numeric values that are stored
1403  * in the filesystem.
1404  */
i_uid_read(const struct inode * inode)1405 static inline uid_t i_uid_read(const struct inode *inode)
1406 {
1407 	return from_kuid(i_user_ns(inode), inode->i_uid);
1408 }
1409 
i_gid_read(const struct inode * inode)1410 static inline gid_t i_gid_read(const struct inode *inode)
1411 {
1412 	return from_kgid(i_user_ns(inode), inode->i_gid);
1413 }
1414 
i_uid_write(struct inode * inode,uid_t uid)1415 static inline void i_uid_write(struct inode *inode, uid_t uid)
1416 {
1417 	inode->i_uid = make_kuid(i_user_ns(inode), uid);
1418 }
1419 
i_gid_write(struct inode * inode,gid_t gid)1420 static inline void i_gid_write(struct inode *inode, gid_t gid)
1421 {
1422 	inode->i_gid = make_kgid(i_user_ns(inode), gid);
1423 }
1424 
1425 /**
1426  * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping
1427  * @idmap: idmap of the mount the inode was found from
1428  * @inode: inode to map
1429  *
1430  * Return: whe inode's i_uid mapped down according to @idmap.
1431  * If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1432  */
i_uid_into_vfsuid(struct mnt_idmap * idmap,const struct inode * inode)1433 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap,
1434 					 const struct inode *inode)
1435 {
1436 	return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid);
1437 }
1438 
1439 /**
1440  * i_uid_needs_update - check whether inode's i_uid needs to be updated
1441  * @idmap: idmap of the mount the inode was found from
1442  * @attr: the new attributes of @inode
1443  * @inode: the inode to update
1444  *
1445  * Check whether the $inode's i_uid field needs to be updated taking idmapped
1446  * mounts into account if the filesystem supports it.
1447  *
1448  * Return: true if @inode's i_uid field needs to be updated, false if not.
1449  */
i_uid_needs_update(struct mnt_idmap * idmap,const struct iattr * attr,const struct inode * inode)1450 static inline bool i_uid_needs_update(struct mnt_idmap *idmap,
1451 				      const struct iattr *attr,
1452 				      const struct inode *inode)
1453 {
1454 	return ((attr->ia_valid & ATTR_UID) &&
1455 		!vfsuid_eq(attr->ia_vfsuid,
1456 			   i_uid_into_vfsuid(idmap, inode)));
1457 }
1458 
1459 /**
1460  * i_uid_update - update @inode's i_uid field
1461  * @idmap: idmap of the mount the inode was found from
1462  * @attr: the new attributes of @inode
1463  * @inode: the inode to update
1464  *
1465  * Safely update @inode's i_uid field translating the vfsuid of any idmapped
1466  * mount into the filesystem kuid.
1467  */
i_uid_update(struct mnt_idmap * idmap,const struct iattr * attr,struct inode * inode)1468 static inline void i_uid_update(struct mnt_idmap *idmap,
1469 				const struct iattr *attr,
1470 				struct inode *inode)
1471 {
1472 	if (attr->ia_valid & ATTR_UID)
1473 		inode->i_uid = from_vfsuid(idmap, i_user_ns(inode),
1474 					   attr->ia_vfsuid);
1475 }
1476 
1477 /**
1478  * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping
1479  * @idmap: idmap of the mount the inode was found from
1480  * @inode: inode to map
1481  *
1482  * Return: the inode's i_gid mapped down according to @idmap.
1483  * If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1484  */
i_gid_into_vfsgid(struct mnt_idmap * idmap,const struct inode * inode)1485 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap,
1486 					 const struct inode *inode)
1487 {
1488 	return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid);
1489 }
1490 
1491 /**
1492  * i_gid_needs_update - check whether inode's i_gid needs to be updated
1493  * @idmap: idmap of the mount the inode was found from
1494  * @attr: the new attributes of @inode
1495  * @inode: the inode to update
1496  *
1497  * Check whether the $inode's i_gid field needs to be updated taking idmapped
1498  * mounts into account if the filesystem supports it.
1499  *
1500  * Return: true if @inode's i_gid field needs to be updated, false if not.
1501  */
i_gid_needs_update(struct mnt_idmap * idmap,const struct iattr * attr,const struct inode * inode)1502 static inline bool i_gid_needs_update(struct mnt_idmap *idmap,
1503 				      const struct iattr *attr,
1504 				      const struct inode *inode)
1505 {
1506 	return ((attr->ia_valid & ATTR_GID) &&
1507 		!vfsgid_eq(attr->ia_vfsgid,
1508 			   i_gid_into_vfsgid(idmap, inode)));
1509 }
1510 
1511 /**
1512  * i_gid_update - update @inode's i_gid field
1513  * @idmap: idmap of the mount the inode was found from
1514  * @attr: the new attributes of @inode
1515  * @inode: the inode to update
1516  *
1517  * Safely update @inode's i_gid field translating the vfsgid of any idmapped
1518  * mount into the filesystem kgid.
1519  */
i_gid_update(struct mnt_idmap * idmap,const struct iattr * attr,struct inode * inode)1520 static inline void i_gid_update(struct mnt_idmap *idmap,
1521 				const struct iattr *attr,
1522 				struct inode *inode)
1523 {
1524 	if (attr->ia_valid & ATTR_GID)
1525 		inode->i_gid = from_vfsgid(idmap, i_user_ns(inode),
1526 					   attr->ia_vfsgid);
1527 }
1528 
1529 /**
1530  * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1531  * @inode: inode to initialize
1532  * @idmap: idmap of the mount the inode was found from
1533  *
1534  * Initialize the i_uid field of @inode. If the inode was found/created via
1535  * an idmapped mount map the caller's fsuid according to @idmap.
1536  */
inode_fsuid_set(struct inode * inode,struct mnt_idmap * idmap)1537 static inline void inode_fsuid_set(struct inode *inode,
1538 				   struct mnt_idmap *idmap)
1539 {
1540 	inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode));
1541 }
1542 
1543 /**
1544  * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1545  * @inode: inode to initialize
1546  * @idmap: idmap of the mount the inode was found from
1547  *
1548  * Initialize the i_gid field of @inode. If the inode was found/created via
1549  * an idmapped mount map the caller's fsgid according to @idmap.
1550  */
inode_fsgid_set(struct inode * inode,struct mnt_idmap * idmap)1551 static inline void inode_fsgid_set(struct inode *inode,
1552 				   struct mnt_idmap *idmap)
1553 {
1554 	inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode));
1555 }
1556 
1557 /**
1558  * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1559  * @sb: the superblock we want a mapping in
1560  * @idmap: idmap of the relevant mount
1561  *
1562  * Check whether the caller's fsuid and fsgid have a valid mapping in the
1563  * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1564  * the caller's fsuid and fsgid according to the @idmap first.
1565  *
1566  * Return: true if fsuid and fsgid is mapped, false if not.
1567  */
fsuidgid_has_mapping(struct super_block * sb,struct mnt_idmap * idmap)1568 static inline bool fsuidgid_has_mapping(struct super_block *sb,
1569 					struct mnt_idmap *idmap)
1570 {
1571 	struct user_namespace *fs_userns = sb->s_user_ns;
1572 	kuid_t kuid;
1573 	kgid_t kgid;
1574 
1575 	kuid = mapped_fsuid(idmap, fs_userns);
1576 	if (!uid_valid(kuid))
1577 		return false;
1578 	kgid = mapped_fsgid(idmap, fs_userns);
1579 	if (!gid_valid(kgid))
1580 		return false;
1581 	return kuid_has_mapping(fs_userns, kuid) &&
1582 	       kgid_has_mapping(fs_userns, kgid);
1583 }
1584 
1585 struct timespec64 current_time(struct inode *inode);
1586 struct timespec64 inode_set_ctime_current(struct inode *inode);
1587 
inode_get_atime_sec(const struct inode * inode)1588 static inline time64_t inode_get_atime_sec(const struct inode *inode)
1589 {
1590 	return inode->i_atime_sec;
1591 }
1592 
inode_get_atime_nsec(const struct inode * inode)1593 static inline long inode_get_atime_nsec(const struct inode *inode)
1594 {
1595 	return inode->i_atime_nsec;
1596 }
1597 
inode_get_atime(const struct inode * inode)1598 static inline struct timespec64 inode_get_atime(const struct inode *inode)
1599 {
1600 	struct timespec64 ts = { .tv_sec  = inode_get_atime_sec(inode),
1601 				 .tv_nsec = inode_get_atime_nsec(inode) };
1602 
1603 	return ts;
1604 }
1605 
inode_set_atime_to_ts(struct inode * inode,struct timespec64 ts)1606 static inline struct timespec64 inode_set_atime_to_ts(struct inode *inode,
1607 						      struct timespec64 ts)
1608 {
1609 	inode->i_atime_sec = ts.tv_sec;
1610 	inode->i_atime_nsec = ts.tv_nsec;
1611 	return ts;
1612 }
1613 
inode_set_atime(struct inode * inode,time64_t sec,long nsec)1614 static inline struct timespec64 inode_set_atime(struct inode *inode,
1615 						time64_t sec, long nsec)
1616 {
1617 	struct timespec64 ts = { .tv_sec  = sec,
1618 				 .tv_nsec = nsec };
1619 
1620 	return inode_set_atime_to_ts(inode, ts);
1621 }
1622 
inode_get_mtime_sec(const struct inode * inode)1623 static inline time64_t inode_get_mtime_sec(const struct inode *inode)
1624 {
1625 	return inode->i_mtime_sec;
1626 }
1627 
inode_get_mtime_nsec(const struct inode * inode)1628 static inline long inode_get_mtime_nsec(const struct inode *inode)
1629 {
1630 	return inode->i_mtime_nsec;
1631 }
1632 
inode_get_mtime(const struct inode * inode)1633 static inline struct timespec64 inode_get_mtime(const struct inode *inode)
1634 {
1635 	struct timespec64 ts = { .tv_sec  = inode_get_mtime_sec(inode),
1636 				 .tv_nsec = inode_get_mtime_nsec(inode) };
1637 	return ts;
1638 }
1639 
inode_set_mtime_to_ts(struct inode * inode,struct timespec64 ts)1640 static inline struct timespec64 inode_set_mtime_to_ts(struct inode *inode,
1641 						      struct timespec64 ts)
1642 {
1643 	inode->i_mtime_sec = ts.tv_sec;
1644 	inode->i_mtime_nsec = ts.tv_nsec;
1645 	return ts;
1646 }
1647 
inode_set_mtime(struct inode * inode,time64_t sec,long nsec)1648 static inline struct timespec64 inode_set_mtime(struct inode *inode,
1649 						time64_t sec, long nsec)
1650 {
1651 	struct timespec64 ts = { .tv_sec  = sec,
1652 				 .tv_nsec = nsec };
1653 	return inode_set_mtime_to_ts(inode, ts);
1654 }
1655 
inode_get_ctime_sec(const struct inode * inode)1656 static inline time64_t inode_get_ctime_sec(const struct inode *inode)
1657 {
1658 	return inode->i_ctime_sec;
1659 }
1660 
inode_get_ctime_nsec(const struct inode * inode)1661 static inline long inode_get_ctime_nsec(const struct inode *inode)
1662 {
1663 	return inode->i_ctime_nsec;
1664 }
1665 
inode_get_ctime(const struct inode * inode)1666 static inline struct timespec64 inode_get_ctime(const struct inode *inode)
1667 {
1668 	struct timespec64 ts = { .tv_sec  = inode_get_ctime_sec(inode),
1669 				 .tv_nsec = inode_get_ctime_nsec(inode) };
1670 
1671 	return ts;
1672 }
1673 
inode_set_ctime_to_ts(struct inode * inode,struct timespec64 ts)1674 static inline struct timespec64 inode_set_ctime_to_ts(struct inode *inode,
1675 						      struct timespec64 ts)
1676 {
1677 	inode->i_ctime_sec = ts.tv_sec;
1678 	inode->i_ctime_nsec = ts.tv_nsec;
1679 	return ts;
1680 }
1681 
1682 /**
1683  * inode_set_ctime - set the ctime in the inode
1684  * @inode: inode in which to set the ctime
1685  * @sec: tv_sec value to set
1686  * @nsec: tv_nsec value to set
1687  *
1688  * Set the ctime in @inode to { @sec, @nsec }
1689  */
inode_set_ctime(struct inode * inode,time64_t sec,long nsec)1690 static inline struct timespec64 inode_set_ctime(struct inode *inode,
1691 						time64_t sec, long nsec)
1692 {
1693 	struct timespec64 ts = { .tv_sec  = sec,
1694 				 .tv_nsec = nsec };
1695 
1696 	return inode_set_ctime_to_ts(inode, ts);
1697 }
1698 
1699 struct timespec64 simple_inode_init_ts(struct inode *inode);
1700 
1701 /*
1702  * Snapshotting support.
1703  */
1704 
1705 /*
1706  * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1707  * instead.
1708  */
__sb_end_write(struct super_block * sb,int level)1709 static inline void __sb_end_write(struct super_block *sb, int level)
1710 {
1711 	percpu_up_read(sb->s_writers.rw_sem + level-1);
1712 }
1713 
__sb_start_write(struct super_block * sb,int level)1714 static inline void __sb_start_write(struct super_block *sb, int level)
1715 {
1716 	percpu_down_read(sb->s_writers.rw_sem + level - 1);
1717 }
1718 
__sb_start_write_trylock(struct super_block * sb,int level)1719 static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1720 {
1721 	return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1722 }
1723 
1724 #define __sb_writers_acquired(sb, lev)	\
1725 	percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1726 #define __sb_writers_release(sb, lev)	\
1727 	percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], _THIS_IP_)
1728 
1729 /**
1730  * __sb_write_started - check if sb freeze level is held
1731  * @sb: the super we write to
1732  * @level: the freeze level
1733  *
1734  * * > 0 - sb freeze level is held
1735  * *   0 - sb freeze level is not held
1736  * * < 0 - !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN
1737  */
__sb_write_started(const struct super_block * sb,int level)1738 static inline int __sb_write_started(const struct super_block *sb, int level)
1739 {
1740 	return lockdep_is_held_type(sb->s_writers.rw_sem + level - 1, 1);
1741 }
1742 
1743 /**
1744  * sb_write_started - check if SB_FREEZE_WRITE is held
1745  * @sb: the super we write to
1746  *
1747  * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1748  */
sb_write_started(const struct super_block * sb)1749 static inline bool sb_write_started(const struct super_block *sb)
1750 {
1751 	return __sb_write_started(sb, SB_FREEZE_WRITE);
1752 }
1753 
1754 /**
1755  * sb_write_not_started - check if SB_FREEZE_WRITE is not held
1756  * @sb: the super we write to
1757  *
1758  * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1759  */
sb_write_not_started(const struct super_block * sb)1760 static inline bool sb_write_not_started(const struct super_block *sb)
1761 {
1762 	return __sb_write_started(sb, SB_FREEZE_WRITE) <= 0;
1763 }
1764 
1765 /**
1766  * file_write_started - check if SB_FREEZE_WRITE is held
1767  * @file: the file we write to
1768  *
1769  * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1770  * May be false positive with !S_ISREG, because file_start_write() has
1771  * no effect on !S_ISREG.
1772  */
file_write_started(const struct file * file)1773 static inline bool file_write_started(const struct file *file)
1774 {
1775 	if (!S_ISREG(file_inode(file)->i_mode))
1776 		return true;
1777 	return sb_write_started(file_inode(file)->i_sb);
1778 }
1779 
1780 /**
1781  * file_write_not_started - check if SB_FREEZE_WRITE is not held
1782  * @file: the file we write to
1783  *
1784  * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1785  * May be false positive with !S_ISREG, because file_start_write() has
1786  * no effect on !S_ISREG.
1787  */
file_write_not_started(const struct file * file)1788 static inline bool file_write_not_started(const struct file *file)
1789 {
1790 	if (!S_ISREG(file_inode(file)->i_mode))
1791 		return true;
1792 	return sb_write_not_started(file_inode(file)->i_sb);
1793 }
1794 
1795 /**
1796  * sb_end_write - drop write access to a superblock
1797  * @sb: the super we wrote to
1798  *
1799  * Decrement number of writers to the filesystem. Wake up possible waiters
1800  * wanting to freeze the filesystem.
1801  */
sb_end_write(struct super_block * sb)1802 static inline void sb_end_write(struct super_block *sb)
1803 {
1804 	__sb_end_write(sb, SB_FREEZE_WRITE);
1805 }
1806 
1807 /**
1808  * sb_end_pagefault - drop write access to a superblock from a page fault
1809  * @sb: the super we wrote to
1810  *
1811  * Decrement number of processes handling write page fault to the filesystem.
1812  * Wake up possible waiters wanting to freeze the filesystem.
1813  */
sb_end_pagefault(struct super_block * sb)1814 static inline void sb_end_pagefault(struct super_block *sb)
1815 {
1816 	__sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1817 }
1818 
1819 /**
1820  * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1821  * @sb: the super we wrote to
1822  *
1823  * Decrement fs-internal number of writers to the filesystem.  Wake up possible
1824  * waiters wanting to freeze the filesystem.
1825  */
sb_end_intwrite(struct super_block * sb)1826 static inline void sb_end_intwrite(struct super_block *sb)
1827 {
1828 	__sb_end_write(sb, SB_FREEZE_FS);
1829 }
1830 
1831 /**
1832  * sb_start_write - get write access to a superblock
1833  * @sb: the super we write to
1834  *
1835  * When a process wants to write data or metadata to a file system (i.e. dirty
1836  * a page or an inode), it should embed the operation in a sb_start_write() -
1837  * sb_end_write() pair to get exclusion against file system freezing. This
1838  * function increments number of writers preventing freezing. If the file
1839  * system is already frozen, the function waits until the file system is
1840  * thawed.
1841  *
1842  * Since freeze protection behaves as a lock, users have to preserve
1843  * ordering of freeze protection and other filesystem locks. Generally,
1844  * freeze protection should be the outermost lock. In particular, we have:
1845  *
1846  * sb_start_write
1847  *   -> i_mutex			(write path, truncate, directory ops, ...)
1848  *   -> s_umount		(freeze_super, thaw_super)
1849  */
sb_start_write(struct super_block * sb)1850 static inline void sb_start_write(struct super_block *sb)
1851 {
1852 	__sb_start_write(sb, SB_FREEZE_WRITE);
1853 }
1854 
sb_start_write_trylock(struct super_block * sb)1855 static inline bool sb_start_write_trylock(struct super_block *sb)
1856 {
1857 	return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1858 }
1859 
1860 /**
1861  * sb_start_pagefault - get write access to a superblock from a page fault
1862  * @sb: the super we write to
1863  *
1864  * When a process starts handling write page fault, it should embed the
1865  * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1866  * exclusion against file system freezing. This is needed since the page fault
1867  * is going to dirty a page. This function increments number of running page
1868  * faults preventing freezing. If the file system is already frozen, the
1869  * function waits until the file system is thawed.
1870  *
1871  * Since page fault freeze protection behaves as a lock, users have to preserve
1872  * ordering of freeze protection and other filesystem locks. It is advised to
1873  * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1874  * handling code implies lock dependency:
1875  *
1876  * mmap_lock
1877  *   -> sb_start_pagefault
1878  */
sb_start_pagefault(struct super_block * sb)1879 static inline void sb_start_pagefault(struct super_block *sb)
1880 {
1881 	__sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1882 }
1883 
1884 /**
1885  * sb_start_intwrite - get write access to a superblock for internal fs purposes
1886  * @sb: the super we write to
1887  *
1888  * This is the third level of protection against filesystem freezing. It is
1889  * free for use by a filesystem. The only requirement is that it must rank
1890  * below sb_start_pagefault.
1891  *
1892  * For example filesystem can call sb_start_intwrite() when starting a
1893  * transaction which somewhat eases handling of freezing for internal sources
1894  * of filesystem changes (internal fs threads, discarding preallocation on file
1895  * close, etc.).
1896  */
sb_start_intwrite(struct super_block * sb)1897 static inline void sb_start_intwrite(struct super_block *sb)
1898 {
1899 	__sb_start_write(sb, SB_FREEZE_FS);
1900 }
1901 
sb_start_intwrite_trylock(struct super_block * sb)1902 static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1903 {
1904 	return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1905 }
1906 
1907 bool inode_owner_or_capable(struct mnt_idmap *idmap,
1908 			    const struct inode *inode);
1909 
1910 /*
1911  * VFS helper functions..
1912  */
1913 int vfs_create(struct mnt_idmap *, struct inode *,
1914 	       struct dentry *, umode_t, bool);
1915 int vfs_mkdir(struct mnt_idmap *, struct inode *,
1916 	      struct dentry *, umode_t);
1917 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *,
1918               umode_t, dev_t);
1919 int vfs_symlink(struct mnt_idmap *, struct inode *,
1920 		struct dentry *, const char *);
1921 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *,
1922 	     struct dentry *, struct inode **);
1923 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *);
1924 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *,
1925 	       struct inode **);
1926 
1927 /**
1928  * struct renamedata - contains all information required for renaming
1929  * @old_mnt_idmap:     idmap of the old mount the inode was found from
1930  * @old_dir:           parent of source
1931  * @old_dentry:                source
1932  * @new_mnt_idmap:     idmap of the new mount the inode was found from
1933  * @new_dir:           parent of destination
1934  * @new_dentry:                destination
1935  * @delegated_inode:   returns an inode needing a delegation break
1936  * @flags:             rename flags
1937  */
1938 struct renamedata {
1939 	struct mnt_idmap *old_mnt_idmap;
1940 	struct inode *old_dir;
1941 	struct dentry *old_dentry;
1942 	struct mnt_idmap *new_mnt_idmap;
1943 	struct inode *new_dir;
1944 	struct dentry *new_dentry;
1945 	struct inode **delegated_inode;
1946 	unsigned int flags;
1947 } __randomize_layout;
1948 
1949 int vfs_rename(struct renamedata *);
1950 
vfs_whiteout(struct mnt_idmap * idmap,struct inode * dir,struct dentry * dentry)1951 static inline int vfs_whiteout(struct mnt_idmap *idmap,
1952 			       struct inode *dir, struct dentry *dentry)
1953 {
1954 	return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE,
1955 			 WHITEOUT_DEV);
1956 }
1957 
1958 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap,
1959 				 const struct path *parentpath,
1960 				 umode_t mode, int open_flag,
1961 				 const struct cred *cred);
1962 struct file *kernel_file_open(const struct path *path, int flags,
1963 			      const struct cred *cred);
1964 
1965 int vfs_mkobj(struct dentry *, umode_t,
1966 		int (*f)(struct dentry *, umode_t, void *),
1967 		void *);
1968 
1969 int vfs_fchown(struct file *file, uid_t user, gid_t group);
1970 int vfs_fchmod(struct file *file, umode_t mode);
1971 int vfs_utimes(const struct path *path, struct timespec64 *times);
1972 
1973 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1974 
1975 #ifdef CONFIG_COMPAT
1976 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
1977 					unsigned long arg);
1978 #else
1979 #define compat_ptr_ioctl NULL
1980 #endif
1981 
1982 /*
1983  * VFS file helper functions.
1984  */
1985 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
1986 		      const struct inode *dir, umode_t mode);
1987 extern bool may_open_dev(const struct path *path);
1988 umode_t mode_strip_sgid(struct mnt_idmap *idmap,
1989 			const struct inode *dir, umode_t mode);
1990 bool in_group_or_capable(struct mnt_idmap *idmap,
1991 			 const struct inode *inode, vfsgid_t vfsgid);
1992 
1993 /*
1994  * This is the "filldir" function type, used by readdir() to let
1995  * the kernel specify what kind of dirent layout it wants to have.
1996  * This allows the kernel to read directories into kernel space or
1997  * to have different dirent layouts depending on the binary type.
1998  * Return 'true' to keep going and 'false' if there are no more entries.
1999  */
2000 struct dir_context;
2001 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
2002 			 unsigned);
2003 
2004 struct dir_context {
2005 	filldir_t actor;
2006 	loff_t pos;
2007 };
2008 
2009 /*
2010  * These flags let !MMU mmap() govern direct device mapping vs immediate
2011  * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
2012  *
2013  * NOMMU_MAP_COPY:	Copy can be mapped (MAP_PRIVATE)
2014  * NOMMU_MAP_DIRECT:	Can be mapped directly (MAP_SHARED)
2015  * NOMMU_MAP_READ:	Can be mapped for reading
2016  * NOMMU_MAP_WRITE:	Can be mapped for writing
2017  * NOMMU_MAP_EXEC:	Can be mapped for execution
2018  */
2019 #define NOMMU_MAP_COPY		0x00000001
2020 #define NOMMU_MAP_DIRECT	0x00000008
2021 #define NOMMU_MAP_READ		VM_MAYREAD
2022 #define NOMMU_MAP_WRITE		VM_MAYWRITE
2023 #define NOMMU_MAP_EXEC		VM_MAYEXEC
2024 
2025 #define NOMMU_VMFLAGS \
2026 	(NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
2027 
2028 /*
2029  * These flags control the behavior of the remap_file_range function pointer.
2030  * If it is called with len == 0 that means "remap to end of source file".
2031  * See Documentation/filesystems/vfs.rst for more details about this call.
2032  *
2033  * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
2034  * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
2035  */
2036 #define REMAP_FILE_DEDUP		(1 << 0)
2037 #define REMAP_FILE_CAN_SHORTEN		(1 << 1)
2038 
2039 /*
2040  * These flags signal that the caller is ok with altering various aspects of
2041  * the behavior of the remap operation.  The changes must be made by the
2042  * implementation; the vfs remap helper functions can take advantage of them.
2043  * Flags in this category exist to preserve the quirky behavior of the hoisted
2044  * btrfs clone/dedupe ioctls.
2045  */
2046 #define REMAP_FILE_ADVISORY		(REMAP_FILE_CAN_SHORTEN)
2047 
2048 /*
2049  * These flags control the behavior of vfs_copy_file_range().
2050  * They are not available to the user via syscall.
2051  *
2052  * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops
2053  */
2054 #define COPY_FILE_SPLICE		(1 << 0)
2055 
2056 struct iov_iter;
2057 struct io_uring_cmd;
2058 struct offset_ctx;
2059 
2060 typedef unsigned int __bitwise fop_flags_t;
2061 
2062 struct file_operations {
2063 	struct module *owner;
2064 	fop_flags_t fop_flags;
2065 	loff_t (*llseek) (struct file *, loff_t, int);
2066 	ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
2067 	ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
2068 	ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
2069 	ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
2070 	int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
2071 			unsigned int flags);
2072 	int (*iterate_shared) (struct file *, struct dir_context *);
2073 	__poll_t (*poll) (struct file *, struct poll_table_struct *);
2074 	long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
2075 	long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
2076 	int (*mmap) (struct file *, struct vm_area_struct *);
2077 	int (*open) (struct inode *, struct file *);
2078 	int (*flush) (struct file *, fl_owner_t id);
2079 	int (*release) (struct inode *, struct file *);
2080 	int (*fsync) (struct file *, loff_t, loff_t, int datasync);
2081 	int (*fasync) (int, struct file *, int);
2082 	int (*lock) (struct file *, int, struct file_lock *);
2083 	unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2084 	int (*check_flags)(int);
2085 	int (*flock) (struct file *, int, struct file_lock *);
2086 	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
2087 	ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
2088 	void (*splice_eof)(struct file *file);
2089 	int (*setlease)(struct file *, int, struct file_lease **, void **);
2090 	long (*fallocate)(struct file *file, int mode, loff_t offset,
2091 			  loff_t len);
2092 	void (*show_fdinfo)(struct seq_file *m, struct file *f);
2093 #ifndef CONFIG_MMU
2094 	unsigned (*mmap_capabilities)(struct file *);
2095 #endif
2096 	ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
2097 			loff_t, size_t, unsigned int);
2098 	loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
2099 				   struct file *file_out, loff_t pos_out,
2100 				   loff_t len, unsigned int remap_flags);
2101 	int (*fadvise)(struct file *, loff_t, loff_t, int);
2102 	int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
2103 	int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *,
2104 				unsigned int poll_flags);
2105 } __randomize_layout;
2106 
2107 /* Supports async buffered reads */
2108 #define FOP_BUFFER_RASYNC	((__force fop_flags_t)(1 << 0))
2109 /* Supports async buffered writes */
2110 #define FOP_BUFFER_WASYNC	((__force fop_flags_t)(1 << 1))
2111 /* Supports synchronous page faults for mappings */
2112 #define FOP_MMAP_SYNC		((__force fop_flags_t)(1 << 2))
2113 /* Supports non-exclusive O_DIRECT writes from multiple threads */
2114 #define FOP_DIO_PARALLEL_WRITE	((__force fop_flags_t)(1 << 3))
2115 /* Contains huge pages */
2116 #define FOP_HUGE_PAGES		((__force fop_flags_t)(1 << 4))
2117 /* Treat loff_t as unsigned (e.g., /dev/mem) */
2118 #define FOP_UNSIGNED_OFFSET	((__force fop_flags_t)(1 << 5))
2119 
2120 /* Wrap a directory iterator that needs exclusive inode access */
2121 int wrap_directory_iterator(struct file *, struct dir_context *,
2122 			    int (*) (struct file *, struct dir_context *));
2123 #define WRAP_DIR_ITER(x) \
2124 	static int shared_##x(struct file *file , struct dir_context *ctx) \
2125 	{ return wrap_directory_iterator(file, ctx, x); }
2126 
2127 struct inode_operations {
2128 	struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
2129 	const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
2130 	int (*permission) (struct mnt_idmap *, struct inode *, int);
2131 	struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
2132 
2133 	int (*readlink) (struct dentry *, char __user *,int);
2134 
2135 	int (*create) (struct mnt_idmap *, struct inode *,struct dentry *,
2136 		       umode_t, bool);
2137 	int (*link) (struct dentry *,struct inode *,struct dentry *);
2138 	int (*unlink) (struct inode *,struct dentry *);
2139 	int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,
2140 			const char *);
2141 	int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,
2142 		      umode_t);
2143 	int (*rmdir) (struct inode *,struct dentry *);
2144 	int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,
2145 		      umode_t,dev_t);
2146 	int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
2147 			struct inode *, struct dentry *, unsigned int);
2148 	int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
2149 	int (*getattr) (struct mnt_idmap *, const struct path *,
2150 			struct kstat *, u32, unsigned int);
2151 	ssize_t (*listxattr) (struct dentry *, char *, size_t);
2152 	int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
2153 		      u64 len);
2154 	int (*update_time)(struct inode *, int);
2155 	int (*atomic_open)(struct inode *, struct dentry *,
2156 			   struct file *, unsigned open_flag,
2157 			   umode_t create_mode);
2158 	int (*tmpfile) (struct mnt_idmap *, struct inode *,
2159 			struct file *, umode_t);
2160 	struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *,
2161 				     int);
2162 	int (*set_acl)(struct mnt_idmap *, struct dentry *,
2163 		       struct posix_acl *, int);
2164 	int (*fileattr_set)(struct mnt_idmap *idmap,
2165 			    struct dentry *dentry, struct fileattr *fa);
2166 	int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
2167 	struct offset_ctx *(*get_offset_ctx)(struct inode *inode);
2168 } ____cacheline_aligned;
2169 
call_mmap(struct file * file,struct vm_area_struct * vma)2170 static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
2171 {
2172 	return file->f_op->mmap(file, vma);
2173 }
2174 
2175 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
2176 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
2177 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
2178 				   loff_t, size_t, unsigned int);
2179 int remap_verify_area(struct file *file, loff_t pos, loff_t len, bool write);
2180 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2181 				    struct file *file_out, loff_t pos_out,
2182 				    loff_t *len, unsigned int remap_flags,
2183 				    const struct iomap_ops *dax_read_ops);
2184 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2185 				  struct file *file_out, loff_t pos_out,
2186 				  loff_t *count, unsigned int remap_flags);
2187 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
2188 				   struct file *file_out, loff_t pos_out,
2189 				   loff_t len, unsigned int remap_flags);
2190 extern int vfs_dedupe_file_range(struct file *file,
2191 				 struct file_dedupe_range *same);
2192 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
2193 					struct file *dst_file, loff_t dst_pos,
2194 					loff_t len, unsigned int remap_flags);
2195 
2196 /**
2197  * enum freeze_holder - holder of the freeze
2198  * @FREEZE_HOLDER_KERNEL: kernel wants to freeze or thaw filesystem
2199  * @FREEZE_HOLDER_USERSPACE: userspace wants to freeze or thaw filesystem
2200  * @FREEZE_MAY_NEST: whether nesting freeze and thaw requests is allowed
2201  *
2202  * Indicate who the owner of the freeze or thaw request is and whether
2203  * the freeze needs to be exclusive or can nest.
2204  * Without @FREEZE_MAY_NEST, multiple freeze and thaw requests from the
2205  * same holder aren't allowed. It is however allowed to hold a single
2206  * @FREEZE_HOLDER_USERSPACE and a single @FREEZE_HOLDER_KERNEL freeze at
2207  * the same time. This is relied upon by some filesystems during online
2208  * repair or similar.
2209  */
2210 enum freeze_holder {
2211 	FREEZE_HOLDER_KERNEL	= (1U << 0),
2212 	FREEZE_HOLDER_USERSPACE	= (1U << 1),
2213 	FREEZE_MAY_NEST		= (1U << 2),
2214 };
2215 
2216 struct super_operations {
2217    	struct inode *(*alloc_inode)(struct super_block *sb);
2218 	void (*destroy_inode)(struct inode *);
2219 	void (*free_inode)(struct inode *);
2220 
2221    	void (*dirty_inode) (struct inode *, int flags);
2222 	int (*write_inode) (struct inode *, struct writeback_control *wbc);
2223 	int (*drop_inode) (struct inode *);
2224 	void (*evict_inode) (struct inode *);
2225 	void (*put_super) (struct super_block *);
2226 	int (*sync_fs)(struct super_block *sb, int wait);
2227 	int (*freeze_super) (struct super_block *, enum freeze_holder who);
2228 	int (*freeze_fs) (struct super_block *);
2229 	int (*thaw_super) (struct super_block *, enum freeze_holder who);
2230 	int (*unfreeze_fs) (struct super_block *);
2231 	int (*statfs) (struct dentry *, struct kstatfs *);
2232 	int (*remount_fs) (struct super_block *, int *, char *);
2233 	void (*umount_begin) (struct super_block *);
2234 
2235 	int (*show_options)(struct seq_file *, struct dentry *);
2236 	int (*show_devname)(struct seq_file *, struct dentry *);
2237 	int (*show_path)(struct seq_file *, struct dentry *);
2238 	int (*show_stats)(struct seq_file *, struct dentry *);
2239 #ifdef CONFIG_QUOTA
2240 	ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
2241 	ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
2242 	struct dquot __rcu **(*get_dquots)(struct inode *);
2243 #endif
2244 	long (*nr_cached_objects)(struct super_block *,
2245 				  struct shrink_control *);
2246 	long (*free_cached_objects)(struct super_block *,
2247 				    struct shrink_control *);
2248 	void (*shutdown)(struct super_block *sb);
2249 };
2250 
2251 /*
2252  * Inode flags - they have no relation to superblock flags now
2253  */
2254 #define S_SYNC		(1 << 0)  /* Writes are synced at once */
2255 #define S_NOATIME	(1 << 1)  /* Do not update access times */
2256 #define S_APPEND	(1 << 2)  /* Append-only file */
2257 #define S_IMMUTABLE	(1 << 3)  /* Immutable file */
2258 #define S_DEAD		(1 << 4)  /* removed, but still open directory */
2259 #define S_NOQUOTA	(1 << 5)  /* Inode is not counted to quota */
2260 #define S_DIRSYNC	(1 << 6)  /* Directory modifications are synchronous */
2261 #define S_NOCMTIME	(1 << 7)  /* Do not update file c/mtime */
2262 #define S_SWAPFILE	(1 << 8)  /* Do not truncate: swapon got its bmaps */
2263 #define S_PRIVATE	(1 << 9)  /* Inode is fs-internal */
2264 #define S_IMA		(1 << 10) /* Inode has an associated IMA struct */
2265 #define S_AUTOMOUNT	(1 << 11) /* Automount/referral quasi-directory */
2266 #define S_NOSEC		(1 << 12) /* no suid or xattr security attributes */
2267 #ifdef CONFIG_FS_DAX
2268 #define S_DAX		(1 << 13) /* Direct Access, avoiding the page cache */
2269 #else
2270 #define S_DAX		0	  /* Make all the DAX code disappear */
2271 #endif
2272 #define S_ENCRYPTED	(1 << 14) /* Encrypted file (using fs/crypto/) */
2273 #define S_CASEFOLD	(1 << 15) /* Casefolded file */
2274 #define S_VERITY	(1 << 16) /* Verity file (using fs/verity/) */
2275 #define S_KERNEL_FILE	(1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
2276 
2277 /*
2278  * Note that nosuid etc flags are inode-specific: setting some file-system
2279  * flags just means all the inodes inherit those flags by default. It might be
2280  * possible to override it selectively if you really wanted to with some
2281  * ioctl() that is not currently implemented.
2282  *
2283  * Exception: SB_RDONLY is always applied to the entire file system.
2284  *
2285  * Unfortunately, it is possible to change a filesystems flags with it mounted
2286  * with files in use.  This means that all of the inodes will not have their
2287  * i_flags updated.  Hence, i_flags no longer inherit the superblock mount
2288  * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
2289  */
2290 #define __IS_FLG(inode, flg)	((inode)->i_sb->s_flags & (flg))
2291 
sb_rdonly(const struct super_block * sb)2292 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
2293 #define IS_RDONLY(inode)	sb_rdonly((inode)->i_sb)
2294 #define IS_SYNC(inode)		(__IS_FLG(inode, SB_SYNCHRONOUS) || \
2295 					((inode)->i_flags & S_SYNC))
2296 #define IS_DIRSYNC(inode)	(__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
2297 					((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
2298 #define IS_MANDLOCK(inode)	__IS_FLG(inode, SB_MANDLOCK)
2299 #define IS_NOATIME(inode)	__IS_FLG(inode, SB_RDONLY|SB_NOATIME)
2300 #define IS_I_VERSION(inode)	__IS_FLG(inode, SB_I_VERSION)
2301 
2302 #define IS_NOQUOTA(inode)	((inode)->i_flags & S_NOQUOTA)
2303 #define IS_APPEND(inode)	((inode)->i_flags & S_APPEND)
2304 #define IS_IMMUTABLE(inode)	((inode)->i_flags & S_IMMUTABLE)
2305 
2306 #ifdef CONFIG_FS_POSIX_ACL
2307 #define IS_POSIXACL(inode)	__IS_FLG(inode, SB_POSIXACL)
2308 #else
2309 #define IS_POSIXACL(inode)	0
2310 #endif
2311 
2312 #define IS_DEADDIR(inode)	((inode)->i_flags & S_DEAD)
2313 #define IS_NOCMTIME(inode)	((inode)->i_flags & S_NOCMTIME)
2314 
2315 #ifdef CONFIG_SWAP
2316 #define IS_SWAPFILE(inode)	((inode)->i_flags & S_SWAPFILE)
2317 #else
2318 #define IS_SWAPFILE(inode)	((void)(inode), 0U)
2319 #endif
2320 
2321 #define IS_PRIVATE(inode)	((inode)->i_flags & S_PRIVATE)
2322 #define IS_IMA(inode)		((inode)->i_flags & S_IMA)
2323 #define IS_AUTOMOUNT(inode)	((inode)->i_flags & S_AUTOMOUNT)
2324 #define IS_NOSEC(inode)		((inode)->i_flags & S_NOSEC)
2325 #define IS_DAX(inode)		((inode)->i_flags & S_DAX)
2326 #define IS_ENCRYPTED(inode)	((inode)->i_flags & S_ENCRYPTED)
2327 #define IS_CASEFOLDED(inode)	((inode)->i_flags & S_CASEFOLD)
2328 #define IS_VERITY(inode)	((inode)->i_flags & S_VERITY)
2329 
2330 #define IS_WHITEOUT(inode)	(S_ISCHR(inode->i_mode) && \
2331 				 (inode)->i_rdev == WHITEOUT_DEV)
2332 
HAS_UNMAPPED_ID(struct mnt_idmap * idmap,struct inode * inode)2333 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap,
2334 				   struct inode *inode)
2335 {
2336 	return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) ||
2337 	       !vfsgid_valid(i_gid_into_vfsgid(idmap, inode));
2338 }
2339 
init_sync_kiocb(struct kiocb * kiocb,struct file * filp)2340 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
2341 {
2342 	*kiocb = (struct kiocb) {
2343 		.ki_filp = filp,
2344 		.ki_flags = filp->f_iocb_flags,
2345 		.ki_ioprio = get_current_ioprio(),
2346 	};
2347 }
2348 
kiocb_clone(struct kiocb * kiocb,struct kiocb * kiocb_src,struct file * filp)2349 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2350 			       struct file *filp)
2351 {
2352 	*kiocb = (struct kiocb) {
2353 		.ki_filp = filp,
2354 		.ki_flags = kiocb_src->ki_flags,
2355 		.ki_ioprio = kiocb_src->ki_ioprio,
2356 		.ki_pos = kiocb_src->ki_pos,
2357 	};
2358 }
2359 
2360 /*
2361  * Inode state bits.  Protected by inode->i_lock
2362  *
2363  * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2364  * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2365  *
2366  * Four bits define the lifetime of an inode.  Initially, inodes are I_NEW,
2367  * until that flag is cleared.  I_WILL_FREE, I_FREEING and I_CLEAR are set at
2368  * various stages of removing an inode.
2369  *
2370  * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2371  *
2372  * I_DIRTY_SYNC		Inode is dirty, but doesn't have to be written on
2373  *			fdatasync() (unless I_DIRTY_DATASYNC is also set).
2374  *			Timestamp updates are the usual cause.
2375  * I_DIRTY_DATASYNC	Data-related inode changes pending.  We keep track of
2376  *			these changes separately from I_DIRTY_SYNC so that we
2377  *			don't have to write inode on fdatasync() when only
2378  *			e.g. the timestamps have changed.
2379  * I_DIRTY_PAGES	Inode has dirty pages.  Inode itself may be clean.
2380  * I_DIRTY_TIME		The inode itself has dirty timestamps, and the
2381  *			lazytime mount option is enabled.  We keep track of this
2382  *			separately from I_DIRTY_SYNC in order to implement
2383  *			lazytime.  This gets cleared if I_DIRTY_INODE
2384  *			(I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But
2385  *			I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already
2386  *			in place because writeback might already be in progress
2387  *			and we don't want to lose the time update
2388  * I_NEW		Serves as both a mutex and completion notification.
2389  *			New inodes set I_NEW.  If two processes both create
2390  *			the same inode, one of them will release its inode and
2391  *			wait for I_NEW to be released before returning.
2392  *			Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2393  *			also cause waiting on I_NEW, without I_NEW actually
2394  *			being set.  find_inode() uses this to prevent returning
2395  *			nearly-dead inodes.
2396  * I_WILL_FREE		Must be set when calling write_inode_now() if i_count
2397  *			is zero.  I_FREEING must be set when I_WILL_FREE is
2398  *			cleared.
2399  * I_FREEING		Set when inode is about to be freed but still has dirty
2400  *			pages or buffers attached or the inode itself is still
2401  *			dirty.
2402  * I_CLEAR		Added by clear_inode().  In this state the inode is
2403  *			clean and can be destroyed.  Inode keeps I_FREEING.
2404  *
2405  *			Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2406  *			prohibited for many purposes.  iget() must wait for
2407  *			the inode to be completely released, then create it
2408  *			anew.  Other functions will just ignore such inodes,
2409  *			if appropriate.  I_NEW is used for waiting.
2410  *
2411  * I_SYNC		Writeback of inode is running. The bit is set during
2412  *			data writeback, and cleared with a wakeup on the bit
2413  *			address once it is done. The bit is also used to pin
2414  *			the inode in memory for flusher thread.
2415  *
2416  * I_REFERENCED		Marks the inode as recently references on the LRU list.
2417  *
2418  * I_WB_SWITCH		Cgroup bdi_writeback switching in progress.  Used to
2419  *			synchronize competing switching instances and to tell
2420  *			wb stat updates to grab the i_pages lock.  See
2421  *			inode_switch_wbs_work_fn() for details.
2422  *
2423  * I_OVL_INUSE		Used by overlayfs to get exclusive ownership on upper
2424  *			and work dirs among overlayfs mounts.
2425  *
2426  * I_CREATING		New object's inode in the middle of setting up.
2427  *
2428  * I_DONTCACHE		Evict inode as soon as it is not used anymore.
2429  *
2430  * I_SYNC_QUEUED	Inode is queued in b_io or b_more_io writeback lists.
2431  *			Used to detect that mark_inode_dirty() should not move
2432  * 			inode between dirty lists.
2433  *
2434  * I_PINNING_FSCACHE_WB	Inode is pinning an fscache object for writeback.
2435  *
2436  * I_LRU_ISOLATING	Inode is pinned being isolated from LRU without holding
2437  *			i_count.
2438  *
2439  * Q: What is the difference between I_WILL_FREE and I_FREEING?
2440  *
2441  * __I_{SYNC,NEW,LRU_ISOLATING} are used to derive unique addresses to wait
2442  * upon. There's one free address left.
2443  */
2444 #define __I_NEW			0
2445 #define I_NEW			(1 << __I_NEW)
2446 #define __I_SYNC		1
2447 #define I_SYNC			(1 << __I_SYNC)
2448 #define __I_LRU_ISOLATING	2
2449 #define I_LRU_ISOLATING		(1 << __I_LRU_ISOLATING)
2450 
2451 #define I_DIRTY_SYNC		(1 << 3)
2452 #define I_DIRTY_DATASYNC	(1 << 4)
2453 #define I_DIRTY_PAGES		(1 << 5)
2454 #define I_WILL_FREE		(1 << 6)
2455 #define I_FREEING		(1 << 7)
2456 #define I_CLEAR			(1 << 8)
2457 #define I_REFERENCED		(1 << 9)
2458 #define I_LINKABLE		(1 << 10)
2459 #define I_DIRTY_TIME		(1 << 11)
2460 #define I_WB_SWITCH		(1 << 12)
2461 #define I_OVL_INUSE		(1 << 13)
2462 #define I_CREATING		(1 << 14)
2463 #define I_DONTCACHE		(1 << 15)
2464 #define I_SYNC_QUEUED		(1 << 16)
2465 #define I_PINNING_NETFS_WB	(1 << 17)
2466 
2467 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2468 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2469 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2470 
2471 extern void __mark_inode_dirty(struct inode *, int);
mark_inode_dirty(struct inode * inode)2472 static inline void mark_inode_dirty(struct inode *inode)
2473 {
2474 	__mark_inode_dirty(inode, I_DIRTY);
2475 }
2476 
mark_inode_dirty_sync(struct inode * inode)2477 static inline void mark_inode_dirty_sync(struct inode *inode)
2478 {
2479 	__mark_inode_dirty(inode, I_DIRTY_SYNC);
2480 }
2481 
2482 /*
2483  * Returns true if the given inode itself only has dirty timestamps (its pages
2484  * may still be dirty) and isn't currently being allocated or freed.
2485  * Filesystems should call this if when writing an inode when lazytime is
2486  * enabled, they want to opportunistically write the timestamps of other inodes
2487  * located very nearby on-disk, e.g. in the same inode block.  This returns true
2488  * if the given inode is in need of such an opportunistic update.  Requires
2489  * i_lock, or at least later re-checking under i_lock.
2490  */
inode_is_dirtytime_only(struct inode * inode)2491 static inline bool inode_is_dirtytime_only(struct inode *inode)
2492 {
2493 	return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2494 				  I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2495 }
2496 
2497 extern void inc_nlink(struct inode *inode);
2498 extern void drop_nlink(struct inode *inode);
2499 extern void clear_nlink(struct inode *inode);
2500 extern void set_nlink(struct inode *inode, unsigned int nlink);
2501 
inode_inc_link_count(struct inode * inode)2502 static inline void inode_inc_link_count(struct inode *inode)
2503 {
2504 	inc_nlink(inode);
2505 	mark_inode_dirty(inode);
2506 }
2507 
inode_dec_link_count(struct inode * inode)2508 static inline void inode_dec_link_count(struct inode *inode)
2509 {
2510 	drop_nlink(inode);
2511 	mark_inode_dirty(inode);
2512 }
2513 
2514 enum file_time_flags {
2515 	S_ATIME = 1,
2516 	S_MTIME = 2,
2517 	S_CTIME = 4,
2518 	S_VERSION = 8,
2519 };
2520 
2521 extern bool atime_needs_update(const struct path *, struct inode *);
2522 extern void touch_atime(const struct path *);
2523 int inode_update_time(struct inode *inode, int flags);
2524 
file_accessed(struct file * file)2525 static inline void file_accessed(struct file *file)
2526 {
2527 	if (!(file->f_flags & O_NOATIME))
2528 		touch_atime(&file->f_path);
2529 }
2530 
2531 extern int file_modified(struct file *file);
2532 int kiocb_modified(struct kiocb *iocb);
2533 
2534 int sync_inode_metadata(struct inode *inode, int wait);
2535 
2536 struct file_system_type {
2537 	const char *name;
2538 	int fs_flags;
2539 #define FS_REQUIRES_DEV		1
2540 #define FS_BINARY_MOUNTDATA	2
2541 #define FS_HAS_SUBTYPE		4
2542 #define FS_USERNS_MOUNT		8	/* Can be mounted by userns root */
2543 #define FS_DISALLOW_NOTIFY_PERM	16	/* Disable fanotify permission events */
2544 #define FS_ALLOW_IDMAP         32      /* FS has been updated to handle vfs idmappings. */
2545 #define FS_RENAME_DOES_D_MOVE	32768	/* FS will handle d_move() during rename() internally. */
2546 	int (*init_fs_context)(struct fs_context *);
2547 	const struct fs_parameter_spec *parameters;
2548 	struct dentry *(*mount) (struct file_system_type *, int,
2549 		       const char *, void *);
2550 	void (*kill_sb) (struct super_block *);
2551 	struct module *owner;
2552 	struct file_system_type * next;
2553 	struct hlist_head fs_supers;
2554 
2555 	struct lock_class_key s_lock_key;
2556 	struct lock_class_key s_umount_key;
2557 	struct lock_class_key s_vfs_rename_key;
2558 	struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2559 
2560 	struct lock_class_key i_lock_key;
2561 	struct lock_class_key i_mutex_key;
2562 	struct lock_class_key invalidate_lock_key;
2563 	struct lock_class_key i_mutex_dir_key;
2564 };
2565 
2566 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2567 
2568 extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2569 	int flags, const char *dev_name, void *data,
2570 	int (*fill_super)(struct super_block *, void *, int));
2571 extern struct dentry *mount_single(struct file_system_type *fs_type,
2572 	int flags, void *data,
2573 	int (*fill_super)(struct super_block *, void *, int));
2574 extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2575 	int flags, void *data,
2576 	int (*fill_super)(struct super_block *, void *, int));
2577 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2578 void retire_super(struct super_block *sb);
2579 void generic_shutdown_super(struct super_block *sb);
2580 void kill_block_super(struct super_block *sb);
2581 void kill_anon_super(struct super_block *sb);
2582 void kill_litter_super(struct super_block *sb);
2583 void deactivate_super(struct super_block *sb);
2584 void deactivate_locked_super(struct super_block *sb);
2585 int set_anon_super(struct super_block *s, void *data);
2586 int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2587 int get_anon_bdev(dev_t *);
2588 void free_anon_bdev(dev_t);
2589 struct super_block *sget_fc(struct fs_context *fc,
2590 			    int (*test)(struct super_block *, struct fs_context *),
2591 			    int (*set)(struct super_block *, struct fs_context *));
2592 struct super_block *sget(struct file_system_type *type,
2593 			int (*test)(struct super_block *,void *),
2594 			int (*set)(struct super_block *,void *),
2595 			int flags, void *data);
2596 struct super_block *sget_dev(struct fs_context *fc, dev_t dev);
2597 
2598 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2599 #define fops_get(fops) ({						\
2600 	const struct file_operations *_fops = (fops);			\
2601 	(((_fops) && try_module_get((_fops)->owner) ? (_fops) : NULL));	\
2602 })
2603 
2604 #define fops_put(fops) ({						\
2605 	const struct file_operations *_fops = (fops);			\
2606 	if (_fops)							\
2607 		module_put((_fops)->owner);				\
2608 })
2609 
2610 /*
2611  * This one is to be used *ONLY* from ->open() instances.
2612  * fops must be non-NULL, pinned down *and* module dependencies
2613  * should be sufficient to pin the caller down as well.
2614  */
2615 #define replace_fops(f, fops) \
2616 	do {	\
2617 		struct file *__file = (f); \
2618 		fops_put(__file->f_op); \
2619 		BUG_ON(!(__file->f_op = (fops))); \
2620 	} while(0)
2621 
2622 extern int register_filesystem(struct file_system_type *);
2623 extern int unregister_filesystem(struct file_system_type *);
2624 extern int vfs_statfs(const struct path *, struct kstatfs *);
2625 extern int user_statfs(const char __user *, struct kstatfs *);
2626 extern int fd_statfs(int, struct kstatfs *);
2627 int freeze_super(struct super_block *super, enum freeze_holder who);
2628 int thaw_super(struct super_block *super, enum freeze_holder who);
2629 extern __printf(2, 3)
2630 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2631 extern int super_setup_bdi(struct super_block *sb);
2632 
super_set_uuid(struct super_block * sb,const u8 * uuid,unsigned len)2633 static inline void super_set_uuid(struct super_block *sb, const u8 *uuid, unsigned len)
2634 {
2635 	if (WARN_ON(len > sizeof(sb->s_uuid)))
2636 		len = sizeof(sb->s_uuid);
2637 	sb->s_uuid_len = len;
2638 	memcpy(&sb->s_uuid, uuid, len);
2639 }
2640 
2641 /* set sb sysfs name based on sb->s_bdev */
super_set_sysfs_name_bdev(struct super_block * sb)2642 static inline void super_set_sysfs_name_bdev(struct super_block *sb)
2643 {
2644 	snprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), "%pg", sb->s_bdev);
2645 }
2646 
2647 /* set sb sysfs name based on sb->s_uuid */
super_set_sysfs_name_uuid(struct super_block * sb)2648 static inline void super_set_sysfs_name_uuid(struct super_block *sb)
2649 {
2650 	WARN_ON(sb->s_uuid_len != sizeof(sb->s_uuid));
2651 	snprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), "%pU", sb->s_uuid.b);
2652 }
2653 
2654 /* set sb sysfs name based on sb->s_id */
super_set_sysfs_name_id(struct super_block * sb)2655 static inline void super_set_sysfs_name_id(struct super_block *sb)
2656 {
2657 	strscpy(sb->s_sysfs_name, sb->s_id, sizeof(sb->s_sysfs_name));
2658 }
2659 
2660 /* try to use something standard before you use this */
2661 __printf(2, 3)
super_set_sysfs_name_generic(struct super_block * sb,const char * fmt,...)2662 static inline void super_set_sysfs_name_generic(struct super_block *sb, const char *fmt, ...)
2663 {
2664 	va_list args;
2665 
2666 	va_start(args, fmt);
2667 	vsnprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), fmt, args);
2668 	va_end(args);
2669 }
2670 
2671 extern int current_umask(void);
2672 
2673 extern void ihold(struct inode * inode);
2674 extern void iput(struct inode *);
2675 int inode_update_timestamps(struct inode *inode, int flags);
2676 int generic_update_time(struct inode *, int);
2677 
2678 /* /sys/fs */
2679 extern struct kobject *fs_kobj;
2680 
2681 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2682 
2683 /* fs/open.c */
2684 struct audit_names;
2685 struct filename {
2686 	const char		*name;	/* pointer to actual string */
2687 	const __user char	*uptr;	/* original userland pointer */
2688 	atomic_t		refcnt;
2689 	struct audit_names	*aname;
2690 	const char		iname[];
2691 };
2692 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2693 
file_mnt_idmap(const struct file * file)2694 static inline struct mnt_idmap *file_mnt_idmap(const struct file *file)
2695 {
2696 	return mnt_idmap(file->f_path.mnt);
2697 }
2698 
2699 /**
2700  * is_idmapped_mnt - check whether a mount is mapped
2701  * @mnt: the mount to check
2702  *
2703  * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped.
2704  *
2705  * Return: true if mount is mapped, false if not.
2706  */
is_idmapped_mnt(const struct vfsmount * mnt)2707 static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2708 {
2709 	return mnt_idmap(mnt) != &nop_mnt_idmap;
2710 }
2711 
2712 extern long vfs_truncate(const struct path *, loff_t);
2713 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start,
2714 		unsigned int time_attrs, struct file *filp);
2715 extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2716 			loff_t len);
2717 extern long do_sys_open(int dfd, const char __user *filename, int flags,
2718 			umode_t mode);
2719 extern struct file *file_open_name(struct filename *, int, umode_t);
2720 extern struct file *filp_open(const char *, int, umode_t);
2721 extern struct file *file_open_root(const struct path *,
2722 				   const char *, int, umode_t);
file_open_root_mnt(struct vfsmount * mnt,const char * name,int flags,umode_t mode)2723 static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2724 				   const char *name, int flags, umode_t mode)
2725 {
2726 	return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2727 			      name, flags, mode);
2728 }
2729 struct file *dentry_open(const struct path *path, int flags,
2730 			 const struct cred *creds);
2731 struct file *dentry_create(const struct path *path, int flags, umode_t mode,
2732 			   const struct cred *cred);
2733 struct path *backing_file_user_path(struct file *f);
2734 
2735 /*
2736  * When mmapping a file on a stackable filesystem (e.g., overlayfs), the file
2737  * stored in ->vm_file is a backing file whose f_inode is on the underlying
2738  * filesystem.  When the mapped file path and inode number are displayed to
2739  * user (e.g. via /proc/<pid>/maps), these helpers should be used to get the
2740  * path and inode number to display to the user, which is the path of the fd
2741  * that user has requested to map and the inode number that would be returned
2742  * by fstat() on that same fd.
2743  */
2744 /* Get the path to display in /proc/<pid>/maps */
file_user_path(struct file * f)2745 static inline const struct path *file_user_path(struct file *f)
2746 {
2747 	if (unlikely(f->f_mode & FMODE_BACKING))
2748 		return backing_file_user_path(f);
2749 	return &f->f_path;
2750 }
2751 /* Get the inode whose inode number to display in /proc/<pid>/maps */
file_user_inode(struct file * f)2752 static inline const struct inode *file_user_inode(struct file *f)
2753 {
2754 	if (unlikely(f->f_mode & FMODE_BACKING))
2755 		return d_inode(backing_file_user_path(f)->dentry);
2756 	return file_inode(f);
2757 }
2758 
file_clone_open(struct file * file)2759 static inline struct file *file_clone_open(struct file *file)
2760 {
2761 	return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2762 }
2763 extern int filp_close(struct file *, fl_owner_t id);
2764 
2765 extern struct filename *getname_flags(const char __user *, int);
2766 extern struct filename *getname_uflags(const char __user *, int);
2767 extern struct filename *getname(const char __user *);
2768 extern struct filename *getname_kernel(const char *);
2769 extern void putname(struct filename *name);
2770 
2771 extern int finish_open(struct file *file, struct dentry *dentry,
2772 			int (*open)(struct inode *, struct file *));
2773 extern int finish_no_open(struct file *file, struct dentry *dentry);
2774 
2775 /* Helper for the simple case when original dentry is used */
finish_open_simple(struct file * file,int error)2776 static inline int finish_open_simple(struct file *file, int error)
2777 {
2778 	if (error)
2779 		return error;
2780 
2781 	return finish_open(file, file->f_path.dentry, NULL);
2782 }
2783 
2784 /* fs/dcache.c */
2785 extern void __init vfs_caches_init_early(void);
2786 extern void __init vfs_caches_init(void);
2787 
2788 extern struct kmem_cache *names_cachep;
2789 
2790 #define __getname()		kmem_cache_alloc(names_cachep, GFP_KERNEL)
2791 #define __putname(name)		kmem_cache_free(names_cachep, (void *)(name))
2792 
2793 extern struct super_block *blockdev_superblock;
sb_is_blkdev_sb(struct super_block * sb)2794 static inline bool sb_is_blkdev_sb(struct super_block *sb)
2795 {
2796 	return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2797 }
2798 
2799 void emergency_thaw_all(void);
2800 extern int sync_filesystem(struct super_block *);
2801 extern const struct file_operations def_blk_fops;
2802 extern const struct file_operations def_chr_fops;
2803 
2804 /* fs/char_dev.c */
2805 #define CHRDEV_MAJOR_MAX 512
2806 /* Marks the bottom of the first segment of free char majors */
2807 #define CHRDEV_MAJOR_DYN_END 234
2808 /* Marks the top and bottom of the second segment of free char majors */
2809 #define CHRDEV_MAJOR_DYN_EXT_START 511
2810 #define CHRDEV_MAJOR_DYN_EXT_END 384
2811 
2812 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2813 extern int register_chrdev_region(dev_t, unsigned, const char *);
2814 extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2815 			     unsigned int count, const char *name,
2816 			     const struct file_operations *fops);
2817 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2818 				unsigned int count, const char *name);
2819 extern void unregister_chrdev_region(dev_t, unsigned);
2820 extern void chrdev_show(struct seq_file *,off_t);
2821 
register_chrdev(unsigned int major,const char * name,const struct file_operations * fops)2822 static inline int register_chrdev(unsigned int major, const char *name,
2823 				  const struct file_operations *fops)
2824 {
2825 	return __register_chrdev(major, 0, 256, name, fops);
2826 }
2827 
unregister_chrdev(unsigned int major,const char * name)2828 static inline void unregister_chrdev(unsigned int major, const char *name)
2829 {
2830 	__unregister_chrdev(major, 0, 256, name);
2831 }
2832 
2833 extern void init_special_inode(struct inode *, umode_t, dev_t);
2834 
2835 /* Invalid inode operations -- fs/bad_inode.c */
2836 extern void make_bad_inode(struct inode *);
2837 extern bool is_bad_inode(struct inode *);
2838 
2839 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2840 						loff_t lend);
2841 extern int __must_check file_check_and_advance_wb_err(struct file *file);
2842 extern int __must_check file_write_and_wait_range(struct file *file,
2843 						loff_t start, loff_t end);
2844 
file_write_and_wait(struct file * file)2845 static inline int file_write_and_wait(struct file *file)
2846 {
2847 	return file_write_and_wait_range(file, 0, LLONG_MAX);
2848 }
2849 
2850 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2851 			   int datasync);
2852 extern int vfs_fsync(struct file *file, int datasync);
2853 
2854 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2855 				unsigned int flags);
2856 
iocb_is_dsync(const struct kiocb * iocb)2857 static inline bool iocb_is_dsync(const struct kiocb *iocb)
2858 {
2859 	return (iocb->ki_flags & IOCB_DSYNC) ||
2860 		IS_SYNC(iocb->ki_filp->f_mapping->host);
2861 }
2862 
2863 /*
2864  * Sync the bytes written if this was a synchronous write.  Expect ki_pos
2865  * to already be updated for the write, and will return either the amount
2866  * of bytes passed in, or an error if syncing the file failed.
2867  */
generic_write_sync(struct kiocb * iocb,ssize_t count)2868 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2869 {
2870 	if (iocb_is_dsync(iocb)) {
2871 		int ret = vfs_fsync_range(iocb->ki_filp,
2872 				iocb->ki_pos - count, iocb->ki_pos - 1,
2873 				(iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2874 		if (ret)
2875 			return ret;
2876 	}
2877 
2878 	return count;
2879 }
2880 
2881 extern void emergency_sync(void);
2882 extern void emergency_remount(void);
2883 
2884 #ifdef CONFIG_BLOCK
2885 extern int bmap(struct inode *inode, sector_t *block);
2886 #else
bmap(struct inode * inode,sector_t * block)2887 static inline int bmap(struct inode *inode,  sector_t *block)
2888 {
2889 	return -EINVAL;
2890 }
2891 #endif
2892 
2893 int notify_change(struct mnt_idmap *, struct dentry *,
2894 		  struct iattr *, struct inode **);
2895 int inode_permission(struct mnt_idmap *, struct inode *, int);
2896 int generic_permission(struct mnt_idmap *, struct inode *, int);
file_permission(struct file * file,int mask)2897 static inline int file_permission(struct file *file, int mask)
2898 {
2899 	return inode_permission(file_mnt_idmap(file),
2900 				file_inode(file), mask);
2901 }
path_permission(const struct path * path,int mask)2902 static inline int path_permission(const struct path *path, int mask)
2903 {
2904 	return inode_permission(mnt_idmap(path->mnt),
2905 				d_inode(path->dentry), mask);
2906 }
2907 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir,
2908 		   struct inode *inode);
2909 
execute_ok(struct inode * inode)2910 static inline bool execute_ok(struct inode *inode)
2911 {
2912 	return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2913 }
2914 
inode_wrong_type(const struct inode * inode,umode_t mode)2915 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2916 {
2917 	return (inode->i_mode ^ mode) & S_IFMT;
2918 }
2919 
2920 /**
2921  * file_start_write - get write access to a superblock for regular file io
2922  * @file: the file we want to write to
2923  *
2924  * This is a variant of sb_start_write() which is a noop on non-regualr file.
2925  * Should be matched with a call to file_end_write().
2926  */
file_start_write(struct file * file)2927 static inline void file_start_write(struct file *file)
2928 {
2929 	if (!S_ISREG(file_inode(file)->i_mode))
2930 		return;
2931 	sb_start_write(file_inode(file)->i_sb);
2932 }
2933 
file_start_write_trylock(struct file * file)2934 static inline bool file_start_write_trylock(struct file *file)
2935 {
2936 	if (!S_ISREG(file_inode(file)->i_mode))
2937 		return true;
2938 	return sb_start_write_trylock(file_inode(file)->i_sb);
2939 }
2940 
2941 /**
2942  * file_end_write - drop write access to a superblock of a regular file
2943  * @file: the file we wrote to
2944  *
2945  * Should be matched with a call to file_start_write().
2946  */
file_end_write(struct file * file)2947 static inline void file_end_write(struct file *file)
2948 {
2949 	if (!S_ISREG(file_inode(file)->i_mode))
2950 		return;
2951 	sb_end_write(file_inode(file)->i_sb);
2952 }
2953 
2954 /**
2955  * kiocb_start_write - get write access to a superblock for async file io
2956  * @iocb: the io context we want to submit the write with
2957  *
2958  * This is a variant of sb_start_write() for async io submission.
2959  * Should be matched with a call to kiocb_end_write().
2960  */
kiocb_start_write(struct kiocb * iocb)2961 static inline void kiocb_start_write(struct kiocb *iocb)
2962 {
2963 	struct inode *inode = file_inode(iocb->ki_filp);
2964 
2965 	sb_start_write(inode->i_sb);
2966 	/*
2967 	 * Fool lockdep by telling it the lock got released so that it
2968 	 * doesn't complain about the held lock when we return to userspace.
2969 	 */
2970 	__sb_writers_release(inode->i_sb, SB_FREEZE_WRITE);
2971 }
2972 
2973 /**
2974  * kiocb_end_write - drop write access to a superblock after async file io
2975  * @iocb: the io context we sumbitted the write with
2976  *
2977  * Should be matched with a call to kiocb_start_write().
2978  */
kiocb_end_write(struct kiocb * iocb)2979 static inline void kiocb_end_write(struct kiocb *iocb)
2980 {
2981 	struct inode *inode = file_inode(iocb->ki_filp);
2982 
2983 	/*
2984 	 * Tell lockdep we inherited freeze protection from submission thread.
2985 	 */
2986 	__sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
2987 	sb_end_write(inode->i_sb);
2988 }
2989 
2990 /*
2991  * This is used for regular files where some users -- especially the
2992  * currently executed binary in a process, previously handled via
2993  * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2994  * read-write shared) accesses.
2995  *
2996  * get_write_access() gets write permission for a file.
2997  * put_write_access() releases this write permission.
2998  * deny_write_access() denies write access to a file.
2999  * allow_write_access() re-enables write access to a file.
3000  *
3001  * The i_writecount field of an inode can have the following values:
3002  * 0: no write access, no denied write access
3003  * < 0: (-i_writecount) users that denied write access to the file.
3004  * > 0: (i_writecount) users that have write access to the file.
3005  *
3006  * Normally we operate on that counter with atomic_{inc,dec} and it's safe
3007  * except for the cases where we don't hold i_writecount yet. Then we need to
3008  * use {get,deny}_write_access() - these functions check the sign and refuse
3009  * to do the change if sign is wrong.
3010  */
get_write_access(struct inode * inode)3011 static inline int get_write_access(struct inode *inode)
3012 {
3013 	return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
3014 }
deny_write_access(struct file * file)3015 static inline int deny_write_access(struct file *file)
3016 {
3017 	struct inode *inode = file_inode(file);
3018 	return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
3019 }
put_write_access(struct inode * inode)3020 static inline void put_write_access(struct inode * inode)
3021 {
3022 	atomic_dec(&inode->i_writecount);
3023 }
allow_write_access(struct file * file)3024 static inline void allow_write_access(struct file *file)
3025 {
3026 	if (file)
3027 		atomic_inc(&file_inode(file)->i_writecount);
3028 }
inode_is_open_for_write(const struct inode * inode)3029 static inline bool inode_is_open_for_write(const struct inode *inode)
3030 {
3031 	return atomic_read(&inode->i_writecount) > 0;
3032 }
3033 
3034 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
i_readcount_dec(struct inode * inode)3035 static inline void i_readcount_dec(struct inode *inode)
3036 {
3037 	BUG_ON(atomic_dec_return(&inode->i_readcount) < 0);
3038 }
i_readcount_inc(struct inode * inode)3039 static inline void i_readcount_inc(struct inode *inode)
3040 {
3041 	atomic_inc(&inode->i_readcount);
3042 }
3043 #else
i_readcount_dec(struct inode * inode)3044 static inline void i_readcount_dec(struct inode *inode)
3045 {
3046 	return;
3047 }
i_readcount_inc(struct inode * inode)3048 static inline void i_readcount_inc(struct inode *inode)
3049 {
3050 	return;
3051 }
3052 #endif
3053 extern int do_pipe_flags(int *, int);
3054 
3055 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
3056 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
3057 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
3058 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
3059 extern struct file * open_exec(const char *);
3060 
3061 /* fs/dcache.c -- generic fs support functions */
3062 extern bool is_subdir(struct dentry *, struct dentry *);
3063 extern bool path_is_under(const struct path *, const struct path *);
3064 
3065 extern char *file_path(struct file *, char *, int);
3066 
3067 /**
3068  * is_dot_dotdot - returns true only if @name is "." or ".."
3069  * @name: file name to check
3070  * @len: length of file name, in bytes
3071  */
is_dot_dotdot(const char * name,size_t len)3072 static inline bool is_dot_dotdot(const char *name, size_t len)
3073 {
3074 	return len && unlikely(name[0] == '.') &&
3075 		(len == 1 || (len == 2 && name[1] == '.'));
3076 }
3077 
3078 #include <linux/err.h>
3079 
3080 /* needed for stackable file system support */
3081 extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
3082 
3083 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
3084 
3085 extern int inode_init_always(struct super_block *, struct inode *);
3086 extern void inode_init_once(struct inode *);
3087 extern void address_space_init_once(struct address_space *mapping);
3088 extern struct inode * igrab(struct inode *);
3089 extern ino_t iunique(struct super_block *, ino_t);
3090 extern int inode_needs_sync(struct inode *inode);
3091 extern int generic_delete_inode(struct inode *inode);
generic_drop_inode(struct inode * inode)3092 static inline int generic_drop_inode(struct inode *inode)
3093 {
3094 	return !inode->i_nlink || inode_unhashed(inode);
3095 }
3096 extern void d_mark_dontcache(struct inode *inode);
3097 
3098 extern struct inode *ilookup5_nowait(struct super_block *sb,
3099 		unsigned long hashval, int (*test)(struct inode *, void *),
3100 		void *data);
3101 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
3102 		int (*test)(struct inode *, void *), void *data);
3103 extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
3104 
3105 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
3106 		int (*test)(struct inode *, void *),
3107 		int (*set)(struct inode *, void *),
3108 		void *data);
3109 struct inode *iget5_locked(struct super_block *, unsigned long,
3110 			   int (*test)(struct inode *, void *),
3111 			   int (*set)(struct inode *, void *), void *);
3112 struct inode *iget5_locked_rcu(struct super_block *, unsigned long,
3113 			       int (*test)(struct inode *, void *),
3114 			       int (*set)(struct inode *, void *), void *);
3115 extern struct inode * iget_locked(struct super_block *, unsigned long);
3116 extern struct inode *find_inode_nowait(struct super_block *,
3117 				       unsigned long,
3118 				       int (*match)(struct inode *,
3119 						    unsigned long, void *),
3120 				       void *data);
3121 extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
3122 				    int (*)(struct inode *, void *), void *);
3123 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
3124 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
3125 extern int insert_inode_locked(struct inode *);
3126 #ifdef CONFIG_DEBUG_LOCK_ALLOC
3127 extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
3128 #else
lockdep_annotate_inode_mutex_key(struct inode * inode)3129 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
3130 #endif
3131 extern void unlock_new_inode(struct inode *);
3132 extern void discard_new_inode(struct inode *);
3133 extern unsigned int get_next_ino(void);
3134 extern void evict_inodes(struct super_block *sb);
3135 void dump_mapping(const struct address_space *);
3136 
3137 /*
3138  * Userspace may rely on the inode number being non-zero. For example, glibc
3139  * simply ignores files with zero i_ino in unlink() and other places.
3140  *
3141  * As an additional complication, if userspace was compiled with
3142  * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
3143  * lower 32 bits, so we need to check that those aren't zero explicitly. With
3144  * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
3145  * better safe than sorry.
3146  */
is_zero_ino(ino_t ino)3147 static inline bool is_zero_ino(ino_t ino)
3148 {
3149 	return (u32)ino == 0;
3150 }
3151 
3152 /*
3153  * inode->i_lock must be held
3154  */
__iget(struct inode * inode)3155 static inline void __iget(struct inode *inode)
3156 {
3157 	atomic_inc(&inode->i_count);
3158 }
3159 
3160 extern void iget_failed(struct inode *);
3161 extern void clear_inode(struct inode *);
3162 extern void __destroy_inode(struct inode *);
3163 extern struct inode *new_inode_pseudo(struct super_block *sb);
3164 extern struct inode *new_inode(struct super_block *sb);
3165 extern void free_inode_nonrcu(struct inode *inode);
3166 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
3167 extern int file_remove_privs_flags(struct file *file, unsigned int flags);
3168 extern int file_remove_privs(struct file *);
3169 int setattr_should_drop_sgid(struct mnt_idmap *idmap,
3170 			     const struct inode *inode);
3171 
3172 /*
3173  * This must be used for allocating filesystems specific inodes to set
3174  * up the inode reclaim context correctly.
3175  */
3176 #define alloc_inode_sb(_sb, _cache, _gfp) kmem_cache_alloc_lru(_cache, &_sb->s_inode_lru, _gfp)
3177 
3178 extern void __insert_inode_hash(struct inode *, unsigned long hashval);
insert_inode_hash(struct inode * inode)3179 static inline void insert_inode_hash(struct inode *inode)
3180 {
3181 	__insert_inode_hash(inode, inode->i_ino);
3182 }
3183 
3184 extern void __remove_inode_hash(struct inode *);
remove_inode_hash(struct inode * inode)3185 static inline void remove_inode_hash(struct inode *inode)
3186 {
3187 	if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
3188 		__remove_inode_hash(inode);
3189 }
3190 
3191 extern void inode_sb_list_add(struct inode *inode);
3192 extern void inode_add_lru(struct inode *inode);
3193 
3194 extern int sb_set_blocksize(struct super_block *, int);
3195 extern int sb_min_blocksize(struct super_block *, int);
3196 
3197 extern int generic_file_mmap(struct file *, struct vm_area_struct *);
3198 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
3199 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
3200 int generic_write_checks_count(struct kiocb *iocb, loff_t *count);
3201 extern int generic_write_check_limits(struct file *file, loff_t pos,
3202 		loff_t *count);
3203 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
3204 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
3205 		ssize_t already_read);
3206 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
3207 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
3208 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
3209 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
3210 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *);
3211 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
3212 		ssize_t direct_written, ssize_t buffered_written);
3213 
3214 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
3215 		rwf_t flags);
3216 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
3217 		rwf_t flags);
3218 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
3219 			   struct iov_iter *iter);
3220 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
3221 			    struct iov_iter *iter);
3222 
3223 /* fs/splice.c */
3224 ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
3225 			    struct pipe_inode_info *pipe,
3226 			    size_t len, unsigned int flags);
3227 ssize_t copy_splice_read(struct file *in, loff_t *ppos,
3228 			 struct pipe_inode_info *pipe,
3229 			 size_t len, unsigned int flags);
3230 extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
3231 		struct file *, loff_t *, size_t, unsigned int);
3232 
3233 
3234 extern void
3235 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
3236 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3237 #define no_llseek NULL
3238 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3239 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3240 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3241 		int whence, loff_t maxsize, loff_t eof);
3242 loff_t generic_llseek_cookie(struct file *file, loff_t offset, int whence,
3243 			     u64 *cookie);
3244 extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3245 		int whence, loff_t size);
3246 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3247 extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3248 int rw_verify_area(int, struct file *, const loff_t *, size_t);
3249 extern int generic_file_open(struct inode * inode, struct file * filp);
3250 extern int nonseekable_open(struct inode * inode, struct file * filp);
3251 extern int stream_open(struct inode * inode, struct file * filp);
3252 
3253 #ifdef CONFIG_BLOCK
3254 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
3255 			    loff_t file_offset);
3256 
3257 enum {
3258 	/* need locking between buffered and direct access */
3259 	DIO_LOCKING	= 0x01,
3260 
3261 	/* filesystem does not support filling holes */
3262 	DIO_SKIP_HOLES	= 0x02,
3263 };
3264 
3265 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
3266 			     struct block_device *bdev, struct iov_iter *iter,
3267 			     get_block_t get_block,
3268 			     dio_iodone_t end_io,
3269 			     int flags);
3270 
blockdev_direct_IO(struct kiocb * iocb,struct inode * inode,struct iov_iter * iter,get_block_t get_block)3271 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3272 					 struct inode *inode,
3273 					 struct iov_iter *iter,
3274 					 get_block_t get_block)
3275 {
3276 	return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
3277 			get_block, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
3278 }
3279 #endif
3280 
3281 bool inode_dio_finished(const struct inode *inode);
3282 void inode_dio_wait(struct inode *inode);
3283 void inode_dio_wait_interruptible(struct inode *inode);
3284 
3285 /**
3286  * inode_dio_begin - signal start of a direct I/O requests
3287  * @inode: inode the direct I/O happens on
3288  *
3289  * This is called once we've finished processing a direct I/O request,
3290  * and is used to wake up callers waiting for direct I/O to be quiesced.
3291  */
inode_dio_begin(struct inode * inode)3292 static inline void inode_dio_begin(struct inode *inode)
3293 {
3294 	atomic_inc(&inode->i_dio_count);
3295 }
3296 
3297 /**
3298  * inode_dio_end - signal finish of a direct I/O requests
3299  * @inode: inode the direct I/O happens on
3300  *
3301  * This is called once we've finished processing a direct I/O request,
3302  * and is used to wake up callers waiting for direct I/O to be quiesced.
3303  */
inode_dio_end(struct inode * inode)3304 static inline void inode_dio_end(struct inode *inode)
3305 {
3306 	if (atomic_dec_and_test(&inode->i_dio_count))
3307 		wake_up_var(&inode->i_dio_count);
3308 }
3309 
3310 extern void inode_set_flags(struct inode *inode, unsigned int flags,
3311 			    unsigned int mask);
3312 
3313 extern const struct file_operations generic_ro_fops;
3314 
3315 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
3316 
3317 extern int readlink_copy(char __user *, int, const char *);
3318 extern int page_readlink(struct dentry *, char __user *, int);
3319 extern const char *page_get_link(struct dentry *, struct inode *,
3320 				 struct delayed_call *);
3321 extern void page_put_link(void *);
3322 extern int page_symlink(struct inode *inode, const char *symname, int len);
3323 extern const struct inode_operations page_symlink_inode_operations;
3324 extern void kfree_link(void *);
3325 void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *);
3326 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
3327 void generic_fill_statx_atomic_writes(struct kstat *stat,
3328 				      unsigned int unit_min,
3329 				      unsigned int unit_max);
3330 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
3331 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
3332 void __inode_add_bytes(struct inode *inode, loff_t bytes);
3333 void inode_add_bytes(struct inode *inode, loff_t bytes);
3334 void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3335 void inode_sub_bytes(struct inode *inode, loff_t bytes);
__inode_get_bytes(struct inode * inode)3336 static inline loff_t __inode_get_bytes(struct inode *inode)
3337 {
3338 	return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3339 }
3340 loff_t inode_get_bytes(struct inode *inode);
3341 void inode_set_bytes(struct inode *inode, loff_t bytes);
3342 const char *simple_get_link(struct dentry *, struct inode *,
3343 			    struct delayed_call *);
3344 extern const struct inode_operations simple_symlink_inode_operations;
3345 
3346 extern int iterate_dir(struct file *, struct dir_context *);
3347 
3348 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
3349 		int flags);
3350 int vfs_fstat(int fd, struct kstat *stat);
3351 
vfs_stat(const char __user * filename,struct kstat * stat)3352 static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3353 {
3354 	return vfs_fstatat(AT_FDCWD, filename, stat, 0);
3355 }
vfs_lstat(const char __user * name,struct kstat * stat)3356 static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3357 {
3358 	return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
3359 }
3360 
3361 extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3362 extern int vfs_readlink(struct dentry *, char __user *, int);
3363 
3364 extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3365 extern void put_filesystem(struct file_system_type *fs);
3366 extern struct file_system_type *get_fs_type(const char *name);
3367 extern void drop_super(struct super_block *sb);
3368 extern void drop_super_exclusive(struct super_block *sb);
3369 extern void iterate_supers(void (*)(struct super_block *, void *), void *);
3370 extern void iterate_supers_type(struct file_system_type *,
3371 			        void (*)(struct super_block *, void *), void *);
3372 
3373 extern int dcache_dir_open(struct inode *, struct file *);
3374 extern int dcache_dir_close(struct inode *, struct file *);
3375 extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3376 extern int dcache_readdir(struct file *, struct dir_context *);
3377 extern int simple_setattr(struct mnt_idmap *, struct dentry *,
3378 			  struct iattr *);
3379 extern int simple_getattr(struct mnt_idmap *, const struct path *,
3380 			  struct kstat *, u32, unsigned int);
3381 extern int simple_statfs(struct dentry *, struct kstatfs *);
3382 extern int simple_open(struct inode *inode, struct file *file);
3383 extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3384 extern int simple_unlink(struct inode *, struct dentry *);
3385 extern int simple_rmdir(struct inode *, struct dentry *);
3386 void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry,
3387 			     struct inode *new_dir, struct dentry *new_dentry);
3388 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
3389 				  struct inode *new_dir, struct dentry *new_dentry);
3390 extern int simple_rename(struct mnt_idmap *, struct inode *,
3391 			 struct dentry *, struct inode *, struct dentry *,
3392 			 unsigned int);
3393 extern void simple_recursive_removal(struct dentry *,
3394                               void (*callback)(struct dentry *));
3395 extern int noop_fsync(struct file *, loff_t, loff_t, int);
3396 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
3397 extern int simple_empty(struct dentry *);
3398 extern int simple_write_begin(struct file *file, struct address_space *mapping,
3399 			loff_t pos, unsigned len,
3400 			struct folio **foliop, void **fsdata);
3401 extern const struct address_space_operations ram_aops;
3402 extern int always_delete_dentry(const struct dentry *);
3403 extern struct inode *alloc_anon_inode(struct super_block *);
3404 extern int simple_nosetlease(struct file *, int, struct file_lease **, void **);
3405 extern const struct dentry_operations simple_dentry_operations;
3406 
3407 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3408 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3409 extern const struct file_operations simple_dir_operations;
3410 extern const struct inode_operations simple_dir_inode_operations;
3411 extern void make_empty_dir_inode(struct inode *inode);
3412 extern bool is_empty_dir_inode(struct inode *inode);
3413 struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3414 struct dentry *d_alloc_name(struct dentry *, const char *);
3415 extern int simple_fill_super(struct super_block *, unsigned long,
3416 			     const struct tree_descr *);
3417 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3418 extern void simple_release_fs(struct vfsmount **mount, int *count);
3419 
3420 extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3421 			loff_t *ppos, const void *from, size_t available);
3422 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3423 		const void __user *from, size_t count);
3424 
3425 struct offset_ctx {
3426 	struct maple_tree	mt;
3427 	unsigned long		next_offset;
3428 };
3429 
3430 void simple_offset_init(struct offset_ctx *octx);
3431 int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry);
3432 void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry);
3433 int simple_offset_empty(struct dentry *dentry);
3434 int simple_offset_rename(struct inode *old_dir, struct dentry *old_dentry,
3435 			 struct inode *new_dir, struct dentry *new_dentry);
3436 int simple_offset_rename_exchange(struct inode *old_dir,
3437 				  struct dentry *old_dentry,
3438 				  struct inode *new_dir,
3439 				  struct dentry *new_dentry);
3440 void simple_offset_destroy(struct offset_ctx *octx);
3441 
3442 extern const struct file_operations simple_offset_dir_operations;
3443 
3444 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3445 extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3446 
3447 extern int generic_check_addressable(unsigned, u64);
3448 
3449 extern void generic_set_sb_d_ops(struct super_block *sb);
3450 extern int generic_ci_match(const struct inode *parent,
3451 			    const struct qstr *name,
3452 			    const struct qstr *folded_name,
3453 			    const u8 *de_name, u32 de_name_len);
3454 
sb_has_encoding(const struct super_block * sb)3455 static inline bool sb_has_encoding(const struct super_block *sb)
3456 {
3457 #if IS_ENABLED(CONFIG_UNICODE)
3458 	return !!sb->s_encoding;
3459 #else
3460 	return false;
3461 #endif
3462 }
3463 
3464 int may_setattr(struct mnt_idmap *idmap, struct inode *inode,
3465 		unsigned int ia_valid);
3466 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *);
3467 extern int inode_newsize_ok(const struct inode *, loff_t offset);
3468 void setattr_copy(struct mnt_idmap *, struct inode *inode,
3469 		  const struct iattr *attr);
3470 
3471 extern int file_update_time(struct file *file);
3472 
vma_is_dax(const struct vm_area_struct * vma)3473 static inline bool vma_is_dax(const struct vm_area_struct *vma)
3474 {
3475 	return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3476 }
3477 
vma_is_fsdax(struct vm_area_struct * vma)3478 static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3479 {
3480 	struct inode *inode;
3481 
3482 	if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
3483 		return false;
3484 	if (!vma_is_dax(vma))
3485 		return false;
3486 	inode = file_inode(vma->vm_file);
3487 	if (S_ISCHR(inode->i_mode))
3488 		return false; /* device-dax */
3489 	return true;
3490 }
3491 
iocb_flags(struct file * file)3492 static inline int iocb_flags(struct file *file)
3493 {
3494 	int res = 0;
3495 	if (file->f_flags & O_APPEND)
3496 		res |= IOCB_APPEND;
3497 	if (file->f_flags & O_DIRECT)
3498 		res |= IOCB_DIRECT;
3499 	if (file->f_flags & O_DSYNC)
3500 		res |= IOCB_DSYNC;
3501 	if (file->f_flags & __O_SYNC)
3502 		res |= IOCB_SYNC;
3503 	return res;
3504 }
3505 
kiocb_set_rw_flags(struct kiocb * ki,rwf_t flags,int rw_type)3506 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags,
3507 				     int rw_type)
3508 {
3509 	int kiocb_flags = 0;
3510 
3511 	/* make sure there's no overlap between RWF and private IOCB flags */
3512 	BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3513 
3514 	if (!flags)
3515 		return 0;
3516 	if (unlikely(flags & ~RWF_SUPPORTED))
3517 		return -EOPNOTSUPP;
3518 	if (unlikely((flags & RWF_APPEND) && (flags & RWF_NOAPPEND)))
3519 		return -EINVAL;
3520 
3521 	if (flags & RWF_NOWAIT) {
3522 		if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3523 			return -EOPNOTSUPP;
3524 	}
3525 	if (flags & RWF_ATOMIC) {
3526 		if (rw_type != WRITE)
3527 			return -EOPNOTSUPP;
3528 		if (!(ki->ki_filp->f_mode & FMODE_CAN_ATOMIC_WRITE))
3529 			return -EOPNOTSUPP;
3530 	}
3531 	kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3532 	if (flags & RWF_SYNC)
3533 		kiocb_flags |= IOCB_DSYNC;
3534 
3535 	if ((flags & RWF_NOAPPEND) && (ki->ki_flags & IOCB_APPEND)) {
3536 		if (IS_APPEND(file_inode(ki->ki_filp)))
3537 			return -EPERM;
3538 		ki->ki_flags &= ~IOCB_APPEND;
3539 	}
3540 
3541 	ki->ki_flags |= kiocb_flags;
3542 	return 0;
3543 }
3544 
3545 /* Transaction based IO helpers */
3546 
3547 /*
3548  * An argresp is stored in an allocated page and holds the
3549  * size of the argument or response, along with its content
3550  */
3551 struct simple_transaction_argresp {
3552 	ssize_t size;
3553 	char data[];
3554 };
3555 
3556 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3557 
3558 char *simple_transaction_get(struct file *file, const char __user *buf,
3559 				size_t size);
3560 ssize_t simple_transaction_read(struct file *file, char __user *buf,
3561 				size_t size, loff_t *pos);
3562 int simple_transaction_release(struct inode *inode, struct file *file);
3563 
3564 void simple_transaction_set(struct file *file, size_t n);
3565 
3566 /*
3567  * simple attribute files
3568  *
3569  * These attributes behave similar to those in sysfs:
3570  *
3571  * Writing to an attribute immediately sets a value, an open file can be
3572  * written to multiple times.
3573  *
3574  * Reading from an attribute creates a buffer from the value that might get
3575  * read with multiple read calls. When the attribute has been read
3576  * completely, no further read calls are possible until the file is opened
3577  * again.
3578  *
3579  * All attributes contain a text representation of a numeric value
3580  * that are accessed with the get() and set() functions.
3581  */
3582 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed)	\
3583 static int __fops ## _open(struct inode *inode, struct file *file)	\
3584 {									\
3585 	__simple_attr_check_format(__fmt, 0ull);			\
3586 	return simple_attr_open(inode, file, __get, __set, __fmt);	\
3587 }									\
3588 static const struct file_operations __fops = {				\
3589 	.owner	 = THIS_MODULE,						\
3590 	.open	 = __fops ## _open,					\
3591 	.release = simple_attr_release,					\
3592 	.read	 = simple_attr_read,					\
3593 	.write	 = (__is_signed) ? simple_attr_write_signed : simple_attr_write,	\
3594 	.llseek	 = generic_file_llseek,					\
3595 }
3596 
3597 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt)		\
3598 	DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false)
3599 
3600 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt)	\
3601 	DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true)
3602 
3603 static inline __printf(1, 2)
__simple_attr_check_format(const char * fmt,...)3604 void __simple_attr_check_format(const char *fmt, ...)
3605 {
3606 	/* don't do anything, just let the compiler check the arguments; */
3607 }
3608 
3609 int simple_attr_open(struct inode *inode, struct file *file,
3610 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
3611 		     const char *fmt);
3612 int simple_attr_release(struct inode *inode, struct file *file);
3613 ssize_t simple_attr_read(struct file *file, char __user *buf,
3614 			 size_t len, loff_t *ppos);
3615 ssize_t simple_attr_write(struct file *file, const char __user *buf,
3616 			  size_t len, loff_t *ppos);
3617 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
3618 				 size_t len, loff_t *ppos);
3619 
3620 struct ctl_table;
3621 int __init list_bdev_fs_names(char *buf, size_t size);
3622 
3623 #define __FMODE_EXEC		((__force int) FMODE_EXEC)
3624 #define __FMODE_NONOTIFY	((__force int) FMODE_NONOTIFY)
3625 
3626 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3627 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3628 					    (flag & __FMODE_NONOTIFY)))
3629 
is_sxid(umode_t mode)3630 static inline bool is_sxid(umode_t mode)
3631 {
3632 	return mode & (S_ISUID | S_ISGID);
3633 }
3634 
check_sticky(struct mnt_idmap * idmap,struct inode * dir,struct inode * inode)3635 static inline int check_sticky(struct mnt_idmap *idmap,
3636 			       struct inode *dir, struct inode *inode)
3637 {
3638 	if (!(dir->i_mode & S_ISVTX))
3639 		return 0;
3640 
3641 	return __check_sticky(idmap, dir, inode);
3642 }
3643 
inode_has_no_xattr(struct inode * inode)3644 static inline void inode_has_no_xattr(struct inode *inode)
3645 {
3646 	if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3647 		inode->i_flags |= S_NOSEC;
3648 }
3649 
is_root_inode(struct inode * inode)3650 static inline bool is_root_inode(struct inode *inode)
3651 {
3652 	return inode == inode->i_sb->s_root->d_inode;
3653 }
3654 
dir_emit(struct dir_context * ctx,const char * name,int namelen,u64 ino,unsigned type)3655 static inline bool dir_emit(struct dir_context *ctx,
3656 			    const char *name, int namelen,
3657 			    u64 ino, unsigned type)
3658 {
3659 	return ctx->actor(ctx, name, namelen, ctx->pos, ino, type);
3660 }
dir_emit_dot(struct file * file,struct dir_context * ctx)3661 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3662 {
3663 	return ctx->actor(ctx, ".", 1, ctx->pos,
3664 			  file->f_path.dentry->d_inode->i_ino, DT_DIR);
3665 }
dir_emit_dotdot(struct file * file,struct dir_context * ctx)3666 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3667 {
3668 	return ctx->actor(ctx, "..", 2, ctx->pos,
3669 			  d_parent_ino(file->f_path.dentry), DT_DIR);
3670 }
dir_emit_dots(struct file * file,struct dir_context * ctx)3671 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3672 {
3673 	if (ctx->pos == 0) {
3674 		if (!dir_emit_dot(file, ctx))
3675 			return false;
3676 		ctx->pos = 1;
3677 	}
3678 	if (ctx->pos == 1) {
3679 		if (!dir_emit_dotdot(file, ctx))
3680 			return false;
3681 		ctx->pos = 2;
3682 	}
3683 	return true;
3684 }
dir_relax(struct inode * inode)3685 static inline bool dir_relax(struct inode *inode)
3686 {
3687 	inode_unlock(inode);
3688 	inode_lock(inode);
3689 	return !IS_DEADDIR(inode);
3690 }
3691 
dir_relax_shared(struct inode * inode)3692 static inline bool dir_relax_shared(struct inode *inode)
3693 {
3694 	inode_unlock_shared(inode);
3695 	inode_lock_shared(inode);
3696 	return !IS_DEADDIR(inode);
3697 }
3698 
3699 extern bool path_noexec(const struct path *path);
3700 extern void inode_nohighmem(struct inode *inode);
3701 
3702 /* mm/fadvise.c */
3703 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3704 		       int advice);
3705 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3706 			   int advice);
3707 
vfs_empty_path(int dfd,const char __user * path)3708 static inline bool vfs_empty_path(int dfd, const char __user *path)
3709 {
3710 	char c;
3711 
3712 	if (dfd < 0)
3713 		return false;
3714 
3715 	/* We now allow NULL to be used for empty path. */
3716 	if (!path)
3717 		return true;
3718 
3719 	if (unlikely(get_user(c, path)))
3720 		return false;
3721 
3722 	return !c;
3723 }
3724 
3725 bool generic_atomic_write_valid(struct iov_iter *iter, loff_t pos);
3726 
3727 #endif /* _LINUX_FS_H */
3728