xref: /openbmc/linux/include/linux/fs.h (revision ef836204)
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 
47 #include <asm/byteorder.h>
48 #include <uapi/linux/fs.h>
49 
50 struct backing_dev_info;
51 struct bdi_writeback;
52 struct bio;
53 struct io_comp_batch;
54 struct export_operations;
55 struct fiemap_extent_info;
56 struct hd_geometry;
57 struct iovec;
58 struct kiocb;
59 struct kobject;
60 struct pipe_inode_info;
61 struct poll_table_struct;
62 struct kstatfs;
63 struct vm_area_struct;
64 struct vfsmount;
65 struct cred;
66 struct swap_info_struct;
67 struct seq_file;
68 struct workqueue_struct;
69 struct iov_iter;
70 struct fscrypt_info;
71 struct fscrypt_operations;
72 struct fsverity_info;
73 struct fsverity_operations;
74 struct fs_context;
75 struct fs_parameter_spec;
76 struct fileattr;
77 struct iomap_ops;
78 
79 extern void __init inode_init(void);
80 extern void __init inode_init_early(void);
81 extern void __init files_init(void);
82 extern void __init files_maxfiles_init(void);
83 
84 extern unsigned long get_max_files(void);
85 extern unsigned int sysctl_nr_open;
86 
87 typedef __kernel_rwf_t rwf_t;
88 
89 struct buffer_head;
90 typedef int (get_block_t)(struct inode *inode, sector_t iblock,
91 			struct buffer_head *bh_result, int create);
92 typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
93 			ssize_t bytes, void *private);
94 
95 #define MAY_EXEC		0x00000001
96 #define MAY_WRITE		0x00000002
97 #define MAY_READ		0x00000004
98 #define MAY_APPEND		0x00000008
99 #define MAY_ACCESS		0x00000010
100 #define MAY_OPEN		0x00000020
101 #define MAY_CHDIR		0x00000040
102 /* called from RCU mode, don't block */
103 #define MAY_NOT_BLOCK		0x00000080
104 
105 /*
106  * flags in file.f_mode.  Note that FMODE_READ and FMODE_WRITE must correspond
107  * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
108  */
109 
110 /* file is open for reading */
111 #define FMODE_READ		((__force fmode_t)0x1)
112 /* file is open for writing */
113 #define FMODE_WRITE		((__force fmode_t)0x2)
114 /* file is seekable */
115 #define FMODE_LSEEK		((__force fmode_t)0x4)
116 /* file can be accessed using pread */
117 #define FMODE_PREAD		((__force fmode_t)0x8)
118 /* file can be accessed using pwrite */
119 #define FMODE_PWRITE		((__force fmode_t)0x10)
120 /* File is opened for execution with sys_execve / sys_uselib */
121 #define FMODE_EXEC		((__force fmode_t)0x20)
122 /* 32bit hashes as llseek() offset (for directories) */
123 #define FMODE_32BITHASH         ((__force fmode_t)0x200)
124 /* 64bit hashes as llseek() offset (for directories) */
125 #define FMODE_64BITHASH         ((__force fmode_t)0x400)
126 
127 /*
128  * Don't update ctime and mtime.
129  *
130  * Currently a special hack for the XFS open_by_handle ioctl, but we'll
131  * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
132  */
133 #define FMODE_NOCMTIME		((__force fmode_t)0x800)
134 
135 /* Expect random access pattern */
136 #define FMODE_RANDOM		((__force fmode_t)0x1000)
137 
138 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */
139 #define FMODE_UNSIGNED_OFFSET	((__force fmode_t)0x2000)
140 
141 /* File is opened with O_PATH; almost nothing can be done with it */
142 #define FMODE_PATH		((__force fmode_t)0x4000)
143 
144 /* File needs atomic accesses to f_pos */
145 #define FMODE_ATOMIC_POS	((__force fmode_t)0x8000)
146 /* Write access to underlying fs */
147 #define FMODE_WRITER		((__force fmode_t)0x10000)
148 /* Has read method(s) */
149 #define FMODE_CAN_READ          ((__force fmode_t)0x20000)
150 /* Has write method(s) */
151 #define FMODE_CAN_WRITE         ((__force fmode_t)0x40000)
152 
153 #define FMODE_OPENED		((__force fmode_t)0x80000)
154 #define FMODE_CREATED		((__force fmode_t)0x100000)
155 
156 /* File is stream-like */
157 #define FMODE_STREAM		((__force fmode_t)0x200000)
158 
159 /* File supports DIRECT IO */
160 #define	FMODE_CAN_ODIRECT	((__force fmode_t)0x400000)
161 
162 #define	FMODE_NOREUSE		((__force fmode_t)0x800000)
163 
164 /* File supports non-exclusive O_DIRECT writes from multiple threads */
165 #define FMODE_DIO_PARALLEL_WRITE	((__force fmode_t)0x1000000)
166 
167 /* File is embedded in backing_file object */
168 #define FMODE_BACKING		((__force fmode_t)0x2000000)
169 
170 /* File was opened by fanotify and shouldn't generate fanotify events */
171 #define FMODE_NONOTIFY		((__force fmode_t)0x4000000)
172 
173 /* File is capable of returning -EAGAIN if I/O will block */
174 #define FMODE_NOWAIT		((__force fmode_t)0x8000000)
175 
176 /* File represents mount that needs unmounting */
177 #define FMODE_NEED_UNMOUNT	((__force fmode_t)0x10000000)
178 
179 /* File does not contribute to nr_files count */
180 #define FMODE_NOACCOUNT		((__force fmode_t)0x20000000)
181 
182 /* File supports async buffered reads */
183 #define FMODE_BUF_RASYNC	((__force fmode_t)0x40000000)
184 
185 /* File supports async nowait buffered writes */
186 #define FMODE_BUF_WASYNC	((__force fmode_t)0x80000000)
187 
188 /*
189  * Attribute flags.  These should be or-ed together to figure out what
190  * has been changed!
191  */
192 #define ATTR_MODE	(1 << 0)
193 #define ATTR_UID	(1 << 1)
194 #define ATTR_GID	(1 << 2)
195 #define ATTR_SIZE	(1 << 3)
196 #define ATTR_ATIME	(1 << 4)
197 #define ATTR_MTIME	(1 << 5)
198 #define ATTR_CTIME	(1 << 6)
199 #define ATTR_ATIME_SET	(1 << 7)
200 #define ATTR_MTIME_SET	(1 << 8)
201 #define ATTR_FORCE	(1 << 9) /* Not a change, but a change it */
202 #define ATTR_KILL_SUID	(1 << 11)
203 #define ATTR_KILL_SGID	(1 << 12)
204 #define ATTR_FILE	(1 << 13)
205 #define ATTR_KILL_PRIV	(1 << 14)
206 #define ATTR_OPEN	(1 << 15) /* Truncating from open(O_TRUNC) */
207 #define ATTR_TIMES_SET	(1 << 16)
208 #define ATTR_TOUCH	(1 << 17)
209 
210 /*
211  * Whiteout is represented by a char device.  The following constants define the
212  * mode and device number to use.
213  */
214 #define WHITEOUT_MODE 0
215 #define WHITEOUT_DEV 0
216 
217 /*
218  * This is the Inode Attributes structure, used for notify_change().  It
219  * uses the above definitions as flags, to know which values have changed.
220  * Also, in this manner, a Filesystem can look at only the values it cares
221  * about.  Basically, these are the attributes that the VFS layer can
222  * request to change from the FS layer.
223  *
224  * Derek Atkins <warlord@MIT.EDU> 94-10-20
225  */
226 struct iattr {
227 	unsigned int	ia_valid;
228 	umode_t		ia_mode;
229 	/*
230 	 * The two anonymous unions wrap structures with the same member.
231 	 *
232 	 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which
233 	 * are a dedicated type requiring the filesystem to use the dedicated
234 	 * helpers. Other filesystem can continue to use ia_{g,u}id until they
235 	 * have been ported.
236 	 *
237 	 * They always contain the same value. In other words FS_ALLOW_IDMAP
238 	 * pass down the same value on idmapped mounts as they would on regular
239 	 * mounts.
240 	 */
241 	union {
242 		kuid_t		ia_uid;
243 		vfsuid_t	ia_vfsuid;
244 	};
245 	union {
246 		kgid_t		ia_gid;
247 		vfsgid_t	ia_vfsgid;
248 	};
249 	loff_t		ia_size;
250 	struct timespec64 ia_atime;
251 	struct timespec64 ia_mtime;
252 	struct timespec64 ia_ctime;
253 
254 	/*
255 	 * Not an attribute, but an auxiliary info for filesystems wanting to
256 	 * implement an ftruncate() like method.  NOTE: filesystem should
257 	 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
258 	 */
259 	struct file	*ia_file;
260 };
261 
262 /*
263  * Includes for diskquotas.
264  */
265 #include <linux/quota.h>
266 
267 /*
268  * Maximum number of layers of fs stack.  Needs to be limited to
269  * prevent kernel stack overflow
270  */
271 #define FILESYSTEM_MAX_STACK_DEPTH 2
272 
273 /**
274  * enum positive_aop_returns - aop return codes with specific semantics
275  *
276  * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
277  * 			    completed, that the page is still locked, and
278  * 			    should be considered active.  The VM uses this hint
279  * 			    to return the page to the active list -- it won't
280  * 			    be a candidate for writeback again in the near
281  * 			    future.  Other callers must be careful to unlock
282  * 			    the page if they get this return.  Returned by
283  * 			    writepage();
284  *
285  * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
286  *  			unlocked it and the page might have been truncated.
287  *  			The caller should back up to acquiring a new page and
288  *  			trying again.  The aop will be taking reasonable
289  *  			precautions not to livelock.  If the caller held a page
290  *  			reference, it should drop it before retrying.  Returned
291  *  			by read_folio().
292  *
293  * address_space_operation functions return these large constants to indicate
294  * special semantics to the caller.  These are much larger than the bytes in a
295  * page to allow for functions that return the number of bytes operated on in a
296  * given page.
297  */
298 
299 enum positive_aop_returns {
300 	AOP_WRITEPAGE_ACTIVATE	= 0x80000,
301 	AOP_TRUNCATED_PAGE	= 0x80001,
302 };
303 
304 /*
305  * oh the beauties of C type declarations.
306  */
307 struct page;
308 struct address_space;
309 struct writeback_control;
310 struct readahead_control;
311 
312 /*
313  * Write life time hint values.
314  * Stored in struct inode as u8.
315  */
316 enum rw_hint {
317 	WRITE_LIFE_NOT_SET	= 0,
318 	WRITE_LIFE_NONE		= RWH_WRITE_LIFE_NONE,
319 	WRITE_LIFE_SHORT	= RWH_WRITE_LIFE_SHORT,
320 	WRITE_LIFE_MEDIUM	= RWH_WRITE_LIFE_MEDIUM,
321 	WRITE_LIFE_LONG		= RWH_WRITE_LIFE_LONG,
322 	WRITE_LIFE_EXTREME	= RWH_WRITE_LIFE_EXTREME,
323 };
324 
325 /* Match RWF_* bits to IOCB bits */
326 #define IOCB_HIPRI		(__force int) RWF_HIPRI
327 #define IOCB_DSYNC		(__force int) RWF_DSYNC
328 #define IOCB_SYNC		(__force int) RWF_SYNC
329 #define IOCB_NOWAIT		(__force int) RWF_NOWAIT
330 #define IOCB_APPEND		(__force int) RWF_APPEND
331 
332 /* non-RWF related bits - start at 16 */
333 #define IOCB_EVENTFD		(1 << 16)
334 #define IOCB_DIRECT		(1 << 17)
335 #define IOCB_WRITE		(1 << 18)
336 /* iocb->ki_waitq is valid */
337 #define IOCB_WAITQ		(1 << 19)
338 #define IOCB_NOIO		(1 << 20)
339 /* can use bio alloc cache */
340 #define IOCB_ALLOC_CACHE	(1 << 21)
341 /*
342  * IOCB_DIO_CALLER_COMP can be set by the iocb owner, to indicate that the
343  * iocb completion can be passed back to the owner for execution from a safe
344  * context rather than needing to be punted through a workqueue. If this
345  * flag is set, the bio completion handling may set iocb->dio_complete to a
346  * handler function and iocb->private to context information for that handler.
347  * The issuer should call the handler with that context information from task
348  * context to complete the processing of the iocb. Note that while this
349  * provides a task context for the dio_complete() callback, it should only be
350  * used on the completion side for non-IO generating completions. It's fine to
351  * call blocking functions from this callback, but they should not wait for
352  * unrelated IO (like cache flushing, new IO generation, etc).
353  */
354 #define IOCB_DIO_CALLER_COMP	(1 << 22)
355 /* kiocb is a read or write operation submitted by fs/aio.c. */
356 #define IOCB_AIO_RW		(1 << 23)
357 
358 /* for use in trace events */
359 #define TRACE_IOCB_STRINGS \
360 	{ IOCB_HIPRI,		"HIPRI" }, \
361 	{ IOCB_DSYNC,		"DSYNC" }, \
362 	{ IOCB_SYNC,		"SYNC" }, \
363 	{ IOCB_NOWAIT,		"NOWAIT" }, \
364 	{ IOCB_APPEND,		"APPEND" }, \
365 	{ IOCB_EVENTFD,		"EVENTFD"}, \
366 	{ IOCB_DIRECT,		"DIRECT" }, \
367 	{ IOCB_WRITE,		"WRITE" }, \
368 	{ IOCB_WAITQ,		"WAITQ" }, \
369 	{ IOCB_NOIO,		"NOIO" }, \
370 	{ IOCB_ALLOC_CACHE,	"ALLOC_CACHE" }, \
371 	{ IOCB_DIO_CALLER_COMP,	"CALLER_COMP" }
372 
373 struct kiocb {
374 	struct file		*ki_filp;
375 	loff_t			ki_pos;
376 	void (*ki_complete)(struct kiocb *iocb, long ret);
377 	void			*private;
378 	int			ki_flags;
379 	u16			ki_ioprio; /* See linux/ioprio.h */
380 	union {
381 		/*
382 		 * Only used for async buffered reads, where it denotes the
383 		 * page waitqueue associated with completing the read. Valid
384 		 * IFF IOCB_WAITQ is set.
385 		 */
386 		struct wait_page_queue	*ki_waitq;
387 		/*
388 		 * Can be used for O_DIRECT IO, where the completion handling
389 		 * is punted back to the issuer of the IO. May only be set
390 		 * if IOCB_DIO_CALLER_COMP is set by the issuer, and the issuer
391 		 * must then check for presence of this handler when ki_complete
392 		 * is invoked. The data passed in to this handler must be
393 		 * assigned to ->private when dio_complete is assigned.
394 		 */
395 		ssize_t (*dio_complete)(void *data);
396 	};
397 };
398 
is_sync_kiocb(struct kiocb * kiocb)399 static inline bool is_sync_kiocb(struct kiocb *kiocb)
400 {
401 	return kiocb->ki_complete == NULL;
402 }
403 
404 struct address_space_operations {
405 	int (*writepage)(struct page *page, struct writeback_control *wbc);
406 	int (*read_folio)(struct file *, struct folio *);
407 
408 	/* Write back some dirty pages from this mapping. */
409 	int (*writepages)(struct address_space *, struct writeback_control *);
410 
411 	/* Mark a folio dirty.  Return true if this dirtied it */
412 	bool (*dirty_folio)(struct address_space *, struct folio *);
413 
414 	void (*readahead)(struct readahead_control *);
415 
416 	int (*write_begin)(struct file *, struct address_space *mapping,
417 				loff_t pos, unsigned len,
418 				struct page **pagep, void **fsdata);
419 	int (*write_end)(struct file *, struct address_space *mapping,
420 				loff_t pos, unsigned len, unsigned copied,
421 				struct page *page, void *fsdata);
422 
423 	/* Unfortunately this kludge is needed for FIBMAP. Don't use it */
424 	sector_t (*bmap)(struct address_space *, sector_t);
425 	void (*invalidate_folio) (struct folio *, size_t offset, size_t len);
426 	bool (*release_folio)(struct folio *, gfp_t);
427 	void (*free_folio)(struct folio *folio);
428 	ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
429 	/*
430 	 * migrate the contents of a folio to the specified target. If
431 	 * migrate_mode is MIGRATE_ASYNC, it must not block.
432 	 */
433 	int (*migrate_folio)(struct address_space *, struct folio *dst,
434 			struct folio *src, enum migrate_mode);
435 	int (*launder_folio)(struct folio *);
436 	bool (*is_partially_uptodate) (struct folio *, size_t from,
437 			size_t count);
438 	void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb);
439 	int (*error_remove_page)(struct address_space *, struct page *);
440 
441 	/* swapfile support */
442 	int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
443 				sector_t *span);
444 	void (*swap_deactivate)(struct file *file);
445 	int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
446 };
447 
448 extern const struct address_space_operations empty_aops;
449 
450 /**
451  * struct address_space - Contents of a cacheable, mappable object.
452  * @host: Owner, either the inode or the block_device.
453  * @i_pages: Cached pages.
454  * @invalidate_lock: Guards coherency between page cache contents and
455  *   file offset->disk block mappings in the filesystem during invalidates.
456  *   It is also used to block modification of page cache contents through
457  *   memory mappings.
458  * @gfp_mask: Memory allocation flags to use for allocating pages.
459  * @i_mmap_writable: Number of VM_SHARED mappings.
460  * @nr_thps: Number of THPs in the pagecache (non-shmem only).
461  * @i_mmap: Tree of private and shared mappings.
462  * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
463  * @nrpages: Number of page entries, protected by the i_pages lock.
464  * @writeback_index: Writeback starts here.
465  * @a_ops: Methods.
466  * @flags: Error bits and flags (AS_*).
467  * @wb_err: The most recent error which has occurred.
468  * @private_lock: For use by the owner of the address_space.
469  * @private_list: For use by the owner of the address_space.
470  * @private_data: For use by the owner of the address_space.
471  */
472 struct address_space {
473 	struct inode		*host;
474 	struct xarray		i_pages;
475 	struct rw_semaphore	invalidate_lock;
476 	gfp_t			gfp_mask;
477 	atomic_t		i_mmap_writable;
478 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
479 	/* number of thp, only for non-shmem files */
480 	atomic_t		nr_thps;
481 #endif
482 	struct rb_root_cached	i_mmap;
483 	unsigned long		nrpages;
484 	pgoff_t			writeback_index;
485 	const struct address_space_operations *a_ops;
486 	unsigned long		flags;
487 	struct rw_semaphore	i_mmap_rwsem;
488 	errseq_t		wb_err;
489 	spinlock_t		private_lock;
490 	struct list_head	private_list;
491 	void			*private_data;
492 } __attribute__((aligned(sizeof(long)))) __randomize_layout;
493 	/*
494 	 * On most architectures that alignment is already the case; but
495 	 * must be enforced here for CRIS, to let the least significant bit
496 	 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
497 	 */
498 
499 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */
500 #define PAGECACHE_TAG_DIRTY	XA_MARK_0
501 #define PAGECACHE_TAG_WRITEBACK	XA_MARK_1
502 #define PAGECACHE_TAG_TOWRITE	XA_MARK_2
503 
504 /*
505  * Returns true if any of the pages in the mapping are marked with the tag.
506  */
mapping_tagged(struct address_space * mapping,xa_mark_t tag)507 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
508 {
509 	return xa_marked(&mapping->i_pages, tag);
510 }
511 
i_mmap_lock_write(struct address_space * mapping)512 static inline void i_mmap_lock_write(struct address_space *mapping)
513 {
514 	down_write(&mapping->i_mmap_rwsem);
515 }
516 
i_mmap_trylock_write(struct address_space * mapping)517 static inline int i_mmap_trylock_write(struct address_space *mapping)
518 {
519 	return down_write_trylock(&mapping->i_mmap_rwsem);
520 }
521 
i_mmap_unlock_write(struct address_space * mapping)522 static inline void i_mmap_unlock_write(struct address_space *mapping)
523 {
524 	up_write(&mapping->i_mmap_rwsem);
525 }
526 
i_mmap_trylock_read(struct address_space * mapping)527 static inline int i_mmap_trylock_read(struct address_space *mapping)
528 {
529 	return down_read_trylock(&mapping->i_mmap_rwsem);
530 }
531 
i_mmap_lock_read(struct address_space * mapping)532 static inline void i_mmap_lock_read(struct address_space *mapping)
533 {
534 	down_read(&mapping->i_mmap_rwsem);
535 }
536 
i_mmap_unlock_read(struct address_space * mapping)537 static inline void i_mmap_unlock_read(struct address_space *mapping)
538 {
539 	up_read(&mapping->i_mmap_rwsem);
540 }
541 
i_mmap_assert_locked(struct address_space * mapping)542 static inline void i_mmap_assert_locked(struct address_space *mapping)
543 {
544 	lockdep_assert_held(&mapping->i_mmap_rwsem);
545 }
546 
i_mmap_assert_write_locked(struct address_space * mapping)547 static inline void i_mmap_assert_write_locked(struct address_space *mapping)
548 {
549 	lockdep_assert_held_write(&mapping->i_mmap_rwsem);
550 }
551 
552 /*
553  * Might pages of this file be mapped into userspace?
554  */
mapping_mapped(struct address_space * mapping)555 static inline int mapping_mapped(struct address_space *mapping)
556 {
557 	return	!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
558 }
559 
560 /*
561  * Might pages of this file have been modified in userspace?
562  * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap
563  * marks vma as VM_SHARED if it is shared, and the file was opened for
564  * writing i.e. vma may be mprotected writable even if now readonly.
565  *
566  * If i_mmap_writable is negative, no new writable mappings are allowed. You
567  * can only deny writable mappings, if none exists right now.
568  */
mapping_writably_mapped(struct address_space * mapping)569 static inline int mapping_writably_mapped(struct address_space *mapping)
570 {
571 	return atomic_read(&mapping->i_mmap_writable) > 0;
572 }
573 
mapping_map_writable(struct address_space * mapping)574 static inline int mapping_map_writable(struct address_space *mapping)
575 {
576 	return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
577 		0 : -EPERM;
578 }
579 
mapping_unmap_writable(struct address_space * mapping)580 static inline void mapping_unmap_writable(struct address_space *mapping)
581 {
582 	atomic_dec(&mapping->i_mmap_writable);
583 }
584 
mapping_deny_writable(struct address_space * mapping)585 static inline int mapping_deny_writable(struct address_space *mapping)
586 {
587 	return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
588 		0 : -EBUSY;
589 }
590 
mapping_allow_writable(struct address_space * mapping)591 static inline void mapping_allow_writable(struct address_space *mapping)
592 {
593 	atomic_inc(&mapping->i_mmap_writable);
594 }
595 
596 /*
597  * Use sequence counter to get consistent i_size on 32-bit processors.
598  */
599 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
600 #include <linux/seqlock.h>
601 #define __NEED_I_SIZE_ORDERED
602 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
603 #else
604 #define i_size_ordered_init(inode) do { } while (0)
605 #endif
606 
607 struct posix_acl;
608 #define ACL_NOT_CACHED ((void *)(-1))
609 /*
610  * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
611  * cache the ACL.  This also means that ->get_inode_acl() can be called in RCU
612  * mode with the LOOKUP_RCU flag.
613  */
614 #define ACL_DONT_CACHE ((void *)(-3))
615 
616 static inline struct posix_acl *
uncached_acl_sentinel(struct task_struct * task)617 uncached_acl_sentinel(struct task_struct *task)
618 {
619 	return (void *)task + 1;
620 }
621 
622 static inline bool
is_uncached_acl(struct posix_acl * acl)623 is_uncached_acl(struct posix_acl *acl)
624 {
625 	return (long)acl & 1;
626 }
627 
628 #define IOP_FASTPERM	0x0001
629 #define IOP_LOOKUP	0x0002
630 #define IOP_NOFOLLOW	0x0004
631 #define IOP_XATTR	0x0008
632 #define IOP_DEFAULT_READLINK	0x0010
633 
634 struct fsnotify_mark_connector;
635 
636 /*
637  * Keep mostly read-only and often accessed (especially for
638  * the RCU path lookup and 'stat' data) fields at the beginning
639  * of the 'struct inode'
640  */
641 struct inode {
642 	umode_t			i_mode;
643 	unsigned short		i_opflags;
644 	kuid_t			i_uid;
645 	struct list_head	i_lru;		/* inode LRU list */
646 	kgid_t			i_gid;
647 	unsigned int		i_flags;
648 
649 #ifdef CONFIG_FS_POSIX_ACL
650 	struct posix_acl	*i_acl;
651 	struct posix_acl	*i_default_acl;
652 #endif
653 
654 	const struct inode_operations	*i_op;
655 	struct super_block	*i_sb;
656 	struct address_space	*i_mapping;
657 
658 #ifdef CONFIG_SECURITY
659 	void			*i_security;
660 #endif
661 
662 	/* Stat data, not accessed from path walking */
663 	unsigned long		i_ino;
664 	/*
665 	 * Filesystems may only read i_nlink directly.  They shall use the
666 	 * following functions for modification:
667 	 *
668 	 *    (set|clear|inc|drop)_nlink
669 	 *    inode_(inc|dec)_link_count
670 	 */
671 	union {
672 		const unsigned int i_nlink;
673 		unsigned int __i_nlink;
674 	};
675 	dev_t			i_rdev;
676 	loff_t			i_size;
677 	struct timespec64	i_atime;
678 	struct timespec64	i_mtime;
679 	struct timespec64	__i_ctime; /* use inode_*_ctime accessors! */
680 	spinlock_t		i_lock;	/* i_blocks, i_bytes, maybe i_size */
681 	unsigned short          i_bytes;
682 	u8			i_blkbits;
683 	u8			i_write_hint;
684 	blkcnt_t		i_blocks;
685 
686 #ifdef __NEED_I_SIZE_ORDERED
687 	seqcount_t		i_size_seqcount;
688 #endif
689 
690 	/* Misc */
691 	unsigned long		i_state;
692 	struct rw_semaphore	i_rwsem;
693 
694 	unsigned long		dirtied_when;	/* jiffies of first dirtying */
695 	unsigned long		dirtied_time_when;
696 
697 	struct hlist_node	i_hash;
698 	struct list_head	i_io_list;	/* backing dev IO list */
699 #ifdef CONFIG_CGROUP_WRITEBACK
700 	struct bdi_writeback	*i_wb;		/* the associated cgroup wb */
701 
702 	/* foreign inode detection, see wbc_detach_inode() */
703 	int			i_wb_frn_winner;
704 	u16			i_wb_frn_avg_time;
705 	u16			i_wb_frn_history;
706 #endif
707 	struct list_head	i_sb_list;
708 	struct list_head	i_wb_list;	/* backing dev writeback list */
709 	union {
710 		struct hlist_head	i_dentry;
711 		struct rcu_head		i_rcu;
712 	};
713 	atomic64_t		i_version;
714 	atomic64_t		i_sequence; /* see futex */
715 	atomic_t		i_count;
716 	atomic_t		i_dio_count;
717 	atomic_t		i_writecount;
718 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
719 	atomic_t		i_readcount; /* struct files open RO */
720 #endif
721 	union {
722 		const struct file_operations	*i_fop;	/* former ->i_op->default_file_ops */
723 		void (*free_inode)(struct inode *);
724 	};
725 	struct file_lock_context	*i_flctx;
726 	struct address_space	i_data;
727 	struct list_head	i_devices;
728 	union {
729 		struct pipe_inode_info	*i_pipe;
730 		struct cdev		*i_cdev;
731 		char			*i_link;
732 		unsigned		i_dir_seq;
733 	};
734 
735 	__u32			i_generation;
736 
737 #ifdef CONFIG_FSNOTIFY
738 	__u32			i_fsnotify_mask; /* all events this inode cares about */
739 	struct fsnotify_mark_connector __rcu	*i_fsnotify_marks;
740 #endif
741 
742 #ifdef CONFIG_FS_ENCRYPTION
743 	struct fscrypt_info	*i_crypt_info;
744 #endif
745 
746 #ifdef CONFIG_FS_VERITY
747 	struct fsverity_info	*i_verity_info;
748 #endif
749 
750 	void			*i_private; /* fs or device private pointer */
751 } __randomize_layout;
752 
753 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
754 
i_blocksize(const struct inode * node)755 static inline unsigned int i_blocksize(const struct inode *node)
756 {
757 	return (1 << node->i_blkbits);
758 }
759 
inode_unhashed(struct inode * inode)760 static inline int inode_unhashed(struct inode *inode)
761 {
762 	return hlist_unhashed(&inode->i_hash);
763 }
764 
765 /*
766  * __mark_inode_dirty expects inodes to be hashed.  Since we don't
767  * want special inodes in the fileset inode space, we make them
768  * appear hashed, but do not put on any lists.  hlist_del()
769  * will work fine and require no locking.
770  */
inode_fake_hash(struct inode * inode)771 static inline void inode_fake_hash(struct inode *inode)
772 {
773 	hlist_add_fake(&inode->i_hash);
774 }
775 
776 /*
777  * inode->i_mutex nesting subclasses for the lock validator:
778  *
779  * 0: the object of the current VFS operation
780  * 1: parent
781  * 2: child/target
782  * 3: xattr
783  * 4: second non-directory
784  * 5: second parent (when locking independent directories in rename)
785  *
786  * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
787  * non-directories at once.
788  *
789  * The locking order between these classes is
790  * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
791  */
792 enum inode_i_mutex_lock_class
793 {
794 	I_MUTEX_NORMAL,
795 	I_MUTEX_PARENT,
796 	I_MUTEX_CHILD,
797 	I_MUTEX_XATTR,
798 	I_MUTEX_NONDIR2,
799 	I_MUTEX_PARENT2,
800 };
801 
inode_lock(struct inode * inode)802 static inline void inode_lock(struct inode *inode)
803 {
804 	down_write(&inode->i_rwsem);
805 }
806 
inode_unlock(struct inode * inode)807 static inline void inode_unlock(struct inode *inode)
808 {
809 	up_write(&inode->i_rwsem);
810 }
811 
inode_lock_shared(struct inode * inode)812 static inline void inode_lock_shared(struct inode *inode)
813 {
814 	down_read(&inode->i_rwsem);
815 }
816 
inode_unlock_shared(struct inode * inode)817 static inline void inode_unlock_shared(struct inode *inode)
818 {
819 	up_read(&inode->i_rwsem);
820 }
821 
inode_trylock(struct inode * inode)822 static inline int inode_trylock(struct inode *inode)
823 {
824 	return down_write_trylock(&inode->i_rwsem);
825 }
826 
inode_trylock_shared(struct inode * inode)827 static inline int inode_trylock_shared(struct inode *inode)
828 {
829 	return down_read_trylock(&inode->i_rwsem);
830 }
831 
inode_is_locked(struct inode * inode)832 static inline int inode_is_locked(struct inode *inode)
833 {
834 	return rwsem_is_locked(&inode->i_rwsem);
835 }
836 
inode_lock_nested(struct inode * inode,unsigned subclass)837 static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
838 {
839 	down_write_nested(&inode->i_rwsem, subclass);
840 }
841 
inode_lock_shared_nested(struct inode * inode,unsigned subclass)842 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
843 {
844 	down_read_nested(&inode->i_rwsem, subclass);
845 }
846 
filemap_invalidate_lock(struct address_space * mapping)847 static inline void filemap_invalidate_lock(struct address_space *mapping)
848 {
849 	down_write(&mapping->invalidate_lock);
850 }
851 
filemap_invalidate_unlock(struct address_space * mapping)852 static inline void filemap_invalidate_unlock(struct address_space *mapping)
853 {
854 	up_write(&mapping->invalidate_lock);
855 }
856 
filemap_invalidate_lock_shared(struct address_space * mapping)857 static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
858 {
859 	down_read(&mapping->invalidate_lock);
860 }
861 
filemap_invalidate_trylock_shared(struct address_space * mapping)862 static inline int filemap_invalidate_trylock_shared(
863 					struct address_space *mapping)
864 {
865 	return down_read_trylock(&mapping->invalidate_lock);
866 }
867 
filemap_invalidate_unlock_shared(struct address_space * mapping)868 static inline void filemap_invalidate_unlock_shared(
869 					struct address_space *mapping)
870 {
871 	up_read(&mapping->invalidate_lock);
872 }
873 
874 void lock_two_nondirectories(struct inode *, struct inode*);
875 void unlock_two_nondirectories(struct inode *, struct inode*);
876 
877 void filemap_invalidate_lock_two(struct address_space *mapping1,
878 				 struct address_space *mapping2);
879 void filemap_invalidate_unlock_two(struct address_space *mapping1,
880 				   struct address_space *mapping2);
881 
882 
883 /*
884  * NOTE: in a 32bit arch with a preemptable kernel and
885  * an UP compile the i_size_read/write must be atomic
886  * with respect to the local cpu (unlike with preempt disabled),
887  * but they don't need to be atomic with respect to other cpus like in
888  * true SMP (so they need either to either locally disable irq around
889  * the read or for example on x86 they can be still implemented as a
890  * cmpxchg8b without the need of the lock prefix). For SMP compiles
891  * and 64bit archs it makes no difference if preempt is enabled or not.
892  */
i_size_read(const struct inode * inode)893 static inline loff_t i_size_read(const struct inode *inode)
894 {
895 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
896 	loff_t i_size;
897 	unsigned int seq;
898 
899 	do {
900 		seq = read_seqcount_begin(&inode->i_size_seqcount);
901 		i_size = inode->i_size;
902 	} while (read_seqcount_retry(&inode->i_size_seqcount, seq));
903 	return i_size;
904 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
905 	loff_t i_size;
906 
907 	preempt_disable();
908 	i_size = inode->i_size;
909 	preempt_enable();
910 	return i_size;
911 #else
912 	return inode->i_size;
913 #endif
914 }
915 
916 /*
917  * NOTE: unlike i_size_read(), i_size_write() does need locking around it
918  * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
919  * can be lost, resulting in subsequent i_size_read() calls spinning forever.
920  */
i_size_write(struct inode * inode,loff_t i_size)921 static inline void i_size_write(struct inode *inode, loff_t i_size)
922 {
923 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
924 	preempt_disable();
925 	write_seqcount_begin(&inode->i_size_seqcount);
926 	inode->i_size = i_size;
927 	write_seqcount_end(&inode->i_size_seqcount);
928 	preempt_enable();
929 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
930 	preempt_disable();
931 	inode->i_size = i_size;
932 	preempt_enable();
933 #else
934 	inode->i_size = i_size;
935 #endif
936 }
937 
iminor(const struct inode * inode)938 static inline unsigned iminor(const struct inode *inode)
939 {
940 	return MINOR(inode->i_rdev);
941 }
942 
imajor(const struct inode * inode)943 static inline unsigned imajor(const struct inode *inode)
944 {
945 	return MAJOR(inode->i_rdev);
946 }
947 
948 struct fown_struct {
949 	rwlock_t lock;          /* protects pid, uid, euid fields */
950 	struct pid *pid;	/* pid or -pgrp where SIGIO should be sent */
951 	enum pid_type pid_type;	/* Kind of process group SIGIO should be sent to */
952 	kuid_t uid, euid;	/* uid/euid of process setting the owner */
953 	int signum;		/* posix.1b rt signal to be delivered on IO */
954 };
955 
956 /**
957  * struct file_ra_state - Track a file's readahead state.
958  * @start: Where the most recent readahead started.
959  * @size: Number of pages read in the most recent readahead.
960  * @async_size: Numer of pages that were/are not needed immediately
961  *      and so were/are genuinely "ahead".  Start next readahead when
962  *      the first of these pages is accessed.
963  * @ra_pages: Maximum size of a readahead request, copied from the bdi.
964  * @mmap_miss: How many mmap accesses missed in the page cache.
965  * @prev_pos: The last byte in the most recent read request.
966  *
967  * When this structure is passed to ->readahead(), the "most recent"
968  * readahead means the current readahead.
969  */
970 struct file_ra_state {
971 	pgoff_t start;
972 	unsigned int size;
973 	unsigned int async_size;
974 	unsigned int ra_pages;
975 	unsigned int mmap_miss;
976 	loff_t prev_pos;
977 };
978 
979 /*
980  * Check if @index falls in the readahead windows.
981  */
ra_has_index(struct file_ra_state * ra,pgoff_t index)982 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
983 {
984 	return (index >= ra->start &&
985 		index <  ra->start + ra->size);
986 }
987 
988 /*
989  * f_{lock,count,pos_lock} members can be highly contended and share
990  * the same cacheline. f_{lock,mode} are very frequently used together
991  * and so share the same cacheline as well. The read-mostly
992  * f_{path,inode,op} are kept on a separate cacheline.
993  */
994 struct file {
995 	union {
996 		struct llist_node	f_llist;
997 		struct rcu_head 	f_rcuhead;
998 		unsigned int 		f_iocb_flags;
999 	};
1000 
1001 	/*
1002 	 * Protects f_ep, f_flags.
1003 	 * Must not be taken from IRQ context.
1004 	 */
1005 	spinlock_t		f_lock;
1006 	fmode_t			f_mode;
1007 	atomic_long_t		f_count;
1008 	struct mutex		f_pos_lock;
1009 	loff_t			f_pos;
1010 	unsigned int		f_flags;
1011 	struct fown_struct	f_owner;
1012 	const struct cred	*f_cred;
1013 	struct file_ra_state	f_ra;
1014 	struct path		f_path;
1015 	struct inode		*f_inode;	/* cached value */
1016 	const struct file_operations	*f_op;
1017 
1018 	u64			f_version;
1019 #ifdef CONFIG_SECURITY
1020 	void			*f_security;
1021 #endif
1022 	/* needed for tty driver, and maybe others */
1023 	void			*private_data;
1024 
1025 #ifdef CONFIG_EPOLL
1026 	/* Used by fs/eventpoll.c to link all the hooks to this file */
1027 	struct hlist_head	*f_ep;
1028 #endif /* #ifdef CONFIG_EPOLL */
1029 	struct address_space	*f_mapping;
1030 	errseq_t		f_wb_err;
1031 	errseq_t		f_sb_err; /* for syncfs */
1032 } __randomize_layout
1033   __attribute__((aligned(4)));	/* lest something weird decides that 2 is OK */
1034 
1035 struct file_handle {
1036 	__u32 handle_bytes;
1037 	int handle_type;
1038 	/* file identifier */
1039 	unsigned char f_handle[] __counted_by(handle_bytes);
1040 };
1041 
get_file(struct file * f)1042 static inline struct file *get_file(struct file *f)
1043 {
1044 	atomic_long_inc(&f->f_count);
1045 	return f;
1046 }
1047 #define get_file_rcu(x) atomic_long_inc_not_zero(&(x)->f_count)
1048 #define file_count(x)	atomic_long_read(&(x)->f_count)
1049 
1050 #define	MAX_NON_LFS	((1UL<<31) - 1)
1051 
1052 /* Page cache limit. The filesystems should put that into their s_maxbytes
1053    limits, otherwise bad things can happen in VM. */
1054 #if BITS_PER_LONG==32
1055 #define MAX_LFS_FILESIZE	((loff_t)ULONG_MAX << PAGE_SHIFT)
1056 #elif BITS_PER_LONG==64
1057 #define MAX_LFS_FILESIZE 	((loff_t)LLONG_MAX)
1058 #endif
1059 
1060 /* legacy typedef, should eventually be removed */
1061 typedef void *fl_owner_t;
1062 
1063 struct file_lock;
1064 
1065 /* The following constant reflects the upper bound of the file/locking space */
1066 #ifndef OFFSET_MAX
1067 #define OFFSET_MAX	type_max(loff_t)
1068 #define OFFT_OFFSET_MAX	type_max(off_t)
1069 #endif
1070 
1071 extern void send_sigio(struct fown_struct *fown, int fd, int band);
1072 
file_inode(const struct file * f)1073 static inline struct inode *file_inode(const struct file *f)
1074 {
1075 	return f->f_inode;
1076 }
1077 
file_dentry(const struct file * file)1078 static inline struct dentry *file_dentry(const struct file *file)
1079 {
1080 	return d_real(file->f_path.dentry, file_inode(file));
1081 }
1082 
1083 struct fasync_struct {
1084 	rwlock_t		fa_lock;
1085 	int			magic;
1086 	int			fa_fd;
1087 	struct fasync_struct	*fa_next; /* singly linked list */
1088 	struct file		*fa_file;
1089 	struct rcu_head		fa_rcu;
1090 };
1091 
1092 #define FASYNC_MAGIC 0x4601
1093 
1094 /* SMP safe fasync helpers: */
1095 extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1096 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1097 extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1098 extern struct fasync_struct *fasync_alloc(void);
1099 extern void fasync_free(struct fasync_struct *);
1100 
1101 /* can be called from interrupts */
1102 extern void kill_fasync(struct fasync_struct **, int, int);
1103 
1104 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1105 extern int f_setown(struct file *filp, int who, int force);
1106 extern void f_delown(struct file *filp);
1107 extern pid_t f_getown(struct file *filp);
1108 extern int send_sigurg(struct fown_struct *fown);
1109 
1110 /*
1111  * sb->s_flags.  Note that these mirror the equivalent MS_* flags where
1112  * represented in both.
1113  */
1114 #define SB_RDONLY       BIT(0)	/* Mount read-only */
1115 #define SB_NOSUID       BIT(1)	/* Ignore suid and sgid bits */
1116 #define SB_NODEV        BIT(2)	/* Disallow access to device special files */
1117 #define SB_NOEXEC       BIT(3)	/* Disallow program execution */
1118 #define SB_SYNCHRONOUS  BIT(4)	/* Writes are synced at once */
1119 #define SB_MANDLOCK     BIT(6)	/* Allow mandatory locks on an FS */
1120 #define SB_DIRSYNC      BIT(7)	/* Directory modifications are synchronous */
1121 #define SB_NOATIME      BIT(10)	/* Do not update access times. */
1122 #define SB_NODIRATIME   BIT(11)	/* Do not update directory access times */
1123 #define SB_SILENT       BIT(15)
1124 #define SB_POSIXACL     BIT(16)	/* VFS does not apply the umask */
1125 #define SB_INLINECRYPT  BIT(17)	/* Use blk-crypto for encrypted files */
1126 #define SB_KERNMOUNT    BIT(22)	/* this is a kern_mount call */
1127 #define SB_I_VERSION    BIT(23)	/* Update inode I_version field */
1128 #define SB_LAZYTIME     BIT(25)	/* Update the on-disk [acm]times lazily */
1129 
1130 /* These sb flags are internal to the kernel */
1131 #define SB_DEAD         BIT(21)
1132 #define SB_DYING        BIT(24)
1133 #define SB_SUBMOUNT     BIT(26)
1134 #define SB_FORCE        BIT(27)
1135 #define SB_NOSEC        BIT(28)
1136 #define SB_BORN         BIT(29)
1137 #define SB_ACTIVE       BIT(30)
1138 #define SB_NOUSER       BIT(31)
1139 
1140 /* These flags relate to encoding and casefolding */
1141 #define SB_ENC_STRICT_MODE_FL	(1 << 0)
1142 
1143 #define sb_has_strict_encoding(sb) \
1144 	(sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1145 
1146 /*
1147  *	Umount options
1148  */
1149 
1150 #define MNT_FORCE	0x00000001	/* Attempt to forcibily umount */
1151 #define MNT_DETACH	0x00000002	/* Just detach from the tree */
1152 #define MNT_EXPIRE	0x00000004	/* Mark for expiry */
1153 #define UMOUNT_NOFOLLOW	0x00000008	/* Don't follow symlink on umount */
1154 #define UMOUNT_UNUSED	0x80000000	/* Flag guaranteed to be unused */
1155 
1156 /* sb->s_iflags */
1157 #define SB_I_CGROUPWB	0x00000001	/* cgroup-aware writeback enabled */
1158 #define SB_I_NOEXEC	0x00000002	/* Ignore executables on this fs */
1159 #define SB_I_NODEV	0x00000004	/* Ignore devices on this fs */
1160 #define SB_I_STABLE_WRITES 0x00000008	/* don't modify blks until WB is done */
1161 
1162 /* sb->s_iflags to limit user namespace mounts */
1163 #define SB_I_USERNS_VISIBLE		0x00000010 /* fstype already mounted */
1164 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE	0x00000020
1165 #define SB_I_UNTRUSTED_MOUNTER		0x00000040
1166 
1167 #define SB_I_SKIP_SYNC	0x00000100	/* Skip superblock at global sync */
1168 #define SB_I_PERSB_BDI	0x00000200	/* has a per-sb bdi */
1169 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1170 #define SB_I_RETIRED	0x00000800	/* superblock shouldn't be reused */
1171 
1172 /* Possible states of 'frozen' field */
1173 enum {
1174 	SB_UNFROZEN = 0,		/* FS is unfrozen */
1175 	SB_FREEZE_WRITE	= 1,		/* Writes, dir ops, ioctls frozen */
1176 	SB_FREEZE_PAGEFAULT = 2,	/* Page faults stopped as well */
1177 	SB_FREEZE_FS = 3,		/* For internal FS use (e.g. to stop
1178 					 * internal threads if needed) */
1179 	SB_FREEZE_COMPLETE = 4,		/* ->freeze_fs finished successfully */
1180 };
1181 
1182 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1183 
1184 struct sb_writers {
1185 	unsigned short			frozen;		/* Is sb frozen? */
1186 	unsigned short			freeze_holders;	/* Who froze fs? */
1187 	struct percpu_rw_semaphore	rw_sem[SB_FREEZE_LEVELS];
1188 };
1189 
1190 struct super_block {
1191 	struct list_head	s_list;		/* Keep this first */
1192 	dev_t			s_dev;		/* search index; _not_ kdev_t */
1193 	unsigned char		s_blocksize_bits;
1194 	unsigned long		s_blocksize;
1195 	loff_t			s_maxbytes;	/* Max file size */
1196 	struct file_system_type	*s_type;
1197 	const struct super_operations	*s_op;
1198 	const struct dquot_operations	*dq_op;
1199 	const struct quotactl_ops	*s_qcop;
1200 	const struct export_operations *s_export_op;
1201 	unsigned long		s_flags;
1202 	unsigned long		s_iflags;	/* internal SB_I_* flags */
1203 	unsigned long		s_magic;
1204 	struct dentry		*s_root;
1205 	struct rw_semaphore	s_umount;
1206 	int			s_count;
1207 	atomic_t		s_active;
1208 #ifdef CONFIG_SECURITY
1209 	void                    *s_security;
1210 #endif
1211 	const struct xattr_handler **s_xattr;
1212 #ifdef CONFIG_FS_ENCRYPTION
1213 	const struct fscrypt_operations	*s_cop;
1214 	struct fscrypt_keyring	*s_master_keys; /* master crypto keys in use */
1215 #endif
1216 #ifdef CONFIG_FS_VERITY
1217 	const struct fsverity_operations *s_vop;
1218 #endif
1219 #if IS_ENABLED(CONFIG_UNICODE)
1220 	struct unicode_map *s_encoding;
1221 	__u16 s_encoding_flags;
1222 #endif
1223 	struct hlist_bl_head	s_roots;	/* alternate root dentries for NFS */
1224 	struct list_head	s_mounts;	/* list of mounts; _not_ for fs use */
1225 	struct block_device	*s_bdev;
1226 	struct bdev_handle	*s_bdev_handle;
1227 	struct backing_dev_info *s_bdi;
1228 	struct mtd_info		*s_mtd;
1229 	struct hlist_node	s_instances;
1230 	unsigned int		s_quota_types;	/* Bitmask of supported quota types */
1231 	struct quota_info	s_dquot;	/* Diskquota specific options */
1232 
1233 	struct sb_writers	s_writers;
1234 
1235 	/*
1236 	 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1237 	 * s_fsnotify_marks together for cache efficiency. They are frequently
1238 	 * accessed and rarely modified.
1239 	 */
1240 	void			*s_fs_info;	/* Filesystem private info */
1241 
1242 	/* Granularity of c/m/atime in ns (cannot be worse than a second) */
1243 	u32			s_time_gran;
1244 	/* Time limits for c/m/atime in seconds */
1245 	time64_t		   s_time_min;
1246 	time64_t		   s_time_max;
1247 #ifdef CONFIG_FSNOTIFY
1248 	__u32			s_fsnotify_mask;
1249 	struct fsnotify_mark_connector __rcu	*s_fsnotify_marks;
1250 #endif
1251 
1252 	char			s_id[32];	/* Informational name */
1253 	uuid_t			s_uuid;		/* UUID */
1254 
1255 	unsigned int		s_max_links;
1256 
1257 	/*
1258 	 * The next field is for VFS *only*. No filesystems have any business
1259 	 * even looking at it. You had been warned.
1260 	 */
1261 	struct mutex s_vfs_rename_mutex;	/* Kludge */
1262 
1263 	/*
1264 	 * Filesystem subtype.  If non-empty the filesystem type field
1265 	 * in /proc/mounts will be "type.subtype"
1266 	 */
1267 	const char *s_subtype;
1268 
1269 	const struct dentry_operations *s_d_op; /* default d_op for dentries */
1270 
1271 	struct shrinker s_shrink;	/* per-sb shrinker handle */
1272 
1273 	/* Number of inodes with nlink == 0 but still referenced */
1274 	atomic_long_t s_remove_count;
1275 
1276 	/*
1277 	 * Number of inode/mount/sb objects that are being watched, note that
1278 	 * inodes objects are currently double-accounted.
1279 	 */
1280 	atomic_long_t s_fsnotify_connectors;
1281 
1282 	/* Read-only state of the superblock is being changed */
1283 	int s_readonly_remount;
1284 
1285 	/* per-sb errseq_t for reporting writeback errors via syncfs */
1286 	errseq_t s_wb_err;
1287 
1288 	/* AIO completions deferred from interrupt context */
1289 	struct workqueue_struct *s_dio_done_wq;
1290 	struct hlist_head s_pins;
1291 
1292 	/*
1293 	 * Owning user namespace and default context in which to
1294 	 * interpret filesystem uids, gids, quotas, device nodes,
1295 	 * xattrs and security labels.
1296 	 */
1297 	struct user_namespace *s_user_ns;
1298 
1299 	/*
1300 	 * The list_lru structure is essentially just a pointer to a table
1301 	 * of per-node lru lists, each of which has its own spinlock.
1302 	 * There is no need to put them into separate cachelines.
1303 	 */
1304 	struct list_lru		s_dentry_lru;
1305 	struct list_lru		s_inode_lru;
1306 	struct rcu_head		rcu;
1307 	struct work_struct	destroy_work;
1308 
1309 	struct mutex		s_sync_lock;	/* sync serialisation lock */
1310 
1311 	/*
1312 	 * Indicates how deep in a filesystem stack this SB is
1313 	 */
1314 	int s_stack_depth;
1315 
1316 	/* s_inode_list_lock protects s_inodes */
1317 	spinlock_t		s_inode_list_lock ____cacheline_aligned_in_smp;
1318 	struct list_head	s_inodes;	/* all inodes */
1319 
1320 	spinlock_t		s_inode_wblist_lock;
1321 	struct list_head	s_inodes_wb;	/* writeback inodes */
1322 } __randomize_layout;
1323 
i_user_ns(const struct inode * inode)1324 static inline struct user_namespace *i_user_ns(const struct inode *inode)
1325 {
1326 	return inode->i_sb->s_user_ns;
1327 }
1328 
1329 /* Helper functions so that in most cases filesystems will
1330  * not need to deal directly with kuid_t and kgid_t and can
1331  * instead deal with the raw numeric values that are stored
1332  * in the filesystem.
1333  */
i_uid_read(const struct inode * inode)1334 static inline uid_t i_uid_read(const struct inode *inode)
1335 {
1336 	return from_kuid(i_user_ns(inode), inode->i_uid);
1337 }
1338 
i_gid_read(const struct inode * inode)1339 static inline gid_t i_gid_read(const struct inode *inode)
1340 {
1341 	return from_kgid(i_user_ns(inode), inode->i_gid);
1342 }
1343 
i_uid_write(struct inode * inode,uid_t uid)1344 static inline void i_uid_write(struct inode *inode, uid_t uid)
1345 {
1346 	inode->i_uid = make_kuid(i_user_ns(inode), uid);
1347 }
1348 
i_gid_write(struct inode * inode,gid_t gid)1349 static inline void i_gid_write(struct inode *inode, gid_t gid)
1350 {
1351 	inode->i_gid = make_kgid(i_user_ns(inode), gid);
1352 }
1353 
1354 /**
1355  * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping
1356  * @idmap: idmap of the mount the inode was found from
1357  * @inode: inode to map
1358  *
1359  * Return: whe inode's i_uid mapped down according to @idmap.
1360  * If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1361  */
i_uid_into_vfsuid(struct mnt_idmap * idmap,const struct inode * inode)1362 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap,
1363 					 const struct inode *inode)
1364 {
1365 	return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid);
1366 }
1367 
1368 /**
1369  * i_uid_needs_update - check whether inode's i_uid needs to be updated
1370  * @idmap: idmap of the mount the inode was found from
1371  * @attr: the new attributes of @inode
1372  * @inode: the inode to update
1373  *
1374  * Check whether the $inode's i_uid field needs to be updated taking idmapped
1375  * mounts into account if the filesystem supports it.
1376  *
1377  * Return: true if @inode's i_uid field needs to be updated, false if not.
1378  */
i_uid_needs_update(struct mnt_idmap * idmap,const struct iattr * attr,const struct inode * inode)1379 static inline bool i_uid_needs_update(struct mnt_idmap *idmap,
1380 				      const struct iattr *attr,
1381 				      const struct inode *inode)
1382 {
1383 	return ((attr->ia_valid & ATTR_UID) &&
1384 		!vfsuid_eq(attr->ia_vfsuid,
1385 			   i_uid_into_vfsuid(idmap, inode)));
1386 }
1387 
1388 /**
1389  * i_uid_update - update @inode's i_uid field
1390  * @idmap: idmap of the mount the inode was found from
1391  * @attr: the new attributes of @inode
1392  * @inode: the inode to update
1393  *
1394  * Safely update @inode's i_uid field translating the vfsuid of any idmapped
1395  * mount into the filesystem kuid.
1396  */
i_uid_update(struct mnt_idmap * idmap,const struct iattr * attr,struct inode * inode)1397 static inline void i_uid_update(struct mnt_idmap *idmap,
1398 				const struct iattr *attr,
1399 				struct inode *inode)
1400 {
1401 	if (attr->ia_valid & ATTR_UID)
1402 		inode->i_uid = from_vfsuid(idmap, i_user_ns(inode),
1403 					   attr->ia_vfsuid);
1404 }
1405 
1406 /**
1407  * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping
1408  * @idmap: idmap of the mount the inode was found from
1409  * @inode: inode to map
1410  *
1411  * Return: the inode's i_gid mapped down according to @idmap.
1412  * If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1413  */
i_gid_into_vfsgid(struct mnt_idmap * idmap,const struct inode * inode)1414 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap,
1415 					 const struct inode *inode)
1416 {
1417 	return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid);
1418 }
1419 
1420 /**
1421  * i_gid_needs_update - check whether inode's i_gid needs to be updated
1422  * @idmap: idmap of the mount the inode was found from
1423  * @attr: the new attributes of @inode
1424  * @inode: the inode to update
1425  *
1426  * Check whether the $inode's i_gid field needs to be updated taking idmapped
1427  * mounts into account if the filesystem supports it.
1428  *
1429  * Return: true if @inode's i_gid field needs to be updated, false if not.
1430  */
i_gid_needs_update(struct mnt_idmap * idmap,const struct iattr * attr,const struct inode * inode)1431 static inline bool i_gid_needs_update(struct mnt_idmap *idmap,
1432 				      const struct iattr *attr,
1433 				      const struct inode *inode)
1434 {
1435 	return ((attr->ia_valid & ATTR_GID) &&
1436 		!vfsgid_eq(attr->ia_vfsgid,
1437 			   i_gid_into_vfsgid(idmap, inode)));
1438 }
1439 
1440 /**
1441  * i_gid_update - update @inode's i_gid field
1442  * @idmap: idmap of the mount the inode was found from
1443  * @attr: the new attributes of @inode
1444  * @inode: the inode to update
1445  *
1446  * Safely update @inode's i_gid field translating the vfsgid of any idmapped
1447  * mount into the filesystem kgid.
1448  */
i_gid_update(struct mnt_idmap * idmap,const struct iattr * attr,struct inode * inode)1449 static inline void i_gid_update(struct mnt_idmap *idmap,
1450 				const struct iattr *attr,
1451 				struct inode *inode)
1452 {
1453 	if (attr->ia_valid & ATTR_GID)
1454 		inode->i_gid = from_vfsgid(idmap, i_user_ns(inode),
1455 					   attr->ia_vfsgid);
1456 }
1457 
1458 /**
1459  * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1460  * @inode: inode to initialize
1461  * @idmap: idmap of the mount the inode was found from
1462  *
1463  * Initialize the i_uid field of @inode. If the inode was found/created via
1464  * an idmapped mount map the caller's fsuid according to @idmap.
1465  */
inode_fsuid_set(struct inode * inode,struct mnt_idmap * idmap)1466 static inline void inode_fsuid_set(struct inode *inode,
1467 				   struct mnt_idmap *idmap)
1468 {
1469 	inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode));
1470 }
1471 
1472 /**
1473  * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1474  * @inode: inode to initialize
1475  * @idmap: idmap of the mount the inode was found from
1476  *
1477  * Initialize the i_gid field of @inode. If the inode was found/created via
1478  * an idmapped mount map the caller's fsgid according to @idmap.
1479  */
inode_fsgid_set(struct inode * inode,struct mnt_idmap * idmap)1480 static inline void inode_fsgid_set(struct inode *inode,
1481 				   struct mnt_idmap *idmap)
1482 {
1483 	inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode));
1484 }
1485 
1486 /**
1487  * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1488  * @sb: the superblock we want a mapping in
1489  * @idmap: idmap of the relevant mount
1490  *
1491  * Check whether the caller's fsuid and fsgid have a valid mapping in the
1492  * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1493  * the caller's fsuid and fsgid according to the @idmap first.
1494  *
1495  * Return: true if fsuid and fsgid is mapped, false if not.
1496  */
fsuidgid_has_mapping(struct super_block * sb,struct mnt_idmap * idmap)1497 static inline bool fsuidgid_has_mapping(struct super_block *sb,
1498 					struct mnt_idmap *idmap)
1499 {
1500 	struct user_namespace *fs_userns = sb->s_user_ns;
1501 	kuid_t kuid;
1502 	kgid_t kgid;
1503 
1504 	kuid = mapped_fsuid(idmap, fs_userns);
1505 	if (!uid_valid(kuid))
1506 		return false;
1507 	kgid = mapped_fsgid(idmap, fs_userns);
1508 	if (!gid_valid(kgid))
1509 		return false;
1510 	return kuid_has_mapping(fs_userns, kuid) &&
1511 	       kgid_has_mapping(fs_userns, kgid);
1512 }
1513 
1514 struct timespec64 current_time(struct inode *inode);
1515 struct timespec64 inode_set_ctime_current(struct inode *inode);
1516 
inode_get_atime_sec(const struct inode * inode)1517 static inline time64_t inode_get_atime_sec(const struct inode *inode)
1518 {
1519 	return inode->i_atime.tv_sec;
1520 }
1521 
inode_get_atime_nsec(const struct inode * inode)1522 static inline long inode_get_atime_nsec(const struct inode *inode)
1523 {
1524 	return inode->i_atime.tv_nsec;
1525 }
1526 
inode_get_atime(const struct inode * inode)1527 static inline struct timespec64 inode_get_atime(const struct inode *inode)
1528 {
1529 	return inode->i_atime;
1530 }
1531 
inode_set_atime_to_ts(struct inode * inode,struct timespec64 ts)1532 static inline struct timespec64 inode_set_atime_to_ts(struct inode *inode,
1533 						      struct timespec64 ts)
1534 {
1535 	inode->i_atime = ts;
1536 	return ts;
1537 }
1538 
inode_set_atime(struct inode * inode,time64_t sec,long nsec)1539 static inline struct timespec64 inode_set_atime(struct inode *inode,
1540 						time64_t sec, long nsec)
1541 {
1542 	struct timespec64 ts = { .tv_sec  = sec,
1543 				 .tv_nsec = nsec };
1544 	return inode_set_atime_to_ts(inode, ts);
1545 }
1546 
inode_get_mtime_sec(const struct inode * inode)1547 static inline time64_t inode_get_mtime_sec(const struct inode *inode)
1548 {
1549 	return inode->i_mtime.tv_sec;
1550 }
1551 
inode_get_mtime_nsec(const struct inode * inode)1552 static inline long inode_get_mtime_nsec(const struct inode *inode)
1553 {
1554 	return inode->i_mtime.tv_nsec;
1555 }
1556 
inode_get_mtime(const struct inode * inode)1557 static inline struct timespec64 inode_get_mtime(const struct inode *inode)
1558 {
1559 	return inode->i_mtime;
1560 }
1561 
inode_set_mtime_to_ts(struct inode * inode,struct timespec64 ts)1562 static inline struct timespec64 inode_set_mtime_to_ts(struct inode *inode,
1563 						      struct timespec64 ts)
1564 {
1565 	inode->i_mtime = ts;
1566 	return ts;
1567 }
1568 
inode_set_mtime(struct inode * inode,time64_t sec,long nsec)1569 static inline struct timespec64 inode_set_mtime(struct inode *inode,
1570 						time64_t sec, long nsec)
1571 {
1572 	struct timespec64 ts = { .tv_sec  = sec,
1573 				 .tv_nsec = nsec };
1574 	return inode_set_mtime_to_ts(inode, ts);
1575 }
1576 
inode_get_ctime_sec(const struct inode * inode)1577 static inline time64_t inode_get_ctime_sec(const struct inode *inode)
1578 {
1579 	return inode->__i_ctime.tv_sec;
1580 }
1581 
inode_get_ctime_nsec(const struct inode * inode)1582 static inline long inode_get_ctime_nsec(const struct inode *inode)
1583 {
1584 	return inode->__i_ctime.tv_nsec;
1585 }
1586 
inode_get_ctime(const struct inode * inode)1587 static inline struct timespec64 inode_get_ctime(const struct inode *inode)
1588 {
1589 	return inode->__i_ctime;
1590 }
1591 
inode_set_ctime_to_ts(struct inode * inode,struct timespec64 ts)1592 static inline struct timespec64 inode_set_ctime_to_ts(struct inode *inode,
1593 						      struct timespec64 ts)
1594 {
1595 	inode->__i_ctime = ts;
1596 	return ts;
1597 }
1598 
1599 /**
1600  * inode_set_ctime - set the ctime in the inode
1601  * @inode: inode in which to set the ctime
1602  * @sec: tv_sec value to set
1603  * @nsec: tv_nsec value to set
1604  *
1605  * Set the ctime in @inode to { @sec, @nsec }
1606  */
inode_set_ctime(struct inode * inode,time64_t sec,long nsec)1607 static inline struct timespec64 inode_set_ctime(struct inode *inode,
1608 						time64_t sec, long nsec)
1609 {
1610 	struct timespec64 ts = { .tv_sec  = sec,
1611 				 .tv_nsec = nsec };
1612 
1613 	return inode_set_ctime_to_ts(inode, ts);
1614 }
1615 
1616 struct timespec64 simple_inode_init_ts(struct inode *inode);
1617 
1618 /*
1619  * Snapshotting support.
1620  */
1621 
1622 /*
1623  * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1624  * instead.
1625  */
__sb_end_write(struct super_block * sb,int level)1626 static inline void __sb_end_write(struct super_block *sb, int level)
1627 {
1628 	percpu_up_read(sb->s_writers.rw_sem + level-1);
1629 }
1630 
__sb_start_write(struct super_block * sb,int level)1631 static inline void __sb_start_write(struct super_block *sb, int level)
1632 {
1633 	percpu_down_read(sb->s_writers.rw_sem + level - 1);
1634 }
1635 
__sb_start_write_trylock(struct super_block * sb,int level)1636 static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1637 {
1638 	return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1639 }
1640 
1641 #define __sb_writers_acquired(sb, lev)	\
1642 	percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1643 #define __sb_writers_release(sb, lev)	\
1644 	percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1645 
sb_write_started(const struct super_block * sb)1646 static inline bool sb_write_started(const struct super_block *sb)
1647 {
1648 	return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1);
1649 }
1650 
1651 /**
1652  * sb_end_write - drop write access to a superblock
1653  * @sb: the super we wrote to
1654  *
1655  * Decrement number of writers to the filesystem. Wake up possible waiters
1656  * wanting to freeze the filesystem.
1657  */
sb_end_write(struct super_block * sb)1658 static inline void sb_end_write(struct super_block *sb)
1659 {
1660 	__sb_end_write(sb, SB_FREEZE_WRITE);
1661 }
1662 
1663 /**
1664  * sb_end_pagefault - drop write access to a superblock from a page fault
1665  * @sb: the super we wrote to
1666  *
1667  * Decrement number of processes handling write page fault to the filesystem.
1668  * Wake up possible waiters wanting to freeze the filesystem.
1669  */
sb_end_pagefault(struct super_block * sb)1670 static inline void sb_end_pagefault(struct super_block *sb)
1671 {
1672 	__sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1673 }
1674 
1675 /**
1676  * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1677  * @sb: the super we wrote to
1678  *
1679  * Decrement fs-internal number of writers to the filesystem.  Wake up possible
1680  * waiters wanting to freeze the filesystem.
1681  */
sb_end_intwrite(struct super_block * sb)1682 static inline void sb_end_intwrite(struct super_block *sb)
1683 {
1684 	__sb_end_write(sb, SB_FREEZE_FS);
1685 }
1686 
1687 /**
1688  * sb_start_write - get write access to a superblock
1689  * @sb: the super we write to
1690  *
1691  * When a process wants to write data or metadata to a file system (i.e. dirty
1692  * a page or an inode), it should embed the operation in a sb_start_write() -
1693  * sb_end_write() pair to get exclusion against file system freezing. This
1694  * function increments number of writers preventing freezing. If the file
1695  * system is already frozen, the function waits until the file system is
1696  * thawed.
1697  *
1698  * Since freeze protection behaves as a lock, users have to preserve
1699  * ordering of freeze protection and other filesystem locks. Generally,
1700  * freeze protection should be the outermost lock. In particular, we have:
1701  *
1702  * sb_start_write
1703  *   -> i_mutex			(write path, truncate, directory ops, ...)
1704  *   -> s_umount		(freeze_super, thaw_super)
1705  */
sb_start_write(struct super_block * sb)1706 static inline void sb_start_write(struct super_block *sb)
1707 {
1708 	__sb_start_write(sb, SB_FREEZE_WRITE);
1709 }
1710 
sb_start_write_trylock(struct super_block * sb)1711 static inline bool sb_start_write_trylock(struct super_block *sb)
1712 {
1713 	return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1714 }
1715 
1716 /**
1717  * sb_start_pagefault - get write access to a superblock from a page fault
1718  * @sb: the super we write to
1719  *
1720  * When a process starts handling write page fault, it should embed the
1721  * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1722  * exclusion against file system freezing. This is needed since the page fault
1723  * is going to dirty a page. This function increments number of running page
1724  * faults preventing freezing. If the file system is already frozen, the
1725  * function waits until the file system is thawed.
1726  *
1727  * Since page fault freeze protection behaves as a lock, users have to preserve
1728  * ordering of freeze protection and other filesystem locks. It is advised to
1729  * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1730  * handling code implies lock dependency:
1731  *
1732  * mmap_lock
1733  *   -> sb_start_pagefault
1734  */
sb_start_pagefault(struct super_block * sb)1735 static inline void sb_start_pagefault(struct super_block *sb)
1736 {
1737 	__sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1738 }
1739 
1740 /**
1741  * sb_start_intwrite - get write access to a superblock for internal fs purposes
1742  * @sb: the super we write to
1743  *
1744  * This is the third level of protection against filesystem freezing. It is
1745  * free for use by a filesystem. The only requirement is that it must rank
1746  * below sb_start_pagefault.
1747  *
1748  * For example filesystem can call sb_start_intwrite() when starting a
1749  * transaction which somewhat eases handling of freezing for internal sources
1750  * of filesystem changes (internal fs threads, discarding preallocation on file
1751  * close, etc.).
1752  */
sb_start_intwrite(struct super_block * sb)1753 static inline void sb_start_intwrite(struct super_block *sb)
1754 {
1755 	__sb_start_write(sb, SB_FREEZE_FS);
1756 }
1757 
sb_start_intwrite_trylock(struct super_block * sb)1758 static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1759 {
1760 	return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1761 }
1762 
1763 bool inode_owner_or_capable(struct mnt_idmap *idmap,
1764 			    const struct inode *inode);
1765 
1766 /*
1767  * VFS helper functions..
1768  */
1769 int vfs_create(struct mnt_idmap *, struct inode *,
1770 	       struct dentry *, umode_t, bool);
1771 int vfs_mkdir(struct mnt_idmap *, struct inode *,
1772 	      struct dentry *, umode_t);
1773 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *,
1774               umode_t, dev_t);
1775 int vfs_symlink(struct mnt_idmap *, struct inode *,
1776 		struct dentry *, const char *);
1777 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *,
1778 	     struct dentry *, struct inode **);
1779 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *);
1780 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *,
1781 	       struct inode **);
1782 
1783 /**
1784  * struct renamedata - contains all information required for renaming
1785  * @old_mnt_idmap:     idmap of the old mount the inode was found from
1786  * @old_dir:           parent of source
1787  * @old_dentry:                source
1788  * @new_mnt_idmap:     idmap of the new mount the inode was found from
1789  * @new_dir:           parent of destination
1790  * @new_dentry:                destination
1791  * @delegated_inode:   returns an inode needing a delegation break
1792  * @flags:             rename flags
1793  */
1794 struct renamedata {
1795 	struct mnt_idmap *old_mnt_idmap;
1796 	struct inode *old_dir;
1797 	struct dentry *old_dentry;
1798 	struct mnt_idmap *new_mnt_idmap;
1799 	struct inode *new_dir;
1800 	struct dentry *new_dentry;
1801 	struct inode **delegated_inode;
1802 	unsigned int flags;
1803 } __randomize_layout;
1804 
1805 int vfs_rename(struct renamedata *);
1806 
vfs_whiteout(struct mnt_idmap * idmap,struct inode * dir,struct dentry * dentry)1807 static inline int vfs_whiteout(struct mnt_idmap *idmap,
1808 			       struct inode *dir, struct dentry *dentry)
1809 {
1810 	return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE,
1811 			 WHITEOUT_DEV);
1812 }
1813 
1814 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap,
1815 				 const struct path *parentpath,
1816 				 umode_t mode, int open_flag,
1817 				 const struct cred *cred);
1818 struct file *kernel_file_open(const struct path *path, int flags,
1819 			      struct inode *inode, const struct cred *cred);
1820 
1821 int vfs_mkobj(struct dentry *, umode_t,
1822 		int (*f)(struct dentry *, umode_t, void *),
1823 		void *);
1824 
1825 int vfs_fchown(struct file *file, uid_t user, gid_t group);
1826 int vfs_fchmod(struct file *file, umode_t mode);
1827 int vfs_utimes(const struct path *path, struct timespec64 *times);
1828 
1829 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1830 
1831 #ifdef CONFIG_COMPAT
1832 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
1833 					unsigned long arg);
1834 #else
1835 #define compat_ptr_ioctl NULL
1836 #endif
1837 
1838 /*
1839  * VFS file helper functions.
1840  */
1841 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
1842 		      const struct inode *dir, umode_t mode);
1843 extern bool may_open_dev(const struct path *path);
1844 umode_t mode_strip_sgid(struct mnt_idmap *idmap,
1845 			const struct inode *dir, umode_t mode);
1846 
1847 /*
1848  * This is the "filldir" function type, used by readdir() to let
1849  * the kernel specify what kind of dirent layout it wants to have.
1850  * This allows the kernel to read directories into kernel space or
1851  * to have different dirent layouts depending on the binary type.
1852  * Return 'true' to keep going and 'false' if there are no more entries.
1853  */
1854 struct dir_context;
1855 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1856 			 unsigned);
1857 
1858 struct dir_context {
1859 	filldir_t actor;
1860 	loff_t pos;
1861 };
1862 
1863 /*
1864  * These flags let !MMU mmap() govern direct device mapping vs immediate
1865  * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1866  *
1867  * NOMMU_MAP_COPY:	Copy can be mapped (MAP_PRIVATE)
1868  * NOMMU_MAP_DIRECT:	Can be mapped directly (MAP_SHARED)
1869  * NOMMU_MAP_READ:	Can be mapped for reading
1870  * NOMMU_MAP_WRITE:	Can be mapped for writing
1871  * NOMMU_MAP_EXEC:	Can be mapped for execution
1872  */
1873 #define NOMMU_MAP_COPY		0x00000001
1874 #define NOMMU_MAP_DIRECT	0x00000008
1875 #define NOMMU_MAP_READ		VM_MAYREAD
1876 #define NOMMU_MAP_WRITE		VM_MAYWRITE
1877 #define NOMMU_MAP_EXEC		VM_MAYEXEC
1878 
1879 #define NOMMU_VMFLAGS \
1880 	(NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
1881 
1882 /*
1883  * These flags control the behavior of the remap_file_range function pointer.
1884  * If it is called with len == 0 that means "remap to end of source file".
1885  * See Documentation/filesystems/vfs.rst for more details about this call.
1886  *
1887  * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
1888  * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
1889  */
1890 #define REMAP_FILE_DEDUP		(1 << 0)
1891 #define REMAP_FILE_CAN_SHORTEN		(1 << 1)
1892 
1893 /*
1894  * These flags signal that the caller is ok with altering various aspects of
1895  * the behavior of the remap operation.  The changes must be made by the
1896  * implementation; the vfs remap helper functions can take advantage of them.
1897  * Flags in this category exist to preserve the quirky behavior of the hoisted
1898  * btrfs clone/dedupe ioctls.
1899  */
1900 #define REMAP_FILE_ADVISORY		(REMAP_FILE_CAN_SHORTEN)
1901 
1902 /*
1903  * These flags control the behavior of vfs_copy_file_range().
1904  * They are not available to the user via syscall.
1905  *
1906  * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops
1907  */
1908 #define COPY_FILE_SPLICE		(1 << 0)
1909 
1910 struct iov_iter;
1911 struct io_uring_cmd;
1912 struct offset_ctx;
1913 
1914 struct file_operations {
1915 	struct module *owner;
1916 	loff_t (*llseek) (struct file *, loff_t, int);
1917 	ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1918 	ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1919 	ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1920 	ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1921 	int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
1922 			unsigned int flags);
1923 	int (*iterate_shared) (struct file *, struct dir_context *);
1924 	__poll_t (*poll) (struct file *, struct poll_table_struct *);
1925 	long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1926 	long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1927 	int (*mmap) (struct file *, struct vm_area_struct *);
1928 	unsigned long mmap_supported_flags;
1929 	int (*open) (struct inode *, struct file *);
1930 	int (*flush) (struct file *, fl_owner_t id);
1931 	int (*release) (struct inode *, struct file *);
1932 	int (*fsync) (struct file *, loff_t, loff_t, int datasync);
1933 	int (*fasync) (int, struct file *, int);
1934 	int (*lock) (struct file *, int, struct file_lock *);
1935 	unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1936 	int (*check_flags)(int);
1937 	int (*flock) (struct file *, int, struct file_lock *);
1938 	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
1939 	ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
1940 	void (*splice_eof)(struct file *file);
1941 	int (*setlease)(struct file *, int, struct file_lock **, void **);
1942 	long (*fallocate)(struct file *file, int mode, loff_t offset,
1943 			  loff_t len);
1944 	void (*show_fdinfo)(struct seq_file *m, struct file *f);
1945 #ifndef CONFIG_MMU
1946 	unsigned (*mmap_capabilities)(struct file *);
1947 #endif
1948 	ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
1949 			loff_t, size_t, unsigned int);
1950 	loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
1951 				   struct file *file_out, loff_t pos_out,
1952 				   loff_t len, unsigned int remap_flags);
1953 	int (*fadvise)(struct file *, loff_t, loff_t, int);
1954 	int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
1955 	int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *,
1956 				unsigned int poll_flags);
1957 } __randomize_layout;
1958 
1959 /* Wrap a directory iterator that needs exclusive inode access */
1960 int wrap_directory_iterator(struct file *, struct dir_context *,
1961 			    int (*) (struct file *, struct dir_context *));
1962 #define WRAP_DIR_ITER(x) \
1963 	static int shared_##x(struct file *file , struct dir_context *ctx) \
1964 	{ return wrap_directory_iterator(file, ctx, x); }
1965 
1966 struct inode_operations {
1967 	struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
1968 	const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
1969 	int (*permission) (struct mnt_idmap *, struct inode *, int);
1970 	struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
1971 
1972 	int (*readlink) (struct dentry *, char __user *,int);
1973 
1974 	int (*create) (struct mnt_idmap *, struct inode *,struct dentry *,
1975 		       umode_t, bool);
1976 	int (*link) (struct dentry *,struct inode *,struct dentry *);
1977 	int (*unlink) (struct inode *,struct dentry *);
1978 	int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,
1979 			const char *);
1980 	int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,
1981 		      umode_t);
1982 	int (*rmdir) (struct inode *,struct dentry *);
1983 	int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,
1984 		      umode_t,dev_t);
1985 	int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
1986 			struct inode *, struct dentry *, unsigned int);
1987 	int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
1988 	int (*getattr) (struct mnt_idmap *, const struct path *,
1989 			struct kstat *, u32, unsigned int);
1990 	ssize_t (*listxattr) (struct dentry *, char *, size_t);
1991 	int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
1992 		      u64 len);
1993 	int (*update_time)(struct inode *, int);
1994 	int (*atomic_open)(struct inode *, struct dentry *,
1995 			   struct file *, unsigned open_flag,
1996 			   umode_t create_mode);
1997 	int (*tmpfile) (struct mnt_idmap *, struct inode *,
1998 			struct file *, umode_t);
1999 	struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *,
2000 				     int);
2001 	int (*set_acl)(struct mnt_idmap *, struct dentry *,
2002 		       struct posix_acl *, int);
2003 	int (*fileattr_set)(struct mnt_idmap *idmap,
2004 			    struct dentry *dentry, struct fileattr *fa);
2005 	int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
2006 	struct offset_ctx *(*get_offset_ctx)(struct inode *inode);
2007 } ____cacheline_aligned;
2008 
call_read_iter(struct file * file,struct kiocb * kio,struct iov_iter * iter)2009 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
2010 				     struct iov_iter *iter)
2011 {
2012 	return file->f_op->read_iter(kio, iter);
2013 }
2014 
call_write_iter(struct file * file,struct kiocb * kio,struct iov_iter * iter)2015 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
2016 				      struct iov_iter *iter)
2017 {
2018 	return file->f_op->write_iter(kio, iter);
2019 }
2020 
call_mmap(struct file * file,struct vm_area_struct * vma)2021 static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
2022 {
2023 	return file->f_op->mmap(file, vma);
2024 }
2025 
2026 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
2027 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
2028 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
2029 				   loff_t, size_t, unsigned int);
2030 extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
2031 				       struct file *file_out, loff_t pos_out,
2032 				       size_t len, unsigned int flags);
2033 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2034 				    struct file *file_out, loff_t pos_out,
2035 				    loff_t *len, unsigned int remap_flags,
2036 				    const struct iomap_ops *dax_read_ops);
2037 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2038 				  struct file *file_out, loff_t pos_out,
2039 				  loff_t *count, unsigned int remap_flags);
2040 extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in,
2041 				  struct file *file_out, loff_t pos_out,
2042 				  loff_t len, unsigned int remap_flags);
2043 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
2044 				   struct file *file_out, loff_t pos_out,
2045 				   loff_t len, unsigned int remap_flags);
2046 extern int vfs_dedupe_file_range(struct file *file,
2047 				 struct file_dedupe_range *same);
2048 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
2049 					struct file *dst_file, loff_t dst_pos,
2050 					loff_t len, unsigned int remap_flags);
2051 
2052 enum freeze_holder {
2053 	FREEZE_HOLDER_KERNEL	= (1U << 0),
2054 	FREEZE_HOLDER_USERSPACE	= (1U << 1),
2055 };
2056 
2057 struct super_operations {
2058    	struct inode *(*alloc_inode)(struct super_block *sb);
2059 	void (*destroy_inode)(struct inode *);
2060 	void (*free_inode)(struct inode *);
2061 
2062    	void (*dirty_inode) (struct inode *, int flags);
2063 	int (*write_inode) (struct inode *, struct writeback_control *wbc);
2064 	int (*drop_inode) (struct inode *);
2065 	void (*evict_inode) (struct inode *);
2066 	void (*put_super) (struct super_block *);
2067 	int (*sync_fs)(struct super_block *sb, int wait);
2068 	int (*freeze_super) (struct super_block *, enum freeze_holder who);
2069 	int (*freeze_fs) (struct super_block *);
2070 	int (*thaw_super) (struct super_block *, enum freeze_holder who);
2071 	int (*unfreeze_fs) (struct super_block *);
2072 	int (*statfs) (struct dentry *, struct kstatfs *);
2073 	int (*remount_fs) (struct super_block *, int *, char *);
2074 	void (*umount_begin) (struct super_block *);
2075 
2076 	int (*show_options)(struct seq_file *, struct dentry *);
2077 	int (*show_devname)(struct seq_file *, struct dentry *);
2078 	int (*show_path)(struct seq_file *, struct dentry *);
2079 	int (*show_stats)(struct seq_file *, struct dentry *);
2080 #ifdef CONFIG_QUOTA
2081 	ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
2082 	ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
2083 	struct dquot __rcu **(*get_dquots)(struct inode *);
2084 #endif
2085 	long (*nr_cached_objects)(struct super_block *,
2086 				  struct shrink_control *);
2087 	long (*free_cached_objects)(struct super_block *,
2088 				    struct shrink_control *);
2089 	void (*shutdown)(struct super_block *sb);
2090 };
2091 
2092 /*
2093  * Inode flags - they have no relation to superblock flags now
2094  */
2095 #define S_SYNC		(1 << 0)  /* Writes are synced at once */
2096 #define S_NOATIME	(1 << 1)  /* Do not update access times */
2097 #define S_APPEND	(1 << 2)  /* Append-only file */
2098 #define S_IMMUTABLE	(1 << 3)  /* Immutable file */
2099 #define S_DEAD		(1 << 4)  /* removed, but still open directory */
2100 #define S_NOQUOTA	(1 << 5)  /* Inode is not counted to quota */
2101 #define S_DIRSYNC	(1 << 6)  /* Directory modifications are synchronous */
2102 #define S_NOCMTIME	(1 << 7)  /* Do not update file c/mtime */
2103 #define S_SWAPFILE	(1 << 8)  /* Do not truncate: swapon got its bmaps */
2104 #define S_PRIVATE	(1 << 9)  /* Inode is fs-internal */
2105 #define S_IMA		(1 << 10) /* Inode has an associated IMA struct */
2106 #define S_AUTOMOUNT	(1 << 11) /* Automount/referral quasi-directory */
2107 #define S_NOSEC		(1 << 12) /* no suid or xattr security attributes */
2108 #ifdef CONFIG_FS_DAX
2109 #define S_DAX		(1 << 13) /* Direct Access, avoiding the page cache */
2110 #else
2111 #define S_DAX		0	  /* Make all the DAX code disappear */
2112 #endif
2113 #define S_ENCRYPTED	(1 << 14) /* Encrypted file (using fs/crypto/) */
2114 #define S_CASEFOLD	(1 << 15) /* Casefolded file */
2115 #define S_VERITY	(1 << 16) /* Verity file (using fs/verity/) */
2116 #define S_KERNEL_FILE	(1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
2117 
2118 /*
2119  * Note that nosuid etc flags are inode-specific: setting some file-system
2120  * flags just means all the inodes inherit those flags by default. It might be
2121  * possible to override it selectively if you really wanted to with some
2122  * ioctl() that is not currently implemented.
2123  *
2124  * Exception: SB_RDONLY is always applied to the entire file system.
2125  *
2126  * Unfortunately, it is possible to change a filesystems flags with it mounted
2127  * with files in use.  This means that all of the inodes will not have their
2128  * i_flags updated.  Hence, i_flags no longer inherit the superblock mount
2129  * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
2130  */
2131 #define __IS_FLG(inode, flg)	((inode)->i_sb->s_flags & (flg))
2132 
sb_rdonly(const struct super_block * sb)2133 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
2134 #define IS_RDONLY(inode)	sb_rdonly((inode)->i_sb)
2135 #define IS_SYNC(inode)		(__IS_FLG(inode, SB_SYNCHRONOUS) || \
2136 					((inode)->i_flags & S_SYNC))
2137 #define IS_DIRSYNC(inode)	(__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
2138 					((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
2139 #define IS_MANDLOCK(inode)	__IS_FLG(inode, SB_MANDLOCK)
2140 #define IS_NOATIME(inode)	__IS_FLG(inode, SB_RDONLY|SB_NOATIME)
2141 #define IS_I_VERSION(inode)	__IS_FLG(inode, SB_I_VERSION)
2142 
2143 #define IS_NOQUOTA(inode)	((inode)->i_flags & S_NOQUOTA)
2144 #define IS_APPEND(inode)	((inode)->i_flags & S_APPEND)
2145 #define IS_IMMUTABLE(inode)	((inode)->i_flags & S_IMMUTABLE)
2146 #define IS_POSIXACL(inode)	__IS_FLG(inode, SB_POSIXACL)
2147 
2148 #define IS_DEADDIR(inode)	((inode)->i_flags & S_DEAD)
2149 #define IS_NOCMTIME(inode)	((inode)->i_flags & S_NOCMTIME)
2150 #define IS_SWAPFILE(inode)	((inode)->i_flags & S_SWAPFILE)
2151 #define IS_PRIVATE(inode)	((inode)->i_flags & S_PRIVATE)
2152 #define IS_IMA(inode)		((inode)->i_flags & S_IMA)
2153 #define IS_AUTOMOUNT(inode)	((inode)->i_flags & S_AUTOMOUNT)
2154 #define IS_NOSEC(inode)		((inode)->i_flags & S_NOSEC)
2155 #define IS_DAX(inode)		((inode)->i_flags & S_DAX)
2156 #define IS_ENCRYPTED(inode)	((inode)->i_flags & S_ENCRYPTED)
2157 #define IS_CASEFOLDED(inode)	((inode)->i_flags & S_CASEFOLD)
2158 #define IS_VERITY(inode)	((inode)->i_flags & S_VERITY)
2159 
2160 #define IS_WHITEOUT(inode)	(S_ISCHR(inode->i_mode) && \
2161 				 (inode)->i_rdev == WHITEOUT_DEV)
2162 
HAS_UNMAPPED_ID(struct mnt_idmap * idmap,struct inode * inode)2163 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap,
2164 				   struct inode *inode)
2165 {
2166 	return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) ||
2167 	       !vfsgid_valid(i_gid_into_vfsgid(idmap, inode));
2168 }
2169 
init_sync_kiocb(struct kiocb * kiocb,struct file * filp)2170 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
2171 {
2172 	*kiocb = (struct kiocb) {
2173 		.ki_filp = filp,
2174 		.ki_flags = filp->f_iocb_flags,
2175 		.ki_ioprio = get_current_ioprio(),
2176 	};
2177 }
2178 
kiocb_clone(struct kiocb * kiocb,struct kiocb * kiocb_src,struct file * filp)2179 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2180 			       struct file *filp)
2181 {
2182 	*kiocb = (struct kiocb) {
2183 		.ki_filp = filp,
2184 		.ki_flags = kiocb_src->ki_flags,
2185 		.ki_ioprio = kiocb_src->ki_ioprio,
2186 		.ki_pos = kiocb_src->ki_pos,
2187 	};
2188 }
2189 
2190 /*
2191  * Inode state bits.  Protected by inode->i_lock
2192  *
2193  * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2194  * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2195  *
2196  * Four bits define the lifetime of an inode.  Initially, inodes are I_NEW,
2197  * until that flag is cleared.  I_WILL_FREE, I_FREEING and I_CLEAR are set at
2198  * various stages of removing an inode.
2199  *
2200  * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2201  *
2202  * I_DIRTY_SYNC		Inode is dirty, but doesn't have to be written on
2203  *			fdatasync() (unless I_DIRTY_DATASYNC is also set).
2204  *			Timestamp updates are the usual cause.
2205  * I_DIRTY_DATASYNC	Data-related inode changes pending.  We keep track of
2206  *			these changes separately from I_DIRTY_SYNC so that we
2207  *			don't have to write inode on fdatasync() when only
2208  *			e.g. the timestamps have changed.
2209  * I_DIRTY_PAGES	Inode has dirty pages.  Inode itself may be clean.
2210  * I_DIRTY_TIME		The inode itself has dirty timestamps, and the
2211  *			lazytime mount option is enabled.  We keep track of this
2212  *			separately from I_DIRTY_SYNC in order to implement
2213  *			lazytime.  This gets cleared if I_DIRTY_INODE
2214  *			(I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But
2215  *			I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already
2216  *			in place because writeback might already be in progress
2217  *			and we don't want to lose the time update
2218  * I_NEW		Serves as both a mutex and completion notification.
2219  *			New inodes set I_NEW.  If two processes both create
2220  *			the same inode, one of them will release its inode and
2221  *			wait for I_NEW to be released before returning.
2222  *			Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2223  *			also cause waiting on I_NEW, without I_NEW actually
2224  *			being set.  find_inode() uses this to prevent returning
2225  *			nearly-dead inodes.
2226  * I_WILL_FREE		Must be set when calling write_inode_now() if i_count
2227  *			is zero.  I_FREEING must be set when I_WILL_FREE is
2228  *			cleared.
2229  * I_FREEING		Set when inode is about to be freed but still has dirty
2230  *			pages or buffers attached or the inode itself is still
2231  *			dirty.
2232  * I_CLEAR		Added by clear_inode().  In this state the inode is
2233  *			clean and can be destroyed.  Inode keeps I_FREEING.
2234  *
2235  *			Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2236  *			prohibited for many purposes.  iget() must wait for
2237  *			the inode to be completely released, then create it
2238  *			anew.  Other functions will just ignore such inodes,
2239  *			if appropriate.  I_NEW is used for waiting.
2240  *
2241  * I_SYNC		Writeback of inode is running. The bit is set during
2242  *			data writeback, and cleared with a wakeup on the bit
2243  *			address once it is done. The bit is also used to pin
2244  *			the inode in memory for flusher thread.
2245  *
2246  * I_REFERENCED		Marks the inode as recently references on the LRU list.
2247  *
2248  * I_DIO_WAKEUP		Never set.  Only used as a key for wait_on_bit().
2249  *
2250  * I_WB_SWITCH		Cgroup bdi_writeback switching in progress.  Used to
2251  *			synchronize competing switching instances and to tell
2252  *			wb stat updates to grab the i_pages lock.  See
2253  *			inode_switch_wbs_work_fn() for details.
2254  *
2255  * I_OVL_INUSE		Used by overlayfs to get exclusive ownership on upper
2256  *			and work dirs among overlayfs mounts.
2257  *
2258  * I_CREATING		New object's inode in the middle of setting up.
2259  *
2260  * I_DONTCACHE		Evict inode as soon as it is not used anymore.
2261  *
2262  * I_SYNC_QUEUED	Inode is queued in b_io or b_more_io writeback lists.
2263  *			Used to detect that mark_inode_dirty() should not move
2264  * 			inode between dirty lists.
2265  *
2266  * I_PINNING_FSCACHE_WB	Inode is pinning an fscache object for writeback.
2267  *
2268  * I_LRU_ISOLATING	Inode is pinned being isolated from LRU without holding
2269  *			i_count.
2270  *
2271  * Q: What is the difference between I_WILL_FREE and I_FREEING?
2272  */
2273 #define I_DIRTY_SYNC		(1 << 0)
2274 #define I_DIRTY_DATASYNC	(1 << 1)
2275 #define I_DIRTY_PAGES		(1 << 2)
2276 #define __I_NEW			3
2277 #define I_NEW			(1 << __I_NEW)
2278 #define I_WILL_FREE		(1 << 4)
2279 #define I_FREEING		(1 << 5)
2280 #define I_CLEAR			(1 << 6)
2281 #define __I_SYNC		7
2282 #define I_SYNC			(1 << __I_SYNC)
2283 #define I_REFERENCED		(1 << 8)
2284 #define __I_DIO_WAKEUP		9
2285 #define I_DIO_WAKEUP		(1 << __I_DIO_WAKEUP)
2286 #define I_LINKABLE		(1 << 10)
2287 #define I_DIRTY_TIME		(1 << 11)
2288 #define I_WB_SWITCH		(1 << 13)
2289 #define I_OVL_INUSE		(1 << 14)
2290 #define I_CREATING		(1 << 15)
2291 #define I_DONTCACHE		(1 << 16)
2292 #define I_SYNC_QUEUED		(1 << 17)
2293 #define I_PINNING_FSCACHE_WB	(1 << 18)
2294 #define __I_LRU_ISOLATING	19
2295 #define I_LRU_ISOLATING		(1 << __I_LRU_ISOLATING)
2296 
2297 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2298 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2299 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2300 
2301 extern void __mark_inode_dirty(struct inode *, int);
mark_inode_dirty(struct inode * inode)2302 static inline void mark_inode_dirty(struct inode *inode)
2303 {
2304 	__mark_inode_dirty(inode, I_DIRTY);
2305 }
2306 
mark_inode_dirty_sync(struct inode * inode)2307 static inline void mark_inode_dirty_sync(struct inode *inode)
2308 {
2309 	__mark_inode_dirty(inode, I_DIRTY_SYNC);
2310 }
2311 
2312 /*
2313  * Returns true if the given inode itself only has dirty timestamps (its pages
2314  * may still be dirty) and isn't currently being allocated or freed.
2315  * Filesystems should call this if when writing an inode when lazytime is
2316  * enabled, they want to opportunistically write the timestamps of other inodes
2317  * located very nearby on-disk, e.g. in the same inode block.  This returns true
2318  * if the given inode is in need of such an opportunistic update.  Requires
2319  * i_lock, or at least later re-checking under i_lock.
2320  */
inode_is_dirtytime_only(struct inode * inode)2321 static inline bool inode_is_dirtytime_only(struct inode *inode)
2322 {
2323 	return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2324 				  I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2325 }
2326 
2327 extern void inc_nlink(struct inode *inode);
2328 extern void drop_nlink(struct inode *inode);
2329 extern void clear_nlink(struct inode *inode);
2330 extern void set_nlink(struct inode *inode, unsigned int nlink);
2331 
inode_inc_link_count(struct inode * inode)2332 static inline void inode_inc_link_count(struct inode *inode)
2333 {
2334 	inc_nlink(inode);
2335 	mark_inode_dirty(inode);
2336 }
2337 
inode_dec_link_count(struct inode * inode)2338 static inline void inode_dec_link_count(struct inode *inode)
2339 {
2340 	drop_nlink(inode);
2341 	mark_inode_dirty(inode);
2342 }
2343 
2344 enum file_time_flags {
2345 	S_ATIME = 1,
2346 	S_MTIME = 2,
2347 	S_CTIME = 4,
2348 	S_VERSION = 8,
2349 };
2350 
2351 extern bool atime_needs_update(const struct path *, struct inode *);
2352 extern void touch_atime(const struct path *);
2353 int inode_update_time(struct inode *inode, int flags);
2354 
file_accessed(struct file * file)2355 static inline void file_accessed(struct file *file)
2356 {
2357 	if (!(file->f_flags & O_NOATIME))
2358 		touch_atime(&file->f_path);
2359 }
2360 
2361 extern int file_modified(struct file *file);
2362 int kiocb_modified(struct kiocb *iocb);
2363 
2364 int sync_inode_metadata(struct inode *inode, int wait);
2365 
2366 struct file_system_type {
2367 	const char *name;
2368 	int fs_flags;
2369 #define FS_REQUIRES_DEV		1
2370 #define FS_BINARY_MOUNTDATA	2
2371 #define FS_HAS_SUBTYPE		4
2372 #define FS_USERNS_MOUNT		8	/* Can be mounted by userns root */
2373 #define FS_DISALLOW_NOTIFY_PERM	16	/* Disable fanotify permission events */
2374 #define FS_ALLOW_IDMAP         32      /* FS has been updated to handle vfs idmappings. */
2375 #define FS_RENAME_DOES_D_MOVE	32768	/* FS will handle d_move() during rename() internally. */
2376 	int (*init_fs_context)(struct fs_context *);
2377 	const struct fs_parameter_spec *parameters;
2378 	struct dentry *(*mount) (struct file_system_type *, int,
2379 		       const char *, void *);
2380 	void (*kill_sb) (struct super_block *);
2381 	struct module *owner;
2382 	struct file_system_type * next;
2383 	struct hlist_head fs_supers;
2384 
2385 	struct lock_class_key s_lock_key;
2386 	struct lock_class_key s_umount_key;
2387 	struct lock_class_key s_vfs_rename_key;
2388 	struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2389 
2390 	struct lock_class_key i_lock_key;
2391 	struct lock_class_key i_mutex_key;
2392 	struct lock_class_key invalidate_lock_key;
2393 	struct lock_class_key i_mutex_dir_key;
2394 };
2395 
2396 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2397 
2398 extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2399 	int flags, const char *dev_name, void *data,
2400 	int (*fill_super)(struct super_block *, void *, int));
2401 extern struct dentry *mount_single(struct file_system_type *fs_type,
2402 	int flags, void *data,
2403 	int (*fill_super)(struct super_block *, void *, int));
2404 extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2405 	int flags, void *data,
2406 	int (*fill_super)(struct super_block *, void *, int));
2407 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2408 void retire_super(struct super_block *sb);
2409 void generic_shutdown_super(struct super_block *sb);
2410 void kill_block_super(struct super_block *sb);
2411 void kill_anon_super(struct super_block *sb);
2412 void kill_litter_super(struct super_block *sb);
2413 void deactivate_super(struct super_block *sb);
2414 void deactivate_locked_super(struct super_block *sb);
2415 int set_anon_super(struct super_block *s, void *data);
2416 int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2417 int get_anon_bdev(dev_t *);
2418 void free_anon_bdev(dev_t);
2419 struct super_block *sget_fc(struct fs_context *fc,
2420 			    int (*test)(struct super_block *, struct fs_context *),
2421 			    int (*set)(struct super_block *, struct fs_context *));
2422 struct super_block *sget(struct file_system_type *type,
2423 			int (*test)(struct super_block *,void *),
2424 			int (*set)(struct super_block *,void *),
2425 			int flags, void *data);
2426 struct super_block *sget_dev(struct fs_context *fc, dev_t dev);
2427 
2428 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2429 #define fops_get(fops) \
2430 	(((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2431 #define fops_put(fops) \
2432 	do { if (fops) module_put((fops)->owner); } while(0)
2433 /*
2434  * This one is to be used *ONLY* from ->open() instances.
2435  * fops must be non-NULL, pinned down *and* module dependencies
2436  * should be sufficient to pin the caller down as well.
2437  */
2438 #define replace_fops(f, fops) \
2439 	do {	\
2440 		struct file *__file = (f); \
2441 		fops_put(__file->f_op); \
2442 		BUG_ON(!(__file->f_op = (fops))); \
2443 	} while(0)
2444 
2445 extern int register_filesystem(struct file_system_type *);
2446 extern int unregister_filesystem(struct file_system_type *);
2447 extern int vfs_statfs(const struct path *, struct kstatfs *);
2448 extern int user_statfs(const char __user *, struct kstatfs *);
2449 extern int fd_statfs(int, struct kstatfs *);
2450 int freeze_super(struct super_block *super, enum freeze_holder who);
2451 int thaw_super(struct super_block *super, enum freeze_holder who);
2452 extern __printf(2, 3)
2453 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2454 extern int super_setup_bdi(struct super_block *sb);
2455 
2456 extern int current_umask(void);
2457 
2458 extern void ihold(struct inode * inode);
2459 extern void iput(struct inode *);
2460 int inode_update_timestamps(struct inode *inode, int flags);
2461 int generic_update_time(struct inode *, int);
2462 
2463 /* /sys/fs */
2464 extern struct kobject *fs_kobj;
2465 
2466 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2467 
2468 /* fs/open.c */
2469 struct audit_names;
2470 struct filename {
2471 	const char		*name;	/* pointer to actual string */
2472 	const __user char	*uptr;	/* original userland pointer */
2473 	atomic_t		refcnt;
2474 	struct audit_names	*aname;
2475 	const char		iname[];
2476 };
2477 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2478 
file_mnt_idmap(struct file * file)2479 static inline struct mnt_idmap *file_mnt_idmap(struct file *file)
2480 {
2481 	return mnt_idmap(file->f_path.mnt);
2482 }
2483 
2484 /**
2485  * is_idmapped_mnt - check whether a mount is mapped
2486  * @mnt: the mount to check
2487  *
2488  * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped.
2489  *
2490  * Return: true if mount is mapped, false if not.
2491  */
is_idmapped_mnt(const struct vfsmount * mnt)2492 static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2493 {
2494 	return mnt_idmap(mnt) != &nop_mnt_idmap;
2495 }
2496 
2497 extern long vfs_truncate(const struct path *, loff_t);
2498 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start,
2499 		unsigned int time_attrs, struct file *filp);
2500 extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2501 			loff_t len);
2502 extern long do_sys_open(int dfd, const char __user *filename, int flags,
2503 			umode_t mode);
2504 extern struct file *file_open_name(struct filename *, int, umode_t);
2505 extern struct file *filp_open(const char *, int, umode_t);
2506 extern struct file *file_open_root(const struct path *,
2507 				   const char *, int, umode_t);
file_open_root_mnt(struct vfsmount * mnt,const char * name,int flags,umode_t mode)2508 static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2509 				   const char *name, int flags, umode_t mode)
2510 {
2511 	return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2512 			      name, flags, mode);
2513 }
2514 struct file *dentry_open(const struct path *path, int flags,
2515 			 const struct cred *creds);
2516 struct file *dentry_create(const struct path *path, int flags, umode_t mode,
2517 			   const struct cred *cred);
2518 struct file *backing_file_open(const struct path *path, int flags,
2519 			       const struct path *real_path,
2520 			       const struct cred *cred);
2521 struct path *backing_file_real_path(struct file *f);
2522 
2523 /*
2524  * file_real_path - get the path corresponding to f_inode
2525  *
2526  * When opening a backing file for a stackable filesystem (e.g.,
2527  * overlayfs) f_path may be on the stackable filesystem and f_inode on
2528  * the underlying filesystem.  When the path associated with f_inode is
2529  * needed, this helper should be used instead of accessing f_path
2530  * directly.
2531 */
file_real_path(struct file * f)2532 static inline const struct path *file_real_path(struct file *f)
2533 {
2534 	if (unlikely(f->f_mode & FMODE_BACKING))
2535 		return backing_file_real_path(f);
2536 	return &f->f_path;
2537 }
2538 
file_clone_open(struct file * file)2539 static inline struct file *file_clone_open(struct file *file)
2540 {
2541 	return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2542 }
2543 extern int filp_close(struct file *, fl_owner_t id);
2544 
2545 extern struct filename *getname_flags(const char __user *, int, int *);
2546 extern struct filename *getname_uflags(const char __user *, int);
2547 extern struct filename *getname(const char __user *);
2548 extern struct filename *getname_kernel(const char *);
2549 extern void putname(struct filename *name);
2550 
2551 extern int finish_open(struct file *file, struct dentry *dentry,
2552 			int (*open)(struct inode *, struct file *));
2553 extern int finish_no_open(struct file *file, struct dentry *dentry);
2554 
2555 /* Helper for the simple case when original dentry is used */
finish_open_simple(struct file * file,int error)2556 static inline int finish_open_simple(struct file *file, int error)
2557 {
2558 	if (error)
2559 		return error;
2560 
2561 	return finish_open(file, file->f_path.dentry, NULL);
2562 }
2563 
2564 /* fs/dcache.c */
2565 extern void __init vfs_caches_init_early(void);
2566 extern void __init vfs_caches_init(void);
2567 
2568 extern struct kmem_cache *names_cachep;
2569 
2570 #define __getname()		kmem_cache_alloc(names_cachep, GFP_KERNEL)
2571 #define __putname(name)		kmem_cache_free(names_cachep, (void *)(name))
2572 
2573 extern struct super_block *blockdev_superblock;
sb_is_blkdev_sb(struct super_block * sb)2574 static inline bool sb_is_blkdev_sb(struct super_block *sb)
2575 {
2576 	return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2577 }
2578 
2579 void emergency_thaw_all(void);
2580 extern int sync_filesystem(struct super_block *);
2581 extern const struct file_operations def_blk_fops;
2582 extern const struct file_operations def_chr_fops;
2583 
2584 /* fs/char_dev.c */
2585 #define CHRDEV_MAJOR_MAX 512
2586 /* Marks the bottom of the first segment of free char majors */
2587 #define CHRDEV_MAJOR_DYN_END 234
2588 /* Marks the top and bottom of the second segment of free char majors */
2589 #define CHRDEV_MAJOR_DYN_EXT_START 511
2590 #define CHRDEV_MAJOR_DYN_EXT_END 384
2591 
2592 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2593 extern int register_chrdev_region(dev_t, unsigned, const char *);
2594 extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2595 			     unsigned int count, const char *name,
2596 			     const struct file_operations *fops);
2597 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2598 				unsigned int count, const char *name);
2599 extern void unregister_chrdev_region(dev_t, unsigned);
2600 extern void chrdev_show(struct seq_file *,off_t);
2601 
register_chrdev(unsigned int major,const char * name,const struct file_operations * fops)2602 static inline int register_chrdev(unsigned int major, const char *name,
2603 				  const struct file_operations *fops)
2604 {
2605 	return __register_chrdev(major, 0, 256, name, fops);
2606 }
2607 
unregister_chrdev(unsigned int major,const char * name)2608 static inline void unregister_chrdev(unsigned int major, const char *name)
2609 {
2610 	__unregister_chrdev(major, 0, 256, name);
2611 }
2612 
2613 extern void init_special_inode(struct inode *, umode_t, dev_t);
2614 
2615 /* Invalid inode operations -- fs/bad_inode.c */
2616 extern void make_bad_inode(struct inode *);
2617 extern bool is_bad_inode(struct inode *);
2618 
2619 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2620 						loff_t lend);
2621 extern int __must_check file_check_and_advance_wb_err(struct file *file);
2622 extern int __must_check file_write_and_wait_range(struct file *file,
2623 						loff_t start, loff_t end);
2624 
file_write_and_wait(struct file * file)2625 static inline int file_write_and_wait(struct file *file)
2626 {
2627 	return file_write_and_wait_range(file, 0, LLONG_MAX);
2628 }
2629 
2630 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2631 			   int datasync);
2632 extern int vfs_fsync(struct file *file, int datasync);
2633 
2634 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2635 				unsigned int flags);
2636 
iocb_is_dsync(const struct kiocb * iocb)2637 static inline bool iocb_is_dsync(const struct kiocb *iocb)
2638 {
2639 	return (iocb->ki_flags & IOCB_DSYNC) ||
2640 		IS_SYNC(iocb->ki_filp->f_mapping->host);
2641 }
2642 
2643 /*
2644  * Sync the bytes written if this was a synchronous write.  Expect ki_pos
2645  * to already be updated for the write, and will return either the amount
2646  * of bytes passed in, or an error if syncing the file failed.
2647  */
generic_write_sync(struct kiocb * iocb,ssize_t count)2648 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2649 {
2650 	if (iocb_is_dsync(iocb)) {
2651 		int ret = vfs_fsync_range(iocb->ki_filp,
2652 				iocb->ki_pos - count, iocb->ki_pos - 1,
2653 				(iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2654 		if (ret)
2655 			return ret;
2656 	}
2657 
2658 	return count;
2659 }
2660 
2661 extern void emergency_sync(void);
2662 extern void emergency_remount(void);
2663 
2664 #ifdef CONFIG_BLOCK
2665 extern int bmap(struct inode *inode, sector_t *block);
2666 #else
bmap(struct inode * inode,sector_t * block)2667 static inline int bmap(struct inode *inode,  sector_t *block)
2668 {
2669 	return -EINVAL;
2670 }
2671 #endif
2672 
2673 int notify_change(struct mnt_idmap *, struct dentry *,
2674 		  struct iattr *, struct inode **);
2675 int inode_permission(struct mnt_idmap *, struct inode *, int);
2676 int generic_permission(struct mnt_idmap *, struct inode *, int);
file_permission(struct file * file,int mask)2677 static inline int file_permission(struct file *file, int mask)
2678 {
2679 	return inode_permission(file_mnt_idmap(file),
2680 				file_inode(file), mask);
2681 }
path_permission(const struct path * path,int mask)2682 static inline int path_permission(const struct path *path, int mask)
2683 {
2684 	return inode_permission(mnt_idmap(path->mnt),
2685 				d_inode(path->dentry), mask);
2686 }
2687 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir,
2688 		   struct inode *inode);
2689 
execute_ok(struct inode * inode)2690 static inline bool execute_ok(struct inode *inode)
2691 {
2692 	return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2693 }
2694 
inode_wrong_type(const struct inode * inode,umode_t mode)2695 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2696 {
2697 	return (inode->i_mode ^ mode) & S_IFMT;
2698 }
2699 
2700 /**
2701  * file_start_write - get write access to a superblock for regular file io
2702  * @file: the file we want to write to
2703  *
2704  * This is a variant of sb_start_write() which is a noop on non-regualr file.
2705  * Should be matched with a call to file_end_write().
2706  */
file_start_write(struct file * file)2707 static inline void file_start_write(struct file *file)
2708 {
2709 	if (!S_ISREG(file_inode(file)->i_mode))
2710 		return;
2711 	sb_start_write(file_inode(file)->i_sb);
2712 }
2713 
file_start_write_trylock(struct file * file)2714 static inline bool file_start_write_trylock(struct file *file)
2715 {
2716 	if (!S_ISREG(file_inode(file)->i_mode))
2717 		return true;
2718 	return sb_start_write_trylock(file_inode(file)->i_sb);
2719 }
2720 
2721 /**
2722  * file_end_write - drop write access to a superblock of a regular file
2723  * @file: the file we wrote to
2724  *
2725  * Should be matched with a call to file_start_write().
2726  */
file_end_write(struct file * file)2727 static inline void file_end_write(struct file *file)
2728 {
2729 	if (!S_ISREG(file_inode(file)->i_mode))
2730 		return;
2731 	sb_end_write(file_inode(file)->i_sb);
2732 }
2733 
2734 /**
2735  * kiocb_start_write - get write access to a superblock for async file io
2736  * @iocb: the io context we want to submit the write with
2737  *
2738  * This is a variant of sb_start_write() for async io submission.
2739  * Should be matched with a call to kiocb_end_write().
2740  */
kiocb_start_write(struct kiocb * iocb)2741 static inline void kiocb_start_write(struct kiocb *iocb)
2742 {
2743 	struct inode *inode = file_inode(iocb->ki_filp);
2744 
2745 	sb_start_write(inode->i_sb);
2746 	/*
2747 	 * Fool lockdep by telling it the lock got released so that it
2748 	 * doesn't complain about the held lock when we return to userspace.
2749 	 */
2750 	__sb_writers_release(inode->i_sb, SB_FREEZE_WRITE);
2751 }
2752 
2753 /**
2754  * kiocb_end_write - drop write access to a superblock after async file io
2755  * @iocb: the io context we sumbitted the write with
2756  *
2757  * Should be matched with a call to kiocb_start_write().
2758  */
kiocb_end_write(struct kiocb * iocb)2759 static inline void kiocb_end_write(struct kiocb *iocb)
2760 {
2761 	struct inode *inode = file_inode(iocb->ki_filp);
2762 
2763 	/*
2764 	 * Tell lockdep we inherited freeze protection from submission thread.
2765 	 */
2766 	__sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
2767 	sb_end_write(inode->i_sb);
2768 }
2769 
2770 /*
2771  * This is used for regular files where some users -- especially the
2772  * currently executed binary in a process, previously handled via
2773  * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2774  * read-write shared) accesses.
2775  *
2776  * get_write_access() gets write permission for a file.
2777  * put_write_access() releases this write permission.
2778  * deny_write_access() denies write access to a file.
2779  * allow_write_access() re-enables write access to a file.
2780  *
2781  * The i_writecount field of an inode can have the following values:
2782  * 0: no write access, no denied write access
2783  * < 0: (-i_writecount) users that denied write access to the file.
2784  * > 0: (i_writecount) users that have write access to the file.
2785  *
2786  * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2787  * except for the cases where we don't hold i_writecount yet. Then we need to
2788  * use {get,deny}_write_access() - these functions check the sign and refuse
2789  * to do the change if sign is wrong.
2790  */
get_write_access(struct inode * inode)2791 static inline int get_write_access(struct inode *inode)
2792 {
2793 	return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2794 }
deny_write_access(struct file * file)2795 static inline int deny_write_access(struct file *file)
2796 {
2797 	struct inode *inode = file_inode(file);
2798 	return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2799 }
put_write_access(struct inode * inode)2800 static inline void put_write_access(struct inode * inode)
2801 {
2802 	atomic_dec(&inode->i_writecount);
2803 }
allow_write_access(struct file * file)2804 static inline void allow_write_access(struct file *file)
2805 {
2806 	if (file)
2807 		atomic_inc(&file_inode(file)->i_writecount);
2808 }
inode_is_open_for_write(const struct inode * inode)2809 static inline bool inode_is_open_for_write(const struct inode *inode)
2810 {
2811 	return atomic_read(&inode->i_writecount) > 0;
2812 }
2813 
2814 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
i_readcount_dec(struct inode * inode)2815 static inline void i_readcount_dec(struct inode *inode)
2816 {
2817 	BUG_ON(atomic_dec_return(&inode->i_readcount) < 0);
2818 }
i_readcount_inc(struct inode * inode)2819 static inline void i_readcount_inc(struct inode *inode)
2820 {
2821 	atomic_inc(&inode->i_readcount);
2822 }
2823 #else
i_readcount_dec(struct inode * inode)2824 static inline void i_readcount_dec(struct inode *inode)
2825 {
2826 	return;
2827 }
i_readcount_inc(struct inode * inode)2828 static inline void i_readcount_inc(struct inode *inode)
2829 {
2830 	return;
2831 }
2832 #endif
2833 extern int do_pipe_flags(int *, int);
2834 
2835 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
2836 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
2837 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
2838 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
2839 extern struct file * open_exec(const char *);
2840 
2841 /* fs/dcache.c -- generic fs support functions */
2842 extern bool is_subdir(struct dentry *, struct dentry *);
2843 extern bool path_is_under(const struct path *, const struct path *);
2844 
2845 extern char *file_path(struct file *, char *, int);
2846 
2847 /**
2848  * is_dot_dotdot - returns true only if @name is "." or ".."
2849  * @name: file name to check
2850  * @len: length of file name, in bytes
2851  */
is_dot_dotdot(const char * name,size_t len)2852 static inline bool is_dot_dotdot(const char *name, size_t len)
2853 {
2854 	return len && unlikely(name[0] == '.') &&
2855 		(len == 1 || (len == 2 && name[1] == '.'));
2856 }
2857 
2858 #include <linux/err.h>
2859 
2860 /* needed for stackable file system support */
2861 extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
2862 
2863 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
2864 
2865 extern int inode_init_always(struct super_block *, struct inode *);
2866 extern void inode_init_once(struct inode *);
2867 extern void address_space_init_once(struct address_space *mapping);
2868 extern struct inode * igrab(struct inode *);
2869 extern ino_t iunique(struct super_block *, ino_t);
2870 extern int inode_needs_sync(struct inode *inode);
2871 extern int generic_delete_inode(struct inode *inode);
generic_drop_inode(struct inode * inode)2872 static inline int generic_drop_inode(struct inode *inode)
2873 {
2874 	return !inode->i_nlink || inode_unhashed(inode);
2875 }
2876 extern void d_mark_dontcache(struct inode *inode);
2877 
2878 extern struct inode *ilookup5_nowait(struct super_block *sb,
2879 		unsigned long hashval, int (*test)(struct inode *, void *),
2880 		void *data);
2881 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
2882 		int (*test)(struct inode *, void *), void *data);
2883 extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
2884 
2885 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
2886 		int (*test)(struct inode *, void *),
2887 		int (*set)(struct inode *, void *),
2888 		void *data);
2889 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
2890 extern struct inode * iget_locked(struct super_block *, unsigned long);
2891 extern struct inode *find_inode_nowait(struct super_block *,
2892 				       unsigned long,
2893 				       int (*match)(struct inode *,
2894 						    unsigned long, void *),
2895 				       void *data);
2896 extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
2897 				    int (*)(struct inode *, void *), void *);
2898 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
2899 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
2900 extern int insert_inode_locked(struct inode *);
2901 #ifdef CONFIG_DEBUG_LOCK_ALLOC
2902 extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
2903 #else
lockdep_annotate_inode_mutex_key(struct inode * inode)2904 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
2905 #endif
2906 extern void unlock_new_inode(struct inode *);
2907 extern void discard_new_inode(struct inode *);
2908 extern unsigned int get_next_ino(void);
2909 extern void evict_inodes(struct super_block *sb);
2910 void dump_mapping(const struct address_space *);
2911 
2912 /*
2913  * Userspace may rely on the inode number being non-zero. For example, glibc
2914  * simply ignores files with zero i_ino in unlink() and other places.
2915  *
2916  * As an additional complication, if userspace was compiled with
2917  * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
2918  * lower 32 bits, so we need to check that those aren't zero explicitly. With
2919  * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
2920  * better safe than sorry.
2921  */
is_zero_ino(ino_t ino)2922 static inline bool is_zero_ino(ino_t ino)
2923 {
2924 	return (u32)ino == 0;
2925 }
2926 
2927 extern void __iget(struct inode * inode);
2928 extern void iget_failed(struct inode *);
2929 extern void clear_inode(struct inode *);
2930 extern void __destroy_inode(struct inode *);
2931 extern struct inode *new_inode_pseudo(struct super_block *sb);
2932 extern struct inode *new_inode(struct super_block *sb);
2933 extern void free_inode_nonrcu(struct inode *inode);
2934 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
2935 extern int file_remove_privs(struct file *);
2936 int setattr_should_drop_sgid(struct mnt_idmap *idmap,
2937 			     const struct inode *inode);
2938 
2939 /*
2940  * This must be used for allocating filesystems specific inodes to set
2941  * up the inode reclaim context correctly.
2942  */
2943 static inline void *
alloc_inode_sb(struct super_block * sb,struct kmem_cache * cache,gfp_t gfp)2944 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp)
2945 {
2946 	return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp);
2947 }
2948 
2949 extern void __insert_inode_hash(struct inode *, unsigned long hashval);
insert_inode_hash(struct inode * inode)2950 static inline void insert_inode_hash(struct inode *inode)
2951 {
2952 	__insert_inode_hash(inode, inode->i_ino);
2953 }
2954 
2955 extern void __remove_inode_hash(struct inode *);
remove_inode_hash(struct inode * inode)2956 static inline void remove_inode_hash(struct inode *inode)
2957 {
2958 	if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
2959 		__remove_inode_hash(inode);
2960 }
2961 
2962 extern void inode_sb_list_add(struct inode *inode);
2963 extern void inode_add_lru(struct inode *inode);
2964 
2965 extern int sb_set_blocksize(struct super_block *, int);
2966 extern int sb_min_blocksize(struct super_block *, int);
2967 
2968 extern int generic_file_mmap(struct file *, struct vm_area_struct *);
2969 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
2970 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
2971 int generic_write_checks_count(struct kiocb *iocb, loff_t *count);
2972 extern int generic_write_check_limits(struct file *file, loff_t pos,
2973 		loff_t *count);
2974 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
2975 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
2976 		ssize_t already_read);
2977 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
2978 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
2979 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
2980 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
2981 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *);
2982 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
2983 		ssize_t direct_written, ssize_t buffered_written);
2984 
2985 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
2986 		rwf_t flags);
2987 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
2988 		rwf_t flags);
2989 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
2990 			   struct iov_iter *iter);
2991 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
2992 			    struct iov_iter *iter);
2993 
2994 /* fs/splice.c */
2995 ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
2996 			    struct pipe_inode_info *pipe,
2997 			    size_t len, unsigned int flags);
2998 ssize_t copy_splice_read(struct file *in, loff_t *ppos,
2999 			 struct pipe_inode_info *pipe,
3000 			 size_t len, unsigned int flags);
3001 extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
3002 		struct file *, loff_t *, size_t, unsigned int);
3003 extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
3004 		loff_t *opos, size_t len, unsigned int flags);
3005 
3006 
3007 extern void
3008 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
3009 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3010 #define no_llseek NULL
3011 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3012 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3013 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3014 		int whence, loff_t maxsize, loff_t eof);
3015 extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3016 		int whence, loff_t size);
3017 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3018 extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3019 int rw_verify_area(int, struct file *, const loff_t *, size_t);
3020 extern int generic_file_open(struct inode * inode, struct file * filp);
3021 extern int nonseekable_open(struct inode * inode, struct file * filp);
3022 extern int stream_open(struct inode * inode, struct file * filp);
3023 
3024 #ifdef CONFIG_BLOCK
3025 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
3026 			    loff_t file_offset);
3027 
3028 enum {
3029 	/* need locking between buffered and direct access */
3030 	DIO_LOCKING	= 0x01,
3031 
3032 	/* filesystem does not support filling holes */
3033 	DIO_SKIP_HOLES	= 0x02,
3034 };
3035 
3036 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
3037 			     struct block_device *bdev, struct iov_iter *iter,
3038 			     get_block_t get_block,
3039 			     dio_iodone_t end_io,
3040 			     int flags);
3041 
blockdev_direct_IO(struct kiocb * iocb,struct inode * inode,struct iov_iter * iter,get_block_t get_block)3042 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3043 					 struct inode *inode,
3044 					 struct iov_iter *iter,
3045 					 get_block_t get_block)
3046 {
3047 	return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
3048 			get_block, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
3049 }
3050 #endif
3051 
3052 void inode_dio_wait(struct inode *inode);
3053 
3054 /**
3055  * inode_dio_begin - signal start of a direct I/O requests
3056  * @inode: inode the direct I/O happens on
3057  *
3058  * This is called once we've finished processing a direct I/O request,
3059  * and is used to wake up callers waiting for direct I/O to be quiesced.
3060  */
inode_dio_begin(struct inode * inode)3061 static inline void inode_dio_begin(struct inode *inode)
3062 {
3063 	atomic_inc(&inode->i_dio_count);
3064 }
3065 
3066 /**
3067  * inode_dio_end - signal finish of a direct I/O requests
3068  * @inode: inode the direct I/O happens on
3069  *
3070  * This is called once we've finished processing a direct I/O request,
3071  * and is used to wake up callers waiting for direct I/O to be quiesced.
3072  */
inode_dio_end(struct inode * inode)3073 static inline void inode_dio_end(struct inode *inode)
3074 {
3075 	if (atomic_dec_and_test(&inode->i_dio_count))
3076 		wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
3077 }
3078 
3079 extern void inode_set_flags(struct inode *inode, unsigned int flags,
3080 			    unsigned int mask);
3081 
3082 extern const struct file_operations generic_ro_fops;
3083 
3084 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
3085 
3086 extern int readlink_copy(char __user *, int, const char *);
3087 extern int page_readlink(struct dentry *, char __user *, int);
3088 extern const char *page_get_link(struct dentry *, struct inode *,
3089 				 struct delayed_call *);
3090 extern void page_put_link(void *);
3091 extern int page_symlink(struct inode *inode, const char *symname, int len);
3092 extern const struct inode_operations page_symlink_inode_operations;
3093 extern void kfree_link(void *);
3094 void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *);
3095 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
3096 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
3097 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
3098 void __inode_add_bytes(struct inode *inode, loff_t bytes);
3099 void inode_add_bytes(struct inode *inode, loff_t bytes);
3100 void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3101 void inode_sub_bytes(struct inode *inode, loff_t bytes);
__inode_get_bytes(struct inode * inode)3102 static inline loff_t __inode_get_bytes(struct inode *inode)
3103 {
3104 	return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3105 }
3106 loff_t inode_get_bytes(struct inode *inode);
3107 void inode_set_bytes(struct inode *inode, loff_t bytes);
3108 const char *simple_get_link(struct dentry *, struct inode *,
3109 			    struct delayed_call *);
3110 extern const struct inode_operations simple_symlink_inode_operations;
3111 
3112 extern int iterate_dir(struct file *, struct dir_context *);
3113 
3114 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
3115 		int flags);
3116 int vfs_fstat(int fd, struct kstat *stat);
3117 
vfs_stat(const char __user * filename,struct kstat * stat)3118 static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3119 {
3120 	return vfs_fstatat(AT_FDCWD, filename, stat, 0);
3121 }
vfs_lstat(const char __user * name,struct kstat * stat)3122 static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3123 {
3124 	return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
3125 }
3126 
3127 extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3128 extern int vfs_readlink(struct dentry *, char __user *, int);
3129 
3130 extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3131 extern void put_filesystem(struct file_system_type *fs);
3132 extern struct file_system_type *get_fs_type(const char *name);
3133 extern struct super_block *get_active_super(struct block_device *bdev);
3134 extern void drop_super(struct super_block *sb);
3135 extern void drop_super_exclusive(struct super_block *sb);
3136 extern void iterate_supers(void (*)(struct super_block *, void *), void *);
3137 extern void iterate_supers_type(struct file_system_type *,
3138 			        void (*)(struct super_block *, void *), void *);
3139 
3140 extern int dcache_dir_open(struct inode *, struct file *);
3141 extern int dcache_dir_close(struct inode *, struct file *);
3142 extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3143 extern int dcache_readdir(struct file *, struct dir_context *);
3144 extern int simple_setattr(struct mnt_idmap *, struct dentry *,
3145 			  struct iattr *);
3146 extern int simple_getattr(struct mnt_idmap *, const struct path *,
3147 			  struct kstat *, u32, unsigned int);
3148 extern int simple_statfs(struct dentry *, struct kstatfs *);
3149 extern int simple_open(struct inode *inode, struct file *file);
3150 extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3151 extern int simple_unlink(struct inode *, struct dentry *);
3152 extern int simple_rmdir(struct inode *, struct dentry *);
3153 void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry,
3154 			     struct inode *new_dir, struct dentry *new_dentry);
3155 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
3156 				  struct inode *new_dir, struct dentry *new_dentry);
3157 extern int simple_rename(struct mnt_idmap *, struct inode *,
3158 			 struct dentry *, struct inode *, struct dentry *,
3159 			 unsigned int);
3160 extern void simple_recursive_removal(struct dentry *,
3161                               void (*callback)(struct dentry *));
3162 extern int noop_fsync(struct file *, loff_t, loff_t, int);
3163 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
3164 extern int simple_empty(struct dentry *);
3165 extern int simple_write_begin(struct file *file, struct address_space *mapping,
3166 			loff_t pos, unsigned len,
3167 			struct page **pagep, void **fsdata);
3168 extern const struct address_space_operations ram_aops;
3169 extern int always_delete_dentry(const struct dentry *);
3170 extern struct inode *alloc_anon_inode(struct super_block *);
3171 extern int simple_nosetlease(struct file *, int, struct file_lock **, void **);
3172 extern const struct dentry_operations simple_dentry_operations;
3173 
3174 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3175 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3176 extern const struct file_operations simple_dir_operations;
3177 extern const struct inode_operations simple_dir_inode_operations;
3178 extern void make_empty_dir_inode(struct inode *inode);
3179 extern bool is_empty_dir_inode(struct inode *inode);
3180 struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3181 struct dentry *d_alloc_name(struct dentry *, const char *);
3182 extern int simple_fill_super(struct super_block *, unsigned long,
3183 			     const struct tree_descr *);
3184 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3185 extern void simple_release_fs(struct vfsmount **mount, int *count);
3186 
3187 extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3188 			loff_t *ppos, const void *from, size_t available);
3189 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3190 		const void __user *from, size_t count);
3191 
3192 struct offset_ctx {
3193 	struct xarray		xa;
3194 	u32			next_offset;
3195 };
3196 
3197 void simple_offset_init(struct offset_ctx *octx);
3198 int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry);
3199 void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry);
3200 int simple_offset_rename_exchange(struct inode *old_dir,
3201 				  struct dentry *old_dentry,
3202 				  struct inode *new_dir,
3203 				  struct dentry *new_dentry);
3204 void simple_offset_destroy(struct offset_ctx *octx);
3205 
3206 extern const struct file_operations simple_offset_dir_operations;
3207 
3208 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3209 extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3210 
3211 extern int generic_check_addressable(unsigned, u64);
3212 
3213 extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry);
3214 
sb_has_encoding(const struct super_block * sb)3215 static inline bool sb_has_encoding(const struct super_block *sb)
3216 {
3217 #if IS_ENABLED(CONFIG_UNICODE)
3218 	return !!sb->s_encoding;
3219 #else
3220 	return false;
3221 #endif
3222 }
3223 
3224 int may_setattr(struct mnt_idmap *idmap, struct inode *inode,
3225 		unsigned int ia_valid);
3226 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *);
3227 extern int inode_newsize_ok(const struct inode *, loff_t offset);
3228 void setattr_copy(struct mnt_idmap *, struct inode *inode,
3229 		  const struct iattr *attr);
3230 
3231 extern int file_update_time(struct file *file);
3232 
vma_is_dax(const struct vm_area_struct * vma)3233 static inline bool vma_is_dax(const struct vm_area_struct *vma)
3234 {
3235 	return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3236 }
3237 
vma_is_fsdax(struct vm_area_struct * vma)3238 static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3239 {
3240 	struct inode *inode;
3241 
3242 	if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
3243 		return false;
3244 	if (!vma_is_dax(vma))
3245 		return false;
3246 	inode = file_inode(vma->vm_file);
3247 	if (S_ISCHR(inode->i_mode))
3248 		return false; /* device-dax */
3249 	return true;
3250 }
3251 
iocb_flags(struct file * file)3252 static inline int iocb_flags(struct file *file)
3253 {
3254 	int res = 0;
3255 	if (file->f_flags & O_APPEND)
3256 		res |= IOCB_APPEND;
3257 	if (file->f_flags & O_DIRECT)
3258 		res |= IOCB_DIRECT;
3259 	if (file->f_flags & O_DSYNC)
3260 		res |= IOCB_DSYNC;
3261 	if (file->f_flags & __O_SYNC)
3262 		res |= IOCB_SYNC;
3263 	return res;
3264 }
3265 
kiocb_set_rw_flags(struct kiocb * ki,rwf_t flags)3266 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
3267 {
3268 	int kiocb_flags = 0;
3269 
3270 	/* make sure there's no overlap between RWF and private IOCB flags */
3271 	BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3272 
3273 	if (!flags)
3274 		return 0;
3275 	if (unlikely(flags & ~RWF_SUPPORTED))
3276 		return -EOPNOTSUPP;
3277 
3278 	if (flags & RWF_NOWAIT) {
3279 		if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3280 			return -EOPNOTSUPP;
3281 		kiocb_flags |= IOCB_NOIO;
3282 	}
3283 	kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3284 	if (flags & RWF_SYNC)
3285 		kiocb_flags |= IOCB_DSYNC;
3286 
3287 	ki->ki_flags |= kiocb_flags;
3288 	return 0;
3289 }
3290 
parent_ino(struct dentry * dentry)3291 static inline ino_t parent_ino(struct dentry *dentry)
3292 {
3293 	ino_t res;
3294 
3295 	/*
3296 	 * Don't strictly need d_lock here? If the parent ino could change
3297 	 * then surely we'd have a deeper race in the caller?
3298 	 */
3299 	spin_lock(&dentry->d_lock);
3300 	res = dentry->d_parent->d_inode->i_ino;
3301 	spin_unlock(&dentry->d_lock);
3302 	return res;
3303 }
3304 
3305 /* Transaction based IO helpers */
3306 
3307 /*
3308  * An argresp is stored in an allocated page and holds the
3309  * size of the argument or response, along with its content
3310  */
3311 struct simple_transaction_argresp {
3312 	ssize_t size;
3313 	char data[];
3314 };
3315 
3316 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3317 
3318 char *simple_transaction_get(struct file *file, const char __user *buf,
3319 				size_t size);
3320 ssize_t simple_transaction_read(struct file *file, char __user *buf,
3321 				size_t size, loff_t *pos);
3322 int simple_transaction_release(struct inode *inode, struct file *file);
3323 
3324 void simple_transaction_set(struct file *file, size_t n);
3325 
3326 /*
3327  * simple attribute files
3328  *
3329  * These attributes behave similar to those in sysfs:
3330  *
3331  * Writing to an attribute immediately sets a value, an open file can be
3332  * written to multiple times.
3333  *
3334  * Reading from an attribute creates a buffer from the value that might get
3335  * read with multiple read calls. When the attribute has been read
3336  * completely, no further read calls are possible until the file is opened
3337  * again.
3338  *
3339  * All attributes contain a text representation of a numeric value
3340  * that are accessed with the get() and set() functions.
3341  */
3342 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed)	\
3343 static int __fops ## _open(struct inode *inode, struct file *file)	\
3344 {									\
3345 	__simple_attr_check_format(__fmt, 0ull);			\
3346 	return simple_attr_open(inode, file, __get, __set, __fmt);	\
3347 }									\
3348 static const struct file_operations __fops = {				\
3349 	.owner	 = THIS_MODULE,						\
3350 	.open	 = __fops ## _open,					\
3351 	.release = simple_attr_release,					\
3352 	.read	 = simple_attr_read,					\
3353 	.write	 = (__is_signed) ? simple_attr_write_signed : simple_attr_write,	\
3354 	.llseek	 = generic_file_llseek,					\
3355 }
3356 
3357 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt)		\
3358 	DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false)
3359 
3360 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt)	\
3361 	DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true)
3362 
3363 static inline __printf(1, 2)
__simple_attr_check_format(const char * fmt,...)3364 void __simple_attr_check_format(const char *fmt, ...)
3365 {
3366 	/* don't do anything, just let the compiler check the arguments; */
3367 }
3368 
3369 int simple_attr_open(struct inode *inode, struct file *file,
3370 		     int (*get)(void *, u64 *), int (*set)(void *, u64),
3371 		     const char *fmt);
3372 int simple_attr_release(struct inode *inode, struct file *file);
3373 ssize_t simple_attr_read(struct file *file, char __user *buf,
3374 			 size_t len, loff_t *ppos);
3375 ssize_t simple_attr_write(struct file *file, const char __user *buf,
3376 			  size_t len, loff_t *ppos);
3377 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
3378 				 size_t len, loff_t *ppos);
3379 
3380 struct ctl_table;
3381 int __init list_bdev_fs_names(char *buf, size_t size);
3382 
3383 #define __FMODE_EXEC		((__force int) FMODE_EXEC)
3384 #define __FMODE_NONOTIFY	((__force int) FMODE_NONOTIFY)
3385 
3386 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3387 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3388 					    (flag & __FMODE_NONOTIFY)))
3389 
is_sxid(umode_t mode)3390 static inline bool is_sxid(umode_t mode)
3391 {
3392 	return mode & (S_ISUID | S_ISGID);
3393 }
3394 
check_sticky(struct mnt_idmap * idmap,struct inode * dir,struct inode * inode)3395 static inline int check_sticky(struct mnt_idmap *idmap,
3396 			       struct inode *dir, struct inode *inode)
3397 {
3398 	if (!(dir->i_mode & S_ISVTX))
3399 		return 0;
3400 
3401 	return __check_sticky(idmap, dir, inode);
3402 }
3403 
inode_has_no_xattr(struct inode * inode)3404 static inline void inode_has_no_xattr(struct inode *inode)
3405 {
3406 	if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3407 		inode->i_flags |= S_NOSEC;
3408 }
3409 
is_root_inode(struct inode * inode)3410 static inline bool is_root_inode(struct inode *inode)
3411 {
3412 	return inode == inode->i_sb->s_root->d_inode;
3413 }
3414 
dir_emit(struct dir_context * ctx,const char * name,int namelen,u64 ino,unsigned type)3415 static inline bool dir_emit(struct dir_context *ctx,
3416 			    const char *name, int namelen,
3417 			    u64 ino, unsigned type)
3418 {
3419 	return ctx->actor(ctx, name, namelen, ctx->pos, ino, type);
3420 }
dir_emit_dot(struct file * file,struct dir_context * ctx)3421 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3422 {
3423 	return ctx->actor(ctx, ".", 1, ctx->pos,
3424 			  file->f_path.dentry->d_inode->i_ino, DT_DIR);
3425 }
dir_emit_dotdot(struct file * file,struct dir_context * ctx)3426 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3427 {
3428 	return ctx->actor(ctx, "..", 2, ctx->pos,
3429 			  parent_ino(file->f_path.dentry), DT_DIR);
3430 }
dir_emit_dots(struct file * file,struct dir_context * ctx)3431 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3432 {
3433 	if (ctx->pos == 0) {
3434 		if (!dir_emit_dot(file, ctx))
3435 			return false;
3436 		ctx->pos = 1;
3437 	}
3438 	if (ctx->pos == 1) {
3439 		if (!dir_emit_dotdot(file, ctx))
3440 			return false;
3441 		ctx->pos = 2;
3442 	}
3443 	return true;
3444 }
dir_relax(struct inode * inode)3445 static inline bool dir_relax(struct inode *inode)
3446 {
3447 	inode_unlock(inode);
3448 	inode_lock(inode);
3449 	return !IS_DEADDIR(inode);
3450 }
3451 
dir_relax_shared(struct inode * inode)3452 static inline bool dir_relax_shared(struct inode *inode)
3453 {
3454 	inode_unlock_shared(inode);
3455 	inode_lock_shared(inode);
3456 	return !IS_DEADDIR(inode);
3457 }
3458 
3459 extern bool path_noexec(const struct path *path);
3460 extern void inode_nohighmem(struct inode *inode);
3461 
3462 /* mm/fadvise.c */
3463 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3464 		       int advice);
3465 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3466 			   int advice);
3467 
3468 #endif /* _LINUX_FS_H */
3469