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
46#include <asm/byteorder.h>
47#include <uapi/linux/fs.h>
48
49struct backing_dev_info;
50struct bdi_writeback;
51struct bio;
52struct io_comp_batch;
53struct export_operations;
54struct fiemap_extent_info;
55struct hd_geometry;
56struct iovec;
57struct kiocb;
58struct kobject;
59struct pipe_inode_info;
60struct poll_table_struct;
61struct kstatfs;
62struct vm_area_struct;
63struct vfsmount;
64struct cred;
65struct swap_info_struct;
66struct seq_file;
67struct workqueue_struct;
68struct iov_iter;
69struct fscrypt_info;
70struct fscrypt_operations;
71struct fsverity_info;
72struct fsverity_operations;
73struct fs_context;
74struct fs_parameter_spec;
75struct fileattr;
76
77extern void __init inode_init(void);
78extern void __init inode_init_early(void);
79extern void __init files_init(void);
80extern void __init files_maxfiles_init(void);
81
82extern unsigned long get_max_files(void);
83extern unsigned int sysctl_nr_open;
84
85typedef __kernel_rwf_t rwf_t;
86
87struct buffer_head;
88typedef int (get_block_t)(struct inode *inode, sector_t iblock,
89			struct buffer_head *bh_result, int create);
90typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
91			ssize_t bytes, void *private);
92
93#define MAY_EXEC		0x00000001
94#define MAY_WRITE		0x00000002
95#define MAY_READ		0x00000004
96#define MAY_APPEND		0x00000008
97#define MAY_ACCESS		0x00000010
98#define MAY_OPEN		0x00000020
99#define MAY_CHDIR		0x00000040
100/* called from RCU mode, don't block */
101#define MAY_NOT_BLOCK		0x00000080
102
103/*
104 * flags in file.f_mode.  Note that FMODE_READ and FMODE_WRITE must correspond
105 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
106 */
107
108/* file is open for reading */
109#define FMODE_READ		((__force fmode_t)0x1)
110/* file is open for writing */
111#define FMODE_WRITE		((__force fmode_t)0x2)
112/* file is seekable */
113#define FMODE_LSEEK		((__force fmode_t)0x4)
114/* file can be accessed using pread */
115#define FMODE_PREAD		((__force fmode_t)0x8)
116/* file can be accessed using pwrite */
117#define FMODE_PWRITE		((__force fmode_t)0x10)
118/* File is opened for execution with sys_execve / sys_uselib */
119#define FMODE_EXEC		((__force fmode_t)0x20)
120/* File is opened with O_NDELAY (only set for block devices) */
121#define FMODE_NDELAY		((__force fmode_t)0x40)
122/* File is opened with O_EXCL (only set for block devices) */
123#define FMODE_EXCL		((__force fmode_t)0x80)
124/* File is opened using open(.., 3, ..) and is writeable only for ioctls
125   (specialy hack for floppy.c) */
126#define FMODE_WRITE_IOCTL	((__force fmode_t)0x100)
127/* 32bit hashes as llseek() offset (for directories) */
128#define FMODE_32BITHASH         ((__force fmode_t)0x200)
129/* 64bit hashes as llseek() offset (for directories) */
130#define FMODE_64BITHASH         ((__force fmode_t)0x400)
131
132/*
133 * Don't update ctime and mtime.
134 *
135 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
136 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
137 */
138#define FMODE_NOCMTIME		((__force fmode_t)0x800)
139
140/* Expect random access pattern */
141#define FMODE_RANDOM		((__force fmode_t)0x1000)
142
143/* File is huge (eg. /dev/mem): treat loff_t as unsigned */
144#define FMODE_UNSIGNED_OFFSET	((__force fmode_t)0x2000)
145
146/* File is opened with O_PATH; almost nothing can be done with it */
147#define FMODE_PATH		((__force fmode_t)0x4000)
148
149/* File needs atomic accesses to f_pos */
150#define FMODE_ATOMIC_POS	((__force fmode_t)0x8000)
151/* Write access to underlying fs */
152#define FMODE_WRITER		((__force fmode_t)0x10000)
153/* Has read method(s) */
154#define FMODE_CAN_READ          ((__force fmode_t)0x20000)
155/* Has write method(s) */
156#define FMODE_CAN_WRITE         ((__force fmode_t)0x40000)
157
158#define FMODE_OPENED		((__force fmode_t)0x80000)
159#define FMODE_CREATED		((__force fmode_t)0x100000)
160
161/* File is stream-like */
162#define FMODE_STREAM		((__force fmode_t)0x200000)
163
164/* File was opened by fanotify and shouldn't generate fanotify events */
165#define FMODE_NONOTIFY		((__force fmode_t)0x4000000)
166
167/* File is capable of returning -EAGAIN if I/O will block */
168#define FMODE_NOWAIT		((__force fmode_t)0x8000000)
169
170/* File represents mount that needs unmounting */
171#define FMODE_NEED_UNMOUNT	((__force fmode_t)0x10000000)
172
173/* File does not contribute to nr_files count */
174#define FMODE_NOACCOUNT		((__force fmode_t)0x20000000)
175
176/* File supports async buffered reads */
177#define FMODE_BUF_RASYNC	((__force fmode_t)0x40000000)
178
179/*
180 * Attribute flags.  These should be or-ed together to figure out what
181 * has been changed!
182 */
183#define ATTR_MODE	(1 << 0)
184#define ATTR_UID	(1 << 1)
185#define ATTR_GID	(1 << 2)
186#define ATTR_SIZE	(1 << 3)
187#define ATTR_ATIME	(1 << 4)
188#define ATTR_MTIME	(1 << 5)
189#define ATTR_CTIME	(1 << 6)
190#define ATTR_ATIME_SET	(1 << 7)
191#define ATTR_MTIME_SET	(1 << 8)
192#define ATTR_FORCE	(1 << 9) /* Not a change, but a change it */
193#define ATTR_KILL_SUID	(1 << 11)
194#define ATTR_KILL_SGID	(1 << 12)
195#define ATTR_FILE	(1 << 13)
196#define ATTR_KILL_PRIV	(1 << 14)
197#define ATTR_OPEN	(1 << 15) /* Truncating from open(O_TRUNC) */
198#define ATTR_TIMES_SET	(1 << 16)
199#define ATTR_TOUCH	(1 << 17)
200
201/*
202 * Whiteout is represented by a char device.  The following constants define the
203 * mode and device number to use.
204 */
205#define WHITEOUT_MODE 0
206#define WHITEOUT_DEV 0
207
208/*
209 * This is the Inode Attributes structure, used for notify_change().  It
210 * uses the above definitions as flags, to know which values have changed.
211 * Also, in this manner, a Filesystem can look at only the values it cares
212 * about.  Basically, these are the attributes that the VFS layer can
213 * request to change from the FS layer.
214 *
215 * Derek Atkins <warlord@MIT.EDU> 94-10-20
216 */
217struct iattr {
218	unsigned int	ia_valid;
219	umode_t		ia_mode;
220	kuid_t		ia_uid;
221	kgid_t		ia_gid;
222	loff_t		ia_size;
223	struct timespec64 ia_atime;
224	struct timespec64 ia_mtime;
225	struct timespec64 ia_ctime;
226
227	/*
228	 * Not an attribute, but an auxiliary info for filesystems wanting to
229	 * implement an ftruncate() like method.  NOTE: filesystem should
230	 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
231	 */
232	struct file	*ia_file;
233};
234
235/*
236 * Includes for diskquotas.
237 */
238#include <linux/quota.h>
239
240/*
241 * Maximum number of layers of fs stack.  Needs to be limited to
242 * prevent kernel stack overflow
243 */
244#define FILESYSTEM_MAX_STACK_DEPTH 2
245
246/**
247 * enum positive_aop_returns - aop return codes with specific semantics
248 *
249 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
250 * 			    completed, that the page is still locked, and
251 * 			    should be considered active.  The VM uses this hint
252 * 			    to return the page to the active list -- it won't
253 * 			    be a candidate for writeback again in the near
254 * 			    future.  Other callers must be careful to unlock
255 * 			    the page if they get this return.  Returned by
256 * 			    writepage();
257 *
258 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
259 *  			unlocked it and the page might have been truncated.
260 *  			The caller should back up to acquiring a new page and
261 *  			trying again.  The aop will be taking reasonable
262 *  			precautions not to livelock.  If the caller held a page
263 *  			reference, it should drop it before retrying.  Returned
264 *  			by readpage().
265 *
266 * address_space_operation functions return these large constants to indicate
267 * special semantics to the caller.  These are much larger than the bytes in a
268 * page to allow for functions that return the number of bytes operated on in a
269 * given page.
270 */
271
272enum positive_aop_returns {
273	AOP_WRITEPAGE_ACTIVATE	= 0x80000,
274	AOP_TRUNCATED_PAGE	= 0x80001,
275};
276
277#define AOP_FLAG_CONT_EXPAND		0x0001 /* called from cont_expand */
278#define AOP_FLAG_NOFS			0x0002 /* used by filesystem to direct
279						* helper code (eg buffer layer)
280						* to clear GFP_FS from alloc */
281
282/*
283 * oh the beauties of C type declarations.
284 */
285struct page;
286struct address_space;
287struct writeback_control;
288struct readahead_control;
289
290/*
291 * Write life time hint values.
292 * Stored in struct inode as u8.
293 */
294enum rw_hint {
295	WRITE_LIFE_NOT_SET	= 0,
296	WRITE_LIFE_NONE		= RWH_WRITE_LIFE_NONE,
297	WRITE_LIFE_SHORT	= RWH_WRITE_LIFE_SHORT,
298	WRITE_LIFE_MEDIUM	= RWH_WRITE_LIFE_MEDIUM,
299	WRITE_LIFE_LONG		= RWH_WRITE_LIFE_LONG,
300	WRITE_LIFE_EXTREME	= RWH_WRITE_LIFE_EXTREME,
301};
302
303/* Match RWF_* bits to IOCB bits */
304#define IOCB_HIPRI		(__force int) RWF_HIPRI
305#define IOCB_DSYNC		(__force int) RWF_DSYNC
306#define IOCB_SYNC		(__force int) RWF_SYNC
307#define IOCB_NOWAIT		(__force int) RWF_NOWAIT
308#define IOCB_APPEND		(__force int) RWF_APPEND
309
310/* non-RWF related bits - start at 16 */
311#define IOCB_EVENTFD		(1 << 16)
312#define IOCB_DIRECT		(1 << 17)
313#define IOCB_WRITE		(1 << 18)
314/* iocb->ki_waitq is valid */
315#define IOCB_WAITQ		(1 << 19)
316#define IOCB_NOIO		(1 << 20)
317/* can use bio alloc cache */
318#define IOCB_ALLOC_CACHE	(1 << 21)
319
320struct kiocb {
321	struct file		*ki_filp;
322
323	/* The 'ki_filp' pointer is shared in a union for aio */
324	randomized_struct_fields_start
325
326	loff_t			ki_pos;
327	void (*ki_complete)(struct kiocb *iocb, long ret);
328	void			*private;
329	int			ki_flags;
330	u16			ki_hint;
331	u16			ki_ioprio; /* See linux/ioprio.h */
332	struct wait_page_queue	*ki_waitq; /* for async buffered IO */
333	randomized_struct_fields_end
334};
335
336static inline bool is_sync_kiocb(struct kiocb *kiocb)
337{
338	return kiocb->ki_complete == NULL;
339}
340
341/*
342 * "descriptor" for what we're up to with a read.
343 * This allows us to use the same read code yet
344 * have multiple different users of the data that
345 * we read from a file.
346 *
347 * The simplest case just copies the data to user
348 * mode.
349 */
350typedef struct {
351	size_t written;
352	size_t count;
353	union {
354		char __user *buf;
355		void *data;
356	} arg;
357	int error;
358} read_descriptor_t;
359
360typedef int (*read_actor_t)(read_descriptor_t *, struct page *,
361		unsigned long, unsigned long);
362
363struct address_space_operations {
364	int (*writepage)(struct page *page, struct writeback_control *wbc);
365	int (*readpage)(struct file *, struct page *);
366
367	/* Write back some dirty pages from this mapping. */
368	int (*writepages)(struct address_space *, struct writeback_control *);
369
370	/* Set a page dirty.  Return true if this dirtied it */
371	int (*set_page_dirty)(struct page *page);
372
373	/*
374	 * Reads in the requested pages. Unlike ->readpage(), this is
375	 * PURELY used for read-ahead!.
376	 */
377	int (*readpages)(struct file *filp, struct address_space *mapping,
378			struct list_head *pages, unsigned nr_pages);
379	void (*readahead)(struct readahead_control *);
380
381	int (*write_begin)(struct file *, struct address_space *mapping,
382				loff_t pos, unsigned len, unsigned flags,
383				struct page **pagep, void **fsdata);
384	int (*write_end)(struct file *, struct address_space *mapping,
385				loff_t pos, unsigned len, unsigned copied,
386				struct page *page, void *fsdata);
387
388	/* Unfortunately this kludge is needed for FIBMAP. Don't use it */
389	sector_t (*bmap)(struct address_space *, sector_t);
390	void (*invalidatepage) (struct page *, unsigned int, unsigned int);
391	int (*releasepage) (struct page *, gfp_t);
392	void (*freepage)(struct page *);
393	ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
394	/*
395	 * migrate the contents of a page to the specified target. If
396	 * migrate_mode is MIGRATE_ASYNC, it must not block.
397	 */
398	int (*migratepage) (struct address_space *,
399			struct page *, struct page *, enum migrate_mode);
400	bool (*isolate_page)(struct page *, isolate_mode_t);
401	void (*putback_page)(struct page *);
402	int (*launder_page) (struct page *);
403	int (*is_partially_uptodate) (struct page *, unsigned long,
404					unsigned long);
405	void (*is_dirty_writeback) (struct page *, bool *, bool *);
406	int (*error_remove_page)(struct address_space *, struct page *);
407
408	/* swapfile support */
409	int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
410				sector_t *span);
411	void (*swap_deactivate)(struct file *file);
412};
413
414extern const struct address_space_operations empty_aops;
415
416/*
417 * pagecache_write_begin/pagecache_write_end must be used by general code
418 * to write into the pagecache.
419 */
420int pagecache_write_begin(struct file *, struct address_space *mapping,
421				loff_t pos, unsigned len, unsigned flags,
422				struct page **pagep, void **fsdata);
423
424int pagecache_write_end(struct file *, struct address_space *mapping,
425				loff_t pos, unsigned len, unsigned copied,
426				struct page *page, void *fsdata);
427
428/**
429 * struct address_space - Contents of a cacheable, mappable object.
430 * @host: Owner, either the inode or the block_device.
431 * @i_pages: Cached pages.
432 * @invalidate_lock: Guards coherency between page cache contents and
433 *   file offset->disk block mappings in the filesystem during invalidates.
434 *   It is also used to block modification of page cache contents through
435 *   memory mappings.
436 * @gfp_mask: Memory allocation flags to use for allocating pages.
437 * @i_mmap_writable: Number of VM_SHARED mappings.
438 * @nr_thps: Number of THPs in the pagecache (non-shmem only).
439 * @i_mmap: Tree of private and shared mappings.
440 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
441 * @nrpages: Number of page entries, protected by the i_pages lock.
442 * @writeback_index: Writeback starts here.
443 * @a_ops: Methods.
444 * @flags: Error bits and flags (AS_*).
445 * @wb_err: The most recent error which has occurred.
446 * @private_lock: For use by the owner of the address_space.
447 * @private_list: For use by the owner of the address_space.
448 * @private_data: For use by the owner of the address_space.
449 */
450struct address_space {
451	struct inode		*host;
452	struct xarray		i_pages;
453	struct rw_semaphore	invalidate_lock;
454	gfp_t			gfp_mask;
455	atomic_t		i_mmap_writable;
456#ifdef CONFIG_READ_ONLY_THP_FOR_FS
457	/* number of thp, only for non-shmem files */
458	atomic_t		nr_thps;
459#endif
460	struct rb_root_cached	i_mmap;
461	struct rw_semaphore	i_mmap_rwsem;
462	unsigned long		nrpages;
463	pgoff_t			writeback_index;
464	const struct address_space_operations *a_ops;
465	unsigned long		flags;
466	errseq_t		wb_err;
467	spinlock_t		private_lock;
468	struct list_head	private_list;
469	void			*private_data;
470} __attribute__((aligned(sizeof(long)))) __randomize_layout;
471	/*
472	 * On most architectures that alignment is already the case; but
473	 * must be enforced here for CRIS, to let the least significant bit
474	 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
475	 */
476
477/* XArray tags, for tagging dirty and writeback pages in the pagecache. */
478#define PAGECACHE_TAG_DIRTY	XA_MARK_0
479#define PAGECACHE_TAG_WRITEBACK	XA_MARK_1
480#define PAGECACHE_TAG_TOWRITE	XA_MARK_2
481
482/*
483 * Returns true if any of the pages in the mapping are marked with the tag.
484 */
485static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
486{
487	return xa_marked(&mapping->i_pages, tag);
488}
489
490static inline void i_mmap_lock_write(struct address_space *mapping)
491{
492	down_write(&mapping->i_mmap_rwsem);
493}
494
495static inline int i_mmap_trylock_write(struct address_space *mapping)
496{
497	return down_write_trylock(&mapping->i_mmap_rwsem);
498}
499
500static inline void i_mmap_unlock_write(struct address_space *mapping)
501{
502	up_write(&mapping->i_mmap_rwsem);
503}
504
505static inline void i_mmap_lock_read(struct address_space *mapping)
506{
507	down_read(&mapping->i_mmap_rwsem);
508}
509
510static inline void i_mmap_unlock_read(struct address_space *mapping)
511{
512	up_read(&mapping->i_mmap_rwsem);
513}
514
515static inline void i_mmap_assert_locked(struct address_space *mapping)
516{
517	lockdep_assert_held(&mapping->i_mmap_rwsem);
518}
519
520static inline void i_mmap_assert_write_locked(struct address_space *mapping)
521{
522	lockdep_assert_held_write(&mapping->i_mmap_rwsem);
523}
524
525/*
526 * Might pages of this file be mapped into userspace?
527 */
528static inline int mapping_mapped(struct address_space *mapping)
529{
530	return	!RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
531}
532
533/*
534 * Might pages of this file have been modified in userspace?
535 * Note that i_mmap_writable counts all VM_SHARED vmas: do_mmap
536 * marks vma as VM_SHARED if it is shared, and the file was opened for
537 * writing i.e. vma may be mprotected writable even if now readonly.
538 *
539 * If i_mmap_writable is negative, no new writable mappings are allowed. You
540 * can only deny writable mappings, if none exists right now.
541 */
542static inline int mapping_writably_mapped(struct address_space *mapping)
543{
544	return atomic_read(&mapping->i_mmap_writable) > 0;
545}
546
547static inline int mapping_map_writable(struct address_space *mapping)
548{
549	return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
550		0 : -EPERM;
551}
552
553static inline void mapping_unmap_writable(struct address_space *mapping)
554{
555	atomic_dec(&mapping->i_mmap_writable);
556}
557
558static inline int mapping_deny_writable(struct address_space *mapping)
559{
560	return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
561		0 : -EBUSY;
562}
563
564static inline void mapping_allow_writable(struct address_space *mapping)
565{
566	atomic_inc(&mapping->i_mmap_writable);
567}
568
569/*
570 * Use sequence counter to get consistent i_size on 32-bit processors.
571 */
572#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
573#include <linux/seqlock.h>
574#define __NEED_I_SIZE_ORDERED
575#define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
576#else
577#define i_size_ordered_init(inode) do { } while (0)
578#endif
579
580struct posix_acl;
581#define ACL_NOT_CACHED ((void *)(-1))
582/*
583 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
584 * cache the ACL.  This also means that ->get_acl() can be called in RCU mode
585 * with the LOOKUP_RCU flag.
586 */
587#define ACL_DONT_CACHE ((void *)(-3))
588
589static inline struct posix_acl *
590uncached_acl_sentinel(struct task_struct *task)
591{
592	return (void *)task + 1;
593}
594
595static inline bool
596is_uncached_acl(struct posix_acl *acl)
597{
598	return (long)acl & 1;
599}
600
601#define IOP_FASTPERM	0x0001
602#define IOP_LOOKUP	0x0002
603#define IOP_NOFOLLOW	0x0004
604#define IOP_XATTR	0x0008
605#define IOP_DEFAULT_READLINK	0x0010
606
607struct fsnotify_mark_connector;
608
609/*
610 * Keep mostly read-only and often accessed (especially for
611 * the RCU path lookup and 'stat' data) fields at the beginning
612 * of the 'struct inode'
613 */
614struct inode {
615	umode_t			i_mode;
616	unsigned short		i_opflags;
617	kuid_t			i_uid;
618	kgid_t			i_gid;
619	unsigned int		i_flags;
620
621#ifdef CONFIG_FS_POSIX_ACL
622	struct posix_acl	*i_acl;
623	struct posix_acl	*i_default_acl;
624#endif
625
626	const struct inode_operations	*i_op;
627	struct super_block	*i_sb;
628	struct address_space	*i_mapping;
629
630#ifdef CONFIG_SECURITY
631	void			*i_security;
632#endif
633
634	/* Stat data, not accessed from path walking */
635	unsigned long		i_ino;
636	/*
637	 * Filesystems may only read i_nlink directly.  They shall use the
638	 * following functions for modification:
639	 *
640	 *    (set|clear|inc|drop)_nlink
641	 *    inode_(inc|dec)_link_count
642	 */
643	union {
644		const unsigned int i_nlink;
645		unsigned int __i_nlink;
646	};
647	dev_t			i_rdev;
648	loff_t			i_size;
649	struct timespec64	i_atime;
650	struct timespec64	i_mtime;
651	struct timespec64	i_ctime;
652	spinlock_t		i_lock;	/* i_blocks, i_bytes, maybe i_size */
653	unsigned short          i_bytes;
654	u8			i_blkbits;
655	u8			i_write_hint;
656	blkcnt_t		i_blocks;
657
658#ifdef __NEED_I_SIZE_ORDERED
659	seqcount_t		i_size_seqcount;
660#endif
661
662	/* Misc */
663	unsigned long		i_state;
664	struct rw_semaphore	i_rwsem;
665
666	unsigned long		dirtied_when;	/* jiffies of first dirtying */
667	unsigned long		dirtied_time_when;
668
669	struct hlist_node	i_hash;
670	struct list_head	i_io_list;	/* backing dev IO list */
671#ifdef CONFIG_CGROUP_WRITEBACK
672	struct bdi_writeback	*i_wb;		/* the associated cgroup wb */
673
674	/* foreign inode detection, see wbc_detach_inode() */
675	int			i_wb_frn_winner;
676	u16			i_wb_frn_avg_time;
677	u16			i_wb_frn_history;
678#endif
679	struct list_head	i_lru;		/* inode LRU list */
680	struct list_head	i_sb_list;
681	struct list_head	i_wb_list;	/* backing dev writeback list */
682	union {
683		struct hlist_head	i_dentry;
684		struct rcu_head		i_rcu;
685	};
686	atomic64_t		i_version;
687	atomic64_t		i_sequence; /* see futex */
688	atomic_t		i_count;
689	atomic_t		i_dio_count;
690	atomic_t		i_writecount;
691#if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
692	atomic_t		i_readcount; /* struct files open RO */
693#endif
694	union {
695		const struct file_operations	*i_fop;	/* former ->i_op->default_file_ops */
696		void (*free_inode)(struct inode *);
697	};
698	struct file_lock_context	*i_flctx;
699	struct address_space	i_data;
700	struct list_head	i_devices;
701	union {
702		struct pipe_inode_info	*i_pipe;
703		struct cdev		*i_cdev;
704		char			*i_link;
705		unsigned		i_dir_seq;
706	};
707
708	__u32			i_generation;
709
710#ifdef CONFIG_FSNOTIFY
711	__u32			i_fsnotify_mask; /* all events this inode cares about */
712	struct fsnotify_mark_connector __rcu	*i_fsnotify_marks;
713#endif
714
715#ifdef CONFIG_FS_ENCRYPTION
716	struct fscrypt_info	*i_crypt_info;
717#endif
718
719#ifdef CONFIG_FS_VERITY
720	struct fsverity_info	*i_verity_info;
721#endif
722
723	void			*i_private; /* fs or device private pointer */
724} __randomize_layout;
725
726struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
727
728static inline unsigned int i_blocksize(const struct inode *node)
729{
730	return (1 << node->i_blkbits);
731}
732
733static inline int inode_unhashed(struct inode *inode)
734{
735	return hlist_unhashed(&inode->i_hash);
736}
737
738/*
739 * __mark_inode_dirty expects inodes to be hashed.  Since we don't
740 * want special inodes in the fileset inode space, we make them
741 * appear hashed, but do not put on any lists.  hlist_del()
742 * will work fine and require no locking.
743 */
744static inline void inode_fake_hash(struct inode *inode)
745{
746	hlist_add_fake(&inode->i_hash);
747}
748
749/*
750 * inode->i_mutex nesting subclasses for the lock validator:
751 *
752 * 0: the object of the current VFS operation
753 * 1: parent
754 * 2: child/target
755 * 3: xattr
756 * 4: second non-directory
757 * 5: second parent (when locking independent directories in rename)
758 *
759 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
760 * non-directories at once.
761 *
762 * The locking order between these classes is
763 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
764 */
765enum inode_i_mutex_lock_class
766{
767	I_MUTEX_NORMAL,
768	I_MUTEX_PARENT,
769	I_MUTEX_CHILD,
770	I_MUTEX_XATTR,
771	I_MUTEX_NONDIR2,
772	I_MUTEX_PARENT2,
773};
774
775static inline void inode_lock(struct inode *inode)
776{
777	down_write(&inode->i_rwsem);
778}
779
780static inline void inode_unlock(struct inode *inode)
781{
782	up_write(&inode->i_rwsem);
783}
784
785static inline void inode_lock_shared(struct inode *inode)
786{
787	down_read(&inode->i_rwsem);
788}
789
790static inline void inode_unlock_shared(struct inode *inode)
791{
792	up_read(&inode->i_rwsem);
793}
794
795static inline int inode_trylock(struct inode *inode)
796{
797	return down_write_trylock(&inode->i_rwsem);
798}
799
800static inline int inode_trylock_shared(struct inode *inode)
801{
802	return down_read_trylock(&inode->i_rwsem);
803}
804
805static inline int inode_is_locked(struct inode *inode)
806{
807	return rwsem_is_locked(&inode->i_rwsem);
808}
809
810static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
811{
812	down_write_nested(&inode->i_rwsem, subclass);
813}
814
815static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
816{
817	down_read_nested(&inode->i_rwsem, subclass);
818}
819
820static inline void filemap_invalidate_lock(struct address_space *mapping)
821{
822	down_write(&mapping->invalidate_lock);
823}
824
825static inline void filemap_invalidate_unlock(struct address_space *mapping)
826{
827	up_write(&mapping->invalidate_lock);
828}
829
830static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
831{
832	down_read(&mapping->invalidate_lock);
833}
834
835static inline int filemap_invalidate_trylock_shared(
836					struct address_space *mapping)
837{
838	return down_read_trylock(&mapping->invalidate_lock);
839}
840
841static inline void filemap_invalidate_unlock_shared(
842					struct address_space *mapping)
843{
844	up_read(&mapping->invalidate_lock);
845}
846
847void lock_two_nondirectories(struct inode *, struct inode*);
848void unlock_two_nondirectories(struct inode *, struct inode*);
849
850void filemap_invalidate_lock_two(struct address_space *mapping1,
851				 struct address_space *mapping2);
852void filemap_invalidate_unlock_two(struct address_space *mapping1,
853				   struct address_space *mapping2);
854
855
856/*
857 * NOTE: in a 32bit arch with a preemptable kernel and
858 * an UP compile the i_size_read/write must be atomic
859 * with respect to the local cpu (unlike with preempt disabled),
860 * but they don't need to be atomic with respect to other cpus like in
861 * true SMP (so they need either to either locally disable irq around
862 * the read or for example on x86 they can be still implemented as a
863 * cmpxchg8b without the need of the lock prefix). For SMP compiles
864 * and 64bit archs it makes no difference if preempt is enabled or not.
865 */
866static inline loff_t i_size_read(const struct inode *inode)
867{
868#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
869	loff_t i_size;
870	unsigned int seq;
871
872	do {
873		seq = read_seqcount_begin(&inode->i_size_seqcount);
874		i_size = inode->i_size;
875	} while (read_seqcount_retry(&inode->i_size_seqcount, seq));
876	return i_size;
877#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
878	loff_t i_size;
879
880	preempt_disable();
881	i_size = inode->i_size;
882	preempt_enable();
883	return i_size;
884#else
885	return inode->i_size;
886#endif
887}
888
889/*
890 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
891 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
892 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
893 */
894static inline void i_size_write(struct inode *inode, loff_t i_size)
895{
896#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
897	preempt_disable();
898	write_seqcount_begin(&inode->i_size_seqcount);
899	inode->i_size = i_size;
900	write_seqcount_end(&inode->i_size_seqcount);
901	preempt_enable();
902#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
903	preempt_disable();
904	inode->i_size = i_size;
905	preempt_enable();
906#else
907	inode->i_size = i_size;
908#endif
909}
910
911static inline unsigned iminor(const struct inode *inode)
912{
913	return MINOR(inode->i_rdev);
914}
915
916static inline unsigned imajor(const struct inode *inode)
917{
918	return MAJOR(inode->i_rdev);
919}
920
921struct fown_struct {
922	rwlock_t lock;          /* protects pid, uid, euid fields */
923	struct pid *pid;	/* pid or -pgrp where SIGIO should be sent */
924	enum pid_type pid_type;	/* Kind of process group SIGIO should be sent to */
925	kuid_t uid, euid;	/* uid/euid of process setting the owner */
926	int signum;		/* posix.1b rt signal to be delivered on IO */
927};
928
929/**
930 * struct file_ra_state - Track a file's readahead state.
931 * @start: Where the most recent readahead started.
932 * @size: Number of pages read in the most recent readahead.
933 * @async_size: Start next readahead when this many pages are left.
934 * @ra_pages: Maximum size of a readahead request.
935 * @mmap_miss: How many mmap accesses missed in the page cache.
936 * @prev_pos: The last byte in the most recent read request.
937 */
938struct file_ra_state {
939	pgoff_t start;
940	unsigned int size;
941	unsigned int async_size;
942	unsigned int ra_pages;
943	unsigned int mmap_miss;
944	loff_t prev_pos;
945};
946
947/*
948 * Check if @index falls in the readahead windows.
949 */
950static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
951{
952	return (index >= ra->start &&
953		index <  ra->start + ra->size);
954}
955
956struct file {
957	union {
958		struct llist_node	fu_llist;
959		struct rcu_head 	fu_rcuhead;
960	} f_u;
961	struct path		f_path;
962	struct inode		*f_inode;	/* cached value */
963	const struct file_operations	*f_op;
964
965	/*
966	 * Protects f_ep, f_flags.
967	 * Must not be taken from IRQ context.
968	 */
969	spinlock_t		f_lock;
970	enum rw_hint		f_write_hint;
971	atomic_long_t		f_count;
972	unsigned int 		f_flags;
973	fmode_t			f_mode;
974	struct mutex		f_pos_lock;
975	loff_t			f_pos;
976	struct fown_struct	f_owner;
977	const struct cred	*f_cred;
978	struct file_ra_state	f_ra;
979
980	u64			f_version;
981#ifdef CONFIG_SECURITY
982	void			*f_security;
983#endif
984	/* needed for tty driver, and maybe others */
985	void			*private_data;
986
987#ifdef CONFIG_EPOLL
988	/* Used by fs/eventpoll.c to link all the hooks to this file */
989	struct hlist_head	*f_ep;
990#endif /* #ifdef CONFIG_EPOLL */
991	struct address_space	*f_mapping;
992	errseq_t		f_wb_err;
993	errseq_t		f_sb_err; /* for syncfs */
994} __randomize_layout
995  __attribute__((aligned(4)));	/* lest something weird decides that 2 is OK */
996
997struct file_handle {
998	__u32 handle_bytes;
999	int handle_type;
1000	/* file identifier */
1001	unsigned char f_handle[];
1002};
1003
1004static inline struct file *get_file(struct file *f)
1005{
1006	atomic_long_inc(&f->f_count);
1007	return f;
1008}
1009#define get_file_rcu_many(x, cnt)	\
1010	atomic_long_add_unless(&(x)->f_count, (cnt), 0)
1011#define get_file_rcu(x) get_file_rcu_many((x), 1)
1012#define file_count(x)	atomic_long_read(&(x)->f_count)
1013
1014#define	MAX_NON_LFS	((1UL<<31) - 1)
1015
1016/* Page cache limit. The filesystems should put that into their s_maxbytes
1017   limits, otherwise bad things can happen in VM. */
1018#if BITS_PER_LONG==32
1019#define MAX_LFS_FILESIZE	((loff_t)ULONG_MAX << PAGE_SHIFT)
1020#elif BITS_PER_LONG==64
1021#define MAX_LFS_FILESIZE 	((loff_t)LLONG_MAX)
1022#endif
1023
1024#define FL_POSIX	1
1025#define FL_FLOCK	2
1026#define FL_DELEG	4	/* NFSv4 delegation */
1027#define FL_ACCESS	8	/* not trying to lock, just looking */
1028#define FL_EXISTS	16	/* when unlocking, test for existence */
1029#define FL_LEASE	32	/* lease held on this file */
1030#define FL_CLOSE	64	/* unlock on close */
1031#define FL_SLEEP	128	/* A blocking lock */
1032#define FL_DOWNGRADE_PENDING	256 /* Lease is being downgraded */
1033#define FL_UNLOCK_PENDING	512 /* Lease is being broken */
1034#define FL_OFDLCK	1024	/* lock is "owned" by struct file */
1035#define FL_LAYOUT	2048	/* outstanding pNFS layout */
1036#define FL_RECLAIM	4096	/* reclaiming from a reboot server */
1037
1038#define FL_CLOSE_POSIX (FL_POSIX | FL_CLOSE)
1039
1040/*
1041 * Special return value from posix_lock_file() and vfs_lock_file() for
1042 * asynchronous locking.
1043 */
1044#define FILE_LOCK_DEFERRED 1
1045
1046/* legacy typedef, should eventually be removed */
1047typedef void *fl_owner_t;
1048
1049struct file_lock;
1050
1051struct file_lock_operations {
1052	void (*fl_copy_lock)(struct file_lock *, struct file_lock *);
1053	void (*fl_release_private)(struct file_lock *);
1054};
1055
1056struct lock_manager_operations {
1057	fl_owner_t (*lm_get_owner)(fl_owner_t);
1058	void (*lm_put_owner)(fl_owner_t);
1059	void (*lm_notify)(struct file_lock *);	/* unblock callback */
1060	int (*lm_grant)(struct file_lock *, int);
1061	bool (*lm_break)(struct file_lock *);
1062	int (*lm_change)(struct file_lock *, int, struct list_head *);
1063	void (*lm_setup)(struct file_lock *, void **);
1064	bool (*lm_breaker_owns_lease)(struct file_lock *);
1065};
1066
1067struct lock_manager {
1068	struct list_head list;
1069	/*
1070	 * NFSv4 and up also want opens blocked during the grace period;
1071	 * NLM doesn't care:
1072	 */
1073	bool block_opens;
1074};
1075
1076struct net;
1077void locks_start_grace(struct net *, struct lock_manager *);
1078void locks_end_grace(struct lock_manager *);
1079bool locks_in_grace(struct net *);
1080bool opens_in_grace(struct net *);
1081
1082/* that will die - we need it for nfs_lock_info */
1083#include <linux/nfs_fs_i.h>
1084
1085/*
1086 * struct file_lock represents a generic "file lock". It's used to represent
1087 * POSIX byte range locks, BSD (flock) locks, and leases. It's important to
1088 * note that the same struct is used to represent both a request for a lock and
1089 * the lock itself, but the same object is never used for both.
1090 *
1091 * FIXME: should we create a separate "struct lock_request" to help distinguish
1092 * these two uses?
1093 *
1094 * The varous i_flctx lists are ordered by:
1095 *
1096 * 1) lock owner
1097 * 2) lock range start
1098 * 3) lock range end
1099 *
1100 * Obviously, the last two criteria only matter for POSIX locks.
1101 */
1102struct file_lock {
1103	struct file_lock *fl_blocker;	/* The lock, that is blocking us */
1104	struct list_head fl_list;	/* link into file_lock_context */
1105	struct hlist_node fl_link;	/* node in global lists */
1106	struct list_head fl_blocked_requests;	/* list of requests with
1107						 * ->fl_blocker pointing here
1108						 */
1109	struct list_head fl_blocked_member;	/* node in
1110						 * ->fl_blocker->fl_blocked_requests
1111						 */
1112	fl_owner_t fl_owner;
1113	unsigned int fl_flags;
1114	unsigned char fl_type;
1115	unsigned int fl_pid;
1116	int fl_link_cpu;		/* what cpu's list is this on? */
1117	wait_queue_head_t fl_wait;
1118	struct file *fl_file;
1119	loff_t fl_start;
1120	loff_t fl_end;
1121
1122	struct fasync_struct *	fl_fasync; /* for lease break notifications */
1123	/* for lease breaks: */
1124	unsigned long fl_break_time;
1125	unsigned long fl_downgrade_time;
1126
1127	const struct file_lock_operations *fl_ops;	/* Callbacks for filesystems */
1128	const struct lock_manager_operations *fl_lmops;	/* Callbacks for lockmanagers */
1129	union {
1130		struct nfs_lock_info	nfs_fl;
1131		struct nfs4_lock_info	nfs4_fl;
1132		struct {
1133			struct list_head link;	/* link in AFS vnode's pending_locks list */
1134			int state;		/* state of grant or error if -ve */
1135			unsigned int	debug_id;
1136		} afs;
1137	} fl_u;
1138} __randomize_layout;
1139
1140struct file_lock_context {
1141	spinlock_t		flc_lock;
1142	struct list_head	flc_flock;
1143	struct list_head	flc_posix;
1144	struct list_head	flc_lease;
1145};
1146
1147/* The following constant reflects the upper bound of the file/locking space */
1148#ifndef OFFSET_MAX
1149#define INT_LIMIT(x)	(~((x)1 << (sizeof(x)*8 - 1)))
1150#define OFFSET_MAX	INT_LIMIT(loff_t)
1151#define OFFT_OFFSET_MAX	INT_LIMIT(off_t)
1152#endif
1153
1154extern void send_sigio(struct fown_struct *fown, int fd, int band);
1155
1156#define locks_inode(f) file_inode(f)
1157
1158#ifdef CONFIG_FILE_LOCKING
1159extern int fcntl_getlk(struct file *, unsigned int, struct flock *);
1160extern int fcntl_setlk(unsigned int, struct file *, unsigned int,
1161			struct flock *);
1162
1163#if BITS_PER_LONG == 32
1164extern int fcntl_getlk64(struct file *, unsigned int, struct flock64 *);
1165extern int fcntl_setlk64(unsigned int, struct file *, unsigned int,
1166			struct flock64 *);
1167#endif
1168
1169extern int fcntl_setlease(unsigned int fd, struct file *filp, long arg);
1170extern int fcntl_getlease(struct file *filp);
1171
1172/* fs/locks.c */
1173void locks_free_lock_context(struct inode *inode);
1174void locks_free_lock(struct file_lock *fl);
1175extern void locks_init_lock(struct file_lock *);
1176extern struct file_lock * locks_alloc_lock(void);
1177extern void locks_copy_lock(struct file_lock *, struct file_lock *);
1178extern void locks_copy_conflock(struct file_lock *, struct file_lock *);
1179extern void locks_remove_posix(struct file *, fl_owner_t);
1180extern void locks_remove_file(struct file *);
1181extern void locks_release_private(struct file_lock *);
1182extern void posix_test_lock(struct file *, struct file_lock *);
1183extern int posix_lock_file(struct file *, struct file_lock *, struct file_lock *);
1184extern int locks_delete_block(struct file_lock *);
1185extern int vfs_test_lock(struct file *, struct file_lock *);
1186extern int vfs_lock_file(struct file *, unsigned int, struct file_lock *, struct file_lock *);
1187extern int vfs_cancel_lock(struct file *filp, struct file_lock *fl);
1188extern int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl);
1189extern int __break_lease(struct inode *inode, unsigned int flags, unsigned int type);
1190extern void lease_get_mtime(struct inode *, struct timespec64 *time);
1191extern int generic_setlease(struct file *, long, struct file_lock **, void **priv);
1192extern int vfs_setlease(struct file *, long, struct file_lock **, void **);
1193extern int lease_modify(struct file_lock *, int, struct list_head *);
1194
1195struct notifier_block;
1196extern int lease_register_notifier(struct notifier_block *);
1197extern void lease_unregister_notifier(struct notifier_block *);
1198
1199struct files_struct;
1200extern void show_fd_locks(struct seq_file *f,
1201			 struct file *filp, struct files_struct *files);
1202#else /* !CONFIG_FILE_LOCKING */
1203static inline int fcntl_getlk(struct file *file, unsigned int cmd,
1204			      struct flock __user *user)
1205{
1206	return -EINVAL;
1207}
1208
1209static inline int fcntl_setlk(unsigned int fd, struct file *file,
1210			      unsigned int cmd, struct flock __user *user)
1211{
1212	return -EACCES;
1213}
1214
1215#if BITS_PER_LONG == 32
1216static inline int fcntl_getlk64(struct file *file, unsigned int cmd,
1217				struct flock64 *user)
1218{
1219	return -EINVAL;
1220}
1221
1222static inline int fcntl_setlk64(unsigned int fd, struct file *file,
1223				unsigned int cmd, struct flock64 *user)
1224{
1225	return -EACCES;
1226}
1227#endif
1228static inline int fcntl_setlease(unsigned int fd, struct file *filp, long arg)
1229{
1230	return -EINVAL;
1231}
1232
1233static inline int fcntl_getlease(struct file *filp)
1234{
1235	return F_UNLCK;
1236}
1237
1238static inline void
1239locks_free_lock_context(struct inode *inode)
1240{
1241}
1242
1243static inline void locks_init_lock(struct file_lock *fl)
1244{
1245	return;
1246}
1247
1248static inline void locks_copy_conflock(struct file_lock *new, struct file_lock *fl)
1249{
1250	return;
1251}
1252
1253static inline void locks_copy_lock(struct file_lock *new, struct file_lock *fl)
1254{
1255	return;
1256}
1257
1258static inline void locks_remove_posix(struct file *filp, fl_owner_t owner)
1259{
1260	return;
1261}
1262
1263static inline void locks_remove_file(struct file *filp)
1264{
1265	return;
1266}
1267
1268static inline void posix_test_lock(struct file *filp, struct file_lock *fl)
1269{
1270	return;
1271}
1272
1273static inline int posix_lock_file(struct file *filp, struct file_lock *fl,
1274				  struct file_lock *conflock)
1275{
1276	return -ENOLCK;
1277}
1278
1279static inline int locks_delete_block(struct file_lock *waiter)
1280{
1281	return -ENOENT;
1282}
1283
1284static inline int vfs_test_lock(struct file *filp, struct file_lock *fl)
1285{
1286	return 0;
1287}
1288
1289static inline int vfs_lock_file(struct file *filp, unsigned int cmd,
1290				struct file_lock *fl, struct file_lock *conf)
1291{
1292	return -ENOLCK;
1293}
1294
1295static inline int vfs_cancel_lock(struct file *filp, struct file_lock *fl)
1296{
1297	return 0;
1298}
1299
1300static inline int locks_lock_inode_wait(struct inode *inode, struct file_lock *fl)
1301{
1302	return -ENOLCK;
1303}
1304
1305static inline int __break_lease(struct inode *inode, unsigned int mode, unsigned int type)
1306{
1307	return 0;
1308}
1309
1310static inline void lease_get_mtime(struct inode *inode,
1311				   struct timespec64 *time)
1312{
1313	return;
1314}
1315
1316static inline int generic_setlease(struct file *filp, long arg,
1317				    struct file_lock **flp, void **priv)
1318{
1319	return -EINVAL;
1320}
1321
1322static inline int vfs_setlease(struct file *filp, long arg,
1323			       struct file_lock **lease, void **priv)
1324{
1325	return -EINVAL;
1326}
1327
1328static inline int lease_modify(struct file_lock *fl, int arg,
1329			       struct list_head *dispose)
1330{
1331	return -EINVAL;
1332}
1333
1334struct files_struct;
1335static inline void show_fd_locks(struct seq_file *f,
1336			struct file *filp, struct files_struct *files) {}
1337#endif /* !CONFIG_FILE_LOCKING */
1338
1339static inline struct inode *file_inode(const struct file *f)
1340{
1341	return f->f_inode;
1342}
1343
1344static inline struct dentry *file_dentry(const struct file *file)
1345{
1346	return d_real(file->f_path.dentry, file_inode(file));
1347}
1348
1349static inline int locks_lock_file_wait(struct file *filp, struct file_lock *fl)
1350{
1351	return locks_lock_inode_wait(locks_inode(filp), fl);
1352}
1353
1354struct fasync_struct {
1355	rwlock_t		fa_lock;
1356	int			magic;
1357	int			fa_fd;
1358	struct fasync_struct	*fa_next; /* singly linked list */
1359	struct file		*fa_file;
1360	struct rcu_head		fa_rcu;
1361};
1362
1363#define FASYNC_MAGIC 0x4601
1364
1365/* SMP safe fasync helpers: */
1366extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1367extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1368extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1369extern struct fasync_struct *fasync_alloc(void);
1370extern void fasync_free(struct fasync_struct *);
1371
1372/* can be called from interrupts */
1373extern void kill_fasync(struct fasync_struct **, int, int);
1374
1375extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1376extern int f_setown(struct file *filp, unsigned long arg, int force);
1377extern void f_delown(struct file *filp);
1378extern pid_t f_getown(struct file *filp);
1379extern int send_sigurg(struct fown_struct *fown);
1380
1381/*
1382 * sb->s_flags.  Note that these mirror the equivalent MS_* flags where
1383 * represented in both.
1384 */
1385#define SB_RDONLY	 1	/* Mount read-only */
1386#define SB_NOSUID	 2	/* Ignore suid and sgid bits */
1387#define SB_NODEV	 4	/* Disallow access to device special files */
1388#define SB_NOEXEC	 8	/* Disallow program execution */
1389#define SB_SYNCHRONOUS	16	/* Writes are synced at once */
1390#define SB_MANDLOCK	64	/* Allow mandatory locks on an FS */
1391#define SB_DIRSYNC	128	/* Directory modifications are synchronous */
1392#define SB_NOATIME	1024	/* Do not update access times. */
1393#define SB_NODIRATIME	2048	/* Do not update directory access times */
1394#define SB_SILENT	32768
1395#define SB_POSIXACL	(1<<16)	/* VFS does not apply the umask */
1396#define SB_INLINECRYPT	(1<<17)	/* Use blk-crypto for encrypted files */
1397#define SB_KERNMOUNT	(1<<22) /* this is a kern_mount call */
1398#define SB_I_VERSION	(1<<23) /* Update inode I_version field */
1399#define SB_LAZYTIME	(1<<25) /* Update the on-disk [acm]times lazily */
1400
1401/* These sb flags are internal to the kernel */
1402#define SB_SUBMOUNT     (1<<26)
1403#define SB_FORCE    	(1<<27)
1404#define SB_NOSEC	(1<<28)
1405#define SB_BORN		(1<<29)
1406#define SB_ACTIVE	(1<<30)
1407#define SB_NOUSER	(1<<31)
1408
1409/* These flags relate to encoding and casefolding */
1410#define SB_ENC_STRICT_MODE_FL	(1 << 0)
1411
1412#define sb_has_strict_encoding(sb) \
1413	(sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1414
1415/*
1416 *	Umount options
1417 */
1418
1419#define MNT_FORCE	0x00000001	/* Attempt to forcibily umount */
1420#define MNT_DETACH	0x00000002	/* Just detach from the tree */
1421#define MNT_EXPIRE	0x00000004	/* Mark for expiry */
1422#define UMOUNT_NOFOLLOW	0x00000008	/* Don't follow symlink on umount */
1423#define UMOUNT_UNUSED	0x80000000	/* Flag guaranteed to be unused */
1424
1425/* sb->s_iflags */
1426#define SB_I_CGROUPWB	0x00000001	/* cgroup-aware writeback enabled */
1427#define SB_I_NOEXEC	0x00000002	/* Ignore executables on this fs */
1428#define SB_I_NODEV	0x00000004	/* Ignore devices on this fs */
1429#define SB_I_STABLE_WRITES 0x00000008	/* don't modify blks until WB is done */
1430
1431/* sb->s_iflags to limit user namespace mounts */
1432#define SB_I_USERNS_VISIBLE		0x00000010 /* fstype already mounted */
1433#define SB_I_IMA_UNVERIFIABLE_SIGNATURE	0x00000020
1434#define SB_I_UNTRUSTED_MOUNTER		0x00000040
1435
1436#define SB_I_SKIP_SYNC	0x00000100	/* Skip superblock at global sync */
1437#define SB_I_PERSB_BDI	0x00000200	/* has a per-sb bdi */
1438
1439/* Possible states of 'frozen' field */
1440enum {
1441	SB_UNFROZEN = 0,		/* FS is unfrozen */
1442	SB_FREEZE_WRITE	= 1,		/* Writes, dir ops, ioctls frozen */
1443	SB_FREEZE_PAGEFAULT = 2,	/* Page faults stopped as well */
1444	SB_FREEZE_FS = 3,		/* For internal FS use (e.g. to stop
1445					 * internal threads if needed) */
1446	SB_FREEZE_COMPLETE = 4,		/* ->freeze_fs finished successfully */
1447};
1448
1449#define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1450
1451struct sb_writers {
1452	int				frozen;		/* Is sb frozen? */
1453	wait_queue_head_t		wait_unfrozen;	/* wait for thaw */
1454	struct percpu_rw_semaphore	rw_sem[SB_FREEZE_LEVELS];
1455};
1456
1457struct super_block {
1458	struct list_head	s_list;		/* Keep this first */
1459	dev_t			s_dev;		/* search index; _not_ kdev_t */
1460	unsigned char		s_blocksize_bits;
1461	unsigned long		s_blocksize;
1462	loff_t			s_maxbytes;	/* Max file size */
1463	struct file_system_type	*s_type;
1464	const struct super_operations	*s_op;
1465	const struct dquot_operations	*dq_op;
1466	const struct quotactl_ops	*s_qcop;
1467	const struct export_operations *s_export_op;
1468	unsigned long		s_flags;
1469	unsigned long		s_iflags;	/* internal SB_I_* flags */
1470	unsigned long		s_magic;
1471	struct dentry		*s_root;
1472	struct rw_semaphore	s_umount;
1473	int			s_count;
1474	atomic_t		s_active;
1475#ifdef CONFIG_SECURITY
1476	void                    *s_security;
1477#endif
1478	const struct xattr_handler **s_xattr;
1479#ifdef CONFIG_FS_ENCRYPTION
1480	const struct fscrypt_operations	*s_cop;
1481	struct key		*s_master_keys; /* master crypto keys in use */
1482#endif
1483#ifdef CONFIG_FS_VERITY
1484	const struct fsverity_operations *s_vop;
1485#endif
1486#ifdef CONFIG_UNICODE
1487	struct unicode_map *s_encoding;
1488	__u16 s_encoding_flags;
1489#endif
1490	struct hlist_bl_head	s_roots;	/* alternate root dentries for NFS */
1491	struct list_head	s_mounts;	/* list of mounts; _not_ for fs use */
1492	struct block_device	*s_bdev;
1493	struct backing_dev_info *s_bdi;
1494	struct mtd_info		*s_mtd;
1495	struct hlist_node	s_instances;
1496	unsigned int		s_quota_types;	/* Bitmask of supported quota types */
1497	struct quota_info	s_dquot;	/* Diskquota specific options */
1498
1499	struct sb_writers	s_writers;
1500
1501	/*
1502	 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1503	 * s_fsnotify_marks together for cache efficiency. They are frequently
1504	 * accessed and rarely modified.
1505	 */
1506	void			*s_fs_info;	/* Filesystem private info */
1507
1508	/* Granularity of c/m/atime in ns (cannot be worse than a second) */
1509	u32			s_time_gran;
1510	/* Time limits for c/m/atime in seconds */
1511	time64_t		   s_time_min;
1512	time64_t		   s_time_max;
1513#ifdef CONFIG_FSNOTIFY
1514	__u32			s_fsnotify_mask;
1515	struct fsnotify_mark_connector __rcu	*s_fsnotify_marks;
1516#endif
1517
1518	char			s_id[32];	/* Informational name */
1519	uuid_t			s_uuid;		/* UUID */
1520
1521	unsigned int		s_max_links;
1522	fmode_t			s_mode;
1523
1524	/*
1525	 * The next field is for VFS *only*. No filesystems have any business
1526	 * even looking at it. You had been warned.
1527	 */
1528	struct mutex s_vfs_rename_mutex;	/* Kludge */
1529
1530	/*
1531	 * Filesystem subtype.  If non-empty the filesystem type field
1532	 * in /proc/mounts will be "type.subtype"
1533	 */
1534	const char *s_subtype;
1535
1536	const struct dentry_operations *s_d_op; /* default d_op for dentries */
1537
1538	struct shrinker s_shrink;	/* per-sb shrinker handle */
1539
1540	/* Number of inodes with nlink == 0 but still referenced */
1541	atomic_long_t s_remove_count;
1542
1543	/*
1544	 * Number of inode/mount/sb objects that are being watched, note that
1545	 * inodes objects are currently double-accounted.
1546	 */
1547	atomic_long_t s_fsnotify_connectors;
1548
1549	/* Being remounted read-only */
1550	int s_readonly_remount;
1551
1552	/* per-sb errseq_t for reporting writeback errors via syncfs */
1553	errseq_t s_wb_err;
1554
1555	/* AIO completions deferred from interrupt context */
1556	struct workqueue_struct *s_dio_done_wq;
1557	struct hlist_head s_pins;
1558
1559	/*
1560	 * Owning user namespace and default context in which to
1561	 * interpret filesystem uids, gids, quotas, device nodes,
1562	 * xattrs and security labels.
1563	 */
1564	struct user_namespace *s_user_ns;
1565
1566	/*
1567	 * The list_lru structure is essentially just a pointer to a table
1568	 * of per-node lru lists, each of which has its own spinlock.
1569	 * There is no need to put them into separate cachelines.
1570	 */
1571	struct list_lru		s_dentry_lru;
1572	struct list_lru		s_inode_lru;
1573	struct rcu_head		rcu;
1574	struct work_struct	destroy_work;
1575
1576	struct mutex		s_sync_lock;	/* sync serialisation lock */
1577
1578	/*
1579	 * Indicates how deep in a filesystem stack this SB is
1580	 */
1581	int s_stack_depth;
1582
1583	/* s_inode_list_lock protects s_inodes */
1584	spinlock_t		s_inode_list_lock ____cacheline_aligned_in_smp;
1585	struct list_head	s_inodes;	/* all inodes */
1586
1587	spinlock_t		s_inode_wblist_lock;
1588	struct list_head	s_inodes_wb;	/* writeback inodes */
1589} __randomize_layout;
1590
1591static inline struct user_namespace *i_user_ns(const struct inode *inode)
1592{
1593	return inode->i_sb->s_user_ns;
1594}
1595
1596/* Helper functions so that in most cases filesystems will
1597 * not need to deal directly with kuid_t and kgid_t and can
1598 * instead deal with the raw numeric values that are stored
1599 * in the filesystem.
1600 */
1601static inline uid_t i_uid_read(const struct inode *inode)
1602{
1603	return from_kuid(i_user_ns(inode), inode->i_uid);
1604}
1605
1606static inline gid_t i_gid_read(const struct inode *inode)
1607{
1608	return from_kgid(i_user_ns(inode), inode->i_gid);
1609}
1610
1611static inline void i_uid_write(struct inode *inode, uid_t uid)
1612{
1613	inode->i_uid = make_kuid(i_user_ns(inode), uid);
1614}
1615
1616static inline void i_gid_write(struct inode *inode, gid_t gid)
1617{
1618	inode->i_gid = make_kgid(i_user_ns(inode), gid);
1619}
1620
1621/**
1622 * i_uid_into_mnt - map an inode's i_uid down into a mnt_userns
1623 * @mnt_userns: user namespace of the mount the inode was found from
1624 * @inode: inode to map
1625 *
1626 * Return: the inode's i_uid mapped down according to @mnt_userns.
1627 * If the inode's i_uid has no mapping INVALID_UID is returned.
1628 */
1629static inline kuid_t i_uid_into_mnt(struct user_namespace *mnt_userns,
1630				    const struct inode *inode)
1631{
1632	return mapped_kuid_fs(mnt_userns, i_user_ns(inode), inode->i_uid);
1633}
1634
1635/**
1636 * i_gid_into_mnt - map an inode's i_gid down into a mnt_userns
1637 * @mnt_userns: user namespace of the mount the inode was found from
1638 * @inode: inode to map
1639 *
1640 * Return: the inode's i_gid mapped down according to @mnt_userns.
1641 * If the inode's i_gid has no mapping INVALID_GID is returned.
1642 */
1643static inline kgid_t i_gid_into_mnt(struct user_namespace *mnt_userns,
1644				    const struct inode *inode)
1645{
1646	return mapped_kgid_fs(mnt_userns, i_user_ns(inode), inode->i_gid);
1647}
1648
1649/**
1650 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1651 * @inode: inode to initialize
1652 * @mnt_userns: user namespace of the mount the inode was found from
1653 *
1654 * Initialize the i_uid field of @inode. If the inode was found/created via
1655 * an idmapped mount map the caller's fsuid according to @mnt_users.
1656 */
1657static inline void inode_fsuid_set(struct inode *inode,
1658				   struct user_namespace *mnt_userns)
1659{
1660	inode->i_uid = mapped_fsuid(mnt_userns, i_user_ns(inode));
1661}
1662
1663/**
1664 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1665 * @inode: inode to initialize
1666 * @mnt_userns: user namespace of the mount the inode was found from
1667 *
1668 * Initialize the i_gid field of @inode. If the inode was found/created via
1669 * an idmapped mount map the caller's fsgid according to @mnt_users.
1670 */
1671static inline void inode_fsgid_set(struct inode *inode,
1672				   struct user_namespace *mnt_userns)
1673{
1674	inode->i_gid = mapped_fsgid(mnt_userns, i_user_ns(inode));
1675}
1676
1677/**
1678 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1679 * @sb: the superblock we want a mapping in
1680 * @mnt_userns: user namespace of the relevant mount
1681 *
1682 * Check whether the caller's fsuid and fsgid have a valid mapping in the
1683 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1684 * the caller's fsuid and fsgid according to the @mnt_userns first.
1685 *
1686 * Return: true if fsuid and fsgid is mapped, false if not.
1687 */
1688static inline bool fsuidgid_has_mapping(struct super_block *sb,
1689					struct user_namespace *mnt_userns)
1690{
1691	struct user_namespace *fs_userns = sb->s_user_ns;
1692	kuid_t kuid;
1693	kgid_t kgid;
1694
1695	kuid = mapped_fsuid(mnt_userns, fs_userns);
1696	if (!uid_valid(kuid))
1697		return false;
1698	kgid = mapped_fsgid(mnt_userns, fs_userns);
1699	if (!gid_valid(kgid))
1700		return false;
1701	return kuid_has_mapping(fs_userns, kuid) &&
1702	       kgid_has_mapping(fs_userns, kgid);
1703}
1704
1705extern struct timespec64 current_time(struct inode *inode);
1706
1707/*
1708 * Snapshotting support.
1709 */
1710
1711/*
1712 * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1713 * instead.
1714 */
1715static inline void __sb_end_write(struct super_block *sb, int level)
1716{
1717	percpu_up_read(sb->s_writers.rw_sem + level-1);
1718}
1719
1720static inline void __sb_start_write(struct super_block *sb, int level)
1721{
1722	percpu_down_read(sb->s_writers.rw_sem + level - 1);
1723}
1724
1725static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1726{
1727	return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1728}
1729
1730#define __sb_writers_acquired(sb, lev)	\
1731	percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1732#define __sb_writers_release(sb, lev)	\
1733	percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1734
1735/**
1736 * sb_end_write - drop write access to a superblock
1737 * @sb: the super we wrote to
1738 *
1739 * Decrement number of writers to the filesystem. Wake up possible waiters
1740 * wanting to freeze the filesystem.
1741 */
1742static inline void sb_end_write(struct super_block *sb)
1743{
1744	__sb_end_write(sb, SB_FREEZE_WRITE);
1745}
1746
1747/**
1748 * sb_end_pagefault - drop write access to a superblock from a page fault
1749 * @sb: the super we wrote to
1750 *
1751 * Decrement number of processes handling write page fault to the filesystem.
1752 * Wake up possible waiters wanting to freeze the filesystem.
1753 */
1754static inline void sb_end_pagefault(struct super_block *sb)
1755{
1756	__sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1757}
1758
1759/**
1760 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1761 * @sb: the super we wrote to
1762 *
1763 * Decrement fs-internal number of writers to the filesystem.  Wake up possible
1764 * waiters wanting to freeze the filesystem.
1765 */
1766static inline void sb_end_intwrite(struct super_block *sb)
1767{
1768	__sb_end_write(sb, SB_FREEZE_FS);
1769}
1770
1771/**
1772 * sb_start_write - get write access to a superblock
1773 * @sb: the super we write to
1774 *
1775 * When a process wants to write data or metadata to a file system (i.e. dirty
1776 * a page or an inode), it should embed the operation in a sb_start_write() -
1777 * sb_end_write() pair to get exclusion against file system freezing. This
1778 * function increments number of writers preventing freezing. If the file
1779 * system is already frozen, the function waits until the file system is
1780 * thawed.
1781 *
1782 * Since freeze protection behaves as a lock, users have to preserve
1783 * ordering of freeze protection and other filesystem locks. Generally,
1784 * freeze protection should be the outermost lock. In particular, we have:
1785 *
1786 * sb_start_write
1787 *   -> i_mutex			(write path, truncate, directory ops, ...)
1788 *   -> s_umount		(freeze_super, thaw_super)
1789 */
1790static inline void sb_start_write(struct super_block *sb)
1791{
1792	__sb_start_write(sb, SB_FREEZE_WRITE);
1793}
1794
1795static inline bool sb_start_write_trylock(struct super_block *sb)
1796{
1797	return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1798}
1799
1800/**
1801 * sb_start_pagefault - get write access to a superblock from a page fault
1802 * @sb: the super we write to
1803 *
1804 * When a process starts handling write page fault, it should embed the
1805 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1806 * exclusion against file system freezing. This is needed since the page fault
1807 * is going to dirty a page. This function increments number of running page
1808 * faults preventing freezing. If the file system is already frozen, the
1809 * function waits until the file system is thawed.
1810 *
1811 * Since page fault freeze protection behaves as a lock, users have to preserve
1812 * ordering of freeze protection and other filesystem locks. It is advised to
1813 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1814 * handling code implies lock dependency:
1815 *
1816 * mmap_lock
1817 *   -> sb_start_pagefault
1818 */
1819static inline void sb_start_pagefault(struct super_block *sb)
1820{
1821	__sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1822}
1823
1824/**
1825 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1826 * @sb: the super we write to
1827 *
1828 * This is the third level of protection against filesystem freezing. It is
1829 * free for use by a filesystem. The only requirement is that it must rank
1830 * below sb_start_pagefault.
1831 *
1832 * For example filesystem can call sb_start_intwrite() when starting a
1833 * transaction which somewhat eases handling of freezing for internal sources
1834 * of filesystem changes (internal fs threads, discarding preallocation on file
1835 * close, etc.).
1836 */
1837static inline void sb_start_intwrite(struct super_block *sb)
1838{
1839	__sb_start_write(sb, SB_FREEZE_FS);
1840}
1841
1842static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1843{
1844	return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1845}
1846
1847bool inode_owner_or_capable(struct user_namespace *mnt_userns,
1848			    const struct inode *inode);
1849
1850/*
1851 * VFS helper functions..
1852 */
1853int vfs_create(struct user_namespace *, struct inode *,
1854	       struct dentry *, umode_t, bool);
1855int vfs_mkdir(struct user_namespace *, struct inode *,
1856	      struct dentry *, umode_t);
1857int vfs_mknod(struct user_namespace *, struct inode *, struct dentry *,
1858              umode_t, dev_t);
1859int vfs_symlink(struct user_namespace *, struct inode *,
1860		struct dentry *, const char *);
1861int vfs_link(struct dentry *, struct user_namespace *, struct inode *,
1862	     struct dentry *, struct inode **);
1863int vfs_rmdir(struct user_namespace *, struct inode *, struct dentry *);
1864int vfs_unlink(struct user_namespace *, struct inode *, struct dentry *,
1865	       struct inode **);
1866
1867/**
1868 * struct renamedata - contains all information required for renaming
1869 * @old_mnt_userns:    old user namespace of the mount the inode was found from
1870 * @old_dir:           parent of source
1871 * @old_dentry:                source
1872 * @new_mnt_userns:    new user namespace of the mount the inode was found from
1873 * @new_dir:           parent of destination
1874 * @new_dentry:                destination
1875 * @delegated_inode:   returns an inode needing a delegation break
1876 * @flags:             rename flags
1877 */
1878struct renamedata {
1879	struct user_namespace *old_mnt_userns;
1880	struct inode *old_dir;
1881	struct dentry *old_dentry;
1882	struct user_namespace *new_mnt_userns;
1883	struct inode *new_dir;
1884	struct dentry *new_dentry;
1885	struct inode **delegated_inode;
1886	unsigned int flags;
1887} __randomize_layout;
1888
1889int vfs_rename(struct renamedata *);
1890
1891static inline int vfs_whiteout(struct user_namespace *mnt_userns,
1892			       struct inode *dir, struct dentry *dentry)
1893{
1894	return vfs_mknod(mnt_userns, dir, dentry, S_IFCHR | WHITEOUT_MODE,
1895			 WHITEOUT_DEV);
1896}
1897
1898struct dentry *vfs_tmpfile(struct user_namespace *mnt_userns,
1899			   struct dentry *dentry, umode_t mode, int open_flag);
1900
1901int vfs_mkobj(struct dentry *, umode_t,
1902		int (*f)(struct dentry *, umode_t, void *),
1903		void *);
1904
1905int vfs_fchown(struct file *file, uid_t user, gid_t group);
1906int vfs_fchmod(struct file *file, umode_t mode);
1907int vfs_utimes(const struct path *path, struct timespec64 *times);
1908
1909extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1910
1911#ifdef CONFIG_COMPAT
1912extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
1913					unsigned long arg);
1914#else
1915#define compat_ptr_ioctl NULL
1916#endif
1917
1918/*
1919 * VFS file helper functions.
1920 */
1921void inode_init_owner(struct user_namespace *mnt_userns, struct inode *inode,
1922		      const struct inode *dir, umode_t mode);
1923extern bool may_open_dev(const struct path *path);
1924
1925/*
1926 * This is the "filldir" function type, used by readdir() to let
1927 * the kernel specify what kind of dirent layout it wants to have.
1928 * This allows the kernel to read directories into kernel space or
1929 * to have different dirent layouts depending on the binary type.
1930 */
1931struct dir_context;
1932typedef int (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1933			 unsigned);
1934
1935struct dir_context {
1936	filldir_t actor;
1937	loff_t pos;
1938};
1939
1940/*
1941 * These flags let !MMU mmap() govern direct device mapping vs immediate
1942 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1943 *
1944 * NOMMU_MAP_COPY:	Copy can be mapped (MAP_PRIVATE)
1945 * NOMMU_MAP_DIRECT:	Can be mapped directly (MAP_SHARED)
1946 * NOMMU_MAP_READ:	Can be mapped for reading
1947 * NOMMU_MAP_WRITE:	Can be mapped for writing
1948 * NOMMU_MAP_EXEC:	Can be mapped for execution
1949 */
1950#define NOMMU_MAP_COPY		0x00000001
1951#define NOMMU_MAP_DIRECT	0x00000008
1952#define NOMMU_MAP_READ		VM_MAYREAD
1953#define NOMMU_MAP_WRITE		VM_MAYWRITE
1954#define NOMMU_MAP_EXEC		VM_MAYEXEC
1955
1956#define NOMMU_VMFLAGS \
1957	(NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
1958
1959/*
1960 * These flags control the behavior of the remap_file_range function pointer.
1961 * If it is called with len == 0 that means "remap to end of source file".
1962 * See Documentation/filesystems/vfs.rst for more details about this call.
1963 *
1964 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
1965 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
1966 */
1967#define REMAP_FILE_DEDUP		(1 << 0)
1968#define REMAP_FILE_CAN_SHORTEN		(1 << 1)
1969
1970/*
1971 * These flags signal that the caller is ok with altering various aspects of
1972 * the behavior of the remap operation.  The changes must be made by the
1973 * implementation; the vfs remap helper functions can take advantage of them.
1974 * Flags in this category exist to preserve the quirky behavior of the hoisted
1975 * btrfs clone/dedupe ioctls.
1976 */
1977#define REMAP_FILE_ADVISORY		(REMAP_FILE_CAN_SHORTEN)
1978
1979struct iov_iter;
1980
1981struct file_operations {
1982	struct module *owner;
1983	loff_t (*llseek) (struct file *, loff_t, int);
1984	ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
1985	ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
1986	ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
1987	ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
1988	int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
1989			unsigned int flags);
1990	int (*iterate) (struct file *, struct dir_context *);
1991	int (*iterate_shared) (struct file *, struct dir_context *);
1992	__poll_t (*poll) (struct file *, struct poll_table_struct *);
1993	long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
1994	long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
1995	int (*mmap) (struct file *, struct vm_area_struct *);
1996	unsigned long mmap_supported_flags;
1997	int (*open) (struct inode *, struct file *);
1998	int (*flush) (struct file *, fl_owner_t id);
1999	int (*release) (struct inode *, struct file *);
2000	int (*fsync) (struct file *, loff_t, loff_t, int datasync);
2001	int (*fasync) (int, struct file *, int);
2002	int (*lock) (struct file *, int, struct file_lock *);
2003	ssize_t (*sendpage) (struct file *, struct page *, int, size_t, loff_t *, int);
2004	unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2005	int (*check_flags)(int);
2006	int (*flock) (struct file *, int, struct file_lock *);
2007	ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
2008	ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
2009	int (*setlease)(struct file *, long, struct file_lock **, void **);
2010	long (*fallocate)(struct file *file, int mode, loff_t offset,
2011			  loff_t len);
2012	void (*show_fdinfo)(struct seq_file *m, struct file *f);
2013#ifndef CONFIG_MMU
2014	unsigned (*mmap_capabilities)(struct file *);
2015#endif
2016	ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
2017			loff_t, size_t, unsigned int);
2018	loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
2019				   struct file *file_out, loff_t pos_out,
2020				   loff_t len, unsigned int remap_flags);
2021	int (*fadvise)(struct file *, loff_t, loff_t, int);
2022} __randomize_layout;
2023
2024struct inode_operations {
2025	struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
2026	const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
2027	int (*permission) (struct user_namespace *, struct inode *, int);
2028	struct posix_acl * (*get_acl)(struct inode *, int, bool);
2029
2030	int (*readlink) (struct dentry *, char __user *,int);
2031
2032	int (*create) (struct user_namespace *, struct inode *,struct dentry *,
2033		       umode_t, bool);
2034	int (*link) (struct dentry *,struct inode *,struct dentry *);
2035	int (*unlink) (struct inode *,struct dentry *);
2036	int (*symlink) (struct user_namespace *, struct inode *,struct dentry *,
2037			const char *);
2038	int (*mkdir) (struct user_namespace *, struct inode *,struct dentry *,
2039		      umode_t);
2040	int (*rmdir) (struct inode *,struct dentry *);
2041	int (*mknod) (struct user_namespace *, struct inode *,struct dentry *,
2042		      umode_t,dev_t);
2043	int (*rename) (struct user_namespace *, struct inode *, struct dentry *,
2044			struct inode *, struct dentry *, unsigned int);
2045	int (*setattr) (struct user_namespace *, struct dentry *,
2046			struct iattr *);
2047	int (*getattr) (struct user_namespace *, const struct path *,
2048			struct kstat *, u32, unsigned int);
2049	ssize_t (*listxattr) (struct dentry *, char *, size_t);
2050	int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
2051		      u64 len);
2052	int (*update_time)(struct inode *, struct timespec64 *, int);
2053	int (*atomic_open)(struct inode *, struct dentry *,
2054			   struct file *, unsigned open_flag,
2055			   umode_t create_mode);
2056	int (*tmpfile) (struct user_namespace *, struct inode *,
2057			struct dentry *, umode_t);
2058	int (*set_acl)(struct user_namespace *, struct inode *,
2059		       struct posix_acl *, int);
2060	int (*fileattr_set)(struct user_namespace *mnt_userns,
2061			    struct dentry *dentry, struct fileattr *fa);
2062	int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
2063} ____cacheline_aligned;
2064
2065static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
2066				     struct iov_iter *iter)
2067{
2068	return file->f_op->read_iter(kio, iter);
2069}
2070
2071static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
2072				      struct iov_iter *iter)
2073{
2074	return file->f_op->write_iter(kio, iter);
2075}
2076
2077static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
2078{
2079	return file->f_op->mmap(file, vma);
2080}
2081
2082extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
2083extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
2084extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
2085				   loff_t, size_t, unsigned int);
2086extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
2087				       struct file *file_out, loff_t pos_out,
2088				       size_t len, unsigned int flags);
2089extern int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2090					 struct file *file_out, loff_t pos_out,
2091					 loff_t *count,
2092					 unsigned int remap_flags);
2093extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in,
2094				  struct file *file_out, loff_t pos_out,
2095				  loff_t len, unsigned int remap_flags);
2096extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
2097				   struct file *file_out, loff_t pos_out,
2098				   loff_t len, unsigned int remap_flags);
2099extern int vfs_dedupe_file_range(struct file *file,
2100				 struct file_dedupe_range *same);
2101extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
2102					struct file *dst_file, loff_t dst_pos,
2103					loff_t len, unsigned int remap_flags);
2104
2105
2106struct super_operations {
2107   	struct inode *(*alloc_inode)(struct super_block *sb);
2108	void (*destroy_inode)(struct inode *);
2109	void (*free_inode)(struct inode *);
2110
2111   	void (*dirty_inode) (struct inode *, int flags);
2112	int (*write_inode) (struct inode *, struct writeback_control *wbc);
2113	int (*drop_inode) (struct inode *);
2114	void (*evict_inode) (struct inode *);
2115	void (*put_super) (struct super_block *);
2116	int (*sync_fs)(struct super_block *sb, int wait);
2117	int (*freeze_super) (struct super_block *);
2118	int (*freeze_fs) (struct super_block *);
2119	int (*thaw_super) (struct super_block *);
2120	int (*unfreeze_fs) (struct super_block *);
2121	int (*statfs) (struct dentry *, struct kstatfs *);
2122	int (*remount_fs) (struct super_block *, int *, char *);
2123	void (*umount_begin) (struct super_block *);
2124
2125	int (*show_options)(struct seq_file *, struct dentry *);
2126	int (*show_devname)(struct seq_file *, struct dentry *);
2127	int (*show_path)(struct seq_file *, struct dentry *);
2128	int (*show_stats)(struct seq_file *, struct dentry *);
2129#ifdef CONFIG_QUOTA
2130	ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
2131	ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
2132	struct dquot **(*get_dquots)(struct inode *);
2133#endif
2134	long (*nr_cached_objects)(struct super_block *,
2135				  struct shrink_control *);
2136	long (*free_cached_objects)(struct super_block *,
2137				    struct shrink_control *);
2138};
2139
2140/*
2141 * Inode flags - they have no relation to superblock flags now
2142 */
2143#define S_SYNC		(1 << 0)  /* Writes are synced at once */
2144#define S_NOATIME	(1 << 1)  /* Do not update access times */
2145#define S_APPEND	(1 << 2)  /* Append-only file */
2146#define S_IMMUTABLE	(1 << 3)  /* Immutable file */
2147#define S_DEAD		(1 << 4)  /* removed, but still open directory */
2148#define S_NOQUOTA	(1 << 5)  /* Inode is not counted to quota */
2149#define S_DIRSYNC	(1 << 6)  /* Directory modifications are synchronous */
2150#define S_NOCMTIME	(1 << 7)  /* Do not update file c/mtime */
2151#define S_SWAPFILE	(1 << 8)  /* Do not truncate: swapon got its bmaps */
2152#define S_PRIVATE	(1 << 9)  /* Inode is fs-internal */
2153#define S_IMA		(1 << 10) /* Inode has an associated IMA struct */
2154#define S_AUTOMOUNT	(1 << 11) /* Automount/referral quasi-directory */
2155#define S_NOSEC		(1 << 12) /* no suid or xattr security attributes */
2156#ifdef CONFIG_FS_DAX
2157#define S_DAX		(1 << 13) /* Direct Access, avoiding the page cache */
2158#else
2159#define S_DAX		0	  /* Make all the DAX code disappear */
2160#endif
2161#define S_ENCRYPTED	(1 << 14) /* Encrypted file (using fs/crypto/) */
2162#define S_CASEFOLD	(1 << 15) /* Casefolded file */
2163#define S_VERITY	(1 << 16) /* Verity file (using fs/verity/) */
2164#define S_KERNEL_FILE	(1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
2165
2166/*
2167 * Note that nosuid etc flags are inode-specific: setting some file-system
2168 * flags just means all the inodes inherit those flags by default. It might be
2169 * possible to override it selectively if you really wanted to with some
2170 * ioctl() that is not currently implemented.
2171 *
2172 * Exception: SB_RDONLY is always applied to the entire file system.
2173 *
2174 * Unfortunately, it is possible to change a filesystems flags with it mounted
2175 * with files in use.  This means that all of the inodes will not have their
2176 * i_flags updated.  Hence, i_flags no longer inherit the superblock mount
2177 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
2178 */
2179#define __IS_FLG(inode, flg)	((inode)->i_sb->s_flags & (flg))
2180
2181static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
2182#define IS_RDONLY(inode)	sb_rdonly((inode)->i_sb)
2183#define IS_SYNC(inode)		(__IS_FLG(inode, SB_SYNCHRONOUS) || \
2184					((inode)->i_flags & S_SYNC))
2185#define IS_DIRSYNC(inode)	(__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
2186					((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
2187#define IS_MANDLOCK(inode)	__IS_FLG(inode, SB_MANDLOCK)
2188#define IS_NOATIME(inode)	__IS_FLG(inode, SB_RDONLY|SB_NOATIME)
2189#define IS_I_VERSION(inode)	__IS_FLG(inode, SB_I_VERSION)
2190
2191#define IS_NOQUOTA(inode)	((inode)->i_flags & S_NOQUOTA)
2192#define IS_APPEND(inode)	((inode)->i_flags & S_APPEND)
2193#define IS_IMMUTABLE(inode)	((inode)->i_flags & S_IMMUTABLE)
2194#define IS_POSIXACL(inode)	__IS_FLG(inode, SB_POSIXACL)
2195
2196#define IS_DEADDIR(inode)	((inode)->i_flags & S_DEAD)
2197#define IS_NOCMTIME(inode)	((inode)->i_flags & S_NOCMTIME)
2198#define IS_SWAPFILE(inode)	((inode)->i_flags & S_SWAPFILE)
2199#define IS_PRIVATE(inode)	((inode)->i_flags & S_PRIVATE)
2200#define IS_IMA(inode)		((inode)->i_flags & S_IMA)
2201#define IS_AUTOMOUNT(inode)	((inode)->i_flags & S_AUTOMOUNT)
2202#define IS_NOSEC(inode)		((inode)->i_flags & S_NOSEC)
2203#define IS_DAX(inode)		((inode)->i_flags & S_DAX)
2204#define IS_ENCRYPTED(inode)	((inode)->i_flags & S_ENCRYPTED)
2205#define IS_CASEFOLDED(inode)	((inode)->i_flags & S_CASEFOLD)
2206#define IS_VERITY(inode)	((inode)->i_flags & S_VERITY)
2207
2208#define IS_WHITEOUT(inode)	(S_ISCHR(inode->i_mode) && \
2209				 (inode)->i_rdev == WHITEOUT_DEV)
2210
2211static inline bool HAS_UNMAPPED_ID(struct user_namespace *mnt_userns,
2212				   struct inode *inode)
2213{
2214	return !uid_valid(i_uid_into_mnt(mnt_userns, inode)) ||
2215	       !gid_valid(i_gid_into_mnt(mnt_userns, inode));
2216}
2217
2218static inline enum rw_hint file_write_hint(struct file *file)
2219{
2220	if (file->f_write_hint != WRITE_LIFE_NOT_SET)
2221		return file->f_write_hint;
2222
2223	return file_inode(file)->i_write_hint;
2224}
2225
2226static inline int iocb_flags(struct file *file);
2227
2228static inline u16 ki_hint_validate(enum rw_hint hint)
2229{
2230	typeof(((struct kiocb *)0)->ki_hint) max_hint = -1;
2231
2232	if (hint <= max_hint)
2233		return hint;
2234	return 0;
2235}
2236
2237static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
2238{
2239	*kiocb = (struct kiocb) {
2240		.ki_filp = filp,
2241		.ki_flags = iocb_flags(filp),
2242		.ki_hint = ki_hint_validate(file_write_hint(filp)),
2243		.ki_ioprio = get_current_ioprio(),
2244	};
2245}
2246
2247static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2248			       struct file *filp)
2249{
2250	*kiocb = (struct kiocb) {
2251		.ki_filp = filp,
2252		.ki_flags = kiocb_src->ki_flags,
2253		.ki_hint = kiocb_src->ki_hint,
2254		.ki_ioprio = kiocb_src->ki_ioprio,
2255		.ki_pos = kiocb_src->ki_pos,
2256	};
2257}
2258
2259/*
2260 * Inode state bits.  Protected by inode->i_lock
2261 *
2262 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2263 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2264 *
2265 * Four bits define the lifetime of an inode.  Initially, inodes are I_NEW,
2266 * until that flag is cleared.  I_WILL_FREE, I_FREEING and I_CLEAR are set at
2267 * various stages of removing an inode.
2268 *
2269 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2270 *
2271 * I_DIRTY_SYNC		Inode is dirty, but doesn't have to be written on
2272 *			fdatasync() (unless I_DIRTY_DATASYNC is also set).
2273 *			Timestamp updates are the usual cause.
2274 * I_DIRTY_DATASYNC	Data-related inode changes pending.  We keep track of
2275 *			these changes separately from I_DIRTY_SYNC so that we
2276 *			don't have to write inode on fdatasync() when only
2277 *			e.g. the timestamps have changed.
2278 * I_DIRTY_PAGES	Inode has dirty pages.  Inode itself may be clean.
2279 * I_DIRTY_TIME		The inode itself only has dirty timestamps, and the
2280 *			lazytime mount option is enabled.  We keep track of this
2281 *			separately from I_DIRTY_SYNC in order to implement
2282 *			lazytime.  This gets cleared if I_DIRTY_INODE
2283 *			(I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set.  I.e.
2284 *			either I_DIRTY_TIME *or* I_DIRTY_INODE can be set in
2285 *			i_state, but not both.  I_DIRTY_PAGES may still be set.
2286 * I_NEW		Serves as both a mutex and completion notification.
2287 *			New inodes set I_NEW.  If two processes both create
2288 *			the same inode, one of them will release its inode and
2289 *			wait for I_NEW to be released before returning.
2290 *			Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2291 *			also cause waiting on I_NEW, without I_NEW actually
2292 *			being set.  find_inode() uses this to prevent returning
2293 *			nearly-dead inodes.
2294 * I_WILL_FREE		Must be set when calling write_inode_now() if i_count
2295 *			is zero.  I_FREEING must be set when I_WILL_FREE is
2296 *			cleared.
2297 * I_FREEING		Set when inode is about to be freed but still has dirty
2298 *			pages or buffers attached or the inode itself is still
2299 *			dirty.
2300 * I_CLEAR		Added by clear_inode().  In this state the inode is
2301 *			clean and can be destroyed.  Inode keeps I_FREEING.
2302 *
2303 *			Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2304 *			prohibited for many purposes.  iget() must wait for
2305 *			the inode to be completely released, then create it
2306 *			anew.  Other functions will just ignore such inodes,
2307 *			if appropriate.  I_NEW is used for waiting.
2308 *
2309 * I_SYNC		Writeback of inode is running. The bit is set during
2310 *			data writeback, and cleared with a wakeup on the bit
2311 *			address once it is done. The bit is also used to pin
2312 *			the inode in memory for flusher thread.
2313 *
2314 * I_REFERENCED		Marks the inode as recently references on the LRU list.
2315 *
2316 * I_DIO_WAKEUP		Never set.  Only used as a key for wait_on_bit().
2317 *
2318 * I_WB_SWITCH		Cgroup bdi_writeback switching in progress.  Used to
2319 *			synchronize competing switching instances and to tell
2320 *			wb stat updates to grab the i_pages lock.  See
2321 *			inode_switch_wbs_work_fn() for details.
2322 *
2323 * I_OVL_INUSE		Used by overlayfs to get exclusive ownership on upper
2324 *			and work dirs among overlayfs mounts.
2325 *
2326 * I_CREATING		New object's inode in the middle of setting up.
2327 *
2328 * I_DONTCACHE		Evict inode as soon as it is not used anymore.
2329 *
2330 * I_SYNC_QUEUED	Inode is queued in b_io or b_more_io writeback lists.
2331 *			Used to detect that mark_inode_dirty() should not move
2332 * 			inode between dirty lists.
2333 *
2334 * I_PINNING_FSCACHE_WB	Inode is pinning an fscache object for writeback.
2335 *
2336 * Q: What is the difference between I_WILL_FREE and I_FREEING?
2337 */
2338#define I_DIRTY_SYNC		(1 << 0)
2339#define I_DIRTY_DATASYNC	(1 << 1)
2340#define I_DIRTY_PAGES		(1 << 2)
2341#define __I_NEW			3
2342#define I_NEW			(1 << __I_NEW)
2343#define I_WILL_FREE		(1 << 4)
2344#define I_FREEING		(1 << 5)
2345#define I_CLEAR			(1 << 6)
2346#define __I_SYNC		7
2347#define I_SYNC			(1 << __I_SYNC)
2348#define I_REFERENCED		(1 << 8)
2349#define __I_DIO_WAKEUP		9
2350#define I_DIO_WAKEUP		(1 << __I_DIO_WAKEUP)
2351#define I_LINKABLE		(1 << 10)
2352#define I_DIRTY_TIME		(1 << 11)
2353#define I_WB_SWITCH		(1 << 13)
2354#define I_OVL_INUSE		(1 << 14)
2355#define I_CREATING		(1 << 15)
2356#define I_DONTCACHE		(1 << 16)
2357#define I_SYNC_QUEUED		(1 << 17)
2358#define I_PINNING_FSCACHE_WB	(1 << 18)
2359
2360#define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2361#define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2362#define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2363
2364extern void __mark_inode_dirty(struct inode *, int);
2365static inline void mark_inode_dirty(struct inode *inode)
2366{
2367	__mark_inode_dirty(inode, I_DIRTY);
2368}
2369
2370static inline void mark_inode_dirty_sync(struct inode *inode)
2371{
2372	__mark_inode_dirty(inode, I_DIRTY_SYNC);
2373}
2374
2375/*
2376 * Returns true if the given inode itself only has dirty timestamps (its pages
2377 * may still be dirty) and isn't currently being allocated or freed.
2378 * Filesystems should call this if when writing an inode when lazytime is
2379 * enabled, they want to opportunistically write the timestamps of other inodes
2380 * located very nearby on-disk, e.g. in the same inode block.  This returns true
2381 * if the given inode is in need of such an opportunistic update.  Requires
2382 * i_lock, or at least later re-checking under i_lock.
2383 */
2384static inline bool inode_is_dirtytime_only(struct inode *inode)
2385{
2386	return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2387				  I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2388}
2389
2390extern void inc_nlink(struct inode *inode);
2391extern void drop_nlink(struct inode *inode);
2392extern void clear_nlink(struct inode *inode);
2393extern void set_nlink(struct inode *inode, unsigned int nlink);
2394
2395static inline void inode_inc_link_count(struct inode *inode)
2396{
2397	inc_nlink(inode);
2398	mark_inode_dirty(inode);
2399}
2400
2401static inline void inode_dec_link_count(struct inode *inode)
2402{
2403	drop_nlink(inode);
2404	mark_inode_dirty(inode);
2405}
2406
2407enum file_time_flags {
2408	S_ATIME = 1,
2409	S_MTIME = 2,
2410	S_CTIME = 4,
2411	S_VERSION = 8,
2412};
2413
2414extern bool atime_needs_update(const struct path *, struct inode *);
2415extern void touch_atime(const struct path *);
2416int inode_update_time(struct inode *inode, struct timespec64 *time, int flags);
2417
2418static inline void file_accessed(struct file *file)
2419{
2420	if (!(file->f_flags & O_NOATIME))
2421		touch_atime(&file->f_path);
2422}
2423
2424extern int file_modified(struct file *file);
2425
2426int sync_inode_metadata(struct inode *inode, int wait);
2427
2428struct file_system_type {
2429	const char *name;
2430	int fs_flags;
2431#define FS_REQUIRES_DEV		1
2432#define FS_BINARY_MOUNTDATA	2
2433#define FS_HAS_SUBTYPE		4
2434#define FS_USERNS_MOUNT		8	/* Can be mounted by userns root */
2435#define FS_DISALLOW_NOTIFY_PERM	16	/* Disable fanotify permission events */
2436#define FS_ALLOW_IDMAP         32      /* FS has been updated to handle vfs idmappings. */
2437#define FS_RENAME_DOES_D_MOVE	32768	/* FS will handle d_move() during rename() internally. */
2438	int (*init_fs_context)(struct fs_context *);
2439	const struct fs_parameter_spec *parameters;
2440	struct dentry *(*mount) (struct file_system_type *, int,
2441		       const char *, void *);
2442	void (*kill_sb) (struct super_block *);
2443	struct module *owner;
2444	struct file_system_type * next;
2445	struct hlist_head fs_supers;
2446
2447	struct lock_class_key s_lock_key;
2448	struct lock_class_key s_umount_key;
2449	struct lock_class_key s_vfs_rename_key;
2450	struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2451
2452	struct lock_class_key i_lock_key;
2453	struct lock_class_key i_mutex_key;
2454	struct lock_class_key invalidate_lock_key;
2455	struct lock_class_key i_mutex_dir_key;
2456};
2457
2458#define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2459
2460extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2461	int flags, const char *dev_name, void *data,
2462	int (*fill_super)(struct super_block *, void *, int));
2463extern struct dentry *mount_single(struct file_system_type *fs_type,
2464	int flags, void *data,
2465	int (*fill_super)(struct super_block *, void *, int));
2466extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2467	int flags, void *data,
2468	int (*fill_super)(struct super_block *, void *, int));
2469extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2470void generic_shutdown_super(struct super_block *sb);
2471void kill_block_super(struct super_block *sb);
2472void kill_anon_super(struct super_block *sb);
2473void kill_litter_super(struct super_block *sb);
2474void deactivate_super(struct super_block *sb);
2475void deactivate_locked_super(struct super_block *sb);
2476int set_anon_super(struct super_block *s, void *data);
2477int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2478int get_anon_bdev(dev_t *);
2479void free_anon_bdev(dev_t);
2480struct super_block *sget_fc(struct fs_context *fc,
2481			    int (*test)(struct super_block *, struct fs_context *),
2482			    int (*set)(struct super_block *, struct fs_context *));
2483struct super_block *sget(struct file_system_type *type,
2484			int (*test)(struct super_block *,void *),
2485			int (*set)(struct super_block *,void *),
2486			int flags, void *data);
2487
2488/* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2489#define fops_get(fops) \
2490	(((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2491#define fops_put(fops) \
2492	do { if (fops) module_put((fops)->owner); } while(0)
2493/*
2494 * This one is to be used *ONLY* from ->open() instances.
2495 * fops must be non-NULL, pinned down *and* module dependencies
2496 * should be sufficient to pin the caller down as well.
2497 */
2498#define replace_fops(f, fops) \
2499	do {	\
2500		struct file *__file = (f); \
2501		fops_put(__file->f_op); \
2502		BUG_ON(!(__file->f_op = (fops))); \
2503	} while(0)
2504
2505extern int register_filesystem(struct file_system_type *);
2506extern int unregister_filesystem(struct file_system_type *);
2507extern struct vfsmount *kern_mount(struct file_system_type *);
2508extern void kern_unmount(struct vfsmount *mnt);
2509extern int may_umount_tree(struct vfsmount *);
2510extern int may_umount(struct vfsmount *);
2511extern long do_mount(const char *, const char __user *,
2512		     const char *, unsigned long, void *);
2513extern struct vfsmount *collect_mounts(const struct path *);
2514extern void drop_collected_mounts(struct vfsmount *);
2515extern int iterate_mounts(int (*)(struct vfsmount *, void *), void *,
2516			  struct vfsmount *);
2517extern int vfs_statfs(const struct path *, struct kstatfs *);
2518extern int user_statfs(const char __user *, struct kstatfs *);
2519extern int fd_statfs(int, struct kstatfs *);
2520extern int freeze_super(struct super_block *super);
2521extern int thaw_super(struct super_block *super);
2522extern bool our_mnt(struct vfsmount *mnt);
2523extern __printf(2, 3)
2524int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2525extern int super_setup_bdi(struct super_block *sb);
2526
2527extern int current_umask(void);
2528
2529extern void ihold(struct inode * inode);
2530extern void iput(struct inode *);
2531extern int generic_update_time(struct inode *, struct timespec64 *, int);
2532
2533/* /sys/fs */
2534extern struct kobject *fs_kobj;
2535
2536#define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2537
2538#ifdef CONFIG_FILE_LOCKING
2539static inline int break_lease(struct inode *inode, unsigned int mode)
2540{
2541	/*
2542	 * Since this check is lockless, we must ensure that any refcounts
2543	 * taken are done before checking i_flctx->flc_lease. Otherwise, we
2544	 * could end up racing with tasks trying to set a new lease on this
2545	 * file.
2546	 */
2547	smp_mb();
2548	if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2549		return __break_lease(inode, mode, FL_LEASE);
2550	return 0;
2551}
2552
2553static inline int break_deleg(struct inode *inode, unsigned int mode)
2554{
2555	/*
2556	 * Since this check is lockless, we must ensure that any refcounts
2557	 * taken are done before checking i_flctx->flc_lease. Otherwise, we
2558	 * could end up racing with tasks trying to set a new lease on this
2559	 * file.
2560	 */
2561	smp_mb();
2562	if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2563		return __break_lease(inode, mode, FL_DELEG);
2564	return 0;
2565}
2566
2567static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2568{
2569	int ret;
2570
2571	ret = break_deleg(inode, O_WRONLY|O_NONBLOCK);
2572	if (ret == -EWOULDBLOCK && delegated_inode) {
2573		*delegated_inode = inode;
2574		ihold(inode);
2575	}
2576	return ret;
2577}
2578
2579static inline int break_deleg_wait(struct inode **delegated_inode)
2580{
2581	int ret;
2582
2583	ret = break_deleg(*delegated_inode, O_WRONLY);
2584	iput(*delegated_inode);
2585	*delegated_inode = NULL;
2586	return ret;
2587}
2588
2589static inline int break_layout(struct inode *inode, bool wait)
2590{
2591	smp_mb();
2592	if (inode->i_flctx && !list_empty_careful(&inode->i_flctx->flc_lease))
2593		return __break_lease(inode,
2594				wait ? O_WRONLY : O_WRONLY | O_NONBLOCK,
2595				FL_LAYOUT);
2596	return 0;
2597}
2598
2599#else /* !CONFIG_FILE_LOCKING */
2600static inline int break_lease(struct inode *inode, unsigned int mode)
2601{
2602	return 0;
2603}
2604
2605static inline int break_deleg(struct inode *inode, unsigned int mode)
2606{
2607	return 0;
2608}
2609
2610static inline int try_break_deleg(struct inode *inode, struct inode **delegated_inode)
2611{
2612	return 0;
2613}
2614
2615static inline int break_deleg_wait(struct inode **delegated_inode)
2616{
2617	BUG();
2618	return 0;
2619}
2620
2621static inline int break_layout(struct inode *inode, bool wait)
2622{
2623	return 0;
2624}
2625
2626#endif /* CONFIG_FILE_LOCKING */
2627
2628/* fs/open.c */
2629struct audit_names;
2630struct filename {
2631	const char		*name;	/* pointer to actual string */
2632	const __user char	*uptr;	/* original userland pointer */
2633	int			refcnt;
2634	struct audit_names	*aname;
2635	const char		iname[];
2636};
2637static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2638
2639static inline struct user_namespace *file_mnt_user_ns(struct file *file)
2640{
2641	return mnt_user_ns(file->f_path.mnt);
2642}
2643
2644/**
2645 * is_idmapped_mnt - check whether a mount is mapped
2646 * @mnt: the mount to check
2647 *
2648 * If @mnt has an idmapping attached different from the
2649 * filesystem's idmapping then @mnt is mapped.
2650 *
2651 * Return: true if mount is mapped, false if not.
2652 */
2653static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2654{
2655	return mnt_user_ns(mnt) != mnt->mnt_sb->s_user_ns;
2656}
2657
2658extern long vfs_truncate(const struct path *, loff_t);
2659int do_truncate(struct user_namespace *, struct dentry *, loff_t start,
2660		unsigned int time_attrs, struct file *filp);
2661extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2662			loff_t len);
2663extern long do_sys_open(int dfd, const char __user *filename, int flags,
2664			umode_t mode);
2665extern struct file *file_open_name(struct filename *, int, umode_t);
2666extern struct file *filp_open(const char *, int, umode_t);
2667extern struct file *file_open_root(const struct path *,
2668				   const char *, int, umode_t);
2669static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2670				   const char *name, int flags, umode_t mode)
2671{
2672	return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2673			      name, flags, mode);
2674}
2675extern struct file * dentry_open(const struct path *, int, const struct cred *);
2676extern struct file * open_with_fake_path(const struct path *, int,
2677					 struct inode*, const struct cred *);
2678static inline struct file *file_clone_open(struct file *file)
2679{
2680	return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2681}
2682extern int filp_close(struct file *, fl_owner_t id);
2683
2684extern struct filename *getname_flags(const char __user *, int, int *);
2685extern struct filename *getname_uflags(const char __user *, int);
2686extern struct filename *getname(const char __user *);
2687extern struct filename *getname_kernel(const char *);
2688extern void putname(struct filename *name);
2689
2690extern int finish_open(struct file *file, struct dentry *dentry,
2691			int (*open)(struct inode *, struct file *));
2692extern int finish_no_open(struct file *file, struct dentry *dentry);
2693
2694/* fs/dcache.c */
2695extern void __init vfs_caches_init_early(void);
2696extern void __init vfs_caches_init(void);
2697
2698extern struct kmem_cache *names_cachep;
2699
2700#define __getname()		kmem_cache_alloc(names_cachep, GFP_KERNEL)
2701#define __putname(name)		kmem_cache_free(names_cachep, (void *)(name))
2702
2703extern struct super_block *blockdev_superblock;
2704static inline bool sb_is_blkdev_sb(struct super_block *sb)
2705{
2706	return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2707}
2708
2709void emergency_thaw_all(void);
2710extern int sync_filesystem(struct super_block *);
2711extern const struct file_operations def_blk_fops;
2712extern const struct file_operations def_chr_fops;
2713
2714/* fs/char_dev.c */
2715#define CHRDEV_MAJOR_MAX 512
2716/* Marks the bottom of the first segment of free char majors */
2717#define CHRDEV_MAJOR_DYN_END 234
2718/* Marks the top and bottom of the second segment of free char majors */
2719#define CHRDEV_MAJOR_DYN_EXT_START 511
2720#define CHRDEV_MAJOR_DYN_EXT_END 384
2721
2722extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2723extern int register_chrdev_region(dev_t, unsigned, const char *);
2724extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2725			     unsigned int count, const char *name,
2726			     const struct file_operations *fops);
2727extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2728				unsigned int count, const char *name);
2729extern void unregister_chrdev_region(dev_t, unsigned);
2730extern void chrdev_show(struct seq_file *,off_t);
2731
2732static inline int register_chrdev(unsigned int major, const char *name,
2733				  const struct file_operations *fops)
2734{
2735	return __register_chrdev(major, 0, 256, name, fops);
2736}
2737
2738static inline void unregister_chrdev(unsigned int major, const char *name)
2739{
2740	__unregister_chrdev(major, 0, 256, name);
2741}
2742
2743extern void init_special_inode(struct inode *, umode_t, dev_t);
2744
2745/* Invalid inode operations -- fs/bad_inode.c */
2746extern void make_bad_inode(struct inode *);
2747extern bool is_bad_inode(struct inode *);
2748
2749unsigned long invalidate_mapping_pages(struct address_space *mapping,
2750					pgoff_t start, pgoff_t end);
2751
2752void invalidate_mapping_pagevec(struct address_space *mapping,
2753				pgoff_t start, pgoff_t end,
2754				unsigned long *nr_pagevec);
2755
2756static inline void invalidate_remote_inode(struct inode *inode)
2757{
2758	if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
2759	    S_ISLNK(inode->i_mode))
2760		invalidate_mapping_pages(inode->i_mapping, 0, -1);
2761}
2762extern int invalidate_inode_pages2(struct address_space *mapping);
2763extern int invalidate_inode_pages2_range(struct address_space *mapping,
2764					 pgoff_t start, pgoff_t end);
2765extern int write_inode_now(struct inode *, int);
2766extern int filemap_fdatawrite(struct address_space *);
2767extern int filemap_flush(struct address_space *);
2768extern int filemap_fdatawait_keep_errors(struct address_space *mapping);
2769extern int filemap_fdatawait_range(struct address_space *, loff_t lstart,
2770				   loff_t lend);
2771extern int filemap_fdatawait_range_keep_errors(struct address_space *mapping,
2772		loff_t start_byte, loff_t end_byte);
2773
2774static inline int filemap_fdatawait(struct address_space *mapping)
2775{
2776	return filemap_fdatawait_range(mapping, 0, LLONG_MAX);
2777}
2778
2779extern bool filemap_range_has_page(struct address_space *, loff_t lstart,
2780				  loff_t lend);
2781extern int filemap_write_and_wait_range(struct address_space *mapping,
2782				        loff_t lstart, loff_t lend);
2783extern int __filemap_fdatawrite_range(struct address_space *mapping,
2784				loff_t start, loff_t end, int sync_mode);
2785extern int filemap_fdatawrite_range(struct address_space *mapping,
2786				loff_t start, loff_t end);
2787extern int filemap_check_errors(struct address_space *mapping);
2788extern void __filemap_set_wb_err(struct address_space *mapping, int err);
2789int filemap_fdatawrite_wbc(struct address_space *mapping,
2790			   struct writeback_control *wbc);
2791
2792static inline int filemap_write_and_wait(struct address_space *mapping)
2793{
2794	return filemap_write_and_wait_range(mapping, 0, LLONG_MAX);
2795}
2796
2797extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2798						loff_t lend);
2799extern int __must_check file_check_and_advance_wb_err(struct file *file);
2800extern int __must_check file_write_and_wait_range(struct file *file,
2801						loff_t start, loff_t end);
2802
2803static inline int file_write_and_wait(struct file *file)
2804{
2805	return file_write_and_wait_range(file, 0, LLONG_MAX);
2806}
2807
2808/**
2809 * filemap_set_wb_err - set a writeback error on an address_space
2810 * @mapping: mapping in which to set writeback error
2811 * @err: error to be set in mapping
2812 *
2813 * When writeback fails in some way, we must record that error so that
2814 * userspace can be informed when fsync and the like are called.  We endeavor
2815 * to report errors on any file that was open at the time of the error.  Some
2816 * internal callers also need to know when writeback errors have occurred.
2817 *
2818 * When a writeback error occurs, most filesystems will want to call
2819 * filemap_set_wb_err to record the error in the mapping so that it will be
2820 * automatically reported whenever fsync is called on the file.
2821 */
2822static inline void filemap_set_wb_err(struct address_space *mapping, int err)
2823{
2824	/* Fastpath for common case of no error */
2825	if (unlikely(err))
2826		__filemap_set_wb_err(mapping, err);
2827}
2828
2829/**
2830 * filemap_check_wb_err - has an error occurred since the mark was sampled?
2831 * @mapping: mapping to check for writeback errors
2832 * @since: previously-sampled errseq_t
2833 *
2834 * Grab the errseq_t value from the mapping, and see if it has changed "since"
2835 * the given value was sampled.
2836 *
2837 * If it has then report the latest error set, otherwise return 0.
2838 */
2839static inline int filemap_check_wb_err(struct address_space *mapping,
2840					errseq_t since)
2841{
2842	return errseq_check(&mapping->wb_err, since);
2843}
2844
2845/**
2846 * filemap_sample_wb_err - sample the current errseq_t to test for later errors
2847 * @mapping: mapping to be sampled
2848 *
2849 * Writeback errors are always reported relative to a particular sample point
2850 * in the past. This function provides those sample points.
2851 */
2852static inline errseq_t filemap_sample_wb_err(struct address_space *mapping)
2853{
2854	return errseq_sample(&mapping->wb_err);
2855}
2856
2857/**
2858 * file_sample_sb_err - sample the current errseq_t to test for later errors
2859 * @file: file pointer to be sampled
2860 *
2861 * Grab the most current superblock-level errseq_t value for the given
2862 * struct file.
2863 */
2864static inline errseq_t file_sample_sb_err(struct file *file)
2865{
2866	return errseq_sample(&file->f_path.dentry->d_sb->s_wb_err);
2867}
2868
2869extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2870			   int datasync);
2871extern int vfs_fsync(struct file *file, int datasync);
2872
2873extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2874				unsigned int flags);
2875
2876/*
2877 * Sync the bytes written if this was a synchronous write.  Expect ki_pos
2878 * to already be updated for the write, and will return either the amount
2879 * of bytes passed in, or an error if syncing the file failed.
2880 */
2881static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2882{
2883	if (iocb->ki_flags & IOCB_DSYNC) {
2884		int ret = vfs_fsync_range(iocb->ki_filp,
2885				iocb->ki_pos - count, iocb->ki_pos - 1,
2886				(iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2887		if (ret)
2888			return ret;
2889	}
2890
2891	return count;
2892}
2893
2894extern void emergency_sync(void);
2895extern void emergency_remount(void);
2896
2897#ifdef CONFIG_BLOCK
2898extern int bmap(struct inode *inode, sector_t *block);
2899#else
2900static inline int bmap(struct inode *inode,  sector_t *block)
2901{
2902	return -EINVAL;
2903}
2904#endif
2905
2906int notify_change(struct user_namespace *, struct dentry *,
2907		  struct iattr *, struct inode **);
2908int inode_permission(struct user_namespace *, struct inode *, int);
2909int generic_permission(struct user_namespace *, struct inode *, int);
2910static inline int file_permission(struct file *file, int mask)
2911{
2912	return inode_permission(file_mnt_user_ns(file),
2913				file_inode(file), mask);
2914}
2915static inline int path_permission(const struct path *path, int mask)
2916{
2917	return inode_permission(mnt_user_ns(path->mnt),
2918				d_inode(path->dentry), mask);
2919}
2920int __check_sticky(struct user_namespace *mnt_userns, struct inode *dir,
2921		   struct inode *inode);
2922
2923static inline bool execute_ok(struct inode *inode)
2924{
2925	return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2926}
2927
2928static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2929{
2930	return (inode->i_mode ^ mode) & S_IFMT;
2931}
2932
2933static inline void file_start_write(struct file *file)
2934{
2935	if (!S_ISREG(file_inode(file)->i_mode))
2936		return;
2937	sb_start_write(file_inode(file)->i_sb);
2938}
2939
2940static inline bool file_start_write_trylock(struct file *file)
2941{
2942	if (!S_ISREG(file_inode(file)->i_mode))
2943		return true;
2944	return sb_start_write_trylock(file_inode(file)->i_sb);
2945}
2946
2947static inline void file_end_write(struct file *file)
2948{
2949	if (!S_ISREG(file_inode(file)->i_mode))
2950		return;
2951	__sb_end_write(file_inode(file)->i_sb, SB_FREEZE_WRITE);
2952}
2953
2954/*
2955 * This is used for regular files where some users -- especially the
2956 * currently executed binary in a process, previously handled via
2957 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2958 * read-write shared) accesses.
2959 *
2960 * get_write_access() gets write permission for a file.
2961 * put_write_access() releases this write permission.
2962 * deny_write_access() denies write access to a file.
2963 * allow_write_access() re-enables write access to a file.
2964 *
2965 * The i_writecount field of an inode can have the following values:
2966 * 0: no write access, no denied write access
2967 * < 0: (-i_writecount) users that denied write access to the file.
2968 * > 0: (i_writecount) users that have write access to the file.
2969 *
2970 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2971 * except for the cases where we don't hold i_writecount yet. Then we need to
2972 * use {get,deny}_write_access() - these functions check the sign and refuse
2973 * to do the change if sign is wrong.
2974 */
2975static inline int get_write_access(struct inode *inode)
2976{
2977	return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2978}
2979static inline int deny_write_access(struct file *file)
2980{
2981	struct inode *inode = file_inode(file);
2982	return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2983}
2984static inline void put_write_access(struct inode * inode)
2985{
2986	atomic_dec(&inode->i_writecount);
2987}
2988static inline void allow_write_access(struct file *file)
2989{
2990	if (file)
2991		atomic_inc(&file_inode(file)->i_writecount);
2992}
2993static inline bool inode_is_open_for_write(const struct inode *inode)
2994{
2995	return atomic_read(&inode->i_writecount) > 0;
2996}
2997
2998#if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
2999static inline void i_readcount_dec(struct inode *inode)
3000{
3001	BUG_ON(!atomic_read(&inode->i_readcount));
3002	atomic_dec(&inode->i_readcount);
3003}
3004static inline void i_readcount_inc(struct inode *inode)
3005{
3006	atomic_inc(&inode->i_readcount);
3007}
3008#else
3009static inline void i_readcount_dec(struct inode *inode)
3010{
3011	return;
3012}
3013static inline void i_readcount_inc(struct inode *inode)
3014{
3015	return;
3016}
3017#endif
3018extern int do_pipe_flags(int *, int);
3019
3020extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
3021ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
3022extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
3023extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
3024extern struct file * open_exec(const char *);
3025
3026/* fs/dcache.c -- generic fs support functions */
3027extern bool is_subdir(struct dentry *, struct dentry *);
3028extern bool path_is_under(const struct path *, const struct path *);
3029
3030extern char *file_path(struct file *, char *, int);
3031
3032#include <linux/err.h>
3033
3034/* needed for stackable file system support */
3035extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
3036
3037extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
3038
3039extern int inode_init_always(struct super_block *, struct inode *);
3040extern void inode_init_once(struct inode *);
3041extern void address_space_init_once(struct address_space *mapping);
3042extern struct inode * igrab(struct inode *);
3043extern ino_t iunique(struct super_block *, ino_t);
3044extern int inode_needs_sync(struct inode *inode);
3045extern int generic_delete_inode(struct inode *inode);
3046static inline int generic_drop_inode(struct inode *inode)
3047{
3048	return !inode->i_nlink || inode_unhashed(inode);
3049}
3050extern void d_mark_dontcache(struct inode *inode);
3051
3052extern struct inode *ilookup5_nowait(struct super_block *sb,
3053		unsigned long hashval, int (*test)(struct inode *, void *),
3054		void *data);
3055extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
3056		int (*test)(struct inode *, void *), void *data);
3057extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
3058
3059extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
3060		int (*test)(struct inode *, void *),
3061		int (*set)(struct inode *, void *),
3062		void *data);
3063extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
3064extern struct inode * iget_locked(struct super_block *, unsigned long);
3065extern struct inode *find_inode_nowait(struct super_block *,
3066				       unsigned long,
3067				       int (*match)(struct inode *,
3068						    unsigned long, void *),
3069				       void *data);
3070extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
3071				    int (*)(struct inode *, void *), void *);
3072extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
3073extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
3074extern int insert_inode_locked(struct inode *);
3075#ifdef CONFIG_DEBUG_LOCK_ALLOC
3076extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
3077#else
3078static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
3079#endif
3080extern void unlock_new_inode(struct inode *);
3081extern void discard_new_inode(struct inode *);
3082extern unsigned int get_next_ino(void);
3083extern void evict_inodes(struct super_block *sb);
3084void dump_mapping(const struct address_space *);
3085
3086/*
3087 * Userspace may rely on the the inode number being non-zero. For example, glibc
3088 * simply ignores files with zero i_ino in unlink() and other places.
3089 *
3090 * As an additional complication, if userspace was compiled with
3091 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
3092 * lower 32 bits, so we need to check that those aren't zero explicitly. With
3093 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
3094 * better safe than sorry.
3095 */
3096static inline bool is_zero_ino(ino_t ino)
3097{
3098	return (u32)ino == 0;
3099}
3100
3101extern void __iget(struct inode * inode);
3102extern void iget_failed(struct inode *);
3103extern void clear_inode(struct inode *);
3104extern void __destroy_inode(struct inode *);
3105extern struct inode *new_inode_pseudo(struct super_block *sb);
3106extern struct inode *new_inode(struct super_block *sb);
3107extern void free_inode_nonrcu(struct inode *inode);
3108extern int should_remove_suid(struct dentry *);
3109extern int file_remove_privs(struct file *);
3110
3111extern void __insert_inode_hash(struct inode *, unsigned long hashval);
3112static inline void insert_inode_hash(struct inode *inode)
3113{
3114	__insert_inode_hash(inode, inode->i_ino);
3115}
3116
3117extern void __remove_inode_hash(struct inode *);
3118static inline void remove_inode_hash(struct inode *inode)
3119{
3120	if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
3121		__remove_inode_hash(inode);
3122}
3123
3124extern void inode_sb_list_add(struct inode *inode);
3125extern void inode_add_lru(struct inode *inode);
3126
3127extern int sb_set_blocksize(struct super_block *, int);
3128extern int sb_min_blocksize(struct super_block *, int);
3129
3130extern int generic_file_mmap(struct file *, struct vm_area_struct *);
3131extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
3132extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
3133extern int generic_write_check_limits(struct file *file, loff_t pos,
3134		loff_t *count);
3135extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
3136ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
3137		ssize_t already_read);
3138extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
3139extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
3140extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
3141extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
3142extern ssize_t generic_perform_write(struct file *, struct iov_iter *, loff_t);
3143
3144ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
3145		rwf_t flags);
3146ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
3147		rwf_t flags);
3148ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
3149			   struct iov_iter *iter);
3150ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
3151			    struct iov_iter *iter);
3152
3153/* fs/splice.c */
3154extern ssize_t generic_file_splice_read(struct file *, loff_t *,
3155		struct pipe_inode_info *, size_t, unsigned int);
3156extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
3157		struct file *, loff_t *, size_t, unsigned int);
3158extern ssize_t generic_splice_sendpage(struct pipe_inode_info *pipe,
3159		struct file *out, loff_t *, size_t len, unsigned int flags);
3160extern long do_splice_direct(struct file *in, loff_t *ppos, struct file *out,
3161		loff_t *opos, size_t len, unsigned int flags);
3162
3163
3164extern void
3165file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
3166extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3167extern loff_t no_llseek(struct file *file, loff_t offset, int whence);
3168extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3169extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3170extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3171		int whence, loff_t maxsize, loff_t eof);
3172extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3173		int whence, loff_t size);
3174extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3175extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3176extern int generic_file_open(struct inode * inode, struct file * filp);
3177extern int nonseekable_open(struct inode * inode, struct file * filp);
3178extern int stream_open(struct inode * inode, struct file * filp);
3179
3180#ifdef CONFIG_BLOCK
3181typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
3182			    loff_t file_offset);
3183
3184enum {
3185	/* need locking between buffered and direct access */
3186	DIO_LOCKING	= 0x01,
3187
3188	/* filesystem does not support filling holes */
3189	DIO_SKIP_HOLES	= 0x02,
3190};
3191
3192ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
3193			     struct block_device *bdev, struct iov_iter *iter,
3194			     get_block_t get_block,
3195			     dio_iodone_t end_io, dio_submit_t submit_io,
3196			     int flags);
3197
3198static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3199					 struct inode *inode,
3200					 struct iov_iter *iter,
3201					 get_block_t get_block)
3202{
3203	return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
3204			get_block, NULL, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
3205}
3206#endif
3207
3208void inode_dio_wait(struct inode *inode);
3209
3210/**
3211 * inode_dio_begin - signal start of a direct I/O requests
3212 * @inode: inode the direct I/O happens on
3213 *
3214 * This is called once we've finished processing a direct I/O request,
3215 * and is used to wake up callers waiting for direct I/O to be quiesced.
3216 */
3217static inline void inode_dio_begin(struct inode *inode)
3218{
3219	atomic_inc(&inode->i_dio_count);
3220}
3221
3222/**
3223 * inode_dio_end - signal finish of a direct I/O requests
3224 * @inode: inode the direct I/O happens on
3225 *
3226 * This is called once we've finished processing a direct I/O request,
3227 * and is used to wake up callers waiting for direct I/O to be quiesced.
3228 */
3229static inline void inode_dio_end(struct inode *inode)
3230{
3231	if (atomic_dec_and_test(&inode->i_dio_count))
3232		wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
3233}
3234
3235/*
3236 * Warn about a page cache invalidation failure diring a direct I/O write.
3237 */
3238void dio_warn_stale_pagecache(struct file *filp);
3239
3240extern void inode_set_flags(struct inode *inode, unsigned int flags,
3241			    unsigned int mask);
3242
3243extern const struct file_operations generic_ro_fops;
3244
3245#define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
3246
3247extern int readlink_copy(char __user *, int, const char *);
3248extern int page_readlink(struct dentry *, char __user *, int);
3249extern const char *page_get_link(struct dentry *, struct inode *,
3250				 struct delayed_call *);
3251extern void page_put_link(void *);
3252extern int __page_symlink(struct inode *inode, const char *symname, int len,
3253		int nofs);
3254extern int page_symlink(struct inode *inode, const char *symname, int len);
3255extern const struct inode_operations page_symlink_inode_operations;
3256extern void kfree_link(void *);
3257void generic_fillattr(struct user_namespace *, struct inode *, struct kstat *);
3258void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
3259extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
3260extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
3261void __inode_add_bytes(struct inode *inode, loff_t bytes);
3262void inode_add_bytes(struct inode *inode, loff_t bytes);
3263void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3264void inode_sub_bytes(struct inode *inode, loff_t bytes);
3265static inline loff_t __inode_get_bytes(struct inode *inode)
3266{
3267	return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3268}
3269loff_t inode_get_bytes(struct inode *inode);
3270void inode_set_bytes(struct inode *inode, loff_t bytes);
3271const char *simple_get_link(struct dentry *, struct inode *,
3272			    struct delayed_call *);
3273extern const struct inode_operations simple_symlink_inode_operations;
3274
3275extern int iterate_dir(struct file *, struct dir_context *);
3276
3277int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
3278		int flags);
3279int vfs_fstat(int fd, struct kstat *stat);
3280
3281static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3282{
3283	return vfs_fstatat(AT_FDCWD, filename, stat, 0);
3284}
3285static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3286{
3287	return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
3288}
3289
3290extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3291extern int vfs_readlink(struct dentry *, char __user *, int);
3292
3293extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3294extern void put_filesystem(struct file_system_type *fs);
3295extern struct file_system_type *get_fs_type(const char *name);
3296extern struct super_block *get_super(struct block_device *);
3297extern struct super_block *get_active_super(struct block_device *bdev);
3298extern void drop_super(struct super_block *sb);
3299extern void drop_super_exclusive(struct super_block *sb);
3300extern void iterate_supers(void (*)(struct super_block *, void *), void *);
3301extern void iterate_supers_type(struct file_system_type *,
3302			        void (*)(struct super_block *, void *), void *);
3303
3304extern int dcache_dir_open(struct inode *, struct file *);
3305extern int dcache_dir_close(struct inode *, struct file *);
3306extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3307extern int dcache_readdir(struct file *, struct dir_context *);
3308extern int simple_setattr(struct user_namespace *, struct dentry *,
3309			  struct iattr *);
3310extern int simple_getattr(struct user_namespace *, const struct path *,
3311			  struct kstat *, u32, unsigned int);
3312extern int simple_statfs(struct dentry *, struct kstatfs *);
3313extern int simple_open(struct inode *inode, struct file *file);
3314extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3315extern int simple_unlink(struct inode *, struct dentry *);
3316extern int simple_rmdir(struct inode *, struct dentry *);
3317extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
3318				  struct inode *new_dir, struct dentry *new_dentry);
3319extern int simple_rename(struct user_namespace *, struct inode *,
3320			 struct dentry *, struct inode *, struct dentry *,
3321			 unsigned int);
3322extern void simple_recursive_removal(struct dentry *,
3323                              void (*callback)(struct dentry *));
3324extern int noop_fsync(struct file *, loff_t, loff_t, int);
3325extern void noop_invalidatepage(struct page *page, unsigned int offset,
3326		unsigned int length);
3327extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
3328extern int simple_empty(struct dentry *);
3329extern int simple_write_begin(struct file *file, struct address_space *mapping,
3330			loff_t pos, unsigned len, unsigned flags,
3331			struct page **pagep, void **fsdata);
3332extern const struct address_space_operations ram_aops;
3333extern int always_delete_dentry(const struct dentry *);
3334extern struct inode *alloc_anon_inode(struct super_block *);
3335extern int simple_nosetlease(struct file *, long, struct file_lock **, void **);
3336extern const struct dentry_operations simple_dentry_operations;
3337
3338extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3339extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3340extern const struct file_operations simple_dir_operations;
3341extern const struct inode_operations simple_dir_inode_operations;
3342extern void make_empty_dir_inode(struct inode *inode);
3343extern bool is_empty_dir_inode(struct inode *inode);
3344struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3345struct dentry *d_alloc_name(struct dentry *, const char *);
3346extern int simple_fill_super(struct super_block *, unsigned long,
3347			     const struct tree_descr *);
3348extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3349extern void simple_release_fs(struct vfsmount **mount, int *count);
3350
3351extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3352			loff_t *ppos, const void *from, size_t available);
3353extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3354		const void __user *from, size_t count);
3355
3356extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3357extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3358
3359extern int generic_check_addressable(unsigned, u64);
3360
3361extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry);
3362
3363#ifdef CONFIG_MIGRATION
3364extern int buffer_migrate_page(struct address_space *,
3365				struct page *, struct page *,
3366				enum migrate_mode);
3367extern int buffer_migrate_page_norefs(struct address_space *,
3368				struct page *, struct page *,
3369				enum migrate_mode);
3370#else
3371#define buffer_migrate_page NULL
3372#define buffer_migrate_page_norefs NULL
3373#endif
3374
3375int may_setattr(struct user_namespace *mnt_userns, struct inode *inode,
3376		unsigned int ia_valid);
3377int setattr_prepare(struct user_namespace *, struct dentry *, struct iattr *);
3378extern int inode_newsize_ok(const struct inode *, loff_t offset);
3379void setattr_copy(struct user_namespace *, struct inode *inode,
3380		  const struct iattr *attr);
3381
3382extern int file_update_time(struct file *file);
3383
3384static inline bool vma_is_dax(const struct vm_area_struct *vma)
3385{
3386	return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3387}
3388
3389static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3390{
3391	struct inode *inode;
3392
3393	if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
3394		return false;
3395	if (!vma_is_dax(vma))
3396		return false;
3397	inode = file_inode(vma->vm_file);
3398	if (S_ISCHR(inode->i_mode))
3399		return false; /* device-dax */
3400	return true;
3401}
3402
3403static inline int iocb_flags(struct file *file)
3404{
3405	int res = 0;
3406	if (file->f_flags & O_APPEND)
3407		res |= IOCB_APPEND;
3408	if (file->f_flags & O_DIRECT)
3409		res |= IOCB_DIRECT;
3410	if ((file->f_flags & O_DSYNC) || IS_SYNC(file->f_mapping->host))
3411		res |= IOCB_DSYNC;
3412	if (file->f_flags & __O_SYNC)
3413		res |= IOCB_SYNC;
3414	return res;
3415}
3416
3417static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
3418{
3419	int kiocb_flags = 0;
3420
3421	/* make sure there's no overlap between RWF and private IOCB flags */
3422	BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3423
3424	if (!flags)
3425		return 0;
3426	if (unlikely(flags & ~RWF_SUPPORTED))
3427		return -EOPNOTSUPP;
3428
3429	if (flags & RWF_NOWAIT) {
3430		if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3431			return -EOPNOTSUPP;
3432		kiocb_flags |= IOCB_NOIO;
3433	}
3434	kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3435	if (flags & RWF_SYNC)
3436		kiocb_flags |= IOCB_DSYNC;
3437
3438	ki->ki_flags |= kiocb_flags;
3439	return 0;
3440}
3441
3442static inline ino_t parent_ino(struct dentry *dentry)
3443{
3444	ino_t res;
3445
3446	/*
3447	 * Don't strictly need d_lock here? If the parent ino could change
3448	 * then surely we'd have a deeper race in the caller?
3449	 */
3450	spin_lock(&dentry->d_lock);
3451	res = dentry->d_parent->d_inode->i_ino;
3452	spin_unlock(&dentry->d_lock);
3453	return res;
3454}
3455
3456/* Transaction based IO helpers */
3457
3458/*
3459 * An argresp is stored in an allocated page and holds the
3460 * size of the argument or response, along with its content
3461 */
3462struct simple_transaction_argresp {
3463	ssize_t size;
3464	char data[];
3465};
3466
3467#define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3468
3469char *simple_transaction_get(struct file *file, const char __user *buf,
3470				size_t size);
3471ssize_t simple_transaction_read(struct file *file, char __user *buf,
3472				size_t size, loff_t *pos);
3473int simple_transaction_release(struct inode *inode, struct file *file);
3474
3475void simple_transaction_set(struct file *file, size_t n);
3476
3477/*
3478 * simple attribute files
3479 *
3480 * These attributes behave similar to those in sysfs:
3481 *
3482 * Writing to an attribute immediately sets a value, an open file can be
3483 * written to multiple times.
3484 *
3485 * Reading from an attribute creates a buffer from the value that might get
3486 * read with multiple read calls. When the attribute has been read
3487 * completely, no further read calls are possible until the file is opened
3488 * again.
3489 *
3490 * All attributes contain a text representation of a numeric value
3491 * that are accessed with the get() and set() functions.
3492 */
3493#define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt)		\
3494static int __fops ## _open(struct inode *inode, struct file *file)	\
3495{									\
3496	__simple_attr_check_format(__fmt, 0ull);			\
3497	return simple_attr_open(inode, file, __get, __set, __fmt);	\
3498}									\
3499static const struct file_operations __fops = {				\
3500	.owner	 = THIS_MODULE,						\
3501	.open	 = __fops ## _open,					\
3502	.release = simple_attr_release,					\
3503	.read	 = simple_attr_read,					\
3504	.write	 = simple_attr_write,					\
3505	.llseek	 = generic_file_llseek,					\
3506}
3507
3508static inline __printf(1, 2)
3509void __simple_attr_check_format(const char *fmt, ...)
3510{
3511	/* don't do anything, just let the compiler check the arguments; */
3512}
3513
3514int simple_attr_open(struct inode *inode, struct file *file,
3515		     int (*get)(void *, u64 *), int (*set)(void *, u64),
3516		     const char *fmt);
3517int simple_attr_release(struct inode *inode, struct file *file);
3518ssize_t simple_attr_read(struct file *file, char __user *buf,
3519			 size_t len, loff_t *ppos);
3520ssize_t simple_attr_write(struct file *file, const char __user *buf,
3521			  size_t len, loff_t *ppos);
3522
3523struct ctl_table;
3524int __init list_bdev_fs_names(char *buf, size_t size);
3525
3526#define __FMODE_EXEC		((__force int) FMODE_EXEC)
3527#define __FMODE_NONOTIFY	((__force int) FMODE_NONOTIFY)
3528
3529#define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3530#define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3531					    (flag & __FMODE_NONOTIFY)))
3532
3533static inline bool is_sxid(umode_t mode)
3534{
3535	return (mode & S_ISUID) || ((mode & S_ISGID) && (mode & S_IXGRP));
3536}
3537
3538static inline int check_sticky(struct user_namespace *mnt_userns,
3539			       struct inode *dir, struct inode *inode)
3540{
3541	if (!(dir->i_mode & S_ISVTX))
3542		return 0;
3543
3544	return __check_sticky(mnt_userns, dir, inode);
3545}
3546
3547static inline void inode_has_no_xattr(struct inode *inode)
3548{
3549	if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3550		inode->i_flags |= S_NOSEC;
3551}
3552
3553static inline bool is_root_inode(struct inode *inode)
3554{
3555	return inode == inode->i_sb->s_root->d_inode;
3556}
3557
3558static inline bool dir_emit(struct dir_context *ctx,
3559			    const char *name, int namelen,
3560			    u64 ino, unsigned type)
3561{
3562	return ctx->actor(ctx, name, namelen, ctx->pos, ino, type) == 0;
3563}
3564static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3565{
3566	return ctx->actor(ctx, ".", 1, ctx->pos,
3567			  file->f_path.dentry->d_inode->i_ino, DT_DIR) == 0;
3568}
3569static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3570{
3571	return ctx->actor(ctx, "..", 2, ctx->pos,
3572			  parent_ino(file->f_path.dentry), DT_DIR) == 0;
3573}
3574static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3575{
3576	if (ctx->pos == 0) {
3577		if (!dir_emit_dot(file, ctx))
3578			return false;
3579		ctx->pos = 1;
3580	}
3581	if (ctx->pos == 1) {
3582		if (!dir_emit_dotdot(file, ctx))
3583			return false;
3584		ctx->pos = 2;
3585	}
3586	return true;
3587}
3588static inline bool dir_relax(struct inode *inode)
3589{
3590	inode_unlock(inode);
3591	inode_lock(inode);
3592	return !IS_DEADDIR(inode);
3593}
3594
3595static inline bool dir_relax_shared(struct inode *inode)
3596{
3597	inode_unlock_shared(inode);
3598	inode_lock_shared(inode);
3599	return !IS_DEADDIR(inode);
3600}
3601
3602extern bool path_noexec(const struct path *path);
3603extern void inode_nohighmem(struct inode *inode);
3604
3605/* mm/fadvise.c */
3606extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3607		       int advice);
3608extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3609			   int advice);
3610
3611/*
3612 * Flush file data before changing attributes.  Caller must hold any locks
3613 * required to prevent further writes to this file until we're done setting
3614 * flags.
3615 */
3616static inline int inode_drain_writes(struct inode *inode)
3617{
3618	inode_dio_wait(inode);
3619	return filemap_write_and_wait(inode->i_mapping);
3620}
3621
3622#endif /* _LINUX_FS_H */
3623