linux - include/linux/fs.h

3450 lines · 113.4 KiB

1	/* SPDX-License-Identifier: GPL-2.0 */
2	#ifndef _LINUX_FS_H
3	#define _LINUX_FS_H
4
5	#include <linux/linkage.h>
6	#include <linux/wait_bit.h>
7	#include <linux/kdev_t.h>
8	#include <linux/dcache.h>
9	#include <linux/path.h>
10	#include <linux/stat.h>
11	#include <linux/cache.h>
12	#include <linux/list.h>
13	#include <linux/list_lru.h>
14	#include <linux/llist.h>
15	#include <linux/radix-tree.h>
16	#include <linux/xarray.h>
17	#include <linux/rbtree.h>
18	#include <linux/init.h>
19	#include <linux/pid.h>
20	#include <linux/bug.h>
21	#include <linux/mutex.h>
22	#include <linux/rwsem.h>
23	#include <linux/mm_types.h>
24	#include <linux/capability.h>
25	#include <linux/semaphore.h>
26	#include <linux/fcntl.h>
27	#include <linux/rculist_bl.h>
28	#include <linux/atomic.h>
29	#include <linux/shrinker.h>
30	#include <linux/migrate_mode.h>
31	#include <linux/uidgid.h>
32	#include <linux/lockdep.h>
33	#include <linux/percpu-rwsem.h>
34	#include <linux/workqueue.h>
35	#include <linux/delayed_call.h>
36	#include <linux/uuid.h>
37	#include <linux/errseq.h>
38	#include <linux/ioprio.h>
39	#include <linux/fs_types.h>
40	#include <linux/build_bug.h>
41	#include <linux/stddef.h>
42	#include <linux/mount.h>
43	#include <linux/cred.h>
44	#include <linux/mnt_idmapping.h>
45	#include <linux/slab.h>
46
47	#include <asm/byteorder.h>
48	#include <uapi/linux/fs.h>
49
50	struct backing_dev_info;
51	struct bdi_writeback;
52	struct bio;
53	struct io_comp_batch;
54	struct export_operations;
55	struct fiemap_extent_info;
56	struct hd_geometry;
57	struct iovec;
58	struct kiocb;
59	struct kobject;
60	struct pipe_inode_info;
61	struct poll_table_struct;
62	struct kstatfs;
63	struct vm_area_struct;
64	struct vfsmount;
65	struct cred;
66	struct swap_info_struct;
67	struct seq_file;
68	struct workqueue_struct;
69	struct iov_iter;
70	struct fscrypt_inode_info;
71	struct fscrypt_operations;
72	struct fsverity_info;
73	struct fsverity_operations;
74	struct fs_context;
75	struct fs_parameter_spec;
76	struct fileattr;
77	struct iomap_ops;
78
79	extern void __init inode_init(void);
80	extern void __init inode_init_early(void);
81	extern void __init files_init(void);
82	extern void __init files_maxfiles_init(void);
83
84	extern unsigned long get_max_files(void);
85	extern unsigned int sysctl_nr_open;
86
87	typedef __kernel_rwf_t rwf_t;
88
89	struct buffer_head;
90	typedef int (get_block_t)(struct inode *inode, sector_t iblock,
91	struct buffer_head *bh_result, int create);
92	typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
93	ssize_t bytes, void *private);
94
95	#define MAY_EXEC 0x00000001
96	#define MAY_WRITE 0x00000002
97	#define MAY_READ 0x00000004
98	#define MAY_APPEND 0x00000008
99	#define MAY_ACCESS 0x00000010
100	#define MAY_OPEN 0x00000020
101	#define MAY_CHDIR 0x00000040
102	/* called from RCU mode, don't block */
103	#define MAY_NOT_BLOCK 0x00000080
104
105	/*
106	* flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond
107	* to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
108	*/
109
110	/* file is open for reading */
111	#define FMODE_READ ((__force fmode_t)0x1)
112	/* file is open for writing */
113	#define FMODE_WRITE ((__force fmode_t)0x2)
114	/* file is seekable */
115	#define FMODE_LSEEK ((__force fmode_t)0x4)
116	/* file can be accessed using pread */
117	#define FMODE_PREAD ((__force fmode_t)0x8)
118	/* file can be accessed using pwrite */
119	#define FMODE_PWRITE ((__force fmode_t)0x10)
120	/* File is opened for execution with sys_execve / sys_uselib */
121	#define FMODE_EXEC ((__force fmode_t)0x20)
122	/* 32bit hashes as llseek() offset (for directories) */
123	#define FMODE_32BITHASH ((__force fmode_t)0x200)
124	/* 64bit hashes as llseek() offset (for directories) */
125	#define FMODE_64BITHASH ((__force fmode_t)0x400)
126
127	/*
128	* Don't update ctime and mtime.
129	*
130	* Currently a special hack for the XFS open_by_handle ioctl, but we'll
131	* hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
132	*/
133	#define FMODE_NOCMTIME ((__force fmode_t)0x800)
134
135	/* Expect random access pattern */
136	#define FMODE_RANDOM ((__force fmode_t)0x1000)
137
138	/* File is huge (eg. /dev/mem): treat loff_t as unsigned */
139	#define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000)
140
141	/* File is opened with O_PATH; almost nothing can be done with it */
142	#define FMODE_PATH ((__force fmode_t)0x4000)
143
144	/* File needs atomic accesses to f_pos */
145	#define FMODE_ATOMIC_POS ((__force fmode_t)0x8000)
146	/* Write access to underlying fs */
147	#define FMODE_WRITER ((__force fmode_t)0x10000)
148	/* Has read method(s) */
149	#define FMODE_CAN_READ ((__force fmode_t)0x20000)
150	/* Has write method(s) */
151	#define FMODE_CAN_WRITE ((__force fmode_t)0x40000)
152
153	#define FMODE_OPENED ((__force fmode_t)0x80000)
154	#define FMODE_CREATED ((__force fmode_t)0x100000)
155
156	/* File is stream-like */
157	#define FMODE_STREAM ((__force fmode_t)0x200000)
158
159	/* File supports DIRECT IO */
160	#define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000)
161
162	#define FMODE_NOREUSE ((__force fmode_t)0x800000)
163
164	/* File supports non-exclusive O_DIRECT writes from multiple threads */
165	#define FMODE_DIO_PARALLEL_WRITE ((__force fmode_t)0x1000000)
166
167	/* File is embedded in backing_file object */
168	#define FMODE_BACKING ((__force fmode_t)0x2000000)
169
170	/* File was opened by fanotify and shouldn't generate fanotify events */
171	#define FMODE_NONOTIFY ((__force fmode_t)0x4000000)
172
173	/* File is capable of returning -EAGAIN if I/O will block */
174	#define FMODE_NOWAIT ((__force fmode_t)0x8000000)
175
176	/* File represents mount that needs unmounting */
177	#define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000)
178
179	/* File does not contribute to nr_files count */
180	#define FMODE_NOACCOUNT ((__force fmode_t)0x20000000)
181
182	/* File supports async buffered reads */
183	#define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000)
184
185	/* File supports async nowait buffered writes */
186	#define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000)
187
188	/*
189	* Attribute flags. These should be or-ed together to figure out what
190	* has been changed!
191	*/
192	#define ATTR_MODE (1 << 0)
193	#define ATTR_UID (1 << 1)
194	#define ATTR_GID (1 << 2)
195	#define ATTR_SIZE (1 << 3)
196	#define ATTR_ATIME (1 << 4)
197	#define ATTR_MTIME (1 << 5)
198	#define ATTR_CTIME (1 << 6)
199	#define ATTR_ATIME_SET (1 << 7)
200	#define ATTR_MTIME_SET (1 << 8)
201	#define ATTR_FORCE (1 << 9) /* Not a change, but a change it */
202	#define ATTR_KILL_SUID (1 << 11)
203	#define ATTR_KILL_SGID (1 << 12)
204	#define ATTR_FILE (1 << 13)
205	#define ATTR_KILL_PRIV (1 << 14)
206	#define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */
207	#define ATTR_TIMES_SET (1 << 16)
208	#define ATTR_TOUCH (1 << 17)
209
210	/*
211	* Whiteout is represented by a char device. The following constants define the
212	* mode and device number to use.
213	*/
214	#define WHITEOUT_MODE 0
215	#define WHITEOUT_DEV 0
216
217	/*
218	* This is the Inode Attributes structure, used for notify_change(). It
219	* uses the above definitions as flags, to know which values have changed.
220	* Also, in this manner, a Filesystem can look at only the values it cares
221	* about. Basically, these are the attributes that the VFS layer can
222	* request to change from the FS layer.
223	*
224	* Derek Atkins <[email protected]> 94-10-20
225	*/
226	struct iattr {
227	unsigned int ia_valid;
228	umode_t ia_mode;
229	/*
230	* The two anonymous unions wrap structures with the same member.
231	*
232	* Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which
233	* are a dedicated type requiring the filesystem to use the dedicated
234	* helpers. Other filesystem can continue to use ia_{g,u}id until they
235	* have been ported.
236	*
237	* They always contain the same value. In other words FS_ALLOW_IDMAP
238	* pass down the same value on idmapped mounts as they would on regular
239	* mounts.
240	*/
241	union {
242	kuid_t ia_uid;
243	vfsuid_t ia_vfsuid;
244	};
245	union {
246	kgid_t ia_gid;
247	vfsgid_t ia_vfsgid;
248	};
249	loff_t ia_size;
250	struct timespec64 ia_atime;
251	struct timespec64 ia_mtime;
252	struct timespec64 ia_ctime;
253
254	/*
255	* Not an attribute, but an auxiliary info for filesystems wanting to
256	* implement an ftruncate() like method. NOTE: filesystem should
257	* check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
258	*/
259	struct file *ia_file;
260	};
261
262	/*
263	* Includes for diskquotas.
264	*/
265	#include <linux/quota.h>
266
267	/*
268	* Maximum number of layers of fs stack. Needs to be limited to
269	* prevent kernel stack overflow
270	*/
271	#define FILESYSTEM_MAX_STACK_DEPTH 2
272
273	/**
274	* enum positive_aop_returns - aop return codes with specific semantics
275	*
276	* @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
277	* completed, that the page is still locked, and
278	* should be considered active. The VM uses this hint
279	* to return the page to the active list -- it won't
280	* be a candidate for writeback again in the near
281	* future. Other callers must be careful to unlock
282	* the page if they get this return. Returned by
283	* writepage();
284	*
285	* @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
286	* unlocked it and the page might have been truncated.
287	* The caller should back up to acquiring a new page and
288	* trying again. The aop will be taking reasonable
289	* precautions not to livelock. If the caller held a page
290	* reference, it should drop it before retrying. Returned
291	* by read_folio().
292	*
293	* address_space_operation functions return these large constants to indicate
294	* special semantics to the caller. These are much larger than the bytes in a
295	* page to allow for functions that return the number of bytes operated on in a
296	* given page.
297	*/
298
299	enum positive_aop_returns {
300	AOP_WRITEPAGE_ACTIVATE = 0x80000,
301	AOP_TRUNCATED_PAGE = 0x80001,
302	};
303
304	/*
305	* oh the beauties of C type declarations.
306	*/
307	struct page;
308	struct address_space;
309	struct writeback_control;
310	struct readahead_control;
311
312	/*
313	* Write life time hint values.
314	* Stored in struct inode as u8.
315	*/
316	enum rw_hint {
317	WRITE_LIFE_NOT_SET = 0,
318	WRITE_LIFE_NONE = RWH_WRITE_LIFE_NONE,
319	WRITE_LIFE_SHORT = RWH_WRITE_LIFE_SHORT,
320	WRITE_LIFE_MEDIUM = RWH_WRITE_LIFE_MEDIUM,
321	WRITE_LIFE_LONG = RWH_WRITE_LIFE_LONG,
322	WRITE_LIFE_EXTREME = RWH_WRITE_LIFE_EXTREME,
323	};
324
325	/* Match RWF_* bits to IOCB bits */
326	#define IOCB_HIPRI (__force int) RWF_HIPRI
327	#define IOCB_DSYNC (__force int) RWF_DSYNC
328	#define IOCB_SYNC (__force int) RWF_SYNC
329	#define IOCB_NOWAIT (__force int) RWF_NOWAIT
330	#define IOCB_APPEND (__force int) RWF_APPEND
331
332	/* non-RWF related bits - start at 16 */
333	#define IOCB_EVENTFD (1 << 16)
334	#define IOCB_DIRECT (1 << 17)
335	#define IOCB_WRITE (1 << 18)
336	/* iocb->ki_waitq is valid */
337	#define IOCB_WAITQ (1 << 19)
338	#define IOCB_NOIO (1 << 20)
339	/* can use bio alloc cache */
340	#define IOCB_ALLOC_CACHE (1 << 21)
341	/*
342	* IOCB_DIO_CALLER_COMP can be set by the iocb owner, to indicate that the
343	* iocb completion can be passed back to the owner for execution from a safe
344	* context rather than needing to be punted through a workqueue. If this
345	* flag is set, the bio completion handling may set iocb->dio_complete to a
346	* handler function and iocb->private to context information for that handler.
347	* The issuer should call the handler with that context information from task
348	* context to complete the processing of the iocb. Note that while this
349	* provides a task context for the dio_complete() callback, it should only be
350	* used on the completion side for non-IO generating completions. It's fine to
351	* call blocking functions from this callback, but they should not wait for
352	* unrelated IO (like cache flushing, new IO generation, etc).
353	*/
354	#define IOCB_DIO_CALLER_COMP (1 << 22)
355
356	/* for use in trace events */
357	#define TRACE_IOCB_STRINGS \
358	{ IOCB_HIPRI, "HIPRI" }, \
359	{ IOCB_DSYNC, "DSYNC" }, \
360	{ IOCB_SYNC, "SYNC" }, \
361	{ IOCB_NOWAIT, "NOWAIT" }, \
362	{ IOCB_APPEND, "APPEND" }, \
363	{ IOCB_EVENTFD, "EVENTFD"}, \
364	{ IOCB_DIRECT, "DIRECT" }, \
365	{ IOCB_WRITE, "WRITE" }, \
366	{ IOCB_WAITQ, "WAITQ" }, \
367	{ IOCB_NOIO, "NOIO" }, \
368	{ IOCB_ALLOC_CACHE, "ALLOC_CACHE" }, \
369	{ IOCB_DIO_CALLER_COMP, "CALLER_COMP" }
370
371	struct kiocb {
372	struct file *ki_filp;
373	loff_t ki_pos;
374	void (ki_complete)(struct kiocb iocb, long ret);
375	void *private;
376	int ki_flags;
377	u16 ki_ioprio; /* See linux/ioprio.h */
378	union {
379	/*
380	* Only used for async buffered reads, where it denotes the
381	* page waitqueue associated with completing the read. Valid
382	* IFF IOCB_WAITQ is set.
383	*/
384	struct wait_page_queue *ki_waitq;
385	/*
386	* Can be used for O_DIRECT IO, where the completion handling
387	* is punted back to the issuer of the IO. May only be set
388	* if IOCB_DIO_CALLER_COMP is set by the issuer, and the issuer
389	* must then check for presence of this handler when ki_complete
390	* is invoked. The data passed in to this handler must be
391	* assigned to ->private when dio_complete is assigned.
392	*/
393	ssize_t (dio_complete)(void data);
394	};
395	};
396
397	static inline bool is_sync_kiocb(struct kiocb *kiocb)
398	{
399	return kiocb->ki_complete == NULL;
400	}
401
402	struct address_space_operations {
403	int (writepage)(struct page page, struct writeback_control *wbc);
404	int (read_folio)(struct file , struct folio *);
405
406	/* Write back some dirty pages from this mapping. */
407	int (writepages)(struct address_space , struct writeback_control *);
408
409	/* Mark a folio dirty. Return true if this dirtied it */
410	bool (dirty_folio)(struct address_space , struct folio *);
411
412	void (readahead)(struct readahead_control );
413
414	int (write_begin)(struct file , struct address_space *mapping,
415	loff_t pos, unsigned len,
416	struct page pagep, void fsdata);
417	int (write_end)(struct file , struct address_space *mapping,
418	loff_t pos, unsigned len, unsigned copied,
419	struct page page, void fsdata);
420
421	/* Unfortunately this kludge is needed for FIBMAP. Don't use it */
422	sector_t (bmap)(struct address_space , sector_t);
423	void (invalidate_folio) (struct folio , size_t offset, size_t len);
424	bool (release_folio)(struct folio , gfp_t);
425	void (free_folio)(struct folio folio);
426	ssize_t (direct_IO)(struct kiocb , struct iov_iter *iter);
427	/*
428	* migrate the contents of a folio to the specified target. If
429	* migrate_mode is MIGRATE_ASYNC, it must not block.
430	*/
431	int (migrate_folio)(struct address_space , struct folio *dst,
432	struct folio *src, enum migrate_mode);
433	int (launder_folio)(struct folio );
434	bool (is_partially_uptodate) (struct folio , size_t from,
435	size_t count);
436	void (is_dirty_writeback) (struct folio , bool dirty, bool wb);
437	int (error_remove_page)(struct address_space , struct page *);
438
439	/* swapfile support */
440	int (swap_activate)(struct swap_info_struct sis, struct file *file,
441	sector_t *span);
442	void (swap_deactivate)(struct file file);
443	int (swap_rw)(struct kiocb iocb, struct iov_iter *iter);
444	};
445
446	extern const struct address_space_operations empty_aops;
447
448	/**
449	* struct address_space - Contents of a cacheable, mappable object.
450	* @host: Owner, either the inode or the block_device.
451	* @i_pages: Cached pages.
452	* @invalidate_lock: Guards coherency between page cache contents and
453	* file offset->disk block mappings in the filesystem during invalidates.
454	* It is also used to block modification of page cache contents through
455	* memory mappings.
456	* @gfp_mask: Memory allocation flags to use for allocating pages.
457	* @i_mmap_writable: Number of VM_SHARED, VM_MAYWRITE mappings.
458	* @nr_thps: Number of THPs in the pagecache (non-shmem only).
459	* @i_mmap: Tree of private and shared mappings.
460	* @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
461	* @nrpages: Number of page entries, protected by the i_pages lock.
462	* @writeback_index: Writeback starts here.
463	* @a_ops: Methods.
464	* @flags: Error bits and flags (AS_*).
465	* @wb_err: The most recent error which has occurred.
466	* @private_lock: For use by the owner of the address_space.
467	* @private_list: For use by the owner of the address_space.
468	* @private_data: For use by the owner of the address_space.
469	*/
470	struct address_space {
471	struct inode *host;
472	struct xarray i_pages;
473	struct rw_semaphore invalidate_lock;
474	gfp_t gfp_mask;
475	atomic_t i_mmap_writable;
476	#ifdef CONFIG_READ_ONLY_THP_FOR_FS
477	/* number of thp, only for non-shmem files */
478	atomic_t nr_thps;
479	#endif
480	struct rb_root_cached i_mmap;
481	unsigned long nrpages;
482	pgoff_t writeback_index;
483	const struct address_space_operations *a_ops;
484	unsigned long flags;
485	struct rw_semaphore i_mmap_rwsem;
486	errseq_t wb_err;
487	spinlock_t private_lock;
488	struct list_head private_list;
489	void *private_data;
490	} __attribute__((aligned(sizeof(long)))) __randomize_layout;
491	/*
492	* On most architectures that alignment is already the case; but
493	* must be enforced here for CRIS, to let the least significant bit
494	* of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
495	*/
496
497	/* XArray tags, for tagging dirty and writeback pages in the pagecache. */
498	#define PAGECACHE_TAG_DIRTY XA_MARK_0
499	#define PAGECACHE_TAG_WRITEBACK XA_MARK_1
500	#define PAGECACHE_TAG_TOWRITE XA_MARK_2
501
502	/*
503	* Returns true if any of the pages in the mapping are marked with the tag.
504	*/
505	static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
506	{
507	return xa_marked(&mapping->i_pages, tag);
508	}
509
510	static inline void i_mmap_lock_write(struct address_space *mapping)
511	{
512	down_write(&mapping->i_mmap_rwsem);
513	}
514
515	static inline int i_mmap_trylock_write(struct address_space *mapping)
516	{
517	return down_write_trylock(&mapping->i_mmap_rwsem);
518	}
519
520	static inline void i_mmap_unlock_write(struct address_space *mapping)
521	{
522	up_write(&mapping->i_mmap_rwsem);
523	}
524
525	static inline int i_mmap_trylock_read(struct address_space *mapping)
526	{
527	return down_read_trylock(&mapping->i_mmap_rwsem);
528	}
529
530	static inline void i_mmap_lock_read(struct address_space *mapping)
531	{
532	down_read(&mapping->i_mmap_rwsem);
533	}
534
535	static inline void i_mmap_unlock_read(struct address_space *mapping)
536	{
537	up_read(&mapping->i_mmap_rwsem);
538	}
539
540	static inline void i_mmap_assert_locked(struct address_space *mapping)
541	{
542	lockdep_assert_held(&mapping->i_mmap_rwsem);
543	}
544
545	static inline void i_mmap_assert_write_locked(struct address_space *mapping)
546	{
547	lockdep_assert_held_write(&mapping->i_mmap_rwsem);
548	}
549
550	/*
551	* Might pages of this file be mapped into userspace?
552	*/
553	static inline int mapping_mapped(struct address_space *mapping)
554	{
555	return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
556	}
557
558	/*
559	* Might pages of this file have been modified in userspace?
560	* Note that i_mmap_writable counts all VM_SHARED, VM_MAYWRITE vmas: do_mmap
561	* marks vma as VM_SHARED if it is shared, and the file was opened for
562	* writing i.e. vma may be mprotected writable even if now readonly.
563	*
564	* If i_mmap_writable is negative, no new writable mappings are allowed. You
565	* can only deny writable mappings, if none exists right now.
566	*/
567	static inline int mapping_writably_mapped(struct address_space *mapping)
568	{
569	return atomic_read(&mapping->i_mmap_writable) > 0;
570	}
571
572	static inline int mapping_map_writable(struct address_space *mapping)
573	{
574	return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
575	0 : -EPERM;
576	}
577
578	static inline void mapping_unmap_writable(struct address_space *mapping)
579	{
580	atomic_dec(&mapping->i_mmap_writable);
581	}
582
583	static inline int mapping_deny_writable(struct address_space *mapping)
584	{
585	return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
586	0 : -EBUSY;
587	}
588
589	static inline void mapping_allow_writable(struct address_space *mapping)
590	{
591	atomic_inc(&mapping->i_mmap_writable);
592	}
593
594	/*
595	* Use sequence counter to get consistent i_size on 32-bit processors.
596	*/
597	#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
598	#include <linux/seqlock.h>
599	#define __NEED_I_SIZE_ORDERED
600	#define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
601	#else
602	#define i_size_ordered_init(inode) do { } while (0)
603	#endif
604
605	struct posix_acl;
606	#define ACL_NOT_CACHED ((void *)(-1))
607	/*
608	* ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
609	* cache the ACL. This also means that ->get_inode_acl() can be called in RCU
610	* mode with the LOOKUP_RCU flag.
611	*/
612	#define ACL_DONT_CACHE ((void *)(-3))
613
614	static inline struct posix_acl *
615	uncached_acl_sentinel(struct task_struct *task)
616	{
617	return (void *)task + 1;
618	}
619
620	static inline bool
621	is_uncached_acl(struct posix_acl *acl)
622	{
623	return (long)acl & 1;
624	}
625
626	#define IOP_FASTPERM 0x0001
627	#define IOP_LOOKUP 0x0002
628	#define IOP_NOFOLLOW 0x0004
629	#define IOP_XATTR 0x0008
630	#define IOP_DEFAULT_READLINK 0x0010
631
632	struct fsnotify_mark_connector;
633
634	/*
635	* Keep mostly read-only and often accessed (especially for
636	* the RCU path lookup and 'stat' data) fields at the beginning
637	* of the 'struct inode'
638	*/
639	struct inode {
640	umode_t i_mode;
641	unsigned short i_opflags;
642	kuid_t i_uid;
643	kgid_t i_gid;
644	unsigned int i_flags;
645
646	#ifdef CONFIG_FS_POSIX_ACL
647	struct posix_acl *i_acl;
648	struct posix_acl *i_default_acl;
649	#endif
650
651	const struct inode_operations *i_op;
652	struct super_block *i_sb;
653	struct address_space *i_mapping;
654
655	#ifdef CONFIG_SECURITY
656	void *i_security;
657	#endif
658
659	/* Stat data, not accessed from path walking */
660	unsigned long i_ino;
661	/*
662	* Filesystems may only read i_nlink directly. They shall use the
663	* following functions for modification:
664	*
665	* (set\|clear\|inc\|drop)_nlink
666	* inode_(inc\|dec)_link_count
667	*/
668	union {
669	const unsigned int i_nlink;
670	unsigned int __i_nlink;
671	};
672	dev_t i_rdev;
673	loff_t i_size;
674	struct timespec64 __i_atime;
675	struct timespec64 __i_mtime;
676	struct timespec64 __i_ctime; /* use inode__ctime accessors! /
677	spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
678	unsigned short i_bytes;
679	u8 i_blkbits;
680	u8 i_write_hint;
681	blkcnt_t i_blocks;
682
683	#ifdef __NEED_I_SIZE_ORDERED
684	seqcount_t i_size_seqcount;
685	#endif
686
687	/* Misc */
688	unsigned long i_state;
689	struct rw_semaphore i_rwsem;
690
691	unsigned long dirtied_when; /* jiffies of first dirtying */
692	unsigned long dirtied_time_when;
693
694	struct hlist_node i_hash;
695	struct list_head i_io_list; /* backing dev IO list */
696	#ifdef CONFIG_CGROUP_WRITEBACK
697	struct bdi_writeback i_wb; / the associated cgroup wb */
698
699	/* foreign inode detection, see wbc_detach_inode() */
700	int i_wb_frn_winner;
701	u16 i_wb_frn_avg_time;
702	u16 i_wb_frn_history;
703	#endif
704	struct list_head i_lru; /* inode LRU list */
705	struct list_head i_sb_list;
706	struct list_head i_wb_list; /* backing dev writeback list */
707	union {
708	struct hlist_head i_dentry;
709	struct rcu_head i_rcu;
710	};
711	atomic64_t i_version;
712	atomic64_t i_sequence; /* see futex */
713	atomic_t i_count;
714	atomic_t i_dio_count;
715	atomic_t i_writecount;
716	#if defined(CONFIG_IMA) \|\| defined(CONFIG_FILE_LOCKING)
717	atomic_t i_readcount; /* struct files open RO */
718	#endif
719	union {
720	const struct file_operations i_fop; / former ->i_op->default_file_ops */
721	void (free_inode)(struct inode );
722	};
723	struct file_lock_context *i_flctx;
724	struct address_space i_data;
725	struct list_head i_devices;
726	union {
727	struct pipe_inode_info *i_pipe;
728	struct cdev *i_cdev;
729	char *i_link;
730	unsigned i_dir_seq;
731	};
732
733	__u32 i_generation;
734
735	#ifdef CONFIG_FSNOTIFY
736	__u32 i_fsnotify_mask; /* all events this inode cares about */
737	struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
738	#endif
739
740	#ifdef CONFIG_FS_ENCRYPTION
741	struct fscrypt_inode_info *i_crypt_info;
742	#endif
743
744	#ifdef CONFIG_FS_VERITY
745	struct fsverity_info *i_verity_info;
746	#endif
747
748	void i_private; / fs or device private pointer */
749	} __randomize_layout;
750
751	struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
752
753	static inline unsigned int i_blocksize(const struct inode *node)
754	{
755	return (1 << node->i_blkbits);
756	}
757
758	static inline int inode_unhashed(struct inode *inode)
759	{
760	return hlist_unhashed(&inode->i_hash);
761	}
762
763	/*
764	* __mark_inode_dirty expects inodes to be hashed. Since we don't
765	* want special inodes in the fileset inode space, we make them
766	* appear hashed, but do not put on any lists. hlist_del()
767	* will work fine and require no locking.
768	*/
769	static inline void inode_fake_hash(struct inode *inode)
770	{
771	hlist_add_fake(&inode->i_hash);
772	}
773
774	/*
775	* inode->i_mutex nesting subclasses for the lock validator:
776	*
777	* 0: the object of the current VFS operation
778	* 1: parent
779	* 2: child/target
780	* 3: xattr
781	* 4: second non-directory
782	* 5: second parent (when locking independent directories in rename)
783	*
784	* I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
785	* non-directories at once.
786	*
787	* The locking order between these classes is
788	* parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
789	*/
790	enum inode_i_mutex_lock_class
791	{
792	I_MUTEX_NORMAL,
793	I_MUTEX_PARENT,
794	I_MUTEX_CHILD,
795	I_MUTEX_XATTR,
796	I_MUTEX_NONDIR2,
797	I_MUTEX_PARENT2,
798	};
799
800	static inline void inode_lock(struct inode *inode)
801	{
802	down_write(&inode->i_rwsem);
803	}
804
805	static inline void inode_unlock(struct inode *inode)
806	{
807	up_write(&inode->i_rwsem);
808	}
809
810	static inline void inode_lock_shared(struct inode *inode)
811	{
812	down_read(&inode->i_rwsem);
813	}
814
815	static inline void inode_unlock_shared(struct inode *inode)
816	{
817	up_read(&inode->i_rwsem);
818	}
819
820	static inline int inode_trylock(struct inode *inode)
821	{
822	return down_write_trylock(&inode->i_rwsem);
823	}
824
825	static inline int inode_trylock_shared(struct inode *inode)
826	{
827	return down_read_trylock(&inode->i_rwsem);
828	}
829
830	static inline int inode_is_locked(struct inode *inode)
831	{
832	return rwsem_is_locked(&inode->i_rwsem);
833	}
834
835	static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
836	{
837	down_write_nested(&inode->i_rwsem, subclass);
838	}
839
840	static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
841	{
842	down_read_nested(&inode->i_rwsem, subclass);
843	}
844
845	static inline void filemap_invalidate_lock(struct address_space *mapping)
846	{
847	down_write(&mapping->invalidate_lock);
848	}
849
850	static inline void filemap_invalidate_unlock(struct address_space *mapping)
851	{
852	up_write(&mapping->invalidate_lock);
853	}
854
855	static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
856	{
857	down_read(&mapping->invalidate_lock);
858	}
859
860	static inline int filemap_invalidate_trylock_shared(
861	struct address_space *mapping)
862	{
863	return down_read_trylock(&mapping->invalidate_lock);
864	}
865
866	static inline void filemap_invalidate_unlock_shared(
867	struct address_space *mapping)
868	{
869	up_read(&mapping->invalidate_lock);
870	}
871
872	void lock_two_nondirectories(struct inode , struct inode);
873	void unlock_two_nondirectories(struct inode , struct inode);
874
875	void filemap_invalidate_lock_two(struct address_space *mapping1,
876	struct address_space *mapping2);
877	void filemap_invalidate_unlock_two(struct address_space *mapping1,
878	struct address_space *mapping2);
879
880
881	/*
882	* NOTE: in a 32bit arch with a preemptable kernel and
883	* an UP compile the i_size_read/write must be atomic
884	* with respect to the local cpu (unlike with preempt disabled),
885	* but they don't need to be atomic with respect to other cpus like in
886	* true SMP (so they need either to either locally disable irq around
887	* the read or for example on x86 they can be still implemented as a
888	* cmpxchg8b without the need of the lock prefix). For SMP compiles
889	* and 64bit archs it makes no difference if preempt is enabled or not.
890	*/
891	static inline loff_t i_size_read(const struct inode *inode)
892	{
893	#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
894	loff_t i_size;
895	unsigned int seq;
896
897	do {
898	seq = read_seqcount_begin(&inode->i_size_seqcount);
899	i_size = inode->i_size;
900	} while (read_seqcount_retry(&inode->i_size_seqcount, seq));
901	return i_size;
902	#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
903	loff_t i_size;
904
905	preempt_disable();
906	i_size = inode->i_size;
907	preempt_enable();
908	return i_size;
909	#else
910	return inode->i_size;
911	#endif
912	}
913
914	/*
915	* NOTE: unlike i_size_read(), i_size_write() does need locking around it
916	* (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
917	* can be lost, resulting in subsequent i_size_read() calls spinning forever.
918	*/
919	static inline void i_size_write(struct inode *inode, loff_t i_size)
920	{
921	#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
922	preempt_disable();
923	write_seqcount_begin(&inode->i_size_seqcount);
924	inode->i_size = i_size;
925	write_seqcount_end(&inode->i_size_seqcount);
926	preempt_enable();
927	#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
928	preempt_disable();
929	inode->i_size = i_size;
930	preempt_enable();
931	#else
932	inode->i_size = i_size;
933	#endif
934	}
935
936	static inline unsigned iminor(const struct inode *inode)
937	{
938	return MINOR(inode->i_rdev);
939	}
940
941	static inline unsigned imajor(const struct inode *inode)
942	{
943	return MAJOR(inode->i_rdev);
944	}
945
946	struct fown_struct {
947	rwlock_t lock; /* protects pid, uid, euid fields */
948	struct pid pid; / pid or -pgrp where SIGIO should be sent */
949	enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
950	kuid_t uid, euid; /* uid/euid of process setting the owner */
951	int signum; /* posix.1b rt signal to be delivered on IO */
952	};
953
954	/**
955	* struct file_ra_state - Track a file's readahead state.
956	* @start: Where the most recent readahead started.
957	* @size: Number of pages read in the most recent readahead.
958	* @async_size: Numer of pages that were/are not needed immediately
959	* and so were/are genuinely "ahead". Start next readahead when
960	* the first of these pages is accessed.
961	* @ra_pages: Maximum size of a readahead request, copied from the bdi.
962	* @mmap_miss: How many mmap accesses missed in the page cache.
963	* @prev_pos: The last byte in the most recent read request.
964	*
965	* When this structure is passed to ->readahead(), the "most recent"
966	* readahead means the current readahead.
967	*/
968	struct file_ra_state {
969	pgoff_t start;
970	unsigned int size;
971	unsigned int async_size;
972	unsigned int ra_pages;
973	unsigned int mmap_miss;
974	loff_t prev_pos;
975	};
976
977	/*
978	* Check if @index falls in the readahead windows.
979	*/
980	static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
981	{
982	return (index >= ra->start &&
983	index < ra->start + ra->size);
984	}
985
986	/*
987	* f_{lock,count,pos_lock} members can be highly contended and share
988	* the same cacheline. f_{lock,mode} are very frequently used together
989	* and so share the same cacheline as well. The read-mostly
990	* f_{path,inode,op} are kept on a separate cacheline.
991	*/
992	struct file {
993	union {
994	struct llist_node f_llist;
995	struct rcu_head f_rcuhead;
996	unsigned int f_iocb_flags;
997	};
998
999	/*
1000	* Protects f_ep, f_flags.
1001	* Must not be taken from IRQ context.
1002	*/
1003	spinlock_t f_lock;
1004	fmode_t f_mode;
1005	atomic_long_t f_count;
1006	struct mutex f_pos_lock;
1007	loff_t f_pos;
1008	unsigned int f_flags;
1009	struct fown_struct f_owner;
1010	const struct cred *f_cred;
1011	struct file_ra_state f_ra;
1012	struct path f_path;
1013	struct inode f_inode; / cached value */
1014	const struct file_operations *f_op;
1015
1016	u64 f_version;
1017	#ifdef CONFIG_SECURITY
1018	void *f_security;
1019	#endif
1020	/* needed for tty driver, and maybe others */
1021	void *private_data;
1022
1023	#ifdef CONFIG_EPOLL
1024	/* Used by fs/eventpoll.c to link all the hooks to this file */
1025	struct hlist_head *f_ep;
1026	#endif /* #ifdef CONFIG_EPOLL */
1027	struct address_space *f_mapping;
1028	errseq_t f_wb_err;
1029	errseq_t f_sb_err; /* for syncfs */
1030	} __randomize_layout
1031	__attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
1032
1033	struct file_handle {
1034	__u32 handle_bytes;
1035	int handle_type;
1036	/* file identifier */
1037	unsigned char f_handle[];
1038	};
1039
1040	static inline struct file get_file(struct file f)
1041	{
1042	atomic_long_inc(&f->f_count);
1043	return f;
1044	}
1045
1046	struct file get_file_rcu(struct file __rcu *f);
1047	struct file get_file_active(struct file *f);
1048
1049	#define file_count(x) atomic_long_read(&(x)->f_count)
1050
1051	#define MAX_NON_LFS ((1UL<<31) - 1)
1052
1053	/* Page cache limit. The filesystems should put that into their s_maxbytes
1054	limits, otherwise bad things can happen in VM. */
1055	#if BITS_PER_LONG==32
1056	#define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT)
1057	#elif BITS_PER_LONG==64
1058	#define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX)
1059	#endif
1060
1061	/* legacy typedef, should eventually be removed */
1062	typedef void *fl_owner_t;
1063
1064	struct file_lock;
1065
1066	/* The following constant reflects the upper bound of the file/locking space */
1067	#ifndef OFFSET_MAX
1068	#define OFFSET_MAX type_max(loff_t)
1069	#define OFFT_OFFSET_MAX type_max(off_t)
1070	#endif
1071
1072	extern void send_sigio(struct fown_struct *fown, int fd, int band);
1073
1074	static inline struct inode file_inode(const struct file f)
1075	{
1076	return f->f_inode;
1077	}
1078
1079	static inline struct dentry file_dentry(const struct file file)
1080	{
1081	return d_real(file->f_path.dentry, file_inode(file));
1082	}
1083
1084	struct fasync_struct {
1085	rwlock_t fa_lock;
1086	int magic;
1087	int fa_fd;
1088	struct fasync_struct fa_next; / singly linked list */
1089	struct file *fa_file;
1090	struct rcu_head fa_rcu;
1091	};
1092
1093	#define FASYNC_MAGIC 0x4601
1094
1095	/* SMP safe fasync helpers: */
1096	extern int fasync_helper(int, struct file , int, struct fasync_struct *);
1097	extern struct fasync_struct fasync_insert_entry(int, struct file , struct fasync_struct *, struct fasync_struct );
1098	extern int fasync_remove_entry(struct file , struct fasync_struct *);
1099	extern struct fasync_struct *fasync_alloc(void);
1100	extern void fasync_free(struct fasync_struct *);
1101
1102	/* can be called from interrupts */
1103	extern void kill_fasync(struct fasync_struct **, int, int);
1104
1105	extern void __f_setown(struct file filp, struct pid , enum pid_type, int force);
1106	extern int f_setown(struct file *filp, int who, int force);
1107	extern void f_delown(struct file *filp);
1108	extern pid_t f_getown(struct file *filp);
1109	extern int send_sigurg(struct fown_struct *fown);
1110
1111	/*
1112	* sb->s_flags. Note that these mirror the equivalent MS_* flags where
1113	* represented in both.
1114	*/
1115	#define SB_RDONLY BIT(0) /* Mount read-only */
1116	#define SB_NOSUID BIT(1) /* Ignore suid and sgid bits */
1117	#define SB_NODEV BIT(2) /* Disallow access to device special files */
1118	#define SB_NOEXEC BIT(3) /* Disallow program execution */
1119	#define SB_SYNCHRONOUS BIT(4) /* Writes are synced at once */
1120	#define SB_MANDLOCK BIT(6) /* Allow mandatory locks on an FS */
1121	#define SB_DIRSYNC BIT(7) /* Directory modifications are synchronous */
1122	#define SB_NOATIME BIT(10) /* Do not update access times. */
1123	#define SB_NODIRATIME BIT(11) /* Do not update directory access times */
1124	#define SB_SILENT BIT(15)
1125	#define SB_POSIXACL BIT(16) /* Supports POSIX ACLs */
1126	#define SB_INLINECRYPT BIT(17) /* Use blk-crypto for encrypted files */
1127	#define SB_KERNMOUNT BIT(22) /* this is a kern_mount call */
1128	#define SB_I_VERSION BIT(23) /* Update inode I_version field */
1129	#define SB_LAZYTIME BIT(25) /* Update the on-disk [acm]times lazily */
1130
1131	/* These sb flags are internal to the kernel */
1132	#define SB_DEAD BIT(21)
1133	#define SB_DYING BIT(24)
1134	#define SB_SUBMOUNT BIT(26)
1135	#define SB_FORCE BIT(27)
1136	#define SB_NOSEC BIT(28)
1137	#define SB_BORN BIT(29)
1138	#define SB_ACTIVE BIT(30)
1139	#define SB_NOUSER BIT(31)
1140
1141	/* These flags relate to encoding and casefolding */
1142	#define SB_ENC_STRICT_MODE_FL (1 << 0)
1143
1144	#define sb_has_strict_encoding(sb) \
1145	(sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1146
1147	/*
1148	* Umount options
1149	*/
1150
1151	#define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1152	#define MNT_DETACH 0x00000002 /* Just detach from the tree */
1153	#define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1154	#define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1155	#define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1156
1157	/* sb->s_iflags */
1158	#define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */
1159	#define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */
1160	#define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */
1161	#define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */
1162
1163	/* sb->s_iflags to limit user namespace mounts */
1164	#define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */
1165	#define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020
1166	#define SB_I_UNTRUSTED_MOUNTER 0x00000040
1167
1168	#define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */
1169	#define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */
1170	#define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1171	#define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */
1172	#define SB_I_NOUMASK 0x00001000 /* VFS does not apply umask */
1173
1174	/* Possible states of 'frozen' field */
1175	enum {
1176	SB_UNFROZEN = 0, /* FS is unfrozen */
1177	SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1178	SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1179	SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1180	* internal threads if needed) */
1181	SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1182	};
1183
1184	#define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1185
1186	struct sb_writers {
1187	unsigned short frozen; /* Is sb frozen? */
1188	unsigned short freeze_holders; /* Who froze fs? */
1189	struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS];
1190	};
1191
1192	struct super_block {
1193	struct list_head s_list; /* Keep this first */
1194	dev_t s_dev; /* search index; _not_ kdev_t */
1195	unsigned char s_blocksize_bits;
1196	unsigned long s_blocksize;
1197	loff_t s_maxbytes; /* Max file size */
1198	struct file_system_type *s_type;
1199	const struct super_operations *s_op;
1200	const struct dquot_operations *dq_op;
1201	const struct quotactl_ops *s_qcop;
1202	const struct export_operations *s_export_op;
1203	unsigned long s_flags;
1204	unsigned long s_iflags; /* internal SB_I_* flags */
1205	unsigned long s_magic;
1206	struct dentry *s_root;
1207	struct rw_semaphore s_umount;
1208	int s_count;
1209	atomic_t s_active;
1210	#ifdef CONFIG_SECURITY
1211	void *s_security;
1212	#endif
1213	const struct xattr_handler * const *s_xattr;
1214	#ifdef CONFIG_FS_ENCRYPTION
1215	const struct fscrypt_operations *s_cop;
1216	struct fscrypt_keyring s_master_keys; / master crypto keys in use */
1217	#endif
1218	#ifdef CONFIG_FS_VERITY
1219	const struct fsverity_operations *s_vop;
1220	#endif
1221	#if IS_ENABLED(CONFIG_UNICODE)
1222	struct unicode_map *s_encoding;
1223	__u16 s_encoding_flags;
1224	#endif
1225	struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
1226	struct list_head s_mounts; /* list of mounts; _not_ for fs use */
1227	struct block_device *s_bdev;
1228	struct bdev_handle *s_bdev_handle;
1229	struct backing_dev_info *s_bdi;
1230	struct mtd_info *s_mtd;
1231	struct hlist_node s_instances;
1232	unsigned int s_quota_types; /* Bitmask of supported quota types */
1233	struct quota_info s_dquot; /* Diskquota specific options */
1234
1235	struct sb_writers s_writers;
1236
1237	/*
1238	* Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1239	* s_fsnotify_marks together for cache efficiency. They are frequently
1240	* accessed and rarely modified.
1241	*/
1242	void s_fs_info; / Filesystem private info */
1243
1244	/* Granularity of c/m/atime in ns (cannot be worse than a second) */
1245	u32 s_time_gran;
1246	/* Time limits for c/m/atime in seconds */
1247	time64_t s_time_min;
1248	time64_t s_time_max;
1249	#ifdef CONFIG_FSNOTIFY
1250	__u32 s_fsnotify_mask;
1251	struct fsnotify_mark_connector __rcu *s_fsnotify_marks;
1252	#endif
1253
1254	char s_id[32]; /* Informational name */
1255	uuid_t s_uuid; /* UUID */
1256
1257	unsigned int s_max_links;
1258
1259	/*
1260	* The next field is for VFS only. No filesystems have any business
1261	* even looking at it. You had been warned.
1262	*/
1263	struct mutex s_vfs_rename_mutex; /* Kludge */
1264
1265	/*
1266	* Filesystem subtype. If non-empty the filesystem type field
1267	* in /proc/mounts will be "type.subtype"
1268	*/
1269	const char *s_subtype;
1270
1271	const struct dentry_operations s_d_op; / default d_op for dentries */
1272
1273	struct shrinker s_shrink; / per-sb shrinker handle */
1274
1275	/* Number of inodes with nlink == 0 but still referenced */
1276	atomic_long_t s_remove_count;
1277
1278	/*
1279	* Number of inode/mount/sb objects that are being watched, note that
1280	* inodes objects are currently double-accounted.
1281	*/
1282	atomic_long_t s_fsnotify_connectors;
1283
1284	/* Read-only state of the superblock is being changed */
1285	int s_readonly_remount;
1286
1287	/* per-sb errseq_t for reporting writeback errors via syncfs */
1288	errseq_t s_wb_err;
1289
1290	/* AIO completions deferred from interrupt context */
1291	struct workqueue_struct *s_dio_done_wq;
1292	struct hlist_head s_pins;
1293
1294	/*
1295	* Owning user namespace and default context in which to
1296	* interpret filesystem uids, gids, quotas, device nodes,
1297	* xattrs and security labels.
1298	*/
1299	struct user_namespace *s_user_ns;
1300
1301	/*
1302	* The list_lru structure is essentially just a pointer to a table
1303	* of per-node lru lists, each of which has its own spinlock.
1304	* There is no need to put them into separate cachelines.
1305	*/
1306	struct list_lru s_dentry_lru;
1307	struct list_lru s_inode_lru;
1308	struct rcu_head rcu;
1309	struct work_struct destroy_work;
1310
1311	struct mutex s_sync_lock; /* sync serialisation lock */
1312
1313	/*
1314	* Indicates how deep in a filesystem stack this SB is
1315	*/
1316	int s_stack_depth;
1317
1318	/* s_inode_list_lock protects s_inodes */
1319	spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
1320	struct list_head s_inodes; /* all inodes */
1321
1322	spinlock_t s_inode_wblist_lock;
1323	struct list_head s_inodes_wb; /* writeback inodes */
1324	} __randomize_layout;
1325
1326	static inline struct user_namespace i_user_ns(const struct inode inode)
1327	{
1328	return inode->i_sb->s_user_ns;
1329	}
1330
1331	/* Helper functions so that in most cases filesystems will
1332	* not need to deal directly with kuid_t and kgid_t and can
1333	* instead deal with the raw numeric values that are stored
1334	* in the filesystem.
1335	*/
1336	static inline uid_t i_uid_read(const struct inode *inode)
1337	{
1338	return from_kuid(i_user_ns(inode), inode->i_uid);
1339	}
1340
1341	static inline gid_t i_gid_read(const struct inode *inode)
1342	{
1343	return from_kgid(i_user_ns(inode), inode->i_gid);
1344	}
1345
1346	static inline void i_uid_write(struct inode *inode, uid_t uid)
1347	{
1348	inode->i_uid = make_kuid(i_user_ns(inode), uid);
1349	}
1350
1351	static inline void i_gid_write(struct inode *inode, gid_t gid)
1352	{
1353	inode->i_gid = make_kgid(i_user_ns(inode), gid);
1354	}
1355
1356	/**
1357	* i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping
1358	* @idmap: idmap of the mount the inode was found from
1359	* @inode: inode to map
1360	*
1361	* Return: whe inode's i_uid mapped down according to @idmap.
1362	* If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1363	*/
1364	static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap,
1365	const struct inode *inode)
1366	{
1367	return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid);
1368	}
1369
1370	/**
1371	* i_uid_needs_update - check whether inode's i_uid needs to be updated
1372	* @idmap: idmap of the mount the inode was found from
1373	* @attr: the new attributes of @inode
1374	* @inode: the inode to update
1375	*
1376	* Check whether the $inode's i_uid field needs to be updated taking idmapped
1377	* mounts into account if the filesystem supports it.
1378	*
1379	* Return: true if @inode's i_uid field needs to be updated, false if not.
1380	*/
1381	static inline bool i_uid_needs_update(struct mnt_idmap *idmap,
1382	const struct iattr *attr,
1383	const struct inode *inode)
1384	{
1385	return ((attr->ia_valid & ATTR_UID) &&
1386	!vfsuid_eq(attr->ia_vfsuid,
1387	i_uid_into_vfsuid(idmap, inode)));
1388	}
1389
1390	/**
1391	* i_uid_update - update @inode's i_uid field
1392	* @idmap: idmap of the mount the inode was found from
1393	* @attr: the new attributes of @inode
1394	* @inode: the inode to update
1395	*
1396	* Safely update @inode's i_uid field translating the vfsuid of any idmapped
1397	* mount into the filesystem kuid.
1398	*/
1399	static inline void i_uid_update(struct mnt_idmap *idmap,
1400	const struct iattr *attr,
1401	struct inode *inode)
1402	{
1403	if (attr->ia_valid & ATTR_UID)
1404	inode->i_uid = from_vfsuid(idmap, i_user_ns(inode),
1405	attr->ia_vfsuid);
1406	}
1407
1408	/**
1409	* i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping
1410	* @idmap: idmap of the mount the inode was found from
1411	* @inode: inode to map
1412	*
1413	* Return: the inode's i_gid mapped down according to @idmap.
1414	* If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1415	*/
1416	static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap,
1417	const struct inode *inode)
1418	{
1419	return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid);
1420	}
1421
1422	/**
1423	* i_gid_needs_update - check whether inode's i_gid needs to be updated
1424	* @idmap: idmap of the mount the inode was found from
1425	* @attr: the new attributes of @inode
1426	* @inode: the inode to update
1427	*
1428	* Check whether the $inode's i_gid field needs to be updated taking idmapped
1429	* mounts into account if the filesystem supports it.
1430	*
1431	* Return: true if @inode's i_gid field needs to be updated, false if not.
1432	*/
1433	static inline bool i_gid_needs_update(struct mnt_idmap *idmap,
1434	const struct iattr *attr,
1435	const struct inode *inode)
1436	{
1437	return ((attr->ia_valid & ATTR_GID) &&
1438	!vfsgid_eq(attr->ia_vfsgid,
1439	i_gid_into_vfsgid(idmap, inode)));
1440	}
1441
1442	/**
1443	* i_gid_update - update @inode's i_gid field
1444	* @idmap: idmap of the mount the inode was found from
1445	* @attr: the new attributes of @inode
1446	* @inode: the inode to update
1447	*
1448	* Safely update @inode's i_gid field translating the vfsgid of any idmapped
1449	* mount into the filesystem kgid.
1450	*/
1451	static inline void i_gid_update(struct mnt_idmap *idmap,
1452	const struct iattr *attr,
1453	struct inode *inode)
1454	{
1455	if (attr->ia_valid & ATTR_GID)
1456	inode->i_gid = from_vfsgid(idmap, i_user_ns(inode),
1457	attr->ia_vfsgid);
1458	}
1459
1460	/**
1461	* inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1462	* @inode: inode to initialize
1463	* @idmap: idmap of the mount the inode was found from
1464	*
1465	* Initialize the i_uid field of @inode. If the inode was found/created via
1466	* an idmapped mount map the caller's fsuid according to @idmap.
1467	*/
1468	static inline void inode_fsuid_set(struct inode *inode,
1469	struct mnt_idmap *idmap)
1470	{
1471	inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode));
1472	}
1473
1474	/**
1475	* inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1476	* @inode: inode to initialize
1477	* @idmap: idmap of the mount the inode was found from
1478	*
1479	* Initialize the i_gid field of @inode. If the inode was found/created via
1480	* an idmapped mount map the caller's fsgid according to @idmap.
1481	*/
1482	static inline void inode_fsgid_set(struct inode *inode,
1483	struct mnt_idmap *idmap)
1484	{
1485	inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode));
1486	}
1487
1488	/**
1489	* fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1490	* @sb: the superblock we want a mapping in
1491	* @idmap: idmap of the relevant mount
1492	*
1493	* Check whether the caller's fsuid and fsgid have a valid mapping in the
1494	* s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1495	* the caller's fsuid and fsgid according to the @idmap first.
1496	*
1497	* Return: true if fsuid and fsgid is mapped, false if not.
1498	*/
1499	static inline bool fsuidgid_has_mapping(struct super_block *sb,
1500	struct mnt_idmap *idmap)
1501	{
1502	struct user_namespace *fs_userns = sb->s_user_ns;
1503	kuid_t kuid;
1504	kgid_t kgid;
1505
1506	kuid = mapped_fsuid(idmap, fs_userns);
1507	if (!uid_valid(kuid))
1508	return false;
1509	kgid = mapped_fsgid(idmap, fs_userns);
1510	if (!gid_valid(kgid))
1511	return false;
1512	return kuid_has_mapping(fs_userns, kuid) &&
1513	kgid_has_mapping(fs_userns, kgid);
1514	}
1515
1516	struct timespec64 current_time(struct inode *inode);
1517	struct timespec64 inode_set_ctime_current(struct inode *inode);
1518
1519	static inline time64_t inode_get_atime_sec(const struct inode *inode)
1520	{
1521	return inode->__i_atime.tv_sec;
1522	}
1523
1524	static inline long inode_get_atime_nsec(const struct inode *inode)
1525	{
1526	return inode->__i_atime.tv_nsec;
1527	}
1528
1529	static inline struct timespec64 inode_get_atime(const struct inode *inode)
1530	{
1531	return inode->__i_atime;
1532	}
1533
1534	static inline struct timespec64 inode_set_atime_to_ts(struct inode *inode,
1535	struct timespec64 ts)
1536	{
1537	inode->__i_atime = ts;
1538	return ts;
1539	}
1540
1541	static inline struct timespec64 inode_set_atime(struct inode *inode,
1542	time64_t sec, long nsec)
1543	{
1544	struct timespec64 ts = { .tv_sec = sec,
1545	.tv_nsec = nsec };
1546	return inode_set_atime_to_ts(inode, ts);
1547	}
1548
1549	static inline time64_t inode_get_mtime_sec(const struct inode *inode)
1550	{
1551	return inode->__i_mtime.tv_sec;
1552	}
1553
1554	static inline long inode_get_mtime_nsec(const struct inode *inode)
1555	{
1556	return inode->__i_mtime.tv_nsec;
1557	}
1558
1559	static inline struct timespec64 inode_get_mtime(const struct inode *inode)
1560	{
1561	return inode->__i_mtime;
1562	}
1563
1564	static inline struct timespec64 inode_set_mtime_to_ts(struct inode *inode,
1565	struct timespec64 ts)
1566	{
1567	inode->__i_mtime = ts;
1568	return ts;
1569	}
1570
1571	static inline struct timespec64 inode_set_mtime(struct inode *inode,
1572	time64_t sec, long nsec)
1573	{
1574	struct timespec64 ts = { .tv_sec = sec,
1575	.tv_nsec = nsec };
1576	return inode_set_mtime_to_ts(inode, ts);
1577	}
1578
1579	static inline time64_t inode_get_ctime_sec(const struct inode *inode)
1580	{
1581	return inode->__i_ctime.tv_sec;
1582	}
1583
1584	static inline long inode_get_ctime_nsec(const struct inode *inode)
1585	{
1586	return inode->__i_ctime.tv_nsec;
1587	}
1588
1589	static inline struct timespec64 inode_get_ctime(const struct inode *inode)
1590	{
1591	return inode->__i_ctime;
1592	}
1593
1594	static inline struct timespec64 inode_set_ctime_to_ts(struct inode *inode,
1595	struct timespec64 ts)
1596	{
1597	inode->__i_ctime = ts;
1598	return ts;
1599	}
1600
1601	/**
1602	* inode_set_ctime - set the ctime in the inode
1603	* @inode: inode in which to set the ctime
1604	* @sec: tv_sec value to set
1605	* @nsec: tv_nsec value to set
1606	*
1607	* Set the ctime in @inode to { @sec, @nsec }
1608	*/
1609	static inline struct timespec64 inode_set_ctime(struct inode *inode,
1610	time64_t sec, long nsec)
1611	{
1612	struct timespec64 ts = { .tv_sec = sec,
1613	.tv_nsec = nsec };
1614
1615	return inode_set_ctime_to_ts(inode, ts);
1616	}
1617
1618	struct timespec64 simple_inode_init_ts(struct inode *inode);
1619
1620	/*
1621	* Snapshotting support.
1622	*/
1623
1624	/*
1625	* These are internal functions, please use sb_start_{write,pagefault,intwrite}
1626	* instead.
1627	*/
1628	static inline void __sb_end_write(struct super_block *sb, int level)
1629	{
1630	percpu_up_read(sb->s_writers.rw_sem + level-1);
1631	}
1632
1633	static inline void __sb_start_write(struct super_block *sb, int level)
1634	{
1635	percpu_down_read(sb->s_writers.rw_sem + level - 1);
1636	}
1637
1638	static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1639	{
1640	return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1641	}
1642
1643	#define __sb_writers_acquired(sb, lev) \
1644	percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1645	#define __sb_writers_release(sb, lev) \
1646	percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1647
1648	static inline bool sb_write_started(const struct super_block *sb)
1649	{
1650	return lockdep_is_held_type(sb->s_writers.rw_sem + SB_FREEZE_WRITE - 1, 1);
1651	}
1652
1653	/**
1654	* sb_end_write - drop write access to a superblock
1655	* @sb: the super we wrote to
1656	*
1657	* Decrement number of writers to the filesystem. Wake up possible waiters
1658	* wanting to freeze the filesystem.
1659	*/
1660	static inline void sb_end_write(struct super_block *sb)
1661	{
1662	__sb_end_write(sb, SB_FREEZE_WRITE);
1663	}
1664
1665	/**
1666	* sb_end_pagefault - drop write access to a superblock from a page fault
1667	* @sb: the super we wrote to
1668	*
1669	* Decrement number of processes handling write page fault to the filesystem.
1670	* Wake up possible waiters wanting to freeze the filesystem.
1671	*/
1672	static inline void sb_end_pagefault(struct super_block *sb)
1673	{
1674	__sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1675	}
1676
1677	/**
1678	* sb_end_intwrite - drop write access to a superblock for internal fs purposes
1679	* @sb: the super we wrote to
1680	*
1681	* Decrement fs-internal number of writers to the filesystem. Wake up possible
1682	* waiters wanting to freeze the filesystem.
1683	*/
1684	static inline void sb_end_intwrite(struct super_block *sb)
1685	{
1686	__sb_end_write(sb, SB_FREEZE_FS);
1687	}
1688
1689	/**
1690	* sb_start_write - get write access to a superblock
1691	* @sb: the super we write to
1692	*
1693	* When a process wants to write data or metadata to a file system (i.e. dirty
1694	* a page or an inode), it should embed the operation in a sb_start_write() -
1695	* sb_end_write() pair to get exclusion against file system freezing. This
1696	* function increments number of writers preventing freezing. If the file
1697	* system is already frozen, the function waits until the file system is
1698	* thawed.
1699	*
1700	* Since freeze protection behaves as a lock, users have to preserve
1701	* ordering of freeze protection and other filesystem locks. Generally,
1702	* freeze protection should be the outermost lock. In particular, we have:
1703	*
1704	* sb_start_write
1705	* -> i_mutex (write path, truncate, directory ops, ...)
1706	* -> s_umount (freeze_super, thaw_super)
1707	*/
1708	static inline void sb_start_write(struct super_block *sb)
1709	{
1710	__sb_start_write(sb, SB_FREEZE_WRITE);
1711	}
1712
1713	static inline bool sb_start_write_trylock(struct super_block *sb)
1714	{
1715	return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1716	}
1717
1718	/**
1719	* sb_start_pagefault - get write access to a superblock from a page fault
1720	* @sb: the super we write to
1721	*
1722	* When a process starts handling write page fault, it should embed the
1723	* operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1724	* exclusion against file system freezing. This is needed since the page fault
1725	* is going to dirty a page. This function increments number of running page
1726	* faults preventing freezing. If the file system is already frozen, the
1727	* function waits until the file system is thawed.
1728	*
1729	* Since page fault freeze protection behaves as a lock, users have to preserve
1730	* ordering of freeze protection and other filesystem locks. It is advised to
1731	* put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1732	* handling code implies lock dependency:
1733	*
1734	* mmap_lock
1735	* -> sb_start_pagefault
1736	*/
1737	static inline void sb_start_pagefault(struct super_block *sb)
1738	{
1739	__sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1740	}
1741
1742	/**
1743	* sb_start_intwrite - get write access to a superblock for internal fs purposes
1744	* @sb: the super we write to
1745	*
1746	* This is the third level of protection against filesystem freezing. It is
1747	* free for use by a filesystem. The only requirement is that it must rank
1748	* below sb_start_pagefault.
1749	*
1750	* For example filesystem can call sb_start_intwrite() when starting a
1751	* transaction which somewhat eases handling of freezing for internal sources
1752	* of filesystem changes (internal fs threads, discarding preallocation on file
1753	* close, etc.).
1754	*/
1755	static inline void sb_start_intwrite(struct super_block *sb)
1756	{
1757	__sb_start_write(sb, SB_FREEZE_FS);
1758	}
1759
1760	static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1761	{
1762	return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1763	}
1764
1765	bool inode_owner_or_capable(struct mnt_idmap *idmap,
1766	const struct inode *inode);
1767
1768	/*
1769	* VFS helper functions..
1770	*/
1771	int vfs_create(struct mnt_idmap , struct inode ,
1772	struct dentry *, umode_t, bool);
1773	int vfs_mkdir(struct mnt_idmap , struct inode ,
1774	struct dentry *, umode_t);
1775	int vfs_mknod(struct mnt_idmap , struct inode , struct dentry *,
1776	umode_t, dev_t);
1777	int vfs_symlink(struct mnt_idmap , struct inode ,
1778	struct dentry , const char );
1779	int vfs_link(struct dentry , struct mnt_idmap , struct inode *,
1780	struct dentry , struct inode *);
1781	int vfs_rmdir(struct mnt_idmap , struct inode , struct dentry *);
1782	int vfs_unlink(struct mnt_idmap , struct inode , struct dentry *,
1783	struct inode **);
1784
1785	/**
1786	* struct renamedata - contains all information required for renaming
1787	* @old_mnt_idmap: idmap of the old mount the inode was found from
1788	* @old_dir: parent of source
1789	* @old_dentry: source
1790	* @new_mnt_idmap: idmap of the new mount the inode was found from
1791	* @new_dir: parent of destination
1792	* @new_dentry: destination
1793	* @delegated_inode: returns an inode needing a delegation break
1794	* @flags: rename flags
1795	*/
1796	struct renamedata {
1797	struct mnt_idmap *old_mnt_idmap;
1798	struct inode *old_dir;
1799	struct dentry *old_dentry;
1800	struct mnt_idmap *new_mnt_idmap;
1801	struct inode *new_dir;
1802	struct dentry *new_dentry;
1803	struct inode **delegated_inode;
1804	unsigned int flags;
1805	} __randomize_layout;
1806
1807	int vfs_rename(struct renamedata *);
1808
1809	static inline int vfs_whiteout(struct mnt_idmap *idmap,
1810	struct inode dir, struct dentry dentry)
1811	{
1812	return vfs_mknod(idmap, dir, dentry, S_IFCHR \| WHITEOUT_MODE,
1813	WHITEOUT_DEV);
1814	}
1815
1816	struct file kernel_tmpfile_open(struct mnt_idmap idmap,
1817	const struct path *parentpath,
1818	umode_t mode, int open_flag,
1819	const struct cred *cred);
1820	struct file kernel_file_open(const struct path path, int flags,
1821	struct inode inode, const struct cred cred);
1822
1823	int vfs_mkobj(struct dentry *, umode_t,
1824	int (f)(struct dentry , umode_t, void *),
1825	void *);
1826
1827	int vfs_fchown(struct file *file, uid_t user, gid_t group);
1828	int vfs_fchmod(struct file *file, umode_t mode);
1829	int vfs_utimes(const struct path path, struct timespec64 times);
1830
1831	extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1832
1833	#ifdef CONFIG_COMPAT
1834	extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
1835	unsigned long arg);
1836	#else
1837	#define compat_ptr_ioctl NULL
1838	#endif
1839
1840	/*
1841	* VFS file helper functions.
1842	*/
1843	void inode_init_owner(struct mnt_idmap idmap, struct inode inode,
1844	const struct inode *dir, umode_t mode);
1845	extern bool may_open_dev(const struct path *path);
1846	umode_t mode_strip_sgid(struct mnt_idmap *idmap,
1847	const struct inode *dir, umode_t mode);
1848
1849	/*
1850	* This is the "filldir" function type, used by readdir() to let
1851	* the kernel specify what kind of dirent layout it wants to have.
1852	* This allows the kernel to read directories into kernel space or
1853	* to have different dirent layouts depending on the binary type.
1854	* Return 'true' to keep going and 'false' if there are no more entries.
1855	*/
1856	struct dir_context;
1857	typedef bool (filldir_t)(struct dir_context , const char *, int, loff_t, u64,
1858	unsigned);
1859
1860	struct dir_context {
1861	filldir_t actor;
1862	loff_t pos;
1863	};
1864
1865	/*
1866	* These flags let !MMU mmap() govern direct device mapping vs immediate
1867	* copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1868	*
1869	* NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE)
1870	* NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED)
1871	* NOMMU_MAP_READ: Can be mapped for reading
1872	* NOMMU_MAP_WRITE: Can be mapped for writing
1873	* NOMMU_MAP_EXEC: Can be mapped for execution
1874	*/
1875	#define NOMMU_MAP_COPY 0x00000001
1876	#define NOMMU_MAP_DIRECT 0x00000008
1877	#define NOMMU_MAP_READ VM_MAYREAD
1878	#define NOMMU_MAP_WRITE VM_MAYWRITE
1879	#define NOMMU_MAP_EXEC VM_MAYEXEC
1880
1881	#define NOMMU_VMFLAGS \
1882	(NOMMU_MAP_READ \| NOMMU_MAP_WRITE \| NOMMU_MAP_EXEC)
1883
1884	/*
1885	* These flags control the behavior of the remap_file_range function pointer.
1886	* If it is called with len == 0 that means "remap to end of source file".
1887	* See Documentation/filesystems/vfs.rst for more details about this call.
1888	*
1889	* REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
1890	* REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
1891	*/
1892	#define REMAP_FILE_DEDUP (1 << 0)
1893	#define REMAP_FILE_CAN_SHORTEN (1 << 1)
1894
1895	/*
1896	* These flags signal that the caller is ok with altering various aspects of
1897	* the behavior of the remap operation. The changes must be made by the
1898	* implementation; the vfs remap helper functions can take advantage of them.
1899	* Flags in this category exist to preserve the quirky behavior of the hoisted
1900	* btrfs clone/dedupe ioctls.
1901	*/
1902	#define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN)
1903
1904	/*
1905	* These flags control the behavior of vfs_copy_file_range().
1906	* They are not available to the user via syscall.
1907	*
1908	* COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops
1909	*/
1910	#define COPY_FILE_SPLICE (1 << 0)
1911
1912	struct iov_iter;
1913	struct io_uring_cmd;
1914	struct offset_ctx;
1915
1916	struct file_operations {
1917	struct module *owner;
1918	loff_t (llseek) (struct file , loff_t, int);
1919	ssize_t (read) (struct file , char __user , size_t, loff_t );
1920	ssize_t (write) (struct file , const char __user , size_t, loff_t );
1921	ssize_t (read_iter) (struct kiocb , struct iov_iter *);
1922	ssize_t (write_iter) (struct kiocb , struct iov_iter *);
1923	int (iopoll)(struct kiocb kiocb, struct io_comp_batch *,
1924	unsigned int flags);
1925	int (iterate_shared) (struct file , struct dir_context *);
1926	__poll_t (poll) (struct file , struct poll_table_struct *);
1927	long (unlocked_ioctl) (struct file , unsigned int, unsigned long);
1928	long (compat_ioctl) (struct file , unsigned int, unsigned long);
1929	int (mmap) (struct file , struct vm_area_struct *);
1930	unsigned long mmap_supported_flags;
1931	int (open) (struct inode , struct file *);
1932	int (flush) (struct file , fl_owner_t id);
1933	int (release) (struct inode , struct file *);
1934	int (fsync) (struct file , loff_t, loff_t, int datasync);
1935	int (fasync) (int, struct file , int);
1936	int (lock) (struct file , int, struct file_lock *);
1937	unsigned long (get_unmapped_area)(struct file , unsigned long, unsigned long, unsigned long, unsigned long);
1938	int (*check_flags)(int);
1939	int (flock) (struct file , int, struct file_lock *);
1940	ssize_t (splice_write)(struct pipe_inode_info , struct file , loff_t , size_t, unsigned int);
1941	ssize_t (splice_read)(struct file , loff_t , struct pipe_inode_info , size_t, unsigned int);
1942	void (splice_eof)(struct file file);
1943	int (setlease)(struct file , int, struct file_lock , void );
1944	long (fallocate)(struct file file, int mode, loff_t offset,
1945	loff_t len);
1946	void (show_fdinfo)(struct seq_file m, struct file *f);
1947	#ifndef CONFIG_MMU
1948	unsigned (mmap_capabilities)(struct file );
1949	#endif
1950	ssize_t (copy_file_range)(struct file , loff_t, struct file *,
1951	loff_t, size_t, unsigned int);
1952	loff_t (remap_file_range)(struct file file_in, loff_t pos_in,
1953	struct file *file_out, loff_t pos_out,
1954	loff_t len, unsigned int remap_flags);
1955	int (fadvise)(struct file , loff_t, loff_t, int);
1956	int (uring_cmd)(struct io_uring_cmd ioucmd, unsigned int issue_flags);
1957	int (uring_cmd_iopoll)(struct io_uring_cmd , struct io_comp_batch *,
1958	unsigned int poll_flags);
1959	} __randomize_layout;
1960
1961	/* Wrap a directory iterator that needs exclusive inode access */
1962	int wrap_directory_iterator(struct file , struct dir_context ,
1963	int () (struct file , struct dir_context *));
1964	#define WRAP_DIR_ITER(x) \
1965	static int shared_##x(struct file file , struct dir_context ctx) \
1966	{ return wrap_directory_iterator(file, ctx, x); }
1967
1968	struct inode_operations {
1969	struct dentry * (lookup) (struct inode ,struct dentry *, unsigned int);
1970	const char * (get_link) (struct dentry , struct inode , struct delayed_call );
1971	int (permission) (struct mnt_idmap , struct inode *, int);
1972	struct posix_acl * (get_inode_acl)(struct inode , int, bool);
1973
1974	int (readlink) (struct dentry , char __user *,int);
1975
1976	int (create) (struct mnt_idmap , struct inode ,struct dentry ,
1977	umode_t, bool);
1978	int (link) (struct dentry ,struct inode ,struct dentry );
1979	int (unlink) (struct inode ,struct dentry *);
1980	int (symlink) (struct mnt_idmap , struct inode ,struct dentry ,
1981	const char *);
1982	int (mkdir) (struct mnt_idmap , struct inode ,struct dentry ,
1983	umode_t);
1984	int (rmdir) (struct inode ,struct dentry *);
1985	int (mknod) (struct mnt_idmap , struct inode ,struct dentry ,
1986	umode_t,dev_t);
1987	int (rename) (struct mnt_idmap , struct inode , struct dentry ,
1988	struct inode , struct dentry , unsigned int);
1989	int (setattr) (struct mnt_idmap , struct dentry , struct iattr );
1990	int (getattr) (struct mnt_idmap , const struct path *,
1991	struct kstat *, u32, unsigned int);
1992	ssize_t (listxattr) (struct dentry , char *, size_t);
1993	int (fiemap)(struct inode , struct fiemap_extent_info *, u64 start,
1994	u64 len);
1995	int (update_time)(struct inode , int);
1996	int (atomic_open)(struct inode , struct dentry *,
1997	struct file *, unsigned open_flag,
1998	umode_t create_mode);
1999	int (tmpfile) (struct mnt_idmap , struct inode *,
2000	struct file *, umode_t);
2001	struct posix_acl (get_acl)(struct mnt_idmap , struct dentry ,
2002	int);
2003	int (set_acl)(struct mnt_idmap , struct dentry *,
2004	struct posix_acl *, int);
2005	int (fileattr_set)(struct mnt_idmap idmap,
2006	struct dentry dentry, struct fileattr fa);
2007	int (fileattr_get)(struct dentry dentry, struct fileattr *fa);
2008	struct offset_ctx (get_offset_ctx)(struct inode *inode);
2009	} ____cacheline_aligned;
2010
2011	static inline ssize_t call_read_iter(struct file file, struct kiocb kio,
2012	struct iov_iter *iter)
2013	{
2014	return file->f_op->read_iter(kio, iter);
2015	}
2016
2017	static inline ssize_t call_write_iter(struct file file, struct kiocb kio,
2018	struct iov_iter *iter)
2019	{
2020	return file->f_op->write_iter(kio, iter);
2021	}
2022
2023	static inline int call_mmap(struct file file, struct vm_area_struct vma)
2024	{
2025	return file->f_op->mmap(file, vma);
2026	}
2027
2028	extern ssize_t vfs_read(struct file , char __user , size_t, loff_t *);
2029	extern ssize_t vfs_write(struct file , const char __user , size_t, loff_t *);
2030	extern ssize_t vfs_copy_file_range(struct file , loff_t , struct file ,
2031	loff_t, size_t, unsigned int);
2032	extern ssize_t generic_copy_file_range(struct file *file_in, loff_t pos_in,
2033	struct file *file_out, loff_t pos_out,
2034	size_t len, unsigned int flags);
2035	int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2036	struct file *file_out, loff_t pos_out,
2037	loff_t *len, unsigned int remap_flags,
2038	const struct iomap_ops *dax_read_ops);
2039	int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2040	struct file *file_out, loff_t pos_out,
2041	loff_t *count, unsigned int remap_flags);
2042	extern loff_t do_clone_file_range(struct file *file_in, loff_t pos_in,
2043	struct file *file_out, loff_t pos_out,
2044	loff_t len, unsigned int remap_flags);
2045	extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
2046	struct file *file_out, loff_t pos_out,
2047	loff_t len, unsigned int remap_flags);
2048	extern int vfs_dedupe_file_range(struct file *file,
2049	struct file_dedupe_range *same);
2050	extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
2051	struct file *dst_file, loff_t dst_pos,
2052	loff_t len, unsigned int remap_flags);
2053
2054	enum freeze_holder {
2055	FREEZE_HOLDER_KERNEL = (1U << 0),
2056	FREEZE_HOLDER_USERSPACE = (1U << 1),
2057	};
2058
2059	struct super_operations {
2060	struct inode (alloc_inode)(struct super_block *sb);
2061	void (destroy_inode)(struct inode );
2062	void (free_inode)(struct inode );
2063
2064	void (dirty_inode) (struct inode , int flags);
2065	int (write_inode) (struct inode , struct writeback_control *wbc);
2066	int (drop_inode) (struct inode );
2067	void (evict_inode) (struct inode );
2068	void (put_super) (struct super_block );
2069	int (sync_fs)(struct super_block sb, int wait);
2070	int (freeze_super) (struct super_block , enum freeze_holder who);
2071	int (freeze_fs) (struct super_block );
2072	int (thaw_super) (struct super_block , enum freeze_holder who);
2073	int (unfreeze_fs) (struct super_block );
2074	int (statfs) (struct dentry , struct kstatfs *);
2075	int (remount_fs) (struct super_block , int , char );
2076	void (umount_begin) (struct super_block );
2077
2078	int (show_options)(struct seq_file , struct dentry *);
2079	int (show_devname)(struct seq_file , struct dentry *);
2080	int (show_path)(struct seq_file , struct dentry *);
2081	int (show_stats)(struct seq_file , struct dentry *);
2082	#ifdef CONFIG_QUOTA
2083	ssize_t (quota_read)(struct super_block , int, char *, size_t, loff_t);
2084	ssize_t (quota_write)(struct super_block , int, const char *, size_t, loff_t);
2085	struct dquot *(get_dquots)(struct inode *);
2086	#endif
2087	long (nr_cached_objects)(struct super_block ,
2088	struct shrink_control *);
2089	long (free_cached_objects)(struct super_block ,
2090	struct shrink_control *);
2091	void (shutdown)(struct super_block sb);
2092	};
2093
2094	/*
2095	* Inode flags - they have no relation to superblock flags now
2096	*/
2097	#define S_SYNC (1 << 0) /* Writes are synced at once */
2098	#define S_NOATIME (1 << 1) /* Do not update access times */
2099	#define S_APPEND (1 << 2) /* Append-only file */
2100	#define S_IMMUTABLE (1 << 3) /* Immutable file */
2101	#define S_DEAD (1 << 4) /* removed, but still open directory */
2102	#define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */
2103	#define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */
2104	#define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */
2105	#define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */
2106	#define S_PRIVATE (1 << 9) /* Inode is fs-internal */
2107	#define S_IMA (1 << 10) /* Inode has an associated IMA struct */
2108	#define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */
2109	#define S_NOSEC (1 << 12) /* no suid or xattr security attributes */
2110	#ifdef CONFIG_FS_DAX
2111	#define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */
2112	#else
2113	#define S_DAX 0 /* Make all the DAX code disappear */
2114	#endif
2115	#define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */
2116	#define S_CASEFOLD (1 << 15) /* Casefolded file */
2117	#define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */
2118	#define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
2119
2120	/*
2121	* Note that nosuid etc flags are inode-specific: setting some file-system
2122	* flags just means all the inodes inherit those flags by default. It might be
2123	* possible to override it selectively if you really wanted to with some
2124	* ioctl() that is not currently implemented.
2125	*
2126	* Exception: SB_RDONLY is always applied to the entire file system.
2127	*
2128	* Unfortunately, it is possible to change a filesystems flags with it mounted
2129	* with files in use. This means that all of the inodes will not have their
2130	* i_flags updated. Hence, i_flags no longer inherit the superblock mount
2131	* flags, so these have to be checked separately. -- [email protected]
2132	*/
2133	#define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
2134
2135	static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
2136	#define IS_RDONLY(inode) sb_rdonly((inode)->i_sb)
2137	#define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) \|\| \
2138	((inode)->i_flags & S_SYNC))
2139	#define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS\|SB_DIRSYNC) \|\| \
2140	((inode)->i_flags & (S_SYNC\|S_DIRSYNC)))
2141	#define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK)
2142	#define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY\|SB_NOATIME)
2143	#define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION)
2144
2145	#define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
2146	#define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
2147	#define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
2148
2149	#ifdef CONFIG_FS_POSIX_ACL
2150	#define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL)
2151	#else
2152	#define IS_POSIXACL(inode) 0
2153	#endif
2154
2155	#define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
2156	#define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
2157	#define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
2158	#define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
2159	#define IS_IMA(inode) ((inode)->i_flags & S_IMA)
2160	#define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
2161	#define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
2162	#define IS_DAX(inode) ((inode)->i_flags & S_DAX)
2163	#define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED)
2164	#define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD)
2165	#define IS_VERITY(inode) ((inode)->i_flags & S_VERITY)
2166
2167	#define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
2168	(inode)->i_rdev == WHITEOUT_DEV)
2169
2170	static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap,
2171	struct inode *inode)
2172	{
2173	return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) \|\|
2174	!vfsgid_valid(i_gid_into_vfsgid(idmap, inode));
2175	}
2176
2177	static inline void init_sync_kiocb(struct kiocb kiocb, struct file filp)
2178	{
2179	*kiocb = (struct kiocb) {
2180	.ki_filp = filp,
2181	.ki_flags = filp->f_iocb_flags,
2182	.ki_ioprio = get_current_ioprio(),
2183	};
2184	}
2185
2186	static inline void kiocb_clone(struct kiocb kiocb, struct kiocb kiocb_src,
2187	struct file *filp)
2188	{
2189	*kiocb = (struct kiocb) {
2190	.ki_filp = filp,
2191	.ki_flags = kiocb_src->ki_flags,
2192	.ki_ioprio = kiocb_src->ki_ioprio,
2193	.ki_pos = kiocb_src->ki_pos,
2194	};
2195	}
2196
2197	/*
2198	* Inode state bits. Protected by inode->i_lock
2199	*
2200	* Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2201	* I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2202	*
2203	* Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
2204	* until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
2205	* various stages of removing an inode.
2206	*
2207	* Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2208	*
2209	* I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
2210	* fdatasync() (unless I_DIRTY_DATASYNC is also set).
2211	* Timestamp updates are the usual cause.
2212	* I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
2213	* these changes separately from I_DIRTY_SYNC so that we
2214	* don't have to write inode on fdatasync() when only
2215	* e.g. the timestamps have changed.
2216	* I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
2217	* I_DIRTY_TIME The inode itself has dirty timestamps, and the
2218	* lazytime mount option is enabled. We keep track of this
2219	* separately from I_DIRTY_SYNC in order to implement
2220	* lazytime. This gets cleared if I_DIRTY_INODE
2221	* (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But
2222	* I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already
2223	* in place because writeback might already be in progress
2224	* and we don't want to lose the time update
2225	* I_NEW Serves as both a mutex and completion notification.
2226	* New inodes set I_NEW. If two processes both create
2227	* the same inode, one of them will release its inode and
2228	* wait for I_NEW to be released before returning.
2229	* Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2230	* also cause waiting on I_NEW, without I_NEW actually
2231	* being set. find_inode() uses this to prevent returning
2232	* nearly-dead inodes.
2233	* I_WILL_FREE Must be set when calling write_inode_now() if i_count
2234	* is zero. I_FREEING must be set when I_WILL_FREE is
2235	* cleared.
2236	* I_FREEING Set when inode is about to be freed but still has dirty
2237	* pages or buffers attached or the inode itself is still
2238	* dirty.
2239	* I_CLEAR Added by clear_inode(). In this state the inode is
2240	* clean and can be destroyed. Inode keeps I_FREEING.
2241	*
2242	* Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2243	* prohibited for many purposes. iget() must wait for
2244	* the inode to be completely released, then create it
2245	* anew. Other functions will just ignore such inodes,
2246	* if appropriate. I_NEW is used for waiting.
2247	*
2248	* I_SYNC Writeback of inode is running. The bit is set during
2249	* data writeback, and cleared with a wakeup on the bit
2250	* address once it is done. The bit is also used to pin
2251	* the inode in memory for flusher thread.
2252	*
2253	* I_REFERENCED Marks the inode as recently references on the LRU list.
2254	*
2255	* I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit().
2256	*
2257	* I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to
2258	* synchronize competing switching instances and to tell
2259	* wb stat updates to grab the i_pages lock. See
2260	* inode_switch_wbs_work_fn() for details.
2261	*
2262	* I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper
2263	* and work dirs among overlayfs mounts.
2264	*
2265	* I_CREATING New object's inode in the middle of setting up.
2266	*
2267	* I_DONTCACHE Evict inode as soon as it is not used anymore.
2268	*
2269	* I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists.
2270	* Used to detect that mark_inode_dirty() should not move
2271	* inode between dirty lists.
2272	*
2273	* I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback.
2274	*
2275	* Q: What is the difference between I_WILL_FREE and I_FREEING?
2276	*/
2277	#define I_DIRTY_SYNC (1 << 0)
2278	#define I_DIRTY_DATASYNC (1 << 1)
2279	#define I_DIRTY_PAGES (1 << 2)
2280	#define __I_NEW 3
2281	#define I_NEW (1 << __I_NEW)
2282	#define I_WILL_FREE (1 << 4)
2283	#define I_FREEING (1 << 5)
2284	#define I_CLEAR (1 << 6)
2285	#define __I_SYNC 7
2286	#define I_SYNC (1 << __I_SYNC)
2287	#define I_REFERENCED (1 << 8)
2288	#define __I_DIO_WAKEUP 9
2289	#define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP)
2290	#define I_LINKABLE (1 << 10)
2291	#define I_DIRTY_TIME (1 << 11)
2292	#define I_WB_SWITCH (1 << 13)
2293	#define I_OVL_INUSE (1 << 14)
2294	#define I_CREATING (1 << 15)
2295	#define I_DONTCACHE (1 << 16)
2296	#define I_SYNC_QUEUED (1 << 17)
2297	#define I_PINNING_FSCACHE_WB (1 << 18)
2298
2299	#define I_DIRTY_INODE (I_DIRTY_SYNC \| I_DIRTY_DATASYNC)
2300	#define I_DIRTY (I_DIRTY_INODE \| I_DIRTY_PAGES)
2301	#define I_DIRTY_ALL (I_DIRTY \| I_DIRTY_TIME)
2302
2303	extern void __mark_inode_dirty(struct inode *, int);
2304	static inline void mark_inode_dirty(struct inode *inode)
2305	{
2306	__mark_inode_dirty(inode, I_DIRTY);
2307	}
2308
2309	static inline void mark_inode_dirty_sync(struct inode *inode)
2310	{
2311	__mark_inode_dirty(inode, I_DIRTY_SYNC);
2312	}
2313
2314	/*
2315	* Returns true if the given inode itself only has dirty timestamps (its pages
2316	* may still be dirty) and isn't currently being allocated or freed.
2317	* Filesystems should call this if when writing an inode when lazytime is
2318	* enabled, they want to opportunistically write the timestamps of other inodes
2319	* located very nearby on-disk, e.g. in the same inode block. This returns true
2320	* if the given inode is in need of such an opportunistic update. Requires
2321	* i_lock, or at least later re-checking under i_lock.
2322	*/
2323	static inline bool inode_is_dirtytime_only(struct inode *inode)
2324	{
2325	return (inode->i_state & (I_DIRTY_TIME \| I_NEW \|
2326	I_FREEING \| I_WILL_FREE)) == I_DIRTY_TIME;
2327	}
2328
2329	extern void inc_nlink(struct inode *inode);
2330	extern void drop_nlink(struct inode *inode);
2331	extern void clear_nlink(struct inode *inode);
2332	extern void set_nlink(struct inode *inode, unsigned int nlink);
2333
2334	static inline void inode_inc_link_count(struct inode *inode)
2335	{
2336	inc_nlink(inode);
2337	mark_inode_dirty(inode);
2338	}
2339
2340	static inline void inode_dec_link_count(struct inode *inode)
2341	{
2342	drop_nlink(inode);
2343	mark_inode_dirty(inode);
2344	}
2345
2346	enum file_time_flags {
2347	S_ATIME = 1,
2348	S_MTIME = 2,
2349	S_CTIME = 4,
2350	S_VERSION = 8,
2351	};
2352
2353	extern bool atime_needs_update(const struct path , struct inode );
2354	extern void touch_atime(const struct path *);
2355	int inode_update_time(struct inode *inode, int flags);
2356
2357	static inline void file_accessed(struct file *file)
2358	{
2359	if (!(file->f_flags & O_NOATIME))
2360	touch_atime(&file->f_path);
2361	}
2362
2363	extern int file_modified(struct file *file);
2364	int kiocb_modified(struct kiocb *iocb);
2365
2366	int sync_inode_metadata(struct inode *inode, int wait);
2367
2368	struct file_system_type {
2369	const char *name;
2370	int fs_flags;
2371	#define FS_REQUIRES_DEV 1
2372	#define FS_BINARY_MOUNTDATA 2
2373	#define FS_HAS_SUBTYPE 4
2374	#define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
2375	#define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */
2376	#define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */
2377	#define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
2378	int (init_fs_context)(struct fs_context );
2379	const struct fs_parameter_spec *parameters;
2380	struct dentry (mount) (struct file_system_type *, int,
2381	const char , void );
2382	void (kill_sb) (struct super_block );
2383	struct module *owner;
2384	struct file_system_type * next;
2385	struct hlist_head fs_supers;
2386
2387	struct lock_class_key s_lock_key;
2388	struct lock_class_key s_umount_key;
2389	struct lock_class_key s_vfs_rename_key;
2390	struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2391
2392	struct lock_class_key i_lock_key;
2393	struct lock_class_key i_mutex_key;
2394	struct lock_class_key invalidate_lock_key;
2395	struct lock_class_key i_mutex_dir_key;
2396	};
2397
2398	#define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2399
2400	extern struct dentry mount_bdev(struct file_system_type fs_type,
2401	int flags, const char dev_name, void data,
2402	int (fill_super)(struct super_block , void *, int));
2403	extern struct dentry mount_single(struct file_system_type fs_type,
2404	int flags, void *data,
2405	int (fill_super)(struct super_block , void *, int));
2406	extern struct dentry mount_nodev(struct file_system_type fs_type,
2407	int flags, void *data,
2408	int (fill_super)(struct super_block , void *, int));
2409	extern struct dentry mount_subtree(struct vfsmount mnt, const char *path);
2410	void retire_super(struct super_block *sb);
2411	void generic_shutdown_super(struct super_block *sb);
2412	void kill_block_super(struct super_block *sb);
2413	void kill_anon_super(struct super_block *sb);
2414	void kill_litter_super(struct super_block *sb);
2415	void deactivate_super(struct super_block *sb);
2416	void deactivate_locked_super(struct super_block *sb);
2417	int set_anon_super(struct super_block s, void data);
2418	int set_anon_super_fc(struct super_block s, struct fs_context fc);
2419	int get_anon_bdev(dev_t *);
2420	void free_anon_bdev(dev_t);
2421	struct super_block sget_fc(struct fs_context fc,
2422	int (test)(struct super_block , struct fs_context *),
2423	int (set)(struct super_block , struct fs_context *));
2424	struct super_block sget(struct file_system_type type,
2425	int (test)(struct super_block ,void *),
2426	int (set)(struct super_block ,void *),
2427	int flags, void *data);
2428	struct super_block sget_dev(struct fs_context fc, dev_t dev);
2429
2430	/* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2431	#define fops_get(fops) \
2432	(((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2433	#define fops_put(fops) \
2434	do { if (fops) module_put((fops)->owner); } while(0)
2435	/*
2436	* This one is to be used ONLY from ->open() instances.
2437	* fops must be non-NULL, pinned down and module dependencies
2438	* should be sufficient to pin the caller down as well.
2439	*/
2440	#define replace_fops(f, fops) \
2441	do { \
2442	struct file *__file = (f); \
2443	fops_put(__file->f_op); \
2444	BUG_ON(!(__file->f_op = (fops))); \
2445	} while(0)
2446
2447	extern int register_filesystem(struct file_system_type *);
2448	extern int unregister_filesystem(struct file_system_type *);
2449	extern int vfs_statfs(const struct path , struct kstatfs );
2450	extern int user_statfs(const char __user , struct kstatfs );
2451	extern int fd_statfs(int, struct kstatfs *);
2452	int freeze_super(struct super_block *super, enum freeze_holder who);
2453	int thaw_super(struct super_block *super, enum freeze_holder who);
2454	extern __printf(2, 3)
2455	int super_setup_bdi_name(struct super_block sb, char fmt, ...);
2456	extern int super_setup_bdi(struct super_block *sb);
2457
2458	extern int current_umask(void);
2459
2460	extern void ihold(struct inode * inode);
2461	extern void iput(struct inode *);
2462	int inode_update_timestamps(struct inode *inode, int flags);
2463	int generic_update_time(struct inode *, int);
2464
2465	/* /sys/fs */
2466	extern struct kobject *fs_kobj;
2467
2468	#define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2469
2470	/* fs/open.c */
2471	struct audit_names;
2472	struct filename {
2473	const char name; / pointer to actual string */
2474	const __user char uptr; / original userland pointer */
2475	atomic_t refcnt;
2476	struct audit_names *aname;
2477	const char iname[];
2478	};
2479	static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2480
2481	static inline struct mnt_idmap file_mnt_idmap(const struct file file)
2482	{
2483	return mnt_idmap(file->f_path.mnt);
2484	}
2485
2486	/**
2487	* is_idmapped_mnt - check whether a mount is mapped
2488	* @mnt: the mount to check
2489	*
2490	* If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped.
2491	*
2492	* Return: true if mount is mapped, false if not.
2493	*/
2494	static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2495	{
2496	return mnt_idmap(mnt) != &nop_mnt_idmap;
2497	}
2498
2499	extern long vfs_truncate(const struct path *, loff_t);
2500	int do_truncate(struct mnt_idmap , struct dentry , loff_t start,
2501	unsigned int time_attrs, struct file *filp);
2502	extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2503	loff_t len);
2504	extern long do_sys_open(int dfd, const char __user *filename, int flags,
2505	umode_t mode);
2506	extern struct file file_open_name(struct filename , int, umode_t);
2507	extern struct file filp_open(const char , int, umode_t);
2508	extern struct file file_open_root(const struct path ,
2509	const char *, int, umode_t);
2510	static inline struct file file_open_root_mnt(struct vfsmount mnt,
2511	const char *name, int flags, umode_t mode)
2512	{
2513	return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2514	name, flags, mode);
2515	}
2516	struct file dentry_open(const struct path path, int flags,
2517	const struct cred *creds);
2518	struct file dentry_create(const struct path path, int flags, umode_t mode,
2519	const struct cred *cred);
2520	struct file backing_file_open(const struct path user_path, int flags,
2521	const struct path *real_path,
2522	const struct cred *cred);
2523	struct path backing_file_user_path(struct file f);
2524
2525	/*
2526	* file_user_path - get the path to display for memory mapped file
2527	*
2528	* When mmapping a file on a stackable filesystem (e.g., overlayfs), the file
2529	* stored in ->vm_file is a backing file whose f_inode is on the underlying
2530	* filesystem. When the mapped file path is displayed to user (e.g. via
2531	* /proc/<pid>/maps), this helper should be used to get the path to display
2532	* to the user, which is the path of the fd that user has requested to map.
2533	*/
2534	static inline const struct path file_user_path(struct file f)
2535	{
2536	if (unlikely(f->f_mode & FMODE_BACKING))
2537	return backing_file_user_path(f);
2538	return &f->f_path;
2539	}
2540
2541	static inline struct file file_clone_open(struct file file)
2542	{
2543	return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2544	}
2545	extern int filp_close(struct file *, fl_owner_t id);
2546
2547	extern struct filename getname_flags(const char __user , int, int *);
2548	extern struct filename getname_uflags(const char __user , int);
2549	extern struct filename getname(const char __user );
2550	extern struct filename getname_kernel(const char );
2551	extern void putname(struct filename *name);
2552
2553	extern int finish_open(struct file file, struct dentry dentry,
2554	int (open)(struct inode , struct file *));
2555	extern int finish_no_open(struct file file, struct dentry dentry);
2556
2557	/* Helper for the simple case when original dentry is used */
2558	static inline int finish_open_simple(struct file *file, int error)
2559	{
2560	if (error)
2561	return error;
2562
2563	return finish_open(file, file->f_path.dentry, NULL);
2564	}
2565
2566	/* fs/dcache.c */
2567	extern void __init vfs_caches_init_early(void);
2568	extern void __init vfs_caches_init(void);
2569
2570	extern struct kmem_cache *names_cachep;
2571
2572	#define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2573	#define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2574
2575	extern struct super_block *blockdev_superblock;
2576	static inline bool sb_is_blkdev_sb(struct super_block *sb)
2577	{
2578	return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2579	}
2580
2581	void emergency_thaw_all(void);
2582	extern int sync_filesystem(struct super_block *);
2583	extern const struct file_operations def_blk_fops;
2584	extern const struct file_operations def_chr_fops;
2585
2586	/* fs/char_dev.c */
2587	#define CHRDEV_MAJOR_MAX 512
2588	/* Marks the bottom of the first segment of free char majors */
2589	#define CHRDEV_MAJOR_DYN_END 234
2590	/* Marks the top and bottom of the second segment of free char majors */
2591	#define CHRDEV_MAJOR_DYN_EXT_START 511
2592	#define CHRDEV_MAJOR_DYN_EXT_END 384
2593
2594	extern int alloc_chrdev_region(dev_t , unsigned, unsigned, const char );
2595	extern int register_chrdev_region(dev_t, unsigned, const char *);
2596	extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2597	unsigned int count, const char *name,
2598	const struct file_operations *fops);
2599	extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2600	unsigned int count, const char *name);
2601	extern void unregister_chrdev_region(dev_t, unsigned);
2602	extern void chrdev_show(struct seq_file *,off_t);
2603
2604	static inline int register_chrdev(unsigned int major, const char *name,
2605	const struct file_operations *fops)
2606	{
2607	return __register_chrdev(major, 0, 256, name, fops);
2608	}
2609
2610	static inline void unregister_chrdev(unsigned int major, const char *name)
2611	{
2612	__unregister_chrdev(major, 0, 256, name);
2613	}
2614
2615	extern void init_special_inode(struct inode *, umode_t, dev_t);
2616
2617	/* Invalid inode operations -- fs/bad_inode.c */
2618	extern void make_bad_inode(struct inode *);
2619	extern bool is_bad_inode(struct inode *);
2620
2621	extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2622	loff_t lend);
2623	extern int __must_check file_check_and_advance_wb_err(struct file *file);
2624	extern int __must_check file_write_and_wait_range(struct file *file,
2625	loff_t start, loff_t end);
2626
2627	static inline int file_write_and_wait(struct file *file)
2628	{
2629	return file_write_and_wait_range(file, 0, LLONG_MAX);
2630	}
2631
2632	extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2633	int datasync);
2634	extern int vfs_fsync(struct file *file, int datasync);
2635
2636	extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2637	unsigned int flags);
2638
2639	static inline bool iocb_is_dsync(const struct kiocb *iocb)
2640	{
2641	return (iocb->ki_flags & IOCB_DSYNC) \|\|
2642	IS_SYNC(iocb->ki_filp->f_mapping->host);
2643	}
2644
2645	/*
2646	* Sync the bytes written if this was a synchronous write. Expect ki_pos
2647	* to already be updated for the write, and will return either the amount
2648	* of bytes passed in, or an error if syncing the file failed.
2649	*/
2650	static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2651	{
2652	if (iocb_is_dsync(iocb)) {
2653	int ret = vfs_fsync_range(iocb->ki_filp,
2654	iocb->ki_pos - count, iocb->ki_pos - 1,
2655	(iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2656	if (ret)
2657	return ret;
2658	}
2659
2660	return count;
2661	}
2662
2663	extern void emergency_sync(void);
2664	extern void emergency_remount(void);
2665
2666	#ifdef CONFIG_BLOCK
2667	extern int bmap(struct inode inode, sector_t block);
2668	#else
2669	static inline int bmap(struct inode inode, sector_t block)
2670	{
2671	return -EINVAL;
2672	}
2673	#endif
2674
2675	int notify_change(struct mnt_idmap , struct dentry ,
2676	struct iattr , struct inode *);
2677	int inode_permission(struct mnt_idmap , struct inode , int);
2678	int generic_permission(struct mnt_idmap , struct inode , int);
2679	static inline int file_permission(struct file *file, int mask)
2680	{
2681	return inode_permission(file_mnt_idmap(file),
2682	file_inode(file), mask);
2683	}
2684	static inline int path_permission(const struct path *path, int mask)
2685	{
2686	return inode_permission(mnt_idmap(path->mnt),
2687	d_inode(path->dentry), mask);
2688	}
2689	int __check_sticky(struct mnt_idmap idmap, struct inode dir,
2690	struct inode *inode);
2691
2692	static inline bool execute_ok(struct inode *inode)
2693	{
2694	return (inode->i_mode & S_IXUGO) \|\| S_ISDIR(inode->i_mode);
2695	}
2696
2697	static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2698	{
2699	return (inode->i_mode ^ mode) & S_IFMT;
2700	}
2701
2702	/**
2703	* file_start_write - get write access to a superblock for regular file io
2704	* @file: the file we want to write to
2705	*
2706	* This is a variant of sb_start_write() which is a noop on non-regualr file.
2707	* Should be matched with a call to file_end_write().
2708	*/
2709	static inline void file_start_write(struct file *file)
2710	{
2711	if (!S_ISREG(file_inode(file)->i_mode))
2712	return;
2713	sb_start_write(file_inode(file)->i_sb);
2714	}
2715
2716	static inline bool file_start_write_trylock(struct file *file)
2717	{
2718	if (!S_ISREG(file_inode(file)->i_mode))
2719	return true;
2720	return sb_start_write_trylock(file_inode(file)->i_sb);
2721	}
2722
2723	/**
2724	* file_end_write - drop write access to a superblock of a regular file
2725	* @file: the file we wrote to
2726	*
2727	* Should be matched with a call to file_start_write().
2728	*/
2729	static inline void file_end_write(struct file *file)
2730	{
2731	if (!S_ISREG(file_inode(file)->i_mode))
2732	return;
2733	sb_end_write(file_inode(file)->i_sb);
2734	}
2735
2736	/**
2737	* kiocb_start_write - get write access to a superblock for async file io
2738	* @iocb: the io context we want to submit the write with
2739	*
2740	* This is a variant of sb_start_write() for async io submission.
2741	* Should be matched with a call to kiocb_end_write().
2742	*/
2743	static inline void kiocb_start_write(struct kiocb *iocb)
2744	{
2745	struct inode *inode = file_inode(iocb->ki_filp);
2746
2747	sb_start_write(inode->i_sb);
2748	/*
2749	* Fool lockdep by telling it the lock got released so that it
2750	* doesn't complain about the held lock when we return to userspace.
2751	*/
2752	__sb_writers_release(inode->i_sb, SB_FREEZE_WRITE);
2753	}
2754
2755	/**
2756	* kiocb_end_write - drop write access to a superblock after async file io
2757	* @iocb: the io context we sumbitted the write with
2758	*
2759	* Should be matched with a call to kiocb_start_write().
2760	*/
2761	static inline void kiocb_end_write(struct kiocb *iocb)
2762	{
2763	struct inode *inode = file_inode(iocb->ki_filp);
2764
2765	/*
2766	* Tell lockdep we inherited freeze protection from submission thread.
2767	*/
2768	__sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
2769	sb_end_write(inode->i_sb);
2770	}
2771
2772	/*
2773	* This is used for regular files where some users -- especially the
2774	* currently executed binary in a process, previously handled via
2775	* VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2776	* read-write shared) accesses.
2777	*
2778	* get_write_access() gets write permission for a file.
2779	* put_write_access() releases this write permission.
2780	* deny_write_access() denies write access to a file.
2781	* allow_write_access() re-enables write access to a file.
2782	*
2783	* The i_writecount field of an inode can have the following values:
2784	* 0: no write access, no denied write access
2785	* < 0: (-i_writecount) users that denied write access to the file.
2786	* > 0: (i_writecount) users that have write access to the file.
2787	*
2788	* Normally we operate on that counter with atomic_{inc,dec} and it's safe
2789	* except for the cases where we don't hold i_writecount yet. Then we need to
2790	* use {get,deny}_write_access() - these functions check the sign and refuse
2791	* to do the change if sign is wrong.
2792	*/
2793	static inline int get_write_access(struct inode *inode)
2794	{
2795	return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2796	}
2797	static inline int deny_write_access(struct file *file)
2798	{
2799	struct inode *inode = file_inode(file);
2800	return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2801	}
2802	static inline void put_write_access(struct inode * inode)
2803	{
2804	atomic_dec(&inode->i_writecount);
2805	}
2806	static inline void allow_write_access(struct file *file)
2807	{
2808	if (file)
2809	atomic_inc(&file_inode(file)->i_writecount);
2810	}
2811	static inline bool inode_is_open_for_write(const struct inode *inode)
2812	{
2813	return atomic_read(&inode->i_writecount) > 0;
2814	}
2815
2816	#if defined(CONFIG_IMA) \|\| defined(CONFIG_FILE_LOCKING)
2817	static inline void i_readcount_dec(struct inode *inode)
2818	{
2819	BUG_ON(atomic_dec_return(&inode->i_readcount) < 0);
2820	}
2821	static inline void i_readcount_inc(struct inode *inode)
2822	{
2823	atomic_inc(&inode->i_readcount);
2824	}
2825	#else
2826	static inline void i_readcount_dec(struct inode *inode)
2827	{
2828	return;
2829	}
2830	static inline void i_readcount_inc(struct inode *inode)
2831	{
2832	return;
2833	}
2834	#endif
2835	extern int do_pipe_flags(int *, int);
2836
2837	extern ssize_t kernel_read(struct file , void , size_t, loff_t *);
2838	ssize_t __kernel_read(struct file file, void buf, size_t count, loff_t *pos);
2839	extern ssize_t kernel_write(struct file , const void , size_t, loff_t *);
2840	extern ssize_t __kernel_write(struct file , const void , size_t, loff_t *);
2841	extern struct file * open_exec(const char *);
2842
2843	/* fs/dcache.c -- generic fs support functions */
2844	extern bool is_subdir(struct dentry , struct dentry );
2845	extern bool path_is_under(const struct path , const struct path );
2846
2847	extern char file_path(struct file , char *, int);
2848
2849	#include <linux/err.h>
2850
2851	/* needed for stackable file system support */
2852	extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
2853
2854	extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
2855
2856	extern int inode_init_always(struct super_block , struct inode );
2857	extern void inode_init_once(struct inode *);
2858	extern void address_space_init_once(struct address_space *mapping);
2859	extern struct inode * igrab(struct inode *);
2860	extern ino_t iunique(struct super_block *, ino_t);
2861	extern int inode_needs_sync(struct inode *inode);
2862	extern int generic_delete_inode(struct inode *inode);
2863	static inline int generic_drop_inode(struct inode *inode)
2864	{
2865	return !inode->i_nlink \|\| inode_unhashed(inode);
2866	}
2867	extern void d_mark_dontcache(struct inode *inode);
2868
2869	extern struct inode ilookup5_nowait(struct super_block sb,
2870	unsigned long hashval, int (test)(struct inode , void *),
2871	void *data);
2872	extern struct inode ilookup5(struct super_block sb, unsigned long hashval,
2873	int (test)(struct inode , void ), void data);
2874	extern struct inode ilookup(struct super_block sb, unsigned long ino);
2875
2876	extern struct inode inode_insert5(struct inode inode, unsigned long hashval,
2877	int (test)(struct inode , void *),
2878	int (set)(struct inode , void *),
2879	void *data);
2880	extern struct inode * iget5_locked(struct super_block , unsigned long, int (test)(struct inode , void ), int (set)(struct inode , void ), void );
2881	extern struct inode * iget_locked(struct super_block *, unsigned long);
2882	extern struct inode find_inode_nowait(struct super_block ,
2883	unsigned long,
2884	int (match)(struct inode ,
2885	unsigned long, void *),
2886	void *data);
2887	extern struct inode find_inode_rcu(struct super_block , unsigned long,
2888	int ()(struct inode , void ), void );
2889	extern struct inode find_inode_by_ino_rcu(struct super_block , unsigned long);
2890	extern int insert_inode_locked4(struct inode , unsigned long, int (test)(struct inode , void ), void *);
2891	extern int insert_inode_locked(struct inode *);
2892	#ifdef CONFIG_DEBUG_LOCK_ALLOC
2893	extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
2894	#else
2895	static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
2896	#endif
2897	extern void unlock_new_inode(struct inode *);
2898	extern void discard_new_inode(struct inode *);
2899	extern unsigned int get_next_ino(void);
2900	extern void evict_inodes(struct super_block *sb);
2901	void dump_mapping(const struct address_space *);
2902
2903	/*
2904	* Userspace may rely on the inode number being non-zero. For example, glibc
2905	* simply ignores files with zero i_ino in unlink() and other places.
2906	*
2907	* As an additional complication, if userspace was compiled with
2908	* _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
2909	* lower 32 bits, so we need to check that those aren't zero explicitly. With
2910	* _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
2911	* better safe than sorry.
2912	*/
2913	static inline bool is_zero_ino(ino_t ino)
2914	{
2915	return (u32)ino == 0;
2916	}
2917
2918	extern void __iget(struct inode * inode);
2919	extern void iget_failed(struct inode *);
2920	extern void clear_inode(struct inode *);
2921	extern void __destroy_inode(struct inode *);
2922	extern struct inode new_inode_pseudo(struct super_block sb);
2923	extern struct inode new_inode(struct super_block sb);
2924	extern void free_inode_nonrcu(struct inode *inode);
2925	extern int setattr_should_drop_suidgid(struct mnt_idmap , struct inode );
2926	extern int file_remove_privs(struct file *);
2927	int setattr_should_drop_sgid(struct mnt_idmap *idmap,
2928	const struct inode *inode);
2929
2930	/*
2931	* This must be used for allocating filesystems specific inodes to set
2932	* up the inode reclaim context correctly.
2933	*/
2934	static inline void *
2935	alloc_inode_sb(struct super_block sb, struct kmem_cache cache, gfp_t gfp)
2936	{
2937	return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp);
2938	}
2939
2940	extern void __insert_inode_hash(struct inode *, unsigned long hashval);
2941	static inline void insert_inode_hash(struct inode *inode)
2942	{
2943	__insert_inode_hash(inode, inode->i_ino);
2944	}
2945
2946	extern void __remove_inode_hash(struct inode *);
2947	static inline void remove_inode_hash(struct inode *inode)
2948	{
2949	if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
2950	__remove_inode_hash(inode);
2951	}
2952
2953	extern void inode_sb_list_add(struct inode *inode);
2954	extern void inode_add_lru(struct inode *inode);
2955
2956	extern int sb_set_blocksize(struct super_block *, int);
2957	extern int sb_min_blocksize(struct super_block *, int);
2958
2959	extern int generic_file_mmap(struct file , struct vm_area_struct );
2960	extern int generic_file_readonly_mmap(struct file , struct vm_area_struct );
2961	extern ssize_t generic_write_checks(struct kiocb , struct iov_iter );
2962	int generic_write_checks_count(struct kiocb iocb, loff_t count);
2963	extern int generic_write_check_limits(struct file *file, loff_t pos,
2964	loff_t *count);
2965	extern int generic_file_rw_checks(struct file file_in, struct file file_out);
2966	ssize_t filemap_read(struct kiocb iocb, struct iov_iter to,
2967	ssize_t already_read);
2968	extern ssize_t generic_file_read_iter(struct kiocb , struct iov_iter );
2969	extern ssize_t __generic_file_write_iter(struct kiocb , struct iov_iter );
2970	extern ssize_t generic_file_write_iter(struct kiocb , struct iov_iter );
2971	extern ssize_t generic_file_direct_write(struct kiocb , struct iov_iter );
2972	ssize_t generic_perform_write(struct kiocb , struct iov_iter );
2973	ssize_t direct_write_fallback(struct kiocb iocb, struct iov_iter iter,
2974	ssize_t direct_written, ssize_t buffered_written);
2975
2976	ssize_t vfs_iter_read(struct file file, struct iov_iter iter, loff_t *ppos,
2977	rwf_t flags);
2978	ssize_t vfs_iter_write(struct file file, struct iov_iter iter, loff_t *ppos,
2979	rwf_t flags);
2980	ssize_t vfs_iocb_iter_read(struct file file, struct kiocb iocb,
2981	struct iov_iter *iter);
2982	ssize_t vfs_iocb_iter_write(struct file file, struct kiocb iocb,
2983	struct iov_iter *iter);
2984
2985	/* fs/splice.c */
2986	ssize_t filemap_splice_read(struct file in, loff_t ppos,
2987	struct pipe_inode_info *pipe,
2988	size_t len, unsigned int flags);
2989	ssize_t copy_splice_read(struct file in, loff_t ppos,
2990	struct pipe_inode_info *pipe,
2991	size_t len, unsigned int flags);
2992	extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
2993	struct file , loff_t , size_t, unsigned int);
2994	extern long do_splice_direct(struct file in, loff_t ppos, struct file *out,
2995	loff_t *opos, size_t len, unsigned int flags);
2996
2997
2998	extern void
2999	file_ra_state_init(struct file_ra_state ra, struct address_space mapping);
3000	extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3001	#define no_llseek NULL
3002	extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3003	extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3004	extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3005	int whence, loff_t maxsize, loff_t eof);
3006	extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3007	int whence, loff_t size);
3008	extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3009	extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3010	int rw_verify_area(int, struct file , const loff_t , size_t);
3011	extern int generic_file_open(struct inode * inode, struct file * filp);
3012	extern int nonseekable_open(struct inode * inode, struct file * filp);
3013	extern int stream_open(struct inode * inode, struct file * filp);
3014
3015	#ifdef CONFIG_BLOCK
3016	typedef void (dio_submit_t)(struct bio bio, struct inode inode,
3017	loff_t file_offset);
3018
3019	enum {
3020	/* need locking between buffered and direct access */
3021	DIO_LOCKING = 0x01,
3022
3023	/* filesystem does not support filling holes */
3024	DIO_SKIP_HOLES = 0x02,
3025	};
3026
3027	ssize_t __blockdev_direct_IO(struct kiocb iocb, struct inode inode,
3028	struct block_device bdev, struct iov_iter iter,
3029	get_block_t get_block,
3030	dio_iodone_t end_io,
3031	int flags);
3032
3033	static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3034	struct inode *inode,
3035	struct iov_iter *iter,
3036	get_block_t get_block)
3037	{
3038	return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
3039	get_block, NULL, DIO_LOCKING \| DIO_SKIP_HOLES);
3040	}
3041	#endif
3042
3043	void inode_dio_wait(struct inode *inode);
3044
3045	/**
3046	* inode_dio_begin - signal start of a direct I/O requests
3047	* @inode: inode the direct I/O happens on
3048	*
3049	* This is called once we've finished processing a direct I/O request,
3050	* and is used to wake up callers waiting for direct I/O to be quiesced.
3051	*/
3052	static inline void inode_dio_begin(struct inode *inode)
3053	{
3054	atomic_inc(&inode->i_dio_count);
3055	}
3056
3057	/**
3058	* inode_dio_end - signal finish of a direct I/O requests
3059	* @inode: inode the direct I/O happens on
3060	*
3061	* This is called once we've finished processing a direct I/O request,
3062	* and is used to wake up callers waiting for direct I/O to be quiesced.
3063	*/
3064	static inline void inode_dio_end(struct inode *inode)
3065	{
3066	if (atomic_dec_and_test(&inode->i_dio_count))
3067	wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
3068	}
3069
3070	extern void inode_set_flags(struct inode *inode, unsigned int flags,
3071	unsigned int mask);
3072
3073	extern const struct file_operations generic_ro_fops;
3074
3075	#define special_file(m) (S_ISCHR(m)\|\|S_ISBLK(m)\|\|S_ISFIFO(m)\|\|S_ISSOCK(m))
3076
3077	extern int readlink_copy(char __user , int, const char );
3078	extern int page_readlink(struct dentry , char __user , int);
3079	extern const char page_get_link(struct dentry , struct inode *,
3080	struct delayed_call *);
3081	extern void page_put_link(void *);
3082	extern int page_symlink(struct inode inode, const char symname, int len);
3083	extern const struct inode_operations page_symlink_inode_operations;
3084	extern void kfree_link(void *);
3085	void generic_fillattr(struct mnt_idmap , u32, struct inode , struct kstat *);
3086	void generic_fill_statx_attr(struct inode inode, struct kstat stat);
3087	extern int vfs_getattr_nosec(const struct path , struct kstat , u32, unsigned int);
3088	extern int vfs_getattr(const struct path , struct kstat , u32, unsigned int);
3089	void __inode_add_bytes(struct inode *inode, loff_t bytes);
3090	void inode_add_bytes(struct inode *inode, loff_t bytes);
3091	void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3092	void inode_sub_bytes(struct inode *inode, loff_t bytes);
3093	static inline loff_t __inode_get_bytes(struct inode *inode)
3094	{
3095	return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3096	}
3097	loff_t inode_get_bytes(struct inode *inode);
3098	void inode_set_bytes(struct inode *inode, loff_t bytes);
3099	const char simple_get_link(struct dentry , struct inode *,
3100	struct delayed_call *);
3101	extern const struct inode_operations simple_symlink_inode_operations;
3102
3103	extern int iterate_dir(struct file , struct dir_context );
3104
3105	int vfs_fstatat(int dfd, const char __user filename, struct kstat stat,
3106	int flags);
3107	int vfs_fstat(int fd, struct kstat *stat);
3108
3109	static inline int vfs_stat(const char __user filename, struct kstat stat)
3110	{
3111	return vfs_fstatat(AT_FDCWD, filename, stat, 0);
3112	}
3113	static inline int vfs_lstat(const char __user name, struct kstat stat)
3114	{
3115	return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
3116	}
3117
3118	extern const char vfs_get_link(struct dentry , struct delayed_call *);
3119	extern int vfs_readlink(struct dentry , char __user , int);
3120
3121	extern struct file_system_type get_filesystem(struct file_system_type fs);
3122	extern void put_filesystem(struct file_system_type *fs);
3123	extern struct file_system_type get_fs_type(const char name);
3124	extern struct super_block get_active_super(struct block_device bdev);
3125	extern void drop_super(struct super_block *sb);
3126	extern void drop_super_exclusive(struct super_block *sb);
3127	extern void iterate_supers(void ()(struct super_block , void ), void );
3128	extern void iterate_supers_type(struct file_system_type *,
3129	void ()(struct super_block , void ), void );
3130
3131	extern int dcache_dir_open(struct inode , struct file );
3132	extern int dcache_dir_close(struct inode , struct file );
3133	extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3134	extern int dcache_readdir(struct file , struct dir_context );
3135	extern int simple_setattr(struct mnt_idmap , struct dentry ,
3136	struct iattr *);
3137	extern int simple_getattr(struct mnt_idmap , const struct path ,
3138	struct kstat *, u32, unsigned int);
3139	extern int simple_statfs(struct dentry , struct kstatfs );
3140	extern int simple_open(struct inode inode, struct file file);
3141	extern int simple_link(struct dentry , struct inode , struct dentry *);
3142	extern int simple_unlink(struct inode , struct dentry );
3143	extern int simple_rmdir(struct inode , struct dentry );
3144	void simple_rename_timestamp(struct inode old_dir, struct dentry old_dentry,
3145	struct inode new_dir, struct dentry new_dentry);
3146	extern int simple_rename_exchange(struct inode old_dir, struct dentry old_dentry,
3147	struct inode new_dir, struct dentry new_dentry);
3148	extern int simple_rename(struct mnt_idmap , struct inode ,
3149	struct dentry , struct inode , struct dentry *,
3150	unsigned int);
3151	extern void simple_recursive_removal(struct dentry *,
3152	void (callback)(struct dentry ));
3153	extern int noop_fsync(struct file *, loff_t, loff_t, int);
3154	extern ssize_t noop_direct_IO(struct kiocb iocb, struct iov_iter iter);
3155	extern int simple_empty(struct dentry *);
3156	extern int simple_write_begin(struct file file, struct address_space mapping,
3157	loff_t pos, unsigned len,
3158	struct page pagep, void fsdata);
3159	extern const struct address_space_operations ram_aops;
3160	extern int always_delete_dentry(const struct dentry *);
3161	extern struct inode alloc_anon_inode(struct super_block );
3162	extern int simple_nosetlease(struct file , int, struct file_lock , void *);
3163	extern const struct dentry_operations simple_dentry_operations;
3164
3165	extern struct dentry simple_lookup(struct inode , struct dentry *, unsigned int flags);
3166	extern ssize_t generic_read_dir(struct file , char __user , size_t, loff_t *);
3167	extern const struct file_operations simple_dir_operations;
3168	extern const struct inode_operations simple_dir_inode_operations;
3169	extern void make_empty_dir_inode(struct inode *inode);
3170	extern bool is_empty_dir_inode(struct inode *inode);
3171	struct tree_descr { const char name; const struct file_operations ops; int mode; };
3172	struct dentry d_alloc_name(struct dentry , const char *);
3173	extern int simple_fill_super(struct super_block *, unsigned long,
3174	const struct tree_descr *);
3175	extern int simple_pin_fs(struct file_system_type , struct vfsmount mount, int count);
3176	extern void simple_release_fs(struct vfsmount *mount, int count);
3177
3178	extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3179	loff_t ppos, const void from, size_t available);
3180	extern ssize_t simple_write_to_buffer(void to, size_t available, loff_t ppos,
3181	const void __user *from, size_t count);
3182
3183	struct offset_ctx {
3184	struct xarray xa;
3185	u32 next_offset;
3186	};
3187
3188	void simple_offset_init(struct offset_ctx *octx);
3189	int simple_offset_add(struct offset_ctx octx, struct dentry dentry);
3190	void simple_offset_remove(struct offset_ctx octx, struct dentry dentry);
3191	int simple_offset_rename_exchange(struct inode *old_dir,
3192	struct dentry *old_dentry,
3193	struct inode *new_dir,
3194	struct dentry *new_dentry);
3195	void simple_offset_destroy(struct offset_ctx *octx);
3196
3197	extern const struct file_operations simple_offset_dir_operations;
3198
3199	extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3200	extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3201
3202	extern int generic_check_addressable(unsigned, u64);
3203
3204	extern void generic_set_encrypted_ci_d_ops(struct dentry *dentry);
3205
3206	int may_setattr(struct mnt_idmap idmap, struct inode inode,
3207	unsigned int ia_valid);
3208	int setattr_prepare(struct mnt_idmap , struct dentry , struct iattr *);
3209	extern int inode_newsize_ok(const struct inode *, loff_t offset);
3210	void setattr_copy(struct mnt_idmap , struct inode inode,
3211	const struct iattr *attr);
3212
3213	extern int file_update_time(struct file *file);
3214
3215	static inline bool vma_is_dax(const struct vm_area_struct *vma)
3216	{
3217	return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3218	}
3219
3220	static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3221	{
3222	struct inode *inode;
3223
3224	if (!IS_ENABLED(CONFIG_FS_DAX) \|\| !vma->vm_file)
3225	return false;
3226	if (!vma_is_dax(vma))
3227	return false;
3228	inode = file_inode(vma->vm_file);
3229	if (S_ISCHR(inode->i_mode))
3230	return false; /* device-dax */
3231	return true;
3232	}
3233
3234	static inline int iocb_flags(struct file *file)
3235	{
3236	int res = 0;
3237	if (file->f_flags & O_APPEND)
3238	res \|= IOCB_APPEND;
3239	if (file->f_flags & O_DIRECT)
3240	res \|= IOCB_DIRECT;
3241	if (file->f_flags & O_DSYNC)
3242	res \|= IOCB_DSYNC;
3243	if (file->f_flags & __O_SYNC)
3244	res \|= IOCB_SYNC;
3245	return res;
3246	}
3247
3248	static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
3249	{
3250	int kiocb_flags = 0;
3251
3252	/* make sure there's no overlap between RWF and private IOCB flags */
3253	BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3254
3255	if (!flags)
3256	return 0;
3257	if (unlikely(flags & ~RWF_SUPPORTED))
3258	return -EOPNOTSUPP;
3259
3260	if (flags & RWF_NOWAIT) {
3261	if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3262	return -EOPNOTSUPP;
3263	kiocb_flags \|= IOCB_NOIO;
3264	}
3265	kiocb_flags \|= (__force int) (flags & RWF_SUPPORTED);
3266	if (flags & RWF_SYNC)
3267	kiocb_flags \|= IOCB_DSYNC;
3268
3269	ki->ki_flags \|= kiocb_flags;
3270	return 0;
3271	}
3272
3273	static inline ino_t parent_ino(struct dentry *dentry)
3274	{
3275	ino_t res;
3276
3277	/*
3278	* Don't strictly need d_lock here? If the parent ino could change
3279	* then surely we'd have a deeper race in the caller?
3280	*/
3281	spin_lock(&dentry->d_lock);
3282	res = dentry->d_parent->d_inode->i_ino;
3283	spin_unlock(&dentry->d_lock);
3284	return res;
3285	}
3286
3287	/* Transaction based IO helpers */
3288
3289	/*
3290	* An argresp is stored in an allocated page and holds the
3291	* size of the argument or response, along with its content
3292	*/
3293	struct simple_transaction_argresp {
3294	ssize_t size;
3295	char data[];
3296	};
3297
3298	#define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3299
3300	char simple_transaction_get(struct file file, const char __user *buf,
3301	size_t size);
3302	ssize_t simple_transaction_read(struct file file, char __user buf,
3303	size_t size, loff_t *pos);
3304	int simple_transaction_release(struct inode inode, struct file file);
3305
3306	void simple_transaction_set(struct file *file, size_t n);
3307
3308	/*
3309	* simple attribute files
3310	*
3311	* These attributes behave similar to those in sysfs:
3312	*
3313	* Writing to an attribute immediately sets a value, an open file can be
3314	* written to multiple times.
3315	*
3316	* Reading from an attribute creates a buffer from the value that might get
3317	* read with multiple read calls. When the attribute has been read
3318	* completely, no further read calls are possible until the file is opened
3319	* again.
3320	*
3321	* All attributes contain a text representation of a numeric value
3322	* that are accessed with the get() and set() functions.
3323	*/
3324	#define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \
3325	static int __fops ## _open(struct inode inode, struct file file) \
3326	{ \
3327	__simple_attr_check_format(__fmt, 0ull); \
3328	return simple_attr_open(inode, file, __get, __set, __fmt); \
3329	} \
3330	static const struct file_operations __fops = { \
3331	.owner = THIS_MODULE, \
3332	.open = __fops ## _open, \
3333	.release = simple_attr_release, \
3334	.read = simple_attr_read, \
3335	.write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \
3336	.llseek = generic_file_llseek, \
3337	}
3338
3339	#define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
3340	DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false)
3341
3342	#define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \
3343	DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true)
3344
3345	static inline __printf(1, 2)
3346	void __simple_attr_check_format(const char *fmt, ...)
3347	{
3348	/* don't do anything, just let the compiler check the arguments; */
3349	}
3350
3351	int simple_attr_open(struct inode inode, struct file file,
3352	int (get)(void , u64 ), int (set)(void *, u64),
3353	const char *fmt);
3354	int simple_attr_release(struct inode inode, struct file file);
3355	ssize_t simple_attr_read(struct file file, char __user buf,
3356	size_t len, loff_t *ppos);
3357	ssize_t simple_attr_write(struct file file, const char __user buf,
3358	size_t len, loff_t *ppos);
3359	ssize_t simple_attr_write_signed(struct file file, const char __user buf,
3360	size_t len, loff_t *ppos);
3361
3362	struct ctl_table;
3363	int __init list_bdev_fs_names(char *buf, size_t size);
3364
3365	#define __FMODE_EXEC ((__force int) FMODE_EXEC)
3366	#define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY)
3367
3368	#define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3369	#define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) \| \
3370	(flag & __FMODE_NONOTIFY)))
3371
3372	static inline bool is_sxid(umode_t mode)
3373	{
3374	return mode & (S_ISUID \| S_ISGID);
3375	}
3376
3377	static inline int check_sticky(struct mnt_idmap *idmap,
3378	struct inode dir, struct inode inode)
3379	{
3380	if (!(dir->i_mode & S_ISVTX))
3381	return 0;
3382
3383	return __check_sticky(idmap, dir, inode);
3384	}
3385
3386	static inline void inode_has_no_xattr(struct inode *inode)
3387	{
3388	if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3389	inode->i_flags \|= S_NOSEC;
3390	}
3391
3392	static inline bool is_root_inode(struct inode *inode)
3393	{
3394	return inode == inode->i_sb->s_root->d_inode;
3395	}
3396
3397	static inline bool dir_emit(struct dir_context *ctx,
3398	const char *name, int namelen,
3399	u64 ino, unsigned type)
3400	{
3401	return ctx->actor(ctx, name, namelen, ctx->pos, ino, type);
3402	}
3403	static inline bool dir_emit_dot(struct file file, struct dir_context ctx)
3404	{
3405	return ctx->actor(ctx, ".", 1, ctx->pos,
3406	file->f_path.dentry->d_inode->i_ino, DT_DIR);
3407	}
3408	static inline bool dir_emit_dotdot(struct file file, struct dir_context ctx)
3409	{
3410	return ctx->actor(ctx, "..", 2, ctx->pos,
3411	parent_ino(file->f_path.dentry), DT_DIR);
3412	}
3413	static inline bool dir_emit_dots(struct file file, struct dir_context ctx)
3414	{
3415	if (ctx->pos == 0) {
3416	if (!dir_emit_dot(file, ctx))
3417	return false;
3418	ctx->pos = 1;
3419	}
3420	if (ctx->pos == 1) {
3421	if (!dir_emit_dotdot(file, ctx))
3422	return false;
3423	ctx->pos = 2;
3424	}
3425	return true;
3426	}
3427	static inline bool dir_relax(struct inode *inode)
3428	{
3429	inode_unlock(inode);
3430	inode_lock(inode);
3431	return !IS_DEADDIR(inode);
3432	}
3433
3434	static inline bool dir_relax_shared(struct inode *inode)
3435	{
3436	inode_unlock_shared(inode);
3437	inode_lock_shared(inode);
3438	return !IS_DEADDIR(inode);
3439	}
3440
3441	extern bool path_noexec(const struct path *path);
3442	extern void inode_nohighmem(struct inode *inode);
3443
3444	/* mm/fadvise.c */
3445	extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3446	int advice);
3447	extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3448	int advice);
3449
3450	#endif /* _LINUX_FS_H */