1/* SPDX-License-Identifier: GPL-2.0 */
2#ifndef _LINUX_FS_H
3#define _LINUX_FS_H
4
5#include <linux/vfsdebug.h>
6#include <linux/linkage.h>
7#include <linux/wait_bit.h>
8#include <linux/kdev_t.h>
9#include <linux/dcache.h>
10#include <linux/path.h>
11#include <linux/stat.h>
12#include <linux/cache.h>
13#include <linux/list.h>
14#include <linux/list_lru.h>
15#include <linux/llist.h>
16#include <linux/radix-tree.h>
17#include <linux/xarray.h>
18#include <linux/rbtree.h>
19#include <linux/init.h>
20#include <linux/pid.h>
21#include <linux/bug.h>
22#include <linux/mutex.h>
23#include <linux/rwsem.h>
24#include <linux/mm_types.h>
25#include <linux/capability.h>
26#include <linux/semaphore.h>
27#include <linux/fcntl.h>
28#include <linux/rculist_bl.h>
29#include <linux/atomic.h>
30#include <linux/shrinker.h>
31#include <linux/migrate_mode.h>
32#include <linux/uidgid.h>
33#include <linux/lockdep.h>
34#include <linux/percpu-rwsem.h>
35#include <linux/workqueue.h>
36#include <linux/delayed_call.h>
37#include <linux/uuid.h>
38#include <linux/errseq.h>
39#include <linux/ioprio.h>
40#include <linux/fs_types.h>
41#include <linux/build_bug.h>
42#include <linux/stddef.h>
43#include <linux/mount.h>
44#include <linux/cred.h>
45#include <linux/mnt_idmapping.h>
46#include <linux/slab.h>
47#include <linux/maple_tree.h>
48#include <linux/rw_hint.h>
49#include <linux/file_ref.h>
50#include <linux/unicode.h>
51
52#include <asm/byteorder.h>
53#include <uapi/linux/fs.h>
54
55struct backing_dev_info;
56struct bdi_writeback;
57struct bio;
58struct io_comp_batch;
59struct export_operations;
60struct fiemap_extent_info;
61struct hd_geometry;
62struct iovec;
63struct kiocb;
64struct kobject;
65struct pipe_inode_info;
66struct poll_table_struct;
67struct kstatfs;
68struct vm_area_struct;
69struct vfsmount;
70struct cred;
71struct swap_info_struct;
72struct seq_file;
73struct workqueue_struct;
74struct iov_iter;
75struct fscrypt_operations;
76struct fsverity_operations;
77struct fsnotify_mark_connector;
78struct fsnotify_sb_info;
79struct fs_context;
80struct fs_parameter_spec;
81struct file_kattr;
82struct iomap_ops;
83
84extern void __init inode_init(void);
85extern void __init inode_init_early(void);
86extern void __init files_init(void);
87extern void __init files_maxfiles_init(void);
88
89extern unsigned long get_max_files(void);
90extern unsigned int sysctl_nr_open;
91
92typedef __kernel_rwf_t rwf_t;
93
94struct buffer_head;
95typedef int (get_block_t)(struct inode *inode, sector_t iblock,
96 struct buffer_head *bh_result, int create);
97typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
98 ssize_t bytes, void *private);
99
100#define MAY_EXEC 0x00000001
101#define MAY_WRITE 0x00000002
102#define MAY_READ 0x00000004
103#define MAY_APPEND 0x00000008
104#define MAY_ACCESS 0x00000010
105#define MAY_OPEN 0x00000020
106#define MAY_CHDIR 0x00000040
107/* called from RCU mode, don't block */
108#define MAY_NOT_BLOCK 0x00000080
109
110/*
111 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond
112 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
113 */
114
115/* file is open for reading */
116#define FMODE_READ ((__force fmode_t)(1 << 0))
117/* file is open for writing */
118#define FMODE_WRITE ((__force fmode_t)(1 << 1))
119/* file is seekable */
120#define FMODE_LSEEK ((__force fmode_t)(1 << 2))
121/* file can be accessed using pread */
122#define FMODE_PREAD ((__force fmode_t)(1 << 3))
123/* file can be accessed using pwrite */
124#define FMODE_PWRITE ((__force fmode_t)(1 << 4))
125/* File is opened for execution with sys_execve / sys_uselib */
126#define FMODE_EXEC ((__force fmode_t)(1 << 5))
127/* File writes are restricted (block device specific) */
128#define FMODE_WRITE_RESTRICTED ((__force fmode_t)(1 << 6))
129/* File supports atomic writes */
130#define FMODE_CAN_ATOMIC_WRITE ((__force fmode_t)(1 << 7))
131
132/* FMODE_* bit 8 */
133
134/* 32bit hashes as llseek() offset (for directories) */
135#define FMODE_32BITHASH ((__force fmode_t)(1 << 9))
136/* 64bit hashes as llseek() offset (for directories) */
137#define FMODE_64BITHASH ((__force fmode_t)(1 << 10))
138
139/*
140 * Don't update ctime and mtime.
141 *
142 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
143 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
144 */
145#define FMODE_NOCMTIME ((__force fmode_t)(1 << 11))
146
147/* Expect random access pattern */
148#define FMODE_RANDOM ((__force fmode_t)(1 << 12))
149
150/* Supports IOCB_HAS_METADATA */
151#define FMODE_HAS_METADATA ((__force fmode_t)(1 << 13))
152
153/* File is opened with O_PATH; almost nothing can be done with it */
154#define FMODE_PATH ((__force fmode_t)(1 << 14))
155
156/* File needs atomic accesses to f_pos */
157#define FMODE_ATOMIC_POS ((__force fmode_t)(1 << 15))
158/* Write access to underlying fs */
159#define FMODE_WRITER ((__force fmode_t)(1 << 16))
160/* Has read method(s) */
161#define FMODE_CAN_READ ((__force fmode_t)(1 << 17))
162/* Has write method(s) */
163#define FMODE_CAN_WRITE ((__force fmode_t)(1 << 18))
164
165#define FMODE_OPENED ((__force fmode_t)(1 << 19))
166#define FMODE_CREATED ((__force fmode_t)(1 << 20))
167
168/* File is stream-like */
169#define FMODE_STREAM ((__force fmode_t)(1 << 21))
170
171/* File supports DIRECT IO */
172#define FMODE_CAN_ODIRECT ((__force fmode_t)(1 << 22))
173
174#define FMODE_NOREUSE ((__force fmode_t)(1 << 23))
175
176/* File is embedded in backing_file object */
177#define FMODE_BACKING ((__force fmode_t)(1 << 24))
178
179/*
180 * Together with FMODE_NONOTIFY_PERM defines which fsnotify events shouldn't be
181 * generated (see below)
182 */
183#define FMODE_NONOTIFY ((__force fmode_t)(1 << 25))
184
185/*
186 * Together with FMODE_NONOTIFY defines which fsnotify events shouldn't be
187 * generated (see below)
188 */
189#define FMODE_NONOTIFY_PERM ((__force fmode_t)(1 << 26))
190
191/* File is capable of returning -EAGAIN if I/O will block */
192#define FMODE_NOWAIT ((__force fmode_t)(1 << 27))
193
194/* File represents mount that needs unmounting */
195#define FMODE_NEED_UNMOUNT ((__force fmode_t)(1 << 28))
196
197/* File does not contribute to nr_files count */
198#define FMODE_NOACCOUNT ((__force fmode_t)(1 << 29))
199
200/*
201 * The two FMODE_NONOTIFY* define which fsnotify events should not be generated
202 * for an open file. These are the possible values of
203 * (f->f_mode & FMODE_FSNOTIFY_MASK) and their meaning:
204 *
205 * FMODE_NONOTIFY - suppress all (incl. non-permission) events.
206 * FMODE_NONOTIFY_PERM - suppress permission (incl. pre-content) events.
207 * FMODE_NONOTIFY | FMODE_NONOTIFY_PERM - suppress only FAN_ACCESS_PERM.
208 */
209#define FMODE_FSNOTIFY_MASK \
210 (FMODE_NONOTIFY | FMODE_NONOTIFY_PERM)
211
212#define FMODE_FSNOTIFY_NONE(mode) \
213 ((mode & FMODE_FSNOTIFY_MASK) == FMODE_NONOTIFY)
214#ifdef CONFIG_FANOTIFY_ACCESS_PERMISSIONS
215#define FMODE_FSNOTIFY_HSM(mode) \
216 ((mode & FMODE_FSNOTIFY_MASK) == 0 || \
217 (mode & FMODE_FSNOTIFY_MASK) == (FMODE_NONOTIFY | FMODE_NONOTIFY_PERM))
218#define FMODE_FSNOTIFY_ACCESS_PERM(mode) \
219 ((mode & FMODE_FSNOTIFY_MASK) == 0)
220#else
221#define FMODE_FSNOTIFY_ACCESS_PERM(mode) 0
222#define FMODE_FSNOTIFY_HSM(mode) 0
223#endif
224
225/*
226 * Attribute flags. These should be or-ed together to figure out what
227 * has been changed!
228 */
229#define ATTR_MODE (1 << 0)
230#define ATTR_UID (1 << 1)
231#define ATTR_GID (1 << 2)
232#define ATTR_SIZE (1 << 3)
233#define ATTR_ATIME (1 << 4)
234#define ATTR_MTIME (1 << 5)
235#define ATTR_CTIME (1 << 6)
236#define ATTR_ATIME_SET (1 << 7)
237#define ATTR_MTIME_SET (1 << 8)
238#define ATTR_FORCE (1 << 9) /* Not a change, but a change it */
239#define ATTR_CTIME_SET (1 << 10)
240#define ATTR_KILL_SUID (1 << 11)
241#define ATTR_KILL_SGID (1 << 12)
242#define ATTR_FILE (1 << 13)
243#define ATTR_KILL_PRIV (1 << 14)
244#define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */
245#define ATTR_TIMES_SET (1 << 16)
246#define ATTR_TOUCH (1 << 17)
247#define ATTR_DELEG (1 << 18) /* Delegated attrs. Don't break write delegations */
248
249/*
250 * Whiteout is represented by a char device. The following constants define the
251 * mode and device number to use.
252 */
253#define WHITEOUT_MODE 0
254#define WHITEOUT_DEV 0
255
256/*
257 * This is the Inode Attributes structure, used for notify_change(). It
258 * uses the above definitions as flags, to know which values have changed.
259 * Also, in this manner, a Filesystem can look at only the values it cares
260 * about. Basically, these are the attributes that the VFS layer can
261 * request to change from the FS layer.
262 *
263 * Derek Atkins <warlord@MIT.EDU> 94-10-20
264 */
265struct iattr {
266 unsigned int ia_valid;
267 umode_t ia_mode;
268 /*
269 * The two anonymous unions wrap structures with the same member.
270 *
271 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which
272 * are a dedicated type requiring the filesystem to use the dedicated
273 * helpers. Other filesystem can continue to use ia_{g,u}id until they
274 * have been ported.
275 *
276 * They always contain the same value. In other words FS_ALLOW_IDMAP
277 * pass down the same value on idmapped mounts as they would on regular
278 * mounts.
279 */
280 union {
281 kuid_t ia_uid;
282 vfsuid_t ia_vfsuid;
283 };
284 union {
285 kgid_t ia_gid;
286 vfsgid_t ia_vfsgid;
287 };
288 loff_t ia_size;
289 struct timespec64 ia_atime;
290 struct timespec64 ia_mtime;
291 struct timespec64 ia_ctime;
292
293 /*
294 * Not an attribute, but an auxiliary info for filesystems wanting to
295 * implement an ftruncate() like method. NOTE: filesystem should
296 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
297 */
298 struct file *ia_file;
299};
300
301/*
302 * Includes for diskquotas.
303 */
304#include <linux/quota.h>
305
306/*
307 * Maximum number of layers of fs stack. Needs to be limited to
308 * prevent kernel stack overflow
309 */
310#define FILESYSTEM_MAX_STACK_DEPTH 2
311
312/**
313 * enum positive_aop_returns - aop return codes with specific semantics
314 *
315 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
316 * completed, that the page is still locked, and
317 * should be considered active. The VM uses this hint
318 * to return the page to the active list -- it won't
319 * be a candidate for writeback again in the near
320 * future. Other callers must be careful to unlock
321 * the page if they get this return. Returned by
322 * writepage();
323 *
324 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
325 * unlocked it and the page might have been truncated.
326 * The caller should back up to acquiring a new page and
327 * trying again. The aop will be taking reasonable
328 * precautions not to livelock. If the caller held a page
329 * reference, it should drop it before retrying. Returned
330 * by read_folio().
331 *
332 * address_space_operation functions return these large constants to indicate
333 * special semantics to the caller. These are much larger than the bytes in a
334 * page to allow for functions that return the number of bytes operated on in a
335 * given page.
336 */
337
338enum positive_aop_returns {
339 AOP_WRITEPAGE_ACTIVATE = 0x80000,
340 AOP_TRUNCATED_PAGE = 0x80001,
341};
342
343/*
344 * oh the beauties of C type declarations.
345 */
346struct page;
347struct address_space;
348struct writeback_control;
349struct readahead_control;
350
351/* Match RWF_* bits to IOCB bits */
352#define IOCB_HIPRI (__force int) RWF_HIPRI
353#define IOCB_DSYNC (__force int) RWF_DSYNC
354#define IOCB_SYNC (__force int) RWF_SYNC
355#define IOCB_NOWAIT (__force int) RWF_NOWAIT
356#define IOCB_APPEND (__force int) RWF_APPEND
357#define IOCB_ATOMIC (__force int) RWF_ATOMIC
358#define IOCB_DONTCACHE (__force int) RWF_DONTCACHE
359#define IOCB_NOSIGNAL (__force int) RWF_NOSIGNAL
360
361/* non-RWF related bits - start at 16 */
362#define IOCB_EVENTFD (1 << 16)
363#define IOCB_DIRECT (1 << 17)
364#define IOCB_WRITE (1 << 18)
365/* iocb->ki_waitq is valid */
366#define IOCB_WAITQ (1 << 19)
367#define IOCB_NOIO (1 << 20)
368/* can use bio alloc cache */
369#define IOCB_ALLOC_CACHE (1 << 21)
370/*
371 * IOCB_DIO_CALLER_COMP can be set by the iocb owner, to indicate that the
372 * iocb completion can be passed back to the owner for execution from a safe
373 * context rather than needing to be punted through a workqueue. If this
374 * flag is set, the bio completion handling may set iocb->dio_complete to a
375 * handler function and iocb->private to context information for that handler.
376 * The issuer should call the handler with that context information from task
377 * context to complete the processing of the iocb. Note that while this
378 * provides a task context for the dio_complete() callback, it should only be
379 * used on the completion side for non-IO generating completions. It's fine to
380 * call blocking functions from this callback, but they should not wait for
381 * unrelated IO (like cache flushing, new IO generation, etc).
382 */
383#define IOCB_DIO_CALLER_COMP (1 << 22)
384/* kiocb is a read or write operation submitted by fs/aio.c. */
385#define IOCB_AIO_RW (1 << 23)
386#define IOCB_HAS_METADATA (1 << 24)
387
388/* for use in trace events */
389#define TRACE_IOCB_STRINGS \
390 { IOCB_HIPRI, "HIPRI" }, \
391 { IOCB_DSYNC, "DSYNC" }, \
392 { IOCB_SYNC, "SYNC" }, \
393 { IOCB_NOWAIT, "NOWAIT" }, \
394 { IOCB_APPEND, "APPEND" }, \
395 { IOCB_ATOMIC, "ATOMIC" }, \
396 { IOCB_DONTCACHE, "DONTCACHE" }, \
397 { IOCB_EVENTFD, "EVENTFD"}, \
398 { IOCB_DIRECT, "DIRECT" }, \
399 { IOCB_WRITE, "WRITE" }, \
400 { IOCB_WAITQ, "WAITQ" }, \
401 { IOCB_NOIO, "NOIO" }, \
402 { IOCB_ALLOC_CACHE, "ALLOC_CACHE" }, \
403 { IOCB_DIO_CALLER_COMP, "CALLER_COMP" }, \
404 { IOCB_AIO_RW, "AIO_RW" }, \
405 { IOCB_HAS_METADATA, "AIO_HAS_METADATA" }
406
407struct kiocb {
408 struct file *ki_filp;
409 loff_t ki_pos;
410 void (*ki_complete)(struct kiocb *iocb, long ret);
411 void *private;
412 int ki_flags;
413 u16 ki_ioprio; /* See linux/ioprio.h */
414 u8 ki_write_stream;
415 union {
416 /*
417 * Only used for async buffered reads, where it denotes the
418 * page waitqueue associated with completing the read. Valid
419 * IFF IOCB_WAITQ is set.
420 */
421 struct wait_page_queue *ki_waitq;
422 /*
423 * Can be used for O_DIRECT IO, where the completion handling
424 * is punted back to the issuer of the IO. May only be set
425 * if IOCB_DIO_CALLER_COMP is set by the issuer, and the issuer
426 * must then check for presence of this handler when ki_complete
427 * is invoked. The data passed in to this handler must be
428 * assigned to ->private when dio_complete is assigned.
429 */
430 ssize_t (*dio_complete)(void *data);
431 };
432};
433
434static inline bool is_sync_kiocb(struct kiocb *kiocb)
435{
436 return kiocb->ki_complete == NULL;
437}
438
439struct address_space_operations {
440 int (*read_folio)(struct file *, struct folio *);
441
442 /* Write back some dirty pages from this mapping. */
443 int (*writepages)(struct address_space *, struct writeback_control *);
444
445 /* Mark a folio dirty. Return true if this dirtied it */
446 bool (*dirty_folio)(struct address_space *, struct folio *);
447
448 void (*readahead)(struct readahead_control *);
449
450 int (*write_begin)(const struct kiocb *, struct address_space *mapping,
451 loff_t pos, unsigned len,
452 struct folio **foliop, void **fsdata);
453 int (*write_end)(const struct kiocb *, struct address_space *mapping,
454 loff_t pos, unsigned len, unsigned copied,
455 struct folio *folio, void *fsdata);
456
457 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
458 sector_t (*bmap)(struct address_space *, sector_t);
459 void (*invalidate_folio) (struct folio *, size_t offset, size_t len);
460 bool (*release_folio)(struct folio *, gfp_t);
461 void (*free_folio)(struct folio *folio);
462 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
463 /*
464 * migrate the contents of a folio to the specified target. If
465 * migrate_mode is MIGRATE_ASYNC, it must not block.
466 */
467 int (*migrate_folio)(struct address_space *, struct folio *dst,
468 struct folio *src, enum migrate_mode);
469 int (*launder_folio)(struct folio *);
470 bool (*is_partially_uptodate) (struct folio *, size_t from,
471 size_t count);
472 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb);
473 int (*error_remove_folio)(struct address_space *, struct folio *);
474
475 /* swapfile support */
476 int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
477 sector_t *span);
478 void (*swap_deactivate)(struct file *file);
479 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
480};
481
482extern const struct address_space_operations empty_aops;
483
484/**
485 * struct address_space - Contents of a cacheable, mappable object.
486 * @host: Owner, either the inode or the block_device.
487 * @i_pages: Cached pages.
488 * @invalidate_lock: Guards coherency between page cache contents and
489 * file offset->disk block mappings in the filesystem during invalidates.
490 * It is also used to block modification of page cache contents through
491 * memory mappings.
492 * @gfp_mask: Memory allocation flags to use for allocating pages.
493 * @i_mmap_writable: Number of VM_SHARED, VM_MAYWRITE mappings.
494 * @nr_thps: Number of THPs in the pagecache (non-shmem only).
495 * @i_mmap: Tree of private and shared mappings.
496 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
497 * @nrpages: Number of page entries, protected by the i_pages lock.
498 * @writeback_index: Writeback starts here.
499 * @a_ops: Methods.
500 * @flags: Error bits and flags (AS_*).
501 * @wb_err: The most recent error which has occurred.
502 * @i_private_lock: For use by the owner of the address_space.
503 * @i_private_list: For use by the owner of the address_space.
504 * @i_private_data: For use by the owner of the address_space.
505 */
506struct address_space {
507 struct inode *host;
508 struct xarray i_pages;
509 struct rw_semaphore invalidate_lock;
510 gfp_t gfp_mask;
511 atomic_t i_mmap_writable;
512#ifdef CONFIG_READ_ONLY_THP_FOR_FS
513 /* number of thp, only for non-shmem files */
514 atomic_t nr_thps;
515#endif
516 struct rb_root_cached i_mmap;
517 unsigned long nrpages;
518 pgoff_t writeback_index;
519 const struct address_space_operations *a_ops;
520 unsigned long flags;
521 errseq_t wb_err;
522 spinlock_t i_private_lock;
523 struct list_head i_private_list;
524 struct rw_semaphore i_mmap_rwsem;
525 void * i_private_data;
526} __attribute__((aligned(sizeof(long)))) __randomize_layout;
527 /*
528 * On most architectures that alignment is already the case; but
529 * must be enforced here for CRIS, to let the least significant bit
530 * of struct folio's "mapping" pointer be used for FOLIO_MAPPING_ANON.
531 */
532
533/* XArray tags, for tagging dirty and writeback pages in the pagecache. */
534#define PAGECACHE_TAG_DIRTY XA_MARK_0
535#define PAGECACHE_TAG_WRITEBACK XA_MARK_1
536#define PAGECACHE_TAG_TOWRITE XA_MARK_2
537
538/*
539 * Returns true if any of the pages in the mapping are marked with the tag.
540 */
541static inline bool mapping_tagged(const struct address_space *mapping, xa_mark_t tag)
542{
543 return xa_marked(xa: &mapping->i_pages, mark: tag);
544}
545
546static inline void i_mmap_lock_write(struct address_space *mapping)
547{
548 down_write(sem: &mapping->i_mmap_rwsem);
549}
550
551static inline int i_mmap_trylock_write(struct address_space *mapping)
552{
553 return down_write_trylock(sem: &mapping->i_mmap_rwsem);
554}
555
556static inline void i_mmap_unlock_write(struct address_space *mapping)
557{
558 up_write(sem: &mapping->i_mmap_rwsem);
559}
560
561static inline int i_mmap_trylock_read(struct address_space *mapping)
562{
563 return down_read_trylock(sem: &mapping->i_mmap_rwsem);
564}
565
566static inline void i_mmap_lock_read(struct address_space *mapping)
567{
568 down_read(sem: &mapping->i_mmap_rwsem);
569}
570
571static inline void i_mmap_unlock_read(struct address_space *mapping)
572{
573 up_read(sem: &mapping->i_mmap_rwsem);
574}
575
576static inline void i_mmap_assert_locked(struct address_space *mapping)
577{
578 lockdep_assert_held(&mapping->i_mmap_rwsem);
579}
580
581static inline void i_mmap_assert_write_locked(struct address_space *mapping)
582{
583 lockdep_assert_held_write(&mapping->i_mmap_rwsem);
584}
585
586/*
587 * Might pages of this file be mapped into userspace?
588 */
589static inline int mapping_mapped(const struct address_space *mapping)
590{
591 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
592}
593
594/*
595 * Might pages of this file have been modified in userspace?
596 * Note that i_mmap_writable counts all VM_SHARED, VM_MAYWRITE vmas: do_mmap
597 * marks vma as VM_SHARED if it is shared, and the file was opened for
598 * writing i.e. vma may be mprotected writable even if now readonly.
599 *
600 * If i_mmap_writable is negative, no new writable mappings are allowed. You
601 * can only deny writable mappings, if none exists right now.
602 */
603static inline int mapping_writably_mapped(const struct address_space *mapping)
604{
605 return atomic_read(v: &mapping->i_mmap_writable) > 0;
606}
607
608static inline int mapping_map_writable(struct address_space *mapping)
609{
610 return atomic_inc_unless_negative(v: &mapping->i_mmap_writable) ?
611 0 : -EPERM;
612}
613
614static inline void mapping_unmap_writable(struct address_space *mapping)
615{
616 atomic_dec(v: &mapping->i_mmap_writable);
617}
618
619static inline int mapping_deny_writable(struct address_space *mapping)
620{
621 return atomic_dec_unless_positive(v: &mapping->i_mmap_writable) ?
622 0 : -EBUSY;
623}
624
625static inline void mapping_allow_writable(struct address_space *mapping)
626{
627 atomic_inc(v: &mapping->i_mmap_writable);
628}
629
630/*
631 * Use sequence counter to get consistent i_size on 32-bit processors.
632 */
633#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
634#include <linux/seqlock.h>
635#define __NEED_I_SIZE_ORDERED
636#define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
637#else
638#define i_size_ordered_init(inode) do { } while (0)
639#endif
640
641struct posix_acl;
642#define ACL_NOT_CACHED ((void *)(-1))
643/*
644 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
645 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU
646 * mode with the LOOKUP_RCU flag.
647 */
648#define ACL_DONT_CACHE ((void *)(-3))
649
650static inline struct posix_acl *
651uncached_acl_sentinel(struct task_struct *task)
652{
653 return (void *)task + 1;
654}
655
656static inline bool
657is_uncached_acl(struct posix_acl *acl)
658{
659 return (long)acl & 1;
660}
661
662#define IOP_FASTPERM 0x0001
663#define IOP_LOOKUP 0x0002
664#define IOP_NOFOLLOW 0x0004
665#define IOP_XATTR 0x0008
666#define IOP_DEFAULT_READLINK 0x0010
667#define IOP_MGTIME 0x0020
668#define IOP_CACHED_LINK 0x0040
669
670/*
671 * Inode state bits. Protected by inode->i_lock
672 *
673 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
674 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
675 *
676 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
677 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
678 * various stages of removing an inode.
679 *
680 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
681 *
682 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
683 * fdatasync() (unless I_DIRTY_DATASYNC is also set).
684 * Timestamp updates are the usual cause.
685 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
686 * these changes separately from I_DIRTY_SYNC so that we
687 * don't have to write inode on fdatasync() when only
688 * e.g. the timestamps have changed.
689 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
690 * I_DIRTY_TIME The inode itself has dirty timestamps, and the
691 * lazytime mount option is enabled. We keep track of this
692 * separately from I_DIRTY_SYNC in order to implement
693 * lazytime. This gets cleared if I_DIRTY_INODE
694 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But
695 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already
696 * in place because writeback might already be in progress
697 * and we don't want to lose the time update
698 * I_NEW Serves as both a mutex and completion notification.
699 * New inodes set I_NEW. If two processes both create
700 * the same inode, one of them will release its inode and
701 * wait for I_NEW to be released before returning.
702 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
703 * also cause waiting on I_NEW, without I_NEW actually
704 * being set. find_inode() uses this to prevent returning
705 * nearly-dead inodes.
706 * I_WILL_FREE Must be set when calling write_inode_now() if i_count
707 * is zero. I_FREEING must be set when I_WILL_FREE is
708 * cleared.
709 * I_FREEING Set when inode is about to be freed but still has dirty
710 * pages or buffers attached or the inode itself is still
711 * dirty.
712 * I_CLEAR Added by clear_inode(). In this state the inode is
713 * clean and can be destroyed. Inode keeps I_FREEING.
714 *
715 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
716 * prohibited for many purposes. iget() must wait for
717 * the inode to be completely released, then create it
718 * anew. Other functions will just ignore such inodes,
719 * if appropriate. I_NEW is used for waiting.
720 *
721 * I_SYNC Writeback of inode is running. The bit is set during
722 * data writeback, and cleared with a wakeup on the bit
723 * address once it is done. The bit is also used to pin
724 * the inode in memory for flusher thread.
725 *
726 * I_REFERENCED Marks the inode as recently references on the LRU list.
727 *
728 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to
729 * synchronize competing switching instances and to tell
730 * wb stat updates to grab the i_pages lock. See
731 * inode_switch_wbs_work_fn() for details.
732 *
733 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper
734 * and work dirs among overlayfs mounts.
735 *
736 * I_CREATING New object's inode in the middle of setting up.
737 *
738 * I_DONTCACHE Evict inode as soon as it is not used anymore.
739 *
740 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists.
741 * Used to detect that mark_inode_dirty() should not move
742 * inode between dirty lists.
743 *
744 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback.
745 *
746 * I_LRU_ISOLATING Inode is pinned being isolated from LRU without holding
747 * i_count.
748 *
749 * Q: What is the difference between I_WILL_FREE and I_FREEING?
750 *
751 * __I_{SYNC,NEW,LRU_ISOLATING} are used to derive unique addresses to wait
752 * upon. There's one free address left.
753 */
754
755enum inode_state_bits {
756 __I_NEW = 0U,
757 __I_SYNC = 1U,
758 __I_LRU_ISOLATING = 2U
759 /* reserved wait address bit 3 */
760};
761
762enum inode_state_flags_t {
763 I_NEW = (1U << __I_NEW),
764 I_SYNC = (1U << __I_SYNC),
765 I_LRU_ISOLATING = (1U << __I_LRU_ISOLATING),
766 /* reserved flag bit 3 */
767 I_DIRTY_SYNC = (1U << 4),
768 I_DIRTY_DATASYNC = (1U << 5),
769 I_DIRTY_PAGES = (1U << 6),
770 I_WILL_FREE = (1U << 7),
771 I_FREEING = (1U << 8),
772 I_CLEAR = (1U << 9),
773 I_REFERENCED = (1U << 10),
774 I_LINKABLE = (1U << 11),
775 I_DIRTY_TIME = (1U << 12),
776 I_WB_SWITCH = (1U << 13),
777 I_OVL_INUSE = (1U << 14),
778 I_CREATING = (1U << 15),
779 I_DONTCACHE = (1U << 16),
780 I_SYNC_QUEUED = (1U << 17),
781 I_PINNING_NETFS_WB = (1U << 18)
782};
783
784#define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
785#define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
786#define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
787
788/*
789 * Keep mostly read-only and often accessed (especially for
790 * the RCU path lookup and 'stat' data) fields at the beginning
791 * of the 'struct inode'
792 */
793struct inode {
794 umode_t i_mode;
795 unsigned short i_opflags;
796 kuid_t i_uid;
797 kgid_t i_gid;
798 unsigned int i_flags;
799
800#ifdef CONFIG_FS_POSIX_ACL
801 struct posix_acl *i_acl;
802 struct posix_acl *i_default_acl;
803#endif
804
805 const struct inode_operations *i_op;
806 struct super_block *i_sb;
807 struct address_space *i_mapping;
808
809#ifdef CONFIG_SECURITY
810 void *i_security;
811#endif
812
813 /* Stat data, not accessed from path walking */
814 unsigned long i_ino;
815 /*
816 * Filesystems may only read i_nlink directly. They shall use the
817 * following functions for modification:
818 *
819 * (set|clear|inc|drop)_nlink
820 * inode_(inc|dec)_link_count
821 */
822 union {
823 const unsigned int i_nlink;
824 unsigned int __i_nlink;
825 };
826 dev_t i_rdev;
827 loff_t i_size;
828 time64_t i_atime_sec;
829 time64_t i_mtime_sec;
830 time64_t i_ctime_sec;
831 u32 i_atime_nsec;
832 u32 i_mtime_nsec;
833 u32 i_ctime_nsec;
834 u32 i_generation;
835 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
836 unsigned short i_bytes;
837 u8 i_blkbits;
838 enum rw_hint i_write_hint;
839 blkcnt_t i_blocks;
840
841#ifdef __NEED_I_SIZE_ORDERED
842 seqcount_t i_size_seqcount;
843#endif
844
845 /* Misc */
846 enum inode_state_flags_t i_state;
847 /* 32-bit hole */
848 struct rw_semaphore i_rwsem;
849
850 unsigned long dirtied_when; /* jiffies of first dirtying */
851 unsigned long dirtied_time_when;
852
853 struct hlist_node i_hash;
854 struct list_head i_io_list; /* backing dev IO list */
855#ifdef CONFIG_CGROUP_WRITEBACK
856 struct bdi_writeback *i_wb; /* the associated cgroup wb */
857
858 /* foreign inode detection, see wbc_detach_inode() */
859 int i_wb_frn_winner;
860 u16 i_wb_frn_avg_time;
861 u16 i_wb_frn_history;
862#endif
863 struct list_head i_lru; /* inode LRU list */
864 struct list_head i_sb_list;
865 struct list_head i_wb_list; /* backing dev writeback list */
866 union {
867 struct hlist_head i_dentry;
868 struct rcu_head i_rcu;
869 };
870 atomic64_t i_version;
871 atomic64_t i_sequence; /* see futex */
872 atomic_t i_count;
873 atomic_t i_dio_count;
874 atomic_t i_writecount;
875#if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
876 atomic_t i_readcount; /* struct files open RO */
877#endif
878 union {
879 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
880 void (*free_inode)(struct inode *);
881 };
882 struct file_lock_context *i_flctx;
883 struct address_space i_data;
884 union {
885 struct list_head i_devices;
886 int i_linklen;
887 };
888 union {
889 struct pipe_inode_info *i_pipe;
890 struct cdev *i_cdev;
891 char *i_link;
892 unsigned i_dir_seq;
893 };
894
895
896#ifdef CONFIG_FSNOTIFY
897 __u32 i_fsnotify_mask; /* all events this inode cares about */
898 /* 32-bit hole reserved for expanding i_fsnotify_mask */
899 struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
900#endif
901
902 void *i_private; /* fs or device private pointer */
903} __randomize_layout;
904
905static inline void inode_set_cached_link(struct inode *inode, char *link, int linklen)
906{
907 VFS_WARN_ON_INODE(strlen(link) != linklen, inode);
908 VFS_WARN_ON_INODE(inode->i_opflags & IOP_CACHED_LINK, inode);
909 inode->i_link = link;
910 inode->i_linklen = linklen;
911 inode->i_opflags |= IOP_CACHED_LINK;
912}
913
914/*
915 * Get bit address from inode->i_state to use with wait_var_event()
916 * infrastructre.
917 */
918#define inode_state_wait_address(inode, bit) ((char *)&(inode)->i_state + (bit))
919
920struct wait_queue_head *inode_bit_waitqueue(struct wait_bit_queue_entry *wqe,
921 struct inode *inode, u32 bit);
922
923static inline void inode_wake_up_bit(struct inode *inode, u32 bit)
924{
925 /* Caller is responsible for correct memory barriers. */
926 wake_up_var(inode_state_wait_address(inode, bit));
927}
928
929struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
930
931static inline unsigned int i_blocksize(const struct inode *node)
932{
933 return (1 << node->i_blkbits);
934}
935
936static inline int inode_unhashed(struct inode *inode)
937{
938 return hlist_unhashed(h: &inode->i_hash);
939}
940
941/*
942 * __mark_inode_dirty expects inodes to be hashed. Since we don't
943 * want special inodes in the fileset inode space, we make them
944 * appear hashed, but do not put on any lists. hlist_del()
945 * will work fine and require no locking.
946 */
947static inline void inode_fake_hash(struct inode *inode)
948{
949 hlist_add_fake(n: &inode->i_hash);
950}
951
952/*
953 * inode->i_rwsem nesting subclasses for the lock validator:
954 *
955 * 0: the object of the current VFS operation
956 * 1: parent
957 * 2: child/target
958 * 3: xattr
959 * 4: second non-directory
960 * 5: second parent (when locking independent directories in rename)
961 *
962 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
963 * non-directories at once.
964 *
965 * The locking order between these classes is
966 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
967 */
968enum inode_i_mutex_lock_class
969{
970 I_MUTEX_NORMAL,
971 I_MUTEX_PARENT,
972 I_MUTEX_CHILD,
973 I_MUTEX_XATTR,
974 I_MUTEX_NONDIR2,
975 I_MUTEX_PARENT2,
976};
977
978static inline void inode_lock(struct inode *inode)
979{
980 down_write(sem: &inode->i_rwsem);
981}
982
983static inline __must_check int inode_lock_killable(struct inode *inode)
984{
985 return down_write_killable(sem: &inode->i_rwsem);
986}
987
988static inline void inode_unlock(struct inode *inode)
989{
990 up_write(sem: &inode->i_rwsem);
991}
992
993static inline void inode_lock_shared(struct inode *inode)
994{
995 down_read(sem: &inode->i_rwsem);
996}
997
998static inline __must_check int inode_lock_shared_killable(struct inode *inode)
999{
1000 return down_read_killable(sem: &inode->i_rwsem);
1001}
1002
1003static inline void inode_unlock_shared(struct inode *inode)
1004{
1005 up_read(sem: &inode->i_rwsem);
1006}
1007
1008static inline int inode_trylock(struct inode *inode)
1009{
1010 return down_write_trylock(sem: &inode->i_rwsem);
1011}
1012
1013static inline int inode_trylock_shared(struct inode *inode)
1014{
1015 return down_read_trylock(sem: &inode->i_rwsem);
1016}
1017
1018static inline int inode_is_locked(struct inode *inode)
1019{
1020 return rwsem_is_locked(sem: &inode->i_rwsem);
1021}
1022
1023static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
1024{
1025 down_write_nested(&inode->i_rwsem, subclass);
1026}
1027
1028static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
1029{
1030 down_read_nested(&inode->i_rwsem, subclass);
1031}
1032
1033static inline void filemap_invalidate_lock(struct address_space *mapping)
1034{
1035 down_write(sem: &mapping->invalidate_lock);
1036}
1037
1038static inline void filemap_invalidate_unlock(struct address_space *mapping)
1039{
1040 up_write(sem: &mapping->invalidate_lock);
1041}
1042
1043static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
1044{
1045 down_read(sem: &mapping->invalidate_lock);
1046}
1047
1048static inline int filemap_invalidate_trylock_shared(
1049 struct address_space *mapping)
1050{
1051 return down_read_trylock(sem: &mapping->invalidate_lock);
1052}
1053
1054static inline void filemap_invalidate_unlock_shared(
1055 struct address_space *mapping)
1056{
1057 up_read(sem: &mapping->invalidate_lock);
1058}
1059
1060void lock_two_nondirectories(struct inode *, struct inode*);
1061void unlock_two_nondirectories(struct inode *, struct inode*);
1062
1063void filemap_invalidate_lock_two(struct address_space *mapping1,
1064 struct address_space *mapping2);
1065void filemap_invalidate_unlock_two(struct address_space *mapping1,
1066 struct address_space *mapping2);
1067
1068
1069/*
1070 * NOTE: in a 32bit arch with a preemptable kernel and
1071 * an UP compile the i_size_read/write must be atomic
1072 * with respect to the local cpu (unlike with preempt disabled),
1073 * but they don't need to be atomic with respect to other cpus like in
1074 * true SMP (so they need either to either locally disable irq around
1075 * the read or for example on x86 they can be still implemented as a
1076 * cmpxchg8b without the need of the lock prefix). For SMP compiles
1077 * and 64bit archs it makes no difference if preempt is enabled or not.
1078 */
1079static inline loff_t i_size_read(const struct inode *inode)
1080{
1081#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
1082 loff_t i_size;
1083 unsigned int seq;
1084
1085 do {
1086 seq = read_seqcount_begin(&inode->i_size_seqcount);
1087 i_size = inode->i_size;
1088 } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
1089 return i_size;
1090#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
1091 loff_t i_size;
1092
1093 preempt_disable();
1094 i_size = inode->i_size;
1095 preempt_enable();
1096 return i_size;
1097#else
1098 /* Pairs with smp_store_release() in i_size_write() */
1099 return smp_load_acquire(&inode->i_size);
1100#endif
1101}
1102
1103/*
1104 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
1105 * (normally i_rwsem), otherwise on 32bit/SMP an update of i_size_seqcount
1106 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
1107 */
1108static inline void i_size_write(struct inode *inode, loff_t i_size)
1109{
1110#if BITS_PER_LONG==32 && defined(CONFIG_SMP)
1111 preempt_disable();
1112 write_seqcount_begin(&inode->i_size_seqcount);
1113 inode->i_size = i_size;
1114 write_seqcount_end(&inode->i_size_seqcount);
1115 preempt_enable();
1116#elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
1117 preempt_disable();
1118 inode->i_size = i_size;
1119 preempt_enable();
1120#else
1121 /*
1122 * Pairs with smp_load_acquire() in i_size_read() to ensure
1123 * changes related to inode size (such as page contents) are
1124 * visible before we see the changed inode size.
1125 */
1126 smp_store_release(&inode->i_size, i_size);
1127#endif
1128}
1129
1130static inline unsigned iminor(const struct inode *inode)
1131{
1132 return MINOR(inode->i_rdev);
1133}
1134
1135static inline unsigned imajor(const struct inode *inode)
1136{
1137 return MAJOR(inode->i_rdev);
1138}
1139
1140struct fown_struct {
1141 struct file *file; /* backpointer for security modules */
1142 rwlock_t lock; /* protects pid, uid, euid fields */
1143 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */
1144 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
1145 kuid_t uid, euid; /* uid/euid of process setting the owner */
1146 int signum; /* posix.1b rt signal to be delivered on IO */
1147};
1148
1149/**
1150 * struct file_ra_state - Track a file's readahead state.
1151 * @start: Where the most recent readahead started.
1152 * @size: Number of pages read in the most recent readahead.
1153 * @async_size: Numer of pages that were/are not needed immediately
1154 * and so were/are genuinely "ahead". Start next readahead when
1155 * the first of these pages is accessed.
1156 * @ra_pages: Maximum size of a readahead request, copied from the bdi.
1157 * @order: Preferred folio order used for most recent readahead.
1158 * @mmap_miss: How many mmap accesses missed in the page cache.
1159 * @prev_pos: The last byte in the most recent read request.
1160 *
1161 * When this structure is passed to ->readahead(), the "most recent"
1162 * readahead means the current readahead.
1163 */
1164struct file_ra_state {
1165 pgoff_t start;
1166 unsigned int size;
1167 unsigned int async_size;
1168 unsigned int ra_pages;
1169 unsigned short order;
1170 unsigned short mmap_miss;
1171 loff_t prev_pos;
1172};
1173
1174/*
1175 * Check if @index falls in the readahead windows.
1176 */
1177static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
1178{
1179 return (index >= ra->start &&
1180 index < ra->start + ra->size);
1181}
1182
1183/**
1184 * struct file - Represents a file
1185 * @f_lock: Protects f_ep, f_flags. Must not be taken from IRQ context.
1186 * @f_mode: FMODE_* flags often used in hotpaths
1187 * @f_op: file operations
1188 * @f_mapping: Contents of a cacheable, mappable object.
1189 * @private_data: filesystem or driver specific data
1190 * @f_inode: cached inode
1191 * @f_flags: file flags
1192 * @f_iocb_flags: iocb flags
1193 * @f_cred: stashed credentials of creator/opener
1194 * @f_owner: file owner
1195 * @f_path: path of the file
1196 * @__f_path: writable alias for @f_path; *ONLY* for core VFS and only before
1197 * the file gets open
1198 * @f_pos_lock: lock protecting file position
1199 * @f_pipe: specific to pipes
1200 * @f_pos: file position
1201 * @f_security: LSM security context of this file
1202 * @f_wb_err: writeback error
1203 * @f_sb_err: per sb writeback errors
1204 * @f_ep: link of all epoll hooks for this file
1205 * @f_task_work: task work entry point
1206 * @f_llist: work queue entrypoint
1207 * @f_ra: file's readahead state
1208 * @f_freeptr: Pointer used by SLAB_TYPESAFE_BY_RCU file cache (don't touch.)
1209 * @f_ref: reference count
1210 */
1211struct file {
1212 spinlock_t f_lock;
1213 fmode_t f_mode;
1214 const struct file_operations *f_op;
1215 struct address_space *f_mapping;
1216 void *private_data;
1217 struct inode *f_inode;
1218 unsigned int f_flags;
1219 unsigned int f_iocb_flags;
1220 const struct cred *f_cred;
1221 struct fown_struct *f_owner;
1222 /* --- cacheline 1 boundary (64 bytes) --- */
1223 union {
1224 const struct path f_path;
1225 struct path __f_path;
1226 };
1227 union {
1228 /* regular files (with FMODE_ATOMIC_POS) and directories */
1229 struct mutex f_pos_lock;
1230 /* pipes */
1231 u64 f_pipe;
1232 };
1233 loff_t f_pos;
1234#ifdef CONFIG_SECURITY
1235 void *f_security;
1236#endif
1237 /* --- cacheline 2 boundary (128 bytes) --- */
1238 errseq_t f_wb_err;
1239 errseq_t f_sb_err;
1240#ifdef CONFIG_EPOLL
1241 struct hlist_head *f_ep;
1242#endif
1243 union {
1244 struct callback_head f_task_work;
1245 struct llist_node f_llist;
1246 struct file_ra_state f_ra;
1247 freeptr_t f_freeptr;
1248 };
1249 file_ref_t f_ref;
1250 /* --- cacheline 3 boundary (192 bytes) --- */
1251} __randomize_layout
1252 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
1253
1254struct file_handle {
1255 __u32 handle_bytes;
1256 int handle_type;
1257 /* file identifier */
1258 unsigned char f_handle[] __counted_by(handle_bytes);
1259};
1260
1261static inline struct file *get_file(struct file *f)
1262{
1263 file_ref_inc(ref: &f->f_ref);
1264 return f;
1265}
1266
1267struct file *get_file_rcu(struct file __rcu **f);
1268struct file *get_file_active(struct file **f);
1269
1270#define file_count(f) file_ref_read(&(f)->f_ref)
1271
1272#define MAX_NON_LFS ((1UL<<31) - 1)
1273
1274/* Page cache limit. The filesystems should put that into their s_maxbytes
1275 limits, otherwise bad things can happen in VM. */
1276#if BITS_PER_LONG==32
1277#define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT)
1278#elif BITS_PER_LONG==64
1279#define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX)
1280#endif
1281
1282/* legacy typedef, should eventually be removed */
1283typedef void *fl_owner_t;
1284
1285struct file_lock;
1286struct file_lease;
1287
1288/* The following constant reflects the upper bound of the file/locking space */
1289#ifndef OFFSET_MAX
1290#define OFFSET_MAX type_max(loff_t)
1291#define OFFT_OFFSET_MAX type_max(off_t)
1292#endif
1293
1294int file_f_owner_allocate(struct file *file);
1295static inline struct fown_struct *file_f_owner(const struct file *file)
1296{
1297 return READ_ONCE(file->f_owner);
1298}
1299
1300extern void send_sigio(struct fown_struct *fown, int fd, int band);
1301
1302static inline struct inode *file_inode(const struct file *f)
1303{
1304 return f->f_inode;
1305}
1306
1307/*
1308 * file_dentry() is a relic from the days that overlayfs was using files with a
1309 * "fake" path, meaning, f_path on overlayfs and f_inode on underlying fs.
1310 * In those days, file_dentry() was needed to get the underlying fs dentry that
1311 * matches f_inode.
1312 * Files with "fake" path should not exist nowadays, so use an assertion to make
1313 * sure that file_dentry() was not papering over filesystem bugs.
1314 */
1315static inline struct dentry *file_dentry(const struct file *file)
1316{
1317 struct dentry *dentry = file->f_path.dentry;
1318
1319 WARN_ON_ONCE(d_inode(dentry) != file_inode(file));
1320 return dentry;
1321}
1322
1323struct fasync_struct {
1324 rwlock_t fa_lock;
1325 int magic;
1326 int fa_fd;
1327 struct fasync_struct *fa_next; /* singly linked list */
1328 struct file *fa_file;
1329 struct rcu_head fa_rcu;
1330};
1331
1332#define FASYNC_MAGIC 0x4601
1333
1334/* SMP safe fasync helpers: */
1335extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1336extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1337extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1338extern struct fasync_struct *fasync_alloc(void);
1339extern void fasync_free(struct fasync_struct *);
1340
1341/* can be called from interrupts */
1342extern void kill_fasync(struct fasync_struct **, int, int);
1343
1344extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1345extern int f_setown(struct file *filp, int who, int force);
1346extern void f_delown(struct file *filp);
1347extern pid_t f_getown(struct file *filp);
1348extern int send_sigurg(struct file *file);
1349
1350/*
1351 * sb->s_flags. Note that these mirror the equivalent MS_* flags where
1352 * represented in both.
1353 */
1354#define SB_RDONLY BIT(0) /* Mount read-only */
1355#define SB_NOSUID BIT(1) /* Ignore suid and sgid bits */
1356#define SB_NODEV BIT(2) /* Disallow access to device special files */
1357#define SB_NOEXEC BIT(3) /* Disallow program execution */
1358#define SB_SYNCHRONOUS BIT(4) /* Writes are synced at once */
1359#define SB_MANDLOCK BIT(6) /* Allow mandatory locks on an FS */
1360#define SB_DIRSYNC BIT(7) /* Directory modifications are synchronous */
1361#define SB_NOATIME BIT(10) /* Do not update access times. */
1362#define SB_NODIRATIME BIT(11) /* Do not update directory access times */
1363#define SB_SILENT BIT(15)
1364#define SB_POSIXACL BIT(16) /* Supports POSIX ACLs */
1365#define SB_INLINECRYPT BIT(17) /* Use blk-crypto for encrypted files */
1366#define SB_KERNMOUNT BIT(22) /* this is a kern_mount call */
1367#define SB_I_VERSION BIT(23) /* Update inode I_version field */
1368#define SB_LAZYTIME BIT(25) /* Update the on-disk [acm]times lazily */
1369
1370/* These sb flags are internal to the kernel */
1371#define SB_DEAD BIT(21)
1372#define SB_DYING BIT(24)
1373#define SB_FORCE BIT(27)
1374#define SB_NOSEC BIT(28)
1375#define SB_BORN BIT(29)
1376#define SB_ACTIVE BIT(30)
1377#define SB_NOUSER BIT(31)
1378
1379/* These flags relate to encoding and casefolding */
1380#define SB_ENC_STRICT_MODE_FL (1 << 0)
1381#define SB_ENC_NO_COMPAT_FALLBACK_FL (1 << 1)
1382
1383#define sb_has_strict_encoding(sb) \
1384 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1385
1386#if IS_ENABLED(CONFIG_UNICODE)
1387#define sb_no_casefold_compat_fallback(sb) \
1388 (sb->s_encoding_flags & SB_ENC_NO_COMPAT_FALLBACK_FL)
1389#else
1390#define sb_no_casefold_compat_fallback(sb) (1)
1391#endif
1392
1393/*
1394 * Umount options
1395 */
1396
1397#define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1398#define MNT_DETACH 0x00000002 /* Just detach from the tree */
1399#define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1400#define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1401#define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1402
1403/* sb->s_iflags */
1404#define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */
1405#define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */
1406#define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */
1407#define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */
1408
1409/* sb->s_iflags to limit user namespace mounts */
1410#define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */
1411#define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020
1412#define SB_I_UNTRUSTED_MOUNTER 0x00000040
1413#define SB_I_EVM_HMAC_UNSUPPORTED 0x00000080
1414
1415#define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */
1416#define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */
1417#define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1418#define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */
1419#define SB_I_NOUMASK 0x00001000 /* VFS does not apply umask */
1420#define SB_I_NOIDMAP 0x00002000 /* No idmapped mounts on this superblock */
1421#define SB_I_ALLOW_HSM 0x00004000 /* Allow HSM events on this superblock */
1422
1423/* Possible states of 'frozen' field */
1424enum {
1425 SB_UNFROZEN = 0, /* FS is unfrozen */
1426 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1427 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1428 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1429 * internal threads if needed) */
1430 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1431};
1432
1433#define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1434
1435struct sb_writers {
1436 unsigned short frozen; /* Is sb frozen? */
1437 int freeze_kcount; /* How many kernel freeze requests? */
1438 int freeze_ucount; /* How many userspace freeze requests? */
1439 const void *freeze_owner; /* Owner of the freeze */
1440 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS];
1441};
1442
1443struct mount;
1444
1445struct super_block {
1446 struct list_head s_list; /* Keep this first */
1447 dev_t s_dev; /* search index; _not_ kdev_t */
1448 unsigned char s_blocksize_bits;
1449 unsigned long s_blocksize;
1450 loff_t s_maxbytes; /* Max file size */
1451 struct file_system_type *s_type;
1452 const struct super_operations *s_op;
1453 const struct dquot_operations *dq_op;
1454 const struct quotactl_ops *s_qcop;
1455 const struct export_operations *s_export_op;
1456 unsigned long s_flags;
1457 unsigned long s_iflags; /* internal SB_I_* flags */
1458 unsigned long s_magic;
1459 struct dentry *s_root;
1460 struct rw_semaphore s_umount;
1461 int s_count;
1462 atomic_t s_active;
1463#ifdef CONFIG_SECURITY
1464 void *s_security;
1465#endif
1466 const struct xattr_handler * const *s_xattr;
1467#ifdef CONFIG_FS_ENCRYPTION
1468 const struct fscrypt_operations *s_cop;
1469 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */
1470#endif
1471#ifdef CONFIG_FS_VERITY
1472 const struct fsverity_operations *s_vop;
1473#endif
1474#if IS_ENABLED(CONFIG_UNICODE)
1475 struct unicode_map *s_encoding;
1476 __u16 s_encoding_flags;
1477#endif
1478 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
1479 struct mount *s_mounts; /* list of mounts; _not_ for fs use */
1480 struct block_device *s_bdev; /* can go away once we use an accessor for @s_bdev_file */
1481 struct file *s_bdev_file;
1482 struct backing_dev_info *s_bdi;
1483 struct mtd_info *s_mtd;
1484 struct hlist_node s_instances;
1485 unsigned int s_quota_types; /* Bitmask of supported quota types */
1486 struct quota_info s_dquot; /* Diskquota specific options */
1487
1488 struct sb_writers s_writers;
1489
1490 /*
1491 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1492 * s_fsnotify_info together for cache efficiency. They are frequently
1493 * accessed and rarely modified.
1494 */
1495 void *s_fs_info; /* Filesystem private info */
1496
1497 /* Granularity of c/m/atime in ns (cannot be worse than a second) */
1498 u32 s_time_gran;
1499 /* Time limits for c/m/atime in seconds */
1500 time64_t s_time_min;
1501 time64_t s_time_max;
1502#ifdef CONFIG_FSNOTIFY
1503 u32 s_fsnotify_mask;
1504 struct fsnotify_sb_info *s_fsnotify_info;
1505#endif
1506
1507 /*
1508 * q: why are s_id and s_sysfs_name not the same? both are human
1509 * readable strings that identify the filesystem
1510 * a: s_id is allowed to change at runtime; it's used in log messages,
1511 * and we want to when a device starts out as single device (s_id is dev
1512 * name) but then a device is hot added and we have to switch to
1513 * identifying it by UUID
1514 * but s_sysfs_name is a handle for programmatic access, and can't
1515 * change at runtime
1516 */
1517 char s_id[32]; /* Informational name */
1518 uuid_t s_uuid; /* UUID */
1519 u8 s_uuid_len; /* Default 16, possibly smaller for weird filesystems */
1520
1521 /* if set, fs shows up under sysfs at /sys/fs/$FSTYP/s_sysfs_name */
1522 char s_sysfs_name[UUID_STRING_LEN + 1];
1523
1524 unsigned int s_max_links;
1525 unsigned int s_d_flags; /* default d_flags for dentries */
1526
1527 /*
1528 * The next field is for VFS *only*. No filesystems have any business
1529 * even looking at it. You had been warned.
1530 */
1531 struct mutex s_vfs_rename_mutex; /* Kludge */
1532
1533 /*
1534 * Filesystem subtype. If non-empty the filesystem type field
1535 * in /proc/mounts will be "type.subtype"
1536 */
1537 const char *s_subtype;
1538
1539 const struct dentry_operations *__s_d_op; /* default d_op for dentries */
1540
1541 struct shrinker *s_shrink; /* per-sb shrinker handle */
1542
1543 /* Number of inodes with nlink == 0 but still referenced */
1544 atomic_long_t s_remove_count;
1545
1546 /* Read-only state of the superblock is being changed */
1547 int s_readonly_remount;
1548
1549 /* per-sb errseq_t for reporting writeback errors via syncfs */
1550 errseq_t s_wb_err;
1551
1552 /* AIO completions deferred from interrupt context */
1553 struct workqueue_struct *s_dio_done_wq;
1554 struct hlist_head s_pins;
1555
1556 /*
1557 * Owning user namespace and default context in which to
1558 * interpret filesystem uids, gids, quotas, device nodes,
1559 * xattrs and security labels.
1560 */
1561 struct user_namespace *s_user_ns;
1562
1563 /*
1564 * The list_lru structure is essentially just a pointer to a table
1565 * of per-node lru lists, each of which has its own spinlock.
1566 * There is no need to put them into separate cachelines.
1567 */
1568 struct list_lru s_dentry_lru;
1569 struct list_lru s_inode_lru;
1570 struct rcu_head rcu;
1571 struct work_struct destroy_work;
1572
1573 struct mutex s_sync_lock; /* sync serialisation lock */
1574
1575 /*
1576 * Indicates how deep in a filesystem stack this SB is
1577 */
1578 int s_stack_depth;
1579
1580 /* s_inode_list_lock protects s_inodes */
1581 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
1582 struct list_head s_inodes; /* all inodes */
1583
1584 spinlock_t s_inode_wblist_lock;
1585 struct list_head s_inodes_wb; /* writeback inodes */
1586} __randomize_layout;
1587
1588static inline struct user_namespace *i_user_ns(const struct inode *inode)
1589{
1590 return inode->i_sb->s_user_ns;
1591}
1592
1593/* Helper functions so that in most cases filesystems will
1594 * not need to deal directly with kuid_t and kgid_t and can
1595 * instead deal with the raw numeric values that are stored
1596 * in the filesystem.
1597 */
1598static inline uid_t i_uid_read(const struct inode *inode)
1599{
1600 return from_kuid(to: i_user_ns(inode), kuid: inode->i_uid);
1601}
1602
1603static inline gid_t i_gid_read(const struct inode *inode)
1604{
1605 return from_kgid(to: i_user_ns(inode), kgid: inode->i_gid);
1606}
1607
1608static inline void i_uid_write(struct inode *inode, uid_t uid)
1609{
1610 inode->i_uid = make_kuid(from: i_user_ns(inode), uid);
1611}
1612
1613static inline void i_gid_write(struct inode *inode, gid_t gid)
1614{
1615 inode->i_gid = make_kgid(from: i_user_ns(inode), gid);
1616}
1617
1618/**
1619 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping
1620 * @idmap: idmap of the mount the inode was found from
1621 * @inode: inode to map
1622 *
1623 * Return: whe inode's i_uid mapped down according to @idmap.
1624 * If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1625 */
1626static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap,
1627 const struct inode *inode)
1628{
1629 return make_vfsuid(idmap, fs_userns: i_user_ns(inode), kuid: inode->i_uid);
1630}
1631
1632/**
1633 * i_uid_needs_update - check whether inode's i_uid needs to be updated
1634 * @idmap: idmap of the mount the inode was found from
1635 * @attr: the new attributes of @inode
1636 * @inode: the inode to update
1637 *
1638 * Check whether the $inode's i_uid field needs to be updated taking idmapped
1639 * mounts into account if the filesystem supports it.
1640 *
1641 * Return: true if @inode's i_uid field needs to be updated, false if not.
1642 */
1643static inline bool i_uid_needs_update(struct mnt_idmap *idmap,
1644 const struct iattr *attr,
1645 const struct inode *inode)
1646{
1647 return ((attr->ia_valid & ATTR_UID) &&
1648 !vfsuid_eq(left: attr->ia_vfsuid,
1649 right: i_uid_into_vfsuid(idmap, inode)));
1650}
1651
1652/**
1653 * i_uid_update - update @inode's i_uid field
1654 * @idmap: idmap of the mount the inode was found from
1655 * @attr: the new attributes of @inode
1656 * @inode: the inode to update
1657 *
1658 * Safely update @inode's i_uid field translating the vfsuid of any idmapped
1659 * mount into the filesystem kuid.
1660 */
1661static inline void i_uid_update(struct mnt_idmap *idmap,
1662 const struct iattr *attr,
1663 struct inode *inode)
1664{
1665 if (attr->ia_valid & ATTR_UID)
1666 inode->i_uid = from_vfsuid(idmap, fs_userns: i_user_ns(inode),
1667 vfsuid: attr->ia_vfsuid);
1668}
1669
1670/**
1671 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping
1672 * @idmap: idmap of the mount the inode was found from
1673 * @inode: inode to map
1674 *
1675 * Return: the inode's i_gid mapped down according to @idmap.
1676 * If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1677 */
1678static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap,
1679 const struct inode *inode)
1680{
1681 return make_vfsgid(idmap, fs_userns: i_user_ns(inode), kgid: inode->i_gid);
1682}
1683
1684/**
1685 * i_gid_needs_update - check whether inode's i_gid needs to be updated
1686 * @idmap: idmap of the mount the inode was found from
1687 * @attr: the new attributes of @inode
1688 * @inode: the inode to update
1689 *
1690 * Check whether the $inode's i_gid field needs to be updated taking idmapped
1691 * mounts into account if the filesystem supports it.
1692 *
1693 * Return: true if @inode's i_gid field needs to be updated, false if not.
1694 */
1695static inline bool i_gid_needs_update(struct mnt_idmap *idmap,
1696 const struct iattr *attr,
1697 const struct inode *inode)
1698{
1699 return ((attr->ia_valid & ATTR_GID) &&
1700 !vfsgid_eq(left: attr->ia_vfsgid,
1701 right: i_gid_into_vfsgid(idmap, inode)));
1702}
1703
1704/**
1705 * i_gid_update - update @inode's i_gid field
1706 * @idmap: idmap of the mount the inode was found from
1707 * @attr: the new attributes of @inode
1708 * @inode: the inode to update
1709 *
1710 * Safely update @inode's i_gid field translating the vfsgid of any idmapped
1711 * mount into the filesystem kgid.
1712 */
1713static inline void i_gid_update(struct mnt_idmap *idmap,
1714 const struct iattr *attr,
1715 struct inode *inode)
1716{
1717 if (attr->ia_valid & ATTR_GID)
1718 inode->i_gid = from_vfsgid(idmap, fs_userns: i_user_ns(inode),
1719 vfsgid: attr->ia_vfsgid);
1720}
1721
1722/**
1723 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1724 * @inode: inode to initialize
1725 * @idmap: idmap of the mount the inode was found from
1726 *
1727 * Initialize the i_uid field of @inode. If the inode was found/created via
1728 * an idmapped mount map the caller's fsuid according to @idmap.
1729 */
1730static inline void inode_fsuid_set(struct inode *inode,
1731 struct mnt_idmap *idmap)
1732{
1733 inode->i_uid = mapped_fsuid(idmap, fs_userns: i_user_ns(inode));
1734}
1735
1736/**
1737 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1738 * @inode: inode to initialize
1739 * @idmap: idmap of the mount the inode was found from
1740 *
1741 * Initialize the i_gid field of @inode. If the inode was found/created via
1742 * an idmapped mount map the caller's fsgid according to @idmap.
1743 */
1744static inline void inode_fsgid_set(struct inode *inode,
1745 struct mnt_idmap *idmap)
1746{
1747 inode->i_gid = mapped_fsgid(idmap, fs_userns: i_user_ns(inode));
1748}
1749
1750/**
1751 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1752 * @sb: the superblock we want a mapping in
1753 * @idmap: idmap of the relevant mount
1754 *
1755 * Check whether the caller's fsuid and fsgid have a valid mapping in the
1756 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1757 * the caller's fsuid and fsgid according to the @idmap first.
1758 *
1759 * Return: true if fsuid and fsgid is mapped, false if not.
1760 */
1761static inline bool fsuidgid_has_mapping(struct super_block *sb,
1762 struct mnt_idmap *idmap)
1763{
1764 struct user_namespace *fs_userns = sb->s_user_ns;
1765 kuid_t kuid;
1766 kgid_t kgid;
1767
1768 kuid = mapped_fsuid(idmap, fs_userns);
1769 if (!uid_valid(uid: kuid))
1770 return false;
1771 kgid = mapped_fsgid(idmap, fs_userns);
1772 if (!gid_valid(gid: kgid))
1773 return false;
1774 return kuid_has_mapping(ns: fs_userns, uid: kuid) &&
1775 kgid_has_mapping(ns: fs_userns, gid: kgid);
1776}
1777
1778struct timespec64 current_time(struct inode *inode);
1779struct timespec64 inode_set_ctime_current(struct inode *inode);
1780struct timespec64 inode_set_ctime_deleg(struct inode *inode,
1781 struct timespec64 update);
1782
1783static inline time64_t inode_get_atime_sec(const struct inode *inode)
1784{
1785 return inode->i_atime_sec;
1786}
1787
1788static inline long inode_get_atime_nsec(const struct inode *inode)
1789{
1790 return inode->i_atime_nsec;
1791}
1792
1793static inline struct timespec64 inode_get_atime(const struct inode *inode)
1794{
1795 struct timespec64 ts = { .tv_sec = inode_get_atime_sec(inode),
1796 .tv_nsec = inode_get_atime_nsec(inode) };
1797
1798 return ts;
1799}
1800
1801static inline struct timespec64 inode_set_atime_to_ts(struct inode *inode,
1802 struct timespec64 ts)
1803{
1804 inode->i_atime_sec = ts.tv_sec;
1805 inode->i_atime_nsec = ts.tv_nsec;
1806 return ts;
1807}
1808
1809static inline struct timespec64 inode_set_atime(struct inode *inode,
1810 time64_t sec, long nsec)
1811{
1812 struct timespec64 ts = { .tv_sec = sec,
1813 .tv_nsec = nsec };
1814
1815 return inode_set_atime_to_ts(inode, ts);
1816}
1817
1818static inline time64_t inode_get_mtime_sec(const struct inode *inode)
1819{
1820 return inode->i_mtime_sec;
1821}
1822
1823static inline long inode_get_mtime_nsec(const struct inode *inode)
1824{
1825 return inode->i_mtime_nsec;
1826}
1827
1828static inline struct timespec64 inode_get_mtime(const struct inode *inode)
1829{
1830 struct timespec64 ts = { .tv_sec = inode_get_mtime_sec(inode),
1831 .tv_nsec = inode_get_mtime_nsec(inode) };
1832 return ts;
1833}
1834
1835static inline struct timespec64 inode_set_mtime_to_ts(struct inode *inode,
1836 struct timespec64 ts)
1837{
1838 inode->i_mtime_sec = ts.tv_sec;
1839 inode->i_mtime_nsec = ts.tv_nsec;
1840 return ts;
1841}
1842
1843static inline struct timespec64 inode_set_mtime(struct inode *inode,
1844 time64_t sec, long nsec)
1845{
1846 struct timespec64 ts = { .tv_sec = sec,
1847 .tv_nsec = nsec };
1848 return inode_set_mtime_to_ts(inode, ts);
1849}
1850
1851/*
1852 * Multigrain timestamps
1853 *
1854 * Conditionally use fine-grained ctime and mtime timestamps when there
1855 * are users actively observing them via getattr. The primary use-case
1856 * for this is NFS clients that use the ctime to distinguish between
1857 * different states of the file, and that are often fooled by multiple
1858 * operations that occur in the same coarse-grained timer tick.
1859 */
1860#define I_CTIME_QUERIED ((u32)BIT(31))
1861
1862static inline time64_t inode_get_ctime_sec(const struct inode *inode)
1863{
1864 return inode->i_ctime_sec;
1865}
1866
1867static inline long inode_get_ctime_nsec(const struct inode *inode)
1868{
1869 return inode->i_ctime_nsec & ~I_CTIME_QUERIED;
1870}
1871
1872static inline struct timespec64 inode_get_ctime(const struct inode *inode)
1873{
1874 struct timespec64 ts = { .tv_sec = inode_get_ctime_sec(inode),
1875 .tv_nsec = inode_get_ctime_nsec(inode) };
1876
1877 return ts;
1878}
1879
1880struct timespec64 inode_set_ctime_to_ts(struct inode *inode, struct timespec64 ts);
1881
1882/**
1883 * inode_set_ctime - set the ctime in the inode
1884 * @inode: inode in which to set the ctime
1885 * @sec: tv_sec value to set
1886 * @nsec: tv_nsec value to set
1887 *
1888 * Set the ctime in @inode to { @sec, @nsec }
1889 */
1890static inline struct timespec64 inode_set_ctime(struct inode *inode,
1891 time64_t sec, long nsec)
1892{
1893 struct timespec64 ts = { .tv_sec = sec,
1894 .tv_nsec = nsec };
1895
1896 return inode_set_ctime_to_ts(inode, ts);
1897}
1898
1899struct timespec64 simple_inode_init_ts(struct inode *inode);
1900
1901/*
1902 * Snapshotting support.
1903 */
1904
1905/*
1906 * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1907 * instead.
1908 */
1909static inline void __sb_end_write(struct super_block *sb, int level)
1910{
1911 percpu_up_read(sem: sb->s_writers.rw_sem + level-1);
1912}
1913
1914static inline void __sb_start_write(struct super_block *sb, int level)
1915{
1916 percpu_down_read_freezable(sem: sb->s_writers.rw_sem + level - 1, freeze: true);
1917}
1918
1919static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1920{
1921 return percpu_down_read_trylock(sem: sb->s_writers.rw_sem + level - 1);
1922}
1923
1924#define __sb_writers_acquired(sb, lev) \
1925 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1926#define __sb_writers_release(sb, lev) \
1927 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], _THIS_IP_)
1928
1929/**
1930 * __sb_write_started - check if sb freeze level is held
1931 * @sb: the super we write to
1932 * @level: the freeze level
1933 *
1934 * * > 0 - sb freeze level is held
1935 * * 0 - sb freeze level is not held
1936 * * < 0 - !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN
1937 */
1938static inline int __sb_write_started(const struct super_block *sb, int level)
1939{
1940 return lockdep_is_held_type(sb->s_writers.rw_sem + level - 1, 1);
1941}
1942
1943/**
1944 * sb_write_started - check if SB_FREEZE_WRITE is held
1945 * @sb: the super we write to
1946 *
1947 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1948 */
1949static inline bool sb_write_started(const struct super_block *sb)
1950{
1951 return __sb_write_started(sb, level: SB_FREEZE_WRITE);
1952}
1953
1954/**
1955 * sb_write_not_started - check if SB_FREEZE_WRITE is not held
1956 * @sb: the super we write to
1957 *
1958 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1959 */
1960static inline bool sb_write_not_started(const struct super_block *sb)
1961{
1962 return __sb_write_started(sb, level: SB_FREEZE_WRITE) <= 0;
1963}
1964
1965/**
1966 * file_write_started - check if SB_FREEZE_WRITE is held
1967 * @file: the file we write to
1968 *
1969 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1970 * May be false positive with !S_ISREG, because file_start_write() has
1971 * no effect on !S_ISREG.
1972 */
1973static inline bool file_write_started(const struct file *file)
1974{
1975 if (!S_ISREG(file_inode(file)->i_mode))
1976 return true;
1977 return sb_write_started(sb: file_inode(f: file)->i_sb);
1978}
1979
1980/**
1981 * file_write_not_started - check if SB_FREEZE_WRITE is not held
1982 * @file: the file we write to
1983 *
1984 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1985 * May be false positive with !S_ISREG, because file_start_write() has
1986 * no effect on !S_ISREG.
1987 */
1988static inline bool file_write_not_started(const struct file *file)
1989{
1990 if (!S_ISREG(file_inode(file)->i_mode))
1991 return true;
1992 return sb_write_not_started(sb: file_inode(f: file)->i_sb);
1993}
1994
1995/**
1996 * sb_end_write - drop write access to a superblock
1997 * @sb: the super we wrote to
1998 *
1999 * Decrement number of writers to the filesystem. Wake up possible waiters
2000 * wanting to freeze the filesystem.
2001 */
2002static inline void sb_end_write(struct super_block *sb)
2003{
2004 __sb_end_write(sb, level: SB_FREEZE_WRITE);
2005}
2006
2007/**
2008 * sb_end_pagefault - drop write access to a superblock from a page fault
2009 * @sb: the super we wrote to
2010 *
2011 * Decrement number of processes handling write page fault to the filesystem.
2012 * Wake up possible waiters wanting to freeze the filesystem.
2013 */
2014static inline void sb_end_pagefault(struct super_block *sb)
2015{
2016 __sb_end_write(sb, level: SB_FREEZE_PAGEFAULT);
2017}
2018
2019/**
2020 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
2021 * @sb: the super we wrote to
2022 *
2023 * Decrement fs-internal number of writers to the filesystem. Wake up possible
2024 * waiters wanting to freeze the filesystem.
2025 */
2026static inline void sb_end_intwrite(struct super_block *sb)
2027{
2028 __sb_end_write(sb, level: SB_FREEZE_FS);
2029}
2030
2031/**
2032 * sb_start_write - get write access to a superblock
2033 * @sb: the super we write to
2034 *
2035 * When a process wants to write data or metadata to a file system (i.e. dirty
2036 * a page or an inode), it should embed the operation in a sb_start_write() -
2037 * sb_end_write() pair to get exclusion against file system freezing. This
2038 * function increments number of writers preventing freezing. If the file
2039 * system is already frozen, the function waits until the file system is
2040 * thawed.
2041 *
2042 * Since freeze protection behaves as a lock, users have to preserve
2043 * ordering of freeze protection and other filesystem locks. Generally,
2044 * freeze protection should be the outermost lock. In particular, we have:
2045 *
2046 * sb_start_write
2047 * -> i_rwsem (write path, truncate, directory ops, ...)
2048 * -> s_umount (freeze_super, thaw_super)
2049 */
2050static inline void sb_start_write(struct super_block *sb)
2051{
2052 __sb_start_write(sb, level: SB_FREEZE_WRITE);
2053}
2054
2055static inline bool sb_start_write_trylock(struct super_block *sb)
2056{
2057 return __sb_start_write_trylock(sb, level: SB_FREEZE_WRITE);
2058}
2059
2060/**
2061 * sb_start_pagefault - get write access to a superblock from a page fault
2062 * @sb: the super we write to
2063 *
2064 * When a process starts handling write page fault, it should embed the
2065 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
2066 * exclusion against file system freezing. This is needed since the page fault
2067 * is going to dirty a page. This function increments number of running page
2068 * faults preventing freezing. If the file system is already frozen, the
2069 * function waits until the file system is thawed.
2070 *
2071 * Since page fault freeze protection behaves as a lock, users have to preserve
2072 * ordering of freeze protection and other filesystem locks. It is advised to
2073 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
2074 * handling code implies lock dependency:
2075 *
2076 * mmap_lock
2077 * -> sb_start_pagefault
2078 */
2079static inline void sb_start_pagefault(struct super_block *sb)
2080{
2081 __sb_start_write(sb, level: SB_FREEZE_PAGEFAULT);
2082}
2083
2084/**
2085 * sb_start_intwrite - get write access to a superblock for internal fs purposes
2086 * @sb: the super we write to
2087 *
2088 * This is the third level of protection against filesystem freezing. It is
2089 * free for use by a filesystem. The only requirement is that it must rank
2090 * below sb_start_pagefault.
2091 *
2092 * For example filesystem can call sb_start_intwrite() when starting a
2093 * transaction which somewhat eases handling of freezing for internal sources
2094 * of filesystem changes (internal fs threads, discarding preallocation on file
2095 * close, etc.).
2096 */
2097static inline void sb_start_intwrite(struct super_block *sb)
2098{
2099 __sb_start_write(sb, level: SB_FREEZE_FS);
2100}
2101
2102static inline bool sb_start_intwrite_trylock(struct super_block *sb)
2103{
2104 return __sb_start_write_trylock(sb, level: SB_FREEZE_FS);
2105}
2106
2107bool inode_owner_or_capable(struct mnt_idmap *idmap,
2108 const struct inode *inode);
2109
2110/*
2111 * VFS helper functions..
2112 */
2113int vfs_create(struct mnt_idmap *, struct inode *,
2114 struct dentry *, umode_t, bool);
2115struct dentry *vfs_mkdir(struct mnt_idmap *, struct inode *,
2116 struct dentry *, umode_t);
2117int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *,
2118 umode_t, dev_t);
2119int vfs_symlink(struct mnt_idmap *, struct inode *,
2120 struct dentry *, const char *);
2121int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *,
2122 struct dentry *, struct inode **);
2123int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *);
2124int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *,
2125 struct inode **);
2126
2127/**
2128 * struct renamedata - contains all information required for renaming
2129 * @mnt_idmap: idmap of the mount in which the rename is happening.
2130 * @old_parent: parent of source
2131 * @old_dentry: source
2132 * @new_parent: parent of destination
2133 * @new_dentry: destination
2134 * @delegated_inode: returns an inode needing a delegation break
2135 * @flags: rename flags
2136 */
2137struct renamedata {
2138 struct mnt_idmap *mnt_idmap;
2139 struct dentry *old_parent;
2140 struct dentry *old_dentry;
2141 struct dentry *new_parent;
2142 struct dentry *new_dentry;
2143 struct inode **delegated_inode;
2144 unsigned int flags;
2145} __randomize_layout;
2146
2147int vfs_rename(struct renamedata *);
2148
2149static inline int vfs_whiteout(struct mnt_idmap *idmap,
2150 struct inode *dir, struct dentry *dentry)
2151{
2152 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE,
2153 WHITEOUT_DEV);
2154}
2155
2156struct file *kernel_tmpfile_open(struct mnt_idmap *idmap,
2157 const struct path *parentpath,
2158 umode_t mode, int open_flag,
2159 const struct cred *cred);
2160struct file *kernel_file_open(const struct path *path, int flags,
2161 const struct cred *cred);
2162
2163int vfs_mkobj(struct dentry *, umode_t,
2164 int (*f)(struct dentry *, umode_t, void *),
2165 void *);
2166
2167int vfs_fchown(struct file *file, uid_t user, gid_t group);
2168int vfs_fchmod(struct file *file, umode_t mode);
2169int vfs_utimes(const struct path *path, struct timespec64 *times);
2170
2171#ifdef CONFIG_COMPAT
2172extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
2173 unsigned long arg);
2174#else
2175#define compat_ptr_ioctl NULL
2176#endif
2177
2178/*
2179 * VFS file helper functions.
2180 */
2181void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
2182 const struct inode *dir, umode_t mode);
2183extern bool may_open_dev(const struct path *path);
2184umode_t mode_strip_sgid(struct mnt_idmap *idmap,
2185 const struct inode *dir, umode_t mode);
2186bool in_group_or_capable(struct mnt_idmap *idmap,
2187 const struct inode *inode, vfsgid_t vfsgid);
2188
2189/*
2190 * This is the "filldir" function type, used by readdir() to let
2191 * the kernel specify what kind of dirent layout it wants to have.
2192 * This allows the kernel to read directories into kernel space or
2193 * to have different dirent layouts depending on the binary type.
2194 * Return 'true' to keep going and 'false' if there are no more entries.
2195 */
2196struct dir_context;
2197typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
2198 unsigned);
2199
2200struct dir_context {
2201 filldir_t actor;
2202 loff_t pos;
2203 /*
2204 * Filesystems MUST NOT MODIFY count, but may use as a hint:
2205 * 0 unknown
2206 * > 0 space in buffer (assume at least one entry)
2207 * INT_MAX unlimited
2208 */
2209 int count;
2210};
2211
2212/* If OR-ed with d_type, pending signals are not checked */
2213#define FILLDIR_FLAG_NOINTR 0x1000
2214
2215/*
2216 * These flags let !MMU mmap() govern direct device mapping vs immediate
2217 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
2218 *
2219 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE)
2220 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED)
2221 * NOMMU_MAP_READ: Can be mapped for reading
2222 * NOMMU_MAP_WRITE: Can be mapped for writing
2223 * NOMMU_MAP_EXEC: Can be mapped for execution
2224 */
2225#define NOMMU_MAP_COPY 0x00000001
2226#define NOMMU_MAP_DIRECT 0x00000008
2227#define NOMMU_MAP_READ VM_MAYREAD
2228#define NOMMU_MAP_WRITE VM_MAYWRITE
2229#define NOMMU_MAP_EXEC VM_MAYEXEC
2230
2231#define NOMMU_VMFLAGS \
2232 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
2233
2234/*
2235 * These flags control the behavior of the remap_file_range function pointer.
2236 * If it is called with len == 0 that means "remap to end of source file".
2237 * See Documentation/filesystems/vfs.rst for more details about this call.
2238 *
2239 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
2240 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
2241 */
2242#define REMAP_FILE_DEDUP (1 << 0)
2243#define REMAP_FILE_CAN_SHORTEN (1 << 1)
2244
2245/*
2246 * These flags signal that the caller is ok with altering various aspects of
2247 * the behavior of the remap operation. The changes must be made by the
2248 * implementation; the vfs remap helper functions can take advantage of them.
2249 * Flags in this category exist to preserve the quirky behavior of the hoisted
2250 * btrfs clone/dedupe ioctls.
2251 */
2252#define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN)
2253
2254/*
2255 * These flags control the behavior of vfs_copy_file_range().
2256 * They are not available to the user via syscall.
2257 *
2258 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops
2259 */
2260#define COPY_FILE_SPLICE (1 << 0)
2261
2262struct iov_iter;
2263struct io_uring_cmd;
2264struct offset_ctx;
2265
2266typedef unsigned int __bitwise fop_flags_t;
2267
2268struct file_operations {
2269 struct module *owner;
2270 fop_flags_t fop_flags;
2271 loff_t (*llseek) (struct file *, loff_t, int);
2272 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
2273 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
2274 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
2275 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
2276 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
2277 unsigned int flags);
2278 int (*iterate_shared) (struct file *, struct dir_context *);
2279 __poll_t (*poll) (struct file *, struct poll_table_struct *);
2280 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
2281 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
2282 int (*mmap) (struct file *, struct vm_area_struct *);
2283 int (*open) (struct inode *, struct file *);
2284 int (*flush) (struct file *, fl_owner_t id);
2285 int (*release) (struct inode *, struct file *);
2286 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
2287 int (*fasync) (int, struct file *, int);
2288 int (*lock) (struct file *, int, struct file_lock *);
2289 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2290 int (*check_flags)(int);
2291 int (*flock) (struct file *, int, struct file_lock *);
2292 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
2293 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
2294 void (*splice_eof)(struct file *file);
2295 int (*setlease)(struct file *, int, struct file_lease **, void **);
2296 long (*fallocate)(struct file *file, int mode, loff_t offset,
2297 loff_t len);
2298 void (*show_fdinfo)(struct seq_file *m, struct file *f);
2299#ifndef CONFIG_MMU
2300 unsigned (*mmap_capabilities)(struct file *);
2301#endif
2302 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
2303 loff_t, size_t, unsigned int);
2304 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
2305 struct file *file_out, loff_t pos_out,
2306 loff_t len, unsigned int remap_flags);
2307 int (*fadvise)(struct file *, loff_t, loff_t, int);
2308 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
2309 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *,
2310 unsigned int poll_flags);
2311 int (*mmap_prepare)(struct vm_area_desc *);
2312} __randomize_layout;
2313
2314/* Supports async buffered reads */
2315#define FOP_BUFFER_RASYNC ((__force fop_flags_t)(1 << 0))
2316/* Supports async buffered writes */
2317#define FOP_BUFFER_WASYNC ((__force fop_flags_t)(1 << 1))
2318/* Supports synchronous page faults for mappings */
2319#define FOP_MMAP_SYNC ((__force fop_flags_t)(1 << 2))
2320/* Supports non-exclusive O_DIRECT writes from multiple threads */
2321#define FOP_DIO_PARALLEL_WRITE ((__force fop_flags_t)(1 << 3))
2322/* Contains huge pages */
2323#define FOP_HUGE_PAGES ((__force fop_flags_t)(1 << 4))
2324/* Treat loff_t as unsigned (e.g., /dev/mem) */
2325#define FOP_UNSIGNED_OFFSET ((__force fop_flags_t)(1 << 5))
2326/* Supports asynchronous lock callbacks */
2327#define FOP_ASYNC_LOCK ((__force fop_flags_t)(1 << 6))
2328/* File system supports uncached read/write buffered IO */
2329#define FOP_DONTCACHE ((__force fop_flags_t)(1 << 7))
2330
2331/* Wrap a directory iterator that needs exclusive inode access */
2332int wrap_directory_iterator(struct file *, struct dir_context *,
2333 int (*) (struct file *, struct dir_context *));
2334#define WRAP_DIR_ITER(x) \
2335 static int shared_##x(struct file *file , struct dir_context *ctx) \
2336 { return wrap_directory_iterator(file, ctx, x); }
2337
2338struct inode_operations {
2339 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
2340 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
2341 int (*permission) (struct mnt_idmap *, struct inode *, int);
2342 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
2343
2344 int (*readlink) (struct dentry *, char __user *,int);
2345
2346 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *,
2347 umode_t, bool);
2348 int (*link) (struct dentry *,struct inode *,struct dentry *);
2349 int (*unlink) (struct inode *,struct dentry *);
2350 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,
2351 const char *);
2352 struct dentry *(*mkdir) (struct mnt_idmap *, struct inode *,
2353 struct dentry *, umode_t);
2354 int (*rmdir) (struct inode *,struct dentry *);
2355 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,
2356 umode_t,dev_t);
2357 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
2358 struct inode *, struct dentry *, unsigned int);
2359 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
2360 int (*getattr) (struct mnt_idmap *, const struct path *,
2361 struct kstat *, u32, unsigned int);
2362 ssize_t (*listxattr) (struct dentry *, char *, size_t);
2363 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
2364 u64 len);
2365 int (*update_time)(struct inode *, int);
2366 int (*atomic_open)(struct inode *, struct dentry *,
2367 struct file *, unsigned open_flag,
2368 umode_t create_mode);
2369 int (*tmpfile) (struct mnt_idmap *, struct inode *,
2370 struct file *, umode_t);
2371 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *,
2372 int);
2373 int (*set_acl)(struct mnt_idmap *, struct dentry *,
2374 struct posix_acl *, int);
2375 int (*fileattr_set)(struct mnt_idmap *idmap,
2376 struct dentry *dentry, struct file_kattr *fa);
2377 int (*fileattr_get)(struct dentry *dentry, struct file_kattr *fa);
2378 struct offset_ctx *(*get_offset_ctx)(struct inode *inode);
2379} ____cacheline_aligned;
2380
2381/* Did the driver provide valid mmap hook configuration? */
2382static inline bool can_mmap_file(struct file *file)
2383{
2384 bool has_mmap = file->f_op->mmap;
2385 bool has_mmap_prepare = file->f_op->mmap_prepare;
2386
2387 /* Hooks are mutually exclusive. */
2388 if (WARN_ON_ONCE(has_mmap && has_mmap_prepare))
2389 return false;
2390 if (!has_mmap && !has_mmap_prepare)
2391 return false;
2392
2393 return true;
2394}
2395
2396int __compat_vma_mmap_prepare(const struct file_operations *f_op,
2397 struct file *file, struct vm_area_struct *vma);
2398int compat_vma_mmap_prepare(struct file *file, struct vm_area_struct *vma);
2399
2400static inline int vfs_mmap(struct file *file, struct vm_area_struct *vma)
2401{
2402 if (file->f_op->mmap_prepare)
2403 return compat_vma_mmap_prepare(file, vma);
2404
2405 return file->f_op->mmap(file, vma);
2406}
2407
2408static inline int vfs_mmap_prepare(struct file *file, struct vm_area_desc *desc)
2409{
2410 return file->f_op->mmap_prepare(desc);
2411}
2412
2413extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
2414extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
2415extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
2416 loff_t, size_t, unsigned int);
2417int remap_verify_area(struct file *file, loff_t pos, loff_t len, bool write);
2418int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2419 struct file *file_out, loff_t pos_out,
2420 loff_t *len, unsigned int remap_flags,
2421 const struct iomap_ops *dax_read_ops);
2422int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2423 struct file *file_out, loff_t pos_out,
2424 loff_t *count, unsigned int remap_flags);
2425extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
2426 struct file *file_out, loff_t pos_out,
2427 loff_t len, unsigned int remap_flags);
2428extern int vfs_dedupe_file_range(struct file *file,
2429 struct file_dedupe_range *same);
2430extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
2431 struct file *dst_file, loff_t dst_pos,
2432 loff_t len, unsigned int remap_flags);
2433
2434/**
2435 * enum freeze_holder - holder of the freeze
2436 * @FREEZE_HOLDER_KERNEL: kernel wants to freeze or thaw filesystem
2437 * @FREEZE_HOLDER_USERSPACE: userspace wants to freeze or thaw filesystem
2438 * @FREEZE_MAY_NEST: whether nesting freeze and thaw requests is allowed
2439 * @FREEZE_EXCL: a freeze that can only be undone by the owner
2440 *
2441 * Indicate who the owner of the freeze or thaw request is and whether
2442 * the freeze needs to be exclusive or can nest.
2443 * Without @FREEZE_MAY_NEST, multiple freeze and thaw requests from the
2444 * same holder aren't allowed. It is however allowed to hold a single
2445 * @FREEZE_HOLDER_USERSPACE and a single @FREEZE_HOLDER_KERNEL freeze at
2446 * the same time. This is relied upon by some filesystems during online
2447 * repair or similar.
2448 */
2449enum freeze_holder {
2450 FREEZE_HOLDER_KERNEL = (1U << 0),
2451 FREEZE_HOLDER_USERSPACE = (1U << 1),
2452 FREEZE_MAY_NEST = (1U << 2),
2453 FREEZE_EXCL = (1U << 3),
2454};
2455
2456struct super_operations {
2457 struct inode *(*alloc_inode)(struct super_block *sb);
2458 void (*destroy_inode)(struct inode *);
2459 void (*free_inode)(struct inode *);
2460
2461 void (*dirty_inode) (struct inode *, int flags);
2462 int (*write_inode) (struct inode *, struct writeback_control *wbc);
2463 int (*drop_inode) (struct inode *);
2464 void (*evict_inode) (struct inode *);
2465 void (*put_super) (struct super_block *);
2466 int (*sync_fs)(struct super_block *sb, int wait);
2467 int (*freeze_super) (struct super_block *, enum freeze_holder who, const void *owner);
2468 int (*freeze_fs) (struct super_block *);
2469 int (*thaw_super) (struct super_block *, enum freeze_holder who, const void *owner);
2470 int (*unfreeze_fs) (struct super_block *);
2471 int (*statfs) (struct dentry *, struct kstatfs *);
2472 int (*remount_fs) (struct super_block *, int *, char *);
2473 void (*umount_begin) (struct super_block *);
2474
2475 int (*show_options)(struct seq_file *, struct dentry *);
2476 int (*show_devname)(struct seq_file *, struct dentry *);
2477 int (*show_path)(struct seq_file *, struct dentry *);
2478 int (*show_stats)(struct seq_file *, struct dentry *);
2479#ifdef CONFIG_QUOTA
2480 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
2481 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
2482 struct dquot __rcu **(*get_dquots)(struct inode *);
2483#endif
2484 long (*nr_cached_objects)(struct super_block *,
2485 struct shrink_control *);
2486 long (*free_cached_objects)(struct super_block *,
2487 struct shrink_control *);
2488 /*
2489 * If a filesystem can support graceful removal of a device and
2490 * continue read-write operations, implement this callback.
2491 *
2492 * Return 0 if the filesystem can continue read-write.
2493 * Non-zero return value or no such callback means the fs will be shutdown
2494 * as usual.
2495 */
2496 int (*remove_bdev)(struct super_block *sb, struct block_device *bdev);
2497 void (*shutdown)(struct super_block *sb);
2498};
2499
2500/*
2501 * Inode flags - they have no relation to superblock flags now
2502 */
2503#define S_SYNC (1 << 0) /* Writes are synced at once */
2504#define S_NOATIME (1 << 1) /* Do not update access times */
2505#define S_APPEND (1 << 2) /* Append-only file */
2506#define S_IMMUTABLE (1 << 3) /* Immutable file */
2507#define S_DEAD (1 << 4) /* removed, but still open directory */
2508#define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */
2509#define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */
2510#define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */
2511#define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */
2512#define S_PRIVATE (1 << 9) /* Inode is fs-internal */
2513#define S_IMA (1 << 10) /* Inode has an associated IMA struct */
2514#define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */
2515#define S_NOSEC (1 << 12) /* no suid or xattr security attributes */
2516#ifdef CONFIG_FS_DAX
2517#define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */
2518#else
2519#define S_DAX 0 /* Make all the DAX code disappear */
2520#endif
2521#define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */
2522#define S_CASEFOLD (1 << 15) /* Casefolded file */
2523#define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */
2524#define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
2525#define S_ANON_INODE (1 << 19) /* Inode is an anonymous inode */
2526
2527/*
2528 * Note that nosuid etc flags are inode-specific: setting some file-system
2529 * flags just means all the inodes inherit those flags by default. It might be
2530 * possible to override it selectively if you really wanted to with some
2531 * ioctl() that is not currently implemented.
2532 *
2533 * Exception: SB_RDONLY is always applied to the entire file system.
2534 *
2535 * Unfortunately, it is possible to change a filesystems flags with it mounted
2536 * with files in use. This means that all of the inodes will not have their
2537 * i_flags updated. Hence, i_flags no longer inherit the superblock mount
2538 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
2539 */
2540#define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
2541
2542static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
2543#define IS_RDONLY(inode) sb_rdonly((inode)->i_sb)
2544#define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \
2545 ((inode)->i_flags & S_SYNC))
2546#define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
2547 ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
2548#define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK)
2549#define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME)
2550#define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION)
2551
2552#define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
2553#define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
2554#define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
2555
2556#ifdef CONFIG_FS_POSIX_ACL
2557#define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL)
2558#else
2559#define IS_POSIXACL(inode) 0
2560#endif
2561
2562#define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
2563#define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
2564
2565#ifdef CONFIG_SWAP
2566#define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
2567#else
2568#define IS_SWAPFILE(inode) ((void)(inode), 0U)
2569#endif
2570
2571#define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
2572#define IS_IMA(inode) ((inode)->i_flags & S_IMA)
2573#define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
2574#define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
2575#define IS_DAX(inode) ((inode)->i_flags & S_DAX)
2576#define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED)
2577#define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD)
2578#define IS_VERITY(inode) ((inode)->i_flags & S_VERITY)
2579
2580#define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
2581 (inode)->i_rdev == WHITEOUT_DEV)
2582#define IS_ANON_FILE(inode) ((inode)->i_flags & S_ANON_INODE)
2583
2584static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap,
2585 struct inode *inode)
2586{
2587 return !vfsuid_valid(uid: i_uid_into_vfsuid(idmap, inode)) ||
2588 !vfsgid_valid(gid: i_gid_into_vfsgid(idmap, inode));
2589}
2590
2591static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
2592{
2593 *kiocb = (struct kiocb) {
2594 .ki_filp = filp,
2595 .ki_flags = filp->f_iocb_flags,
2596 .ki_ioprio = get_current_ioprio(),
2597 };
2598}
2599
2600static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2601 struct file *filp)
2602{
2603 *kiocb = (struct kiocb) {
2604 .ki_filp = filp,
2605 .ki_flags = kiocb_src->ki_flags,
2606 .ki_ioprio = kiocb_src->ki_ioprio,
2607 .ki_pos = kiocb_src->ki_pos,
2608 };
2609}
2610
2611extern void __mark_inode_dirty(struct inode *, int);
2612static inline void mark_inode_dirty(struct inode *inode)
2613{
2614 __mark_inode_dirty(inode, I_DIRTY);
2615}
2616
2617static inline void mark_inode_dirty_sync(struct inode *inode)
2618{
2619 __mark_inode_dirty(inode, I_DIRTY_SYNC);
2620}
2621
2622static inline int icount_read(const struct inode *inode)
2623{
2624 return atomic_read(v: &inode->i_count);
2625}
2626
2627/*
2628 * Returns true if the given inode itself only has dirty timestamps (its pages
2629 * may still be dirty) and isn't currently being allocated or freed.
2630 * Filesystems should call this if when writing an inode when lazytime is
2631 * enabled, they want to opportunistically write the timestamps of other inodes
2632 * located very nearby on-disk, e.g. in the same inode block. This returns true
2633 * if the given inode is in need of such an opportunistic update. Requires
2634 * i_lock, or at least later re-checking under i_lock.
2635 */
2636static inline bool inode_is_dirtytime_only(struct inode *inode)
2637{
2638 return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2639 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2640}
2641
2642extern void inc_nlink(struct inode *inode);
2643extern void drop_nlink(struct inode *inode);
2644extern void clear_nlink(struct inode *inode);
2645extern void set_nlink(struct inode *inode, unsigned int nlink);
2646
2647static inline void inode_inc_link_count(struct inode *inode)
2648{
2649 inc_nlink(inode);
2650 mark_inode_dirty(inode);
2651}
2652
2653static inline void inode_dec_link_count(struct inode *inode)
2654{
2655 drop_nlink(inode);
2656 mark_inode_dirty(inode);
2657}
2658
2659enum file_time_flags {
2660 S_ATIME = 1,
2661 S_MTIME = 2,
2662 S_CTIME = 4,
2663 S_VERSION = 8,
2664};
2665
2666extern bool atime_needs_update(const struct path *, struct inode *);
2667extern void touch_atime(const struct path *);
2668int inode_update_time(struct inode *inode, int flags);
2669
2670static inline void file_accessed(struct file *file)
2671{
2672 if (!(file->f_flags & O_NOATIME))
2673 touch_atime(&file->f_path);
2674}
2675
2676extern int file_modified(struct file *file);
2677int kiocb_modified(struct kiocb *iocb);
2678
2679int sync_inode_metadata(struct inode *inode, int wait);
2680
2681struct file_system_type {
2682 const char *name;
2683 int fs_flags;
2684#define FS_REQUIRES_DEV 1
2685#define FS_BINARY_MOUNTDATA 2
2686#define FS_HAS_SUBTYPE 4
2687#define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
2688#define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */
2689#define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */
2690#define FS_MGTIME 64 /* FS uses multigrain timestamps */
2691#define FS_LBS 128 /* FS supports LBS */
2692#define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
2693 int (*init_fs_context)(struct fs_context *);
2694 const struct fs_parameter_spec *parameters;
2695 struct dentry *(*mount) (struct file_system_type *, int,
2696 const char *, void *);
2697 void (*kill_sb) (struct super_block *);
2698 struct module *owner;
2699 struct file_system_type * next;
2700 struct hlist_head fs_supers;
2701
2702 struct lock_class_key s_lock_key;
2703 struct lock_class_key s_umount_key;
2704 struct lock_class_key s_vfs_rename_key;
2705 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2706
2707 struct lock_class_key i_lock_key;
2708 struct lock_class_key i_mutex_key;
2709 struct lock_class_key invalidate_lock_key;
2710 struct lock_class_key i_mutex_dir_key;
2711};
2712
2713#define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2714
2715/**
2716 * is_mgtime: is this inode using multigrain timestamps
2717 * @inode: inode to test for multigrain timestamps
2718 *
2719 * Return true if the inode uses multigrain timestamps, false otherwise.
2720 */
2721static inline bool is_mgtime(const struct inode *inode)
2722{
2723 return inode->i_opflags & IOP_MGTIME;
2724}
2725
2726extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2727void retire_super(struct super_block *sb);
2728void generic_shutdown_super(struct super_block *sb);
2729void kill_block_super(struct super_block *sb);
2730void kill_anon_super(struct super_block *sb);
2731void kill_litter_super(struct super_block *sb);
2732void deactivate_super(struct super_block *sb);
2733void deactivate_locked_super(struct super_block *sb);
2734int set_anon_super(struct super_block *s, void *data);
2735int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2736int get_anon_bdev(dev_t *);
2737void free_anon_bdev(dev_t);
2738struct super_block *sget_fc(struct fs_context *fc,
2739 int (*test)(struct super_block *, struct fs_context *),
2740 int (*set)(struct super_block *, struct fs_context *));
2741struct super_block *sget(struct file_system_type *type,
2742 int (*test)(struct super_block *,void *),
2743 int (*set)(struct super_block *,void *),
2744 int flags, void *data);
2745struct super_block *sget_dev(struct fs_context *fc, dev_t dev);
2746
2747/* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2748#define fops_get(fops) ({ \
2749 const struct file_operations *_fops = (fops); \
2750 (((_fops) && try_module_get((_fops)->owner) ? (_fops) : NULL)); \
2751})
2752
2753#define fops_put(fops) ({ \
2754 const struct file_operations *_fops = (fops); \
2755 if (_fops) \
2756 module_put((_fops)->owner); \
2757})
2758
2759/*
2760 * This one is to be used *ONLY* from ->open() instances.
2761 * fops must be non-NULL, pinned down *and* module dependencies
2762 * should be sufficient to pin the caller down as well.
2763 */
2764#define replace_fops(f, fops) \
2765 do { \
2766 struct file *__file = (f); \
2767 fops_put(__file->f_op); \
2768 BUG_ON(!(__file->f_op = (fops))); \
2769 } while(0)
2770
2771extern int register_filesystem(struct file_system_type *);
2772extern int unregister_filesystem(struct file_system_type *);
2773extern int vfs_statfs(const struct path *, struct kstatfs *);
2774extern int user_statfs(const char __user *, struct kstatfs *);
2775extern int fd_statfs(int, struct kstatfs *);
2776int freeze_super(struct super_block *super, enum freeze_holder who,
2777 const void *freeze_owner);
2778int thaw_super(struct super_block *super, enum freeze_holder who,
2779 const void *freeze_owner);
2780extern __printf(2, 3)
2781int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2782extern int super_setup_bdi(struct super_block *sb);
2783
2784static inline void super_set_uuid(struct super_block *sb, const u8 *uuid, unsigned len)
2785{
2786 if (WARN_ON(len > sizeof(sb->s_uuid)))
2787 len = sizeof(sb->s_uuid);
2788 sb->s_uuid_len = len;
2789 memcpy(to: &sb->s_uuid, from: uuid, len);
2790}
2791
2792/* set sb sysfs name based on sb->s_bdev */
2793static inline void super_set_sysfs_name_bdev(struct super_block *sb)
2794{
2795 snprintf(buf: sb->s_sysfs_name, size: sizeof(sb->s_sysfs_name), fmt: "%pg", sb->s_bdev);
2796}
2797
2798/* set sb sysfs name based on sb->s_uuid */
2799static inline void super_set_sysfs_name_uuid(struct super_block *sb)
2800{
2801 WARN_ON(sb->s_uuid_len != sizeof(sb->s_uuid));
2802 snprintf(buf: sb->s_sysfs_name, size: sizeof(sb->s_sysfs_name), fmt: "%pU", sb->s_uuid.b);
2803}
2804
2805/* set sb sysfs name based on sb->s_id */
2806static inline void super_set_sysfs_name_id(struct super_block *sb)
2807{
2808 strscpy(sb->s_sysfs_name, sb->s_id, sizeof(sb->s_sysfs_name));
2809}
2810
2811/* try to use something standard before you use this */
2812__printf(2, 3)
2813static inline void super_set_sysfs_name_generic(struct super_block *sb, const char *fmt, ...)
2814{
2815 va_list args;
2816
2817 va_start(args, fmt);
2818 vsnprintf(buf: sb->s_sysfs_name, size: sizeof(sb->s_sysfs_name), fmt, args);
2819 va_end(args);
2820}
2821
2822extern int current_umask(void);
2823
2824extern void ihold(struct inode * inode);
2825extern void iput(struct inode *);
2826int inode_update_timestamps(struct inode *inode, int flags);
2827int generic_update_time(struct inode *, int);
2828
2829/* /sys/fs */
2830extern struct kobject *fs_kobj;
2831
2832#define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2833
2834/* fs/open.c */
2835struct audit_names;
2836struct filename {
2837 const char *name; /* pointer to actual string */
2838 const __user char *uptr; /* original userland pointer */
2839 atomic_t refcnt;
2840 struct audit_names *aname;
2841 const char iname[];
2842};
2843static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2844
2845static inline struct mnt_idmap *file_mnt_idmap(const struct file *file)
2846{
2847 return mnt_idmap(mnt: file->f_path.mnt);
2848}
2849
2850/**
2851 * is_idmapped_mnt - check whether a mount is mapped
2852 * @mnt: the mount to check
2853 *
2854 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped.
2855 *
2856 * Return: true if mount is mapped, false if not.
2857 */
2858static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2859{
2860 return mnt_idmap(mnt) != &nop_mnt_idmap;
2861}
2862
2863int vfs_truncate(const struct path *, loff_t);
2864int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start,
2865 unsigned int time_attrs, struct file *filp);
2866extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2867 loff_t len);
2868int do_sys_open(int dfd, const char __user *filename, int flags,
2869 umode_t mode);
2870extern struct file *file_open_name(struct filename *, int, umode_t);
2871extern struct file *filp_open(const char *, int, umode_t);
2872extern struct file *file_open_root(const struct path *,
2873 const char *, int, umode_t);
2874static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2875 const char *name, int flags, umode_t mode)
2876{
2877 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2878 name, flags, mode);
2879}
2880struct file *dentry_open(const struct path *path, int flags,
2881 const struct cred *creds);
2882struct file *dentry_open_nonotify(const struct path *path, int flags,
2883 const struct cred *cred);
2884struct file *dentry_create(const struct path *path, int flags, umode_t mode,
2885 const struct cred *cred);
2886const struct path *backing_file_user_path(const struct file *f);
2887
2888/*
2889 * When mmapping a file on a stackable filesystem (e.g., overlayfs), the file
2890 * stored in ->vm_file is a backing file whose f_inode is on the underlying
2891 * filesystem. When the mapped file path and inode number are displayed to
2892 * user (e.g. via /proc/<pid>/maps), these helpers should be used to get the
2893 * path and inode number to display to the user, which is the path of the fd
2894 * that user has requested to map and the inode number that would be returned
2895 * by fstat() on that same fd.
2896 */
2897/* Get the path to display in /proc/<pid>/maps */
2898static inline const struct path *file_user_path(const struct file *f)
2899{
2900 if (unlikely(f->f_mode & FMODE_BACKING))
2901 return backing_file_user_path(f);
2902 return &f->f_path;
2903}
2904/* Get the inode whose inode number to display in /proc/<pid>/maps */
2905static inline const struct inode *file_user_inode(const struct file *f)
2906{
2907 if (unlikely(f->f_mode & FMODE_BACKING))
2908 return d_inode(dentry: backing_file_user_path(f)->dentry);
2909 return file_inode(f);
2910}
2911
2912static inline struct file *file_clone_open(struct file *file)
2913{
2914 return dentry_open(path: &file->f_path, flags: file->f_flags, creds: file->f_cred);
2915}
2916extern int filp_close(struct file *, fl_owner_t id);
2917
2918extern struct filename *getname_flags(const char __user *, int);
2919extern struct filename *getname_uflags(const char __user *, int);
2920static inline struct filename *getname(const char __user *name)
2921{
2922 return getname_flags(name, 0);
2923}
2924extern struct filename *getname_kernel(const char *);
2925extern struct filename *__getname_maybe_null(const char __user *);
2926static inline struct filename *getname_maybe_null(const char __user *name, int flags)
2927{
2928 if (!(flags & AT_EMPTY_PATH))
2929 return getname(name);
2930
2931 if (!name)
2932 return NULL;
2933 return __getname_maybe_null(name);
2934}
2935extern void putname(struct filename *name);
2936DEFINE_FREE(putname, struct filename *, if (!IS_ERR_OR_NULL(_T)) putname(_T))
2937
2938static inline struct filename *refname(struct filename *name)
2939{
2940 atomic_inc(v: &name->refcnt);
2941 return name;
2942}
2943
2944extern int finish_open(struct file *file, struct dentry *dentry,
2945 int (*open)(struct inode *, struct file *));
2946extern int finish_no_open(struct file *file, struct dentry *dentry);
2947
2948/* Helper for the simple case when original dentry is used */
2949static inline int finish_open_simple(struct file *file, int error)
2950{
2951 if (error)
2952 return error;
2953
2954 return finish_open(file, dentry: file->f_path.dentry, NULL);
2955}
2956
2957/* fs/dcache.c */
2958extern void __init vfs_caches_init_early(void);
2959extern void __init vfs_caches_init(void);
2960
2961extern struct kmem_cache *names_cachep;
2962
2963#define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2964#define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2965
2966extern struct super_block *blockdev_superblock;
2967static inline bool sb_is_blkdev_sb(struct super_block *sb)
2968{
2969 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2970}
2971
2972void emergency_thaw_all(void);
2973extern int sync_filesystem(struct super_block *);
2974extern const struct file_operations def_blk_fops;
2975extern const struct file_operations def_chr_fops;
2976
2977/* fs/char_dev.c */
2978#define CHRDEV_MAJOR_MAX 512
2979/* Marks the bottom of the first segment of free char majors */
2980#define CHRDEV_MAJOR_DYN_END 234
2981/* Marks the top and bottom of the second segment of free char majors */
2982#define CHRDEV_MAJOR_DYN_EXT_START 511
2983#define CHRDEV_MAJOR_DYN_EXT_END 384
2984
2985extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2986extern int register_chrdev_region(dev_t, unsigned, const char *);
2987extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2988 unsigned int count, const char *name,
2989 const struct file_operations *fops);
2990extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2991 unsigned int count, const char *name);
2992extern void unregister_chrdev_region(dev_t, unsigned);
2993extern void chrdev_show(struct seq_file *,off_t);
2994
2995static inline int register_chrdev(unsigned int major, const char *name,
2996 const struct file_operations *fops)
2997{
2998 return __register_chrdev(major, baseminor: 0, count: 256, name, fops);
2999}
3000
3001static inline void unregister_chrdev(unsigned int major, const char *name)
3002{
3003 __unregister_chrdev(major, baseminor: 0, count: 256, name);
3004}
3005
3006extern void init_special_inode(struct inode *, umode_t, dev_t);
3007
3008/* Invalid inode operations -- fs/bad_inode.c */
3009extern void make_bad_inode(struct inode *);
3010extern bool is_bad_inode(struct inode *);
3011
3012extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
3013 loff_t lend);
3014extern int __must_check file_check_and_advance_wb_err(struct file *file);
3015extern int __must_check file_write_and_wait_range(struct file *file,
3016 loff_t start, loff_t end);
3017int filemap_fdatawrite_range_kick(struct address_space *mapping, loff_t start,
3018 loff_t end);
3019
3020static inline int file_write_and_wait(struct file *file)
3021{
3022 return file_write_and_wait_range(file, start: 0, LLONG_MAX);
3023}
3024
3025extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
3026 int datasync);
3027extern int vfs_fsync(struct file *file, int datasync);
3028
3029extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
3030 unsigned int flags);
3031
3032static inline bool iocb_is_dsync(const struct kiocb *iocb)
3033{
3034 return (iocb->ki_flags & IOCB_DSYNC) ||
3035 IS_SYNC(iocb->ki_filp->f_mapping->host);
3036}
3037
3038/*
3039 * Sync the bytes written if this was a synchronous write. Expect ki_pos
3040 * to already be updated for the write, and will return either the amount
3041 * of bytes passed in, or an error if syncing the file failed.
3042 */
3043static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
3044{
3045 if (iocb_is_dsync(iocb)) {
3046 int ret = vfs_fsync_range(file: iocb->ki_filp,
3047 start: iocb->ki_pos - count, end: iocb->ki_pos - 1,
3048 datasync: (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
3049 if (ret)
3050 return ret;
3051 } else if (iocb->ki_flags & IOCB_DONTCACHE) {
3052 struct address_space *mapping = iocb->ki_filp->f_mapping;
3053
3054 filemap_fdatawrite_range_kick(mapping, start: iocb->ki_pos - count,
3055 end: iocb->ki_pos - 1);
3056 }
3057
3058 return count;
3059}
3060
3061extern void emergency_sync(void);
3062extern void emergency_remount(void);
3063
3064#ifdef CONFIG_BLOCK
3065extern int bmap(struct inode *inode, sector_t *block);
3066#else
3067static inline int bmap(struct inode *inode, sector_t *block)
3068{
3069 return -EINVAL;
3070}
3071#endif
3072
3073int notify_change(struct mnt_idmap *, struct dentry *,
3074 struct iattr *, struct inode **);
3075int inode_permission(struct mnt_idmap *, struct inode *, int);
3076int generic_permission(struct mnt_idmap *, struct inode *, int);
3077static inline int file_permission(struct file *file, int mask)
3078{
3079 return inode_permission(file_mnt_idmap(file),
3080 file_inode(f: file), mask);
3081}
3082static inline int path_permission(const struct path *path, int mask)
3083{
3084 return inode_permission(mnt_idmap(mnt: path->mnt),
3085 d_inode(dentry: path->dentry), mask);
3086}
3087int __check_sticky(struct mnt_idmap *idmap, struct inode *dir,
3088 struct inode *inode);
3089
3090static inline bool execute_ok(struct inode *inode)
3091{
3092 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
3093}
3094
3095static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
3096{
3097 return (inode->i_mode ^ mode) & S_IFMT;
3098}
3099
3100/**
3101 * file_start_write - get write access to a superblock for regular file io
3102 * @file: the file we want to write to
3103 *
3104 * This is a variant of sb_start_write() which is a noop on non-regualr file.
3105 * Should be matched with a call to file_end_write().
3106 */
3107static inline void file_start_write(struct file *file)
3108{
3109 if (!S_ISREG(file_inode(file)->i_mode))
3110 return;
3111 sb_start_write(sb: file_inode(f: file)->i_sb);
3112}
3113
3114static inline bool file_start_write_trylock(struct file *file)
3115{
3116 if (!S_ISREG(file_inode(file)->i_mode))
3117 return true;
3118 return sb_start_write_trylock(sb: file_inode(f: file)->i_sb);
3119}
3120
3121/**
3122 * file_end_write - drop write access to a superblock of a regular file
3123 * @file: the file we wrote to
3124 *
3125 * Should be matched with a call to file_start_write().
3126 */
3127static inline void file_end_write(struct file *file)
3128{
3129 if (!S_ISREG(file_inode(file)->i_mode))
3130 return;
3131 sb_end_write(sb: file_inode(f: file)->i_sb);
3132}
3133
3134/**
3135 * kiocb_start_write - get write access to a superblock for async file io
3136 * @iocb: the io context we want to submit the write with
3137 *
3138 * This is a variant of sb_start_write() for async io submission.
3139 * Should be matched with a call to kiocb_end_write().
3140 */
3141static inline void kiocb_start_write(struct kiocb *iocb)
3142{
3143 struct inode *inode = file_inode(f: iocb->ki_filp);
3144
3145 sb_start_write(sb: inode->i_sb);
3146 /*
3147 * Fool lockdep by telling it the lock got released so that it
3148 * doesn't complain about the held lock when we return to userspace.
3149 */
3150 __sb_writers_release(inode->i_sb, SB_FREEZE_WRITE);
3151}
3152
3153/**
3154 * kiocb_end_write - drop write access to a superblock after async file io
3155 * @iocb: the io context we sumbitted the write with
3156 *
3157 * Should be matched with a call to kiocb_start_write().
3158 */
3159static inline void kiocb_end_write(struct kiocb *iocb)
3160{
3161 struct inode *inode = file_inode(f: iocb->ki_filp);
3162
3163 /*
3164 * Tell lockdep we inherited freeze protection from submission thread.
3165 */
3166 __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
3167 sb_end_write(sb: inode->i_sb);
3168}
3169
3170/*
3171 * This is used for regular files where some users -- especially the
3172 * currently executed binary in a process, previously handled via
3173 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
3174 * read-write shared) accesses.
3175 *
3176 * get_write_access() gets write permission for a file.
3177 * put_write_access() releases this write permission.
3178 * deny_write_access() denies write access to a file.
3179 * allow_write_access() re-enables write access to a file.
3180 *
3181 * The i_writecount field of an inode can have the following values:
3182 * 0: no write access, no denied write access
3183 * < 0: (-i_writecount) users that denied write access to the file.
3184 * > 0: (i_writecount) users that have write access to the file.
3185 *
3186 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
3187 * except for the cases where we don't hold i_writecount yet. Then we need to
3188 * use {get,deny}_write_access() - these functions check the sign and refuse
3189 * to do the change if sign is wrong.
3190 */
3191static inline int get_write_access(struct inode *inode)
3192{
3193 return atomic_inc_unless_negative(v: &inode->i_writecount) ? 0 : -ETXTBSY;
3194}
3195static inline int deny_write_access(struct file *file)
3196{
3197 struct inode *inode = file_inode(f: file);
3198 return atomic_dec_unless_positive(v: &inode->i_writecount) ? 0 : -ETXTBSY;
3199}
3200static inline void put_write_access(struct inode * inode)
3201{
3202 atomic_dec(v: &inode->i_writecount);
3203}
3204static inline void allow_write_access(struct file *file)
3205{
3206 if (file)
3207 atomic_inc(v: &file_inode(f: file)->i_writecount);
3208}
3209
3210/*
3211 * Do not prevent write to executable file when watched by pre-content events.
3212 *
3213 * Note that FMODE_FSNOTIFY_HSM mode is set depending on pre-content watches at
3214 * the time of file open and remains constant for entire lifetime of the file,
3215 * so if pre-content watches are added post execution or removed before the end
3216 * of the execution, it will not cause i_writecount reference leak.
3217 */
3218static inline int exe_file_deny_write_access(struct file *exe_file)
3219{
3220 if (unlikely(FMODE_FSNOTIFY_HSM(exe_file->f_mode)))
3221 return 0;
3222 return deny_write_access(file: exe_file);
3223}
3224static inline void exe_file_allow_write_access(struct file *exe_file)
3225{
3226 if (unlikely(!exe_file || FMODE_FSNOTIFY_HSM(exe_file->f_mode)))
3227 return;
3228 allow_write_access(file: exe_file);
3229}
3230
3231static inline void file_set_fsnotify_mode(struct file *file, fmode_t mode)
3232{
3233 file->f_mode &= ~FMODE_FSNOTIFY_MASK;
3234 file->f_mode |= mode;
3235}
3236
3237static inline bool inode_is_open_for_write(const struct inode *inode)
3238{
3239 return atomic_read(v: &inode->i_writecount) > 0;
3240}
3241
3242#if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
3243static inline void i_readcount_dec(struct inode *inode)
3244{
3245 BUG_ON(atomic_dec_return(&inode->i_readcount) < 0);
3246}
3247static inline void i_readcount_inc(struct inode *inode)
3248{
3249 atomic_inc(v: &inode->i_readcount);
3250}
3251#else
3252static inline void i_readcount_dec(struct inode *inode)
3253{
3254 return;
3255}
3256static inline void i_readcount_inc(struct inode *inode)
3257{
3258 return;
3259}
3260#endif
3261extern int do_pipe_flags(int *, int);
3262
3263extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
3264ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
3265extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
3266extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
3267extern struct file * open_exec(const char *);
3268
3269/* fs/dcache.c -- generic fs support functions */
3270extern bool is_subdir(struct dentry *, struct dentry *);
3271extern bool path_is_under(const struct path *, const struct path *);
3272
3273extern char *file_path(struct file *, char *, int);
3274
3275/**
3276 * is_dot_dotdot - returns true only if @name is "." or ".."
3277 * @name: file name to check
3278 * @len: length of file name, in bytes
3279 */
3280static inline bool is_dot_dotdot(const char *name, size_t len)
3281{
3282 return len && unlikely(name[0] == '.') &&
3283 (len == 1 || (len == 2 && name[1] == '.'));
3284}
3285
3286/**
3287 * name_contains_dotdot - check if a file name contains ".." path components
3288 * @name: File path string to check
3289 * Search for ".." surrounded by either '/' or start/end of string.
3290 */
3291static inline bool name_contains_dotdot(const char *name)
3292{
3293 size_t name_len;
3294
3295 name_len = strlen(name);
3296 return strcmp(name, "..") == 0 ||
3297 strncmp(name, "../", 3) == 0 ||
3298 strstr(name, "/../") != NULL ||
3299 (name_len >= 3 && strcmp(name + name_len - 3, "/..") == 0);
3300}
3301
3302#include <linux/err.h>
3303
3304/* needed for stackable file system support */
3305extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
3306
3307extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
3308
3309extern int inode_init_always_gfp(struct super_block *, struct inode *, gfp_t);
3310static inline int inode_init_always(struct super_block *sb, struct inode *inode)
3311{
3312 return inode_init_always_gfp(sb, inode, GFP_NOFS);
3313}
3314
3315extern void inode_init_once(struct inode *);
3316extern void address_space_init_once(struct address_space *mapping);
3317extern struct inode * igrab(struct inode *);
3318extern ino_t iunique(struct super_block *, ino_t);
3319extern int inode_needs_sync(struct inode *inode);
3320extern int inode_just_drop(struct inode *inode);
3321static inline int inode_generic_drop(struct inode *inode)
3322{
3323 return !inode->i_nlink || inode_unhashed(inode);
3324}
3325extern void d_mark_dontcache(struct inode *inode);
3326
3327extern struct inode *ilookup5_nowait(struct super_block *sb,
3328 unsigned long hashval, int (*test)(struct inode *, void *),
3329 void *data);
3330extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
3331 int (*test)(struct inode *, void *), void *data);
3332extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
3333
3334extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
3335 int (*test)(struct inode *, void *),
3336 int (*set)(struct inode *, void *),
3337 void *data);
3338struct inode *iget5_locked(struct super_block *, unsigned long,
3339 int (*test)(struct inode *, void *),
3340 int (*set)(struct inode *, void *), void *);
3341struct inode *iget5_locked_rcu(struct super_block *, unsigned long,
3342 int (*test)(struct inode *, void *),
3343 int (*set)(struct inode *, void *), void *);
3344extern struct inode * iget_locked(struct super_block *, unsigned long);
3345extern struct inode *find_inode_nowait(struct super_block *,
3346 unsigned long,
3347 int (*match)(struct inode *,
3348 unsigned long, void *),
3349 void *data);
3350extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
3351 int (*)(struct inode *, void *), void *);
3352extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
3353extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
3354extern int insert_inode_locked(struct inode *);
3355#ifdef CONFIG_DEBUG_LOCK_ALLOC
3356extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
3357#else
3358static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
3359#endif
3360extern void unlock_new_inode(struct inode *);
3361extern void discard_new_inode(struct inode *);
3362extern unsigned int get_next_ino(void);
3363extern void evict_inodes(struct super_block *sb);
3364void dump_mapping(const struct address_space *);
3365
3366/*
3367 * Userspace may rely on the inode number being non-zero. For example, glibc
3368 * simply ignores files with zero i_ino in unlink() and other places.
3369 *
3370 * As an additional complication, if userspace was compiled with
3371 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
3372 * lower 32 bits, so we need to check that those aren't zero explicitly. With
3373 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
3374 * better safe than sorry.
3375 */
3376static inline bool is_zero_ino(ino_t ino)
3377{
3378 return (u32)ino == 0;
3379}
3380
3381/*
3382 * inode->i_lock must be held
3383 */
3384static inline void __iget(struct inode *inode)
3385{
3386 atomic_inc(v: &inode->i_count);
3387}
3388
3389extern void iget_failed(struct inode *);
3390extern void clear_inode(struct inode *);
3391extern void __destroy_inode(struct inode *);
3392struct inode *alloc_inode(struct super_block *sb);
3393static inline struct inode *new_inode_pseudo(struct super_block *sb)
3394{
3395 return alloc_inode(sb);
3396}
3397extern struct inode *new_inode(struct super_block *sb);
3398extern void free_inode_nonrcu(struct inode *inode);
3399extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
3400extern int file_remove_privs(struct file *);
3401int setattr_should_drop_sgid(struct mnt_idmap *idmap,
3402 const struct inode *inode);
3403
3404/*
3405 * This must be used for allocating filesystems specific inodes to set
3406 * up the inode reclaim context correctly.
3407 */
3408#define alloc_inode_sb(_sb, _cache, _gfp) kmem_cache_alloc_lru(_cache, &_sb->s_inode_lru, _gfp)
3409
3410extern void __insert_inode_hash(struct inode *, unsigned long hashval);
3411static inline void insert_inode_hash(struct inode *inode)
3412{
3413 __insert_inode_hash(inode, hashval: inode->i_ino);
3414}
3415
3416extern void __remove_inode_hash(struct inode *);
3417static inline void remove_inode_hash(struct inode *inode)
3418{
3419 if (!inode_unhashed(inode) && !hlist_fake(h: &inode->i_hash))
3420 __remove_inode_hash(inode);
3421}
3422
3423extern void inode_sb_list_add(struct inode *inode);
3424extern void inode_add_lru(struct inode *inode);
3425
3426extern int sb_set_blocksize(struct super_block *, int);
3427extern int sb_min_blocksize(struct super_block *, int);
3428
3429int generic_file_mmap(struct file *, struct vm_area_struct *);
3430int generic_file_mmap_prepare(struct vm_area_desc *desc);
3431int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
3432int generic_file_readonly_mmap_prepare(struct vm_area_desc *desc);
3433extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
3434int generic_write_checks_count(struct kiocb *iocb, loff_t *count);
3435extern int generic_write_check_limits(struct file *file, loff_t pos,
3436 loff_t *count);
3437extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
3438ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
3439 ssize_t already_read);
3440extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
3441extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
3442extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
3443extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
3444ssize_t generic_perform_write(struct kiocb *, struct iov_iter *);
3445ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
3446 ssize_t direct_written, ssize_t buffered_written);
3447
3448ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
3449 rwf_t flags);
3450ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
3451 rwf_t flags);
3452ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
3453 struct iov_iter *iter);
3454ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
3455 struct iov_iter *iter);
3456
3457/* fs/splice.c */
3458ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
3459 struct pipe_inode_info *pipe,
3460 size_t len, unsigned int flags);
3461ssize_t copy_splice_read(struct file *in, loff_t *ppos,
3462 struct pipe_inode_info *pipe,
3463 size_t len, unsigned int flags);
3464extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
3465 struct file *, loff_t *, size_t, unsigned int);
3466
3467
3468extern void
3469file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
3470extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3471extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3472extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3473extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3474 int whence, loff_t maxsize, loff_t eof);
3475loff_t generic_llseek_cookie(struct file *file, loff_t offset, int whence,
3476 u64 *cookie);
3477extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3478 int whence, loff_t size);
3479extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3480extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3481int rw_verify_area(int, struct file *, const loff_t *, size_t);
3482extern int generic_file_open(struct inode * inode, struct file * filp);
3483extern int nonseekable_open(struct inode * inode, struct file * filp);
3484extern int stream_open(struct inode * inode, struct file * filp);
3485
3486#ifdef CONFIG_BLOCK
3487typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
3488 loff_t file_offset);
3489
3490enum {
3491 /* need locking between buffered and direct access */
3492 DIO_LOCKING = 0x01,
3493
3494 /* filesystem does not support filling holes */
3495 DIO_SKIP_HOLES = 0x02,
3496};
3497
3498ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
3499 struct block_device *bdev, struct iov_iter *iter,
3500 get_block_t get_block,
3501 dio_iodone_t end_io,
3502 int flags);
3503
3504static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3505 struct inode *inode,
3506 struct iov_iter *iter,
3507 get_block_t get_block)
3508{
3509 return __blockdev_direct_IO(iocb, inode, bdev: inode->i_sb->s_bdev, iter,
3510 get_block, NULL, flags: DIO_LOCKING | DIO_SKIP_HOLES);
3511}
3512#endif
3513
3514bool inode_dio_finished(const struct inode *inode);
3515void inode_dio_wait(struct inode *inode);
3516void inode_dio_wait_interruptible(struct inode *inode);
3517
3518/**
3519 * inode_dio_begin - signal start of a direct I/O requests
3520 * @inode: inode the direct I/O happens on
3521 *
3522 * This is called once we've finished processing a direct I/O request,
3523 * and is used to wake up callers waiting for direct I/O to be quiesced.
3524 */
3525static inline void inode_dio_begin(struct inode *inode)
3526{
3527 atomic_inc(v: &inode->i_dio_count);
3528}
3529
3530/**
3531 * inode_dio_end - signal finish of a direct I/O requests
3532 * @inode: inode the direct I/O happens on
3533 *
3534 * This is called once we've finished processing a direct I/O request,
3535 * and is used to wake up callers waiting for direct I/O to be quiesced.
3536 */
3537static inline void inode_dio_end(struct inode *inode)
3538{
3539 if (atomic_dec_and_test(v: &inode->i_dio_count))
3540 wake_up_var(var: &inode->i_dio_count);
3541}
3542
3543extern void inode_set_flags(struct inode *inode, unsigned int flags,
3544 unsigned int mask);
3545
3546extern const struct file_operations generic_ro_fops;
3547
3548#define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
3549
3550extern int readlink_copy(char __user *, int, const char *, int);
3551extern int page_readlink(struct dentry *, char __user *, int);
3552extern const char *page_get_link_raw(struct dentry *, struct inode *,
3553 struct delayed_call *);
3554extern const char *page_get_link(struct dentry *, struct inode *,
3555 struct delayed_call *);
3556extern void page_put_link(void *);
3557extern int page_symlink(struct inode *inode, const char *symname, int len);
3558extern const struct inode_operations page_symlink_inode_operations;
3559extern void kfree_link(void *);
3560void fill_mg_cmtime(struct kstat *stat, u32 request_mask, struct inode *inode);
3561void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *);
3562void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
3563void generic_fill_statx_atomic_writes(struct kstat *stat,
3564 unsigned int unit_min,
3565 unsigned int unit_max,
3566 unsigned int unit_max_opt);
3567extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
3568extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
3569void __inode_add_bytes(struct inode *inode, loff_t bytes);
3570void inode_add_bytes(struct inode *inode, loff_t bytes);
3571void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3572void inode_sub_bytes(struct inode *inode, loff_t bytes);
3573static inline loff_t __inode_get_bytes(struct inode *inode)
3574{
3575 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3576}
3577loff_t inode_get_bytes(struct inode *inode);
3578void inode_set_bytes(struct inode *inode, loff_t bytes);
3579const char *simple_get_link(struct dentry *, struct inode *,
3580 struct delayed_call *);
3581extern const struct inode_operations simple_symlink_inode_operations;
3582
3583extern int iterate_dir(struct file *, struct dir_context *);
3584
3585int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
3586 int flags);
3587int vfs_fstat(int fd, struct kstat *stat);
3588
3589static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3590{
3591 return vfs_fstatat(AT_FDCWD, filename, stat, flags: 0);
3592}
3593static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3594{
3595 return vfs_fstatat(AT_FDCWD, filename: name, stat, AT_SYMLINK_NOFOLLOW);
3596}
3597
3598extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3599extern int vfs_readlink(struct dentry *, char __user *, int);
3600
3601extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3602extern void put_filesystem(struct file_system_type *fs);
3603extern struct file_system_type *get_fs_type(const char *name);
3604extern void drop_super(struct super_block *sb);
3605extern void drop_super_exclusive(struct super_block *sb);
3606extern void iterate_supers(void (*f)(struct super_block *, void *), void *arg);
3607extern void iterate_supers_type(struct file_system_type *,
3608 void (*)(struct super_block *, void *), void *);
3609void filesystems_freeze(void);
3610void filesystems_thaw(void);
3611
3612extern int dcache_dir_open(struct inode *, struct file *);
3613extern int dcache_dir_close(struct inode *, struct file *);
3614extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3615extern int dcache_readdir(struct file *, struct dir_context *);
3616extern int simple_setattr(struct mnt_idmap *, struct dentry *,
3617 struct iattr *);
3618extern int simple_getattr(struct mnt_idmap *, const struct path *,
3619 struct kstat *, u32, unsigned int);
3620extern int simple_statfs(struct dentry *, struct kstatfs *);
3621extern int simple_open(struct inode *inode, struct file *file);
3622extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3623extern int simple_unlink(struct inode *, struct dentry *);
3624extern int simple_rmdir(struct inode *, struct dentry *);
3625void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry,
3626 struct inode *new_dir, struct dentry *new_dentry);
3627extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
3628 struct inode *new_dir, struct dentry *new_dentry);
3629extern int simple_rename(struct mnt_idmap *, struct inode *,
3630 struct dentry *, struct inode *, struct dentry *,
3631 unsigned int);
3632extern void simple_recursive_removal(struct dentry *,
3633 void (*callback)(struct dentry *));
3634extern void locked_recursive_removal(struct dentry *,
3635 void (*callback)(struct dentry *));
3636extern int noop_fsync(struct file *, loff_t, loff_t, int);
3637extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
3638extern int simple_empty(struct dentry *);
3639extern int simple_write_begin(const struct kiocb *iocb,
3640 struct address_space *mapping,
3641 loff_t pos, unsigned len,
3642 struct folio **foliop, void **fsdata);
3643extern const struct address_space_operations ram_aops;
3644extern int always_delete_dentry(const struct dentry *);
3645extern struct inode *alloc_anon_inode(struct super_block *);
3646struct inode *anon_inode_make_secure_inode(struct super_block *sb, const char *name,
3647 const struct inode *context_inode);
3648extern int simple_nosetlease(struct file *, int, struct file_lease **, void **);
3649
3650extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3651extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3652extern const struct file_operations simple_dir_operations;
3653extern const struct inode_operations simple_dir_inode_operations;
3654extern void make_empty_dir_inode(struct inode *inode);
3655extern bool is_empty_dir_inode(struct inode *inode);
3656struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3657struct dentry *d_alloc_name(struct dentry *, const char *);
3658extern int simple_fill_super(struct super_block *, unsigned long,
3659 const struct tree_descr *);
3660extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3661extern void simple_release_fs(struct vfsmount **mount, int *count);
3662struct dentry *simple_start_creating(struct dentry *, const char *);
3663
3664extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3665 loff_t *ppos, const void *from, size_t available);
3666extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3667 const void __user *from, size_t count);
3668
3669struct offset_ctx {
3670 struct maple_tree mt;
3671 unsigned long next_offset;
3672};
3673
3674void simple_offset_init(struct offset_ctx *octx);
3675int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry);
3676void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry);
3677int simple_offset_rename(struct inode *old_dir, struct dentry *old_dentry,
3678 struct inode *new_dir, struct dentry *new_dentry);
3679int simple_offset_rename_exchange(struct inode *old_dir,
3680 struct dentry *old_dentry,
3681 struct inode *new_dir,
3682 struct dentry *new_dentry);
3683void simple_offset_destroy(struct offset_ctx *octx);
3684
3685extern const struct file_operations simple_offset_dir_operations;
3686
3687extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3688extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3689
3690extern int generic_check_addressable(unsigned, u64);
3691
3692extern void generic_set_sb_d_ops(struct super_block *sb);
3693extern int generic_ci_match(const struct inode *parent,
3694 const struct qstr *name,
3695 const struct qstr *folded_name,
3696 const u8 *de_name, u32 de_name_len);
3697
3698#if IS_ENABLED(CONFIG_UNICODE)
3699int generic_ci_d_hash(const struct dentry *dentry, struct qstr *str);
3700int generic_ci_d_compare(const struct dentry *dentry, unsigned int len,
3701 const char *str, const struct qstr *name);
3702
3703/**
3704 * generic_ci_validate_strict_name - Check if a given name is suitable
3705 * for a directory
3706 *
3707 * This functions checks if the proposed filename is valid for the
3708 * parent directory. That means that only valid UTF-8 filenames will be
3709 * accepted for casefold directories from filesystems created with the
3710 * strict encoding flag. That also means that any name will be
3711 * accepted for directories that doesn't have casefold enabled, or
3712 * aren't being strict with the encoding.
3713 *
3714 * @dir: inode of the directory where the new file will be created
3715 * @name: name of the new file
3716 *
3717 * Return:
3718 * * True: if the filename is suitable for this directory. It can be
3719 * true if a given name is not suitable for a strict encoding
3720 * directory, but the directory being used isn't strict
3721 * * False if the filename isn't suitable for this directory. This only
3722 * happens when a directory is casefolded and the filesystem is strict
3723 * about its encoding.
3724 */
3725static inline bool generic_ci_validate_strict_name(struct inode *dir,
3726 const struct qstr *name)
3727{
3728 if (!IS_CASEFOLDED(dir) || !sb_has_strict_encoding(dir->i_sb))
3729 return true;
3730
3731 /*
3732 * A casefold dir must have a encoding set, unless the filesystem
3733 * is corrupted
3734 */
3735 if (WARN_ON_ONCE(!dir->i_sb->s_encoding))
3736 return true;
3737
3738 return !utf8_validate(dir->i_sb->s_encoding, name);
3739}
3740#else
3741static inline bool generic_ci_validate_strict_name(struct inode *dir,
3742 const struct qstr *name)
3743{
3744 return true;
3745}
3746#endif
3747
3748static inline struct unicode_map *sb_encoding(const struct super_block *sb)
3749{
3750#if IS_ENABLED(CONFIG_UNICODE)
3751 return sb->s_encoding;
3752#else
3753 return NULL;
3754#endif
3755}
3756
3757static inline bool sb_has_encoding(const struct super_block *sb)
3758{
3759 return !!sb_encoding(sb);
3760}
3761
3762/*
3763 * Compare if two super blocks have the same encoding and flags
3764 */
3765static inline bool sb_same_encoding(const struct super_block *sb1,
3766 const struct super_block *sb2)
3767{
3768#if IS_ENABLED(CONFIG_UNICODE)
3769 if (sb1->s_encoding == sb2->s_encoding)
3770 return true;
3771
3772 return (sb1->s_encoding && sb2->s_encoding &&
3773 (sb1->s_encoding->version == sb2->s_encoding->version) &&
3774 (sb1->s_encoding_flags == sb2->s_encoding_flags));
3775#else
3776 return true;
3777#endif
3778}
3779
3780int may_setattr(struct mnt_idmap *idmap, struct inode *inode,
3781 unsigned int ia_valid);
3782int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *);
3783extern int inode_newsize_ok(const struct inode *, loff_t offset);
3784void setattr_copy(struct mnt_idmap *, struct inode *inode,
3785 const struct iattr *attr);
3786
3787extern int file_update_time(struct file *file);
3788
3789static inline bool file_is_dax(const struct file *file)
3790{
3791 return file && IS_DAX(file->f_mapping->host);
3792}
3793
3794static inline bool vma_is_dax(const struct vm_area_struct *vma)
3795{
3796 return file_is_dax(file: vma->vm_file);
3797}
3798
3799static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3800{
3801 struct inode *inode;
3802
3803 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
3804 return false;
3805 if (!vma_is_dax(vma))
3806 return false;
3807 inode = file_inode(f: vma->vm_file);
3808 if (S_ISCHR(inode->i_mode))
3809 return false; /* device-dax */
3810 return true;
3811}
3812
3813static inline int iocb_flags(struct file *file)
3814{
3815 int res = 0;
3816 if (file->f_flags & O_APPEND)
3817 res |= IOCB_APPEND;
3818 if (file->f_flags & O_DIRECT)
3819 res |= IOCB_DIRECT;
3820 if (file->f_flags & O_DSYNC)
3821 res |= IOCB_DSYNC;
3822 if (file->f_flags & __O_SYNC)
3823 res |= IOCB_SYNC;
3824 return res;
3825}
3826
3827static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags,
3828 int rw_type)
3829{
3830 int kiocb_flags = 0;
3831
3832 /* make sure there's no overlap between RWF and private IOCB flags */
3833 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3834
3835 if (!flags)
3836 return 0;
3837 if (unlikely(flags & ~RWF_SUPPORTED))
3838 return -EOPNOTSUPP;
3839 if (unlikely((flags & RWF_APPEND) && (flags & RWF_NOAPPEND)))
3840 return -EINVAL;
3841
3842 if (flags & RWF_NOWAIT) {
3843 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3844 return -EOPNOTSUPP;
3845 }
3846 if (flags & RWF_ATOMIC) {
3847 if (rw_type != WRITE)
3848 return -EOPNOTSUPP;
3849 if (!(ki->ki_filp->f_mode & FMODE_CAN_ATOMIC_WRITE))
3850 return -EOPNOTSUPP;
3851 }
3852 if (flags & RWF_DONTCACHE) {
3853 /* file system must support it */
3854 if (!(ki->ki_filp->f_op->fop_flags & FOP_DONTCACHE))
3855 return -EOPNOTSUPP;
3856 /* DAX mappings not supported */
3857 if (IS_DAX(ki->ki_filp->f_mapping->host))
3858 return -EOPNOTSUPP;
3859 }
3860 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3861 if (flags & RWF_SYNC)
3862 kiocb_flags |= IOCB_DSYNC;
3863
3864 if ((flags & RWF_NOAPPEND) && (ki->ki_flags & IOCB_APPEND)) {
3865 if (IS_APPEND(file_inode(ki->ki_filp)))
3866 return -EPERM;
3867 ki->ki_flags &= ~IOCB_APPEND;
3868 }
3869
3870 ki->ki_flags |= kiocb_flags;
3871 return 0;
3872}
3873
3874/* Transaction based IO helpers */
3875
3876/*
3877 * An argresp is stored in an allocated page and holds the
3878 * size of the argument or response, along with its content
3879 */
3880struct simple_transaction_argresp {
3881 ssize_t size;
3882 char data[];
3883};
3884
3885#define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3886
3887char *simple_transaction_get(struct file *file, const char __user *buf,
3888 size_t size);
3889ssize_t simple_transaction_read(struct file *file, char __user *buf,
3890 size_t size, loff_t *pos);
3891int simple_transaction_release(struct inode *inode, struct file *file);
3892
3893void simple_transaction_set(struct file *file, size_t n);
3894
3895/*
3896 * simple attribute files
3897 *
3898 * These attributes behave similar to those in sysfs:
3899 *
3900 * Writing to an attribute immediately sets a value, an open file can be
3901 * written to multiple times.
3902 *
3903 * Reading from an attribute creates a buffer from the value that might get
3904 * read with multiple read calls. When the attribute has been read
3905 * completely, no further read calls are possible until the file is opened
3906 * again.
3907 *
3908 * All attributes contain a text representation of a numeric value
3909 * that are accessed with the get() and set() functions.
3910 */
3911#define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \
3912static int __fops ## _open(struct inode *inode, struct file *file) \
3913{ \
3914 __simple_attr_check_format(__fmt, 0ull); \
3915 return simple_attr_open(inode, file, __get, __set, __fmt); \
3916} \
3917static const struct file_operations __fops = { \
3918 .owner = THIS_MODULE, \
3919 .open = __fops ## _open, \
3920 .release = simple_attr_release, \
3921 .read = simple_attr_read, \
3922 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \
3923 .llseek = generic_file_llseek, \
3924}
3925
3926#define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
3927 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false)
3928
3929#define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \
3930 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true)
3931
3932static inline __printf(1, 2)
3933void __simple_attr_check_format(const char *fmt, ...)
3934{
3935 /* don't do anything, just let the compiler check the arguments; */
3936}
3937
3938int simple_attr_open(struct inode *inode, struct file *file,
3939 int (*get)(void *, u64 *), int (*set)(void *, u64),
3940 const char *fmt);
3941int simple_attr_release(struct inode *inode, struct file *file);
3942ssize_t simple_attr_read(struct file *file, char __user *buf,
3943 size_t len, loff_t *ppos);
3944ssize_t simple_attr_write(struct file *file, const char __user *buf,
3945 size_t len, loff_t *ppos);
3946ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
3947 size_t len, loff_t *ppos);
3948
3949struct ctl_table;
3950int __init list_bdev_fs_names(char *buf, size_t size);
3951
3952#define __FMODE_EXEC ((__force int) FMODE_EXEC)
3953
3954#define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3955#define OPEN_FMODE(flag) ((__force fmode_t)((flag + 1) & O_ACCMODE))
3956
3957static inline bool is_sxid(umode_t mode)
3958{
3959 return mode & (S_ISUID | S_ISGID);
3960}
3961
3962static inline int check_sticky(struct mnt_idmap *idmap,
3963 struct inode *dir, struct inode *inode)
3964{
3965 if (!(dir->i_mode & S_ISVTX))
3966 return 0;
3967
3968 return __check_sticky(idmap, dir, inode);
3969}
3970
3971static inline void inode_has_no_xattr(struct inode *inode)
3972{
3973 if (!is_sxid(mode: inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3974 inode->i_flags |= S_NOSEC;
3975}
3976
3977static inline bool is_root_inode(struct inode *inode)
3978{
3979 return inode == inode->i_sb->s_root->d_inode;
3980}
3981
3982static inline bool dir_emit(struct dir_context *ctx,
3983 const char *name, int namelen,
3984 u64 ino, unsigned type)
3985{
3986 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type);
3987}
3988static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3989{
3990 return ctx->actor(ctx, ".", 1, ctx->pos,
3991 file->f_path.dentry->d_inode->i_ino, DT_DIR);
3992}
3993static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3994{
3995 return ctx->actor(ctx, "..", 2, ctx->pos,
3996 d_parent_ino(dentry: file->f_path.dentry), DT_DIR);
3997}
3998static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3999{
4000 if (ctx->pos == 0) {
4001 if (!dir_emit_dot(file, ctx))
4002 return false;
4003 ctx->pos = 1;
4004 }
4005 if (ctx->pos == 1) {
4006 if (!dir_emit_dotdot(file, ctx))
4007 return false;
4008 ctx->pos = 2;
4009 }
4010 return true;
4011}
4012static inline bool dir_relax(struct inode *inode)
4013{
4014 inode_unlock(inode);
4015 inode_lock(inode);
4016 return !IS_DEADDIR(inode);
4017}
4018
4019static inline bool dir_relax_shared(struct inode *inode)
4020{
4021 inode_unlock_shared(inode);
4022 inode_lock_shared(inode);
4023 return !IS_DEADDIR(inode);
4024}
4025
4026extern bool path_noexec(const struct path *path);
4027extern void inode_nohighmem(struct inode *inode);
4028
4029/* mm/fadvise.c */
4030extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
4031 int advice);
4032extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
4033 int advice);
4034
4035static inline bool vfs_empty_path(int dfd, const char __user *path)
4036{
4037 char c;
4038
4039 if (dfd < 0)
4040 return false;
4041
4042 /* We now allow NULL to be used for empty path. */
4043 if (!path)
4044 return true;
4045
4046 if (unlikely(get_user(c, path)))
4047 return false;
4048
4049 return !c;
4050}
4051
4052int generic_atomic_write_valid(struct kiocb *iocb, struct iov_iter *iter);
4053
4054static inline bool extensible_ioctl_valid(unsigned int cmd_a,
4055 unsigned int cmd_b, size_t min_size)
4056{
4057 if (_IOC_DIR(cmd_a) != _IOC_DIR(cmd_b))
4058 return false;
4059 if (_IOC_TYPE(cmd_a) != _IOC_TYPE(cmd_b))
4060 return false;
4061 if (_IOC_NR(cmd_a) != _IOC_NR(cmd_b))
4062 return false;
4063 if (_IOC_SIZE(cmd_a) < min_size)
4064 return false;
4065 return true;
4066}
4067
4068#endif /* _LINUX_FS_H */
4069