1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/linkage.h>
6 #include <linux/wait_bit.h>
7 #include <linux/kdev_t.h>
8 #include <linux/dcache.h>
9 #include <linux/path.h>
10 #include <linux/stat.h>
11 #include <linux/cache.h>
12 #include <linux/list.h>
13 #include <linux/list_lru.h>
14 #include <linux/llist.h>
15 #include <linux/radix-tree.h>
16 #include <linux/xarray.h>
17 #include <linux/rbtree.h>
18 #include <linux/init.h>
19 #include <linux/pid.h>
20 #include <linux/bug.h>
21 #include <linux/mutex.h>
22 #include <linux/rwsem.h>
23 #include <linux/mm_types.h>
24 #include <linux/capability.h>
25 #include <linux/semaphore.h>
26 #include <linux/fcntl.h>
27 #include <linux/rculist_bl.h>
28 #include <linux/atomic.h>
29 #include <linux/shrinker.h>
30 #include <linux/migrate_mode.h>
31 #include <linux/uidgid.h>
32 #include <linux/lockdep.h>
33 #include <linux/percpu-rwsem.h>
34 #include <linux/workqueue.h>
35 #include <linux/delayed_call.h>
36 #include <linux/uuid.h>
37 #include <linux/errseq.h>
38 #include <linux/ioprio.h>
39 #include <linux/fs_types.h>
40 #include <linux/build_bug.h>
41 #include <linux/stddef.h>
42 #include <linux/mount.h>
43 #include <linux/cred.h>
44 #include <linux/mnt_idmapping.h>
45 #include <linux/slab.h>
46 #include <linux/maple_tree.h>
47 #include <linux/rw_hint.h>
49 #include <asm/byteorder.h>
50 #include <uapi/linux/fs.h>
52 struct backing_dev_info;
56 struct export_operations;
57 struct fiemap_extent_info;
62 struct pipe_inode_info;
63 struct poll_table_struct;
65 struct vm_area_struct;
68 struct swap_info_struct;
70 struct workqueue_struct;
72 struct fscrypt_inode_info;
73 struct fscrypt_operations;
75 struct fsverity_operations;
77 struct fs_parameter_spec;
81 extern void __init inode_init(void);
82 extern void __init inode_init_early(void);
83 extern void __init files_init(void);
84 extern void __init files_maxfiles_init(void);
86 extern unsigned long get_max_files(void);
87 extern unsigned int sysctl_nr_open;
89 typedef __kernel_rwf_t rwf_t;
92 typedef int (get_block_t)(struct inode *inode, sector_t iblock,
93 struct buffer_head *bh_result, int create);
94 typedef int (dio_iodone_t)(struct kiocb *iocb, loff_t offset,
95 ssize_t bytes, void *private);
97 #define MAY_EXEC 0x00000001
98 #define MAY_WRITE 0x00000002
99 #define MAY_READ 0x00000004
100 #define MAY_APPEND 0x00000008
101 #define MAY_ACCESS 0x00000010
102 #define MAY_OPEN 0x00000020
103 #define MAY_CHDIR 0x00000040
104 /* called from RCU mode, don't block */
105 #define MAY_NOT_BLOCK 0x00000080
108 * flags in file.f_mode. Note that FMODE_READ and FMODE_WRITE must correspond
109 * to O_WRONLY and O_RDWR via the strange trick in do_dentry_open()
112 /* file is open for reading */
113 #define FMODE_READ ((__force fmode_t)0x1)
114 /* file is open for writing */
115 #define FMODE_WRITE ((__force fmode_t)0x2)
116 /* file is seekable */
117 #define FMODE_LSEEK ((__force fmode_t)0x4)
118 /* file can be accessed using pread */
119 #define FMODE_PREAD ((__force fmode_t)0x8)
120 /* file can be accessed using pwrite */
121 #define FMODE_PWRITE ((__force fmode_t)0x10)
122 /* File is opened for execution with sys_execve / sys_uselib */
123 #define FMODE_EXEC ((__force fmode_t)0x20)
124 /* 32bit hashes as llseek() offset (for directories) */
125 #define FMODE_32BITHASH ((__force fmode_t)0x200)
126 /* 64bit hashes as llseek() offset (for directories) */
127 #define FMODE_64BITHASH ((__force fmode_t)0x400)
130 * Don't update ctime and mtime.
132 * Currently a special hack for the XFS open_by_handle ioctl, but we'll
133 * hopefully graduate it to a proper O_CMTIME flag supported by open(2) soon.
135 #define FMODE_NOCMTIME ((__force fmode_t)0x800)
137 /* Expect random access pattern */
138 #define FMODE_RANDOM ((__force fmode_t)0x1000)
140 /* File is huge (eg. /dev/mem): treat loff_t as unsigned */
141 #define FMODE_UNSIGNED_OFFSET ((__force fmode_t)0x2000)
143 /* File is opened with O_PATH; almost nothing can be done with it */
144 #define FMODE_PATH ((__force fmode_t)0x4000)
146 /* File needs atomic accesses to f_pos */
147 #define FMODE_ATOMIC_POS ((__force fmode_t)0x8000)
148 /* Write access to underlying fs */
149 #define FMODE_WRITER ((__force fmode_t)0x10000)
150 /* Has read method(s) */
151 #define FMODE_CAN_READ ((__force fmode_t)0x20000)
152 /* Has write method(s) */
153 #define FMODE_CAN_WRITE ((__force fmode_t)0x40000)
155 #define FMODE_OPENED ((__force fmode_t)0x80000)
156 #define FMODE_CREATED ((__force fmode_t)0x100000)
158 /* File is stream-like */
159 #define FMODE_STREAM ((__force fmode_t)0x200000)
161 /* File supports DIRECT IO */
162 #define FMODE_CAN_ODIRECT ((__force fmode_t)0x400000)
164 #define FMODE_NOREUSE ((__force fmode_t)0x800000)
166 /* File supports non-exclusive O_DIRECT writes from multiple threads */
167 #define FMODE_DIO_PARALLEL_WRITE ((__force fmode_t)0x1000000)
169 /* File is embedded in backing_file object */
170 #define FMODE_BACKING ((__force fmode_t)0x2000000)
172 /* File was opened by fanotify and shouldn't generate fanotify events */
173 #define FMODE_NONOTIFY ((__force fmode_t)0x4000000)
175 /* File is capable of returning -EAGAIN if I/O will block */
176 #define FMODE_NOWAIT ((__force fmode_t)0x8000000)
178 /* File represents mount that needs unmounting */
179 #define FMODE_NEED_UNMOUNT ((__force fmode_t)0x10000000)
181 /* File does not contribute to nr_files count */
182 #define FMODE_NOACCOUNT ((__force fmode_t)0x20000000)
184 /* File supports async buffered reads */
185 #define FMODE_BUF_RASYNC ((__force fmode_t)0x40000000)
187 /* File supports async nowait buffered writes */
188 #define FMODE_BUF_WASYNC ((__force fmode_t)0x80000000)
191 * Attribute flags. These should be or-ed together to figure out what
194 #define ATTR_MODE (1 << 0)
195 #define ATTR_UID (1 << 1)
196 #define ATTR_GID (1 << 2)
197 #define ATTR_SIZE (1 << 3)
198 #define ATTR_ATIME (1 << 4)
199 #define ATTR_MTIME (1 << 5)
200 #define ATTR_CTIME (1 << 6)
201 #define ATTR_ATIME_SET (1 << 7)
202 #define ATTR_MTIME_SET (1 << 8)
203 #define ATTR_FORCE (1 << 9) /* Not a change, but a change it */
204 #define ATTR_KILL_SUID (1 << 11)
205 #define ATTR_KILL_SGID (1 << 12)
206 #define ATTR_FILE (1 << 13)
207 #define ATTR_KILL_PRIV (1 << 14)
208 #define ATTR_OPEN (1 << 15) /* Truncating from open(O_TRUNC) */
209 #define ATTR_TIMES_SET (1 << 16)
210 #define ATTR_TOUCH (1 << 17)
213 * Whiteout is represented by a char device. The following constants define the
214 * mode and device number to use.
216 #define WHITEOUT_MODE 0
217 #define WHITEOUT_DEV 0
220 * This is the Inode Attributes structure, used for notify_change(). It
221 * uses the above definitions as flags, to know which values have changed.
222 * Also, in this manner, a Filesystem can look at only the values it cares
223 * about. Basically, these are the attributes that the VFS layer can
224 * request to change from the FS layer.
226 * Derek Atkins <warlord@MIT.EDU> 94-10-20
229 unsigned int ia_valid;
232 * The two anonymous unions wrap structures with the same member.
234 * Filesystems raising FS_ALLOW_IDMAP need to use ia_vfs{g,u}id which
235 * are a dedicated type requiring the filesystem to use the dedicated
236 * helpers. Other filesystem can continue to use ia_{g,u}id until they
239 * They always contain the same value. In other words FS_ALLOW_IDMAP
240 * pass down the same value on idmapped mounts as they would on regular
252 struct timespec64 ia_atime;
253 struct timespec64 ia_mtime;
254 struct timespec64 ia_ctime;
257 * Not an attribute, but an auxiliary info for filesystems wanting to
258 * implement an ftruncate() like method. NOTE: filesystem should
259 * check for (ia_valid & ATTR_FILE), and not for (ia_file != NULL).
261 struct file *ia_file;
265 * Includes for diskquotas.
267 #include <linux/quota.h>
270 * Maximum number of layers of fs stack. Needs to be limited to
271 * prevent kernel stack overflow
273 #define FILESYSTEM_MAX_STACK_DEPTH 2
276 * enum positive_aop_returns - aop return codes with specific semantics
278 * @AOP_WRITEPAGE_ACTIVATE: Informs the caller that page writeback has
279 * completed, that the page is still locked, and
280 * should be considered active. The VM uses this hint
281 * to return the page to the active list -- it won't
282 * be a candidate for writeback again in the near
283 * future. Other callers must be careful to unlock
284 * the page if they get this return. Returned by
287 * @AOP_TRUNCATED_PAGE: The AOP method that was handed a locked page has
288 * unlocked it and the page might have been truncated.
289 * The caller should back up to acquiring a new page and
290 * trying again. The aop will be taking reasonable
291 * precautions not to livelock. If the caller held a page
292 * reference, it should drop it before retrying. Returned
295 * address_space_operation functions return these large constants to indicate
296 * special semantics to the caller. These are much larger than the bytes in a
297 * page to allow for functions that return the number of bytes operated on in a
301 enum positive_aop_returns {
302 AOP_WRITEPAGE_ACTIVATE = 0x80000,
303 AOP_TRUNCATED_PAGE = 0x80001,
307 * oh the beauties of C type declarations.
310 struct address_space;
311 struct writeback_control;
312 struct readahead_control;
314 /* Match RWF_* bits to IOCB bits */
315 #define IOCB_HIPRI (__force int) RWF_HIPRI
316 #define IOCB_DSYNC (__force int) RWF_DSYNC
317 #define IOCB_SYNC (__force int) RWF_SYNC
318 #define IOCB_NOWAIT (__force int) RWF_NOWAIT
319 #define IOCB_APPEND (__force int) RWF_APPEND
321 /* non-RWF related bits - start at 16 */
322 #define IOCB_EVENTFD (1 << 16)
323 #define IOCB_DIRECT (1 << 17)
324 #define IOCB_WRITE (1 << 18)
325 /* iocb->ki_waitq is valid */
326 #define IOCB_WAITQ (1 << 19)
327 #define IOCB_NOIO (1 << 20)
328 /* can use bio alloc cache */
329 #define IOCB_ALLOC_CACHE (1 << 21)
331 * IOCB_DIO_CALLER_COMP can be set by the iocb owner, to indicate that the
332 * iocb completion can be passed back to the owner for execution from a safe
333 * context rather than needing to be punted through a workqueue. If this
334 * flag is set, the bio completion handling may set iocb->dio_complete to a
335 * handler function and iocb->private to context information for that handler.
336 * The issuer should call the handler with that context information from task
337 * context to complete the processing of the iocb. Note that while this
338 * provides a task context for the dio_complete() callback, it should only be
339 * used on the completion side for non-IO generating completions. It's fine to
340 * call blocking functions from this callback, but they should not wait for
341 * unrelated IO (like cache flushing, new IO generation, etc).
343 #define IOCB_DIO_CALLER_COMP (1 << 22)
344 /* kiocb is a read or write operation submitted by fs/aio.c. */
345 #define IOCB_AIO_RW (1 << 23)
347 /* for use in trace events */
348 #define TRACE_IOCB_STRINGS \
349 { IOCB_HIPRI, "HIPRI" }, \
350 { IOCB_DSYNC, "DSYNC" }, \
351 { IOCB_SYNC, "SYNC" }, \
352 { IOCB_NOWAIT, "NOWAIT" }, \
353 { IOCB_APPEND, "APPEND" }, \
354 { IOCB_EVENTFD, "EVENTFD"}, \
355 { IOCB_DIRECT, "DIRECT" }, \
356 { IOCB_WRITE, "WRITE" }, \
357 { IOCB_WAITQ, "WAITQ" }, \
358 { IOCB_NOIO, "NOIO" }, \
359 { IOCB_ALLOC_CACHE, "ALLOC_CACHE" }, \
360 { IOCB_DIO_CALLER_COMP, "CALLER_COMP" }
363 struct file *ki_filp;
365 void (*ki_complete)(struct kiocb *iocb, long ret);
368 u16 ki_ioprio; /* See linux/ioprio.h */
371 * Only used for async buffered reads, where it denotes the
372 * page waitqueue associated with completing the read. Valid
373 * IFF IOCB_WAITQ is set.
375 struct wait_page_queue *ki_waitq;
377 * Can be used for O_DIRECT IO, where the completion handling
378 * is punted back to the issuer of the IO. May only be set
379 * if IOCB_DIO_CALLER_COMP is set by the issuer, and the issuer
380 * must then check for presence of this handler when ki_complete
381 * is invoked. The data passed in to this handler must be
382 * assigned to ->private when dio_complete is assigned.
384 ssize_t (*dio_complete)(void *data);
388 static inline bool is_sync_kiocb(struct kiocb *kiocb)
390 return kiocb->ki_complete == NULL;
393 struct address_space_operations {
394 int (*writepage)(struct page *page, struct writeback_control *wbc);
395 int (*read_folio)(struct file *, struct folio *);
397 /* Write back some dirty pages from this mapping. */
398 int (*writepages)(struct address_space *, struct writeback_control *);
400 /* Mark a folio dirty. Return true if this dirtied it */
401 bool (*dirty_folio)(struct address_space *, struct folio *);
403 void (*readahead)(struct readahead_control *);
405 int (*write_begin)(struct file *, struct address_space *mapping,
406 loff_t pos, unsigned len,
407 struct page **pagep, void **fsdata);
408 int (*write_end)(struct file *, struct address_space *mapping,
409 loff_t pos, unsigned len, unsigned copied,
410 struct page *page, void *fsdata);
412 /* Unfortunately this kludge is needed for FIBMAP. Don't use it */
413 sector_t (*bmap)(struct address_space *, sector_t);
414 void (*invalidate_folio) (struct folio *, size_t offset, size_t len);
415 bool (*release_folio)(struct folio *, gfp_t);
416 void (*free_folio)(struct folio *folio);
417 ssize_t (*direct_IO)(struct kiocb *, struct iov_iter *iter);
419 * migrate the contents of a folio to the specified target. If
420 * migrate_mode is MIGRATE_ASYNC, it must not block.
422 int (*migrate_folio)(struct address_space *, struct folio *dst,
423 struct folio *src, enum migrate_mode);
424 int (*launder_folio)(struct folio *);
425 bool (*is_partially_uptodate) (struct folio *, size_t from,
427 void (*is_dirty_writeback) (struct folio *, bool *dirty, bool *wb);
428 int (*error_remove_folio)(struct address_space *, struct folio *);
430 /* swapfile support */
431 int (*swap_activate)(struct swap_info_struct *sis, struct file *file,
433 void (*swap_deactivate)(struct file *file);
434 int (*swap_rw)(struct kiocb *iocb, struct iov_iter *iter);
437 extern const struct address_space_operations empty_aops;
440 * struct address_space - Contents of a cacheable, mappable object.
441 * @host: Owner, either the inode or the block_device.
442 * @i_pages: Cached pages.
443 * @invalidate_lock: Guards coherency between page cache contents and
444 * file offset->disk block mappings in the filesystem during invalidates.
445 * It is also used to block modification of page cache contents through
447 * @gfp_mask: Memory allocation flags to use for allocating pages.
448 * @i_mmap_writable: Number of VM_SHARED, VM_MAYWRITE mappings.
449 * @nr_thps: Number of THPs in the pagecache (non-shmem only).
450 * @i_mmap: Tree of private and shared mappings.
451 * @i_mmap_rwsem: Protects @i_mmap and @i_mmap_writable.
452 * @nrpages: Number of page entries, protected by the i_pages lock.
453 * @writeback_index: Writeback starts here.
455 * @flags: Error bits and flags (AS_*).
456 * @wb_err: The most recent error which has occurred.
457 * @i_private_lock: For use by the owner of the address_space.
458 * @i_private_list: For use by the owner of the address_space.
459 * @i_private_data: For use by the owner of the address_space.
461 struct address_space {
463 struct xarray i_pages;
464 struct rw_semaphore invalidate_lock;
466 atomic_t i_mmap_writable;
467 #ifdef CONFIG_READ_ONLY_THP_FOR_FS
468 /* number of thp, only for non-shmem files */
471 struct rb_root_cached i_mmap;
472 unsigned long nrpages;
473 pgoff_t writeback_index;
474 const struct address_space_operations *a_ops;
477 spinlock_t i_private_lock;
478 struct list_head i_private_list;
479 struct rw_semaphore i_mmap_rwsem;
480 void * i_private_data;
481 } __attribute__((aligned(sizeof(long)))) __randomize_layout;
483 * On most architectures that alignment is already the case; but
484 * must be enforced here for CRIS, to let the least significant bit
485 * of struct page's "mapping" pointer be used for PAGE_MAPPING_ANON.
488 /* XArray tags, for tagging dirty and writeback pages in the pagecache. */
489 #define PAGECACHE_TAG_DIRTY XA_MARK_0
490 #define PAGECACHE_TAG_WRITEBACK XA_MARK_1
491 #define PAGECACHE_TAG_TOWRITE XA_MARK_2
494 * Returns true if any of the pages in the mapping are marked with the tag.
496 static inline bool mapping_tagged(struct address_space *mapping, xa_mark_t tag)
498 return xa_marked(&mapping->i_pages, tag);
501 static inline void i_mmap_lock_write(struct address_space *mapping)
503 down_write(&mapping->i_mmap_rwsem);
506 static inline int i_mmap_trylock_write(struct address_space *mapping)
508 return down_write_trylock(&mapping->i_mmap_rwsem);
511 static inline void i_mmap_unlock_write(struct address_space *mapping)
513 up_write(&mapping->i_mmap_rwsem);
516 static inline int i_mmap_trylock_read(struct address_space *mapping)
518 return down_read_trylock(&mapping->i_mmap_rwsem);
521 static inline void i_mmap_lock_read(struct address_space *mapping)
523 down_read(&mapping->i_mmap_rwsem);
526 static inline void i_mmap_unlock_read(struct address_space *mapping)
528 up_read(&mapping->i_mmap_rwsem);
531 static inline void i_mmap_assert_locked(struct address_space *mapping)
533 lockdep_assert_held(&mapping->i_mmap_rwsem);
536 static inline void i_mmap_assert_write_locked(struct address_space *mapping)
538 lockdep_assert_held_write(&mapping->i_mmap_rwsem);
542 * Might pages of this file be mapped into userspace?
544 static inline int mapping_mapped(struct address_space *mapping)
546 return !RB_EMPTY_ROOT(&mapping->i_mmap.rb_root);
550 * Might pages of this file have been modified in userspace?
551 * Note that i_mmap_writable counts all VM_SHARED, VM_MAYWRITE vmas: do_mmap
552 * marks vma as VM_SHARED if it is shared, and the file was opened for
553 * writing i.e. vma may be mprotected writable even if now readonly.
555 * If i_mmap_writable is negative, no new writable mappings are allowed. You
556 * can only deny writable mappings, if none exists right now.
558 static inline int mapping_writably_mapped(struct address_space *mapping)
560 return atomic_read(&mapping->i_mmap_writable) > 0;
563 static inline int mapping_map_writable(struct address_space *mapping)
565 return atomic_inc_unless_negative(&mapping->i_mmap_writable) ?
569 static inline void mapping_unmap_writable(struct address_space *mapping)
571 atomic_dec(&mapping->i_mmap_writable);
574 static inline int mapping_deny_writable(struct address_space *mapping)
576 return atomic_dec_unless_positive(&mapping->i_mmap_writable) ?
580 static inline void mapping_allow_writable(struct address_space *mapping)
582 atomic_inc(&mapping->i_mmap_writable);
586 * Use sequence counter to get consistent i_size on 32-bit processors.
588 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
589 #include <linux/seqlock.h>
590 #define __NEED_I_SIZE_ORDERED
591 #define i_size_ordered_init(inode) seqcount_init(&inode->i_size_seqcount)
593 #define i_size_ordered_init(inode) do { } while (0)
597 #define ACL_NOT_CACHED ((void *)(-1))
599 * ACL_DONT_CACHE is for stacked filesystems, that rely on underlying fs to
600 * cache the ACL. This also means that ->get_inode_acl() can be called in RCU
601 * mode with the LOOKUP_RCU flag.
603 #define ACL_DONT_CACHE ((void *)(-3))
605 static inline struct posix_acl *
606 uncached_acl_sentinel(struct task_struct *task)
608 return (void *)task + 1;
612 is_uncached_acl(struct posix_acl *acl)
614 return (long)acl & 1;
617 #define IOP_FASTPERM 0x0001
618 #define IOP_LOOKUP 0x0002
619 #define IOP_NOFOLLOW 0x0004
620 #define IOP_XATTR 0x0008
621 #define IOP_DEFAULT_READLINK 0x0010
623 struct fsnotify_mark_connector;
626 * Keep mostly read-only and often accessed (especially for
627 * the RCU path lookup and 'stat' data) fields at the beginning
628 * of the 'struct inode'
632 unsigned short i_opflags;
635 unsigned int i_flags;
637 #ifdef CONFIG_FS_POSIX_ACL
638 struct posix_acl *i_acl;
639 struct posix_acl *i_default_acl;
642 const struct inode_operations *i_op;
643 struct super_block *i_sb;
644 struct address_space *i_mapping;
646 #ifdef CONFIG_SECURITY
650 /* Stat data, not accessed from path walking */
653 * Filesystems may only read i_nlink directly. They shall use the
654 * following functions for modification:
656 * (set|clear|inc|drop)_nlink
657 * inode_(inc|dec)_link_count
660 const unsigned int i_nlink;
661 unsigned int __i_nlink;
665 struct timespec64 __i_atime;
666 struct timespec64 __i_mtime;
667 struct timespec64 __i_ctime; /* use inode_*_ctime accessors! */
668 spinlock_t i_lock; /* i_blocks, i_bytes, maybe i_size */
669 unsigned short i_bytes;
671 enum rw_hint i_write_hint;
674 #ifdef __NEED_I_SIZE_ORDERED
675 seqcount_t i_size_seqcount;
679 unsigned long i_state;
680 struct rw_semaphore i_rwsem;
682 unsigned long dirtied_when; /* jiffies of first dirtying */
683 unsigned long dirtied_time_when;
685 struct hlist_node i_hash;
686 struct list_head i_io_list; /* backing dev IO list */
687 #ifdef CONFIG_CGROUP_WRITEBACK
688 struct bdi_writeback *i_wb; /* the associated cgroup wb */
690 /* foreign inode detection, see wbc_detach_inode() */
692 u16 i_wb_frn_avg_time;
693 u16 i_wb_frn_history;
695 struct list_head i_lru; /* inode LRU list */
696 struct list_head i_sb_list;
697 struct list_head i_wb_list; /* backing dev writeback list */
699 struct hlist_head i_dentry;
700 struct rcu_head i_rcu;
702 atomic64_t i_version;
703 atomic64_t i_sequence; /* see futex */
705 atomic_t i_dio_count;
706 atomic_t i_writecount;
707 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
708 atomic_t i_readcount; /* struct files open RO */
711 const struct file_operations *i_fop; /* former ->i_op->default_file_ops */
712 void (*free_inode)(struct inode *);
714 struct file_lock_context *i_flctx;
715 struct address_space i_data;
716 struct list_head i_devices;
718 struct pipe_inode_info *i_pipe;
726 #ifdef CONFIG_FSNOTIFY
727 __u32 i_fsnotify_mask; /* all events this inode cares about */
728 struct fsnotify_mark_connector __rcu *i_fsnotify_marks;
731 #ifdef CONFIG_FS_ENCRYPTION
732 struct fscrypt_inode_info *i_crypt_info;
735 #ifdef CONFIG_FS_VERITY
736 struct fsverity_info *i_verity_info;
739 void *i_private; /* fs or device private pointer */
740 } __randomize_layout;
742 struct timespec64 timestamp_truncate(struct timespec64 t, struct inode *inode);
744 static inline unsigned int i_blocksize(const struct inode *node)
746 return (1 << node->i_blkbits);
749 static inline int inode_unhashed(struct inode *inode)
751 return hlist_unhashed(&inode->i_hash);
755 * __mark_inode_dirty expects inodes to be hashed. Since we don't
756 * want special inodes in the fileset inode space, we make them
757 * appear hashed, but do not put on any lists. hlist_del()
758 * will work fine and require no locking.
760 static inline void inode_fake_hash(struct inode *inode)
762 hlist_add_fake(&inode->i_hash);
766 * inode->i_mutex nesting subclasses for the lock validator:
768 * 0: the object of the current VFS operation
772 * 4: second non-directory
773 * 5: second parent (when locking independent directories in rename)
775 * I_MUTEX_NONDIR2 is for certain operations (such as rename) which lock two
776 * non-directories at once.
778 * The locking order between these classes is
779 * parent[2] -> child -> grandchild -> normal -> xattr -> second non-directory
781 enum inode_i_mutex_lock_class
791 static inline void inode_lock(struct inode *inode)
793 down_write(&inode->i_rwsem);
796 static inline void inode_unlock(struct inode *inode)
798 up_write(&inode->i_rwsem);
801 static inline void inode_lock_shared(struct inode *inode)
803 down_read(&inode->i_rwsem);
806 static inline void inode_unlock_shared(struct inode *inode)
808 up_read(&inode->i_rwsem);
811 static inline int inode_trylock(struct inode *inode)
813 return down_write_trylock(&inode->i_rwsem);
816 static inline int inode_trylock_shared(struct inode *inode)
818 return down_read_trylock(&inode->i_rwsem);
821 static inline int inode_is_locked(struct inode *inode)
823 return rwsem_is_locked(&inode->i_rwsem);
826 static inline void inode_lock_nested(struct inode *inode, unsigned subclass)
828 down_write_nested(&inode->i_rwsem, subclass);
831 static inline void inode_lock_shared_nested(struct inode *inode, unsigned subclass)
833 down_read_nested(&inode->i_rwsem, subclass);
836 static inline void filemap_invalidate_lock(struct address_space *mapping)
838 down_write(&mapping->invalidate_lock);
841 static inline void filemap_invalidate_unlock(struct address_space *mapping)
843 up_write(&mapping->invalidate_lock);
846 static inline void filemap_invalidate_lock_shared(struct address_space *mapping)
848 down_read(&mapping->invalidate_lock);
851 static inline int filemap_invalidate_trylock_shared(
852 struct address_space *mapping)
854 return down_read_trylock(&mapping->invalidate_lock);
857 static inline void filemap_invalidate_unlock_shared(
858 struct address_space *mapping)
860 up_read(&mapping->invalidate_lock);
863 void lock_two_nondirectories(struct inode *, struct inode*);
864 void unlock_two_nondirectories(struct inode *, struct inode*);
866 void filemap_invalidate_lock_two(struct address_space *mapping1,
867 struct address_space *mapping2);
868 void filemap_invalidate_unlock_two(struct address_space *mapping1,
869 struct address_space *mapping2);
873 * NOTE: in a 32bit arch with a preemptable kernel and
874 * an UP compile the i_size_read/write must be atomic
875 * with respect to the local cpu (unlike with preempt disabled),
876 * but they don't need to be atomic with respect to other cpus like in
877 * true SMP (so they need either to either locally disable irq around
878 * the read or for example on x86 they can be still implemented as a
879 * cmpxchg8b without the need of the lock prefix). For SMP compiles
880 * and 64bit archs it makes no difference if preempt is enabled or not.
882 static inline loff_t i_size_read(const struct inode *inode)
884 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
889 seq = read_seqcount_begin(&inode->i_size_seqcount);
890 i_size = inode->i_size;
891 } while (read_seqcount_retry(&inode->i_size_seqcount, seq));
893 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
897 i_size = inode->i_size;
901 /* Pairs with smp_store_release() in i_size_write() */
902 return smp_load_acquire(&inode->i_size);
907 * NOTE: unlike i_size_read(), i_size_write() does need locking around it
908 * (normally i_mutex), otherwise on 32bit/SMP an update of i_size_seqcount
909 * can be lost, resulting in subsequent i_size_read() calls spinning forever.
911 static inline void i_size_write(struct inode *inode, loff_t i_size)
913 #if BITS_PER_LONG==32 && defined(CONFIG_SMP)
915 write_seqcount_begin(&inode->i_size_seqcount);
916 inode->i_size = i_size;
917 write_seqcount_end(&inode->i_size_seqcount);
919 #elif BITS_PER_LONG==32 && defined(CONFIG_PREEMPTION)
921 inode->i_size = i_size;
925 * Pairs with smp_load_acquire() in i_size_read() to ensure
926 * changes related to inode size (such as page contents) are
927 * visible before we see the changed inode size.
929 smp_store_release(&inode->i_size, i_size);
933 static inline unsigned iminor(const struct inode *inode)
935 return MINOR(inode->i_rdev);
938 static inline unsigned imajor(const struct inode *inode)
940 return MAJOR(inode->i_rdev);
944 rwlock_t lock; /* protects pid, uid, euid fields */
945 struct pid *pid; /* pid or -pgrp where SIGIO should be sent */
946 enum pid_type pid_type; /* Kind of process group SIGIO should be sent to */
947 kuid_t uid, euid; /* uid/euid of process setting the owner */
948 int signum; /* posix.1b rt signal to be delivered on IO */
952 * struct file_ra_state - Track a file's readahead state.
953 * @start: Where the most recent readahead started.
954 * @size: Number of pages read in the most recent readahead.
955 * @async_size: Numer of pages that were/are not needed immediately
956 * and so were/are genuinely "ahead". Start next readahead when
957 * the first of these pages is accessed.
958 * @ra_pages: Maximum size of a readahead request, copied from the bdi.
959 * @mmap_miss: How many mmap accesses missed in the page cache.
960 * @prev_pos: The last byte in the most recent read request.
962 * When this structure is passed to ->readahead(), the "most recent"
963 * readahead means the current readahead.
965 struct file_ra_state {
968 unsigned int async_size;
969 unsigned int ra_pages;
970 unsigned int mmap_miss;
975 * Check if @index falls in the readahead windows.
977 static inline int ra_has_index(struct file_ra_state *ra, pgoff_t index)
979 return (index >= ra->start &&
980 index < ra->start + ra->size);
984 * f_{lock,count,pos_lock} members can be highly contended and share
985 * the same cacheline. f_{lock,mode} are very frequently used together
986 * and so share the same cacheline as well. The read-mostly
987 * f_{path,inode,op} are kept on a separate cacheline.
991 /* fput() uses task work when closing and freeing file (default). */
992 struct callback_head f_task_work;
993 /* fput() must use workqueue (most kernel threads). */
994 struct llist_node f_llist;
995 unsigned int f_iocb_flags;
999 * Protects f_ep, f_flags.
1000 * Must not be taken from IRQ context.
1004 atomic_long_t f_count;
1005 struct mutex f_pos_lock;
1007 unsigned int f_flags;
1008 struct fown_struct f_owner;
1009 const struct cred *f_cred;
1010 struct file_ra_state f_ra;
1012 struct inode *f_inode; /* cached value */
1013 const struct file_operations *f_op;
1016 #ifdef CONFIG_SECURITY
1019 /* needed for tty driver, and maybe others */
1023 /* Used by fs/eventpoll.c to link all the hooks to this file */
1024 struct hlist_head *f_ep;
1025 #endif /* #ifdef CONFIG_EPOLL */
1026 struct address_space *f_mapping;
1028 errseq_t f_sb_err; /* for syncfs */
1029 } __randomize_layout
1030 __attribute__((aligned(4))); /* lest something weird decides that 2 is OK */
1032 struct file_handle {
1035 /* file identifier */
1036 unsigned char f_handle[];
1039 static inline struct file *get_file(struct file *f)
1041 atomic_long_inc(&f->f_count);
1045 struct file *get_file_rcu(struct file __rcu **f);
1046 struct file *get_file_active(struct file **f);
1048 #define file_count(x) atomic_long_read(&(x)->f_count)
1050 #define MAX_NON_LFS ((1UL<<31) - 1)
1052 /* Page cache limit. The filesystems should put that into their s_maxbytes
1053 limits, otherwise bad things can happen in VM. */
1054 #if BITS_PER_LONG==32
1055 #define MAX_LFS_FILESIZE ((loff_t)ULONG_MAX << PAGE_SHIFT)
1056 #elif BITS_PER_LONG==64
1057 #define MAX_LFS_FILESIZE ((loff_t)LLONG_MAX)
1060 /* legacy typedef, should eventually be removed */
1061 typedef void *fl_owner_t;
1066 /* The following constant reflects the upper bound of the file/locking space */
1068 #define OFFSET_MAX type_max(loff_t)
1069 #define OFFT_OFFSET_MAX type_max(off_t)
1072 extern void send_sigio(struct fown_struct *fown, int fd, int band);
1074 static inline struct inode *file_inode(const struct file *f)
1080 * file_dentry() is a relic from the days that overlayfs was using files with a
1081 * "fake" path, meaning, f_path on overlayfs and f_inode on underlying fs.
1082 * In those days, file_dentry() was needed to get the underlying fs dentry that
1084 * Files with "fake" path should not exist nowadays, so use an assertion to make
1085 * sure that file_dentry() was not papering over filesystem bugs.
1087 static inline struct dentry *file_dentry(const struct file *file)
1089 struct dentry *dentry = file->f_path.dentry;
1091 WARN_ON_ONCE(d_inode(dentry) != file_inode(file));
1095 struct fasync_struct {
1099 struct fasync_struct *fa_next; /* singly linked list */
1100 struct file *fa_file;
1101 struct rcu_head fa_rcu;
1104 #define FASYNC_MAGIC 0x4601
1106 /* SMP safe fasync helpers: */
1107 extern int fasync_helper(int, struct file *, int, struct fasync_struct **);
1108 extern struct fasync_struct *fasync_insert_entry(int, struct file *, struct fasync_struct **, struct fasync_struct *);
1109 extern int fasync_remove_entry(struct file *, struct fasync_struct **);
1110 extern struct fasync_struct *fasync_alloc(void);
1111 extern void fasync_free(struct fasync_struct *);
1113 /* can be called from interrupts */
1114 extern void kill_fasync(struct fasync_struct **, int, int);
1116 extern void __f_setown(struct file *filp, struct pid *, enum pid_type, int force);
1117 extern int f_setown(struct file *filp, int who, int force);
1118 extern void f_delown(struct file *filp);
1119 extern pid_t f_getown(struct file *filp);
1120 extern int send_sigurg(struct fown_struct *fown);
1123 * sb->s_flags. Note that these mirror the equivalent MS_* flags where
1124 * represented in both.
1126 #define SB_RDONLY BIT(0) /* Mount read-only */
1127 #define SB_NOSUID BIT(1) /* Ignore suid and sgid bits */
1128 #define SB_NODEV BIT(2) /* Disallow access to device special files */
1129 #define SB_NOEXEC BIT(3) /* Disallow program execution */
1130 #define SB_SYNCHRONOUS BIT(4) /* Writes are synced at once */
1131 #define SB_MANDLOCK BIT(6) /* Allow mandatory locks on an FS */
1132 #define SB_DIRSYNC BIT(7) /* Directory modifications are synchronous */
1133 #define SB_NOATIME BIT(10) /* Do not update access times. */
1134 #define SB_NODIRATIME BIT(11) /* Do not update directory access times */
1135 #define SB_SILENT BIT(15)
1136 #define SB_POSIXACL BIT(16) /* Supports POSIX ACLs */
1137 #define SB_INLINECRYPT BIT(17) /* Use blk-crypto for encrypted files */
1138 #define SB_KERNMOUNT BIT(22) /* this is a kern_mount call */
1139 #define SB_I_VERSION BIT(23) /* Update inode I_version field */
1140 #define SB_LAZYTIME BIT(25) /* Update the on-disk [acm]times lazily */
1142 /* These sb flags are internal to the kernel */
1143 #define SB_DEAD BIT(21)
1144 #define SB_DYING BIT(24)
1145 #define SB_SUBMOUNT BIT(26)
1146 #define SB_FORCE BIT(27)
1147 #define SB_NOSEC BIT(28)
1148 #define SB_BORN BIT(29)
1149 #define SB_ACTIVE BIT(30)
1150 #define SB_NOUSER BIT(31)
1152 /* These flags relate to encoding and casefolding */
1153 #define SB_ENC_STRICT_MODE_FL (1 << 0)
1155 #define sb_has_strict_encoding(sb) \
1156 (sb->s_encoding_flags & SB_ENC_STRICT_MODE_FL)
1162 #define MNT_FORCE 0x00000001 /* Attempt to forcibily umount */
1163 #define MNT_DETACH 0x00000002 /* Just detach from the tree */
1164 #define MNT_EXPIRE 0x00000004 /* Mark for expiry */
1165 #define UMOUNT_NOFOLLOW 0x00000008 /* Don't follow symlink on umount */
1166 #define UMOUNT_UNUSED 0x80000000 /* Flag guaranteed to be unused */
1169 #define SB_I_CGROUPWB 0x00000001 /* cgroup-aware writeback enabled */
1170 #define SB_I_NOEXEC 0x00000002 /* Ignore executables on this fs */
1171 #define SB_I_NODEV 0x00000004 /* Ignore devices on this fs */
1172 #define SB_I_STABLE_WRITES 0x00000008 /* don't modify blks until WB is done */
1174 /* sb->s_iflags to limit user namespace mounts */
1175 #define SB_I_USERNS_VISIBLE 0x00000010 /* fstype already mounted */
1176 #define SB_I_IMA_UNVERIFIABLE_SIGNATURE 0x00000020
1177 #define SB_I_UNTRUSTED_MOUNTER 0x00000040
1178 #define SB_I_EVM_UNSUPPORTED 0x00000080
1180 #define SB_I_SKIP_SYNC 0x00000100 /* Skip superblock at global sync */
1181 #define SB_I_PERSB_BDI 0x00000200 /* has a per-sb bdi */
1182 #define SB_I_TS_EXPIRY_WARNED 0x00000400 /* warned about timestamp range expiry */
1183 #define SB_I_RETIRED 0x00000800 /* superblock shouldn't be reused */
1184 #define SB_I_NOUMASK 0x00001000 /* VFS does not apply umask */
1186 /* Possible states of 'frozen' field */
1188 SB_UNFROZEN = 0, /* FS is unfrozen */
1189 SB_FREEZE_WRITE = 1, /* Writes, dir ops, ioctls frozen */
1190 SB_FREEZE_PAGEFAULT = 2, /* Page faults stopped as well */
1191 SB_FREEZE_FS = 3, /* For internal FS use (e.g. to stop
1192 * internal threads if needed) */
1193 SB_FREEZE_COMPLETE = 4, /* ->freeze_fs finished successfully */
1196 #define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
1199 unsigned short frozen; /* Is sb frozen? */
1200 int freeze_kcount; /* How many kernel freeze requests? */
1201 int freeze_ucount; /* How many userspace freeze requests? */
1202 struct percpu_rw_semaphore rw_sem[SB_FREEZE_LEVELS];
1205 struct super_block {
1206 struct list_head s_list; /* Keep this first */
1207 dev_t s_dev; /* search index; _not_ kdev_t */
1208 unsigned char s_blocksize_bits;
1209 unsigned long s_blocksize;
1210 loff_t s_maxbytes; /* Max file size */
1211 struct file_system_type *s_type;
1212 const struct super_operations *s_op;
1213 const struct dquot_operations *dq_op;
1214 const struct quotactl_ops *s_qcop;
1215 const struct export_operations *s_export_op;
1216 unsigned long s_flags;
1217 unsigned long s_iflags; /* internal SB_I_* flags */
1218 unsigned long s_magic;
1219 struct dentry *s_root;
1220 struct rw_semaphore s_umount;
1223 #ifdef CONFIG_SECURITY
1226 const struct xattr_handler * const *s_xattr;
1227 #ifdef CONFIG_FS_ENCRYPTION
1228 const struct fscrypt_operations *s_cop;
1229 struct fscrypt_keyring *s_master_keys; /* master crypto keys in use */
1231 #ifdef CONFIG_FS_VERITY
1232 const struct fsverity_operations *s_vop;
1234 #if IS_ENABLED(CONFIG_UNICODE)
1235 struct unicode_map *s_encoding;
1236 __u16 s_encoding_flags;
1238 struct hlist_bl_head s_roots; /* alternate root dentries for NFS */
1239 struct list_head s_mounts; /* list of mounts; _not_ for fs use */
1240 struct block_device *s_bdev; /* can go away once we use an accessor for @s_bdev_file */
1241 struct file *s_bdev_file;
1242 struct backing_dev_info *s_bdi;
1243 struct mtd_info *s_mtd;
1244 struct hlist_node s_instances;
1245 unsigned int s_quota_types; /* Bitmask of supported quota types */
1246 struct quota_info s_dquot; /* Diskquota specific options */
1248 struct sb_writers s_writers;
1251 * Keep s_fs_info, s_time_gran, s_fsnotify_mask, and
1252 * s_fsnotify_marks together for cache efficiency. They are frequently
1253 * accessed and rarely modified.
1255 void *s_fs_info; /* Filesystem private info */
1257 /* Granularity of c/m/atime in ns (cannot be worse than a second) */
1259 /* Time limits for c/m/atime in seconds */
1260 time64_t s_time_min;
1261 time64_t s_time_max;
1262 #ifdef CONFIG_FSNOTIFY
1263 __u32 s_fsnotify_mask;
1264 struct fsnotify_mark_connector __rcu *s_fsnotify_marks;
1268 * q: why are s_id and s_sysfs_name not the same? both are human
1269 * readable strings that identify the filesystem
1270 * a: s_id is allowed to change at runtime; it's used in log messages,
1271 * and we want to when a device starts out as single device (s_id is dev
1272 * name) but then a device is hot added and we have to switch to
1273 * identifying it by UUID
1274 * but s_sysfs_name is a handle for programmatic access, and can't
1277 char s_id[32]; /* Informational name */
1278 uuid_t s_uuid; /* UUID */
1279 u8 s_uuid_len; /* Default 16, possibly smaller for weird filesystems */
1281 /* if set, fs shows up under sysfs at /sys/fs/$FSTYP/s_sysfs_name */
1282 char s_sysfs_name[UUID_STRING_LEN + 1];
1284 unsigned int s_max_links;
1287 * The next field is for VFS *only*. No filesystems have any business
1288 * even looking at it. You had been warned.
1290 struct mutex s_vfs_rename_mutex; /* Kludge */
1293 * Filesystem subtype. If non-empty the filesystem type field
1294 * in /proc/mounts will be "type.subtype"
1296 const char *s_subtype;
1298 const struct dentry_operations *s_d_op; /* default d_op for dentries */
1300 struct shrinker *s_shrink; /* per-sb shrinker handle */
1302 /* Number of inodes with nlink == 0 but still referenced */
1303 atomic_long_t s_remove_count;
1306 * Number of inode/mount/sb objects that are being watched, note that
1307 * inodes objects are currently double-accounted.
1309 atomic_long_t s_fsnotify_connectors;
1311 /* Read-only state of the superblock is being changed */
1312 int s_readonly_remount;
1314 /* per-sb errseq_t for reporting writeback errors via syncfs */
1317 /* AIO completions deferred from interrupt context */
1318 struct workqueue_struct *s_dio_done_wq;
1319 struct hlist_head s_pins;
1322 * Owning user namespace and default context in which to
1323 * interpret filesystem uids, gids, quotas, device nodes,
1324 * xattrs and security labels.
1326 struct user_namespace *s_user_ns;
1329 * The list_lru structure is essentially just a pointer to a table
1330 * of per-node lru lists, each of which has its own spinlock.
1331 * There is no need to put them into separate cachelines.
1333 struct list_lru s_dentry_lru;
1334 struct list_lru s_inode_lru;
1335 struct rcu_head rcu;
1336 struct work_struct destroy_work;
1338 struct mutex s_sync_lock; /* sync serialisation lock */
1341 * Indicates how deep in a filesystem stack this SB is
1345 /* s_inode_list_lock protects s_inodes */
1346 spinlock_t s_inode_list_lock ____cacheline_aligned_in_smp;
1347 struct list_head s_inodes; /* all inodes */
1349 spinlock_t s_inode_wblist_lock;
1350 struct list_head s_inodes_wb; /* writeback inodes */
1351 } __randomize_layout;
1353 static inline struct user_namespace *i_user_ns(const struct inode *inode)
1355 return inode->i_sb->s_user_ns;
1358 /* Helper functions so that in most cases filesystems will
1359 * not need to deal directly with kuid_t and kgid_t and can
1360 * instead deal with the raw numeric values that are stored
1361 * in the filesystem.
1363 static inline uid_t i_uid_read(const struct inode *inode)
1365 return from_kuid(i_user_ns(inode), inode->i_uid);
1368 static inline gid_t i_gid_read(const struct inode *inode)
1370 return from_kgid(i_user_ns(inode), inode->i_gid);
1373 static inline void i_uid_write(struct inode *inode, uid_t uid)
1375 inode->i_uid = make_kuid(i_user_ns(inode), uid);
1378 static inline void i_gid_write(struct inode *inode, gid_t gid)
1380 inode->i_gid = make_kgid(i_user_ns(inode), gid);
1384 * i_uid_into_vfsuid - map an inode's i_uid down according to an idmapping
1385 * @idmap: idmap of the mount the inode was found from
1386 * @inode: inode to map
1388 * Return: whe inode's i_uid mapped down according to @idmap.
1389 * If the inode's i_uid has no mapping INVALID_VFSUID is returned.
1391 static inline vfsuid_t i_uid_into_vfsuid(struct mnt_idmap *idmap,
1392 const struct inode *inode)
1394 return make_vfsuid(idmap, i_user_ns(inode), inode->i_uid);
1398 * i_uid_needs_update - check whether inode's i_uid needs to be updated
1399 * @idmap: idmap of the mount the inode was found from
1400 * @attr: the new attributes of @inode
1401 * @inode: the inode to update
1403 * Check whether the $inode's i_uid field needs to be updated taking idmapped
1404 * mounts into account if the filesystem supports it.
1406 * Return: true if @inode's i_uid field needs to be updated, false if not.
1408 static inline bool i_uid_needs_update(struct mnt_idmap *idmap,
1409 const struct iattr *attr,
1410 const struct inode *inode)
1412 return ((attr->ia_valid & ATTR_UID) &&
1413 !vfsuid_eq(attr->ia_vfsuid,
1414 i_uid_into_vfsuid(idmap, inode)));
1418 * i_uid_update - update @inode's i_uid field
1419 * @idmap: idmap of the mount the inode was found from
1420 * @attr: the new attributes of @inode
1421 * @inode: the inode to update
1423 * Safely update @inode's i_uid field translating the vfsuid of any idmapped
1424 * mount into the filesystem kuid.
1426 static inline void i_uid_update(struct mnt_idmap *idmap,
1427 const struct iattr *attr,
1428 struct inode *inode)
1430 if (attr->ia_valid & ATTR_UID)
1431 inode->i_uid = from_vfsuid(idmap, i_user_ns(inode),
1436 * i_gid_into_vfsgid - map an inode's i_gid down according to an idmapping
1437 * @idmap: idmap of the mount the inode was found from
1438 * @inode: inode to map
1440 * Return: the inode's i_gid mapped down according to @idmap.
1441 * If the inode's i_gid has no mapping INVALID_VFSGID is returned.
1443 static inline vfsgid_t i_gid_into_vfsgid(struct mnt_idmap *idmap,
1444 const struct inode *inode)
1446 return make_vfsgid(idmap, i_user_ns(inode), inode->i_gid);
1450 * i_gid_needs_update - check whether inode's i_gid needs to be updated
1451 * @idmap: idmap of the mount the inode was found from
1452 * @attr: the new attributes of @inode
1453 * @inode: the inode to update
1455 * Check whether the $inode's i_gid field needs to be updated taking idmapped
1456 * mounts into account if the filesystem supports it.
1458 * Return: true if @inode's i_gid field needs to be updated, false if not.
1460 static inline bool i_gid_needs_update(struct mnt_idmap *idmap,
1461 const struct iattr *attr,
1462 const struct inode *inode)
1464 return ((attr->ia_valid & ATTR_GID) &&
1465 !vfsgid_eq(attr->ia_vfsgid,
1466 i_gid_into_vfsgid(idmap, inode)));
1470 * i_gid_update - update @inode's i_gid field
1471 * @idmap: idmap of the mount the inode was found from
1472 * @attr: the new attributes of @inode
1473 * @inode: the inode to update
1475 * Safely update @inode's i_gid field translating the vfsgid of any idmapped
1476 * mount into the filesystem kgid.
1478 static inline void i_gid_update(struct mnt_idmap *idmap,
1479 const struct iattr *attr,
1480 struct inode *inode)
1482 if (attr->ia_valid & ATTR_GID)
1483 inode->i_gid = from_vfsgid(idmap, i_user_ns(inode),
1488 * inode_fsuid_set - initialize inode's i_uid field with callers fsuid
1489 * @inode: inode to initialize
1490 * @idmap: idmap of the mount the inode was found from
1492 * Initialize the i_uid field of @inode. If the inode was found/created via
1493 * an idmapped mount map the caller's fsuid according to @idmap.
1495 static inline void inode_fsuid_set(struct inode *inode,
1496 struct mnt_idmap *idmap)
1498 inode->i_uid = mapped_fsuid(idmap, i_user_ns(inode));
1502 * inode_fsgid_set - initialize inode's i_gid field with callers fsgid
1503 * @inode: inode to initialize
1504 * @idmap: idmap of the mount the inode was found from
1506 * Initialize the i_gid field of @inode. If the inode was found/created via
1507 * an idmapped mount map the caller's fsgid according to @idmap.
1509 static inline void inode_fsgid_set(struct inode *inode,
1510 struct mnt_idmap *idmap)
1512 inode->i_gid = mapped_fsgid(idmap, i_user_ns(inode));
1516 * fsuidgid_has_mapping() - check whether caller's fsuid/fsgid is mapped
1517 * @sb: the superblock we want a mapping in
1518 * @idmap: idmap of the relevant mount
1520 * Check whether the caller's fsuid and fsgid have a valid mapping in the
1521 * s_user_ns of the superblock @sb. If the caller is on an idmapped mount map
1522 * the caller's fsuid and fsgid according to the @idmap first.
1524 * Return: true if fsuid and fsgid is mapped, false if not.
1526 static inline bool fsuidgid_has_mapping(struct super_block *sb,
1527 struct mnt_idmap *idmap)
1529 struct user_namespace *fs_userns = sb->s_user_ns;
1533 kuid = mapped_fsuid(idmap, fs_userns);
1534 if (!uid_valid(kuid))
1536 kgid = mapped_fsgid(idmap, fs_userns);
1537 if (!gid_valid(kgid))
1539 return kuid_has_mapping(fs_userns, kuid) &&
1540 kgid_has_mapping(fs_userns, kgid);
1543 struct timespec64 current_time(struct inode *inode);
1544 struct timespec64 inode_set_ctime_current(struct inode *inode);
1546 static inline time64_t inode_get_atime_sec(const struct inode *inode)
1548 return inode->__i_atime.tv_sec;
1551 static inline long inode_get_atime_nsec(const struct inode *inode)
1553 return inode->__i_atime.tv_nsec;
1556 static inline struct timespec64 inode_get_atime(const struct inode *inode)
1558 return inode->__i_atime;
1561 static inline struct timespec64 inode_set_atime_to_ts(struct inode *inode,
1562 struct timespec64 ts)
1564 inode->__i_atime = ts;
1568 static inline struct timespec64 inode_set_atime(struct inode *inode,
1569 time64_t sec, long nsec)
1571 struct timespec64 ts = { .tv_sec = sec,
1573 return inode_set_atime_to_ts(inode, ts);
1576 static inline time64_t inode_get_mtime_sec(const struct inode *inode)
1578 return inode->__i_mtime.tv_sec;
1581 static inline long inode_get_mtime_nsec(const struct inode *inode)
1583 return inode->__i_mtime.tv_nsec;
1586 static inline struct timespec64 inode_get_mtime(const struct inode *inode)
1588 return inode->__i_mtime;
1591 static inline struct timespec64 inode_set_mtime_to_ts(struct inode *inode,
1592 struct timespec64 ts)
1594 inode->__i_mtime = ts;
1598 static inline struct timespec64 inode_set_mtime(struct inode *inode,
1599 time64_t sec, long nsec)
1601 struct timespec64 ts = { .tv_sec = sec,
1603 return inode_set_mtime_to_ts(inode, ts);
1606 static inline time64_t inode_get_ctime_sec(const struct inode *inode)
1608 return inode->__i_ctime.tv_sec;
1611 static inline long inode_get_ctime_nsec(const struct inode *inode)
1613 return inode->__i_ctime.tv_nsec;
1616 static inline struct timespec64 inode_get_ctime(const struct inode *inode)
1618 return inode->__i_ctime;
1621 static inline struct timespec64 inode_set_ctime_to_ts(struct inode *inode,
1622 struct timespec64 ts)
1624 inode->__i_ctime = ts;
1629 * inode_set_ctime - set the ctime in the inode
1630 * @inode: inode in which to set the ctime
1631 * @sec: tv_sec value to set
1632 * @nsec: tv_nsec value to set
1634 * Set the ctime in @inode to { @sec, @nsec }
1636 static inline struct timespec64 inode_set_ctime(struct inode *inode,
1637 time64_t sec, long nsec)
1639 struct timespec64 ts = { .tv_sec = sec,
1642 return inode_set_ctime_to_ts(inode, ts);
1645 struct timespec64 simple_inode_init_ts(struct inode *inode);
1648 * Snapshotting support.
1652 * These are internal functions, please use sb_start_{write,pagefault,intwrite}
1655 static inline void __sb_end_write(struct super_block *sb, int level)
1657 percpu_up_read(sb->s_writers.rw_sem + level-1);
1660 static inline void __sb_start_write(struct super_block *sb, int level)
1662 percpu_down_read(sb->s_writers.rw_sem + level - 1);
1665 static inline bool __sb_start_write_trylock(struct super_block *sb, int level)
1667 return percpu_down_read_trylock(sb->s_writers.rw_sem + level - 1);
1670 #define __sb_writers_acquired(sb, lev) \
1671 percpu_rwsem_acquire(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1672 #define __sb_writers_release(sb, lev) \
1673 percpu_rwsem_release(&(sb)->s_writers.rw_sem[(lev)-1], 1, _THIS_IP_)
1676 * __sb_write_started - check if sb freeze level is held
1677 * @sb: the super we write to
1678 * @level: the freeze level
1680 * * > 0 - sb freeze level is held
1681 * * 0 - sb freeze level is not held
1682 * * < 0 - !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN
1684 static inline int __sb_write_started(const struct super_block *sb, int level)
1686 return lockdep_is_held_type(sb->s_writers.rw_sem + level - 1, 1);
1690 * sb_write_started - check if SB_FREEZE_WRITE is held
1691 * @sb: the super we write to
1693 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1695 static inline bool sb_write_started(const struct super_block *sb)
1697 return __sb_write_started(sb, SB_FREEZE_WRITE);
1701 * sb_write_not_started - check if SB_FREEZE_WRITE is not held
1702 * @sb: the super we write to
1704 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1706 static inline bool sb_write_not_started(const struct super_block *sb)
1708 return __sb_write_started(sb, SB_FREEZE_WRITE) <= 0;
1712 * file_write_started - check if SB_FREEZE_WRITE is held
1713 * @file: the file we write to
1715 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1716 * May be false positive with !S_ISREG, because file_start_write() has
1717 * no effect on !S_ISREG.
1719 static inline bool file_write_started(const struct file *file)
1721 if (!S_ISREG(file_inode(file)->i_mode))
1723 return sb_write_started(file_inode(file)->i_sb);
1727 * file_write_not_started - check if SB_FREEZE_WRITE is not held
1728 * @file: the file we write to
1730 * May be false positive with !CONFIG_LOCKDEP/LOCK_STATE_UNKNOWN.
1731 * May be false positive with !S_ISREG, because file_start_write() has
1732 * no effect on !S_ISREG.
1734 static inline bool file_write_not_started(const struct file *file)
1736 if (!S_ISREG(file_inode(file)->i_mode))
1738 return sb_write_not_started(file_inode(file)->i_sb);
1742 * sb_end_write - drop write access to a superblock
1743 * @sb: the super we wrote to
1745 * Decrement number of writers to the filesystem. Wake up possible waiters
1746 * wanting to freeze the filesystem.
1748 static inline void sb_end_write(struct super_block *sb)
1750 __sb_end_write(sb, SB_FREEZE_WRITE);
1754 * sb_end_pagefault - drop write access to a superblock from a page fault
1755 * @sb: the super we wrote to
1757 * Decrement number of processes handling write page fault to the filesystem.
1758 * Wake up possible waiters wanting to freeze the filesystem.
1760 static inline void sb_end_pagefault(struct super_block *sb)
1762 __sb_end_write(sb, SB_FREEZE_PAGEFAULT);
1766 * sb_end_intwrite - drop write access to a superblock for internal fs purposes
1767 * @sb: the super we wrote to
1769 * Decrement fs-internal number of writers to the filesystem. Wake up possible
1770 * waiters wanting to freeze the filesystem.
1772 static inline void sb_end_intwrite(struct super_block *sb)
1774 __sb_end_write(sb, SB_FREEZE_FS);
1778 * sb_start_write - get write access to a superblock
1779 * @sb: the super we write to
1781 * When a process wants to write data or metadata to a file system (i.e. dirty
1782 * a page or an inode), it should embed the operation in a sb_start_write() -
1783 * sb_end_write() pair to get exclusion against file system freezing. This
1784 * function increments number of writers preventing freezing. If the file
1785 * system is already frozen, the function waits until the file system is
1788 * Since freeze protection behaves as a lock, users have to preserve
1789 * ordering of freeze protection and other filesystem locks. Generally,
1790 * freeze protection should be the outermost lock. In particular, we have:
1793 * -> i_mutex (write path, truncate, directory ops, ...)
1794 * -> s_umount (freeze_super, thaw_super)
1796 static inline void sb_start_write(struct super_block *sb)
1798 __sb_start_write(sb, SB_FREEZE_WRITE);
1801 static inline bool sb_start_write_trylock(struct super_block *sb)
1803 return __sb_start_write_trylock(sb, SB_FREEZE_WRITE);
1807 * sb_start_pagefault - get write access to a superblock from a page fault
1808 * @sb: the super we write to
1810 * When a process starts handling write page fault, it should embed the
1811 * operation into sb_start_pagefault() - sb_end_pagefault() pair to get
1812 * exclusion against file system freezing. This is needed since the page fault
1813 * is going to dirty a page. This function increments number of running page
1814 * faults preventing freezing. If the file system is already frozen, the
1815 * function waits until the file system is thawed.
1817 * Since page fault freeze protection behaves as a lock, users have to preserve
1818 * ordering of freeze protection and other filesystem locks. It is advised to
1819 * put sb_start_pagefault() close to mmap_lock in lock ordering. Page fault
1820 * handling code implies lock dependency:
1823 * -> sb_start_pagefault
1825 static inline void sb_start_pagefault(struct super_block *sb)
1827 __sb_start_write(sb, SB_FREEZE_PAGEFAULT);
1831 * sb_start_intwrite - get write access to a superblock for internal fs purposes
1832 * @sb: the super we write to
1834 * This is the third level of protection against filesystem freezing. It is
1835 * free for use by a filesystem. The only requirement is that it must rank
1836 * below sb_start_pagefault.
1838 * For example filesystem can call sb_start_intwrite() when starting a
1839 * transaction which somewhat eases handling of freezing for internal sources
1840 * of filesystem changes (internal fs threads, discarding preallocation on file
1843 static inline void sb_start_intwrite(struct super_block *sb)
1845 __sb_start_write(sb, SB_FREEZE_FS);
1848 static inline bool sb_start_intwrite_trylock(struct super_block *sb)
1850 return __sb_start_write_trylock(sb, SB_FREEZE_FS);
1853 bool inode_owner_or_capable(struct mnt_idmap *idmap,
1854 const struct inode *inode);
1857 * VFS helper functions..
1859 int vfs_create(struct mnt_idmap *, struct inode *,
1860 struct dentry *, umode_t, bool);
1861 int vfs_mkdir(struct mnt_idmap *, struct inode *,
1862 struct dentry *, umode_t);
1863 int vfs_mknod(struct mnt_idmap *, struct inode *, struct dentry *,
1865 int vfs_symlink(struct mnt_idmap *, struct inode *,
1866 struct dentry *, const char *);
1867 int vfs_link(struct dentry *, struct mnt_idmap *, struct inode *,
1868 struct dentry *, struct inode **);
1869 int vfs_rmdir(struct mnt_idmap *, struct inode *, struct dentry *);
1870 int vfs_unlink(struct mnt_idmap *, struct inode *, struct dentry *,
1874 * struct renamedata - contains all information required for renaming
1875 * @old_mnt_idmap: idmap of the old mount the inode was found from
1876 * @old_dir: parent of source
1877 * @old_dentry: source
1878 * @new_mnt_idmap: idmap of the new mount the inode was found from
1879 * @new_dir: parent of destination
1880 * @new_dentry: destination
1881 * @delegated_inode: returns an inode needing a delegation break
1882 * @flags: rename flags
1885 struct mnt_idmap *old_mnt_idmap;
1886 struct inode *old_dir;
1887 struct dentry *old_dentry;
1888 struct mnt_idmap *new_mnt_idmap;
1889 struct inode *new_dir;
1890 struct dentry *new_dentry;
1891 struct inode **delegated_inode;
1893 } __randomize_layout;
1895 int vfs_rename(struct renamedata *);
1897 static inline int vfs_whiteout(struct mnt_idmap *idmap,
1898 struct inode *dir, struct dentry *dentry)
1900 return vfs_mknod(idmap, dir, dentry, S_IFCHR | WHITEOUT_MODE,
1904 struct file *kernel_tmpfile_open(struct mnt_idmap *idmap,
1905 const struct path *parentpath,
1906 umode_t mode, int open_flag,
1907 const struct cred *cred);
1908 struct file *kernel_file_open(const struct path *path, int flags,
1909 struct inode *inode, const struct cred *cred);
1911 int vfs_mkobj(struct dentry *, umode_t,
1912 int (*f)(struct dentry *, umode_t, void *),
1915 int vfs_fchown(struct file *file, uid_t user, gid_t group);
1916 int vfs_fchmod(struct file *file, umode_t mode);
1917 int vfs_utimes(const struct path *path, struct timespec64 *times);
1919 extern long vfs_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
1921 #ifdef CONFIG_COMPAT
1922 extern long compat_ptr_ioctl(struct file *file, unsigned int cmd,
1925 #define compat_ptr_ioctl NULL
1929 * VFS file helper functions.
1931 void inode_init_owner(struct mnt_idmap *idmap, struct inode *inode,
1932 const struct inode *dir, umode_t mode);
1933 extern bool may_open_dev(const struct path *path);
1934 umode_t mode_strip_sgid(struct mnt_idmap *idmap,
1935 const struct inode *dir, umode_t mode);
1938 * This is the "filldir" function type, used by readdir() to let
1939 * the kernel specify what kind of dirent layout it wants to have.
1940 * This allows the kernel to read directories into kernel space or
1941 * to have different dirent layouts depending on the binary type.
1942 * Return 'true' to keep going and 'false' if there are no more entries.
1945 typedef bool (*filldir_t)(struct dir_context *, const char *, int, loff_t, u64,
1948 struct dir_context {
1954 * These flags let !MMU mmap() govern direct device mapping vs immediate
1955 * copying more easily for MAP_PRIVATE, especially for ROM filesystems.
1957 * NOMMU_MAP_COPY: Copy can be mapped (MAP_PRIVATE)
1958 * NOMMU_MAP_DIRECT: Can be mapped directly (MAP_SHARED)
1959 * NOMMU_MAP_READ: Can be mapped for reading
1960 * NOMMU_MAP_WRITE: Can be mapped for writing
1961 * NOMMU_MAP_EXEC: Can be mapped for execution
1963 #define NOMMU_MAP_COPY 0x00000001
1964 #define NOMMU_MAP_DIRECT 0x00000008
1965 #define NOMMU_MAP_READ VM_MAYREAD
1966 #define NOMMU_MAP_WRITE VM_MAYWRITE
1967 #define NOMMU_MAP_EXEC VM_MAYEXEC
1969 #define NOMMU_VMFLAGS \
1970 (NOMMU_MAP_READ | NOMMU_MAP_WRITE | NOMMU_MAP_EXEC)
1973 * These flags control the behavior of the remap_file_range function pointer.
1974 * If it is called with len == 0 that means "remap to end of source file".
1975 * See Documentation/filesystems/vfs.rst for more details about this call.
1977 * REMAP_FILE_DEDUP: only remap if contents identical (i.e. deduplicate)
1978 * REMAP_FILE_CAN_SHORTEN: caller can handle a shortened request
1980 #define REMAP_FILE_DEDUP (1 << 0)
1981 #define REMAP_FILE_CAN_SHORTEN (1 << 1)
1984 * These flags signal that the caller is ok with altering various aspects of
1985 * the behavior of the remap operation. The changes must be made by the
1986 * implementation; the vfs remap helper functions can take advantage of them.
1987 * Flags in this category exist to preserve the quirky behavior of the hoisted
1988 * btrfs clone/dedupe ioctls.
1990 #define REMAP_FILE_ADVISORY (REMAP_FILE_CAN_SHORTEN)
1993 * These flags control the behavior of vfs_copy_file_range().
1994 * They are not available to the user via syscall.
1996 * COPY_FILE_SPLICE: call splice direct instead of fs clone/copy ops
1998 #define COPY_FILE_SPLICE (1 << 0)
2001 struct io_uring_cmd;
2004 struct file_operations {
2005 struct module *owner;
2006 loff_t (*llseek) (struct file *, loff_t, int);
2007 ssize_t (*read) (struct file *, char __user *, size_t, loff_t *);
2008 ssize_t (*write) (struct file *, const char __user *, size_t, loff_t *);
2009 ssize_t (*read_iter) (struct kiocb *, struct iov_iter *);
2010 ssize_t (*write_iter) (struct kiocb *, struct iov_iter *);
2011 int (*iopoll)(struct kiocb *kiocb, struct io_comp_batch *,
2012 unsigned int flags);
2013 int (*iterate_shared) (struct file *, struct dir_context *);
2014 __poll_t (*poll) (struct file *, struct poll_table_struct *);
2015 long (*unlocked_ioctl) (struct file *, unsigned int, unsigned long);
2016 long (*compat_ioctl) (struct file *, unsigned int, unsigned long);
2017 int (*mmap) (struct file *, struct vm_area_struct *);
2018 unsigned long mmap_supported_flags;
2019 int (*open) (struct inode *, struct file *);
2020 int (*flush) (struct file *, fl_owner_t id);
2021 int (*release) (struct inode *, struct file *);
2022 int (*fsync) (struct file *, loff_t, loff_t, int datasync);
2023 int (*fasync) (int, struct file *, int);
2024 int (*lock) (struct file *, int, struct file_lock *);
2025 unsigned long (*get_unmapped_area)(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2026 int (*check_flags)(int);
2027 int (*flock) (struct file *, int, struct file_lock *);
2028 ssize_t (*splice_write)(struct pipe_inode_info *, struct file *, loff_t *, size_t, unsigned int);
2029 ssize_t (*splice_read)(struct file *, loff_t *, struct pipe_inode_info *, size_t, unsigned int);
2030 void (*splice_eof)(struct file *file);
2031 int (*setlease)(struct file *, int, struct file_lease **, void **);
2032 long (*fallocate)(struct file *file, int mode, loff_t offset,
2034 void (*show_fdinfo)(struct seq_file *m, struct file *f);
2036 unsigned (*mmap_capabilities)(struct file *);
2038 ssize_t (*copy_file_range)(struct file *, loff_t, struct file *,
2039 loff_t, size_t, unsigned int);
2040 loff_t (*remap_file_range)(struct file *file_in, loff_t pos_in,
2041 struct file *file_out, loff_t pos_out,
2042 loff_t len, unsigned int remap_flags);
2043 int (*fadvise)(struct file *, loff_t, loff_t, int);
2044 int (*uring_cmd)(struct io_uring_cmd *ioucmd, unsigned int issue_flags);
2045 int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *,
2046 unsigned int poll_flags);
2047 } __randomize_layout;
2049 /* Wrap a directory iterator that needs exclusive inode access */
2050 int wrap_directory_iterator(struct file *, struct dir_context *,
2051 int (*) (struct file *, struct dir_context *));
2052 #define WRAP_DIR_ITER(x) \
2053 static int shared_##x(struct file *file , struct dir_context *ctx) \
2054 { return wrap_directory_iterator(file, ctx, x); }
2056 struct inode_operations {
2057 struct dentry * (*lookup) (struct inode *,struct dentry *, unsigned int);
2058 const char * (*get_link) (struct dentry *, struct inode *, struct delayed_call *);
2059 int (*permission) (struct mnt_idmap *, struct inode *, int);
2060 struct posix_acl * (*get_inode_acl)(struct inode *, int, bool);
2062 int (*readlink) (struct dentry *, char __user *,int);
2064 int (*create) (struct mnt_idmap *, struct inode *,struct dentry *,
2066 int (*link) (struct dentry *,struct inode *,struct dentry *);
2067 int (*unlink) (struct inode *,struct dentry *);
2068 int (*symlink) (struct mnt_idmap *, struct inode *,struct dentry *,
2070 int (*mkdir) (struct mnt_idmap *, struct inode *,struct dentry *,
2072 int (*rmdir) (struct inode *,struct dentry *);
2073 int (*mknod) (struct mnt_idmap *, struct inode *,struct dentry *,
2075 int (*rename) (struct mnt_idmap *, struct inode *, struct dentry *,
2076 struct inode *, struct dentry *, unsigned int);
2077 int (*setattr) (struct mnt_idmap *, struct dentry *, struct iattr *);
2078 int (*getattr) (struct mnt_idmap *, const struct path *,
2079 struct kstat *, u32, unsigned int);
2080 ssize_t (*listxattr) (struct dentry *, char *, size_t);
2081 int (*fiemap)(struct inode *, struct fiemap_extent_info *, u64 start,
2083 int (*update_time)(struct inode *, int);
2084 int (*atomic_open)(struct inode *, struct dentry *,
2085 struct file *, unsigned open_flag,
2086 umode_t create_mode);
2087 int (*tmpfile) (struct mnt_idmap *, struct inode *,
2088 struct file *, umode_t);
2089 struct posix_acl *(*get_acl)(struct mnt_idmap *, struct dentry *,
2091 int (*set_acl)(struct mnt_idmap *, struct dentry *,
2092 struct posix_acl *, int);
2093 int (*fileattr_set)(struct mnt_idmap *idmap,
2094 struct dentry *dentry, struct fileattr *fa);
2095 int (*fileattr_get)(struct dentry *dentry, struct fileattr *fa);
2096 struct offset_ctx *(*get_offset_ctx)(struct inode *inode);
2097 } ____cacheline_aligned;
2099 static inline ssize_t call_read_iter(struct file *file, struct kiocb *kio,
2100 struct iov_iter *iter)
2102 return file->f_op->read_iter(kio, iter);
2105 static inline ssize_t call_write_iter(struct file *file, struct kiocb *kio,
2106 struct iov_iter *iter)
2108 return file->f_op->write_iter(kio, iter);
2111 static inline int call_mmap(struct file *file, struct vm_area_struct *vma)
2113 return file->f_op->mmap(file, vma);
2116 extern ssize_t vfs_read(struct file *, char __user *, size_t, loff_t *);
2117 extern ssize_t vfs_write(struct file *, const char __user *, size_t, loff_t *);
2118 extern ssize_t vfs_copy_file_range(struct file *, loff_t , struct file *,
2119 loff_t, size_t, unsigned int);
2120 int __generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2121 struct file *file_out, loff_t pos_out,
2122 loff_t *len, unsigned int remap_flags,
2123 const struct iomap_ops *dax_read_ops);
2124 int generic_remap_file_range_prep(struct file *file_in, loff_t pos_in,
2125 struct file *file_out, loff_t pos_out,
2126 loff_t *count, unsigned int remap_flags);
2127 extern loff_t vfs_clone_file_range(struct file *file_in, loff_t pos_in,
2128 struct file *file_out, loff_t pos_out,
2129 loff_t len, unsigned int remap_flags);
2130 extern int vfs_dedupe_file_range(struct file *file,
2131 struct file_dedupe_range *same);
2132 extern loff_t vfs_dedupe_file_range_one(struct file *src_file, loff_t src_pos,
2133 struct file *dst_file, loff_t dst_pos,
2134 loff_t len, unsigned int remap_flags);
2137 * enum freeze_holder - holder of the freeze
2138 * @FREEZE_HOLDER_KERNEL: kernel wants to freeze or thaw filesystem
2139 * @FREEZE_HOLDER_USERSPACE: userspace wants to freeze or thaw filesystem
2140 * @FREEZE_MAY_NEST: whether nesting freeze and thaw requests is allowed
2142 * Indicate who the owner of the freeze or thaw request is and whether
2143 * the freeze needs to be exclusive or can nest.
2144 * Without @FREEZE_MAY_NEST, multiple freeze and thaw requests from the
2145 * same holder aren't allowed. It is however allowed to hold a single
2146 * @FREEZE_HOLDER_USERSPACE and a single @FREEZE_HOLDER_KERNEL freeze at
2147 * the same time. This is relied upon by some filesystems during online
2148 * repair or similar.
2150 enum freeze_holder {
2151 FREEZE_HOLDER_KERNEL = (1U << 0),
2152 FREEZE_HOLDER_USERSPACE = (1U << 1),
2153 FREEZE_MAY_NEST = (1U << 2),
2156 struct super_operations {
2157 struct inode *(*alloc_inode)(struct super_block *sb);
2158 void (*destroy_inode)(struct inode *);
2159 void (*free_inode)(struct inode *);
2161 void (*dirty_inode) (struct inode *, int flags);
2162 int (*write_inode) (struct inode *, struct writeback_control *wbc);
2163 int (*drop_inode) (struct inode *);
2164 void (*evict_inode) (struct inode *);
2165 void (*put_super) (struct super_block *);
2166 int (*sync_fs)(struct super_block *sb, int wait);
2167 int (*freeze_super) (struct super_block *, enum freeze_holder who);
2168 int (*freeze_fs) (struct super_block *);
2169 int (*thaw_super) (struct super_block *, enum freeze_holder who);
2170 int (*unfreeze_fs) (struct super_block *);
2171 int (*statfs) (struct dentry *, struct kstatfs *);
2172 int (*remount_fs) (struct super_block *, int *, char *);
2173 void (*umount_begin) (struct super_block *);
2175 int (*show_options)(struct seq_file *, struct dentry *);
2176 int (*show_devname)(struct seq_file *, struct dentry *);
2177 int (*show_path)(struct seq_file *, struct dentry *);
2178 int (*show_stats)(struct seq_file *, struct dentry *);
2180 ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
2181 ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
2182 struct dquot __rcu **(*get_dquots)(struct inode *);
2184 long (*nr_cached_objects)(struct super_block *,
2185 struct shrink_control *);
2186 long (*free_cached_objects)(struct super_block *,
2187 struct shrink_control *);
2188 void (*shutdown)(struct super_block *sb);
2192 * Inode flags - they have no relation to superblock flags now
2194 #define S_SYNC (1 << 0) /* Writes are synced at once */
2195 #define S_NOATIME (1 << 1) /* Do not update access times */
2196 #define S_APPEND (1 << 2) /* Append-only file */
2197 #define S_IMMUTABLE (1 << 3) /* Immutable file */
2198 #define S_DEAD (1 << 4) /* removed, but still open directory */
2199 #define S_NOQUOTA (1 << 5) /* Inode is not counted to quota */
2200 #define S_DIRSYNC (1 << 6) /* Directory modifications are synchronous */
2201 #define S_NOCMTIME (1 << 7) /* Do not update file c/mtime */
2202 #define S_SWAPFILE (1 << 8) /* Do not truncate: swapon got its bmaps */
2203 #define S_PRIVATE (1 << 9) /* Inode is fs-internal */
2204 #define S_IMA (1 << 10) /* Inode has an associated IMA struct */
2205 #define S_AUTOMOUNT (1 << 11) /* Automount/referral quasi-directory */
2206 #define S_NOSEC (1 << 12) /* no suid or xattr security attributes */
2207 #ifdef CONFIG_FS_DAX
2208 #define S_DAX (1 << 13) /* Direct Access, avoiding the page cache */
2210 #define S_DAX 0 /* Make all the DAX code disappear */
2212 #define S_ENCRYPTED (1 << 14) /* Encrypted file (using fs/crypto/) */
2213 #define S_CASEFOLD (1 << 15) /* Casefolded file */
2214 #define S_VERITY (1 << 16) /* Verity file (using fs/verity/) */
2215 #define S_KERNEL_FILE (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
2218 * Note that nosuid etc flags are inode-specific: setting some file-system
2219 * flags just means all the inodes inherit those flags by default. It might be
2220 * possible to override it selectively if you really wanted to with some
2221 * ioctl() that is not currently implemented.
2223 * Exception: SB_RDONLY is always applied to the entire file system.
2225 * Unfortunately, it is possible to change a filesystems flags with it mounted
2226 * with files in use. This means that all of the inodes will not have their
2227 * i_flags updated. Hence, i_flags no longer inherit the superblock mount
2228 * flags, so these have to be checked separately. -- rmk@arm.uk.linux.org
2230 #define __IS_FLG(inode, flg) ((inode)->i_sb->s_flags & (flg))
2232 static inline bool sb_rdonly(const struct super_block *sb) { return sb->s_flags & SB_RDONLY; }
2233 #define IS_RDONLY(inode) sb_rdonly((inode)->i_sb)
2234 #define IS_SYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS) || \
2235 ((inode)->i_flags & S_SYNC))
2236 #define IS_DIRSYNC(inode) (__IS_FLG(inode, SB_SYNCHRONOUS|SB_DIRSYNC) || \
2237 ((inode)->i_flags & (S_SYNC|S_DIRSYNC)))
2238 #define IS_MANDLOCK(inode) __IS_FLG(inode, SB_MANDLOCK)
2239 #define IS_NOATIME(inode) __IS_FLG(inode, SB_RDONLY|SB_NOATIME)
2240 #define IS_I_VERSION(inode) __IS_FLG(inode, SB_I_VERSION)
2242 #define IS_NOQUOTA(inode) ((inode)->i_flags & S_NOQUOTA)
2243 #define IS_APPEND(inode) ((inode)->i_flags & S_APPEND)
2244 #define IS_IMMUTABLE(inode) ((inode)->i_flags & S_IMMUTABLE)
2246 #ifdef CONFIG_FS_POSIX_ACL
2247 #define IS_POSIXACL(inode) __IS_FLG(inode, SB_POSIXACL)
2249 #define IS_POSIXACL(inode) 0
2252 #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD)
2253 #define IS_NOCMTIME(inode) ((inode)->i_flags & S_NOCMTIME)
2254 #define IS_SWAPFILE(inode) ((inode)->i_flags & S_SWAPFILE)
2255 #define IS_PRIVATE(inode) ((inode)->i_flags & S_PRIVATE)
2256 #define IS_IMA(inode) ((inode)->i_flags & S_IMA)
2257 #define IS_AUTOMOUNT(inode) ((inode)->i_flags & S_AUTOMOUNT)
2258 #define IS_NOSEC(inode) ((inode)->i_flags & S_NOSEC)
2259 #define IS_DAX(inode) ((inode)->i_flags & S_DAX)
2260 #define IS_ENCRYPTED(inode) ((inode)->i_flags & S_ENCRYPTED)
2261 #define IS_CASEFOLDED(inode) ((inode)->i_flags & S_CASEFOLD)
2262 #define IS_VERITY(inode) ((inode)->i_flags & S_VERITY)
2264 #define IS_WHITEOUT(inode) (S_ISCHR(inode->i_mode) && \
2265 (inode)->i_rdev == WHITEOUT_DEV)
2267 static inline bool HAS_UNMAPPED_ID(struct mnt_idmap *idmap,
2268 struct inode *inode)
2270 return !vfsuid_valid(i_uid_into_vfsuid(idmap, inode)) ||
2271 !vfsgid_valid(i_gid_into_vfsgid(idmap, inode));
2274 static inline void init_sync_kiocb(struct kiocb *kiocb, struct file *filp)
2276 *kiocb = (struct kiocb) {
2278 .ki_flags = filp->f_iocb_flags,
2279 .ki_ioprio = get_current_ioprio(),
2283 static inline void kiocb_clone(struct kiocb *kiocb, struct kiocb *kiocb_src,
2286 *kiocb = (struct kiocb) {
2288 .ki_flags = kiocb_src->ki_flags,
2289 .ki_ioprio = kiocb_src->ki_ioprio,
2290 .ki_pos = kiocb_src->ki_pos,
2295 * Inode state bits. Protected by inode->i_lock
2297 * Four bits determine the dirty state of the inode: I_DIRTY_SYNC,
2298 * I_DIRTY_DATASYNC, I_DIRTY_PAGES, and I_DIRTY_TIME.
2300 * Four bits define the lifetime of an inode. Initially, inodes are I_NEW,
2301 * until that flag is cleared. I_WILL_FREE, I_FREEING and I_CLEAR are set at
2302 * various stages of removing an inode.
2304 * Two bits are used for locking and completion notification, I_NEW and I_SYNC.
2306 * I_DIRTY_SYNC Inode is dirty, but doesn't have to be written on
2307 * fdatasync() (unless I_DIRTY_DATASYNC is also set).
2308 * Timestamp updates are the usual cause.
2309 * I_DIRTY_DATASYNC Data-related inode changes pending. We keep track of
2310 * these changes separately from I_DIRTY_SYNC so that we
2311 * don't have to write inode on fdatasync() when only
2312 * e.g. the timestamps have changed.
2313 * I_DIRTY_PAGES Inode has dirty pages. Inode itself may be clean.
2314 * I_DIRTY_TIME The inode itself has dirty timestamps, and the
2315 * lazytime mount option is enabled. We keep track of this
2316 * separately from I_DIRTY_SYNC in order to implement
2317 * lazytime. This gets cleared if I_DIRTY_INODE
2318 * (I_DIRTY_SYNC and/or I_DIRTY_DATASYNC) gets set. But
2319 * I_DIRTY_TIME can still be set if I_DIRTY_SYNC is already
2320 * in place because writeback might already be in progress
2321 * and we don't want to lose the time update
2322 * I_NEW Serves as both a mutex and completion notification.
2323 * New inodes set I_NEW. If two processes both create
2324 * the same inode, one of them will release its inode and
2325 * wait for I_NEW to be released before returning.
2326 * Inodes in I_WILL_FREE, I_FREEING or I_CLEAR state can
2327 * also cause waiting on I_NEW, without I_NEW actually
2328 * being set. find_inode() uses this to prevent returning
2329 * nearly-dead inodes.
2330 * I_WILL_FREE Must be set when calling write_inode_now() if i_count
2331 * is zero. I_FREEING must be set when I_WILL_FREE is
2333 * I_FREEING Set when inode is about to be freed but still has dirty
2334 * pages or buffers attached or the inode itself is still
2336 * I_CLEAR Added by clear_inode(). In this state the inode is
2337 * clean and can be destroyed. Inode keeps I_FREEING.
2339 * Inodes that are I_WILL_FREE, I_FREEING or I_CLEAR are
2340 * prohibited for many purposes. iget() must wait for
2341 * the inode to be completely released, then create it
2342 * anew. Other functions will just ignore such inodes,
2343 * if appropriate. I_NEW is used for waiting.
2345 * I_SYNC Writeback of inode is running. The bit is set during
2346 * data writeback, and cleared with a wakeup on the bit
2347 * address once it is done. The bit is also used to pin
2348 * the inode in memory for flusher thread.
2350 * I_REFERENCED Marks the inode as recently references on the LRU list.
2352 * I_DIO_WAKEUP Never set. Only used as a key for wait_on_bit().
2354 * I_WB_SWITCH Cgroup bdi_writeback switching in progress. Used to
2355 * synchronize competing switching instances and to tell
2356 * wb stat updates to grab the i_pages lock. See
2357 * inode_switch_wbs_work_fn() for details.
2359 * I_OVL_INUSE Used by overlayfs to get exclusive ownership on upper
2360 * and work dirs among overlayfs mounts.
2362 * I_CREATING New object's inode in the middle of setting up.
2364 * I_DONTCACHE Evict inode as soon as it is not used anymore.
2366 * I_SYNC_QUEUED Inode is queued in b_io or b_more_io writeback lists.
2367 * Used to detect that mark_inode_dirty() should not move
2368 * inode between dirty lists.
2370 * I_PINNING_FSCACHE_WB Inode is pinning an fscache object for writeback.
2372 * Q: What is the difference between I_WILL_FREE and I_FREEING?
2374 #define I_DIRTY_SYNC (1 << 0)
2375 #define I_DIRTY_DATASYNC (1 << 1)
2376 #define I_DIRTY_PAGES (1 << 2)
2378 #define I_NEW (1 << __I_NEW)
2379 #define I_WILL_FREE (1 << 4)
2380 #define I_FREEING (1 << 5)
2381 #define I_CLEAR (1 << 6)
2383 #define I_SYNC (1 << __I_SYNC)
2384 #define I_REFERENCED (1 << 8)
2385 #define __I_DIO_WAKEUP 9
2386 #define I_DIO_WAKEUP (1 << __I_DIO_WAKEUP)
2387 #define I_LINKABLE (1 << 10)
2388 #define I_DIRTY_TIME (1 << 11)
2389 #define I_WB_SWITCH (1 << 13)
2390 #define I_OVL_INUSE (1 << 14)
2391 #define I_CREATING (1 << 15)
2392 #define I_DONTCACHE (1 << 16)
2393 #define I_SYNC_QUEUED (1 << 17)
2394 #define I_PINNING_NETFS_WB (1 << 18)
2396 #define I_DIRTY_INODE (I_DIRTY_SYNC | I_DIRTY_DATASYNC)
2397 #define I_DIRTY (I_DIRTY_INODE | I_DIRTY_PAGES)
2398 #define I_DIRTY_ALL (I_DIRTY | I_DIRTY_TIME)
2400 extern void __mark_inode_dirty(struct inode *, int);
2401 static inline void mark_inode_dirty(struct inode *inode)
2403 __mark_inode_dirty(inode, I_DIRTY);
2406 static inline void mark_inode_dirty_sync(struct inode *inode)
2408 __mark_inode_dirty(inode, I_DIRTY_SYNC);
2412 * Returns true if the given inode itself only has dirty timestamps (its pages
2413 * may still be dirty) and isn't currently being allocated or freed.
2414 * Filesystems should call this if when writing an inode when lazytime is
2415 * enabled, they want to opportunistically write the timestamps of other inodes
2416 * located very nearby on-disk, e.g. in the same inode block. This returns true
2417 * if the given inode is in need of such an opportunistic update. Requires
2418 * i_lock, or at least later re-checking under i_lock.
2420 static inline bool inode_is_dirtytime_only(struct inode *inode)
2422 return (inode->i_state & (I_DIRTY_TIME | I_NEW |
2423 I_FREEING | I_WILL_FREE)) == I_DIRTY_TIME;
2426 extern void inc_nlink(struct inode *inode);
2427 extern void drop_nlink(struct inode *inode);
2428 extern void clear_nlink(struct inode *inode);
2429 extern void set_nlink(struct inode *inode, unsigned int nlink);
2431 static inline void inode_inc_link_count(struct inode *inode)
2434 mark_inode_dirty(inode);
2437 static inline void inode_dec_link_count(struct inode *inode)
2440 mark_inode_dirty(inode);
2443 enum file_time_flags {
2450 extern bool atime_needs_update(const struct path *, struct inode *);
2451 extern void touch_atime(const struct path *);
2452 int inode_update_time(struct inode *inode, int flags);
2454 static inline void file_accessed(struct file *file)
2456 if (!(file->f_flags & O_NOATIME))
2457 touch_atime(&file->f_path);
2460 extern int file_modified(struct file *file);
2461 int kiocb_modified(struct kiocb *iocb);
2463 int sync_inode_metadata(struct inode *inode, int wait);
2465 struct file_system_type {
2468 #define FS_REQUIRES_DEV 1
2469 #define FS_BINARY_MOUNTDATA 2
2470 #define FS_HAS_SUBTYPE 4
2471 #define FS_USERNS_MOUNT 8 /* Can be mounted by userns root */
2472 #define FS_DISALLOW_NOTIFY_PERM 16 /* Disable fanotify permission events */
2473 #define FS_ALLOW_IDMAP 32 /* FS has been updated to handle vfs idmappings. */
2474 #define FS_RENAME_DOES_D_MOVE 32768 /* FS will handle d_move() during rename() internally. */
2475 int (*init_fs_context)(struct fs_context *);
2476 const struct fs_parameter_spec *parameters;
2477 struct dentry *(*mount) (struct file_system_type *, int,
2478 const char *, void *);
2479 void (*kill_sb) (struct super_block *);
2480 struct module *owner;
2481 struct file_system_type * next;
2482 struct hlist_head fs_supers;
2484 struct lock_class_key s_lock_key;
2485 struct lock_class_key s_umount_key;
2486 struct lock_class_key s_vfs_rename_key;
2487 struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
2489 struct lock_class_key i_lock_key;
2490 struct lock_class_key i_mutex_key;
2491 struct lock_class_key invalidate_lock_key;
2492 struct lock_class_key i_mutex_dir_key;
2495 #define MODULE_ALIAS_FS(NAME) MODULE_ALIAS("fs-" NAME)
2497 extern struct dentry *mount_bdev(struct file_system_type *fs_type,
2498 int flags, const char *dev_name, void *data,
2499 int (*fill_super)(struct super_block *, void *, int));
2500 extern struct dentry *mount_single(struct file_system_type *fs_type,
2501 int flags, void *data,
2502 int (*fill_super)(struct super_block *, void *, int));
2503 extern struct dentry *mount_nodev(struct file_system_type *fs_type,
2504 int flags, void *data,
2505 int (*fill_super)(struct super_block *, void *, int));
2506 extern struct dentry *mount_subtree(struct vfsmount *mnt, const char *path);
2507 void retire_super(struct super_block *sb);
2508 void generic_shutdown_super(struct super_block *sb);
2509 void kill_block_super(struct super_block *sb);
2510 void kill_anon_super(struct super_block *sb);
2511 void kill_litter_super(struct super_block *sb);
2512 void deactivate_super(struct super_block *sb);
2513 void deactivate_locked_super(struct super_block *sb);
2514 int set_anon_super(struct super_block *s, void *data);
2515 int set_anon_super_fc(struct super_block *s, struct fs_context *fc);
2516 int get_anon_bdev(dev_t *);
2517 void free_anon_bdev(dev_t);
2518 struct super_block *sget_fc(struct fs_context *fc,
2519 int (*test)(struct super_block *, struct fs_context *),
2520 int (*set)(struct super_block *, struct fs_context *));
2521 struct super_block *sget(struct file_system_type *type,
2522 int (*test)(struct super_block *,void *),
2523 int (*set)(struct super_block *,void *),
2524 int flags, void *data);
2525 struct super_block *sget_dev(struct fs_context *fc, dev_t dev);
2527 /* Alas, no aliases. Too much hassle with bringing module.h everywhere */
2528 #define fops_get(fops) \
2529 (((fops) && try_module_get((fops)->owner) ? (fops) : NULL))
2530 #define fops_put(fops) \
2531 do { if (fops) module_put((fops)->owner); } while(0)
2533 * This one is to be used *ONLY* from ->open() instances.
2534 * fops must be non-NULL, pinned down *and* module dependencies
2535 * should be sufficient to pin the caller down as well.
2537 #define replace_fops(f, fops) \
2539 struct file *__file = (f); \
2540 fops_put(__file->f_op); \
2541 BUG_ON(!(__file->f_op = (fops))); \
2544 extern int register_filesystem(struct file_system_type *);
2545 extern int unregister_filesystem(struct file_system_type *);
2546 extern int vfs_statfs(const struct path *, struct kstatfs *);
2547 extern int user_statfs(const char __user *, struct kstatfs *);
2548 extern int fd_statfs(int, struct kstatfs *);
2549 int freeze_super(struct super_block *super, enum freeze_holder who);
2550 int thaw_super(struct super_block *super, enum freeze_holder who);
2551 extern __printf(2, 3)
2552 int super_setup_bdi_name(struct super_block *sb, char *fmt, ...);
2553 extern int super_setup_bdi(struct super_block *sb);
2555 static inline void super_set_uuid(struct super_block *sb, const u8 *uuid, unsigned len)
2557 if (WARN_ON(len > sizeof(sb->s_uuid)))
2558 len = sizeof(sb->s_uuid);
2559 sb->s_uuid_len = len;
2560 memcpy(&sb->s_uuid, uuid, len);
2563 /* set sb sysfs name based on sb->s_bdev */
2564 static inline void super_set_sysfs_name_bdev(struct super_block *sb)
2566 snprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), "%pg", sb->s_bdev);
2569 /* set sb sysfs name based on sb->s_uuid */
2570 static inline void super_set_sysfs_name_uuid(struct super_block *sb)
2572 WARN_ON(sb->s_uuid_len != sizeof(sb->s_uuid));
2573 snprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), "%pU", sb->s_uuid.b);
2576 /* set sb sysfs name based on sb->s_id */
2577 static inline void super_set_sysfs_name_id(struct super_block *sb)
2579 strscpy(sb->s_sysfs_name, sb->s_id, sizeof(sb->s_sysfs_name));
2582 /* try to use something standard before you use this */
2584 static inline void super_set_sysfs_name_generic(struct super_block *sb, const char *fmt, ...)
2588 va_start(args, fmt);
2589 vsnprintf(sb->s_sysfs_name, sizeof(sb->s_sysfs_name), fmt, args);
2593 extern int current_umask(void);
2595 extern void ihold(struct inode * inode);
2596 extern void iput(struct inode *);
2597 int inode_update_timestamps(struct inode *inode, int flags);
2598 int generic_update_time(struct inode *, int);
2601 extern struct kobject *fs_kobj;
2603 #define MAX_RW_COUNT (INT_MAX & PAGE_MASK)
2608 const char *name; /* pointer to actual string */
2609 const __user char *uptr; /* original userland pointer */
2611 struct audit_names *aname;
2614 static_assert(offsetof(struct filename, iname) % sizeof(long) == 0);
2616 static inline struct mnt_idmap *file_mnt_idmap(const struct file *file)
2618 return mnt_idmap(file->f_path.mnt);
2622 * is_idmapped_mnt - check whether a mount is mapped
2623 * @mnt: the mount to check
2625 * If @mnt has an non @nop_mnt_idmap attached to it then @mnt is mapped.
2627 * Return: true if mount is mapped, false if not.
2629 static inline bool is_idmapped_mnt(const struct vfsmount *mnt)
2631 return mnt_idmap(mnt) != &nop_mnt_idmap;
2634 extern long vfs_truncate(const struct path *, loff_t);
2635 int do_truncate(struct mnt_idmap *, struct dentry *, loff_t start,
2636 unsigned int time_attrs, struct file *filp);
2637 extern int vfs_fallocate(struct file *file, int mode, loff_t offset,
2639 extern long do_sys_open(int dfd, const char __user *filename, int flags,
2641 extern struct file *file_open_name(struct filename *, int, umode_t);
2642 extern struct file *filp_open(const char *, int, umode_t);
2643 extern struct file *file_open_root(const struct path *,
2644 const char *, int, umode_t);
2645 static inline struct file *file_open_root_mnt(struct vfsmount *mnt,
2646 const char *name, int flags, umode_t mode)
2648 return file_open_root(&(struct path){.mnt = mnt, .dentry = mnt->mnt_root},
2651 struct file *dentry_open(const struct path *path, int flags,
2652 const struct cred *creds);
2653 struct file *dentry_create(const struct path *path, int flags, umode_t mode,
2654 const struct cred *cred);
2655 struct path *backing_file_user_path(struct file *f);
2658 * When mmapping a file on a stackable filesystem (e.g., overlayfs), the file
2659 * stored in ->vm_file is a backing file whose f_inode is on the underlying
2660 * filesystem. When the mapped file path and inode number are displayed to
2661 * user (e.g. via /proc/<pid>/maps), these helpers should be used to get the
2662 * path and inode number to display to the user, which is the path of the fd
2663 * that user has requested to map and the inode number that would be returned
2664 * by fstat() on that same fd.
2666 /* Get the path to display in /proc/<pid>/maps */
2667 static inline const struct path *file_user_path(struct file *f)
2669 if (unlikely(f->f_mode & FMODE_BACKING))
2670 return backing_file_user_path(f);
2673 /* Get the inode whose inode number to display in /proc/<pid>/maps */
2674 static inline const struct inode *file_user_inode(struct file *f)
2676 if (unlikely(f->f_mode & FMODE_BACKING))
2677 return d_inode(backing_file_user_path(f)->dentry);
2678 return file_inode(f);
2681 static inline struct file *file_clone_open(struct file *file)
2683 return dentry_open(&file->f_path, file->f_flags, file->f_cred);
2685 extern int filp_close(struct file *, fl_owner_t id);
2687 extern struct filename *getname_flags(const char __user *, int, int *);
2688 extern struct filename *getname_uflags(const char __user *, int);
2689 extern struct filename *getname(const char __user *);
2690 extern struct filename *getname_kernel(const char *);
2691 extern void putname(struct filename *name);
2693 extern int finish_open(struct file *file, struct dentry *dentry,
2694 int (*open)(struct inode *, struct file *));
2695 extern int finish_no_open(struct file *file, struct dentry *dentry);
2697 /* Helper for the simple case when original dentry is used */
2698 static inline int finish_open_simple(struct file *file, int error)
2703 return finish_open(file, file->f_path.dentry, NULL);
2707 extern void __init vfs_caches_init_early(void);
2708 extern void __init vfs_caches_init(void);
2710 extern struct kmem_cache *names_cachep;
2712 #define __getname() kmem_cache_alloc(names_cachep, GFP_KERNEL)
2713 #define __putname(name) kmem_cache_free(names_cachep, (void *)(name))
2715 extern struct super_block *blockdev_superblock;
2716 static inline bool sb_is_blkdev_sb(struct super_block *sb)
2718 return IS_ENABLED(CONFIG_BLOCK) && sb == blockdev_superblock;
2721 void emergency_thaw_all(void);
2722 extern int sync_filesystem(struct super_block *);
2723 extern const struct file_operations def_blk_fops;
2724 extern const struct file_operations def_chr_fops;
2727 #define CHRDEV_MAJOR_MAX 512
2728 /* Marks the bottom of the first segment of free char majors */
2729 #define CHRDEV_MAJOR_DYN_END 234
2730 /* Marks the top and bottom of the second segment of free char majors */
2731 #define CHRDEV_MAJOR_DYN_EXT_START 511
2732 #define CHRDEV_MAJOR_DYN_EXT_END 384
2734 extern int alloc_chrdev_region(dev_t *, unsigned, unsigned, const char *);
2735 extern int register_chrdev_region(dev_t, unsigned, const char *);
2736 extern int __register_chrdev(unsigned int major, unsigned int baseminor,
2737 unsigned int count, const char *name,
2738 const struct file_operations *fops);
2739 extern void __unregister_chrdev(unsigned int major, unsigned int baseminor,
2740 unsigned int count, const char *name);
2741 extern void unregister_chrdev_region(dev_t, unsigned);
2742 extern void chrdev_show(struct seq_file *,off_t);
2744 static inline int register_chrdev(unsigned int major, const char *name,
2745 const struct file_operations *fops)
2747 return __register_chrdev(major, 0, 256, name, fops);
2750 static inline void unregister_chrdev(unsigned int major, const char *name)
2752 __unregister_chrdev(major, 0, 256, name);
2755 extern void init_special_inode(struct inode *, umode_t, dev_t);
2757 /* Invalid inode operations -- fs/bad_inode.c */
2758 extern void make_bad_inode(struct inode *);
2759 extern bool is_bad_inode(struct inode *);
2761 extern int __must_check file_fdatawait_range(struct file *file, loff_t lstart,
2763 extern int __must_check file_check_and_advance_wb_err(struct file *file);
2764 extern int __must_check file_write_and_wait_range(struct file *file,
2765 loff_t start, loff_t end);
2767 static inline int file_write_and_wait(struct file *file)
2769 return file_write_and_wait_range(file, 0, LLONG_MAX);
2772 extern int vfs_fsync_range(struct file *file, loff_t start, loff_t end,
2774 extern int vfs_fsync(struct file *file, int datasync);
2776 extern int sync_file_range(struct file *file, loff_t offset, loff_t nbytes,
2777 unsigned int flags);
2779 static inline bool iocb_is_dsync(const struct kiocb *iocb)
2781 return (iocb->ki_flags & IOCB_DSYNC) ||
2782 IS_SYNC(iocb->ki_filp->f_mapping->host);
2786 * Sync the bytes written if this was a synchronous write. Expect ki_pos
2787 * to already be updated for the write, and will return either the amount
2788 * of bytes passed in, or an error if syncing the file failed.
2790 static inline ssize_t generic_write_sync(struct kiocb *iocb, ssize_t count)
2792 if (iocb_is_dsync(iocb)) {
2793 int ret = vfs_fsync_range(iocb->ki_filp,
2794 iocb->ki_pos - count, iocb->ki_pos - 1,
2795 (iocb->ki_flags & IOCB_SYNC) ? 0 : 1);
2803 extern void emergency_sync(void);
2804 extern void emergency_remount(void);
2807 extern int bmap(struct inode *inode, sector_t *block);
2809 static inline int bmap(struct inode *inode, sector_t *block)
2815 int notify_change(struct mnt_idmap *, struct dentry *,
2816 struct iattr *, struct inode **);
2817 int inode_permission(struct mnt_idmap *, struct inode *, int);
2818 int generic_permission(struct mnt_idmap *, struct inode *, int);
2819 static inline int file_permission(struct file *file, int mask)
2821 return inode_permission(file_mnt_idmap(file),
2822 file_inode(file), mask);
2824 static inline int path_permission(const struct path *path, int mask)
2826 return inode_permission(mnt_idmap(path->mnt),
2827 d_inode(path->dentry), mask);
2829 int __check_sticky(struct mnt_idmap *idmap, struct inode *dir,
2830 struct inode *inode);
2832 static inline bool execute_ok(struct inode *inode)
2834 return (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode);
2837 static inline bool inode_wrong_type(const struct inode *inode, umode_t mode)
2839 return (inode->i_mode ^ mode) & S_IFMT;
2843 * file_start_write - get write access to a superblock for regular file io
2844 * @file: the file we want to write to
2846 * This is a variant of sb_start_write() which is a noop on non-regualr file.
2847 * Should be matched with a call to file_end_write().
2849 static inline void file_start_write(struct file *file)
2851 if (!S_ISREG(file_inode(file)->i_mode))
2853 sb_start_write(file_inode(file)->i_sb);
2856 static inline bool file_start_write_trylock(struct file *file)
2858 if (!S_ISREG(file_inode(file)->i_mode))
2860 return sb_start_write_trylock(file_inode(file)->i_sb);
2864 * file_end_write - drop write access to a superblock of a regular file
2865 * @file: the file we wrote to
2867 * Should be matched with a call to file_start_write().
2869 static inline void file_end_write(struct file *file)
2871 if (!S_ISREG(file_inode(file)->i_mode))
2873 sb_end_write(file_inode(file)->i_sb);
2877 * kiocb_start_write - get write access to a superblock for async file io
2878 * @iocb: the io context we want to submit the write with
2880 * This is a variant of sb_start_write() for async io submission.
2881 * Should be matched with a call to kiocb_end_write().
2883 static inline void kiocb_start_write(struct kiocb *iocb)
2885 struct inode *inode = file_inode(iocb->ki_filp);
2887 sb_start_write(inode->i_sb);
2889 * Fool lockdep by telling it the lock got released so that it
2890 * doesn't complain about the held lock when we return to userspace.
2892 __sb_writers_release(inode->i_sb, SB_FREEZE_WRITE);
2896 * kiocb_end_write - drop write access to a superblock after async file io
2897 * @iocb: the io context we sumbitted the write with
2899 * Should be matched with a call to kiocb_start_write().
2901 static inline void kiocb_end_write(struct kiocb *iocb)
2903 struct inode *inode = file_inode(iocb->ki_filp);
2906 * Tell lockdep we inherited freeze protection from submission thread.
2908 __sb_writers_acquired(inode->i_sb, SB_FREEZE_WRITE);
2909 sb_end_write(inode->i_sb);
2913 * This is used for regular files where some users -- especially the
2914 * currently executed binary in a process, previously handled via
2915 * VM_DENYWRITE -- cannot handle concurrent write (and maybe mmap
2916 * read-write shared) accesses.
2918 * get_write_access() gets write permission for a file.
2919 * put_write_access() releases this write permission.
2920 * deny_write_access() denies write access to a file.
2921 * allow_write_access() re-enables write access to a file.
2923 * The i_writecount field of an inode can have the following values:
2924 * 0: no write access, no denied write access
2925 * < 0: (-i_writecount) users that denied write access to the file.
2926 * > 0: (i_writecount) users that have write access to the file.
2928 * Normally we operate on that counter with atomic_{inc,dec} and it's safe
2929 * except for the cases where we don't hold i_writecount yet. Then we need to
2930 * use {get,deny}_write_access() - these functions check the sign and refuse
2931 * to do the change if sign is wrong.
2933 static inline int get_write_access(struct inode *inode)
2935 return atomic_inc_unless_negative(&inode->i_writecount) ? 0 : -ETXTBSY;
2937 static inline int deny_write_access(struct file *file)
2939 struct inode *inode = file_inode(file);
2940 return atomic_dec_unless_positive(&inode->i_writecount) ? 0 : -ETXTBSY;
2942 static inline void put_write_access(struct inode * inode)
2944 atomic_dec(&inode->i_writecount);
2946 static inline void allow_write_access(struct file *file)
2949 atomic_inc(&file_inode(file)->i_writecount);
2951 static inline bool inode_is_open_for_write(const struct inode *inode)
2953 return atomic_read(&inode->i_writecount) > 0;
2956 #if defined(CONFIG_IMA) || defined(CONFIG_FILE_LOCKING)
2957 static inline void i_readcount_dec(struct inode *inode)
2959 BUG_ON(atomic_dec_return(&inode->i_readcount) < 0);
2961 static inline void i_readcount_inc(struct inode *inode)
2963 atomic_inc(&inode->i_readcount);
2966 static inline void i_readcount_dec(struct inode *inode)
2970 static inline void i_readcount_inc(struct inode *inode)
2975 extern int do_pipe_flags(int *, int);
2977 extern ssize_t kernel_read(struct file *, void *, size_t, loff_t *);
2978 ssize_t __kernel_read(struct file *file, void *buf, size_t count, loff_t *pos);
2979 extern ssize_t kernel_write(struct file *, const void *, size_t, loff_t *);
2980 extern ssize_t __kernel_write(struct file *, const void *, size_t, loff_t *);
2981 extern struct file * open_exec(const char *);
2983 /* fs/dcache.c -- generic fs support functions */
2984 extern bool is_subdir(struct dentry *, struct dentry *);
2985 extern bool path_is_under(const struct path *, const struct path *);
2987 extern char *file_path(struct file *, char *, int);
2990 * is_dot_dotdot - returns true only if @name is "." or ".."
2991 * @name: file name to check
2992 * @len: length of file name, in bytes
2994 static inline bool is_dot_dotdot(const char *name, size_t len)
2996 return len && unlikely(name[0] == '.') &&
2997 (len == 1 || (len == 2 && name[1] == '.'));
3000 #include <linux/err.h>
3002 /* needed for stackable file system support */
3003 extern loff_t default_llseek(struct file *file, loff_t offset, int whence);
3005 extern loff_t vfs_llseek(struct file *file, loff_t offset, int whence);
3007 extern int inode_init_always(struct super_block *, struct inode *);
3008 extern void inode_init_once(struct inode *);
3009 extern void address_space_init_once(struct address_space *mapping);
3010 extern struct inode * igrab(struct inode *);
3011 extern ino_t iunique(struct super_block *, ino_t);
3012 extern int inode_needs_sync(struct inode *inode);
3013 extern int generic_delete_inode(struct inode *inode);
3014 static inline int generic_drop_inode(struct inode *inode)
3016 return !inode->i_nlink || inode_unhashed(inode);
3018 extern void d_mark_dontcache(struct inode *inode);
3020 extern struct inode *ilookup5_nowait(struct super_block *sb,
3021 unsigned long hashval, int (*test)(struct inode *, void *),
3023 extern struct inode *ilookup5(struct super_block *sb, unsigned long hashval,
3024 int (*test)(struct inode *, void *), void *data);
3025 extern struct inode *ilookup(struct super_block *sb, unsigned long ino);
3027 extern struct inode *inode_insert5(struct inode *inode, unsigned long hashval,
3028 int (*test)(struct inode *, void *),
3029 int (*set)(struct inode *, void *),
3031 extern struct inode * iget5_locked(struct super_block *, unsigned long, int (*test)(struct inode *, void *), int (*set)(struct inode *, void *), void *);
3032 extern struct inode * iget_locked(struct super_block *, unsigned long);
3033 extern struct inode *find_inode_nowait(struct super_block *,
3035 int (*match)(struct inode *,
3036 unsigned long, void *),
3038 extern struct inode *find_inode_rcu(struct super_block *, unsigned long,
3039 int (*)(struct inode *, void *), void *);
3040 extern struct inode *find_inode_by_ino_rcu(struct super_block *, unsigned long);
3041 extern int insert_inode_locked4(struct inode *, unsigned long, int (*test)(struct inode *, void *), void *);
3042 extern int insert_inode_locked(struct inode *);
3043 #ifdef CONFIG_DEBUG_LOCK_ALLOC
3044 extern void lockdep_annotate_inode_mutex_key(struct inode *inode);
3046 static inline void lockdep_annotate_inode_mutex_key(struct inode *inode) { };
3048 extern void unlock_new_inode(struct inode *);
3049 extern void discard_new_inode(struct inode *);
3050 extern unsigned int get_next_ino(void);
3051 extern void evict_inodes(struct super_block *sb);
3052 void dump_mapping(const struct address_space *);
3055 * Userspace may rely on the inode number being non-zero. For example, glibc
3056 * simply ignores files with zero i_ino in unlink() and other places.
3058 * As an additional complication, if userspace was compiled with
3059 * _FILE_OFFSET_BITS=32 on a 64-bit kernel we'll only end up reading out the
3060 * lower 32 bits, so we need to check that those aren't zero explicitly. With
3061 * _FILE_OFFSET_BITS=64, this may cause some harmless false-negatives, but
3062 * better safe than sorry.
3064 static inline bool is_zero_ino(ino_t ino)
3066 return (u32)ino == 0;
3069 extern void __iget(struct inode * inode);
3070 extern void iget_failed(struct inode *);
3071 extern void clear_inode(struct inode *);
3072 extern void __destroy_inode(struct inode *);
3073 extern struct inode *new_inode_pseudo(struct super_block *sb);
3074 extern struct inode *new_inode(struct super_block *sb);
3075 extern void free_inode_nonrcu(struct inode *inode);
3076 extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
3077 extern int file_remove_privs_flags(struct file *file, unsigned int flags);
3078 extern int file_remove_privs(struct file *);
3079 int setattr_should_drop_sgid(struct mnt_idmap *idmap,
3080 const struct inode *inode);
3083 * This must be used for allocating filesystems specific inodes to set
3084 * up the inode reclaim context correctly.
3086 static inline void *
3087 alloc_inode_sb(struct super_block *sb, struct kmem_cache *cache, gfp_t gfp)
3089 return kmem_cache_alloc_lru(cache, &sb->s_inode_lru, gfp);
3092 extern void __insert_inode_hash(struct inode *, unsigned long hashval);
3093 static inline void insert_inode_hash(struct inode *inode)
3095 __insert_inode_hash(inode, inode->i_ino);
3098 extern void __remove_inode_hash(struct inode *);
3099 static inline void remove_inode_hash(struct inode *inode)
3101 if (!inode_unhashed(inode) && !hlist_fake(&inode->i_hash))
3102 __remove_inode_hash(inode);
3105 extern void inode_sb_list_add(struct inode *inode);
3106 extern void inode_add_lru(struct inode *inode);
3108 extern int sb_set_blocksize(struct super_block *, int);
3109 extern int sb_min_blocksize(struct super_block *, int);
3111 extern int generic_file_mmap(struct file *, struct vm_area_struct *);
3112 extern int generic_file_readonly_mmap(struct file *, struct vm_area_struct *);
3113 extern ssize_t generic_write_checks(struct kiocb *, struct iov_iter *);
3114 int generic_write_checks_count(struct kiocb *iocb, loff_t *count);
3115 extern int generic_write_check_limits(struct file *file, loff_t pos,
3117 extern int generic_file_rw_checks(struct file *file_in, struct file *file_out);
3118 ssize_t filemap_read(struct kiocb *iocb, struct iov_iter *to,
3119 ssize_t already_read);
3120 extern ssize_t generic_file_read_iter(struct kiocb *, struct iov_iter *);
3121 extern ssize_t __generic_file_write_iter(struct kiocb *, struct iov_iter *);
3122 extern ssize_t generic_file_write_iter(struct kiocb *, struct iov_iter *);
3123 extern ssize_t generic_file_direct_write(struct kiocb *, struct iov_iter *);
3124 ssize_t generic_perform_write(struct kiocb *, struct iov_iter *);
3125 ssize_t direct_write_fallback(struct kiocb *iocb, struct iov_iter *iter,
3126 ssize_t direct_written, ssize_t buffered_written);
3128 ssize_t vfs_iter_read(struct file *file, struct iov_iter *iter, loff_t *ppos,
3130 ssize_t vfs_iter_write(struct file *file, struct iov_iter *iter, loff_t *ppos,
3132 ssize_t vfs_iocb_iter_read(struct file *file, struct kiocb *iocb,
3133 struct iov_iter *iter);
3134 ssize_t vfs_iocb_iter_write(struct file *file, struct kiocb *iocb,
3135 struct iov_iter *iter);
3138 ssize_t filemap_splice_read(struct file *in, loff_t *ppos,
3139 struct pipe_inode_info *pipe,
3140 size_t len, unsigned int flags);
3141 ssize_t copy_splice_read(struct file *in, loff_t *ppos,
3142 struct pipe_inode_info *pipe,
3143 size_t len, unsigned int flags);
3144 extern ssize_t iter_file_splice_write(struct pipe_inode_info *,
3145 struct file *, loff_t *, size_t, unsigned int);
3149 file_ra_state_init(struct file_ra_state *ra, struct address_space *mapping);
3150 extern loff_t noop_llseek(struct file *file, loff_t offset, int whence);
3151 #define no_llseek NULL
3152 extern loff_t vfs_setpos(struct file *file, loff_t offset, loff_t maxsize);
3153 extern loff_t generic_file_llseek(struct file *file, loff_t offset, int whence);
3154 extern loff_t generic_file_llseek_size(struct file *file, loff_t offset,
3155 int whence, loff_t maxsize, loff_t eof);
3156 extern loff_t fixed_size_llseek(struct file *file, loff_t offset,
3157 int whence, loff_t size);
3158 extern loff_t no_seek_end_llseek_size(struct file *, loff_t, int, loff_t);
3159 extern loff_t no_seek_end_llseek(struct file *, loff_t, int);
3160 int rw_verify_area(int, struct file *, const loff_t *, size_t);
3161 extern int generic_file_open(struct inode * inode, struct file * filp);
3162 extern int nonseekable_open(struct inode * inode, struct file * filp);
3163 extern int stream_open(struct inode * inode, struct file * filp);
3166 typedef void (dio_submit_t)(struct bio *bio, struct inode *inode,
3167 loff_t file_offset);
3170 /* need locking between buffered and direct access */
3173 /* filesystem does not support filling holes */
3174 DIO_SKIP_HOLES = 0x02,
3177 ssize_t __blockdev_direct_IO(struct kiocb *iocb, struct inode *inode,
3178 struct block_device *bdev, struct iov_iter *iter,
3179 get_block_t get_block,
3180 dio_iodone_t end_io,
3183 static inline ssize_t blockdev_direct_IO(struct kiocb *iocb,
3184 struct inode *inode,
3185 struct iov_iter *iter,
3186 get_block_t get_block)
3188 return __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev, iter,
3189 get_block, NULL, DIO_LOCKING | DIO_SKIP_HOLES);
3193 void inode_dio_wait(struct inode *inode);
3196 * inode_dio_begin - signal start of a direct I/O requests
3197 * @inode: inode the direct I/O happens on
3199 * This is called once we've finished processing a direct I/O request,
3200 * and is used to wake up callers waiting for direct I/O to be quiesced.
3202 static inline void inode_dio_begin(struct inode *inode)
3204 atomic_inc(&inode->i_dio_count);
3208 * inode_dio_end - signal finish of a direct I/O requests
3209 * @inode: inode the direct I/O happens on
3211 * This is called once we've finished processing a direct I/O request,
3212 * and is used to wake up callers waiting for direct I/O to be quiesced.
3214 static inline void inode_dio_end(struct inode *inode)
3216 if (atomic_dec_and_test(&inode->i_dio_count))
3217 wake_up_bit(&inode->i_state, __I_DIO_WAKEUP);
3220 extern void inode_set_flags(struct inode *inode, unsigned int flags,
3223 extern const struct file_operations generic_ro_fops;
3225 #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m))
3227 extern int readlink_copy(char __user *, int, const char *);
3228 extern int page_readlink(struct dentry *, char __user *, int);
3229 extern const char *page_get_link(struct dentry *, struct inode *,
3230 struct delayed_call *);
3231 extern void page_put_link(void *);
3232 extern int page_symlink(struct inode *inode, const char *symname, int len);
3233 extern const struct inode_operations page_symlink_inode_operations;
3234 extern void kfree_link(void *);
3235 void generic_fillattr(struct mnt_idmap *, u32, struct inode *, struct kstat *);
3236 void generic_fill_statx_attr(struct inode *inode, struct kstat *stat);
3237 extern int vfs_getattr_nosec(const struct path *, struct kstat *, u32, unsigned int);
3238 extern int vfs_getattr(const struct path *, struct kstat *, u32, unsigned int);
3239 void __inode_add_bytes(struct inode *inode, loff_t bytes);
3240 void inode_add_bytes(struct inode *inode, loff_t bytes);
3241 void __inode_sub_bytes(struct inode *inode, loff_t bytes);
3242 void inode_sub_bytes(struct inode *inode, loff_t bytes);
3243 static inline loff_t __inode_get_bytes(struct inode *inode)
3245 return (((loff_t)inode->i_blocks) << 9) + inode->i_bytes;
3247 loff_t inode_get_bytes(struct inode *inode);
3248 void inode_set_bytes(struct inode *inode, loff_t bytes);
3249 const char *simple_get_link(struct dentry *, struct inode *,
3250 struct delayed_call *);
3251 extern const struct inode_operations simple_symlink_inode_operations;
3253 extern int iterate_dir(struct file *, struct dir_context *);
3255 int vfs_fstatat(int dfd, const char __user *filename, struct kstat *stat,
3257 int vfs_fstat(int fd, struct kstat *stat);
3259 static inline int vfs_stat(const char __user *filename, struct kstat *stat)
3261 return vfs_fstatat(AT_FDCWD, filename, stat, 0);
3263 static inline int vfs_lstat(const char __user *name, struct kstat *stat)
3265 return vfs_fstatat(AT_FDCWD, name, stat, AT_SYMLINK_NOFOLLOW);
3268 extern const char *vfs_get_link(struct dentry *, struct delayed_call *);
3269 extern int vfs_readlink(struct dentry *, char __user *, int);
3271 extern struct file_system_type *get_filesystem(struct file_system_type *fs);
3272 extern void put_filesystem(struct file_system_type *fs);
3273 extern struct file_system_type *get_fs_type(const char *name);
3274 extern void drop_super(struct super_block *sb);
3275 extern void drop_super_exclusive(struct super_block *sb);
3276 extern void iterate_supers(void (*)(struct super_block *, void *), void *);
3277 extern void iterate_supers_type(struct file_system_type *,
3278 void (*)(struct super_block *, void *), void *);
3280 extern int dcache_dir_open(struct inode *, struct file *);
3281 extern int dcache_dir_close(struct inode *, struct file *);
3282 extern loff_t dcache_dir_lseek(struct file *, loff_t, int);
3283 extern int dcache_readdir(struct file *, struct dir_context *);
3284 extern int simple_setattr(struct mnt_idmap *, struct dentry *,
3286 extern int simple_getattr(struct mnt_idmap *, const struct path *,
3287 struct kstat *, u32, unsigned int);
3288 extern int simple_statfs(struct dentry *, struct kstatfs *);
3289 extern int simple_open(struct inode *inode, struct file *file);
3290 extern int simple_link(struct dentry *, struct inode *, struct dentry *);
3291 extern int simple_unlink(struct inode *, struct dentry *);
3292 extern int simple_rmdir(struct inode *, struct dentry *);
3293 void simple_rename_timestamp(struct inode *old_dir, struct dentry *old_dentry,
3294 struct inode *new_dir, struct dentry *new_dentry);
3295 extern int simple_rename_exchange(struct inode *old_dir, struct dentry *old_dentry,
3296 struct inode *new_dir, struct dentry *new_dentry);
3297 extern int simple_rename(struct mnt_idmap *, struct inode *,
3298 struct dentry *, struct inode *, struct dentry *,
3300 extern void simple_recursive_removal(struct dentry *,
3301 void (*callback)(struct dentry *));
3302 extern int noop_fsync(struct file *, loff_t, loff_t, int);
3303 extern ssize_t noop_direct_IO(struct kiocb *iocb, struct iov_iter *iter);
3304 extern int simple_empty(struct dentry *);
3305 extern int simple_write_begin(struct file *file, struct address_space *mapping,
3306 loff_t pos, unsigned len,
3307 struct page **pagep, void **fsdata);
3308 extern const struct address_space_operations ram_aops;
3309 extern int always_delete_dentry(const struct dentry *);
3310 extern struct inode *alloc_anon_inode(struct super_block *);
3311 extern int simple_nosetlease(struct file *, int, struct file_lease **, void **);
3312 extern const struct dentry_operations simple_dentry_operations;
3314 extern struct dentry *simple_lookup(struct inode *, struct dentry *, unsigned int flags);
3315 extern ssize_t generic_read_dir(struct file *, char __user *, size_t, loff_t *);
3316 extern const struct file_operations simple_dir_operations;
3317 extern const struct inode_operations simple_dir_inode_operations;
3318 extern void make_empty_dir_inode(struct inode *inode);
3319 extern bool is_empty_dir_inode(struct inode *inode);
3320 struct tree_descr { const char *name; const struct file_operations *ops; int mode; };
3321 struct dentry *d_alloc_name(struct dentry *, const char *);
3322 extern int simple_fill_super(struct super_block *, unsigned long,
3323 const struct tree_descr *);
3324 extern int simple_pin_fs(struct file_system_type *, struct vfsmount **mount, int *count);
3325 extern void simple_release_fs(struct vfsmount **mount, int *count);
3327 extern ssize_t simple_read_from_buffer(void __user *to, size_t count,
3328 loff_t *ppos, const void *from, size_t available);
3329 extern ssize_t simple_write_to_buffer(void *to, size_t available, loff_t *ppos,
3330 const void __user *from, size_t count);
3333 struct maple_tree mt;
3334 unsigned long next_offset;
3337 void simple_offset_init(struct offset_ctx *octx);
3338 int simple_offset_add(struct offset_ctx *octx, struct dentry *dentry);
3339 void simple_offset_remove(struct offset_ctx *octx, struct dentry *dentry);
3340 int simple_offset_empty(struct dentry *dentry);
3341 int simple_offset_rename_exchange(struct inode *old_dir,
3342 struct dentry *old_dentry,
3343 struct inode *new_dir,
3344 struct dentry *new_dentry);
3345 void simple_offset_destroy(struct offset_ctx *octx);
3347 extern const struct file_operations simple_offset_dir_operations;
3349 extern int __generic_file_fsync(struct file *, loff_t, loff_t, int);
3350 extern int generic_file_fsync(struct file *, loff_t, loff_t, int);
3352 extern int generic_check_addressable(unsigned, u64);
3354 extern void generic_set_sb_d_ops(struct super_block *sb);
3356 static inline bool sb_has_encoding(const struct super_block *sb)
3358 #if IS_ENABLED(CONFIG_UNICODE)
3359 return !!sb->s_encoding;
3365 int may_setattr(struct mnt_idmap *idmap, struct inode *inode,
3366 unsigned int ia_valid);
3367 int setattr_prepare(struct mnt_idmap *, struct dentry *, struct iattr *);
3368 extern int inode_newsize_ok(const struct inode *, loff_t offset);
3369 void setattr_copy(struct mnt_idmap *, struct inode *inode,
3370 const struct iattr *attr);
3372 extern int file_update_time(struct file *file);
3374 static inline bool vma_is_dax(const struct vm_area_struct *vma)
3376 return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
3379 static inline bool vma_is_fsdax(struct vm_area_struct *vma)
3381 struct inode *inode;
3383 if (!IS_ENABLED(CONFIG_FS_DAX) || !vma->vm_file)
3385 if (!vma_is_dax(vma))
3387 inode = file_inode(vma->vm_file);
3388 if (S_ISCHR(inode->i_mode))
3389 return false; /* device-dax */
3393 static inline int iocb_flags(struct file *file)
3396 if (file->f_flags & O_APPEND)
3398 if (file->f_flags & O_DIRECT)
3400 if (file->f_flags & O_DSYNC)
3402 if (file->f_flags & __O_SYNC)
3407 static inline int kiocb_set_rw_flags(struct kiocb *ki, rwf_t flags)
3409 int kiocb_flags = 0;
3411 /* make sure there's no overlap between RWF and private IOCB flags */
3412 BUILD_BUG_ON((__force int) RWF_SUPPORTED & IOCB_EVENTFD);
3416 if (unlikely(flags & ~RWF_SUPPORTED))
3418 if (unlikely((flags & RWF_APPEND) && (flags & RWF_NOAPPEND)))
3421 if (flags & RWF_NOWAIT) {
3422 if (!(ki->ki_filp->f_mode & FMODE_NOWAIT))
3424 kiocb_flags |= IOCB_NOIO;
3426 kiocb_flags |= (__force int) (flags & RWF_SUPPORTED);
3427 if (flags & RWF_SYNC)
3428 kiocb_flags |= IOCB_DSYNC;
3430 if ((flags & RWF_NOAPPEND) && (ki->ki_flags & IOCB_APPEND)) {
3431 if (IS_APPEND(file_inode(ki->ki_filp)))
3433 ki->ki_flags &= ~IOCB_APPEND;
3436 ki->ki_flags |= kiocb_flags;
3440 static inline ino_t parent_ino(struct dentry *dentry)
3445 * Don't strictly need d_lock here? If the parent ino could change
3446 * then surely we'd have a deeper race in the caller?
3448 spin_lock(&dentry->d_lock);
3449 res = dentry->d_parent->d_inode->i_ino;
3450 spin_unlock(&dentry->d_lock);
3454 /* Transaction based IO helpers */
3457 * An argresp is stored in an allocated page and holds the
3458 * size of the argument or response, along with its content
3460 struct simple_transaction_argresp {
3465 #define SIMPLE_TRANSACTION_LIMIT (PAGE_SIZE - sizeof(struct simple_transaction_argresp))
3467 char *simple_transaction_get(struct file *file, const char __user *buf,
3469 ssize_t simple_transaction_read(struct file *file, char __user *buf,
3470 size_t size, loff_t *pos);
3471 int simple_transaction_release(struct inode *inode, struct file *file);
3473 void simple_transaction_set(struct file *file, size_t n);
3476 * simple attribute files
3478 * These attributes behave similar to those in sysfs:
3480 * Writing to an attribute immediately sets a value, an open file can be
3481 * written to multiple times.
3483 * Reading from an attribute creates a buffer from the value that might get
3484 * read with multiple read calls. When the attribute has been read
3485 * completely, no further read calls are possible until the file is opened
3488 * All attributes contain a text representation of a numeric value
3489 * that are accessed with the get() and set() functions.
3491 #define DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, __is_signed) \
3492 static int __fops ## _open(struct inode *inode, struct file *file) \
3494 __simple_attr_check_format(__fmt, 0ull); \
3495 return simple_attr_open(inode, file, __get, __set, __fmt); \
3497 static const struct file_operations __fops = { \
3498 .owner = THIS_MODULE, \
3499 .open = __fops ## _open, \
3500 .release = simple_attr_release, \
3501 .read = simple_attr_read, \
3502 .write = (__is_signed) ? simple_attr_write_signed : simple_attr_write, \
3503 .llseek = generic_file_llseek, \
3506 #define DEFINE_SIMPLE_ATTRIBUTE(__fops, __get, __set, __fmt) \
3507 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, false)
3509 #define DEFINE_SIMPLE_ATTRIBUTE_SIGNED(__fops, __get, __set, __fmt) \
3510 DEFINE_SIMPLE_ATTRIBUTE_XSIGNED(__fops, __get, __set, __fmt, true)
3512 static inline __printf(1, 2)
3513 void __simple_attr_check_format(const char *fmt, ...)
3515 /* don't do anything, just let the compiler check the arguments; */
3518 int simple_attr_open(struct inode *inode, struct file *file,
3519 int (*get)(void *, u64 *), int (*set)(void *, u64),
3521 int simple_attr_release(struct inode *inode, struct file *file);
3522 ssize_t simple_attr_read(struct file *file, char __user *buf,
3523 size_t len, loff_t *ppos);
3524 ssize_t simple_attr_write(struct file *file, const char __user *buf,
3525 size_t len, loff_t *ppos);
3526 ssize_t simple_attr_write_signed(struct file *file, const char __user *buf,
3527 size_t len, loff_t *ppos);
3530 int __init list_bdev_fs_names(char *buf, size_t size);
3532 #define __FMODE_EXEC ((__force int) FMODE_EXEC)
3533 #define __FMODE_NONOTIFY ((__force int) FMODE_NONOTIFY)
3535 #define ACC_MODE(x) ("\004\002\006\006"[(x)&O_ACCMODE])
3536 #define OPEN_FMODE(flag) ((__force fmode_t)(((flag + 1) & O_ACCMODE) | \
3537 (flag & __FMODE_NONOTIFY)))
3539 static inline bool is_sxid(umode_t mode)
3541 return mode & (S_ISUID | S_ISGID);
3544 static inline int check_sticky(struct mnt_idmap *idmap,
3545 struct inode *dir, struct inode *inode)
3547 if (!(dir->i_mode & S_ISVTX))
3550 return __check_sticky(idmap, dir, inode);
3553 static inline void inode_has_no_xattr(struct inode *inode)
3555 if (!is_sxid(inode->i_mode) && (inode->i_sb->s_flags & SB_NOSEC))
3556 inode->i_flags |= S_NOSEC;
3559 static inline bool is_root_inode(struct inode *inode)
3561 return inode == inode->i_sb->s_root->d_inode;
3564 static inline bool dir_emit(struct dir_context *ctx,
3565 const char *name, int namelen,
3566 u64 ino, unsigned type)
3568 return ctx->actor(ctx, name, namelen, ctx->pos, ino, type);
3570 static inline bool dir_emit_dot(struct file *file, struct dir_context *ctx)
3572 return ctx->actor(ctx, ".", 1, ctx->pos,
3573 file->f_path.dentry->d_inode->i_ino, DT_DIR);
3575 static inline bool dir_emit_dotdot(struct file *file, struct dir_context *ctx)
3577 return ctx->actor(ctx, "..", 2, ctx->pos,
3578 parent_ino(file->f_path.dentry), DT_DIR);
3580 static inline bool dir_emit_dots(struct file *file, struct dir_context *ctx)
3582 if (ctx->pos == 0) {
3583 if (!dir_emit_dot(file, ctx))
3587 if (ctx->pos == 1) {
3588 if (!dir_emit_dotdot(file, ctx))
3594 static inline bool dir_relax(struct inode *inode)
3596 inode_unlock(inode);
3598 return !IS_DEADDIR(inode);
3601 static inline bool dir_relax_shared(struct inode *inode)
3603 inode_unlock_shared(inode);
3604 inode_lock_shared(inode);
3605 return !IS_DEADDIR(inode);
3608 extern bool path_noexec(const struct path *path);
3609 extern void inode_nohighmem(struct inode *inode);
3612 extern int vfs_fadvise(struct file *file, loff_t offset, loff_t len,
3614 extern int generic_fadvise(struct file *file, loff_t offset, loff_t len,
3617 #endif /* _LINUX_FS_H */